Page MenuHomeFreeBSD
Files F84250121

No OneTemporary
Actions

View Options

This file is larger than 256 KB, so syntax highlighting was skipped.
diff --git a/cmd/zhack/zhack.c b/cmd/zhack/zhack.c
index 96871d2aab94..73ce888c0b1d 100644
--- a/cmd/zhack/zhack.c
+++ b/cmd/zhack/zhack.c
@@ -1,703 +1,702 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2011, 2015 by Delphix. All rights reserved.
* Copyright (c) 2013 Steven Hartland. All rights reserved.
*/
/*
* zhack is a debugging tool that can write changes to ZFS pool using libzpool
* for testing purposes. Altering pools with zhack is unsupported and may
* result in corrupted pools.
*/
#include <stdio.h>
#include <stdlib.h>
#include <ctype.h>
#include <sys/stat.h>
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/dmu.h>
#include <sys/zap.h>
#include <sys/zfs_znode.h>
#include <sys/dsl_synctask.h>
#include <sys/vdev.h>
#include <sys/vdev_impl.h>
#include <sys/fs/zfs.h>
#include <sys/dmu_objset.h>
#include <sys/dsl_pool.h>
#include <sys/zio_checksum.h>
#include <sys/zio_compress.h>
#include <sys/zfeature.h>
#include <sys/dmu_tx.h>
#include <zfeature_common.h>
#include <libzutil.h>
-const char cmdname[] = "zhack";
static importargs_t g_importargs;
static char *g_pool;
static boolean_t g_readonly;
static __attribute__((noreturn)) void
usage(void)
{
(void) fprintf(stderr,
- "Usage: %s [-c cachefile] [-d dir] <subcommand> <args> ...\n"
+ "Usage: zhack [-c cachefile] [-d dir] <subcommand> <args> ...\n"
"where <subcommand> <args> is one of the following:\n"
- "\n", cmdname);
+ "\n");
(void) fprintf(stderr,
" feature stat <pool>\n"
" print information about enabled features\n"
" feature enable [-r] [-d desc] <pool> <feature>\n"
" add a new enabled feature to the pool\n"
" -d <desc> sets the feature's description\n"
" -r set read-only compatible flag for feature\n"
" feature ref [-md] <pool> <feature>\n"
" change the refcount on the given feature\n"
" -d decrease instead of increase the refcount\n"
" -m add the feature to the label if increasing refcount\n"
"\n"
" <feature> : should be a feature guid\n"
"\n"
" label repair <device>\n"
" repair corrupted label checksums\n"
"\n"
" <device> : path to vdev\n");
exit(1);
}
static __attribute__((noreturn)) __attribute__((format(printf, 3, 4))) void
fatal(spa_t *spa, void *tag, const char *fmt, ...)
{
va_list ap;
if (spa != NULL) {
spa_close(spa, tag);
(void) spa_export(g_pool, NULL, B_TRUE, B_FALSE);
}
va_start(ap, fmt);
- (void) fprintf(stderr, "%s: ", cmdname);
+ (void) fputs("zhack: ", stderr);
(void) vfprintf(stderr, fmt, ap);
va_end(ap);
- (void) fprintf(stderr, "\n");
+ (void) fputc('\n', stderr);
exit(1);
}
static int
space_delta_cb(dmu_object_type_t bonustype, const void *data,
zfs_file_info_t *zoi)
{
(void) data, (void) zoi;
/*
* Is it a valid type of object to track?
*/
if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA)
return (ENOENT);
(void) fprintf(stderr, "modifying object that needs user accounting");
abort();
}
/*
* Target is the dataset whose pool we want to open.
*/
static void
zhack_import(char *target, boolean_t readonly)
{
nvlist_t *config;
nvlist_t *props;
int error;
kernel_init(readonly ? SPA_MODE_READ :
(SPA_MODE_READ | SPA_MODE_WRITE));
dmu_objset_register_type(DMU_OST_ZFS, space_delta_cb);
g_readonly = readonly;
g_importargs.can_be_active = readonly;
g_pool = strdup(target);
error = zpool_find_config(NULL, target, &config, &g_importargs,
&libzpool_config_ops);
if (error)
fatal(NULL, FTAG, "cannot import '%s'", target);
props = NULL;
if (readonly) {
VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0);
VERIFY(nvlist_add_uint64(props,
zpool_prop_to_name(ZPOOL_PROP_READONLY), 1) == 0);
}
zfeature_checks_disable = B_TRUE;
error = spa_import(target, config, props,
(readonly ? ZFS_IMPORT_SKIP_MMP : ZFS_IMPORT_NORMAL));
fnvlist_free(config);
zfeature_checks_disable = B_FALSE;
if (error == EEXIST)
error = 0;
if (error)
fatal(NULL, FTAG, "can't import '%s': %s", target,
strerror(error));
}
static void
zhack_spa_open(char *target, boolean_t readonly, void *tag, spa_t **spa)
{
int err;
zhack_import(target, readonly);
zfeature_checks_disable = B_TRUE;
err = spa_open(target, spa, tag);
zfeature_checks_disable = B_FALSE;
if (err != 0)
fatal(*spa, FTAG, "cannot open '%s': %s", target,
strerror(err));
if (spa_version(*spa) < SPA_VERSION_FEATURES) {
fatal(*spa, FTAG, "'%s' has version %d, features not enabled",
target, (int)spa_version(*spa));
}
}
static void
dump_obj(objset_t *os, uint64_t obj, const char *name)
{
zap_cursor_t zc;
zap_attribute_t za;
(void) printf("%s_obj:\n", name);
for (zap_cursor_init(&zc, os, obj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
if (za.za_integer_length == 8) {
ASSERT(za.za_num_integers == 1);
(void) printf("\t%s = %llu\n",
za.za_name, (u_longlong_t)za.za_first_integer);
} else {
ASSERT(za.za_integer_length == 1);
char val[1024];
VERIFY(zap_lookup(os, obj, za.za_name,
1, sizeof (val), val) == 0);
(void) printf("\t%s = %s\n", za.za_name, val);
}
}
zap_cursor_fini(&zc);
}
static void
dump_mos(spa_t *spa)
{
nvlist_t *nv = spa->spa_label_features;
nvpair_t *pair;
(void) printf("label config:\n");
for (pair = nvlist_next_nvpair(nv, NULL);
pair != NULL;
pair = nvlist_next_nvpair(nv, pair)) {
(void) printf("\t%s\n", nvpair_name(pair));
}
}
static void
zhack_do_feature_stat(int argc, char **argv)
{
spa_t *spa;
objset_t *os;
char *target;
argc--;
argv++;
if (argc < 1) {
(void) fprintf(stderr, "error: missing pool name\n");
usage();
}
target = argv[0];
zhack_spa_open(target, B_TRUE, FTAG, &spa);
os = spa->spa_meta_objset;
dump_obj(os, spa->spa_feat_for_read_obj, "for_read");
dump_obj(os, spa->spa_feat_for_write_obj, "for_write");
dump_obj(os, spa->spa_feat_desc_obj, "descriptions");
if (spa_feature_is_active(spa, SPA_FEATURE_ENABLED_TXG)) {
dump_obj(os, spa->spa_feat_enabled_txg_obj, "enabled_txg");
}
dump_mos(spa);
spa_close(spa, FTAG);
}
static void
zhack_feature_enable_sync(void *arg, dmu_tx_t *tx)
{
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
zfeature_info_t *feature = arg;
feature_enable_sync(spa, feature, tx);
spa_history_log_internal(spa, "zhack enable feature", tx,
"name=%s flags=%u",
feature->fi_guid, feature->fi_flags);
}
static void
zhack_do_feature_enable(int argc, char **argv)
{
int c;
char *desc, *target;
spa_t *spa;
objset_t *mos;
zfeature_info_t feature;
- spa_feature_t nodeps[] = { SPA_FEATURE_NONE };
+ const spa_feature_t nodeps[] = { SPA_FEATURE_NONE };
/*
* Features are not added to the pool's label until their refcounts
* are incremented, so fi_mos can just be left as false for now.
*/
desc = NULL;
feature.fi_uname = "zhack";
feature.fi_flags = 0;
feature.fi_depends = nodeps;
feature.fi_feature = SPA_FEATURE_NONE;
optind = 1;
while ((c = getopt(argc, argv, "+rd:")) != -1) {
switch (c) {
case 'r':
feature.fi_flags |= ZFEATURE_FLAG_READONLY_COMPAT;
break;
case 'd':
desc = strdup(optarg);
break;
default:
usage();
break;
}
}
if (desc == NULL)
desc = strdup("zhack injected");
feature.fi_desc = desc;
argc -= optind;
argv += optind;
if (argc < 2) {
(void) fprintf(stderr, "error: missing feature or pool name\n");
usage();
}
target = argv[0];
feature.fi_guid = argv[1];
if (!zfeature_is_valid_guid(feature.fi_guid))
fatal(NULL, FTAG, "invalid feature guid: %s", feature.fi_guid);
zhack_spa_open(target, B_FALSE, FTAG, &spa);
mos = spa->spa_meta_objset;
if (zfeature_is_supported(feature.fi_guid))
fatal(spa, FTAG, "'%s' is a real feature, will not enable",
feature.fi_guid);
if (0 == zap_contains(mos, spa->spa_feat_desc_obj, feature.fi_guid))
fatal(spa, FTAG, "feature already enabled: %s",
feature.fi_guid);
VERIFY0(dsl_sync_task(spa_name(spa), NULL,
zhack_feature_enable_sync, &feature, 5, ZFS_SPACE_CHECK_NORMAL));
spa_close(spa, FTAG);
free(desc);
}
static void
feature_incr_sync(void *arg, dmu_tx_t *tx)
{
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
zfeature_info_t *feature = arg;
uint64_t refcount;
VERIFY0(feature_get_refcount_from_disk(spa, feature, &refcount));
feature_sync(spa, feature, refcount + 1, tx);
spa_history_log_internal(spa, "zhack feature incr", tx,
"name=%s", feature->fi_guid);
}
static void
feature_decr_sync(void *arg, dmu_tx_t *tx)
{
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
zfeature_info_t *feature = arg;
uint64_t refcount;
VERIFY0(feature_get_refcount_from_disk(spa, feature, &refcount));
feature_sync(spa, feature, refcount - 1, tx);
spa_history_log_internal(spa, "zhack feature decr", tx,
"name=%s", feature->fi_guid);
}
static void
zhack_do_feature_ref(int argc, char **argv)
{
int c;
char *target;
boolean_t decr = B_FALSE;
spa_t *spa;
objset_t *mos;
zfeature_info_t feature;
- spa_feature_t nodeps[] = { SPA_FEATURE_NONE };
+ const spa_feature_t nodeps[] = { SPA_FEATURE_NONE };
/*
* fi_desc does not matter here because it was written to disk
* when the feature was enabled, but we need to properly set the
* feature for read or write based on the information we read off
* disk later.
*/
feature.fi_uname = "zhack";
feature.fi_flags = 0;
feature.fi_desc = NULL;
feature.fi_depends = nodeps;
feature.fi_feature = SPA_FEATURE_NONE;
optind = 1;
while ((c = getopt(argc, argv, "+md")) != -1) {
switch (c) {
case 'm':
feature.fi_flags |= ZFEATURE_FLAG_MOS;
break;
case 'd':
decr = B_TRUE;
break;
default:
usage();
break;
}
}
argc -= optind;
argv += optind;
if (argc < 2) {
(void) fprintf(stderr, "error: missing feature or pool name\n");
usage();
}
target = argv[0];
feature.fi_guid = argv[1];
if (!zfeature_is_valid_guid(feature.fi_guid))
fatal(NULL, FTAG, "invalid feature guid: %s", feature.fi_guid);
zhack_spa_open(target, B_FALSE, FTAG, &spa);
mos = spa->spa_meta_objset;
if (zfeature_is_supported(feature.fi_guid)) {
fatal(spa, FTAG,
"'%s' is a real feature, will not change refcount",
feature.fi_guid);
}
if (0 == zap_contains(mos, spa->spa_feat_for_read_obj,
feature.fi_guid)) {
feature.fi_flags &= ~ZFEATURE_FLAG_READONLY_COMPAT;
} else if (0 == zap_contains(mos, spa->spa_feat_for_write_obj,
feature.fi_guid)) {
feature.fi_flags |= ZFEATURE_FLAG_READONLY_COMPAT;
} else {
fatal(spa, FTAG, "feature is not enabled: %s", feature.fi_guid);
}
if (decr) {
uint64_t count;
if (feature_get_refcount_from_disk(spa, &feature,
&count) == 0 && count == 0) {
fatal(spa, FTAG, "feature refcount already 0: %s",
feature.fi_guid);
}
}
VERIFY0(dsl_sync_task(spa_name(spa), NULL,
decr ? feature_decr_sync : feature_incr_sync, &feature,
5, ZFS_SPACE_CHECK_NORMAL));
spa_close(spa, FTAG);
}
static int
zhack_do_feature(int argc, char **argv)
{
char *subcommand;
argc--;
argv++;
if (argc == 0) {
(void) fprintf(stderr,
"error: no feature operation specified\n");
usage();
}
subcommand = argv[0];
if (strcmp(subcommand, "stat") == 0) {
zhack_do_feature_stat(argc, argv);
} else if (strcmp(subcommand, "enable") == 0) {
zhack_do_feature_enable(argc, argv);
} else if (strcmp(subcommand, "ref") == 0) {
zhack_do_feature_ref(argc, argv);
} else {
(void) fprintf(stderr, "error: unknown subcommand: %s\n",
subcommand);
usage();
}
return (0);
}
static int
zhack_repair_label_cksum(int argc, char **argv)
{
zio_checksum_info_t *ci = &zio_checksum_table[ZIO_CHECKSUM_LABEL];
const char *cfg_keys[] = { ZPOOL_CONFIG_VERSION,
ZPOOL_CONFIG_POOL_STATE, ZPOOL_CONFIG_GUID };
boolean_t labels_repaired[VDEV_LABELS];
boolean_t repaired = B_FALSE;
vdev_label_t labels[VDEV_LABELS];
struct stat st;
int fd;
bzero(labels_repaired, sizeof (labels_repaired));
bzero(labels, sizeof (labels));
abd_init();
argc -= 1;
argv += 1;
if (argc < 1) {
(void) fprintf(stderr, "error: missing device\n");
usage();
}
if ((fd = open(argv[0], O_RDWR)) == -1)
fatal(NULL, FTAG, "cannot open '%s': %s", argv[0],
strerror(errno));
if (stat(argv[0], &st) != 0)
fatal(NULL, FTAG, "cannot stat '%s': %s", argv[0],
strerror(errno));
for (int l = 0; l < VDEV_LABELS; l++) {
uint64_t label_offset, offset;
zio_cksum_t expected_cksum;
zio_cksum_t actual_cksum;
zio_cksum_t verifier;
zio_eck_t *eck;
nvlist_t *cfg;
int byteswap;
uint64_t val;
ssize_t err;
vdev_label_t *vl = &labels[l];
label_offset = vdev_label_offset(st.st_size, l, 0);
err = pread64(fd, vl, sizeof (vdev_label_t), label_offset);
if (err == -1) {
(void) fprintf(stderr, "error: cannot read "
"label %d: %s\n", l, strerror(errno));
continue;
} else if (err != sizeof (vdev_label_t)) {
(void) fprintf(stderr, "error: bad label %d read size "
"\n", l);
continue;
}
err = nvlist_unpack(vl->vl_vdev_phys.vp_nvlist,
VDEV_PHYS_SIZE - sizeof (zio_eck_t), &cfg, 0);
if (err) {
(void) fprintf(stderr, "error: cannot unpack nvlist "
"label %d\n", l);
continue;
}
for (int i = 0; i < ARRAY_SIZE(cfg_keys); i++) {
err = nvlist_lookup_uint64(cfg, cfg_keys[i], &val);
if (err) {
(void) fprintf(stderr, "error: label %d: "
"cannot find nvlist key %s\n",
l, cfg_keys[i]);
continue;
}
}
void *data = (char *)vl + offsetof(vdev_label_t, vl_vdev_phys);
eck = (zio_eck_t *)((char *)(data) + VDEV_PHYS_SIZE) - 1;
offset = label_offset + offsetof(vdev_label_t, vl_vdev_phys);
ZIO_SET_CHECKSUM(&verifier, offset, 0, 0, 0);
byteswap = (eck->zec_magic == BSWAP_64(ZEC_MAGIC));
if (byteswap)
byteswap_uint64_array(&verifier, sizeof (zio_cksum_t));
expected_cksum = eck->zec_cksum;
eck->zec_cksum = verifier;
abd_t *abd = abd_get_from_buf(data, VDEV_PHYS_SIZE);
ci->ci_func[byteswap](abd, VDEV_PHYS_SIZE, NULL, &actual_cksum);
abd_free(abd);
if (byteswap)
byteswap_uint64_array(&expected_cksum,
sizeof (zio_cksum_t));
if (ZIO_CHECKSUM_EQUAL(actual_cksum, expected_cksum))
continue;
eck->zec_cksum = actual_cksum;
err = pwrite64(fd, data, VDEV_PHYS_SIZE, offset);
if (err == -1) {
(void) fprintf(stderr, "error: cannot write "
"label %d: %s\n", l, strerror(errno));
continue;
} else if (err != VDEV_PHYS_SIZE) {
(void) fprintf(stderr, "error: bad write size "
"label %d\n", l);
continue;
}
fsync(fd);
labels_repaired[l] = B_TRUE;
}
close(fd);
abd_fini();
for (int l = 0; l < VDEV_LABELS; l++) {
(void) printf("label %d: %s\n", l,
labels_repaired[l] ? "repaired" : "skipped");
repaired |= labels_repaired[l];
}
if (repaired)
return (0);
return (1);
}
static int
zhack_do_label(int argc, char **argv)
{
char *subcommand;
int err;
argc--;
argv++;
if (argc == 0) {
(void) fprintf(stderr,
"error: no label operation specified\n");
usage();
}
subcommand = argv[0];
if (strcmp(subcommand, "repair") == 0) {
err = zhack_repair_label_cksum(argc, argv);
} else {
(void) fprintf(stderr, "error: unknown subcommand: %s\n",
subcommand);
usage();
}
return (err);
}
#define MAX_NUM_PATHS 1024
int
main(int argc, char **argv)
{
extern void zfs_prop_init(void);
char *path[MAX_NUM_PATHS];
const char *subcommand;
int rv = 0;
int c;
g_importargs.path = path;
dprintf_setup(&argc, argv);
zfs_prop_init();
while ((c = getopt(argc, argv, "+c:d:")) != -1) {
switch (c) {
case 'c':
g_importargs.cachefile = optarg;
break;
case 'd':
assert(g_importargs.paths < MAX_NUM_PATHS);
g_importargs.path[g_importargs.paths++] = optarg;
break;
default:
usage();
break;
}
}
argc -= optind;
argv += optind;
optind = 1;
if (argc == 0) {
(void) fprintf(stderr, "error: no command specified\n");
usage();
}
subcommand = argv[0];
if (strcmp(subcommand, "feature") == 0) {
rv = zhack_do_feature(argc, argv);
} else if (strcmp(subcommand, "label") == 0) {
return (zhack_do_label(argc, argv));
} else {
(void) fprintf(stderr, "error: unknown subcommand: %s\n",
subcommand);
usage();
}
if (!g_readonly && spa_export(g_pool, NULL, B_TRUE, B_FALSE) != 0) {
fatal(NULL, FTAG, "pool export failed; "
"changes may not be committed to disk\n");
}
kernel_fini();
return (rv);
}
diff --git a/cmd/ztest/ztest.c b/cmd/ztest/ztest.c
index a99f60964aec..bb2f14298279 100644
--- a/cmd/ztest/ztest.c
+++ b/cmd/ztest/ztest.c
@@ -1,8208 +1,8208 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2013 Steven Hartland. All rights reserved.
* Copyright (c) 2014 Integros [integros.com]
* Copyright 2017 Joyent, Inc.
* Copyright (c) 2017, Intel Corporation.
*/
/*
* The objective of this program is to provide a DMU/ZAP/SPA stress test
* that runs entirely in userland, is easy to use, and easy to extend.
*
* The overall design of the ztest program is as follows:
*
* (1) For each major functional area (e.g. adding vdevs to a pool,
* creating and destroying datasets, reading and writing objects, etc)
* we have a simple routine to test that functionality. These
* individual routines do not have to do anything "stressful".
*
* (2) We turn these simple functionality tests into a stress test by
* running them all in parallel, with as many threads as desired,
* and spread across as many datasets, objects, and vdevs as desired.
*
* (3) While all this is happening, we inject faults into the pool to
* verify that self-healing data really works.
*
* (4) Every time we open a dataset, we change its checksum and compression
* functions. Thus even individual objects vary from block to block
* in which checksum they use and whether they're compressed.
*
* (5) To verify that we never lose on-disk consistency after a crash,
* we run the entire test in a child of the main process.
* At random times, the child self-immolates with a SIGKILL.
* This is the software equivalent of pulling the power cord.
* The parent then runs the test again, using the existing
* storage pool, as many times as desired. If backwards compatibility
* testing is enabled ztest will sometimes run the "older" version
* of ztest after a SIGKILL.
*
* (6) To verify that we don't have future leaks or temporal incursions,
* many of the functional tests record the transaction group number
* as part of their data. When reading old data, they verify that
* the transaction group number is less than the current, open txg.
* If you add a new test, please do this if applicable.
*
* (7) Threads are created with a reduced stack size, for sanity checking.
* Therefore, it's important not to allocate huge buffers on the stack.
*
* When run with no arguments, ztest runs for about five minutes and
* produces no output if successful. To get a little bit of information,
* specify -V. To get more information, specify -VV, and so on.
*
* To turn this into an overnight stress test, use -T to specify run time.
*
* You can ask more vdevs [-v], datasets [-d], or threads [-t]
* to increase the pool capacity, fanout, and overall stress level.
*
* Use the -k option to set the desired frequency of kills.
*
* When ztest invokes itself it passes all relevant information through a
* temporary file which is mmap-ed in the child process. This allows shared
* memory to survive the exec syscall. The ztest_shared_hdr_t struct is always
* stored at offset 0 of this file and contains information on the size and
* number of shared structures in the file. The information stored in this file
* must remain backwards compatible with older versions of ztest so that
* ztest can invoke them during backwards compatibility testing (-B).
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/dmu.h>
#include <sys/txg.h>
#include <sys/dbuf.h>
#include <sys/zap.h>
#include <sys/dmu_objset.h>
#include <sys/poll.h>
#include <sys/stat.h>
#include <sys/time.h>
#include <sys/wait.h>
#include <sys/mman.h>
#include <sys/resource.h>
#include <sys/zio.h>
#include <sys/zil.h>
#include <sys/zil_impl.h>
#include <sys/vdev_draid.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_file.h>
#include <sys/vdev_initialize.h>
#include <sys/vdev_raidz.h>
#include <sys/vdev_trim.h>
#include <sys/spa_impl.h>
#include <sys/metaslab_impl.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_destroy.h>
#include <sys/dsl_scan.h>
#include <sys/zio_checksum.h>
#include <sys/zfs_refcount.h>
#include <sys/zfeature.h>
#include <sys/dsl_userhold.h>
#include <sys/abd.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <getopt.h>
#include <signal.h>
#include <umem.h>
#include <ctype.h>
#include <math.h>
#include <sys/fs/zfs.h>
#include <zfs_fletcher.h>
#include <libnvpair.h>
#include <libzutil.h>
#include <sys/crypto/icp.h>
#if (__GLIBC__ && !__UCLIBC__)
#include <execinfo.h> /* for backtrace() */
#endif
static int ztest_fd_data = -1;
static int ztest_fd_rand = -1;
typedef struct ztest_shared_hdr {
uint64_t zh_hdr_size;
uint64_t zh_opts_size;
uint64_t zh_size;
uint64_t zh_stats_size;
uint64_t zh_stats_count;
uint64_t zh_ds_size;
uint64_t zh_ds_count;
} ztest_shared_hdr_t;
static ztest_shared_hdr_t *ztest_shared_hdr;
enum ztest_class_state {
ZTEST_VDEV_CLASS_OFF,
ZTEST_VDEV_CLASS_ON,
ZTEST_VDEV_CLASS_RND
};
#define ZO_GVARS_MAX_ARGLEN ((size_t)64)
#define ZO_GVARS_MAX_COUNT ((size_t)10)
typedef struct ztest_shared_opts {
char zo_pool[ZFS_MAX_DATASET_NAME_LEN];
char zo_dir[ZFS_MAX_DATASET_NAME_LEN];
char zo_alt_ztest[MAXNAMELEN];
char zo_alt_libpath[MAXNAMELEN];
uint64_t zo_vdevs;
uint64_t zo_vdevtime;
size_t zo_vdev_size;
int zo_ashift;
int zo_mirrors;
int zo_raid_children;
int zo_raid_parity;
char zo_raid_type[8];
int zo_draid_data;
int zo_draid_spares;
int zo_datasets;
int zo_threads;
uint64_t zo_passtime;
uint64_t zo_killrate;
int zo_verbose;
int zo_init;
uint64_t zo_time;
uint64_t zo_maxloops;
uint64_t zo_metaslab_force_ganging;
int zo_mmp_test;
int zo_special_vdevs;
int zo_dump_dbgmsg;
int zo_gvars_count;
char zo_gvars[ZO_GVARS_MAX_COUNT][ZO_GVARS_MAX_ARGLEN];
} ztest_shared_opts_t;
/* Default values for command line options. */
#define DEFAULT_POOL "ztest"
#define DEFAULT_VDEV_DIR "/tmp"
#define DEFAULT_VDEV_COUNT 5
#define DEFAULT_VDEV_SIZE (SPA_MINDEVSIZE * 4) /* 256m default size */
#define DEFAULT_VDEV_SIZE_STR "256M"
#define DEFAULT_ASHIFT SPA_MINBLOCKSHIFT
#define DEFAULT_MIRRORS 2
#define DEFAULT_RAID_CHILDREN 4
#define DEFAULT_RAID_PARITY 1
#define DEFAULT_DRAID_DATA 4
#define DEFAULT_DRAID_SPARES 1
#define DEFAULT_DATASETS_COUNT 7
#define DEFAULT_THREADS 23
#define DEFAULT_RUN_TIME 300 /* 300 seconds */
#define DEFAULT_RUN_TIME_STR "300 sec"
#define DEFAULT_PASS_TIME 60 /* 60 seconds */
#define DEFAULT_PASS_TIME_STR "60 sec"
#define DEFAULT_KILL_RATE 70 /* 70% kill rate */
#define DEFAULT_KILLRATE_STR "70%"
#define DEFAULT_INITS 1
#define DEFAULT_MAX_LOOPS 50 /* 5 minutes */
#define DEFAULT_FORCE_GANGING (64 << 10)
#define DEFAULT_FORCE_GANGING_STR "64K"
/* Simplifying assumption: -1 is not a valid default. */
#define NO_DEFAULT -1
static const ztest_shared_opts_t ztest_opts_defaults = {
.zo_pool = DEFAULT_POOL,
.zo_dir = DEFAULT_VDEV_DIR,
.zo_alt_ztest = { '\0' },
.zo_alt_libpath = { '\0' },
.zo_vdevs = DEFAULT_VDEV_COUNT,
.zo_ashift = DEFAULT_ASHIFT,
.zo_mirrors = DEFAULT_MIRRORS,
.zo_raid_children = DEFAULT_RAID_CHILDREN,
.zo_raid_parity = DEFAULT_RAID_PARITY,
.zo_raid_type = VDEV_TYPE_RAIDZ,
.zo_vdev_size = DEFAULT_VDEV_SIZE,
.zo_draid_data = DEFAULT_DRAID_DATA, /* data drives */
.zo_draid_spares = DEFAULT_DRAID_SPARES, /* distributed spares */
.zo_datasets = DEFAULT_DATASETS_COUNT,
.zo_threads = DEFAULT_THREADS,
.zo_passtime = DEFAULT_PASS_TIME,
.zo_killrate = DEFAULT_KILL_RATE,
.zo_verbose = 0,
.zo_mmp_test = 0,
.zo_init = DEFAULT_INITS,
.zo_time = DEFAULT_RUN_TIME,
.zo_maxloops = DEFAULT_MAX_LOOPS, /* max loops during spa_freeze() */
.zo_metaslab_force_ganging = DEFAULT_FORCE_GANGING,
.zo_special_vdevs = ZTEST_VDEV_CLASS_RND,
.zo_gvars_count = 0,
};
extern uint64_t metaslab_force_ganging;
extern uint64_t metaslab_df_alloc_threshold;
extern unsigned long zfs_deadman_synctime_ms;
extern int metaslab_preload_limit;
-extern boolean_t zfs_compressed_arc_enabled;
+extern int zfs_compressed_arc_enabled;
extern int zfs_abd_scatter_enabled;
extern int dmu_object_alloc_chunk_shift;
extern boolean_t zfs_force_some_double_word_sm_entries;
extern unsigned long zio_decompress_fail_fraction;
extern unsigned long zfs_reconstruct_indirect_damage_fraction;
static ztest_shared_opts_t *ztest_shared_opts;
static ztest_shared_opts_t ztest_opts;
static char *ztest_wkeydata = "abcdefghijklmnopqrstuvwxyz012345";
typedef struct ztest_shared_ds {
uint64_t zd_seq;
} ztest_shared_ds_t;
static ztest_shared_ds_t *ztest_shared_ds;
#define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d])
#define BT_MAGIC 0x123456789abcdefULL
#define MAXFAULTS(zs) \
(MAX((zs)->zs_mirrors, 1) * (ztest_opts.zo_raid_parity + 1) - 1)
enum ztest_io_type {
ZTEST_IO_WRITE_TAG,
ZTEST_IO_WRITE_PATTERN,
ZTEST_IO_WRITE_ZEROES,
ZTEST_IO_TRUNCATE,
ZTEST_IO_SETATTR,
ZTEST_IO_REWRITE,
ZTEST_IO_TYPES
};
typedef struct ztest_block_tag {
uint64_t bt_magic;
uint64_t bt_objset;
uint64_t bt_object;
uint64_t bt_dnodesize;
uint64_t bt_offset;
uint64_t bt_gen;
uint64_t bt_txg;
uint64_t bt_crtxg;
} ztest_block_tag_t;
typedef struct bufwad {
uint64_t bw_index;
uint64_t bw_txg;
uint64_t bw_data;
} bufwad_t;
/*
* It would be better to use a rangelock_t per object. Unfortunately
* the rangelock_t is not a drop-in replacement for rl_t, because we
* still need to map from object ID to rangelock_t.
*/
typedef enum {
RL_READER,
RL_WRITER,
RL_APPEND
} rl_type_t;
typedef struct rll {
void *rll_writer;
int rll_readers;
kmutex_t rll_lock;
kcondvar_t rll_cv;
} rll_t;
typedef struct rl {
uint64_t rl_object;
uint64_t rl_offset;
uint64_t rl_size;
rll_t *rl_lock;
} rl_t;
#define ZTEST_RANGE_LOCKS 64
#define ZTEST_OBJECT_LOCKS 64
/*
* Object descriptor. Used as a template for object lookup/create/remove.
*/
typedef struct ztest_od {
uint64_t od_dir;
uint64_t od_object;
dmu_object_type_t od_type;
dmu_object_type_t od_crtype;
uint64_t od_blocksize;
uint64_t od_crblocksize;
uint64_t od_crdnodesize;
uint64_t od_gen;
uint64_t od_crgen;
char od_name[ZFS_MAX_DATASET_NAME_LEN];
} ztest_od_t;
/*
* Per-dataset state.
*/
typedef struct ztest_ds {
ztest_shared_ds_t *zd_shared;
objset_t *zd_os;
pthread_rwlock_t zd_zilog_lock;
zilog_t *zd_zilog;
ztest_od_t *zd_od; /* debugging aid */
char zd_name[ZFS_MAX_DATASET_NAME_LEN];
kmutex_t zd_dirobj_lock;
rll_t zd_object_lock[ZTEST_OBJECT_LOCKS];
rll_t zd_range_lock[ZTEST_RANGE_LOCKS];
} ztest_ds_t;
/*
* Per-iteration state.
*/
typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id);
typedef struct ztest_info {
ztest_func_t *zi_func; /* test function */
uint64_t zi_iters; /* iterations per execution */
uint64_t *zi_interval; /* execute every <interval> seconds */
const char *zi_funcname; /* name of test function */
} ztest_info_t;
typedef struct ztest_shared_callstate {
uint64_t zc_count; /* per-pass count */
uint64_t zc_time; /* per-pass time */
uint64_t zc_next; /* next time to call this function */
} ztest_shared_callstate_t;
static ztest_shared_callstate_t *ztest_shared_callstate;
#define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c])
ztest_func_t ztest_dmu_read_write;
ztest_func_t ztest_dmu_write_parallel;
ztest_func_t ztest_dmu_object_alloc_free;
ztest_func_t ztest_dmu_object_next_chunk;
ztest_func_t ztest_dmu_commit_callbacks;
ztest_func_t ztest_zap;
ztest_func_t ztest_zap_parallel;
ztest_func_t ztest_zil_commit;
ztest_func_t ztest_zil_remount;
ztest_func_t ztest_dmu_read_write_zcopy;
ztest_func_t ztest_dmu_objset_create_destroy;
ztest_func_t ztest_dmu_prealloc;
ztest_func_t ztest_fzap;
ztest_func_t ztest_dmu_snapshot_create_destroy;
ztest_func_t ztest_dsl_prop_get_set;
ztest_func_t ztest_spa_prop_get_set;
ztest_func_t ztest_spa_create_destroy;
ztest_func_t ztest_fault_inject;
ztest_func_t ztest_dmu_snapshot_hold;
ztest_func_t ztest_mmp_enable_disable;
ztest_func_t ztest_scrub;
ztest_func_t ztest_dsl_dataset_promote_busy;
ztest_func_t ztest_vdev_attach_detach;
ztest_func_t ztest_vdev_LUN_growth;
ztest_func_t ztest_vdev_add_remove;
ztest_func_t ztest_vdev_class_add;
ztest_func_t ztest_vdev_aux_add_remove;
ztest_func_t ztest_split_pool;
ztest_func_t ztest_reguid;
ztest_func_t ztest_spa_upgrade;
ztest_func_t ztest_device_removal;
ztest_func_t ztest_spa_checkpoint_create_discard;
ztest_func_t ztest_initialize;
ztest_func_t ztest_trim;
ztest_func_t ztest_fletcher;
ztest_func_t ztest_fletcher_incr;
ztest_func_t ztest_verify_dnode_bt;
uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */
uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */
uint64_t zopt_often = 1ULL * NANOSEC; /* every second */
uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */
uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */
#define ZTI_INIT(func, iters, interval) \
{ .zi_func = (func), \
.zi_iters = (iters), \
.zi_interval = (interval), \
.zi_funcname = # func }
ztest_info_t ztest_info[] = {
ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always),
ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always),
ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always),
ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes),
ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always),
ZTI_INIT(ztest_zap, 30, &zopt_always),
ZTI_INIT(ztest_zap_parallel, 100, &zopt_always),
ZTI_INIT(ztest_split_pool, 1, &zopt_always),
ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant),
ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes),
ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often),
ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often),
ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often),
ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes),
#if 0
ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes),
#endif
ZTI_INIT(ztest_fzap, 1, &zopt_sometimes),
ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes),
ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes),
ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes),
ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes),
ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes),
ZTI_INIT(ztest_reguid, 1, &zopt_rarely),
ZTI_INIT(ztest_scrub, 1, &zopt_rarely),
ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely),
ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely),
ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes),
ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely),
ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime),
ZTI_INIT(ztest_vdev_class_add, 1, &ztest_opts.zo_vdevtime),
ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime),
ZTI_INIT(ztest_device_removal, 1, &zopt_sometimes),
ZTI_INIT(ztest_spa_checkpoint_create_discard, 1, &zopt_rarely),
ZTI_INIT(ztest_initialize, 1, &zopt_sometimes),
ZTI_INIT(ztest_trim, 1, &zopt_sometimes),
ZTI_INIT(ztest_fletcher, 1, &zopt_rarely),
ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely),
ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes),
};
#define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t))
/*
* The following struct is used to hold a list of uncalled commit callbacks.
* The callbacks are ordered by txg number.
*/
typedef struct ztest_cb_list {
kmutex_t zcl_callbacks_lock;
list_t zcl_callbacks;
} ztest_cb_list_t;
/*
* Stuff we need to share writably between parent and child.
*/
typedef struct ztest_shared {
boolean_t zs_do_init;
hrtime_t zs_proc_start;
hrtime_t zs_proc_stop;
hrtime_t zs_thread_start;
hrtime_t zs_thread_stop;
hrtime_t zs_thread_kill;
uint64_t zs_enospc_count;
uint64_t zs_vdev_next_leaf;
uint64_t zs_vdev_aux;
uint64_t zs_alloc;
uint64_t zs_space;
uint64_t zs_splits;
uint64_t zs_mirrors;
uint64_t zs_metaslab_sz;
uint64_t zs_metaslab_df_alloc_threshold;
uint64_t zs_guid;
} ztest_shared_t;
#define ID_PARALLEL -1ULL
static char ztest_dev_template[] = "%s/%s.%llua";
static char ztest_aux_template[] = "%s/%s.%s.%llu";
ztest_shared_t *ztest_shared;
static spa_t *ztest_spa = NULL;
static ztest_ds_t *ztest_ds;
static kmutex_t ztest_vdev_lock;
static boolean_t ztest_device_removal_active = B_FALSE;
static boolean_t ztest_pool_scrubbed = B_FALSE;
static kmutex_t ztest_checkpoint_lock;
/*
* The ztest_name_lock protects the pool and dataset namespace used by
* the individual tests. To modify the namespace, consumers must grab
* this lock as writer. Grabbing the lock as reader will ensure that the
* namespace does not change while the lock is held.
*/
static pthread_rwlock_t ztest_name_lock;
static boolean_t ztest_dump_core = B_TRUE;
static boolean_t ztest_exiting;
/* Global commit callback list */
static ztest_cb_list_t zcl;
/* Commit cb delay */
static uint64_t zc_min_txg_delay = UINT64_MAX;
static int zc_cb_counter = 0;
/*
* Minimum number of commit callbacks that need to be registered for us to check
* whether the minimum txg delay is acceptable.
*/
#define ZTEST_COMMIT_CB_MIN_REG 100
/*
* If a number of txgs equal to this threshold have been created after a commit
* callback has been registered but not called, then we assume there is an
* implementation bug.
*/
#define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000)
enum ztest_object {
ZTEST_META_DNODE = 0,
ZTEST_DIROBJ,
ZTEST_OBJECTS
};
static void usage(boolean_t) __NORETURN;
static int ztest_scrub_impl(spa_t *spa);
/*
* These libumem hooks provide a reasonable set of defaults for the allocator's
* debugging facilities.
*/
const char *
_umem_debug_init(void)
{
return ("default,verbose"); /* $UMEM_DEBUG setting */
}
const char *
_umem_logging_init(void)
{
return ("fail,contents"); /* $UMEM_LOGGING setting */
}
static void
dump_debug_buffer(void)
{
ssize_t ret __attribute__((unused));
if (!ztest_opts.zo_dump_dbgmsg)
return;
/*
* We use write() instead of printf() so that this function
* is safe to call from a signal handler.
*/
ret = write(STDOUT_FILENO, "\n", 1);
zfs_dbgmsg_print("ztest");
}
#define BACKTRACE_SZ 100
static void sig_handler(int signo)
{
struct sigaction action;
#if (__GLIBC__ && !__UCLIBC__) /* backtrace() is a GNU extension */
int nptrs;
void *buffer[BACKTRACE_SZ];
nptrs = backtrace(buffer, BACKTRACE_SZ);
backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO);
#endif
dump_debug_buffer();
/*
* Restore default action and re-raise signal so SIGSEGV and
* SIGABRT can trigger a core dump.
*/
action.sa_handler = SIG_DFL;
sigemptyset(&action.sa_mask);
action.sa_flags = 0;
(void) sigaction(signo, &action, NULL);
raise(signo);
}
#define FATAL_MSG_SZ 1024
char *fatal_msg;
static __attribute__((noreturn)) __attribute__((format(printf, 2, 3))) void
fatal(int do_perror, char *message, ...)
{
va_list args;
int save_errno = errno;
char *buf;
(void) fflush(stdout);
buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL);
va_start(args, message);
(void) sprintf(buf, "ztest: ");
/* LINTED */
(void) vsprintf(buf + strlen(buf), message, args);
va_end(args);
if (do_perror) {
(void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf),
": %s", strerror(save_errno));
}
(void) fprintf(stderr, "%s\n", buf);
fatal_msg = buf; /* to ease debugging */
if (ztest_dump_core)
abort();
else
dump_debug_buffer();
exit(3);
}
static int
str2shift(const char *buf)
{
const char *ends = "BKMGTPEZ";
int i;
if (buf[0] == '\0')
return (0);
for (i = 0; i < strlen(ends); i++) {
if (toupper(buf[0]) == ends[i])
break;
}
if (i == strlen(ends)) {
(void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n",
buf);
usage(B_FALSE);
}
if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) {
return (10*i);
}
(void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf);
usage(B_FALSE);
}
static uint64_t
nicenumtoull(const char *buf)
{
char *end;
uint64_t val;
val = strtoull(buf, &end, 0);
if (end == buf) {
(void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf);
usage(B_FALSE);
} else if (end[0] == '.') {
double fval = strtod(buf, &end);
fval *= pow(2, str2shift(end));
/*
* UINT64_MAX is not exactly representable as a double.
* The closest representation is UINT64_MAX + 1, so we
* use a >= comparison instead of > for the bounds check.
*/
if (fval >= (double)UINT64_MAX) {
(void) fprintf(stderr, "ztest: value too large: %s\n",
buf);
usage(B_FALSE);
}
val = (uint64_t)fval;
} else {
int shift = str2shift(end);
if (shift >= 64 || (val << shift) >> shift != val) {
(void) fprintf(stderr, "ztest: value too large: %s\n",
buf);
usage(B_FALSE);
}
val <<= shift;
}
return (val);
}
typedef struct ztest_option {
const char short_opt;
const char *long_opt;
const char *long_opt_param;
const char *comment;
unsigned int default_int;
char *default_str;
} ztest_option_t;
/*
* The following option_table is used for generating the usage info as well as
* the long and short option information for calling getopt_long().
*/
static ztest_option_t option_table[] = {
{ 'v', "vdevs", "INTEGER", "Number of vdevs", DEFAULT_VDEV_COUNT,
NULL},
{ 's', "vdev-size", "INTEGER", "Size of each vdev",
NO_DEFAULT, DEFAULT_VDEV_SIZE_STR},
{ 'a', "alignment-shift", "INTEGER",
"Alignment shift; use 0 for random", DEFAULT_ASHIFT, NULL},
{ 'm', "mirror-copies", "INTEGER", "Number of mirror copies",
DEFAULT_MIRRORS, NULL},
{ 'r', "raid-disks", "INTEGER", "Number of raidz/draid disks",
DEFAULT_RAID_CHILDREN, NULL},
{ 'R', "raid-parity", "INTEGER", "Raid parity",
DEFAULT_RAID_PARITY, NULL},
{ 'K', "raid-kind", "raidz|draid|random", "Raid kind",
NO_DEFAULT, "random"},
{ 'D', "draid-data", "INTEGER", "Number of draid data drives",
DEFAULT_DRAID_DATA, NULL},
{ 'S', "draid-spares", "INTEGER", "Number of draid spares",
DEFAULT_DRAID_SPARES, NULL},
{ 'd', "datasets", "INTEGER", "Number of datasets",
DEFAULT_DATASETS_COUNT, NULL},
{ 't', "threads", "INTEGER", "Number of ztest threads",
DEFAULT_THREADS, NULL},
{ 'g', "gang-block-threshold", "INTEGER",
"Metaslab gang block threshold",
NO_DEFAULT, DEFAULT_FORCE_GANGING_STR},
{ 'i', "init-count", "INTEGER", "Number of times to initialize pool",
DEFAULT_INITS, NULL},
{ 'k', "kill-percentage", "INTEGER", "Kill percentage",
NO_DEFAULT, DEFAULT_KILLRATE_STR},
{ 'p', "pool-name", "STRING", "Pool name",
NO_DEFAULT, DEFAULT_POOL},
{ 'f', "vdev-file-directory", "PATH", "File directory for vdev files",
NO_DEFAULT, DEFAULT_VDEV_DIR},
{ 'M', "multi-host", NULL,
"Multi-host; simulate pool imported on remote host",
NO_DEFAULT, NULL},
{ 'E', "use-existing-pool", NULL,
"Use existing pool instead of creating new one", NO_DEFAULT, NULL},
{ 'T', "run-time", "INTEGER", "Total run time",
NO_DEFAULT, DEFAULT_RUN_TIME_STR},
{ 'P', "pass-time", "INTEGER", "Time per pass",
NO_DEFAULT, DEFAULT_PASS_TIME_STR},
{ 'F', "freeze-loops", "INTEGER", "Max loops in spa_freeze()",
DEFAULT_MAX_LOOPS, NULL},
{ 'B', "alt-ztest", "PATH", "Alternate ztest path",
NO_DEFAULT, NULL},
{ 'C', "vdev-class-state", "on|off|random", "vdev class state",
NO_DEFAULT, "random"},
{ 'o', "option", "\"OPTION=INTEGER\"",
"Set global variable to an unsigned 32-bit integer value",
NO_DEFAULT, NULL},
{ 'G', "dump-debug-msg", NULL,
"Dump zfs_dbgmsg buffer before exiting due to an error",
NO_DEFAULT, NULL},
{ 'V', "verbose", NULL,
"Verbose (use multiple times for ever more verbosity)",
NO_DEFAULT, NULL},
{ 'h', "help", NULL, "Show this help",
NO_DEFAULT, NULL},
{0, 0, 0, 0, 0, 0}
};
static struct option *long_opts = NULL;
static char *short_opts = NULL;
static void
init_options(void)
{
ASSERT3P(long_opts, ==, NULL);
ASSERT3P(short_opts, ==, NULL);
int count = sizeof (option_table) / sizeof (option_table[0]);
long_opts = umem_alloc(sizeof (struct option) * count, UMEM_NOFAIL);
short_opts = umem_alloc(sizeof (char) * 2 * count, UMEM_NOFAIL);
int short_opt_index = 0;
for (int i = 0; i < count; i++) {
long_opts[i].val = option_table[i].short_opt;
long_opts[i].name = option_table[i].long_opt;
long_opts[i].has_arg = option_table[i].long_opt_param != NULL
? required_argument : no_argument;
long_opts[i].flag = NULL;
short_opts[short_opt_index++] = option_table[i].short_opt;
if (option_table[i].long_opt_param != NULL) {
short_opts[short_opt_index++] = ':';
}
}
}
static void
fini_options(void)
{
int count = sizeof (option_table) / sizeof (option_table[0]);
umem_free(long_opts, sizeof (struct option) * count);
umem_free(short_opts, sizeof (char) * 2 * count);
long_opts = NULL;
short_opts = NULL;
}
static void
usage(boolean_t requested)
{
char option[80];
FILE *fp = requested ? stdout : stderr;
(void) fprintf(fp, "Usage: %s [OPTIONS...]\n", DEFAULT_POOL);
for (int i = 0; option_table[i].short_opt != 0; i++) {
if (option_table[i].long_opt_param != NULL) {
(void) sprintf(option, " -%c --%s=%s",
option_table[i].short_opt,
option_table[i].long_opt,
option_table[i].long_opt_param);
} else {
(void) sprintf(option, " -%c --%s",
option_table[i].short_opt,
option_table[i].long_opt);
}
(void) fprintf(fp, " %-40s%s", option,
option_table[i].comment);
if (option_table[i].long_opt_param != NULL) {
if (option_table[i].default_str != NULL) {
(void) fprintf(fp, " (default: %s)",
option_table[i].default_str);
} else if (option_table[i].default_int != NO_DEFAULT) {
(void) fprintf(fp, " (default: %u)",
option_table[i].default_int);
}
}
(void) fprintf(fp, "\n");
}
exit(requested ? 0 : 1);
}
static uint64_t
ztest_random(uint64_t range)
{
uint64_t r;
ASSERT3S(ztest_fd_rand, >=, 0);
if (range == 0)
return (0);
if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r))
fatal(B_TRUE, "short read from /dev/urandom");
return (r % range);
}
static void
ztest_parse_name_value(const char *input, ztest_shared_opts_t *zo)
{
char name[32];
char *value;
int state = ZTEST_VDEV_CLASS_RND;
(void) strlcpy(name, input, sizeof (name));
value = strchr(name, '=');
if (value == NULL) {
(void) fprintf(stderr, "missing value in property=value "
"'-C' argument (%s)\n", input);
usage(B_FALSE);
}
*(value) = '\0';
value++;
if (strcmp(value, "on") == 0) {
state = ZTEST_VDEV_CLASS_ON;
} else if (strcmp(value, "off") == 0) {
state = ZTEST_VDEV_CLASS_OFF;
} else if (strcmp(value, "random") == 0) {
state = ZTEST_VDEV_CLASS_RND;
} else {
(void) fprintf(stderr, "invalid property value '%s'\n", value);
usage(B_FALSE);
}
if (strcmp(name, "special") == 0) {
zo->zo_special_vdevs = state;
} else {
(void) fprintf(stderr, "invalid property name '%s'\n", name);
usage(B_FALSE);
}
if (zo->zo_verbose >= 3)
(void) printf("%s vdev state is '%s'\n", name, value);
}
static void
process_options(int argc, char **argv)
{
char *path;
ztest_shared_opts_t *zo = &ztest_opts;
int opt;
uint64_t value;
char altdir[MAXNAMELEN] = { 0 };
char raid_kind[8] = { "random" };
bcopy(&ztest_opts_defaults, zo, sizeof (*zo));
init_options();
while ((opt = getopt_long(argc, argv, short_opts, long_opts,
NULL)) != EOF) {
value = 0;
switch (opt) {
case 'v':
case 's':
case 'a':
case 'm':
case 'r':
case 'R':
case 'D':
case 'S':
case 'd':
case 't':
case 'g':
case 'i':
case 'k':
case 'T':
case 'P':
case 'F':
value = nicenumtoull(optarg);
}
switch (opt) {
case 'v':
zo->zo_vdevs = value;
break;
case 's':
zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value);
break;
case 'a':
zo->zo_ashift = value;
break;
case 'm':
zo->zo_mirrors = value;
break;
case 'r':
zo->zo_raid_children = MAX(1, value);
break;
case 'R':
zo->zo_raid_parity = MIN(MAX(value, 1), 3);
break;
case 'K':
(void) strlcpy(raid_kind, optarg, sizeof (raid_kind));
break;
case 'D':
zo->zo_draid_data = MAX(1, value);
break;
case 'S':
zo->zo_draid_spares = MAX(1, value);
break;
case 'd':
zo->zo_datasets = MAX(1, value);
break;
case 't':
zo->zo_threads = MAX(1, value);
break;
case 'g':
zo->zo_metaslab_force_ganging =
MAX(SPA_MINBLOCKSIZE << 1, value);
break;
case 'i':
zo->zo_init = value;
break;
case 'k':
zo->zo_killrate = value;
break;
case 'p':
(void) strlcpy(zo->zo_pool, optarg,
sizeof (zo->zo_pool));
break;
case 'f':
path = realpath(optarg, NULL);
if (path == NULL) {
(void) fprintf(stderr, "error: %s: %s\n",
optarg, strerror(errno));
usage(B_FALSE);
} else {
(void) strlcpy(zo->zo_dir, path,
sizeof (zo->zo_dir));
free(path);
}
break;
case 'M':
zo->zo_mmp_test = 1;
break;
case 'V':
zo->zo_verbose++;
break;
case 'E':
zo->zo_init = 0;
break;
case 'T':
zo->zo_time = value;
break;
case 'P':
zo->zo_passtime = MAX(1, value);
break;
case 'F':
zo->zo_maxloops = MAX(1, value);
break;
case 'B':
(void) strlcpy(altdir, optarg, sizeof (altdir));
break;
case 'C':
ztest_parse_name_value(optarg, zo);
break;
case 'o':
if (zo->zo_gvars_count >= ZO_GVARS_MAX_COUNT) {
(void) fprintf(stderr,
"max global var count (%zu) exceeded\n",
ZO_GVARS_MAX_COUNT);
usage(B_FALSE);
}
char *v = zo->zo_gvars[zo->zo_gvars_count];
if (strlcpy(v, optarg, ZO_GVARS_MAX_ARGLEN) >=
ZO_GVARS_MAX_ARGLEN) {
(void) fprintf(stderr,
"global var option '%s' is too long\n",
optarg);
usage(B_FALSE);
}
zo->zo_gvars_count++;
break;
case 'G':
zo->zo_dump_dbgmsg = 1;
break;
case 'h':
usage(B_TRUE);
break;
case '?':
default:
usage(B_FALSE);
break;
}
}
fini_options();
/* When raid choice is 'random' add a draid pool 50% of the time */
if (strcmp(raid_kind, "random") == 0) {
(void) strlcpy(raid_kind, (ztest_random(2) == 0) ?
"draid" : "raidz", sizeof (raid_kind));
if (ztest_opts.zo_verbose >= 3)
(void) printf("choosing RAID type '%s'\n", raid_kind);
}
if (strcmp(raid_kind, "draid") == 0) {
uint64_t min_devsize;
/* With fewer disk use 256M, otherwise 128M is OK */
min_devsize = (ztest_opts.zo_raid_children < 16) ?
(256ULL << 20) : (128ULL << 20);
/* No top-level mirrors with dRAID for now */
zo->zo_mirrors = 0;
/* Use more appropriate defaults for dRAID */
if (zo->zo_vdevs == ztest_opts_defaults.zo_vdevs)
zo->zo_vdevs = 1;
if (zo->zo_raid_children ==
ztest_opts_defaults.zo_raid_children)
zo->zo_raid_children = 16;
if (zo->zo_ashift < 12)
zo->zo_ashift = 12;
if (zo->zo_vdev_size < min_devsize)
zo->zo_vdev_size = min_devsize;
if (zo->zo_draid_data + zo->zo_raid_parity >
zo->zo_raid_children - zo->zo_draid_spares) {
(void) fprintf(stderr, "error: too few draid "
"children (%d) for stripe width (%d)\n",
zo->zo_raid_children,
zo->zo_draid_data + zo->zo_raid_parity);
usage(B_FALSE);
}
(void) strlcpy(zo->zo_raid_type, VDEV_TYPE_DRAID,
sizeof (zo->zo_raid_type));
} else /* using raidz */ {
ASSERT0(strcmp(raid_kind, "raidz"));
zo->zo_raid_parity = MIN(zo->zo_raid_parity,
zo->zo_raid_children - 1);
}
zo->zo_vdevtime =
(zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs :
UINT64_MAX >> 2);
if (strlen(altdir) > 0) {
char *cmd;
char *realaltdir;
char *bin;
char *ztest;
char *isa;
int isalen;
cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
VERIFY3P(NULL, !=, realpath(getexecname(), cmd));
if (0 != access(altdir, F_OK)) {
ztest_dump_core = B_FALSE;
fatal(B_TRUE, "invalid alternate ztest path: %s",
altdir);
}
VERIFY3P(NULL, !=, realpath(altdir, realaltdir));
/*
* 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest".
* We want to extract <isa> to determine if we should use
* 32 or 64 bit binaries.
*/
bin = strstr(cmd, "/usr/bin/");
ztest = strstr(bin, "/ztest");
isa = bin + 9;
isalen = ztest - isa;
(void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest),
"%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa);
(void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath),
"%s/usr/lib/%.*s", realaltdir, isalen, isa);
if (0 != access(zo->zo_alt_ztest, X_OK)) {
ztest_dump_core = B_FALSE;
fatal(B_TRUE, "invalid alternate ztest: %s",
zo->zo_alt_ztest);
} else if (0 != access(zo->zo_alt_libpath, X_OK)) {
ztest_dump_core = B_FALSE;
fatal(B_TRUE, "invalid alternate lib directory %s",
zo->zo_alt_libpath);
}
umem_free(cmd, MAXPATHLEN);
umem_free(realaltdir, MAXPATHLEN);
}
}
static void
ztest_kill(ztest_shared_t *zs)
{
zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa));
zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa));
/*
* Before we kill off ztest, make sure that the config is updated.
* See comment above spa_write_cachefile().
*/
mutex_enter(&spa_namespace_lock);
spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE);
mutex_exit(&spa_namespace_lock);
(void) kill(getpid(), SIGKILL);
}
static void
ztest_record_enospc(const char *s)
{
(void) s;
ztest_shared->zs_enospc_count++;
}
static uint64_t
ztest_get_ashift(void)
{
if (ztest_opts.zo_ashift == 0)
return (SPA_MINBLOCKSHIFT + ztest_random(5));
return (ztest_opts.zo_ashift);
}
static boolean_t
ztest_is_draid_spare(const char *name)
{
uint64_t spare_id = 0, parity = 0, vdev_id = 0;
if (sscanf(name, VDEV_TYPE_DRAID "%"PRIu64"-%"PRIu64"-%"PRIu64"",
&parity, &vdev_id, &spare_id) == 3) {
return (B_TRUE);
}
return (B_FALSE);
}
static nvlist_t *
make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift)
{
char *pathbuf;
uint64_t vdev;
nvlist_t *file;
boolean_t draid_spare = B_FALSE;
pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
if (ashift == 0)
ashift = ztest_get_ashift();
if (path == NULL) {
path = pathbuf;
if (aux != NULL) {
vdev = ztest_shared->zs_vdev_aux;
(void) snprintf(path, MAXPATHLEN,
ztest_aux_template, ztest_opts.zo_dir,
pool == NULL ? ztest_opts.zo_pool : pool,
aux, vdev);
} else {
vdev = ztest_shared->zs_vdev_next_leaf++;
(void) snprintf(path, MAXPATHLEN,
ztest_dev_template, ztest_opts.zo_dir,
pool == NULL ? ztest_opts.zo_pool : pool, vdev);
}
} else {
draid_spare = ztest_is_draid_spare(path);
}
if (size != 0 && !draid_spare) {
int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666);
if (fd == -1)
fatal(B_TRUE, "can't open %s", path);
if (ftruncate(fd, size) != 0)
fatal(B_TRUE, "can't ftruncate %s", path);
(void) close(fd);
}
file = fnvlist_alloc();
fnvlist_add_string(file, ZPOOL_CONFIG_TYPE,
draid_spare ? VDEV_TYPE_DRAID_SPARE : VDEV_TYPE_FILE);
fnvlist_add_string(file, ZPOOL_CONFIG_PATH, path);
fnvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift);
umem_free(pathbuf, MAXPATHLEN);
return (file);
}
static nvlist_t *
make_vdev_raid(char *path, char *aux, char *pool, size_t size,
uint64_t ashift, int r)
{
nvlist_t *raid, **child;
int c;
if (r < 2)
return (make_vdev_file(path, aux, pool, size, ashift));
child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL);
for (c = 0; c < r; c++)
child[c] = make_vdev_file(path, aux, pool, size, ashift);
raid = fnvlist_alloc();
fnvlist_add_string(raid, ZPOOL_CONFIG_TYPE,
ztest_opts.zo_raid_type);
fnvlist_add_uint64(raid, ZPOOL_CONFIG_NPARITY,
ztest_opts.zo_raid_parity);
fnvlist_add_nvlist_array(raid, ZPOOL_CONFIG_CHILDREN,
(const nvlist_t **)child, r);
if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0) {
uint64_t ndata = ztest_opts.zo_draid_data;
uint64_t nparity = ztest_opts.zo_raid_parity;
uint64_t nspares = ztest_opts.zo_draid_spares;
uint64_t children = ztest_opts.zo_raid_children;
uint64_t ngroups = 1;
/*
* Calculate the minimum number of groups required to fill a
* slice. This is the LCM of the stripe width (data + parity)
* and the number of data drives (children - spares).
*/
while (ngroups * (ndata + nparity) % (children - nspares) != 0)
ngroups++;
/* Store the basic dRAID configuration. */
fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NDATA, ndata);
fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NSPARES, nspares);
fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NGROUPS, ngroups);
}
for (c = 0; c < r; c++)
fnvlist_free(child[c]);
umem_free(child, r * sizeof (nvlist_t *));
return (raid);
}
static nvlist_t *
make_vdev_mirror(char *path, char *aux, char *pool, size_t size,
uint64_t ashift, int r, int m)
{
nvlist_t *mirror, **child;
int c;
if (m < 1)
return (make_vdev_raid(path, aux, pool, size, ashift, r));
child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL);
for (c = 0; c < m; c++)
child[c] = make_vdev_raid(path, aux, pool, size, ashift, r);
mirror = fnvlist_alloc();
fnvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, VDEV_TYPE_MIRROR);
fnvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN,
(const nvlist_t **)child, m);
for (c = 0; c < m; c++)
fnvlist_free(child[c]);
umem_free(child, m * sizeof (nvlist_t *));
return (mirror);
}
static nvlist_t *
make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift,
const char *class, int r, int m, int t)
{
nvlist_t *root, **child;
int c;
boolean_t log;
ASSERT3S(t, >, 0);
log = (class != NULL && strcmp(class, "log") == 0);
child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL);
for (c = 0; c < t; c++) {
child[c] = make_vdev_mirror(path, aux, pool, size, ashift,
r, m);
fnvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, log);
if (class != NULL && class[0] != '\0') {
ASSERT(m > 1 || log); /* expecting a mirror */
fnvlist_add_string(child[c],
ZPOOL_CONFIG_ALLOCATION_BIAS, class);
}
}
root = fnvlist_alloc();
fnvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT);
fnvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN,
(const nvlist_t **)child, t);
for (c = 0; c < t; c++)
fnvlist_free(child[c]);
umem_free(child, t * sizeof (nvlist_t *));
return (root);
}
/*
* Find a random spa version. Returns back a random spa version in the
* range [initial_version, SPA_VERSION_FEATURES].
*/
static uint64_t
ztest_random_spa_version(uint64_t initial_version)
{
uint64_t version = initial_version;
if (version <= SPA_VERSION_BEFORE_FEATURES) {
version = version +
ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1);
}
if (version > SPA_VERSION_BEFORE_FEATURES)
version = SPA_VERSION_FEATURES;
ASSERT(SPA_VERSION_IS_SUPPORTED(version));
return (version);
}
static int
ztest_random_blocksize(void)
{
ASSERT3U(ztest_spa->spa_max_ashift, !=, 0);
/*
* Choose a block size >= the ashift.
* If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks.
*/
int maxbs = SPA_OLD_MAXBLOCKSHIFT;
if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE)
maxbs = 20;
uint64_t block_shift =
ztest_random(maxbs - ztest_spa->spa_max_ashift + 1);
return (1 << (SPA_MINBLOCKSHIFT + block_shift));
}
static int
ztest_random_dnodesize(void)
{
int slots;
int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT;
if (max_slots == DNODE_MIN_SLOTS)
return (DNODE_MIN_SIZE);
/*
* Weight the random distribution more heavily toward smaller
* dnode sizes since that is more likely to reflect real-world
* usage.
*/
ASSERT3U(max_slots, >, 4);
switch (ztest_random(10)) {
case 0:
slots = 5 + ztest_random(max_slots - 4);
break;
case 1 ... 4:
slots = 2 + ztest_random(3);
break;
default:
slots = 1;
break;
}
return (slots << DNODE_SHIFT);
}
static int
ztest_random_ibshift(void)
{
return (DN_MIN_INDBLKSHIFT +
ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1));
}
static uint64_t
ztest_random_vdev_top(spa_t *spa, boolean_t log_ok)
{
uint64_t top;
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *tvd;
ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);
do {
top = ztest_random(rvd->vdev_children);
tvd = rvd->vdev_child[top];
} while (!vdev_is_concrete(tvd) || (tvd->vdev_islog && !log_ok) ||
tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL);
return (top);
}
static uint64_t
ztest_random_dsl_prop(zfs_prop_t prop)
{
uint64_t value;
do {
value = zfs_prop_random_value(prop, ztest_random(-1ULL));
} while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF);
return (value);
}
static int
ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value,
boolean_t inherit)
{
const char *propname = zfs_prop_to_name(prop);
const char *valname;
char *setpoint;
uint64_t curval;
int error;
error = dsl_prop_set_int(osname, propname,
(inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value);
if (error == ENOSPC) {
ztest_record_enospc(FTAG);
return (error);
}
ASSERT0(error);
setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint));
if (ztest_opts.zo_verbose >= 6) {
int err;
err = zfs_prop_index_to_string(prop, curval, &valname);
if (err)
(void) printf("%s %s = %llu at '%s'\n", osname,
propname, (unsigned long long)curval, setpoint);
else
(void) printf("%s %s = %s at '%s'\n",
osname, propname, valname, setpoint);
}
umem_free(setpoint, MAXPATHLEN);
return (error);
}
static int
ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value)
{
spa_t *spa = ztest_spa;
nvlist_t *props = NULL;
int error;
props = fnvlist_alloc();
fnvlist_add_uint64(props, zpool_prop_to_name(prop), value);
error = spa_prop_set(spa, props);
fnvlist_free(props);
if (error == ENOSPC) {
ztest_record_enospc(FTAG);
return (error);
}
ASSERT0(error);
return (error);
}
static int
ztest_dmu_objset_own(const char *name, dmu_objset_type_t type,
boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp)
{
int err;
char *cp = NULL;
char ddname[ZFS_MAX_DATASET_NAME_LEN];
strcpy(ddname, name);
cp = strchr(ddname, '@');
if (cp != NULL)
*cp = '\0';
err = dmu_objset_own(name, type, readonly, decrypt, tag, osp);
while (decrypt && err == EACCES) {
dsl_crypto_params_t *dcp;
nvlist_t *crypto_args = fnvlist_alloc();
fnvlist_add_uint8_array(crypto_args, "wkeydata",
(uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN);
VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
crypto_args, &dcp));
err = spa_keystore_load_wkey(ddname, dcp, B_FALSE);
/*
* Note: if there was an error loading, the wkey was not
* consumed, and needs to be freed.
*/
dsl_crypto_params_free(dcp, (err != 0));
fnvlist_free(crypto_args);
if (err == EINVAL) {
/*
* We couldn't load a key for this dataset so try
* the parent. This loop will eventually hit the
* encryption root since ztest only makes clones
* as children of their origin datasets.
*/
cp = strrchr(ddname, '/');
if (cp == NULL)
return (err);
*cp = '\0';
err = EACCES;
continue;
} else if (err != 0) {
break;
}
err = dmu_objset_own(name, type, readonly, decrypt, tag, osp);
break;
}
return (err);
}
static void
ztest_rll_init(rll_t *rll)
{
rll->rll_writer = NULL;
rll->rll_readers = 0;
mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL);
}
static void
ztest_rll_destroy(rll_t *rll)
{
ASSERT3P(rll->rll_writer, ==, NULL);
ASSERT0(rll->rll_readers);
mutex_destroy(&rll->rll_lock);
cv_destroy(&rll->rll_cv);
}
static void
ztest_rll_lock(rll_t *rll, rl_type_t type)
{
mutex_enter(&rll->rll_lock);
if (type == RL_READER) {
while (rll->rll_writer != NULL)
(void) cv_wait(&rll->rll_cv, &rll->rll_lock);
rll->rll_readers++;
} else {
while (rll->rll_writer != NULL || rll->rll_readers)
(void) cv_wait(&rll->rll_cv, &rll->rll_lock);
rll->rll_writer = curthread;
}
mutex_exit(&rll->rll_lock);
}
static void
ztest_rll_unlock(rll_t *rll)
{
mutex_enter(&rll->rll_lock);
if (rll->rll_writer) {
ASSERT0(rll->rll_readers);
rll->rll_writer = NULL;
} else {
ASSERT3S(rll->rll_readers, >, 0);
ASSERT3P(rll->rll_writer, ==, NULL);
rll->rll_readers--;
}
if (rll->rll_writer == NULL && rll->rll_readers == 0)
cv_broadcast(&rll->rll_cv);
mutex_exit(&rll->rll_lock);
}
static void
ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type)
{
rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)];
ztest_rll_lock(rll, type);
}
static void
ztest_object_unlock(ztest_ds_t *zd, uint64_t object)
{
rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)];
ztest_rll_unlock(rll);
}
static rl_t *
ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset,
uint64_t size, rl_type_t type)
{
uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1));
rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)];
rl_t *rl;
rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL);
rl->rl_object = object;
rl->rl_offset = offset;
rl->rl_size = size;
rl->rl_lock = rll;
ztest_rll_lock(rll, type);
return (rl);
}
static void
ztest_range_unlock(rl_t *rl)
{
rll_t *rll = rl->rl_lock;
ztest_rll_unlock(rll);
umem_free(rl, sizeof (*rl));
}
static void
ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os)
{
zd->zd_os = os;
zd->zd_zilog = dmu_objset_zil(os);
zd->zd_shared = szd;
dmu_objset_name(os, zd->zd_name);
int l;
if (zd->zd_shared != NULL)
zd->zd_shared->zd_seq = 0;
VERIFY0(pthread_rwlock_init(&zd->zd_zilog_lock, NULL));
mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL);
for (l = 0; l < ZTEST_OBJECT_LOCKS; l++)
ztest_rll_init(&zd->zd_object_lock[l]);
for (l = 0; l < ZTEST_RANGE_LOCKS; l++)
ztest_rll_init(&zd->zd_range_lock[l]);
}
static void
ztest_zd_fini(ztest_ds_t *zd)
{
int l;
mutex_destroy(&zd->zd_dirobj_lock);
(void) pthread_rwlock_destroy(&zd->zd_zilog_lock);
for (l = 0; l < ZTEST_OBJECT_LOCKS; l++)
ztest_rll_destroy(&zd->zd_object_lock[l]);
for (l = 0; l < ZTEST_RANGE_LOCKS; l++)
ztest_rll_destroy(&zd->zd_range_lock[l]);
}
#define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT)
static uint64_t
ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag)
{
uint64_t txg;
int error;
/*
* Attempt to assign tx to some transaction group.
*/
error = dmu_tx_assign(tx, txg_how);
if (error) {
if (error == ERESTART) {
ASSERT3U(txg_how, ==, TXG_NOWAIT);
dmu_tx_wait(tx);
} else {
ASSERT3U(error, ==, ENOSPC);
ztest_record_enospc(tag);
}
dmu_tx_abort(tx);
return (0);
}
txg = dmu_tx_get_txg(tx);
ASSERT3U(txg, !=, 0);
return (txg);
}
static void
ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object,
uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg,
uint64_t crtxg)
{
bt->bt_magic = BT_MAGIC;
bt->bt_objset = dmu_objset_id(os);
bt->bt_object = object;
bt->bt_dnodesize = dnodesize;
bt->bt_offset = offset;
bt->bt_gen = gen;
bt->bt_txg = txg;
bt->bt_crtxg = crtxg;
}
static void
ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object,
uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg,
uint64_t crtxg)
{
ASSERT3U(bt->bt_magic, ==, BT_MAGIC);
ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os));
ASSERT3U(bt->bt_object, ==, object);
ASSERT3U(bt->bt_dnodesize, ==, dnodesize);
ASSERT3U(bt->bt_offset, ==, offset);
ASSERT3U(bt->bt_gen, <=, gen);
ASSERT3U(bt->bt_txg, <=, txg);
ASSERT3U(bt->bt_crtxg, ==, crtxg);
}
static ztest_block_tag_t *
ztest_bt_bonus(dmu_buf_t *db)
{
dmu_object_info_t doi;
ztest_block_tag_t *bt;
dmu_object_info_from_db(db, &doi);
ASSERT3U(doi.doi_bonus_size, <=, db->db_size);
ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt));
bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt));
return (bt);
}
/*
* Generate a token to fill up unused bonus buffer space. Try to make
* it unique to the object, generation, and offset to verify that data
* is not getting overwritten by data from other dnodes.
*/
#define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \
(((ds) << 48) | ((gen) << 32) | ((obj) << 8) | (offset))
/*
* Fill up the unused bonus buffer region before the block tag with a
* verifiable pattern. Filling the whole bonus area with non-zero data
* helps ensure that all dnode traversal code properly skips the
* interior regions of large dnodes.
*/
static void
ztest_fill_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj,
objset_t *os, uint64_t gen)
{
uint64_t *bonusp;
ASSERT(IS_P2ALIGNED((char *)end - (char *)db->db_data, 8));
for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) {
uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os),
gen, bonusp - (uint64_t *)db->db_data);
*bonusp = token;
}
}
/*
* Verify that the unused area of a bonus buffer is filled with the
* expected tokens.
*/
static void
ztest_verify_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj,
objset_t *os, uint64_t gen)
{
uint64_t *bonusp;
for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) {
uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os),
gen, bonusp - (uint64_t *)db->db_data);
VERIFY3U(*bonusp, ==, token);
}
}
/*
* ZIL logging ops
*/
#define lrz_type lr_mode
#define lrz_blocksize lr_uid
#define lrz_ibshift lr_gid
#define lrz_bonustype lr_rdev
#define lrz_dnodesize lr_crtime[1]
static void
ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr)
{
char *name = (void *)(lr + 1); /* name follows lr */
size_t namesize = strlen(name) + 1;
itx_t *itx;
if (zil_replaying(zd->zd_zilog, tx))
return;
itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize);
bcopy(&lr->lr_common + 1, &itx->itx_lr + 1,
sizeof (*lr) + namesize - sizeof (lr_t));
zil_itx_assign(zd->zd_zilog, itx, tx);
}
static void
ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object)
{
char *name = (void *)(lr + 1); /* name follows lr */
size_t namesize = strlen(name) + 1;
itx_t *itx;
if (zil_replaying(zd->zd_zilog, tx))
return;
itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize);
bcopy(&lr->lr_common + 1, &itx->itx_lr + 1,
sizeof (*lr) + namesize - sizeof (lr_t));
itx->itx_oid = object;
zil_itx_assign(zd->zd_zilog, itx, tx);
}
static void
ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr)
{
itx_t *itx;
itx_wr_state_t write_state = ztest_random(WR_NUM_STATES);
if (zil_replaying(zd->zd_zilog, tx))
return;
if (lr->lr_length > zil_max_log_data(zd->zd_zilog))
write_state = WR_INDIRECT;
itx = zil_itx_create(TX_WRITE,
sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0));
if (write_state == WR_COPIED &&
dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length,
((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) {
zil_itx_destroy(itx);
itx = zil_itx_create(TX_WRITE, sizeof (*lr));
write_state = WR_NEED_COPY;
}
itx->itx_private = zd;
itx->itx_wr_state = write_state;
itx->itx_sync = (ztest_random(8) == 0);
bcopy(&lr->lr_common + 1, &itx->itx_lr + 1,
sizeof (*lr) - sizeof (lr_t));
zil_itx_assign(zd->zd_zilog, itx, tx);
}
static void
ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr)
{
itx_t *itx;
if (zil_replaying(zd->zd_zilog, tx))
return;
itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr));
bcopy(&lr->lr_common + 1, &itx->itx_lr + 1,
sizeof (*lr) - sizeof (lr_t));
itx->itx_sync = B_FALSE;
zil_itx_assign(zd->zd_zilog, itx, tx);
}
static void
ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr)
{
itx_t *itx;
if (zil_replaying(zd->zd_zilog, tx))
return;
itx = zil_itx_create(TX_SETATTR, sizeof (*lr));
bcopy(&lr->lr_common + 1, &itx->itx_lr + 1,
sizeof (*lr) - sizeof (lr_t));
itx->itx_sync = B_FALSE;
zil_itx_assign(zd->zd_zilog, itx, tx);
}
/*
* ZIL replay ops
*/
static int
ztest_replay_create(void *arg1, void *arg2, boolean_t byteswap)
{
ztest_ds_t *zd = arg1;
lr_create_t *lr = arg2;
char *name = (void *)(lr + 1); /* name follows lr */
objset_t *os = zd->zd_os;
ztest_block_tag_t *bbt;
dmu_buf_t *db;
dmu_tx_t *tx;
uint64_t txg;
int error = 0;
int bonuslen;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
ASSERT3U(lr->lr_doid, ==, ZTEST_DIROBJ);
ASSERT3S(name[0], !=, '\0');
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name);
if (lr->lrz_type == DMU_OT_ZAP_OTHER) {
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL);
} else {
dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
}
txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
if (txg == 0)
return (ENOSPC);
ASSERT3U(dmu_objset_zil(os)->zl_replay, ==, !!lr->lr_foid);
bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize);
if (lr->lrz_type == DMU_OT_ZAP_OTHER) {
if (lr->lr_foid == 0) {
lr->lr_foid = zap_create_dnsize(os,
lr->lrz_type, lr->lrz_bonustype,
bonuslen, lr->lrz_dnodesize, tx);
} else {
error = zap_create_claim_dnsize(os, lr->lr_foid,
lr->lrz_type, lr->lrz_bonustype,
bonuslen, lr->lrz_dnodesize, tx);
}
} else {
if (lr->lr_foid == 0) {
lr->lr_foid = dmu_object_alloc_dnsize(os,
lr->lrz_type, 0, lr->lrz_bonustype,
bonuslen, lr->lrz_dnodesize, tx);
} else {
error = dmu_object_claim_dnsize(os, lr->lr_foid,
lr->lrz_type, 0, lr->lrz_bonustype,
bonuslen, lr->lrz_dnodesize, tx);
}
}
if (error) {
ASSERT3U(error, ==, EEXIST);
ASSERT(zd->zd_zilog->zl_replay);
dmu_tx_commit(tx);
return (error);
}
ASSERT3U(lr->lr_foid, !=, 0);
if (lr->lrz_type != DMU_OT_ZAP_OTHER)
VERIFY0(dmu_object_set_blocksize(os, lr->lr_foid,
lr->lrz_blocksize, lr->lrz_ibshift, tx));
VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));
bbt = ztest_bt_bonus(db);
dmu_buf_will_dirty(db, tx);
ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL,
lr->lr_gen, txg, txg);
ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen);
dmu_buf_rele(db, FTAG);
VERIFY0(zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1,
&lr->lr_foid, tx));
(void) ztest_log_create(zd, tx, lr);
dmu_tx_commit(tx);
return (0);
}
static int
ztest_replay_remove(void *arg1, void *arg2, boolean_t byteswap)
{
ztest_ds_t *zd = arg1;
lr_remove_t *lr = arg2;
char *name = (void *)(lr + 1); /* name follows lr */
objset_t *os = zd->zd_os;
dmu_object_info_t doi;
dmu_tx_t *tx;
uint64_t object, txg;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
ASSERT3U(lr->lr_doid, ==, ZTEST_DIROBJ);
ASSERT3S(name[0], !=, '\0');
VERIFY0(
zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object));
ASSERT3U(object, !=, 0);
ztest_object_lock(zd, object, RL_WRITER);
VERIFY0(dmu_object_info(os, object, &doi));
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name);
dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END);
txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
if (txg == 0) {
ztest_object_unlock(zd, object);
return (ENOSPC);
}
if (doi.doi_type == DMU_OT_ZAP_OTHER) {
VERIFY0(zap_destroy(os, object, tx));
} else {
VERIFY0(dmu_object_free(os, object, tx));
}
VERIFY0(zap_remove(os, lr->lr_doid, name, tx));
(void) ztest_log_remove(zd, tx, lr, object);
dmu_tx_commit(tx);
ztest_object_unlock(zd, object);
return (0);
}
static int
ztest_replay_write(void *arg1, void *arg2, boolean_t byteswap)
{
ztest_ds_t *zd = arg1;
lr_write_t *lr = arg2;
objset_t *os = zd->zd_os;
void *data = lr + 1; /* data follows lr */
uint64_t offset, length;
ztest_block_tag_t *bt = data;
ztest_block_tag_t *bbt;
uint64_t gen, txg, lrtxg, crtxg;
dmu_object_info_t doi;
dmu_tx_t *tx;
dmu_buf_t *db;
arc_buf_t *abuf = NULL;
rl_t *rl;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
offset = lr->lr_offset;
length = lr->lr_length;
/* If it's a dmu_sync() block, write the whole block */
if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
if (length < blocksize) {
offset -= offset % blocksize;
length = blocksize;
}
}
if (bt->bt_magic == BSWAP_64(BT_MAGIC))
byteswap_uint64_array(bt, sizeof (*bt));
if (bt->bt_magic != BT_MAGIC)
bt = NULL;
ztest_object_lock(zd, lr->lr_foid, RL_READER);
rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER);
VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));
dmu_object_info_from_db(db, &doi);
bbt = ztest_bt_bonus(db);
ASSERT3U(bbt->bt_magic, ==, BT_MAGIC);
gen = bbt->bt_gen;
crtxg = bbt->bt_crtxg;
lrtxg = lr->lr_common.lrc_txg;
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, lr->lr_foid, offset, length);
if (ztest_random(8) == 0 && length == doi.doi_data_block_size &&
P2PHASE(offset, length) == 0)
abuf = dmu_request_arcbuf(db, length);
txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
if (txg == 0) {
if (abuf != NULL)
dmu_return_arcbuf(abuf);
dmu_buf_rele(db, FTAG);
ztest_range_unlock(rl);
ztest_object_unlock(zd, lr->lr_foid);
return (ENOSPC);
}
if (bt != NULL) {
/*
* Usually, verify the old data before writing new data --
* but not always, because we also want to verify correct
* behavior when the data was not recently read into cache.
*/
ASSERT0(offset % doi.doi_data_block_size);
if (ztest_random(4) != 0) {
int prefetch = ztest_random(2) ?
DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH;
ztest_block_tag_t rbt;
VERIFY(dmu_read(os, lr->lr_foid, offset,
sizeof (rbt), &rbt, prefetch) == 0);
if (rbt.bt_magic == BT_MAGIC) {
ztest_bt_verify(&rbt, os, lr->lr_foid, 0,
offset, gen, txg, crtxg);
}
}
/*
* Writes can appear to be newer than the bonus buffer because
* the ztest_get_data() callback does a dmu_read() of the
* open-context data, which may be different than the data
* as it was when the write was generated.
*/
if (zd->zd_zilog->zl_replay) {
ztest_bt_verify(bt, os, lr->lr_foid, 0, offset,
MAX(gen, bt->bt_gen), MAX(txg, lrtxg),
bt->bt_crtxg);
}
/*
* Set the bt's gen/txg to the bonus buffer's gen/txg
* so that all of the usual ASSERTs will work.
*/
ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg,
crtxg);
}
if (abuf == NULL) {
dmu_write(os, lr->lr_foid, offset, length, data, tx);
} else {
bcopy(data, abuf->b_data, length);
dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx);
}
(void) ztest_log_write(zd, tx, lr);
dmu_buf_rele(db, FTAG);
dmu_tx_commit(tx);
ztest_range_unlock(rl);
ztest_object_unlock(zd, lr->lr_foid);
return (0);
}
static int
ztest_replay_truncate(void *arg1, void *arg2, boolean_t byteswap)
{
ztest_ds_t *zd = arg1;
lr_truncate_t *lr = arg2;
objset_t *os = zd->zd_os;
dmu_tx_t *tx;
uint64_t txg;
rl_t *rl;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
ztest_object_lock(zd, lr->lr_foid, RL_READER);
rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length,
RL_WRITER);
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length);
txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
if (txg == 0) {
ztest_range_unlock(rl);
ztest_object_unlock(zd, lr->lr_foid);
return (ENOSPC);
}
VERIFY0(dmu_free_range(os, lr->lr_foid, lr->lr_offset,
lr->lr_length, tx));
(void) ztest_log_truncate(zd, tx, lr);
dmu_tx_commit(tx);
ztest_range_unlock(rl);
ztest_object_unlock(zd, lr->lr_foid);
return (0);
}
static int
ztest_replay_setattr(void *arg1, void *arg2, boolean_t byteswap)
{
ztest_ds_t *zd = arg1;
lr_setattr_t *lr = arg2;
objset_t *os = zd->zd_os;
dmu_tx_t *tx;
dmu_buf_t *db;
ztest_block_tag_t *bbt;
uint64_t txg, lrtxg, crtxg, dnodesize;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
ztest_object_lock(zd, lr->lr_foid, RL_WRITER);
VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));
tx = dmu_tx_create(os);
dmu_tx_hold_bonus(tx, lr->lr_foid);
txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
if (txg == 0) {
dmu_buf_rele(db, FTAG);
ztest_object_unlock(zd, lr->lr_foid);
return (ENOSPC);
}
bbt = ztest_bt_bonus(db);
ASSERT3U(bbt->bt_magic, ==, BT_MAGIC);
crtxg = bbt->bt_crtxg;
lrtxg = lr->lr_common.lrc_txg;
dnodesize = bbt->bt_dnodesize;
if (zd->zd_zilog->zl_replay) {
ASSERT3U(lr->lr_size, !=, 0);
ASSERT3U(lr->lr_mode, !=, 0);
ASSERT3U(lrtxg, !=, 0);
} else {
/*
* Randomly change the size and increment the generation.
*/
lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) *
sizeof (*bbt);
lr->lr_mode = bbt->bt_gen + 1;
ASSERT0(lrtxg);
}
/*
* Verify that the current bonus buffer is not newer than our txg.
*/
ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode,
MAX(txg, lrtxg), crtxg);
dmu_buf_will_dirty(db, tx);
ASSERT3U(lr->lr_size, >=, sizeof (*bbt));
ASSERT3U(lr->lr_size, <=, db->db_size);
VERIFY0(dmu_set_bonus(db, lr->lr_size, tx));
bbt = ztest_bt_bonus(db);
ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode,
txg, crtxg);
ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen);
dmu_buf_rele(db, FTAG);
(void) ztest_log_setattr(zd, tx, lr);
dmu_tx_commit(tx);
ztest_object_unlock(zd, lr->lr_foid);
return (0);
}
zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = {
NULL, /* 0 no such transaction type */
ztest_replay_create, /* TX_CREATE */
NULL, /* TX_MKDIR */
NULL, /* TX_MKXATTR */
NULL, /* TX_SYMLINK */
ztest_replay_remove, /* TX_REMOVE */
NULL, /* TX_RMDIR */
NULL, /* TX_LINK */
NULL, /* TX_RENAME */
ztest_replay_write, /* TX_WRITE */
ztest_replay_truncate, /* TX_TRUNCATE */
ztest_replay_setattr, /* TX_SETATTR */
NULL, /* TX_ACL */
NULL, /* TX_CREATE_ACL */
NULL, /* TX_CREATE_ATTR */
NULL, /* TX_CREATE_ACL_ATTR */
NULL, /* TX_MKDIR_ACL */
NULL, /* TX_MKDIR_ATTR */
NULL, /* TX_MKDIR_ACL_ATTR */
NULL, /* TX_WRITE2 */
};
/*
* ZIL get_data callbacks
*/
static void
ztest_get_done(zgd_t *zgd, int error)
{
(void) error;
ztest_ds_t *zd = zgd->zgd_private;
uint64_t object = ((rl_t *)zgd->zgd_lr)->rl_object;
if (zgd->zgd_db)
dmu_buf_rele(zgd->zgd_db, zgd);
ztest_range_unlock((rl_t *)zgd->zgd_lr);
ztest_object_unlock(zd, object);
umem_free(zgd, sizeof (*zgd));
}
static int
ztest_get_data(void *arg, uint64_t arg2, lr_write_t *lr, char *buf,
struct lwb *lwb, zio_t *zio)
{
(void) arg2;
ztest_ds_t *zd = arg;
objset_t *os = zd->zd_os;
uint64_t object = lr->lr_foid;
uint64_t offset = lr->lr_offset;
uint64_t size = lr->lr_length;
uint64_t txg = lr->lr_common.lrc_txg;
uint64_t crtxg;
dmu_object_info_t doi;
dmu_buf_t *db;
zgd_t *zgd;
int error;
ASSERT3P(lwb, !=, NULL);
ASSERT3P(zio, !=, NULL);
ASSERT3U(size, !=, 0);
ztest_object_lock(zd, object, RL_READER);
error = dmu_bonus_hold(os, object, FTAG, &db);
if (error) {
ztest_object_unlock(zd, object);
return (error);
}
crtxg = ztest_bt_bonus(db)->bt_crtxg;
if (crtxg == 0 || crtxg > txg) {
dmu_buf_rele(db, FTAG);
ztest_object_unlock(zd, object);
return (ENOENT);
}
dmu_object_info_from_db(db, &doi);
dmu_buf_rele(db, FTAG);
db = NULL;
zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL);
zgd->zgd_lwb = lwb;
zgd->zgd_private = zd;
if (buf != NULL) { /* immediate write */
zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd,
object, offset, size, RL_READER);
error = dmu_read(os, object, offset, size, buf,
DMU_READ_NO_PREFETCH);
ASSERT0(error);
} else {
size = doi.doi_data_block_size;
if (ISP2(size)) {
offset = P2ALIGN(offset, size);
} else {
ASSERT3U(offset, <, size);
offset = 0;
}
zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd,
object, offset, size, RL_READER);
error = dmu_buf_hold(os, object, offset, zgd, &db,
DMU_READ_NO_PREFETCH);
if (error == 0) {
blkptr_t *bp = &lr->lr_blkptr;
zgd->zgd_db = db;
zgd->zgd_bp = bp;
ASSERT3U(db->db_offset, ==, offset);
ASSERT3U(db->db_size, ==, size);
error = dmu_sync(zio, lr->lr_common.lrc_txg,
ztest_get_done, zgd);
if (error == 0)
return (0);
}
}
ztest_get_done(zgd, error);
return (error);
}
static void *
ztest_lr_alloc(size_t lrsize, char *name)
{
char *lr;
size_t namesize = name ? strlen(name) + 1 : 0;
lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL);
if (name)
bcopy(name, lr + lrsize, namesize);
return (lr);
}
static void
ztest_lr_free(void *lr, size_t lrsize, char *name)
{
size_t namesize = name ? strlen(name) + 1 : 0;
umem_free(lr, lrsize + namesize);
}
/*
* Lookup a bunch of objects. Returns the number of objects not found.
*/
static int
ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count)
{
int missing = 0;
int error;
int i;
ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock));
for (i = 0; i < count; i++, od++) {
od->od_object = 0;
error = zap_lookup(zd->zd_os, od->od_dir, od->od_name,
sizeof (uint64_t), 1, &od->od_object);
if (error) {
ASSERT3S(error, ==, ENOENT);
ASSERT0(od->od_object);
missing++;
} else {
dmu_buf_t *db;
ztest_block_tag_t *bbt;
dmu_object_info_t doi;
ASSERT3U(od->od_object, !=, 0);
ASSERT0(missing); /* there should be no gaps */
ztest_object_lock(zd, od->od_object, RL_READER);
VERIFY0(dmu_bonus_hold(zd->zd_os, od->od_object,
FTAG, &db));
dmu_object_info_from_db(db, &doi);
bbt = ztest_bt_bonus(db);
ASSERT3U(bbt->bt_magic, ==, BT_MAGIC);
od->od_type = doi.doi_type;
od->od_blocksize = doi.doi_data_block_size;
od->od_gen = bbt->bt_gen;
dmu_buf_rele(db, FTAG);
ztest_object_unlock(zd, od->od_object);
}
}
return (missing);
}
static int
ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count)
{
int missing = 0;
int i;
ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock));
for (i = 0; i < count; i++, od++) {
if (missing) {
od->od_object = 0;
missing++;
continue;
}
lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name);
lr->lr_doid = od->od_dir;
lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */
lr->lrz_type = od->od_crtype;
lr->lrz_blocksize = od->od_crblocksize;
lr->lrz_ibshift = ztest_random_ibshift();
lr->lrz_bonustype = DMU_OT_UINT64_OTHER;
lr->lrz_dnodesize = od->od_crdnodesize;
lr->lr_gen = od->od_crgen;
lr->lr_crtime[0] = time(NULL);
if (ztest_replay_create(zd, lr, B_FALSE) != 0) {
ASSERT0(missing);
od->od_object = 0;
missing++;
} else {
od->od_object = lr->lr_foid;
od->od_type = od->od_crtype;
od->od_blocksize = od->od_crblocksize;
od->od_gen = od->od_crgen;
ASSERT3U(od->od_object, !=, 0);
}
ztest_lr_free(lr, sizeof (*lr), od->od_name);
}
return (missing);
}
static int
ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count)
{
int missing = 0;
int error;
int i;
ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock));
od += count - 1;
for (i = count - 1; i >= 0; i--, od--) {
if (missing) {
missing++;
continue;
}
/*
* No object was found.
*/
if (od->od_object == 0)
continue;
lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name);
lr->lr_doid = od->od_dir;
if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) {
ASSERT3U(error, ==, ENOSPC);
missing++;
} else {
od->od_object = 0;
}
ztest_lr_free(lr, sizeof (*lr), od->od_name);
}
return (missing);
}
static int
ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size,
void *data)
{
lr_write_t *lr;
int error;
lr = ztest_lr_alloc(sizeof (*lr) + size, NULL);
lr->lr_foid = object;
lr->lr_offset = offset;
lr->lr_length = size;
lr->lr_blkoff = 0;
BP_ZERO(&lr->lr_blkptr);
bcopy(data, lr + 1, size);
error = ztest_replay_write(zd, lr, B_FALSE);
ztest_lr_free(lr, sizeof (*lr) + size, NULL);
return (error);
}
static int
ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size)
{
lr_truncate_t *lr;
int error;
lr = ztest_lr_alloc(sizeof (*lr), NULL);
lr->lr_foid = object;
lr->lr_offset = offset;
lr->lr_length = size;
error = ztest_replay_truncate(zd, lr, B_FALSE);
ztest_lr_free(lr, sizeof (*lr), NULL);
return (error);
}
static int
ztest_setattr(ztest_ds_t *zd, uint64_t object)
{
lr_setattr_t *lr;
int error;
lr = ztest_lr_alloc(sizeof (*lr), NULL);
lr->lr_foid = object;
lr->lr_size = 0;
lr->lr_mode = 0;
error = ztest_replay_setattr(zd, lr, B_FALSE);
ztest_lr_free(lr, sizeof (*lr), NULL);
return (error);
}
static void
ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size)
{
objset_t *os = zd->zd_os;
dmu_tx_t *tx;
uint64_t txg;
rl_t *rl;
txg_wait_synced(dmu_objset_pool(os), 0);
ztest_object_lock(zd, object, RL_READER);
rl = ztest_range_lock(zd, object, offset, size, RL_WRITER);
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, object, offset, size);
txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
if (txg != 0) {
dmu_prealloc(os, object, offset, size, tx);
dmu_tx_commit(tx);
txg_wait_synced(dmu_objset_pool(os), txg);
} else {
(void) dmu_free_long_range(os, object, offset, size);
}
ztest_range_unlock(rl);
ztest_object_unlock(zd, object);
}
static void
ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset)
{
int err;
ztest_block_tag_t wbt;
dmu_object_info_t doi;
enum ztest_io_type io_type;
uint64_t blocksize;
void *data;
VERIFY0(dmu_object_info(zd->zd_os, object, &doi));
blocksize = doi.doi_data_block_size;
data = umem_alloc(blocksize, UMEM_NOFAIL);
/*
* Pick an i/o type at random, biased toward writing block tags.
*/
io_type = ztest_random(ZTEST_IO_TYPES);
if (ztest_random(2) == 0)
io_type = ZTEST_IO_WRITE_TAG;
(void) pthread_rwlock_rdlock(&zd->zd_zilog_lock);
switch (io_type) {
case ZTEST_IO_WRITE_TAG:
ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize,
offset, 0, 0, 0);
(void) ztest_write(zd, object, offset, sizeof (wbt), &wbt);
break;
case ZTEST_IO_WRITE_PATTERN:
(void) memset(data, 'a' + (object + offset) % 5, blocksize);
if (ztest_random(2) == 0) {
/*
* Induce fletcher2 collisions to ensure that
* zio_ddt_collision() detects and resolves them
* when using fletcher2-verify for deduplication.
*/
((uint64_t *)data)[0] ^= 1ULL << 63;
((uint64_t *)data)[4] ^= 1ULL << 63;
}
(void) ztest_write(zd, object, offset, blocksize, data);
break;
case ZTEST_IO_WRITE_ZEROES:
bzero(data, blocksize);
(void) ztest_write(zd, object, offset, blocksize, data);
break;
case ZTEST_IO_TRUNCATE:
(void) ztest_truncate(zd, object, offset, blocksize);
break;
case ZTEST_IO_SETATTR:
(void) ztest_setattr(zd, object);
break;
default:
break;
case ZTEST_IO_REWRITE:
(void) pthread_rwlock_rdlock(&ztest_name_lock);
err = ztest_dsl_prop_set_uint64(zd->zd_name,
ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa),
B_FALSE);
VERIFY(err == 0 || err == ENOSPC);
err = ztest_dsl_prop_set_uint64(zd->zd_name,
ZFS_PROP_COMPRESSION,
ztest_random_dsl_prop(ZFS_PROP_COMPRESSION),
B_FALSE);
VERIFY(err == 0 || err == ENOSPC);
(void) pthread_rwlock_unlock(&ztest_name_lock);
VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data,
DMU_READ_NO_PREFETCH));
(void) ztest_write(zd, object, offset, blocksize, data);
break;
}
(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);
umem_free(data, blocksize);
}
/*
* Initialize an object description template.
*/
static void
ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index,
dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize,
uint64_t gen)
{
od->od_dir = ZTEST_DIROBJ;
od->od_object = 0;
od->od_crtype = type;
od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize();
od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize();
od->od_crgen = gen;
od->od_type = DMU_OT_NONE;
od->od_blocksize = 0;
od->od_gen = 0;
(void) snprintf(od->od_name, sizeof (od->od_name),
"%s(%"PRId64")[%"PRIu64"]",
tag, id, index);
}
/*
* Lookup or create the objects for a test using the od template.
* If the objects do not all exist, or if 'remove' is specified,
* remove any existing objects and create new ones. Otherwise,
* use the existing objects.
*/
static int
ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove)
{
int count = size / sizeof (*od);
int rv = 0;
mutex_enter(&zd->zd_dirobj_lock);
if ((ztest_lookup(zd, od, count) != 0 || remove) &&
(ztest_remove(zd, od, count) != 0 ||
ztest_create(zd, od, count) != 0))
rv = -1;
zd->zd_od = od;
mutex_exit(&zd->zd_dirobj_lock);
return (rv);
}
void
ztest_zil_commit(ztest_ds_t *zd, uint64_t id)
{
(void) id;
zilog_t *zilog = zd->zd_zilog;
(void) pthread_rwlock_rdlock(&zd->zd_zilog_lock);
zil_commit(zilog, ztest_random(ZTEST_OBJECTS));
/*
* Remember the committed values in zd, which is in parent/child
* shared memory. If we die, the next iteration of ztest_run()
* will verify that the log really does contain this record.
*/
mutex_enter(&zilog->zl_lock);
ASSERT3P(zd->zd_shared, !=, NULL);
ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq);
zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq;
mutex_exit(&zilog->zl_lock);
(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);
}
/*
* This function is designed to simulate the operations that occur during a
* mount/unmount operation. We hold the dataset across these operations in an
* attempt to expose any implicit assumptions about ZIL management.
*/
void
ztest_zil_remount(ztest_ds_t *zd, uint64_t id)
{
(void) id;
objset_t *os = zd->zd_os;
/*
* We hold the ztest_vdev_lock so we don't cause problems with
* other threads that wish to remove a log device, such as
* ztest_device_removal().
*/
mutex_enter(&ztest_vdev_lock);
/*
* We grab the zd_dirobj_lock to ensure that no other thread is
* updating the zil (i.e. adding in-memory log records) and the
* zd_zilog_lock to block any I/O.
*/
mutex_enter(&zd->zd_dirobj_lock);
(void) pthread_rwlock_wrlock(&zd->zd_zilog_lock);
/* zfsvfs_teardown() */
zil_close(zd->zd_zilog);
/* zfsvfs_setup() */
VERIFY3P(zil_open(os, ztest_get_data), ==, zd->zd_zilog);
zil_replay(os, zd, ztest_replay_vector);
(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);
mutex_exit(&zd->zd_dirobj_lock);
mutex_exit(&ztest_vdev_lock);
}
/*
* Verify that we can't destroy an active pool, create an existing pool,
* or create a pool with a bad vdev spec.
*/
void
ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
ztest_shared_opts_t *zo = &ztest_opts;
spa_t *spa;
nvlist_t *nvroot;
if (zo->zo_mmp_test)
return;
/*
* Attempt to create using a bad file.
*/
nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1);
VERIFY3U(ENOENT, ==,
spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL));
fnvlist_free(nvroot);
/*
* Attempt to create using a bad mirror.
*/
nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 2, 1);
VERIFY3U(ENOENT, ==,
spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL));
fnvlist_free(nvroot);
/*
* Attempt to create an existing pool. It shouldn't matter
* what's in the nvroot; we should fail with EEXIST.
*/
(void) pthread_rwlock_rdlock(&ztest_name_lock);
nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1);
VERIFY3U(EEXIST, ==,
spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL));
fnvlist_free(nvroot);
/*
* We open a reference to the spa and then we try to export it
* expecting one of the following errors:
*
* EBUSY
* Because of the reference we just opened.
*
* ZFS_ERR_EXPORT_IN_PROGRESS
* For the case that there is another ztest thread doing
* an export concurrently.
*/
VERIFY0(spa_open(zo->zo_pool, &spa, FTAG));
int error = spa_destroy(zo->zo_pool);
if (error != EBUSY && error != ZFS_ERR_EXPORT_IN_PROGRESS) {
fatal(B_FALSE, "spa_destroy(%s) returned unexpected value %d",
spa->spa_name, error);
}
spa_close(spa, FTAG);
(void) pthread_rwlock_unlock(&ztest_name_lock);
}
/*
* Start and then stop the MMP threads to ensure the startup and shutdown code
* works properly. Actual protection and property-related code tested via ZTS.
*/
void
ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
ztest_shared_opts_t *zo = &ztest_opts;
spa_t *spa = ztest_spa;
if (zo->zo_mmp_test)
return;
/*
* Since enabling MMP involves setting a property, it could not be done
* while the pool is suspended.
*/
if (spa_suspended(spa))
return;
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
mutex_enter(&spa->spa_props_lock);
zfs_multihost_fail_intervals = 0;
if (!spa_multihost(spa)) {
spa->spa_multihost = B_TRUE;
mmp_thread_start(spa);
}
mutex_exit(&spa->spa_props_lock);
spa_config_exit(spa, SCL_CONFIG, FTAG);
txg_wait_synced(spa_get_dsl(spa), 0);
mmp_signal_all_threads();
txg_wait_synced(spa_get_dsl(spa), 0);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
mutex_enter(&spa->spa_props_lock);
if (spa_multihost(spa)) {
mmp_thread_stop(spa);
spa->spa_multihost = B_FALSE;
}
mutex_exit(&spa->spa_props_lock);
spa_config_exit(spa, SCL_CONFIG, FTAG);
}
void
ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
spa_t *spa;
uint64_t initial_version = SPA_VERSION_INITIAL;
uint64_t version, newversion;
nvlist_t *nvroot, *props;
char *name;
if (ztest_opts.zo_mmp_test)
return;
/* dRAID added after feature flags, skip upgrade test. */
if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0)
return;
mutex_enter(&ztest_vdev_lock);
name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool);
/*
* Clean up from previous runs.
*/
(void) spa_destroy(name);
nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0,
NULL, ztest_opts.zo_raid_children, ztest_opts.zo_mirrors, 1);
/*
* If we're configuring a RAIDZ device then make sure that the
* initial version is capable of supporting that feature.
*/
switch (ztest_opts.zo_raid_parity) {
case 0:
case 1:
initial_version = SPA_VERSION_INITIAL;
break;
case 2:
initial_version = SPA_VERSION_RAIDZ2;
break;
case 3:
initial_version = SPA_VERSION_RAIDZ3;
break;
}
/*
* Create a pool with a spa version that can be upgraded. Pick
* a value between initial_version and SPA_VERSION_BEFORE_FEATURES.
*/
do {
version = ztest_random_spa_version(initial_version);
} while (version > SPA_VERSION_BEFORE_FEATURES);
props = fnvlist_alloc();
fnvlist_add_uint64(props,
zpool_prop_to_name(ZPOOL_PROP_VERSION), version);
VERIFY0(spa_create(name, nvroot, props, NULL, NULL));
fnvlist_free(nvroot);
fnvlist_free(props);
VERIFY0(spa_open(name, &spa, FTAG));
VERIFY3U(spa_version(spa), ==, version);
newversion = ztest_random_spa_version(version + 1);
if (ztest_opts.zo_verbose >= 4) {
(void) printf("upgrading spa version from "
"%"PRIu64" to %"PRIu64"\n",
version, newversion);
}
spa_upgrade(spa, newversion);
VERIFY3U(spa_version(spa), >, version);
VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config,
zpool_prop_to_name(ZPOOL_PROP_VERSION)));
spa_close(spa, FTAG);
kmem_strfree(name);
mutex_exit(&ztest_vdev_lock);
}
static void
ztest_spa_checkpoint(spa_t *spa)
{
ASSERT(MUTEX_HELD(&ztest_checkpoint_lock));
int error = spa_checkpoint(spa->spa_name);
switch (error) {
case 0:
case ZFS_ERR_DEVRM_IN_PROGRESS:
case ZFS_ERR_DISCARDING_CHECKPOINT:
case ZFS_ERR_CHECKPOINT_EXISTS:
break;
case ENOSPC:
ztest_record_enospc(FTAG);
break;
default:
fatal(B_FALSE, "spa_checkpoint(%s) = %d", spa->spa_name, error);
}
}
static void
ztest_spa_discard_checkpoint(spa_t *spa)
{
ASSERT(MUTEX_HELD(&ztest_checkpoint_lock));
int error = spa_checkpoint_discard(spa->spa_name);
switch (error) {
case 0:
case ZFS_ERR_DISCARDING_CHECKPOINT:
case ZFS_ERR_NO_CHECKPOINT:
break;
default:
fatal(B_FALSE, "spa_discard_checkpoint(%s) = %d",
spa->spa_name, error);
}
}
void
ztest_spa_checkpoint_create_discard(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
spa_t *spa = ztest_spa;
mutex_enter(&ztest_checkpoint_lock);
if (ztest_random(2) == 0) {
ztest_spa_checkpoint(spa);
} else {
ztest_spa_discard_checkpoint(spa);
}
mutex_exit(&ztest_checkpoint_lock);
}
static vdev_t *
vdev_lookup_by_path(vdev_t *vd, const char *path)
{
vdev_t *mvd;
int c;
if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0)
return (vd);
for (c = 0; c < vd->vdev_children; c++)
if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) !=
NULL)
return (mvd);
return (NULL);
}
static int
spa_num_top_vdevs(spa_t *spa)
{
vdev_t *rvd = spa->spa_root_vdev;
ASSERT3U(spa_config_held(spa, SCL_VDEV, RW_READER), ==, SCL_VDEV);
return (rvd->vdev_children);
}
/*
* Verify that vdev_add() works as expected.
*/
void
ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
ztest_shared_t *zs = ztest_shared;
spa_t *spa = ztest_spa;
uint64_t leaves;
uint64_t guid;
nvlist_t *nvroot;
int error;
if (ztest_opts.zo_mmp_test)
return;
mutex_enter(&ztest_vdev_lock);
leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) *
ztest_opts.zo_raid_children;
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves;
/*
* If we have slogs then remove them 1/4 of the time.
*/
if (spa_has_slogs(spa) && ztest_random(4) == 0) {
metaslab_group_t *mg;
/*
* find the first real slog in log allocation class
*/
mg = spa_log_class(spa)->mc_allocator[0].mca_rotor;
while (!mg->mg_vd->vdev_islog)
mg = mg->mg_next;
guid = mg->mg_vd->vdev_guid;
spa_config_exit(spa, SCL_VDEV, FTAG);
/*
* We have to grab the zs_name_lock as writer to
* prevent a race between removing a slog (dmu_objset_find)
* and destroying a dataset. Removing the slog will
* grab a reference on the dataset which may cause
* dsl_destroy_head() to fail with EBUSY thus
* leaving the dataset in an inconsistent state.
*/
pthread_rwlock_wrlock(&ztest_name_lock);
error = spa_vdev_remove(spa, guid, B_FALSE);
pthread_rwlock_unlock(&ztest_name_lock);
switch (error) {
case 0:
case EEXIST: /* Generic zil_reset() error */
case EBUSY: /* Replay required */
case EACCES: /* Crypto key not loaded */
case ZFS_ERR_CHECKPOINT_EXISTS:
case ZFS_ERR_DISCARDING_CHECKPOINT:
break;
default:
fatal(B_FALSE, "spa_vdev_remove() = %d", error);
}
} else {
spa_config_exit(spa, SCL_VDEV, FTAG);
/*
* Make 1/4 of the devices be log devices
*/
nvroot = make_vdev_root(NULL, NULL, NULL,
ztest_opts.zo_vdev_size, 0, (ztest_random(4) == 0) ?
"log" : NULL, ztest_opts.zo_raid_children, zs->zs_mirrors,
1);
error = spa_vdev_add(spa, nvroot);
fnvlist_free(nvroot);
switch (error) {
case 0:
break;
case ENOSPC:
ztest_record_enospc("spa_vdev_add");
break;
default:
fatal(B_FALSE, "spa_vdev_add() = %d", error);
}
}
mutex_exit(&ztest_vdev_lock);
}
void
ztest_vdev_class_add(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
ztest_shared_t *zs = ztest_shared;
spa_t *spa = ztest_spa;
uint64_t leaves;
nvlist_t *nvroot;
const char *class = (ztest_random(2) == 0) ?
VDEV_ALLOC_BIAS_SPECIAL : VDEV_ALLOC_BIAS_DEDUP;
int error;
/*
* By default add a special vdev 50% of the time
*/
if ((ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_OFF) ||
(ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_RND &&
ztest_random(2) == 0)) {
return;
}
mutex_enter(&ztest_vdev_lock);
/* Only test with mirrors */
if (zs->zs_mirrors < 2) {
mutex_exit(&ztest_vdev_lock);
return;
}
/* requires feature@allocation_classes */
if (!spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)) {
mutex_exit(&ztest_vdev_lock);
return;
}
leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) *
ztest_opts.zo_raid_children;
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves;
spa_config_exit(spa, SCL_VDEV, FTAG);
nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0,
class, ztest_opts.zo_raid_children, zs->zs_mirrors, 1);
error = spa_vdev_add(spa, nvroot);
fnvlist_free(nvroot);
if (error == ENOSPC)
ztest_record_enospc("spa_vdev_add");
else if (error != 0)
fatal(B_FALSE, "spa_vdev_add() = %d", error);
/*
* 50% of the time allow small blocks in the special class
*/
if (error == 0 &&
spa_special_class(spa)->mc_groups == 1 && ztest_random(2) == 0) {
if (ztest_opts.zo_verbose >= 3)
(void) printf("Enabling special VDEV small blocks\n");
(void) ztest_dsl_prop_set_uint64(zd->zd_name,
ZFS_PROP_SPECIAL_SMALL_BLOCKS, 32768, B_FALSE);
}
mutex_exit(&ztest_vdev_lock);
if (ztest_opts.zo_verbose >= 3) {
metaslab_class_t *mc;
if (strcmp(class, VDEV_ALLOC_BIAS_SPECIAL) == 0)
mc = spa_special_class(spa);
else
mc = spa_dedup_class(spa);
(void) printf("Added a %s mirrored vdev (of %d)\n",
class, (int)mc->mc_groups);
}
}
/*
* Verify that adding/removing aux devices (l2arc, hot spare) works as expected.
*/
void
ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
ztest_shared_t *zs = ztest_shared;
spa_t *spa = ztest_spa;
vdev_t *rvd = spa->spa_root_vdev;
spa_aux_vdev_t *sav;
char *aux;
char *path;
uint64_t guid = 0;
int error, ignore_err = 0;
if (ztest_opts.zo_mmp_test)
return;
path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
if (ztest_random(2) == 0) {
sav = &spa->spa_spares;
aux = ZPOOL_CONFIG_SPARES;
} else {
sav = &spa->spa_l2cache;
aux = ZPOOL_CONFIG_L2CACHE;
}
mutex_enter(&ztest_vdev_lock);
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
if (sav->sav_count != 0 && ztest_random(4) == 0) {
/*
* Pick a random device to remove.
*/
vdev_t *svd = sav->sav_vdevs[ztest_random(sav->sav_count)];
/* dRAID spares cannot be removed; try anyways to see ENOTSUP */
if (strstr(svd->vdev_path, VDEV_TYPE_DRAID) != NULL)
ignore_err = ENOTSUP;
guid = svd->vdev_guid;
} else {
/*
* Find an unused device we can add.
*/
zs->zs_vdev_aux = 0;
for (;;) {
int c;
(void) snprintf(path, MAXPATHLEN, ztest_aux_template,
ztest_opts.zo_dir, ztest_opts.zo_pool, aux,
zs->zs_vdev_aux);
for (c = 0; c < sav->sav_count; c++)
if (strcmp(sav->sav_vdevs[c]->vdev_path,
path) == 0)
break;
if (c == sav->sav_count &&
vdev_lookup_by_path(rvd, path) == NULL)
break;
zs->zs_vdev_aux++;
}
}
spa_config_exit(spa, SCL_VDEV, FTAG);
if (guid == 0) {
/*
* Add a new device.
*/
nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL,
(ztest_opts.zo_vdev_size * 5) / 4, 0, NULL, 0, 0, 1);
error = spa_vdev_add(spa, nvroot);
switch (error) {
case 0:
break;
default:
fatal(B_FALSE, "spa_vdev_add(%p) = %d", nvroot, error);
}
fnvlist_free(nvroot);
} else {
/*
* Remove an existing device. Sometimes, dirty its
* vdev state first to make sure we handle removal
* of devices that have pending state changes.
*/
if (ztest_random(2) == 0)
(void) vdev_online(spa, guid, 0, NULL);
error = spa_vdev_remove(spa, guid, B_FALSE);
switch (error) {
case 0:
case EBUSY:
case ZFS_ERR_CHECKPOINT_EXISTS:
case ZFS_ERR_DISCARDING_CHECKPOINT:
break;
default:
if (error != ignore_err)
fatal(B_FALSE,
"spa_vdev_remove(%"PRIu64") = %d",
guid, error);
}
}
mutex_exit(&ztest_vdev_lock);
umem_free(path, MAXPATHLEN);
}
/*
* split a pool if it has mirror tlvdevs
*/
void
ztest_split_pool(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
ztest_shared_t *zs = ztest_shared;
spa_t *spa = ztest_spa;
vdev_t *rvd = spa->spa_root_vdev;
nvlist_t *tree, **child, *config, *split, **schild;
uint_t c, children, schildren = 0, lastlogid = 0;
int error = 0;
if (ztest_opts.zo_mmp_test)
return;
mutex_enter(&ztest_vdev_lock);
/* ensure we have a usable config; mirrors of raidz aren't supported */
if (zs->zs_mirrors < 3 || ztest_opts.zo_raid_children > 1) {
mutex_exit(&ztest_vdev_lock);
return;
}
/* clean up the old pool, if any */
(void) spa_destroy("splitp");
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
/* generate a config from the existing config */
mutex_enter(&spa->spa_props_lock);
tree = fnvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE);
mutex_exit(&spa->spa_props_lock);
VERIFY0(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN,
&child, &children));
schild = malloc(rvd->vdev_children * sizeof (nvlist_t *));
for (c = 0; c < children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
nvlist_t **mchild;
uint_t mchildren;
if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) {
schild[schildren] = fnvlist_alloc();
fnvlist_add_string(schild[schildren],
ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE);
fnvlist_add_uint64(schild[schildren],
ZPOOL_CONFIG_IS_HOLE, 1);
if (lastlogid == 0)
lastlogid = schildren;
++schildren;
continue;
}
lastlogid = 0;
VERIFY0(nvlist_lookup_nvlist_array(child[c],
ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren));
schild[schildren++] = fnvlist_dup(mchild[0]);
}
/* OK, create a config that can be used to split */
split = fnvlist_alloc();
fnvlist_add_string(split, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT);
fnvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN,
(const nvlist_t **)schild, lastlogid != 0 ? lastlogid : schildren);
config = fnvlist_alloc();
fnvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split);
for (c = 0; c < schildren; c++)
fnvlist_free(schild[c]);
free(schild);
fnvlist_free(split);
spa_config_exit(spa, SCL_VDEV, FTAG);
(void) pthread_rwlock_wrlock(&ztest_name_lock);
error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE);
(void) pthread_rwlock_unlock(&ztest_name_lock);
fnvlist_free(config);
if (error == 0) {
(void) printf("successful split - results:\n");
mutex_enter(&spa_namespace_lock);
show_pool_stats(spa);
show_pool_stats(spa_lookup("splitp"));
mutex_exit(&spa_namespace_lock);
++zs->zs_splits;
--zs->zs_mirrors;
}
mutex_exit(&ztest_vdev_lock);
}
/*
* Verify that we can attach and detach devices.
*/
void
ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
ztest_shared_t *zs = ztest_shared;
spa_t *spa = ztest_spa;
spa_aux_vdev_t *sav = &spa->spa_spares;
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *oldvd, *newvd, *pvd;
nvlist_t *root;
uint64_t leaves;
uint64_t leaf, top;
uint64_t ashift = ztest_get_ashift();
uint64_t oldguid, pguid;
uint64_t oldsize, newsize;
char *oldpath, *newpath;
int replacing;
int oldvd_has_siblings = B_FALSE;
int newvd_is_spare = B_FALSE;
int newvd_is_dspare = B_FALSE;
int oldvd_is_log;
int error, expected_error;
if (ztest_opts.zo_mmp_test)
return;
oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
mutex_enter(&ztest_vdev_lock);
leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raid_children;
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
/*
* If a vdev is in the process of being removed, its removal may
* finish while we are in progress, leading to an unexpected error
* value. Don't bother trying to attach while we are in the middle
* of removal.
*/
if (ztest_device_removal_active) {
spa_config_exit(spa, SCL_ALL, FTAG);
goto out;
}
/*
* Decide whether to do an attach or a replace.
*/
replacing = ztest_random(2);
/*
* Pick a random top-level vdev.
*/
top = ztest_random_vdev_top(spa, B_TRUE);
/*
* Pick a random leaf within it.
*/
leaf = ztest_random(leaves);
/*
* Locate this vdev.
*/
oldvd = rvd->vdev_child[top];
/* pick a child from the mirror */
if (zs->zs_mirrors >= 1) {
ASSERT3P(oldvd->vdev_ops, ==, &vdev_mirror_ops);
ASSERT3U(oldvd->vdev_children, >=, zs->zs_mirrors);
oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raid_children];
}
/* pick a child out of the raidz group */
if (ztest_opts.zo_raid_children > 1) {
if (strcmp(oldvd->vdev_ops->vdev_op_type, "raidz") == 0)
ASSERT3P(oldvd->vdev_ops, ==, &vdev_raidz_ops);
else
ASSERT3P(oldvd->vdev_ops, ==, &vdev_draid_ops);
ASSERT3U(oldvd->vdev_children, ==, ztest_opts.zo_raid_children);
oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raid_children];
}
/*
* If we're already doing an attach or replace, oldvd may be a
* mirror vdev -- in which case, pick a random child.
*/
while (oldvd->vdev_children != 0) {
oldvd_has_siblings = B_TRUE;
ASSERT3U(oldvd->vdev_children, >=, 2);
oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)];
}
oldguid = oldvd->vdev_guid;
oldsize = vdev_get_min_asize(oldvd);
oldvd_is_log = oldvd->vdev_top->vdev_islog;
(void) strcpy(oldpath, oldvd->vdev_path);
pvd = oldvd->vdev_parent;
pguid = pvd->vdev_guid;
/*
* If oldvd has siblings, then half of the time, detach it. Prior
* to the detach the pool is scrubbed in order to prevent creating
* unrepairable blocks as a result of the data corruption injection.
*/
if (oldvd_has_siblings && ztest_random(2) == 0) {
spa_config_exit(spa, SCL_ALL, FTAG);
error = ztest_scrub_impl(spa);
if (error)
goto out;
error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE);
if (error != 0 && error != ENODEV && error != EBUSY &&
error != ENOTSUP && error != ZFS_ERR_CHECKPOINT_EXISTS &&
error != ZFS_ERR_DISCARDING_CHECKPOINT)
fatal(B_FALSE, "detach (%s) returned %d",
oldpath, error);
goto out;
}
/*
* For the new vdev, choose with equal probability between the two
* standard paths (ending in either 'a' or 'b') or a random hot spare.
*/
if (sav->sav_count != 0 && ztest_random(3) == 0) {
newvd = sav->sav_vdevs[ztest_random(sav->sav_count)];
newvd_is_spare = B_TRUE;
if (newvd->vdev_ops == &vdev_draid_spare_ops)
newvd_is_dspare = B_TRUE;
(void) strcpy(newpath, newvd->vdev_path);
} else {
(void) snprintf(newpath, MAXPATHLEN, ztest_dev_template,
ztest_opts.zo_dir, ztest_opts.zo_pool,
top * leaves + leaf);
if (ztest_random(2) == 0)
newpath[strlen(newpath) - 1] = 'b';
newvd = vdev_lookup_by_path(rvd, newpath);
}
if (newvd) {
/*
* Reopen to ensure the vdev's asize field isn't stale.
*/
vdev_reopen(newvd);
newsize = vdev_get_min_asize(newvd);
} else {
/*
* Make newsize a little bigger or smaller than oldsize.
* If it's smaller, the attach should fail.
* If it's larger, and we're doing a replace,
* we should get dynamic LUN growth when we're done.
*/
newsize = 10 * oldsize / (9 + ztest_random(3));
}
/*
* If pvd is not a mirror or root, the attach should fail with ENOTSUP,
* unless it's a replace; in that case any non-replacing parent is OK.
*
* If newvd is already part of the pool, it should fail with EBUSY.
*
* If newvd is too small, it should fail with EOVERFLOW.
*
* If newvd is a distributed spare and it's being attached to a
* dRAID which is not its parent it should fail with EINVAL.
*/
if (pvd->vdev_ops != &vdev_mirror_ops &&
pvd->vdev_ops != &vdev_root_ops && (!replacing ||
pvd->vdev_ops == &vdev_replacing_ops ||
pvd->vdev_ops == &vdev_spare_ops))
expected_error = ENOTSUP;
else if (newvd_is_spare && (!replacing || oldvd_is_log))
expected_error = ENOTSUP;
else if (newvd == oldvd)
expected_error = replacing ? 0 : EBUSY;
else if (vdev_lookup_by_path(rvd, newpath) != NULL)
expected_error = EBUSY;
else if (!newvd_is_dspare && newsize < oldsize)
expected_error = EOVERFLOW;
else if (ashift > oldvd->vdev_top->vdev_ashift)
expected_error = EDOM;
else if (newvd_is_dspare && pvd != vdev_draid_spare_get_parent(newvd))
expected_error = ENOTSUP;
else
expected_error = 0;
spa_config_exit(spa, SCL_ALL, FTAG);
/*
* Build the nvlist describing newpath.
*/
root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0,
ashift, NULL, 0, 0, 1);
/*
* When supported select either a healing or sequential resilver.
*/
boolean_t rebuilding = B_FALSE;
if (pvd->vdev_ops == &vdev_mirror_ops ||
pvd->vdev_ops == &vdev_root_ops) {
rebuilding = !!ztest_random(2);
}
error = spa_vdev_attach(spa, oldguid, root, replacing, rebuilding);
fnvlist_free(root);
/*
* If our parent was the replacing vdev, but the replace completed,
* then instead of failing with ENOTSUP we may either succeed,
* fail with ENODEV, or fail with EOVERFLOW.
*/
if (expected_error == ENOTSUP &&
(error == 0 || error == ENODEV || error == EOVERFLOW))
expected_error = error;
/*
* If someone grew the LUN, the replacement may be too small.
*/
if (error == EOVERFLOW || error == EBUSY)
expected_error = error;
if (error == ZFS_ERR_CHECKPOINT_EXISTS ||
error == ZFS_ERR_DISCARDING_CHECKPOINT ||
error == ZFS_ERR_RESILVER_IN_PROGRESS ||
error == ZFS_ERR_REBUILD_IN_PROGRESS)
expected_error = error;
if (error != expected_error && expected_error != EBUSY) {
fatal(B_FALSE, "attach (%s %"PRIu64", %s %"PRIu64", %d) "
"returned %d, expected %d",
oldpath, oldsize, newpath,
newsize, replacing, error, expected_error);
}
out:
mutex_exit(&ztest_vdev_lock);
umem_free(oldpath, MAXPATHLEN);
umem_free(newpath, MAXPATHLEN);
}
void
ztest_device_removal(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
spa_t *spa = ztest_spa;
vdev_t *vd;
uint64_t guid;
int error;
mutex_enter(&ztest_vdev_lock);
if (ztest_device_removal_active) {
mutex_exit(&ztest_vdev_lock);
return;
}
/*
* Remove a random top-level vdev and wait for removal to finish.
*/
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
vd = vdev_lookup_top(spa, ztest_random_vdev_top(spa, B_FALSE));
guid = vd->vdev_guid;
spa_config_exit(spa, SCL_VDEV, FTAG);
error = spa_vdev_remove(spa, guid, B_FALSE);
if (error == 0) {
ztest_device_removal_active = B_TRUE;
mutex_exit(&ztest_vdev_lock);
/*
* spa->spa_vdev_removal is created in a sync task that
* is initiated via dsl_sync_task_nowait(). Since the
* task may not run before spa_vdev_remove() returns, we
* must wait at least 1 txg to ensure that the removal
* struct has been created.
*/
txg_wait_synced(spa_get_dsl(spa), 0);
while (spa->spa_removing_phys.sr_state == DSS_SCANNING)
txg_wait_synced(spa_get_dsl(spa), 0);
} else {
mutex_exit(&ztest_vdev_lock);
return;
}
/*
* The pool needs to be scrubbed after completing device removal.
* Failure to do so may result in checksum errors due to the
* strategy employed by ztest_fault_inject() when selecting which
* offset are redundant and can be damaged.
*/
error = spa_scan(spa, POOL_SCAN_SCRUB);
if (error == 0) {
while (dsl_scan_scrubbing(spa_get_dsl(spa)))
txg_wait_synced(spa_get_dsl(spa), 0);
}
mutex_enter(&ztest_vdev_lock);
ztest_device_removal_active = B_FALSE;
mutex_exit(&ztest_vdev_lock);
}
/*
* Callback function which expands the physical size of the vdev.
*/
static vdev_t *
grow_vdev(vdev_t *vd, void *arg)
{
spa_t *spa __maybe_unused = vd->vdev_spa;
size_t *newsize = arg;
size_t fsize;
int fd;
ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), ==, SCL_STATE);
ASSERT(vd->vdev_ops->vdev_op_leaf);
if ((fd = open(vd->vdev_path, O_RDWR)) == -1)
return (vd);
fsize = lseek(fd, 0, SEEK_END);
VERIFY0(ftruncate(fd, *newsize));
if (ztest_opts.zo_verbose >= 6) {
(void) printf("%s grew from %lu to %lu bytes\n",
vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize);
}
(void) close(fd);
return (NULL);
}
/*
* Callback function which expands a given vdev by calling vdev_online().
*/
static vdev_t *
online_vdev(vdev_t *vd, void *arg)
{
(void) arg;
spa_t *spa = vd->vdev_spa;
vdev_t *tvd = vd->vdev_top;
uint64_t guid = vd->vdev_guid;
uint64_t generation = spa->spa_config_generation + 1;
vdev_state_t newstate = VDEV_STATE_UNKNOWN;
int error;
ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), ==, SCL_STATE);
ASSERT(vd->vdev_ops->vdev_op_leaf);
/* Calling vdev_online will initialize the new metaslabs */
spa_config_exit(spa, SCL_STATE, spa);
error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate);
spa_config_enter(spa, SCL_STATE, spa, RW_READER);
/*
* If vdev_online returned an error or the underlying vdev_open
* failed then we abort the expand. The only way to know that
* vdev_open fails is by checking the returned newstate.
*/
if (error || newstate != VDEV_STATE_HEALTHY) {
if (ztest_opts.zo_verbose >= 5) {
(void) printf("Unable to expand vdev, state %u, "
"error %d\n", newstate, error);
}
return (vd);
}
ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY);
/*
* Since we dropped the lock we need to ensure that we're
* still talking to the original vdev. It's possible this
* vdev may have been detached/replaced while we were
* trying to online it.
*/
if (generation != spa->spa_config_generation) {
if (ztest_opts.zo_verbose >= 5) {
(void) printf("vdev configuration has changed, "
"guid %"PRIu64", state %"PRIu64", "
"expected gen %"PRIu64", got gen %"PRIu64"\n",
guid,
tvd->vdev_state,
generation,
spa->spa_config_generation);
}
return (vd);
}
return (NULL);
}
/*
* Traverse the vdev tree calling the supplied function.
* We continue to walk the tree until we either have walked all
* children or we receive a non-NULL return from the callback.
* If a NULL callback is passed, then we just return back the first
* leaf vdev we encounter.
*/
static vdev_t *
vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg)
{
uint_t c;
if (vd->vdev_ops->vdev_op_leaf) {
if (func == NULL)
return (vd);
else
return (func(vd, arg));
}
for (c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL)
return (cvd);
}
return (NULL);
}
/*
* Verify that dynamic LUN growth works as expected.
*/
void
ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
spa_t *spa = ztest_spa;
vdev_t *vd, *tvd;
metaslab_class_t *mc;
metaslab_group_t *mg;
size_t psize, newsize;
uint64_t top;
uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count;
mutex_enter(&ztest_checkpoint_lock);
mutex_enter(&ztest_vdev_lock);
spa_config_enter(spa, SCL_STATE, spa, RW_READER);
/*
* If there is a vdev removal in progress, it could complete while
* we are running, in which case we would not be able to verify
* that the metaslab_class space increased (because it decreases
* when the device removal completes).
*/
if (ztest_device_removal_active) {
spa_config_exit(spa, SCL_STATE, spa);
mutex_exit(&ztest_vdev_lock);
mutex_exit(&ztest_checkpoint_lock);
return;
}
top = ztest_random_vdev_top(spa, B_TRUE);
tvd = spa->spa_root_vdev->vdev_child[top];
mg = tvd->vdev_mg;
mc = mg->mg_class;
old_ms_count = tvd->vdev_ms_count;
old_class_space = metaslab_class_get_space(mc);
/*
* Determine the size of the first leaf vdev associated with
* our top-level device.
*/
vd = vdev_walk_tree(tvd, NULL, NULL);
ASSERT3P(vd, !=, NULL);
ASSERT(vd->vdev_ops->vdev_op_leaf);
psize = vd->vdev_psize;
/*
* We only try to expand the vdev if it's healthy, less than 4x its
* original size, and it has a valid psize.
*/
if (tvd->vdev_state != VDEV_STATE_HEALTHY ||
psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) {
spa_config_exit(spa, SCL_STATE, spa);
mutex_exit(&ztest_vdev_lock);
mutex_exit(&ztest_checkpoint_lock);
return;
}
ASSERT3U(psize, >, 0);
newsize = psize + MAX(psize / 8, SPA_MAXBLOCKSIZE);
ASSERT3U(newsize, >, psize);
if (ztest_opts.zo_verbose >= 6) {
(void) printf("Expanding LUN %s from %lu to %lu\n",
vd->vdev_path, (ulong_t)psize, (ulong_t)newsize);
}
/*
* Growing the vdev is a two step process:
* 1). expand the physical size (i.e. relabel)
* 2). online the vdev to create the new metaslabs
*/
if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL ||
vdev_walk_tree(tvd, online_vdev, NULL) != NULL ||
tvd->vdev_state != VDEV_STATE_HEALTHY) {
if (ztest_opts.zo_verbose >= 5) {
(void) printf("Could not expand LUN because "
"the vdev configuration changed.\n");
}
spa_config_exit(spa, SCL_STATE, spa);
mutex_exit(&ztest_vdev_lock);
mutex_exit(&ztest_checkpoint_lock);
return;
}
spa_config_exit(spa, SCL_STATE, spa);
/*
* Expanding the LUN will update the config asynchronously,
* thus we must wait for the async thread to complete any
* pending tasks before proceeding.
*/
for (;;) {
boolean_t done;
mutex_enter(&spa->spa_async_lock);
done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks);
mutex_exit(&spa->spa_async_lock);
if (done)
break;
txg_wait_synced(spa_get_dsl(spa), 0);
(void) poll(NULL, 0, 100);
}
spa_config_enter(spa, SCL_STATE, spa, RW_READER);
tvd = spa->spa_root_vdev->vdev_child[top];
new_ms_count = tvd->vdev_ms_count;
new_class_space = metaslab_class_get_space(mc);
if (tvd->vdev_mg != mg || mg->mg_class != mc) {
if (ztest_opts.zo_verbose >= 5) {
(void) printf("Could not verify LUN expansion due to "
"intervening vdev offline or remove.\n");
}
spa_config_exit(spa, SCL_STATE, spa);
mutex_exit(&ztest_vdev_lock);
mutex_exit(&ztest_checkpoint_lock);
return;
}
/*
* Make sure we were able to grow the vdev.
*/
if (new_ms_count <= old_ms_count) {
fatal(B_FALSE,
"LUN expansion failed: ms_count %"PRIu64" < %"PRIu64"\n",
old_ms_count, new_ms_count);
}
/*
* Make sure we were able to grow the pool.
*/
if (new_class_space <= old_class_space) {
fatal(B_FALSE,
"LUN expansion failed: class_space %"PRIu64" < %"PRIu64"\n",
old_class_space, new_class_space);
}
if (ztest_opts.zo_verbose >= 5) {
char oldnumbuf[NN_NUMBUF_SZ], newnumbuf[NN_NUMBUF_SZ];
nicenum(old_class_space, oldnumbuf, sizeof (oldnumbuf));
nicenum(new_class_space, newnumbuf, sizeof (newnumbuf));
(void) printf("%s grew from %s to %s\n",
spa->spa_name, oldnumbuf, newnumbuf);
}
spa_config_exit(spa, SCL_STATE, spa);
mutex_exit(&ztest_vdev_lock);
mutex_exit(&ztest_checkpoint_lock);
}
/*
* Verify that dmu_objset_{create,destroy,open,close} work as expected.
*/
static void
ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
{
(void) arg, (void) cr;
/*
* Create the objects common to all ztest datasets.
*/
VERIFY0(zap_create_claim(os, ZTEST_DIROBJ,
DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx));
}
static int
ztest_dataset_create(char *dsname)
{
int err;
uint64_t rand;
dsl_crypto_params_t *dcp = NULL;
/*
* 50% of the time, we create encrypted datasets
* using a random cipher suite and a hard-coded
* wrapping key.
*/
rand = ztest_random(2);
if (rand != 0) {
nvlist_t *crypto_args = fnvlist_alloc();
nvlist_t *props = fnvlist_alloc();
/* slight bias towards the default cipher suite */
rand = ztest_random(ZIO_CRYPT_FUNCTIONS);
if (rand < ZIO_CRYPT_AES_128_CCM)
rand = ZIO_CRYPT_ON;
fnvlist_add_uint64(props,
zfs_prop_to_name(ZFS_PROP_ENCRYPTION), rand);
fnvlist_add_uint8_array(crypto_args, "wkeydata",
(uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN);
/*
* These parameters aren't really used by the kernel. They
* are simply stored so that userspace knows how to load
* the wrapping key.
*/
fnvlist_add_uint64(props,
zfs_prop_to_name(ZFS_PROP_KEYFORMAT), ZFS_KEYFORMAT_RAW);
fnvlist_add_string(props,
zfs_prop_to_name(ZFS_PROP_KEYLOCATION), "prompt");
fnvlist_add_uint64(props,
zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), 0ULL);
fnvlist_add_uint64(props,
zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 0ULL);
VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, props,
crypto_args, &dcp));
/*
* Cycle through all available encryption implementations
* to verify interoperability.
*/
VERIFY0(gcm_impl_set("cycle"));
VERIFY0(aes_impl_set("cycle"));
fnvlist_free(crypto_args);
fnvlist_free(props);
}
err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, dcp,
ztest_objset_create_cb, NULL);
dsl_crypto_params_free(dcp, !!err);
rand = ztest_random(100);
if (err || rand < 80)
return (err);
if (ztest_opts.zo_verbose >= 5)
(void) printf("Setting dataset %s to sync always\n", dsname);
return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC,
ZFS_SYNC_ALWAYS, B_FALSE));
}
static int
ztest_objset_destroy_cb(const char *name, void *arg)
{
(void) arg;
objset_t *os;
dmu_object_info_t doi;
int error;
/*
* Verify that the dataset contains a directory object.
*/
VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE,
B_TRUE, FTAG, &os));
error = dmu_object_info(os, ZTEST_DIROBJ, &doi);
if (error != ENOENT) {
/* We could have crashed in the middle of destroying it */
ASSERT0(error);
ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER);
ASSERT3S(doi.doi_physical_blocks_512, >=, 0);
}
dmu_objset_disown(os, B_TRUE, FTAG);
/*
* Destroy the dataset.
*/
if (strchr(name, '@') != NULL) {
VERIFY0(dsl_destroy_snapshot(name, B_TRUE));
} else {
error = dsl_destroy_head(name);
if (error == ENOSPC) {
/* There could be checkpoint or insufficient slop */
ztest_record_enospc(FTAG);
} else if (error != EBUSY) {
/* There could be a hold on this dataset */
ASSERT0(error);
}
}
return (0);
}
static boolean_t
ztest_snapshot_create(char *osname, uint64_t id)
{
char snapname[ZFS_MAX_DATASET_NAME_LEN];
int error;
(void) snprintf(snapname, sizeof (snapname), "%"PRIu64"", id);
error = dmu_objset_snapshot_one(osname, snapname);
if (error == ENOSPC) {
ztest_record_enospc(FTAG);
return (B_FALSE);
}
if (error != 0 && error != EEXIST) {
fatal(B_FALSE, "ztest_snapshot_create(%s@%s) = %d", osname,
snapname, error);
}
return (B_TRUE);
}
static boolean_t
ztest_snapshot_destroy(char *osname, uint64_t id)
{
char snapname[ZFS_MAX_DATASET_NAME_LEN];
int error;
(void) snprintf(snapname, sizeof (snapname), "%s@%"PRIu64"",
osname, id);
error = dsl_destroy_snapshot(snapname, B_FALSE);
if (error != 0 && error != ENOENT)
fatal(B_FALSE, "ztest_snapshot_destroy(%s) = %d",
snapname, error);
return (B_TRUE);
}
void
ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id)
{
(void) zd;
ztest_ds_t *zdtmp;
int iters;
int error;
objset_t *os, *os2;
char name[ZFS_MAX_DATASET_NAME_LEN];
zilog_t *zilog;
int i;
zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL);
(void) pthread_rwlock_rdlock(&ztest_name_lock);
(void) snprintf(name, sizeof (name), "%s/temp_%"PRIu64"",
ztest_opts.zo_pool, id);
/*
* If this dataset exists from a previous run, process its replay log
* half of the time. If we don't replay it, then dsl_destroy_head()
* (invoked from ztest_objset_destroy_cb()) should just throw it away.
*/
if (ztest_random(2) == 0 &&
ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE,
B_TRUE, FTAG, &os) == 0) {
ztest_zd_init(zdtmp, NULL, os);
zil_replay(os, zdtmp, ztest_replay_vector);
ztest_zd_fini(zdtmp);
dmu_objset_disown(os, B_TRUE, FTAG);
}
/*
* There may be an old instance of the dataset we're about to
* create lying around from a previous run. If so, destroy it
* and all of its snapshots.
*/
(void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL,
DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS);
/*
* Verify that the destroyed dataset is no longer in the namespace.
*/
VERIFY3U(ENOENT, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE,
B_TRUE, FTAG, &os));
/*
* Verify that we can create a new dataset.
*/
error = ztest_dataset_create(name);
if (error) {
if (error == ENOSPC) {
ztest_record_enospc(FTAG);
goto out;
}
fatal(B_FALSE, "dmu_objset_create(%s) = %d", name, error);
}
VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE,
FTAG, &os));
ztest_zd_init(zdtmp, NULL, os);
/*
* Open the intent log for it.
*/
zilog = zil_open(os, ztest_get_data);
/*
* Put some objects in there, do a little I/O to them,
* and randomly take a couple of snapshots along the way.
*/
iters = ztest_random(5);
for (i = 0; i < iters; i++) {
ztest_dmu_object_alloc_free(zdtmp, id);
if (ztest_random(iters) == 0)
(void) ztest_snapshot_create(name, i);
}
/*
* Verify that we cannot create an existing dataset.
*/
VERIFY3U(EEXIST, ==,
dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL));
/*
* Verify that we can hold an objset that is also owned.
*/
VERIFY0(dmu_objset_hold(name, FTAG, &os2));
dmu_objset_rele(os2, FTAG);
/*
* Verify that we cannot own an objset that is already owned.
*/
VERIFY3U(EBUSY, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER,
B_FALSE, B_TRUE, FTAG, &os2));
zil_close(zilog);
dmu_objset_disown(os, B_TRUE, FTAG);
ztest_zd_fini(zdtmp);
out:
(void) pthread_rwlock_unlock(&ztest_name_lock);
umem_free(zdtmp, sizeof (ztest_ds_t));
}
/*
* Verify that dmu_snapshot_{create,destroy,open,close} work as expected.
*/
void
ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id)
{
(void) pthread_rwlock_rdlock(&ztest_name_lock);
(void) ztest_snapshot_destroy(zd->zd_name, id);
(void) ztest_snapshot_create(zd->zd_name, id);
(void) pthread_rwlock_unlock(&ztest_name_lock);
}
/*
* Cleanup non-standard snapshots and clones.
*/
static void
ztest_dsl_dataset_cleanup(char *osname, uint64_t id)
{
char *snap1name;
char *clone1name;
char *snap2name;
char *clone2name;
char *snap3name;
int error;
snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
(void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, "%s@s1_%"PRIu64"",
osname, id);
(void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, "%s/c1_%"PRIu64"",
osname, id);
(void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, "%s@s2_%"PRIu64"",
clone1name, id);
(void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, "%s/c2_%"PRIu64"",
osname, id);
(void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, "%s@s3_%"PRIu64"",
clone1name, id);
error = dsl_destroy_head(clone2name);
if (error && error != ENOENT)
fatal(B_FALSE, "dsl_destroy_head(%s) = %d", clone2name, error);
error = dsl_destroy_snapshot(snap3name, B_FALSE);
if (error && error != ENOENT)
fatal(B_FALSE, "dsl_destroy_snapshot(%s) = %d",
snap3name, error);
error = dsl_destroy_snapshot(snap2name, B_FALSE);
if (error && error != ENOENT)
fatal(B_FALSE, "dsl_destroy_snapshot(%s) = %d",
snap2name, error);
error = dsl_destroy_head(clone1name);
if (error && error != ENOENT)
fatal(B_FALSE, "dsl_destroy_head(%s) = %d", clone1name, error);
error = dsl_destroy_snapshot(snap1name, B_FALSE);
if (error && error != ENOENT)
fatal(B_FALSE, "dsl_destroy_snapshot(%s) = %d",
snap1name, error);
umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN);
umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN);
umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN);
umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN);
umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN);
}
/*
* Verify dsl_dataset_promote handles EBUSY
*/
void
ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id)
{
objset_t *os;
char *snap1name;
char *clone1name;
char *snap2name;
char *clone2name;
char *snap3name;
char *osname = zd->zd_name;
int error;
snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
(void) pthread_rwlock_rdlock(&ztest_name_lock);
ztest_dsl_dataset_cleanup(osname, id);
(void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, "%s@s1_%"PRIu64"",
osname, id);
(void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, "%s/c1_%"PRIu64"",
osname, id);
(void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, "%s@s2_%"PRIu64"",
clone1name, id);
(void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, "%s/c2_%"PRIu64"",
osname, id);
(void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, "%s@s3_%"PRIu64"",
clone1name, id);
error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1);
if (error && error != EEXIST) {
if (error == ENOSPC) {
ztest_record_enospc(FTAG);
goto out;
}
fatal(B_FALSE, "dmu_take_snapshot(%s) = %d", snap1name, error);
}
error = dmu_objset_clone(clone1name, snap1name);
if (error) {
if (error == ENOSPC) {
ztest_record_enospc(FTAG);
goto out;
}
fatal(B_FALSE, "dmu_objset_create(%s) = %d", clone1name, error);
}
error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1);
if (error && error != EEXIST) {
if (error == ENOSPC) {
ztest_record_enospc(FTAG);
goto out;
}
fatal(B_FALSE, "dmu_open_snapshot(%s) = %d", snap2name, error);
}
error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1);
if (error && error != EEXIST) {
if (error == ENOSPC) {
ztest_record_enospc(FTAG);
goto out;
}
fatal(B_FALSE, "dmu_open_snapshot(%s) = %d", snap3name, error);
}
error = dmu_objset_clone(clone2name, snap3name);
if (error) {
if (error == ENOSPC) {
ztest_record_enospc(FTAG);
goto out;
}
fatal(B_FALSE, "dmu_objset_create(%s) = %d", clone2name, error);
}
error = ztest_dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE,
FTAG, &os);
if (error)
fatal(B_FALSE, "dmu_objset_own(%s) = %d", snap2name, error);
error = dsl_dataset_promote(clone2name, NULL);
if (error == ENOSPC) {
dmu_objset_disown(os, B_TRUE, FTAG);
ztest_record_enospc(FTAG);
goto out;
}
if (error != EBUSY)
fatal(B_FALSE, "dsl_dataset_promote(%s), %d, not EBUSY",
clone2name, error);
dmu_objset_disown(os, B_TRUE, FTAG);
out:
ztest_dsl_dataset_cleanup(osname, id);
(void) pthread_rwlock_unlock(&ztest_name_lock);
umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN);
umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN);
umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN);
umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN);
umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN);
}
#undef OD_ARRAY_SIZE
#define OD_ARRAY_SIZE 4
/*
* Verify that dmu_object_{alloc,free} work as expected.
*/
void
ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id)
{
ztest_od_t *od;
int batchsize;
int size;
int b;
size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
od = umem_alloc(size, UMEM_NOFAIL);
batchsize = OD_ARRAY_SIZE;
for (b = 0; b < batchsize; b++)
ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER,
0, 0, 0);
/*
* Destroy the previous batch of objects, create a new batch,
* and do some I/O on the new objects.
*/
if (ztest_object_init(zd, od, size, B_TRUE) != 0)
return;
while (ztest_random(4 * batchsize) != 0)
ztest_io(zd, od[ztest_random(batchsize)].od_object,
ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
umem_free(od, size);
}
/*
* Rewind the global allocator to verify object allocation backfilling.
*/
void
ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id)
{
(void) id;
objset_t *os = zd->zd_os;
int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift;
uint64_t object;
/*
* Rewind the global allocator randomly back to a lower object number
* to force backfilling and reclamation of recently freed dnodes.
*/
mutex_enter(&os->os_obj_lock);
object = ztest_random(os->os_obj_next_chunk);
os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk);
mutex_exit(&os->os_obj_lock);
}
#undef OD_ARRAY_SIZE
#define OD_ARRAY_SIZE 2
/*
* Verify that dmu_{read,write} work as expected.
*/
void
ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id)
{
int size;
ztest_od_t *od;
objset_t *os = zd->zd_os;
size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
od = umem_alloc(size, UMEM_NOFAIL);
dmu_tx_t *tx;
int freeit, error;
uint64_t i, n, s, txg;
bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT;
uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize;
uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t);
uint64_t regions = 997;
uint64_t stride = 123456789ULL;
uint64_t width = 40;
int free_percent = 5;
/*
* This test uses two objects, packobj and bigobj, that are always
* updated together (i.e. in the same tx) so that their contents are
* in sync and can be compared. Their contents relate to each other
* in a simple way: packobj is a dense array of 'bufwad' structures,
* while bigobj is a sparse array of the same bufwads. Specifically,
* for any index n, there are three bufwads that should be identical:
*
* packobj, at offset n * sizeof (bufwad_t)
* bigobj, at the head of the nth chunk
* bigobj, at the tail of the nth chunk
*
* The chunk size is arbitrary. It doesn't have to be a power of two,
* and it doesn't have any relation to the object blocksize.
* The only requirement is that it can hold at least two bufwads.
*
* Normally, we write the bufwad to each of these locations.
* However, free_percent of the time we instead write zeroes to
* packobj and perform a dmu_free_range() on bigobj. By comparing
* bigobj to packobj, we can verify that the DMU is correctly
* tracking which parts of an object are allocated and free,
* and that the contents of the allocated blocks are correct.
*/
/*
* Read the directory info. If it's the first time, set things up.
*/
ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize);
ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0,
chunksize);
if (ztest_object_init(zd, od, size, B_FALSE) != 0) {
umem_free(od, size);
return;
}
bigobj = od[0].od_object;
packobj = od[1].od_object;
chunksize = od[0].od_gen;
ASSERT3U(chunksize, ==, od[1].od_gen);
/*
* Prefetch a random chunk of the big object.
* Our aim here is to get some async reads in flight
* for blocks that we may free below; the DMU should
* handle this race correctly.
*/
n = ztest_random(regions) * stride + ztest_random(width);
s = 1 + ztest_random(2 * width - 1);
dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize,
ZIO_PRIORITY_SYNC_READ);
/*
* Pick a random index and compute the offsets into packobj and bigobj.
*/
n = ztest_random(regions) * stride + ztest_random(width);
s = 1 + ztest_random(width - 1);
packoff = n * sizeof (bufwad_t);
packsize = s * sizeof (bufwad_t);
bigoff = n * chunksize;
bigsize = s * chunksize;
packbuf = umem_alloc(packsize, UMEM_NOFAIL);
bigbuf = umem_alloc(bigsize, UMEM_NOFAIL);
/*
* free_percent of the time, free a range of bigobj rather than
* overwriting it.
*/
freeit = (ztest_random(100) < free_percent);
/*
* Read the current contents of our objects.
*/
error = dmu_read(os, packobj, packoff, packsize, packbuf,
DMU_READ_PREFETCH);
ASSERT0(error);
error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf,
DMU_READ_PREFETCH);
ASSERT0(error);
/*
* Get a tx for the mods to both packobj and bigobj.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, packobj, packoff, packsize);
if (freeit)
dmu_tx_hold_free(tx, bigobj, bigoff, bigsize);
else
dmu_tx_hold_write(tx, bigobj, bigoff, bigsize);
/* This accounts for setting the checksum/compression. */
dmu_tx_hold_bonus(tx, bigobj);
txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
if (txg == 0) {
umem_free(packbuf, packsize);
umem_free(bigbuf, bigsize);
umem_free(od, size);
return;
}
enum zio_checksum cksum;
do {
cksum = (enum zio_checksum)
ztest_random_dsl_prop(ZFS_PROP_CHECKSUM);
} while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS);
dmu_object_set_checksum(os, bigobj, cksum, tx);
enum zio_compress comp;
do {
comp = (enum zio_compress)
ztest_random_dsl_prop(ZFS_PROP_COMPRESSION);
} while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS);
dmu_object_set_compress(os, bigobj, comp, tx);
/*
* For each index from n to n + s, verify that the existing bufwad
* in packobj matches the bufwads at the head and tail of the
* corresponding chunk in bigobj. Then update all three bufwads
* with the new values we want to write out.
*/
for (i = 0; i < s; i++) {
/* LINTED */
pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t));
/* LINTED */
bigH = (bufwad_t *)((char *)bigbuf + i * chunksize);
/* LINTED */
bigT = (bufwad_t *)((char *)bigH + chunksize) - 1;
ASSERT3U((uintptr_t)bigH - (uintptr_t)bigbuf, <, bigsize);
ASSERT3U((uintptr_t)bigT - (uintptr_t)bigbuf, <, bigsize);
if (pack->bw_txg > txg)
fatal(B_FALSE,
"future leak: got %"PRIx64", open txg is %"PRIx64"",
pack->bw_txg, txg);
if (pack->bw_data != 0 && pack->bw_index != n + i)
fatal(B_FALSE, "wrong index: "
"got %"PRIx64", wanted %"PRIx64"+%"PRIx64"",
pack->bw_index, n, i);
if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0)
fatal(B_FALSE, "pack/bigH mismatch in %p/%p",
pack, bigH);
if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0)
fatal(B_FALSE, "pack/bigT mismatch in %p/%p",
pack, bigT);
if (freeit) {
bzero(pack, sizeof (bufwad_t));
} else {
pack->bw_index = n + i;
pack->bw_txg = txg;
pack->bw_data = 1 + ztest_random(-2ULL);
}
*bigH = *pack;
*bigT = *pack;
}
/*
* We've verified all the old bufwads, and made new ones.
* Now write them out.
*/
dmu_write(os, packobj, packoff, packsize, packbuf, tx);
if (freeit) {
if (ztest_opts.zo_verbose >= 7) {
(void) printf("freeing offset %"PRIx64" size %"PRIx64""
" txg %"PRIx64"\n",
bigoff, bigsize, txg);
}
VERIFY0(dmu_free_range(os, bigobj, bigoff, bigsize, tx));
} else {
if (ztest_opts.zo_verbose >= 7) {
(void) printf("writing offset %"PRIx64" size %"PRIx64""
" txg %"PRIx64"\n",
bigoff, bigsize, txg);
}
dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx);
}
dmu_tx_commit(tx);
/*
* Sanity check the stuff we just wrote.
*/
{
void *packcheck = umem_alloc(packsize, UMEM_NOFAIL);
void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL);
VERIFY0(dmu_read(os, packobj, packoff,
packsize, packcheck, DMU_READ_PREFETCH));
VERIFY0(dmu_read(os, bigobj, bigoff,
bigsize, bigcheck, DMU_READ_PREFETCH));
ASSERT0(bcmp(packbuf, packcheck, packsize));
ASSERT0(bcmp(bigbuf, bigcheck, bigsize));
umem_free(packcheck, packsize);
umem_free(bigcheck, bigsize);
}
umem_free(packbuf, packsize);
umem_free(bigbuf, bigsize);
umem_free(od, size);
}
static void
compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf,
uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg)
{
uint64_t i;
bufwad_t *pack;
bufwad_t *bigH;
bufwad_t *bigT;
/*
* For each index from n to n + s, verify that the existing bufwad
* in packobj matches the bufwads at the head and tail of the
* corresponding chunk in bigobj. Then update all three bufwads
* with the new values we want to write out.
*/
for (i = 0; i < s; i++) {
/* LINTED */
pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t));
/* LINTED */
bigH = (bufwad_t *)((char *)bigbuf + i * chunksize);
/* LINTED */
bigT = (bufwad_t *)((char *)bigH + chunksize) - 1;
ASSERT3U((uintptr_t)bigH - (uintptr_t)bigbuf, <, bigsize);
ASSERT3U((uintptr_t)bigT - (uintptr_t)bigbuf, <, bigsize);
if (pack->bw_txg > txg)
fatal(B_FALSE,
"future leak: got %"PRIx64", open txg is %"PRIx64"",
pack->bw_txg, txg);
if (pack->bw_data != 0 && pack->bw_index != n + i)
fatal(B_FALSE, "wrong index: "
"got %"PRIx64", wanted %"PRIx64"+%"PRIx64"",
pack->bw_index, n, i);
if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0)
fatal(B_FALSE, "pack/bigH mismatch in %p/%p",
pack, bigH);
if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0)
fatal(B_FALSE, "pack/bigT mismatch in %p/%p",
pack, bigT);
pack->bw_index = n + i;
pack->bw_txg = txg;
pack->bw_data = 1 + ztest_random(-2ULL);
*bigH = *pack;
*bigT = *pack;
}
}
#undef OD_ARRAY_SIZE
#define OD_ARRAY_SIZE 2
void
ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id)
{
objset_t *os = zd->zd_os;
ztest_od_t *od;
dmu_tx_t *tx;
uint64_t i;
int error;
int size;
uint64_t n, s, txg;
bufwad_t *packbuf, *bigbuf;
uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize;
uint64_t blocksize = ztest_random_blocksize();
uint64_t chunksize = blocksize;
uint64_t regions = 997;
uint64_t stride = 123456789ULL;
uint64_t width = 9;
dmu_buf_t *bonus_db;
arc_buf_t **bigbuf_arcbufs;
dmu_object_info_t doi;
size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
od = umem_alloc(size, UMEM_NOFAIL);
/*
* This test uses two objects, packobj and bigobj, that are always
* updated together (i.e. in the same tx) so that their contents are
* in sync and can be compared. Their contents relate to each other
* in a simple way: packobj is a dense array of 'bufwad' structures,
* while bigobj is a sparse array of the same bufwads. Specifically,
* for any index n, there are three bufwads that should be identical:
*
* packobj, at offset n * sizeof (bufwad_t)
* bigobj, at the head of the nth chunk
* bigobj, at the tail of the nth chunk
*
* The chunk size is set equal to bigobj block size so that
* dmu_assign_arcbuf_by_dbuf() can be tested for object updates.
*/
/*
* Read the directory info. If it's the first time, set things up.
*/
ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0);
ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0,
chunksize);
if (ztest_object_init(zd, od, size, B_FALSE) != 0) {
umem_free(od, size);
return;
}
bigobj = od[0].od_object;
packobj = od[1].od_object;
blocksize = od[0].od_blocksize;
chunksize = blocksize;
ASSERT3U(chunksize, ==, od[1].od_gen);
VERIFY0(dmu_object_info(os, bigobj, &doi));
VERIFY(ISP2(doi.doi_data_block_size));
VERIFY3U(chunksize, ==, doi.doi_data_block_size);
VERIFY3U(chunksize, >=, 2 * sizeof (bufwad_t));
/*
* Pick a random index and compute the offsets into packobj and bigobj.
*/
n = ztest_random(regions) * stride + ztest_random(width);
s = 1 + ztest_random(width - 1);
packoff = n * sizeof (bufwad_t);
packsize = s * sizeof (bufwad_t);
bigoff = n * chunksize;
bigsize = s * chunksize;
packbuf = umem_zalloc(packsize, UMEM_NOFAIL);
bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL);
VERIFY0(dmu_bonus_hold(os, bigobj, FTAG, &bonus_db));
bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL);
/*
* Iteration 0 test zcopy for DB_UNCACHED dbufs.
* Iteration 1 test zcopy to already referenced dbufs.
* Iteration 2 test zcopy to dirty dbuf in the same txg.
* Iteration 3 test zcopy to dbuf dirty in previous txg.
* Iteration 4 test zcopy when dbuf is no longer dirty.
* Iteration 5 test zcopy when it can't be done.
* Iteration 6 one more zcopy write.
*/
for (i = 0; i < 7; i++) {
uint64_t j;
uint64_t off;
/*
* In iteration 5 (i == 5) use arcbufs
* that don't match bigobj blksz to test
* dmu_assign_arcbuf_by_dbuf() when it can't directly
* assign an arcbuf to a dbuf.
*/
for (j = 0; j < s; j++) {
if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) {
bigbuf_arcbufs[j] =
dmu_request_arcbuf(bonus_db, chunksize);
} else {
bigbuf_arcbufs[2 * j] =
dmu_request_arcbuf(bonus_db, chunksize / 2);
bigbuf_arcbufs[2 * j + 1] =
dmu_request_arcbuf(bonus_db, chunksize / 2);
}
}
/*
* Get a tx for the mods to both packobj and bigobj.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, packobj, packoff, packsize);
dmu_tx_hold_write(tx, bigobj, bigoff, bigsize);
txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
if (txg == 0) {
umem_free(packbuf, packsize);
umem_free(bigbuf, bigsize);
for (j = 0; j < s; j++) {
if (i != 5 ||
chunksize < (SPA_MINBLOCKSIZE * 2)) {
dmu_return_arcbuf(bigbuf_arcbufs[j]);
} else {
dmu_return_arcbuf(
bigbuf_arcbufs[2 * j]);
dmu_return_arcbuf(
bigbuf_arcbufs[2 * j + 1]);
}
}
umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
umem_free(od, size);
dmu_buf_rele(bonus_db, FTAG);
return;
}
/*
* 50% of the time don't read objects in the 1st iteration to
* test dmu_assign_arcbuf_by_dbuf() for the case when there are
* no existing dbufs for the specified offsets.
*/
if (i != 0 || ztest_random(2) != 0) {
error = dmu_read(os, packobj, packoff,
packsize, packbuf, DMU_READ_PREFETCH);
ASSERT0(error);
error = dmu_read(os, bigobj, bigoff, bigsize,
bigbuf, DMU_READ_PREFETCH);
ASSERT0(error);
}
compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize,
n, chunksize, txg);
/*
* We've verified all the old bufwads, and made new ones.
* Now write them out.
*/
dmu_write(os, packobj, packoff, packsize, packbuf, tx);
if (ztest_opts.zo_verbose >= 7) {
(void) printf("writing offset %"PRIx64" size %"PRIx64""
" txg %"PRIx64"\n",
bigoff, bigsize, txg);
}
for (off = bigoff, j = 0; j < s; j++, off += chunksize) {
dmu_buf_t *dbt;
if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) {
bcopy((caddr_t)bigbuf + (off - bigoff),
bigbuf_arcbufs[j]->b_data, chunksize);
} else {
bcopy((caddr_t)bigbuf + (off - bigoff),
bigbuf_arcbufs[2 * j]->b_data,
chunksize / 2);
bcopy((caddr_t)bigbuf + (off - bigoff) +
chunksize / 2,
bigbuf_arcbufs[2 * j + 1]->b_data,
chunksize / 2);
}
if (i == 1) {
VERIFY(dmu_buf_hold(os, bigobj, off,
FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0);
}
if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) {
VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db,
off, bigbuf_arcbufs[j], tx));
} else {
VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db,
off, bigbuf_arcbufs[2 * j], tx));
VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db,
off + chunksize / 2,
bigbuf_arcbufs[2 * j + 1], tx));
}
if (i == 1) {
dmu_buf_rele(dbt, FTAG);
}
}
dmu_tx_commit(tx);
/*
* Sanity check the stuff we just wrote.
*/
{
void *packcheck = umem_alloc(packsize, UMEM_NOFAIL);
void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL);
VERIFY0(dmu_read(os, packobj, packoff,
packsize, packcheck, DMU_READ_PREFETCH));
VERIFY0(dmu_read(os, bigobj, bigoff,
bigsize, bigcheck, DMU_READ_PREFETCH));
ASSERT0(bcmp(packbuf, packcheck, packsize));
ASSERT0(bcmp(bigbuf, bigcheck, bigsize));
umem_free(packcheck, packsize);
umem_free(bigcheck, bigsize);
}
if (i == 2) {
txg_wait_open(dmu_objset_pool(os), 0, B_TRUE);
} else if (i == 3) {
txg_wait_synced(dmu_objset_pool(os), 0);
}
}
dmu_buf_rele(bonus_db, FTAG);
umem_free(packbuf, packsize);
umem_free(bigbuf, bigsize);
umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
umem_free(od, size);
}
void
ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id)
{
(void) id;
ztest_od_t *od;
od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
uint64_t offset = (1ULL << (ztest_random(20) + 43)) +
(ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
/*
* Have multiple threads write to large offsets in an object
* to verify that parallel writes to an object -- even to the
* same blocks within the object -- doesn't cause any trouble.
*/
ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0);
if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0)
return;
while (ztest_random(10) != 0)
ztest_io(zd, od->od_object, offset);
umem_free(od, sizeof (ztest_od_t));
}
void
ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id)
{
ztest_od_t *od;
uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) +
(ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
uint64_t count = ztest_random(20) + 1;
uint64_t blocksize = ztest_random_blocksize();
void *data;
od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0);
if (ztest_object_init(zd, od, sizeof (ztest_od_t),
!ztest_random(2)) != 0) {
umem_free(od, sizeof (ztest_od_t));
return;
}
if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) {
umem_free(od, sizeof (ztest_od_t));
return;
}
ztest_prealloc(zd, od->od_object, offset, count * blocksize);
data = umem_zalloc(blocksize, UMEM_NOFAIL);
while (ztest_random(count) != 0) {
uint64_t randoff = offset + (ztest_random(count) * blocksize);
if (ztest_write(zd, od->od_object, randoff, blocksize,
data) != 0)
break;
while (ztest_random(4) != 0)
ztest_io(zd, od->od_object, randoff);
}
umem_free(data, blocksize);
umem_free(od, sizeof (ztest_od_t));
}
/*
* Verify that zap_{create,destroy,add,remove,update} work as expected.
*/
#define ZTEST_ZAP_MIN_INTS 1
#define ZTEST_ZAP_MAX_INTS 4
#define ZTEST_ZAP_MAX_PROPS 1000
void
ztest_zap(ztest_ds_t *zd, uint64_t id)
{
objset_t *os = zd->zd_os;
ztest_od_t *od;
uint64_t object;
uint64_t txg, last_txg;
uint64_t value[ZTEST_ZAP_MAX_INTS];
uint64_t zl_ints, zl_intsize, prop;
int i, ints;
dmu_tx_t *tx;
char propname[100], txgname[100];
int error;
char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" };
od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0);
if (ztest_object_init(zd, od, sizeof (ztest_od_t),
!ztest_random(2)) != 0)
goto out;
object = od->od_object;
/*
* Generate a known hash collision, and verify that
* we can lookup and remove both entries.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
if (txg == 0)
goto out;
for (i = 0; i < 2; i++) {
value[i] = i;
VERIFY0(zap_add(os, object, hc[i], sizeof (uint64_t),
1, &value[i], tx));
}
for (i = 0; i < 2; i++) {
VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i],
sizeof (uint64_t), 1, &value[i], tx));
VERIFY0(
zap_length(os, object, hc[i], &zl_intsize, &zl_ints));
ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
ASSERT3U(zl_ints, ==, 1);
}
for (i = 0; i < 2; i++) {
VERIFY0(zap_remove(os, object, hc[i], tx));
}
dmu_tx_commit(tx);
/*
* Generate a bunch of random entries.
*/
ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS);
prop = ztest_random(ZTEST_ZAP_MAX_PROPS);
(void) sprintf(propname, "prop_%"PRIu64"", prop);
(void) sprintf(txgname, "txg_%"PRIu64"", prop);
bzero(value, sizeof (value));
last_txg = 0;
/*
* If these zap entries already exist, validate their contents.
*/
error = zap_length(os, object, txgname, &zl_intsize, &zl_ints);
if (error == 0) {
ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
ASSERT3U(zl_ints, ==, 1);
VERIFY0(zap_lookup(os, object, txgname, zl_intsize,
zl_ints, &last_txg));
VERIFY0(zap_length(os, object, propname, &zl_intsize,
&zl_ints));
ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
ASSERT3U(zl_ints, ==, ints);
VERIFY0(zap_lookup(os, object, propname, zl_intsize,
zl_ints, value));
for (i = 0; i < ints; i++) {
ASSERT3U(value[i], ==, last_txg + object + i);
}
} else {
ASSERT3U(error, ==, ENOENT);
}
/*
* Atomically update two entries in our zap object.
* The first is named txg_%llu, and contains the txg
* in which the property was last updated. The second
* is named prop_%llu, and the nth element of its value
* should be txg + object + n.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
if (txg == 0)
goto out;
if (last_txg > txg)
fatal(B_FALSE, "zap future leak: old %"PRIu64" new %"PRIu64"",
last_txg, txg);
for (i = 0; i < ints; i++)
value[i] = txg + object + i;
VERIFY0(zap_update(os, object, txgname, sizeof (uint64_t),
1, &txg, tx));
VERIFY0(zap_update(os, object, propname, sizeof (uint64_t),
ints, value, tx));
dmu_tx_commit(tx);
/*
* Remove a random pair of entries.
*/
prop = ztest_random(ZTEST_ZAP_MAX_PROPS);
(void) sprintf(propname, "prop_%"PRIu64"", prop);
(void) sprintf(txgname, "txg_%"PRIu64"", prop);
error = zap_length(os, object, txgname, &zl_intsize, &zl_ints);
if (error == ENOENT)
goto out;
ASSERT0(error);
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
if (txg == 0)
goto out;
VERIFY0(zap_remove(os, object, txgname, tx));
VERIFY0(zap_remove(os, object, propname, tx));
dmu_tx_commit(tx);
out:
umem_free(od, sizeof (ztest_od_t));
}
/*
* Test case to test the upgrading of a microzap to fatzap.
*/
void
ztest_fzap(ztest_ds_t *zd, uint64_t id)
{
objset_t *os = zd->zd_os;
ztest_od_t *od;
uint64_t object, txg, value;
od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0);
if (ztest_object_init(zd, od, sizeof (ztest_od_t),
!ztest_random(2)) != 0)
goto out;
object = od->od_object;
/*
* Add entries to this ZAP and make sure it spills over
* and gets upgraded to a fatzap. Also, since we are adding
* 2050 entries we should see ptrtbl growth and leaf-block split.
*/
for (value = 0; value < 2050; value++) {
char name[ZFS_MAX_DATASET_NAME_LEN];
dmu_tx_t *tx;
int error;
(void) snprintf(name, sizeof (name), "fzap-%"PRIu64"-%"PRIu64"",
id, value);
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, object, B_TRUE, name);
txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
if (txg == 0)
goto out;
error = zap_add(os, object, name, sizeof (uint64_t), 1,
&value, tx);
ASSERT(error == 0 || error == EEXIST);
dmu_tx_commit(tx);
}
out:
umem_free(od, sizeof (ztest_od_t));
}
void
ztest_zap_parallel(ztest_ds_t *zd, uint64_t id)
{
(void) id;
objset_t *os = zd->zd_os;
ztest_od_t *od;
uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc;
dmu_tx_t *tx;
int i, namelen, error;
int micro = ztest_random(2);
char name[20], string_value[20];
void *data;
od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0);
if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) {
umem_free(od, sizeof (ztest_od_t));
return;
}
object = od->od_object;
/*
* Generate a random name of the form 'xxx.....' where each
* x is a random printable character and the dots are dots.
* There are 94 such characters, and the name length goes from
* 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names.
*/
namelen = ztest_random(sizeof (name) - 5) + 5 + 1;
for (i = 0; i < 3; i++)
name[i] = '!' + ztest_random('~' - '!' + 1);
for (; i < namelen - 1; i++)
name[i] = '.';
name[i] = '\0';
if ((namelen & 1) || micro) {
wsize = sizeof (txg);
wc = 1;
data = &txg;
} else {
wsize = 1;
wc = namelen;
data = string_value;
}
count = -1ULL;
VERIFY0(zap_count(os, object, &count));
ASSERT3S(count, !=, -1ULL);
/*
* Select an operation: length, lookup, add, update, remove.
*/
i = ztest_random(5);
if (i >= 2) {
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
if (txg == 0) {
umem_free(od, sizeof (ztest_od_t));
return;
}
bcopy(name, string_value, namelen);
} else {
tx = NULL;
txg = 0;
bzero(string_value, namelen);
}
switch (i) {
case 0:
error = zap_length(os, object, name, &zl_wsize, &zl_wc);
if (error == 0) {
ASSERT3U(wsize, ==, zl_wsize);
ASSERT3U(wc, ==, zl_wc);
} else {
ASSERT3U(error, ==, ENOENT);
}
break;
case 1:
error = zap_lookup(os, object, name, wsize, wc, data);
if (error == 0) {
if (data == string_value &&
bcmp(name, data, namelen) != 0)
fatal(B_FALSE, "name '%s' != val '%s' len %d",
name, (char *)data, namelen);
} else {
ASSERT3U(error, ==, ENOENT);
}
break;
case 2:
error = zap_add(os, object, name, wsize, wc, data, tx);
ASSERT(error == 0 || error == EEXIST);
break;
case 3:
VERIFY0(zap_update(os, object, name, wsize, wc, data, tx));
break;
case 4:
error = zap_remove(os, object, name, tx);
ASSERT(error == 0 || error == ENOENT);
break;
}
if (tx != NULL)
dmu_tx_commit(tx);
umem_free(od, sizeof (ztest_od_t));
}
/*
* Commit callback data.
*/
typedef struct ztest_cb_data {
list_node_t zcd_node;
uint64_t zcd_txg;
int zcd_expected_err;
boolean_t zcd_added;
boolean_t zcd_called;
spa_t *zcd_spa;
} ztest_cb_data_t;
/* This is the actual commit callback function */
static void
ztest_commit_callback(void *arg, int error)
{
ztest_cb_data_t *data = arg;
uint64_t synced_txg;
VERIFY3P(data, !=, NULL);
VERIFY3S(data->zcd_expected_err, ==, error);
VERIFY(!data->zcd_called);
synced_txg = spa_last_synced_txg(data->zcd_spa);
if (data->zcd_txg > synced_txg)
fatal(B_FALSE,
"commit callback of txg %"PRIu64" called prematurely, "
"last synced txg = %"PRIu64"\n",
data->zcd_txg, synced_txg);
data->zcd_called = B_TRUE;
if (error == ECANCELED) {
ASSERT0(data->zcd_txg);
ASSERT(!data->zcd_added);
/*
* The private callback data should be destroyed here, but
* since we are going to check the zcd_called field after
* dmu_tx_abort(), we will destroy it there.
*/
return;
}
ASSERT(data->zcd_added);
ASSERT3U(data->zcd_txg, !=, 0);
(void) mutex_enter(&zcl.zcl_callbacks_lock);
/* See if this cb was called more quickly */
if ((synced_txg - data->zcd_txg) < zc_min_txg_delay)
zc_min_txg_delay = synced_txg - data->zcd_txg;
/* Remove our callback from the list */
list_remove(&zcl.zcl_callbacks, data);
(void) mutex_exit(&zcl.zcl_callbacks_lock);
umem_free(data, sizeof (ztest_cb_data_t));
}
/* Allocate and initialize callback data structure */
static ztest_cb_data_t *
ztest_create_cb_data(objset_t *os, uint64_t txg)
{
ztest_cb_data_t *cb_data;
cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL);
cb_data->zcd_txg = txg;
cb_data->zcd_spa = dmu_objset_spa(os);
list_link_init(&cb_data->zcd_node);
return (cb_data);
}
/*
* Commit callback test.
*/
void
ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id)
{
objset_t *os = zd->zd_os;
ztest_od_t *od;
dmu_tx_t *tx;
ztest_cb_data_t *cb_data[3], *tmp_cb;
uint64_t old_txg, txg;
int i, error = 0;
od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0);
if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) {
umem_free(od, sizeof (ztest_od_t));
return;
}
tx = dmu_tx_create(os);
cb_data[0] = ztest_create_cb_data(os, 0);
dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]);
dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t));
/* Every once in a while, abort the transaction on purpose */
if (ztest_random(100) == 0)
error = -1;
if (!error)
error = dmu_tx_assign(tx, TXG_NOWAIT);
txg = error ? 0 : dmu_tx_get_txg(tx);
cb_data[0]->zcd_txg = txg;
cb_data[1] = ztest_create_cb_data(os, txg);
dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]);
if (error) {
/*
* It's not a strict requirement to call the registered
* callbacks from inside dmu_tx_abort(), but that's what
* it's supposed to happen in the current implementation
* so we will check for that.
*/
for (i = 0; i < 2; i++) {
cb_data[i]->zcd_expected_err = ECANCELED;
VERIFY(!cb_data[i]->zcd_called);
}
dmu_tx_abort(tx);
for (i = 0; i < 2; i++) {
VERIFY(cb_data[i]->zcd_called);
umem_free(cb_data[i], sizeof (ztest_cb_data_t));
}
umem_free(od, sizeof (ztest_od_t));
return;
}
cb_data[2] = ztest_create_cb_data(os, txg);
dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]);
/*
* Read existing data to make sure there isn't a future leak.
*/
VERIFY0(dmu_read(os, od->od_object, 0, sizeof (uint64_t),
&old_txg, DMU_READ_PREFETCH));
if (old_txg > txg)
fatal(B_FALSE,
"future leak: got %"PRIu64", open txg is %"PRIu64"",
old_txg, txg);
dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx);
(void) mutex_enter(&zcl.zcl_callbacks_lock);
/*
* Since commit callbacks don't have any ordering requirement and since
* it is theoretically possible for a commit callback to be called
* after an arbitrary amount of time has elapsed since its txg has been
* synced, it is difficult to reliably determine whether a commit
* callback hasn't been called due to high load or due to a flawed
* implementation.
*
* In practice, we will assume that if after a certain number of txgs a
* commit callback hasn't been called, then most likely there's an
* implementation bug..
*/
tmp_cb = list_head(&zcl.zcl_callbacks);
if (tmp_cb != NULL &&
tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) {
fatal(B_FALSE,
"Commit callback threshold exceeded, "
"oldest txg: %"PRIu64", open txg: %"PRIu64"\n",
tmp_cb->zcd_txg, txg);
}
/*
* Let's find the place to insert our callbacks.
*
* Even though the list is ordered by txg, it is possible for the
* insertion point to not be the end because our txg may already be
* quiescing at this point and other callbacks in the open txg
* (from other objsets) may have sneaked in.
*/
tmp_cb = list_tail(&zcl.zcl_callbacks);
while (tmp_cb != NULL && tmp_cb->zcd_txg > txg)
tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb);
/* Add the 3 callbacks to the list */
for (i = 0; i < 3; i++) {
if (tmp_cb == NULL)
list_insert_head(&zcl.zcl_callbacks, cb_data[i]);
else
list_insert_after(&zcl.zcl_callbacks, tmp_cb,
cb_data[i]);
cb_data[i]->zcd_added = B_TRUE;
VERIFY(!cb_data[i]->zcd_called);
tmp_cb = cb_data[i];
}
zc_cb_counter += 3;
(void) mutex_exit(&zcl.zcl_callbacks_lock);
dmu_tx_commit(tx);
umem_free(od, sizeof (ztest_od_t));
}
/*
* Visit each object in the dataset. Verify that its properties
* are consistent what was stored in the block tag when it was created,
* and that its unused bonus buffer space has not been overwritten.
*/
void
ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id)
{
(void) id;
objset_t *os = zd->zd_os;
uint64_t obj;
int err = 0;
for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) {
ztest_block_tag_t *bt = NULL;
dmu_object_info_t doi;
dmu_buf_t *db;
ztest_object_lock(zd, obj, RL_READER);
if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) {
ztest_object_unlock(zd, obj);
continue;
}
dmu_object_info_from_db(db, &doi);
if (doi.doi_bonus_size >= sizeof (*bt))
bt = ztest_bt_bonus(db);
if (bt && bt->bt_magic == BT_MAGIC) {
ztest_bt_verify(bt, os, obj, doi.doi_dnodesize,
bt->bt_offset, bt->bt_gen, bt->bt_txg,
bt->bt_crtxg);
ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen);
}
dmu_buf_rele(db, FTAG);
ztest_object_unlock(zd, obj);
}
}
void
ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id)
{
(void) id;
zfs_prop_t proplist[] = {
ZFS_PROP_CHECKSUM,
ZFS_PROP_COMPRESSION,
ZFS_PROP_COPIES,
ZFS_PROP_DEDUP
};
(void) pthread_rwlock_rdlock(&ztest_name_lock);
for (int p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++)
(void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p],
ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2));
VERIFY0(ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE,
ztest_random_blocksize(), (int)ztest_random(2)));
(void) pthread_rwlock_unlock(&ztest_name_lock);
}
void
ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
nvlist_t *props = NULL;
(void) pthread_rwlock_rdlock(&ztest_name_lock);
(void) ztest_spa_prop_set_uint64(ZPOOL_PROP_AUTOTRIM, ztest_random(2));
VERIFY0(spa_prop_get(ztest_spa, &props));
if (ztest_opts.zo_verbose >= 6)
dump_nvlist(props, 4);
fnvlist_free(props);
(void) pthread_rwlock_unlock(&ztest_name_lock);
}
static int
user_release_one(const char *snapname, const char *holdname)
{
nvlist_t *snaps, *holds;
int error;
snaps = fnvlist_alloc();
holds = fnvlist_alloc();
fnvlist_add_boolean(holds, holdname);
fnvlist_add_nvlist(snaps, snapname, holds);
fnvlist_free(holds);
error = dsl_dataset_user_release(snaps, NULL);
fnvlist_free(snaps);
return (error);
}
/*
* Test snapshot hold/release and deferred destroy.
*/
void
ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id)
{
int error;
objset_t *os = zd->zd_os;
objset_t *origin;
char snapname[100];
char fullname[100];
char clonename[100];
char tag[100];
char osname[ZFS_MAX_DATASET_NAME_LEN];
nvlist_t *holds;
(void) pthread_rwlock_rdlock(&ztest_name_lock);
dmu_objset_name(os, osname);
(void) snprintf(snapname, sizeof (snapname), "sh1_%"PRIu64"", id);
(void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname);
(void) snprintf(clonename, sizeof (clonename), "%s/ch1_%"PRIu64"",
osname, id);
(void) snprintf(tag, sizeof (tag), "tag_%"PRIu64"", id);
/*
* Clean up from any previous run.
*/
error = dsl_destroy_head(clonename);
if (error != ENOENT)
ASSERT0(error);
error = user_release_one(fullname, tag);
if (error != ESRCH && error != ENOENT)
ASSERT0(error);
error = dsl_destroy_snapshot(fullname, B_FALSE);
if (error != ENOENT)
ASSERT0(error);
/*
* Create snapshot, clone it, mark snap for deferred destroy,
* destroy clone, verify snap was also destroyed.
*/
error = dmu_objset_snapshot_one(osname, snapname);
if (error) {
if (error == ENOSPC) {
ztest_record_enospc("dmu_objset_snapshot");
goto out;
}
fatal(B_FALSE, "dmu_objset_snapshot(%s) = %d", fullname, error);
}
error = dmu_objset_clone(clonename, fullname);
if (error) {
if (error == ENOSPC) {
ztest_record_enospc("dmu_objset_clone");
goto out;
}
fatal(B_FALSE, "dmu_objset_clone(%s) = %d", clonename, error);
}
error = dsl_destroy_snapshot(fullname, B_TRUE);
if (error) {
fatal(B_FALSE, "dsl_destroy_snapshot(%s, B_TRUE) = %d",
fullname, error);
}
error = dsl_destroy_head(clonename);
if (error)
fatal(B_FALSE, "dsl_destroy_head(%s) = %d", clonename, error);
error = dmu_objset_hold(fullname, FTAG, &origin);
if (error != ENOENT)
fatal(B_FALSE, "dmu_objset_hold(%s) = %d", fullname, error);
/*
* Create snapshot, add temporary hold, verify that we can't
* destroy a held snapshot, mark for deferred destroy,
* release hold, verify snapshot was destroyed.
*/
error = dmu_objset_snapshot_one(osname, snapname);
if (error) {
if (error == ENOSPC) {
ztest_record_enospc("dmu_objset_snapshot");
goto out;
}
fatal(B_FALSE, "dmu_objset_snapshot(%s) = %d", fullname, error);
}
holds = fnvlist_alloc();
fnvlist_add_string(holds, fullname, tag);
error = dsl_dataset_user_hold(holds, 0, NULL);
fnvlist_free(holds);
if (error == ENOSPC) {
ztest_record_enospc("dsl_dataset_user_hold");
goto out;
} else if (error) {
fatal(B_FALSE, "dsl_dataset_user_hold(%s, %s) = %u",
fullname, tag, error);
}
error = dsl_destroy_snapshot(fullname, B_FALSE);
if (error != EBUSY) {
fatal(B_FALSE, "dsl_destroy_snapshot(%s, B_FALSE) = %d",
fullname, error);
}
error = dsl_destroy_snapshot(fullname, B_TRUE);
if (error) {
fatal(B_FALSE, "dsl_destroy_snapshot(%s, B_TRUE) = %d",
fullname, error);
}
error = user_release_one(fullname, tag);
if (error)
fatal(B_FALSE, "user_release_one(%s, %s) = %d",
fullname, tag, error);
VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT);
out:
(void) pthread_rwlock_unlock(&ztest_name_lock);
}
/*
* Inject random faults into the on-disk data.
*/
void
ztest_fault_inject(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
ztest_shared_t *zs = ztest_shared;
spa_t *spa = ztest_spa;
int fd;
uint64_t offset;
uint64_t leaves;
uint64_t bad = 0x1990c0ffeedecadeull;
uint64_t top, leaf;
char *path0;
char *pathrand;
size_t fsize;
int bshift = SPA_MAXBLOCKSHIFT + 2;
int iters = 1000;
int maxfaults;
int mirror_save;
vdev_t *vd0 = NULL;
uint64_t guid0 = 0;
boolean_t islog = B_FALSE;
path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
mutex_enter(&ztest_vdev_lock);
/*
* Device removal is in progress, fault injection must be disabled
* until it completes and the pool is scrubbed. The fault injection
* strategy for damaging blocks does not take in to account evacuated
* blocks which may have already been damaged.
*/
if (ztest_device_removal_active) {
mutex_exit(&ztest_vdev_lock);
goto out;
}
maxfaults = MAXFAULTS(zs);
leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raid_children;
mirror_save = zs->zs_mirrors;
mutex_exit(&ztest_vdev_lock);
ASSERT3U(leaves, >=, 1);
/*
* While ztest is running the number of leaves will not change. This
* is critical for the fault injection logic as it determines where
* errors can be safely injected such that they are always repairable.
*
* When restarting ztest a different number of leaves may be requested
* which will shift the regions to be damaged. This is fine as long
* as the pool has been scrubbed prior to using the new mapping.
* Failure to do can result in non-repairable damage being injected.
*/
if (ztest_pool_scrubbed == B_FALSE)
goto out;
/*
* Grab the name lock as reader. There are some operations
* which don't like to have their vdevs changed while
* they are in progress (i.e. spa_change_guid). Those
* operations will have grabbed the name lock as writer.
*/
(void) pthread_rwlock_rdlock(&ztest_name_lock);
/*
* We need SCL_STATE here because we're going to look at vd0->vdev_tsd.
*/
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
if (ztest_random(2) == 0) {
/*
* Inject errors on a normal data device or slog device.
*/
top = ztest_random_vdev_top(spa, B_TRUE);
leaf = ztest_random(leaves) + zs->zs_splits;
/*
* Generate paths to the first leaf in this top-level vdev,
* and to the random leaf we selected. We'll induce transient
* write failures and random online/offline activity on leaf 0,
* and we'll write random garbage to the randomly chosen leaf.
*/
(void) snprintf(path0, MAXPATHLEN, ztest_dev_template,
ztest_opts.zo_dir, ztest_opts.zo_pool,
top * leaves + zs->zs_splits);
(void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template,
ztest_opts.zo_dir, ztest_opts.zo_pool,
top * leaves + leaf);
vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0);
if (vd0 != NULL && vd0->vdev_top->vdev_islog)
islog = B_TRUE;
/*
* If the top-level vdev needs to be resilvered
* then we only allow faults on the device that is
* resilvering.
*/
if (vd0 != NULL && maxfaults != 1 &&
(!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) ||
vd0->vdev_resilver_txg != 0)) {
/*
* Make vd0 explicitly claim to be unreadable,
* or unwritable, or reach behind its back
* and close the underlying fd. We can do this if
* maxfaults == 0 because we'll fail and reexecute,
* and we can do it if maxfaults >= 2 because we'll
* have enough redundancy. If maxfaults == 1, the
* combination of this with injection of random data
* corruption below exceeds the pool's fault tolerance.
*/
vdev_file_t *vf = vd0->vdev_tsd;
zfs_dbgmsg("injecting fault to vdev %llu; maxfaults=%d",
(long long)vd0->vdev_id, (int)maxfaults);
if (vf != NULL && ztest_random(3) == 0) {
(void) close(vf->vf_file->f_fd);
vf->vf_file->f_fd = -1;
} else if (ztest_random(2) == 0) {
vd0->vdev_cant_read = B_TRUE;
} else {
vd0->vdev_cant_write = B_TRUE;
}
guid0 = vd0->vdev_guid;
}
} else {
/*
* Inject errors on an l2cache device.
*/
spa_aux_vdev_t *sav = &spa->spa_l2cache;
if (sav->sav_count == 0) {
spa_config_exit(spa, SCL_STATE, FTAG);
(void) pthread_rwlock_unlock(&ztest_name_lock);
goto out;
}
vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)];
guid0 = vd0->vdev_guid;
(void) strcpy(path0, vd0->vdev_path);
(void) strcpy(pathrand, vd0->vdev_path);
leaf = 0;
leaves = 1;
maxfaults = INT_MAX; /* no limit on cache devices */
}
spa_config_exit(spa, SCL_STATE, FTAG);
(void) pthread_rwlock_unlock(&ztest_name_lock);
/*
* If we can tolerate two or more faults, or we're dealing
* with a slog, randomly online/offline vd0.
*/
if ((maxfaults >= 2 || islog) && guid0 != 0) {
if (ztest_random(10) < 6) {
int flags = (ztest_random(2) == 0 ?
ZFS_OFFLINE_TEMPORARY : 0);
/*
* We have to grab the zs_name_lock as writer to
* prevent a race between offlining a slog and
* destroying a dataset. Offlining the slog will
* grab a reference on the dataset which may cause
* dsl_destroy_head() to fail with EBUSY thus
* leaving the dataset in an inconsistent state.
*/
if (islog)
(void) pthread_rwlock_wrlock(&ztest_name_lock);
VERIFY3U(vdev_offline(spa, guid0, flags), !=, EBUSY);
if (islog)
(void) pthread_rwlock_unlock(&ztest_name_lock);
} else {
/*
* Ideally we would like to be able to randomly
* call vdev_[on|off]line without holding locks
* to force unpredictable failures but the side
* effects of vdev_[on|off]line prevent us from
* doing so. We grab the ztest_vdev_lock here to
* prevent a race between injection testing and
* aux_vdev removal.
*/
mutex_enter(&ztest_vdev_lock);
(void) vdev_online(spa, guid0, 0, NULL);
mutex_exit(&ztest_vdev_lock);
}
}
if (maxfaults == 0)
goto out;
/*
* We have at least single-fault tolerance, so inject data corruption.
*/
fd = open(pathrand, O_RDWR);
if (fd == -1) /* we hit a gap in the device namespace */
goto out;
fsize = lseek(fd, 0, SEEK_END);
while (--iters != 0) {
/*
* The offset must be chosen carefully to ensure that
* we do not inject a given logical block with errors
* on two different leaf devices, because ZFS can not
* tolerate that (if maxfaults==1).
*
* To achieve this we divide each leaf device into
* chunks of size (# leaves * SPA_MAXBLOCKSIZE * 4).
* Each chunk is further divided into error-injection
* ranges (can accept errors) and clear ranges (we do
* not inject errors in those). Each error-injection
* range can accept errors only for a single leaf vdev.
* Error-injection ranges are separated by clear ranges.
*
* For example, with 3 leaves, each chunk looks like:
* 0 to 32M: injection range for leaf 0
* 32M to 64M: clear range - no injection allowed
* 64M to 96M: injection range for leaf 1
* 96M to 128M: clear range - no injection allowed
* 128M to 160M: injection range for leaf 2
* 160M to 192M: clear range - no injection allowed
*
* Each clear range must be large enough such that a
* single block cannot straddle it. This way a block
* can't be a target in two different injection ranges
* (on different leaf vdevs).
*/
offset = ztest_random(fsize / (leaves << bshift)) *
(leaves << bshift) + (leaf << bshift) +
(ztest_random(1ULL << (bshift - 1)) & -8ULL);
/*
* Only allow damage to the labels at one end of the vdev.
*
* If all labels are damaged, the device will be totally
* inaccessible, which will result in loss of data,
* because we also damage (parts of) the other side of
* the mirror/raidz.
*
* Additionally, we will always have both an even and an
* odd label, so that we can handle crashes in the
* middle of vdev_config_sync().
*/
if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE)
continue;
/*
* The two end labels are stored at the "end" of the disk, but
* the end of the disk (vdev_psize) is aligned to
* sizeof (vdev_label_t).
*/
uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t));
if ((leaf & 1) == 1 &&
offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE)
continue;
mutex_enter(&ztest_vdev_lock);
if (mirror_save != zs->zs_mirrors) {
mutex_exit(&ztest_vdev_lock);
(void) close(fd);
goto out;
}
if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad))
fatal(B_TRUE,
"can't inject bad word at 0x%"PRIx64" in %s",
offset, pathrand);
mutex_exit(&ztest_vdev_lock);
if (ztest_opts.zo_verbose >= 7)
(void) printf("injected bad word into %s,"
" offset 0x%"PRIx64"\n", pathrand, offset);
}
(void) close(fd);
out:
umem_free(path0, MAXPATHLEN);
umem_free(pathrand, MAXPATHLEN);
}
/*
* By design ztest will never inject uncorrectable damage in to the pool.
* Issue a scrub, wait for it to complete, and verify there is never any
* persistent damage.
*
* Only after a full scrub has been completed is it safe to start injecting
* data corruption. See the comment in zfs_fault_inject().
*/
static int
ztest_scrub_impl(spa_t *spa)
{
int error = spa_scan(spa, POOL_SCAN_SCRUB);
if (error)
return (error);
while (dsl_scan_scrubbing(spa_get_dsl(spa)))
txg_wait_synced(spa_get_dsl(spa), 0);
if (spa_get_errlog_size(spa) > 0)
return (ECKSUM);
ztest_pool_scrubbed = B_TRUE;
return (0);
}
/*
* Scrub the pool.
*/
void
ztest_scrub(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
spa_t *spa = ztest_spa;
int error;
/*
* Scrub in progress by device removal.
*/
if (ztest_device_removal_active)
return;
/*
* Start a scrub, wait a moment, then force a restart.
*/
(void) spa_scan(spa, POOL_SCAN_SCRUB);
(void) poll(NULL, 0, 100);
error = ztest_scrub_impl(spa);
if (error == EBUSY)
error = 0;
ASSERT0(error);
}
/*
* Change the guid for the pool.
*/
void
ztest_reguid(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
spa_t *spa = ztest_spa;
uint64_t orig, load;
int error;
if (ztest_opts.zo_mmp_test)
return;
orig = spa_guid(spa);
load = spa_load_guid(spa);
(void) pthread_rwlock_wrlock(&ztest_name_lock);
error = spa_change_guid(spa);
(void) pthread_rwlock_unlock(&ztest_name_lock);
if (error != 0)
return;
if (ztest_opts.zo_verbose >= 4) {
(void) printf("Changed guid old %"PRIu64" -> %"PRIu64"\n",
orig, spa_guid(spa));
}
VERIFY3U(orig, !=, spa_guid(spa));
VERIFY3U(load, ==, spa_load_guid(spa));
}
void
ztest_fletcher(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
hrtime_t end = gethrtime() + NANOSEC;
while (gethrtime() <= end) {
int run_count = 100;
void *buf;
struct abd *abd_data, *abd_meta;
uint32_t size;
int *ptr;
int i;
zio_cksum_t zc_ref;
zio_cksum_t zc_ref_byteswap;
size = ztest_random_blocksize();
buf = umem_alloc(size, UMEM_NOFAIL);
abd_data = abd_alloc(size, B_FALSE);
abd_meta = abd_alloc(size, B_TRUE);
for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++)
*ptr = ztest_random(UINT_MAX);
abd_copy_from_buf_off(abd_data, buf, 0, size);
abd_copy_from_buf_off(abd_meta, buf, 0, size);
VERIFY0(fletcher_4_impl_set("scalar"));
fletcher_4_native(buf, size, NULL, &zc_ref);
fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap);
VERIFY0(fletcher_4_impl_set("cycle"));
while (run_count-- > 0) {
zio_cksum_t zc;
zio_cksum_t zc_byteswap;
fletcher_4_byteswap(buf, size, NULL, &zc_byteswap);
fletcher_4_native(buf, size, NULL, &zc);
VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc)));
VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap,
sizeof (zc_byteswap)));
/* Test ABD - data */
abd_fletcher_4_byteswap(abd_data, size, NULL,
&zc_byteswap);
abd_fletcher_4_native(abd_data, size, NULL, &zc);
VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc)));
VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap,
sizeof (zc_byteswap)));
/* Test ABD - metadata */
abd_fletcher_4_byteswap(abd_meta, size, NULL,
&zc_byteswap);
abd_fletcher_4_native(abd_meta, size, NULL, &zc);
VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc)));
VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap,
sizeof (zc_byteswap)));
}
umem_free(buf, size);
abd_free(abd_data);
abd_free(abd_meta);
}
}
void
ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
void *buf;
size_t size;
int *ptr;
int i;
zio_cksum_t zc_ref;
zio_cksum_t zc_ref_bswap;
hrtime_t end = gethrtime() + NANOSEC;
while (gethrtime() <= end) {
int run_count = 100;
size = ztest_random_blocksize();
buf = umem_alloc(size, UMEM_NOFAIL);
for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++)
*ptr = ztest_random(UINT_MAX);
VERIFY0(fletcher_4_impl_set("scalar"));
fletcher_4_native(buf, size, NULL, &zc_ref);
fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap);
VERIFY0(fletcher_4_impl_set("cycle"));
while (run_count-- > 0) {
zio_cksum_t zc;
zio_cksum_t zc_bswap;
size_t pos = 0;
ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0);
ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0);
while (pos < size) {
size_t inc = 64 * ztest_random(size / 67);
/* sometimes add few bytes to test non-simd */
if (ztest_random(100) < 10)
inc += P2ALIGN(ztest_random(64),
sizeof (uint32_t));
if (inc > (size - pos))
inc = size - pos;
fletcher_4_incremental_native(buf + pos, inc,
&zc);
fletcher_4_incremental_byteswap(buf + pos, inc,
&zc_bswap);
pos += inc;
}
VERIFY3U(pos, ==, size);
VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref));
VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap));
/*
* verify if incremental on the whole buffer is
* equivalent to non-incremental version
*/
ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0);
ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0);
fletcher_4_incremental_native(buf, size, &zc);
fletcher_4_incremental_byteswap(buf, size, &zc_bswap);
VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref));
VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap));
}
umem_free(buf, size);
}
}
static int
ztest_set_global_vars(void)
{
for (size_t i = 0; i < ztest_opts.zo_gvars_count; i++) {
char *kv = ztest_opts.zo_gvars[i];
VERIFY3U(strlen(kv), <=, ZO_GVARS_MAX_ARGLEN);
VERIFY3U(strlen(kv), >, 0);
int err = set_global_var(kv);
if (ztest_opts.zo_verbose > 0) {
(void) printf("setting global var %s ... %s\n", kv,
err ? "failed" : "ok");
}
if (err != 0) {
(void) fprintf(stderr,
"failed to set global var '%s'\n", kv);
return (err);
}
}
return (0);
}
static char **
ztest_global_vars_to_zdb_args(void)
{
char **args = calloc(2*ztest_opts.zo_gvars_count + 1, sizeof (char *));
char **cur = args;
for (size_t i = 0; i < ztest_opts.zo_gvars_count; i++) {
char *kv = ztest_opts.zo_gvars[i];
*cur = "-o";
cur++;
*cur = strdup(kv);
cur++;
}
ASSERT3P(cur, ==, &args[2*ztest_opts.zo_gvars_count]);
*cur = NULL;
return (args);
}
/* The end of strings is indicated by a NULL element */
static char *
join_strings(char **strings, const char *sep)
{
size_t totallen = 0;
for (char **sp = strings; *sp != NULL; sp++) {
totallen += strlen(*sp);
totallen += strlen(sep);
}
if (totallen > 0) {
ASSERT(totallen >= strlen(sep));
totallen -= strlen(sep);
}
size_t buflen = totallen + 1;
char *o = malloc(buflen); /* trailing 0 byte */
o[0] = '\0';
for (char **sp = strings; *sp != NULL; sp++) {
size_t would;
would = strlcat(o, *sp, buflen);
VERIFY3U(would, <, buflen);
if (*(sp+1) == NULL) {
break;
}
would = strlcat(o, sep, buflen);
VERIFY3U(would, <, buflen);
}
ASSERT3S(strlen(o), ==, totallen);
return (o);
}
static int
ztest_check_path(char *path)
{
struct stat s;
/* return true on success */
return (!stat(path, &s));
}
static void
ztest_get_zdb_bin(char *bin, int len)
{
char *zdb_path;
/*
* Try to use ZDB_PATH and in-tree zdb path. If not successful, just
* let popen to search through PATH.
*/
if ((zdb_path = getenv("ZDB_PATH"))) {
strlcpy(bin, zdb_path, len); /* In env */
if (!ztest_check_path(bin)) {
ztest_dump_core = 0;
fatal(B_TRUE, "invalid ZDB_PATH '%s'", bin);
}
return;
}
VERIFY3P(realpath(getexecname(), bin), !=, NULL);
if (strstr(bin, "/ztest/")) {
strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */
strcat(bin, "/zdb/zdb");
if (ztest_check_path(bin))
return;
}
strcpy(bin, "zdb");
}
static vdev_t *
ztest_random_concrete_vdev_leaf(vdev_t *vd)
{
if (vd == NULL)
return (NULL);
if (vd->vdev_children == 0)
return (vd);
vdev_t *eligible[vd->vdev_children];
int eligible_idx = 0, i;
for (i = 0; i < vd->vdev_children; i++) {
vdev_t *cvd = vd->vdev_child[i];
if (cvd->vdev_top->vdev_removing)
continue;
if (cvd->vdev_children > 0 ||
(vdev_is_concrete(cvd) && !cvd->vdev_detached)) {
eligible[eligible_idx++] = cvd;
}
}
VERIFY3S(eligible_idx, >, 0);
uint64_t child_no = ztest_random(eligible_idx);
return (ztest_random_concrete_vdev_leaf(eligible[child_no]));
}
void
ztest_initialize(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
spa_t *spa = ztest_spa;
int error = 0;
mutex_enter(&ztest_vdev_lock);
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
/* Random leaf vdev */
vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev);
if (rand_vd == NULL) {
spa_config_exit(spa, SCL_VDEV, FTAG);
mutex_exit(&ztest_vdev_lock);
return;
}
/*
* The random vdev we've selected may change as soon as we
* drop the spa_config_lock. We create local copies of things
* we're interested in.
*/
uint64_t guid = rand_vd->vdev_guid;
char *path = strdup(rand_vd->vdev_path);
boolean_t active = rand_vd->vdev_initialize_thread != NULL;
zfs_dbgmsg("vd %px, guid %llu", rand_vd, (u_longlong_t)guid);
spa_config_exit(spa, SCL_VDEV, FTAG);
uint64_t cmd = ztest_random(POOL_INITIALIZE_FUNCS);
nvlist_t *vdev_guids = fnvlist_alloc();
nvlist_t *vdev_errlist = fnvlist_alloc();
fnvlist_add_uint64(vdev_guids, path, guid);
error = spa_vdev_initialize(spa, vdev_guids, cmd, vdev_errlist);
fnvlist_free(vdev_guids);
fnvlist_free(vdev_errlist);
switch (cmd) {
case POOL_INITIALIZE_CANCEL:
if (ztest_opts.zo_verbose >= 4) {
(void) printf("Cancel initialize %s", path);
if (!active)
(void) printf(" failed (no initialize active)");
(void) printf("\n");
}
break;
case POOL_INITIALIZE_START:
if (ztest_opts.zo_verbose >= 4) {
(void) printf("Start initialize %s", path);
if (active && error == 0)
(void) printf(" failed (already active)");
else if (error != 0)
(void) printf(" failed (error %d)", error);
(void) printf("\n");
}
break;
case POOL_INITIALIZE_SUSPEND:
if (ztest_opts.zo_verbose >= 4) {
(void) printf("Suspend initialize %s", path);
if (!active)
(void) printf(" failed (no initialize active)");
(void) printf("\n");
}
break;
}
free(path);
mutex_exit(&ztest_vdev_lock);
}
void
ztest_trim(ztest_ds_t *zd, uint64_t id)
{
(void) zd, (void) id;
spa_t *spa = ztest_spa;
int error = 0;
mutex_enter(&ztest_vdev_lock);
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
/* Random leaf vdev */
vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev);
if (rand_vd == NULL) {
spa_config_exit(spa, SCL_VDEV, FTAG);
mutex_exit(&ztest_vdev_lock);
return;
}
/*
* The random vdev we've selected may change as soon as we
* drop the spa_config_lock. We create local copies of things
* we're interested in.
*/
uint64_t guid = rand_vd->vdev_guid;
char *path = strdup(rand_vd->vdev_path);
boolean_t active = rand_vd->vdev_trim_thread != NULL;
zfs_dbgmsg("vd %p, guid %llu", rand_vd, (u_longlong_t)guid);
spa_config_exit(spa, SCL_VDEV, FTAG);
uint64_t cmd = ztest_random(POOL_TRIM_FUNCS);
uint64_t rate = 1 << ztest_random(30);
boolean_t partial = (ztest_random(5) > 0);
boolean_t secure = (ztest_random(5) > 0);
nvlist_t *vdev_guids = fnvlist_alloc();
nvlist_t *vdev_errlist = fnvlist_alloc();
fnvlist_add_uint64(vdev_guids, path, guid);
error = spa_vdev_trim(spa, vdev_guids, cmd, rate, partial,
secure, vdev_errlist);
fnvlist_free(vdev_guids);
fnvlist_free(vdev_errlist);
switch (cmd) {
case POOL_TRIM_CANCEL:
if (ztest_opts.zo_verbose >= 4) {
(void) printf("Cancel TRIM %s", path);
if (!active)
(void) printf(" failed (no TRIM active)");
(void) printf("\n");
}
break;
case POOL_TRIM_START:
if (ztest_opts.zo_verbose >= 4) {
(void) printf("Start TRIM %s", path);
if (active && error == 0)
(void) printf(" failed (already active)");
else if (error != 0)
(void) printf(" failed (error %d)", error);
(void) printf("\n");
}
break;
case POOL_TRIM_SUSPEND:
if (ztest_opts.zo_verbose >= 4) {
(void) printf("Suspend TRIM %s", path);
if (!active)
(void) printf(" failed (no TRIM active)");
(void) printf("\n");
}
break;
}
free(path);
mutex_exit(&ztest_vdev_lock);
}
/*
* Verify pool integrity by running zdb.
*/
static void
ztest_run_zdb(char *pool)
{
int status;
char *bin;
char *zdb;
char *zbuf;
const int len = MAXPATHLEN + MAXNAMELEN + 20;
FILE *fp;
bin = umem_alloc(len, UMEM_NOFAIL);
zdb = umem_alloc(len, UMEM_NOFAIL);
zbuf = umem_alloc(1024, UMEM_NOFAIL);
ztest_get_zdb_bin(bin, len);
char **set_gvars_args = ztest_global_vars_to_zdb_args();
char *set_gvars_args_joined = join_strings(set_gvars_args, " ");
free(set_gvars_args);
size_t would = snprintf(zdb, len,
"%s -bcc%s%s -G -d -Y -e -y %s -p %s %s",
bin,
ztest_opts.zo_verbose >= 3 ? "s" : "",
ztest_opts.zo_verbose >= 4 ? "v" : "",
set_gvars_args_joined,
ztest_opts.zo_dir,
pool);
ASSERT3U(would, <, len);
free(set_gvars_args_joined);
if (ztest_opts.zo_verbose >= 5)
(void) printf("Executing %s\n", strstr(zdb, "zdb "));
fp = popen(zdb, "r");
while (fgets(zbuf, 1024, fp) != NULL)
if (ztest_opts.zo_verbose >= 3)
(void) printf("%s", zbuf);
status = pclose(fp);
if (status == 0)
goto out;
ztest_dump_core = 0;
if (WIFEXITED(status))
fatal(B_FALSE, "'%s' exit code %d", zdb, WEXITSTATUS(status));
else
fatal(B_FALSE, "'%s' died with signal %d",
zdb, WTERMSIG(status));
out:
umem_free(bin, len);
umem_free(zdb, len);
umem_free(zbuf, 1024);
}
static void
ztest_walk_pool_directory(char *header)
{
spa_t *spa = NULL;
if (ztest_opts.zo_verbose >= 6)
(void) printf("%s\n", header);
mutex_enter(&spa_namespace_lock);
while ((spa = spa_next(spa)) != NULL)
if (ztest_opts.zo_verbose >= 6)
(void) printf("\t%s\n", spa_name(spa));
mutex_exit(&spa_namespace_lock);
}
static void
ztest_spa_import_export(char *oldname, char *newname)
{
nvlist_t *config, *newconfig;
uint64_t pool_guid;
spa_t *spa;
int error;
if (ztest_opts.zo_verbose >= 4) {
(void) printf("import/export: old = %s, new = %s\n",
oldname, newname);
}
/*
* Clean up from previous runs.
*/
(void) spa_destroy(newname);
/*
* Get the pool's configuration and guid.
*/
VERIFY0(spa_open(oldname, &spa, FTAG));
/*
* Kick off a scrub to tickle scrub/export races.
*/
if (ztest_random(2) == 0)
(void) spa_scan(spa, POOL_SCAN_SCRUB);
pool_guid = spa_guid(spa);
spa_close(spa, FTAG);
ztest_walk_pool_directory("pools before export");
/*
* Export it.
*/
VERIFY0(spa_export(oldname, &config, B_FALSE, B_FALSE));
ztest_walk_pool_directory("pools after export");
/*
* Try to import it.
*/
newconfig = spa_tryimport(config);
ASSERT3P(newconfig, !=, NULL);
fnvlist_free(newconfig);
/*
* Import it under the new name.
*/
error = spa_import(newname, config, NULL, 0);
if (error != 0) {
dump_nvlist(config, 0);
fatal(B_FALSE, "couldn't import pool %s as %s: error %u",
oldname, newname, error);
}
ztest_walk_pool_directory("pools after import");
/*
* Try to import it again -- should fail with EEXIST.
*/
VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0));
/*
* Try to import it under a different name -- should fail with EEXIST.
*/
VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0));
/*
* Verify that the pool is no longer visible under the old name.
*/
VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG));
/*
* Verify that we can open and close the pool using the new name.
*/
VERIFY0(spa_open(newname, &spa, FTAG));
ASSERT3U(pool_guid, ==, spa_guid(spa));
spa_close(spa, FTAG);
fnvlist_free(config);
}
static void
ztest_resume(spa_t *spa)
{
if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6)
(void) printf("resuming from suspended state\n");
spa_vdev_state_enter(spa, SCL_NONE);
vdev_clear(spa, NULL);
(void) spa_vdev_state_exit(spa, NULL, 0);
(void) zio_resume(spa);
}
static void
ztest_resume_thread(void *arg)
{
spa_t *spa = arg;
while (!ztest_exiting) {
if (spa_suspended(spa))
ztest_resume(spa);
(void) poll(NULL, 0, 100);
/*
* Periodically change the zfs_compressed_arc_enabled setting.
*/
if (ztest_random(10) == 0)
zfs_compressed_arc_enabled = ztest_random(2);
/*
* Periodically change the zfs_abd_scatter_enabled setting.
*/
if (ztest_random(10) == 0)
zfs_abd_scatter_enabled = ztest_random(2);
}
thread_exit();
}
static void
ztest_deadman_thread(void *arg)
{
ztest_shared_t *zs = arg;
spa_t *spa = ztest_spa;
hrtime_t delay, overdue, last_run = gethrtime();
delay = (zs->zs_thread_stop - zs->zs_thread_start) +
MSEC2NSEC(zfs_deadman_synctime_ms);
while (!ztest_exiting) {
/*
* Wait for the delay timer while checking occasionally
* if we should stop.
*/
if (gethrtime() < last_run + delay) {
(void) poll(NULL, 0, 1000);
continue;
}
/*
* If the pool is suspended then fail immediately. Otherwise,
* check to see if the pool is making any progress. If
* vdev_deadman() discovers that there hasn't been any recent
* I/Os then it will end up aborting the tests.
*/
if (spa_suspended(spa) || spa->spa_root_vdev == NULL) {
fatal(B_FALSE,
"aborting test after %lu seconds because "
"pool has transitioned to a suspended state.",
zfs_deadman_synctime_ms / 1000);
}
vdev_deadman(spa->spa_root_vdev, FTAG);
/*
* If the process doesn't complete within a grace period of
* zfs_deadman_synctime_ms over the expected finish time,
* then it may be hung and is terminated.
*/
overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms);
if (gethrtime() > overdue) {
fatal(B_FALSE,
"aborting test after %llu seconds because "
"the process is overdue for termination.",
(gethrtime() - zs->zs_proc_start) / NANOSEC);
}
(void) printf("ztest has been running for %lld seconds\n",
(gethrtime() - zs->zs_proc_start) / NANOSEC);
last_run = gethrtime();
delay = MSEC2NSEC(zfs_deadman_checktime_ms);
}
thread_exit();
}
static void
ztest_execute(int test, ztest_info_t *zi, uint64_t id)
{
ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets];
ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test);
hrtime_t functime = gethrtime();
int i;
for (i = 0; i < zi->zi_iters; i++)
zi->zi_func(zd, id);
functime = gethrtime() - functime;
atomic_add_64(&zc->zc_count, 1);
atomic_add_64(&zc->zc_time, functime);
if (ztest_opts.zo_verbose >= 4)
(void) printf("%6.2f sec in %s\n",
(double)functime / NANOSEC, zi->zi_funcname);
}
static void
ztest_thread(void *arg)
{
int rand;
uint64_t id = (uintptr_t)arg;
ztest_shared_t *zs = ztest_shared;
uint64_t call_next;
hrtime_t now;
ztest_info_t *zi;
ztest_shared_callstate_t *zc;
while ((now = gethrtime()) < zs->zs_thread_stop) {
/*
* See if it's time to force a crash.
*/
if (now > zs->zs_thread_kill)
ztest_kill(zs);
/*
* If we're getting ENOSPC with some regularity, stop.
*/
if (zs->zs_enospc_count > 10)
break;
/*
* Pick a random function to execute.
*/
rand = ztest_random(ZTEST_FUNCS);
zi = &ztest_info[rand];
zc = ZTEST_GET_SHARED_CALLSTATE(rand);
call_next = zc->zc_next;
if (now >= call_next &&
atomic_cas_64(&zc->zc_next, call_next, call_next +
ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) {
ztest_execute(rand, zi, id);
}
}
thread_exit();
}
static void
ztest_dataset_name(char *dsname, char *pool, int d)
{
(void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d);
}
static void
ztest_dataset_destroy(int d)
{
char name[ZFS_MAX_DATASET_NAME_LEN];
int t;
ztest_dataset_name(name, ztest_opts.zo_pool, d);
if (ztest_opts.zo_verbose >= 3)
(void) printf("Destroying %s to free up space\n", name);
/*
* Cleanup any non-standard clones and snapshots. In general,
* ztest thread t operates on dataset (t % zopt_datasets),
* so there may be more than one thing to clean up.
*/
for (t = d; t < ztest_opts.zo_threads;
t += ztest_opts.zo_datasets)
ztest_dsl_dataset_cleanup(name, t);
(void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL,
DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN);
}
static void
ztest_dataset_dirobj_verify(ztest_ds_t *zd)
{
uint64_t usedobjs, dirobjs, scratch;
/*
* ZTEST_DIROBJ is the object directory for the entire dataset.
* Therefore, the number of objects in use should equal the
* number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself.
* If not, we have an object leak.
*
* Note that we can only check this in ztest_dataset_open(),
* when the open-context and syncing-context values agree.
* That's because zap_count() returns the open-context value,
* while dmu_objset_space() returns the rootbp fill count.
*/
VERIFY0(zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs));
dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch);
ASSERT3U(dirobjs + 1, ==, usedobjs);
}
static int
ztest_dataset_open(int d)
{
ztest_ds_t *zd = &ztest_ds[d];
uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq;
objset_t *os;
zilog_t *zilog;
char name[ZFS_MAX_DATASET_NAME_LEN];
int error;
ztest_dataset_name(name, ztest_opts.zo_pool, d);
(void) pthread_rwlock_rdlock(&ztest_name_lock);
error = ztest_dataset_create(name);
if (error == ENOSPC) {
(void) pthread_rwlock_unlock(&ztest_name_lock);
ztest_record_enospc(FTAG);
return (error);
}
ASSERT(error == 0 || error == EEXIST);
VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE,
B_TRUE, zd, &os));
(void) pthread_rwlock_unlock(&ztest_name_lock);
ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os);
zilog = zd->zd_zilog;
if (zilog->zl_header->zh_claim_lr_seq != 0 &&
zilog->zl_header->zh_claim_lr_seq < committed_seq)
fatal(B_FALSE, "missing log records: "
"claimed %"PRIu64" < committed %"PRIu64"",
zilog->zl_header->zh_claim_lr_seq, committed_seq);
ztest_dataset_dirobj_verify(zd);
zil_replay(os, zd, ztest_replay_vector);
ztest_dataset_dirobj_verify(zd);
if (ztest_opts.zo_verbose >= 6)
(void) printf("%s replay %"PRIu64" blocks, "
"%"PRIu64" records, seq %"PRIu64"\n",
zd->zd_name,
zilog->zl_parse_blk_count,
zilog->zl_parse_lr_count,
zilog->zl_replaying_seq);
zilog = zil_open(os, ztest_get_data);
if (zilog->zl_replaying_seq != 0 &&
zilog->zl_replaying_seq < committed_seq)
fatal(B_FALSE, "missing log records: "
"replayed %"PRIu64" < committed %"PRIu64"",
zilog->zl_replaying_seq, committed_seq);
return (0);
}
static void
ztest_dataset_close(int d)
{
ztest_ds_t *zd = &ztest_ds[d];
zil_close(zd->zd_zilog);
dmu_objset_disown(zd->zd_os, B_TRUE, zd);
ztest_zd_fini(zd);
}
static int
ztest_replay_zil_cb(const char *name, void *arg)
{
(void) arg;
objset_t *os;
ztest_ds_t *zdtmp;
VERIFY0(ztest_dmu_objset_own(name, DMU_OST_ANY, B_TRUE,
B_TRUE, FTAG, &os));
zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL);
ztest_zd_init(zdtmp, NULL, os);
zil_replay(os, zdtmp, ztest_replay_vector);
ztest_zd_fini(zdtmp);
if (dmu_objset_zil(os)->zl_parse_lr_count != 0 &&
ztest_opts.zo_verbose >= 6) {
zilog_t *zilog = dmu_objset_zil(os);
(void) printf("%s replay %"PRIu64" blocks, "
"%"PRIu64" records, seq %"PRIu64"\n",
name,
zilog->zl_parse_blk_count,
zilog->zl_parse_lr_count,
zilog->zl_replaying_seq);
}
umem_free(zdtmp, sizeof (ztest_ds_t));
dmu_objset_disown(os, B_TRUE, FTAG);
return (0);
}
static void
ztest_freeze(void)
{
ztest_ds_t *zd = &ztest_ds[0];
spa_t *spa;
int numloops = 0;
if (ztest_opts.zo_verbose >= 3)
(void) printf("testing spa_freeze()...\n");
kernel_init(SPA_MODE_READ | SPA_MODE_WRITE);
VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
VERIFY0(ztest_dataset_open(0));
ztest_spa = spa;
/*
* Force the first log block to be transactionally allocated.
* We have to do this before we freeze the pool -- otherwise
* the log chain won't be anchored.
*/
while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) {
ztest_dmu_object_alloc_free(zd, 0);
zil_commit(zd->zd_zilog, 0);
}
txg_wait_synced(spa_get_dsl(spa), 0);
/*
* Freeze the pool. This stops spa_sync() from doing anything,
* so that the only way to record changes from now on is the ZIL.
*/
spa_freeze(spa);
/*
* Because it is hard to predict how much space a write will actually
* require beforehand, we leave ourselves some fudge space to write over
* capacity.
*/
uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2;
/*
* Run tests that generate log records but don't alter the pool config
* or depend on DSL sync tasks (snapshots, objset create/destroy, etc).
* We do a txg_wait_synced() after each iteration to force the txg
* to increase well beyond the last synced value in the uberblock.
* The ZIL should be OK with that.
*
* Run a random number of times less than zo_maxloops and ensure we do
* not run out of space on the pool.
*/
while (ztest_random(10) != 0 &&
numloops++ < ztest_opts.zo_maxloops &&
metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) {
ztest_od_t od;
ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0);
VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE));
ztest_io(zd, od.od_object,
ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
txg_wait_synced(spa_get_dsl(spa), 0);
}
/*
* Commit all of the changes we just generated.
*/
zil_commit(zd->zd_zilog, 0);
txg_wait_synced(spa_get_dsl(spa), 0);
/*
* Close our dataset and close the pool.
*/
ztest_dataset_close(0);
spa_close(spa, FTAG);
kernel_fini();
/*
* Open and close the pool and dataset to induce log replay.
*/
kernel_init(SPA_MODE_READ | SPA_MODE_WRITE);
VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
ASSERT3U(spa_freeze_txg(spa), ==, UINT64_MAX);
VERIFY0(ztest_dataset_open(0));
ztest_spa = spa;
txg_wait_synced(spa_get_dsl(spa), 0);
ztest_dataset_close(0);
ztest_reguid(NULL, 0);
spa_close(spa, FTAG);
kernel_fini();
}
static void
ztest_import_impl(void)
{
importargs_t args = { 0 };
nvlist_t *cfg = NULL;
int nsearch = 1;
char *searchdirs[nsearch];
int flags = ZFS_IMPORT_MISSING_LOG;
searchdirs[0] = ztest_opts.zo_dir;
args.paths = nsearch;
args.path = searchdirs;
args.can_be_active = B_FALSE;
VERIFY0(zpool_find_config(NULL, ztest_opts.zo_pool, &cfg, &args,
&libzpool_config_ops));
VERIFY0(spa_import(ztest_opts.zo_pool, cfg, NULL, flags));
fnvlist_free(cfg);
}
/*
* Import a storage pool with the given name.
*/
static void
ztest_import(ztest_shared_t *zs)
{
spa_t *spa;
mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL);
VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));
kernel_init(SPA_MODE_READ | SPA_MODE_WRITE);
ztest_import_impl();
VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
zs->zs_metaslab_sz =
1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
spa_close(spa, FTAG);
kernel_fini();
if (!ztest_opts.zo_mmp_test) {
ztest_run_zdb(ztest_opts.zo_pool);
ztest_freeze();
ztest_run_zdb(ztest_opts.zo_pool);
}
(void) pthread_rwlock_destroy(&ztest_name_lock);
mutex_destroy(&ztest_vdev_lock);
mutex_destroy(&ztest_checkpoint_lock);
}
/*
* Kick off threads to run tests on all datasets in parallel.
*/
static void
ztest_run(ztest_shared_t *zs)
{
spa_t *spa;
objset_t *os;
kthread_t *resume_thread, *deadman_thread;
kthread_t **run_threads;
uint64_t object;
int error;
int t, d;
ztest_exiting = B_FALSE;
/*
* Initialize parent/child shared state.
*/
mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL);
VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));
zs->zs_thread_start = gethrtime();
zs->zs_thread_stop =
zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC;
zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop);
zs->zs_thread_kill = zs->zs_thread_stop;
if (ztest_random(100) < ztest_opts.zo_killrate) {
zs->zs_thread_kill -=
ztest_random(ztest_opts.zo_passtime * NANOSEC);
}
mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t),
offsetof(ztest_cb_data_t, zcd_node));
/*
* Open our pool. It may need to be imported first depending on
* what tests were running when the previous pass was terminated.
*/
kernel_init(SPA_MODE_READ | SPA_MODE_WRITE);
error = spa_open(ztest_opts.zo_pool, &spa, FTAG);
if (error) {
VERIFY3S(error, ==, ENOENT);
ztest_import_impl();
VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
zs->zs_metaslab_sz =
1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
}
metaslab_preload_limit = ztest_random(20) + 1;
ztest_spa = spa;
VERIFY0(vdev_raidz_impl_set("cycle"));
dmu_objset_stats_t dds;
VERIFY0(ztest_dmu_objset_own(ztest_opts.zo_pool,
DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os));
dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
dmu_objset_fast_stat(os, &dds);
dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
zs->zs_guid = dds.dds_guid;
dmu_objset_disown(os, B_TRUE, FTAG);
/*
* Create a thread to periodically resume suspended I/O.
*/
resume_thread = thread_create(NULL, 0, ztest_resume_thread,
spa, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri);
/*
* Create a deadman thread and set to panic if we hang.
*/
deadman_thread = thread_create(NULL, 0, ztest_deadman_thread,
zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri);
spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC;
/*
* Verify that we can safely inquire about any object,
* whether it's allocated or not. To make it interesting,
* we probe a 5-wide window around each power of two.
* This hits all edge cases, including zero and the max.
*/
for (t = 0; t < 64; t++) {
for (d = -5; d <= 5; d++) {
error = dmu_object_info(spa->spa_meta_objset,
(1ULL << t) + d, NULL);
ASSERT(error == 0 || error == ENOENT ||
error == EINVAL);
}
}
/*
* If we got any ENOSPC errors on the previous run, destroy something.
*/
if (zs->zs_enospc_count != 0) {
int d = ztest_random(ztest_opts.zo_datasets);
ztest_dataset_destroy(d);
}
zs->zs_enospc_count = 0;
/*
* If we were in the middle of ztest_device_removal() and were killed
* we need to ensure the removal and scrub complete before running
* any tests that check ztest_device_removal_active. The removal will
* be restarted automatically when the spa is opened, but we need to
* initiate the scrub manually if it is not already in progress. Note
* that we always run the scrub whenever an indirect vdev exists
* because we have no way of knowing for sure if ztest_device_removal()
* fully completed its scrub before the pool was reimported.
*/
if (spa->spa_removing_phys.sr_state == DSS_SCANNING ||
spa->spa_removing_phys.sr_prev_indirect_vdev != -1) {
while (spa->spa_removing_phys.sr_state == DSS_SCANNING)
txg_wait_synced(spa_get_dsl(spa), 0);
error = ztest_scrub_impl(spa);
if (error == EBUSY)
error = 0;
ASSERT0(error);
}
run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *),
UMEM_NOFAIL);
if (ztest_opts.zo_verbose >= 4)
(void) printf("starting main threads...\n");
/*
* Replay all logs of all datasets in the pool. This is primarily for
* temporary datasets which wouldn't otherwise get replayed, which
* can trigger failures when attempting to offline a SLOG in
* ztest_fault_inject().
*/
(void) dmu_objset_find(ztest_opts.zo_pool, ztest_replay_zil_cb,
NULL, DS_FIND_CHILDREN);
/*
* Kick off all the tests that run in parallel.
*/
for (t = 0; t < ztest_opts.zo_threads; t++) {
if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) {
umem_free(run_threads, ztest_opts.zo_threads *
sizeof (kthread_t *));
return;
}
run_threads[t] = thread_create(NULL, 0, ztest_thread,
(void *)(uintptr_t)t, 0, NULL, TS_RUN | TS_JOINABLE,
defclsyspri);
}
/*
* Wait for all of the tests to complete.
*/
for (t = 0; t < ztest_opts.zo_threads; t++)
VERIFY0(thread_join(run_threads[t]));
/*
* Close all datasets. This must be done after all the threads
* are joined so we can be sure none of the datasets are in-use
* by any of the threads.
*/
for (t = 0; t < ztest_opts.zo_threads; t++) {
if (t < ztest_opts.zo_datasets)
ztest_dataset_close(t);
}
txg_wait_synced(spa_get_dsl(spa), 0);
zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa));
zs->zs_space = metaslab_class_get_space(spa_normal_class(spa));
umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *));
/* Kill the resume and deadman threads */
ztest_exiting = B_TRUE;
VERIFY0(thread_join(resume_thread));
VERIFY0(thread_join(deadman_thread));
ztest_resume(spa);
/*
* Right before closing the pool, kick off a bunch of async I/O;
* spa_close() should wait for it to complete.
*/
for (object = 1; object < 50; object++) {
dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20,
ZIO_PRIORITY_SYNC_READ);
}
/* Verify that at least one commit cb was called in a timely fashion */
if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG)
VERIFY0(zc_min_txg_delay);
spa_close(spa, FTAG);
/*
* Verify that we can loop over all pools.
*/
mutex_enter(&spa_namespace_lock);
for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa))
if (ztest_opts.zo_verbose > 3)
(void) printf("spa_next: found %s\n", spa_name(spa));
mutex_exit(&spa_namespace_lock);
/*
* Verify that we can export the pool and reimport it under a
* different name.
*/
if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) {
char name[ZFS_MAX_DATASET_NAME_LEN];
(void) snprintf(name, sizeof (name), "%s_import",
ztest_opts.zo_pool);
ztest_spa_import_export(ztest_opts.zo_pool, name);
ztest_spa_import_export(name, ztest_opts.zo_pool);
}
kernel_fini();
list_destroy(&zcl.zcl_callbacks);
mutex_destroy(&zcl.zcl_callbacks_lock);
(void) pthread_rwlock_destroy(&ztest_name_lock);
mutex_destroy(&ztest_vdev_lock);
mutex_destroy(&ztest_checkpoint_lock);
}
static void
print_time(hrtime_t t, char *timebuf)
{
hrtime_t s = t / NANOSEC;
hrtime_t m = s / 60;
hrtime_t h = m / 60;
hrtime_t d = h / 24;
s -= m * 60;
m -= h * 60;
h -= d * 24;
timebuf[0] = '\0';
if (d)
(void) sprintf(timebuf,
"%llud%02lluh%02llum%02llus", d, h, m, s);
else if (h)
(void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s);
else if (m)
(void) sprintf(timebuf, "%llum%02llus", m, s);
else
(void) sprintf(timebuf, "%llus", s);
}
static nvlist_t *
make_random_props(void)
{
nvlist_t *props;
props = fnvlist_alloc();
if (ztest_random(2) == 0)
return (props);
fnvlist_add_uint64(props,
zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1);
return (props);
}
/*
* Create a storage pool with the given name and initial vdev size.
* Then test spa_freeze() functionality.
*/
static void
ztest_init(ztest_shared_t *zs)
{
spa_t *spa;
nvlist_t *nvroot, *props;
int i;
mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL);
VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));
kernel_init(SPA_MODE_READ | SPA_MODE_WRITE);
/*
* Create the storage pool.
*/
(void) spa_destroy(ztest_opts.zo_pool);
ztest_shared->zs_vdev_next_leaf = 0;
zs->zs_splits = 0;
zs->zs_mirrors = ztest_opts.zo_mirrors;
nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0,
NULL, ztest_opts.zo_raid_children, zs->zs_mirrors, 1);
props = make_random_props();
/*
* We don't expect the pool to suspend unless maxfaults == 0,
* in which case ztest_fault_inject() temporarily takes away
* the only valid replica.
*/
fnvlist_add_uint64(props,
zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE),
MAXFAULTS(zs) ? ZIO_FAILURE_MODE_PANIC : ZIO_FAILURE_MODE_WAIT);
for (i = 0; i < SPA_FEATURES; i++) {
char *buf;
if (!spa_feature_table[i].fi_zfs_mod_supported)
continue;
/*
* 75% chance of using the log space map feature. We want ztest
* to exercise both the code paths that use the log space map
* feature and the ones that don't.
*/
if (i == SPA_FEATURE_LOG_SPACEMAP && ztest_random(4) == 0)
continue;
VERIFY3S(-1, !=, asprintf(&buf, "feature@%s",
spa_feature_table[i].fi_uname));
fnvlist_add_uint64(props, buf, 0);
free(buf);
}
VERIFY0(spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL));
fnvlist_free(nvroot);
fnvlist_free(props);
VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
zs->zs_metaslab_sz =
1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
spa_close(spa, FTAG);
kernel_fini();
if (!ztest_opts.zo_mmp_test) {
ztest_run_zdb(ztest_opts.zo_pool);
ztest_freeze();
ztest_run_zdb(ztest_opts.zo_pool);
}
(void) pthread_rwlock_destroy(&ztest_name_lock);
mutex_destroy(&ztest_vdev_lock);
mutex_destroy(&ztest_checkpoint_lock);
}
static void
setup_data_fd(void)
{
static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX";
ztest_fd_data = mkstemp(ztest_name_data);
ASSERT3S(ztest_fd_data, >=, 0);
(void) unlink(ztest_name_data);
}
static int
shared_data_size(ztest_shared_hdr_t *hdr)
{
int size;
size = hdr->zh_hdr_size;
size += hdr->zh_opts_size;
size += hdr->zh_size;
size += hdr->zh_stats_size * hdr->zh_stats_count;
size += hdr->zh_ds_size * hdr->zh_ds_count;
return (size);
}
static void
setup_hdr(void)
{
int size;
ztest_shared_hdr_t *hdr;
hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()),
PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0);
ASSERT3P(hdr, !=, MAP_FAILED);
VERIFY0(ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t)));
hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t);
hdr->zh_opts_size = sizeof (ztest_shared_opts_t);
hdr->zh_size = sizeof (ztest_shared_t);
hdr->zh_stats_size = sizeof (ztest_shared_callstate_t);
hdr->zh_stats_count = ZTEST_FUNCS;
hdr->zh_ds_size = sizeof (ztest_shared_ds_t);
hdr->zh_ds_count = ztest_opts.zo_datasets;
size = shared_data_size(hdr);
VERIFY0(ftruncate(ztest_fd_data, size));
(void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize()));
}
static void
setup_data(void)
{
int size, offset;
ztest_shared_hdr_t *hdr;
uint8_t *buf;
hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()),
PROT_READ, MAP_SHARED, ztest_fd_data, 0);
ASSERT3P(hdr, !=, MAP_FAILED);
size = shared_data_size(hdr);
(void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize()));
hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()),
PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0);
ASSERT3P(hdr, !=, MAP_FAILED);
buf = (uint8_t *)hdr;
offset = hdr->zh_hdr_size;
ztest_shared_opts = (void *)&buf[offset];
offset += hdr->zh_opts_size;
ztest_shared = (void *)&buf[offset];
offset += hdr->zh_size;
ztest_shared_callstate = (void *)&buf[offset];
offset += hdr->zh_stats_size * hdr->zh_stats_count;
ztest_shared_ds = (void *)&buf[offset];
}
static boolean_t
exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp)
{
pid_t pid;
int status;
char *cmdbuf = NULL;
pid = fork();
if (cmd == NULL) {
cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
(void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN);
cmd = cmdbuf;
}
if (pid == -1)
fatal(B_TRUE, "fork failed");
if (pid == 0) { /* child */
char *emptyargv[2] = { cmd, NULL };
char fd_data_str[12];
struct rlimit rl = { 1024, 1024 };
(void) setrlimit(RLIMIT_NOFILE, &rl);
(void) close(ztest_fd_rand);
VERIFY3S(11, >=,
snprintf(fd_data_str, 12, "%d", ztest_fd_data));
VERIFY0(setenv("ZTEST_FD_DATA", fd_data_str, 1));
(void) enable_extended_FILE_stdio(-1, -1);
if (libpath != NULL)
VERIFY0(setenv("LD_LIBRARY_PATH", libpath, 1));
(void) execv(cmd, emptyargv);
ztest_dump_core = B_FALSE;
fatal(B_TRUE, "exec failed: %s", cmd);
}
if (cmdbuf != NULL) {
umem_free(cmdbuf, MAXPATHLEN);
cmd = NULL;
}
while (waitpid(pid, &status, 0) != pid)
continue;
if (statusp != NULL)
*statusp = status;
if (WIFEXITED(status)) {
if (WEXITSTATUS(status) != 0) {
(void) fprintf(stderr, "child exited with code %d\n",
WEXITSTATUS(status));
exit(2);
}
return (B_FALSE);
} else if (WIFSIGNALED(status)) {
if (!ignorekill || WTERMSIG(status) != SIGKILL) {
(void) fprintf(stderr, "child died with signal %d\n",
WTERMSIG(status));
exit(3);
}
return (B_TRUE);
} else {
(void) fprintf(stderr, "something strange happened to child\n");
exit(4);
}
}
static void
ztest_run_init(void)
{
int i;
ztest_shared_t *zs = ztest_shared;
/*
* Blow away any existing copy of zpool.cache
*/
(void) remove(spa_config_path);
if (ztest_opts.zo_init == 0) {
if (ztest_opts.zo_verbose >= 1)
(void) printf("Importing pool %s\n",
ztest_opts.zo_pool);
ztest_import(zs);
return;
}
/*
* Create and initialize our storage pool.
*/
for (i = 1; i <= ztest_opts.zo_init; i++) {
bzero(zs, sizeof (ztest_shared_t));
if (ztest_opts.zo_verbose >= 3 &&
ztest_opts.zo_init != 1) {
(void) printf("ztest_init(), pass %d\n", i);
}
ztest_init(zs);
}
}
int
main(int argc, char **argv)
{
int kills = 0;
int iters = 0;
int older = 0;
int newer = 0;
ztest_shared_t *zs;
ztest_info_t *zi;
ztest_shared_callstate_t *zc;
char timebuf[100];
char numbuf[NN_NUMBUF_SZ];
char *cmd;
boolean_t hasalt;
int f, err;
char *fd_data_str = getenv("ZTEST_FD_DATA");
struct sigaction action;
(void) setvbuf(stdout, NULL, _IOLBF, 0);
dprintf_setup(&argc, argv);
zfs_deadman_synctime_ms = 300000;
zfs_deadman_checktime_ms = 30000;
/*
* As two-word space map entries may not come up often (especially
* if pool and vdev sizes are small) we want to force at least some
* of them so the feature get tested.
*/
zfs_force_some_double_word_sm_entries = B_TRUE;
/*
* Verify that even extensively damaged split blocks with many
* segments can be reconstructed in a reasonable amount of time
* when reconstruction is known to be possible.
*
* Note: the lower this value is, the more damage we inflict, and
* the more time ztest spends in recovering that damage. We chose
* to induce damage 1/100th of the time so recovery is tested but
* not so frequently that ztest doesn't get to test other code paths.
*/
zfs_reconstruct_indirect_damage_fraction = 100;
action.sa_handler = sig_handler;
sigemptyset(&action.sa_mask);
action.sa_flags = 0;
if (sigaction(SIGSEGV, &action, NULL) < 0) {
(void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n",
strerror(errno));
exit(EXIT_FAILURE);
}
if (sigaction(SIGABRT, &action, NULL) < 0) {
(void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n",
strerror(errno));
exit(EXIT_FAILURE);
}
/*
* Force random_get_bytes() to use /dev/urandom in order to prevent
* ztest from needlessly depleting the system entropy pool.
*/
random_path = "/dev/urandom";
ztest_fd_rand = open(random_path, O_RDONLY);
ASSERT3S(ztest_fd_rand, >=, 0);
if (!fd_data_str) {
process_options(argc, argv);
setup_data_fd();
setup_hdr();
setup_data();
bcopy(&ztest_opts, ztest_shared_opts,
sizeof (*ztest_shared_opts));
} else {
ztest_fd_data = atoi(fd_data_str);
setup_data();
bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts));
}
ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count);
err = ztest_set_global_vars();
if (err != 0 && !fd_data_str) {
/* error message done by ztest_set_global_vars */
exit(EXIT_FAILURE);
} else {
/* children should not be spawned if setting gvars fails */
VERIFY3S(err, ==, 0);
}
/* Override location of zpool.cache */
VERIFY3S(asprintf((char **)&spa_config_path, "%s/zpool.cache",
ztest_opts.zo_dir), !=, -1);
ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t),
UMEM_NOFAIL);
zs = ztest_shared;
if (fd_data_str) {
metaslab_force_ganging = ztest_opts.zo_metaslab_force_ganging;
metaslab_df_alloc_threshold =
zs->zs_metaslab_df_alloc_threshold;
if (zs->zs_do_init)
ztest_run_init();
else
ztest_run(zs);
exit(0);
}
hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0);
if (ztest_opts.zo_verbose >= 1) {
(void) printf("%"PRIu64" vdevs, %d datasets, %d threads,"
"%d %s disks, %"PRIu64" seconds...\n\n",
ztest_opts.zo_vdevs,
ztest_opts.zo_datasets,
ztest_opts.zo_threads,
ztest_opts.zo_raid_children,
ztest_opts.zo_raid_type,
ztest_opts.zo_time);
}
cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL);
(void) strlcpy(cmd, getexecname(), MAXNAMELEN);
zs->zs_do_init = B_TRUE;
if (strlen(ztest_opts.zo_alt_ztest) != 0) {
if (ztest_opts.zo_verbose >= 1) {
(void) printf("Executing older ztest for "
"initialization: %s\n", ztest_opts.zo_alt_ztest);
}
VERIFY(!exec_child(ztest_opts.zo_alt_ztest,
ztest_opts.zo_alt_libpath, B_FALSE, NULL));
} else {
VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL));
}
zs->zs_do_init = B_FALSE;
zs->zs_proc_start = gethrtime();
zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC;
for (f = 0; f < ZTEST_FUNCS; f++) {
zi = &ztest_info[f];
zc = ZTEST_GET_SHARED_CALLSTATE(f);
if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop)
zc->zc_next = UINT64_MAX;
else
zc->zc_next = zs->zs_proc_start +
ztest_random(2 * zi->zi_interval[0] + 1);
}
/*
* Run the tests in a loop. These tests include fault injection
* to verify that self-healing data works, and forced crashes
* to verify that we never lose on-disk consistency.
*/
while (gethrtime() < zs->zs_proc_stop) {
int status;
boolean_t killed;
/*
* Initialize the workload counters for each function.
*/
for (f = 0; f < ZTEST_FUNCS; f++) {
zc = ZTEST_GET_SHARED_CALLSTATE(f);
zc->zc_count = 0;
zc->zc_time = 0;
}
/* Set the allocation switch size */
zs->zs_metaslab_df_alloc_threshold =
ztest_random(zs->zs_metaslab_sz / 4) + 1;
if (!hasalt || ztest_random(2) == 0) {
if (hasalt && ztest_opts.zo_verbose >= 1) {
(void) printf("Executing newer ztest: %s\n",
cmd);
}
newer++;
killed = exec_child(cmd, NULL, B_TRUE, &status);
} else {
if (hasalt && ztest_opts.zo_verbose >= 1) {
(void) printf("Executing older ztest: %s\n",
ztest_opts.zo_alt_ztest);
}
older++;
killed = exec_child(ztest_opts.zo_alt_ztest,
ztest_opts.zo_alt_libpath, B_TRUE, &status);
}
if (killed)
kills++;
iters++;
if (ztest_opts.zo_verbose >= 1) {
hrtime_t now = gethrtime();
now = MIN(now, zs->zs_proc_stop);
print_time(zs->zs_proc_stop - now, timebuf);
nicenum(zs->zs_space, numbuf, sizeof (numbuf));
(void) printf("Pass %3d, %8s, %3"PRIu64" ENOSPC, "
"%4.1f%% of %5s used, %3.0f%% done, %8s to go\n",
iters,
WIFEXITED(status) ? "Complete" : "SIGKILL",
zs->zs_enospc_count,
100.0 * zs->zs_alloc / zs->zs_space,
numbuf,
100.0 * (now - zs->zs_proc_start) /
(ztest_opts.zo_time * NANOSEC), timebuf);
}
if (ztest_opts.zo_verbose >= 2) {
(void) printf("\nWorkload summary:\n\n");
(void) printf("%7s %9s %s\n",
"Calls", "Time", "Function");
(void) printf("%7s %9s %s\n",
"-----", "----", "--------");
for (f = 0; f < ZTEST_FUNCS; f++) {
zi = &ztest_info[f];
zc = ZTEST_GET_SHARED_CALLSTATE(f);
print_time(zc->zc_time, timebuf);
(void) printf("%7"PRIu64" %9s %s\n",
zc->zc_count, timebuf,
zi->zi_funcname);
}
(void) printf("\n");
}
if (!ztest_opts.zo_mmp_test)
ztest_run_zdb(ztest_opts.zo_pool);
}
if (ztest_opts.zo_verbose >= 1) {
if (hasalt) {
(void) printf("%d runs of older ztest: %s\n", older,
ztest_opts.zo_alt_ztest);
(void) printf("%d runs of newer ztest: %s\n", newer,
cmd);
}
(void) printf("%d killed, %d completed, %.0f%% kill rate\n",
kills, iters - kills, (100.0 * kills) / MAX(1, iters));
}
umem_free(cmd, MAXNAMELEN);
return (0);
}
diff --git a/include/libzfs.h b/include/libzfs.h
index e135ae2ee066..fe70a5b3a6fd 100644
--- a/include/libzfs.h
+++ b/include/libzfs.h
@@ -1,999 +1,999 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright Joyent, Inc.
* Copyright (c) 2013 Steven Hartland. All rights reserved.
* Copyright (c) 2016, Intel Corporation.
* Copyright 2016 Nexenta Systems, Inc.
* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
* Copyright (c) 2019 Datto Inc.
* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
*/
#ifndef _LIBZFS_H
#define _LIBZFS_H extern __attribute__((visibility("default")))
#include <assert.h>
#include <libnvpair.h>
#include <sys/mnttab.h>
#include <sys/param.h>
#include <sys/types.h>
#include <sys/varargs.h>
#include <sys/fs/zfs.h>
#include <sys/avl.h>
#include <ucred.h>
#include <libzfs_core.h>
#ifdef __cplusplus
extern "C" {
#endif
/*
* Miscellaneous ZFS constants
*/
#define ZFS_MAXPROPLEN MAXPATHLEN
#define ZPOOL_MAXPROPLEN MAXPATHLEN
/*
* libzfs errors
*/
typedef enum zfs_error {
EZFS_SUCCESS = 0, /* no error -- success */
EZFS_NOMEM = 2000, /* out of memory */
EZFS_BADPROP, /* invalid property value */
EZFS_PROPREADONLY, /* cannot set readonly property */
EZFS_PROPTYPE, /* property does not apply to dataset type */
EZFS_PROPNONINHERIT, /* property is not inheritable */
EZFS_PROPSPACE, /* bad quota or reservation */
EZFS_BADTYPE, /* dataset is not of appropriate type */
EZFS_BUSY, /* pool or dataset is busy */
EZFS_EXISTS, /* pool or dataset already exists */
EZFS_NOENT, /* no such pool or dataset */
EZFS_BADSTREAM, /* bad backup stream */
EZFS_DSREADONLY, /* dataset is readonly */
EZFS_VOLTOOBIG, /* volume is too large for 32-bit system */
EZFS_INVALIDNAME, /* invalid dataset name */
EZFS_BADRESTORE, /* unable to restore to destination */
EZFS_BADBACKUP, /* backup failed */
EZFS_BADTARGET, /* bad attach/detach/replace target */
EZFS_NODEVICE, /* no such device in pool */
EZFS_BADDEV, /* invalid device to add */
EZFS_NOREPLICAS, /* no valid replicas */
EZFS_RESILVERING, /* resilvering (healing reconstruction) */
EZFS_BADVERSION, /* unsupported version */
EZFS_POOLUNAVAIL, /* pool is currently unavailable */
EZFS_DEVOVERFLOW, /* too many devices in one vdev */
EZFS_BADPATH, /* must be an absolute path */
EZFS_CROSSTARGET, /* rename or clone across pool or dataset */
EZFS_ZONED, /* used improperly in local zone */
EZFS_MOUNTFAILED, /* failed to mount dataset */
EZFS_UMOUNTFAILED, /* failed to unmount dataset */
EZFS_UNSHARENFSFAILED, /* failed to unshare over nfs */
EZFS_SHARENFSFAILED, /* failed to share over nfs */
EZFS_PERM, /* permission denied */
EZFS_NOSPC, /* out of space */
EZFS_FAULT, /* bad address */
EZFS_IO, /* I/O error */
EZFS_INTR, /* signal received */
EZFS_ISSPARE, /* device is a hot spare */
EZFS_INVALCONFIG, /* invalid vdev configuration */
EZFS_RECURSIVE, /* recursive dependency */
EZFS_NOHISTORY, /* no history object */
EZFS_POOLPROPS, /* couldn't retrieve pool props */
EZFS_POOL_NOTSUP, /* ops not supported for this type of pool */
EZFS_POOL_INVALARG, /* invalid argument for this pool operation */
EZFS_NAMETOOLONG, /* dataset name is too long */
EZFS_OPENFAILED, /* open of device failed */
EZFS_NOCAP, /* couldn't get capacity */
EZFS_LABELFAILED, /* write of label failed */
EZFS_BADWHO, /* invalid permission who */
EZFS_BADPERM, /* invalid permission */
EZFS_BADPERMSET, /* invalid permission set name */
EZFS_NODELEGATION, /* delegated administration is disabled */
EZFS_UNSHARESMBFAILED, /* failed to unshare over smb */
EZFS_SHARESMBFAILED, /* failed to share over smb */
EZFS_BADCACHE, /* bad cache file */
EZFS_ISL2CACHE, /* device is for the level 2 ARC */
EZFS_VDEVNOTSUP, /* unsupported vdev type */
EZFS_NOTSUP, /* ops not supported on this dataset */
EZFS_ACTIVE_SPARE, /* pool has active shared spare devices */
EZFS_UNPLAYED_LOGS, /* log device has unplayed logs */
EZFS_REFTAG_RELE, /* snapshot release: tag not found */
EZFS_REFTAG_HOLD, /* snapshot hold: tag already exists */
EZFS_TAGTOOLONG, /* snapshot hold/rele: tag too long */
EZFS_PIPEFAILED, /* pipe create failed */
EZFS_THREADCREATEFAILED, /* thread create failed */
EZFS_POSTSPLIT_ONLINE, /* onlining a disk after splitting it */
EZFS_SCRUBBING, /* currently scrubbing */
EZFS_NO_SCRUB, /* no active scrub */
EZFS_DIFF, /* general failure of zfs diff */
EZFS_DIFFDATA, /* bad zfs diff data */
EZFS_POOLREADONLY, /* pool is in read-only mode */
EZFS_SCRUB_PAUSED, /* scrub currently paused */
EZFS_ACTIVE_POOL, /* pool is imported on a different system */
EZFS_CRYPTOFAILED, /* failed to setup encryption */
EZFS_NO_PENDING, /* cannot cancel, no operation is pending */
EZFS_CHECKPOINT_EXISTS, /* checkpoint exists */
EZFS_DISCARDING_CHECKPOINT, /* currently discarding a checkpoint */
EZFS_NO_CHECKPOINT, /* pool has no checkpoint */
EZFS_DEVRM_IN_PROGRESS, /* a device is currently being removed */
EZFS_VDEV_TOO_BIG, /* a device is too big to be used */
EZFS_IOC_NOTSUPPORTED, /* operation not supported by zfs module */
EZFS_TOOMANY, /* argument list too long */
EZFS_INITIALIZING, /* currently initializing */
EZFS_NO_INITIALIZE, /* no active initialize */
EZFS_WRONG_PARENT, /* invalid parent dataset (e.g ZVOL) */
EZFS_TRIMMING, /* currently trimming */
EZFS_NO_TRIM, /* no active trim */
EZFS_TRIM_NOTSUP, /* device does not support trim */
EZFS_NO_RESILVER_DEFER, /* pool doesn't support resilver_defer */
EZFS_EXPORT_IN_PROGRESS, /* currently exporting the pool */
EZFS_REBUILDING, /* resilvering (sequential reconstrution) */
EZFS_VDEV_NOTSUP, /* ops not supported for this type of vdev */
EZFS_UNKNOWN
} zfs_error_t;
/*
* The following data structures are all part
* of the zfs_allow_t data structure which is
* used for printing 'allow' permissions.
* It is a linked list of zfs_allow_t's which
* then contain avl tree's for user/group/sets/...
* and each one of the entries in those trees have
* avl tree's for the permissions they belong to and
* whether they are local,descendent or local+descendent
* permissions. The AVL trees are used primarily for
* sorting purposes, but also so that we can quickly find
* a given user and or permission.
*/
typedef struct zfs_perm_node {
avl_node_t z_node;
char z_pname[MAXPATHLEN];
} zfs_perm_node_t;
typedef struct zfs_allow_node {
avl_node_t z_node;
char z_key[MAXPATHLEN]; /* name, such as joe */
avl_tree_t z_localdescend; /* local+descendent perms */
avl_tree_t z_local; /* local permissions */
avl_tree_t z_descend; /* descendent permissions */
} zfs_allow_node_t;
typedef struct zfs_allow {
struct zfs_allow *z_next;
char z_setpoint[MAXPATHLEN];
avl_tree_t z_sets;
avl_tree_t z_crperms;
avl_tree_t z_user;
avl_tree_t z_group;
avl_tree_t z_everyone;
} zfs_allow_t;
/*
* Basic handle types
*/
typedef struct zfs_handle zfs_handle_t;
typedef struct zpool_handle zpool_handle_t;
typedef struct libzfs_handle libzfs_handle_t;
_LIBZFS_H int zpool_wait(zpool_handle_t *, zpool_wait_activity_t);
_LIBZFS_H int zpool_wait_status(zpool_handle_t *, zpool_wait_activity_t,
boolean_t *, boolean_t *);
/*
* Library initialization
*/
_LIBZFS_H libzfs_handle_t *libzfs_init(void);
_LIBZFS_H void libzfs_fini(libzfs_handle_t *);
_LIBZFS_H libzfs_handle_t *zpool_get_handle(zpool_handle_t *);
_LIBZFS_H libzfs_handle_t *zfs_get_handle(zfs_handle_t *);
_LIBZFS_H void libzfs_print_on_error(libzfs_handle_t *, boolean_t);
_LIBZFS_H void zfs_save_arguments(int argc, char **, char *, int);
_LIBZFS_H int zpool_log_history(libzfs_handle_t *, const char *);
_LIBZFS_H int libzfs_errno(libzfs_handle_t *);
_LIBZFS_H const char *libzfs_error_init(int);
_LIBZFS_H const char *libzfs_error_action(libzfs_handle_t *);
_LIBZFS_H const char *libzfs_error_description(libzfs_handle_t *);
_LIBZFS_H int zfs_standard_error(libzfs_handle_t *, int, const char *);
_LIBZFS_H void libzfs_mnttab_init(libzfs_handle_t *);
_LIBZFS_H void libzfs_mnttab_fini(libzfs_handle_t *);
_LIBZFS_H void libzfs_mnttab_cache(libzfs_handle_t *, boolean_t);
_LIBZFS_H int libzfs_mnttab_find(libzfs_handle_t *, const char *,
struct mnttab *);
_LIBZFS_H void libzfs_mnttab_add(libzfs_handle_t *, const char *,
const char *, const char *);
_LIBZFS_H void libzfs_mnttab_remove(libzfs_handle_t *, const char *);
/*
* Basic handle functions
*/
_LIBZFS_H zpool_handle_t *zpool_open(libzfs_handle_t *, const char *);
_LIBZFS_H zpool_handle_t *zpool_open_canfail(libzfs_handle_t *, const char *);
_LIBZFS_H void zpool_close(zpool_handle_t *);
_LIBZFS_H const char *zpool_get_name(zpool_handle_t *);
_LIBZFS_H int zpool_get_state(zpool_handle_t *);
_LIBZFS_H const char *zpool_state_to_name(vdev_state_t, vdev_aux_t);
_LIBZFS_H const char *zpool_pool_state_to_name(pool_state_t);
_LIBZFS_H void zpool_free_handles(libzfs_handle_t *);
/*
* Iterate over all active pools in the system.
*/
typedef int (*zpool_iter_f)(zpool_handle_t *, void *);
_LIBZFS_H int zpool_iter(libzfs_handle_t *, zpool_iter_f, void *);
_LIBZFS_H boolean_t zpool_skip_pool(const char *);
/*
* Functions to create and destroy pools
*/
_LIBZFS_H int zpool_create(libzfs_handle_t *, const char *, nvlist_t *,
nvlist_t *, nvlist_t *);
_LIBZFS_H int zpool_destroy(zpool_handle_t *, const char *);
_LIBZFS_H int zpool_add(zpool_handle_t *, nvlist_t *);
typedef struct splitflags {
/* do not split, but return the config that would be split off */
int dryrun : 1;
/* after splitting, import the pool */
int import : 1;
int name_flags;
} splitflags_t;
typedef struct trimflags {
/* requested vdevs are for the entire pool */
boolean_t fullpool;
/* request a secure trim, requires support from device */
boolean_t secure;
/* after starting trim, block until trim completes */
boolean_t wait;
/* trim at the requested rate in bytes/second */
uint64_t rate;
} trimflags_t;
/*
* Functions to manipulate pool and vdev state
*/
_LIBZFS_H int zpool_scan(zpool_handle_t *, pool_scan_func_t, pool_scrub_cmd_t);
_LIBZFS_H int zpool_initialize(zpool_handle_t *, pool_initialize_func_t,
nvlist_t *);
_LIBZFS_H int zpool_initialize_wait(zpool_handle_t *, pool_initialize_func_t,
nvlist_t *);
_LIBZFS_H int zpool_trim(zpool_handle_t *, pool_trim_func_t, nvlist_t *,
trimflags_t *);
_LIBZFS_H int zpool_clear(zpool_handle_t *, const char *, nvlist_t *);
_LIBZFS_H int zpool_reguid(zpool_handle_t *);
_LIBZFS_H int zpool_reopen_one(zpool_handle_t *, void *);
_LIBZFS_H int zpool_sync_one(zpool_handle_t *, void *);
_LIBZFS_H int zpool_vdev_online(zpool_handle_t *, const char *, int,
vdev_state_t *);
_LIBZFS_H int zpool_vdev_offline(zpool_handle_t *, const char *, boolean_t);
_LIBZFS_H int zpool_vdev_attach(zpool_handle_t *, const char *,
const char *, nvlist_t *, int, boolean_t);
_LIBZFS_H int zpool_vdev_detach(zpool_handle_t *, const char *);
_LIBZFS_H int zpool_vdev_remove(zpool_handle_t *, const char *);
_LIBZFS_H int zpool_vdev_remove_cancel(zpool_handle_t *);
_LIBZFS_H int zpool_vdev_indirect_size(zpool_handle_t *, const char *,
uint64_t *);
_LIBZFS_H int zpool_vdev_split(zpool_handle_t *, char *, nvlist_t **,
nvlist_t *, splitflags_t);
_LIBZFS_H int zpool_vdev_fault(zpool_handle_t *, uint64_t, vdev_aux_t);
_LIBZFS_H int zpool_vdev_degrade(zpool_handle_t *, uint64_t, vdev_aux_t);
_LIBZFS_H int zpool_vdev_clear(zpool_handle_t *, uint64_t);
_LIBZFS_H nvlist_t *zpool_find_vdev(zpool_handle_t *, const char *, boolean_t *,
boolean_t *, boolean_t *);
_LIBZFS_H nvlist_t *zpool_find_vdev_by_physpath(zpool_handle_t *, const char *,
boolean_t *, boolean_t *, boolean_t *);
_LIBZFS_H int zpool_label_disk(libzfs_handle_t *, zpool_handle_t *,
const char *);
_LIBZFS_H uint64_t zpool_vdev_path_to_guid(zpool_handle_t *zhp,
const char *path);
_LIBZFS_H const char *zpool_get_state_str(zpool_handle_t *);
/*
* Functions to manage pool properties
*/
_LIBZFS_H int zpool_set_prop(zpool_handle_t *, const char *, const char *);
_LIBZFS_H int zpool_get_prop(zpool_handle_t *, zpool_prop_t, char *,
size_t proplen, zprop_source_t *, boolean_t literal);
_LIBZFS_H uint64_t zpool_get_prop_int(zpool_handle_t *, zpool_prop_t,
zprop_source_t *);
_LIBZFS_H int zpool_props_refresh(zpool_handle_t *);
_LIBZFS_H const char *zpool_prop_to_name(zpool_prop_t);
_LIBZFS_H const char *zpool_prop_values(zpool_prop_t);
/*
* Functions to manage vdev properties
*/
_LIBZFS_H int zpool_get_vdev_prop_value(nvlist_t *, vdev_prop_t, char *, char *,
size_t, zprop_source_t *, boolean_t);
_LIBZFS_H int zpool_get_vdev_prop(zpool_handle_t *, const char *, vdev_prop_t,
char *, char *, size_t, zprop_source_t *, boolean_t);
_LIBZFS_H int zpool_get_all_vdev_props(zpool_handle_t *, const char *,
nvlist_t **);
_LIBZFS_H int zpool_set_vdev_prop(zpool_handle_t *, const char *, const char *,
const char *);
_LIBZFS_H const char *vdev_prop_to_name(vdev_prop_t);
_LIBZFS_H const char *vdev_prop_values(vdev_prop_t);
_LIBZFS_H boolean_t vdev_prop_user(const char *name);
_LIBZFS_H const char *vdev_prop_column_name(vdev_prop_t);
_LIBZFS_H boolean_t vdev_prop_align_right(vdev_prop_t);
/*
* Pool health statistics.
*/
typedef enum {
/*
* The following correspond to faults as defined in the (fault.fs.zfs.*)
* event namespace. Each is associated with a corresponding message ID.
* This must be kept in sync with the zfs_msgid_table in
* lib/libzfs/libzfs_status.c.
*/
ZPOOL_STATUS_CORRUPT_CACHE, /* corrupt /kernel/drv/zpool.cache */
ZPOOL_STATUS_MISSING_DEV_R, /* missing device with replicas */
ZPOOL_STATUS_MISSING_DEV_NR, /* missing device with no replicas */
ZPOOL_STATUS_CORRUPT_LABEL_R, /* bad device label with replicas */
ZPOOL_STATUS_CORRUPT_LABEL_NR, /* bad device label with no replicas */
ZPOOL_STATUS_BAD_GUID_SUM, /* sum of device guids didn't match */
ZPOOL_STATUS_CORRUPT_POOL, /* pool metadata is corrupted */
ZPOOL_STATUS_CORRUPT_DATA, /* data errors in user (meta)data */
ZPOOL_STATUS_FAILING_DEV, /* device experiencing errors */
ZPOOL_STATUS_VERSION_NEWER, /* newer on-disk version */
ZPOOL_STATUS_HOSTID_MISMATCH, /* last accessed by another system */
ZPOOL_STATUS_HOSTID_ACTIVE, /* currently active on another system */
ZPOOL_STATUS_HOSTID_REQUIRED, /* multihost=on and hostid=0 */
ZPOOL_STATUS_IO_FAILURE_WAIT, /* failed I/O, failmode 'wait' */
ZPOOL_STATUS_IO_FAILURE_CONTINUE, /* failed I/O, failmode 'continue' */
ZPOOL_STATUS_IO_FAILURE_MMP, /* failed MMP, failmode not 'panic' */
ZPOOL_STATUS_BAD_LOG, /* cannot read log chain(s) */
ZPOOL_STATUS_ERRATA, /* informational errata available */
/*
* If the pool has unsupported features but can still be opened in
* read-only mode, its status is ZPOOL_STATUS_UNSUP_FEAT_WRITE. If the
* pool has unsupported features but cannot be opened at all, its
* status is ZPOOL_STATUS_UNSUP_FEAT_READ.
*/
ZPOOL_STATUS_UNSUP_FEAT_READ, /* unsupported features for read */
ZPOOL_STATUS_UNSUP_FEAT_WRITE, /* unsupported features for write */
/*
* These faults have no corresponding message ID. At the time we are
* checking the status, the original reason for the FMA fault (I/O or
* checksum errors) has been lost.
*/
ZPOOL_STATUS_FAULTED_DEV_R, /* faulted device with replicas */
ZPOOL_STATUS_FAULTED_DEV_NR, /* faulted device with no replicas */
/*
* The following are not faults per se, but still an error possibly
* requiring administrative attention. There is no corresponding
* message ID.
*/
ZPOOL_STATUS_VERSION_OLDER, /* older legacy on-disk version */
ZPOOL_STATUS_FEAT_DISABLED, /* supported features are disabled */
ZPOOL_STATUS_RESILVERING, /* device being resilvered */
ZPOOL_STATUS_OFFLINE_DEV, /* device offline */
ZPOOL_STATUS_REMOVED_DEV, /* removed device */
ZPOOL_STATUS_REBUILDING, /* device being rebuilt */
ZPOOL_STATUS_REBUILD_SCRUB, /* recommend scrubbing the pool */
ZPOOL_STATUS_NON_NATIVE_ASHIFT, /* (e.g. 512e dev with ashift of 9) */
ZPOOL_STATUS_COMPATIBILITY_ERR, /* bad 'compatibility' property */
ZPOOL_STATUS_INCOMPATIBLE_FEAT, /* feature set outside compatibility */
/*
* Finally, the following indicates a healthy pool.
*/
ZPOOL_STATUS_OK
} zpool_status_t;
_LIBZFS_H zpool_status_t zpool_get_status(zpool_handle_t *, char **,
zpool_errata_t *);
_LIBZFS_H zpool_status_t zpool_import_status(nvlist_t *, char **,
zpool_errata_t *);
/*
* Statistics and configuration functions.
*/
_LIBZFS_H nvlist_t *zpool_get_config(zpool_handle_t *, nvlist_t **);
_LIBZFS_H nvlist_t *zpool_get_features(zpool_handle_t *);
_LIBZFS_H int zpool_refresh_stats(zpool_handle_t *, boolean_t *);
_LIBZFS_H int zpool_get_errlog(zpool_handle_t *, nvlist_t **);
/*
* Import and export functions
*/
_LIBZFS_H int zpool_export(zpool_handle_t *, boolean_t, const char *);
_LIBZFS_H int zpool_export_force(zpool_handle_t *, const char *);
_LIBZFS_H int zpool_import(libzfs_handle_t *, nvlist_t *, const char *,
char *altroot);
_LIBZFS_H int zpool_import_props(libzfs_handle_t *, nvlist_t *, const char *,
nvlist_t *, int);
_LIBZFS_H void zpool_print_unsup_feat(nvlist_t *config);
/*
* Miscellaneous pool functions
*/
struct zfs_cmd;
-_LIBZFS_H const char *zfs_history_event_names[];
+_LIBZFS_H const char *const zfs_history_event_names[];
typedef enum {
VDEV_NAME_PATH = 1 << 0,
VDEV_NAME_GUID = 1 << 1,
VDEV_NAME_FOLLOW_LINKS = 1 << 2,
VDEV_NAME_TYPE_ID = 1 << 3,
} vdev_name_t;
_LIBZFS_H char *zpool_vdev_name(libzfs_handle_t *, zpool_handle_t *, nvlist_t *,
int name_flags);
_LIBZFS_H int zpool_upgrade(zpool_handle_t *, uint64_t);
_LIBZFS_H int zpool_get_history(zpool_handle_t *, nvlist_t **, uint64_t *,
boolean_t *);
_LIBZFS_H int zpool_events_next(libzfs_handle_t *, nvlist_t **, int *, unsigned,
int);
_LIBZFS_H int zpool_events_clear(libzfs_handle_t *, int *);
_LIBZFS_H int zpool_events_seek(libzfs_handle_t *, uint64_t, int);
_LIBZFS_H void zpool_obj_to_path_ds(zpool_handle_t *, uint64_t, uint64_t,
char *, size_t);
_LIBZFS_H void zpool_obj_to_path(zpool_handle_t *, uint64_t, uint64_t, char *,
size_t);
_LIBZFS_H int zfs_ioctl(libzfs_handle_t *, int, struct zfs_cmd *);
_LIBZFS_H int zpool_get_physpath(zpool_handle_t *, char *, size_t);
_LIBZFS_H void zpool_explain_recover(libzfs_handle_t *, const char *, int,
nvlist_t *);
_LIBZFS_H int zpool_checkpoint(zpool_handle_t *);
_LIBZFS_H int zpool_discard_checkpoint(zpool_handle_t *);
_LIBZFS_H boolean_t zpool_is_draid_spare(const char *);
/*
* Basic handle manipulations. These functions do not create or destroy the
* underlying datasets, only the references to them.
*/
_LIBZFS_H zfs_handle_t *zfs_open(libzfs_handle_t *, const char *, int);
_LIBZFS_H zfs_handle_t *zfs_handle_dup(zfs_handle_t *);
_LIBZFS_H void zfs_close(zfs_handle_t *);
_LIBZFS_H zfs_type_t zfs_get_type(const zfs_handle_t *);
_LIBZFS_H zfs_type_t zfs_get_underlying_type(const zfs_handle_t *);
_LIBZFS_H const char *zfs_get_name(const zfs_handle_t *);
_LIBZFS_H zpool_handle_t *zfs_get_pool_handle(const zfs_handle_t *);
_LIBZFS_H const char *zfs_get_pool_name(const zfs_handle_t *);
/*
* Property management functions. Some functions are shared with the kernel,
* and are found in sys/fs/zfs.h.
*/
/*
* zfs dataset property management
*/
_LIBZFS_H const char *zfs_prop_default_string(zfs_prop_t);
_LIBZFS_H uint64_t zfs_prop_default_numeric(zfs_prop_t);
_LIBZFS_H const char *zfs_prop_column_name(zfs_prop_t);
_LIBZFS_H boolean_t zfs_prop_align_right(zfs_prop_t);
_LIBZFS_H nvlist_t *zfs_valid_proplist(libzfs_handle_t *, zfs_type_t,
nvlist_t *, uint64_t, zfs_handle_t *, zpool_handle_t *, boolean_t,
const char *);
_LIBZFS_H const char *zfs_prop_to_name(zfs_prop_t);
_LIBZFS_H int zfs_prop_set(zfs_handle_t *, const char *, const char *);
_LIBZFS_H int zfs_prop_set_list(zfs_handle_t *, nvlist_t *);
_LIBZFS_H int zfs_prop_get(zfs_handle_t *, zfs_prop_t, char *, size_t,
zprop_source_t *, char *, size_t, boolean_t);
_LIBZFS_H int zfs_prop_get_recvd(zfs_handle_t *, const char *, char *, size_t,
boolean_t);
_LIBZFS_H int zfs_prop_get_numeric(zfs_handle_t *, zfs_prop_t, uint64_t *,
zprop_source_t *, char *, size_t);
_LIBZFS_H int zfs_prop_get_userquota_int(zfs_handle_t *zhp,
const char *propname, uint64_t *propvalue);
_LIBZFS_H int zfs_prop_get_userquota(zfs_handle_t *zhp, const char *propname,
char *propbuf, int proplen, boolean_t literal);
_LIBZFS_H int zfs_prop_get_written_int(zfs_handle_t *zhp, const char *propname,
uint64_t *propvalue);
_LIBZFS_H int zfs_prop_get_written(zfs_handle_t *zhp, const char *propname,
char *propbuf, int proplen, boolean_t literal);
_LIBZFS_H int zfs_prop_get_feature(zfs_handle_t *zhp, const char *propname,
char *buf, size_t len);
_LIBZFS_H uint64_t getprop_uint64(zfs_handle_t *, zfs_prop_t, char **);
_LIBZFS_H uint64_t zfs_prop_get_int(zfs_handle_t *, zfs_prop_t);
_LIBZFS_H int zfs_prop_inherit(zfs_handle_t *, const char *, boolean_t);
_LIBZFS_H const char *zfs_prop_values(zfs_prop_t);
_LIBZFS_H int zfs_prop_is_string(zfs_prop_t prop);
_LIBZFS_H nvlist_t *zfs_get_all_props(zfs_handle_t *);
_LIBZFS_H nvlist_t *zfs_get_user_props(zfs_handle_t *);
_LIBZFS_H nvlist_t *zfs_get_recvd_props(zfs_handle_t *);
_LIBZFS_H nvlist_t *zfs_get_clones_nvl(zfs_handle_t *);
_LIBZFS_H int zfs_wait_status(zfs_handle_t *, zfs_wait_activity_t,
boolean_t *, boolean_t *);
/*
* zfs encryption management
*/
_LIBZFS_H int zfs_crypto_get_encryption_root(zfs_handle_t *, boolean_t *,
char *);
_LIBZFS_H int zfs_crypto_create(libzfs_handle_t *, char *, nvlist_t *,
nvlist_t *, boolean_t stdin_available, uint8_t **, uint_t *);
_LIBZFS_H int zfs_crypto_clone_check(libzfs_handle_t *, zfs_handle_t *, char *,
nvlist_t *);
_LIBZFS_H int zfs_crypto_attempt_load_keys(libzfs_handle_t *, char *);
_LIBZFS_H int zfs_crypto_load_key(zfs_handle_t *, boolean_t, char *);
_LIBZFS_H int zfs_crypto_unload_key(zfs_handle_t *);
_LIBZFS_H int zfs_crypto_rewrap(zfs_handle_t *, nvlist_t *, boolean_t);
typedef struct zprop_list {
int pl_prop;
char *pl_user_prop;
struct zprop_list *pl_next;
boolean_t pl_all;
size_t pl_width;
size_t pl_recvd_width;
boolean_t pl_fixed;
} zprop_list_t;
_LIBZFS_H int zfs_expand_proplist(zfs_handle_t *, zprop_list_t **, boolean_t,
boolean_t);
_LIBZFS_H void zfs_prune_proplist(zfs_handle_t *, uint8_t *);
_LIBZFS_H int vdev_expand_proplist(zpool_handle_t *, const char *,
zprop_list_t **);
#define ZFS_MOUNTPOINT_NONE "none"
#define ZFS_MOUNTPOINT_LEGACY "legacy"
#define ZFS_FEATURE_DISABLED "disabled"
#define ZFS_FEATURE_ENABLED "enabled"
#define ZFS_FEATURE_ACTIVE "active"
#define ZFS_UNSUPPORTED_INACTIVE "inactive"
#define ZFS_UNSUPPORTED_READONLY "readonly"
/*
* zpool property management
*/
_LIBZFS_H int zpool_expand_proplist(zpool_handle_t *, zprop_list_t **,
zfs_type_t, boolean_t);
_LIBZFS_H int zpool_prop_get_feature(zpool_handle_t *, const char *, char *,
size_t);
_LIBZFS_H const char *zpool_prop_default_string(zpool_prop_t);
_LIBZFS_H uint64_t zpool_prop_default_numeric(zpool_prop_t);
_LIBZFS_H const char *zpool_prop_column_name(zpool_prop_t);
_LIBZFS_H boolean_t zpool_prop_align_right(zpool_prop_t);
/*
* Functions shared by zfs and zpool property management.
*/
_LIBZFS_H int zprop_iter(zprop_func func, void *cb, boolean_t show_all,
boolean_t ordered, zfs_type_t type);
_LIBZFS_H int zprop_get_list(libzfs_handle_t *, char *, zprop_list_t **,
zfs_type_t);
_LIBZFS_H void zprop_free_list(zprop_list_t *);
#define ZFS_GET_NCOLS 5
typedef enum {
GET_COL_NONE,
GET_COL_NAME,
GET_COL_PROPERTY,
GET_COL_VALUE,
GET_COL_RECVD,
GET_COL_SOURCE
} zfs_get_column_t;
/*
* Functions for printing zfs or zpool properties
*/
typedef struct vdev_cbdata {
int cb_name_flags;
char **cb_names;
unsigned int cb_names_count;
} vdev_cbdata_t;
typedef struct zprop_get_cbdata {
int cb_sources;
zfs_get_column_t cb_columns[ZFS_GET_NCOLS];
int cb_colwidths[ZFS_GET_NCOLS + 1];
boolean_t cb_scripted;
boolean_t cb_literal;
boolean_t cb_first;
zprop_list_t *cb_proplist;
zfs_type_t cb_type;
vdev_cbdata_t cb_vdevs;
} zprop_get_cbdata_t;
_LIBZFS_H void zprop_print_one_property(const char *, zprop_get_cbdata_t *,
const char *, const char *, zprop_source_t, const char *,
const char *);
/*
* Iterator functions.
*/
typedef int (*zfs_iter_f)(zfs_handle_t *, void *);
_LIBZFS_H int zfs_iter_root(libzfs_handle_t *, zfs_iter_f, void *);
_LIBZFS_H int zfs_iter_children(zfs_handle_t *, zfs_iter_f, void *);
_LIBZFS_H int zfs_iter_dependents(zfs_handle_t *, boolean_t, zfs_iter_f,
void *);
_LIBZFS_H int zfs_iter_filesystems(zfs_handle_t *, zfs_iter_f, void *);
_LIBZFS_H int zfs_iter_snapshots(zfs_handle_t *, boolean_t, zfs_iter_f, void *,
uint64_t, uint64_t);
_LIBZFS_H int zfs_iter_snapshots_sorted(zfs_handle_t *, zfs_iter_f, void *,
uint64_t, uint64_t);
_LIBZFS_H int zfs_iter_snapspec(zfs_handle_t *, const char *, zfs_iter_f,
void *);
_LIBZFS_H int zfs_iter_bookmarks(zfs_handle_t *, zfs_iter_f, void *);
_LIBZFS_H int zfs_iter_mounted(zfs_handle_t *, zfs_iter_f, void *);
typedef struct get_all_cb {
zfs_handle_t **cb_handles;
size_t cb_alloc;
size_t cb_used;
} get_all_cb_t;
_LIBZFS_H void zfs_foreach_mountpoint(libzfs_handle_t *, zfs_handle_t **,
size_t, zfs_iter_f, void *, boolean_t);
_LIBZFS_H void libzfs_add_handle(get_all_cb_t *, zfs_handle_t *);
/*
* Functions to create and destroy datasets.
*/
_LIBZFS_H int zfs_create(libzfs_handle_t *, const char *, zfs_type_t,
nvlist_t *);
_LIBZFS_H int zfs_create_ancestors(libzfs_handle_t *, const char *);
_LIBZFS_H int zfs_destroy(zfs_handle_t *, boolean_t);
_LIBZFS_H int zfs_destroy_snaps(zfs_handle_t *, char *, boolean_t);
_LIBZFS_H int zfs_destroy_snaps_nvl(libzfs_handle_t *, nvlist_t *, boolean_t);
_LIBZFS_H int zfs_destroy_snaps_nvl_os(libzfs_handle_t *, nvlist_t *);
_LIBZFS_H int zfs_clone(zfs_handle_t *, const char *, nvlist_t *);
_LIBZFS_H int zfs_snapshot(libzfs_handle_t *, const char *, boolean_t,
nvlist_t *);
_LIBZFS_H int zfs_snapshot_nvl(libzfs_handle_t *hdl, nvlist_t *snaps,
nvlist_t *props);
_LIBZFS_H int zfs_rollback(zfs_handle_t *, zfs_handle_t *, boolean_t);
typedef struct renameflags {
/* recursive rename */
int recursive : 1;
/* don't unmount file systems */
int nounmount : 1;
/* force unmount file systems */
int forceunmount : 1;
} renameflags_t;
_LIBZFS_H int zfs_rename(zfs_handle_t *, const char *, renameflags_t);
typedef struct sendflags {
/* Amount of extra information to print. */
int verbosity;
/* recursive send (ie, -R) */
boolean_t replicate;
/* for recursive send, skip sending missing snapshots */
boolean_t skipmissing;
/* for incrementals, do all intermediate snapshots */
boolean_t doall;
/* if dataset is a clone, do incremental from its origin */
boolean_t fromorigin;
/* field no longer used, maintained for backwards compatibility */
boolean_t pad;
/* send properties (ie, -p) */
boolean_t props;
/* do not send (no-op, ie. -n) */
boolean_t dryrun;
/* parsable verbose output (ie. -P) */
boolean_t parsable;
/* show progress (ie. -v) */
boolean_t progress;
/* large blocks (>128K) are permitted */
boolean_t largeblock;
/* WRITE_EMBEDDED records of type DATA are permitted */
boolean_t embed_data;
/* compressed WRITE records are permitted */
boolean_t compress;
/* raw encrypted records are permitted */
boolean_t raw;
/* only send received properties (ie. -b) */
boolean_t backup;
/* include snapshot holds in send stream */
boolean_t holds;
/* stream represents a partially received dataset */
boolean_t saved;
} sendflags_t;
typedef boolean_t (snapfilter_cb_t)(zfs_handle_t *, void *);
_LIBZFS_H int zfs_send(zfs_handle_t *, const char *, const char *,
sendflags_t *, int, snapfilter_cb_t, void *, nvlist_t **);
_LIBZFS_H int zfs_send_one(zfs_handle_t *, const char *, int, sendflags_t *,
const char *);
_LIBZFS_H int zfs_send_progress(zfs_handle_t *, int, uint64_t *, uint64_t *);
_LIBZFS_H int zfs_send_resume(libzfs_handle_t *, sendflags_t *, int outfd,
const char *);
_LIBZFS_H int zfs_send_saved(zfs_handle_t *, sendflags_t *, int, const char *);
_LIBZFS_H nvlist_t *zfs_send_resume_token_to_nvlist(libzfs_handle_t *hdl,
const char *token);
_LIBZFS_H int zfs_promote(zfs_handle_t *);
_LIBZFS_H int zfs_hold(zfs_handle_t *, const char *, const char *,
boolean_t, int);
_LIBZFS_H int zfs_hold_nvl(zfs_handle_t *, int, nvlist_t *);
_LIBZFS_H int zfs_release(zfs_handle_t *, const char *, const char *,
boolean_t);
_LIBZFS_H int zfs_get_holds(zfs_handle_t *, nvlist_t **);
_LIBZFS_H uint64_t zvol_volsize_to_reservation(zpool_handle_t *, uint64_t,
nvlist_t *);
typedef int (*zfs_userspace_cb_t)(void *arg, const char *domain,
uid_t rid, uint64_t space);
_LIBZFS_H int zfs_userspace(zfs_handle_t *, zfs_userquota_prop_t,
zfs_userspace_cb_t, void *);
_LIBZFS_H int zfs_get_fsacl(zfs_handle_t *, nvlist_t **);
_LIBZFS_H int zfs_set_fsacl(zfs_handle_t *, boolean_t, nvlist_t *);
typedef struct recvflags {
/* print informational messages (ie, -v was specified) */
boolean_t verbose;
/* the destination is a prefix, not the exact fs (ie, -d) */
boolean_t isprefix;
/*
* Only the tail of the sent snapshot path is appended to the
* destination to determine the received snapshot name (ie, -e).
*/
boolean_t istail;
/* do not actually do the recv, just check if it would work (ie, -n) */
boolean_t dryrun;
/* rollback/destroy filesystems as necessary (eg, -F) */
boolean_t force;
/* set "canmount=off" on all modified filesystems */
boolean_t canmountoff;
/*
* Mark the file systems as "resumable" and do not destroy them if the
* receive is interrupted
*/
boolean_t resumable;
/* byteswap flag is used internally; callers need not specify */
boolean_t byteswap;
/* do not mount file systems as they are extracted (private) */
boolean_t nomount;
/* Was holds flag set in the compound header? */
boolean_t holds;
/* skip receive of snapshot holds */
boolean_t skipholds;
/* mount the filesystem unless nomount is specified */
boolean_t domount;
/* force unmount while recv snapshot (private) */
boolean_t forceunmount;
} recvflags_t;
_LIBZFS_H int zfs_receive(libzfs_handle_t *, const char *, nvlist_t *,
recvflags_t *, int, avl_tree_t *);
typedef enum diff_flags {
ZFS_DIFF_PARSEABLE = 1 << 0,
ZFS_DIFF_TIMESTAMP = 1 << 1,
ZFS_DIFF_CLASSIFY = 1 << 2,
ZFS_DIFF_NO_MANGLE = 1 << 3
} diff_flags_t;
_LIBZFS_H int zfs_show_diffs(zfs_handle_t *, int, const char *, const char *,
int);
/*
* Miscellaneous functions.
*/
_LIBZFS_H const char *zfs_type_to_name(zfs_type_t);
_LIBZFS_H void zfs_refresh_properties(zfs_handle_t *);
_LIBZFS_H int zfs_name_valid(const char *, zfs_type_t);
_LIBZFS_H zfs_handle_t *zfs_path_to_zhandle(libzfs_handle_t *, const char *,
zfs_type_t);
_LIBZFS_H int zfs_parent_name(zfs_handle_t *, char *, size_t);
_LIBZFS_H boolean_t zfs_dataset_exists(libzfs_handle_t *, const char *,
zfs_type_t);
_LIBZFS_H int zfs_spa_version(zfs_handle_t *, int *);
_LIBZFS_H boolean_t zfs_bookmark_exists(const char *path);
/*
* Mount support functions.
*/
_LIBZFS_H boolean_t is_mounted(libzfs_handle_t *, const char *special, char **);
_LIBZFS_H boolean_t zfs_is_mounted(zfs_handle_t *, char **);
_LIBZFS_H int zfs_mount(zfs_handle_t *, const char *, int);
_LIBZFS_H int zfs_mount_at(zfs_handle_t *, const char *, int, const char *);
_LIBZFS_H int zfs_unmount(zfs_handle_t *, const char *, int);
_LIBZFS_H int zfs_unmountall(zfs_handle_t *, int);
_LIBZFS_H int zfs_mount_delegation_check(void);
#if defined(__linux__) || defined(__APPLE__)
_LIBZFS_H int zfs_parse_mount_options(char *mntopts, unsigned long *mntflags,
unsigned long *zfsflags, int sloppy, char *badopt, char *mtabopt);
_LIBZFS_H void zfs_adjust_mount_options(zfs_handle_t *zhp, const char *mntpoint,
char *mntopts, char *mtabopt);
#endif
/*
* Share support functions.
*/
_LIBZFS_H boolean_t zfs_is_shared(zfs_handle_t *);
_LIBZFS_H int zfs_share(zfs_handle_t *);
_LIBZFS_H int zfs_unshare(zfs_handle_t *);
/*
* Protocol-specific share support functions.
*/
_LIBZFS_H boolean_t zfs_is_shared_nfs(zfs_handle_t *, char **);
_LIBZFS_H boolean_t zfs_is_shared_smb(zfs_handle_t *, char **);
_LIBZFS_H int zfs_share_nfs(zfs_handle_t *);
_LIBZFS_H int zfs_share_smb(zfs_handle_t *);
_LIBZFS_H int zfs_shareall(zfs_handle_t *);
_LIBZFS_H int zfs_unshare_nfs(zfs_handle_t *, const char *);
_LIBZFS_H int zfs_unshare_smb(zfs_handle_t *, const char *);
_LIBZFS_H int zfs_unshareall_nfs(zfs_handle_t *);
_LIBZFS_H int zfs_unshareall_smb(zfs_handle_t *);
_LIBZFS_H int zfs_unshareall_bypath(zfs_handle_t *, const char *);
_LIBZFS_H int zfs_unshareall_bytype(zfs_handle_t *, const char *, const char *);
_LIBZFS_H int zfs_unshareall(zfs_handle_t *);
_LIBZFS_H int zfs_deleg_share_nfs(libzfs_handle_t *, char *, char *, char *,
void *, void *, int, zfs_share_op_t);
_LIBZFS_H void zfs_commit_nfs_shares(void);
_LIBZFS_H void zfs_commit_smb_shares(void);
_LIBZFS_H void zfs_commit_all_shares(void);
_LIBZFS_H void zfs_commit_shares(const char *);
_LIBZFS_H int zfs_nicestrtonum(libzfs_handle_t *, const char *, uint64_t *);
/*
* Utility functions to run an _LIBZFS_Hal process.
*/
#define STDOUT_VERBOSE 0x01
#define STDERR_VERBOSE 0x02
#define NO_DEFAULT_PATH 0x04 /* Don't use $PATH to lookup the command */
_LIBZFS_H int libzfs_run_process(const char *, char **, int);
_LIBZFS_H int libzfs_run_process_get_stdout(const char *, char *[], char *[],
char **[], int *);
_LIBZFS_H int libzfs_run_process_get_stdout_nopath(const char *, char *[],
char *[], char **[], int *);
_LIBZFS_H void libzfs_free_str_array(char **, int);
_LIBZFS_H int libzfs_envvar_is_set(char *);
/*
* Utility functions for zfs version
*/
_LIBZFS_H void zfs_version_userland(char *, int);
_LIBZFS_H int zfs_version_kernel(char *, int);
_LIBZFS_H int zfs_version_print(void);
/*
* Given a device or file, determine if it is part of a pool.
*/
_LIBZFS_H int zpool_in_use(libzfs_handle_t *, int, pool_state_t *, char **,
boolean_t *);
/*
* Label manipulation.
*/
_LIBZFS_H int zpool_clear_label(int);
_LIBZFS_H int zpool_set_bootenv(zpool_handle_t *, const nvlist_t *);
_LIBZFS_H int zpool_get_bootenv(zpool_handle_t *, nvlist_t **);
/*
* Management interfaces for SMB ACL files
*/
_LIBZFS_H int zfs_smb_acl_add(libzfs_handle_t *, char *, char *, char *);
_LIBZFS_H int zfs_smb_acl_remove(libzfs_handle_t *, char *, char *, char *);
_LIBZFS_H int zfs_smb_acl_purge(libzfs_handle_t *, char *, char *);
_LIBZFS_H int zfs_smb_acl_rename(libzfs_handle_t *, char *, char *, char *,
char *);
/*
* Enable and disable datasets within a pool by mounting/unmounting and
* sharing/unsharing them.
*/
_LIBZFS_H int zpool_enable_datasets(zpool_handle_t *, const char *, int);
_LIBZFS_H int zpool_disable_datasets(zpool_handle_t *, boolean_t);
_LIBZFS_H void zpool_disable_datasets_os(zpool_handle_t *, boolean_t);
_LIBZFS_H void zpool_disable_volume_os(const char *);
/*
* Parse a features file for -o compatibility
*/
typedef enum {
ZPOOL_COMPATIBILITY_OK,
ZPOOL_COMPATIBILITY_WARNTOKEN,
ZPOOL_COMPATIBILITY_BADTOKEN,
ZPOOL_COMPATIBILITY_BADFILE,
ZPOOL_COMPATIBILITY_NOFILES
} zpool_compat_status_t;
_LIBZFS_H zpool_compat_status_t zpool_load_compat(const char *,
boolean_t *, char *, size_t);
#ifdef __FreeBSD__
/*
* Attach/detach the given filesystem to/from the given jail.
*/
_LIBZFS_H int zfs_jail(zfs_handle_t *zhp, int jailid, int attach);
/*
* Set loader options for next boot.
*/
_LIBZFS_H int zpool_nextboot(libzfs_handle_t *, uint64_t, uint64_t,
const char *);
#endif /* __FreeBSD__ */
#ifdef __cplusplus
}
#endif
#endif /* _LIBZFS_H */
diff --git a/include/os/freebsd/zfs/sys/freebsd_crypto.h b/include/os/freebsd/zfs/sys/freebsd_crypto.h
index e240f5b0ddca..a3ed4182656c 100644
--- a/include/os/freebsd/zfs/sys/freebsd_crypto.h
+++ b/include/os/freebsd/zfs/sys/freebsd_crypto.h
@@ -1,98 +1,98 @@
/*
* Copyright (c) 2018 Sean Eric Fagan <sef@ixsystems.com>
* Portions Copyright (c) 2005-2011 Pawel Jakub Dawidek <pawel@dawidek.net>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* Portions of this file were taken from GELI's implementation of hmac.
*
* $FreeBSD$
*/
#ifndef _ZFS_FREEBSD_CRYPTO_H
#define _ZFS_FREEBSD_CRYPTO_H
#include <sys/errno.h>
#include <sys/mutex.h>
#include <opencrypto/cryptodev.h>
#include <crypto/sha2/sha256.h>
#include <crypto/sha2/sha512.h>
#define SUN_CKM_AES_CCM "CKM_AES_CCM"
#define SUN_CKM_AES_GCM "CKM_AES_GCM"
#define SUN_CKM_SHA512_HMAC "CKM_SHA512_HMAC"
#define CRYPTO_KEY_RAW 1
#define CRYPTO_BITS2BYTES(n) ((n) == 0 ? 0 : (((n) - 1) >> 3) + 1)
#define CRYPTO_BYTES2BITS(n) ((n) << 3)
struct zio_crypt_info;
typedef struct freebsd_crypt_session {
struct mtx fs_lock;
crypto_session_t fs_sid;
boolean_t fs_done;
} freebsd_crypt_session_t;
/*
* Unused types to minimize code differences.
*/
typedef void *crypto_mechanism_t;
typedef void *crypto_ctx_template_t;
/*
* Unlike the ICP crypto_key type, this only
* supports <data, length> (the equivalent of
* CRYPTO_KEY_RAW).
*/
typedef struct crypto_key {
int ck_format; /* Unused, but minimizes code diff */
void *ck_data;
size_t ck_length;
} crypto_key_t;
typedef struct hmac_ctx {
SHA512_CTX innerctx;
SHA512_CTX outerctx;
} *crypto_context_t;
/*
* The only algorithm ZFS uses for hashing is SHA512_HMAC.
*/
void crypto_mac(const crypto_key_t *key, const void *in_data,
size_t in_data_size, void *out_data, size_t out_data_size);
void crypto_mac_init(struct hmac_ctx *ctx, const crypto_key_t *key);
void crypto_mac_update(struct hmac_ctx *ctx, const void *data,
size_t data_size);
void crypto_mac_final(struct hmac_ctx *ctx, void *out_data,
size_t out_data_size);
int freebsd_crypt_newsession(freebsd_crypt_session_t *sessp,
- struct zio_crypt_info *, crypto_key_t *);
+ const struct zio_crypt_info *, crypto_key_t *);
void freebsd_crypt_freesession(freebsd_crypt_session_t *sessp);
int freebsd_crypt_uio(boolean_t, freebsd_crypt_session_t *,
- struct zio_crypt_info *, zfs_uio_t *, crypto_key_t *, uint8_t *,
+ const struct zio_crypt_info *, zfs_uio_t *, crypto_key_t *, uint8_t *,
size_t, size_t);
#endif /* _ZFS_FREEBSD_CRYPTO_H */
diff --git a/include/os/freebsd/zfs/sys/zfs_context_os.h b/include/os/freebsd/zfs/sys/zfs_context_os.h
index a32eb52c53c1..867199501396 100644
--- a/include/os/freebsd/zfs/sys/zfs_context_os.h
+++ b/include/os/freebsd/zfs/sys/zfs_context_os.h
@@ -1,90 +1,90 @@
/*
* Copyright (c) 2020 iXsystems, Inc.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* $FreeBSD$
*/
#ifndef ZFS_CONTEXT_OS_H_
#define ZFS_CONTEXT_OS_H_
#include <sys/condvar.h>
#include <sys/rwlock.h>
#include <sys/sig.h>
#include_next <sys/sdt.h>
#include <sys/misc.h>
#include <sys/kdb.h>
#include <sys/pathname.h>
#include <sys/conf.h>
#include <sys/types.h>
#include <sys/ccompat.h>
#include <linux/types.h>
#if KSTACK_PAGES * PAGE_SIZE >= 16384
#define HAVE_LARGE_STACKS 1
#endif
#define cond_resched() kern_yield(PRI_USER)
#define taskq_create_sysdc(a, b, d, e, p, dc, f) \
- (taskq_create(a, b, maxclsyspri, d, e, f))
+ ((void) sizeof (dc), taskq_create(a, b, maxclsyspri, d, e, f))
#define tsd_create(keyp, destructor) do { \
*(keyp) = osd_thread_register((destructor)); \
KASSERT(*(keyp) > 0, ("cannot register OSD")); \
} while (0)
#define tsd_destroy(keyp) osd_thread_deregister(*(keyp))
#define tsd_get(key) osd_thread_get(curthread, (key))
#define tsd_set(key, value) osd_thread_set(curthread, (key), (value))
#define fm_panic panic
extern int zfs_debug_level;
extern struct mtx zfs_debug_mtx;
#define ZFS_LOG(lvl, ...) do { \
if (((lvl) & 0xff) <= zfs_debug_level) { \
mtx_lock(&zfs_debug_mtx); \
printf("%s:%u[%d]: ", \
__func__, __LINE__, (lvl)); \
printf(__VA_ARGS__); \
printf("\n"); \
if ((lvl) & 0x100) \
kdb_backtrace(); \
mtx_unlock(&zfs_debug_mtx); \
} \
} while (0)
#define MSEC_TO_TICK(msec) (howmany((hrtime_t)(msec) * hz, MILLISEC))
extern int hz;
extern int tick;
typedef int fstrans_cookie_t;
#define spl_fstrans_mark() (0)
#define spl_fstrans_unmark(x) (x = 0)
#define signal_pending(x) SIGPENDING(x)
#define current curthread
#define thread_join(x)
typedef struct opensolaris_utsname utsname_t;
extern utsname_t *utsname(void);
extern int spa_import_rootpool(const char *name, bool checkpointrewind);
#endif
diff --git a/include/os/freebsd/zfs/sys/zfs_znode_impl.h b/include/os/freebsd/zfs/sys/zfs_znode_impl.h
index 4456046e6e4c..096c9e16d347 100644
--- a/include/os/freebsd/zfs/sys/zfs_znode_impl.h
+++ b/include/os/freebsd/zfs/sys/zfs_znode_impl.h
@@ -1,187 +1,186 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2015 by Delphix. All rights reserved.
* Copyright (c) 2014 Integros [integros.com]
* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
*/
#ifndef _FREEBSD_ZFS_SYS_ZNODE_IMPL_H
#define _FREEBSD_ZFS_SYS_ZNODE_IMPL_H
#include <sys/list.h>
#include <sys/dmu.h>
#include <sys/sa.h>
#include <sys/zfs_vfsops.h>
#include <sys/rrwlock.h>
#include <sys/zfs_sa.h>
#include <sys/zfs_stat.h>
#include <sys/zfs_rlock.h>
#include <sys/zfs_acl.h>
#include <sys/zil.h>
#include <sys/zfs_project.h>
#include <vm/vm_object.h>
#include <sys/uio.h>
#ifdef __cplusplus
extern "C" {
#endif
/*
* Directory entry locks control access to directory entries.
* They are used to protect creates, deletes, and renames.
* Each directory znode has a mutex and a list of locked names.
*/
#define ZNODE_OS_FIELDS \
struct zfsvfs *z_zfsvfs; \
vnode_t *z_vnode; \
char *z_cached_symlink; \
uint64_t z_uid; \
uint64_t z_gid; \
uint64_t z_gen; \
uint64_t z_atime[2]; \
uint64_t z_links;
#define ZFS_LINK_MAX UINT64_MAX
/*
* ZFS minor numbers can refer to either a control device instance or
* a zvol. Depending on the value of zss_type, zss_data points to either
* a zvol_state_t or a zfs_onexit_t.
*/
enum zfs_soft_state_type {
ZSST_ZVOL,
ZSST_CTLDEV
};
typedef struct zfs_soft_state {
enum zfs_soft_state_type zss_type;
void *zss_data;
} zfs_soft_state_t;
/*
* Range locking rules
* --------------------
* 1. When truncating a file (zfs_create, zfs_setattr, zfs_space) the whole
* file range needs to be locked as RL_WRITER. Only then can the pages be
* freed etc and zp_size reset. zp_size must be set within range lock.
* 2. For writes and punching holes (zfs_write & zfs_space) just the range
* being written or freed needs to be locked as RL_WRITER.
* Multiple writes at the end of the file must coordinate zp_size updates
* to ensure data isn't lost. A compare and swap loop is currently used
* to ensure the file size is at least the offset last written.
* 3. For reads (zfs_read, zfs_get_data & zfs_putapage) just the range being
* read needs to be locked as RL_READER. A check against zp_size can then
* be made for reading beyond end of file.
*/
/*
* Convert between znode pointers and vnode pointers
*/
#define ZTOV(ZP) ((ZP)->z_vnode)
#define ZTOI(ZP) ((ZP)->z_vnode)
#define VTOZ(VP) ((struct znode *)(VP)->v_data)
#define VTOZ_SMR(VP) ((znode_t *)vn_load_v_data_smr(VP))
#define ITOZ(VP) ((struct znode *)(VP)->v_data)
#define zhold(zp) vhold(ZTOV((zp)))
#define zrele(zp) vrele(ZTOV((zp)))
#define ZTOZSB(zp) ((zp)->z_zfsvfs)
#define ITOZSB(vp) (VTOZ(vp)->z_zfsvfs)
#define ZTOTYPE(zp) (ZTOV(zp)->v_type)
#define ZTOGID(zp) ((zp)->z_gid)
#define ZTOUID(zp) ((zp)->z_uid)
#define ZTONLNK(zp) ((zp)->z_links)
#define Z_ISBLK(type) ((type) == VBLK)
#define Z_ISCHR(type) ((type) == VCHR)
#define Z_ISLNK(type) ((type) == VLNK)
#define Z_ISDIR(type) ((type) == VDIR)
#define zn_has_cached_data(zp) vn_has_cached_data(ZTOV(zp))
#define zn_flush_cached_data(zp, sync) vn_flush_cached_data(ZTOV(zp), sync)
#define zn_rlimit_fsize(zp, uio) \
vn_rlimit_fsize(ZTOV(zp), GET_UIO_STRUCT(uio), zfs_uio_td(uio))
/* Called on entry to each ZFS vnode and vfs operation */
#define ZFS_ENTER(zfsvfs) \
{ \
ZFS_TEARDOWN_ENTER_READ((zfsvfs), FTAG); \
if (__predict_false((zfsvfs)->z_unmounted)) { \
ZFS_TEARDOWN_EXIT_READ(zfsvfs, FTAG); \
return (EIO); \
} \
}
/* Must be called before exiting the vop */
#define ZFS_EXIT(zfsvfs) ZFS_TEARDOWN_EXIT_READ(zfsvfs, FTAG)
/* Verifies the znode is valid */
#define ZFS_VERIFY_ZP(zp) \
if (__predict_false((zp)->z_sa_hdl == NULL)) { \
ZFS_EXIT((zp)->z_zfsvfs); \
return (EIO); \
} \
/*
* Macros for dealing with dmu_buf_hold
*/
#define ZFS_OBJ_HASH(obj_num) ((obj_num) & (ZFS_OBJ_MTX_SZ - 1))
#define ZFS_OBJ_MUTEX(zfsvfs, obj_num) \
(&(zfsvfs)->z_hold_mtx[ZFS_OBJ_HASH(obj_num)])
#define ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num) \
mutex_enter(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
#define ZFS_OBJ_HOLD_TRYENTER(zfsvfs, obj_num) \
mutex_tryenter(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
#define ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num) \
mutex_exit(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
/* Encode ZFS stored time values from a struct timespec */
#define ZFS_TIME_ENCODE(tp, stmp) \
{ \
(stmp)[0] = (uint64_t)(tp)->tv_sec; \
(stmp)[1] = (uint64_t)(tp)->tv_nsec; \
}
/* Decode ZFS stored time values to a struct timespec */
#define ZFS_TIME_DECODE(tp, stmp) \
{ \
(tp)->tv_sec = (time_t)(stmp)[0]; \
(tp)->tv_nsec = (long)(stmp)[1]; \
}
#define ZFS_ACCESSTIME_STAMP(zfsvfs, zp) \
if ((zfsvfs)->z_atime && !((zfsvfs)->z_vfs->vfs_flag & VFS_RDONLY)) \
zfs_tstamp_update_setup_ext(zp, ACCESSED, NULL, NULL, B_FALSE);
extern void zfs_tstamp_update_setup_ext(struct znode *,
uint_t, uint64_t [2], uint64_t [2], boolean_t have_tx);
extern void zfs_znode_free(struct znode *);
-extern zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE];
-extern int zfsfstype;
+extern zil_replay_func_t *const zfs_replay_vector[TX_MAX_TYPE];
extern int zfs_znode_parent_and_name(struct znode *zp, struct znode **dzpp,
char *buf);
#ifdef __cplusplus
}
#endif
#endif /* _FREEBSD_SYS_FS_ZFS_ZNODE_H */
diff --git a/include/os/linux/spl/rpc/xdr.h b/include/os/linux/spl/rpc/xdr.h
index c62080a1178a..1c03a42a9961 100644
--- a/include/os/linux/spl/rpc/xdr.h
+++ b/include/os/linux/spl/rpc/xdr.h
@@ -1,155 +1,156 @@
/*
* Copyright (c) 2008 Sun Microsystems, Inc.
* Written by Ricardo Correia <Ricardo.M.Correia@Sun.COM>
*
* This file is part of the SPL, Solaris Porting Layer.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef _SPL_RPC_XDR_H
#define _SPL_RPC_XDR_H
#include <sys/types.h>
typedef int bool_t;
/*
* XDR enums and types.
*/
enum xdr_op {
XDR_ENCODE,
XDR_DECODE
};
struct xdr_ops;
typedef struct {
- struct xdr_ops *x_ops; /* Let caller know xdrmem_create() succeeds */
+ const struct xdr_ops *x_ops;
+ /* Let caller know xdrmem_create() succeeds */
caddr_t x_addr; /* Current buffer addr */
caddr_t x_addr_end; /* End of the buffer */
enum xdr_op x_op; /* Stream direction */
} XDR;
typedef bool_t (*xdrproc_t)(XDR *xdrs, void *ptr);
struct xdr_ops {
bool_t (*xdr_control)(XDR *, int, void *);
bool_t (*xdr_char)(XDR *, char *);
bool_t (*xdr_u_short)(XDR *, unsigned short *);
bool_t (*xdr_u_int)(XDR *, unsigned *);
bool_t (*xdr_u_longlong_t)(XDR *, u_longlong_t *);
bool_t (*xdr_opaque)(XDR *, caddr_t, const uint_t);
bool_t (*xdr_string)(XDR *, char **, const uint_t);
bool_t (*xdr_array)(XDR *, caddr_t *, uint_t *, const uint_t,
const uint_t, const xdrproc_t);
};
/*
* XDR control operator.
*/
#define XDR_GET_BYTES_AVAIL 1
struct xdr_bytesrec {
bool_t xc_is_last_record;
size_t xc_num_avail;
};
/*
* XDR functions.
*/
void xdrmem_create(XDR *xdrs, const caddr_t addr, const uint_t size,
const enum xdr_op op);
/* Currently not needed. If needed later, we'll add it to struct xdr_ops */
#define xdr_destroy(xdrs) ((void) 0)
#define xdr_control(xdrs, req, info) \
(xdrs)->x_ops->xdr_control((xdrs), (req), (info))
/*
* For precaution, the following are defined as static inlines instead of macros
* to get some amount of type safety.
*
* Also, macros wouldn't work in the case where typecasting is done, because it
* must be possible to reference the functions' addresses by these names.
*/
static inline bool_t xdr_char(XDR *xdrs, char *cp)
{
return (xdrs->x_ops->xdr_char(xdrs, cp));
}
static inline bool_t xdr_u_short(XDR *xdrs, unsigned short *usp)
{
return (xdrs->x_ops->xdr_u_short(xdrs, usp));
}
static inline bool_t xdr_short(XDR *xdrs, short *sp)
{
BUILD_BUG_ON(sizeof (short) != 2);
return (xdrs->x_ops->xdr_u_short(xdrs, (unsigned short *) sp));
}
static inline bool_t xdr_u_int(XDR *xdrs, unsigned *up)
{
return (xdrs->x_ops->xdr_u_int(xdrs, up));
}
static inline bool_t xdr_int(XDR *xdrs, int *ip)
{
BUILD_BUG_ON(sizeof (int) != 4);
return (xdrs->x_ops->xdr_u_int(xdrs, (unsigned *)ip));
}
static inline bool_t xdr_u_longlong_t(XDR *xdrs, u_longlong_t *ullp)
{
return (xdrs->x_ops->xdr_u_longlong_t(xdrs, ullp));
}
static inline bool_t xdr_longlong_t(XDR *xdrs, longlong_t *llp)
{
BUILD_BUG_ON(sizeof (longlong_t) != 8);
return (xdrs->x_ops->xdr_u_longlong_t(xdrs, (u_longlong_t *)llp));
}
/*
* Fixed-length opaque data.
*/
static inline bool_t xdr_opaque(XDR *xdrs, caddr_t cp, const uint_t cnt)
{
return (xdrs->x_ops->xdr_opaque(xdrs, cp, cnt));
}
/*
* Variable-length string.
* The *sp buffer must have (maxsize + 1) bytes.
*/
static inline bool_t xdr_string(XDR *xdrs, char **sp, const uint_t maxsize)
{
return (xdrs->x_ops->xdr_string(xdrs, sp, maxsize));
}
/*
* Variable-length arrays.
*/
static inline bool_t xdr_array(XDR *xdrs, caddr_t *arrp, uint_t *sizep,
const uint_t maxsize, const uint_t elsize, const xdrproc_t elproc)
{
return xdrs->x_ops->xdr_array(xdrs, arrp, sizep, maxsize, elsize,
elproc);
}
#endif /* SPL_RPC_XDR_H */
diff --git a/include/os/linux/spl/sys/sysmacros.h b/include/os/linux/spl/sys/sysmacros.h
index 98d1ab1d7f8a..be1f77e43bda 100644
--- a/include/os/linux/spl/sys/sysmacros.h
+++ b/include/os/linux/spl/sys/sysmacros.h
@@ -1,206 +1,205 @@
/*
* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
* Copyright (C) 2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
* UCRL-CODE-235197
*
* This file is part of the SPL, Solaris Porting Layer.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef _SPL_SYSMACROS_H
#define _SPL_SYSMACROS_H
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/sched/rt.h>
#include <linux/cpumask.h>
#include <sys/debug.h>
#include <sys/zone.h>
#include <sys/signal.h>
#include <asm/page.h>
#ifndef _KERNEL
#define _KERNEL __KERNEL__
#endif
#define FALSE 0
#define TRUE 1
#define INT8_MAX (127)
#define INT8_MIN (-128)
#define UINT8_MAX (255)
#define UINT8_MIN (0)
#define INT16_MAX (32767)
#define INT16_MIN (-32768)
#define UINT16_MAX (65535)
#define UINT16_MIN (0)
#define INT32_MAX INT_MAX
#define INT32_MIN INT_MIN
#define UINT32_MAX UINT_MAX
#define UINT32_MIN UINT_MIN
#define INT64_MAX LLONG_MAX
#define INT64_MIN LLONG_MIN
#define UINT64_MAX ULLONG_MAX
#define UINT64_MIN ULLONG_MIN
#define NBBY 8
#define MAXMSGLEN 256
#define MAXNAMELEN 256
#define MAXPATHLEN 4096
#define MAXOFFSET_T LLONG_MAX
#define MAXBSIZE 8192
#define DEV_BSIZE 512
#define DEV_BSHIFT 9 /* log2(DEV_BSIZE) */
#define proc_pageout NULL
#define curproc current
#define max_ncpus num_possible_cpus()
#define boot_ncpus num_online_cpus()
#define CPU_SEQID smp_processor_id()
#define CPU_SEQID_UNSTABLE raw_smp_processor_id()
#define is_system_labeled() 0
#ifndef RLIM64_INFINITY
#define RLIM64_INFINITY (~0ULL)
#endif
/*
* 0..MAX_PRIO-1: Process priority
* 0..MAX_RT_PRIO-1: RT priority tasks
* MAX_RT_PRIO..MAX_PRIO-1: SCHED_NORMAL tasks
*
* Treat shim tasks as SCHED_NORMAL tasks
*/
#define minclsyspri (MAX_PRIO-1)
#define maxclsyspri (MAX_RT_PRIO)
#define defclsyspri (DEFAULT_PRIO)
#ifndef NICE_TO_PRIO
#define NICE_TO_PRIO(nice) (MAX_RT_PRIO + (nice) + 20)
#endif
#ifndef PRIO_TO_NICE
#define PRIO_TO_NICE(prio) ((prio) - MAX_RT_PRIO - 20)
#endif
/*
* Missing macros
*/
#ifndef PAGESIZE
#define PAGESIZE PAGE_SIZE
#endif
#ifndef PAGESHIFT
#define PAGESHIFT PAGE_SHIFT
#endif
/* Missing globals */
-extern char spl_gitrev[64];
extern unsigned long spl_hostid;
/* Missing misc functions */
extern uint32_t zone_get_hostid(void *zone);
extern void spl_setup(void);
extern void spl_cleanup(void);
#define highbit(x) __fls(x)
#define lowbit(x) __ffs(x)
#define highbit64(x) fls64(x)
#define makedevice(maj, min) makedev(maj, min)
/* common macros */
#ifndef MIN
#define MIN(a, b) ((a) < (b) ? (a) : (b))
#endif
#ifndef MAX
#define MAX(a, b) ((a) < (b) ? (b) : (a))
#endif
#ifndef ABS
#define ABS(a) ((a) < 0 ? -(a) : (a))
#endif
#ifndef DIV_ROUND_UP
#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
#endif
#ifndef roundup
#define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
#endif
#ifndef howmany
#define howmany(x, y) (((x) + ((y) - 1)) / (y))
#endif
/*
* Compatibility macros/typedefs needed for Solaris -> Linux port
*/
#define P2ALIGN(x, align) ((x) & -(align))
#define P2CROSS(x, y, align) (((x) ^ (y)) > (align) - 1)
#define P2ROUNDUP(x, align) ((((x) - 1) | ((align) - 1)) + 1)
#define P2PHASE(x, align) ((x) & ((align) - 1))
#define P2NPHASE(x, align) (-(x) & ((align) - 1))
#define ISP2(x) (((x) & ((x) - 1)) == 0)
#define IS_P2ALIGNED(v, a) ((((uintptr_t)(v)) & ((uintptr_t)(a) - 1)) == 0)
#define P2BOUNDARY(off, len, align) \
(((off) ^ ((off) + (len) - 1)) > (align) - 1)
/*
* Typed version of the P2* macros. These macros should be used to ensure
* that the result is correctly calculated based on the data type of (x),
* which is passed in as the last argument, regardless of the data
* type of the alignment. For example, if (x) is of type uint64_t,
* and we want to round it up to a page boundary using "PAGESIZE" as
* the alignment, we can do either
*
* P2ROUNDUP(x, (uint64_t)PAGESIZE)
* or
* P2ROUNDUP_TYPED(x, PAGESIZE, uint64_t)
*/
#define P2ALIGN_TYPED(x, align, type) \
((type)(x) & -(type)(align))
#define P2PHASE_TYPED(x, align, type) \
((type)(x) & ((type)(align) - 1))
#define P2NPHASE_TYPED(x, align, type) \
(-(type)(x) & ((type)(align) - 1))
#define P2ROUNDUP_TYPED(x, align, type) \
((((type)(x) - 1) | ((type)(align) - 1)) + 1)
#define P2END_TYPED(x, align, type) \
(-(~(type)(x) & -(type)(align)))
#define P2PHASEUP_TYPED(x, align, phase, type) \
((type)(phase) - (((type)(phase) - (type)(x)) & -(type)(align)))
#define P2CROSS_TYPED(x, y, align, type) \
(((type)(x) ^ (type)(y)) > (type)(align) - 1)
#define P2SAMEHIGHBIT_TYPED(x, y, type) \
(((type)(x) ^ (type)(y)) < ((type)(x) & (type)(y)))
#define SET_ERROR(err) \
(__set_error(__FILE__, __func__, __LINE__, err), err)
#include <linux/sort.h>
#define qsort(base, num, size, cmp) \
sort(base, num, size, cmp, NULL)
#if !defined(_KMEMUSER) && !defined(offsetof)
/* avoid any possibility of clashing with <stddef.h> version */
#define offsetof(s, m) ((size_t)(&(((s *)0)->m)))
#endif
#endif /* _SPL_SYSMACROS_H */
diff --git a/include/os/linux/spl/sys/taskq.h b/include/os/linux/spl/sys/taskq.h
index b50175a10873..2a6cd8283d16 100644
--- a/include/os/linux/spl/sys/taskq.h
+++ b/include/os/linux/spl/sys/taskq.h
@@ -1,168 +1,169 @@
/*
* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
* Copyright (C) 2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
* UCRL-CODE-235197
*
* This file is part of the SPL, Solaris Porting Layer.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef _SPL_TASKQ_H
#define _SPL_TASKQ_H
#include <linux/module.h>
#include <linux/gfp.h>
#include <linux/slab.h>
#include <linux/interrupt.h>
#include <linux/kthread.h>
#include <sys/types.h>
#include <sys/thread.h>
#include <sys/rwlock.h>
#include <sys/wait.h>
#define TASKQ_NAMELEN 31
#define TASKQ_PREPOPULATE 0x00000001
#define TASKQ_CPR_SAFE 0x00000002
#define TASKQ_DYNAMIC 0x00000004
#define TASKQ_THREADS_CPU_PCT 0x00000008
#define TASKQ_DC_BATCH 0x00000010
#define TASKQ_ACTIVE 0x80000000
/*
* Flags for taskq_dispatch. TQ_SLEEP/TQ_NOSLEEP should be same as
* KM_SLEEP/KM_NOSLEEP. TQ_NOQUEUE/TQ_NOALLOC are set particularly
* large so as not to conflict with already used GFP_* defines.
*/
#define TQ_SLEEP 0x00000000
#define TQ_NOSLEEP 0x00000001
#define TQ_PUSHPAGE 0x00000002
#define TQ_NOQUEUE 0x01000000
#define TQ_NOALLOC 0x02000000
#define TQ_NEW 0x04000000
#define TQ_FRONT 0x08000000
/*
* Reserved taskqid values.
*/
#define TASKQID_INVALID ((taskqid_t)0)
#define TASKQID_INITIAL ((taskqid_t)1)
/*
* spin_lock(lock) and spin_lock_nested(lock,0) are equivalent,
* so TQ_LOCK_DYNAMIC must not evaluate to 0
*/
typedef enum tq_lock_role {
TQ_LOCK_GENERAL = 0,
TQ_LOCK_DYNAMIC = 1,
} tq_lock_role_t;
typedef unsigned long taskqid_t;
typedef void (task_func_t)(void *);
typedef struct taskq {
spinlock_t tq_lock; /* protects taskq_t */
char *tq_name; /* taskq name */
int tq_instance; /* instance of tq_name */
struct list_head tq_thread_list; /* list of all threads */
struct list_head tq_active_list; /* list of active threads */
int tq_nactive; /* # of active threads */
int tq_nthreads; /* # of existing threads */
int tq_nspawn; /* # of threads being spawned */
int tq_maxthreads; /* # of threads maximum */
/* If PERCPU flag is set, percent of NCPUs to have as threads */
int tq_cpu_pct;
int tq_pri; /* priority */
int tq_minalloc; /* min taskq_ent_t pool size */
int tq_maxalloc; /* max taskq_ent_t pool size */
int tq_nalloc; /* cur taskq_ent_t pool size */
uint_t tq_flags; /* flags */
taskqid_t tq_next_id; /* next pend/work id */
taskqid_t tq_lowest_id; /* lowest pend/work id */
struct list_head tq_free_list; /* free taskq_ent_t's */
struct list_head tq_pend_list; /* pending taskq_ent_t's */
struct list_head tq_prio_list; /* priority taskq_ent_t's */
struct list_head tq_delay_list; /* delayed taskq_ent_t's */
struct list_head tq_taskqs; /* all taskq_t's */
spl_wait_queue_head_t tq_work_waitq; /* new work waitq */
spl_wait_queue_head_t tq_wait_waitq; /* wait waitq */
tq_lock_role_t tq_lock_class; /* class when taking tq_lock */
/* list node for the cpu hotplug callback */
struct hlist_node tq_hp_cb_node;
boolean_t tq_hp_support;
} taskq_t;
typedef struct taskq_ent {
spinlock_t tqent_lock;
spl_wait_queue_head_t tqent_waitq;
struct timer_list tqent_timer;
struct list_head tqent_list;
taskqid_t tqent_id;
task_func_t *tqent_func;
void *tqent_arg;
taskq_t *tqent_taskq;
uintptr_t tqent_flags;
unsigned long tqent_birth;
} taskq_ent_t;
#define TQENT_FLAG_PREALLOC 0x1
#define TQENT_FLAG_CANCEL 0x2
typedef struct taskq_thread {
struct list_head tqt_thread_list;
struct list_head tqt_active_list;
struct task_struct *tqt_thread;
taskq_t *tqt_tq;
taskqid_t tqt_id;
taskq_ent_t *tqt_task;
uintptr_t tqt_flags;
} taskq_thread_t;
/* Global system-wide dynamic task queue available for all consumers */
extern taskq_t *system_taskq;
/* Global dynamic task queue for long delay */
extern taskq_t *system_delay_taskq;
/* List of all taskqs */
extern struct list_head tq_list;
extern struct rw_semaphore tq_list_sem;
extern taskqid_t taskq_dispatch(taskq_t *, task_func_t, void *, uint_t);
extern taskqid_t taskq_dispatch_delay(taskq_t *, task_func_t, void *,
uint_t, clock_t);
extern void taskq_dispatch_ent(taskq_t *, task_func_t, void *, uint_t,
taskq_ent_t *);
extern int taskq_empty_ent(taskq_ent_t *);
extern void taskq_init_ent(taskq_ent_t *);
extern taskq_t *taskq_create(const char *, int, pri_t, int, int, uint_t);
extern void taskq_destroy(taskq_t *);
extern void taskq_wait_id(taskq_t *, taskqid_t);
extern void taskq_wait_outstanding(taskq_t *, taskqid_t);
extern void taskq_wait(taskq_t *);
extern int taskq_cancel_id(taskq_t *, taskqid_t);
extern int taskq_member(taskq_t *, kthread_t *);
extern taskq_t *taskq_of_curthread(void);
#define taskq_create_proc(name, nthreads, pri, min, max, proc, flags) \
taskq_create(name, nthreads, pri, min, max, flags)
#define taskq_create_sysdc(name, nthreads, min, max, proc, dc, flags) \
- taskq_create(name, nthreads, maxclsyspri, min, max, flags)
+ ((void) sizeof (dc), \
+ taskq_create(name, nthreads, maxclsyspri, min, max, flags))
int spl_taskq_init(void);
void spl_taskq_fini(void);
#endif /* _SPL_TASKQ_H */
diff --git a/include/os/linux/zfs/sys/zfs_znode_impl.h b/include/os/linux/zfs/sys/zfs_znode_impl.h
index de46fc8f2bd8..2e1bd857614d 100644
--- a/include/os/linux/zfs/sys/zfs_znode_impl.h
+++ b/include/os/linux/zfs/sys/zfs_znode_impl.h
@@ -1,183 +1,182 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
*/
#ifndef _SYS_ZFS_ZNODE_IMPL_H
#define _SYS_ZFS_ZNODE_IMPL_H
#ifndef _KERNEL
#error "no user serviceable parts within"
#endif
#include <sys/isa_defs.h>
#include <sys/types32.h>
#include <sys/list.h>
#include <sys/dmu.h>
#include <sys/sa.h>
#include <sys/time.h>
#include <sys/zfs_vfsops.h>
#include <sys/rrwlock.h>
#include <sys/zfs_sa.h>
#include <sys/zfs_stat.h>
#include <sys/zfs_rlock.h>
#ifdef __cplusplus
extern "C" {
#endif
#define ZNODE_OS_FIELDS \
inode_timespec_t z_btime; /* creation/birth time (cached) */ \
struct inode z_inode;
/*
* Convert between znode pointers and inode pointers
*/
#define ZTOI(znode) (&((znode)->z_inode))
#define ITOZ(inode) (container_of((inode), znode_t, z_inode))
#define ZTOZSB(znode) ((zfsvfs_t *)(ZTOI(znode)->i_sb->s_fs_info))
#define ITOZSB(inode) ((zfsvfs_t *)((inode)->i_sb->s_fs_info))
#define ZTOTYPE(zp) (ZTOI(zp)->i_mode)
#define ZTOGID(zp) (ZTOI(zp)->i_gid)
#define ZTOUID(zp) (ZTOI(zp)->i_uid)
#define ZTONLNK(zp) (ZTOI(zp)->i_nlink)
#define Z_ISBLK(type) S_ISBLK(type)
#define Z_ISCHR(type) S_ISCHR(type)
#define Z_ISLNK(type) S_ISLNK(type)
#define Z_ISDEV(type) (S_ISCHR(type) || S_ISBLK(type) || S_ISFIFO(type))
#define Z_ISDIR(type) S_ISDIR(type)
#define zn_has_cached_data(zp) ((zp)->z_is_mapped)
#define zn_flush_cached_data(zp, sync) write_inode_now(ZTOI(zp), sync)
#define zn_rlimit_fsize(zp, uio) (0)
/*
* zhold() wraps igrab() on Linux, and igrab() may fail when the
* inode is in the process of being deleted. As zhold() must only be
* called when a ref already exists - so the inode cannot be
* mid-deletion - we VERIFY() this.
*/
#define zhold(zp) VERIFY3P(igrab(ZTOI((zp))), !=, NULL)
#define zrele(zp) iput(ZTOI((zp)))
/* Called on entry to each ZFS inode and vfs operation. */
#define ZFS_ENTER_ERROR(zfsvfs, error) \
do { \
ZFS_TEARDOWN_ENTER_READ(zfsvfs, FTAG); \
if (unlikely((zfsvfs)->z_unmounted)) { \
ZFS_TEARDOWN_EXIT_READ(zfsvfs, FTAG); \
return (error); \
} \
} while (0)
#define ZFS_ENTER(zfsvfs) ZFS_ENTER_ERROR(zfsvfs, EIO)
#define ZPL_ENTER(zfsvfs) ZFS_ENTER_ERROR(zfsvfs, -EIO)
/* Must be called before exiting the operation. */
#define ZFS_EXIT(zfsvfs) \
do { \
zfs_exit_fs(zfsvfs); \
ZFS_TEARDOWN_EXIT_READ(zfsvfs, FTAG); \
} while (0)
#define ZPL_EXIT(zfsvfs) \
do { \
rrm_exit(&(zfsvfs)->z_teardown_lock, FTAG); \
} while (0)
/* Verifies the znode is valid. */
#define ZFS_VERIFY_ZP_ERROR(zp, error) \
do { \
if (unlikely((zp)->z_sa_hdl == NULL)) { \
ZFS_EXIT(ZTOZSB(zp)); \
return (error); \
} \
} while (0)
#define ZFS_VERIFY_ZP(zp) ZFS_VERIFY_ZP_ERROR(zp, EIO)
#define ZPL_VERIFY_ZP(zp) ZFS_VERIFY_ZP_ERROR(zp, -EIO)
/*
* Macros for dealing with dmu_buf_hold
*/
#define ZFS_OBJ_MTX_SZ 64
#define ZFS_OBJ_MTX_MAX (1024 * 1024)
#define ZFS_OBJ_HASH(zfsvfs, obj) ((obj) & ((zfsvfs->z_hold_size) - 1))
extern unsigned int zfs_object_mutex_size;
/*
* Encode ZFS stored time values from a struct timespec / struct timespec64.
*/
#define ZFS_TIME_ENCODE(tp, stmp) \
do { \
(stmp)[0] = (uint64_t)(tp)->tv_sec; \
(stmp)[1] = (uint64_t)(tp)->tv_nsec; \
} while (0)
#if defined(HAVE_INODE_TIMESPEC64_TIMES)
/*
* Decode ZFS stored time values to a struct timespec64
* 4.18 and newer kernels.
*/
#define ZFS_TIME_DECODE(tp, stmp) \
do { \
(tp)->tv_sec = (time64_t)(stmp)[0]; \
(tp)->tv_nsec = (long)(stmp)[1]; \
} while (0)
#else
/*
* Decode ZFS stored time values to a struct timespec
* 4.17 and older kernels.
*/
#define ZFS_TIME_DECODE(tp, stmp) \
do { \
(tp)->tv_sec = (time_t)(stmp)[0]; \
(tp)->tv_nsec = (long)(stmp)[1]; \
} while (0)
#endif /* HAVE_INODE_TIMESPEC64_TIMES */
#define ZFS_ACCESSTIME_STAMP(zfsvfs, zp)
struct znode;
extern int zfs_sync(struct super_block *, int, cred_t *);
extern int zfs_inode_alloc(struct super_block *, struct inode **ip);
extern void zfs_inode_destroy(struct inode *);
extern void zfs_mark_inode_dirty(struct inode *);
extern boolean_t zfs_relatime_need_update(const struct inode *);
#if defined(HAVE_UIO_RW)
extern caddr_t zfs_map_page(page_t *, enum seg_rw);
extern void zfs_unmap_page(page_t *, caddr_t);
#endif /* HAVE_UIO_RW */
-extern zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE];
-extern int zfsfstype;
+extern zil_replay_func_t *const zfs_replay_vector[TX_MAX_TYPE];
#ifdef __cplusplus
}
#endif
#endif /* _SYS_ZFS_ZNODE_IMPL_H */
diff --git a/include/os/linux/zfs/sys/zpl.h b/include/os/linux/zfs/sys/zpl.h
index ff86e027bbe2..afb16e5c7907 100644
--- a/include/os/linux/zfs/sys/zpl.h
+++ b/include/os/linux/zfs/sys/zpl.h
@@ -1,201 +1,200 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2011, Lawrence Livermore National Security, LLC.
*/
#ifndef _SYS_ZPL_H
#define _SYS_ZPL_H
#include <sys/mntent.h>
#include <sys/vfs.h>
#include <linux/aio.h>
#include <linux/dcache_compat.h>
#include <linux/exportfs.h>
#include <linux/falloc.h>
#include <linux/parser.h>
#include <linux/task_io_accounting_ops.h>
#include <linux/vfs_compat.h>
#include <linux/writeback.h>
#include <linux/xattr_compat.h>
/* zpl_inode.c */
extern void zpl_vap_init(vattr_t *vap, struct inode *dir,
umode_t mode, cred_t *cr);
extern const struct inode_operations zpl_inode_operations;
extern const struct inode_operations zpl_dir_inode_operations;
extern const struct inode_operations zpl_symlink_inode_operations;
extern const struct inode_operations zpl_special_inode_operations;
extern dentry_operations_t zpl_dentry_operations;
extern const struct address_space_operations zpl_address_space_operations;
extern const struct file_operations zpl_file_operations;
extern const struct file_operations zpl_dir_file_operations;
/* zpl_super.c */
extern void zpl_prune_sb(int64_t nr_to_scan, void *arg);
extern const struct super_operations zpl_super_operations;
extern const struct export_operations zpl_export_operations;
extern struct file_system_type zpl_fs_type;
/* zpl_xattr.c */
extern ssize_t zpl_xattr_list(struct dentry *dentry, char *buf, size_t size);
extern int zpl_xattr_security_init(struct inode *ip, struct inode *dip,
const struct qstr *qstr);
#if defined(CONFIG_FS_POSIX_ACL)
#if defined(HAVE_SET_ACL)
#if defined(HAVE_SET_ACL_USERNS)
extern int zpl_set_acl(struct user_namespace *userns, struct inode *ip,
struct posix_acl *acl, int type);
#else
extern int zpl_set_acl(struct inode *ip, struct posix_acl *acl, int type);
#endif /* HAVE_SET_ACL_USERNS */
#endif /* HAVE_SET_ACL */
#if defined(HAVE_GET_ACL_RCU)
extern struct posix_acl *zpl_get_acl(struct inode *ip, int type, bool rcu);
#elif defined(HAVE_GET_ACL)
extern struct posix_acl *zpl_get_acl(struct inode *ip, int type);
#endif
extern int zpl_init_acl(struct inode *ip, struct inode *dir);
extern int zpl_chmod_acl(struct inode *ip);
#else
static inline int
zpl_init_acl(struct inode *ip, struct inode *dir)
{
return (0);
}
static inline int
zpl_chmod_acl(struct inode *ip)
{
return (0);
}
#endif /* CONFIG_FS_POSIX_ACL */
extern xattr_handler_t *zpl_xattr_handlers[];
/* zpl_ctldir.c */
extern const struct file_operations zpl_fops_root;
extern const struct inode_operations zpl_ops_root;
extern const struct file_operations zpl_fops_snapdir;
extern const struct inode_operations zpl_ops_snapdir;
-extern const struct dentry_operations zpl_dops_snapdirs;
extern const struct file_operations zpl_fops_shares;
extern const struct inode_operations zpl_ops_shares;
#if defined(HAVE_VFS_ITERATE) || defined(HAVE_VFS_ITERATE_SHARED)
#define ZPL_DIR_CONTEXT_INIT(_dirent, _actor, _pos) { \
.actor = _actor, \
.pos = _pos, \
}
typedef struct dir_context zpl_dir_context_t;
#define zpl_dir_emit dir_emit
#define zpl_dir_emit_dot dir_emit_dot
#define zpl_dir_emit_dotdot dir_emit_dotdot
#define zpl_dir_emit_dots dir_emit_dots
#else
typedef struct zpl_dir_context {
void *dirent;
const filldir_t actor;
loff_t pos;
} zpl_dir_context_t;
#define ZPL_DIR_CONTEXT_INIT(_dirent, _actor, _pos) { \
.dirent = _dirent, \
.actor = _actor, \
.pos = _pos, \
}
static inline bool
zpl_dir_emit(zpl_dir_context_t *ctx, const char *name, int namelen,
uint64_t ino, unsigned type)
{
return (!ctx->actor(ctx->dirent, name, namelen, ctx->pos, ino, type));
}
static inline bool
zpl_dir_emit_dot(struct file *file, zpl_dir_context_t *ctx)
{
return (ctx->actor(ctx->dirent, ".", 1, ctx->pos,
file_inode(file)->i_ino, DT_DIR) == 0);
}
static inline bool
zpl_dir_emit_dotdot(struct file *file, zpl_dir_context_t *ctx)
{
return (ctx->actor(ctx->dirent, "..", 2, ctx->pos,
parent_ino(file_dentry(file)), DT_DIR) == 0);
}
static inline bool
zpl_dir_emit_dots(struct file *file, zpl_dir_context_t *ctx)
{
if (ctx->pos == 0) {
if (!zpl_dir_emit_dot(file, ctx))
return (false);
ctx->pos = 1;
}
if (ctx->pos == 1) {
if (!zpl_dir_emit_dotdot(file, ctx))
return (false);
ctx->pos = 2;
}
return (true);
}
#endif /* HAVE_VFS_ITERATE */
#if defined(HAVE_INODE_TIMESTAMP_TRUNCATE)
#define zpl_inode_timestamp_truncate(ts, ip) timestamp_truncate(ts, ip)
#elif defined(HAVE_INODE_TIMESPEC64_TIMES)
#define zpl_inode_timestamp_truncate(ts, ip) \
timespec64_trunc(ts, (ip)->i_sb->s_time_gran)
#else
#define zpl_inode_timestamp_truncate(ts, ip) \
timespec_trunc(ts, (ip)->i_sb->s_time_gran)
#endif
#if defined(HAVE_INODE_OWNER_OR_CAPABLE)
#define zpl_inode_owner_or_capable(ns, ip) inode_owner_or_capable(ip)
#elif defined(HAVE_INODE_OWNER_OR_CAPABLE_IDMAPPED)
#define zpl_inode_owner_or_capable(ns, ip) inode_owner_or_capable(ns, ip)
#else
#error "Unsupported kernel"
#endif
#ifdef HAVE_SETATTR_PREPARE_USERNS
#define zpl_setattr_prepare(ns, dentry, ia) setattr_prepare(ns, dentry, ia)
#else
/*
* Use kernel-provided version, or our own from
* linux/vfs_compat.h
*/
#define zpl_setattr_prepare(ns, dentry, ia) setattr_prepare(dentry, ia)
#endif
#endif /* _SYS_ZPL_H */
diff --git a/include/sys/crypto/api.h b/include/sys/crypto/api.h
index 8aecfeaff0f4..17c9a645922e 100644
--- a/include/sys/crypto/api.h
+++ b/include/sys/crypto/api.h
@@ -1,425 +1,425 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#ifndef _SYS_CRYPTO_API_H
#define _SYS_CRYPTO_API_H
#ifdef __cplusplus
extern "C" {
#endif
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
typedef long crypto_req_id_t;
typedef void *crypto_bc_t;
typedef void *crypto_context_t;
typedef void *crypto_ctx_template_t;
typedef uint32_t crypto_call_flag_t;
/* crypto_call_flag's values */
#define CRYPTO_ALWAYS_QUEUE 0x00000001 /* ALWAYS queue the req. */
#define CRYPTO_NOTIFY_OPDONE 0x00000002 /* Notify intermediate steps */
#define CRYPTO_SKIP_REQID 0x00000004 /* Skip request ID generation */
#define CRYPTO_RESTRICTED 0x00000008 /* cannot use restricted prov */
typedef struct {
crypto_call_flag_t cr_flag;
void (*cr_callback_func)(void *, int);
void *cr_callback_arg;
crypto_req_id_t cr_reqid;
} crypto_call_req_t;
/*
* Returns the mechanism type corresponding to a mechanism name.
*/
#define CRYPTO_MECH_INVALID ((uint64_t)-1)
-extern crypto_mech_type_t crypto_mech2id(char *name);
+extern crypto_mech_type_t crypto_mech2id(const char *name);
/*
* Create and destroy context templates.
*/
extern int crypto_create_ctx_template(crypto_mechanism_t *mech,
crypto_key_t *key, crypto_ctx_template_t *tmpl, int kmflag);
extern void crypto_destroy_ctx_template(crypto_ctx_template_t tmpl);
/*
* Single and multi-part digest operations.
*/
extern int crypto_digest(crypto_mechanism_t *mech, crypto_data_t *data,
crypto_data_t *digest, crypto_call_req_t *cr);
extern int crypto_digest_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_data_t *, crypto_data_t *,
crypto_call_req_t *);
extern int crypto_digest_init(crypto_mechanism_t *mech, crypto_context_t *ctxp,
crypto_call_req_t *cr);
extern int crypto_digest_init_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_context_t *, crypto_call_req_t *);
extern int crypto_digest_update(crypto_context_t ctx, crypto_data_t *data,
crypto_call_req_t *cr);
extern int crypto_digest_final(crypto_context_t ctx, crypto_data_t *digest,
crypto_call_req_t *cr);
/*
* Single and multi-part MAC operations.
*/
extern int crypto_mac(crypto_mechanism_t *mech, crypto_data_t *data,
crypto_key_t *key, crypto_ctx_template_t tmpl, crypto_data_t *mac,
crypto_call_req_t *cr);
extern int crypto_mac_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_data_t *, crypto_key_t *,
crypto_ctx_template_t, crypto_data_t *, crypto_call_req_t *);
extern int crypto_mac_verify(crypto_mechanism_t *mech, crypto_data_t *data,
crypto_key_t *key, crypto_ctx_template_t tmpl, crypto_data_t *mac,
crypto_call_req_t *cr);
extern int crypto_mac_verify_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_data_t *, crypto_key_t *,
crypto_ctx_template_t, crypto_data_t *, crypto_call_req_t *);
extern int crypto_mac_init(crypto_mechanism_t *mech, crypto_key_t *key,
crypto_ctx_template_t tmpl, crypto_context_t *ctxp, crypto_call_req_t *cr);
extern int crypto_mac_init_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_ctx_template_t,
crypto_context_t *, crypto_call_req_t *);
extern int crypto_mac_update(crypto_context_t ctx, crypto_data_t *data,
crypto_call_req_t *cr);
extern int crypto_mac_final(crypto_context_t ctx, crypto_data_t *data,
crypto_call_req_t *cr);
/*
* Single and multi-part sign with private key operations.
*/
extern int crypto_sign(crypto_mechanism_t *mech, crypto_key_t *key,
crypto_data_t *data, crypto_ctx_template_t tmpl,
crypto_data_t *signature, crypto_call_req_t *cr);
extern int crypto_sign_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_ctx_template_t, crypto_data_t *, crypto_call_req_t *);
extern int crypto_sign_init(crypto_mechanism_t *mech, crypto_key_t *key,
crypto_ctx_template_t tmpl, crypto_context_t *ctxp, crypto_call_req_t *cr);
extern int crypto_sign_init_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_ctx_template_t,
crypto_context_t *, crypto_call_req_t *);
extern int crypto_sign_update(crypto_context_t ctx, crypto_data_t *data,
crypto_call_req_t *cr);
extern int crypto_sign_final(crypto_context_t ctx, crypto_data_t *signature,
crypto_call_req_t *cr);
extern int crypto_sign_recover_init_prov(crypto_provider_t,
crypto_session_id_t, crypto_mechanism_t *, crypto_key_t *,
crypto_ctx_template_t tmpl, crypto_context_t *, crypto_call_req_t *);
extern int crypto_sign_recover(crypto_mechanism_t *mech, crypto_key_t *key,
crypto_data_t *data, crypto_ctx_template_t tmpl, crypto_data_t *signature,
crypto_call_req_t *cr);
extern int crypto_sign_recover_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_ctx_template_t, crypto_data_t *, crypto_call_req_t *);
/*
* Single and multi-part verify with public key operations.
*/
extern int crypto_verify(crypto_mechanism_t *mech, crypto_key_t *key,
crypto_data_t *data, crypto_ctx_template_t tmpl, crypto_data_t *signature,
crypto_call_req_t *cr);
extern int crypto_verify_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_ctx_template_t, crypto_data_t *, crypto_call_req_t *);
extern int crypto_verify_init(crypto_mechanism_t *mech, crypto_key_t *key,
crypto_ctx_template_t tmpl, crypto_context_t *ctxp, crypto_call_req_t *cr);
extern int crypto_verify_init_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_ctx_template_t,
crypto_context_t *, crypto_call_req_t *);
extern int crypto_verify_update(crypto_context_t ctx, crypto_data_t *data,
crypto_call_req_t *cr);
extern int crypto_verify_final(crypto_context_t ctx, crypto_data_t *signature,
crypto_call_req_t *cr);
extern int crypto_verify_recover_init_prov(crypto_provider_t,
crypto_session_id_t, crypto_mechanism_t *, crypto_key_t *,
crypto_ctx_template_t tmpl, crypto_context_t *, crypto_call_req_t *);
extern int crypto_verify_recover(crypto_mechanism_t *mech, crypto_key_t *key,
crypto_data_t *signature, crypto_ctx_template_t tmpl, crypto_data_t *data,
crypto_call_req_t *cr);
extern int crypto_verify_recover_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_ctx_template_t, crypto_data_t *, crypto_call_req_t *);
/*
* Single and multi-part encryption operations.
*/
extern int crypto_encrypt(crypto_mechanism_t *mech, crypto_data_t *plaintext,
crypto_key_t *key, crypto_ctx_template_t tmpl, crypto_data_t *ciphertext,
crypto_call_req_t *cr);
extern int crypto_encrypt_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_data_t *, crypto_key_t *,
crypto_ctx_template_t, crypto_data_t *, crypto_call_req_t *);
extern int crypto_encrypt_init(crypto_mechanism_t *mech, crypto_key_t *key,
crypto_ctx_template_t tmpl, crypto_context_t *ctxp, crypto_call_req_t *cr);
extern int crypto_encrypt_init_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_ctx_template_t,
crypto_context_t *, crypto_call_req_t *);
extern int crypto_encrypt_update(crypto_context_t ctx,
crypto_data_t *plaintext, crypto_data_t *ciphertext,
crypto_call_req_t *cr);
extern int crypto_encrypt_final(crypto_context_t ctx,
crypto_data_t *ciphertext, crypto_call_req_t *cr);
/*
* Single and multi-part decryption operations.
*/
extern int crypto_decrypt(crypto_mechanism_t *mech, crypto_data_t *ciphertext,
crypto_key_t *key, crypto_ctx_template_t tmpl, crypto_data_t *plaintext,
crypto_call_req_t *cr);
extern int crypto_decrypt_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_data_t *, crypto_key_t *,
crypto_ctx_template_t, crypto_data_t *, crypto_call_req_t *);
extern int crypto_decrypt_init(crypto_mechanism_t *mech, crypto_key_t *key,
crypto_ctx_template_t tmpl, crypto_context_t *ctxp,
crypto_call_req_t *cr);
extern int crypto_decrypt_init_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_ctx_template_t,
crypto_context_t *, crypto_call_req_t *);
extern int crypto_decrypt_update(crypto_context_t ctx,
crypto_data_t *ciphertext, crypto_data_t *plaintext,
crypto_call_req_t *cr);
extern int crypto_decrypt_final(crypto_context_t ctx, crypto_data_t *plaintext,
crypto_call_req_t *cr);
/*
* Single and multi-part encrypt/MAC dual operations.
*/
extern int crypto_encrypt_mac(crypto_mechanism_t *encr_mech,
crypto_mechanism_t *mac_mech, crypto_data_t *pt,
crypto_key_t *encr_key, crypto_key_t *mac_key,
crypto_ctx_template_t encr_tmpl, crypto_ctx_template_t mac_tmpl,
crypto_dual_data_t *ct, crypto_data_t *mac, crypto_call_req_t *cr);
extern int crypto_encrypt_mac_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_mechanism_t *, crypto_data_t *,
crypto_key_t *, crypto_key_t *, crypto_ctx_template_t,
crypto_ctx_template_t, crypto_dual_data_t *, crypto_data_t *,
crypto_call_req_t *);
extern int crypto_encrypt_mac_init(crypto_mechanism_t *encr_mech,
crypto_mechanism_t *mac_mech, crypto_key_t *encr_key,
crypto_key_t *mac_key, crypto_ctx_template_t encr_tmpl,
crypto_ctx_template_t mac_tmpl, crypto_context_t *ctxp,
crypto_call_req_t *cr);
extern int crypto_encrypt_mac_init_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_mechanism_t *, crypto_key_t *, crypto_key_t *,
crypto_ctx_template_t, crypto_ctx_template_t, crypto_context_t *,
crypto_call_req_t *);
extern int crypto_encrypt_mac_update(crypto_context_t ctx,
crypto_data_t *pt, crypto_dual_data_t *ct, crypto_call_req_t *cr);
extern int crypto_encrypt_mac_final(crypto_context_t ctx,
crypto_dual_data_t *ct, crypto_data_t *mac, crypto_call_req_t *cr);
/*
* Single and multi-part MAC/decrypt dual operations.
*/
extern int crypto_mac_decrypt(crypto_mechanism_t *mac_mech,
crypto_mechanism_t *decr_mech, crypto_dual_data_t *ct,
crypto_key_t *mac_key, crypto_key_t *decr_key,
crypto_ctx_template_t mac_tmpl, crypto_ctx_template_t decr_tmpl,
crypto_data_t *mac, crypto_data_t *pt, crypto_call_req_t *cr);
extern int crypto_mac_decrypt_prov(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *mac_mech, crypto_mechanism_t *decr_mech,
crypto_dual_data_t *ct, crypto_key_t *mac_key, crypto_key_t *decr_key,
crypto_ctx_template_t mac_tmpl, crypto_ctx_template_t decr_tmpl,
crypto_data_t *mac, crypto_data_t *pt, crypto_call_req_t *cr);
extern int crypto_mac_verify_decrypt(crypto_mechanism_t *mac_mech,
crypto_mechanism_t *decr_mech, crypto_dual_data_t *ct,
crypto_key_t *mac_key, crypto_key_t *decr_key,
crypto_ctx_template_t mac_tmpl, crypto_ctx_template_t decr_tmpl,
crypto_data_t *mac, crypto_data_t *pt, crypto_call_req_t *cr);
extern int crypto_mac_verify_decrypt_prov(crypto_provider_t,
crypto_session_id_t, crypto_mechanism_t *mac_mech,
crypto_mechanism_t *decr_mech, crypto_dual_data_t *ct,
crypto_key_t *mac_key, crypto_key_t *decr_key,
crypto_ctx_template_t mac_tmpl, crypto_ctx_template_t decr_tmpl,
crypto_data_t *mac, crypto_data_t *pt, crypto_call_req_t *cr);
extern int crypto_mac_decrypt_init(crypto_mechanism_t *mac_mech,
crypto_mechanism_t *decr_mech, crypto_key_t *mac_key,
crypto_key_t *decr_key, crypto_ctx_template_t mac_tmpl,
crypto_ctx_template_t decr_tmpl, crypto_context_t *ctxp,
crypto_call_req_t *cr);
extern int crypto_mac_decrypt_init_prov(crypto_provider_t,
crypto_session_id_t, crypto_mechanism_t *mac_mech,
crypto_mechanism_t *decr_mech, crypto_key_t *mac_key,
crypto_key_t *decr_key, crypto_ctx_template_t mac_tmpl,
crypto_ctx_template_t decr_tmpl, crypto_context_t *ctxp,
crypto_call_req_t *cr);
extern int crypto_mac_decrypt_update(crypto_context_t ctx,
crypto_dual_data_t *ct, crypto_data_t *pt, crypto_call_req_t *cr);
extern int crypto_mac_decrypt_final(crypto_context_t ctx, crypto_data_t *mac,
crypto_data_t *pt, crypto_call_req_t *cr);
/* Session Management */
extern int crypto_session_open(crypto_provider_t, crypto_session_id_t *,
crypto_call_req_t *);
extern int crypto_session_close(crypto_provider_t, crypto_session_id_t,
crypto_call_req_t *);
extern int crypto_session_login(crypto_provider_t, crypto_session_id_t,
crypto_user_type_t, char *, size_t, crypto_call_req_t *);
extern int crypto_session_logout(crypto_provider_t, crypto_session_id_t,
crypto_call_req_t *);
/* Object Management */
extern int crypto_object_copy(crypto_provider_t, crypto_session_id_t,
crypto_object_id_t, crypto_object_attribute_t *, uint_t,
crypto_object_id_t *, crypto_call_req_t *);
extern int crypto_object_create(crypto_provider_t, crypto_session_id_t,
crypto_object_attribute_t *, uint_t, crypto_object_id_t *,
crypto_call_req_t *);
extern int crypto_object_destroy(crypto_provider_t, crypto_session_id_t,
crypto_object_id_t, crypto_call_req_t *);
extern int crypto_object_get_attribute_value(crypto_provider_t,
crypto_session_id_t, crypto_object_id_t, crypto_object_attribute_t *,
uint_t, crypto_call_req_t *);
extern int crypto_object_get_size(crypto_provider_t, crypto_session_id_t,
crypto_object_id_t, size_t *, crypto_call_req_t *);
extern int crypto_object_find_final(crypto_provider_t, void *,
crypto_call_req_t *);
extern int crypto_object_find_init(crypto_provider_t, crypto_session_id_t,
crypto_object_attribute_t *, uint_t, void **, crypto_call_req_t *);
extern int crypto_object_find(crypto_provider_t, void *, crypto_object_id_t *,
uint_t *, uint_t, crypto_call_req_t *);
extern int crypto_object_set_attribute_value(crypto_provider_t,
crypto_session_id_t, crypto_object_id_t, crypto_object_attribute_t *,
uint_t, crypto_call_req_t *);
/* Key Management */
extern int crypto_key_derive(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_object_attribute_t *,
uint_t, crypto_object_id_t *, crypto_call_req_t *);
extern int crypto_key_generate(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_object_attribute_t *, uint_t,
crypto_object_id_t *, crypto_call_req_t *);
extern int crypto_key_generate_pair(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_object_attribute_t *, uint_t,
crypto_object_attribute_t *, uint_t, crypto_object_id_t *,
crypto_object_id_t *, crypto_call_req_t *);
extern int crypto_key_unwrap(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, uchar_t *, size_t *,
crypto_object_attribute_t *, uint_t, crypto_object_id_t *,
crypto_call_req_t *);
extern int crypto_key_wrap(crypto_provider_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_object_id_t *, uchar_t *,
size_t *, crypto_call_req_t *);
extern int crypto_key_check_prov(crypto_provider_t, crypto_mechanism_t *mech,
crypto_key_t *key);
extern int crypto_key_check(crypto_mechanism_t *mech, crypto_key_t *key);
/*
* Routines to cancel a single asynchronous request or all asynchronous
* requests associated with a particular context.
*/
extern void crypto_cancel_req(crypto_req_id_t req);
extern void crypto_cancel_ctx(crypto_context_t ctx);
/*
* crypto_get_mech_list(9F) allocates and returns the list of currently
* supported cryptographic mechanisms.
*/
extern crypto_mech_name_t *crypto_get_mech_list(uint_t *count, int kmflag);
extern void crypto_free_mech_list(crypto_mech_name_t *mech_names,
uint_t count);
extern crypto_provider_t crypto_get_provider(char *, char *, char *);
extern int crypto_get_provinfo(crypto_provider_t, crypto_provider_ext_info_t *);
extern void crypto_release_provider(crypto_provider_t);
/*
* A kernel consumer can request to be notified when some particular event
* occurs. The valid events, callback function type, and functions to
* be called to register or unregister for notification are defined below.
*/
#define CRYPTO_EVENT_MECHS_CHANGED 0x00000001
#define CRYPTO_EVENT_PROVIDER_REGISTERED 0x00000002
#define CRYPTO_EVENT_PROVIDER_UNREGISTERED 0x00000004
typedef enum {
CRYPTO_MECH_ADDED = 1,
CRYPTO_MECH_REMOVED
} crypto_event_change_t;
/* The event_arg argument structure for CRYPTO_EVENT_PROVIDERS_CHANGE event */
typedef struct crypto_notify_event_change {
crypto_mech_name_t ec_mech_name;
crypto_provider_type_t ec_provider_type;
crypto_event_change_t ec_change;
} crypto_notify_event_change_t;
typedef void *crypto_notify_handle_t;
typedef void (*crypto_notify_callback_t)(uint32_t event_mask, void *event_arg);
extern crypto_notify_handle_t crypto_notify_events(
crypto_notify_callback_t nf, uint32_t event_mask);
extern void crypto_unnotify_events(crypto_notify_handle_t);
/*
* crypto_bufcall(9F) group of routines.
*/
extern crypto_bc_t crypto_bufcall_alloc(void);
extern int crypto_bufcall_free(crypto_bc_t bc);
extern int crypto_bufcall(crypto_bc_t bc, void (*func)(void *arg), void *arg);
extern int crypto_unbufcall(crypto_bc_t bc);
/*
* To obtain the list of key size ranges supported by a mechanism.
*/
#define CRYPTO_MECH_USAGE_ENCRYPT 0x00000001
#define CRYPTO_MECH_USAGE_DECRYPT 0x00000002
#define CRYPTO_MECH_USAGE_MAC 0x00000004
typedef uint32_t crypto_mech_usage_t;
typedef struct crypto_mechanism_info {
size_t mi_min_key_size;
size_t mi_max_key_size;
crypto_keysize_unit_t mi_keysize_unit; /* for mi_xxx_key_size */
crypto_mech_usage_t mi_usage;
} crypto_mechanism_info_t;
#ifdef _SYSCALL32
typedef struct crypto_mechanism_info32 {
size32_t mi_min_key_size;
size32_t mi_max_key_size;
crypto_keysize_unit_t mi_keysize_unit; /* for mi_xxx_key_size */
crypto_mech_usage_t mi_usage;
} crypto_mechanism_info32_t;
#endif /* _SYSCALL32 */
extern int crypto_get_all_mech_info(crypto_mech_type_t,
crypto_mechanism_info_t **, uint_t *, int);
extern void crypto_free_all_mech_info(crypto_mechanism_info_t *, uint_t);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_CRYPTO_API_H */
diff --git a/include/sys/dmu.h b/include/sys/dmu.h
index b9e6731bd41a..1ddff0d4e4e7 100644
--- a/include/sys/dmu.h
+++ b/include/sys/dmu.h
@@ -1,1075 +1,1074 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2012, Joyent, Inc. All rights reserved.
* Copyright 2014 HybridCluster. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright 2013 Saso Kiselkov. All rights reserved.
* Copyright (c) 2017, Intel Corporation.
*/
/* Portions Copyright 2010 Robert Milkowski */
#ifndef _SYS_DMU_H
#define _SYS_DMU_H
/*
* This file describes the interface that the DMU provides for its
* consumers.
*
* The DMU also interacts with the SPA. That interface is described in
* dmu_spa.h.
*/
#include <sys/zfs_context.h>
#include <sys/inttypes.h>
#include <sys/cred.h>
#include <sys/fs/zfs.h>
#include <sys/zio_compress.h>
#include <sys/zio_priority.h>
#include <sys/uio.h>
#include <sys/zfs_file.h>
#ifdef __cplusplus
extern "C" {
#endif
struct page;
struct vnode;
struct spa;
struct zilog;
struct zio;
struct blkptr;
struct zap_cursor;
struct dsl_dataset;
struct dsl_pool;
struct dnode;
struct drr_begin;
struct drr_end;
struct zbookmark_phys;
struct spa;
struct nvlist;
struct arc_buf;
struct zio_prop;
struct sa_handle;
struct dsl_crypto_params;
struct locked_range;
typedef struct objset objset_t;
typedef struct dmu_tx dmu_tx_t;
typedef struct dsl_dir dsl_dir_t;
typedef struct dnode dnode_t;
typedef enum dmu_object_byteswap {
DMU_BSWAP_UINT8,
DMU_BSWAP_UINT16,
DMU_BSWAP_UINT32,
DMU_BSWAP_UINT64,
DMU_BSWAP_ZAP,
DMU_BSWAP_DNODE,
DMU_BSWAP_OBJSET,
DMU_BSWAP_ZNODE,
DMU_BSWAP_OLDACL,
DMU_BSWAP_ACL,
/*
* Allocating a new byteswap type number makes the on-disk format
* incompatible with any other format that uses the same number.
*
* Data can usually be structured to work with one of the
* DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
*/
DMU_BSWAP_NUMFUNCS
} dmu_object_byteswap_t;
#define DMU_OT_NEWTYPE 0x80
#define DMU_OT_METADATA 0x40
#define DMU_OT_ENCRYPTED 0x20
#define DMU_OT_BYTESWAP_MASK 0x1f
/*
* Defines a uint8_t object type. Object types specify if the data
* in the object is metadata (boolean) and how to byteswap the data
* (dmu_object_byteswap_t). All of the types created by this method
* are cached in the dbuf metadata cache.
*/
#define DMU_OT(byteswap, metadata, encrypted) \
(DMU_OT_NEWTYPE | \
((metadata) ? DMU_OT_METADATA : 0) | \
((encrypted) ? DMU_OT_ENCRYPTED : 0) | \
((byteswap) & DMU_OT_BYTESWAP_MASK))
#define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
(ot) < DMU_OT_NUMTYPES)
#define DMU_OT_IS_METADATA_CACHED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
B_TRUE : dmu_ot[(ot)].ot_dbuf_metadata_cache)
/*
* MDB doesn't have dmu_ot; it defines these macros itself.
*/
#ifndef ZFS_MDB
#define DMU_OT_IS_METADATA_IMPL(ot) (dmu_ot[ot].ot_metadata)
#define DMU_OT_IS_ENCRYPTED_IMPL(ot) (dmu_ot[ot].ot_encrypt)
#define DMU_OT_BYTESWAP_IMPL(ot) (dmu_ot[ot].ot_byteswap)
#endif
#define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
((ot) & DMU_OT_METADATA) : \
DMU_OT_IS_METADATA_IMPL(ot))
#define DMU_OT_IS_DDT(ot) \
((ot) == DMU_OT_DDT_ZAP)
/* Note: ztest uses DMU_OT_UINT64_OTHER as a proxy for file blocks */
#define DMU_OT_IS_FILE(ot) \
((ot) == DMU_OT_PLAIN_FILE_CONTENTS || (ot) == DMU_OT_UINT64_OTHER)
#define DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
((ot) & DMU_OT_ENCRYPTED) : \
DMU_OT_IS_ENCRYPTED_IMPL(ot))
/*
* These object types use bp_fill != 1 for their L0 bp's. Therefore they can't
* have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill
* is repurposed for embedded BPs.
*/
#define DMU_OT_HAS_FILL(ot) \
((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET)
#define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
((ot) & DMU_OT_BYTESWAP_MASK) : \
DMU_OT_BYTESWAP_IMPL(ot))
typedef enum dmu_object_type {
DMU_OT_NONE,
/* general: */
DMU_OT_OBJECT_DIRECTORY, /* ZAP */
DMU_OT_OBJECT_ARRAY, /* UINT64 */
DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
DMU_OT_BPOBJ, /* UINT64 */
DMU_OT_BPOBJ_HDR, /* UINT64 */
/* spa: */
DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
DMU_OT_SPACE_MAP, /* UINT64 */
/* zil: */
DMU_OT_INTENT_LOG, /* UINT64 */
/* dmu: */
DMU_OT_DNODE, /* DNODE */
DMU_OT_OBJSET, /* OBJSET */
/* dsl: */
DMU_OT_DSL_DIR, /* UINT64 */
DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
DMU_OT_DSL_PROPS, /* ZAP */
DMU_OT_DSL_DATASET, /* UINT64 */
/* zpl: */
DMU_OT_ZNODE, /* ZNODE */
DMU_OT_OLDACL, /* Old ACL */
DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
DMU_OT_MASTER_NODE, /* ZAP */
DMU_OT_UNLINKED_SET, /* ZAP */
/* zvol: */
DMU_OT_ZVOL, /* UINT8 */
DMU_OT_ZVOL_PROP, /* ZAP */
/* other; for testing only! */
DMU_OT_PLAIN_OTHER, /* UINT8 */
DMU_OT_UINT64_OTHER, /* UINT64 */
DMU_OT_ZAP_OTHER, /* ZAP */
/* new object types: */
DMU_OT_ERROR_LOG, /* ZAP */
DMU_OT_SPA_HISTORY, /* UINT8 */
DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
DMU_OT_POOL_PROPS, /* ZAP */
DMU_OT_DSL_PERMS, /* ZAP */
DMU_OT_ACL, /* ACL */
DMU_OT_SYSACL, /* SYSACL */
DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
DMU_OT_NEXT_CLONES, /* ZAP */
DMU_OT_SCAN_QUEUE, /* ZAP */
DMU_OT_USERGROUP_USED, /* ZAP */
DMU_OT_USERGROUP_QUOTA, /* ZAP */
DMU_OT_USERREFS, /* ZAP */
DMU_OT_DDT_ZAP, /* ZAP */
DMU_OT_DDT_STATS, /* ZAP */
DMU_OT_SA, /* System attr */
DMU_OT_SA_MASTER_NODE, /* ZAP */
DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
DMU_OT_SCAN_XLATE, /* ZAP */
DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
DMU_OT_DEADLIST, /* ZAP */
DMU_OT_DEADLIST_HDR, /* UINT64 */
DMU_OT_DSL_CLONES, /* ZAP */
DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
/*
* Do not allocate new object types here. Doing so makes the on-disk
* format incompatible with any other format that uses the same object
* type number.
*
* When creating an object which does not have one of the above types
* use the DMU_OTN_* type with the correct byteswap and metadata
* values.
*
* The DMU_OTN_* types do not have entries in the dmu_ot table,
* use the DMU_OT_IS_METADATA() and DMU_OT_BYTESWAP() macros instead
* of indexing into dmu_ot directly (this works for both DMU_OT_* types
* and DMU_OTN_* types).
*/
DMU_OT_NUMTYPES,
/*
* Names for valid types declared with DMU_OT().
*/
DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_FALSE),
DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_FALSE),
DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_FALSE),
DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_FALSE),
DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_FALSE),
DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_FALSE),
DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_FALSE),
DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_FALSE),
DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_FALSE),
DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_FALSE),
DMU_OTN_UINT8_ENC_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_TRUE),
DMU_OTN_UINT8_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_TRUE),
DMU_OTN_UINT16_ENC_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_TRUE),
DMU_OTN_UINT16_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_TRUE),
DMU_OTN_UINT32_ENC_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_TRUE),
DMU_OTN_UINT32_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_TRUE),
DMU_OTN_UINT64_ENC_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_TRUE),
DMU_OTN_UINT64_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_TRUE),
DMU_OTN_ZAP_ENC_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_TRUE),
DMU_OTN_ZAP_ENC_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_TRUE),
} dmu_object_type_t;
/*
* These flags are intended to be used to specify the "txg_how"
* parameter when calling the dmu_tx_assign() function. See the comment
* above dmu_tx_assign() for more details on the meaning of these flags.
*/
#define TXG_NOWAIT (0ULL)
#define TXG_WAIT (1ULL<<0)
#define TXG_NOTHROTTLE (1ULL<<1)
void byteswap_uint64_array(void *buf, size_t size);
void byteswap_uint32_array(void *buf, size_t size);
void byteswap_uint16_array(void *buf, size_t size);
void byteswap_uint8_array(void *buf, size_t size);
void zap_byteswap(void *buf, size_t size);
void zfs_oldacl_byteswap(void *buf, size_t size);
void zfs_acl_byteswap(void *buf, size_t size);
void zfs_znode_byteswap(void *buf, size_t size);
#define DS_FIND_SNAPSHOTS (1<<0)
#define DS_FIND_CHILDREN (1<<1)
#define DS_FIND_SERIALIZE (1<<2)
/*
* The maximum number of bytes that can be accessed as part of one
* operation, including metadata.
*/
#define DMU_MAX_ACCESS (64 * 1024 * 1024) /* 64MB */
#define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
#define DMU_USERUSED_OBJECT (-1ULL)
#define DMU_GROUPUSED_OBJECT (-2ULL)
#define DMU_PROJECTUSED_OBJECT (-3ULL)
/*
* Zap prefix for object accounting in DMU_{USER,GROUP,PROJECT}USED_OBJECT.
*/
#define DMU_OBJACCT_PREFIX "obj-"
#define DMU_OBJACCT_PREFIX_LEN 4
/*
* artificial blkids for bonus buffer and spill blocks
*/
#define DMU_BONUS_BLKID (-1ULL)
#define DMU_SPILL_BLKID (-2ULL)
/*
* Public routines to create, destroy, open, and close objsets.
*/
typedef void dmu_objset_create_sync_func_t(objset_t *os, void *arg,
cred_t *cr, dmu_tx_t *tx);
int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
int dmu_objset_own(const char *name, dmu_objset_type_t type,
boolean_t readonly, boolean_t key_required, void *tag, objset_t **osp);
void dmu_objset_rele(objset_t *os, void *tag);
void dmu_objset_disown(objset_t *os, boolean_t key_required, void *tag);
int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp);
void dmu_objset_evict_dbufs(objset_t *os);
int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
struct dsl_crypto_params *dcp, dmu_objset_create_sync_func_t func,
void *arg);
int dmu_objset_clone(const char *name, const char *origin);
int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
struct nvlist *errlist);
int dmu_objset_snapshot_one(const char *fsname, const char *snapname);
int dmu_objset_find(const char *name, int func(const char *, void *), void *arg,
int flags);
void dmu_objset_byteswap(void *buf, size_t size);
int dsl_dataset_rename_snapshot(const char *fsname,
const char *oldsnapname, const char *newsnapname, boolean_t recursive);
typedef struct dmu_buf {
uint64_t db_object; /* object that this buffer is part of */
uint64_t db_offset; /* byte offset in this object */
uint64_t db_size; /* size of buffer in bytes */
void *db_data; /* data in buffer */
} dmu_buf_t;
/*
* The names of zap entries in the DIRECTORY_OBJECT of the MOS.
*/
#define DMU_POOL_DIRECTORY_OBJECT 1
#define DMU_POOL_CONFIG "config"
#define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
#define DMU_POOL_FEATURES_FOR_READ "features_for_read"
#define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
#define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg"
#define DMU_POOL_ROOT_DATASET "root_dataset"
#define DMU_POOL_SYNC_BPOBJ "sync_bplist"
#define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
#define DMU_POOL_ERRLOG_LAST "errlog_last"
#define DMU_POOL_SPARES "spares"
#define DMU_POOL_DEFLATE "deflate"
#define DMU_POOL_HISTORY "history"
#define DMU_POOL_PROPS "pool_props"
#define DMU_POOL_L2CACHE "l2cache"
#define DMU_POOL_TMP_USERREFS "tmp_userrefs"
#define DMU_POOL_DDT "DDT-%s-%s-%s"
#define DMU_POOL_DDT_STATS "DDT-statistics"
#define DMU_POOL_CREATION_VERSION "creation_version"
#define DMU_POOL_SCAN "scan"
#define DMU_POOL_FREE_BPOBJ "free_bpobj"
#define DMU_POOL_BPTREE_OBJ "bptree_obj"
#define DMU_POOL_EMPTY_BPOBJ "empty_bpobj"
#define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt"
#define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map"
#define DMU_POOL_REMOVING "com.delphix:removing"
#define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj"
#define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect"
#define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint"
#define DMU_POOL_LOG_SPACEMAP_ZAP "com.delphix:log_spacemap_zap"
#define DMU_POOL_DELETED_CLONES "com.delphix:deleted_clones"
/*
* Allocate an object from this objset. The range of object numbers
* available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
*
* The transaction must be assigned to a txg. The newly allocated
* object will be "held" in the transaction (ie. you can modify the
* newly allocated object in this transaction).
*
* dmu_object_alloc() chooses an object and returns it in *objectp.
*
* dmu_object_claim() allocates a specific object number. If that
* number is already allocated, it fails and returns EEXIST.
*
* Return 0 on success, or ENOSPC or EEXIST as specified above.
*/
uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
uint64_t dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize,
int indirect_blockshift,
dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot,
int blocksize, dmu_object_type_t bonus_type, int bonus_len,
int dnodesize, dmu_tx_t *tx);
uint64_t dmu_object_alloc_hold(objset_t *os, dmu_object_type_t ot,
int blocksize, int indirect_blockshift, dmu_object_type_t bonustype,
int bonuslen, int dnodesize, dnode_t **allocated_dnode, void *tag,
dmu_tx_t *tx);
int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
int blocksize, dmu_object_type_t bonus_type, int bonus_len,
int dnodesize, dmu_tx_t *tx);
int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp);
int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object,
dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype,
int bonuslen, int dnodesize, boolean_t keep_spill, dmu_tx_t *tx);
int dmu_object_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx);
/*
* Free an object from this objset.
*
* The object's data will be freed as well (ie. you don't need to call
* dmu_free(object, 0, -1, tx)).
*
* The object need not be held in the transaction.
*
* If there are any holds on this object's buffers (via dmu_buf_hold()),
* or tx holds on the object (via dmu_tx_hold_object()), you can not
* free it; it fails and returns EBUSY.
*
* If the object is not allocated, it fails and returns ENOENT.
*
* Return 0 on success, or EBUSY or ENOENT as specified above.
*/
int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx);
/*
* Find the next allocated or free object.
*
* The objectp parameter is in-out. It will be updated to be the next
* object which is allocated. Ignore objects which have not been
* modified since txg.
*
* XXX Can only be called on a objset with no dirty data.
*
* Returns 0 on success, or ENOENT if there are no more objects.
*/
int dmu_object_next(objset_t *os, uint64_t *objectp,
boolean_t hole, uint64_t txg);
/*
* Set the number of levels on a dnode. nlevels must be greater than the
* current number of levels or an EINVAL will be returned.
*/
int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels,
dmu_tx_t *tx);
/*
* Set the data blocksize for an object.
*
* The object cannot have any blocks allocated beyond the first. If
* the first block is allocated already, the new size must be greater
* than the current block size. If these conditions are not met,
* ENOTSUP will be returned.
*
* Returns 0 on success, or EBUSY if there are any holds on the object
* contents, or ENOTSUP as described above.
*/
int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
int ibs, dmu_tx_t *tx);
/*
* Manually set the maxblkid on a dnode. This will adjust nlevels accordingly
* to accommodate the change. When calling this function, the caller must
* ensure that the object's nlevels can sufficiently support the new maxblkid.
*/
int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid,
dmu_tx_t *tx);
/*
* Set the checksum property on a dnode. The new checksum algorithm will
* apply to all newly written blocks; existing blocks will not be affected.
*/
void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
dmu_tx_t *tx);
/*
* Set the compress property on a dnode. The new compression algorithm will
* apply to all newly written blocks; existing blocks will not be affected.
*/
void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
dmu_tx_t *tx);
void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
int compressed_size, int byteorder, dmu_tx_t *tx);
void dmu_redact(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
dmu_tx_t *tx);
/*
* Decide how to write a block: checksum, compression, number of copies, etc.
*/
#define WP_NOFILL 0x1
#define WP_DMU_SYNC 0x2
#define WP_SPILL 0x4
void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp,
struct zio_prop *zp);
/*
* The bonus data is accessed more or less like a regular buffer.
* You must dmu_bonus_hold() to get the buffer, which will give you a
* dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
* data. As with any normal buffer, you must call dmu_buf_will_dirty()
* before modifying it, and the
* object must be held in an assigned transaction before calling
* dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
* buffer as well. You must release what you hold with dmu_buf_rele().
*
* Returns ENOENT, EIO, or 0.
*/
int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp);
int dmu_bonus_hold_by_dnode(dnode_t *dn, void *tag, dmu_buf_t **dbp,
uint32_t flags);
int dmu_bonus_max(void);
int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *);
int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *);
dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *);
/*
* Special spill buffer support used by "SA" framework
*/
int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, uint32_t flags, void *tag,
dmu_buf_t **dbp);
int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags,
void *tag, dmu_buf_t **dbp);
int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp);
/*
* Obtain the DMU buffer from the specified object which contains the
* specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
* that it will remain in memory. You must release the hold with
* dmu_buf_rele(). You must not access the dmu_buf_t after releasing
* what you hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
*
* You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
* on the returned buffer before reading or writing the buffer's
* db_data. The comments for those routines describe what particular
* operations are valid after calling them.
*
* The object number must be a valid, allocated object number.
*/
int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
void *tag, dmu_buf_t **, int flags);
int dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp);
int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
void *tag, dmu_buf_t **dbp, int flags);
int dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset,
uint64_t length, boolean_t read, void *tag, int *numbufsp,
dmu_buf_t ***dbpp, uint32_t flags);
/*
* Add a reference to a dmu buffer that has already been held via
* dmu_buf_hold() in the current context.
*/
void dmu_buf_add_ref(dmu_buf_t *db, void* tag);
/*
* Attempt to add a reference to a dmu buffer that is in an unknown state,
* using a pointer that may have been invalidated by eviction processing.
* The request will succeed if the passed in dbuf still represents the
* same os/object/blkid, is ineligible for eviction, and has at least
* one hold by a user other than the syncer.
*/
boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object,
uint64_t blkid, void *tag);
void dmu_buf_rele(dmu_buf_t *db, void *tag);
uint64_t dmu_buf_refcount(dmu_buf_t *db);
uint64_t dmu_buf_user_refcount(dmu_buf_t *db);
/*
* dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
* range of an object. A pointer to an array of dmu_buf_t*'s is
* returned (in *dbpp).
*
* dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
* frees the array. The hold on the array of buffers MUST be released
* with dmu_buf_rele_array. You can NOT release the hold on each buffer
* individually with dmu_buf_rele.
*/
int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
uint64_t length, boolean_t read, void *tag,
int *numbufsp, dmu_buf_t ***dbpp);
void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag);
typedef void dmu_buf_evict_func_t(void *user_ptr);
/*
* A DMU buffer user object may be associated with a dbuf for the
* duration of its lifetime. This allows the user of a dbuf (client)
* to attach private data to a dbuf (e.g. in-core only data such as a
* dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified
* when that dbuf has been evicted. Clients typically respond to the
* eviction notification by freeing their private data, thus ensuring
* the same lifetime for both dbuf and private data.
*
* The mapping from a dmu_buf_user_t to any client private data is the
* client's responsibility. All current consumers of the API with private
* data embed a dmu_buf_user_t as the first member of the structure for
* their private data. This allows conversions between the two types
* with a simple cast. Since the DMU buf user API never needs access
* to the private data, other strategies can be employed if necessary
* or convenient for the client (e.g. using container_of() to do the
* conversion for private data that cannot have the dmu_buf_user_t as
* its first member).
*
* Eviction callbacks are executed without the dbuf mutex held or any
* other type of mechanism to guarantee that the dbuf is still available.
* For this reason, users must assume the dbuf has already been freed
* and not reference the dbuf from the callback context.
*
* Users requesting "immediate eviction" are notified as soon as the dbuf
* is only referenced by dirty records (dirties == holds). Otherwise the
* notification occurs after eviction processing for the dbuf begins.
*/
typedef struct dmu_buf_user {
/*
* Asynchronous user eviction callback state.
*/
taskq_ent_t dbu_tqent;
/*
* This instance's eviction function pointers.
*
* dbu_evict_func_sync is called synchronously and then
* dbu_evict_func_async is executed asynchronously on a taskq.
*/
dmu_buf_evict_func_t *dbu_evict_func_sync;
dmu_buf_evict_func_t *dbu_evict_func_async;
#ifdef ZFS_DEBUG
/*
* Pointer to user's dbuf pointer. NULL for clients that do
* not associate a dbuf with their user data.
*
* The dbuf pointer is cleared upon eviction so as to catch
* use-after-evict bugs in clients.
*/
dmu_buf_t **dbu_clear_on_evict_dbufp;
#endif
} dmu_buf_user_t;
/*
* Initialize the given dmu_buf_user_t instance with the eviction function
* evict_func, to be called when the user is evicted.
*
* NOTE: This function should only be called once on a given dmu_buf_user_t.
* To allow enforcement of this, dbu must already be zeroed on entry.
*/
static inline void
dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync,
dmu_buf_evict_func_t *evict_func_async,
dmu_buf_t **clear_on_evict_dbufp __maybe_unused)
{
ASSERT(dbu->dbu_evict_func_sync == NULL);
ASSERT(dbu->dbu_evict_func_async == NULL);
/* must have at least one evict func */
IMPLY(evict_func_sync == NULL, evict_func_async != NULL);
dbu->dbu_evict_func_sync = evict_func_sync;
dbu->dbu_evict_func_async = evict_func_async;
taskq_init_ent(&dbu->dbu_tqent);
#ifdef ZFS_DEBUG
dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp;
#endif
}
/*
* Attach user data to a dbuf and mark it for normal (when the dbuf's
* data is cleared or its reference count goes to zero) eviction processing.
*
* Returns NULL on success, or the existing user if another user currently
* owns the buffer.
*/
void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user);
/*
* Attach user data to a dbuf and mark it for immediate (its dirty and
* reference counts are equal) eviction processing.
*
* Returns NULL on success, or the existing user if another user currently
* owns the buffer.
*/
void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user);
/*
* Replace the current user of a dbuf.
*
* If given the current user of a dbuf, replaces the dbuf's user with
* "new_user" and returns the user data pointer that was replaced.
* Otherwise returns the current, and unmodified, dbuf user pointer.
*/
void *dmu_buf_replace_user(dmu_buf_t *db,
dmu_buf_user_t *old_user, dmu_buf_user_t *new_user);
/*
* Remove the specified user data for a DMU buffer.
*
* Returns the user that was removed on success, or the current user if
* another user currently owns the buffer.
*/
void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user);
/*
* Returns the user data (dmu_buf_user_t *) associated with this dbuf.
*/
void *dmu_buf_get_user(dmu_buf_t *db);
objset_t *dmu_buf_get_objset(dmu_buf_t *db);
dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db);
void dmu_buf_dnode_exit(dmu_buf_t *db);
/* Block until any in-progress dmu buf user evictions complete. */
void dmu_buf_user_evict_wait(void);
/*
* Returns the blkptr associated with this dbuf, or NULL if not set.
*/
struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
/*
* Indicate that you are going to modify the buffer's data (db_data).
*
* The transaction (tx) must be assigned to a txg (ie. you've called
* dmu_tx_assign()). The buffer's object must be held in the tx
* (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
*/
void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
boolean_t dmu_buf_is_dirty(dmu_buf_t *db, dmu_tx_t *tx);
void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder,
const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx);
/*
* You must create a transaction, then hold the objects which you will
* (or might) modify as part of this transaction. Then you must assign
* the transaction to a transaction group. Once the transaction has
* been assigned, you can modify buffers which belong to held objects as
* part of this transaction. You can't modify buffers before the
* transaction has been assigned; you can't modify buffers which don't
* belong to objects which this transaction holds; you can't hold
* objects once the transaction has been assigned. You may hold an
* object which you are going to free (with dmu_object_free()), but you
* don't have to.
*
* You can abort the transaction before it has been assigned.
*
* Note that you may hold buffers (with dmu_buf_hold) at any time,
* regardless of transaction state.
*/
#define DMU_NEW_OBJECT (-1ULL)
#define DMU_OBJECT_END (-1ULL)
dmu_tx_t *dmu_tx_create(objset_t *os);
void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off,
int len);
void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
uint64_t len);
void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off,
uint64_t len);
void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name);
void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add,
const char *name);
void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn);
void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow);
void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
void dmu_tx_abort(dmu_tx_t *tx);
int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
void dmu_tx_wait(dmu_tx_t *tx);
void dmu_tx_commit(dmu_tx_t *tx);
void dmu_tx_mark_netfree(dmu_tx_t *tx);
/*
* To register a commit callback, dmu_tx_callback_register() must be called.
*
* dcb_data is a pointer to caller private data that is passed on as a
* callback parameter. The caller is responsible for properly allocating and
* freeing it.
*
* When registering a callback, the transaction must be already created, but
* it cannot be committed or aborted. It can be assigned to a txg or not.
*
* The callback will be called after the transaction has been safely written
* to stable storage and will also be called if the dmu_tx is aborted.
* If there is any error which prevents the transaction from being committed to
* disk, the callback will be called with a value of error != 0.
*
* When multiple callbacks are registered to the transaction, the callbacks
* will be called in reverse order to let Lustre, the only user of commit
* callback currently, take the fast path of its commit callback handling.
*/
typedef void dmu_tx_callback_func_t(void *dcb_data, int error);
void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func,
void *dcb_data);
void dmu_tx_do_callbacks(list_t *cb_list, int error);
/*
* Free up the data blocks for a defined range of a file. If size is
* -1, the range from offset to end-of-file is freed.
*/
int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
uint64_t size, dmu_tx_t *tx);
int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
uint64_t size);
int dmu_free_long_object(objset_t *os, uint64_t object);
/*
* Convenience functions.
*
* Canfail routines will return 0 on success, or an errno if there is a
* nonrecoverable I/O error.
*/
#define DMU_READ_PREFETCH 0 /* prefetch */
#define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
#define DMU_READ_NO_DECRYPT 2 /* don't decrypt */
int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
void *buf, uint32_t flags);
int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf,
uint32_t flags);
void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
const void *buf, dmu_tx_t *tx);
void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size,
const void *buf, dmu_tx_t *tx);
void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
dmu_tx_t *tx);
#ifdef _KERNEL
int dmu_read_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size);
int dmu_read_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size);
int dmu_read_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size);
int dmu_write_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size,
dmu_tx_t *tx);
int dmu_write_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size,
dmu_tx_t *tx);
int dmu_write_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size,
dmu_tx_t *tx);
#endif
struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size);
void dmu_return_arcbuf(struct arc_buf *buf);
int dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset,
struct arc_buf *buf, dmu_tx_t *tx);
int dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset,
struct arc_buf *buf, dmu_tx_t *tx);
#define dmu_assign_arcbuf dmu_assign_arcbuf_by_dbuf
-extern int zfs_prefetch_disable;
extern int zfs_max_recordsize;
/*
* Asynchronously try to read in the data.
*/
void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
uint64_t len, enum zio_priority pri);
typedef struct dmu_object_info {
/* All sizes are in bytes unless otherwise indicated. */
uint32_t doi_data_block_size;
uint32_t doi_metadata_block_size;
dmu_object_type_t doi_type;
dmu_object_type_t doi_bonus_type;
uint64_t doi_bonus_size;
uint8_t doi_indirection; /* 2 = dnode->indirect->data */
uint8_t doi_checksum;
uint8_t doi_compress;
uint8_t doi_nblkptr;
uint8_t doi_pad[4];
uint64_t doi_dnodesize;
uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
uint64_t doi_max_offset;
uint64_t doi_fill_count; /* number of non-empty blocks */
} dmu_object_info_t;
typedef void (*const arc_byteswap_func_t)(void *buf, size_t size);
typedef struct dmu_object_type_info {
dmu_object_byteswap_t ot_byteswap;
boolean_t ot_metadata;
boolean_t ot_dbuf_metadata_cache;
boolean_t ot_encrypt;
char *ot_name;
} dmu_object_type_info_t;
typedef const struct dmu_object_byteswap_info {
arc_byteswap_func_t ob_func;
char *ob_name;
} dmu_object_byteswap_info_t;
extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
/*
* Get information on a DMU object.
*
* Return 0 on success or ENOENT if object is not allocated.
*
* If doi is NULL, just indicates whether the object exists.
*/
int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi);
void __dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi);
/* Like dmu_object_info, but faster if you have a held dnode in hand. */
void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi);
/* Like dmu_object_info, but faster if you have a held dbuf in hand. */
void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi);
/*
* Like dmu_object_info_from_db, but faster still when you only care about
* the size.
*/
void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize,
u_longlong_t *nblk512);
void dmu_object_dnsize_from_db(dmu_buf_t *db, int *dnsize);
typedef struct dmu_objset_stats {
uint64_t dds_num_clones; /* number of clones of this */
uint64_t dds_creation_txg;
uint64_t dds_guid;
dmu_objset_type_t dds_type;
uint8_t dds_is_snapshot;
uint8_t dds_inconsistent;
uint8_t dds_redacted;
char dds_origin[ZFS_MAX_DATASET_NAME_LEN];
} dmu_objset_stats_t;
/*
* Get stats on a dataset.
*/
void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat);
/*
* Add entries to the nvlist for all the objset's properties. See
* zfs_prop_table[] and zfs(1m) for details on the properties.
*/
void dmu_objset_stats(objset_t *os, struct nvlist *nv);
/*
* Get the space usage statistics for statvfs().
*
* refdbytes is the amount of space "referenced" by this objset.
* availbytes is the amount of space available to this objset, taking
* into account quotas & reservations, assuming that no other objsets
* use the space first. These values correspond to the 'referenced' and
* 'available' properties, described in the zfs(1m) manpage.
*
* usedobjs and availobjs are the number of objects currently allocated,
* and available.
*/
void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
uint64_t *usedobjsp, uint64_t *availobjsp);
/*
* The fsid_guid is a 56-bit ID that can change to avoid collisions.
* (Contrast with the ds_guid which is a 64-bit ID that will never
* change, so there is a small probability that it will collide.)
*/
uint64_t dmu_objset_fsid_guid(objset_t *os);
/*
* Get the [cm]time for an objset's snapshot dir
*/
inode_timespec_t dmu_objset_snap_cmtime(objset_t *os);
int dmu_objset_is_snapshot(objset_t *os);
extern struct spa *dmu_objset_spa(objset_t *os);
extern struct zilog *dmu_objset_zil(objset_t *os);
extern struct dsl_pool *dmu_objset_pool(objset_t *os);
extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
extern void dmu_objset_name(objset_t *os, char *buf);
extern dmu_objset_type_t dmu_objset_type(objset_t *os);
extern uint64_t dmu_objset_id(objset_t *os);
extern uint64_t dmu_objset_dnodesize(objset_t *os);
extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os);
extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os);
extern int dmu_objset_blksize(objset_t *os);
extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
extern int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *val);
extern int dmu_snapshot_realname(objset_t *os, const char *name, char *real,
int maxlen, boolean_t *conflict);
extern int dmu_dir_list_next(objset_t *os, int namelen, char *name,
uint64_t *idp, uint64_t *offp);
typedef struct zfs_file_info {
uint64_t zfi_user;
uint64_t zfi_group;
uint64_t zfi_project;
uint64_t zfi_generation;
} zfs_file_info_t;
typedef int file_info_cb_t(dmu_object_type_t bonustype, const void *data,
struct zfs_file_info *zoi);
extern void dmu_objset_register_type(dmu_objset_type_t ost,
file_info_cb_t *cb);
extern void dmu_objset_set_user(objset_t *os, void *user_ptr);
extern void *dmu_objset_get_user(objset_t *os);
/*
* Return the txg number for the given assigned transaction.
*/
uint64_t dmu_tx_get_txg(dmu_tx_t *tx);
/*
* Synchronous write.
* If a parent zio is provided this function initiates a write on the
* provided buffer as a child of the parent zio.
* In the absence of a parent zio, the write is completed synchronously.
* At write completion, blk is filled with the bp of the written block.
* Note that while the data covered by this function will be on stable
* storage when the write completes this new data does not become a
* permanent part of the file until the associated transaction commits.
*/
/*
* {zfs,zvol,ztest}_get_done() args
*/
typedef struct zgd {
struct lwb *zgd_lwb;
struct blkptr *zgd_bp;
dmu_buf_t *zgd_db;
struct zfs_locked_range *zgd_lr;
void *zgd_private;
} zgd_t;
typedef void dmu_sync_cb_t(zgd_t *arg, int error);
int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd);
/*
* Find the next hole or data block in file starting at *off
* Return found offset in *off. Return ESRCH for end of file.
*/
int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
uint64_t *off);
/*
* Initial setup and final teardown.
*/
extern void dmu_init(void);
extern void dmu_fini(void);
typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp,
uint64_t object, uint64_t offset, int len);
void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
dmu_traverse_cb_t cb, void *arg);
int dmu_diff(const char *tosnap_name, const char *fromsnap_name,
zfs_file_t *fp, offset_t *offp);
/* CRC64 table */
#define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
extern uint64_t zfs_crc64_table[256];
#ifdef __cplusplus
}
#endif
#endif /* _SYS_DMU_H */
diff --git a/include/sys/dmu_recv.h b/include/sys/dmu_recv.h
index 7188b2a02248..1fdb986e2ed6 100644
--- a/include/sys/dmu_recv.h
+++ b/include/sys/dmu_recv.h
@@ -1,87 +1,87 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
*/
#ifndef _DMU_RECV_H
#define _DMU_RECV_H
#include <sys/inttypes.h>
#include <sys/dsl_bookmark.h>
#include <sys/dsl_dataset.h>
#include <sys/spa.h>
#include <sys/objlist.h>
-extern const char *recv_clone_name;
+extern const char *const recv_clone_name;
typedef struct dmu_recv_cookie {
struct dsl_dataset *drc_ds;
struct dmu_replay_record *drc_drr_begin;
struct drr_begin *drc_drrb;
const char *drc_tofs;
const char *drc_tosnap;
boolean_t drc_newfs;
boolean_t drc_byteswap;
uint64_t drc_featureflags;
boolean_t drc_force;
boolean_t drc_resumable;
boolean_t drc_should_save;
boolean_t drc_raw;
boolean_t drc_clone;
boolean_t drc_spill;
nvlist_t *drc_keynvl;
uint64_t drc_fromsnapobj;
uint64_t drc_ivset_guid;
void *drc_owner;
cred_t *drc_cred;
proc_t *drc_proc;
nvlist_t *drc_begin_nvl;
objset_t *drc_os;
zfs_file_t *drc_fp; /* The file to read the stream from */
uint64_t drc_voff; /* The current offset in the stream */
uint64_t drc_bytes_read;
/*
* A record that has had its payload read in, but hasn't yet been handed
* off to the worker thread.
*/
struct receive_record_arg *drc_rrd;
/* A record that has had its header read in, but not its payload. */
struct receive_record_arg *drc_next_rrd;
zio_cksum_t drc_cksum;
zio_cksum_t drc_prev_cksum;
/* Sorted list of objects not to issue prefetches for. */
objlist_t *drc_ignore_objlist;
} dmu_recv_cookie_t;
int dmu_recv_begin(char *, char *, dmu_replay_record_t *,
boolean_t, boolean_t, nvlist_t *, nvlist_t *, char *,
dmu_recv_cookie_t *, zfs_file_t *, offset_t *);
int dmu_recv_stream(dmu_recv_cookie_t *, offset_t *);
int dmu_recv_end(dmu_recv_cookie_t *, void *);
boolean_t dmu_objset_is_receiving(objset_t *);
#endif /* _DMU_RECV_H */
diff --git a/include/sys/dsl_dataset.h b/include/sys/dsl_dataset.h
index 3c9199b861c4..29bbf7e1868a 100644
--- a/include/sys/dsl_dataset.h
+++ b/include/sys/dsl_dataset.h
@@ -1,509 +1,508 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
* Copyright (c) 2013 Steven Hartland. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
*/
#ifndef _SYS_DSL_DATASET_H
#define _SYS_DSL_DATASET_H
#include <sys/dmu.h>
#include <sys/spa.h>
#include <sys/txg.h>
#include <sys/zio.h>
#include <sys/bplist.h>
#include <sys/dsl_synctask.h>
#include <sys/zfs_context.h>
#include <sys/dsl_deadlist.h>
#include <sys/zfs_refcount.h>
#include <sys/rrwlock.h>
#include <sys/dsl_crypt.h>
#include <zfeature_common.h>
#ifdef __cplusplus
extern "C" {
#endif
-extern int zfs_allow_redacted_dataset_mount;
struct dsl_dataset;
struct dsl_dir;
struct dsl_pool;
struct dsl_crypto_params;
struct dsl_key_mapping;
struct zfs_bookmark_phys;
#define DS_FLAG_INCONSISTENT (1ULL<<0)
#define DS_IS_INCONSISTENT(ds) \
(dsl_dataset_phys(ds)->ds_flags & DS_FLAG_INCONSISTENT)
/*
* Do not allow this dataset to be promoted.
*/
#define DS_FLAG_NOPROMOTE (1ULL<<1)
/*
* DS_FLAG_UNIQUE_ACCURATE is set if ds_unique_bytes has been correctly
* calculated for head datasets (starting with SPA_VERSION_UNIQUE_ACCURATE,
* refquota/refreservations).
*/
#define DS_FLAG_UNIQUE_ACCURATE (1ULL<<2)
/*
* DS_FLAG_DEFER_DESTROY is set after 'zfs destroy -d' has been called
* on a dataset. This allows the dataset to be destroyed using 'zfs release'.
*/
#define DS_FLAG_DEFER_DESTROY (1ULL<<3)
#define DS_IS_DEFER_DESTROY(ds) \
(dsl_dataset_phys(ds)->ds_flags & DS_FLAG_DEFER_DESTROY)
/*
* DS_FIELD_* are strings that are used in the "extensified" dataset zap object.
* They should be of the format <reverse-dns>:<field>.
*/
/*
* This field's value is the object ID of a zap object which contains the
* bookmarks of this dataset. If it is present, then this dataset is counted
* in the refcount of the SPA_FEATURES_BOOKMARKS feature.
*/
#define DS_FIELD_BOOKMARK_NAMES "com.delphix:bookmarks"
/*
* This field is present (with value=0) if this dataset may contain large
* dnodes (>512B). If it is present, then this dataset is counted in the
* refcount of the SPA_FEATURE_LARGE_DNODE feature.
*/
#define DS_FIELD_LARGE_DNODE "org.zfsonlinux:large_dnode"
/*
* These fields are set on datasets that are in the middle of a resumable
* receive, and allow the sender to resume the send if it is interrupted.
*/
#define DS_FIELD_RESUME_FROMGUID "com.delphix:resume_fromguid"
#define DS_FIELD_RESUME_TONAME "com.delphix:resume_toname"
#define DS_FIELD_RESUME_TOGUID "com.delphix:resume_toguid"
#define DS_FIELD_RESUME_OBJECT "com.delphix:resume_object"
#define DS_FIELD_RESUME_OFFSET "com.delphix:resume_offset"
#define DS_FIELD_RESUME_BYTES "com.delphix:resume_bytes"
#define DS_FIELD_RESUME_LARGEBLOCK "com.delphix:resume_largeblockok"
#define DS_FIELD_RESUME_EMBEDOK "com.delphix:resume_embedok"
#define DS_FIELD_RESUME_COMPRESSOK "com.delphix:resume_compressok"
#define DS_FIELD_RESUME_RAWOK "com.datto:resume_rawok"
/*
* This field is set to the object number of the remap deadlist if one exists.
*/
#define DS_FIELD_REMAP_DEADLIST "com.delphix:remap_deadlist"
/*
* We were receiving an incremental from a redaction bookmark, and these are the
* guids of its snapshots.
*/
#define DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS \
"com.delphix:resume_redact_book_snaps"
/*
* This field is set to the ivset guid for encrypted snapshots. This is used
* for validating raw receives.
*/
#define DS_FIELD_IVSET_GUID "com.datto:ivset_guid"
/*
* DS_FLAG_CI_DATASET is set if the dataset contains a file system whose
* name lookups should be performed case-insensitively.
*/
#define DS_FLAG_CI_DATASET (1ULL<<16)
#define DS_CREATE_FLAG_NODIRTY (1ULL<<24)
typedef struct dsl_dataset_phys {
uint64_t ds_dir_obj; /* DMU_OT_DSL_DIR */
uint64_t ds_prev_snap_obj; /* DMU_OT_DSL_DATASET */
uint64_t ds_prev_snap_txg;
uint64_t ds_next_snap_obj; /* DMU_OT_DSL_DATASET */
uint64_t ds_snapnames_zapobj; /* DMU_OT_DSL_DS_SNAP_MAP 0 for snaps */
uint64_t ds_num_children; /* clone/snap children; ==0 for head */
uint64_t ds_creation_time; /* seconds since 1970 */
uint64_t ds_creation_txg;
uint64_t ds_deadlist_obj; /* DMU_OT_DEADLIST */
/*
* ds_referenced_bytes, ds_compressed_bytes, and ds_uncompressed_bytes
* include all blocks referenced by this dataset, including those
* shared with any other datasets.
*/
uint64_t ds_referenced_bytes;
uint64_t ds_compressed_bytes;
uint64_t ds_uncompressed_bytes;
uint64_t ds_unique_bytes; /* only relevant to snapshots */
/*
* The ds_fsid_guid is a 56-bit ID that can change to avoid
* collisions. The ds_guid is a 64-bit ID that will never
* change, so there is a small probability that it will collide.
*/
uint64_t ds_fsid_guid;
uint64_t ds_guid;
uint64_t ds_flags; /* DS_FLAG_* */
blkptr_t ds_bp;
uint64_t ds_next_clones_obj; /* DMU_OT_DSL_CLONES */
uint64_t ds_props_obj; /* DMU_OT_DSL_PROPS for snaps */
uint64_t ds_userrefs_obj; /* DMU_OT_USERREFS */
uint64_t ds_pad[5]; /* pad out to 320 bytes for good measure */
} dsl_dataset_phys_t;
typedef struct dsl_dataset {
dmu_buf_user_t ds_dbu;
rrwlock_t ds_bp_rwlock; /* Protects ds_phys->ds_bp */
/* Immutable: */
struct dsl_dir *ds_dir;
dmu_buf_t *ds_dbuf;
uint64_t ds_object;
uint64_t ds_fsid_guid;
boolean_t ds_is_snapshot;
struct dsl_key_mapping *ds_key_mapping;
/* only used in syncing context, only valid for non-snapshots: */
struct dsl_dataset *ds_prev;
uint64_t ds_bookmarks_obj; /* DMU_OTN_ZAP_METADATA */
avl_tree_t ds_bookmarks; /* dsl_bookmark_node_t */
/* has internal locking: */
dsl_deadlist_t ds_deadlist;
bplist_t ds_pending_deadlist;
/*
* The remap deadlist contains blocks (DVA's, really) that are
* referenced by the previous snapshot and point to indirect vdevs,
* but in this dataset they have been remapped to point to concrete
* (or at least, less-indirect) vdevs. In other words, the
* physical DVA is referenced by the previous snapshot but not by
* this dataset. Logically, the DVA continues to be referenced,
* but we are using a different (less indirect) physical DVA.
* This deadlist is used to determine when physical DVAs that
* point to indirect vdevs are no longer referenced anywhere,
* and thus should be marked obsolete.
*
* This is only used if SPA_FEATURE_OBSOLETE_COUNTS is enabled.
*/
dsl_deadlist_t ds_remap_deadlist;
/* protects creation of the ds_remap_deadlist */
kmutex_t ds_remap_deadlist_lock;
/* protected by lock on pool's dp_dirty_datasets list */
txg_node_t ds_dirty_link;
list_node_t ds_synced_link;
/*
* ds_phys->ds_<accounting> is also protected by ds_lock.
* Protected by ds_lock:
*/
kmutex_t ds_lock;
objset_t *ds_objset;
uint64_t ds_userrefs;
void *ds_owner;
/*
* Long holds prevent the ds from being destroyed; they allow the
* ds to remain held even after dropping the dp_config_rwlock.
* Owning counts as a long hold. See the comments above
* dsl_pool_hold() for details.
*/
zfs_refcount_t ds_longholds;
/* no locking; only for making guesses */
uint64_t ds_trysnap_txg;
/* for objset_open() */
kmutex_t ds_opening_lock;
uint64_t ds_reserved; /* cached refreservation */
uint64_t ds_quota; /* cached refquota */
kmutex_t ds_sendstream_lock;
list_t ds_sendstreams;
/*
* When in the middle of a resumable receive, tracks how much
* progress we have made.
*/
uint64_t ds_resume_object[TXG_SIZE];
uint64_t ds_resume_offset[TXG_SIZE];
uint64_t ds_resume_bytes[TXG_SIZE];
/* Protected by our dsl_dir's dd_lock */
list_t ds_prop_cbs;
/*
* For ZFEATURE_FLAG_PER_DATASET features, set if this dataset
* uses this feature.
*/
void *ds_feature[SPA_FEATURES];
/*
* Set if we need to activate the feature on this dataset this txg
* (used only in syncing context).
*/
void *ds_feature_activation[SPA_FEATURES];
/* Protected by ds_lock; keep at end of struct for better locality */
char ds_snapname[ZFS_MAX_DATASET_NAME_LEN];
} dsl_dataset_t;
static inline dsl_dataset_phys_t *
dsl_dataset_phys(dsl_dataset_t *ds)
{
return ((dsl_dataset_phys_t *)ds->ds_dbuf->db_data);
}
typedef struct dsl_dataset_promote_arg {
const char *ddpa_clonename;
dsl_dataset_t *ddpa_clone;
list_t shared_snaps, origin_snaps, clone_snaps;
dsl_dataset_t *origin_origin; /* origin of the origin */
uint64_t used, comp, uncomp, unique, cloneusedsnap, originusedsnap;
nvlist_t *err_ds;
cred_t *cr;
proc_t *proc;
} dsl_dataset_promote_arg_t;
typedef struct dsl_dataset_rollback_arg {
const char *ddra_fsname;
const char *ddra_tosnap;
void *ddra_owner;
nvlist_t *ddra_result;
} dsl_dataset_rollback_arg_t;
typedef struct dsl_dataset_snapshot_arg {
nvlist_t *ddsa_snaps;
nvlist_t *ddsa_props;
nvlist_t *ddsa_errors;
cred_t *ddsa_cr;
proc_t *ddsa_proc;
} dsl_dataset_snapshot_arg_t;
/*
* The max length of a temporary tag prefix is the number of hex digits
* required to express UINT64_MAX plus one for the hyphen.
*/
#define MAX_TAG_PREFIX_LEN 17
#define dsl_dataset_is_snapshot(ds) \
(dsl_dataset_phys(ds)->ds_num_children != 0)
#define DS_UNIQUE_IS_ACCURATE(ds) \
((dsl_dataset_phys(ds)->ds_flags & DS_FLAG_UNIQUE_ACCURATE) != 0)
/* flags for holding the dataset */
typedef enum ds_hold_flags {
DS_HOLD_FLAG_NONE = 0 << 0,
DS_HOLD_FLAG_DECRYPT = 1 << 0 /* needs access to encrypted data */
} ds_hold_flags_t;
int dsl_dataset_hold(struct dsl_pool *dp, const char *name, void *tag,
dsl_dataset_t **dsp);
int dsl_dataset_hold_flags(struct dsl_pool *dp, const char *name,
ds_hold_flags_t flags, void *tag, dsl_dataset_t **dsp);
boolean_t dsl_dataset_try_add_ref(struct dsl_pool *dp, dsl_dataset_t *ds,
void *tag);
int dsl_dataset_create_key_mapping(dsl_dataset_t *ds);
int dsl_dataset_hold_obj_flags(struct dsl_pool *dp, uint64_t dsobj,
ds_hold_flags_t flags, void *tag, dsl_dataset_t **);
void dsl_dataset_remove_key_mapping(dsl_dataset_t *ds);
int dsl_dataset_hold_obj(struct dsl_pool *dp, uint64_t dsobj,
void *tag, dsl_dataset_t **);
void dsl_dataset_rele_flags(dsl_dataset_t *ds, ds_hold_flags_t flags,
void *tag);
void dsl_dataset_rele(dsl_dataset_t *ds, void *tag);
int dsl_dataset_own(struct dsl_pool *dp, const char *name,
ds_hold_flags_t flags, void *tag, dsl_dataset_t **dsp);
int dsl_dataset_own_force(struct dsl_pool *dp, const char *name,
ds_hold_flags_t flags, void *tag, dsl_dataset_t **dsp);
int dsl_dataset_own_obj(struct dsl_pool *dp, uint64_t dsobj,
ds_hold_flags_t flags, void *tag, dsl_dataset_t **dsp);
int dsl_dataset_own_obj_force(struct dsl_pool *dp, uint64_t dsobj,
ds_hold_flags_t flags, void *tag, dsl_dataset_t **dsp);
void dsl_dataset_disown(dsl_dataset_t *ds, ds_hold_flags_t flags, void *tag);
void dsl_dataset_name(dsl_dataset_t *ds, char *name);
boolean_t dsl_dataset_tryown(dsl_dataset_t *ds, void *tag, boolean_t override);
int dsl_dataset_namelen(dsl_dataset_t *ds);
boolean_t dsl_dataset_has_owner(dsl_dataset_t *ds);
uint64_t dsl_dataset_create_sync(dsl_dir_t *pds, const char *lastname,
dsl_dataset_t *origin, uint64_t flags, cred_t *,
struct dsl_crypto_params *, dmu_tx_t *);
uint64_t dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin,
struct dsl_crypto_params *dcp, uint64_t flags, dmu_tx_t *tx);
void dsl_dataset_snapshot_sync(void *arg, dmu_tx_t *tx);
int dsl_dataset_snapshot_check(void *arg, dmu_tx_t *tx);
int dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors);
void dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx);
int dsl_dataset_promote_check(void *arg, dmu_tx_t *tx);
int dsl_dataset_promote(const char *name, char *conflsnap);
int dsl_dataset_rename_snapshot(const char *fsname,
const char *oldsnapname, const char *newsnapname, boolean_t recursive);
int dsl_dataset_snapshot_tmp(const char *fsname, const char *snapname,
minor_t cleanup_minor, const char *htag);
blkptr_t *dsl_dataset_get_blkptr(dsl_dataset_t *ds);
spa_t *dsl_dataset_get_spa(dsl_dataset_t *ds);
boolean_t dsl_dataset_modified_since_snap(dsl_dataset_t *ds,
dsl_dataset_t *snap);
void dsl_dataset_sync(dsl_dataset_t *ds, zio_t *zio, dmu_tx_t *tx);
void dsl_dataset_sync_done(dsl_dataset_t *ds, dmu_tx_t *tx);
void dsl_dataset_block_born(dsl_dataset_t *ds, const blkptr_t *bp,
dmu_tx_t *tx);
int dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp,
dmu_tx_t *tx, boolean_t async);
void dsl_dataset_block_remapped(dsl_dataset_t *ds, uint64_t vdev,
uint64_t offset, uint64_t size, uint64_t birth, dmu_tx_t *tx);
int dsl_dataset_snap_lookup(dsl_dataset_t *ds, const char *name,
uint64_t *value);
void dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx);
int get_clones_stat_impl(dsl_dataset_t *ds, nvlist_t *val);
char *get_receive_resume_stats_impl(dsl_dataset_t *ds);
char *get_child_receive_stats(dsl_dataset_t *ds);
uint64_t dsl_get_refratio(dsl_dataset_t *ds);
uint64_t dsl_get_logicalreferenced(dsl_dataset_t *ds);
uint64_t dsl_get_compressratio(dsl_dataset_t *ds);
uint64_t dsl_get_used(dsl_dataset_t *ds);
uint64_t dsl_get_creation(dsl_dataset_t *ds);
uint64_t dsl_get_creationtxg(dsl_dataset_t *ds);
uint64_t dsl_get_refquota(dsl_dataset_t *ds);
uint64_t dsl_get_refreservation(dsl_dataset_t *ds);
uint64_t dsl_get_guid(dsl_dataset_t *ds);
uint64_t dsl_get_unique(dsl_dataset_t *ds);
uint64_t dsl_get_objsetid(dsl_dataset_t *ds);
uint64_t dsl_get_userrefs(dsl_dataset_t *ds);
uint64_t dsl_get_defer_destroy(dsl_dataset_t *ds);
uint64_t dsl_get_referenced(dsl_dataset_t *ds);
uint64_t dsl_get_numclones(dsl_dataset_t *ds);
uint64_t dsl_get_inconsistent(dsl_dataset_t *ds);
uint64_t dsl_get_redacted(dsl_dataset_t *ds);
uint64_t dsl_get_available(dsl_dataset_t *ds);
int dsl_get_written(dsl_dataset_t *ds, uint64_t *written);
int dsl_get_prev_snap(dsl_dataset_t *ds, char *snap);
void dsl_get_redact_snaps(dsl_dataset_t *ds, nvlist_t *propval);
int dsl_get_mountpoint(dsl_dataset_t *ds, const char *dsname, char *value,
char *source);
void get_clones_stat(dsl_dataset_t *ds, nvlist_t *nv);
void dsl_dataset_stats(dsl_dataset_t *os, nvlist_t *nv);
void dsl_dataset_fast_stat(dsl_dataset_t *ds, dmu_objset_stats_t *stat);
void dsl_dataset_space(dsl_dataset_t *ds,
uint64_t *refdbytesp, uint64_t *availbytesp,
uint64_t *usedobjsp, uint64_t *availobjsp);
uint64_t dsl_dataset_fsid_guid(dsl_dataset_t *ds);
int dsl_dataset_space_written(dsl_dataset_t *oldsnap, dsl_dataset_t *newds,
uint64_t *usedp, uint64_t *compp, uint64_t *uncompp);
int dsl_dataset_space_written_bookmark(struct zfs_bookmark_phys *bmp,
dsl_dataset_t *newds, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp);
int dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap, dsl_dataset_t *last,
uint64_t *usedp, uint64_t *compp, uint64_t *uncompp);
int dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf);
int dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota,
uint64_t asize, uint64_t inflight, uint64_t *used,
uint64_t *ref_rsrv);
int dsl_dataset_set_refquota(const char *dsname, zprop_source_t source,
uint64_t quota);
int dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source,
uint64_t reservation);
int dsl_dataset_set_compression(const char *dsname, zprop_source_t source,
uint64_t compression);
boolean_t dsl_dataset_is_before(dsl_dataset_t *later, dsl_dataset_t *earlier,
uint64_t earlier_txg);
-void dsl_dataset_long_hold(dsl_dataset_t *ds, void *tag);
-void dsl_dataset_long_rele(dsl_dataset_t *ds, void *tag);
+void dsl_dataset_long_hold(dsl_dataset_t *ds, const void *tag);
+void dsl_dataset_long_rele(dsl_dataset_t *ds, const void *tag);
boolean_t dsl_dataset_long_held(dsl_dataset_t *ds);
int dsl_dataset_clone_swap_check_impl(dsl_dataset_t *clone,
dsl_dataset_t *origin_head, boolean_t force, void *owner, dmu_tx_t *tx);
void dsl_dataset_clone_swap_sync_impl(dsl_dataset_t *clone,
dsl_dataset_t *origin_head, dmu_tx_t *tx);
int dsl_dataset_snapshot_check_impl(dsl_dataset_t *ds, const char *snapname,
dmu_tx_t *tx, boolean_t recv, uint64_t cnt, cred_t *cr, proc_t *proc);
void dsl_dataset_snapshot_sync_impl(dsl_dataset_t *ds, const char *snapname,
dmu_tx_t *tx);
void dsl_dataset_remove_from_next_clones(dsl_dataset_t *ds, uint64_t obj,
dmu_tx_t *tx);
void dsl_dataset_recalc_head_uniq(dsl_dataset_t *ds);
int dsl_dataset_get_snapname(dsl_dataset_t *ds);
int dsl_dataset_snap_lookup(dsl_dataset_t *ds, const char *name,
uint64_t *value);
int dsl_dataset_snap_remove(dsl_dataset_t *ds, const char *name, dmu_tx_t *tx,
boolean_t adj_cnt);
void dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t *ds,
zprop_source_t source, uint64_t value, dmu_tx_t *tx);
void dsl_dataset_zapify(dsl_dataset_t *ds, dmu_tx_t *tx);
boolean_t dsl_dataset_is_zapified(dsl_dataset_t *ds);
boolean_t dsl_dataset_has_resume_receive_state(dsl_dataset_t *ds);
int dsl_dataset_rollback_check(void *arg, dmu_tx_t *tx);
void dsl_dataset_rollback_sync(void *arg, dmu_tx_t *tx);
int dsl_dataset_rollback(const char *fsname, const char *tosnap, void *owner,
nvlist_t *result);
uint64_t dsl_dataset_get_remap_deadlist_object(dsl_dataset_t *ds);
void dsl_dataset_create_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx);
boolean_t dsl_dataset_remap_deadlist_exists(dsl_dataset_t *ds);
void dsl_dataset_destroy_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx);
void dsl_dataset_activate_feature(uint64_t dsobj, spa_feature_t f, void *arg,
dmu_tx_t *tx);
void dsl_dataset_deactivate_feature(dsl_dataset_t *ds, spa_feature_t f,
dmu_tx_t *tx);
boolean_t dsl_dataset_feature_is_active(dsl_dataset_t *ds, spa_feature_t f);
boolean_t dsl_dataset_get_uint64_array_feature(dsl_dataset_t *ds,
spa_feature_t f, uint64_t *outlength, uint64_t **outp);
void dsl_dataset_activate_redaction(dsl_dataset_t *ds, uint64_t *redact_snaps,
uint64_t num_redact_snaps, dmu_tx_t *tx);
#ifdef ZFS_DEBUG
#define dprintf_ds(ds, fmt, ...) do { \
if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
char *__ds_name = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); \
dsl_dataset_name(ds, __ds_name); \
dprintf("ds=%s " fmt, __ds_name, __VA_ARGS__); \
kmem_free(__ds_name, ZFS_MAX_DATASET_NAME_LEN); \
} \
} while (0)
#else
#define dprintf_ds(dd, fmt, ...)
#endif
#ifdef __cplusplus
}
#endif
#endif /* _SYS_DSL_DATASET_H */
diff --git a/include/sys/dsl_pool.h b/include/sys/dsl_pool.h
index 44900f8ceb2f..0283a8c589c1 100644
--- a/include/sys/dsl_pool.h
+++ b/include/sys/dsl_pool.h
@@ -1,206 +1,205 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013, 2018 by Delphix. All rights reserved.
* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
*/
#ifndef _SYS_DSL_POOL_H
#define _SYS_DSL_POOL_H
#include <sys/spa.h>
#include <sys/txg.h>
#include <sys/txg_impl.h>
#include <sys/zfs_context.h>
#include <sys/zio.h>
#include <sys/dnode.h>
#include <sys/ddt.h>
#include <sys/arc.h>
#include <sys/bpobj.h>
#include <sys/bptree.h>
#include <sys/rrwlock.h>
#include <sys/dsl_synctask.h>
#include <sys/mmp.h>
#include <sys/aggsum.h>
#ifdef __cplusplus
extern "C" {
#endif
extern int zfs_txg_synctime_ms;
struct objset;
struct dsl_dir;
struct dsl_dataset;
struct dsl_pool;
struct dmu_tx;
struct dsl_scan;
struct dsl_crypto_params;
struct dsl_deadlist;
extern unsigned long zfs_dirty_data_max;
extern unsigned long zfs_dirty_data_max_max;
extern unsigned long zfs_wrlog_data_max;
-extern int zfs_dirty_data_sync_percent;
extern int zfs_dirty_data_max_percent;
extern int zfs_dirty_data_max_max_percent;
extern int zfs_delay_min_dirty_percent;
extern unsigned long zfs_delay_scale;
/* These macros are for indexing into the zfs_all_blkstats_t. */
#define DMU_OT_DEFERRED DMU_OT_NONE
#define DMU_OT_OTHER DMU_OT_NUMTYPES /* place holder for DMU_OT() types */
#define DMU_OT_TOTAL (DMU_OT_NUMTYPES + 1)
typedef struct zfs_blkstat {
uint64_t zb_count;
uint64_t zb_asize;
uint64_t zb_lsize;
uint64_t zb_psize;
uint64_t zb_gangs;
uint64_t zb_ditto_2_of_2_samevdev;
uint64_t zb_ditto_2_of_3_samevdev;
uint64_t zb_ditto_3_of_3_samevdev;
} zfs_blkstat_t;
typedef struct zfs_all_blkstats {
zfs_blkstat_t zab_type[DN_MAX_LEVELS + 1][DMU_OT_TOTAL + 1];
kmutex_t zab_lock;
} zfs_all_blkstats_t;
typedef struct dsl_pool {
/* Immutable */
spa_t *dp_spa;
struct objset *dp_meta_objset;
struct dsl_dir *dp_root_dir;
struct dsl_dir *dp_mos_dir;
struct dsl_dir *dp_free_dir;
struct dsl_dir *dp_leak_dir;
struct dsl_dataset *dp_origin_snap;
uint64_t dp_root_dir_obj;
struct taskq *dp_zrele_taskq;
struct taskq *dp_unlinked_drain_taskq;
/* No lock needed - sync context only */
blkptr_t dp_meta_rootbp;
uint64_t dp_tmp_userrefs_obj;
bpobj_t dp_free_bpobj;
uint64_t dp_bptree_obj;
uint64_t dp_empty_bpobj;
bpobj_t dp_obsolete_bpobj;
struct dsl_scan *dp_scan;
/* Uses dp_lock */
kmutex_t dp_lock;
kcondvar_t dp_spaceavail_cv;
uint64_t dp_dirty_pertxg[TXG_SIZE];
uint64_t dp_dirty_total;
uint64_t dp_long_free_dirty_pertxg[TXG_SIZE];
uint64_t dp_mos_used_delta;
uint64_t dp_mos_compressed_delta;
uint64_t dp_mos_uncompressed_delta;
aggsum_t dp_wrlog_pertxg[TXG_SIZE];
aggsum_t dp_wrlog_total;
/*
* Time of most recently scheduled (furthest in the future)
* wakeup for delayed transactions.
*/
hrtime_t dp_last_wakeup;
/* Has its own locking */
tx_state_t dp_tx;
txg_list_t dp_dirty_datasets;
txg_list_t dp_dirty_zilogs;
txg_list_t dp_dirty_dirs;
txg_list_t dp_sync_tasks;
txg_list_t dp_early_sync_tasks;
taskq_t *dp_sync_taskq;
taskq_t *dp_zil_clean_taskq;
/*
* Protects administrative changes (properties, namespace)
*
* It is only held for write in syncing context. Therefore
* syncing context does not need to ever have it for read, since
* nobody else could possibly have it for write.
*/
rrwlock_t dp_config_rwlock;
zfs_all_blkstats_t *dp_blkstats;
} dsl_pool_t;
int dsl_pool_init(spa_t *spa, uint64_t txg, dsl_pool_t **dpp);
int dsl_pool_open(dsl_pool_t *dp);
void dsl_pool_close(dsl_pool_t *dp);
dsl_pool_t *dsl_pool_create(spa_t *spa, nvlist_t *zplprops,
struct dsl_crypto_params *dcp, uint64_t txg);
void dsl_pool_sync(dsl_pool_t *dp, uint64_t txg);
void dsl_pool_sync_done(dsl_pool_t *dp, uint64_t txg);
int dsl_pool_sync_context(dsl_pool_t *dp);
uint64_t dsl_pool_adjustedsize(dsl_pool_t *dp, zfs_space_check_t slop_policy);
uint64_t dsl_pool_unreserved_space(dsl_pool_t *dp,
zfs_space_check_t slop_policy);
void dsl_pool_wrlog_count(dsl_pool_t *dp, int64_t size, uint64_t txg);
boolean_t dsl_pool_wrlog_over_max(dsl_pool_t *dp);
void dsl_pool_dirty_space(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx);
void dsl_pool_undirty_space(dsl_pool_t *dp, int64_t space, uint64_t txg);
void dsl_free(dsl_pool_t *dp, uint64_t txg, const blkptr_t *bpp);
void dsl_free_sync(zio_t *pio, dsl_pool_t *dp, uint64_t txg,
const blkptr_t *bpp);
void dsl_pool_create_origin(dsl_pool_t *dp, dmu_tx_t *tx);
void dsl_pool_upgrade_clones(dsl_pool_t *dp, dmu_tx_t *tx);
void dsl_pool_upgrade_dir_clones(dsl_pool_t *dp, dmu_tx_t *tx);
void dsl_pool_mos_diduse_space(dsl_pool_t *dp,
int64_t used, int64_t comp, int64_t uncomp);
void dsl_pool_ckpoint_diduse_space(dsl_pool_t *dp,
int64_t used, int64_t comp, int64_t uncomp);
boolean_t dsl_pool_need_dirty_delay(dsl_pool_t *dp);
void dsl_pool_config_enter(dsl_pool_t *dp, void *tag);
void dsl_pool_config_enter_prio(dsl_pool_t *dp, void *tag);
void dsl_pool_config_exit(dsl_pool_t *dp, void *tag);
boolean_t dsl_pool_config_held(dsl_pool_t *dp);
boolean_t dsl_pool_config_held_writer(dsl_pool_t *dp);
taskq_t *dsl_pool_zrele_taskq(dsl_pool_t *dp);
taskq_t *dsl_pool_unlinked_drain_taskq(dsl_pool_t *dp);
int dsl_pool_user_hold(dsl_pool_t *dp, uint64_t dsobj,
const char *tag, uint64_t now, dmu_tx_t *tx);
int dsl_pool_user_release(dsl_pool_t *dp, uint64_t dsobj,
const char *tag, dmu_tx_t *tx);
void dsl_pool_clean_tmp_userrefs(dsl_pool_t *dp);
int dsl_pool_open_special_dir(dsl_pool_t *dp, const char *name, dsl_dir_t **);
int dsl_pool_hold(const char *name, void *tag, dsl_pool_t **dp);
void dsl_pool_rele(dsl_pool_t *dp, void *tag);
void dsl_pool_create_obsolete_bpobj(dsl_pool_t *dp, dmu_tx_t *tx);
void dsl_pool_destroy_obsolete_bpobj(dsl_pool_t *dp, dmu_tx_t *tx);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_DSL_POOL_H */
diff --git a/include/sys/fs/zfs.h b/include/sys/fs/zfs.h
index 287b3beae91e..6bc82198df23 100644
--- a/include/sys/fs/zfs.h
+++ b/include/sys/fs/zfs.h
@@ -1,1757 +1,1758 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2013, 2017 Joyent, Inc. All rights reserved.
* Copyright (c) 2014 Integros [integros.com]
* Copyright (c) 2017, Intel Corporation.
* Copyright (c) 2019 Datto Inc.
* Portions Copyright 2010 Robert Milkowski
* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
*/
#ifndef _SYS_FS_ZFS_H
#define _SYS_FS_ZFS_H extern __attribute__((visibility("default")))
#include <sys/time.h>
#include <sys/zio_priority.h>
#ifdef __cplusplus
extern "C" {
#endif
/*
* Types and constants shared between userland and the kernel.
*/
/*
* Each dataset can be one of the following types. These constants can be
* combined into masks that can be passed to various functions.
*/
typedef enum {
ZFS_TYPE_FILESYSTEM = (1 << 0),
ZFS_TYPE_SNAPSHOT = (1 << 1),
ZFS_TYPE_VOLUME = (1 << 2),
ZFS_TYPE_POOL = (1 << 3),
ZFS_TYPE_BOOKMARK = (1 << 4),
ZFS_TYPE_VDEV = (1 << 5),
} zfs_type_t;
/*
* NB: lzc_dataset_type should be updated whenever a new objset type is added,
* if it represents a real type of a dataset that can be created from userland.
*/
typedef enum dmu_objset_type {
DMU_OST_NONE,
DMU_OST_META,
DMU_OST_ZFS,
DMU_OST_ZVOL,
DMU_OST_OTHER, /* For testing only! */
DMU_OST_ANY, /* Be careful! */
DMU_OST_NUMTYPES
} dmu_objset_type_t;
#define ZFS_TYPE_DATASET \
(ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME | ZFS_TYPE_SNAPSHOT)
/*
* All of these include the terminating NUL byte.
*/
#define ZAP_MAXNAMELEN 256
#define ZAP_MAXVALUELEN (1024 * 8)
#define ZAP_OLDMAXVALUELEN 1024
#define ZFS_MAX_DATASET_NAME_LEN 256
/*
* Dataset properties are identified by these constants and must be added to
* the end of this list to ensure that external consumers are not affected
* by the change. If you make any changes to this list, be sure to update
* the property table in module/zcommon/zfs_prop.c.
*/
typedef enum {
ZPROP_CONT = -2,
ZPROP_INVAL = -1,
ZFS_PROP_TYPE = 0,
ZFS_PROP_CREATION,
ZFS_PROP_USED,
ZFS_PROP_AVAILABLE,
ZFS_PROP_REFERENCED,
ZFS_PROP_COMPRESSRATIO,
ZFS_PROP_MOUNTED,
ZFS_PROP_ORIGIN,
ZFS_PROP_QUOTA,
ZFS_PROP_RESERVATION,
ZFS_PROP_VOLSIZE,
ZFS_PROP_VOLBLOCKSIZE,
ZFS_PROP_RECORDSIZE,
ZFS_PROP_MOUNTPOINT,
ZFS_PROP_SHARENFS,
ZFS_PROP_CHECKSUM,
ZFS_PROP_COMPRESSION,
ZFS_PROP_ATIME,
ZFS_PROP_DEVICES,
ZFS_PROP_EXEC,
ZFS_PROP_SETUID,
ZFS_PROP_READONLY,
ZFS_PROP_ZONED,
ZFS_PROP_SNAPDIR,
ZFS_PROP_ACLMODE,
ZFS_PROP_ACLINHERIT,
ZFS_PROP_CREATETXG,
ZFS_PROP_NAME, /* not exposed to the user */
ZFS_PROP_CANMOUNT,
ZFS_PROP_ISCSIOPTIONS, /* not exposed to the user */
ZFS_PROP_XATTR,
ZFS_PROP_NUMCLONES, /* not exposed to the user */
ZFS_PROP_COPIES,
ZFS_PROP_VERSION,
ZFS_PROP_UTF8ONLY,
ZFS_PROP_NORMALIZE,
ZFS_PROP_CASE,
ZFS_PROP_VSCAN,
ZFS_PROP_NBMAND,
ZFS_PROP_SHARESMB,
ZFS_PROP_REFQUOTA,
ZFS_PROP_REFRESERVATION,
ZFS_PROP_GUID,
ZFS_PROP_PRIMARYCACHE,
ZFS_PROP_SECONDARYCACHE,
ZFS_PROP_USEDSNAP,
ZFS_PROP_USEDDS,
ZFS_PROP_USEDCHILD,
ZFS_PROP_USEDREFRESERV,
ZFS_PROP_USERACCOUNTING, /* not exposed to the user */
ZFS_PROP_STMF_SHAREINFO, /* not exposed to the user */
ZFS_PROP_DEFER_DESTROY,
ZFS_PROP_USERREFS,
ZFS_PROP_LOGBIAS,
ZFS_PROP_UNIQUE, /* not exposed to the user */
ZFS_PROP_OBJSETID,
ZFS_PROP_DEDUP,
ZFS_PROP_MLSLABEL,
ZFS_PROP_SYNC,
ZFS_PROP_DNODESIZE,
ZFS_PROP_REFRATIO,
ZFS_PROP_WRITTEN,
ZFS_PROP_CLONES,
ZFS_PROP_LOGICALUSED,
ZFS_PROP_LOGICALREFERENCED,
ZFS_PROP_INCONSISTENT, /* not exposed to the user */
ZFS_PROP_VOLMODE,
ZFS_PROP_FILESYSTEM_LIMIT,
ZFS_PROP_SNAPSHOT_LIMIT,
ZFS_PROP_FILESYSTEM_COUNT,
ZFS_PROP_SNAPSHOT_COUNT,
ZFS_PROP_SNAPDEV,
ZFS_PROP_ACLTYPE,
ZFS_PROP_SELINUX_CONTEXT,
ZFS_PROP_SELINUX_FSCONTEXT,
ZFS_PROP_SELINUX_DEFCONTEXT,
ZFS_PROP_SELINUX_ROOTCONTEXT,
ZFS_PROP_RELATIME,
ZFS_PROP_REDUNDANT_METADATA,
ZFS_PROP_OVERLAY,
ZFS_PROP_PREV_SNAP,
ZFS_PROP_RECEIVE_RESUME_TOKEN,
ZFS_PROP_ENCRYPTION,
ZFS_PROP_KEYLOCATION,
ZFS_PROP_KEYFORMAT,
ZFS_PROP_PBKDF2_SALT,
ZFS_PROP_PBKDF2_ITERS,
ZFS_PROP_ENCRYPTION_ROOT,
ZFS_PROP_KEY_GUID,
ZFS_PROP_KEYSTATUS,
ZFS_PROP_REMAPTXG, /* obsolete - no longer used */
ZFS_PROP_SPECIAL_SMALL_BLOCKS,
ZFS_PROP_IVSET_GUID, /* not exposed to the user */
ZFS_PROP_REDACTED,
ZFS_PROP_REDACT_SNAPS,
ZFS_NUM_PROPS
} zfs_prop_t;
typedef enum {
ZFS_PROP_USERUSED,
ZFS_PROP_USERQUOTA,
ZFS_PROP_GROUPUSED,
ZFS_PROP_GROUPQUOTA,
ZFS_PROP_USEROBJUSED,
ZFS_PROP_USEROBJQUOTA,
ZFS_PROP_GROUPOBJUSED,
ZFS_PROP_GROUPOBJQUOTA,
ZFS_PROP_PROJECTUSED,
ZFS_PROP_PROJECTQUOTA,
ZFS_PROP_PROJECTOBJUSED,
ZFS_PROP_PROJECTOBJQUOTA,
ZFS_NUM_USERQUOTA_PROPS
} zfs_userquota_prop_t;
-_SYS_FS_ZFS_H const char *zfs_userquota_prop_prefixes[ZFS_NUM_USERQUOTA_PROPS];
+_SYS_FS_ZFS_H const char *const zfs_userquota_prop_prefixes[
+ ZFS_NUM_USERQUOTA_PROPS];
/*
* Pool properties are identified by these constants and must be added to the
* end of this list to ensure that external consumers are not affected
* by the change. Properties must be registered in zfs_prop_init().
*/
typedef enum {
ZPOOL_PROP_INVAL = -1,
ZPOOL_PROP_NAME,
ZPOOL_PROP_SIZE,
ZPOOL_PROP_CAPACITY,
ZPOOL_PROP_ALTROOT,
ZPOOL_PROP_HEALTH,
ZPOOL_PROP_GUID,
ZPOOL_PROP_VERSION,
ZPOOL_PROP_BOOTFS,
ZPOOL_PROP_DELEGATION,
ZPOOL_PROP_AUTOREPLACE,
ZPOOL_PROP_CACHEFILE,
ZPOOL_PROP_FAILUREMODE,
ZPOOL_PROP_LISTSNAPS,
ZPOOL_PROP_AUTOEXPAND,
ZPOOL_PROP_DEDUPDITTO,
ZPOOL_PROP_DEDUPRATIO,
ZPOOL_PROP_FREE,
ZPOOL_PROP_ALLOCATED,
ZPOOL_PROP_READONLY,
ZPOOL_PROP_ASHIFT,
ZPOOL_PROP_COMMENT,
ZPOOL_PROP_EXPANDSZ,
ZPOOL_PROP_FREEING,
ZPOOL_PROP_FRAGMENTATION,
ZPOOL_PROP_LEAKED,
ZPOOL_PROP_MAXBLOCKSIZE,
ZPOOL_PROP_TNAME,
ZPOOL_PROP_MAXDNODESIZE,
ZPOOL_PROP_MULTIHOST,
ZPOOL_PROP_CHECKPOINT,
ZPOOL_PROP_LOAD_GUID,
ZPOOL_PROP_AUTOTRIM,
ZPOOL_PROP_COMPATIBILITY,
ZPOOL_NUM_PROPS
} zpool_prop_t;
/* Small enough to not hog a whole line of printout in zpool(8). */
#define ZPROP_MAX_COMMENT 32
#define ZPROP_BOOLEAN_NA 2
#define ZPROP_VALUE "value"
#define ZPROP_SOURCE "source"
typedef enum {
ZPROP_SRC_NONE = 0x1,
ZPROP_SRC_DEFAULT = 0x2,
ZPROP_SRC_TEMPORARY = 0x4,
ZPROP_SRC_LOCAL = 0x8,
ZPROP_SRC_INHERITED = 0x10,
ZPROP_SRC_RECEIVED = 0x20
} zprop_source_t;
#define ZPROP_SRC_ALL 0x3f
#define ZPROP_SOURCE_VAL_RECVD "$recvd"
#define ZPROP_N_MORE_ERRORS "N_MORE_ERRORS"
/*
* Dataset flag implemented as a special entry in the props zap object
* indicating that the dataset has received properties on or after
* SPA_VERSION_RECVD_PROPS. The first such receive blows away local properties
* just as it did in earlier versions, and thereafter, local properties are
* preserved.
*/
#define ZPROP_HAS_RECVD "$hasrecvd"
typedef enum {
ZPROP_ERR_NOCLEAR = 0x1, /* failure to clear existing props */
ZPROP_ERR_NORESTORE = 0x2 /* failure to restore props on error */
} zprop_errflags_t;
typedef int (*zprop_func)(int, void *);
/*
* Properties to be set on the root file system of a new pool
* are stuffed into their own nvlist, which is then included in
* the properties nvlist with the pool properties.
*/
#define ZPOOL_ROOTFS_PROPS "root-props-nvl"
/*
* Length of 'written@' and 'written#'
*/
#define ZFS_WRITTEN_PROP_PREFIX_LEN 8
/*
* VDEV properties are identified by these constants and must be added to the
* end of this list to ensure that external consumers are not affected
* by the change. If you make any changes to this list, be sure to update
* the property table in usr/src/common/zfs/zpool_prop.c.
*/
typedef enum {
VDEV_PROP_INVAL = -1,
#define VDEV_PROP_USER VDEV_PROP_INVAL
VDEV_PROP_NAME,
VDEV_PROP_CAPACITY,
VDEV_PROP_STATE,
VDEV_PROP_GUID,
VDEV_PROP_ASIZE,
VDEV_PROP_PSIZE,
VDEV_PROP_ASHIFT,
VDEV_PROP_SIZE,
VDEV_PROP_FREE,
VDEV_PROP_ALLOCATED,
VDEV_PROP_COMMENT,
VDEV_PROP_EXPANDSZ,
VDEV_PROP_FRAGMENTATION,
VDEV_PROP_BOOTSIZE,
VDEV_PROP_PARITY,
VDEV_PROP_PATH,
VDEV_PROP_DEVID,
VDEV_PROP_PHYS_PATH,
VDEV_PROP_ENC_PATH,
VDEV_PROP_FRU,
VDEV_PROP_PARENT,
VDEV_PROP_CHILDREN,
VDEV_PROP_NUMCHILDREN,
VDEV_PROP_READ_ERRORS,
VDEV_PROP_WRITE_ERRORS,
VDEV_PROP_CHECKSUM_ERRORS,
VDEV_PROP_INITIALIZE_ERRORS,
VDEV_PROP_OPS_NULL,
VDEV_PROP_OPS_READ,
VDEV_PROP_OPS_WRITE,
VDEV_PROP_OPS_FREE,
VDEV_PROP_OPS_CLAIM,
VDEV_PROP_OPS_TRIM,
VDEV_PROP_BYTES_NULL,
VDEV_PROP_BYTES_READ,
VDEV_PROP_BYTES_WRITE,
VDEV_PROP_BYTES_FREE,
VDEV_PROP_BYTES_CLAIM,
VDEV_PROP_BYTES_TRIM,
VDEV_PROP_REMOVING,
VDEV_PROP_ALLOCATING,
VDEV_NUM_PROPS
} vdev_prop_t;
/*
* Dataset property functions shared between libzfs and kernel.
*/
_SYS_FS_ZFS_H const char *zfs_prop_default_string(zfs_prop_t);
_SYS_FS_ZFS_H uint64_t zfs_prop_default_numeric(zfs_prop_t);
_SYS_FS_ZFS_H boolean_t zfs_prop_readonly(zfs_prop_t);
_SYS_FS_ZFS_H boolean_t zfs_prop_visible(zfs_prop_t prop);
_SYS_FS_ZFS_H boolean_t zfs_prop_inheritable(zfs_prop_t);
_SYS_FS_ZFS_H boolean_t zfs_prop_setonce(zfs_prop_t);
_SYS_FS_ZFS_H boolean_t zfs_prop_encryption_key_param(zfs_prop_t);
_SYS_FS_ZFS_H boolean_t zfs_prop_valid_keylocation(const char *, boolean_t);
_SYS_FS_ZFS_H const char *zfs_prop_to_name(zfs_prop_t);
_SYS_FS_ZFS_H zfs_prop_t zfs_name_to_prop(const char *);
_SYS_FS_ZFS_H boolean_t zfs_prop_user(const char *);
_SYS_FS_ZFS_H boolean_t zfs_prop_userquota(const char *);
_SYS_FS_ZFS_H boolean_t zfs_prop_written(const char *);
_SYS_FS_ZFS_H int zfs_prop_index_to_string(zfs_prop_t, uint64_t, const char **);
_SYS_FS_ZFS_H int zfs_prop_string_to_index(zfs_prop_t, const char *,
uint64_t *);
_SYS_FS_ZFS_H uint64_t zfs_prop_random_value(zfs_prop_t, uint64_t seed);
_SYS_FS_ZFS_H boolean_t zfs_prop_valid_for_type(int, zfs_type_t, boolean_t);
/*
* Pool property functions shared between libzfs and kernel.
*/
_SYS_FS_ZFS_H zpool_prop_t zpool_name_to_prop(const char *);
_SYS_FS_ZFS_H const char *zpool_prop_to_name(zpool_prop_t);
_SYS_FS_ZFS_H const char *zpool_prop_default_string(zpool_prop_t);
_SYS_FS_ZFS_H uint64_t zpool_prop_default_numeric(zpool_prop_t);
_SYS_FS_ZFS_H boolean_t zpool_prop_readonly(zpool_prop_t);
_SYS_FS_ZFS_H boolean_t zpool_prop_setonce(zpool_prop_t);
_SYS_FS_ZFS_H boolean_t zpool_prop_feature(const char *);
_SYS_FS_ZFS_H boolean_t zpool_prop_unsupported(const char *);
_SYS_FS_ZFS_H int zpool_prop_index_to_string(zpool_prop_t, uint64_t,
const char **);
_SYS_FS_ZFS_H int zpool_prop_string_to_index(zpool_prop_t, const char *,
uint64_t *);
_SYS_FS_ZFS_H uint64_t zpool_prop_random_value(zpool_prop_t, uint64_t seed);
/*
* VDEV property functions shared between libzfs and kernel.
*/
_SYS_FS_ZFS_H vdev_prop_t vdev_name_to_prop(const char *);
_SYS_FS_ZFS_H boolean_t vdev_prop_user(const char *name);
_SYS_FS_ZFS_H const char *vdev_prop_to_name(vdev_prop_t);
_SYS_FS_ZFS_H const char *vdev_prop_default_string(vdev_prop_t);
_SYS_FS_ZFS_H uint64_t vdev_prop_default_numeric(vdev_prop_t);
_SYS_FS_ZFS_H boolean_t vdev_prop_readonly(vdev_prop_t prop);
_SYS_FS_ZFS_H int vdev_prop_index_to_string(vdev_prop_t, uint64_t,
const char **);
_SYS_FS_ZFS_H int vdev_prop_string_to_index(vdev_prop_t, const char *,
uint64_t *);
_SYS_FS_ZFS_H boolean_t zpool_prop_vdev(const char *name);
_SYS_FS_ZFS_H uint64_t vdev_prop_random_value(vdev_prop_t prop, uint64_t seed);
/*
* Definitions for the Delegation.
*/
typedef enum {
ZFS_DELEG_WHO_UNKNOWN = 0,
ZFS_DELEG_USER = 'u',
ZFS_DELEG_USER_SETS = 'U',
ZFS_DELEG_GROUP = 'g',
ZFS_DELEG_GROUP_SETS = 'G',
ZFS_DELEG_EVERYONE = 'e',
ZFS_DELEG_EVERYONE_SETS = 'E',
ZFS_DELEG_CREATE = 'c',
ZFS_DELEG_CREATE_SETS = 'C',
ZFS_DELEG_NAMED_SET = 's',
ZFS_DELEG_NAMED_SET_SETS = 'S'
} zfs_deleg_who_type_t;
typedef enum {
ZFS_DELEG_NONE = 0,
ZFS_DELEG_PERM_LOCAL = 1,
ZFS_DELEG_PERM_DESCENDENT = 2,
ZFS_DELEG_PERM_LOCALDESCENDENT = 3,
ZFS_DELEG_PERM_CREATE = 4
} zfs_deleg_inherit_t;
#define ZFS_DELEG_PERM_UID "uid"
#define ZFS_DELEG_PERM_GID "gid"
#define ZFS_DELEG_PERM_GROUPS "groups"
#define ZFS_MLSLABEL_DEFAULT "none"
#define ZFS_SMB_ACL_SRC "src"
#define ZFS_SMB_ACL_TARGET "target"
typedef enum {
ZFS_CANMOUNT_OFF = 0,
ZFS_CANMOUNT_ON = 1,
ZFS_CANMOUNT_NOAUTO = 2
} zfs_canmount_type_t;
typedef enum {
ZFS_LOGBIAS_LATENCY = 0,
ZFS_LOGBIAS_THROUGHPUT = 1
} zfs_logbias_op_t;
typedef enum zfs_share_op {
ZFS_SHARE_NFS = 0,
ZFS_UNSHARE_NFS = 1,
ZFS_SHARE_SMB = 2,
ZFS_UNSHARE_SMB = 3
} zfs_share_op_t;
typedef enum zfs_smb_acl_op {
ZFS_SMB_ACL_ADD,
ZFS_SMB_ACL_REMOVE,
ZFS_SMB_ACL_RENAME,
ZFS_SMB_ACL_PURGE
} zfs_smb_acl_op_t;
typedef enum zfs_cache_type {
ZFS_CACHE_NONE = 0,
ZFS_CACHE_METADATA = 1,
ZFS_CACHE_ALL = 2
} zfs_cache_type_t;
typedef enum {
ZFS_SYNC_STANDARD = 0,
ZFS_SYNC_ALWAYS = 1,
ZFS_SYNC_DISABLED = 2
} zfs_sync_type_t;
typedef enum {
ZFS_XATTR_OFF = 0,
ZFS_XATTR_DIR = 1,
ZFS_XATTR_SA = 2
} zfs_xattr_type_t;
typedef enum {
ZFS_DNSIZE_LEGACY = 0,
ZFS_DNSIZE_AUTO = 1,
ZFS_DNSIZE_1K = 1024,
ZFS_DNSIZE_2K = 2048,
ZFS_DNSIZE_4K = 4096,
ZFS_DNSIZE_8K = 8192,
ZFS_DNSIZE_16K = 16384
} zfs_dnsize_type_t;
typedef enum {
ZFS_REDUNDANT_METADATA_ALL,
ZFS_REDUNDANT_METADATA_MOST
} zfs_redundant_metadata_type_t;
typedef enum {
ZFS_VOLMODE_DEFAULT = 0,
ZFS_VOLMODE_GEOM = 1,
ZFS_VOLMODE_DEV = 2,
ZFS_VOLMODE_NONE = 3
} zfs_volmode_t;
typedef enum zfs_keystatus {
ZFS_KEYSTATUS_NONE = 0,
ZFS_KEYSTATUS_UNAVAILABLE,
ZFS_KEYSTATUS_AVAILABLE,
} zfs_keystatus_t;
typedef enum zfs_keyformat {
ZFS_KEYFORMAT_NONE = 0,
ZFS_KEYFORMAT_RAW,
ZFS_KEYFORMAT_HEX,
ZFS_KEYFORMAT_PASSPHRASE,
ZFS_KEYFORMAT_FORMATS
} zfs_keyformat_t;
typedef enum zfs_key_location {
ZFS_KEYLOCATION_NONE = 0,
ZFS_KEYLOCATION_PROMPT,
ZFS_KEYLOCATION_URI,
ZFS_KEYLOCATION_LOCATIONS
} zfs_keylocation_t;
#define DEFAULT_PBKDF2_ITERATIONS 350000
#define MIN_PBKDF2_ITERATIONS 100000
/*
* On-disk version number.
*/
#define SPA_VERSION_1 1ULL
#define SPA_VERSION_2 2ULL
#define SPA_VERSION_3 3ULL
#define SPA_VERSION_4 4ULL
#define SPA_VERSION_5 5ULL
#define SPA_VERSION_6 6ULL
#define SPA_VERSION_7 7ULL
#define SPA_VERSION_8 8ULL
#define SPA_VERSION_9 9ULL
#define SPA_VERSION_10 10ULL
#define SPA_VERSION_11 11ULL
#define SPA_VERSION_12 12ULL
#define SPA_VERSION_13 13ULL
#define SPA_VERSION_14 14ULL
#define SPA_VERSION_15 15ULL
#define SPA_VERSION_16 16ULL
#define SPA_VERSION_17 17ULL
#define SPA_VERSION_18 18ULL
#define SPA_VERSION_19 19ULL
#define SPA_VERSION_20 20ULL
#define SPA_VERSION_21 21ULL
#define SPA_VERSION_22 22ULL
#define SPA_VERSION_23 23ULL
#define SPA_VERSION_24 24ULL
#define SPA_VERSION_25 25ULL
#define SPA_VERSION_26 26ULL
#define SPA_VERSION_27 27ULL
#define SPA_VERSION_28 28ULL
#define SPA_VERSION_5000 5000ULL
/*
* The incrementing pool version number has been replaced by pool feature
* flags. For more details, see zfeature.c.
*/
#define SPA_VERSION SPA_VERSION_5000
#define SPA_VERSION_STRING "5000"
/*
* Symbolic names for the changes that caused a SPA_VERSION switch.
* Used in the code when checking for presence or absence of a feature.
* Feel free to define multiple symbolic names for each version if there
* were multiple changes to on-disk structures during that version.
*
* NOTE: When checking the current SPA_VERSION in your code, be sure
* to use spa_version() since it reports the version of the
* last synced uberblock. Checking the in-flight version can
* be dangerous in some cases.
*/
#define SPA_VERSION_INITIAL SPA_VERSION_1
#define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2
#define SPA_VERSION_SPARES SPA_VERSION_3
#define SPA_VERSION_RAIDZ2 SPA_VERSION_3
#define SPA_VERSION_BPOBJ_ACCOUNT SPA_VERSION_3
#define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3
#define SPA_VERSION_DNODE_BYTES SPA_VERSION_3
#define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4
#define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5
#define SPA_VERSION_BOOTFS SPA_VERSION_6
#define SPA_VERSION_SLOGS SPA_VERSION_7
#define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8
#define SPA_VERSION_FUID SPA_VERSION_9
#define SPA_VERSION_REFRESERVATION SPA_VERSION_9
#define SPA_VERSION_REFQUOTA SPA_VERSION_9
#define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9
#define SPA_VERSION_L2CACHE SPA_VERSION_10
#define SPA_VERSION_NEXT_CLONES SPA_VERSION_11
#define SPA_VERSION_ORIGIN SPA_VERSION_11
#define SPA_VERSION_DSL_SCRUB SPA_VERSION_11
#define SPA_VERSION_SNAP_PROPS SPA_VERSION_12
#define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13
#define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14
#define SPA_VERSION_USERSPACE SPA_VERSION_15
#define SPA_VERSION_STMF_PROP SPA_VERSION_16
#define SPA_VERSION_RAIDZ3 SPA_VERSION_17
#define SPA_VERSION_USERREFS SPA_VERSION_18
#define SPA_VERSION_HOLES SPA_VERSION_19
#define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20
#define SPA_VERSION_DEDUP SPA_VERSION_21
#define SPA_VERSION_RECVD_PROPS SPA_VERSION_22
#define SPA_VERSION_SLIM_ZIL SPA_VERSION_23
#define SPA_VERSION_SA SPA_VERSION_24
#define SPA_VERSION_SCAN SPA_VERSION_25
#define SPA_VERSION_DIR_CLONES SPA_VERSION_26
#define SPA_VERSION_DEADLISTS SPA_VERSION_26
#define SPA_VERSION_FAST_SNAP SPA_VERSION_27
#define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28
#define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28
#define SPA_VERSION_FEATURES SPA_VERSION_5000
#define SPA_VERSION_IS_SUPPORTED(v) \
(((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \
((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))
/*
* ZPL version - rev'd whenever an incompatible on-disk format change
* occurs. This is independent of SPA/DMU/ZAP versioning. You must
* also update the version_table[] and help message in zfs_prop.c.
*/
#define ZPL_VERSION_1 1ULL
#define ZPL_VERSION_2 2ULL
#define ZPL_VERSION_3 3ULL
#define ZPL_VERSION_4 4ULL
#define ZPL_VERSION_5 5ULL
#define ZPL_VERSION ZPL_VERSION_5
#define ZPL_VERSION_STRING "5"
#define ZPL_VERSION_INITIAL ZPL_VERSION_1
#define ZPL_VERSION_DIRENT_TYPE ZPL_VERSION_2
#define ZPL_VERSION_FUID ZPL_VERSION_3
#define ZPL_VERSION_NORMALIZATION ZPL_VERSION_3
#define ZPL_VERSION_SYSATTR ZPL_VERSION_3
#define ZPL_VERSION_USERSPACE ZPL_VERSION_4
#define ZPL_VERSION_SA ZPL_VERSION_5
/* Persistent L2ARC version */
#define L2ARC_PERSISTENT_VERSION_1 1ULL
#define L2ARC_PERSISTENT_VERSION L2ARC_PERSISTENT_VERSION_1
#define L2ARC_PERSISTENT_VERSION_STRING "1"
/* Rewind policy information */
#define ZPOOL_NO_REWIND 1 /* No policy - default behavior */
#define ZPOOL_NEVER_REWIND 2 /* Do not search for best txg or rewind */
#define ZPOOL_TRY_REWIND 4 /* Search for best txg, but do not rewind */
#define ZPOOL_DO_REWIND 8 /* Rewind to best txg w/in deferred frees */
#define ZPOOL_EXTREME_REWIND 16 /* Allow extreme measures to find best txg */
#define ZPOOL_REWIND_MASK 28 /* All the possible rewind bits */
#define ZPOOL_REWIND_POLICIES 31 /* All the possible policy bits */
typedef struct zpool_load_policy {
uint32_t zlp_rewind; /* rewind policy requested */
uint64_t zlp_maxmeta; /* max acceptable meta-data errors */
uint64_t zlp_maxdata; /* max acceptable data errors */
uint64_t zlp_txg; /* specific txg to load */
} zpool_load_policy_t;
/*
* The following are configuration names used in the nvlist describing a pool's
* configuration. New on-disk names should be prefixed with "<reversed-DNS>:"
* (e.g. "org.openzfs:") to avoid conflicting names being developed
* independently.
*/
#define ZPOOL_CONFIG_VERSION "version"
#define ZPOOL_CONFIG_POOL_NAME "name"
#define ZPOOL_CONFIG_POOL_STATE "state"
#define ZPOOL_CONFIG_POOL_TXG "txg"
#define ZPOOL_CONFIG_POOL_GUID "pool_guid"
#define ZPOOL_CONFIG_CREATE_TXG "create_txg"
#define ZPOOL_CONFIG_TOP_GUID "top_guid"
#define ZPOOL_CONFIG_VDEV_TREE "vdev_tree"
#define ZPOOL_CONFIG_TYPE "type"
#define ZPOOL_CONFIG_CHILDREN "children"
#define ZPOOL_CONFIG_ID "id"
#define ZPOOL_CONFIG_GUID "guid"
#define ZPOOL_CONFIG_INDIRECT_OBJECT "com.delphix:indirect_object"
#define ZPOOL_CONFIG_INDIRECT_BIRTHS "com.delphix:indirect_births"
#define ZPOOL_CONFIG_PREV_INDIRECT_VDEV "com.delphix:prev_indirect_vdev"
#define ZPOOL_CONFIG_PATH "path"
#define ZPOOL_CONFIG_DEVID "devid"
#define ZPOOL_CONFIG_SPARE_ID "spareid"
#define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array"
#define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift"
#define ZPOOL_CONFIG_ASHIFT "ashift"
#define ZPOOL_CONFIG_ASIZE "asize"
#define ZPOOL_CONFIG_DTL "DTL"
#define ZPOOL_CONFIG_SCAN_STATS "scan_stats" /* not stored on disk */
#define ZPOOL_CONFIG_REMOVAL_STATS "removal_stats" /* not stored on disk */
#define ZPOOL_CONFIG_CHECKPOINT_STATS "checkpoint_stats" /* not on disk */
#define ZPOOL_CONFIG_VDEV_STATS "vdev_stats" /* not stored on disk */
#define ZPOOL_CONFIG_INDIRECT_SIZE "indirect_size" /* not stored on disk */
/* container nvlist of extended stats */
#define ZPOOL_CONFIG_VDEV_STATS_EX "vdev_stats_ex"
/* Active queue read/write stats */
#define ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE "vdev_sync_r_active_queue"
#define ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE "vdev_sync_w_active_queue"
#define ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE "vdev_async_r_active_queue"
#define ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE "vdev_async_w_active_queue"
#define ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE "vdev_async_scrub_active_queue"
#define ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE "vdev_async_trim_active_queue"
#define ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE "vdev_rebuild_active_queue"
/* Queue sizes */
#define ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE "vdev_sync_r_pend_queue"
#define ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE "vdev_sync_w_pend_queue"
#define ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE "vdev_async_r_pend_queue"
#define ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE "vdev_async_w_pend_queue"
#define ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE "vdev_async_scrub_pend_queue"
#define ZPOOL_CONFIG_VDEV_TRIM_PEND_QUEUE "vdev_async_trim_pend_queue"
#define ZPOOL_CONFIG_VDEV_REBUILD_PEND_QUEUE "vdev_rebuild_pend_queue"
/* Latency read/write histogram stats */
#define ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO "vdev_tot_r_lat_histo"
#define ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO "vdev_tot_w_lat_histo"
#define ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO "vdev_disk_r_lat_histo"
#define ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO "vdev_disk_w_lat_histo"
#define ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO "vdev_sync_r_lat_histo"
#define ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO "vdev_sync_w_lat_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO "vdev_async_r_lat_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO "vdev_async_w_lat_histo"
#define ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO "vdev_scrub_histo"
#define ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO "vdev_trim_histo"
#define ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO "vdev_rebuild_histo"
/* Request size histograms */
#define ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO "vdev_sync_ind_r_histo"
#define ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO "vdev_sync_ind_w_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO "vdev_async_ind_r_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO "vdev_async_ind_w_histo"
#define ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO "vdev_ind_scrub_histo"
#define ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO "vdev_ind_trim_histo"
#define ZPOOL_CONFIG_VDEV_IND_REBUILD_HISTO "vdev_ind_rebuild_histo"
#define ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO "vdev_sync_agg_r_histo"
#define ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO "vdev_sync_agg_w_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO "vdev_async_agg_r_histo"
#define ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO "vdev_async_agg_w_histo"
#define ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO "vdev_agg_scrub_histo"
#define ZPOOL_CONFIG_VDEV_AGG_TRIM_HISTO "vdev_agg_trim_histo"
#define ZPOOL_CONFIG_VDEV_AGG_REBUILD_HISTO "vdev_agg_rebuild_histo"
/* Number of slow IOs */
#define ZPOOL_CONFIG_VDEV_SLOW_IOS "vdev_slow_ios"
/* vdev enclosure sysfs path */
#define ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH "vdev_enc_sysfs_path"
#define ZPOOL_CONFIG_WHOLE_DISK "whole_disk"
#define ZPOOL_CONFIG_ERRCOUNT "error_count"
#define ZPOOL_CONFIG_NOT_PRESENT "not_present"
#define ZPOOL_CONFIG_SPARES "spares"
#define ZPOOL_CONFIG_IS_SPARE "is_spare"
#define ZPOOL_CONFIG_NPARITY "nparity"
#define ZPOOL_CONFIG_HOSTID "hostid"
#define ZPOOL_CONFIG_HOSTNAME "hostname"
#define ZPOOL_CONFIG_LOADED_TIME "initial_load_time"
#define ZPOOL_CONFIG_UNSPARE "unspare"
#define ZPOOL_CONFIG_PHYS_PATH "phys_path"
#define ZPOOL_CONFIG_IS_LOG "is_log"
#define ZPOOL_CONFIG_L2CACHE "l2cache"
#define ZPOOL_CONFIG_HOLE_ARRAY "hole_array"
#define ZPOOL_CONFIG_VDEV_CHILDREN "vdev_children"
#define ZPOOL_CONFIG_IS_HOLE "is_hole"
#define ZPOOL_CONFIG_DDT_HISTOGRAM "ddt_histogram"
#define ZPOOL_CONFIG_DDT_OBJ_STATS "ddt_object_stats"
#define ZPOOL_CONFIG_DDT_STATS "ddt_stats"
#define ZPOOL_CONFIG_SPLIT "splitcfg"
#define ZPOOL_CONFIG_ORIG_GUID "orig_guid"
#define ZPOOL_CONFIG_SPLIT_GUID "split_guid"
#define ZPOOL_CONFIG_SPLIT_LIST "guid_list"
#define ZPOOL_CONFIG_NONALLOCATING "non_allocating"
#define ZPOOL_CONFIG_REMOVING "removing"
#define ZPOOL_CONFIG_RESILVER_TXG "resilver_txg"
#define ZPOOL_CONFIG_REBUILD_TXG "rebuild_txg"
#define ZPOOL_CONFIG_COMMENT "comment"
#define ZPOOL_CONFIG_SUSPENDED "suspended" /* not stored on disk */
#define ZPOOL_CONFIG_SUSPENDED_REASON "suspended_reason" /* not stored */
#define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */
#define ZPOOL_CONFIG_BOOTFS "bootfs" /* not stored on disk */
#define ZPOOL_CONFIG_MISSING_DEVICES "missing_vdevs" /* not stored on disk */
#define ZPOOL_CONFIG_LOAD_INFO "load_info" /* not stored on disk */
#define ZPOOL_CONFIG_REWIND_INFO "rewind_info" /* not stored on disk */
#define ZPOOL_CONFIG_UNSUP_FEAT "unsup_feat" /* not stored on disk */
#define ZPOOL_CONFIG_ENABLED_FEAT "enabled_feat" /* not stored on disk */
#define ZPOOL_CONFIG_CAN_RDONLY "can_rdonly" /* not stored on disk */
#define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read"
#define ZPOOL_CONFIG_FEATURE_STATS "feature_stats" /* not stored on disk */
#define ZPOOL_CONFIG_ERRATA "errata" /* not stored on disk */
#define ZPOOL_CONFIG_VDEV_TOP_ZAP "com.delphix:vdev_zap_top"
#define ZPOOL_CONFIG_VDEV_LEAF_ZAP "com.delphix:vdev_zap_leaf"
#define ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS "com.delphix:has_per_vdev_zaps"
#define ZPOOL_CONFIG_RESILVER_DEFER "com.datto:resilver_defer"
#define ZPOOL_CONFIG_CACHEFILE "cachefile" /* not stored on disk */
#define ZPOOL_CONFIG_MMP_STATE "mmp_state" /* not stored on disk */
#define ZPOOL_CONFIG_MMP_TXG "mmp_txg" /* not stored on disk */
#define ZPOOL_CONFIG_MMP_SEQ "mmp_seq" /* not stored on disk */
#define ZPOOL_CONFIG_MMP_HOSTNAME "mmp_hostname" /* not stored on disk */
#define ZPOOL_CONFIG_MMP_HOSTID "mmp_hostid" /* not stored on disk */
#define ZPOOL_CONFIG_ALLOCATION_BIAS "alloc_bias" /* not stored on disk */
#define ZPOOL_CONFIG_EXPANSION_TIME "expansion_time" /* not stored */
#define ZPOOL_CONFIG_REBUILD_STATS "org.openzfs:rebuild_stats"
#define ZPOOL_CONFIG_COMPATIBILITY "compatibility"
/*
* The persistent vdev state is stored as separate values rather than a single
* 'vdev_state' entry. This is because a device can be in multiple states, such
* as offline and degraded.
*/
#define ZPOOL_CONFIG_OFFLINE "offline"
#define ZPOOL_CONFIG_FAULTED "faulted"
#define ZPOOL_CONFIG_DEGRADED "degraded"
#define ZPOOL_CONFIG_REMOVED "removed"
#define ZPOOL_CONFIG_FRU "fru"
#define ZPOOL_CONFIG_AUX_STATE "aux_state"
/* Pool load policy parameters */
#define ZPOOL_LOAD_POLICY "load-policy"
#define ZPOOL_LOAD_REWIND_POLICY "load-rewind-policy"
#define ZPOOL_LOAD_REQUEST_TXG "load-request-txg"
#define ZPOOL_LOAD_META_THRESH "load-meta-thresh"
#define ZPOOL_LOAD_DATA_THRESH "load-data-thresh"
/* Rewind data discovered */
#define ZPOOL_CONFIG_LOAD_TIME "rewind_txg_ts"
#define ZPOOL_CONFIG_LOAD_DATA_ERRORS "verify_data_errors"
#define ZPOOL_CONFIG_REWIND_TIME "seconds_of_rewind"
/* dRAID configuration */
#define ZPOOL_CONFIG_DRAID_NDATA "draid_ndata"
#define ZPOOL_CONFIG_DRAID_NSPARES "draid_nspares"
#define ZPOOL_CONFIG_DRAID_NGROUPS "draid_ngroups"
#define VDEV_TYPE_ROOT "root"
#define VDEV_TYPE_MIRROR "mirror"
#define VDEV_TYPE_REPLACING "replacing"
#define VDEV_TYPE_RAIDZ "raidz"
#define VDEV_TYPE_DRAID "draid"
#define VDEV_TYPE_DRAID_SPARE "dspare"
#define VDEV_TYPE_DISK "disk"
#define VDEV_TYPE_FILE "file"
#define VDEV_TYPE_MISSING "missing"
#define VDEV_TYPE_HOLE "hole"
#define VDEV_TYPE_SPARE "spare"
#define VDEV_TYPE_LOG "log"
#define VDEV_TYPE_L2CACHE "l2cache"
#define VDEV_TYPE_INDIRECT "indirect"
#define VDEV_RAIDZ_MAXPARITY 3
#define VDEV_DRAID_MAXPARITY 3
#define VDEV_DRAID_MIN_CHILDREN 2
#define VDEV_DRAID_MAX_CHILDREN UINT8_MAX
/* VDEV_TOP_ZAP_* are used in top-level vdev ZAP objects. */
#define VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM \
"com.delphix:indirect_obsolete_sm"
#define VDEV_TOP_ZAP_OBSOLETE_COUNTS_ARE_PRECISE \
"com.delphix:obsolete_counts_are_precise"
#define VDEV_TOP_ZAP_POOL_CHECKPOINT_SM \
"com.delphix:pool_checkpoint_sm"
#define VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS \
"com.delphix:ms_unflushed_phys_txgs"
#define VDEV_TOP_ZAP_VDEV_REBUILD_PHYS \
"org.openzfs:vdev_rebuild"
#define VDEV_TOP_ZAP_ALLOCATION_BIAS \
"org.zfsonlinux:allocation_bias"
/* vdev metaslab allocation bias */
#define VDEV_ALLOC_BIAS_LOG "log"
#define VDEV_ALLOC_BIAS_SPECIAL "special"
#define VDEV_ALLOC_BIAS_DEDUP "dedup"
/* vdev initialize state */
#define VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET \
"com.delphix:next_offset_to_initialize"
#define VDEV_LEAF_ZAP_INITIALIZE_STATE \
"com.delphix:vdev_initialize_state"
#define VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME \
"com.delphix:vdev_initialize_action_time"
/* vdev TRIM state */
#define VDEV_LEAF_ZAP_TRIM_LAST_OFFSET \
"org.zfsonlinux:next_offset_to_trim"
#define VDEV_LEAF_ZAP_TRIM_STATE \
"org.zfsonlinux:vdev_trim_state"
#define VDEV_LEAF_ZAP_TRIM_ACTION_TIME \
"org.zfsonlinux:vdev_trim_action_time"
#define VDEV_LEAF_ZAP_TRIM_RATE \
"org.zfsonlinux:vdev_trim_rate"
#define VDEV_LEAF_ZAP_TRIM_PARTIAL \
"org.zfsonlinux:vdev_trim_partial"
#define VDEV_LEAF_ZAP_TRIM_SECURE \
"org.zfsonlinux:vdev_trim_secure"
/*
* This is needed in userland to report the minimum necessary device size.
*/
#define SPA_MINDEVSIZE (64ULL << 20)
/*
* Set if the fragmentation has not yet been calculated. This can happen
* because the space maps have not been upgraded or the histogram feature
* is not enabled.
*/
#define ZFS_FRAG_INVALID UINT64_MAX
/*
* The location of the pool configuration repository, shared between kernel and
* userland.
*/
#define ZPOOL_CACHE_BOOT "/boot/zfs/zpool.cache"
#define ZPOOL_CACHE "/etc/zfs/zpool.cache"
/*
* Settings for zpool compatibility features files
*/
#define ZPOOL_SYSCONF_COMPAT_D SYSCONFDIR "/zfs/compatibility.d"
#define ZPOOL_DATA_COMPAT_D PKGDATADIR "/compatibility.d"
#define ZPOOL_COMPAT_MAXSIZE 16384
/*
* Hard-wired compatibility settings
*/
#define ZPOOL_COMPAT_LEGACY "legacy"
#define ZPOOL_COMPAT_OFF "off"
/*
* vdev states are ordered from least to most healthy.
* A vdev that's CANT_OPEN or below is considered unusable.
*/
typedef enum vdev_state {
VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */
VDEV_STATE_CLOSED, /* Not currently open */
VDEV_STATE_OFFLINE, /* Not allowed to open */
VDEV_STATE_REMOVED, /* Explicitly removed from system */
VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */
VDEV_STATE_FAULTED, /* External request to fault device */
VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */
VDEV_STATE_HEALTHY /* Presumed good */
} vdev_state_t;
#define VDEV_STATE_ONLINE VDEV_STATE_HEALTHY
/*
* vdev aux states. When a vdev is in the CANT_OPEN state, the aux field
* of the vdev stats structure uses these constants to distinguish why.
*/
typedef enum vdev_aux {
VDEV_AUX_NONE, /* no error */
VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */
VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */
VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */
VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */
VDEV_AUX_TOO_SMALL, /* vdev size is too small */
VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */
VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */
VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */
VDEV_AUX_UNSUP_FEAT, /* unsupported features */
VDEV_AUX_SPARED, /* hot spare used in another pool */
VDEV_AUX_ERR_EXCEEDED, /* too many errors */
VDEV_AUX_IO_FAILURE, /* experienced I/O failure */
VDEV_AUX_BAD_LOG, /* cannot read log chain(s) */
VDEV_AUX_EXTERNAL, /* external diagnosis or forced fault */
VDEV_AUX_SPLIT_POOL, /* vdev was split off into another pool */
VDEV_AUX_BAD_ASHIFT, /* vdev ashift is invalid */
VDEV_AUX_EXTERNAL_PERSIST, /* persistent forced fault */
VDEV_AUX_ACTIVE, /* vdev active on a different host */
VDEV_AUX_CHILDREN_OFFLINE, /* all children are offline */
VDEV_AUX_ASHIFT_TOO_BIG, /* vdev's min block size is too large */
} vdev_aux_t;
/*
* pool state. The following states are written to disk as part of the normal
* SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE, L2CACHE. The remaining
* states are software abstractions used at various levels to communicate
* pool state.
*/
typedef enum pool_state {
POOL_STATE_ACTIVE = 0, /* In active use */
POOL_STATE_EXPORTED, /* Explicitly exported */
POOL_STATE_DESTROYED, /* Explicitly destroyed */
POOL_STATE_SPARE, /* Reserved for hot spare use */
POOL_STATE_L2CACHE, /* Level 2 ARC device */
POOL_STATE_UNINITIALIZED, /* Internal spa_t state */
POOL_STATE_UNAVAIL, /* Internal libzfs state */
POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */
} pool_state_t;
/*
* mmp state. The following states provide additional detail describing
* why a pool couldn't be safely imported.
*/
typedef enum mmp_state {
MMP_STATE_ACTIVE = 0, /* In active use */
MMP_STATE_INACTIVE, /* Inactive and safe to import */
MMP_STATE_NO_HOSTID /* System hostid is not set */
} mmp_state_t;
/*
* Scan Functions.
*/
typedef enum pool_scan_func {
POOL_SCAN_NONE,
POOL_SCAN_SCRUB,
POOL_SCAN_RESILVER,
POOL_SCAN_FUNCS
} pool_scan_func_t;
/*
* Used to control scrub pause and resume.
*/
typedef enum pool_scrub_cmd {
POOL_SCRUB_NORMAL = 0,
POOL_SCRUB_PAUSE,
POOL_SCRUB_FLAGS_END
} pool_scrub_cmd_t;
typedef enum {
CS_NONE,
CS_CHECKPOINT_EXISTS,
CS_CHECKPOINT_DISCARDING,
CS_NUM_STATES
} checkpoint_state_t;
typedef struct pool_checkpoint_stat {
uint64_t pcs_state; /* checkpoint_state_t */
uint64_t pcs_start_time; /* time checkpoint/discard started */
uint64_t pcs_space; /* checkpointed space */
} pool_checkpoint_stat_t;
/*
* ZIO types. Needed to interpret vdev statistics below.
*/
typedef enum zio_type {
ZIO_TYPE_NULL = 0,
ZIO_TYPE_READ,
ZIO_TYPE_WRITE,
ZIO_TYPE_FREE,
ZIO_TYPE_CLAIM,
ZIO_TYPE_IOCTL,
ZIO_TYPE_TRIM,
ZIO_TYPES
} zio_type_t;
/*
* Pool statistics. Note: all fields should be 64-bit because this
* is passed between kernel and userland as an nvlist uint64 array.
*/
typedef struct pool_scan_stat {
/* values stored on disk */
uint64_t pss_func; /* pool_scan_func_t */
uint64_t pss_state; /* dsl_scan_state_t */
uint64_t pss_start_time; /* scan start time */
uint64_t pss_end_time; /* scan end time */
uint64_t pss_to_examine; /* total bytes to scan */
uint64_t pss_examined; /* total bytes located by scanner */
uint64_t pss_to_process; /* total bytes to process */
uint64_t pss_processed; /* total processed bytes */
uint64_t pss_errors; /* scan errors */
/* values not stored on disk */
uint64_t pss_pass_exam; /* examined bytes per scan pass */
uint64_t pss_pass_start; /* start time of a scan pass */
uint64_t pss_pass_scrub_pause; /* pause time of a scrub pass */
/* cumulative time scrub spent paused, needed for rate calculation */
uint64_t pss_pass_scrub_spent_paused;
uint64_t pss_pass_issued; /* issued bytes per scan pass */
uint64_t pss_issued; /* total bytes checked by scanner */
} pool_scan_stat_t;
typedef struct pool_removal_stat {
uint64_t prs_state; /* dsl_scan_state_t */
uint64_t prs_removing_vdev;
uint64_t prs_start_time;
uint64_t prs_end_time;
uint64_t prs_to_copy; /* bytes that need to be copied */
uint64_t prs_copied; /* bytes copied so far */
/*
* bytes of memory used for indirect mappings.
* This includes all removed vdevs.
*/
uint64_t prs_mapping_memory;
} pool_removal_stat_t;
typedef enum dsl_scan_state {
DSS_NONE,
DSS_SCANNING,
DSS_FINISHED,
DSS_CANCELED,
DSS_NUM_STATES
} dsl_scan_state_t;
typedef struct vdev_rebuild_stat {
uint64_t vrs_state; /* vdev_rebuild_state_t */
uint64_t vrs_start_time; /* time_t */
uint64_t vrs_end_time; /* time_t */
uint64_t vrs_scan_time_ms; /* total run time (millisecs) */
uint64_t vrs_bytes_scanned; /* allocated bytes scanned */
uint64_t vrs_bytes_issued; /* read bytes issued */
uint64_t vrs_bytes_rebuilt; /* rebuilt bytes */
uint64_t vrs_bytes_est; /* total bytes to scan */
uint64_t vrs_errors; /* scanning errors */
uint64_t vrs_pass_time_ms; /* pass run time (millisecs) */
uint64_t vrs_pass_bytes_scanned; /* bytes scanned since start/resume */
uint64_t vrs_pass_bytes_issued; /* bytes rebuilt since start/resume */
} vdev_rebuild_stat_t;
/*
* Errata described by https://openzfs.github.io/openzfs-docs/msg/ZFS-8000-ER.
* The ordering of this enum must be maintained to ensure the errata identifiers
* map to the correct documentation. New errata may only be appended to the
* list and must contain corresponding documentation at the above link.
*/
typedef enum zpool_errata {
ZPOOL_ERRATA_NONE,
ZPOOL_ERRATA_ZOL_2094_SCRUB,
ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY,
ZPOOL_ERRATA_ZOL_6845_ENCRYPTION,
ZPOOL_ERRATA_ZOL_8308_ENCRYPTION,
} zpool_errata_t;
/*
* Vdev statistics. Note: all fields should be 64-bit because this
* is passed between kernel and user land as an nvlist uint64 array.
*
* The vs_ops[] and vs_bytes[] arrays must always be an array size of 6 in
* order to keep subsequent members at their known fixed offsets. When
* adding a new field it must be added to the end the structure.
*/
#define VS_ZIO_TYPES 6
typedef struct vdev_stat {
hrtime_t vs_timestamp; /* time since vdev load */
uint64_t vs_state; /* vdev state */
uint64_t vs_aux; /* see vdev_aux_t */
uint64_t vs_alloc; /* space allocated */
uint64_t vs_space; /* total capacity */
uint64_t vs_dspace; /* deflated capacity */
uint64_t vs_rsize; /* replaceable dev size */
uint64_t vs_esize; /* expandable dev size */
uint64_t vs_ops[VS_ZIO_TYPES]; /* operation count */
uint64_t vs_bytes[VS_ZIO_TYPES]; /* bytes read/written */
uint64_t vs_read_errors; /* read errors */
uint64_t vs_write_errors; /* write errors */
uint64_t vs_checksum_errors; /* checksum errors */
uint64_t vs_initialize_errors; /* initializing errors */
uint64_t vs_self_healed; /* self-healed bytes */
uint64_t vs_scan_removing; /* removing? */
uint64_t vs_scan_processed; /* scan processed bytes */
uint64_t vs_fragmentation; /* device fragmentation */
uint64_t vs_initialize_bytes_done; /* bytes initialized */
uint64_t vs_initialize_bytes_est; /* total bytes to initialize */
uint64_t vs_initialize_state; /* vdev_initializing_state_t */
uint64_t vs_initialize_action_time; /* time_t */
uint64_t vs_checkpoint_space; /* checkpoint-consumed space */
uint64_t vs_resilver_deferred; /* resilver deferred */
uint64_t vs_slow_ios; /* slow IOs */
uint64_t vs_trim_errors; /* trimming errors */
uint64_t vs_trim_notsup; /* supported by device */
uint64_t vs_trim_bytes_done; /* bytes trimmed */
uint64_t vs_trim_bytes_est; /* total bytes to trim */
uint64_t vs_trim_state; /* vdev_trim_state_t */
uint64_t vs_trim_action_time; /* time_t */
uint64_t vs_rebuild_processed; /* bytes rebuilt */
uint64_t vs_configured_ashift; /* TLV vdev_ashift */
uint64_t vs_logical_ashift; /* vdev_logical_ashift */
uint64_t vs_physical_ashift; /* vdev_physical_ashift */
uint64_t vs_noalloc; /* allocations halted? */
} vdev_stat_t;
/* BEGIN CSTYLED */
#define VDEV_STAT_VALID(field, uint64_t_field_count) \
((uint64_t_field_count * sizeof (uint64_t)) >= \
(offsetof(vdev_stat_t, field) + sizeof (((vdev_stat_t *)NULL)->field)))
/* END CSTYLED */
/*
* Extended stats
*
* These are stats which aren't included in the original iostat output. For
* convenience, they are grouped together in vdev_stat_ex, although each stat
* is individually exported as an nvlist.
*/
typedef struct vdev_stat_ex {
/* Number of ZIOs issued to disk and waiting to finish */
uint64_t vsx_active_queue[ZIO_PRIORITY_NUM_QUEUEABLE];
/* Number of ZIOs pending to be issued to disk */
uint64_t vsx_pend_queue[ZIO_PRIORITY_NUM_QUEUEABLE];
/*
* Below are the histograms for various latencies. Buckets are in
* units of nanoseconds.
*/
/*
* 2^37 nanoseconds = 134s. Timeouts will probably start kicking in
* before this.
*/
#define VDEV_L_HISTO_BUCKETS 37 /* Latency histo buckets */
#define VDEV_RQ_HISTO_BUCKETS 25 /* Request size histo buckets */
/* Amount of time in ZIO queue (ns) */
uint64_t vsx_queue_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
[VDEV_L_HISTO_BUCKETS];
/* Total ZIO latency (ns). Includes queuing and disk access time */
uint64_t vsx_total_histo[ZIO_TYPES][VDEV_L_HISTO_BUCKETS];
/* Amount of time to read/write the disk (ns) */
uint64_t vsx_disk_histo[ZIO_TYPES][VDEV_L_HISTO_BUCKETS];
/* "lookup the bucket for a value" histogram macros */
#define HISTO(val, buckets) (val != 0 ? MIN(highbit64(val) - 1, \
buckets - 1) : 0)
#define L_HISTO(a) HISTO(a, VDEV_L_HISTO_BUCKETS)
#define RQ_HISTO(a) HISTO(a, VDEV_RQ_HISTO_BUCKETS)
/* Physical IO histogram */
uint64_t vsx_ind_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
[VDEV_RQ_HISTO_BUCKETS];
/* Delegated (aggregated) physical IO histogram */
uint64_t vsx_agg_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
[VDEV_RQ_HISTO_BUCKETS];
} vdev_stat_ex_t;
/*
* Initialize functions.
*/
typedef enum pool_initialize_func {
POOL_INITIALIZE_START,
POOL_INITIALIZE_CANCEL,
POOL_INITIALIZE_SUSPEND,
POOL_INITIALIZE_FUNCS
} pool_initialize_func_t;
/*
* TRIM functions.
*/
typedef enum pool_trim_func {
POOL_TRIM_START,
POOL_TRIM_CANCEL,
POOL_TRIM_SUSPEND,
POOL_TRIM_FUNCS
} pool_trim_func_t;
/*
* DDT statistics. Note: all fields should be 64-bit because this
* is passed between kernel and userland as an nvlist uint64 array.
*/
typedef struct ddt_object {
uint64_t ddo_count; /* number of elements in ddt */
uint64_t ddo_dspace; /* size of ddt on disk */
uint64_t ddo_mspace; /* size of ddt in-core */
} ddt_object_t;
typedef struct ddt_stat {
uint64_t dds_blocks; /* blocks */
uint64_t dds_lsize; /* logical size */
uint64_t dds_psize; /* physical size */
uint64_t dds_dsize; /* deflated allocated size */
uint64_t dds_ref_blocks; /* referenced blocks */
uint64_t dds_ref_lsize; /* referenced lsize * refcnt */
uint64_t dds_ref_psize; /* referenced psize * refcnt */
uint64_t dds_ref_dsize; /* referenced dsize * refcnt */
} ddt_stat_t;
typedef struct ddt_histogram {
ddt_stat_t ddh_stat[64]; /* power-of-two histogram buckets */
} ddt_histogram_t;
#define ZVOL_DRIVER "zvol"
#define ZFS_DRIVER "zfs"
#define ZFS_DEV "/dev/zfs"
#define ZFS_SUPER_MAGIC 0x2fc12fc1
/* general zvol path */
#define ZVOL_DIR "/dev/zvol/"
#define ZVOL_MAJOR 230
#define ZVOL_MINOR_BITS 4
#define ZVOL_MINOR_MASK ((1U << ZVOL_MINOR_BITS) - 1)
#define ZVOL_MINORS (1 << 4)
#define ZVOL_DEV_NAME "zd"
#define ZVOL_PROP_NAME "name"
#define ZVOL_DEFAULT_BLOCKSIZE 16384
typedef enum {
VDEV_INITIALIZE_NONE,
VDEV_INITIALIZE_ACTIVE,
VDEV_INITIALIZE_CANCELED,
VDEV_INITIALIZE_SUSPENDED,
VDEV_INITIALIZE_COMPLETE
} vdev_initializing_state_t;
typedef enum {
VDEV_TRIM_NONE,
VDEV_TRIM_ACTIVE,
VDEV_TRIM_CANCELED,
VDEV_TRIM_SUSPENDED,
VDEV_TRIM_COMPLETE,
} vdev_trim_state_t;
typedef enum {
VDEV_REBUILD_NONE,
VDEV_REBUILD_ACTIVE,
VDEV_REBUILD_CANCELED,
VDEV_REBUILD_COMPLETE,
} vdev_rebuild_state_t;
/*
* nvlist name constants. Facilitate restricting snapshot iteration range for
* the "list next snapshot" ioctl
*/
#define SNAP_ITER_MIN_TXG "snap_iter_min_txg"
#define SNAP_ITER_MAX_TXG "snap_iter_max_txg"
/*
* /dev/zfs ioctl numbers.
*
* These numbers cannot change over time. New ioctl numbers must be appended.
*/
typedef enum zfs_ioc {
/*
* Core features - 81/128 numbers reserved.
*/
#ifdef __FreeBSD__
ZFS_IOC_FIRST = 0,
#else
ZFS_IOC_FIRST = ('Z' << 8),
#endif
ZFS_IOC = ZFS_IOC_FIRST,
ZFS_IOC_POOL_CREATE = ZFS_IOC_FIRST, /* 0x5a00 */
ZFS_IOC_POOL_DESTROY, /* 0x5a01 */
ZFS_IOC_POOL_IMPORT, /* 0x5a02 */
ZFS_IOC_POOL_EXPORT, /* 0x5a03 */
ZFS_IOC_POOL_CONFIGS, /* 0x5a04 */
ZFS_IOC_POOL_STATS, /* 0x5a05 */
ZFS_IOC_POOL_TRYIMPORT, /* 0x5a06 */
ZFS_IOC_POOL_SCAN, /* 0x5a07 */
ZFS_IOC_POOL_FREEZE, /* 0x5a08 */
ZFS_IOC_POOL_UPGRADE, /* 0x5a09 */
ZFS_IOC_POOL_GET_HISTORY, /* 0x5a0a */
ZFS_IOC_VDEV_ADD, /* 0x5a0b */
ZFS_IOC_VDEV_REMOVE, /* 0x5a0c */
ZFS_IOC_VDEV_SET_STATE, /* 0x5a0d */
ZFS_IOC_VDEV_ATTACH, /* 0x5a0e */
ZFS_IOC_VDEV_DETACH, /* 0x5a0f */
ZFS_IOC_VDEV_SETPATH, /* 0x5a10 */
ZFS_IOC_VDEV_SETFRU, /* 0x5a11 */
ZFS_IOC_OBJSET_STATS, /* 0x5a12 */
ZFS_IOC_OBJSET_ZPLPROPS, /* 0x5a13 */
ZFS_IOC_DATASET_LIST_NEXT, /* 0x5a14 */
ZFS_IOC_SNAPSHOT_LIST_NEXT, /* 0x5a15 */
ZFS_IOC_SET_PROP, /* 0x5a16 */
ZFS_IOC_CREATE, /* 0x5a17 */
ZFS_IOC_DESTROY, /* 0x5a18 */
ZFS_IOC_ROLLBACK, /* 0x5a19 */
ZFS_IOC_RENAME, /* 0x5a1a */
ZFS_IOC_RECV, /* 0x5a1b */
ZFS_IOC_SEND, /* 0x5a1c */
ZFS_IOC_INJECT_FAULT, /* 0x5a1d */
ZFS_IOC_CLEAR_FAULT, /* 0x5a1e */
ZFS_IOC_INJECT_LIST_NEXT, /* 0x5a1f */
ZFS_IOC_ERROR_LOG, /* 0x5a20 */
ZFS_IOC_CLEAR, /* 0x5a21 */
ZFS_IOC_PROMOTE, /* 0x5a22 */
ZFS_IOC_SNAPSHOT, /* 0x5a23 */
ZFS_IOC_DSOBJ_TO_DSNAME, /* 0x5a24 */
ZFS_IOC_OBJ_TO_PATH, /* 0x5a25 */
ZFS_IOC_POOL_SET_PROPS, /* 0x5a26 */
ZFS_IOC_POOL_GET_PROPS, /* 0x5a27 */
ZFS_IOC_SET_FSACL, /* 0x5a28 */
ZFS_IOC_GET_FSACL, /* 0x5a29 */
ZFS_IOC_SHARE, /* 0x5a2a */
ZFS_IOC_INHERIT_PROP, /* 0x5a2b */
ZFS_IOC_SMB_ACL, /* 0x5a2c */
ZFS_IOC_USERSPACE_ONE, /* 0x5a2d */
ZFS_IOC_USERSPACE_MANY, /* 0x5a2e */
ZFS_IOC_USERSPACE_UPGRADE, /* 0x5a2f */
ZFS_IOC_HOLD, /* 0x5a30 */
ZFS_IOC_RELEASE, /* 0x5a31 */
ZFS_IOC_GET_HOLDS, /* 0x5a32 */
ZFS_IOC_OBJSET_RECVD_PROPS, /* 0x5a33 */
ZFS_IOC_VDEV_SPLIT, /* 0x5a34 */
ZFS_IOC_NEXT_OBJ, /* 0x5a35 */
ZFS_IOC_DIFF, /* 0x5a36 */
ZFS_IOC_TMP_SNAPSHOT, /* 0x5a37 */
ZFS_IOC_OBJ_TO_STATS, /* 0x5a38 */
ZFS_IOC_SPACE_WRITTEN, /* 0x5a39 */
ZFS_IOC_SPACE_SNAPS, /* 0x5a3a */
ZFS_IOC_DESTROY_SNAPS, /* 0x5a3b */
ZFS_IOC_POOL_REGUID, /* 0x5a3c */
ZFS_IOC_POOL_REOPEN, /* 0x5a3d */
ZFS_IOC_SEND_PROGRESS, /* 0x5a3e */
ZFS_IOC_LOG_HISTORY, /* 0x5a3f */
ZFS_IOC_SEND_NEW, /* 0x5a40 */
ZFS_IOC_SEND_SPACE, /* 0x5a41 */
ZFS_IOC_CLONE, /* 0x5a42 */
ZFS_IOC_BOOKMARK, /* 0x5a43 */
ZFS_IOC_GET_BOOKMARKS, /* 0x5a44 */
ZFS_IOC_DESTROY_BOOKMARKS, /* 0x5a45 */
ZFS_IOC_RECV_NEW, /* 0x5a46 */
ZFS_IOC_POOL_SYNC, /* 0x5a47 */
ZFS_IOC_CHANNEL_PROGRAM, /* 0x5a48 */
ZFS_IOC_LOAD_KEY, /* 0x5a49 */
ZFS_IOC_UNLOAD_KEY, /* 0x5a4a */
ZFS_IOC_CHANGE_KEY, /* 0x5a4b */
ZFS_IOC_REMAP, /* 0x5a4c */
ZFS_IOC_POOL_CHECKPOINT, /* 0x5a4d */
ZFS_IOC_POOL_DISCARD_CHECKPOINT, /* 0x5a4e */
ZFS_IOC_POOL_INITIALIZE, /* 0x5a4f */
ZFS_IOC_POOL_TRIM, /* 0x5a50 */
ZFS_IOC_REDACT, /* 0x5a51 */
ZFS_IOC_GET_BOOKMARK_PROPS, /* 0x5a52 */
ZFS_IOC_WAIT, /* 0x5a53 */
ZFS_IOC_WAIT_FS, /* 0x5a54 */
ZFS_IOC_VDEV_GET_PROPS, /* 0x5a55 */
ZFS_IOC_VDEV_SET_PROPS, /* 0x5a56 */
/*
* Per-platform (Optional) - 8/128 numbers reserved.
*/
ZFS_IOC_PLATFORM = ZFS_IOC_FIRST + 0x80,
ZFS_IOC_EVENTS_NEXT, /* 0x81 (Linux) */
ZFS_IOC_EVENTS_CLEAR, /* 0x82 (Linux) */
ZFS_IOC_EVENTS_SEEK, /* 0x83 (Linux) */
ZFS_IOC_NEXTBOOT, /* 0x84 (FreeBSD) */
ZFS_IOC_JAIL, /* 0x85 (FreeBSD) */
ZFS_IOC_UNJAIL, /* 0x86 (FreeBSD) */
ZFS_IOC_SET_BOOTENV, /* 0x87 */
ZFS_IOC_GET_BOOTENV, /* 0x88 */
ZFS_IOC_LAST
} zfs_ioc_t;
/*
* zvol ioctl to get dataset name
*/
#define BLKZNAME _IOR(0x12, 125, char[ZFS_MAX_DATASET_NAME_LEN])
/*
* ZFS-specific error codes used for returning descriptive errors
* to the userland through zfs ioctls.
*
* The enum implicitly includes all the error codes from errno.h.
* New code should use and extend this enum for errors that are
* not described precisely by generic errno codes.
*
* These numbers should not change over time. New entries should be appended.
*
* (Keep in sync with contrib/pyzfs/libzfs_core/_constants.py)
*/
typedef enum {
ZFS_ERR_CHECKPOINT_EXISTS = 1024,
ZFS_ERR_DISCARDING_CHECKPOINT,
ZFS_ERR_NO_CHECKPOINT,
ZFS_ERR_DEVRM_IN_PROGRESS,
ZFS_ERR_VDEV_TOO_BIG,
ZFS_ERR_IOC_CMD_UNAVAIL,
ZFS_ERR_IOC_ARG_UNAVAIL,
ZFS_ERR_IOC_ARG_REQUIRED,
ZFS_ERR_IOC_ARG_BADTYPE,
ZFS_ERR_WRONG_PARENT,
ZFS_ERR_FROM_IVSET_GUID_MISSING,
ZFS_ERR_FROM_IVSET_GUID_MISMATCH,
ZFS_ERR_SPILL_BLOCK_FLAG_MISSING,
ZFS_ERR_UNKNOWN_SEND_STREAM_FEATURE,
ZFS_ERR_EXPORT_IN_PROGRESS,
ZFS_ERR_BOOKMARK_SOURCE_NOT_ANCESTOR,
ZFS_ERR_STREAM_TRUNCATED,
ZFS_ERR_STREAM_LARGE_BLOCK_MISMATCH,
ZFS_ERR_RESILVER_IN_PROGRESS,
ZFS_ERR_REBUILD_IN_PROGRESS,
ZFS_ERR_BADPROP,
ZFS_ERR_VDEV_NOTSUP,
} zfs_errno_t;
/*
* Internal SPA load state. Used by FMA diagnosis engine.
*/
typedef enum {
SPA_LOAD_NONE, /* no load in progress */
SPA_LOAD_OPEN, /* normal open */
SPA_LOAD_IMPORT, /* import in progress */
SPA_LOAD_TRYIMPORT, /* tryimport in progress */
SPA_LOAD_RECOVER, /* recovery requested */
SPA_LOAD_ERROR, /* load failed */
SPA_LOAD_CREATE /* creation in progress */
} spa_load_state_t;
typedef enum {
ZPOOL_WAIT_CKPT_DISCARD,
ZPOOL_WAIT_FREE,
ZPOOL_WAIT_INITIALIZE,
ZPOOL_WAIT_REPLACE,
ZPOOL_WAIT_REMOVE,
ZPOOL_WAIT_RESILVER,
ZPOOL_WAIT_SCRUB,
ZPOOL_WAIT_TRIM,
ZPOOL_WAIT_NUM_ACTIVITIES
} zpool_wait_activity_t;
typedef enum {
ZFS_WAIT_DELETEQ,
ZFS_WAIT_NUM_ACTIVITIES
} zfs_wait_activity_t;
/*
* Bookmark name values.
*/
#define ZPOOL_ERR_LIST "error list"
#define ZPOOL_ERR_DATASET "dataset"
#define ZPOOL_ERR_OBJECT "object"
#define HIS_MAX_RECORD_LEN (MAXPATHLEN + MAXPATHLEN + 1)
/*
* The following are names used in the nvlist describing
* the pool's history log.
*/
#define ZPOOL_HIST_RECORD "history record"
#define ZPOOL_HIST_TIME "history time"
#define ZPOOL_HIST_CMD "history command"
#define ZPOOL_HIST_WHO "history who"
#define ZPOOL_HIST_ZONE "history zone"
#define ZPOOL_HIST_HOST "history hostname"
#define ZPOOL_HIST_TXG "history txg"
#define ZPOOL_HIST_INT_EVENT "history internal event"
#define ZPOOL_HIST_INT_STR "history internal str"
#define ZPOOL_HIST_INT_NAME "internal_name"
#define ZPOOL_HIST_IOCTL "ioctl"
#define ZPOOL_HIST_INPUT_NVL "in_nvl"
#define ZPOOL_HIST_OUTPUT_NVL "out_nvl"
#define ZPOOL_HIST_OUTPUT_SIZE "out_size"
#define ZPOOL_HIST_DSNAME "dsname"
#define ZPOOL_HIST_DSID "dsid"
#define ZPOOL_HIST_ERRNO "errno"
#define ZPOOL_HIST_ELAPSED_NS "elapsed_ns"
/*
* Special nvlist name that will not have its args recorded in the pool's
* history log.
*/
#define ZPOOL_HIDDEN_ARGS "hidden_args"
/*
* The following are names used when invoking ZFS_IOC_POOL_INITIALIZE.
*/
#define ZPOOL_INITIALIZE_COMMAND "initialize_command"
#define ZPOOL_INITIALIZE_VDEVS "initialize_vdevs"
/*
* The following are names used when invoking ZFS_IOC_POOL_TRIM.
*/
#define ZPOOL_TRIM_COMMAND "trim_command"
#define ZPOOL_TRIM_VDEVS "trim_vdevs"
#define ZPOOL_TRIM_RATE "trim_rate"
#define ZPOOL_TRIM_SECURE "trim_secure"
/*
* The following are names used when invoking ZFS_IOC_POOL_WAIT.
*/
#define ZPOOL_WAIT_ACTIVITY "wait_activity"
#define ZPOOL_WAIT_TAG "wait_tag"
#define ZPOOL_WAIT_WAITED "wait_waited"
/*
* The following are names used when invoking ZFS_IOC_VDEV_GET_PROP.
*/
#define ZPOOL_VDEV_PROPS_GET_VDEV "vdevprops_get_vdev"
#define ZPOOL_VDEV_PROPS_GET_PROPS "vdevprops_get_props"
/*
* The following are names used when invoking ZFS_IOC_VDEV_SET_PROP.
*/
#define ZPOOL_VDEV_PROPS_SET_VDEV "vdevprops_set_vdev"
#define ZPOOL_VDEV_PROPS_SET_PROPS "vdevprops_set_props"
/*
* The following are names used when invoking ZFS_IOC_WAIT_FS.
*/
#define ZFS_WAIT_ACTIVITY "wait_activity"
#define ZFS_WAIT_WAITED "wait_waited"
/*
* Flags for ZFS_IOC_VDEV_SET_STATE
*/
#define ZFS_ONLINE_CHECKREMOVE 0x1
#define ZFS_ONLINE_UNSPARE 0x2
#define ZFS_ONLINE_FORCEFAULT 0x4
#define ZFS_ONLINE_EXPAND 0x8
#define ZFS_OFFLINE_TEMPORARY 0x1
/*
* Flags for ZFS_IOC_POOL_IMPORT
*/
#define ZFS_IMPORT_NORMAL 0x0
#define ZFS_IMPORT_VERBATIM 0x1
#define ZFS_IMPORT_ANY_HOST 0x2
#define ZFS_IMPORT_MISSING_LOG 0x4
#define ZFS_IMPORT_ONLY 0x8
#define ZFS_IMPORT_TEMP_NAME 0x10
#define ZFS_IMPORT_SKIP_MMP 0x20
#define ZFS_IMPORT_LOAD_KEYS 0x40
#define ZFS_IMPORT_CHECKPOINT 0x80
/*
* Channel program argument/return nvlist keys and defaults.
*/
#define ZCP_ARG_PROGRAM "program"
#define ZCP_ARG_ARGLIST "arg"
#define ZCP_ARG_SYNC "sync"
#define ZCP_ARG_INSTRLIMIT "instrlimit"
#define ZCP_ARG_MEMLIMIT "memlimit"
#define ZCP_ARG_CLIARGV "argv"
#define ZCP_RET_ERROR "error"
#define ZCP_RET_RETURN "return"
#define ZCP_DEFAULT_INSTRLIMIT (10 * 1000 * 1000)
#define ZCP_MAX_INSTRLIMIT (10 * ZCP_DEFAULT_INSTRLIMIT)
#define ZCP_DEFAULT_MEMLIMIT (10 * 1024 * 1024)
#define ZCP_MAX_MEMLIMIT (10 * ZCP_DEFAULT_MEMLIMIT)
/*
* Sysevent payload members. ZFS will generate the following sysevents with the
* given payloads:
*
* ESC_ZFS_RESILVER_START
* ESC_ZFS_RESILVER_FINISH
*
* ZFS_EV_POOL_NAME DATA_TYPE_STRING
* ZFS_EV_POOL_GUID DATA_TYPE_UINT64
* ZFS_EV_RESILVER_TYPE DATA_TYPE_STRING
*
* ESC_ZFS_POOL_DESTROY
* ESC_ZFS_POOL_REGUID
*
* ZFS_EV_POOL_NAME DATA_TYPE_STRING
* ZFS_EV_POOL_GUID DATA_TYPE_UINT64
*
* ESC_ZFS_VDEV_REMOVE
* ESC_ZFS_VDEV_CLEAR
* ESC_ZFS_VDEV_CHECK
*
* ZFS_EV_POOL_NAME DATA_TYPE_STRING
* ZFS_EV_POOL_GUID DATA_TYPE_UINT64
* ZFS_EV_VDEV_PATH DATA_TYPE_STRING (optional)
* ZFS_EV_VDEV_GUID DATA_TYPE_UINT64
*
* ESC_ZFS_HISTORY_EVENT
*
* ZFS_EV_POOL_NAME DATA_TYPE_STRING
* ZFS_EV_POOL_GUID DATA_TYPE_UINT64
* ZFS_EV_HIST_TIME DATA_TYPE_UINT64 (optional)
* ZFS_EV_HIST_CMD DATA_TYPE_STRING (optional)
* ZFS_EV_HIST_WHO DATA_TYPE_UINT64 (optional)
* ZFS_EV_HIST_ZONE DATA_TYPE_STRING (optional)
* ZFS_EV_HIST_HOST DATA_TYPE_STRING (optional)
* ZFS_EV_HIST_TXG DATA_TYPE_UINT64 (optional)
* ZFS_EV_HIST_INT_EVENT DATA_TYPE_UINT64 (optional)
* ZFS_EV_HIST_INT_STR DATA_TYPE_STRING (optional)
* ZFS_EV_HIST_INT_NAME DATA_TYPE_STRING (optional)
* ZFS_EV_HIST_IOCTL DATA_TYPE_STRING (optional)
* ZFS_EV_HIST_DSNAME DATA_TYPE_STRING (optional)
* ZFS_EV_HIST_DSID DATA_TYPE_UINT64 (optional)
*
* The ZFS_EV_HIST_* members will correspond to the ZPOOL_HIST_* members in the
* history log nvlist. The keynames will be free of any spaces or other
* characters that could be potentially unexpected to consumers of the
* sysevents.
*/
#define ZFS_EV_POOL_NAME "pool_name"
#define ZFS_EV_POOL_GUID "pool_guid"
#define ZFS_EV_VDEV_PATH "vdev_path"
#define ZFS_EV_VDEV_GUID "vdev_guid"
#define ZFS_EV_HIST_TIME "history_time"
#define ZFS_EV_HIST_CMD "history_command"
#define ZFS_EV_HIST_WHO "history_who"
#define ZFS_EV_HIST_ZONE "history_zone"
#define ZFS_EV_HIST_HOST "history_hostname"
#define ZFS_EV_HIST_TXG "history_txg"
#define ZFS_EV_HIST_INT_EVENT "history_internal_event"
#define ZFS_EV_HIST_INT_STR "history_internal_str"
#define ZFS_EV_HIST_INT_NAME "history_internal_name"
#define ZFS_EV_HIST_IOCTL "history_ioctl"
#define ZFS_EV_HIST_DSNAME "history_dsname"
#define ZFS_EV_HIST_DSID "history_dsid"
#define ZFS_EV_RESILVER_TYPE "resilver_type"
/*
* We currently support block sizes from 512 bytes to 16MB.
* The benefits of larger blocks, and thus larger IO, need to be weighed
* against the cost of COWing a giant block to modify one byte, and the
* large latency of reading or writing a large block.
*
* Note that although blocks up to 16MB are supported, the recordsize
* property can not be set larger than zfs_max_recordsize (default 1MB).
* See the comment near zfs_max_recordsize in dsl_dataset.c for details.
*
* Note that although the LSIZE field of the blkptr_t can store sizes up
* to 32MB, the dnode's dn_datablkszsec can only store sizes up to
* 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
*/
#define SPA_MINBLOCKSHIFT 9
#define SPA_OLD_MAXBLOCKSHIFT 17
#define SPA_MAXBLOCKSHIFT 24
#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
#define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT)
#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
/* supported encryption algorithms */
enum zio_encrypt {
ZIO_CRYPT_INHERIT = 0,
ZIO_CRYPT_ON,
ZIO_CRYPT_OFF,
ZIO_CRYPT_AES_128_CCM,
ZIO_CRYPT_AES_192_CCM,
ZIO_CRYPT_AES_256_CCM,
ZIO_CRYPT_AES_128_GCM,
ZIO_CRYPT_AES_192_GCM,
ZIO_CRYPT_AES_256_GCM,
ZIO_CRYPT_FUNCTIONS
};
#define ZIO_CRYPT_ON_VALUE ZIO_CRYPT_AES_256_GCM
#define ZIO_CRYPT_DEFAULT ZIO_CRYPT_OFF
#ifdef __cplusplus
}
#endif
#endif /* _SYS_FS_ZFS_H */
diff --git a/include/sys/metaslab.h b/include/sys/metaslab.h
index ecff65f13de5..129a68be41c5 100644
--- a/include/sys/metaslab.h
+++ b/include/sys/metaslab.h
@@ -1,148 +1,148 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
* Copyright (c) 2017, Intel Corporation.
*/
#ifndef _SYS_METASLAB_H
#define _SYS_METASLAB_H
#include <sys/spa.h>
#include <sys/space_map.h>
#include <sys/txg.h>
#include <sys/zio.h>
#include <sys/avl.h>
#ifdef __cplusplus
extern "C" {
#endif
typedef struct metaslab_ops {
uint64_t (*msop_alloc)(metaslab_t *, uint64_t);
} metaslab_ops_t;
-extern metaslab_ops_t *zfs_metaslab_ops;
+extern const metaslab_ops_t zfs_metaslab_ops;
int metaslab_init(metaslab_group_t *, uint64_t, uint64_t, uint64_t,
metaslab_t **);
void metaslab_fini(metaslab_t *);
void metaslab_set_unflushed_txg(metaslab_t *, uint64_t, dmu_tx_t *);
void metaslab_set_estimated_condensed_size(metaslab_t *, uint64_t, dmu_tx_t *);
uint64_t metaslab_unflushed_txg(metaslab_t *);
uint64_t metaslab_estimated_condensed_size(metaslab_t *);
int metaslab_sort_by_flushed(const void *, const void *);
uint64_t metaslab_unflushed_changes_memused(metaslab_t *);
int metaslab_load(metaslab_t *);
void metaslab_unload(metaslab_t *);
boolean_t metaslab_flush(metaslab_t *, dmu_tx_t *);
uint64_t metaslab_allocated_space(metaslab_t *);
void metaslab_sync(metaslab_t *, uint64_t);
void metaslab_sync_done(metaslab_t *, uint64_t);
void metaslab_sync_reassess(metaslab_group_t *);
uint64_t metaslab_largest_allocatable(metaslab_t *);
/*
* metaslab alloc flags
*/
#define METASLAB_HINTBP_FAVOR 0x0
#define METASLAB_HINTBP_AVOID 0x1
#define METASLAB_GANG_HEADER 0x2
#define METASLAB_GANG_CHILD 0x4
#define METASLAB_ASYNC_ALLOC 0x8
#define METASLAB_DONT_THROTTLE 0x10
#define METASLAB_MUST_RESERVE 0x20
#define METASLAB_FASTWRITE 0x40
#define METASLAB_ZIL 0x80
int metaslab_alloc(spa_t *, metaslab_class_t *, uint64_t,
blkptr_t *, int, uint64_t, blkptr_t *, int, zio_alloc_list_t *, zio_t *,
int);
int metaslab_alloc_dva(spa_t *, metaslab_class_t *, uint64_t,
dva_t *, int, dva_t *, uint64_t, int, zio_alloc_list_t *, int);
void metaslab_free(spa_t *, const blkptr_t *, uint64_t, boolean_t);
void metaslab_free_concrete(vdev_t *, uint64_t, uint64_t, boolean_t);
void metaslab_free_dva(spa_t *, const dva_t *, boolean_t);
void metaslab_free_impl_cb(uint64_t, vdev_t *, uint64_t, uint64_t, void *);
void metaslab_unalloc_dva(spa_t *, const dva_t *, uint64_t);
int metaslab_claim(spa_t *, const blkptr_t *, uint64_t);
int metaslab_claim_impl(vdev_t *, uint64_t, uint64_t, uint64_t);
void metaslab_check_free(spa_t *, const blkptr_t *);
void metaslab_fastwrite_mark(spa_t *, const blkptr_t *);
void metaslab_fastwrite_unmark(spa_t *, const blkptr_t *);
void metaslab_stat_init(void);
void metaslab_stat_fini(void);
void metaslab_trace_init(zio_alloc_list_t *);
void metaslab_trace_fini(zio_alloc_list_t *);
-metaslab_class_t *metaslab_class_create(spa_t *, metaslab_ops_t *);
+metaslab_class_t *metaslab_class_create(spa_t *, const metaslab_ops_t *);
void metaslab_class_destroy(metaslab_class_t *);
int metaslab_class_validate(metaslab_class_t *);
void metaslab_class_histogram_verify(metaslab_class_t *);
uint64_t metaslab_class_fragmentation(metaslab_class_t *);
uint64_t metaslab_class_expandable_space(metaslab_class_t *);
boolean_t metaslab_class_throttle_reserve(metaslab_class_t *, int, int,
zio_t *, int);
void metaslab_class_throttle_unreserve(metaslab_class_t *, int, int, zio_t *);
void metaslab_class_evict_old(metaslab_class_t *, uint64_t);
uint64_t metaslab_class_get_alloc(metaslab_class_t *);
uint64_t metaslab_class_get_space(metaslab_class_t *);
uint64_t metaslab_class_get_dspace(metaslab_class_t *);
uint64_t metaslab_class_get_deferred(metaslab_class_t *);
void metaslab_space_update(vdev_t *, metaslab_class_t *,
int64_t, int64_t, int64_t);
metaslab_group_t *metaslab_group_create(metaslab_class_t *, vdev_t *, int);
void metaslab_group_destroy(metaslab_group_t *);
void metaslab_group_activate(metaslab_group_t *);
void metaslab_group_passivate(metaslab_group_t *);
boolean_t metaslab_group_initialized(metaslab_group_t *);
uint64_t metaslab_group_get_space(metaslab_group_t *);
void metaslab_group_histogram_verify(metaslab_group_t *);
uint64_t metaslab_group_fragmentation(metaslab_group_t *);
void metaslab_group_histogram_remove(metaslab_group_t *, metaslab_t *);
void metaslab_group_alloc_decrement(spa_t *, uint64_t, void *, int, int,
boolean_t);
void metaslab_group_alloc_verify(spa_t *, const blkptr_t *, void *, int);
void metaslab_recalculate_weight_and_sort(metaslab_t *);
void metaslab_disable(metaslab_t *);
void metaslab_enable(metaslab_t *, boolean_t, boolean_t);
void metaslab_set_selected_txg(metaslab_t *, uint64_t);
extern int metaslab_debug_load;
range_seg_type_t metaslab_calculate_range_tree_type(vdev_t *vdev,
metaslab_t *msp, uint64_t *start, uint64_t *shift);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_METASLAB_H */
diff --git a/include/sys/metaslab_impl.h b/include/sys/metaslab_impl.h
index adf4c03a20db..3dbee4c17fef 100644
--- a/include/sys/metaslab_impl.h
+++ b/include/sys/metaslab_impl.h
@@ -1,572 +1,572 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
*/
#ifndef _SYS_METASLAB_IMPL_H
#define _SYS_METASLAB_IMPL_H
#include <sys/metaslab.h>
#include <sys/space_map.h>
#include <sys/range_tree.h>
#include <sys/vdev.h>
#include <sys/txg.h>
#include <sys/avl.h>
#include <sys/multilist.h>
#ifdef __cplusplus
extern "C" {
#endif
/*
* Metaslab allocation tracing record.
*/
typedef struct metaslab_alloc_trace {
list_node_t mat_list_node;
metaslab_group_t *mat_mg;
metaslab_t *mat_msp;
uint64_t mat_size;
uint64_t mat_weight;
uint32_t mat_dva_id;
uint64_t mat_offset;
int mat_allocator;
} metaslab_alloc_trace_t;
/*
* Used by the metaslab allocation tracing facility to indicate
* error conditions. These errors are stored to the offset member
* of the metaslab_alloc_trace_t record and displayed by mdb.
*/
typedef enum trace_alloc_type {
TRACE_ALLOC_FAILURE = -1ULL,
TRACE_TOO_SMALL = -2ULL,
TRACE_FORCE_GANG = -3ULL,
TRACE_NOT_ALLOCATABLE = -4ULL,
TRACE_GROUP_FAILURE = -5ULL,
TRACE_ENOSPC = -6ULL,
TRACE_CONDENSING = -7ULL,
TRACE_VDEV_ERROR = -8ULL,
TRACE_DISABLED = -9ULL,
} trace_alloc_type_t;
#define METASLAB_WEIGHT_PRIMARY (1ULL << 63)
#define METASLAB_WEIGHT_SECONDARY (1ULL << 62)
#define METASLAB_WEIGHT_CLAIM (1ULL << 61)
#define METASLAB_WEIGHT_TYPE (1ULL << 60)
#define METASLAB_ACTIVE_MASK \
(METASLAB_WEIGHT_PRIMARY | METASLAB_WEIGHT_SECONDARY | \
METASLAB_WEIGHT_CLAIM)
/*
* The metaslab weight is used to encode the amount of free space in a
* metaslab, such that the "best" metaslab appears first when sorting the
* metaslabs by weight. The weight (and therefore the "best" metaslab) can
* be determined in two different ways: by computing a weighted sum of all
* the free space in the metaslab (a space based weight) or by counting only
* the free segments of the largest size (a segment based weight). We prefer
* the segment based weight because it reflects how the free space is
* comprised, but we cannot always use it -- legacy pools do not have the
* space map histogram information necessary to determine the largest
* contiguous regions. Pools that have the space map histogram determine
* the segment weight by looking at each bucket in the histogram and
* determining the free space whose size in bytes is in the range:
* [2^i, 2^(i+1))
* We then encode the largest index, i, that contains regions into the
* segment-weighted value.
*
* Space-based weight:
*
* 64 56 48 40 32 24 16 8 0
* +-------+-------+-------+-------+-------+-------+-------+-------+
* |PSC1| weighted-free space |
* +-------+-------+-------+-------+-------+-------+-------+-------+
*
* PS - indicates primary and secondary activation
* C - indicates activation for claimed block zio
* space - the fragmentation-weighted space
*
* Segment-based weight:
*
* 64 56 48 40 32 24 16 8 0
* +-------+-------+-------+-------+-------+-------+-------+-------+
* |PSC0| idx| count of segments in region |
* +-------+-------+-------+-------+-------+-------+-------+-------+
*
* PS - indicates primary and secondary activation
* C - indicates activation for claimed block zio
* idx - index for the highest bucket in the histogram
* count - number of segments in the specified bucket
*/
#define WEIGHT_GET_ACTIVE(weight) BF64_GET((weight), 61, 3)
#define WEIGHT_SET_ACTIVE(weight, x) BF64_SET((weight), 61, 3, x)
#define WEIGHT_IS_SPACEBASED(weight) \
((weight) == 0 || BF64_GET((weight), 60, 1))
#define WEIGHT_SET_SPACEBASED(weight) BF64_SET((weight), 60, 1, 1)
/*
* These macros are only applicable to segment-based weighting.
*/
#define WEIGHT_GET_INDEX(weight) BF64_GET((weight), 54, 6)
#define WEIGHT_SET_INDEX(weight, x) BF64_SET((weight), 54, 6, x)
#define WEIGHT_GET_COUNT(weight) BF64_GET((weight), 0, 54)
#define WEIGHT_SET_COUNT(weight, x) BF64_SET((weight), 0, 54, x)
/*
* Per-allocator data structure.
*/
typedef struct metaslab_class_allocator {
metaslab_group_t *mca_rotor;
uint64_t mca_aliquot;
/*
* The allocation throttle works on a reservation system. Whenever
* an asynchronous zio wants to perform an allocation it must
* first reserve the number of blocks that it wants to allocate.
* If there aren't sufficient slots available for the pending zio
* then that I/O is throttled until more slots free up. The current
* number of reserved allocations is maintained by the mca_alloc_slots
* refcount. The mca_alloc_max_slots value determines the maximum
* number of allocations that the system allows. Gang blocks are
* allowed to reserve slots even if we've reached the maximum
* number of allocations allowed.
*/
uint64_t mca_alloc_max_slots;
zfs_refcount_t mca_alloc_slots;
} ____cacheline_aligned metaslab_class_allocator_t;
/*
* A metaslab class encompasses a category of allocatable top-level vdevs.
* Each top-level vdev is associated with a metaslab group which defines
* the allocatable region for that vdev. Examples of these categories include
* "normal" for data block allocations (i.e. main pool allocations) or "log"
* for allocations designated for intent log devices (i.e. slog devices).
* When a block allocation is requested from the SPA it is associated with a
* metaslab_class_t, and only top-level vdevs (i.e. metaslab groups) belonging
* to the class can be used to satisfy that request. Allocations are done
* by traversing the metaslab groups that are linked off of the mca_rotor field.
* This rotor points to the next metaslab group where allocations will be
* attempted. Allocating a block is a 3 step process -- select the metaslab
* group, select the metaslab, and then allocate the block. The metaslab
* class defines the low-level block allocator that will be used as the
* final step in allocation. These allocators are pluggable allowing each class
* to use a block allocator that best suits that class.
*/
struct metaslab_class {
kmutex_t mc_lock;
spa_t *mc_spa;
- metaslab_ops_t *mc_ops;
+ const metaslab_ops_t *mc_ops;
/*
* Track the number of metaslab groups that have been initialized
* and can accept allocations. An initialized metaslab group is
* one has been completely added to the config (i.e. we have
* updated the MOS config and the space has been added to the pool).
*/
uint64_t mc_groups;
/*
* Toggle to enable/disable the allocation throttle.
*/
boolean_t mc_alloc_throttle_enabled;
uint64_t mc_alloc_groups; /* # of allocatable groups */
uint64_t mc_alloc; /* total allocated space */
uint64_t mc_deferred; /* total deferred frees */
uint64_t mc_space; /* total space (alloc + free) */
uint64_t mc_dspace; /* total deflated space */
uint64_t mc_histogram[RANGE_TREE_HISTOGRAM_SIZE];
/*
* List of all loaded metaslabs in the class, sorted in order of most
* recent use.
*/
multilist_t mc_metaslab_txg_list;
metaslab_class_allocator_t mc_allocator[];
};
/*
* Per-allocator data structure.
*/
typedef struct metaslab_group_allocator {
uint64_t mga_cur_max_alloc_queue_depth;
zfs_refcount_t mga_alloc_queue_depth;
metaslab_t *mga_primary;
metaslab_t *mga_secondary;
} metaslab_group_allocator_t;
/*
* Metaslab groups encapsulate all the allocatable regions (i.e. metaslabs)
* of a top-level vdev. They are linked together to form a circular linked
* list and can belong to only one metaslab class. Metaslab groups may become
* ineligible for allocations for a number of reasons such as limited free
* space, fragmentation, or going offline. When this happens the allocator will
* simply find the next metaslab group in the linked list and attempt
* to allocate from that group instead.
*/
struct metaslab_group {
kmutex_t mg_lock;
avl_tree_t mg_metaslab_tree;
uint64_t mg_aliquot;
boolean_t mg_allocatable; /* can we allocate? */
uint64_t mg_ms_ready;
/*
* A metaslab group is considered to be initialized only after
* we have updated the MOS config and added the space to the pool.
* We only allow allocation attempts to a metaslab group if it
* has been initialized.
*/
boolean_t mg_initialized;
uint64_t mg_free_capacity; /* percentage free */
int64_t mg_bias;
int64_t mg_activation_count;
metaslab_class_t *mg_class;
vdev_t *mg_vd;
taskq_t *mg_taskq;
metaslab_group_t *mg_prev;
metaslab_group_t *mg_next;
/*
* In order for the allocation throttle to function properly, we cannot
* have too many IOs going to each disk by default; the throttle
* operates by allocating more work to disks that finish quickly, so
* allocating larger chunks to each disk reduces its effectiveness.
* However, if the number of IOs going to each allocator is too small,
* we will not perform proper aggregation at the vdev_queue layer,
* also resulting in decreased performance. Therefore, we will use a
* ramp-up strategy.
*
* Each allocator in each metaslab group has a current queue depth
* (mg_alloc_queue_depth[allocator]) and a current max queue depth
* (mga_cur_max_alloc_queue_depth[allocator]), and each metaslab group
* has an absolute max queue depth (mg_max_alloc_queue_depth). We
* add IOs to an allocator until the mg_alloc_queue_depth for that
* allocator hits the cur_max. Every time an IO completes for a given
* allocator on a given metaslab group, we increment its cur_max until
* it reaches mg_max_alloc_queue_depth. The cur_max resets every txg to
* help protect against disks that decrease in performance over time.
*
* It's possible for an allocator to handle more allocations than
* its max. This can occur when gang blocks are required or when other
* groups are unable to handle their share of allocations.
*/
uint64_t mg_max_alloc_queue_depth;
/*
* A metalab group that can no longer allocate the minimum block
* size will set mg_no_free_space. Once a metaslab group is out
* of space then its share of work must be distributed to other
* groups.
*/
boolean_t mg_no_free_space;
uint64_t mg_allocations;
uint64_t mg_failed_allocations;
uint64_t mg_fragmentation;
uint64_t mg_histogram[RANGE_TREE_HISTOGRAM_SIZE];
int mg_ms_disabled;
boolean_t mg_disabled_updating;
kmutex_t mg_ms_disabled_lock;
kcondvar_t mg_ms_disabled_cv;
int mg_allocators;
metaslab_group_allocator_t mg_allocator[];
};
/*
* This value defines the number of elements in the ms_lbas array. The value
* of 64 was chosen as it covers all power of 2 buckets up to UINT64_MAX.
* This is the equivalent of highbit(UINT64_MAX).
*/
#define MAX_LBAS 64
/*
* Each metaslab maintains a set of in-core trees to track metaslab
* operations. The in-core free tree (ms_allocatable) contains the list of
* free segments which are eligible for allocation. As blocks are
* allocated, the allocated segment are removed from the ms_allocatable and
* added to a per txg allocation tree (ms_allocating). As blocks are
* freed, they are added to the free tree (ms_freeing). These trees
* allow us to process all allocations and frees in syncing context
* where it is safe to update the on-disk space maps. An additional set
* of in-core trees is maintained to track deferred frees
* (ms_defer). Once a block is freed it will move from the
* ms_freed to the ms_defer tree. A deferred free means that a block
* has been freed but cannot be used by the pool until TXG_DEFER_SIZE
* transactions groups later. For example, a block that is freed in txg
* 50 will not be available for reallocation until txg 52 (50 +
* TXG_DEFER_SIZE). This provides a safety net for uberblock rollback.
* A pool could be safely rolled back TXG_DEFERS_SIZE transactions
* groups and ensure that no block has been reallocated.
*
* The simplified transition diagram looks like this:
*
*
* ALLOCATE
* |
* V
* free segment (ms_allocatable) -> ms_allocating[4] -> (write to space map)
* ^
* | ms_freeing <--- FREE
* | |
* | v
* | ms_freed
* | |
* +-------- ms_defer[2] <-------+-------> (write to space map)
*
*
* Each metaslab's space is tracked in a single space map in the MOS,
* which is only updated in syncing context. Each time we sync a txg,
* we append the allocs and frees from that txg to the space map. The
* pool space is only updated once all metaslabs have finished syncing.
*
* To load the in-core free tree we read the space map from disk. This
* object contains a series of alloc and free records that are combined
* to make up the list of all free segments in this metaslab. These
* segments are represented in-core by the ms_allocatable and are stored
* in an AVL tree.
*
* As the space map grows (as a result of the appends) it will
* eventually become space-inefficient. When the metaslab's in-core
* free tree is zfs_condense_pct/100 times the size of the minimal
* on-disk representation, we rewrite it in its minimized form. If a
* metaslab needs to condense then we must set the ms_condensing flag to
* ensure that allocations are not performed on the metaslab that is
* being written.
*/
struct metaslab {
/*
* This is the main lock of the metaslab and its purpose is to
* coordinate our allocations and frees [e.g metaslab_block_alloc(),
* metaslab_free_concrete(), ..etc] with our various syncing
* procedures [e.g. metaslab_sync(), metaslab_sync_done(), ..etc].
*
* The lock is also used during some miscellaneous operations like
* using the metaslab's histogram for the metaslab group's histogram
* aggregation, or marking the metaslab for initialization.
*/
kmutex_t ms_lock;
/*
* Acquired together with the ms_lock whenever we expect to
* write to metaslab data on-disk (i.e flushing entries to
* the metaslab's space map). It helps coordinate readers of
* the metaslab's space map [see spa_vdev_remove_thread()]
* with writers [see metaslab_sync() or metaslab_flush()].
*
* Note that metaslab_load(), even though a reader, uses
* a completely different mechanism to deal with the reading
* of the metaslab's space map based on ms_synced_length. That
* said, the function still uses the ms_sync_lock after it
* has read the ms_sm [see relevant comment in metaslab_load()
* as to why].
*/
kmutex_t ms_sync_lock;
kcondvar_t ms_load_cv;
space_map_t *ms_sm;
uint64_t ms_id;
uint64_t ms_start;
uint64_t ms_size;
uint64_t ms_fragmentation;
range_tree_t *ms_allocating[TXG_SIZE];
range_tree_t *ms_allocatable;
uint64_t ms_allocated_this_txg;
uint64_t ms_allocating_total;
/*
* The following range trees are accessed only from syncing context.
* ms_free*tree only have entries while syncing, and are empty
* between syncs.
*/
range_tree_t *ms_freeing; /* to free this syncing txg */
range_tree_t *ms_freed; /* already freed this syncing txg */
range_tree_t *ms_defer[TXG_DEFER_SIZE];
range_tree_t *ms_checkpointing; /* to add to the checkpoint */
/*
* The ms_trim tree is the set of allocatable segments which are
* eligible for trimming. (When the metaslab is loaded, it's a
* subset of ms_allocatable.) It's kept in-core as long as the
* autotrim property is set and is not vacated when the metaslab
* is unloaded. Its purpose is to aggregate freed ranges to
* facilitate efficient trimming.
*/
range_tree_t *ms_trim;
boolean_t ms_condensing; /* condensing? */
boolean_t ms_condense_wanted;
/*
* The number of consumers which have disabled the metaslab.
*/
uint64_t ms_disabled;
/*
* We must always hold the ms_lock when modifying ms_loaded
* and ms_loading.
*/
boolean_t ms_loaded;
boolean_t ms_loading;
kcondvar_t ms_flush_cv;
boolean_t ms_flushing;
/*
* The following histograms count entries that are in the
* metaslab's space map (and its histogram) but are not in
* ms_allocatable yet, because they are in ms_freed, ms_freeing,
* or ms_defer[].
*
* When the metaslab is not loaded, its ms_weight needs to
* reflect what is allocatable (i.e. what will be part of
* ms_allocatable if it is loaded). The weight is computed from
* the spacemap histogram, but that includes ranges that are
* not yet allocatable (because they are in ms_freed,
* ms_freeing, or ms_defer[]). Therefore, when calculating the
* weight, we need to remove those ranges.
*
* The ranges in the ms_freed and ms_defer[] range trees are all
* present in the spacemap. However, the spacemap may have
* multiple entries to represent a contiguous range, because it
* is written across multiple sync passes, but the changes of
* all sync passes are consolidated into the range trees.
* Adjacent ranges that are freed in different sync passes of
* one txg will be represented separately (as 2 or more entries)
* in the space map (and its histogram), but these adjacent
* ranges will be consolidated (represented as one entry) in the
* ms_freed/ms_defer[] range trees (and their histograms).
*
* When calculating the weight, we can not simply subtract the
* range trees' histograms from the spacemap's histogram,
* because the range trees' histograms may have entries in
* higher buckets than the spacemap, due to consolidation.
* Instead we must subtract the exact entries that were added to
* the spacemap's histogram. ms_synchist and ms_deferhist[]
* represent these exact entries, so we can subtract them from
* the spacemap's histogram when calculating ms_weight.
*
* ms_synchist represents the same ranges as ms_freeing +
* ms_freed, but without consolidation across sync passes.
*
* ms_deferhist[i] represents the same ranges as ms_defer[i],
* but without consolidation across sync passes.
*/
uint64_t ms_synchist[SPACE_MAP_HISTOGRAM_SIZE];
uint64_t ms_deferhist[TXG_DEFER_SIZE][SPACE_MAP_HISTOGRAM_SIZE];
/*
* Tracks the exact amount of allocated space of this metaslab
* (and specifically the metaslab's space map) up to the most
* recently completed sync pass [see usage in metaslab_sync()].
*/
uint64_t ms_allocated_space;
int64_t ms_deferspace; /* sum of ms_defermap[] space */
uint64_t ms_weight; /* weight vs. others in group */
uint64_t ms_activation_weight; /* activation weight */
/*
* Track of whenever a metaslab is selected for loading or allocation.
* We use this value to determine how long the metaslab should
* stay cached.
*/
uint64_t ms_selected_txg;
/*
* ms_load/unload_time can be used for performance monitoring
* (e.g. by dtrace or mdb).
*/
hrtime_t ms_load_time; /* time last loaded */
hrtime_t ms_unload_time; /* time last unloaded */
hrtime_t ms_selected_time; /* time last allocated from */
uint64_t ms_alloc_txg; /* last successful alloc (debug only) */
uint64_t ms_max_size; /* maximum allocatable size */
/*
* -1 if it's not active in an allocator, otherwise set to the allocator
* this metaslab is active for.
*/
int ms_allocator;
boolean_t ms_primary; /* Only valid if ms_allocator is not -1 */
/*
* The metaslab block allocators can optionally use a size-ordered
* range tree and/or an array of LBAs. Not all allocators use
* this functionality. The ms_allocatable_by_size should always
* contain the same number of segments as the ms_allocatable. The
* only difference is that the ms_allocatable_by_size is ordered by
* segment sizes.
*/
zfs_btree_t ms_allocatable_by_size;
zfs_btree_t ms_unflushed_frees_by_size;
uint64_t ms_lbas[MAX_LBAS];
metaslab_group_t *ms_group; /* metaslab group */
avl_node_t ms_group_node; /* node in metaslab group tree */
txg_node_t ms_txg_node; /* per-txg dirty metaslab links */
avl_node_t ms_spa_txg_node; /* node in spa_metaslabs_by_txg */
/*
* Node in metaslab class's selected txg list
*/
multilist_node_t ms_class_txg_node;
/*
* Allocs and frees that are committed to the vdev log spacemap but
* not yet to this metaslab's spacemap.
*/
range_tree_t *ms_unflushed_allocs;
range_tree_t *ms_unflushed_frees;
/*
* We have flushed entries up to but not including this TXG. In
* other words, all changes from this TXG and onward should not
* be in this metaslab's space map and must be read from the
* log space maps.
*/
uint64_t ms_unflushed_txg;
/* updated every time we are done syncing the metaslab's space map */
uint64_t ms_synced_length;
boolean_t ms_new;
};
typedef struct metaslab_unflushed_phys {
/* on-disk counterpart of ms_unflushed_txg */
uint64_t msp_unflushed_txg;
} metaslab_unflushed_phys_t;
#ifdef __cplusplus
}
#endif
#endif /* _SYS_METASLAB_IMPL_H */
diff --git a/include/sys/nvpair.h b/include/sys/nvpair.h
index 340b0d79c0a2..81494b62d7ec 100644
--- a/include/sys/nvpair.h
+++ b/include/sys/nvpair.h
@@ -1,416 +1,416 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2000, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
*/
#ifndef _SYS_NVPAIR_H
#define _SYS_NVPAIR_H extern __attribute__((visibility("default")))
#include <sys/types.h>
#include <sys/time.h>
#include <sys/errno.h>
#ifdef __cplusplus
extern "C" {
#endif
typedef enum {
DATA_TYPE_DONTCARE = -1,
DATA_TYPE_UNKNOWN = 0,
DATA_TYPE_BOOLEAN,
DATA_TYPE_BYTE,
DATA_TYPE_INT16,
DATA_TYPE_UINT16,
DATA_TYPE_INT32,
DATA_TYPE_UINT32,
DATA_TYPE_INT64,
DATA_TYPE_UINT64,
DATA_TYPE_STRING,
DATA_TYPE_BYTE_ARRAY,
DATA_TYPE_INT16_ARRAY,
DATA_TYPE_UINT16_ARRAY,
DATA_TYPE_INT32_ARRAY,
DATA_TYPE_UINT32_ARRAY,
DATA_TYPE_INT64_ARRAY,
DATA_TYPE_UINT64_ARRAY,
DATA_TYPE_STRING_ARRAY,
DATA_TYPE_HRTIME,
DATA_TYPE_NVLIST,
DATA_TYPE_NVLIST_ARRAY,
DATA_TYPE_BOOLEAN_VALUE,
DATA_TYPE_INT8,
DATA_TYPE_UINT8,
DATA_TYPE_BOOLEAN_ARRAY,
DATA_TYPE_INT8_ARRAY,
#if !defined(_KERNEL) && !defined(_STANDALONE)
DATA_TYPE_UINT8_ARRAY,
DATA_TYPE_DOUBLE
#else
DATA_TYPE_UINT8_ARRAY
#endif
} data_type_t;
typedef struct nvpair {
int32_t nvp_size; /* size of this nvpair */
int16_t nvp_name_sz; /* length of name string */
int16_t nvp_reserve; /* not used */
int32_t nvp_value_elem; /* number of elements for array types */
data_type_t nvp_type; /* type of value */
/* name string */
/* aligned ptr array for string arrays */
/* aligned array of data for value */
} nvpair_t;
/* nvlist header */
typedef struct nvlist {
int32_t nvl_version;
uint32_t nvl_nvflag; /* persistent flags */
uint64_t nvl_priv; /* ptr to private data if not packed */
uint32_t nvl_flag;
int32_t nvl_pad; /* currently not used, for alignment */
} nvlist_t;
/* nvp implementation version */
#define NV_VERSION 0
/* nvlist pack encoding */
#define NV_ENCODE_NATIVE 0
#define NV_ENCODE_XDR 1
/* nvlist persistent unique name flags, stored in nvl_nvflags */
#define NV_UNIQUE_NAME 0x1
#define NV_UNIQUE_NAME_TYPE 0x2
/* nvlist lookup pairs related flags */
#define NV_FLAG_NOENTOK 0x1
/* convenience macros */
#define NV_ALIGN(x) (((ulong_t)(x) + 7ul) & ~7ul)
#define NV_ALIGN4(x) (((x) + 3) & ~3)
#define NVP_SIZE(nvp) ((nvp)->nvp_size)
#define NVP_NAME(nvp) ((char *)(nvp) + sizeof (nvpair_t))
#define NVP_TYPE(nvp) ((nvp)->nvp_type)
#define NVP_NELEM(nvp) ((nvp)->nvp_value_elem)
#define NVP_VALUE(nvp) ((char *)(nvp) + NV_ALIGN(sizeof (nvpair_t) \
+ (nvp)->nvp_name_sz))
#define NVL_VERSION(nvl) ((nvl)->nvl_version)
#define NVL_SIZE(nvl) ((nvl)->nvl_size)
#define NVL_FLAG(nvl) ((nvl)->nvl_flag)
/* NV allocator framework */
typedef struct nv_alloc_ops nv_alloc_ops_t;
typedef struct nv_alloc {
const nv_alloc_ops_t *nva_ops;
void *nva_arg;
} nv_alloc_t;
struct nv_alloc_ops {
int (*nv_ao_init)(nv_alloc_t *, va_list);
void (*nv_ao_fini)(nv_alloc_t *);
void *(*nv_ao_alloc)(nv_alloc_t *, size_t);
void (*nv_ao_free)(nv_alloc_t *, void *, size_t);
void (*nv_ao_reset)(nv_alloc_t *);
};
-_SYS_NVPAIR_H const nv_alloc_ops_t *nv_fixed_ops;
-_SYS_NVPAIR_H nv_alloc_t *nv_alloc_nosleep;
+_SYS_NVPAIR_H const nv_alloc_ops_t *const nv_fixed_ops;
+_SYS_NVPAIR_H nv_alloc_t *const nv_alloc_nosleep;
#if defined(_KERNEL)
-_SYS_NVPAIR_H nv_alloc_t *nv_alloc_sleep;
-_SYS_NVPAIR_H nv_alloc_t *nv_alloc_pushpage;
+_SYS_NVPAIR_H nv_alloc_t *const nv_alloc_sleep;
+_SYS_NVPAIR_H nv_alloc_t *const nv_alloc_pushpage;
#endif
_SYS_NVPAIR_H int nv_alloc_init(nv_alloc_t *, const nv_alloc_ops_t *,
/* args */ ...);
_SYS_NVPAIR_H void nv_alloc_reset(nv_alloc_t *);
_SYS_NVPAIR_H void nv_alloc_fini(nv_alloc_t *);
/* list management */
_SYS_NVPAIR_H int nvlist_alloc(nvlist_t **, uint_t, int);
_SYS_NVPAIR_H void nvlist_free(nvlist_t *);
_SYS_NVPAIR_H int nvlist_size(nvlist_t *, size_t *, int);
_SYS_NVPAIR_H int nvlist_pack(nvlist_t *, char **, size_t *, int, int);
_SYS_NVPAIR_H int nvlist_unpack(char *, size_t, nvlist_t **, int);
_SYS_NVPAIR_H int nvlist_dup(const nvlist_t *, nvlist_t **, int);
_SYS_NVPAIR_H int nvlist_merge(nvlist_t *, nvlist_t *, int);
_SYS_NVPAIR_H uint_t nvlist_nvflag(nvlist_t *);
_SYS_NVPAIR_H int nvlist_xalloc(nvlist_t **, uint_t, nv_alloc_t *);
_SYS_NVPAIR_H int nvlist_xpack(nvlist_t *, char **, size_t *, int,
nv_alloc_t *);
_SYS_NVPAIR_H int nvlist_xunpack(char *, size_t, nvlist_t **, nv_alloc_t *);
_SYS_NVPAIR_H int nvlist_xdup(const nvlist_t *, nvlist_t **, nv_alloc_t *);
_SYS_NVPAIR_H nv_alloc_t *nvlist_lookup_nv_alloc(nvlist_t *);
_SYS_NVPAIR_H int nvlist_add_nvpair(nvlist_t *, nvpair_t *);
_SYS_NVPAIR_H int nvlist_add_boolean(nvlist_t *, const char *);
_SYS_NVPAIR_H int nvlist_add_boolean_value(nvlist_t *, const char *, boolean_t);
_SYS_NVPAIR_H int nvlist_add_byte(nvlist_t *, const char *, uchar_t);
_SYS_NVPAIR_H int nvlist_add_int8(nvlist_t *, const char *, int8_t);
_SYS_NVPAIR_H int nvlist_add_uint8(nvlist_t *, const char *, uint8_t);
_SYS_NVPAIR_H int nvlist_add_int16(nvlist_t *, const char *, int16_t);
_SYS_NVPAIR_H int nvlist_add_uint16(nvlist_t *, const char *, uint16_t);
_SYS_NVPAIR_H int nvlist_add_int32(nvlist_t *, const char *, int32_t);
_SYS_NVPAIR_H int nvlist_add_uint32(nvlist_t *, const char *, uint32_t);
_SYS_NVPAIR_H int nvlist_add_int64(nvlist_t *, const char *, int64_t);
_SYS_NVPAIR_H int nvlist_add_uint64(nvlist_t *, const char *, uint64_t);
_SYS_NVPAIR_H int nvlist_add_string(nvlist_t *, const char *, const char *);
_SYS_NVPAIR_H int nvlist_add_nvlist(nvlist_t *, const char *, const nvlist_t *);
_SYS_NVPAIR_H int nvlist_add_boolean_array(nvlist_t *, const char *,
const boolean_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_byte_array(nvlist_t *, const char *,
const uchar_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_int8_array(nvlist_t *, const char *,
const int8_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_uint8_array(nvlist_t *, const char *,
const uint8_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_int16_array(nvlist_t *, const char *,
const int16_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_uint16_array(nvlist_t *, const char *,
const uint16_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_int32_array(nvlist_t *, const char *,
const int32_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_uint32_array(nvlist_t *, const char *,
const uint32_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_int64_array(nvlist_t *, const char *,
const int64_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_uint64_array(nvlist_t *, const char *,
const uint64_t *, uint_t);
_SYS_NVPAIR_H int nvlist_add_string_array(nvlist_t *, const char *,
const char * const *, uint_t);
_SYS_NVPAIR_H int nvlist_add_nvlist_array(nvlist_t *, const char *,
const nvlist_t * const *, uint_t);
_SYS_NVPAIR_H int nvlist_add_hrtime(nvlist_t *, const char *, hrtime_t);
#if !defined(_KERNEL) && !defined(_STANDALONE)
_SYS_NVPAIR_H int nvlist_add_double(nvlist_t *, const char *, double);
#endif
_SYS_NVPAIR_H int nvlist_remove(nvlist_t *, const char *, data_type_t);
_SYS_NVPAIR_H int nvlist_remove_all(nvlist_t *, const char *);
_SYS_NVPAIR_H int nvlist_remove_nvpair(nvlist_t *, nvpair_t *);
_SYS_NVPAIR_H int nvlist_lookup_boolean(const nvlist_t *, const char *);
_SYS_NVPAIR_H int nvlist_lookup_boolean_value(const nvlist_t *, const char *,
boolean_t *);
_SYS_NVPAIR_H int nvlist_lookup_byte(const nvlist_t *, const char *, uchar_t *);
_SYS_NVPAIR_H int nvlist_lookup_int8(const nvlist_t *, const char *, int8_t *);
_SYS_NVPAIR_H int nvlist_lookup_uint8(const nvlist_t *, const char *,
uint8_t *);
_SYS_NVPAIR_H int nvlist_lookup_int16(const nvlist_t *, const char *,
int16_t *);
_SYS_NVPAIR_H int nvlist_lookup_uint16(const nvlist_t *, const char *,
uint16_t *);
_SYS_NVPAIR_H int nvlist_lookup_int32(const nvlist_t *, const char *,
int32_t *);
_SYS_NVPAIR_H int nvlist_lookup_uint32(const nvlist_t *, const char *,
uint32_t *);
_SYS_NVPAIR_H int nvlist_lookup_int64(const nvlist_t *, const char *,
int64_t *);
_SYS_NVPAIR_H int nvlist_lookup_uint64(const nvlist_t *, const char *,
uint64_t *);
_SYS_NVPAIR_H int nvlist_lookup_string(nvlist_t *, const char *, char **);
_SYS_NVPAIR_H int nvlist_lookup_nvlist(nvlist_t *, const char *, nvlist_t **);
_SYS_NVPAIR_H int nvlist_lookup_boolean_array(nvlist_t *, const char *,
boolean_t **, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_byte_array(nvlist_t *, const char *, uchar_t **,
uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_int8_array(nvlist_t *, const char *, int8_t **,
uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_uint8_array(nvlist_t *, const char *,
uint8_t **, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_int16_array(nvlist_t *, const char *,
int16_t **, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_uint16_array(nvlist_t *, const char *,
uint16_t **, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_int32_array(nvlist_t *, const char *,
int32_t **, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_uint32_array(nvlist_t *, const char *,
uint32_t **, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_int64_array(nvlist_t *, const char *,
int64_t **, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_uint64_array(nvlist_t *, const char *,
uint64_t **, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_string_array(nvlist_t *, const char *,
char ***, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_nvlist_array(nvlist_t *, const char *,
nvlist_t ***, uint_t *);
_SYS_NVPAIR_H int nvlist_lookup_hrtime(nvlist_t *, const char *, hrtime_t *);
_SYS_NVPAIR_H int nvlist_lookup_pairs(nvlist_t *, int, ...);
#if !defined(_KERNEL) && !defined(_STANDALONE)
_SYS_NVPAIR_H int nvlist_lookup_double(const nvlist_t *, const char *,
double *);
#endif
_SYS_NVPAIR_H int nvlist_lookup_nvpair(nvlist_t *, const char *, nvpair_t **);
_SYS_NVPAIR_H int nvlist_lookup_nvpair_embedded_index(nvlist_t *, const char *,
nvpair_t **, int *, char **);
_SYS_NVPAIR_H boolean_t nvlist_exists(const nvlist_t *, const char *);
_SYS_NVPAIR_H boolean_t nvlist_empty(const nvlist_t *);
/* processing nvpair */
_SYS_NVPAIR_H nvpair_t *nvlist_next_nvpair(nvlist_t *, const nvpair_t *);
_SYS_NVPAIR_H nvpair_t *nvlist_prev_nvpair(nvlist_t *, const nvpair_t *);
_SYS_NVPAIR_H char *nvpair_name(const nvpair_t *);
_SYS_NVPAIR_H data_type_t nvpair_type(const nvpair_t *);
_SYS_NVPAIR_H int nvpair_type_is_array(const nvpair_t *);
_SYS_NVPAIR_H int nvpair_value_boolean_value(const nvpair_t *, boolean_t *);
_SYS_NVPAIR_H int nvpair_value_byte(const nvpair_t *, uchar_t *);
_SYS_NVPAIR_H int nvpair_value_int8(const nvpair_t *, int8_t *);
_SYS_NVPAIR_H int nvpair_value_uint8(const nvpair_t *, uint8_t *);
_SYS_NVPAIR_H int nvpair_value_int16(const nvpair_t *, int16_t *);
_SYS_NVPAIR_H int nvpair_value_uint16(const nvpair_t *, uint16_t *);
_SYS_NVPAIR_H int nvpair_value_int32(const nvpair_t *, int32_t *);
_SYS_NVPAIR_H int nvpair_value_uint32(const nvpair_t *, uint32_t *);
_SYS_NVPAIR_H int nvpair_value_int64(const nvpair_t *, int64_t *);
_SYS_NVPAIR_H int nvpair_value_uint64(const nvpair_t *, uint64_t *);
_SYS_NVPAIR_H int nvpair_value_string(nvpair_t *, char **);
_SYS_NVPAIR_H int nvpair_value_nvlist(nvpair_t *, nvlist_t **);
_SYS_NVPAIR_H int nvpair_value_boolean_array(nvpair_t *, boolean_t **,
uint_t *);
_SYS_NVPAIR_H int nvpair_value_byte_array(nvpair_t *, uchar_t **, uint_t *);
_SYS_NVPAIR_H int nvpair_value_int8_array(nvpair_t *, int8_t **, uint_t *);
_SYS_NVPAIR_H int nvpair_value_uint8_array(nvpair_t *, uint8_t **, uint_t *);
_SYS_NVPAIR_H int nvpair_value_int16_array(nvpair_t *, int16_t **, uint_t *);
_SYS_NVPAIR_H int nvpair_value_uint16_array(nvpair_t *, uint16_t **, uint_t *);
_SYS_NVPAIR_H int nvpair_value_int32_array(nvpair_t *, int32_t **, uint_t *);
_SYS_NVPAIR_H int nvpair_value_uint32_array(nvpair_t *, uint32_t **, uint_t *);
_SYS_NVPAIR_H int nvpair_value_int64_array(nvpair_t *, int64_t **, uint_t *);
_SYS_NVPAIR_H int nvpair_value_uint64_array(nvpair_t *, uint64_t **, uint_t *);
_SYS_NVPAIR_H int nvpair_value_string_array(nvpair_t *, char ***, uint_t *);
_SYS_NVPAIR_H int nvpair_value_nvlist_array(nvpair_t *, nvlist_t ***, uint_t *);
_SYS_NVPAIR_H int nvpair_value_hrtime(nvpair_t *, hrtime_t *);
#if !defined(_KERNEL) && !defined(_STANDALONE)
_SYS_NVPAIR_H int nvpair_value_double(const nvpair_t *, double *);
#endif
_SYS_NVPAIR_H nvlist_t *fnvlist_alloc(void);
_SYS_NVPAIR_H void fnvlist_free(nvlist_t *);
_SYS_NVPAIR_H size_t fnvlist_size(nvlist_t *);
_SYS_NVPAIR_H char *fnvlist_pack(nvlist_t *, size_t *);
_SYS_NVPAIR_H void fnvlist_pack_free(char *, size_t);
_SYS_NVPAIR_H nvlist_t *fnvlist_unpack(char *, size_t);
_SYS_NVPAIR_H nvlist_t *fnvlist_dup(const nvlist_t *);
_SYS_NVPAIR_H void fnvlist_merge(nvlist_t *, nvlist_t *);
_SYS_NVPAIR_H size_t fnvlist_num_pairs(nvlist_t *);
_SYS_NVPAIR_H void fnvlist_add_boolean(nvlist_t *, const char *);
_SYS_NVPAIR_H void fnvlist_add_boolean_value(nvlist_t *, const char *,
boolean_t);
_SYS_NVPAIR_H void fnvlist_add_byte(nvlist_t *, const char *, uchar_t);
_SYS_NVPAIR_H void fnvlist_add_int8(nvlist_t *, const char *, int8_t);
_SYS_NVPAIR_H void fnvlist_add_uint8(nvlist_t *, const char *, uint8_t);
_SYS_NVPAIR_H void fnvlist_add_int16(nvlist_t *, const char *, int16_t);
_SYS_NVPAIR_H void fnvlist_add_uint16(nvlist_t *, const char *, uint16_t);
_SYS_NVPAIR_H void fnvlist_add_int32(nvlist_t *, const char *, int32_t);
_SYS_NVPAIR_H void fnvlist_add_uint32(nvlist_t *, const char *, uint32_t);
_SYS_NVPAIR_H void fnvlist_add_int64(nvlist_t *, const char *, int64_t);
_SYS_NVPAIR_H void fnvlist_add_uint64(nvlist_t *, const char *, uint64_t);
_SYS_NVPAIR_H void fnvlist_add_string(nvlist_t *, const char *, const char *);
_SYS_NVPAIR_H void fnvlist_add_nvlist(nvlist_t *, const char *, nvlist_t *);
_SYS_NVPAIR_H void fnvlist_add_nvpair(nvlist_t *, nvpair_t *);
_SYS_NVPAIR_H void fnvlist_add_boolean_array(nvlist_t *, const char *,
const boolean_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_byte_array(nvlist_t *, const char *,
const uchar_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_int8_array(nvlist_t *, const char *,
const int8_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_uint8_array(nvlist_t *, const char *,
const uint8_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_int16_array(nvlist_t *, const char *,
const int16_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_uint16_array(nvlist_t *, const char *,
const uint16_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_int32_array(nvlist_t *, const char *,
const int32_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_uint32_array(nvlist_t *, const char *,
const uint32_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_int64_array(nvlist_t *, const char *,
const int64_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_uint64_array(nvlist_t *, const char *,
const uint64_t *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_string_array(nvlist_t *, const char *,
const char * const *, uint_t);
_SYS_NVPAIR_H void fnvlist_add_nvlist_array(nvlist_t *, const char *,
const nvlist_t * const *, uint_t);
_SYS_NVPAIR_H void fnvlist_remove(nvlist_t *, const char *);
_SYS_NVPAIR_H void fnvlist_remove_nvpair(nvlist_t *, nvpair_t *);
_SYS_NVPAIR_H nvpair_t *fnvlist_lookup_nvpair(nvlist_t *, const char *);
_SYS_NVPAIR_H boolean_t fnvlist_lookup_boolean(const nvlist_t *, const char *);
_SYS_NVPAIR_H boolean_t fnvlist_lookup_boolean_value(const nvlist_t *,
const char *);
_SYS_NVPAIR_H uchar_t fnvlist_lookup_byte(const nvlist_t *, const char *);
_SYS_NVPAIR_H int8_t fnvlist_lookup_int8(const nvlist_t *, const char *);
_SYS_NVPAIR_H int16_t fnvlist_lookup_int16(const nvlist_t *, const char *);
_SYS_NVPAIR_H int32_t fnvlist_lookup_int32(const nvlist_t *, const char *);
_SYS_NVPAIR_H int64_t fnvlist_lookup_int64(const nvlist_t *, const char *);
_SYS_NVPAIR_H uint8_t fnvlist_lookup_uint8(const nvlist_t *, const char *);
_SYS_NVPAIR_H uint16_t fnvlist_lookup_uint16(const nvlist_t *, const char *);
_SYS_NVPAIR_H uint32_t fnvlist_lookup_uint32(const nvlist_t *, const char *);
_SYS_NVPAIR_H uint64_t fnvlist_lookup_uint64(const nvlist_t *, const char *);
_SYS_NVPAIR_H char *fnvlist_lookup_string(nvlist_t *, const char *);
_SYS_NVPAIR_H nvlist_t *fnvlist_lookup_nvlist(nvlist_t *, const char *);
_SYS_NVPAIR_H boolean_t *fnvlist_lookup_boolean_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H uchar_t *fnvlist_lookup_byte_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H int8_t *fnvlist_lookup_int8_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H uint8_t *fnvlist_lookup_uint8_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H int16_t *fnvlist_lookup_int16_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H uint16_t *fnvlist_lookup_uint16_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H int32_t *fnvlist_lookup_int32_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H uint32_t *fnvlist_lookup_uint32_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H int64_t *fnvlist_lookup_int64_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H uint64_t *fnvlist_lookup_uint64_array(nvlist_t *, const char *,
uint_t *);
_SYS_NVPAIR_H boolean_t fnvpair_value_boolean_value(const nvpair_t *nvp);
_SYS_NVPAIR_H uchar_t fnvpair_value_byte(const nvpair_t *nvp);
_SYS_NVPAIR_H int8_t fnvpair_value_int8(const nvpair_t *nvp);
_SYS_NVPAIR_H int16_t fnvpair_value_int16(const nvpair_t *nvp);
_SYS_NVPAIR_H int32_t fnvpair_value_int32(const nvpair_t *nvp);
_SYS_NVPAIR_H int64_t fnvpair_value_int64(const nvpair_t *nvp);
_SYS_NVPAIR_H uint8_t fnvpair_value_uint8(const nvpair_t *nvp);
_SYS_NVPAIR_H uint16_t fnvpair_value_uint16(const nvpair_t *nvp);
_SYS_NVPAIR_H uint32_t fnvpair_value_uint32(const nvpair_t *nvp);
_SYS_NVPAIR_H uint64_t fnvpair_value_uint64(const nvpair_t *nvp);
_SYS_NVPAIR_H char *fnvpair_value_string(nvpair_t *nvp);
_SYS_NVPAIR_H nvlist_t *fnvpair_value_nvlist(nvpair_t *nvp);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_NVPAIR_H */
diff --git a/include/sys/range_tree.h b/include/sys/range_tree.h
index fef3d4d7bd21..895d802572d8 100644
--- a/include/sys/range_tree.h
+++ b/include/sys/range_tree.h
@@ -1,330 +1,330 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright (c) 2013, 2019 by Delphix. All rights reserved.
*/
#ifndef _SYS_RANGE_TREE_H
#define _SYS_RANGE_TREE_H
#include <sys/btree.h>
#include <sys/dmu.h>
#ifdef __cplusplus
extern "C" {
#endif
#define RANGE_TREE_HISTOGRAM_SIZE 64
typedef struct range_tree_ops range_tree_ops_t;
typedef enum range_seg_type {
RANGE_SEG32,
RANGE_SEG64,
RANGE_SEG_GAP,
RANGE_SEG_NUM_TYPES,
} range_seg_type_t;
/*
* Note: the range_tree may not be accessed concurrently; consumers
* must provide external locking if required.
*/
typedef struct range_tree {
zfs_btree_t rt_root; /* offset-ordered segment b-tree */
uint64_t rt_space; /* sum of all segments in the map */
range_seg_type_t rt_type; /* type of range_seg_t in use */
/*
* All data that is stored in the range tree must have a start higher
* than or equal to rt_start, and all sizes and offsets must be
* multiples of 1 << rt_shift.
*/
uint8_t rt_shift;
uint64_t rt_start;
- range_tree_ops_t *rt_ops;
+ const range_tree_ops_t *rt_ops;
/* rt_btree_compare should only be set if rt_arg is a b-tree */
void *rt_arg;
int (*rt_btree_compare)(const void *, const void *);
uint64_t rt_gap; /* allowable inter-segment gap */
/*
* The rt_histogram maintains a histogram of ranges. Each bucket,
* rt_histogram[i], contains the number of ranges whose size is:
* 2^i <= size of range in bytes < 2^(i+1)
*/
uint64_t rt_histogram[RANGE_TREE_HISTOGRAM_SIZE];
} range_tree_t;
typedef struct range_seg32 {
uint32_t rs_start; /* starting offset of this segment */
uint32_t rs_end; /* ending offset (non-inclusive) */
} range_seg32_t;
/*
* Extremely large metaslabs, vdev-wide trees, and dnode-wide trees may
* require 64-bit integers for ranges.
*/
typedef struct range_seg64 {
uint64_t rs_start; /* starting offset of this segment */
uint64_t rs_end; /* ending offset (non-inclusive) */
} range_seg64_t;
typedef struct range_seg_gap {
uint64_t rs_start; /* starting offset of this segment */
uint64_t rs_end; /* ending offset (non-inclusive) */
uint64_t rs_fill; /* actual fill if gap mode is on */
} range_seg_gap_t;
/*
* This type needs to be the largest of the range segs, since it will be stack
* allocated and then cast the actual type to do tree operations.
*/
typedef range_seg_gap_t range_seg_max_t;
/*
* This is just for clarity of code purposes, so we can make it clear that a
* pointer is to a range seg of some type; when we need to do the actual math,
* we'll figure out the real type.
*/
typedef void range_seg_t;
struct range_tree_ops {
void (*rtop_create)(range_tree_t *rt, void *arg);
void (*rtop_destroy)(range_tree_t *rt, void *arg);
void (*rtop_add)(range_tree_t *rt, void *rs, void *arg);
void (*rtop_remove)(range_tree_t *rt, void *rs, void *arg);
void (*rtop_vacate)(range_tree_t *rt, void *arg);
};
static inline uint64_t
rs_get_start_raw(const range_seg_t *rs, const range_tree_t *rt)
{
ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
switch (rt->rt_type) {
case RANGE_SEG32:
return (((const range_seg32_t *)rs)->rs_start);
case RANGE_SEG64:
return (((const range_seg64_t *)rs)->rs_start);
case RANGE_SEG_GAP:
return (((const range_seg_gap_t *)rs)->rs_start);
default:
VERIFY(0);
return (0);
}
}
static inline uint64_t
rs_get_end_raw(const range_seg_t *rs, const range_tree_t *rt)
{
ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
switch (rt->rt_type) {
case RANGE_SEG32:
return (((const range_seg32_t *)rs)->rs_end);
case RANGE_SEG64:
return (((const range_seg64_t *)rs)->rs_end);
case RANGE_SEG_GAP:
return (((const range_seg_gap_t *)rs)->rs_end);
default:
VERIFY(0);
return (0);
}
}
static inline uint64_t
rs_get_fill_raw(const range_seg_t *rs, const range_tree_t *rt)
{
ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
switch (rt->rt_type) {
case RANGE_SEG32: {
const range_seg32_t *r32 = (const range_seg32_t *)rs;
return (r32->rs_end - r32->rs_start);
}
case RANGE_SEG64: {
const range_seg64_t *r64 = (const range_seg64_t *)rs;
return (r64->rs_end - r64->rs_start);
}
case RANGE_SEG_GAP:
return (((const range_seg_gap_t *)rs)->rs_fill);
default:
VERIFY(0);
return (0);
}
}
static inline uint64_t
rs_get_start(const range_seg_t *rs, const range_tree_t *rt)
{
return ((rs_get_start_raw(rs, rt) << rt->rt_shift) + rt->rt_start);
}
static inline uint64_t
rs_get_end(const range_seg_t *rs, const range_tree_t *rt)
{
return ((rs_get_end_raw(rs, rt) << rt->rt_shift) + rt->rt_start);
}
static inline uint64_t
rs_get_fill(const range_seg_t *rs, const range_tree_t *rt)
{
return (rs_get_fill_raw(rs, rt) << rt->rt_shift);
}
static inline void
rs_set_start_raw(range_seg_t *rs, range_tree_t *rt, uint64_t start)
{
ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
switch (rt->rt_type) {
case RANGE_SEG32:
ASSERT3U(start, <=, UINT32_MAX);
((range_seg32_t *)rs)->rs_start = (uint32_t)start;
break;
case RANGE_SEG64:
((range_seg64_t *)rs)->rs_start = start;
break;
case RANGE_SEG_GAP:
((range_seg_gap_t *)rs)->rs_start = start;
break;
default:
VERIFY(0);
}
}
static inline void
rs_set_end_raw(range_seg_t *rs, range_tree_t *rt, uint64_t end)
{
ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
switch (rt->rt_type) {
case RANGE_SEG32:
ASSERT3U(end, <=, UINT32_MAX);
((range_seg32_t *)rs)->rs_end = (uint32_t)end;
break;
case RANGE_SEG64:
((range_seg64_t *)rs)->rs_end = end;
break;
case RANGE_SEG_GAP:
((range_seg_gap_t *)rs)->rs_end = end;
break;
default:
VERIFY(0);
}
}
static inline void
rs_set_fill_raw(range_seg_t *rs, range_tree_t *rt, uint64_t fill)
{
ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
switch (rt->rt_type) {
case RANGE_SEG32:
/* fall through */
case RANGE_SEG64:
ASSERT3U(fill, ==, rs_get_end_raw(rs, rt) - rs_get_start_raw(rs,
rt));
break;
case RANGE_SEG_GAP:
((range_seg_gap_t *)rs)->rs_fill = fill;
break;
default:
VERIFY(0);
}
}
static inline void
rs_set_start(range_seg_t *rs, range_tree_t *rt, uint64_t start)
{
ASSERT3U(start, >=, rt->rt_start);
ASSERT(IS_P2ALIGNED(start, 1ULL << rt->rt_shift));
rs_set_start_raw(rs, rt, (start - rt->rt_start) >> rt->rt_shift);
}
static inline void
rs_set_end(range_seg_t *rs, range_tree_t *rt, uint64_t end)
{
ASSERT3U(end, >=, rt->rt_start);
ASSERT(IS_P2ALIGNED(end, 1ULL << rt->rt_shift));
rs_set_end_raw(rs, rt, (end - rt->rt_start) >> rt->rt_shift);
}
static inline void
rs_set_fill(range_seg_t *rs, range_tree_t *rt, uint64_t fill)
{
ASSERT(IS_P2ALIGNED(fill, 1ULL << rt->rt_shift));
rs_set_fill_raw(rs, rt, fill >> rt->rt_shift);
}
typedef void range_tree_func_t(void *arg, uint64_t start, uint64_t size);
-range_tree_t *range_tree_create_impl(range_tree_ops_t *ops,
+range_tree_t *range_tree_create_impl(const range_tree_ops_t *ops,
range_seg_type_t type, void *arg, uint64_t start, uint64_t shift,
int (*zfs_btree_compare) (const void *, const void *), uint64_t gap);
-range_tree_t *range_tree_create(range_tree_ops_t *ops, range_seg_type_t type,
- void *arg, uint64_t start, uint64_t shift);
+range_tree_t *range_tree_create(const range_tree_ops_t *ops,
+ range_seg_type_t type, void *arg, uint64_t start, uint64_t shift);
void range_tree_destroy(range_tree_t *rt);
boolean_t range_tree_contains(range_tree_t *rt, uint64_t start, uint64_t size);
range_seg_t *range_tree_find(range_tree_t *rt, uint64_t start, uint64_t size);
boolean_t range_tree_find_in(range_tree_t *rt, uint64_t start, uint64_t size,
uint64_t *ostart, uint64_t *osize);
void range_tree_verify_not_present(range_tree_t *rt,
uint64_t start, uint64_t size);
void range_tree_resize_segment(range_tree_t *rt, range_seg_t *rs,
uint64_t newstart, uint64_t newsize);
uint64_t range_tree_space(range_tree_t *rt);
uint64_t range_tree_numsegs(range_tree_t *rt);
boolean_t range_tree_is_empty(range_tree_t *rt);
void range_tree_swap(range_tree_t **rtsrc, range_tree_t **rtdst);
void range_tree_stat_verify(range_tree_t *rt);
uint64_t range_tree_min(range_tree_t *rt);
uint64_t range_tree_max(range_tree_t *rt);
uint64_t range_tree_span(range_tree_t *rt);
void range_tree_add(void *arg, uint64_t start, uint64_t size);
void range_tree_remove(void *arg, uint64_t start, uint64_t size);
void range_tree_remove_fill(range_tree_t *rt, uint64_t start, uint64_t size);
void range_tree_adjust_fill(range_tree_t *rt, range_seg_t *rs, int64_t delta);
void range_tree_clear(range_tree_t *rt, uint64_t start, uint64_t size);
void range_tree_vacate(range_tree_t *rt, range_tree_func_t *func, void *arg);
void range_tree_walk(range_tree_t *rt, range_tree_func_t *func, void *arg);
range_seg_t *range_tree_first(range_tree_t *rt);
void range_tree_remove_xor_add_segment(uint64_t start, uint64_t end,
range_tree_t *removefrom, range_tree_t *addto);
void range_tree_remove_xor_add(range_tree_t *rt, range_tree_t *removefrom,
range_tree_t *addto);
void rt_btree_create(range_tree_t *rt, void *arg);
void rt_btree_destroy(range_tree_t *rt, void *arg);
void rt_btree_add(range_tree_t *rt, range_seg_t *rs, void *arg);
void rt_btree_remove(range_tree_t *rt, range_seg_t *rs, void *arg);
void rt_btree_vacate(range_tree_t *rt, void *arg);
-extern range_tree_ops_t rt_btree_ops;
+extern const range_tree_ops_t rt_btree_ops;
#ifdef __cplusplus
}
#endif
#endif /* _SYS_RANGE_TREE_H */
diff --git a/include/sys/sa.h b/include/sys/sa.h
index 98eb8f9cd79f..32f6bd0ccd95 100644
--- a/include/sys/sa.h
+++ b/include/sys/sa.h
@@ -1,174 +1,175 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
*/
#ifndef _SYS_SA_H
#define _SYS_SA_H
#include <sys/dmu.h>
/*
* Currently available byteswap functions.
* If it all possible new attributes should used
* one of the already defined byteswap functions.
* If a new byteswap function is added then the
* ZPL/Pool version will need to be bumped.
*/
typedef enum sa_bswap_type {
SA_UINT64_ARRAY,
SA_UINT32_ARRAY,
SA_UINT16_ARRAY,
SA_UINT8_ARRAY,
SA_ACL,
} sa_bswap_type_t;
typedef uint16_t sa_attr_type_t;
/*
* Attribute to register support for.
*/
typedef struct sa_attr_reg {
char *sa_name; /* attribute name */
uint16_t sa_length;
sa_bswap_type_t sa_byteswap; /* bswap function enum */
sa_attr_type_t sa_attr; /* filled in during registration */
} sa_attr_reg_t;
typedef void (sa_data_locator_t)(void **, uint32_t *, uint32_t,
boolean_t, void *userptr);
/*
* array of attributes to store.
*
* This array should be treated as opaque/private data.
* The SA_BULK_ADD_ATTR() macro should be used for manipulating
* the array.
*
* When sa_replace_all_by_template() is used the attributes
* will be stored in the order defined in the array, except that
* the attributes may be split between the bonus and the spill buffer
*
*/
typedef struct sa_bulk_attr {
void *sa_data;
sa_data_locator_t *sa_data_func;
uint16_t sa_length;
sa_attr_type_t sa_attr;
/* the following are private to the sa framework */
void *sa_addr;
uint16_t sa_buftype;
uint16_t sa_size;
} sa_bulk_attr_t;
/*
* The on-disk format of sa_hdr_phys_t limits SA lengths to 16-bit values.
*/
#define SA_ATTR_MAX_LEN UINT16_MAX
/*
* special macro for adding entries for bulk attr support
* bulk - sa_bulk_attr_t
* count - integer that will be incremented during each add
* attr - attribute to manipulate
* func - function for accessing data.
* data - pointer to data.
* len - length of data
*/
#define SA_ADD_BULK_ATTR(b, idx, attr, func, data, len) \
{ \
ASSERT3U(len, <=, SA_ATTR_MAX_LEN); \
b[idx].sa_attr = attr;\
b[idx].sa_data_func = func; \
b[idx].sa_data = data; \
b[idx++].sa_length = len; \
}
typedef struct sa_os sa_os_t;
typedef enum sa_handle_type {
SA_HDL_SHARED,
SA_HDL_PRIVATE
} sa_handle_type_t;
struct sa_handle;
typedef void *sa_lookup_tab_t;
typedef struct sa_handle sa_handle_t;
typedef void (sa_update_cb_t)(sa_handle_t *, dmu_tx_t *tx);
int sa_handle_get(objset_t *, uint64_t, void *userp,
sa_handle_type_t, sa_handle_t **);
int sa_handle_get_from_db(objset_t *, dmu_buf_t *, void *userp,
sa_handle_type_t, sa_handle_t **);
void sa_handle_destroy(sa_handle_t *);
int sa_buf_hold(objset_t *, uint64_t, void *, dmu_buf_t **);
void sa_buf_rele(dmu_buf_t *, void *);
int sa_lookup(sa_handle_t *, sa_attr_type_t, void *buf, uint32_t buflen);
int sa_update(sa_handle_t *, sa_attr_type_t, void *buf,
uint32_t buflen, dmu_tx_t *);
int sa_remove(sa_handle_t *, sa_attr_type_t, dmu_tx_t *);
int sa_bulk_lookup(sa_handle_t *, sa_bulk_attr_t *, int count);
int sa_bulk_lookup_locked(sa_handle_t *, sa_bulk_attr_t *, int count);
int sa_bulk_update(sa_handle_t *, sa_bulk_attr_t *, int count, dmu_tx_t *);
int sa_size(sa_handle_t *, sa_attr_type_t, int *);
void sa_object_info(sa_handle_t *, dmu_object_info_t *);
void sa_object_size(sa_handle_t *, uint32_t *, u_longlong_t *);
void *sa_get_userdata(sa_handle_t *);
void sa_set_userp(sa_handle_t *, void *);
dmu_buf_t *sa_get_db(sa_handle_t *);
uint64_t sa_handle_object(sa_handle_t *);
boolean_t sa_attr_would_spill(sa_handle_t *, sa_attr_type_t, int size);
void sa_spill_rele(sa_handle_t *);
void sa_register_update_callback(objset_t *, sa_update_cb_t *);
-int sa_setup(objset_t *, uint64_t, sa_attr_reg_t *, int, sa_attr_type_t **);
+int sa_setup(objset_t *, uint64_t, const sa_attr_reg_t *, int,
+ sa_attr_type_t **);
void sa_tear_down(objset_t *);
int sa_replace_all_by_template(sa_handle_t *, sa_bulk_attr_t *,
int, dmu_tx_t *);
int sa_replace_all_by_template_locked(sa_handle_t *, sa_bulk_attr_t *,
int, dmu_tx_t *);
boolean_t sa_enabled(objset_t *);
void sa_cache_init(void);
void sa_cache_fini(void);
int sa_set_sa_object(objset_t *, uint64_t);
int sa_hdrsize(void *);
void sa_handle_lock(sa_handle_t *);
void sa_handle_unlock(sa_handle_t *);
#ifdef _KERNEL
int sa_lookup_uio(sa_handle_t *, sa_attr_type_t, zfs_uio_t *);
int sa_add_projid(sa_handle_t *, dmu_tx_t *, uint64_t);
#endif
#ifdef __cplusplus
extern "C" {
#endif
#ifdef __cplusplus
}
#endif
#endif /* _SYS_SA_H */
diff --git a/include/sys/spa.h b/include/sys/spa.h
index 2e365eabe21d..896b0f9563ae 100644
--- a/include/sys/spa.h
+++ b/include/sys/spa.h
@@ -1,1215 +1,1218 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2021 by Delphix. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright 2013 Saso Kiselkov. All rights reserved.
* Copyright (c) 2014 Integros [integros.com]
* Copyright 2017 Joyent, Inc.
* Copyright (c) 2017, 2019, Datto Inc. All rights reserved.
* Copyright (c) 2017, Intel Corporation.
* Copyright (c) 2019, Allan Jude
* Copyright (c) 2019, Klara Inc.
*/
#ifndef _SYS_SPA_H
#define _SYS_SPA_H
#include <sys/avl.h>
#include <sys/zfs_context.h>
#include <sys/kstat.h>
#include <sys/nvpair.h>
#include <sys/sysmacros.h>
#include <sys/types.h>
#include <sys/fs/zfs.h>
#include <sys/spa_checksum.h>
#include <sys/dmu.h>
#include <sys/space_map.h>
#include <sys/bitops.h>
#ifdef __cplusplus
extern "C" {
#endif
/*
* Forward references that lots of things need.
*/
typedef struct spa spa_t;
typedef struct vdev vdev_t;
typedef struct metaslab metaslab_t;
typedef struct metaslab_group metaslab_group_t;
typedef struct metaslab_class metaslab_class_t;
typedef struct zio zio_t;
typedef struct zilog zilog_t;
typedef struct spa_aux_vdev spa_aux_vdev_t;
typedef struct ddt ddt_t;
typedef struct ddt_entry ddt_entry_t;
typedef struct zbookmark_phys zbookmark_phys_t;
struct bpobj;
struct bplist;
struct dsl_pool;
struct dsl_dataset;
struct dsl_crypto_params;
/*
* Alignment Shift (ashift) is an immutable, internal top-level vdev property
* which can only be set at vdev creation time. Physical writes are always done
* according to it, which makes 2^ashift the smallest possible IO on a vdev.
*
* We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB
* (2^16 = 65,536).
*/
#define ASHIFT_MIN 9
#define ASHIFT_MAX 16
/*
* Size of block to hold the configuration data (a packed nvlist)
*/
#define SPA_CONFIG_BLOCKSIZE (1ULL << 14)
/*
* The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
* The ASIZE encoding should be at least 64 times larger (6 more bits)
* to support up to 4-way RAID-Z mirror mode with worst-case gang block
* overhead, three DVAs per bp, plus one more bit in case we do anything
* else that expands the ASIZE.
*/
#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */
#define SPA_COMPRESSBITS 7
#define SPA_VDEVBITS 24
#define SPA_COMPRESSMASK ((1U << SPA_COMPRESSBITS) - 1)
/*
* All SPA data is represented by 128-bit data virtual addresses (DVAs).
* The members of the dva_t should be considered opaque outside the SPA.
*/
typedef struct dva {
uint64_t dva_word[2];
} dva_t;
/*
* Some checksums/hashes need a 256-bit initialization salt. This salt is kept
* secret and is suitable for use in MAC algorithms as the key.
*/
typedef struct zio_cksum_salt {
uint8_t zcs_bytes[32];
} zio_cksum_salt_t;
/*
* Each block is described by its DVAs, time of birth, checksum, etc.
* The word-by-word, bit-by-bit layout of the blkptr is as follows:
*
* 64 56 48 40 32 24 16 8 0
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 0 | pad | vdev1 | GRID | ASIZE |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 1 |G| offset1 |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 2 | pad | vdev2 | GRID | ASIZE |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 3 |G| offset2 |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 4 | pad | vdev3 | GRID | ASIZE |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 5 |G| offset3 |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 7 | padding |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 8 | padding |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 9 | physical birth txg |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* a | logical birth txg |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* b | fill count |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* c | checksum[0] |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* d | checksum[1] |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* e | checksum[2] |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* f | checksum[3] |
* +-------+-------+-------+-------+-------+-------+-------+-------+
*
* Legend:
*
* vdev virtual device ID
* offset offset into virtual device
* LSIZE logical size
* PSIZE physical size (after compression)
* ASIZE allocated size (including RAID-Z parity and gang block headers)
* GRID RAID-Z layout information (reserved for future use)
* cksum checksum function
* comp compression function
* G gang block indicator
* B byteorder (endianness)
* D dedup
* X encryption
* E blkptr_t contains embedded data (see below)
* lvl level of indirection
* type DMU object type
* phys birth txg when dva[0] was written; zero if same as logical birth txg
* note that typically all the dva's would be written in this
* txg, but they could be different if they were moved by
* device removal.
* log. birth transaction group in which the block was logically born
* fill count number of non-zero blocks under this bp
* checksum[4] 256-bit checksum of the data this bp describes
*/
/*
* The blkptr_t's of encrypted blocks also need to store the encryption
* parameters so that the block can be decrypted. This layout is as follows:
*
* 64 56 48 40 32 24 16 8 0
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 0 | vdev1 | GRID | ASIZE |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 1 |G| offset1 |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 2 | vdev2 | GRID | ASIZE |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 3 |G| offset2 |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 4 | salt |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 5 | IV1 |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 7 | padding |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 8 | padding |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 9 | physical birth txg |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* a | logical birth txg |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* b | IV2 | fill count |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* c | checksum[0] |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* d | checksum[1] |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* e | MAC[0] |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* f | MAC[1] |
* +-------+-------+-------+-------+-------+-------+-------+-------+
*
* Legend:
*
* salt Salt for generating encryption keys
* IV1 First 64 bits of encryption IV
* X Block requires encryption handling (set to 1)
* E blkptr_t contains embedded data (set to 0, see below)
* fill count number of non-zero blocks under this bp (truncated to 32 bits)
* IV2 Last 32 bits of encryption IV
* checksum[2] 128-bit checksum of the data this bp describes
* MAC[2] 128-bit message authentication code for this data
*
* The X bit being set indicates that this block is one of 3 types. If this is
* a level 0 block with an encrypted object type, the block is encrypted
* (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted
* object type, this block is authenticated with an HMAC (see
* BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC
* words to store a checksum-of-MACs from the level below (see
* BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED()
* refers to both encrypted and authenticated blocks and BP_USES_CRYPT()
* refers to any of these 3 kinds of blocks.
*
* The additional encryption parameters are the salt, IV, and MAC which are
* explained in greater detail in the block comment at the top of zio_crypt.c.
* The MAC occupies half of the checksum space since it serves a very similar
* purpose: to prevent data corruption on disk. The only functional difference
* is that the checksum is used to detect on-disk corruption whether or not the
* encryption key is loaded and the MAC provides additional protection against
* malicious disk tampering. We use the 3rd DVA to store the salt and first
* 64 bits of the IV. As a result encrypted blocks can only have 2 copies
* maximum instead of the normal 3. The last 32 bits of the IV are stored in
* the upper bits of what is usually the fill count. Note that only blocks at
* level 0 or -2 are ever encrypted, which allows us to guarantee that these
* 32 bits are not trampled over by other code (see zio_crypt.c for details).
* The salt and IV are not used for authenticated bps or bps with an indirect
* MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits
* for the fill count.
*/
/*
* "Embedded" blkptr_t's don't actually point to a block, instead they
* have a data payload embedded in the blkptr_t itself. See the comment
* in blkptr.c for more details.
*
* The blkptr_t is laid out as follows:
*
* 64 56 48 40 32 24 16 8 0
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 0 | payload |
* 1 | payload |
* 2 | payload |
* 3 | payload |
* 4 | payload |
* 5 | payload |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* 7 | payload |
* 8 | payload |
* 9 | payload |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* a | logical birth txg |
* +-------+-------+-------+-------+-------+-------+-------+-------+
* b | payload |
* c | payload |
* d | payload |
* e | payload |
* f | payload |
* +-------+-------+-------+-------+-------+-------+-------+-------+
*
* Legend:
*
* payload contains the embedded data
* B (byteorder) byteorder (endianness)
* D (dedup) padding (set to zero)
* X encryption (set to zero)
* E (embedded) set to one
* lvl indirection level
* type DMU object type
* etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
* comp compression function of payload
* PSIZE size of payload after compression, in bytes
* LSIZE logical size of payload, in bytes
* note that 25 bits is enough to store the largest
* "normal" BP's LSIZE (2^16 * 2^9) in bytes
* log. birth transaction group in which the block was logically born
*
* Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
* bp's they are stored in units of SPA_MINBLOCKSHIFT.
* Generally, the generic BP_GET_*() macros can be used on embedded BP's.
* The B, D, X, lvl, type, and comp fields are stored the same as with normal
* BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
* be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
* other macros, as they assert that they are only used on BP's of the correct
* "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use
* the payload space for encryption parameters (see the comment above on
* how encryption parameters are stored).
*/
#define BPE_GET_ETYPE(bp) \
(ASSERT(BP_IS_EMBEDDED(bp)), \
BF64_GET((bp)->blk_prop, 40, 8))
#define BPE_SET_ETYPE(bp, t) do { \
ASSERT(BP_IS_EMBEDDED(bp)); \
BF64_SET((bp)->blk_prop, 40, 8, t); \
} while (0)
#define BPE_GET_LSIZE(bp) \
(ASSERT(BP_IS_EMBEDDED(bp)), \
BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
#define BPE_SET_LSIZE(bp, x) do { \
ASSERT(BP_IS_EMBEDDED(bp)); \
BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
} while (0)
#define BPE_GET_PSIZE(bp) \
(ASSERT(BP_IS_EMBEDDED(bp)), \
BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
#define BPE_SET_PSIZE(bp, x) do { \
ASSERT(BP_IS_EMBEDDED(bp)); \
BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
} while (0)
typedef enum bp_embedded_type {
BP_EMBEDDED_TYPE_DATA,
BP_EMBEDDED_TYPE_RESERVED, /* Reserved for Delphix byteswap feature. */
BP_EMBEDDED_TYPE_REDACTED,
NUM_BP_EMBEDDED_TYPES
} bp_embedded_type_t;
#define BPE_NUM_WORDS 14
#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
#define BPE_IS_PAYLOADWORD(bp, wp) \
((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)
#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
#define SPA_SYNC_MIN_VDEVS 3 /* min vdevs to update during sync */
/*
* A block is a hole when it has either 1) never been written to, or
* 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
* without physically allocating disk space. Holes are represented in the
* blkptr_t structure by zeroed blk_dva. Correct checking for holes is
* done through the BP_IS_HOLE macro. For holes, the logical size, level,
* DMU object type, and birth times are all also stored for holes that
* were written to at some point (i.e. were punched after having been filled).
*/
typedef struct blkptr {
dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
uint64_t blk_prop; /* size, compression, type, etc */
uint64_t blk_pad[2]; /* Extra space for the future */
uint64_t blk_phys_birth; /* txg when block was allocated */
uint64_t blk_birth; /* transaction group at birth */
uint64_t blk_fill; /* fill count */
zio_cksum_t blk_cksum; /* 256-bit checksum */
} blkptr_t;
/*
* Macros to get and set fields in a bp or DVA.
*/
/*
* Note, for gang blocks, DVA_GET_ASIZE() is the total space allocated for
* this gang DVA including its children BP's. The space allocated at this
* DVA's vdev/offset is vdev_gang_header_asize(vdev).
*/
#define DVA_GET_ASIZE(dva) \
BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
#define DVA_SET_ASIZE(dva, x) \
BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
SPA_MINBLOCKSHIFT, 0, x)
#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)
#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS)
#define DVA_SET_VDEV(dva, x) \
BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x)
#define DVA_GET_OFFSET(dva) \
BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
#define DVA_SET_OFFSET(dva, x) \
BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)
#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)
#define BP_GET_LSIZE(bp) \
(BP_IS_EMBEDDED(bp) ? \
(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
#define BP_SET_LSIZE(bp, x) do { \
ASSERT(!BP_IS_EMBEDDED(bp)); \
BF64_SET_SB((bp)->blk_prop, \
0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
} while (0)
#define BP_GET_PSIZE(bp) \
(BP_IS_EMBEDDED(bp) ? 0 : \
BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
#define BP_SET_PSIZE(bp, x) do { \
ASSERT(!BP_IS_EMBEDDED(bp)); \
BF64_SET_SB((bp)->blk_prop, \
16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
} while (0)
#define BP_GET_COMPRESS(bp) \
BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
#define BP_SET_COMPRESS(bp, x) \
BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)
#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
#define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x)
#define BP_GET_CHECKSUM(bp) \
(BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
BF64_GET((bp)->blk_prop, 40, 8))
#define BP_SET_CHECKSUM(bp, x) do { \
ASSERT(!BP_IS_EMBEDDED(bp)); \
BF64_SET((bp)->blk_prop, 40, 8, x); \
} while (0)
#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)
#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)
/* encrypted, authenticated, and MAC cksum bps use the same bit */
#define BP_USES_CRYPT(bp) BF64_GET((bp)->blk_prop, 61, 1)
#define BP_SET_CRYPT(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x)
#define BP_IS_ENCRYPTED(bp) \
(BP_USES_CRYPT(bp) && \
BP_GET_LEVEL(bp) <= 0 && \
DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
#define BP_IS_AUTHENTICATED(bp) \
(BP_USES_CRYPT(bp) && \
BP_GET_LEVEL(bp) <= 0 && \
!DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))
#define BP_HAS_INDIRECT_MAC_CKSUM(bp) \
(BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0)
#define BP_IS_PROTECTED(bp) \
(BP_IS_ENCRYPTED(bp) || BP_IS_AUTHENTICATED(bp))
#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)
#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)
#define BP_GET_FREE(bp) BF64_GET((bp)->blk_fill, 0, 1)
#define BP_SET_FREE(bp, x) BF64_SET((bp)->blk_fill, 0, 1, x)
#define BP_PHYSICAL_BIRTH(bp) \
(BP_IS_EMBEDDED(bp) ? 0 : \
(bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
#define BP_SET_BIRTH(bp, logical, physical) \
{ \
ASSERT(!BP_IS_EMBEDDED(bp)); \
(bp)->blk_birth = (logical); \
(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
}
#define BP_GET_FILL(bp) \
((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \
((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill))
#define BP_SET_FILL(bp, fill) \
{ \
if (BP_IS_ENCRYPTED(bp)) \
BF64_SET((bp)->blk_fill, 0, 32, fill); \
else \
(bp)->blk_fill = fill; \
}
#define BP_GET_IV2(bp) \
(ASSERT(BP_IS_ENCRYPTED(bp)), \
BF64_GET((bp)->blk_fill, 32, 32))
#define BP_SET_IV2(bp, iv2) \
{ \
ASSERT(BP_IS_ENCRYPTED(bp)); \
BF64_SET((bp)->blk_fill, 32, 32, iv2); \
}
#define BP_IS_METADATA(bp) \
(BP_GET_LEVEL(bp) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
#define BP_GET_ASIZE(bp) \
(BP_IS_EMBEDDED(bp) ? 0 : \
DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
(DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
#define BP_GET_UCSIZE(bp) \
(BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))
#define BP_GET_NDVAS(bp) \
(BP_IS_EMBEDDED(bp) ? 0 : \
!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
(!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))
#define BP_COUNT_GANG(bp) \
(BP_IS_EMBEDDED(bp) ? 0 : \
(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
DVA_GET_GANG(&(bp)->blk_dva[1]) + \
(DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))))
#define DVA_EQUAL(dva1, dva2) \
((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
(dva1)->dva_word[0] == (dva2)->dva_word[0])
#define BP_EQUAL(bp1, bp2) \
(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \
(bp1)->blk_birth == (bp2)->blk_birth && \
DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \
DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \
DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)
#define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
#define BP_IS_GANG(bp) \
(BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
(dva)->dva_word[1] == 0ULL)
#define BP_IS_HOLE(bp) \
(!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))
#define BP_SET_REDACTED(bp) \
{ \
BP_SET_EMBEDDED(bp, B_TRUE); \
BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_REDACTED); \
}
#define BP_IS_REDACTED(bp) \
(BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_REDACTED)
/* BP_IS_RAIDZ(bp) assumes no block compression */
#define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
BP_GET_PSIZE(bp))
#define BP_ZERO(bp) \
{ \
(bp)->blk_dva[0].dva_word[0] = 0; \
(bp)->blk_dva[0].dva_word[1] = 0; \
(bp)->blk_dva[1].dva_word[0] = 0; \
(bp)->blk_dva[1].dva_word[1] = 0; \
(bp)->blk_dva[2].dva_word[0] = 0; \
(bp)->blk_dva[2].dva_word[1] = 0; \
(bp)->blk_prop = 0; \
(bp)->blk_pad[0] = 0; \
(bp)->blk_pad[1] = 0; \
(bp)->blk_phys_birth = 0; \
(bp)->blk_birth = 0; \
(bp)->blk_fill = 0; \
ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
}
#ifdef _ZFS_BIG_ENDIAN
#define ZFS_HOST_BYTEORDER (0ULL)
#else
#define ZFS_HOST_BYTEORDER (1ULL)
#endif
#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)
#define BP_SPRINTF_LEN 400
/*
* This macro allows code sharing between zfs, libzpool, and mdb.
* 'func' is either snprintf() or mdb_snprintf().
* 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
*/
#define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
{ \
static const char *copyname[] = \
{ "zero", "single", "double", "triple" }; \
int len = 0; \
int copies = 0; \
const char *crypt_type; \
if (bp != NULL) { \
if (BP_IS_ENCRYPTED(bp)) { \
crypt_type = "encrypted"; \
/* LINTED E_SUSPICIOUS_COMPARISON */ \
} else if (BP_IS_AUTHENTICATED(bp)) { \
crypt_type = "authenticated"; \
} else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { \
crypt_type = "indirect-MAC"; \
} else { \
crypt_type = "unencrypted"; \
} \
} \
if (bp == NULL) { \
len += func(buf + len, size - len, "<NULL>"); \
} else if (BP_IS_HOLE(bp)) { \
len += func(buf + len, size - len, \
"HOLE [L%llu %s] " \
"size=%llxL birth=%lluL", \
(u_longlong_t)BP_GET_LEVEL(bp), \
type, \
(u_longlong_t)BP_GET_LSIZE(bp), \
(u_longlong_t)bp->blk_birth); \
} else if (BP_IS_EMBEDDED(bp)) { \
len = func(buf + len, size - len, \
"EMBEDDED [L%llu %s] et=%u %s " \
"size=%llxL/%llxP birth=%lluL", \
(u_longlong_t)BP_GET_LEVEL(bp), \
type, \
(int)BPE_GET_ETYPE(bp), \
compress, \
(u_longlong_t)BPE_GET_LSIZE(bp), \
(u_longlong_t)BPE_GET_PSIZE(bp), \
(u_longlong_t)bp->blk_birth); \
} else if (BP_IS_REDACTED(bp)) { \
len += func(buf + len, size - len, \
"REDACTED [L%llu %s] size=%llxL birth=%lluL", \
(u_longlong_t)BP_GET_LEVEL(bp), \
type, \
(u_longlong_t)BP_GET_LSIZE(bp), \
(u_longlong_t)bp->blk_birth); \
} else { \
for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \
const dva_t *dva = &bp->blk_dva[d]; \
if (DVA_IS_VALID(dva)) \
copies++; \
len += func(buf + len, size - len, \
"DVA[%d]=<%llu:%llx:%llx>%c", d, \
(u_longlong_t)DVA_GET_VDEV(dva), \
(u_longlong_t)DVA_GET_OFFSET(dva), \
(u_longlong_t)DVA_GET_ASIZE(dva), \
ws); \
} \
if (BP_IS_ENCRYPTED(bp)) { \
len += func(buf + len, size - len, \
"salt=%llx iv=%llx:%llx%c", \
(u_longlong_t)bp->blk_dva[2].dva_word[0], \
(u_longlong_t)bp->blk_dva[2].dva_word[1], \
(u_longlong_t)BP_GET_IV2(bp), \
ws); \
} \
if (BP_IS_GANG(bp) && \
DVA_GET_ASIZE(&bp->blk_dva[2]) <= \
DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \
copies--; \
len += func(buf + len, size - len, \
"[L%llu %s] %s %s %s %s %s %s %s%c" \
"size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \
"cksum=%llx:%llx:%llx:%llx", \
(u_longlong_t)BP_GET_LEVEL(bp), \
type, \
checksum, \
compress, \
crypt_type, \
BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \
BP_IS_GANG(bp) ? "gang" : "contiguous", \
BP_GET_DEDUP(bp) ? "dedup" : "unique", \
copyname[copies], \
ws, \
(u_longlong_t)BP_GET_LSIZE(bp), \
(u_longlong_t)BP_GET_PSIZE(bp), \
(u_longlong_t)bp->blk_birth, \
(u_longlong_t)BP_PHYSICAL_BIRTH(bp), \
(u_longlong_t)BP_GET_FILL(bp), \
ws, \
(u_longlong_t)bp->blk_cksum.zc_word[0], \
(u_longlong_t)bp->blk_cksum.zc_word[1], \
(u_longlong_t)bp->blk_cksum.zc_word[2], \
(u_longlong_t)bp->blk_cksum.zc_word[3]); \
} \
ASSERT(len < size); \
}
#define BP_GET_BUFC_TYPE(bp) \
(BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)
typedef enum spa_import_type {
SPA_IMPORT_EXISTING,
SPA_IMPORT_ASSEMBLE
} spa_import_type_t;
typedef enum spa_mode {
SPA_MODE_UNINIT = 0,
SPA_MODE_READ = 1,
SPA_MODE_WRITE = 2,
} spa_mode_t;
/*
* Send TRIM commands in-line during normal pool operation while deleting.
* OFF: no
* ON: yes
* NB: IN_FREEBSD_BASE is defined within the FreeBSD sources.
*/
typedef enum {
SPA_AUTOTRIM_OFF = 0, /* default */
SPA_AUTOTRIM_ON,
#ifdef IN_FREEBSD_BASE
SPA_AUTOTRIM_DEFAULT = SPA_AUTOTRIM_ON,
#else
SPA_AUTOTRIM_DEFAULT = SPA_AUTOTRIM_OFF,
#endif
} spa_autotrim_t;
/*
* Reason TRIM command was issued, used internally for accounting purposes.
*/
typedef enum trim_type {
TRIM_TYPE_MANUAL = 0,
TRIM_TYPE_AUTO = 1,
TRIM_TYPE_SIMPLE = 2
} trim_type_t;
/* state manipulation functions */
extern int spa_open(const char *pool, spa_t **, void *tag);
extern int spa_open_rewind(const char *pool, spa_t **, void *tag,
nvlist_t *policy, nvlist_t **config);
extern int spa_get_stats(const char *pool, nvlist_t **config, char *altroot,
size_t buflen);
extern int spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
nvlist_t *zplprops, struct dsl_crypto_params *dcp);
extern int spa_import(char *pool, nvlist_t *config, nvlist_t *props,
uint64_t flags);
extern nvlist_t *spa_tryimport(nvlist_t *tryconfig);
extern int spa_destroy(const char *pool);
extern int spa_checkpoint(const char *pool);
extern int spa_checkpoint_discard(const char *pool);
extern int spa_export(const char *pool, nvlist_t **oldconfig, boolean_t force,
boolean_t hardforce);
extern int spa_reset(const char *pool);
extern void spa_async_request(spa_t *spa, int flag);
extern void spa_async_unrequest(spa_t *spa, int flag);
extern void spa_async_suspend(spa_t *spa);
extern void spa_async_resume(spa_t *spa);
extern int spa_async_tasks(spa_t *spa);
extern spa_t *spa_inject_addref(char *pool);
extern void spa_inject_delref(spa_t *spa);
extern void spa_scan_stat_init(spa_t *spa);
extern int spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps);
extern int bpobj_enqueue_alloc_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx);
extern int bpobj_enqueue_free_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx);
#define SPA_ASYNC_CONFIG_UPDATE 0x01
#define SPA_ASYNC_REMOVE 0x02
#define SPA_ASYNC_PROBE 0x04
#define SPA_ASYNC_RESILVER_DONE 0x08
#define SPA_ASYNC_RESILVER 0x10
#define SPA_ASYNC_AUTOEXPAND 0x20
#define SPA_ASYNC_REMOVE_DONE 0x40
#define SPA_ASYNC_REMOVE_STOP 0x80
#define SPA_ASYNC_INITIALIZE_RESTART 0x100
#define SPA_ASYNC_TRIM_RESTART 0x200
#define SPA_ASYNC_AUTOTRIM_RESTART 0x400
#define SPA_ASYNC_L2CACHE_REBUILD 0x800
#define SPA_ASYNC_L2CACHE_TRIM 0x1000
#define SPA_ASYNC_REBUILD_DONE 0x2000
/* device manipulation */
extern int spa_vdev_add(spa_t *spa, nvlist_t *nvroot);
extern int spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot,
int replacing, int rebuild);
extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
int replace_done);
extern int spa_vdev_alloc(spa_t *spa, uint64_t guid);
extern int spa_vdev_noalloc(spa_t *spa, uint64_t guid);
extern boolean_t spa_vdev_remove_active(spa_t *spa);
extern int spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type,
nvlist_t *vdev_errlist);
extern int spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type,
uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist);
extern int spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath);
extern int spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru);
extern int spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
nvlist_t *props, boolean_t exp);
/* spare state (which is global across all pools) */
extern void spa_spare_add(vdev_t *vd);
extern void spa_spare_remove(vdev_t *vd);
extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt);
extern void spa_spare_activate(vdev_t *vd);
/* L2ARC state (which is global across all pools) */
extern void spa_l2cache_add(vdev_t *vd);
extern void spa_l2cache_remove(vdev_t *vd);
extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
extern void spa_l2cache_activate(vdev_t *vd);
extern void spa_l2cache_drop(spa_t *spa);
/* scanning */
extern int spa_scan(spa_t *spa, pool_scan_func_t func);
extern int spa_scan_stop(spa_t *spa);
extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);
/* spa syncing */
extern void spa_sync(spa_t *spa, uint64_t txg); /* only for DMU use */
extern void spa_sync_allpools(void);
extern int zfs_sync_pass_deferred_free;
/* spa namespace global mutex */
extern kmutex_t spa_namespace_lock;
/*
* SPA configuration functions in spa_config.c
*/
#define SPA_CONFIG_UPDATE_POOL 0
#define SPA_CONFIG_UPDATE_VDEVS 1
extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t);
extern void spa_config_load(void);
extern nvlist_t *spa_all_configs(uint64_t *);
extern void spa_config_set(spa_t *spa, nvlist_t *config);
extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
int getstats);
extern void spa_config_update(spa_t *spa, int what);
extern int spa_config_parse(spa_t *spa, vdev_t **vdp, nvlist_t *nv,
vdev_t *parent, uint_t id, int atype);
/*
* Miscellaneous SPA routines in spa_misc.c
*/
/* Namespace manipulation */
extern spa_t *spa_lookup(const char *name);
extern spa_t *spa_add(const char *name, nvlist_t *config, const char *altroot);
extern void spa_remove(spa_t *spa);
extern spa_t *spa_next(spa_t *prev);
/* Refcount functions */
extern void spa_open_ref(spa_t *spa, void *tag);
extern void spa_close(spa_t *spa, void *tag);
extern void spa_async_close(spa_t *spa, void *tag);
extern boolean_t spa_refcount_zero(spa_t *spa);
#define SCL_NONE 0x00
#define SCL_CONFIG 0x01
#define SCL_STATE 0x02
#define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
#define SCL_ALLOC 0x08
#define SCL_ZIO 0x10
#define SCL_FREE 0x20
#define SCL_VDEV 0x40
#define SCL_LOCKS 7
#define SCL_ALL ((1 << SCL_LOCKS) - 1)
#define SCL_STATE_ALL (SCL_STATE | SCL_L2ARC | SCL_ZIO)
/* Historical pool statistics */
typedef struct spa_history_kstat {
kmutex_t lock;
uint64_t count;
uint64_t size;
kstat_t *kstat;
void *priv;
list_t list;
} spa_history_kstat_t;
typedef struct spa_history_list {
uint64_t size;
procfs_list_t procfs_list;
} spa_history_list_t;
typedef struct spa_stats {
spa_history_list_t read_history;
spa_history_list_t txg_history;
spa_history_kstat_t tx_assign_histogram;
spa_history_list_t mmp_history;
spa_history_kstat_t state; /* pool state */
spa_history_kstat_t iostats;
} spa_stats_t;
typedef enum txg_state {
TXG_STATE_BIRTH = 0,
TXG_STATE_OPEN = 1,
TXG_STATE_QUIESCED = 2,
TXG_STATE_WAIT_FOR_SYNC = 3,
TXG_STATE_SYNCED = 4,
TXG_STATE_COMMITTED = 5,
} txg_state_t;
typedef struct txg_stat {
vdev_stat_t vs1;
vdev_stat_t vs2;
uint64_t txg;
uint64_t ndirty;
} txg_stat_t;
/* Assorted pool IO kstats */
typedef struct spa_iostats {
kstat_named_t trim_extents_written;
kstat_named_t trim_bytes_written;
kstat_named_t trim_extents_skipped;
kstat_named_t trim_bytes_skipped;
kstat_named_t trim_extents_failed;
kstat_named_t trim_bytes_failed;
kstat_named_t autotrim_extents_written;
kstat_named_t autotrim_bytes_written;
kstat_named_t autotrim_extents_skipped;
kstat_named_t autotrim_bytes_skipped;
kstat_named_t autotrim_extents_failed;
kstat_named_t autotrim_bytes_failed;
kstat_named_t simple_trim_extents_written;
kstat_named_t simple_trim_bytes_written;
kstat_named_t simple_trim_extents_skipped;
kstat_named_t simple_trim_bytes_skipped;
kstat_named_t simple_trim_extents_failed;
kstat_named_t simple_trim_bytes_failed;
} spa_iostats_t;
extern void spa_stats_init(spa_t *spa);
extern void spa_stats_destroy(spa_t *spa);
extern void spa_read_history_add(spa_t *spa, const zbookmark_phys_t *zb,
uint32_t aflags);
extern void spa_txg_history_add(spa_t *spa, uint64_t txg, hrtime_t birth_time);
extern int spa_txg_history_set(spa_t *spa, uint64_t txg,
txg_state_t completed_state, hrtime_t completed_time);
extern txg_stat_t *spa_txg_history_init_io(spa_t *, uint64_t,
struct dsl_pool *);
extern void spa_txg_history_fini_io(spa_t *, txg_stat_t *);
extern void spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs);
extern int spa_mmp_history_set_skip(spa_t *spa, uint64_t mmp_kstat_id);
extern int spa_mmp_history_set(spa_t *spa, uint64_t mmp_kstat_id, int io_error,
hrtime_t duration);
extern void spa_mmp_history_add(spa_t *spa, uint64_t txg, uint64_t timestamp,
uint64_t mmp_delay, vdev_t *vd, int label, uint64_t mmp_kstat_id,
int error);
extern void spa_iostats_trim_add(spa_t *spa, trim_type_t type,
uint64_t extents_written, uint64_t bytes_written,
uint64_t extents_skipped, uint64_t bytes_skipped,
uint64_t extents_failed, uint64_t bytes_failed);
extern void spa_import_progress_add(spa_t *spa);
extern void spa_import_progress_remove(uint64_t spa_guid);
extern int spa_import_progress_set_mmp_check(uint64_t pool_guid,
uint64_t mmp_sec_remaining);
extern int spa_import_progress_set_max_txg(uint64_t pool_guid,
uint64_t max_txg);
extern int spa_import_progress_set_state(uint64_t pool_guid,
spa_load_state_t spa_load_state);
/* Pool configuration locks */
extern int spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw);
extern void spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw);
extern void spa_config_exit(spa_t *spa, int locks, const void *tag);
extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
/* Pool vdev add/remove lock */
extern uint64_t spa_vdev_enter(spa_t *spa);
extern uint64_t spa_vdev_detach_enter(spa_t *spa, uint64_t guid);
extern uint64_t spa_vdev_config_enter(spa_t *spa);
extern void spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg,
int error, char *tag);
extern int spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error);
/* Pool vdev state change lock */
extern void spa_vdev_state_enter(spa_t *spa, int oplock);
extern int spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error);
/* Log state */
typedef enum spa_log_state {
SPA_LOG_UNKNOWN = 0, /* unknown log state */
SPA_LOG_MISSING, /* missing log(s) */
SPA_LOG_CLEAR, /* clear the log(s) */
SPA_LOG_GOOD, /* log(s) are good */
} spa_log_state_t;
extern spa_log_state_t spa_get_log_state(spa_t *spa);
extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
extern int spa_reset_logs(spa_t *spa);
/* Log claim callback */
extern void spa_claim_notify(zio_t *zio);
extern void spa_deadman(void *);
/* Accessor functions */
extern boolean_t spa_shutting_down(spa_t *spa);
extern struct dsl_pool *spa_get_dsl(spa_t *spa);
extern boolean_t spa_is_initializing(spa_t *spa);
extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa);
extern blkptr_t *spa_get_rootblkptr(spa_t *spa);
extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
extern void spa_altroot(spa_t *, char *, size_t);
extern int spa_sync_pass(spa_t *spa);
extern char *spa_name(spa_t *spa);
extern uint64_t spa_guid(spa_t *spa);
extern uint64_t spa_load_guid(spa_t *spa);
extern uint64_t spa_last_synced_txg(spa_t *spa);
extern uint64_t spa_first_txg(spa_t *spa);
extern uint64_t spa_syncing_txg(spa_t *spa);
extern uint64_t spa_final_dirty_txg(spa_t *spa);
extern uint64_t spa_version(spa_t *spa);
extern pool_state_t spa_state(spa_t *spa);
extern spa_load_state_t spa_load_state(spa_t *spa);
extern uint64_t spa_freeze_txg(spa_t *spa);
extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
extern uint64_t spa_get_dspace(spa_t *spa);
extern uint64_t spa_get_checkpoint_space(spa_t *spa);
extern uint64_t spa_get_slop_space(spa_t *spa);
extern void spa_update_dspace(spa_t *spa);
extern uint64_t spa_version(spa_t *spa);
extern boolean_t spa_deflate(spa_t *spa);
extern metaslab_class_t *spa_normal_class(spa_t *spa);
extern metaslab_class_t *spa_log_class(spa_t *spa);
extern metaslab_class_t *spa_embedded_log_class(spa_t *spa);
extern metaslab_class_t *spa_special_class(spa_t *spa);
extern metaslab_class_t *spa_dedup_class(spa_t *spa);
extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size,
dmu_object_type_t objtype, uint_t level, uint_t special_smallblk);
extern void spa_evicting_os_register(spa_t *, objset_t *os);
extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
extern void spa_evicting_os_wait(spa_t *spa);
extern int spa_max_replication(spa_t *spa);
extern int spa_prev_software_version(spa_t *spa);
extern uint64_t spa_get_failmode(spa_t *spa);
extern uint64_t spa_get_deadman_failmode(spa_t *spa);
extern void spa_set_deadman_failmode(spa_t *spa, const char *failmode);
extern boolean_t spa_suspended(spa_t *spa);
extern uint64_t spa_bootfs(spa_t *spa);
extern uint64_t spa_delegation(spa_t *spa);
extern objset_t *spa_meta_objset(spa_t *spa);
extern space_map_t *spa_syncing_log_sm(spa_t *spa);
extern uint64_t spa_deadman_synctime(spa_t *spa);
extern uint64_t spa_deadman_ziotime(spa_t *spa);
extern uint64_t spa_dirty_data(spa_t *spa);
extern spa_autotrim_t spa_get_autotrim(spa_t *spa);
/* Miscellaneous support routines */
extern void spa_load_failed(spa_t *spa, const char *fmt, ...)
__attribute__((format(printf, 2, 3)));
extern void spa_load_note(spa_t *spa, const char *fmt, ...)
__attribute__((format(printf, 2, 3)));
extern void spa_activate_mos_feature(spa_t *spa, const char *feature,
dmu_tx_t *tx);
extern void spa_deactivate_mos_feature(spa_t *spa, const char *feature);
extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
extern char *spa_strdup(const char *);
extern void spa_strfree(char *);
extern uint64_t spa_generate_guid(spa_t *spa);
extern void snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp);
extern void spa_freeze(spa_t *spa);
extern int spa_change_guid(spa_t *spa);
extern void spa_upgrade(spa_t *spa, uint64_t version);
extern void spa_evict_all(void);
extern vdev_t *spa_lookup_by_guid(spa_t *spa, uint64_t guid,
boolean_t l2cache);
extern boolean_t spa_has_l2cache(spa_t *, uint64_t guid);
extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
extern uint64_t dva_get_dsize_sync(spa_t *spa, const dva_t *dva);
extern uint64_t bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp);
extern uint64_t bp_get_dsize(spa_t *spa, const blkptr_t *bp);
extern boolean_t spa_has_slogs(spa_t *spa);
extern boolean_t spa_is_root(spa_t *spa);
extern boolean_t spa_writeable(spa_t *spa);
extern boolean_t spa_has_pending_synctask(spa_t *spa);
extern int spa_maxblocksize(spa_t *spa);
extern int spa_maxdnodesize(spa_t *spa);
extern boolean_t spa_has_checkpoint(spa_t *spa);
extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa);
extern boolean_t spa_suspend_async_destroy(spa_t *spa);
extern uint64_t spa_min_claim_txg(spa_t *spa);
extern boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva,
const blkptr_t *bp);
typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size,
void *arg);
extern boolean_t spa_remap_blkptr(spa_t *spa, blkptr_t *bp,
spa_remap_cb_t callback, void *arg);
extern uint64_t spa_get_last_removal_txg(spa_t *spa);
extern boolean_t spa_trust_config(spa_t *spa);
extern uint64_t spa_missing_tvds_allowed(spa_t *spa);
extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing);
extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa);
extern uint64_t spa_total_metaslabs(spa_t *spa);
extern boolean_t spa_multihost(spa_t *spa);
extern uint32_t spa_get_hostid(spa_t *spa);
extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *);
extern boolean_t spa_livelist_delete_check(spa_t *spa);
extern spa_mode_t spa_mode(spa_t *spa);
extern uint64_t zfs_strtonum(const char *str, char **nptr);
extern char *spa_his_ievent_table[];
extern void spa_history_create_obj(spa_t *spa, dmu_tx_t *tx);
extern int spa_history_get(spa_t *spa, uint64_t *offset, uint64_t *len_read,
char *his_buf);
extern int spa_history_log(spa_t *spa, const char *his_buf);
extern int spa_history_log_nvl(spa_t *spa, nvlist_t *nvl);
extern void spa_history_log_version(spa_t *spa, const char *operation,
dmu_tx_t *tx);
extern void spa_history_log_internal(spa_t *spa, const char *operation,
dmu_tx_t *tx, const char *fmt, ...) __printflike(4, 5);
extern void spa_history_log_internal_ds(struct dsl_dataset *ds, const char *op,
dmu_tx_t *tx, const char *fmt, ...) __printflike(4, 5);
extern void spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
dmu_tx_t *tx, const char *fmt, ...) __printflike(4, 5);
extern const char *spa_state_to_name(spa_t *spa);
/* error handling */
struct zbookmark_phys;
extern void spa_log_error(spa_t *spa, const zbookmark_phys_t *zb);
extern int zfs_ereport_post(const char *clazz, spa_t *spa, vdev_t *vd,
const zbookmark_phys_t *zb, zio_t *zio, uint64_t state);
extern boolean_t zfs_ereport_is_valid(const char *clazz, spa_t *spa, vdev_t *vd,
zio_t *zio);
extern void zfs_ereport_taskq_fini(void);
extern void zfs_ereport_clear(spa_t *spa, vdev_t *vd);
extern nvlist_t *zfs_event_create(spa_t *spa, vdev_t *vd, const char *type,
const char *name, nvlist_t *aux);
extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
extern void zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate);
extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
extern uint64_t spa_get_errlog_size(spa_t *spa);
extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
extern void spa_errlog_rotate(spa_t *spa);
extern void spa_errlog_drain(spa_t *spa);
extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
/* vdev cache */
extern void vdev_cache_stat_init(void);
extern void vdev_cache_stat_fini(void);
/* vdev mirror */
extern void vdev_mirror_stat_init(void);
extern void vdev_mirror_stat_fini(void);
/* Initialization and termination */
extern void spa_init(spa_mode_t mode);
extern void spa_fini(void);
extern void spa_boot_init(void);
/* properties */
extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
extern int spa_prop_get(spa_t *spa, nvlist_t **nvp);
extern void spa_prop_clear_bootfs(spa_t *spa, uint64_t obj, dmu_tx_t *tx);
extern void spa_configfile_set(spa_t *, nvlist_t *, boolean_t);
/* asynchronous event notification */
extern void spa_event_notify(spa_t *spa, vdev_t *vdev, nvlist_t *hist_nvl,
const char *name);
extern void zfs_ereport_zvol_post(const char *subclass, const char *name,
const char *device_name, const char *raw_name);
/* waiting for pool activities to complete */
extern int spa_wait(const char *pool, zpool_wait_activity_t activity,
boolean_t *waited);
extern int spa_wait_tag(const char *name, zpool_wait_activity_t activity,
uint64_t tag, boolean_t *waited);
extern void spa_notify_waiters(spa_t *spa);
extern void spa_wake_waiters(spa_t *spa);
/* module param call functions */
int param_set_deadman_ziotime(ZFS_MODULE_PARAM_ARGS);
int param_set_deadman_synctime(ZFS_MODULE_PARAM_ARGS);
int param_set_slop_shift(ZFS_MODULE_PARAM_ARGS);
int param_set_deadman_failmode(ZFS_MODULE_PARAM_ARGS);
#ifdef ZFS_DEBUG
#define dprintf_bp(bp, fmt, ...) do { \
if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
kmem_free(__blkbuf, BP_SPRINTF_LEN); \
} \
} while (0)
#else
#define dprintf_bp(bp, fmt, ...)
#endif
extern spa_mode_t spa_mode_global;
extern int zfs_deadman_enabled;
extern unsigned long zfs_deadman_synctime_ms;
extern unsigned long zfs_deadman_ziotime_ms;
extern unsigned long zfs_deadman_checktime_ms;
+extern kmem_cache_t *zio_buf_cache[];
+extern kmem_cache_t *zio_data_buf_cache[];
+
#ifdef __cplusplus
}
#endif
#endif /* _SYS_SPA_H */
diff --git a/include/sys/spa_impl.h b/include/sys/spa_impl.h
index eee4783fe3f8..9946c4e3c316 100644
--- a/include/sys/spa_impl.h
+++ b/include/sys/spa_impl.h
@@ -1,467 +1,467 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright 2013 Saso Kiselkov. All rights reserved.
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
* Copyright (c) 2017 Datto Inc.
* Copyright (c) 2017, Intel Corporation.
*/
#ifndef _SYS_SPA_IMPL_H
#define _SYS_SPA_IMPL_H
#include <sys/spa.h>
#include <sys/spa_checkpoint.h>
#include <sys/spa_log_spacemap.h>
#include <sys/vdev.h>
#include <sys/vdev_rebuild.h>
#include <sys/vdev_removal.h>
#include <sys/metaslab.h>
#include <sys/dmu.h>
#include <sys/dsl_pool.h>
#include <sys/uberblock_impl.h>
#include <sys/zfs_context.h>
#include <sys/avl.h>
#include <sys/zfs_refcount.h>
#include <sys/bplist.h>
#include <sys/bpobj.h>
#include <sys/dsl_crypt.h>
#include <sys/zfeature.h>
#include <sys/zthr.h>
#include <sys/dsl_deadlist.h>
#include <zfeature_common.h>
#ifdef __cplusplus
extern "C" {
#endif
typedef struct spa_alloc {
kmutex_t spaa_lock;
avl_tree_t spaa_tree;
} ____cacheline_aligned spa_alloc_t;
typedef struct spa_error_entry {
zbookmark_phys_t se_bookmark;
char *se_name;
avl_node_t se_avl;
} spa_error_entry_t;
typedef struct spa_history_phys {
uint64_t sh_pool_create_len; /* ending offset of zpool create */
uint64_t sh_phys_max_off; /* physical EOF */
uint64_t sh_bof; /* logical BOF */
uint64_t sh_eof; /* logical EOF */
uint64_t sh_records_lost; /* num of records overwritten */
} spa_history_phys_t;
/*
* All members must be uint64_t, for byteswap purposes.
*/
typedef struct spa_removing_phys {
uint64_t sr_state; /* dsl_scan_state_t */
/*
* The vdev ID that we most recently attempted to remove,
* or -1 if no removal has been attempted.
*/
uint64_t sr_removing_vdev;
/*
* The vdev ID that we most recently successfully removed,
* or -1 if no devices have been removed.
*/
uint64_t sr_prev_indirect_vdev;
uint64_t sr_start_time;
uint64_t sr_end_time;
/*
* Note that we can not use the space map's or indirect mapping's
* accounting as a substitute for these values, because we need to
* count frees of not-yet-copied data as though it did the copy.
* Otherwise, we could get into a situation where copied > to_copy,
* or we complete before copied == to_copy.
*/
uint64_t sr_to_copy; /* bytes that need to be copied */
uint64_t sr_copied; /* bytes that have been copied or freed */
} spa_removing_phys_t;
/*
* This struct is stored as an entry in the DMU_POOL_DIRECTORY_OBJECT
* (with key DMU_POOL_CONDENSING_INDIRECT). It is present if a condense
* of an indirect vdev's mapping object is in progress.
*/
typedef struct spa_condensing_indirect_phys {
/*
* The vdev ID of the indirect vdev whose indirect mapping is
* being condensed.
*/
uint64_t scip_vdev;
/*
* The vdev's old obsolete spacemap. This spacemap's contents are
* being integrated into the new mapping.
*/
uint64_t scip_prev_obsolete_sm_object;
/*
* The new mapping object that is being created.
*/
uint64_t scip_next_mapping_object;
} spa_condensing_indirect_phys_t;
struct spa_aux_vdev {
uint64_t sav_object; /* MOS object for device list */
nvlist_t *sav_config; /* cached device config */
vdev_t **sav_vdevs; /* devices */
int sav_count; /* number devices */
boolean_t sav_sync; /* sync the device list */
nvlist_t **sav_pending; /* pending device additions */
uint_t sav_npending; /* # pending devices */
};
typedef struct spa_config_lock {
kmutex_t scl_lock;
kthread_t *scl_writer;
int scl_write_wanted;
int scl_count;
kcondvar_t scl_cv;
} ____cacheline_aligned spa_config_lock_t;
typedef struct spa_config_dirent {
list_node_t scd_link;
char *scd_path;
} spa_config_dirent_t;
typedef enum zio_taskq_type {
ZIO_TASKQ_ISSUE = 0,
ZIO_TASKQ_ISSUE_HIGH,
ZIO_TASKQ_INTERRUPT,
ZIO_TASKQ_INTERRUPT_HIGH,
ZIO_TASKQ_TYPES
} zio_taskq_type_t;
/*
* State machine for the zpool-poolname process. The states transitions
* are done as follows:
*
* From To Routine
* PROC_NONE -> PROC_CREATED spa_activate()
* PROC_CREATED -> PROC_ACTIVE spa_thread()
* PROC_ACTIVE -> PROC_DEACTIVATE spa_deactivate()
* PROC_DEACTIVATE -> PROC_GONE spa_thread()
* PROC_GONE -> PROC_NONE spa_deactivate()
*/
typedef enum spa_proc_state {
SPA_PROC_NONE, /* spa_proc = &p0, no process created */
SPA_PROC_CREATED, /* spa_activate() has proc, is waiting */
SPA_PROC_ACTIVE, /* taskqs created, spa_proc set */
SPA_PROC_DEACTIVATE, /* spa_deactivate() requests process exit */
SPA_PROC_GONE /* spa_thread() is exiting, spa_proc = &p0 */
} spa_proc_state_t;
typedef struct spa_taskqs {
uint_t stqs_count;
taskq_t **stqs_taskq;
} spa_taskqs_t;
typedef enum spa_all_vdev_zap_action {
AVZ_ACTION_NONE = 0,
AVZ_ACTION_DESTROY, /* Destroy all per-vdev ZAPs and the AVZ. */
AVZ_ACTION_REBUILD, /* Populate the new AVZ, see spa_avz_rebuild */
AVZ_ACTION_INITIALIZE
} spa_avz_action_t;
typedef enum spa_config_source {
SPA_CONFIG_SRC_NONE = 0,
SPA_CONFIG_SRC_SCAN, /* scan of path (default: /dev/dsk) */
SPA_CONFIG_SRC_CACHEFILE, /* any cachefile */
SPA_CONFIG_SRC_TRYIMPORT, /* returned from call to tryimport */
SPA_CONFIG_SRC_SPLIT, /* new pool in a pool split */
SPA_CONFIG_SRC_MOS /* MOS, but not always from right txg */
} spa_config_source_t;
struct spa {
/*
* Fields protected by spa_namespace_lock.
*/
char spa_name[ZFS_MAX_DATASET_NAME_LEN]; /* pool name */
char *spa_comment; /* comment */
avl_node_t spa_avl; /* node in spa_namespace_avl */
nvlist_t *spa_config; /* last synced config */
nvlist_t *spa_config_syncing; /* currently syncing config */
nvlist_t *spa_config_splitting; /* config for splitting */
nvlist_t *spa_load_info; /* info and errors from load */
uint64_t spa_config_txg; /* txg of last config change */
int spa_sync_pass; /* iterate-to-convergence */
pool_state_t spa_state; /* pool state */
int spa_inject_ref; /* injection references */
uint8_t spa_sync_on; /* sync threads are running */
spa_load_state_t spa_load_state; /* current load operation */
boolean_t spa_indirect_vdevs_loaded; /* mappings loaded? */
boolean_t spa_trust_config; /* do we trust vdev tree? */
boolean_t spa_is_splitting; /* in the middle of a split? */
spa_config_source_t spa_config_source; /* where config comes from? */
uint64_t spa_import_flags; /* import specific flags */
spa_taskqs_t spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES];
dsl_pool_t *spa_dsl_pool;
boolean_t spa_is_initializing; /* true while opening pool */
boolean_t spa_is_exporting; /* true while exporting pool */
metaslab_class_t *spa_normal_class; /* normal data class */
metaslab_class_t *spa_log_class; /* intent log data class */
metaslab_class_t *spa_embedded_log_class; /* log on normal vdevs */
metaslab_class_t *spa_special_class; /* special allocation class */
metaslab_class_t *spa_dedup_class; /* dedup allocation class */
uint64_t spa_first_txg; /* first txg after spa_open() */
uint64_t spa_final_txg; /* txg of export/destroy */
uint64_t spa_freeze_txg; /* freeze pool at this txg */
uint64_t spa_load_max_txg; /* best initial ub_txg */
uint64_t spa_claim_max_txg; /* highest claimed birth txg */
inode_timespec_t spa_loaded_ts; /* 1st successful open time */
objset_t *spa_meta_objset; /* copy of dp->dp_meta_objset */
kmutex_t spa_evicting_os_lock; /* Evicting objset list lock */
list_t spa_evicting_os_list; /* Objsets being evicted. */
kcondvar_t spa_evicting_os_cv; /* Objset Eviction Completion */
txg_list_t spa_vdev_txg_list; /* per-txg dirty vdev list */
vdev_t *spa_root_vdev; /* top-level vdev container */
uint64_t spa_min_ashift; /* of vdevs in normal class */
uint64_t spa_max_ashift; /* of vdevs in normal class */
uint64_t spa_min_alloc; /* of vdevs in normal class */
uint64_t spa_config_guid; /* config pool guid */
uint64_t spa_load_guid; /* spa_load initialized guid */
uint64_t spa_last_synced_guid; /* last synced guid */
list_t spa_config_dirty_list; /* vdevs with dirty config */
list_t spa_state_dirty_list; /* vdevs with dirty state */
/*
* spa_allocs is an array, whose lengths is stored in spa_alloc_count.
* There is one tree and one lock for each allocator, to help improve
* allocation performance in write-heavy workloads.
*/
spa_alloc_t *spa_allocs;
int spa_alloc_count;
spa_aux_vdev_t spa_spares; /* hot spares */
spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */
nvlist_t *spa_label_features; /* Features for reading MOS */
uint64_t spa_config_object; /* MOS object for pool config */
uint64_t spa_config_generation; /* config generation number */
uint64_t spa_syncing_txg; /* txg currently syncing */
bpobj_t spa_deferred_bpobj; /* deferred-free bplist */
bplist_t spa_free_bplist[TXG_SIZE]; /* bplist of stuff to free */
zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */
/* checksum context templates */
kmutex_t spa_cksum_tmpls_lock;
void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS];
uberblock_t spa_ubsync; /* last synced uberblock */
uberblock_t spa_uberblock; /* current uberblock */
boolean_t spa_extreme_rewind; /* rewind past deferred frees */
kmutex_t spa_scrub_lock; /* resilver/scrub lock */
uint64_t spa_scrub_inflight; /* in-flight scrub bytes */
/* in-flight verification bytes */
uint64_t spa_load_verify_bytes;
kcondvar_t spa_scrub_io_cv; /* scrub I/O completion */
uint8_t spa_scrub_active; /* active or suspended? */
uint8_t spa_scrub_type; /* type of scrub we're doing */
uint8_t spa_scrub_finished; /* indicator to rotate logs */
uint8_t spa_scrub_started; /* started since last boot */
uint8_t spa_scrub_reopen; /* scrub doing vdev_reopen */
uint64_t spa_scan_pass_start; /* start time per pass/reboot */
uint64_t spa_scan_pass_scrub_pause; /* scrub pause time */
uint64_t spa_scan_pass_scrub_spent_paused; /* total paused */
uint64_t spa_scan_pass_exam; /* examined bytes per pass */
uint64_t spa_scan_pass_issued; /* issued bytes per pass */
/*
* We are in the middle of a resilver, and another resilver
* is needed once this one completes. This is set iff any
* vdev_resilver_deferred is set.
*/
boolean_t spa_resilver_deferred;
kmutex_t spa_async_lock; /* protect async state */
kthread_t *spa_async_thread; /* thread doing async task */
int spa_async_suspended; /* async tasks suspended */
kcondvar_t spa_async_cv; /* wait for thread_exit() */
uint16_t spa_async_tasks; /* async task mask */
uint64_t spa_missing_tvds; /* unopenable tvds on load */
uint64_t spa_missing_tvds_allowed; /* allow loading spa? */
uint64_t spa_nonallocating_dspace;
spa_removing_phys_t spa_removing_phys;
spa_vdev_removal_t *spa_vdev_removal;
spa_condensing_indirect_phys_t spa_condensing_indirect_phys;
spa_condensing_indirect_t *spa_condensing_indirect;
zthr_t *spa_condense_zthr; /* zthr doing condense. */
uint64_t spa_checkpoint_txg; /* the txg of the checkpoint */
spa_checkpoint_info_t spa_checkpoint_info; /* checkpoint accounting */
zthr_t *spa_checkpoint_discard_zthr;
space_map_t *spa_syncing_log_sm; /* current log space map */
avl_tree_t spa_sm_logs_by_txg;
kmutex_t spa_flushed_ms_lock; /* for metaslabs_by_flushed */
avl_tree_t spa_metaslabs_by_flushed;
spa_unflushed_stats_t spa_unflushed_stats;
list_t spa_log_summary;
uint64_t spa_log_flushall_txg;
zthr_t *spa_livelist_delete_zthr; /* deleting livelists */
zthr_t *spa_livelist_condense_zthr; /* condensing livelists */
uint64_t spa_livelists_to_delete; /* set of livelists to free */
livelist_condense_entry_t spa_to_condense; /* next to condense */
char *spa_root; /* alternate root directory */
uint64_t spa_ena; /* spa-wide ereport ENA */
int spa_last_open_failed; /* error if last open failed */
uint64_t spa_last_ubsync_txg; /* "best" uberblock txg */
uint64_t spa_last_ubsync_txg_ts; /* timestamp from that ub */
uint64_t spa_load_txg; /* ub txg that loaded */
uint64_t spa_load_txg_ts; /* timestamp from that ub */
uint64_t spa_load_meta_errors; /* verify metadata err count */
uint64_t spa_load_data_errors; /* verify data err count */
uint64_t spa_verify_min_txg; /* start txg of verify scrub */
kmutex_t spa_errlog_lock; /* error log lock */
uint64_t spa_errlog_last; /* last error log object */
uint64_t spa_errlog_scrub; /* scrub error log object */
kmutex_t spa_errlist_lock; /* error list/ereport lock */
avl_tree_t spa_errlist_last; /* last error list */
avl_tree_t spa_errlist_scrub; /* scrub error list */
uint64_t spa_deflate; /* should we deflate? */
uint64_t spa_history; /* history object */
kmutex_t spa_history_lock; /* history lock */
vdev_t *spa_pending_vdev; /* pending vdev additions */
kmutex_t spa_props_lock; /* property lock */
uint64_t spa_pool_props_object; /* object for properties */
uint64_t spa_bootfs; /* default boot filesystem */
uint64_t spa_failmode; /* failure mode for the pool */
uint64_t spa_deadman_failmode; /* failure mode for deadman */
uint64_t spa_delegation; /* delegation on/off */
list_t spa_config_list; /* previous cache file(s) */
/* per-CPU array of root of async I/O: */
zio_t **spa_async_zio_root;
zio_t *spa_suspend_zio_root; /* root of all suspended I/O */
zio_t *spa_txg_zio[TXG_SIZE]; /* spa_sync() waits for this */
kmutex_t spa_suspend_lock; /* protects suspend_zio_root */
kcondvar_t spa_suspend_cv; /* notification of resume */
zio_suspend_reason_t spa_suspended; /* pool is suspended */
uint8_t spa_claiming; /* pool is doing zil_claim() */
boolean_t spa_is_root; /* pool is root */
int spa_minref; /* num refs when first opened */
spa_mode_t spa_mode; /* SPA_MODE_{READ|WRITE} */
boolean_t spa_read_spacemaps; /* spacemaps available if ro */
spa_log_state_t spa_log_state; /* log state */
uint64_t spa_autoexpand; /* lun expansion on/off */
ddt_t *spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; /* in-core DDTs */
uint64_t spa_ddt_stat_object; /* DDT statistics */
uint64_t spa_dedup_dspace; /* Cache get_dedup_dspace() */
uint64_t spa_dedup_checksum; /* default dedup checksum */
uint64_t spa_dspace; /* dspace in normal class */
kmutex_t spa_vdev_top_lock; /* dueling offline/remove */
kmutex_t spa_proc_lock; /* protects spa_proc* */
kcondvar_t spa_proc_cv; /* spa_proc_state transitions */
spa_proc_state_t spa_proc_state; /* see definition */
proc_t *spa_proc; /* "zpool-poolname" process */
uintptr_t spa_did; /* if procp != p0, did of t1 */
boolean_t spa_autoreplace; /* autoreplace set in open */
int spa_vdev_locks; /* locks grabbed */
uint64_t spa_creation_version; /* version at pool creation */
uint64_t spa_prev_software_version; /* See ub_software_version */
uint64_t spa_feat_for_write_obj; /* required to write to pool */
uint64_t spa_feat_for_read_obj; /* required to read from pool */
uint64_t spa_feat_desc_obj; /* Feature descriptions */
uint64_t spa_feat_enabled_txg_obj; /* Feature enabled txg */
kmutex_t spa_feat_stats_lock; /* protects spa_feat_stats */
nvlist_t *spa_feat_stats; /* Cache of enabled features */
/* cache feature refcounts */
uint64_t spa_feat_refcount_cache[SPA_FEATURES];
taskqid_t spa_deadman_tqid; /* Task id */
uint64_t spa_deadman_calls; /* number of deadman calls */
hrtime_t spa_sync_starttime; /* starting time of spa_sync */
uint64_t spa_deadman_synctime; /* deadman sync expiration */
uint64_t spa_deadman_ziotime; /* deadman zio expiration */
uint64_t spa_all_vdev_zaps; /* ZAP of per-vd ZAP obj #s */
spa_avz_action_t spa_avz_action; /* destroy/rebuild AVZ? */
uint64_t spa_autotrim; /* automatic background trim? */
uint64_t spa_errata; /* errata issues detected */
spa_stats_t spa_stats; /* assorted spa statistics */
spa_keystore_t spa_keystore; /* loaded crypto keys */
/* arc_memory_throttle() parameters during low memory condition */
uint64_t spa_lowmem_page_load; /* memory load during txg */
uint64_t spa_lowmem_last_txg; /* txg window start */
hrtime_t spa_ccw_fail_time; /* Conf cache write fail time */
taskq_t *spa_zvol_taskq; /* Taskq for minor management */
taskq_t *spa_prefetch_taskq; /* Taskq for prefetch threads */
uint64_t spa_multihost; /* multihost aware (mmp) */
mmp_thread_t spa_mmp; /* multihost mmp thread */
list_t spa_leaf_list; /* list of leaf vdevs */
uint64_t spa_leaf_list_gen; /* track leaf_list changes */
uint32_t spa_hostid; /* cached system hostid */
/* synchronization for threads in spa_wait */
kmutex_t spa_activities_lock;
kcondvar_t spa_activities_cv;
kcondvar_t spa_waiters_cv;
int spa_waiters; /* number of waiting threads */
boolean_t spa_waiters_cancel; /* waiters should return */
char *spa_compatibility; /* compatibility file(s) */
/*
* spa_refcount & spa_config_lock must be the last elements
* because zfs_refcount_t changes size based on compilation options.
* In order for the MDB module to function correctly, the other
* fields must remain in the same location.
*/
spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */
zfs_refcount_t spa_refcount; /* number of opens */
taskq_t *spa_upgrade_taskq; /* taskq for upgrade jobs */
};
extern char *spa_config_path;
-extern char *zfs_deadman_failmode;
+extern const char *zfs_deadman_failmode;
extern int spa_slop_shift;
extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent);
extern void spa_taskq_dispatch_sync(spa_t *, zio_type_t t, zio_taskq_type_t q,
task_func_t *func, void *arg, uint_t flags);
extern void spa_load_spares(spa_t *spa);
extern void spa_load_l2cache(spa_t *spa);
extern sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl,
const char *name);
extern void spa_event_post(sysevent_t *ev);
extern int param_set_deadman_failmode_common(const char *val);
extern void spa_set_deadman_synctime(hrtime_t ns);
extern void spa_set_deadman_ziotime(hrtime_t ns);
extern const char *spa_history_zone(void);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_SPA_IMPL_H */
diff --git a/include/sys/vdev_impl.h b/include/sys/vdev_impl.h
index 86959725a513..1567182a7b7e 100644
--- a/include/sys/vdev_impl.h
+++ b/include/sys/vdev_impl.h
@@ -1,659 +1,657 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright (c) 2017, Intel Corporation.
*/
#ifndef _SYS_VDEV_IMPL_H
#define _SYS_VDEV_IMPL_H
#include <sys/avl.h>
#include <sys/bpobj.h>
#include <sys/dmu.h>
#include <sys/metaslab.h>
#include <sys/nvpair.h>
#include <sys/space_map.h>
#include <sys/vdev.h>
#include <sys/dkio.h>
#include <sys/uberblock_impl.h>
#include <sys/vdev_indirect_mapping.h>
#include <sys/vdev_indirect_births.h>
#include <sys/vdev_rebuild.h>
#include <sys/vdev_removal.h>
#include <sys/zfs_ratelimit.h>
#ifdef __cplusplus
extern "C" {
#endif
/*
* Virtual device descriptors.
*
* All storage pool operations go through the virtual device framework,
* which provides data replication and I/O scheduling.
*/
/*
* Forward declarations that lots of things need.
*/
typedef struct vdev_queue vdev_queue_t;
typedef struct vdev_cache vdev_cache_t;
typedef struct vdev_cache_entry vdev_cache_entry_t;
struct abd;
extern int zfs_vdev_queue_depth_pct;
extern int zfs_vdev_def_queue_depth;
extern uint32_t zfs_vdev_async_write_max_active;
/*
* Virtual device operations
*/
typedef int vdev_init_func_t(spa_t *spa, nvlist_t *nv, void **tsd);
typedef void vdev_fini_func_t(vdev_t *vd);
typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size,
uint64_t *ashift, uint64_t *pshift);
typedef void vdev_close_func_t(vdev_t *vd);
typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
typedef uint64_t vdev_min_asize_func_t(vdev_t *vd);
typedef uint64_t vdev_min_alloc_func_t(vdev_t *vd);
typedef void vdev_io_start_func_t(zio_t *zio);
typedef void vdev_io_done_func_t(zio_t *zio);
typedef void vdev_state_change_func_t(vdev_t *vd, int, int);
typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, const dva_t *dva,
size_t psize, uint64_t phys_birth);
typedef void vdev_hold_func_t(vdev_t *vd);
typedef void vdev_rele_func_t(vdev_t *vd);
typedef void vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd,
uint64_t offset, uint64_t size, void *arg);
typedef void vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size,
vdev_remap_cb_t callback, void *arg);
/*
* Given a target vdev, translates the logical range "in" to the physical
* range "res"
*/
typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg64_t *logical,
range_seg64_t *physical, range_seg64_t *remain);
typedef uint64_t vdev_rebuild_asize_func_t(vdev_t *vd, uint64_t start,
uint64_t size, uint64_t max_segment);
typedef void vdev_metaslab_init_func_t(vdev_t *vd, uint64_t *startp,
uint64_t *sizep);
typedef void vdev_config_generate_func_t(vdev_t *vd, nvlist_t *nv);
typedef uint64_t vdev_nparity_func_t(vdev_t *vd);
typedef uint64_t vdev_ndisks_func_t(vdev_t *vd);
typedef const struct vdev_ops {
vdev_init_func_t *vdev_op_init;
vdev_fini_func_t *vdev_op_fini;
vdev_open_func_t *vdev_op_open;
vdev_close_func_t *vdev_op_close;
vdev_asize_func_t *vdev_op_asize;
vdev_min_asize_func_t *vdev_op_min_asize;
vdev_min_alloc_func_t *vdev_op_min_alloc;
vdev_io_start_func_t *vdev_op_io_start;
vdev_io_done_func_t *vdev_op_io_done;
vdev_state_change_func_t *vdev_op_state_change;
vdev_need_resilver_func_t *vdev_op_need_resilver;
vdev_hold_func_t *vdev_op_hold;
vdev_rele_func_t *vdev_op_rele;
vdev_remap_func_t *vdev_op_remap;
vdev_xlation_func_t *vdev_op_xlate;
vdev_rebuild_asize_func_t *vdev_op_rebuild_asize;
vdev_metaslab_init_func_t *vdev_op_metaslab_init;
vdev_config_generate_func_t *vdev_op_config_generate;
vdev_nparity_func_t *vdev_op_nparity;
vdev_ndisks_func_t *vdev_op_ndisks;
char vdev_op_type[16];
boolean_t vdev_op_leaf;
} vdev_ops_t;
/*
* Virtual device properties
*/
struct vdev_cache_entry {
struct abd *ve_abd;
uint64_t ve_offset;
clock_t ve_lastused;
avl_node_t ve_offset_node;
avl_node_t ve_lastused_node;
uint32_t ve_hits;
uint16_t ve_missed_update;
zio_t *ve_fill_io;
};
struct vdev_cache {
avl_tree_t vc_offset_tree;
avl_tree_t vc_lastused_tree;
kmutex_t vc_lock;
};
typedef struct vdev_queue_class {
uint32_t vqc_active;
/*
* Sorted by offset or timestamp, depending on if the queue is
* LBA-ordered vs FIFO.
*/
avl_tree_t vqc_queued_tree;
} vdev_queue_class_t;
struct vdev_queue {
vdev_t *vq_vdev;
vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE];
avl_tree_t vq_active_tree;
avl_tree_t vq_read_offset_tree;
avl_tree_t vq_write_offset_tree;
avl_tree_t vq_trim_offset_tree;
uint64_t vq_last_offset;
zio_priority_t vq_last_prio; /* Last sent I/O priority. */
uint32_t vq_ia_active; /* Active interactive I/Os. */
uint32_t vq_nia_credit; /* Non-interactive I/Os credit. */
hrtime_t vq_io_complete_ts; /* time last i/o completed */
hrtime_t vq_io_delta_ts;
zio_t vq_io_search; /* used as local for stack reduction */
kmutex_t vq_lock;
};
typedef enum vdev_alloc_bias {
VDEV_BIAS_NONE,
VDEV_BIAS_LOG, /* dedicated to ZIL data (SLOG) */
VDEV_BIAS_SPECIAL, /* dedicated to ddt, metadata, and small blks */
VDEV_BIAS_DEDUP /* dedicated to dedup metadata */
} vdev_alloc_bias_t;
/*
* On-disk indirect vdev state.
*
* An indirect vdev is described exclusively in the MOS config of a pool.
* The config for an indirect vdev includes several fields, which are
* accessed in memory by a vdev_indirect_config_t.
*/
typedef struct vdev_indirect_config {
/*
* Object (in MOS) which contains the indirect mapping. This object
* contains an array of vdev_indirect_mapping_entry_phys_t ordered by
* vimep_src. The bonus buffer for this object is a
* vdev_indirect_mapping_phys_t. This object is allocated when a vdev
* removal is initiated.
*
* Note that this object can be empty if none of the data on the vdev
* has been copied yet.
*/
uint64_t vic_mapping_object;
/*
* Object (in MOS) which contains the birth times for the mapping
* entries. This object contains an array of
* vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
* buffer for this object is a vdev_indirect_birth_phys_t. This object
* is allocated when a vdev removal is initiated.
*
* Note that this object can be empty if none of the vdev has yet been
* copied.
*/
uint64_t vic_births_object;
/*
* This is the vdev ID which was removed previous to this vdev, or
* UINT64_MAX if there are no previously removed vdevs.
*/
uint64_t vic_prev_indirect_vdev;
} vdev_indirect_config_t;
/*
* Virtual device descriptor
*/
struct vdev {
/*
* Common to all vdev types.
*/
uint64_t vdev_id; /* child number in vdev parent */
uint64_t vdev_guid; /* unique ID for this vdev */
uint64_t vdev_guid_sum; /* self guid + all child guids */
uint64_t vdev_orig_guid; /* orig. guid prior to remove */
uint64_t vdev_asize; /* allocatable device capacity */
uint64_t vdev_min_asize; /* min acceptable asize */
uint64_t vdev_max_asize; /* max acceptable asize */
uint64_t vdev_ashift; /* block alignment shift */
/*
* Logical block alignment shift
*
* The smallest sized/aligned I/O supported by the device.
*/
uint64_t vdev_logical_ashift;
/*
* Physical block alignment shift
*
* The device supports logical I/Os with vdev_logical_ashift
* size/alignment, but optimum performance will be achieved by
* aligning/sizing requests to vdev_physical_ashift. Smaller
* requests may be inflated or incur device level read-modify-write
* operations.
*
* May be 0 to indicate no preference (i.e. use vdev_logical_ashift).
*/
uint64_t vdev_physical_ashift;
uint64_t vdev_state; /* see VDEV_STATE_* #defines */
uint64_t vdev_prevstate; /* used when reopening a vdev */
vdev_ops_t *vdev_ops; /* vdev operations */
spa_t *vdev_spa; /* spa for this vdev */
void *vdev_tsd; /* type-specific data */
vdev_t *vdev_top; /* top-level vdev */
vdev_t *vdev_parent; /* parent vdev */
vdev_t **vdev_child; /* array of children */
uint64_t vdev_children; /* number of children */
vdev_stat_t vdev_stat; /* virtual device statistics */
vdev_stat_ex_t vdev_stat_ex; /* extended statistics */
boolean_t vdev_expanding; /* expand the vdev? */
boolean_t vdev_reopening; /* reopen in progress? */
boolean_t vdev_nonrot; /* true if solid state */
int vdev_load_error; /* error on last load */
int vdev_open_error; /* error on last open */
int vdev_validate_error; /* error on last validate */
kthread_t *vdev_open_thread; /* thread opening children */
kthread_t *vdev_validate_thread; /* thread validating children */
uint64_t vdev_crtxg; /* txg when top-level was added */
/*
* Top-level vdev state.
*/
uint64_t vdev_ms_array; /* metaslab array object */
uint64_t vdev_ms_shift; /* metaslab size shift */
uint64_t vdev_ms_count; /* number of metaslabs */
metaslab_group_t *vdev_mg; /* metaslab group */
metaslab_group_t *vdev_log_mg; /* embedded slog metaslab group */
metaslab_t **vdev_ms; /* metaslab array */
uint64_t vdev_pending_fastwrite; /* allocated fastwrites */
txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */
txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */
txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */
boolean_t vdev_remove_wanted; /* async remove wanted? */
boolean_t vdev_probe_wanted; /* async probe wanted? */
list_node_t vdev_config_dirty_node; /* config dirty list */
list_node_t vdev_state_dirty_node; /* state dirty list */
uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */
uint64_t vdev_islog; /* is an intent log device */
uint64_t vdev_noalloc; /* device is passivated? */
uint64_t vdev_removing; /* device is being removed? */
boolean_t vdev_ishole; /* is a hole in the namespace */
uint64_t vdev_top_zap;
vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias */
/* pool checkpoint related */
space_map_t *vdev_checkpoint_sm; /* contains reserved blocks */
/* Initialize related */
boolean_t vdev_initialize_exit_wanted;
vdev_initializing_state_t vdev_initialize_state;
list_node_t vdev_initialize_node;
kthread_t *vdev_initialize_thread;
/* Protects vdev_initialize_thread and vdev_initialize_state. */
kmutex_t vdev_initialize_lock;
kcondvar_t vdev_initialize_cv;
uint64_t vdev_initialize_offset[TXG_SIZE];
uint64_t vdev_initialize_last_offset;
range_tree_t *vdev_initialize_tree; /* valid while initializing */
uint64_t vdev_initialize_bytes_est;
uint64_t vdev_initialize_bytes_done;
uint64_t vdev_initialize_action_time; /* start and end time */
/* TRIM related */
boolean_t vdev_trim_exit_wanted;
boolean_t vdev_autotrim_exit_wanted;
vdev_trim_state_t vdev_trim_state;
list_node_t vdev_trim_node;
kmutex_t vdev_autotrim_lock;
kcondvar_t vdev_autotrim_cv;
kthread_t *vdev_autotrim_thread;
/* Protects vdev_trim_thread and vdev_trim_state. */
kmutex_t vdev_trim_lock;
kcondvar_t vdev_trim_cv;
kthread_t *vdev_trim_thread;
uint64_t vdev_trim_offset[TXG_SIZE];
uint64_t vdev_trim_last_offset;
uint64_t vdev_trim_bytes_est;
uint64_t vdev_trim_bytes_done;
uint64_t vdev_trim_rate; /* requested rate (bytes/sec) */
uint64_t vdev_trim_partial; /* requested partial TRIM */
uint64_t vdev_trim_secure; /* requested secure TRIM */
uint64_t vdev_trim_action_time; /* start and end time */
/* Rebuild related */
boolean_t vdev_rebuilding;
boolean_t vdev_rebuild_exit_wanted;
boolean_t vdev_rebuild_cancel_wanted;
boolean_t vdev_rebuild_reset_wanted;
kmutex_t vdev_rebuild_lock;
kcondvar_t vdev_rebuild_cv;
kthread_t *vdev_rebuild_thread;
vdev_rebuild_t vdev_rebuild_config;
/* For limiting outstanding I/Os (initialize, TRIM) */
kmutex_t vdev_initialize_io_lock;
kcondvar_t vdev_initialize_io_cv;
uint64_t vdev_initialize_inflight;
kmutex_t vdev_trim_io_lock;
kcondvar_t vdev_trim_io_cv;
uint64_t vdev_trim_inflight[3];
/*
* Values stored in the config for an indirect or removing vdev.
*/
vdev_indirect_config_t vdev_indirect_config;
/*
* The vdev_indirect_rwlock protects the vdev_indirect_mapping
* pointer from changing on indirect vdevs (when it is condensed).
* Note that removing (not yet indirect) vdevs have different
* access patterns (the mapping is not accessed from open context,
* e.g. from zio_read) and locking strategy (e.g. svr_lock).
*/
krwlock_t vdev_indirect_rwlock;
vdev_indirect_mapping_t *vdev_indirect_mapping;
vdev_indirect_births_t *vdev_indirect_births;
/*
* In memory data structures used to manage the obsolete sm, for
* indirect or removing vdevs.
*
* The vdev_obsolete_segments is the in-core record of the segments
* that are no longer referenced anywhere in the pool (due to
* being freed or remapped and not referenced by any snapshots).
* During a sync, segments are added to vdev_obsolete_segments
* via vdev_indirect_mark_obsolete(); at the end of each sync
* pass, this is appended to vdev_obsolete_sm via
* vdev_indirect_sync_obsolete(). The vdev_obsolete_lock
* protects against concurrent modifications of vdev_obsolete_segments
* from multiple zio threads.
*/
kmutex_t vdev_obsolete_lock;
range_tree_t *vdev_obsolete_segments;
space_map_t *vdev_obsolete_sm;
/*
* Protects the vdev_scan_io_queue field itself as well as the
* structure's contents (when present).
*/
kmutex_t vdev_scan_io_queue_lock;
struct dsl_scan_io_queue *vdev_scan_io_queue;
/*
* Leaf vdev state.
*/
range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */
space_map_t *vdev_dtl_sm; /* dirty time log space map */
txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */
uint64_t vdev_dtl_object; /* DTL object */
uint64_t vdev_psize; /* physical device capacity */
uint64_t vdev_wholedisk; /* true if this is a whole disk */
uint64_t vdev_offline; /* persistent offline state */
uint64_t vdev_faulted; /* persistent faulted state */
uint64_t vdev_degraded; /* persistent degraded state */
uint64_t vdev_removed; /* persistent removed state */
uint64_t vdev_resilver_txg; /* persistent resilvering state */
uint64_t vdev_rebuild_txg; /* persistent rebuilding state */
char *vdev_path; /* vdev path (if any) */
char *vdev_devid; /* vdev devid (if any) */
char *vdev_physpath; /* vdev device path (if any) */
char *vdev_enc_sysfs_path; /* enclosure sysfs path */
char *vdev_fru; /* physical FRU location */
uint64_t vdev_not_present; /* not present during import */
uint64_t vdev_unspare; /* unspare when resilvering done */
boolean_t vdev_nowritecache; /* true if flushwritecache failed */
boolean_t vdev_has_trim; /* TRIM is supported */
boolean_t vdev_has_securetrim; /* secure TRIM is supported */
boolean_t vdev_checkremove; /* temporary online test */
boolean_t vdev_forcefault; /* force online fault */
boolean_t vdev_splitting; /* split or repair in progress */
boolean_t vdev_delayed_close; /* delayed device close? */
boolean_t vdev_tmpoffline; /* device taken offline temporarily? */
boolean_t vdev_detached; /* device detached? */
boolean_t vdev_cant_read; /* vdev is failing all reads */
boolean_t vdev_cant_write; /* vdev is failing all writes */
boolean_t vdev_isspare; /* was a hot spare */
boolean_t vdev_isl2cache; /* was a l2cache device */
boolean_t vdev_copy_uberblocks; /* post expand copy uberblocks */
boolean_t vdev_resilver_deferred; /* resilver deferred */
vdev_queue_t vdev_queue; /* I/O deadline schedule queue */
vdev_cache_t vdev_cache; /* physical block cache */
spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */
zio_t *vdev_probe_zio; /* root of current probe */
vdev_aux_t vdev_label_aux; /* on-disk aux state */
uint64_t vdev_leaf_zap;
hrtime_t vdev_mmp_pending; /* 0 if write finished */
uint64_t vdev_mmp_kstat_id; /* to find kstat entry */
uint64_t vdev_expansion_time; /* vdev's last expansion time */
list_node_t vdev_leaf_node; /* leaf vdev list */
/*
* For DTrace to work in userland (libzpool) context, these fields must
* remain at the end of the structure. DTrace will use the kernel's
* CTF definition for 'struct vdev', and since the size of a kmutex_t is
* larger in userland, the offsets for the rest of the fields would be
* incorrect.
*/
kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */
kmutex_t vdev_stat_lock; /* vdev_stat */
kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */
/*
* We rate limit ZIO delay, deadman, and checksum events, since they
* can flood ZED with tons of events when a drive is acting up.
*/
zfs_ratelimit_t vdev_delay_rl;
zfs_ratelimit_t vdev_deadman_rl;
zfs_ratelimit_t vdev_checksum_rl;
};
#define VDEV_PAD_SIZE (8 << 10)
/* 2 padding areas (vl_pad1 and vl_be) to skip */
#define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
#define VDEV_PHYS_SIZE (112 << 10)
#define VDEV_UBERBLOCK_RING (128 << 10)
/*
* MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
* ring when MMP is enabled.
*/
#define MMP_BLOCKS_PER_LABEL 1
/* The largest uberblock we support is 8k. */
#define MAX_UBERBLOCK_SHIFT (13)
#define VDEV_UBERBLOCK_SHIFT(vd) \
MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
MAX_UBERBLOCK_SHIFT)
#define VDEV_UBERBLOCK_COUNT(vd) \
(VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
#define VDEV_UBERBLOCK_OFFSET(vd, n) \
offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
#define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
typedef struct vdev_phys {
char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
zio_eck_t vp_zbt;
} vdev_phys_t;
typedef enum vbe_vers {
/*
* The bootenv file is stored as ascii text in the envblock.
* It is used by the GRUB bootloader used on Linux to store the
* contents of the grubenv file. The file is stored as raw ASCII,
* and is protected by an embedded checksum. By default, GRUB will
* check if the boot filesystem supports storing the environment data
* in a special location, and if so, will invoke filesystem specific
* logic to retrieve it. This can be overridden by a variable, should
* the user so desire.
*/
VB_RAW = 0,
/*
* The bootenv file is converted to an nvlist and then packed into the
* envblock.
*/
VB_NVLIST = 1
} vbe_vers_t;
typedef struct vdev_boot_envblock {
uint64_t vbe_version;
char vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) -
sizeof (zio_eck_t)];
zio_eck_t vbe_zbt;
} vdev_boot_envblock_t;
CTASSERT_GLOBAL(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE);
typedef struct vdev_label {
char vl_pad1[VDEV_PAD_SIZE]; /* 8K */
vdev_boot_envblock_t vl_be; /* 8K */
vdev_phys_t vl_vdev_phys; /* 112K */
char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
} vdev_label_t; /* 256K total */
/*
* vdev_dirty() flags
*/
#define VDD_METASLAB 0x01
#define VDD_DTL 0x02
/* Offset of embedded boot loader region on each label */
#define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
/*
* Size of embedded boot loader region on each label.
* The total size of the first two labels plus the boot area is 4MB.
*/
#define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
/*
* Size of label regions at the start and end of each leaf device.
*/
#define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
#define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
#define VDEV_LABELS 4
#define VDEV_BEST_LABEL VDEV_LABELS
#define VDEV_OFFSET_IS_LABEL(vd, off) \
(((off) < VDEV_LABEL_START_SIZE) || \
((off) >= ((vd)->vdev_psize - VDEV_LABEL_END_SIZE)))
#define VDEV_ALLOC_LOAD 0
#define VDEV_ALLOC_ADD 1
#define VDEV_ALLOC_SPARE 2
#define VDEV_ALLOC_L2CACHE 3
#define VDEV_ALLOC_ROOTPOOL 4
#define VDEV_ALLOC_SPLIT 5
#define VDEV_ALLOC_ATTACH 6
/*
* Allocate or free a vdev
*/
extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid,
vdev_ops_t *ops);
extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
vdev_t *parent, uint_t id, int alloctype);
extern void vdev_free(vdev_t *vd);
/*
* Add or remove children and parents
*/
extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
extern void vdev_compact_children(vdev_t *pvd);
extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
extern void vdev_remove_parent(vdev_t *cvd);
/*
* vdev sync load and sync
*/
extern boolean_t vdev_log_state_valid(vdev_t *vd);
extern int vdev_load(vdev_t *vd);
extern int vdev_dtl_load(vdev_t *vd);
extern void vdev_sync(vdev_t *vd, uint64_t txg);
extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);
/*
* Available vdev types.
*/
extern vdev_ops_t vdev_root_ops;
extern vdev_ops_t vdev_mirror_ops;
extern vdev_ops_t vdev_replacing_ops;
extern vdev_ops_t vdev_raidz_ops;
extern vdev_ops_t vdev_draid_ops;
extern vdev_ops_t vdev_draid_spare_ops;
extern vdev_ops_t vdev_disk_ops;
extern vdev_ops_t vdev_file_ops;
extern vdev_ops_t vdev_missing_ops;
extern vdev_ops_t vdev_hole_ops;
extern vdev_ops_t vdev_spare_ops;
extern vdev_ops_t vdev_indirect_ops;
/*
* Common size functions
*/
extern void vdev_default_xlate(vdev_t *vd, const range_seg64_t *logical_rs,
range_seg64_t *physical_rs, range_seg64_t *remain_rs);
extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
extern uint64_t vdev_default_min_asize(vdev_t *vd);
extern uint64_t vdev_get_min_asize(vdev_t *vd);
extern void vdev_set_min_asize(vdev_t *vd);
extern uint64_t vdev_get_min_alloc(vdev_t *vd);
extern uint64_t vdev_get_nparity(vdev_t *vd);
extern uint64_t vdev_get_ndisks(vdev_t *vd);
/*
* Global variables
*/
extern int zfs_vdev_standard_sm_blksz;
-/* zdb uses this tunable, so it must be declared here to make lint happy. */
-extern int zfs_vdev_cache_size;
/*
* Functions from vdev_indirect.c
*/
extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx);
extern boolean_t vdev_indirect_should_condense(vdev_t *vd);
extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx);
extern int vdev_obsolete_sm_object(vdev_t *vd, uint64_t *sm_obj);
extern int vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise);
/*
* Other miscellaneous functions
*/
int vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj);
void vdev_metaslab_group_create(vdev_t *vd);
/*
* Vdev ashift optimization tunables
*/
extern uint64_t zfs_vdev_min_auto_ashift;
extern uint64_t zfs_vdev_max_auto_ashift;
int param_set_min_auto_ashift(ZFS_MODULE_PARAM_ARGS);
int param_set_max_auto_ashift(ZFS_MODULE_PARAM_ARGS);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_VDEV_IMPL_H */
diff --git a/include/sys/vdev_raidz_impl.h b/include/sys/vdev_raidz_impl.h
index 908723da0c2a..890e725e18d8 100644
--- a/include/sys/vdev_raidz_impl.h
+++ b/include/sys/vdev_raidz_impl.h
@@ -1,389 +1,389 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (C) 2016 Gvozden Nešković. All rights reserved.
*/
#ifndef _VDEV_RAIDZ_H
#define _VDEV_RAIDZ_H
#include <sys/types.h>
#include <sys/debug.h>
#include <sys/kstat.h>
#include <sys/abd.h>
#include <sys/vdev_impl.h>
#ifdef __cplusplus
extern "C" {
#endif
#define CODE_P (0U)
#define CODE_Q (1U)
#define CODE_R (2U)
#define PARITY_P (1U)
#define PARITY_PQ (2U)
#define PARITY_PQR (3U)
#define TARGET_X (0U)
#define TARGET_Y (1U)
#define TARGET_Z (2U)
/*
* Parity generation methods indexes
*/
enum raidz_math_gen_op {
RAIDZ_GEN_P = 0,
RAIDZ_GEN_PQ,
RAIDZ_GEN_PQR,
RAIDZ_GEN_NUM = 3
};
/*
* Data reconstruction methods indexes
*/
enum raidz_rec_op {
RAIDZ_REC_P = 0,
RAIDZ_REC_Q,
RAIDZ_REC_R,
RAIDZ_REC_PQ,
RAIDZ_REC_PR,
RAIDZ_REC_QR,
RAIDZ_REC_PQR,
RAIDZ_REC_NUM = 7
};
-extern const char *raidz_gen_name[RAIDZ_GEN_NUM];
-extern const char *raidz_rec_name[RAIDZ_REC_NUM];
+extern const char *const raidz_gen_name[RAIDZ_GEN_NUM];
+extern const char *const raidz_rec_name[RAIDZ_REC_NUM];
/*
* Methods used to define raidz implementation
*
* @raidz_gen_f Parity generation function
* @par1 pointer to raidz_map
* @raidz_rec_f Data reconstruction function
* @par1 pointer to raidz_map
* @par2 array of reconstruction targets
* @will_work_f Function returns TRUE if impl. is supported on the system
* @init_impl_f Function is called once on init
* @fini_impl_f Function is called once on fini
*/
typedef void (*raidz_gen_f)(void *);
typedef int (*raidz_rec_f)(void *, const int *);
typedef boolean_t (*will_work_f)(void);
typedef void (*init_impl_f)(void);
typedef void (*fini_impl_f)(void);
#define RAIDZ_IMPL_NAME_MAX (20)
typedef struct raidz_impl_ops {
init_impl_f init;
fini_impl_f fini;
raidz_gen_f gen[RAIDZ_GEN_NUM]; /* Parity generate functions */
raidz_rec_f rec[RAIDZ_REC_NUM]; /* Data reconstruction functions */
will_work_f is_supported; /* Support check function */
char name[RAIDZ_IMPL_NAME_MAX]; /* Name of the implementation */
} raidz_impl_ops_t;
typedef struct raidz_col {
uint64_t rc_devidx; /* child device index for I/O */
uint64_t rc_offset; /* device offset */
uint64_t rc_size; /* I/O size */
abd_t rc_abdstruct; /* rc_abd probably points here */
abd_t *rc_abd; /* I/O data */
abd_t *rc_orig_data; /* pre-reconstruction */
int rc_error; /* I/O error for this device */
uint8_t rc_tried; /* Did we attempt this I/O column? */
uint8_t rc_skipped; /* Did we skip this I/O column? */
uint8_t rc_need_orig_restore; /* need to restore from orig_data? */
uint8_t rc_force_repair; /* Write good data to this column */
uint8_t rc_allow_repair; /* Allow repair I/O to this column */
} raidz_col_t;
typedef struct raidz_row {
uint64_t rr_cols; /* Regular column count */
uint64_t rr_scols; /* Count including skipped columns */
uint64_t rr_bigcols; /* Remainder data column count */
uint64_t rr_missingdata; /* Count of missing data devices */
uint64_t rr_missingparity; /* Count of missing parity devices */
uint64_t rr_firstdatacol; /* First data column/parity count */
abd_t *rr_abd_empty; /* dRAID empty sector buffer */
int rr_nempty; /* empty sectors included in parity */
#ifdef ZFS_DEBUG
uint64_t rr_offset; /* Logical offset for *_io_verify() */
uint64_t rr_size; /* Physical size for *_io_verify() */
#endif
raidz_col_t rr_col[0]; /* Flexible array of I/O columns */
} raidz_row_t;
typedef struct raidz_map {
boolean_t rm_ecksuminjected; /* checksum error was injected */
int rm_nrows; /* Regular row count */
int rm_nskip; /* RAIDZ sectors skipped for padding */
int rm_skipstart; /* Column index of padding start */
const raidz_impl_ops_t *rm_ops; /* RAIDZ math operations */
raidz_row_t *rm_row[0]; /* flexible array of rows */
} raidz_map_t;
#define RAIDZ_ORIGINAL_IMPL (INT_MAX)
extern const raidz_impl_ops_t vdev_raidz_scalar_impl;
extern boolean_t raidz_will_scalar_work(void);
#if defined(__x86_64) && defined(HAVE_SSE2) /* only x86_64 for now */
extern const raidz_impl_ops_t vdev_raidz_sse2_impl;
#endif
#if defined(__x86_64) && defined(HAVE_SSSE3) /* only x86_64 for now */
extern const raidz_impl_ops_t vdev_raidz_ssse3_impl;
#endif
#if defined(__x86_64) && defined(HAVE_AVX2) /* only x86_64 for now */
extern const raidz_impl_ops_t vdev_raidz_avx2_impl;
#endif
#if defined(__x86_64) && defined(HAVE_AVX512F) /* only x86_64 for now */
extern const raidz_impl_ops_t vdev_raidz_avx512f_impl;
#endif
#if defined(__x86_64) && defined(HAVE_AVX512BW) /* only x86_64 for now */
extern const raidz_impl_ops_t vdev_raidz_avx512bw_impl;
#endif
#if defined(__aarch64__)
extern const raidz_impl_ops_t vdev_raidz_aarch64_neon_impl;
extern const raidz_impl_ops_t vdev_raidz_aarch64_neonx2_impl;
#endif
#if defined(__powerpc__)
extern const raidz_impl_ops_t vdev_raidz_powerpc_altivec_impl;
#endif
/*
* Commonly used raidz_map helpers
*
* raidz_parity Returns parity of the RAIDZ block
* raidz_ncols Returns number of columns the block spans
* Note, all rows have the same number of columns.
* raidz_nbigcols Returns number of big columns
* raidz_col_p Returns pointer to a column
* raidz_col_size Returns size of a column
* raidz_big_size Returns size of big columns
* raidz_short_size Returns size of short columns
*/
#define raidz_parity(rm) ((rm)->rm_row[0]->rr_firstdatacol)
#define raidz_ncols(rm) ((rm)->rm_row[0]->rr_cols)
#define raidz_nbigcols(rm) ((rm)->rm_bigcols)
#define raidz_col_p(rm, c) ((rm)->rm_col + (c))
#define raidz_col_size(rm, c) ((rm)->rm_col[c].rc_size)
#define raidz_big_size(rm) (raidz_col_size(rm, CODE_P))
#define raidz_short_size(rm) (raidz_col_size(rm, raidz_ncols(rm)-1))
/*
* Macro defines an RAIDZ parity generation method
*
* @code parity the function produce
* @impl name of the implementation
*/
#define _RAIDZ_GEN_WRAP(code, impl) \
static void \
impl ## _gen_ ## code(void *rrp) \
{ \
raidz_row_t *rr = (raidz_row_t *)rrp; \
raidz_generate_## code ## _impl(rr); \
}
/*
* Macro defines an RAIDZ data reconstruction method
*
* @code parity the function produce
* @impl name of the implementation
*/
#define _RAIDZ_REC_WRAP(code, impl) \
static int \
impl ## _rec_ ## code(void *rrp, const int *tgtidx) \
{ \
raidz_row_t *rr = (raidz_row_t *)rrp; \
return (raidz_reconstruct_## code ## _impl(rr, tgtidx)); \
}
/*
* Define all gen methods for an implementation
*
* @impl name of the implementation
*/
#define DEFINE_GEN_METHODS(impl) \
_RAIDZ_GEN_WRAP(p, impl); \
_RAIDZ_GEN_WRAP(pq, impl); \
_RAIDZ_GEN_WRAP(pqr, impl)
/*
* Define all rec functions for an implementation
*
* @impl name of the implementation
*/
#define DEFINE_REC_METHODS(impl) \
_RAIDZ_REC_WRAP(p, impl); \
_RAIDZ_REC_WRAP(q, impl); \
_RAIDZ_REC_WRAP(r, impl); \
_RAIDZ_REC_WRAP(pq, impl); \
_RAIDZ_REC_WRAP(pr, impl); \
_RAIDZ_REC_WRAP(qr, impl); \
_RAIDZ_REC_WRAP(pqr, impl)
#define RAIDZ_GEN_METHODS(impl) \
{ \
[RAIDZ_GEN_P] = & impl ## _gen_p, \
[RAIDZ_GEN_PQ] = & impl ## _gen_pq, \
[RAIDZ_GEN_PQR] = & impl ## _gen_pqr \
}
#define RAIDZ_REC_METHODS(impl) \
{ \
[RAIDZ_REC_P] = & impl ## _rec_p, \
[RAIDZ_REC_Q] = & impl ## _rec_q, \
[RAIDZ_REC_R] = & impl ## _rec_r, \
[RAIDZ_REC_PQ] = & impl ## _rec_pq, \
[RAIDZ_REC_PR] = & impl ## _rec_pr, \
[RAIDZ_REC_QR] = & impl ## _rec_qr, \
[RAIDZ_REC_PQR] = & impl ## _rec_pqr \
}
typedef struct raidz_impl_kstat {
uint64_t gen[RAIDZ_GEN_NUM]; /* gen method speed B/s */
uint64_t rec[RAIDZ_REC_NUM]; /* rec method speed B/s */
} raidz_impl_kstat_t;
/*
* Enumerate various multiplication constants
* used in reconstruction methods
*/
typedef enum raidz_mul_info {
/* Reconstruct Q */
MUL_Q_X = 0,
/* Reconstruct R */
MUL_R_X = 0,
/* Reconstruct PQ */
MUL_PQ_X = 0,
MUL_PQ_Y = 1,
/* Reconstruct PR */
MUL_PR_X = 0,
MUL_PR_Y = 1,
/* Reconstruct QR */
MUL_QR_XQ = 0,
MUL_QR_X = 1,
MUL_QR_YQ = 2,
MUL_QR_Y = 3,
/* Reconstruct PQR */
MUL_PQR_XP = 0,
MUL_PQR_XQ = 1,
MUL_PQR_XR = 2,
MUL_PQR_YU = 3,
MUL_PQR_YP = 4,
MUL_PQR_YQ = 5,
MUL_CNT = 6
} raidz_mul_info_t;
/*
* Powers of 2 in the Galois field.
*/
extern const uint8_t vdev_raidz_pow2[256] __attribute__((aligned(256)));
/* Logs of 2 in the Galois field defined above. */
extern const uint8_t vdev_raidz_log2[256] __attribute__((aligned(256)));
/*
* Multiply a given number by 2 raised to the given power.
*/
static inline uint8_t
vdev_raidz_exp2(const uint8_t a, const unsigned exp)
{
if (a == 0)
return (0);
return (vdev_raidz_pow2[(exp + (unsigned)vdev_raidz_log2[a]) % 255]);
}
/*
* Galois Field operations.
*
* gf_exp2 - computes 2 raised to the given power
* gf_exp2 - computes 4 raised to the given power
* gf_mul - multiplication
* gf_div - division
* gf_inv - multiplicative inverse
*/
typedef unsigned gf_t;
typedef unsigned gf_log_t;
static inline gf_t
gf_mul(const gf_t a, const gf_t b)
{
gf_log_t logsum;
if (a == 0 || b == 0)
return (0);
logsum = (gf_log_t)vdev_raidz_log2[a] + (gf_log_t)vdev_raidz_log2[b];
return ((gf_t)vdev_raidz_pow2[logsum % 255]);
}
static inline gf_t
gf_div(const gf_t a, const gf_t b)
{
gf_log_t logsum;
ASSERT3U(b, >, 0);
if (a == 0)
return (0);
logsum = (gf_log_t)255 + (gf_log_t)vdev_raidz_log2[a] -
(gf_log_t)vdev_raidz_log2[b];
return ((gf_t)vdev_raidz_pow2[logsum % 255]);
}
static inline gf_t
gf_inv(const gf_t a)
{
gf_log_t logsum;
ASSERT3U(a, >, 0);
logsum = (gf_log_t)255 - (gf_log_t)vdev_raidz_log2[a];
return ((gf_t)vdev_raidz_pow2[logsum]);
}
static inline gf_t
gf_exp2(gf_log_t exp)
{
return (vdev_raidz_pow2[exp % 255]);
}
static inline gf_t
gf_exp4(gf_log_t exp)
{
ASSERT3U(exp, <=, 255);
return ((gf_t)vdev_raidz_pow2[(2 * exp) % 255]);
}
#ifdef __cplusplus
}
#endif
#endif /* _VDEV_RAIDZ_H */
diff --git a/include/sys/zfs_acl.h b/include/sys/zfs_acl.h
index 010686a9121b..98387a49adbe 100644
--- a/include/sys/zfs_acl.h
+++ b/include/sys/zfs_acl.h
@@ -1,248 +1,248 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
*/
#ifndef _SYS_FS_ZFS_ACL_H
#define _SYS_FS_ZFS_ACL_H
#ifdef _KERNEL
#include <sys/isa_defs.h>
#include <sys/types32.h>
#include <sys/xvattr.h>
#endif
#include <sys/acl.h>
#include <sys/dmu.h>
#include <sys/zfs_fuid.h>
#include <sys/sa.h>
#ifdef __cplusplus
extern "C" {
#endif
struct znode_phys;
#define ACE_SLOT_CNT 6
#define ZFS_ACL_VERSION_INITIAL 0ULL
#define ZFS_ACL_VERSION_FUID 1ULL
#define ZFS_ACL_VERSION ZFS_ACL_VERSION_FUID
/*
* ZFS ACLs (Access Control Lists) are stored in various forms.
*
* Files created with ACL version ZFS_ACL_VERSION_INITIAL
* will all be created with fixed length ACEs of type
* zfs_oldace_t.
*
* Files with ACL version ZFS_ACL_VERSION_FUID will be created
* with various sized ACEs. The abstraction entries will utilize
* zfs_ace_hdr_t, normal user/group entries will use zfs_ace_t
* and some specialized CIFS ACEs will use zfs_object_ace_t.
*/
/*
* All ACEs have a common hdr. For
* owner@, group@, and everyone@ this is all
* that's needed.
*/
typedef struct zfs_ace_hdr {
uint16_t z_type;
uint16_t z_flags;
uint32_t z_access_mask;
} zfs_ace_hdr_t;
typedef zfs_ace_hdr_t zfs_ace_abstract_t;
/*
* Standard ACE
*/
typedef struct zfs_ace {
zfs_ace_hdr_t z_hdr;
uint64_t z_fuid;
} zfs_ace_t;
/*
* The following type only applies to ACE_ACCESS_ALLOWED|DENIED_OBJECT_ACE_TYPE
* and will only be set/retrieved in a CIFS context.
*/
typedef struct zfs_object_ace {
zfs_ace_t z_ace;
uint8_t z_object_type[16]; /* object type */
uint8_t z_inherit_type[16]; /* inherited object type */
} zfs_object_ace_t;
typedef struct zfs_oldace {
uint32_t z_fuid; /* "who" */
uint32_t z_access_mask; /* access mask */
uint16_t z_flags; /* flags, i.e inheritance */
uint16_t z_type; /* type of entry allow/deny */
} zfs_oldace_t;
typedef struct zfs_acl_phys_v0 {
uint64_t z_acl_extern_obj; /* ext acl pieces */
uint32_t z_acl_count; /* Number of ACEs */
uint16_t z_acl_version; /* acl version */
uint16_t z_acl_pad; /* pad */
zfs_oldace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */
} zfs_acl_phys_v0_t;
#define ZFS_ACE_SPACE (sizeof (zfs_oldace_t) * ACE_SLOT_CNT)
/*
* Size of ACL count is always 2 bytes.
* Necessary to for dealing with both V0 ACL and V1 ACL layout
*/
#define ZFS_ACL_COUNT_SIZE (sizeof (uint16_t))
typedef struct zfs_acl_phys {
uint64_t z_acl_extern_obj; /* ext acl pieces */
uint32_t z_acl_size; /* Number of bytes in ACL */
uint16_t z_acl_version; /* acl version */
uint16_t z_acl_count; /* ace count */
uint8_t z_ace_data[ZFS_ACE_SPACE]; /* space for embedded ACEs */
} zfs_acl_phys_t;
typedef struct acl_ops {
uint32_t (*ace_mask_get) (void *acep); /* get access mask */
void (*ace_mask_set) (void *acep,
uint32_t mask); /* set access mask */
uint16_t (*ace_flags_get) (void *acep); /* get flags */
void (*ace_flags_set) (void *acep,
uint16_t flags); /* set flags */
uint16_t (*ace_type_get)(void *acep); /* get type */
void (*ace_type_set)(void *acep,
uint16_t type); /* set type */
uint64_t (*ace_who_get)(void *acep); /* get who/fuid */
void (*ace_who_set)(void *acep,
uint64_t who); /* set who/fuid */
size_t (*ace_size)(void *acep); /* how big is this ace */
size_t (*ace_abstract_size)(void); /* sizeof abstract entry */
int (*ace_mask_off)(void); /* off of access mask in ace */
/* ptr to data if any */
int (*ace_data)(void *acep, void **datap);
} acl_ops_t;
/*
* A zfs_acl_t structure is composed of a list of zfs_acl_node_t's.
* Each node will have one or more ACEs associated with it. You will
* only have multiple nodes during a chmod operation. Normally only
* one node is required.
*/
typedef struct zfs_acl_node {
list_node_t z_next; /* Next chunk of ACEs */
void *z_acldata; /* pointer into actual ACE(s) */
void *z_allocdata; /* pointer to kmem allocated memory */
size_t z_allocsize; /* Size of blob in bytes */
size_t z_size; /* length of ACL data */
uint64_t z_ace_count; /* number of ACEs in this acl node */
int z_ace_idx; /* ace iterator positioned on */
} zfs_acl_node_t;
typedef struct zfs_acl {
uint64_t z_acl_count; /* Number of ACEs */
size_t z_acl_bytes; /* Number of bytes in ACL */
uint_t z_version; /* version of ACL */
void *z_next_ace; /* pointer to next ACE */
uint64_t z_hints; /* ACL hints (ZFS_INHERIT_ACE ...) */
zfs_acl_node_t *z_curr_node; /* current node iterator is handling */
list_t z_acl; /* chunks of ACE data */
- acl_ops_t *z_ops; /* ACL operations */
+ const acl_ops_t *z_ops; /* ACL operations */
} zfs_acl_t;
typedef struct acl_locator_cb {
zfs_acl_t *cb_aclp;
zfs_acl_node_t *cb_acl_node;
} zfs_acl_locator_cb_t;
#define ACL_DATA_ALLOCED 0x1
#define ZFS_ACL_SIZE(aclcnt) (sizeof (ace_t) * (aclcnt))
struct zfs_fuid_info;
typedef struct zfs_acl_ids {
uint64_t z_fuid; /* file owner fuid */
uint64_t z_fgid; /* file group owner fuid */
uint64_t z_mode; /* mode to set on create */
zfs_acl_t *z_aclp; /* ACL to create with file */
struct zfs_fuid_info *z_fuidp; /* for tracking fuids for log */
} zfs_acl_ids_t;
/*
* Property values for acl_mode and acl_inherit.
*
* acl_mode can take discard, noallow, groupmask and passthrough.
* whereas acl_inherit has secure instead of groupmask.
*/
#define ZFS_ACL_DISCARD 0
#define ZFS_ACL_NOALLOW 1
#define ZFS_ACL_GROUPMASK 2
#define ZFS_ACL_PASSTHROUGH 3
#define ZFS_ACL_RESTRICTED 4
#define ZFS_ACL_PASSTHROUGH_X 5
struct znode;
struct zfsvfs;
#ifdef _KERNEL
int zfs_acl_ids_create(struct znode *, int, vattr_t *,
cred_t *, vsecattr_t *, zfs_acl_ids_t *);
void zfs_acl_ids_free(zfs_acl_ids_t *);
boolean_t zfs_acl_ids_overquota(struct zfsvfs *, zfs_acl_ids_t *, uint64_t);
int zfs_getacl(struct znode *, vsecattr_t *, boolean_t, cred_t *);
int zfs_setacl(struct znode *, vsecattr_t *, boolean_t, cred_t *);
void zfs_acl_rele(void *);
void zfs_oldace_byteswap(ace_t *, int);
void zfs_ace_byteswap(void *, size_t, boolean_t);
extern boolean_t zfs_has_access(struct znode *zp, cred_t *cr);
extern int zfs_zaccess(struct znode *, int, int, boolean_t, cred_t *);
int zfs_fastaccesschk_execute(struct znode *, cred_t *);
extern int zfs_zaccess_rwx(struct znode *, mode_t, int, cred_t *);
extern int zfs_zaccess_unix(struct znode *, mode_t, cred_t *);
extern int zfs_acl_access(struct znode *, int, cred_t *);
int zfs_acl_chmod_setattr(struct znode *, zfs_acl_t **, uint64_t);
int zfs_zaccess_delete(struct znode *, struct znode *, cred_t *);
int zfs_zaccess_rename(struct znode *, struct znode *,
struct znode *, struct znode *, cred_t *cr);
void zfs_acl_free(zfs_acl_t *);
int zfs_vsec_2_aclp(struct zfsvfs *, umode_t, vsecattr_t *, cred_t *,
struct zfs_fuid_info **, zfs_acl_t **);
int zfs_aclset_common(struct znode *, zfs_acl_t *, cred_t *, dmu_tx_t *);
uint64_t zfs_external_acl(struct znode *);
int zfs_znode_acl_version(struct znode *);
int zfs_acl_size(struct znode *, int *);
zfs_acl_t *zfs_acl_alloc(int);
zfs_acl_node_t *zfs_acl_node_alloc(size_t);
void zfs_acl_xform(struct znode *, zfs_acl_t *, cred_t *);
void zfs_acl_data_locator(void **, uint32_t *, uint32_t, boolean_t, void *);
uint64_t zfs_mode_compute(uint64_t, zfs_acl_t *,
uint64_t *, uint64_t, uint64_t);
int zfs_acl_node_read(struct znode *, boolean_t, zfs_acl_t **, boolean_t);
int zfs_acl_chown_setattr(struct znode *);
#endif
#ifdef __cplusplus
}
#endif
#endif /* _SYS_FS_ZFS_ACL_H */
diff --git a/include/sys/zfs_context.h b/include/sys/zfs_context.h
index b1df9f3f3878..6d1fd83df522 100644
--- a/include/sys/zfs_context.h
+++ b/include/sys/zfs_context.h
@@ -1,782 +1,782 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
* Copyright (c) 2012, Joyent, Inc. All rights reserved.
*/
#ifndef _SYS_ZFS_CONTEXT_H
#define _SYS_ZFS_CONTEXT_H
#ifdef __cplusplus
extern "C" {
#endif
/*
* This code compiles in three different contexts. When __KERNEL__ is defined,
* the code uses "unix-like" kernel interfaces. When _STANDALONE is defined, the
* code is running in a reduced capacity environment of the boot loader which is
* generally a subset of both POSIX and kernel interfaces (with a few unique
* interfaces too). When neither are defined, it's in a userland POSIX or
* similar environment.
*/
#if defined(__KERNEL__) || defined(_STANDALONE)
#include <sys/types.h>
#include <sys/atomic.h>
#include <sys/sysmacros.h>
#include <sys/vmsystm.h>
#include <sys/condvar.h>
#include <sys/cmn_err.h>
#include <sys/kmem.h>
#include <sys/kmem_cache.h>
#include <sys/vmem.h>
#include <sys/taskq.h>
#include <sys/param.h>
#include <sys/disp.h>
#include <sys/debug.h>
#include <sys/random.h>
#include <sys/strings.h>
#include <sys/byteorder.h>
#include <sys/list.h>
#include <sys/time.h>
#include <sys/zone.h>
#include <sys/kstat.h>
#include <sys/zfs_debug.h>
#include <sys/sysevent.h>
#include <sys/sysevent/eventdefs.h>
#include <sys/zfs_delay.h>
#include <sys/sunddi.h>
#include <sys/ctype.h>
#include <sys/disp.h>
#include <sys/trace.h>
#include <sys/procfs_list.h>
#include <sys/mod.h>
#include <sys/uio_impl.h>
#include <sys/zfs_context_os.h>
#else /* _KERNEL || _STANDALONE */
#define _SYS_MUTEX_H
#define _SYS_RWLOCK_H
#define _SYS_CONDVAR_H
#define _SYS_VNODE_H
#define _SYS_VFS_H
#define _SYS_SUNDDI_H
#define _SYS_CALLB_H
#include <stdio.h>
#include <stdlib.h>
#include <stddef.h>
#include <stdarg.h>
#include <fcntl.h>
#include <unistd.h>
#include <errno.h>
#include <string.h>
#include <strings.h>
#include <pthread.h>
#include <setjmp.h>
#include <assert.h>
#include <umem.h>
#include <limits.h>
#include <atomic.h>
#include <dirent.h>
#include <time.h>
#include <ctype.h>
#include <signal.h>
#include <sys/mman.h>
#include <sys/types.h>
#include <sys/cred.h>
#include <sys/sysmacros.h>
#include <sys/resource.h>
#include <sys/byteorder.h>
#include <sys/list.h>
#include <sys/mod.h>
#include <sys/uio.h>
#include <sys/zfs_debug.h>
#include <sys/kstat.h>
#include <sys/u8_textprep.h>
#include <sys/sysevent.h>
#include <sys/sysevent/eventdefs.h>
#include <sys/sunddi.h>
#include <sys/debug.h>
#include <sys/utsname.h>
#include <sys/trace_zfs.h>
#include <sys/zfs_context_os.h>
/*
* Stack
*/
#define noinline __attribute__((noinline))
#define likely(x) __builtin_expect((x), 1)
#define unlikely(x) __builtin_expect((x), 0)
/*
* Debugging
*/
/*
* Note that we are not using the debugging levels.
*/
#define CE_CONT 0 /* continuation */
#define CE_NOTE 1 /* notice */
#define CE_WARN 2 /* warning */
#define CE_PANIC 3 /* panic */
#define CE_IGNORE 4 /* print nothing */
/*
* ZFS debugging
*/
extern void dprintf_setup(int *argc, char **argv);
extern void cmn_err(int, const char *, ...);
extern void vcmn_err(int, const char *, va_list);
extern void panic(const char *, ...) __NORETURN;
extern void vpanic(const char *, va_list) __NORETURN;
#define fm_panic panic
/*
* DTrace SDT probes have different signatures in userland than they do in
* the kernel. If they're being used in kernel code, re-define them out of
* existence for their counterparts in libzpool.
*
* Here's an example of how to use the set-error probes in userland:
* zfs$target:::set-error /arg0 == EBUSY/ {stack();}
*
* Here's an example of how to use DTRACE_PROBE probes in userland:
* If there is a probe declared as follows:
* DTRACE_PROBE2(zfs__probe_name, uint64_t, blkid, dnode_t *, dn);
* Then you can use it as follows:
* zfs$target:::probe2 /copyinstr(arg0) == "zfs__probe_name"/
* {printf("%u %p\n", arg1, arg2);}
*/
#ifdef DTRACE_PROBE
#undef DTRACE_PROBE
#endif /* DTRACE_PROBE */
#define DTRACE_PROBE(a)
#ifdef DTRACE_PROBE1
#undef DTRACE_PROBE1
#endif /* DTRACE_PROBE1 */
#define DTRACE_PROBE1(a, b, c)
#ifdef DTRACE_PROBE2
#undef DTRACE_PROBE2
#endif /* DTRACE_PROBE2 */
#define DTRACE_PROBE2(a, b, c, d, e)
#ifdef DTRACE_PROBE3
#undef DTRACE_PROBE3
#endif /* DTRACE_PROBE3 */
#define DTRACE_PROBE3(a, b, c, d, e, f, g)
#ifdef DTRACE_PROBE4
#undef DTRACE_PROBE4
#endif /* DTRACE_PROBE4 */
#define DTRACE_PROBE4(a, b, c, d, e, f, g, h, i)
/*
* Tunables.
*/
typedef struct zfs_kernel_param {
const char *name; /* unused stub */
} zfs_kernel_param_t;
#define ZFS_MODULE_PARAM(scope_prefix, name_prefix, name, type, perm, desc)
#define ZFS_MODULE_PARAM_ARGS void
#define ZFS_MODULE_PARAM_CALL(scope_prefix, name_prefix, name, setfunc, \
getfunc, perm, desc)
/*
* Threads.
*/
typedef pthread_t kthread_t;
#define TS_RUN 0x00000002
#define TS_JOINABLE 0x00000004
#define curthread ((void *)(uintptr_t)pthread_self())
#define kpreempt(x) yield()
#define getcomm() "unknown"
#define thread_create_named(name, stk, stksize, func, arg, len, \
pp, state, pri) \
zk_thread_create(func, arg, stksize, state)
#define thread_create(stk, stksize, func, arg, len, pp, state, pri) \
zk_thread_create(func, arg, stksize, state)
#define thread_exit() pthread_exit(NULL)
#define thread_join(t) pthread_join((pthread_t)(t), NULL)
#define newproc(f, a, cid, pri, ctp, pid) (ENOSYS)
/* in libzpool, p0 exists only to have its address taken */
typedef struct proc {
uintptr_t this_is_never_used_dont_dereference_it;
} proc_t;
extern struct proc p0;
#define curproc (&p0)
#define PS_NONE -1
extern kthread_t *zk_thread_create(void (*func)(void *), void *arg,
size_t stksize, int state);
#define issig(why) (FALSE)
#define ISSIG(thr, why) (FALSE)
#define kpreempt_disable() ((void)0)
#define kpreempt_enable() ((void)0)
#define cond_resched() sched_yield()
/*
* Mutexes
*/
typedef struct kmutex {
pthread_mutex_t m_lock;
pthread_t m_owner;
} kmutex_t;
#define MUTEX_DEFAULT 0
#define MUTEX_NOLOCKDEP MUTEX_DEFAULT
#define MUTEX_HELD(mp) pthread_equal((mp)->m_owner, pthread_self())
#define MUTEX_NOT_HELD(mp) !MUTEX_HELD(mp)
extern void mutex_init(kmutex_t *mp, char *name, int type, void *cookie);
extern void mutex_destroy(kmutex_t *mp);
extern void mutex_enter(kmutex_t *mp);
extern void mutex_exit(kmutex_t *mp);
extern int mutex_tryenter(kmutex_t *mp);
#define NESTED_SINGLE 1
#define mutex_enter_nested(mp, class) mutex_enter(mp)
/*
* RW locks
*/
typedef struct krwlock {
pthread_rwlock_t rw_lock;
pthread_t rw_owner;
uint_t rw_readers;
} krwlock_t;
typedef int krw_t;
#define RW_READER 0
#define RW_WRITER 1
#define RW_DEFAULT RW_READER
#define RW_NOLOCKDEP RW_READER
#define RW_READ_HELD(rw) ((rw)->rw_readers > 0)
#define RW_WRITE_HELD(rw) pthread_equal((rw)->rw_owner, pthread_self())
#define RW_LOCK_HELD(rw) (RW_READ_HELD(rw) || RW_WRITE_HELD(rw))
extern void rw_init(krwlock_t *rwlp, char *name, int type, void *arg);
extern void rw_destroy(krwlock_t *rwlp);
extern void rw_enter(krwlock_t *rwlp, krw_t rw);
extern int rw_tryenter(krwlock_t *rwlp, krw_t rw);
extern int rw_tryupgrade(krwlock_t *rwlp);
extern void rw_exit(krwlock_t *rwlp);
#define rw_downgrade(rwlp) do { } while (0)
/*
* Credentials
*/
extern uid_t crgetuid(cred_t *cr);
extern uid_t crgetruid(cred_t *cr);
extern gid_t crgetgid(cred_t *cr);
extern int crgetngroups(cred_t *cr);
extern gid_t *crgetgroups(cred_t *cr);
/*
* Condition variables
*/
typedef pthread_cond_t kcondvar_t;
#define CV_DEFAULT 0
#define CALLOUT_FLAG_ABSOLUTE 0x2
extern void cv_init(kcondvar_t *cv, char *name, int type, void *arg);
extern void cv_destroy(kcondvar_t *cv);
extern void cv_wait(kcondvar_t *cv, kmutex_t *mp);
extern int cv_wait_sig(kcondvar_t *cv, kmutex_t *mp);
extern int cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime);
extern int cv_timedwait_hires(kcondvar_t *cvp, kmutex_t *mp, hrtime_t tim,
hrtime_t res, int flag);
extern void cv_signal(kcondvar_t *cv);
extern void cv_broadcast(kcondvar_t *cv);
#define cv_timedwait_io(cv, mp, at) cv_timedwait(cv, mp, at)
#define cv_timedwait_idle(cv, mp, at) cv_timedwait(cv, mp, at)
#define cv_timedwait_sig(cv, mp, at) cv_timedwait(cv, mp, at)
#define cv_wait_io(cv, mp) cv_wait(cv, mp)
#define cv_wait_idle(cv, mp) cv_wait(cv, mp)
#define cv_wait_io_sig(cv, mp) cv_wait_sig(cv, mp)
#define cv_timedwait_sig_hires(cv, mp, t, r, f) \
cv_timedwait_hires(cv, mp, t, r, f)
#define cv_timedwait_idle_hires(cv, mp, t, r, f) \
cv_timedwait_hires(cv, mp, t, r, f)
/*
* Thread-specific data
*/
#define tsd_get(k) pthread_getspecific(k)
#define tsd_set(k, v) pthread_setspecific(k, v)
#define tsd_create(kp, d) pthread_key_create((pthread_key_t *)kp, d)
#define tsd_destroy(kp) /* nothing */
#ifdef __FreeBSD__
typedef off_t loff_t;
#endif
/*
* kstat creation, installation and deletion
*/
extern kstat_t *kstat_create(const char *, int,
const char *, const char *, uchar_t, ulong_t, uchar_t);
extern void kstat_install(kstat_t *);
extern void kstat_delete(kstat_t *);
extern void kstat_set_raw_ops(kstat_t *ksp,
int (*headers)(char *buf, size_t size),
int (*data)(char *buf, size_t size, void *data),
void *(*addr)(kstat_t *ksp, loff_t index));
/*
* procfs list manipulation
*/
typedef struct procfs_list {
void *pl_private;
kmutex_t pl_lock;
list_t pl_list;
uint64_t pl_next_id;
size_t pl_node_offset;
} procfs_list_t;
#ifndef __cplusplus
struct seq_file { };
void seq_printf(struct seq_file *m, const char *fmt, ...);
typedef struct procfs_list_node {
list_node_t pln_link;
uint64_t pln_id;
} procfs_list_node_t;
void procfs_list_install(const char *module,
const char *submodule,
const char *name,
mode_t mode,
procfs_list_t *procfs_list,
int (*show)(struct seq_file *f, void *p),
int (*show_header)(struct seq_file *f),
int (*clear)(procfs_list_t *procfs_list),
size_t procfs_list_node_off);
void procfs_list_uninstall(procfs_list_t *procfs_list);
void procfs_list_destroy(procfs_list_t *procfs_list);
void procfs_list_add(procfs_list_t *procfs_list, void *p);
#endif
/*
* Kernel memory
*/
#define KM_SLEEP UMEM_NOFAIL
#define KM_PUSHPAGE KM_SLEEP
#define KM_NOSLEEP UMEM_DEFAULT
#define KM_NORMALPRI 0 /* not needed with UMEM_DEFAULT */
#define KMC_NODEBUG UMC_NODEBUG
#define KMC_KVMEM 0x0
#define kmem_alloc(_s, _f) umem_alloc(_s, _f)
#define kmem_zalloc(_s, _f) umem_zalloc(_s, _f)
#define kmem_free(_b, _s) umem_free(_b, _s)
#define vmem_alloc(_s, _f) kmem_alloc(_s, _f)
#define vmem_zalloc(_s, _f) kmem_zalloc(_s, _f)
#define vmem_free(_b, _s) kmem_free(_b, _s)
#define kmem_cache_create(_a, _b, _c, _d, _e, _f, _g, _h, _i) \
umem_cache_create(_a, _b, _c, _d, _e, _f, _g, _h, _i)
#define kmem_cache_destroy(_c) umem_cache_destroy(_c)
#define kmem_cache_alloc(_c, _f) umem_cache_alloc(_c, _f)
#define kmem_cache_free(_c, _b) umem_cache_free(_c, _b)
#define kmem_debugging() 0
#define kmem_cache_reap_now(_c) umem_cache_reap_now(_c);
#define kmem_cache_set_move(_c, _cb) /* nothing */
#define POINTER_INVALIDATE(_pp) /* nothing */
#define POINTER_IS_VALID(_p) 0
typedef umem_cache_t kmem_cache_t;
typedef enum kmem_cbrc {
KMEM_CBRC_YES,
KMEM_CBRC_NO,
KMEM_CBRC_LATER,
KMEM_CBRC_DONT_NEED,
KMEM_CBRC_DONT_KNOW
} kmem_cbrc_t;
/*
* Task queues
*/
#define TASKQ_NAMELEN 31
typedef uintptr_t taskqid_t;
typedef void (task_func_t)(void *);
typedef struct taskq_ent {
struct taskq_ent *tqent_next;
struct taskq_ent *tqent_prev;
task_func_t *tqent_func;
void *tqent_arg;
uintptr_t tqent_flags;
} taskq_ent_t;
typedef struct taskq {
char tq_name[TASKQ_NAMELEN + 1];
kmutex_t tq_lock;
krwlock_t tq_threadlock;
kcondvar_t tq_dispatch_cv;
kcondvar_t tq_wait_cv;
kthread_t **tq_threadlist;
int tq_flags;
int tq_active;
int tq_nthreads;
int tq_nalloc;
int tq_minalloc;
int tq_maxalloc;
kcondvar_t tq_maxalloc_cv;
int tq_maxalloc_wait;
taskq_ent_t *tq_freelist;
taskq_ent_t tq_task;
} taskq_t;
#define TQENT_FLAG_PREALLOC 0x1 /* taskq_dispatch_ent used */
#define TASKQ_PREPOPULATE 0x0001
#define TASKQ_CPR_SAFE 0x0002 /* Use CPR safe protocol */
#define TASKQ_DYNAMIC 0x0004 /* Use dynamic thread scheduling */
#define TASKQ_THREADS_CPU_PCT 0x0008 /* Scale # threads by # cpus */
#define TASKQ_DC_BATCH 0x0010 /* Mark threads as batch */
#define TQ_SLEEP KM_SLEEP /* Can block for memory */
#define TQ_NOSLEEP KM_NOSLEEP /* cannot block for memory; may fail */
#define TQ_NOQUEUE 0x02 /* Do not enqueue if can't dispatch */
#define TQ_FRONT 0x08 /* Queue in front */
#define TASKQID_INVALID ((taskqid_t)0)
extern taskq_t *system_taskq;
extern taskq_t *system_delay_taskq;
extern taskq_t *taskq_create(const char *, int, pri_t, int, int, uint_t);
#define taskq_create_proc(a, b, c, d, e, p, f) \
(taskq_create(a, b, c, d, e, f))
#define taskq_create_sysdc(a, b, d, e, p, dc, f) \
- (taskq_create(a, b, maxclsyspri, d, e, f))
+ ((void) sizeof (dc), taskq_create(a, b, maxclsyspri, d, e, f))
extern taskqid_t taskq_dispatch(taskq_t *, task_func_t, void *, uint_t);
extern taskqid_t taskq_dispatch_delay(taskq_t *, task_func_t, void *, uint_t,
clock_t);
extern void taskq_dispatch_ent(taskq_t *, task_func_t, void *, uint_t,
taskq_ent_t *);
extern int taskq_empty_ent(taskq_ent_t *);
extern void taskq_init_ent(taskq_ent_t *);
extern void taskq_destroy(taskq_t *);
extern void taskq_wait(taskq_t *);
extern void taskq_wait_id(taskq_t *, taskqid_t);
extern void taskq_wait_outstanding(taskq_t *, taskqid_t);
extern int taskq_member(taskq_t *, kthread_t *);
extern taskq_t *taskq_of_curthread(void);
extern int taskq_cancel_id(taskq_t *, taskqid_t);
extern void system_taskq_init(void);
extern void system_taskq_fini(void);
#define XVA_MAPSIZE 3
#define XVA_MAGIC 0x78766174
extern char *vn_dumpdir;
#define AV_SCANSTAMP_SZ 32 /* length of anti-virus scanstamp */
typedef struct xoptattr {
inode_timespec_t xoa_createtime; /* Create time of file */
uint8_t xoa_archive;
uint8_t xoa_system;
uint8_t xoa_readonly;
uint8_t xoa_hidden;
uint8_t xoa_nounlink;
uint8_t xoa_immutable;
uint8_t xoa_appendonly;
uint8_t xoa_nodump;
uint8_t xoa_settable;
uint8_t xoa_opaque;
uint8_t xoa_av_quarantined;
uint8_t xoa_av_modified;
uint8_t xoa_av_scanstamp[AV_SCANSTAMP_SZ];
uint8_t xoa_reparse;
uint8_t xoa_offline;
uint8_t xoa_sparse;
} xoptattr_t;
typedef struct vattr {
uint_t va_mask; /* bit-mask of attributes */
u_offset_t va_size; /* file size in bytes */
} vattr_t;
typedef struct xvattr {
vattr_t xva_vattr; /* Embedded vattr structure */
uint32_t xva_magic; /* Magic Number */
uint32_t xva_mapsize; /* Size of attr bitmap (32-bit words) */
uint32_t *xva_rtnattrmapp; /* Ptr to xva_rtnattrmap[] */
uint32_t xva_reqattrmap[XVA_MAPSIZE]; /* Requested attrs */
uint32_t xva_rtnattrmap[XVA_MAPSIZE]; /* Returned attrs */
xoptattr_t xva_xoptattrs; /* Optional attributes */
} xvattr_t;
typedef struct vsecattr {
uint_t vsa_mask; /* See below */
int vsa_aclcnt; /* ACL entry count */
void *vsa_aclentp; /* pointer to ACL entries */
int vsa_dfaclcnt; /* default ACL entry count */
void *vsa_dfaclentp; /* pointer to default ACL entries */
size_t vsa_aclentsz; /* ACE size in bytes of vsa_aclentp */
} vsecattr_t;
#define AT_MODE 0x00002
#define AT_UID 0x00004
#define AT_GID 0x00008
#define AT_FSID 0x00010
#define AT_NODEID 0x00020
#define AT_NLINK 0x00040
#define AT_SIZE 0x00080
#define AT_ATIME 0x00100
#define AT_MTIME 0x00200
#define AT_CTIME 0x00400
#define AT_RDEV 0x00800
#define AT_BLKSIZE 0x01000
#define AT_NBLOCKS 0x02000
#define AT_SEQ 0x08000
#define AT_XVATTR 0x10000
#define CRCREAT 0
#define F_FREESP 11
#define FIGNORECASE 0x80000 /* request case-insensitive lookups */
/*
* Random stuff
*/
#define ddi_get_lbolt() (gethrtime() >> 23)
#define ddi_get_lbolt64() (gethrtime() >> 23)
#define hz 119 /* frequency when using gethrtime() >> 23 for lbolt */
#define ddi_time_before(a, b) (a < b)
#define ddi_time_after(a, b) ddi_time_before(b, a)
#define ddi_time_before_eq(a, b) (!ddi_time_after(a, b))
#define ddi_time_after_eq(a, b) ddi_time_before_eq(b, a)
#define ddi_time_before64(a, b) (a < b)
#define ddi_time_after64(a, b) ddi_time_before64(b, a)
#define ddi_time_before_eq64(a, b) (!ddi_time_after64(a, b))
#define ddi_time_after_eq64(a, b) ddi_time_before_eq64(b, a)
extern void delay(clock_t ticks);
#define SEC_TO_TICK(sec) ((sec) * hz)
#define MSEC_TO_TICK(msec) (howmany((hrtime_t)(msec) * hz, MILLISEC))
#define USEC_TO_TICK(usec) (howmany((hrtime_t)(usec) * hz, MICROSEC))
#define NSEC_TO_TICK(nsec) (howmany((hrtime_t)(nsec) * hz, NANOSEC))
#define max_ncpus 64
#define boot_ncpus (sysconf(_SC_NPROCESSORS_ONLN))
/*
* Process priorities as defined by setpriority(2) and getpriority(2).
*/
#define minclsyspri 19
#define maxclsyspri -20
#define defclsyspri 0
#define CPU_SEQID ((uintptr_t)pthread_self() & (max_ncpus - 1))
#define CPU_SEQID_UNSTABLE CPU_SEQID
#define kcred NULL
#define CRED() NULL
#define ptob(x) ((x) * PAGESIZE)
#define NN_DIVISOR_1000 (1U << 0)
#define NN_NUMBUF_SZ (6)
extern uint64_t physmem;
extern const char *random_path;
extern const char *urandom_path;
extern int highbit64(uint64_t i);
extern int lowbit64(uint64_t i);
extern int random_get_bytes(uint8_t *ptr, size_t len);
extern int random_get_pseudo_bytes(uint8_t *ptr, size_t len);
static __inline__ uint32_t
random_in_range(uint32_t range)
{
uint32_t r;
ASSERT(range != 0);
if (range == 1)
return (0);
(void) random_get_pseudo_bytes((uint8_t *)&r, sizeof (r));
return (r % range);
}
extern void kernel_init(int mode);
extern void kernel_fini(void);
extern void random_init(void);
extern void random_fini(void);
struct spa;
extern void show_pool_stats(struct spa *);
extern int set_global_var(char const *arg);
typedef struct callb_cpr {
kmutex_t *cc_lockp;
} callb_cpr_t;
#define CALLB_CPR_INIT(cp, lockp, func, name) { \
(cp)->cc_lockp = lockp; \
}
#define CALLB_CPR_SAFE_BEGIN(cp) { \
ASSERT(MUTEX_HELD((cp)->cc_lockp)); \
}
#define CALLB_CPR_SAFE_END(cp, lockp) { \
ASSERT(MUTEX_HELD((cp)->cc_lockp)); \
}
#define CALLB_CPR_EXIT(cp) { \
ASSERT(MUTEX_HELD((cp)->cc_lockp)); \
mutex_exit((cp)->cc_lockp); \
}
#define zone_dataset_visible(x, y) (1)
#define INGLOBALZONE(z) (1)
extern uint32_t zone_get_hostid(void *zonep);
extern char *kmem_vasprintf(const char *fmt, va_list adx);
extern char *kmem_asprintf(const char *fmt, ...);
#define kmem_strfree(str) kmem_free((str), strlen(str) + 1)
#define kmem_strdup(s) strdup(s)
/*
* Hostname information
*/
extern char hw_serial[]; /* for userland-emulated hostid access */
extern int ddi_strtoul(const char *str, char **nptr, int base,
unsigned long *result);
extern int ddi_strtoull(const char *str, char **nptr, int base,
u_longlong_t *result);
typedef struct utsname utsname_t;
extern utsname_t *utsname(void);
/* ZFS Boot Related stuff. */
struct _buf {
intptr_t _fd;
};
struct bootstat {
uint64_t st_size;
};
typedef struct ace_object {
uid_t a_who;
uint32_t a_access_mask;
uint16_t a_flags;
uint16_t a_type;
uint8_t a_obj_type[16];
uint8_t a_inherit_obj_type[16];
} ace_object_t;
#define ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE 0x05
#define ACE_ACCESS_DENIED_OBJECT_ACE_TYPE 0x06
#define ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE 0x07
#define ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE 0x08
extern int zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr);
extern int zfs_secpolicy_rename_perms(const char *from, const char *to,
cred_t *cr);
extern int zfs_secpolicy_destroy_perms(const char *name, cred_t *cr);
extern int secpolicy_zfs(const cred_t *cr);
extern int secpolicy_zfs_proc(const cred_t *cr, proc_t *proc);
extern zoneid_t getzoneid(void);
/* SID stuff */
typedef struct ksiddomain {
uint_t kd_ref;
uint_t kd_len;
char *kd_name;
} ksiddomain_t;
ksiddomain_t *ksid_lookupdomain(const char *);
void ksiddomain_rele(ksiddomain_t *);
#define DDI_SLEEP KM_SLEEP
#define ddi_log_sysevent(_a, _b, _c, _d, _e, _f, _g) \
sysevent_post_event(_c, _d, _b, "libzpool", _e, _f)
#define zfs_sleep_until(wakeup) \
do { \
hrtime_t delta = wakeup - gethrtime(); \
struct timespec ts; \
ts.tv_sec = delta / NANOSEC; \
ts.tv_nsec = delta % NANOSEC; \
(void) nanosleep(&ts, NULL); \
} while (0)
typedef int fstrans_cookie_t;
extern fstrans_cookie_t spl_fstrans_mark(void);
extern void spl_fstrans_unmark(fstrans_cookie_t);
extern int __spl_pf_fstrans_check(void);
extern int kmem_cache_reap_active(void);
/*
* Kernel modules
*/
#define __init
#define __exit
#endif /* _KERNEL || _STANDALONE */
#ifdef __cplusplus
};
#endif
#endif /* _SYS_ZFS_CONTEXT_H */
diff --git a/include/sys/zfs_ioctl_impl.h b/include/sys/zfs_ioctl_impl.h
index 3db67ae98414..f9e4f6e6c4b2 100644
--- a/include/sys/zfs_ioctl_impl.h
+++ b/include/sys/zfs_ioctl_impl.h
@@ -1,102 +1,101 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
#ifndef _ZFS_IOCTL_IMPL_H_
#define _ZFS_IOCTL_IMPL_H_
extern kmutex_t zfsdev_state_lock;
-extern zfsdev_state_t *zfsdev_state_list;
extern unsigned long zfs_max_nvlist_src_size;
typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
typedef enum {
POOL_CHECK_NONE = 1 << 0,
POOL_CHECK_SUSPENDED = 1 << 1,
POOL_CHECK_READONLY = 1 << 2,
} zfs_ioc_poolcheck_t;
typedef enum {
NO_NAME,
POOL_NAME,
DATASET_NAME,
ENTITY_NAME
} zfs_ioc_namecheck_t;
/*
* IOC Keys are used to document and validate user->kernel interface inputs.
* See zfs_keys_recv_new for an example declaration. Any key name that is not
* listed will be rejected as input.
*
* The keyname 'optional' is always allowed, and must be an nvlist if present.
* Arguments which older kernels can safely ignore can be placed under the
* "optional" key.
*
* When adding new keys to an existing ioc for new functionality, consider:
* - adding an entry into zfs_sysfs.c zfs_features[] list
* - updating the libzfs_input_check.c test utility
*
* Note: in the ZK_WILDCARDLIST case, the name serves as documentation
* for the expected name (bookmark, snapshot, property, etc) but there
* is no validation in the preflight zfs_check_input_nvpairs() check.
*/
typedef enum {
ZK_OPTIONAL = 1 << 0, /* pair is optional */
ZK_WILDCARDLIST = 1 << 1, /* one or more unspecified key names */
} ioc_key_flag_t;
typedef struct zfs_ioc_key {
const char *zkey_name;
data_type_t zkey_type;
ioc_key_flag_t zkey_flags;
} zfs_ioc_key_t;
int zfs_secpolicy_config(zfs_cmd_t *, nvlist_t *, cred_t *);
void zfs_ioctl_register_dataset_nolog(zfs_ioc_t, zfs_ioc_legacy_func_t *,
zfs_secpolicy_func_t *, zfs_ioc_poolcheck_t);
void zfs_ioctl_register(const char *, zfs_ioc_t, zfs_ioc_func_t *,
zfs_secpolicy_func_t *, zfs_ioc_namecheck_t, zfs_ioc_poolcheck_t,
boolean_t, boolean_t, const zfs_ioc_key_t *, size_t);
uint64_t zfs_max_nvlist_src_size_os(void);
void zfs_ioctl_update_mount_cache(const char *dsname);
void zfs_ioctl_init_os(void);
boolean_t zfs_vfs_held(zfsvfs_t *);
int zfs_vfs_ref(zfsvfs_t **);
void zfs_vfs_rele(zfsvfs_t *);
long zfsdev_ioctl_common(uint_t, zfs_cmd_t *, int);
int zfsdev_attach(void);
void zfsdev_detach(void);
void zfsdev_private_set_state(void *, zfsdev_state_t *);
zfsdev_state_t *zfsdev_private_get_state(void *);
int zfsdev_state_init(void *);
void zfsdev_state_destroy(void *);
int zfs_kmod_init(void);
void zfs_kmod_fini(void);
#endif
diff --git a/include/sys/zfs_sa.h b/include/sys/zfs_sa.h
index 1ca7ced331c5..a0c383807aa4 100644
--- a/include/sys/zfs_sa.h
+++ b/include/sys/zfs_sa.h
@@ -1,153 +1,152 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2010 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#ifndef _SYS_ZFS_SA_H
#define _SYS_ZFS_SA_H
#ifdef _KERNEL
#include <sys/types32.h>
#include <sys/list.h>
#include <sys/dmu.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_znode.h>
#include <sys/sa.h>
#include <sys/zil.h>
#endif
#ifdef __cplusplus
extern "C" {
#endif
/*
* This is the list of known attributes
* to the ZPL. The values of the actual
* attributes are not defined by the order
* the enums. It is controlled by the attribute
* registration mechanism. Two different file system
* could have different numeric values for the same
* attributes. this list is only used for dereferencing
* into the table that will hold the actual numeric value.
*/
typedef enum zpl_attr {
ZPL_ATIME,
ZPL_MTIME,
ZPL_CTIME,
ZPL_CRTIME,
ZPL_GEN,
ZPL_MODE,
ZPL_SIZE,
ZPL_PARENT,
ZPL_LINKS,
ZPL_XATTR,
ZPL_RDEV,
ZPL_FLAGS,
ZPL_UID,
ZPL_GID,
ZPL_PAD,
ZPL_ZNODE_ACL,
ZPL_DACL_COUNT,
ZPL_SYMLINK,
ZPL_SCANSTAMP,
ZPL_DACL_ACES,
ZPL_DXATTR,
ZPL_PROJID,
ZPL_END
} zpl_attr_t;
#define ZFS_OLD_ZNODE_PHYS_SIZE 0x108
#define ZFS_SA_BASE_ATTR_SIZE (ZFS_OLD_ZNODE_PHYS_SIZE - \
sizeof (zfs_acl_phys_t))
#define SA_MODE_OFFSET 0
#define SA_SIZE_OFFSET 8
#define SA_GEN_OFFSET 16
#define SA_UID_OFFSET 24
#define SA_GID_OFFSET 32
#define SA_PARENT_OFFSET 40
#define SA_FLAGS_OFFSET 48
#define SA_PROJID_OFFSET 128
-extern sa_attr_reg_t zfs_attr_table[ZPL_END + 1];
-extern sa_attr_reg_t zfs_legacy_attr_table[ZPL_END + 1];
+extern const sa_attr_reg_t zfs_attr_table[ZPL_END + 1];
/*
* This is a deprecated data structure that only exists for
* dealing with file systems create prior to ZPL version 5.
*/
typedef struct znode_phys {
uint64_t zp_atime[2]; /* 0 - last file access time */
uint64_t zp_mtime[2]; /* 16 - last file modification time */
uint64_t zp_ctime[2]; /* 32 - last file change time */
uint64_t zp_crtime[2]; /* 48 - creation time */
uint64_t zp_gen; /* 64 - generation (txg of creation) */
uint64_t zp_mode; /* 72 - file mode bits */
uint64_t zp_size; /* 80 - size of file */
uint64_t zp_parent; /* 88 - directory parent (`..') */
uint64_t zp_links; /* 96 - number of links to file */
uint64_t zp_xattr; /* 104 - DMU object for xattrs */
uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */
uint64_t zp_flags; /* 120 - persistent flags */
uint64_t zp_uid; /* 128 - file owner */
uint64_t zp_gid; /* 136 - owning group */
uint64_t zp_zap; /* 144 - extra attributes */
uint64_t zp_pad[3]; /* 152 - future */
zfs_acl_phys_t zp_acl; /* 176 - 263 ACL */
/*
* Data may pad out any remaining bytes in the znode buffer, eg:
*
* |<---------------------- dnode_phys (512) ------------------------>|
* |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->|
* |<---- znode (264) ---->|<---- data (56) ---->|
*
* At present, we use this space for the following:
* - symbolic links
* - 32-byte anti-virus scanstamp (regular files only)
*/
} znode_phys_t;
#ifdef _KERNEL
#define DXATTR_MAX_ENTRY_SIZE (32768)
#define DXATTR_MAX_SA_SIZE (SPA_OLD_MAXBLOCKSIZE >> 1)
int zfs_sa_readlink(struct znode *, zfs_uio_t *);
void zfs_sa_symlink(struct znode *, char *link, int len, dmu_tx_t *);
void zfs_sa_get_scanstamp(struct znode *, xvattr_t *);
void zfs_sa_set_scanstamp(struct znode *, xvattr_t *, dmu_tx_t *);
int zfs_sa_get_xattr(struct znode *);
int zfs_sa_set_xattr(struct znode *);
void zfs_sa_upgrade(struct sa_handle *, dmu_tx_t *);
void zfs_sa_upgrade_txholds(dmu_tx_t *, struct znode *);
void zfs_sa_init(void);
void zfs_sa_fini(void);
#endif
#ifdef __cplusplus
}
#endif
#endif /* _SYS_ZFS_SA_H */
diff --git a/include/sys/zil.h b/include/sys/zil.h
index cefbccb32f22..8e5a49da2929 100644
--- a/include/sys/zil.h
+++ b/include/sys/zil.h
@@ -1,530 +1,528 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
*/
/* Portions Copyright 2010 Robert Milkowski */
#ifndef _SYS_ZIL_H
#define _SYS_ZIL_H
#include <sys/types.h>
#include <sys/spa.h>
#include <sys/zio.h>
#include <sys/dmu.h>
#include <sys/zio_crypt.h>
#ifdef __cplusplus
extern "C" {
#endif
struct dsl_pool;
struct dsl_dataset;
struct lwb;
/*
* Intent log format:
*
* Each objset has its own intent log. The log header (zil_header_t)
* for objset N's intent log is kept in the Nth object of the SPA's
* intent_log objset. The log header points to a chain of log blocks,
* each of which contains log records (i.e., transactions) followed by
* a log block trailer (zil_trailer_t). The format of a log record
* depends on the record (or transaction) type, but all records begin
* with a common structure that defines the type, length, and txg.
*/
/*
* Intent log header - this on disk structure holds fields to manage
* the log. All fields are 64 bit to easily handle cross architectures.
*/
typedef struct zil_header {
uint64_t zh_claim_txg; /* txg in which log blocks were claimed */
uint64_t zh_replay_seq; /* highest replayed sequence number */
blkptr_t zh_log; /* log chain */
uint64_t zh_claim_blk_seq; /* highest claimed block sequence number */
uint64_t zh_flags; /* header flags */
uint64_t zh_claim_lr_seq; /* highest claimed lr sequence number */
uint64_t zh_pad[3];
} zil_header_t;
/*
* zh_flags bit settings
*/
#define ZIL_REPLAY_NEEDED 0x1 /* replay needed - internal only */
#define ZIL_CLAIM_LR_SEQ_VALID 0x2 /* zh_claim_lr_seq field is valid */
/*
* Log block chaining.
*
* Log blocks are chained together. Originally they were chained at the
* end of the block. For performance reasons the chain was moved to the
* beginning of the block which allows writes for only the data being used.
* The older position is supported for backwards compatibility.
*
* The zio_eck_t contains a zec_cksum which for the intent log is
* the sequence number of this log block. A seq of 0 is invalid.
* The zec_cksum is checked by the SPA against the sequence
* number passed in the blk_cksum field of the blkptr_t
*/
typedef struct zil_chain {
uint64_t zc_pad;
blkptr_t zc_next_blk; /* next block in chain */
uint64_t zc_nused; /* bytes in log block used */
zio_eck_t zc_eck; /* block trailer */
} zil_chain_t;
#define ZIL_MIN_BLKSZ 4096ULL
/*
* ziltest is by and large an ugly hack, but very useful in
* checking replay without tedious work.
* When running ziltest we want to keep all itx's and so maintain
* a single list in the zl_itxg[] that uses a high txg: ZILTEST_TXG
* We subtract TXG_CONCURRENT_STATES to allow for common code.
*/
#define ZILTEST_TXG (UINT64_MAX - TXG_CONCURRENT_STATES)
/*
* The words of a log block checksum.
*/
#define ZIL_ZC_GUID_0 0
#define ZIL_ZC_GUID_1 1
#define ZIL_ZC_OBJSET 2
#define ZIL_ZC_SEQ 3
typedef enum zil_create {
Z_FILE,
Z_DIR,
Z_XATTRDIR,
} zil_create_t;
/*
* size of xvattr log section.
* its composed of lr_attr_t + xvattr bitmap + 2 64 bit timestamps
* for create time and a single 64 bit integer for all of the attributes,
* and 4 64 bit integers (32 bytes) for the scanstamp.
*
*/
#define ZIL_XVAT_SIZE(mapsize) \
sizeof (lr_attr_t) + (sizeof (uint32_t) * (mapsize - 1)) + \
(sizeof (uint64_t) * 7)
/*
* Size of ACL in log. The ACE data is padded out to properly align
* on 8 byte boundary.
*/
#define ZIL_ACE_LENGTH(x) (roundup(x, sizeof (uint64_t)))
/*
* Intent log transaction types and record structures
*/
#define TX_COMMIT 0 /* Commit marker (no on-disk state) */
#define TX_CREATE 1 /* Create file */
#define TX_MKDIR 2 /* Make directory */
#define TX_MKXATTR 3 /* Make XATTR directory */
#define TX_SYMLINK 4 /* Create symbolic link to a file */
#define TX_REMOVE 5 /* Remove file */
#define TX_RMDIR 6 /* Remove directory */
#define TX_LINK 7 /* Create hard link to a file */
#define TX_RENAME 8 /* Rename a file */
#define TX_WRITE 9 /* File write */
#define TX_TRUNCATE 10 /* Truncate a file */
#define TX_SETATTR 11 /* Set file attributes */
#define TX_ACL_V0 12 /* Set old formatted ACL */
#define TX_ACL 13 /* Set ACL */
#define TX_CREATE_ACL 14 /* create with ACL */
#define TX_CREATE_ATTR 15 /* create + attrs */
#define TX_CREATE_ACL_ATTR 16 /* create with ACL + attrs */
#define TX_MKDIR_ACL 17 /* mkdir with ACL */
#define TX_MKDIR_ATTR 18 /* mkdir with attr */
#define TX_MKDIR_ACL_ATTR 19 /* mkdir with ACL + attrs */
#define TX_WRITE2 20 /* dmu_sync EALREADY write */
#define TX_MAX_TYPE 21 /* Max transaction type */
/*
* The transactions for mkdir, symlink, remove, rmdir, link, and rename
* may have the following bit set, indicating the original request
* specified case-insensitive handling of names.
*/
#define TX_CI ((uint64_t)0x1 << 63) /* case-insensitive behavior requested */
/*
* Transactions for write, truncate, setattr, acl_v0, and acl can be logged
* out of order. For convenience in the code, all such records must have
* lr_foid at the same offset.
*/
#define TX_OOO(txtype) \
((txtype) == TX_WRITE || \
(txtype) == TX_TRUNCATE || \
(txtype) == TX_SETATTR || \
(txtype) == TX_ACL_V0 || \
(txtype) == TX_ACL || \
(txtype) == TX_WRITE2)
/*
* The number of dnode slots consumed by the object is stored in the 8
* unused upper bits of the object ID. We subtract 1 from the value
* stored on disk for compatibility with implementations that don't
* support large dnodes. The slot count for a single-slot dnode will
* contain 0 for those bits to preserve the log record format for
* "small" dnodes.
*/
#define LR_FOID_GET_SLOTS(oid) (BF64_GET((oid), 56, 8) + 1)
#define LR_FOID_SET_SLOTS(oid, x) BF64_SET((oid), 56, 8, (x) - 1)
#define LR_FOID_GET_OBJ(oid) BF64_GET((oid), 0, DN_MAX_OBJECT_SHIFT)
#define LR_FOID_SET_OBJ(oid, x) BF64_SET((oid), 0, DN_MAX_OBJECT_SHIFT, (x))
/*
* Format of log records.
* The fields are carefully defined to allow them to be aligned
* and sized the same on sparc & intel architectures.
* Each log record has a common structure at the beginning.
*
* The log record on disk (lrc_seq) holds the sequence number of all log
* records which is used to ensure we don't replay the same record.
*/
typedef struct { /* common log record header */
uint64_t lrc_txtype; /* intent log transaction type */
uint64_t lrc_reclen; /* transaction record length */
uint64_t lrc_txg; /* dmu transaction group number */
uint64_t lrc_seq; /* see comment above */
} lr_t;
/*
* Common start of all out-of-order record types (TX_OOO() above).
*/
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_foid; /* object id */
} lr_ooo_t;
/*
* Handle option extended vattr attributes.
*
* Whenever new attributes are added the version number
* will need to be updated as will code in
* zfs_log.c and zfs_replay.c
*/
typedef struct {
uint32_t lr_attr_masksize; /* number of elements in array */
uint32_t lr_attr_bitmap; /* First entry of array */
/* remainder of array and any additional fields */
} lr_attr_t;
/*
* log record for creates without optional ACL.
* This log record does support optional xvattr_t attributes.
*/
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_doid; /* object id of directory */
uint64_t lr_foid; /* object id of created file object */
uint64_t lr_mode; /* mode of object */
uint64_t lr_uid; /* uid of object */
uint64_t lr_gid; /* gid of object */
uint64_t lr_gen; /* generation (txg of creation) */
uint64_t lr_crtime[2]; /* creation time */
uint64_t lr_rdev; /* rdev of object to create */
/* name of object to create follows this */
/* for symlinks, link content follows name */
/* for creates with xvattr data, the name follows the xvattr info */
} lr_create_t;
/*
* FUID ACL record will be an array of ACEs from the original ACL.
* If this array includes ephemeral IDs, the record will also include
* an array of log-specific FUIDs to replace the ephemeral IDs.
* Only one copy of each unique domain will be present, so the log-specific
* FUIDs will use an index into a compressed domain table. On replay this
* information will be used to construct real FUIDs (and bypass idmap,
* since it may not be available).
*/
/*
* Log record for creates with optional ACL
* This log record is also used for recording any FUID
* information needed for replaying the create. If the
* file doesn't have any actual ACEs then the lr_aclcnt
* would be zero.
*
* After lr_acl_flags, there are a lr_acl_bytes number of variable sized ace's.
* If create is also setting xvattr's, then acl data follows xvattr.
* If ACE FUIDs are needed then they will follow the xvattr_t. Following
* the FUIDs will be the domain table information. The FUIDs for the owner
* and group will be in lr_create. Name follows ACL data.
*/
typedef struct {
lr_create_t lr_create; /* common create portion */
uint64_t lr_aclcnt; /* number of ACEs in ACL */
uint64_t lr_domcnt; /* number of unique domains */
uint64_t lr_fuidcnt; /* number of real fuids */
uint64_t lr_acl_bytes; /* number of bytes in ACL */
uint64_t lr_acl_flags; /* ACL flags */
} lr_acl_create_t;
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_doid; /* obj id of directory */
/* name of object to remove follows this */
} lr_remove_t;
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_doid; /* obj id of directory */
uint64_t lr_link_obj; /* obj id of link */
/* name of object to link follows this */
} lr_link_t;
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_sdoid; /* obj id of source directory */
uint64_t lr_tdoid; /* obj id of target directory */
/* 2 strings: names of source and destination follow this */
} lr_rename_t;
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_foid; /* file object to write */
uint64_t lr_offset; /* offset to write to */
uint64_t lr_length; /* user data length to write */
uint64_t lr_blkoff; /* no longer used */
blkptr_t lr_blkptr; /* spa block pointer for replay */
/* write data will follow for small writes */
} lr_write_t;
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_foid; /* object id of file to truncate */
uint64_t lr_offset; /* offset to truncate from */
uint64_t lr_length; /* length to truncate */
} lr_truncate_t;
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_foid; /* file object to change attributes */
uint64_t lr_mask; /* mask of attributes to set */
uint64_t lr_mode; /* mode to set */
uint64_t lr_uid; /* uid to set */
uint64_t lr_gid; /* gid to set */
uint64_t lr_size; /* size to set */
uint64_t lr_atime[2]; /* access time */
uint64_t lr_mtime[2]; /* modification time */
/* optional attribute lr_attr_t may be here */
} lr_setattr_t;
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_foid; /* obj id of file */
uint64_t lr_aclcnt; /* number of acl entries */
/* lr_aclcnt number of ace_t entries follow this */
} lr_acl_v0_t;
typedef struct {
lr_t lr_common; /* common portion of log record */
uint64_t lr_foid; /* obj id of file */
uint64_t lr_aclcnt; /* number of ACEs in ACL */
uint64_t lr_domcnt; /* number of unique domains */
uint64_t lr_fuidcnt; /* number of real fuids */
uint64_t lr_acl_bytes; /* number of bytes in ACL */
uint64_t lr_acl_flags; /* ACL flags */
/* lr_acl_bytes number of variable sized ace's follows */
} lr_acl_t;
/*
* ZIL structure definitions, interface function prototype and globals.
*/
/*
* Writes are handled in three different ways:
*
* WR_INDIRECT:
* In this mode, if we need to commit the write later, then the block
* is immediately written into the file system (using dmu_sync),
* and a pointer to the block is put into the log record.
* When the txg commits the block is linked in.
* This saves additionally writing the data into the log record.
* There are a few requirements for this to occur:
* - write is greater than zfs/zvol_immediate_write_sz
* - not using slogs (as slogs are assumed to always be faster
* than writing into the main pool)
* - the write occupies only one block
* WR_COPIED:
* If we know we'll immediately be committing the
* transaction (O_SYNC or O_DSYNC), then we allocate a larger
* log record here for the data and copy the data in.
* WR_NEED_COPY:
* Otherwise we don't allocate a buffer, and *if* we need to
* flush the write later then a buffer is allocated and
* we retrieve the data using the dmu.
*/
typedef enum {
WR_INDIRECT, /* indirect - a large write (dmu_sync() data */
/* and put blkptr in log, rather than actual data) */
WR_COPIED, /* immediate - data is copied into lr_write_t */
WR_NEED_COPY, /* immediate - data needs to be copied if pushed */
WR_NUM_STATES /* number of states */
} itx_wr_state_t;
typedef void (*zil_callback_t)(void *data);
typedef struct itx {
list_node_t itx_node; /* linkage on zl_itx_list */
void *itx_private; /* type-specific opaque data */
itx_wr_state_t itx_wr_state; /* write state */
uint8_t itx_sync; /* synchronous transaction */
zil_callback_t itx_callback; /* Called when the itx is persistent */
void *itx_callback_data; /* User data for the callback */
size_t itx_size; /* allocated itx structure size */
uint64_t itx_oid; /* object id */
uint64_t itx_gen; /* gen number for zfs_get_data */
lr_t itx_lr; /* common part of log record */
/* followed by type-specific part of lr_xx_t and its immediate data */
} itx_t;
/*
* Used for zil kstat.
*/
typedef struct zil_stats {
/*
* Number of times a ZIL commit (e.g. fsync) has been requested.
*/
kstat_named_t zil_commit_count;
/*
* Number of times the ZIL has been flushed to stable storage.
* This is less than zil_commit_count when commits are "merged"
* (see the documentation above zil_commit()).
*/
kstat_named_t zil_commit_writer_count;
/*
* Number of transactions (reads, writes, renames, etc.)
* that have been committed.
*/
kstat_named_t zil_itx_count;
/*
* See the documentation for itx_wr_state_t above.
* Note that "bytes" accumulates the length of the transactions
* (i.e. data), not the actual log record sizes.
*/
kstat_named_t zil_itx_indirect_count;
kstat_named_t zil_itx_indirect_bytes;
kstat_named_t zil_itx_copied_count;
kstat_named_t zil_itx_copied_bytes;
kstat_named_t zil_itx_needcopy_count;
kstat_named_t zil_itx_needcopy_bytes;
/*
* Transactions which have been allocated to the "normal"
* (i.e. not slog) storage pool. Note that "bytes" accumulate
* the actual log record sizes - which do not include the actual
* data in case of indirect writes.
*/
kstat_named_t zil_itx_metaslab_normal_count;
kstat_named_t zil_itx_metaslab_normal_bytes;
/*
* Transactions which have been allocated to the "slog" storage pool.
* If there are no separate log devices, this is the same as the
* "normal" pool.
*/
kstat_named_t zil_itx_metaslab_slog_count;
kstat_named_t zil_itx_metaslab_slog_bytes;
} zil_stats_t;
-extern zil_stats_t zil_stats;
-
#define ZIL_STAT_INCR(stat, val) \
atomic_add_64(&zil_stats.stat.value.ui64, (val));
#define ZIL_STAT_BUMP(stat) \
ZIL_STAT_INCR(stat, 1);
typedef int zil_parse_blk_func_t(zilog_t *zilog, const blkptr_t *bp, void *arg,
uint64_t txg);
typedef int zil_parse_lr_func_t(zilog_t *zilog, const lr_t *lr, void *arg,
uint64_t txg);
typedef int zil_replay_func_t(void *arg1, void *arg2, boolean_t byteswap);
typedef int zil_get_data_t(void *arg, uint64_t arg2, lr_write_t *lr, char *dbuf,
struct lwb *lwb, zio_t *zio);
extern int zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func,
zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg,
boolean_t decrypt);
extern void zil_init(void);
extern void zil_fini(void);
extern zilog_t *zil_alloc(objset_t *os, zil_header_t *zh_phys);
extern void zil_free(zilog_t *zilog);
extern zilog_t *zil_open(objset_t *os, zil_get_data_t *get_data);
extern void zil_close(zilog_t *zilog);
extern void zil_replay(objset_t *os, void *arg,
- zil_replay_func_t *replay_func[TX_MAX_TYPE]);
+ zil_replay_func_t *const replay_func[TX_MAX_TYPE]);
extern boolean_t zil_replaying(zilog_t *zilog, dmu_tx_t *tx);
extern void zil_destroy(zilog_t *zilog, boolean_t keep_first);
extern void zil_destroy_sync(zilog_t *zilog, dmu_tx_t *tx);
extern itx_t *zil_itx_create(uint64_t txtype, size_t lrsize);
extern void zil_itx_destroy(itx_t *itx);
extern void zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx);
extern void zil_async_to_sync(zilog_t *zilog, uint64_t oid);
extern void zil_commit(zilog_t *zilog, uint64_t oid);
extern void zil_commit_impl(zilog_t *zilog, uint64_t oid);
extern void zil_remove_async(zilog_t *zilog, uint64_t oid);
extern int zil_reset(const char *osname, void *txarg);
extern int zil_claim(struct dsl_pool *dp,
struct dsl_dataset *ds, void *txarg);
extern int zil_check_log_chain(struct dsl_pool *dp,
struct dsl_dataset *ds, void *tx);
extern void zil_sync(zilog_t *zilog, dmu_tx_t *tx);
extern void zil_clean(zilog_t *zilog, uint64_t synced_txg);
extern int zil_suspend(const char *osname, void **cookiep);
extern void zil_resume(void *cookie);
extern void zil_lwb_add_block(struct lwb *lwb, const blkptr_t *bp);
extern void zil_lwb_add_txg(struct lwb *lwb, uint64_t txg);
extern int zil_bp_tree_add(zilog_t *zilog, const blkptr_t *bp);
extern void zil_set_sync(zilog_t *zilog, uint64_t syncval);
extern void zil_set_logbias(zilog_t *zilog, uint64_t slogval);
extern uint64_t zil_max_copied_data(zilog_t *zilog);
extern uint64_t zil_max_log_data(zilog_t *zilog);
extern int zil_replay_disable;
#ifdef __cplusplus
}
#endif
#endif /* _SYS_ZIL_H */
diff --git a/include/sys/zio.h b/include/sys/zio.h
index 121b58dea58e..07135d1e2a07 100644
--- a/include/sys/zio.h
+++ b/include/sys/zio.h
@@ -1,690 +1,690 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
* Copyright 2016 Toomas Soome <tsoome@me.com>
* Copyright (c) 2019, Allan Jude
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019-2020, Michael Niewöhner
*/
#ifndef _ZIO_H
#define _ZIO_H
#include <sys/zio_priority.h>
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/txg.h>
#include <sys/avl.h>
#include <sys/fs/zfs.h>
#include <sys/zio_impl.h>
#ifdef __cplusplus
extern "C" {
#endif
/*
* Embedded checksum
*/
#define ZEC_MAGIC 0x210da7ab10c7a11ULL
typedef struct zio_eck {
uint64_t zec_magic; /* for validation, endianness */
zio_cksum_t zec_cksum; /* 256-bit checksum */
} zio_eck_t;
/*
* Gang block headers are self-checksumming and contain an array
* of block pointers.
*/
#define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
#define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
sizeof (zio_eck_t)) / sizeof (blkptr_t))
#define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
sizeof (zio_eck_t) - \
(SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
sizeof (uint64_t))
typedef struct zio_gbh {
blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
uint64_t zg_filler[SPA_GBH_FILLER];
zio_eck_t zg_tail;
} zio_gbh_phys_t;
enum zio_checksum {
ZIO_CHECKSUM_INHERIT = 0,
ZIO_CHECKSUM_ON,
ZIO_CHECKSUM_OFF,
ZIO_CHECKSUM_LABEL,
ZIO_CHECKSUM_GANG_HEADER,
ZIO_CHECKSUM_ZILOG,
ZIO_CHECKSUM_FLETCHER_2,
ZIO_CHECKSUM_FLETCHER_4,
ZIO_CHECKSUM_SHA256,
ZIO_CHECKSUM_ZILOG2,
ZIO_CHECKSUM_NOPARITY,
ZIO_CHECKSUM_SHA512,
ZIO_CHECKSUM_SKEIN,
ZIO_CHECKSUM_EDONR,
ZIO_CHECKSUM_FUNCTIONS
};
/*
* The number of "legacy" compression functions which can be set on individual
* objects.
*/
#define ZIO_CHECKSUM_LEGACY_FUNCTIONS ZIO_CHECKSUM_ZILOG2
#define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4
#define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
#define ZIO_CHECKSUM_MASK 0xffULL
#define ZIO_CHECKSUM_VERIFY (1U << 8)
#define ZIO_DEDUPCHECKSUM ZIO_CHECKSUM_SHA256
/* macros defining encryption lengths */
#define ZIO_OBJSET_MAC_LEN 32
#define ZIO_DATA_IV_LEN 12
#define ZIO_DATA_SALT_LEN 8
#define ZIO_DATA_MAC_LEN 16
/*
* The number of "legacy" compression functions which can be set on individual
* objects.
*/
#define ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4
/*
* The meaning of "compress = on" selected by the compression features enabled
* on a given pool.
*/
#define ZIO_COMPRESS_LEGACY_ON_VALUE ZIO_COMPRESS_LZJB
#define ZIO_COMPRESS_LZ4_ON_VALUE ZIO_COMPRESS_LZ4
#define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
#define BOOTFS_COMPRESS_VALID(compress) \
((compress) == ZIO_COMPRESS_LZJB || \
(compress) == ZIO_COMPRESS_LZ4 || \
(compress) == ZIO_COMPRESS_GZIP_1 || \
(compress) == ZIO_COMPRESS_GZIP_2 || \
(compress) == ZIO_COMPRESS_GZIP_3 || \
(compress) == ZIO_COMPRESS_GZIP_4 || \
(compress) == ZIO_COMPRESS_GZIP_5 || \
(compress) == ZIO_COMPRESS_GZIP_6 || \
(compress) == ZIO_COMPRESS_GZIP_7 || \
(compress) == ZIO_COMPRESS_GZIP_8 || \
(compress) == ZIO_COMPRESS_GZIP_9 || \
(compress) == ZIO_COMPRESS_ZLE || \
(compress) == ZIO_COMPRESS_ZSTD || \
(compress) == ZIO_COMPRESS_ON || \
(compress) == ZIO_COMPRESS_OFF)
#define ZIO_COMPRESS_ALGO(x) (x & SPA_COMPRESSMASK)
#define ZIO_COMPRESS_LEVEL(x) ((x & ~SPA_COMPRESSMASK) >> SPA_COMPRESSBITS)
#define ZIO_COMPRESS_RAW(type, level) (type | ((level) << SPA_COMPRESSBITS))
#define ZIO_COMPLEVEL_ZSTD(level) \
ZIO_COMPRESS_RAW(ZIO_COMPRESS_ZSTD, level)
#define ZIO_FAILURE_MODE_WAIT 0
#define ZIO_FAILURE_MODE_CONTINUE 1
#define ZIO_FAILURE_MODE_PANIC 2
typedef enum zio_suspend_reason {
ZIO_SUSPEND_NONE = 0,
ZIO_SUSPEND_IOERR,
ZIO_SUSPEND_MMP,
} zio_suspend_reason_t;
enum zio_flag {
/*
* Flags inherited by gang, ddt, and vdev children,
* and that must be equal for two zios to aggregate
*/
ZIO_FLAG_DONT_AGGREGATE = 1U << 0,
ZIO_FLAG_IO_REPAIR = 1U << 1,
ZIO_FLAG_SELF_HEAL = 1U << 2,
ZIO_FLAG_RESILVER = 1U << 3,
ZIO_FLAG_SCRUB = 1U << 4,
ZIO_FLAG_SCAN_THREAD = 1U << 5,
ZIO_FLAG_PHYSICAL = 1U << 6,
#define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1)
/*
* Flags inherited by ddt, gang, and vdev children.
*/
ZIO_FLAG_CANFAIL = 1U << 7, /* must be first for INHERIT */
ZIO_FLAG_SPECULATIVE = 1U << 8,
ZIO_FLAG_CONFIG_WRITER = 1U << 9,
ZIO_FLAG_DONT_RETRY = 1U << 10,
ZIO_FLAG_DONT_CACHE = 1U << 11,
ZIO_FLAG_NODATA = 1U << 12,
ZIO_FLAG_INDUCE_DAMAGE = 1U << 13,
ZIO_FLAG_IO_ALLOCATING = 1U << 14,
#define ZIO_FLAG_DDT_INHERIT (ZIO_FLAG_IO_RETRY - 1)
#define ZIO_FLAG_GANG_INHERIT (ZIO_FLAG_IO_RETRY - 1)
/*
* Flags inherited by vdev children.
*/
ZIO_FLAG_IO_RETRY = 1U << 15, /* must be first for INHERIT */
ZIO_FLAG_PROBE = 1U << 16,
ZIO_FLAG_TRYHARD = 1U << 17,
ZIO_FLAG_OPTIONAL = 1U << 18,
#define ZIO_FLAG_VDEV_INHERIT (ZIO_FLAG_DONT_QUEUE - 1)
/*
* Flags not inherited by any children.
*/
ZIO_FLAG_DONT_QUEUE = 1U << 19, /* must be first for INHERIT */
ZIO_FLAG_DONT_PROPAGATE = 1U << 20,
ZIO_FLAG_IO_BYPASS = 1U << 21,
ZIO_FLAG_IO_REWRITE = 1U << 22,
ZIO_FLAG_RAW_COMPRESS = 1U << 23,
ZIO_FLAG_RAW_ENCRYPT = 1U << 24,
ZIO_FLAG_GANG_CHILD = 1U << 25,
ZIO_FLAG_DDT_CHILD = 1U << 26,
ZIO_FLAG_GODFATHER = 1U << 27,
ZIO_FLAG_NOPWRITE = 1U << 28,
ZIO_FLAG_REEXECUTED = 1U << 29,
ZIO_FLAG_DELEGATED = 1U << 30,
ZIO_FLAG_FASTWRITE = 1U << 31,
};
#define ZIO_FLAG_MUSTSUCCEED 0
#define ZIO_FLAG_RAW (ZIO_FLAG_RAW_COMPRESS | ZIO_FLAG_RAW_ENCRYPT)
#define ZIO_DDT_CHILD_FLAGS(zio) \
(((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \
ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
#define ZIO_GANG_CHILD_FLAGS(zio) \
(((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \
ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
#define ZIO_VDEV_CHILD_FLAGS(zio) \
(((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \
ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_CANFAIL)
#define ZIO_CHILD_BIT(x) (1U << (x))
#define ZIO_CHILD_BIT_IS_SET(val, x) ((val) & (1U << (x)))
enum zio_child {
ZIO_CHILD_VDEV = 0,
ZIO_CHILD_GANG,
ZIO_CHILD_DDT,
ZIO_CHILD_LOGICAL,
ZIO_CHILD_TYPES
};
#define ZIO_CHILD_VDEV_BIT ZIO_CHILD_BIT(ZIO_CHILD_VDEV)
#define ZIO_CHILD_GANG_BIT ZIO_CHILD_BIT(ZIO_CHILD_GANG)
#define ZIO_CHILD_DDT_BIT ZIO_CHILD_BIT(ZIO_CHILD_DDT)
#define ZIO_CHILD_LOGICAL_BIT ZIO_CHILD_BIT(ZIO_CHILD_LOGICAL)
#define ZIO_CHILD_ALL_BITS \
(ZIO_CHILD_VDEV_BIT | ZIO_CHILD_GANG_BIT | \
ZIO_CHILD_DDT_BIT | ZIO_CHILD_LOGICAL_BIT)
enum zio_wait_type {
ZIO_WAIT_READY = 0,
ZIO_WAIT_DONE,
ZIO_WAIT_TYPES
};
typedef void zio_done_func_t(zio_t *zio);
extern int zio_exclude_metadata;
extern int zio_dva_throttle_enabled;
-extern const char *zio_type_name[ZIO_TYPES];
+extern const char *const zio_type_name[ZIO_TYPES];
/*
* A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
* identifies any block in the pool. By convention, the meta-objset (MOS)
* is objset 0, and the meta-dnode is object 0. This covers all blocks
* except root blocks and ZIL blocks, which are defined as follows:
*
* Root blocks (objset_phys_t) are object 0, level -1: <objset, 0, -1, 0>.
* ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
* dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
* dnode visit bookmarks are <objset, object id of dnode, -3, 0>.
*
* Note: this structure is called a bookmark because its original purpose
* was to remember where to resume a pool-wide traverse.
*
* Note: this structure is passed between userland and the kernel, and is
* stored on disk (by virtue of being incorporated into other on-disk
* structures, e.g. dsl_scan_phys_t).
*/
struct zbookmark_phys {
uint64_t zb_objset;
uint64_t zb_object;
int64_t zb_level;
uint64_t zb_blkid;
};
#define SET_BOOKMARK(zb, objset, object, level, blkid) \
{ \
(zb)->zb_objset = objset; \
(zb)->zb_object = object; \
(zb)->zb_level = level; \
(zb)->zb_blkid = blkid; \
}
#define ZB_DESTROYED_OBJSET (-1ULL)
#define ZB_ROOT_OBJECT (0ULL)
#define ZB_ROOT_LEVEL (-1LL)
#define ZB_ROOT_BLKID (0ULL)
#define ZB_ZIL_OBJECT (0ULL)
#define ZB_ZIL_LEVEL (-2LL)
#define ZB_DNODE_LEVEL (-3LL)
#define ZB_DNODE_BLKID (0ULL)
#define ZB_IS_ZERO(zb) \
((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \
(zb)->zb_level == 0 && (zb)->zb_blkid == 0)
#define ZB_IS_ROOT(zb) \
((zb)->zb_object == ZB_ROOT_OBJECT && \
(zb)->zb_level == ZB_ROOT_LEVEL && \
(zb)->zb_blkid == ZB_ROOT_BLKID)
typedef struct zio_prop {
enum zio_checksum zp_checksum;
enum zio_compress zp_compress;
uint8_t zp_complevel;
dmu_object_type_t zp_type;
uint8_t zp_level;
uint8_t zp_copies;
boolean_t zp_dedup;
boolean_t zp_dedup_verify;
boolean_t zp_nopwrite;
boolean_t zp_encrypt;
boolean_t zp_byteorder;
uint8_t zp_salt[ZIO_DATA_SALT_LEN];
uint8_t zp_iv[ZIO_DATA_IV_LEN];
uint8_t zp_mac[ZIO_DATA_MAC_LEN];
uint32_t zp_zpl_smallblk;
} zio_prop_t;
typedef struct zio_cksum_report zio_cksum_report_t;
typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
const abd_t *good_data);
typedef void zio_cksum_free_f(void *cbdata, size_t size);
struct zio_bad_cksum; /* defined in zio_checksum.h */
struct dnode_phys;
struct abd;
struct zio_cksum_report {
struct zio_cksum_report *zcr_next;
nvlist_t *zcr_ereport;
nvlist_t *zcr_detector;
void *zcr_cbdata;
size_t zcr_cbinfo; /* passed to zcr_free() */
uint64_t zcr_sector;
uint64_t zcr_align;
uint64_t zcr_length;
zio_cksum_finish_f *zcr_finish;
zio_cksum_free_f *zcr_free;
/* internal use only */
struct zio_bad_cksum *zcr_ckinfo; /* information from failure */
};
typedef struct zio_vsd_ops {
zio_done_func_t *vsd_free;
} zio_vsd_ops_t;
typedef struct zio_gang_node {
zio_gbh_phys_t *gn_gbh;
struct zio_gang_node *gn_child[SPA_GBH_NBLKPTRS];
} zio_gang_node_t;
typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp,
zio_gang_node_t *gn, struct abd *data, uint64_t offset);
typedef void zio_transform_func_t(zio_t *zio, struct abd *data, uint64_t size);
typedef struct zio_transform {
struct abd *zt_orig_abd;
uint64_t zt_orig_size;
uint64_t zt_bufsize;
zio_transform_func_t *zt_transform;
struct zio_transform *zt_next;
} zio_transform_t;
typedef zio_t *zio_pipe_stage_t(zio_t *zio);
/*
* The io_reexecute flags are distinct from io_flags because the child must
* be able to propagate them to the parent. The normal io_flags are local
* to the zio, not protected by any lock, and not modifiable by children;
* the reexecute flags are protected by io_lock, modifiable by children,
* and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
*/
#define ZIO_REEXECUTE_NOW 0x01
#define ZIO_REEXECUTE_SUSPEND 0x02
/*
* The io_trim flags are used to specify the type of TRIM to perform. They
* only apply to ZIO_TYPE_TRIM zios are distinct from io_flags.
*/
enum trim_flag {
ZIO_TRIM_SECURE = 1U << 0,
};
typedef struct zio_alloc_list {
list_t zal_list;
uint64_t zal_size;
} zio_alloc_list_t;
typedef struct zio_link {
zio_t *zl_parent;
zio_t *zl_child;
list_node_t zl_parent_node;
list_node_t zl_child_node;
} zio_link_t;
struct zio {
/* Core information about this I/O */
zbookmark_phys_t io_bookmark;
zio_prop_t io_prop;
zio_type_t io_type;
enum zio_child io_child_type;
enum trim_flag io_trim_flags;
int io_cmd;
zio_priority_t io_priority;
uint8_t io_reexecute;
uint8_t io_state[ZIO_WAIT_TYPES];
uint64_t io_txg;
spa_t *io_spa;
blkptr_t *io_bp;
blkptr_t *io_bp_override;
blkptr_t io_bp_copy;
list_t io_parent_list;
list_t io_child_list;
zio_t *io_logical;
zio_transform_t *io_transform_stack;
/* Callback info */
zio_done_func_t *io_ready;
zio_done_func_t *io_children_ready;
zio_done_func_t *io_physdone;
zio_done_func_t *io_done;
void *io_private;
int64_t io_prev_space_delta; /* DMU private */
blkptr_t io_bp_orig;
/* io_lsize != io_orig_size iff this is a raw write */
uint64_t io_lsize;
/* Data represented by this I/O */
struct abd *io_abd;
struct abd *io_orig_abd;
uint64_t io_size;
uint64_t io_orig_size;
/* Stuff for the vdev stack */
vdev_t *io_vd;
void *io_vsd;
const zio_vsd_ops_t *io_vsd_ops;
metaslab_class_t *io_metaslab_class; /* dva throttle class */
uint64_t io_offset;
hrtime_t io_timestamp; /* submitted at */
hrtime_t io_queued_timestamp;
hrtime_t io_target_timestamp;
hrtime_t io_delta; /* vdev queue service delta */
hrtime_t io_delay; /* Device access time (disk or */
/* file). */
avl_node_t io_queue_node;
avl_node_t io_offset_node;
avl_node_t io_alloc_node;
zio_alloc_list_t io_alloc_list;
/* Internal pipeline state */
enum zio_flag io_flags;
enum zio_stage io_stage;
enum zio_stage io_pipeline;
enum zio_flag io_orig_flags;
enum zio_stage io_orig_stage;
enum zio_stage io_orig_pipeline;
enum zio_stage io_pipeline_trace;
int io_error;
int io_child_error[ZIO_CHILD_TYPES];
uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
uint64_t io_child_count;
uint64_t io_phys_children;
uint64_t io_parent_count;
uint64_t *io_stall;
zio_t *io_gang_leader;
zio_gang_node_t *io_gang_tree;
void *io_executor;
void *io_waiter;
void *io_bio;
kmutex_t io_lock;
kcondvar_t io_cv;
int io_allocator;
/* FMA state */
zio_cksum_report_t *io_cksum_report;
uint64_t io_ena;
/* Taskq dispatching state */
taskq_ent_t io_tqent;
};
enum blk_verify_flag {
BLK_VERIFY_ONLY,
BLK_VERIFY_LOG,
BLK_VERIFY_HALT
};
extern int zio_bookmark_compare(const void *, const void *);
extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
zio_done_func_t *done, void *priv, enum zio_flag flags);
extern zio_t *zio_root(spa_t *spa,
zio_done_func_t *done, void *priv, enum zio_flag flags);
extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
struct abd *data, uint64_t lsize, zio_done_func_t *done, void *priv,
zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb);
extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp,
zio_done_func_t *ready, zio_done_func_t *children_ready,
zio_done_func_t *physdone, zio_done_func_t *done,
void *priv, zio_priority_t priority, enum zio_flag flags,
const zbookmark_phys_t *zb);
extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
struct abd *data, uint64_t size, zio_done_func_t *done, void *priv,
zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb);
extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
boolean_t nopwrite);
extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
const blkptr_t *bp,
zio_done_func_t *done, void *priv, enum zio_flag flags);
extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
zio_done_func_t *done, void *priv, enum zio_flag flags);
extern zio_t *zio_trim(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
zio_done_func_t *done, void *priv, zio_priority_t priority,
enum zio_flag flags, enum trim_flag trim_flags);
extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
uint64_t size, struct abd *data, int checksum,
zio_done_func_t *done, void *priv, zio_priority_t priority,
enum zio_flag flags, boolean_t labels);
extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
uint64_t size, struct abd *data, int checksum,
zio_done_func_t *done, void *priv, zio_priority_t priority,
enum zio_flag flags, boolean_t labels);
extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
const blkptr_t *bp, enum zio_flag flags);
extern int zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg,
blkptr_t *new_bp, uint64_t size, boolean_t *slog);
extern void zio_flush(zio_t *zio, vdev_t *vd);
extern void zio_shrink(zio_t *zio, uint64_t size);
extern int zio_wait(zio_t *zio);
extern void zio_nowait(zio_t *zio);
extern void zio_execute(void *zio);
extern void zio_interrupt(void *zio);
extern void zio_delay_init(zio_t *zio);
extern void zio_delay_interrupt(zio_t *zio);
extern void zio_deadman(zio_t *zio, char *tag);
extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **);
extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **);
extern zio_t *zio_unique_parent(zio_t *cio);
extern void zio_add_child(zio_t *pio, zio_t *cio);
extern void *zio_buf_alloc(size_t size);
extern void zio_buf_free(void *buf, size_t size);
extern void *zio_data_buf_alloc(size_t size);
extern void zio_data_buf_free(void *buf, size_t size);
extern void zio_push_transform(zio_t *zio, struct abd *abd, uint64_t size,
uint64_t bufsize, zio_transform_func_t *transform);
extern void zio_pop_transforms(zio_t *zio);
extern void zio_resubmit_stage_async(void *);
extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
uint64_t offset, struct abd *data, uint64_t size, int type,
zio_priority_t priority, enum zio_flag flags,
zio_done_func_t *done, void *priv);
extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset,
struct abd *data, uint64_t size, zio_type_t type, zio_priority_t priority,
enum zio_flag flags, zio_done_func_t *done, void *priv);
extern void zio_vdev_io_bypass(zio_t *zio);
extern void zio_vdev_io_reissue(zio_t *zio);
extern void zio_vdev_io_redone(zio_t *zio);
extern void zio_change_priority(zio_t *pio, zio_priority_t priority);
extern void zio_checksum_verified(zio_t *zio);
extern int zio_worst_error(int e1, int e2);
extern enum zio_checksum zio_checksum_select(enum zio_checksum child,
enum zio_checksum parent);
extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa,
enum zio_checksum child, enum zio_checksum parent);
extern enum zio_compress zio_compress_select(spa_t *spa,
enum zio_compress child, enum zio_compress parent);
extern uint8_t zio_complevel_select(spa_t *spa, enum zio_compress compress,
uint8_t child, uint8_t parent);
extern void zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t);
extern int zio_resume(spa_t *spa);
extern void zio_resume_wait(spa_t *spa);
extern boolean_t zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp,
boolean_t config_held, enum blk_verify_flag blk_verify);
/*
* Initial setup and teardown.
*/
extern void zio_init(void);
extern void zio_fini(void);
/*
* Fault injection
*/
struct zinject_record;
extern uint32_t zio_injection_enabled;
extern int zio_inject_fault(char *name, int flags, int *id,
struct zinject_record *record);
extern int zio_inject_list_next(int *id, char *name, size_t buflen,
struct zinject_record *record);
extern int zio_clear_fault(int id);
extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type);
extern int zio_handle_decrypt_injection(spa_t *spa, const zbookmark_phys_t *zb,
uint64_t type, int error);
extern int zio_handle_fault_injection(zio_t *zio, int error);
extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
extern int zio_handle_device_injections(vdev_t *vd, zio_t *zio, int err1,
int err2);
extern int zio_handle_label_injection(zio_t *zio, int error);
extern void zio_handle_ignored_writes(zio_t *zio);
extern hrtime_t zio_handle_io_delay(zio_t *zio);
/*
* Checksum ereport functions
*/
extern int zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd,
const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset,
uint64_t length, struct zio_bad_cksum *info);
extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
const abd_t *good_data, const abd_t *bad_data, boolean_t drop_if_identical);
extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
/* If we have the good data in hand, this function can be used */
extern int zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset,
uint64_t length, const abd_t *good_data, const abd_t *bad_data,
struct zio_bad_cksum *info);
void zio_vsd_default_cksum_report(zio_t *zio, zio_cksum_report_t *zcr);
extern void zfs_ereport_snapshot_post(const char *subclass, spa_t *spa,
const char *name);
/* Called from spa_sync(), but primarily an injection handler */
extern void spa_handle_ignored_writes(spa_t *spa);
/* zbookmark_phys functions */
boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp,
const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2,
uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);
#ifdef __cplusplus
}
#endif
#endif /* _ZIO_H */
diff --git a/include/sys/zio_crypt.h b/include/sys/zio_crypt.h
index d7a63d69582d..f1edd76f0d8e 100644
--- a/include/sys/zio_crypt.h
+++ b/include/sys/zio_crypt.h
@@ -1,160 +1,160 @@
/*
* CDDL HEADER START
*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2017, Datto, Inc. All rights reserved.
*/
#ifndef _SYS_ZIO_CRYPT_H
#define _SYS_ZIO_CRYPT_H
#include <sys/dmu.h>
#include <sys/zfs_refcount.h>
#if defined(__FreeBSD__) && defined(_KERNEL)
#include <sys/freebsd_crypto.h>
#else
#include <sys/crypto/api.h>
#endif /* __FreeBSD__ */
#include <sys/nvpair.h>
#include <sys/avl.h>
#include <sys/zio.h>
/* forward declarations */
struct zbookmark_phys;
#define WRAPPING_KEY_LEN 32
#define WRAPPING_IV_LEN ZIO_DATA_IV_LEN
#define WRAPPING_MAC_LEN ZIO_DATA_MAC_LEN
#define MASTER_KEY_MAX_LEN 32
#define SHA512_HMAC_KEYLEN 64
#define ZIO_CRYPT_KEY_CURRENT_VERSION 1ULL
typedef enum zio_crypt_type {
ZC_TYPE_NONE = 0,
ZC_TYPE_CCM,
ZC_TYPE_GCM
} zio_crypt_type_t;
/* table of supported crypto algorithms, modes and keylengths. */
typedef struct zio_crypt_info {
/* mechanism name, needed by ICP */
#if defined(__FreeBSD__) && defined(_KERNEL)
/*
* I've deliberately used a different name here, to catch
* ICP-using code.
*/
const char *ci_algname;
#else
crypto_mech_name_t ci_mechname;
#endif
/* cipher mode type (GCM, CCM) */
zio_crypt_type_t ci_crypt_type;
/* length of the encryption key */
size_t ci_keylen;
/* human-readable name of the encryption algorithm */
char *ci_name;
} zio_crypt_info_t;
-extern zio_crypt_info_t zio_crypt_table[ZIO_CRYPT_FUNCTIONS];
+extern const zio_crypt_info_t zio_crypt_table[ZIO_CRYPT_FUNCTIONS];
/* in memory representation of an unwrapped key that is loaded into memory */
typedef struct zio_crypt_key {
/* encryption algorithm */
uint64_t zk_crypt;
/* on-disk format version */
uint64_t zk_version;
/* GUID for uniquely identifying this key. Not encrypted on disk. */
uint64_t zk_guid;
/* buffer for master key */
uint8_t zk_master_keydata[MASTER_KEY_MAX_LEN];
/* buffer for hmac key */
uint8_t zk_hmac_keydata[SHA512_HMAC_KEYLEN];
/* buffer for current encryption key derived from master key */
uint8_t zk_current_keydata[MASTER_KEY_MAX_LEN];
/* current 64 bit salt for deriving an encryption key */
uint8_t zk_salt[ZIO_DATA_SALT_LEN];
/* count of how many times the current salt has been used */
uint64_t zk_salt_count;
/* illumos crypto api current encryption key */
crypto_key_t zk_current_key;
#if defined(__FreeBSD__) && defined(_KERNEL)
/* Session for current encryption key. Must always be set */
freebsd_crypt_session_t zk_session;
#else
/* template of current encryption key for illumos crypto api */
crypto_ctx_template_t zk_current_tmpl;
#endif
/* illumos crypto api current hmac key */
crypto_key_t zk_hmac_key;
/* template of hmac key for illumos crypto api */
crypto_ctx_template_t zk_hmac_tmpl;
/* lock for changing the salt and dependent values */
krwlock_t zk_salt_lock;
} zio_crypt_key_t;
void zio_crypt_key_destroy(zio_crypt_key_t *key);
int zio_crypt_key_init(uint64_t crypt, zio_crypt_key_t *key);
int zio_crypt_key_get_salt(zio_crypt_key_t *key, uint8_t *salt_out);
int zio_crypt_key_wrap(crypto_key_t *cwkey, zio_crypt_key_t *key, uint8_t *iv,
uint8_t *mac, uint8_t *keydata_out, uint8_t *hmac_keydata_out);
int zio_crypt_key_unwrap(crypto_key_t *cwkey, uint64_t crypt, uint64_t version,
uint64_t guid, uint8_t *keydata, uint8_t *hmac_keydata, uint8_t *iv,
uint8_t *mac, zio_crypt_key_t *key);
int zio_crypt_generate_iv(uint8_t *ivbuf);
int zio_crypt_generate_iv_salt_dedup(zio_crypt_key_t *key, uint8_t *data,
uint_t datalen, uint8_t *ivbuf, uint8_t *salt);
void zio_crypt_encode_params_bp(blkptr_t *bp, uint8_t *salt, uint8_t *iv);
void zio_crypt_decode_params_bp(const blkptr_t *bp, uint8_t *salt, uint8_t *iv);
void zio_crypt_encode_mac_bp(blkptr_t *bp, uint8_t *mac);
void zio_crypt_decode_mac_bp(const blkptr_t *bp, uint8_t *mac);
void zio_crypt_encode_mac_zil(void *data, uint8_t *mac);
void zio_crypt_decode_mac_zil(const void *data, uint8_t *mac);
void zio_crypt_copy_dnode_bonus(abd_t *src_abd, uint8_t *dst, uint_t datalen);
int zio_crypt_do_indirect_mac_checksum(boolean_t generate, void *buf,
uint_t datalen, boolean_t byteswap, uint8_t *cksum);
int zio_crypt_do_indirect_mac_checksum_abd(boolean_t generate, abd_t *abd,
uint_t datalen, boolean_t byteswap, uint8_t *cksum);
int zio_crypt_do_hmac(zio_crypt_key_t *key, uint8_t *data, uint_t datalen,
uint8_t *digestbuf, uint_t digestlen);
int zio_crypt_do_objset_hmacs(zio_crypt_key_t *key, void *data, uint_t datalen,
boolean_t byteswap, uint8_t *portable_mac, uint8_t *local_mac);
int zio_do_crypt_data(boolean_t encrypt, zio_crypt_key_t *key,
dmu_object_type_t ot, boolean_t byteswap, uint8_t *salt, uint8_t *iv,
uint8_t *mac, uint_t datalen, uint8_t *plainbuf, uint8_t *cipherbuf,
boolean_t *no_crypt);
int zio_do_crypt_abd(boolean_t encrypt, zio_crypt_key_t *key,
dmu_object_type_t ot, boolean_t byteswap, uint8_t *salt, uint8_t *iv,
uint8_t *mac, uint_t datalen, abd_t *pabd, abd_t *cabd,
boolean_t *no_crypt);
#endif
diff --git a/include/sys/zvol_impl.h b/include/sys/zvol_impl.h
index 89fe59800498..223393de79fd 100644
--- a/include/sys/zvol_impl.h
+++ b/include/sys/zvol_impl.h
@@ -1,110 +1,109 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
#ifndef _SYS_ZVOL_IMPL_H
#define _SYS_ZVOL_IMPL_H
#include <sys/zfs_context.h>
#define ZVOL_RDONLY 0x1
/*
* Whether the zvol has been written to (as opposed to ZVOL_RDONLY, which
* specifies whether or not the zvol _can_ be written to)
*/
#define ZVOL_WRITTEN_TO 0x2
#define ZVOL_DUMPIFIED 0x4
#define ZVOL_EXCL 0x8
/*
* The in-core state of each volume.
*/
typedef struct zvol_state {
char zv_name[MAXNAMELEN]; /* name */
uint64_t zv_volsize; /* advertised space */
uint64_t zv_volblocksize; /* volume block size */
objset_t *zv_objset; /* objset handle */
uint32_t zv_flags; /* ZVOL_* flags */
uint32_t zv_open_count; /* open counts */
uint32_t zv_changed; /* disk changed */
uint32_t zv_volmode; /* volmode */
zilog_t *zv_zilog; /* ZIL handle */
zfs_rangelock_t zv_rangelock; /* for range locking */
dnode_t *zv_dn; /* dnode hold */
dataset_kstats_t zv_kstat; /* zvol kstats */
list_node_t zv_next; /* next zvol_state_t linkage */
uint64_t zv_hash; /* name hash */
struct hlist_node zv_hlink; /* hash link */
kmutex_t zv_state_lock; /* protects zvol_state_t */
atomic_t zv_suspend_ref; /* refcount for suspend */
krwlock_t zv_suspend_lock; /* suspend lock */
struct zvol_state_os *zv_zso; /* private platform state */
} zvol_state_t;
-extern list_t zvol_state_list;
extern krwlock_t zvol_state_lock;
#define ZVOL_HT_SIZE 1024
extern struct hlist_head *zvol_htable;
#define ZVOL_HT_HEAD(hash) (&zvol_htable[(hash) & (ZVOL_HT_SIZE-1)])
-extern zil_replay_func_t *zvol_replay_vector[TX_MAX_TYPE];
+extern zil_replay_func_t *const zvol_replay_vector[TX_MAX_TYPE];
extern unsigned int zvol_volmode;
extern unsigned int zvol_inhibit_dev;
/*
* platform independent functions exported to platform code
*/
zvol_state_t *zvol_find_by_name_hash(const char *name,
uint64_t hash, int mode);
int zvol_first_open(zvol_state_t *zv, boolean_t readonly);
uint64_t zvol_name_hash(const char *name);
void zvol_remove_minors_impl(const char *name);
void zvol_last_close(zvol_state_t *zv);
void zvol_insert(zvol_state_t *zv);
void zvol_log_truncate(zvol_state_t *zv, dmu_tx_t *tx, uint64_t off,
uint64_t len, boolean_t sync);
void zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, uint64_t offset,
uint64_t size, int sync);
int zvol_get_data(void *arg, uint64_t arg2, lr_write_t *lr, char *buf,
struct lwb *lwb, zio_t *zio);
int zvol_init_impl(void);
void zvol_fini_impl(void);
void zvol_wait_close(zvol_state_t *zv);
/*
* platform dependent functions exported to platform independent code
*/
typedef struct zvol_platform_ops {
void (*zv_free)(zvol_state_t *);
void (*zv_rename_minor)(zvol_state_t *, const char *);
int (*zv_create_minor)(const char *);
int (*zv_update_volsize)(zvol_state_t *, uint64_t);
boolean_t (*zv_is_zvol)(const char *);
void (*zv_clear_private)(zvol_state_t *);
void (*zv_set_disk_ro)(zvol_state_t *, int flags);
void (*zv_set_capacity)(zvol_state_t *, uint64_t capacity);
} zvol_platform_ops_t;
void zvol_register_ops(const zvol_platform_ops_t *ops);
#endif
diff --git a/include/zfs_comutil.h b/include/zfs_comutil.h
index 3e4716224a06..ea2da4a15f1d 100644
--- a/include/zfs_comutil.h
+++ b/include/zfs_comutil.h
@@ -1,53 +1,53 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2017 by Delphix. All rights reserved.
*/
#ifndef _ZFS_COMUTIL_H
#define _ZFS_COMUTIL_H extern __attribute__((visibility("default")))
#include <sys/fs/zfs.h>
#include <sys/types.h>
#ifdef __cplusplus
extern "C" {
#endif
_ZFS_COMUTIL_H boolean_t zfs_allocatable_devs(nvlist_t *);
_ZFS_COMUTIL_H boolean_t zfs_special_devs(nvlist_t *, char *);
_ZFS_COMUTIL_H void zpool_get_load_policy(nvlist_t *, zpool_load_policy_t *);
_ZFS_COMUTIL_H int zfs_zpl_version_map(int spa_version);
_ZFS_COMUTIL_H int zfs_spa_version_map(int zpl_version);
_ZFS_COMUTIL_H boolean_t zfs_dataset_name_hidden(const char *);
#define ZFS_NUM_LEGACY_HISTORY_EVENTS 41
-_ZFS_COMUTIL_H const char *
+_ZFS_COMUTIL_H const char *const
zfs_history_event_names[ZFS_NUM_LEGACY_HISTORY_EVENTS];
#ifdef __cplusplus
}
#endif
#endif /* _ZFS_COMUTIL_H */
diff --git a/include/zfs_deleg.h b/include/zfs_deleg.h
index 1ae08850fbe9..77f64786b089 100644
--- a/include/zfs_deleg.h
+++ b/include/zfs_deleg.h
@@ -1,99 +1,99 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2010 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2013, 2015 by Delphix. All rights reserved.
*/
#ifndef _ZFS_DELEG_H
#define _ZFS_DELEG_H extern __attribute__((visibility("default")))
#include <sys/fs/zfs.h>
#ifdef __cplusplus
extern "C" {
#endif
#define ZFS_DELEG_SET_NAME_CHR '@' /* set name lead char */
#define ZFS_DELEG_FIELD_SEP_CHR '$' /* field separator */
/*
* Max name length for a delegation attribute
*/
#define ZFS_MAX_DELEG_NAME 128
#define ZFS_DELEG_LOCAL 'l'
#define ZFS_DELEG_DESCENDENT 'd'
#define ZFS_DELEG_NA '-'
typedef enum {
ZFS_DELEG_NOTE_CREATE,
ZFS_DELEG_NOTE_DESTROY,
ZFS_DELEG_NOTE_SNAPSHOT,
ZFS_DELEG_NOTE_ROLLBACK,
ZFS_DELEG_NOTE_CLONE,
ZFS_DELEG_NOTE_PROMOTE,
ZFS_DELEG_NOTE_RENAME,
ZFS_DELEG_NOTE_SEND,
ZFS_DELEG_NOTE_RECEIVE,
ZFS_DELEG_NOTE_ALLOW,
ZFS_DELEG_NOTE_USERPROP,
ZFS_DELEG_NOTE_MOUNT,
ZFS_DELEG_NOTE_SHARE,
ZFS_DELEG_NOTE_USERQUOTA,
ZFS_DELEG_NOTE_GROUPQUOTA,
ZFS_DELEG_NOTE_USERUSED,
ZFS_DELEG_NOTE_GROUPUSED,
ZFS_DELEG_NOTE_USEROBJQUOTA,
ZFS_DELEG_NOTE_GROUPOBJQUOTA,
ZFS_DELEG_NOTE_USEROBJUSED,
ZFS_DELEG_NOTE_GROUPOBJUSED,
ZFS_DELEG_NOTE_HOLD,
ZFS_DELEG_NOTE_RELEASE,
ZFS_DELEG_NOTE_DIFF,
ZFS_DELEG_NOTE_BOOKMARK,
ZFS_DELEG_NOTE_LOAD_KEY,
ZFS_DELEG_NOTE_CHANGE_KEY,
ZFS_DELEG_NOTE_PROJECTUSED,
ZFS_DELEG_NOTE_PROJECTQUOTA,
ZFS_DELEG_NOTE_PROJECTOBJUSED,
ZFS_DELEG_NOTE_PROJECTOBJQUOTA,
ZFS_DELEG_NOTE_NONE
} zfs_deleg_note_t;
typedef struct zfs_deleg_perm_tab {
char *z_perm;
zfs_deleg_note_t z_note;
} zfs_deleg_perm_tab_t;
-_ZFS_DELEG_H zfs_deleg_perm_tab_t zfs_deleg_perm_tab[];
+_ZFS_DELEG_H const zfs_deleg_perm_tab_t zfs_deleg_perm_tab[];
_ZFS_DELEG_H int zfs_deleg_verify_nvlist(nvlist_t *nvlist);
_ZFS_DELEG_H void zfs_deleg_whokey(char *attr, zfs_deleg_who_type_t type,
char checkflag, void *data);
_ZFS_DELEG_H const char *zfs_deleg_canonicalize_perm(const char *perm);
#ifdef __cplusplus
}
#endif
#endif /* _ZFS_DELEG_H */
diff --git a/lib/libnvpair/libnvpair.abi b/lib/libnvpair/libnvpair.abi
index 4a001c3dc3a6..01be5785a4ae 100644
--- a/lib/libnvpair/libnvpair.abi
+++ b/lib/libnvpair/libnvpair.abi
@@ -1,2805 +1,2806 @@
<abi-corpus version='2.0' architecture='elf-amd-x86_64' soname='libnvpair.so.3'>
<elf-needed>
<dependency name='libc.so.6'/>
</elf-needed>
<elf-function-symbols>
- <elf-symbol name='_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
- <elf-symbol name='_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='dump_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_add_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_alloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_dup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_lookup_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_merge' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_num_pairs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_pack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_pack_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_remove_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_size' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvlist_unpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fnvpair_value_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libspl_assertf' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nv_alloc_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nv_alloc_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nv_alloc_reset' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_double' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_hrtime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_add_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_alloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_dup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_exists' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_double' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_hrtime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_nv_alloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_nvpair_embedded_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_pairs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_lookup_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_merge' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_next_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_nvflag' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_pack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prev_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_print' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_print_json' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctl_alloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctl_dofmt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctl_doindent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctl_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctl_getdest' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctl_setdest' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctl_setfmt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctl_setindent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_double' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_hrtime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_prtctlop_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_remove_all' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_remove_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_size' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_unpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_xalloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_xdup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_xpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvlist_xunpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_type_is_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_double' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_hrtime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_match' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_match_regex' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nvpair_value_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
</elf-function-symbols>
<elf-variable-symbols>
<elf-symbol name='libspl_assert_ok' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nv_alloc_nosleep' size='8' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='nv_fixed_ops' size='8' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
</elf-variable-symbols>
<abi-instr address-size='64' path='../../module/nvpair/fnvpair.c' language='LANG_C99'>
<function-decl name='fnvlist_alloc' mangled-name='fnvlist_alloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_alloc'>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='fnvlist_free' mangled-name='fnvlist_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_free'>
<parameter type-id='5ce45b60' name='nvl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_size' mangled-name='fnvlist_size' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_size'>
<parameter type-id='5ce45b60' name='nvl'/>
<return type-id='b59d7dce'/>
</function-decl>
<function-decl name='fnvlist_pack' mangled-name='fnvlist_pack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_pack'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='78c01427' name='sizep'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='fnvlist_pack_free' mangled-name='fnvlist_pack_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_pack_free'>
<parameter type-id='26a90f95' name='pack'/>
<parameter type-id='b59d7dce' name='size'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_unpack' mangled-name='fnvlist_unpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_unpack'>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='buflen'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='fnvlist_dup' mangled-name='fnvlist_dup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_dup'>
<parameter type-id='22cce67b' name='nvl'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='fnvlist_merge' mangled-name='fnvlist_merge' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_merge'>
<parameter type-id='5ce45b60' name='dst'/>
<parameter type-id='5ce45b60' name='src'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_num_pairs' mangled-name='fnvlist_num_pairs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_num_pairs'>
<parameter type-id='5ce45b60' name='nvl'/>
<return type-id='b59d7dce'/>
</function-decl>
<function-decl name='fnvlist_add_boolean' mangled-name='fnvlist_add_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_boolean'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_boolean_value' mangled-name='fnvlist_add_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_boolean_value'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='c19b74c3' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_byte' mangled-name='fnvlist_add_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_byte'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='d8bf0010' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_int8' mangled-name='fnvlist_add_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int8'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='ee31ee44' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_uint8' mangled-name='fnvlist_add_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint8'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='b96825af' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_int16' mangled-name='fnvlist_add_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int16'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='23bd8cb5' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_uint16' mangled-name='fnvlist_add_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint16'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='149c6638' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_int32' mangled-name='fnvlist_add_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int32'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='3ff5601b' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_uint32' mangled-name='fnvlist_add_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint32'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='8f92235e' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_int64' mangled-name='fnvlist_add_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int64'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9da381c4' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_uint64' mangled-name='fnvlist_add_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint64'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9c313c2d' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_string' mangled-name='fnvlist_add_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_string'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='80f4b756' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_nvlist' mangled-name='fnvlist_add_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_nvlist'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='5ce45b60' name='val'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_nvpair' mangled-name='fnvlist_add_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_nvpair'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='3fa542f0' name='pair'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_boolean_array' mangled-name='fnvlist_add_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_boolean_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='c5f6c15b' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_byte_array' mangled-name='fnvlist_add_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_byte_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='d1db479e' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_int8_array' mangled-name='fnvlist_add_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int8_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='a06445da' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_uint8_array' mangled-name='fnvlist_add_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint8_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9f7200cf' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_int16_array' mangled-name='fnvlist_add_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int16_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='a3eb883d' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_uint16_array' mangled-name='fnvlist_add_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint16_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='1b7d11c6' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_int32_array' mangled-name='fnvlist_add_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int32_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='1f526493' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_uint32_array' mangled-name='fnvlist_add_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint32_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='a6798dcc' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_int64_array' mangled-name='fnvlist_add_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int64_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='505bed1a' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_uint64_array' mangled-name='fnvlist_add_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint64_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='713a56f5' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_string_array' mangled-name='fnvlist_add_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_string_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='13956559' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_add_nvlist_array' mangled-name='fnvlist_add_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_nvlist_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='3bbfee2e' name='val'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_remove' mangled-name='fnvlist_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_remove'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_remove_nvpair' mangled-name='fnvlist_remove_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_remove_nvpair'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='3fa542f0' name='pair'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fnvlist_lookup_nvpair' mangled-name='fnvlist_lookup_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_nvpair'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='3fa542f0'/>
</function-decl>
<function-decl name='fnvlist_lookup_boolean' mangled-name='fnvlist_lookup_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_boolean'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='fnvlist_lookup_boolean_value' mangled-name='fnvlist_lookup_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_boolean_value'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='fnvlist_lookup_byte' mangled-name='fnvlist_lookup_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_byte'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='d8bf0010'/>
</function-decl>
<function-decl name='fnvlist_lookup_int8' mangled-name='fnvlist_lookup_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int8'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='ee31ee44'/>
</function-decl>
<function-decl name='fnvlist_lookup_int16' mangled-name='fnvlist_lookup_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int16'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='23bd8cb5'/>
</function-decl>
<function-decl name='fnvlist_lookup_int32' mangled-name='fnvlist_lookup_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int32'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='3ff5601b'/>
</function-decl>
<function-decl name='fnvlist_lookup_int64' mangled-name='fnvlist_lookup_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int64'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='9da381c4'/>
</function-decl>
<function-decl name='fnvlist_lookup_uint8' mangled-name='fnvlist_lookup_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint8'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='fnvlist_lookup_uint16' mangled-name='fnvlist_lookup_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint16'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='fnvlist_lookup_uint32' mangled-name='fnvlist_lookup_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint32'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='fnvlist_lookup_uint64' mangled-name='fnvlist_lookup_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint64'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='fnvlist_lookup_string' mangled-name='fnvlist_lookup_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_string'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='fnvlist_lookup_nvlist' mangled-name='fnvlist_lookup_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_nvlist'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='fnvlist_lookup_boolean_array' mangled-name='fnvlist_lookup_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_boolean_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='37e3bd22'/>
</function-decl>
<function-decl name='fnvlist_lookup_byte_array' mangled-name='fnvlist_lookup_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_byte_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='45b65157'/>
</function-decl>
<function-decl name='fnvlist_lookup_int8_array' mangled-name='fnvlist_lookup_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int8_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='256d5229'/>
</function-decl>
<function-decl name='fnvlist_lookup_uint8_array' mangled-name='fnvlist_lookup_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint8_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='ae3e8ca6'/>
</function-decl>
<function-decl name='fnvlist_lookup_int16_array' mangled-name='fnvlist_lookup_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int16_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='f76f73d0'/>
</function-decl>
<function-decl name='fnvlist_lookup_uint16_array' mangled-name='fnvlist_lookup_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint16_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='8a121f49'/>
</function-decl>
<function-decl name='fnvlist_lookup_int32_array' mangled-name='fnvlist_lookup_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int32_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='4aafb922'/>
</function-decl>
<function-decl name='fnvlist_lookup_uint32_array' mangled-name='fnvlist_lookup_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint32_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='90421557'/>
</function-decl>
<function-decl name='fnvlist_lookup_int64_array' mangled-name='fnvlist_lookup_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int64_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='cb785ebf'/>
</function-decl>
<function-decl name='fnvlist_lookup_uint64_array' mangled-name='fnvlist_lookup_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint64_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='5d6479ae'/>
</function-decl>
<function-decl name='fnvpair_value_boolean_value' mangled-name='fnvpair_value_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_boolean_value'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='fnvpair_value_byte' mangled-name='fnvpair_value_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_byte'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='d8bf0010'/>
</function-decl>
<function-decl name='fnvpair_value_int8' mangled-name='fnvpair_value_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_int8'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='ee31ee44'/>
</function-decl>
<function-decl name='fnvpair_value_int16' mangled-name='fnvpair_value_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_int16'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='23bd8cb5'/>
</function-decl>
<function-decl name='fnvpair_value_int32' mangled-name='fnvpair_value_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_int32'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='3ff5601b'/>
</function-decl>
<function-decl name='fnvpair_value_int64' mangled-name='fnvpair_value_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_int64'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='9da381c4'/>
</function-decl>
<function-decl name='fnvpair_value_uint8' mangled-name='fnvpair_value_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_uint8'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='fnvpair_value_uint16' mangled-name='fnvpair_value_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_uint16'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='fnvpair_value_uint32' mangled-name='fnvpair_value_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_uint32'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='fnvpair_value_uint64' mangled-name='fnvpair_value_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_uint64'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='fnvpair_value_string' mangled-name='fnvpair_value_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_string'>
<parameter type-id='3fa542f0' name='nvp'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='fnvpair_value_nvlist' mangled-name='fnvpair_value_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_nvlist'>
<parameter type-id='3fa542f0' name='nvp'/>
<return type-id='5ce45b60'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='../../module/nvpair/nvpair.c' language='LANG_C99'>
<pointer-type-def type-id='37e3bd22' size-in-bits='64' id='03829398'/>
<pointer-type-def type-id='9b23c9ad' size-in-bits='64' id='c0563f85'/>
<qualified-type-def type-id='c19b74c3' const='yes' id='12373e33'/>
<pointer-type-def type-id='12373e33' size-in-bits='64' id='c5f6c15b'/>
<qualified-type-def type-id='80f4b756' const='yes' id='b99c00c9'/>
<pointer-type-def type-id='b99c00c9' size-in-bits='64' id='13956559'/>
<qualified-type-def type-id='23bd8cb5' const='yes' id='75f7b0c5'/>
<pointer-type-def type-id='75f7b0c5' size-in-bits='64' id='a3eb883d'/>
<qualified-type-def type-id='3ff5601b' const='yes' id='922df12b'/>
<pointer-type-def type-id='922df12b' size-in-bits='64' id='1f526493'/>
<qualified-type-def type-id='9da381c4' const='yes' id='f07b7694'/>
<pointer-type-def type-id='f07b7694' size-in-bits='64' id='505bed1a'/>
<qualified-type-def type-id='ee31ee44' const='yes' id='721c32d4'/>
<pointer-type-def type-id='721c32d4' size-in-bits='64' id='a06445da'/>
<qualified-type-def type-id='8e8d4be3' const='yes' id='693c3853'/>
<pointer-type-def type-id='693c3853' size-in-bits='64' id='22cce67b'/>
<qualified-type-def type-id='22cce67b' const='yes' id='d2816df0'/>
<pointer-type-def type-id='d2816df0' size-in-bits='64' id='3bbfee2e'/>
<qualified-type-def type-id='57928edf' const='yes' id='642ee20f'/>
<pointer-type-def type-id='642ee20f' size-in-bits='64' id='dace003f'/>
<qualified-type-def type-id='d8bf0010' const='yes' id='a9125480'/>
<pointer-type-def type-id='a9125480' size-in-bits='64' id='d1db479e'/>
<qualified-type-def type-id='149c6638' const='yes' id='b01a5ac8'/>
<pointer-type-def type-id='b01a5ac8' size-in-bits='64' id='1b7d11c6'/>
<qualified-type-def type-id='8f92235e' const='yes' id='b9930aae'/>
<pointer-type-def type-id='b9930aae' size-in-bits='64' id='a6798dcc'/>
<qualified-type-def type-id='9c313c2d' const='yes' id='c3b7ba7d'/>
<pointer-type-def type-id='c3b7ba7d' size-in-bits='64' id='713a56f5'/>
<qualified-type-def type-id='b96825af' const='yes' id='2b61797f'/>
<pointer-type-def type-id='2b61797f' size-in-bits='64' id='9f7200cf'/>
<pointer-type-def type-id='a0eb0f08' size-in-bits='64' id='7408d286'/>
<pointer-type-def type-id='cebdd548' size-in-bits='64' id='e379e62d'/>
<pointer-type-def type-id='95e97e5e' size-in-bits='64' id='7292109c'/>
<pointer-type-def type-id='f76f73d0' size-in-bits='64' id='7e73928e'/>
<pointer-type-def type-id='4aafb922' size-in-bits='64' id='9aa04798'/>
<pointer-type-def type-id='cb785ebf' size-in-bits='64' id='e37ce48f'/>
<pointer-type-def type-id='256d5229' size-in-bits='64' id='ee181ab9'/>
<pointer-type-def type-id='857bb57e' size-in-bits='64' id='75be733c'/>
<pointer-type-def type-id='3fa542f0' size-in-bits='64' id='0b283d2e'/>
<pointer-type-def type-id='b59d7dce' size-in-bits='64' id='78c01427'/>
<pointer-type-def type-id='45b65157' size-in-bits='64' id='3b0247c7'/>
<pointer-type-def type-id='8a121f49' size-in-bits='64' id='bd8768d9'/>
<pointer-type-def type-id='90421557' size-in-bits='64' id='9507d3c7'/>
<pointer-type-def type-id='5d6479ae' size-in-bits='64' id='892b4acc'/>
<pointer-type-def type-id='ae3e8ca6' size-in-bits='64' id='d8774064'/>
<pointer-type-def type-id='3502e3ff' size-in-bits='64' id='4dd26a40'/>
<function-decl name='nv_alloc_init' mangled-name='nv_alloc_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nv_alloc_init'>
<parameter type-id='11871392' name='nva'/>
<parameter type-id='ee1d4944' name='nvo'/>
<parameter is-variadic='yes'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nv_alloc_reset' mangled-name='nv_alloc_reset' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nv_alloc_reset'>
<parameter type-id='11871392' name='nva'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nv_alloc_fini' mangled-name='nv_alloc_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nv_alloc_fini'>
<parameter type-id='11871392' name='nva'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_lookup_nv_alloc' mangled-name='nvlist_lookup_nv_alloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nv_alloc'>
<parameter type-id='5ce45b60' name='nvl'/>
<return type-id='11871392'/>
</function-decl>
<function-decl name='nvlist_nvflag' mangled-name='nvlist_nvflag' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_nvflag'>
<parameter type-id='5ce45b60' name='nvl'/>
<return type-id='3502e3ff'/>
</function-decl>
<function-decl name='nvlist_alloc' mangled-name='nvlist_alloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_alloc'>
<parameter type-id='857bb57e' name='nvlp'/>
<parameter type-id='3502e3ff' name='nvflag'/>
<parameter type-id='95e97e5e' name='kmflag'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_xalloc' mangled-name='nvlist_xalloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_xalloc'>
<parameter type-id='857bb57e' name='nvlp'/>
<parameter type-id='3502e3ff' name='nvflag'/>
<parameter type-id='11871392' name='nva'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_free' mangled-name='nvlist_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_free'>
<parameter type-id='5ce45b60' name='nvl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_dup' mangled-name='nvlist_dup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_dup'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='857bb57e' name='nvlp'/>
<parameter type-id='95e97e5e' name='kmflag'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_xdup' mangled-name='nvlist_xdup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_xdup'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='857bb57e' name='nvlp'/>
<parameter type-id='11871392' name='nva'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_remove_all' mangled-name='nvlist_remove_all' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_remove_all'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_remove' mangled-name='nvlist_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_remove'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='8d0687d2' name='type'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_remove_nvpair' mangled-name='nvlist_remove_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_remove_nvpair'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='3fa542f0' name='nvp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_boolean' mangled-name='nvlist_add_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_boolean'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_boolean_value' mangled-name='nvlist_add_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_boolean_value'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='c19b74c3' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_byte' mangled-name='nvlist_add_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_byte'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='d8bf0010' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_int8' mangled-name='nvlist_add_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int8'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='ee31ee44' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_uint8' mangled-name='nvlist_add_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint8'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='b96825af' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_int16' mangled-name='nvlist_add_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int16'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='23bd8cb5' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_uint16' mangled-name='nvlist_add_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint16'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='149c6638' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_int32' mangled-name='nvlist_add_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int32'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='3ff5601b' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_uint32' mangled-name='nvlist_add_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint32'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='8f92235e' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_int64' mangled-name='nvlist_add_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int64'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9da381c4' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_uint64' mangled-name='nvlist_add_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint64'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9c313c2d' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_double' mangled-name='nvlist_add_double' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_double'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='a0eb0f08' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_string' mangled-name='nvlist_add_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_string'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='80f4b756' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_boolean_array' mangled-name='nvlist_add_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_boolean_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='c5f6c15b' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_byte_array' mangled-name='nvlist_add_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_byte_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='d1db479e' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_int8_array' mangled-name='nvlist_add_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int8_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='a06445da' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_uint8_array' mangled-name='nvlist_add_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint8_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9f7200cf' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_int16_array' mangled-name='nvlist_add_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int16_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='a3eb883d' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_uint16_array' mangled-name='nvlist_add_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint16_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='1b7d11c6' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_int32_array' mangled-name='nvlist_add_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int32_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='1f526493' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_uint32_array' mangled-name='nvlist_add_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint32_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='a6798dcc' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_int64_array' mangled-name='nvlist_add_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int64_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='505bed1a' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_uint64_array' mangled-name='nvlist_add_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint64_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='713a56f5' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_string_array' mangled-name='nvlist_add_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_string_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='13956559' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_hrtime' mangled-name='nvlist_add_hrtime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_hrtime'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='cebdd548' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_nvlist' mangled-name='nvlist_add_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_nvlist'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='22cce67b' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_nvlist_array' mangled-name='nvlist_add_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_nvlist_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='3bbfee2e' name='a'/>
<parameter type-id='3502e3ff' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_next_nvpair' mangled-name='nvlist_next_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_next_nvpair'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='3fa542f0'/>
</function-decl>
<function-decl name='nvlist_prev_nvpair' mangled-name='nvlist_prev_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prev_nvpair'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='3fa542f0'/>
</function-decl>
<function-decl name='nvlist_empty' mangled-name='nvlist_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_empty'>
<parameter type-id='22cce67b' name='nvl'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='nvpair_name' mangled-name='nvpair_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_name'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='nvpair_type' mangled-name='nvpair_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_type'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='8d0687d2'/>
</function-decl>
<function-decl name='nvpair_type_is_array' mangled-name='nvpair_type_is_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_type_is_array'>
<parameter type-id='dace003f' name='nvp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_boolean' mangled-name='nvlist_lookup_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_boolean'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_boolean_value' mangled-name='nvlist_lookup_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_boolean_value'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='37e3bd22' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_byte' mangled-name='nvlist_lookup_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_byte'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='45b65157' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_int8' mangled-name='nvlist_lookup_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int8'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='256d5229' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_uint8' mangled-name='nvlist_lookup_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint8'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='ae3e8ca6' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_int16' mangled-name='nvlist_lookup_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int16'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='f76f73d0' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_uint16' mangled-name='nvlist_lookup_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint16'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='8a121f49' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_int32' mangled-name='nvlist_lookup_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int32'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='4aafb922' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_uint32' mangled-name='nvlist_lookup_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint32'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='90421557' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_int64' mangled-name='nvlist_lookup_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int64'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='cb785ebf' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_uint64' mangled-name='nvlist_lookup_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint64'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='5d6479ae' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_double' mangled-name='nvlist_lookup_double' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_double'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='7408d286' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_string' mangled-name='nvlist_lookup_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_string'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9b23c9ad' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_nvlist' mangled-name='nvlist_lookup_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nvlist'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='857bb57e' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_boolean_array' mangled-name='nvlist_lookup_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_boolean_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='03829398' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_byte_array' mangled-name='nvlist_lookup_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_byte_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='3b0247c7' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_int8_array' mangled-name='nvlist_lookup_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int8_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='ee181ab9' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_uint8_array' mangled-name='nvlist_lookup_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint8_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='d8774064' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_int16_array' mangled-name='nvlist_lookup_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int16_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='7e73928e' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_uint16_array' mangled-name='nvlist_lookup_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint16_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='bd8768d9' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_int32_array' mangled-name='nvlist_lookup_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int32_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9aa04798' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_uint32_array' mangled-name='nvlist_lookup_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint32_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9507d3c7' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_int64_array' mangled-name='nvlist_lookup_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int64_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='e37ce48f' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_uint64_array' mangled-name='nvlist_lookup_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint64_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='892b4acc' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_string_array' mangled-name='nvlist_lookup_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_string_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='c0563f85' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_nvlist_array' mangled-name='nvlist_lookup_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nvlist_array'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='75be733c' name='a'/>
<parameter type-id='4dd26a40' name='n'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_hrtime' mangled-name='nvlist_lookup_hrtime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_hrtime'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='e379e62d' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_pairs' mangled-name='nvlist_lookup_pairs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_pairs'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='95e97e5e' name='flag'/>
<parameter is-variadic='yes'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_nvpair' mangled-name='nvlist_lookup_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nvpair'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='0b283d2e' name='ret'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_lookup_nvpair_embedded_index' mangled-name='nvlist_lookup_nvpair_embedded_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nvpair_embedded_index'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='0b283d2e' name='ret'/>
<parameter type-id='7292109c' name='ip'/>
<parameter type-id='9b23c9ad' name='ep'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_exists' mangled-name='nvlist_exists' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_exists'>
<parameter type-id='22cce67b' name='nvl'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='nvpair_value_boolean_value' mangled-name='nvpair_value_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_boolean_value'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='37e3bd22' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_byte' mangled-name='nvpair_value_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_byte'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='45b65157' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_int8' mangled-name='nvpair_value_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int8'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='256d5229' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_uint8' mangled-name='nvpair_value_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint8'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='ae3e8ca6' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_int16' mangled-name='nvpair_value_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int16'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='f76f73d0' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_uint16' mangled-name='nvpair_value_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint16'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='8a121f49' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_int32' mangled-name='nvpair_value_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int32'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='4aafb922' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_uint32' mangled-name='nvpair_value_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint32'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='90421557' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_int64' mangled-name='nvpair_value_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int64'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='cb785ebf' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_uint64' mangled-name='nvpair_value_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint64'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='5d6479ae' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_double' mangled-name='nvpair_value_double' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_double'>
<parameter type-id='dace003f' name='nvp'/>
<parameter type-id='7408d286' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_string' mangled-name='nvpair_value_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_string'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='9b23c9ad' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_nvlist' mangled-name='nvpair_value_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_nvlist'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='857bb57e' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_boolean_array' mangled-name='nvpair_value_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_boolean_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='03829398' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_byte_array' mangled-name='nvpair_value_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_byte_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='3b0247c7' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_int8_array' mangled-name='nvpair_value_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int8_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='ee181ab9' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_uint8_array' mangled-name='nvpair_value_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint8_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='d8774064' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_int16_array' mangled-name='nvpair_value_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int16_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='7e73928e' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_uint16_array' mangled-name='nvpair_value_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint16_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='bd8768d9' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_int32_array' mangled-name='nvpair_value_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int32_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='9aa04798' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_uint32_array' mangled-name='nvpair_value_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint32_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='9507d3c7' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_int64_array' mangled-name='nvpair_value_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int64_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='e37ce48f' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_uint64_array' mangled-name='nvpair_value_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint64_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='892b4acc' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_string_array' mangled-name='nvpair_value_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_string_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='c0563f85' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_nvlist_array' mangled-name='nvpair_value_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_nvlist_array'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='75be733c' name='val'/>
<parameter type-id='4dd26a40' name='nelem'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_hrtime' mangled-name='nvpair_value_hrtime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_hrtime'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='e379e62d' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_add_nvpair' mangled-name='nvlist_add_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_nvpair'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='3fa542f0' name='nvp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_merge' mangled-name='nvlist_merge' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_merge'>
<parameter type-id='5ce45b60' name='dst'/>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='95e97e5e' name='flag'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_size' mangled-name='nvlist_size' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_size'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='78c01427' name='size'/>
<parameter type-id='95e97e5e' name='encoding'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_pack' mangled-name='nvlist_pack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_pack'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='9b23c9ad' name='bufp'/>
<parameter type-id='78c01427' name='buflen'/>
<parameter type-id='95e97e5e' name='encoding'/>
<parameter type-id='95e97e5e' name='kmflag'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_xpack' mangled-name='nvlist_xpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_xpack'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='9b23c9ad' name='bufp'/>
<parameter type-id='78c01427' name='buflen'/>
<parameter type-id='95e97e5e' name='encoding'/>
<parameter type-id='11871392' name='nva'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_unpack' mangled-name='nvlist_unpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_unpack'>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='buflen'/>
<parameter type-id='857bb57e' name='nvlp'/>
<parameter type-id='95e97e5e' name='kmflag'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvlist_xunpack' mangled-name='nvlist_xunpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_xunpack'>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='buflen'/>
<parameter type-id='857bb57e' name='nvlp'/>
<parameter type-id='11871392' name='nva'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='../../module/nvpair/nvpair_alloc_fixed.c' language='LANG_C99'>
- <var-decl name='nv_fixed_ops' type-id='ee1d4944' mangled-name='nv_fixed_ops' visibility='default' elf-symbol-id='nv_fixed_ops'/>
+ <qualified-type-def type-id='ee1d4944' const='yes' id='4b95388f'/>
+ <var-decl name='nv_fixed_ops' type-id='4b95388f' mangled-name='nv_fixed_ops' visibility='default' elf-symbol-id='nv_fixed_ops'/>
</abi-instr>
<abi-instr address-size='64' path='libnvpair.c' language='LANG_C99'>
<type-decl name='char' size-in-bits='8' id='a84c031d'/>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8' id='89feb1ec'>
<subrange length='1' type-id='7359adad' id='52f813b4'/>
</array-type-def>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='160' id='664ac0b7'>
<subrange length='20' type-id='7359adad' id='fdca39cf'/>
</array-type-def>
+ <class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
+ <class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
+ <class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
+ <class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
<type-decl name='double' size-in-bits='64' id='a0eb0f08'/>
<type-decl name='int' size-in-bits='32' id='95e97e5e'/>
<type-decl name='long int' size-in-bits='64' id='bd54fe1a'/>
<type-decl name='long long int' size-in-bits='64' id='1eb56b1e'/>
<type-decl name='short int' size-in-bits='16' id='a2185560'/>
<type-decl name='signed char' size-in-bits='8' id='28577a57'/>
<type-decl name='unnamed-enum-underlying-type-32' is-anonymous='yes' size-in-bits='32' alignment-in-bits='32' id='9cac1fee'/>
<type-decl name='unsigned char' size-in-bits='8' id='002ac4a6'/>
<type-decl name='unsigned int' size-in-bits='32' id='f0981eeb'/>
<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
<type-decl name='variadic parameter type' id='2c1145c5'/>
+ <type-decl name='void' id='48b5725f'/>
<typedef-decl name='nvlist_prtctl_t' type-id='196db161' id='b0c1ff8d'/>
<enum-decl name='nvlist_indent_mode' id='628aafab'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='NVLIST_INDENT_ABS' value='0'/>
<enumerator name='NVLIST_INDENT_TABBED' value='1'/>
</enum-decl>
<enum-decl name='nvlist_prtctl_fmt' id='c8dcc53a'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='NVLIST_FMT_MEMBER_NAME' value='0'/>
<enumerator name='NVLIST_FMT_MEMBER_POSTAMBLE' value='1'/>
<enumerator name='NVLIST_FMT_BTWN_ARRAY' value='2'/>
</enum-decl>
<enum-decl name='data_type_t' naming-typedef-id='8d0687d2' id='aeeae136'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='DATA_TYPE_DONTCARE' value='-1'/>
<enumerator name='DATA_TYPE_UNKNOWN' value='0'/>
<enumerator name='DATA_TYPE_BOOLEAN' value='1'/>
<enumerator name='DATA_TYPE_BYTE' value='2'/>
<enumerator name='DATA_TYPE_INT16' value='3'/>
<enumerator name='DATA_TYPE_UINT16' value='4'/>
<enumerator name='DATA_TYPE_INT32' value='5'/>
<enumerator name='DATA_TYPE_UINT32' value='6'/>
<enumerator name='DATA_TYPE_INT64' value='7'/>
<enumerator name='DATA_TYPE_UINT64' value='8'/>
<enumerator name='DATA_TYPE_STRING' value='9'/>
<enumerator name='DATA_TYPE_BYTE_ARRAY' value='10'/>
<enumerator name='DATA_TYPE_INT16_ARRAY' value='11'/>
<enumerator name='DATA_TYPE_UINT16_ARRAY' value='12'/>
<enumerator name='DATA_TYPE_INT32_ARRAY' value='13'/>
<enumerator name='DATA_TYPE_UINT32_ARRAY' value='14'/>
<enumerator name='DATA_TYPE_INT64_ARRAY' value='15'/>
<enumerator name='DATA_TYPE_UINT64_ARRAY' value='16'/>
<enumerator name='DATA_TYPE_STRING_ARRAY' value='17'/>
<enumerator name='DATA_TYPE_HRTIME' value='18'/>
<enumerator name='DATA_TYPE_NVLIST' value='19'/>
<enumerator name='DATA_TYPE_NVLIST_ARRAY' value='20'/>
<enumerator name='DATA_TYPE_BOOLEAN_VALUE' value='21'/>
<enumerator name='DATA_TYPE_INT8' value='22'/>
<enumerator name='DATA_TYPE_UINT8' value='23'/>
<enumerator name='DATA_TYPE_BOOLEAN_ARRAY' value='24'/>
<enumerator name='DATA_TYPE_INT8_ARRAY' value='25'/>
<enumerator name='DATA_TYPE_UINT8_ARRAY' value='26'/>
<enumerator name='DATA_TYPE_DOUBLE' value='27'/>
</enum-decl>
<typedef-decl name='data_type_t' type-id='aeeae136' id='8d0687d2'/>
<class-decl name='nvpair' size-in-bits='128' is-struct='yes' visibility='default' id='1c34e459'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='nvp_size' type-id='3ff5601b' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='nvp_name_sz' type-id='23bd8cb5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='48'>
<var-decl name='nvp_reserve' type-id='23bd8cb5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='nvp_value_elem' type-id='3ff5601b' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='96'>
<var-decl name='nvp_type' type-id='8d0687d2' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='nvpair_t' type-id='1c34e459' id='57928edf'/>
<class-decl name='nvlist' size-in-bits='192' is-struct='yes' visibility='default' id='ac266fd9'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='nvl_version' type-id='3ff5601b' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='nvl_nvflag' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='nvl_priv' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='nvl_flag' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='160'>
<var-decl name='nvl_pad' type-id='3ff5601b' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='nvlist_t' type-id='ac266fd9' id='8e8d4be3'/>
<enum-decl name='boolean_t' naming-typedef-id='c19b74c3' id='f58c8277'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='B_FALSE' value='0'/>
<enumerator name='B_TRUE' value='1'/>
</enum-decl>
<typedef-decl name='boolean_t' type-id='f58c8277' id='c19b74c3'/>
<typedef-decl name='uchar_t' type-id='002ac4a6' id='d8bf0010'/>
<typedef-decl name='uint_t' type-id='f0981eeb' id='3502e3ff'/>
<typedef-decl name='hrtime_t' type-id='1eb56b1e' id='cebdd548'/>
<class-decl name='nvlist_printops' size-in-bits='3456' is-struct='yes' visibility='default' id='ebc6735b'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='print_boolean' type-id='e7f43f72' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='print_boolean_value' type-id='e7f43f73' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='print_byte' type-id='e7f43f74' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='print_int8' type-id='e7f43f75' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='512'>
<var-decl name='print_uint8' type-id='e7f43f76' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='640'>
<var-decl name='print_int16' type-id='e7f43f77' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='768'>
<var-decl name='print_uint16' type-id='e7f43f78' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='896'>
<var-decl name='print_int32' type-id='e7f43f79' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1024'>
<var-decl name='print_uint32' type-id='e7f43f7a' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1152'>
<var-decl name='print_int64' type-id='e7f43f7b' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1280'>
<var-decl name='print_uint64' type-id='e7f43f7c' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1408'>
<var-decl name='print_double' type-id='e7f43f7d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1536'>
<var-decl name='print_string' type-id='e7f43f7e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1664'>
<var-decl name='print_hrtime' type-id='e7f43f7f' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1792'>
<var-decl name='print_nvlist' type-id='e7f43f80' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1920'>
<var-decl name='print_boolean_array' type-id='e7f43f81' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2048'>
<var-decl name='print_byte_array' type-id='e7f43f82' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2176'>
<var-decl name='print_int8_array' type-id='e7f43f83' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2304'>
<var-decl name='print_uint8_array' type-id='e7f43f84' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2432'>
<var-decl name='print_int16_array' type-id='e7f43f85' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2560'>
<var-decl name='print_uint16_array' type-id='e7f43f86' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2688'>
<var-decl name='print_int32_array' type-id='e7f43f87' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2816'>
<var-decl name='print_uint32_array' type-id='e7f43f88' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2944'>
<var-decl name='print_int64_array' type-id='e7f43f89' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='3072'>
<var-decl name='print_uint64_array' type-id='e7f43f8a' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='3200'>
<var-decl name='print_string_array' type-id='e7f43f8b' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='3328'>
<var-decl name='print_nvlist_array' type-id='e7f43f8c' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f72'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='6d994334' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__1' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f73'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='6a2f50c1' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__2' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f74'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='8a1fb33a' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__3' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f75'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='506696a8' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__4' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f76'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='39b623f9' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__5' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f77'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='ea6be4eb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__6' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f78'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='f10f1e84' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__7' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f79'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='1708018d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__8' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7a'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='90174072' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__9' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7b'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='d2af7f32' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__10' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7c'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='0b22f759' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__11' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7d'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='3be4d568' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__12' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7e'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='c0d0f877' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__13' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7f'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='e1c54c3c' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__14' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f80'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='19ea27ae' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__15' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f81'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='7ef0e988' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__16' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f82'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='7391ed39' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__17' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f83'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='42257af5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__18' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f84'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='330cc0d0' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__19' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f85'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='506ab59a' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__20' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f86'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='ed6a3a3d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__21' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f87'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='750cc41c' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__22' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f88'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='292cdbcf' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__23' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f89'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='aaea91b5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__24' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f8a'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='7e85a9b6' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__25' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f8b'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='de20bf07' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='__anonymous_struct__26' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f8c'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='op' type-id='2835af80' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='nvlist_prtctl' size-in-bits='576' is-struct='yes' visibility='default' id='d2e8bad9'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='nvprt_fp' type-id='822cd80b' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='nvprt_indent_mode' type-id='628aafab' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='96'>
<var-decl name='nvprt_indent' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='nvprt_indentinc' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='nvprt_nmfmt' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='nvprt_eomfmt' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='nvprt_btwnarrfmt' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='nvprt_btwnarrfmt_nl' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='nvprt_dfltops' type-id='7be54adb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='512'>
<var-decl name='nvprt_custops' type-id='7be54adb' visibility='default'/>
</data-member>
</class-decl>
+ <typedef-decl name='__re_long_size_t' type-id='7359adad' id='ba516949'/>
<typedef-decl name='reg_syntax_t' type-id='7359adad' id='1b72c3b3'/>
<class-decl name='re_pattern_buffer' size-in-bits='512' is-struct='yes' visibility='default' id='19fc9a8c'>
<data-member access='public' layout-offset-in-bits='0'>
- <var-decl name='buffer' type-id='cf536864' visibility='default'/>
+ <var-decl name='buffer' type-id='33976309' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
- <var-decl name='allocated' type-id='7359adad' visibility='default'/>
+ <var-decl name='allocated' type-id='ba516949' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
- <var-decl name='used' type-id='7359adad' visibility='default'/>
+ <var-decl name='used' type-id='ba516949' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='syntax' type-id='1b72c3b3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='fastmap' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='translate' type-id='cf536864' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='re_nsub' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='can_be_null' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='449'>
<var-decl name='regs_allocated' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='451'>
<var-decl name='fastmap_accurate' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='452'>
<var-decl name='no_sub' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='453'>
<var-decl name='not_bol' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='454'>
<var-decl name='not_eol' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='455'>
<var-decl name='newline_anchor' type-id='f0981eeb' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='regex_t' type-id='19fc9a8c' id='aca3bac8'/>
+ <typedef-decl name='int8_t' type-id='2171a512' id='ee31ee44'/>
+ <typedef-decl name='int16_t' type-id='03896e23' id='23bd8cb5'/>
+ <typedef-decl name='int32_t' type-id='33f57a65' id='3ff5601b'/>
+ <typedef-decl name='int64_t' type-id='0c9942d2' id='9da381c4'/>
+ <typedef-decl name='uint8_t' type-id='c51d6389' id='b96825af'/>
+ <typedef-decl name='uint16_t' type-id='253c2d2a' id='149c6638'/>
+ <typedef-decl name='uint32_t' type-id='62f1140c' id='8f92235e'/>
+ <typedef-decl name='uint64_t' type-id='8910171f' id='9c313c2d'/>
+ <typedef-decl name='__int8_t' type-id='28577a57' id='2171a512'/>
+ <typedef-decl name='__uint8_t' type-id='002ac4a6' id='c51d6389'/>
+ <typedef-decl name='__int16_t' type-id='a2185560' id='03896e23'/>
+ <typedef-decl name='__uint16_t' type-id='8efea9e5' id='253c2d2a'/>
+ <typedef-decl name='__int32_t' type-id='95e97e5e' id='33f57a65'/>
+ <typedef-decl name='__uint32_t' type-id='f0981eeb' id='62f1140c'/>
+ <typedef-decl name='__int64_t' type-id='bd54fe1a' id='0c9942d2'/>
+ <typedef-decl name='__uint64_t' type-id='7359adad' id='8910171f'/>
+ <typedef-decl name='__off_t' type-id='bd54fe1a' id='79989e9c'/>
+ <typedef-decl name='__off64_t' type-id='bd54fe1a' id='724e4de6'/>
+ <typedef-decl name='FILE' type-id='ec1ed955' id='aa12d1ba'/>
<typedef-decl name='_IO_lock_t' type-id='48b5725f' id='bb4788fa'/>
- <class-decl name='_IO_marker' size-in-bits='192' is-struct='yes' visibility='default' id='010ae0b9'>
- <data-member access='public' layout-offset-in-bits='0'>
- <var-decl name='_next' type-id='e4c6fa61' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='64'>
- <var-decl name='_sbuf' type-id='dca988a5' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='128'>
- <var-decl name='_pos' type-id='95e97e5e' visibility='default'/>
- </data-member>
- </class-decl>
<class-decl name='_IO_FILE' size-in-bits='1728' is-struct='yes' visibility='default' id='ec1ed955'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='_flags' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='_IO_read_ptr' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='_IO_read_end' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='_IO_read_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='_IO_write_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='_IO_write_ptr' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='_IO_write_end' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='_IO_buf_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='512'>
<var-decl name='_IO_buf_end' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='576'>
<var-decl name='_IO_save_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='640'>
<var-decl name='_IO_backup_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='704'>
<var-decl name='_IO_save_end' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='768'>
<var-decl name='_markers' type-id='e4c6fa61' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='832'>
<var-decl name='_chain' type-id='dca988a5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='896'>
<var-decl name='_fileno' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='928'>
<var-decl name='_flags2' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='960'>
<var-decl name='_old_offset' type-id='79989e9c' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1024'>
<var-decl name='_cur_column' type-id='8efea9e5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1040'>
<var-decl name='_vtable_offset' type-id='28577a57' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1048'>
<var-decl name='_shortbuf' type-id='89feb1ec' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1088'>
<var-decl name='_lock' type-id='cecf4ea7' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1152'>
<var-decl name='_offset' type-id='724e4de6' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1216'>
- <var-decl name='__pad1' type-id='eaa32e2f' visibility='default'/>
+ <var-decl name='_codecvt' type-id='570f8c59' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1280'>
- <var-decl name='__pad2' type-id='eaa32e2f' visibility='default'/>
+ <var-decl name='_wide_data' type-id='c65a1f29' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1344'>
- <var-decl name='__pad3' type-id='eaa32e2f' visibility='default'/>
+ <var-decl name='_freeres_list' type-id='dca988a5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1408'>
- <var-decl name='__pad4' type-id='eaa32e2f' visibility='default'/>
+ <var-decl name='_freeres_buf' type-id='eaa32e2f' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1472'>
<var-decl name='__pad5' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1536'>
<var-decl name='_mode' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1568'>
<var-decl name='_unused2' type-id='664ac0b7' visibility='default'/>
</data-member>
</class-decl>
- <typedef-decl name='int8_t' type-id='2171a512' id='ee31ee44'/>
- <typedef-decl name='int16_t' type-id='03896e23' id='23bd8cb5'/>
- <typedef-decl name='int32_t' type-id='33f57a65' id='3ff5601b'/>
- <typedef-decl name='int64_t' type-id='0c9942d2' id='9da381c4'/>
- <typedef-decl name='uint8_t' type-id='c51d6389' id='b96825af'/>
- <typedef-decl name='uint16_t' type-id='253c2d2a' id='149c6638'/>
- <typedef-decl name='uint32_t' type-id='62f1140c' id='8f92235e'/>
- <typedef-decl name='uint64_t' type-id='8910171f' id='9c313c2d'/>
- <typedef-decl name='__int8_t' type-id='28577a57' id='2171a512'/>
- <typedef-decl name='__uint8_t' type-id='002ac4a6' id='c51d6389'/>
- <typedef-decl name='__int16_t' type-id='a2185560' id='03896e23'/>
- <typedef-decl name='__uint16_t' type-id='8efea9e5' id='253c2d2a'/>
- <typedef-decl name='__int32_t' type-id='95e97e5e' id='33f57a65'/>
- <typedef-decl name='__uint32_t' type-id='f0981eeb' id='62f1140c'/>
- <typedef-decl name='__int64_t' type-id='bd54fe1a' id='0c9942d2'/>
- <typedef-decl name='__uint64_t' type-id='7359adad' id='8910171f'/>
- <typedef-decl name='__off_t' type-id='bd54fe1a' id='79989e9c'/>
- <typedef-decl name='__off64_t' type-id='bd54fe1a' id='724e4de6'/>
- <typedef-decl name='FILE' type-id='ec1ed955' id='aa12d1ba'/>
<typedef-decl name='size_t' type-id='7359adad' id='b59d7dce'/>
<pointer-type-def type-id='aa12d1ba' size-in-bits='64' id='822cd80b'/>
<pointer-type-def type-id='ec1ed955' size-in-bits='64' id='dca988a5'/>
+ <pointer-type-def type-id='a4036571' size-in-bits='64' id='570f8c59'/>
<pointer-type-def type-id='bb4788fa' size-in-bits='64' id='cecf4ea7'/>
<pointer-type-def type-id='010ae0b9' size-in-bits='64' id='e4c6fa61'/>
+ <pointer-type-def type-id='79bd3751' size-in-bits='64' id='c65a1f29'/>
<pointer-type-def type-id='c19b74c3' size-in-bits='64' id='37e3bd22'/>
<pointer-type-def type-id='a84c031d' size-in-bits='64' id='26a90f95'/>
<pointer-type-def type-id='26a90f95' size-in-bits='64' id='9b23c9ad'/>
<qualified-type-def type-id='a84c031d' const='yes' id='9b45d938'/>
<pointer-type-def type-id='9b45d938' size-in-bits='64' id='80f4b756'/>
<pointer-type-def type-id='9f88f76e' size-in-bits='64' id='7ef0e988'/>
<pointer-type-def type-id='c5bb1a2b' size-in-bits='64' id='c0d0f877'/>
<pointer-type-def type-id='573fea1b' size-in-bits='64' id='de20bf07'/>
<pointer-type-def type-id='70284cc6' size-in-bits='64' id='3be4d568'/>
<pointer-type-def type-id='700c3bca' size-in-bits='64' id='6d994334'/>
<pointer-type-def type-id='18ac1860' size-in-bits='64' id='506ab59a'/>
<pointer-type-def type-id='328fee42' size-in-bits='64' id='750cc41c'/>
<pointer-type-def type-id='7ba5cd31' size-in-bits='64' id='aaea91b5'/>
<pointer-type-def type-id='a86d8029' size-in-bits='64' id='42257af5'/>
<pointer-type-def type-id='0b4eb914' size-in-bits='64' id='19ea27ae'/>
<pointer-type-def type-id='c6c8144e' size-in-bits='64' id='2835af80'/>
<pointer-type-def type-id='20f7b475' size-in-bits='64' id='6a2f50c1'/>
<pointer-type-def type-id='102ee17a' size-in-bits='64' id='e1c54c3c'/>
<pointer-type-def type-id='49b69c77' size-in-bits='64' id='ea6be4eb'/>
<pointer-type-def type-id='cb5d50f1' size-in-bits='64' id='1708018d'/>
<pointer-type-def type-id='880d56b8' size-in-bits='64' id='d2af7f32'/>
<pointer-type-def type-id='a739bfc6' size-in-bits='64' id='506696a8'/>
<pointer-type-def type-id='234f35e8' size-in-bits='64' id='8a1fb33a'/>
<pointer-type-def type-id='41f7168a' size-in-bits='64' id='f10f1e84'/>
<pointer-type-def type-id='e8d6e508' size-in-bits='64' id='90174072'/>
<pointer-type-def type-id='f3daafe5' size-in-bits='64' id='0b22f759'/>
<pointer-type-def type-id='17ab04ad' size-in-bits='64' id='39b623f9'/>
<pointer-type-def type-id='256cdd75' size-in-bits='64' id='7391ed39'/>
<pointer-type-def type-id='cc10a041' size-in-bits='64' id='ed6a3a3d'/>
<pointer-type-def type-id='9fd269d3' size-in-bits='64' id='292cdbcf'/>
<pointer-type-def type-id='3bd73b0c' size-in-bits='64' id='7e85a9b6'/>
<pointer-type-def type-id='0d445e26' size-in-bits='64' id='330cc0d0'/>
<pointer-type-def type-id='e4b89f30' size-in-bits='64' id='ed8aa8ba'/>
<pointer-type-def type-id='be7f4941' size-in-bits='64' id='2809de35'/>
<pointer-type-def type-id='fe5ae69d' size-in-bits='64' id='90d5edb9'/>
<pointer-type-def type-id='2783af3c' size-in-bits='64' id='e44553b6'/>
<pointer-type-def type-id='33c6e3d8' size-in-bits='64' id='1263777a'/>
<pointer-type-def type-id='dadb9eca' size-in-bits='64' id='cbda43ac'/>
<pointer-type-def type-id='55b9e070' size-in-bits='64' id='b3fae562'/>
<pointer-type-def type-id='8e63c78b' size-in-bits='64' id='8b41e457'/>
<pointer-type-def type-id='c542ed33' size-in-bits='64' id='f9668a57'/>
<pointer-type-def type-id='5dea179a' size-in-bits='64' id='001d8764'/>
<pointer-type-def type-id='b6f659a0' size-in-bits='64' id='44f188f2'/>
<pointer-type-def type-id='2765bd17' size-in-bits='64' id='976f721b'/>
<pointer-type-def type-id='9e073b5c' size-in-bits='64' id='ee62ad8e'/>
<pointer-type-def type-id='2c785071' size-in-bits='64' id='957d9f35'/>
<pointer-type-def type-id='aad19bf7' size-in-bits='64' id='4db8acf3'/>
<pointer-type-def type-id='0660e71a' size-in-bits='64' id='0ca7b13c'/>
<pointer-type-def type-id='250287b8' size-in-bits='64' id='a91bad5a'/>
<pointer-type-def type-id='e7344862' size-in-bits='64' id='519bf35c'/>
<pointer-type-def type-id='32b6d968' size-in-bits='64' id='92988dea'/>
<pointer-type-def type-id='5c975642' size-in-bits='64' id='7f8ee7e4'/>
<pointer-type-def type-id='0155b993' size-in-bits='64' id='2c8c4457'/>
<pointer-type-def type-id='6e8b02cb' size-in-bits='64' id='eb944897'/>
<pointer-type-def type-id='d434b7d7' size-in-bits='64' id='108e6453'/>
<pointer-type-def type-id='c645e10f' size-in-bits='64' id='5cbe16ab'/>
<pointer-type-def type-id='de41f295' size-in-bits='64' id='d94cdfa1'/>
<pointer-type-def type-id='b2fbf64a' size-in-bits='64' id='470a7fd4'/>
<pointer-type-def type-id='cc22d314' size-in-bits='64' id='eddda806'/>
<pointer-type-def type-id='23bd8cb5' size-in-bits='64' id='f76f73d0'/>
<pointer-type-def type-id='3ff5601b' size-in-bits='64' id='4aafb922'/>
<pointer-type-def type-id='9da381c4' size-in-bits='64' id='cb785ebf'/>
<pointer-type-def type-id='ee31ee44' size-in-bits='64' id='256d5229'/>
<pointer-type-def type-id='ebc6735b' size-in-bits='64' id='7be54adb'/>
<pointer-type-def type-id='d2e8bad9' size-in-bits='64' id='196db161'/>
<pointer-type-def type-id='8e8d4be3' size-in-bits='64' id='5ce45b60'/>
<pointer-type-def type-id='5ce45b60' size-in-bits='64' id='857bb57e'/>
<pointer-type-def type-id='57928edf' size-in-bits='64' id='3fa542f0'/>
+ <pointer-type-def type-id='b48d2441' size-in-bits='64' id='33976309'/>
<pointer-type-def type-id='aca3bac8' size-in-bits='64' id='d33f11cb'/>
<pointer-type-def type-id='d8bf0010' size-in-bits='64' id='45b65157'/>
<pointer-type-def type-id='149c6638' size-in-bits='64' id='8a121f49'/>
<pointer-type-def type-id='8f92235e' size-in-bits='64' id='90421557'/>
<pointer-type-def type-id='9c313c2d' size-in-bits='64' id='5d6479ae'/>
<pointer-type-def type-id='b96825af' size-in-bits='64' id='ae3e8ca6'/>
<pointer-type-def type-id='002ac4a6' size-in-bits='64' id='cf536864'/>
<pointer-type-def type-id='48b5725f' size-in-bits='64' id='eaa32e2f'/>
+ <class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
+ <class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
+ <class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
+ <class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
<function-decl name='nvlist_prtctl_setdest' mangled-name='nvlist_prtctl_setdest' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_setdest'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='822cd80b' name='fp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctl_getdest' mangled-name='nvlist_prtctl_getdest' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_getdest'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<return type-id='822cd80b'/>
</function-decl>
<function-decl name='nvlist_prtctl_setindent' mangled-name='nvlist_prtctl_setindent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_setindent'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='628aafab' name='mode'/>
<parameter type-id='95e97e5e' name='start'/>
<parameter type-id='95e97e5e' name='inc'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctl_doindent' mangled-name='nvlist_prtctl_doindent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_doindent'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='95e97e5e' name='onemore'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctl_setfmt' mangled-name='nvlist_prtctl_setfmt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_setfmt'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='c8dcc53a' name='which'/>
<parameter type-id='80f4b756' name='fmt'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctl_dofmt' mangled-name='nvlist_prtctl_dofmt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_dofmt'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='c8dcc53a' name='which'/>
<parameter is-variadic='yes'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_boolean' mangled-name='nvlist_prtctlop_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_boolean'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='1263777a' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_boolean_value' mangled-name='nvlist_prtctlop_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_boolean_value'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='976f721b' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_byte' mangled-name='nvlist_prtctlop_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_byte'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='519bf35c' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_int8' mangled-name='nvlist_prtctlop_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int8'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='a91bad5a' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_uint8' mangled-name='nvlist_prtctlop_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint8'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='eb944897' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_int16' mangled-name='nvlist_prtctlop_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int16'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='957d9f35' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_uint16' mangled-name='nvlist_prtctlop_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint16'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='92988dea' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_int32' mangled-name='nvlist_prtctlop_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int32'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='4db8acf3' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_uint32' mangled-name='nvlist_prtctlop_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint32'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='7f8ee7e4' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_int64' mangled-name='nvlist_prtctlop_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int64'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='0ca7b13c' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_uint64' mangled-name='nvlist_prtctlop_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint64'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='2c8c4457' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_double' mangled-name='nvlist_prtctlop_double' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_double'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='e44553b6' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_string' mangled-name='nvlist_prtctlop_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_string'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='2809de35' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_hrtime' mangled-name='nvlist_prtctlop_hrtime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_hrtime'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='ee62ad8e' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_nvlist' mangled-name='nvlist_prtctlop_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_nvlist'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='001d8764' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_boolean_array' mangled-name='nvlist_prtctlop_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_boolean_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='ed8aa8ba' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_byte_array' mangled-name='nvlist_prtctlop_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_byte_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='108e6453' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_int8_array' mangled-name='nvlist_prtctlop_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int8_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='f9668a57' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_uint8_array' mangled-name='nvlist_prtctlop_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint8_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='eddda806' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_int16_array' mangled-name='nvlist_prtctlop_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int16_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='cbda43ac' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_uint16_array' mangled-name='nvlist_prtctlop_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint16_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='5cbe16ab' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_int32_array' mangled-name='nvlist_prtctlop_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int32_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='b3fae562' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_uint32_array' mangled-name='nvlist_prtctlop_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint32_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='d94cdfa1' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_int64_array' mangled-name='nvlist_prtctlop_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int64_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='8b41e457' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_uint64_array' mangled-name='nvlist_prtctlop_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint64_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='470a7fd4' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_string_array' mangled-name='nvlist_prtctlop_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_string_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='90d5edb9' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctlop_nvlist_array' mangled-name='nvlist_prtctlop_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_nvlist_array'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<parameter type-id='44f188f2' name='func'/>
<parameter type-id='eaa32e2f' name='private'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prtctl_alloc' mangled-name='nvlist_prtctl_alloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_alloc'>
<return type-id='b0c1ff8d'/>
</function-decl>
<function-decl name='nvlist_prtctl_free' mangled-name='nvlist_prtctl_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_free'>
<parameter type-id='b0c1ff8d' name='pctl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_print' mangled-name='nvlist_print' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_print'>
<parameter type-id='822cd80b' name='fp'/>
<parameter type-id='5ce45b60' name='nvl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvlist_prt' mangled-name='nvlist_prt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prt'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='b0c1ff8d' name='pctl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='dump_nvlist' mangled-name='dump_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='dump_nvlist'>
<parameter type-id='5ce45b60' name='list'/>
<parameter type-id='95e97e5e' name='indent'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='nvpair_value_match_regex' mangled-name='nvpair_value_match_regex' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_match_regex'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='95e97e5e' name='ai'/>
<parameter type-id='26a90f95' name='value'/>
<parameter type-id='d33f11cb' name='value_regex'/>
<parameter type-id='9b23c9ad' name='ep'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='nvpair_value_match' mangled-name='nvpair_value_match' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_match'>
<parameter type-id='3fa542f0' name='nvp'/>
<parameter type-id='95e97e5e' name='ai'/>
<parameter type-id='26a90f95' name='value'/>
<parameter type-id='9b23c9ad' name='ep'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-type size-in-bits='64' id='9f88f76e'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='37e3bd22'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='c5bb1a2b'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='26a90f95'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='573fea1b'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='9b23c9ad'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='70284cc6'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='a0eb0f08'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='700c3bca'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='95e97e5e'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='18ac1860'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='f76f73d0'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='328fee42'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='4aafb922'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='7ba5cd31'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='cb785ebf'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='a86d8029'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='256d5229'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='0b4eb914'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='5ce45b60'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='c6c8144e'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='857bb57e'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='20f7b475'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='c19b74c3'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='102ee17a'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='cebdd548'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='49b69c77'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='23bd8cb5'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='cb5d50f1'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='3ff5601b'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='880d56b8'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='9da381c4'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='a739bfc6'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='ee31ee44'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='234f35e8'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='d8bf0010'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='41f7168a'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='149c6638'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='e8d6e508'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='8f92235e'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='f3daafe5'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='9c313c2d'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='17ab04ad'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='b96825af'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='256cdd75'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='45b65157'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='cc10a041'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='8a121f49'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='9fd269d3'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='90421557'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='3bd73b0c'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='5d6479ae'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='0d445e26'>
<parameter type-id='196db161'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='ae3e8ca6'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='e4b89f30'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='37e3bd22'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='be7f4941'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='26a90f95'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='fe5ae69d'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='9b23c9ad'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='2783af3c'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='a0eb0f08'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='33c6e3d8'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='95e97e5e'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='dadb9eca'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='f76f73d0'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='55b9e070'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='4aafb922'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='8e63c78b'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='cb785ebf'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='c542ed33'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='256d5229'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='5dea179a'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='5ce45b60'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='b6f659a0'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='857bb57e'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='2765bd17'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='c19b74c3'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='9e073b5c'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='cebdd548'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='2c785071'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='23bd8cb5'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='aad19bf7'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='3ff5601b'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='0660e71a'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='9da381c4'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='250287b8'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='ee31ee44'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='e7344862'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='d8bf0010'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='32b6d968'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='149c6638'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='5c975642'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='8f92235e'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='0155b993'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='9c313c2d'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='6e8b02cb'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='b96825af'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='d434b7d7'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='45b65157'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='c645e10f'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='8a121f49'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='de41f295'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='90421557'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='b2fbf64a'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='5d6479ae'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='cc22d314'>
<parameter type-id='b0c1ff8d'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<parameter type-id='80f4b756'/>
<parameter type-id='ae3e8ca6'/>
<parameter type-id='3502e3ff'/>
<return type-id='95e97e5e'/>
</function-type>
- <type-decl name='void' id='48b5725f'/>
</abi-instr>
<abi-instr address-size='64' path='libnvpair_json.c' language='LANG_C99'>
<function-decl name='nvlist_print_json' mangled-name='nvlist_print_json' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_print_json'>
<parameter type-id='822cd80b' name='fp'/>
<parameter type-id='5ce45b60' name='nvl'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='nvpair_alloc_system.c' language='LANG_C99'>
<class-decl name='__va_list_tag' size-in-bits='192' is-struct='yes' visibility='default' id='d5027220'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='gp_offset' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='fp_offset' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='overflow_arg_area' type-id='eaa32e2f' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='reg_save_area' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='nv_alloc_ops_t' type-id='8f6cc4f4' id='03e8ffd6'/>
<class-decl name='nv_alloc' size-in-bits='128' is-struct='yes' visibility='default' id='98213087'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='nva_ops' type-id='ee1d4944' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='nva_arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='nv_alloc_t' type-id='98213087' id='cca08635'/>
<class-decl name='nv_alloc_ops' size-in-bits='320' is-struct='yes' visibility='default' id='8f6cc4f4'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='nv_ao_init' type-id='76da8447' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='nv_ao_fini' type-id='fe356f6f' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='nv_ao_alloc' type-id='9ff7f508' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='nv_ao_free' type-id='520da3f4' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='nv_ao_reset' type-id='fe356f6f' visibility='default'/>
</data-member>
</class-decl>
<pointer-type-def type-id='d5027220' size-in-bits='64' id='b7f2d5e6'/>
<qualified-type-def type-id='03e8ffd6' const='yes' id='aca16c06'/>
<pointer-type-def type-id='aca16c06' size-in-bits='64' id='ee1d4944'/>
<pointer-type-def type-id='e9ff7293' size-in-bits='64' id='76da8447'/>
<pointer-type-def type-id='cca08635' size-in-bits='64' id='11871392'/>
+ <qualified-type-def type-id='11871392' const='yes' id='e1cbfe1d'/>
<pointer-type-def type-id='51a21b4b' size-in-bits='64' id='fe356f6f'/>
<pointer-type-def type-id='1169c032' size-in-bits='64' id='520da3f4'/>
<pointer-type-def type-id='9fff962e' size-in-bits='64' id='9ff7f508'/>
- <var-decl name='nv_alloc_nosleep' type-id='11871392' mangled-name='nv_alloc_nosleep' visibility='default' elf-symbol-id='nv_alloc_nosleep'/>
+ <var-decl name='nv_alloc_nosleep' type-id='e1cbfe1d' mangled-name='nv_alloc_nosleep' visibility='default' elf-symbol-id='nv_alloc_nosleep'/>
<function-type size-in-bits='64' id='e9ff7293'>
<parameter type-id='11871392'/>
<parameter type-id='b7f2d5e6'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='51a21b4b'>
<parameter type-id='11871392'/>
<return type-id='48b5725f'/>
</function-type>
<function-type size-in-bits='64' id='1169c032'>
<parameter type-id='11871392'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='b59d7dce'/>
<return type-id='48b5725f'/>
</function-type>
<function-type size-in-bits='64' id='9fff962e'>
<parameter type-id='11871392'/>
<parameter type-id='b59d7dce'/>
<return type-id='eaa32e2f'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='assert.c' language='LANG_C99'>
<var-decl name='libspl_assert_ok' type-id='95e97e5e' mangled-name='libspl_assert_ok' visibility='default' elf-symbol-id='libspl_assert_ok'/>
<function-decl name='libspl_assertf' mangled-name='libspl_assertf' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_assertf'>
<parameter type-id='80f4b756' name='file'/>
<parameter type-id='80f4b756' name='func'/>
<parameter type-id='95e97e5e' name='line'/>
<parameter type-id='80f4b756' name='format'/>
<parameter is-variadic='yes'/>
<return type-id='48b5725f'/>
</function-decl>
</abi-instr>
</abi-corpus>
diff --git a/lib/libnvpair/nvpair_alloc_system.c b/lib/libnvpair/nvpair_alloc_system.c
index 9771f58f67f3..59806ea4dc93 100644
--- a/lib/libnvpair/nvpair_alloc_system.c
+++ b/lib/libnvpair/nvpair_alloc_system.c
@@ -1,65 +1,65 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License, Version 1.0 only
* (the "License"). You may not use this file except in compliance
* with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2004 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#include <rpc/types.h>
#include <sys/kmem.h>
#include <sys/nvpair.h>
static void *
nv_alloc_sys(nv_alloc_t *nva, size_t size)
{
return (kmem_alloc(size, (int)(uintptr_t)nva->nva_arg));
}
static void
nv_free_sys(nv_alloc_t *nva, void *buf, size_t size)
{
(void) nva;
kmem_free(buf, size);
}
static const nv_alloc_ops_t system_ops = {
NULL, /* nv_ao_init() */
NULL, /* nv_ao_fini() */
nv_alloc_sys, /* nv_ao_alloc() */
nv_free_sys, /* nv_ao_free() */
NULL /* nv_ao_reset() */
};
-nv_alloc_t nv_alloc_sleep_def = {
+static nv_alloc_t nv_alloc_sleep_def = {
&system_ops,
(void *)KM_SLEEP
};
-nv_alloc_t nv_alloc_nosleep_def = {
+static nv_alloc_t nv_alloc_nosleep_def = {
&system_ops,
(void *)KM_NOSLEEP
};
-nv_alloc_t *nv_alloc_sleep = &nv_alloc_sleep_def;
-nv_alloc_t *nv_alloc_nosleep = &nv_alloc_nosleep_def;
+nv_alloc_t *const nv_alloc_sleep = &nv_alloc_sleep_def;
+nv_alloc_t *const nv_alloc_nosleep = &nv_alloc_nosleep_def;
diff --git a/lib/libspl/include/umem.h b/lib/libspl/include/umem.h
index 65f12595e64f..eee0dc97578a 100644
--- a/lib/libspl/include/umem.h
+++ b/lib/libspl/include/umem.h
@@ -1,208 +1,208 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License, Version 1.0 only
* (the "License"). You may not use this file except in compliance
* with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#ifndef _LIBSPL_UMEM_H
#define _LIBSPL_UMEM_H
/*
* XXX: We should use the real portable umem library if it is detected
* at configure time. However, if the library is not available, we can
* use a trivial malloc based implementation. This obviously impacts
* performance, but unless you are using a full userspace build of zpool for
* something other than ztest, you are likely not going to notice or care.
*
* https://labs.omniti.com/trac/portableumem
*/
#include <sys/debug.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#ifdef __cplusplus
extern "C" {
#endif
typedef void vmem_t;
/*
* Flags for umem_alloc/umem_free
*/
#define UMEM_DEFAULT 0x0000 /* normal -- may fail */
#define UMEM_NOFAIL 0x0100 /* Never fails */
/*
* Flags for umem_cache_create()
*/
#define UMC_NODEBUG 0x00020000
#define UMEM_CACHE_NAMELEN 31
typedef int umem_nofail_callback_t(void);
typedef int umem_constructor_t(void *, void *, int);
typedef void umem_destructor_t(void *, void *);
typedef void umem_reclaim_t(void *);
typedef struct umem_cache {
char cache_name[UMEM_CACHE_NAMELEN + 1];
size_t cache_bufsize;
size_t cache_align;
umem_constructor_t *cache_constructor;
umem_destructor_t *cache_destructor;
umem_reclaim_t *cache_reclaim;
void *cache_private;
void *cache_arena;
int cache_cflags;
} umem_cache_t;
/* Prototypes for functions to provide defaults for umem envvars */
const char *_umem_debug_init(void);
const char *_umem_options_init(void);
const char *_umem_logging_init(void);
static inline void *
umem_alloc(size_t size, int flags)
{
void *ptr = NULL;
do {
ptr = malloc(size);
} while (ptr == NULL && (flags & UMEM_NOFAIL));
return (ptr);
}
static inline void *
umem_alloc_aligned(size_t size, size_t align, int flags)
{
void *ptr = NULL;
int rc = EINVAL;
do {
rc = posix_memalign(&ptr, align, size);
} while (rc == ENOMEM && (flags & UMEM_NOFAIL));
if (rc == EINVAL) {
fprintf(stderr, "%s: invalid memory alignment (%zd)\n",
__func__, align);
if (flags & UMEM_NOFAIL)
abort();
return (NULL);
}
return (ptr);
}
static inline void *
umem_zalloc(size_t size, int flags)
{
void *ptr = NULL;
ptr = umem_alloc(size, flags);
if (ptr)
memset(ptr, 0, size);
return (ptr);
}
static inline void
-umem_free(void *ptr, size_t size __maybe_unused)
+umem_free(const void *ptr, size_t size __maybe_unused)
{
- free(ptr);
+ free((void *)ptr);
}
static inline void
umem_nofail_callback(umem_nofail_callback_t *cb __maybe_unused)
{}
static inline umem_cache_t *
umem_cache_create(
char *name, size_t bufsize, size_t align,
umem_constructor_t *constructor,
umem_destructor_t *destructor,
umem_reclaim_t *reclaim,
void *priv, void *vmp, int cflags)
{
umem_cache_t *cp;
cp = (umem_cache_t *)umem_alloc(sizeof (umem_cache_t), UMEM_DEFAULT);
if (cp) {
strlcpy(cp->cache_name, name, UMEM_CACHE_NAMELEN);
cp->cache_bufsize = bufsize;
cp->cache_align = align;
cp->cache_constructor = constructor;
cp->cache_destructor = destructor;
cp->cache_reclaim = reclaim;
cp->cache_private = priv;
cp->cache_arena = vmp;
cp->cache_cflags = cflags;
}
return (cp);
}
static inline void
umem_cache_destroy(umem_cache_t *cp)
{
umem_free(cp, sizeof (umem_cache_t));
}
static inline void *
umem_cache_alloc(umem_cache_t *cp, int flags)
{
void *ptr = NULL;
if (cp->cache_align != 0)
ptr = umem_alloc_aligned(
cp->cache_bufsize, cp->cache_align, flags);
else
ptr = umem_alloc(cp->cache_bufsize, flags);
if (ptr && cp->cache_constructor)
cp->cache_constructor(ptr, cp->cache_private, UMEM_DEFAULT);
return (ptr);
}
static inline void
umem_cache_free(umem_cache_t *cp, void *ptr)
{
if (cp->cache_destructor)
cp->cache_destructor(ptr, cp->cache_private);
umem_free(ptr, cp->cache_bufsize);
}
static inline void
umem_cache_reap_now(umem_cache_t *cp __maybe_unused)
{
}
#ifdef __cplusplus
}
#endif
#endif
diff --git a/lib/libzfs/libzfs.abi b/lib/libzfs/libzfs.abi
index 1defa9c24c00..b682f4f7091f 100644
--- a/lib/libzfs/libzfs.abi
+++ b/lib/libzfs/libzfs.abi
@@ -1,5853 +1,5851 @@
<abi-corpus version='2.0' architecture='elf-amd-x86_64' soname='libzfs.so.4'>
<elf-needed>
<dependency name='libzfs_core.so.3'/>
<dependency name='libnvpair.so.3'/>
<dependency name='libuuid.so.1'/>
<dependency name='librt.so.1'/>
<dependency name='libblkid.so.1'/>
<dependency name='libudev.so.1'/>
<dependency name='libuutil.so.3'/>
<dependency name='libm.so.6'/>
<dependency name='libcrypto.so.1.1'/>
<dependency name='libz.so.1'/>
- <dependency name='libdl.so.2'/>
<dependency name='libpthread.so.0'/>
<dependency name='libc.so.6'/>
<dependency name='ld-linux-x86-64.so.2'/>
</elf-needed>
<elf-function-symbols>
- <elf-symbol name='_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
- <elf-symbol name='_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='_sol_getmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_add_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_and_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_cas_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_cas_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_cas_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_cas_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_cas_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_cas_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_cas_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_cas_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_cas_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_clear_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_dec_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_inc_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_or_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_set_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_sub_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_swap_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_swap_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_swap_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_swap_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_swap_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_swap_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_swap_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_swap_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='atomic_swap_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_destroy_nodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_find' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_first' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_insert' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_insert_here' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_is_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_last' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_nearest' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_numnodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_swap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_update' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_update_gt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_update_lt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='avl_walk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='bookmark_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='cityhash4' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='color_end' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='color_start' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='dataset_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='dataset_nestcheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='efi_alloc_and_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='efi_alloc_and_read' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='efi_err_check' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='efi_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='efi_rescan' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='efi_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='efi_use_whole_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='efi_write' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='entity_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_2_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_2_incremental_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_2_incremental_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_2_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_impl_set' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_incremental_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_incremental_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_native_varsize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='get_dataset_depth' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='get_system_hostid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='getexecname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='getextmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='getmntany' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='getprop_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='getzoneid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='is_mounted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='is_mpath_whole_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libspl_assertf' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_add_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_envvar_is_set' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_errno' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_error_action' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_error_description' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_error_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_free_str_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_mnttab_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_mnttab_cache' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_mnttab_find' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_mnttab_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_mnttab_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_mnttab_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_print_on_error' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_run_process' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_run_process_get_stdout' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_run_process_get_stdout_nopath' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_insert_after' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_insert_before' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_insert_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_insert_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_is_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_link_active' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_link_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_link_replace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_move_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_prev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_remove_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_remove_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='list_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='membar_consumer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='membar_enter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='membar_exit' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='membar_producer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='mkdirp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='mountpoint_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='permset_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='pool_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='print_timestamp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='printf_color' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='sa_commit_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='sa_disable_share' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='sa_enable_share' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='sa_errorstr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='sa_is_shared' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='sa_validate_shareopts' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='snapshot_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='spl_pagesize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='strlcat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='strlcpy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='tpool_abandon' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='tpool_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='tpool_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='tpool_dispatch' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='tpool_member' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='tpool_resume' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='tpool_suspend' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='tpool_suspended' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='tpool_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='update_vdev_config_dev_strs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_expand_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_align_right' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_column_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_default_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_default_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_get_table' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_readonly' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_user' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='vdev_prop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfeature_depends_on' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfeature_is_supported' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfeature_is_valid_guid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfeature_lookup_guid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfeature_lookup_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_adjust_mount_options' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_allocatable_devs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_append_partition' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_basename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_bookmark_exists' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_clone' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_close' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_commit_all_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_commit_nfs_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_commit_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_commit_smb_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_component_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_create_ancestors' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_crypto_attempt_load_keys' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_crypto_clone_check' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_crypto_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_crypto_get_encryption_root' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_crypto_load_key' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_crypto_rewrap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_crypto_unload_key' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_dataset_exists' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_dataset_name_hidden' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_deleg_canonicalize_perm' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_deleg_verify_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_deleg_whokey' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_destroy_snaps' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_destroy_snaps_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_destroy_snaps_nvl_os' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_dev_flush' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_dev_is_dm' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_dev_is_whole_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_device_get_devid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_device_get_physical' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_dirnamelen' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_expand_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_foreach_mountpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_all_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_clones_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_enclosure_sysfs_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_fsacl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_holds' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_pool_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_pool_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_recvd_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_underlying_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_underlying_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_get_user_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_handle_dup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_hold' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_hold_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_ioctl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_is_mounted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_is_shared' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_is_shared_nfs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_is_shared_smb' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_isnumber' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_iter_bookmarks' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_iter_children' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_iter_dependents' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_iter_filesystems' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_iter_mounted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_iter_root' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_iter_snapshots' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_iter_snapshots_sorted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_iter_snapspec' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_mod_supported' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_mount' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_mount_at' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_mount_delegation_check' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_name_valid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_nicebytes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_nicenum' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_nicenum_format' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_niceraw' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_nicestrtonum' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_nicetime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_open' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_parent_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_parse_mount_options' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_path_to_zhandle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_promote' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_align_right' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_column_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_default_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_default_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_delegatable' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_encryption_key_param' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get_recvd' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get_table' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get_userquota' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get_userquota_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get_written' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_get_written_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_inherit' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_inheritable' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_is_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_readonly' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_set' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_set_list' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_setonce' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_user' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_userquota' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_valid_for_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_valid_keylocation' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_visible' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prop_written' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_prune_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_receive' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_refresh_properties' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_release' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_rename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_resolve_shortname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_rollback' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_save_arguments' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_send' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_send_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_send_progress' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_send_resume' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_send_resume_token_to_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_send_saved' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_set_fsacl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_share' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_share_nfs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_share_smb' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_shareall' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_show_diffs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_smb_acl_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_smb_acl_purge' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_smb_acl_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_smb_acl_rename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_snapshot' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_snapshot_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_spa_version' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_spa_version_map' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_special_devs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_standard_error' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_strcmp_pathname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_strip_partition' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_strip_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_type_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unmount' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unmountall' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unshare' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unshare_nfs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unshare_smb' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unshareall' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unshareall_bypath' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unshareall_bytype' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unshareall_nfs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_unshareall_smb' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_userspace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_valid_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_version_kernel' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_version_print' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_version_userland' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_wait_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_zpl_version_map' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_checkpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_clear' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_clear_label' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_close' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_default_search_paths' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_disable_datasets' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_disable_datasets_os' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_disable_volume_os' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_discard_checkpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_dump_ddt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_enable_datasets' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_events_clear' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_events_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_events_seek' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_expand_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_explain_recover' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_export' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_export_force' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_feature_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_find_config' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_find_vdev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_find_vdev_by_physpath' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_free_handles' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_all_vdev_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_bootenv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_config' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_errlog' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_features' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_history' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_load_policy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_physpath' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_prop_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_state' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_state_str' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_vdev_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_get_vdev_prop_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_history_unpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_import' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_import_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_import_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_in_use' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_initialize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_initialize_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_is_draid_spare' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_iter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_label_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_label_disk_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_load_compat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_log_history' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_obj_to_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_obj_to_path_ds' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_open' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_open_canfail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_pool_state_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_print_unsup_feat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_align_right' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_column_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_default_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_default_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_feature' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_get_feature' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_get_table' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_readonly' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_setonce' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_unsupported' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_prop_vdev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_props_refresh' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_read_label' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_refresh_stats' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_reguid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_reopen_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_scan' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_search_import' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_set_bootenv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_set_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_set_vdev_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_skip_pool' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_state_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_sync_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_trim' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_upgrade' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_attach' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_clear' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_degrade' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_detach' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_fault' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_indirect_size' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_offline' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_online' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_path_to_guid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_remove_cancel' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_vdev_split' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zpool_wait_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_free_list' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_get_list' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_iter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_iter_common' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_print_one_property' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_register_hidden' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_register_impl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_register_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_register_number' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_register_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_valid_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_valid_for_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zprop_width' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zvol_volsize_to_reservation' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
</elf-function-symbols>
<elf-variable-symbols>
<elf-symbol name='efi_debug' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_abd_ops' size='24' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_avx2_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_avx512bw_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_avx512f_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_sse2_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_ssse3_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_superscalar4_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='fletcher_4_superscalar_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libspl_assert_ok' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='libzfs_config_ops' size='16' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='spa_feature_table' size='1904' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfeature_checks_disable' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_deleg_perm_tab' size='512' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_history_event_names' size='328' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_max_dataset_nesting' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
<elf-symbol name='zfs_userquota_prop_prefixes' size='96' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
</elf-variable-symbols>
<abi-instr address-size='64' path='../../module/avl/avl.c' language='LANG_C99'>
<typedef-decl name='avl_index_t' type-id='e475ab95' id='fba6cb51'/>
<pointer-type-def type-id='fba6cb51' size-in-bits='64' id='32adbf30'/>
<pointer-type-def type-id='eaa32e2f' size-in-bits='64' id='63e171df'/>
<function-decl name='avl_walk' mangled-name='avl_walk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_walk'>
<parameter type-id='a3681dea' name='tree'/>
<parameter type-id='eaa32e2f' name='oldnode'/>
<parameter type-id='95e97e5e' name='left'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='avl_first' mangled-name='avl_first' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_first'>
<parameter type-id='a3681dea' name='tree'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='avl_last' mangled-name='avl_last' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_last'>
<parameter type-id='a3681dea' name='tree'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='avl_nearest' mangled-name='avl_nearest' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_nearest'>
<parameter type-id='a3681dea' name='tree'/>
<parameter type-id='fba6cb51' name='where'/>
<parameter type-id='95e97e5e' name='direction'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='avl_find' mangled-name='avl_find' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_find'>
<parameter type-id='a3681dea' name='tree'/>
<parameter type-id='eaa32e2f' name='value'/>
<parameter type-id='32adbf30' name='where'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='avl_insert' mangled-name='avl_insert' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_insert'>
<parameter type-id='a3681dea' name='tree'/>
<parameter type-id='eaa32e2f' name='new_data'/>
<parameter type-id='fba6cb51' name='where'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='avl_insert_here' mangled-name='avl_insert_here' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_insert_here'>
<parameter type-id='a3681dea' name='tree'/>
<parameter type-id='eaa32e2f' name='new_data'/>
<parameter type-id='eaa32e2f' name='here'/>
<parameter type-id='95e97e5e' name='direction'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='avl_add' mangled-name='avl_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_add'>
<parameter type-id='a3681dea' name='tree'/>
<parameter type-id='eaa32e2f' name='new_node'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='avl_remove' mangled-name='avl_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_remove'>
<parameter type-id='a3681dea' name='tree'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='avl_update_lt' mangled-name='avl_update_lt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update_lt'>
<parameter type-id='a3681dea' name='t'/>
<parameter type-id='eaa32e2f' name='obj'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='avl_update_gt' mangled-name='avl_update_gt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update_gt'>
<parameter type-id='a3681dea' name='t'/>
<parameter type-id='eaa32e2f' name='obj'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='avl_update' mangled-name='avl_update' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update'>
<parameter type-id='a3681dea' name='t'/>
<parameter type-id='eaa32e2f' name='obj'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='avl_swap' mangled-name='avl_swap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_swap'>
<parameter type-id='a3681dea' name='tree1'/>
<parameter type-id='a3681dea' name='tree2'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='avl_create' mangled-name='avl_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_create'>
<parameter type-id='a3681dea' name='tree'/>
<parameter type-id='585e1de9' name='compar'/>
<parameter type-id='b59d7dce' name='size'/>
<parameter type-id='b59d7dce' name='offset'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='avl_destroy' mangled-name='avl_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_destroy'>
<parameter type-id='a3681dea' name='tree'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='avl_numnodes' mangled-name='avl_numnodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_numnodes'>
<parameter type-id='a3681dea' name='tree'/>
<return type-id='ee1f298e'/>
</function-decl>
<function-decl name='avl_is_empty' mangled-name='avl_is_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_is_empty'>
<parameter type-id='a3681dea' name='tree'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='avl_destroy_nodes' mangled-name='avl_destroy_nodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_destroy_nodes'>
<parameter type-id='a3681dea' name='tree'/>
<parameter type-id='63e171df' name='cookie'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-type size-in-bits='64' id='96ee24a5'>
<parameter type-id='eaa32e2f'/>
<parameter type-id='eaa32e2f'/>
<return type-id='95e97e5e'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='rdwr_efi.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='288' id='16e6f2c6'>
<subrange length='36' type-id='7359adad' id='ae666bde'/>
</array-type-def>
<array-type-def dimensions='1' type-id='a65ae39c' size-in-bits='960' id='fa198beb'>
<subrange length='1' type-id='7359adad' id='52f813b4'/>
</array-type-def>
<array-type-def dimensions='1' type-id='b96825af' size-in-bits='48' id='0f562bd0'>
<subrange length='6' type-id='7359adad' id='52fa524b'/>
</array-type-def>
<array-type-def dimensions='1' type-id='3502e3ff' size-in-bits='384' id='dba89ba3'>
<subrange length='12' type-id='7359adad' id='84827bdc'/>
</array-type-def>
<array-type-def dimensions='1' type-id='3502e3ff' size-in-bits='256' id='01d84ed4'>
<subrange length='8' type-id='7359adad' id='56e0c0b1'/>
</array-type-def>
<class-decl name='dk_part' size-in-bits='960' is-struct='yes' visibility='default' id='a65ae39c'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='p_start' type-id='804dc465' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='p_size' type-id='804dc465' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='p_guid' type-id='214f32ea' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='p_tag' type-id='d908a348' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='272'>
<var-decl name='p_flag' type-id='d908a348' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='288'>
<var-decl name='p_name' type-id='16e6f2c6' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='576'>
<var-decl name='p_uguid' type-id='214f32ea' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='704'>
<var-decl name='p_resv' type-id='01d84ed4' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='dk_gpt' size-in-bits='1920' is-struct='yes' visibility='default' id='dd4a2e5a'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='efi_version' type-id='3502e3ff' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='efi_nparts' type-id='3502e3ff' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='efi_part_size' type-id='3502e3ff' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='96'>
<var-decl name='efi_lbasize' type-id='3502e3ff' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='efi_last_lba' type-id='804dc465' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='efi_first_u_lba' type-id='804dc465' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='efi_last_u_lba' type-id='804dc465' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='efi_disk_uguid' type-id='214f32ea' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='efi_flags' type-id='3502e3ff' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='480'>
<var-decl name='efi_reserved1' type-id='3502e3ff' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='512'>
<var-decl name='efi_altern_lba' type-id='804dc465' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='576'>
<var-decl name='efi_reserved' type-id='dba89ba3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='960'>
<var-decl name='efi_parts' type-id='fa198beb' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='uuid' size-in-bits='128' is-struct='yes' visibility='default' id='214f32ea'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='time_low' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='time_mid' type-id='149c6638' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='48'>
<var-decl name='time_hi_and_version' type-id='149c6638' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='clock_seq_hi_and_reserved' type-id='b96825af' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='72'>
<var-decl name='clock_seq_low' type-id='b96825af' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='80'>
<var-decl name='node_addr' type-id='0f562bd0' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='ushort_t' type-id='8efea9e5' id='d908a348'/>
<pointer-type-def type-id='dd4a2e5a' size-in-bits='64' id='0d8119a8'/>
<pointer-type-def type-id='0d8119a8' size-in-bits='64' id='c43b27a6'/>
<var-decl name='efi_debug' type-id='95e97e5e' mangled-name='efi_debug' visibility='default' elf-symbol-id='efi_debug'/>
<function-decl name='efi_alloc_and_init' mangled-name='efi_alloc_and_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_alloc_and_init'>
<parameter type-id='95e97e5e' name='fd'/>
<parameter type-id='8f92235e' name='nparts'/>
<parameter type-id='c43b27a6' name='vtoc'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='efi_alloc_and_read' mangled-name='efi_alloc_and_read' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_alloc_and_read'>
<parameter type-id='95e97e5e' name='fd'/>
<parameter type-id='c43b27a6' name='vtoc'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='efi_rescan' mangled-name='efi_rescan' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_rescan'>
<parameter type-id='95e97e5e' name='fd'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='efi_use_whole_disk' mangled-name='efi_use_whole_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_use_whole_disk'>
<parameter type-id='95e97e5e' name='fd'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='efi_write' mangled-name='efi_write' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_write'>
<parameter type-id='95e97e5e' name='fd'/>
<parameter type-id='0d8119a8' name='vtoc'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='efi_free' mangled-name='efi_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_free'>
<parameter type-id='0d8119a8' name='ptr'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='efi_type' mangled-name='efi_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_type'>
<parameter type-id='95e97e5e' name='fd'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='efi_err_check' mangled-name='efi_err_check' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_err_check'>
<parameter type-id='0d8119a8' name='vtoc'/>
<return type-id='48b5725f'/>
</function-decl>
<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
</abi-instr>
<abi-instr address-size='64' path='libshare.c' language='LANG_C99'>
<function-decl name='sa_enable_share' mangled-name='sa_enable_share' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_enable_share'>
<parameter type-id='80f4b756' name='zfsname'/>
<parameter type-id='80f4b756' name='mountpoint'/>
<parameter type-id='80f4b756' name='shareopts'/>
<parameter type-id='26a90f95' name='protocol'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='sa_disable_share' mangled-name='sa_disable_share' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_disable_share'>
<parameter type-id='80f4b756' name='mountpoint'/>
<parameter type-id='26a90f95' name='protocol'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='sa_is_shared' mangled-name='sa_is_shared' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_is_shared'>
<parameter type-id='80f4b756' name='mountpoint'/>
<parameter type-id='26a90f95' name='protocol'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='sa_commit_shares' mangled-name='sa_commit_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_commit_shares'>
<parameter type-id='80f4b756' name='protocol'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='sa_errorstr' mangled-name='sa_errorstr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_errorstr'>
<parameter type-id='95e97e5e' name='err'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='sa_validate_shareopts' mangled-name='sa_validate_shareopts' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_validate_shareopts'>
<parameter type-id='26a90f95' name='options'/>
<parameter type-id='26a90f95' name='proto'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='os/linux/smb.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='2040' id='11641789'>
<subrange length='255' type-id='7359adad' id='36e7f891'/>
</array-type-def>
<class-decl name='smb_share_s' size-in-bits='36992' is-struct='yes' visibility='default' id='a75bc907'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='name' type-id='11641789' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2040'>
<var-decl name='path' type-id='d16c6df4' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='34808'>
<var-decl name='comment' type-id='11641789' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='36864'>
<var-decl name='guest_ok' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='36928'>
<var-decl name='next' type-id='05ed1c5f' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='smb_share_t' type-id='a75bc907' id='2d05afd9'/>
<pointer-type-def type-id='a75bc907' size-in-bits='64' id='05ed1c5f'/>
<pointer-type-def type-id='2d05afd9' size-in-bits='64' id='a3e5c654'/>
<var-decl name='smb_shares' type-id='a3e5c654' visibility='default'/>
</abi-instr>
<abi-instr address-size='64' path='assert.c' language='LANG_C99'>
<var-decl name='libspl_assert_ok' type-id='95e97e5e' mangled-name='libspl_assert_ok' visibility='default' elf-symbol-id='libspl_assert_ok'/>
<function-decl name='libspl_assertf' mangled-name='libspl_assertf' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_assertf'>
<parameter type-id='80f4b756' name='file'/>
<parameter type-id='80f4b756' name='func'/>
<parameter type-id='95e97e5e' name='line'/>
<parameter type-id='80f4b756' name='format'/>
<parameter is-variadic='yes'/>
<return type-id='48b5725f'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='atomic.c' language='LANG_C99'>
<type-decl name='signed char' size-in-bits='8' id='28577a57'/>
<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
<typedef-decl name='int8_t' type-id='2171a512' id='ee31ee44'/>
<typedef-decl name='int16_t' type-id='03896e23' id='23bd8cb5'/>
<typedef-decl name='uint16_t' type-id='253c2d2a' id='149c6638'/>
<typedef-decl name='__int8_t' type-id='28577a57' id='2171a512'/>
<typedef-decl name='__int16_t' type-id='a2185560' id='03896e23'/>
<typedef-decl name='__uint16_t' type-id='8efea9e5' id='253c2d2a'/>
<qualified-type-def type-id='149c6638' volatile='yes' id='5120c5f7'/>
<pointer-type-def type-id='5120c5f7' size-in-bits='64' id='93977ae7'/>
<qualified-type-def type-id='8f92235e' volatile='yes' id='430e0681'/>
<pointer-type-def type-id='430e0681' size-in-bits='64' id='3a147f31'/>
<qualified-type-def type-id='b96825af' volatile='yes' id='84ff7d66'/>
<pointer-type-def type-id='84ff7d66' size-in-bits='64' id='aa323ea4'/>
<qualified-type-def type-id='ee1f298e' volatile='yes' id='6f7e09cb'/>
<pointer-type-def type-id='6f7e09cb' size-in-bits='64' id='64698d33'/>
<qualified-type-def type-id='48b5725f' volatile='yes' id='b0b3cbf9'/>
<pointer-type-def type-id='b0b3cbf9' size-in-bits='64' id='fe09dd29'/>
<function-decl name='atomic_inc_8' mangled-name='atomic_inc_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8'>
<parameter type-id='aa323ea4' name='target'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_inc_16' mangled-name='atomic_inc_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16'>
<parameter type-id='93977ae7' name='target'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_inc_32' mangled-name='atomic_inc_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32'>
<parameter type-id='3a147f31' name='target'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_inc_ulong' mangled-name='atomic_inc_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong'>
<parameter type-id='64698d33' name='target'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_dec_8' mangled-name='atomic_dec_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8'>
<parameter type-id='aa323ea4' name='target'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_dec_16' mangled-name='atomic_dec_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16'>
<parameter type-id='93977ae7' name='target'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_dec_32' mangled-name='atomic_dec_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32'>
<parameter type-id='3a147f31' name='target'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_dec_ulong' mangled-name='atomic_dec_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong'>
<parameter type-id='64698d33' name='target'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_add_8' mangled-name='atomic_add_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='ee31ee44' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_add_16' mangled-name='atomic_add_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='23bd8cb5' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_add_32' mangled-name='atomic_add_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='3ff5601b' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_add_long' mangled-name='atomic_add_long' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_long'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='bd54fe1a' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_add_ptr' mangled-name='atomic_add_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr'>
<parameter type-id='fe09dd29' name='target'/>
<parameter type-id='79a0948f' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_sub_8' mangled-name='atomic_sub_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='ee31ee44' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_sub_16' mangled-name='atomic_sub_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='23bd8cb5' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_sub_32' mangled-name='atomic_sub_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='3ff5601b' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_sub_long' mangled-name='atomic_sub_long' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_long'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='bd54fe1a' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_sub_ptr' mangled-name='atomic_sub_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr'>
<parameter type-id='fe09dd29' name='target'/>
<parameter type-id='79a0948f' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_or_8' mangled-name='atomic_or_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='b96825af' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_or_16' mangled-name='atomic_or_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='149c6638' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_or_32' mangled-name='atomic_or_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='8f92235e' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_or_ulong' mangled-name='atomic_or_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='ee1f298e' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_and_8' mangled-name='atomic_and_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='b96825af' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_and_16' mangled-name='atomic_and_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='149c6638' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_and_32' mangled-name='atomic_and_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='8f92235e' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_and_ulong' mangled-name='atomic_and_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='ee1f298e' name='bits'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='atomic_inc_8_nv' mangled-name='atomic_inc_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8_nv'>
<parameter type-id='aa323ea4' name='target'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='atomic_inc_16_nv' mangled-name='atomic_inc_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16_nv'>
<parameter type-id='93977ae7' name='target'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='atomic_inc_32_nv' mangled-name='atomic_inc_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32_nv'>
<parameter type-id='3a147f31' name='target'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='atomic_inc_ulong_nv' mangled-name='atomic_inc_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong_nv'>
<parameter type-id='64698d33' name='target'/>
<return type-id='ee1f298e'/>
</function-decl>
<function-decl name='atomic_dec_8_nv' mangled-name='atomic_dec_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8_nv'>
<parameter type-id='aa323ea4' name='target'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='atomic_dec_16_nv' mangled-name='atomic_dec_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16_nv'>
<parameter type-id='93977ae7' name='target'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='atomic_dec_32_nv' mangled-name='atomic_dec_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32_nv'>
<parameter type-id='3a147f31' name='target'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='atomic_dec_ulong_nv' mangled-name='atomic_dec_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong_nv'>
<parameter type-id='64698d33' name='target'/>
<return type-id='ee1f298e'/>
</function-decl>
<function-decl name='atomic_add_8_nv' mangled-name='atomic_add_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8_nv'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='ee31ee44' name='bits'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='atomic_add_16_nv' mangled-name='atomic_add_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16_nv'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='23bd8cb5' name='bits'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='atomic_add_32_nv' mangled-name='atomic_add_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32_nv'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='3ff5601b' name='bits'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='atomic_add_long_nv' mangled-name='atomic_add_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_long_nv'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='bd54fe1a' name='bits'/>
<return type-id='ee1f298e'/>
</function-decl>
<function-decl name='atomic_add_ptr_nv' mangled-name='atomic_add_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr_nv'>
<parameter type-id='fe09dd29' name='target'/>
<parameter type-id='79a0948f' name='bits'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='atomic_sub_8_nv' mangled-name='atomic_sub_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8_nv'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='ee31ee44' name='bits'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='atomic_sub_16_nv' mangled-name='atomic_sub_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16_nv'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='23bd8cb5' name='bits'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='atomic_sub_32_nv' mangled-name='atomic_sub_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32_nv'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='3ff5601b' name='bits'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='atomic_sub_long_nv' mangled-name='atomic_sub_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_long_nv'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='bd54fe1a' name='bits'/>
<return type-id='ee1f298e'/>
</function-decl>
<function-decl name='atomic_sub_ptr_nv' mangled-name='atomic_sub_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr_nv'>
<parameter type-id='fe09dd29' name='target'/>
<parameter type-id='79a0948f' name='bits'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='atomic_or_8_nv' mangled-name='atomic_or_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8_nv'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='b96825af' name='bits'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='atomic_or_16_nv' mangled-name='atomic_or_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16_nv'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='149c6638' name='bits'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='atomic_or_32_nv' mangled-name='atomic_or_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32_nv'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='8f92235e' name='bits'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='atomic_or_ulong_nv' mangled-name='atomic_or_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong_nv'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='ee1f298e' name='bits'/>
<return type-id='ee1f298e'/>
</function-decl>
<function-decl name='atomic_and_8_nv' mangled-name='atomic_and_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8_nv'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='b96825af' name='bits'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='atomic_and_16_nv' mangled-name='atomic_and_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16_nv'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='149c6638' name='bits'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='atomic_and_32_nv' mangled-name='atomic_and_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32_nv'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='8f92235e' name='bits'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='atomic_and_ulong_nv' mangled-name='atomic_and_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong_nv'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='ee1f298e' name='bits'/>
<return type-id='ee1f298e'/>
</function-decl>
<function-decl name='atomic_cas_8' mangled-name='atomic_cas_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_8'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='b96825af' name='exp'/>
<parameter type-id='b96825af' name='des'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='atomic_cas_16' mangled-name='atomic_cas_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_16'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='149c6638' name='exp'/>
<parameter type-id='149c6638' name='des'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='atomic_cas_32' mangled-name='atomic_cas_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_32'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='8f92235e' name='exp'/>
<parameter type-id='8f92235e' name='des'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='atomic_cas_ulong' mangled-name='atomic_cas_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ulong'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='ee1f298e' name='exp'/>
<parameter type-id='ee1f298e' name='des'/>
<return type-id='ee1f298e'/>
</function-decl>
<function-decl name='atomic_cas_ptr' mangled-name='atomic_cas_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ptr'>
<parameter type-id='fe09dd29' name='target'/>
<parameter type-id='eaa32e2f' name='exp'/>
<parameter type-id='eaa32e2f' name='des'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='atomic_swap_8' mangled-name='atomic_swap_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_8'>
<parameter type-id='aa323ea4' name='target'/>
<parameter type-id='b96825af' name='bits'/>
<return type-id='b96825af'/>
</function-decl>
<function-decl name='atomic_swap_16' mangled-name='atomic_swap_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_16'>
<parameter type-id='93977ae7' name='target'/>
<parameter type-id='149c6638' name='bits'/>
<return type-id='149c6638'/>
</function-decl>
<function-decl name='atomic_swap_32' mangled-name='atomic_swap_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_32'>
<parameter type-id='3a147f31' name='target'/>
<parameter type-id='8f92235e' name='bits'/>
<return type-id='8f92235e'/>
</function-decl>
<function-decl name='atomic_swap_ulong' mangled-name='atomic_swap_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ulong'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='ee1f298e' name='bits'/>
<return type-id='ee1f298e'/>
</function-decl>
<function-decl name='atomic_swap_ptr' mangled-name='atomic_swap_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ptr'>
<parameter type-id='fe09dd29' name='target'/>
<parameter type-id='eaa32e2f' name='bits'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='atomic_set_long_excl' mangled-name='atomic_set_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_set_long_excl'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='3502e3ff' name='value'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='atomic_clear_long_excl' mangled-name='atomic_clear_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_clear_long_excl'>
<parameter type-id='64698d33' name='target'/>
<parameter type-id='3502e3ff' name='value'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='membar_enter' mangled-name='membar_enter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_enter'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='membar_producer' mangled-name='membar_producer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_producer'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='membar_consumer' mangled-name='membar_consumer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_consumer'>
<return type-id='48b5725f'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='getexecname.c' language='LANG_C99'>
<function-decl name='getexecname' mangled-name='getexecname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getexecname'>
<return type-id='80f4b756'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='list.c' language='LANG_C99'>
<typedef-decl name='list_node_t' type-id='b0b5e45e' id='b21843b2'/>
<typedef-decl name='list_t' type-id='e824dae9' id='0899125f'/>
<class-decl name='list_node' size-in-bits='128' is-struct='yes' visibility='default' id='b0b5e45e'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='next' type-id='b03eadb4' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='prev' type-id='b03eadb4' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='list' size-in-bits='256' is-struct='yes' visibility='default' id='e824dae9'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='list_size' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='list_offset' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='list_head' type-id='b0b5e45e' visibility='default'/>
</data-member>
</class-decl>
<pointer-type-def type-id='b0b5e45e' size-in-bits='64' id='b03eadb4'/>
<pointer-type-def type-id='b21843b2' size-in-bits='64' id='ccc38265'/>
<pointer-type-def type-id='0899125f' size-in-bits='64' id='352ec160'/>
<function-decl name='list_create' mangled-name='list_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_create'>
<parameter type-id='352ec160' name='list'/>
<parameter type-id='b59d7dce' name='size'/>
<parameter type-id='b59d7dce' name='offset'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_destroy' mangled-name='list_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_destroy'>
<parameter type-id='352ec160' name='list'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_insert_after' mangled-name='list_insert_after' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_after'>
<parameter type-id='352ec160' name='list'/>
<parameter type-id='eaa32e2f' name='object'/>
<parameter type-id='eaa32e2f' name='nobject'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_insert_before' mangled-name='list_insert_before' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_before'>
<parameter type-id='352ec160' name='list'/>
<parameter type-id='eaa32e2f' name='object'/>
<parameter type-id='eaa32e2f' name='nobject'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_insert_head' mangled-name='list_insert_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_head'>
<parameter type-id='352ec160' name='list'/>
<parameter type-id='eaa32e2f' name='object'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_insert_tail' mangled-name='list_insert_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_tail'>
<parameter type-id='352ec160' name='list'/>
<parameter type-id='eaa32e2f' name='object'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_remove' mangled-name='list_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove'>
<parameter type-id='352ec160' name='list'/>
<parameter type-id='eaa32e2f' name='object'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_remove_head' mangled-name='list_remove_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_head'>
<parameter type-id='352ec160' name='list'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='list_remove_tail' mangled-name='list_remove_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_tail'>
<parameter type-id='352ec160' name='list'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='list_head' mangled-name='list_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_head'>
<parameter type-id='352ec160' name='list'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='list_tail' mangled-name='list_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_tail'>
<parameter type-id='352ec160' name='list'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='list_next' mangled-name='list_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_next'>
<parameter type-id='352ec160' name='list'/>
<parameter type-id='eaa32e2f' name='object'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='list_prev' mangled-name='list_prev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_prev'>
<parameter type-id='352ec160' name='list'/>
<parameter type-id='eaa32e2f' name='object'/>
<return type-id='eaa32e2f'/>
</function-decl>
<function-decl name='list_move_tail' mangled-name='list_move_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_move_tail'>
<parameter type-id='352ec160' name='dst'/>
<parameter type-id='352ec160' name='src'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_link_replace' mangled-name='list_link_replace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_replace'>
<parameter type-id='ccc38265' name='lold'/>
<parameter type-id='ccc38265' name='lnew'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_link_init' mangled-name='list_link_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_init'>
<parameter type-id='ccc38265' name='ln'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='list_link_active' mangled-name='list_link_active' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_active'>
<parameter type-id='ccc38265' name='ln'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='list_is_empty' mangled-name='list_is_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_is_empty'>
<parameter type-id='352ec160' name='list'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='mkdirp.c' language='LANG_C99'>
<typedef-decl name='mode_t' type-id='e1c52942' id='d50d396c'/>
<function-decl name='mkdirp' mangled-name='mkdirp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='mkdirp'>
<parameter type-id='80f4b756' name='d'/>
<parameter type-id='d50d396c' name='mode'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='os/linux/gethostid.c' language='LANG_C99'>
<function-decl name='get_system_hostid' mangled-name='get_system_hostid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='get_system_hostid'>
<return type-id='7359adad'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='os/linux/getmntany.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='03085adc' size-in-bits='192' id='083f8d58'>
<subrange length='3' type-id='7359adad' id='56f209d2'/>
</array-type-def>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8' id='89feb1ec'>
<subrange length='1' type-id='7359adad' id='52f813b4'/>
</array-type-def>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='160' id='664ac0b7'>
<subrange length='20' type-id='7359adad' id='fdca39cf'/>
</array-type-def>
+ <class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
+ <class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
+ <class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
<class-decl name='extmnttab' size-in-bits='320' is-struct='yes' visibility='default' id='0c544dc0'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='mnt_major' type-id='3502e3ff' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='288'>
<var-decl name='mnt_minor' type-id='3502e3ff' visibility='default'/>
</data-member>
</class-decl>
- <typedef-decl name='_IO_lock_t' type-id='48b5725f' id='bb4788fa'/>
- <class-decl name='_IO_marker' size-in-bits='192' is-struct='yes' visibility='default' id='010ae0b9'>
+ <class-decl name='stat64' size-in-bits='1152' is-struct='yes' visibility='default' id='0bbec9cd'>
<data-member access='public' layout-offset-in-bits='0'>
- <var-decl name='_next' type-id='e4c6fa61' visibility='default'/>
+ <var-decl name='st_dev' type-id='35ed8932' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
- <var-decl name='_sbuf' type-id='dca988a5' visibility='default'/>
+ <var-decl name='st_ino' type-id='71288a47' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
- <var-decl name='_pos' type-id='95e97e5e' visibility='default'/>
+ <var-decl name='st_nlink' type-id='80f0b9df' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='192'>
+ <var-decl name='st_mode' type-id='e1c52942' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='224'>
+ <var-decl name='st_uid' type-id='cc5fcceb' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='256'>
+ <var-decl name='st_gid' type-id='d94ec6d9' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='288'>
+ <var-decl name='__pad0' type-id='95e97e5e' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='320'>
+ <var-decl name='st_rdev' type-id='35ed8932' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='384'>
+ <var-decl name='st_size' type-id='79989e9c' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='448'>
+ <var-decl name='st_blksize' type-id='d3f10a7f' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='512'>
+ <var-decl name='st_blocks' type-id='4e711bf1' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='576'>
+ <var-decl name='st_atim' type-id='a9c79a1f' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='704'>
+ <var-decl name='st_mtim' type-id='a9c79a1f' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='832'>
+ <var-decl name='st_ctim' type-id='a9c79a1f' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='960'>
+ <var-decl name='__glibc_reserved' type-id='083f8d58' visibility='default'/>
</data-member>
</class-decl>
+ <typedef-decl name='__dev_t' type-id='7359adad' id='35ed8932'/>
+ <typedef-decl name='__gid_t' type-id='f0981eeb' id='d94ec6d9'/>
+ <typedef-decl name='__ino64_t' type-id='7359adad' id='71288a47'/>
+ <typedef-decl name='__mode_t' type-id='f0981eeb' id='e1c52942'/>
+ <typedef-decl name='__nlink_t' type-id='7359adad' id='80f0b9df'/>
+ <typedef-decl name='__off_t' type-id='bd54fe1a' id='79989e9c'/>
+ <typedef-decl name='__off64_t' type-id='bd54fe1a' id='724e4de6'/>
+ <typedef-decl name='__time_t' type-id='bd54fe1a' id='65eda9c0'/>
+ <typedef-decl name='__blksize_t' type-id='bd54fe1a' id='d3f10a7f'/>
+ <typedef-decl name='__blkcnt64_t' type-id='bd54fe1a' id='4e711bf1'/>
+ <typedef-decl name='__syscall_slong_t' type-id='bd54fe1a' id='03085adc'/>
+ <typedef-decl name='FILE' type-id='ec1ed955' id='aa12d1ba'/>
+ <typedef-decl name='_IO_lock_t' type-id='48b5725f' id='bb4788fa'/>
<class-decl name='_IO_FILE' size-in-bits='1728' is-struct='yes' visibility='default' id='ec1ed955'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='_flags' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='_IO_read_ptr' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='_IO_read_end' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='_IO_read_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='_IO_write_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='_IO_write_ptr' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='_IO_write_end' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='_IO_buf_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='512'>
<var-decl name='_IO_buf_end' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='576'>
<var-decl name='_IO_save_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='640'>
<var-decl name='_IO_backup_base' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='704'>
<var-decl name='_IO_save_end' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='768'>
<var-decl name='_markers' type-id='e4c6fa61' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='832'>
<var-decl name='_chain' type-id='dca988a5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='896'>
<var-decl name='_fileno' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='928'>
<var-decl name='_flags2' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='960'>
<var-decl name='_old_offset' type-id='79989e9c' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1024'>
<var-decl name='_cur_column' type-id='8efea9e5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1040'>
<var-decl name='_vtable_offset' type-id='28577a57' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1048'>
<var-decl name='_shortbuf' type-id='89feb1ec' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1088'>
<var-decl name='_lock' type-id='cecf4ea7' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1152'>
<var-decl name='_offset' type-id='724e4de6' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1216'>
- <var-decl name='__pad1' type-id='eaa32e2f' visibility='default'/>
+ <var-decl name='_codecvt' type-id='570f8c59' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1280'>
- <var-decl name='__pad2' type-id='eaa32e2f' visibility='default'/>
+ <var-decl name='_wide_data' type-id='c65a1f29' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1344'>
- <var-decl name='__pad3' type-id='eaa32e2f' visibility='default'/>
+ <var-decl name='_freeres_list' type-id='dca988a5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1408'>
- <var-decl name='__pad4' type-id='eaa32e2f' visibility='default'/>
+ <var-decl name='_freeres_buf' type-id='eaa32e2f' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1472'>
<var-decl name='__pad5' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1536'>
<var-decl name='_mode' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1568'>
<var-decl name='_unused2' type-id='664ac0b7' visibility='default'/>
</data-member>
</class-decl>
- <class-decl name='stat64' size-in-bits='1152' is-struct='yes' visibility='default' id='0bbec9cd'>
- <data-member access='public' layout-offset-in-bits='0'>
- <var-decl name='st_dev' type-id='35ed8932' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='64'>
- <var-decl name='st_ino' type-id='71288a47' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='128'>
- <var-decl name='st_nlink' type-id='80f0b9df' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='192'>
- <var-decl name='st_mode' type-id='e1c52942' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='224'>
- <var-decl name='st_uid' type-id='cc5fcceb' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='256'>
- <var-decl name='st_gid' type-id='d94ec6d9' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='288'>
- <var-decl name='__pad0' type-id='95e97e5e' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='320'>
- <var-decl name='st_rdev' type-id='35ed8932' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='384'>
- <var-decl name='st_size' type-id='79989e9c' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='448'>
- <var-decl name='st_blksize' type-id='d3f10a7f' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='512'>
- <var-decl name='st_blocks' type-id='4e711bf1' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='576'>
- <var-decl name='st_atim' type-id='a9c79a1f' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='704'>
- <var-decl name='st_mtim' type-id='a9c79a1f' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='832'>
- <var-decl name='st_ctim' type-id='a9c79a1f' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='960'>
- <var-decl name='__glibc_reserved' type-id='083f8d58' visibility='default'/>
- </data-member>
- </class-decl>
- <typedef-decl name='__dev_t' type-id='7359adad' id='35ed8932'/>
- <typedef-decl name='__gid_t' type-id='f0981eeb' id='d94ec6d9'/>
- <typedef-decl name='__ino64_t' type-id='7359adad' id='71288a47'/>
- <typedef-decl name='__mode_t' type-id='f0981eeb' id='e1c52942'/>
- <typedef-decl name='__nlink_t' type-id='7359adad' id='80f0b9df'/>
- <typedef-decl name='__off_t' type-id='bd54fe1a' id='79989e9c'/>
- <typedef-decl name='__off64_t' type-id='bd54fe1a' id='724e4de6'/>
- <typedef-decl name='__time_t' type-id='bd54fe1a' id='65eda9c0'/>
- <typedef-decl name='__blksize_t' type-id='bd54fe1a' id='d3f10a7f'/>
- <typedef-decl name='__blkcnt64_t' type-id='bd54fe1a' id='4e711bf1'/>
- <typedef-decl name='__syscall_slong_t' type-id='bd54fe1a' id='03085adc'/>
- <typedef-decl name='FILE' type-id='ec1ed955' id='aa12d1ba'/>
<class-decl name='timespec' size-in-bits='128' is-struct='yes' visibility='default' id='a9c79a1f'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='tv_sec' type-id='65eda9c0' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='tv_nsec' type-id='03085adc' visibility='default'/>
</data-member>
</class-decl>
<pointer-type-def type-id='aa12d1ba' size-in-bits='64' id='822cd80b'/>
<pointer-type-def type-id='ec1ed955' size-in-bits='64' id='dca988a5'/>
+ <pointer-type-def type-id='a4036571' size-in-bits='64' id='570f8c59'/>
<pointer-type-def type-id='bb4788fa' size-in-bits='64' id='cecf4ea7'/>
<pointer-type-def type-id='010ae0b9' size-in-bits='64' id='e4c6fa61'/>
+ <pointer-type-def type-id='79bd3751' size-in-bits='64' id='c65a1f29'/>
<pointer-type-def type-id='0c544dc0' size-in-bits='64' id='394fc496'/>
<pointer-type-def type-id='0bbec9cd' size-in-bits='64' id='62f7a03d'/>
+ <class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
+ <class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
+ <class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
<function-decl name='getmntany' mangled-name='getmntany' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getmntany'>
<parameter type-id='822cd80b' name='fp'/>
<parameter type-id='9d424d31' name='mgetp'/>
<parameter type-id='9d424d31' name='mrefp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='_sol_getmntent' mangled-name='_sol_getmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='_sol_getmntent'>
<parameter type-id='822cd80b' name='fp'/>
<parameter type-id='9d424d31' name='mgetp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='getextmntent' mangled-name='getextmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getextmntent'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='394fc496' name='entry'/>
<parameter type-id='62f7a03d' name='statbuf'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='os/linux/zone.c' language='LANG_C99'>
<typedef-decl name='zoneid_t' type-id='95e97e5e' id='4da03624'/>
<function-decl name='getzoneid' mangled-name='getzoneid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getzoneid'>
<return type-id='4da03624'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='page.c' language='LANG_C99'>
<function-decl name='spl_pagesize' mangled-name='spl_pagesize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='spl_pagesize'>
<return type-id='b59d7dce'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='strlcat.c' language='LANG_C99'>
<function-decl name='strlcat' mangled-name='strlcat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcat'>
<parameter type-id='26a90f95' name='dst'/>
<parameter type-id='80f4b756' name='src'/>
<parameter type-id='b59d7dce' name='dstsize'/>
<return type-id='b59d7dce'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='strlcpy.c' language='LANG_C99'>
<function-decl name='strlcpy' mangled-name='strlcpy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcpy'>
<parameter type-id='26a90f95' name='dst'/>
<parameter type-id='80f4b756' name='src'/>
<parameter type-id='b59d7dce' name='len'/>
<return type-id='b59d7dce'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='timestamp.c' language='LANG_C99'>
<function-decl name='print_timestamp' mangled-name='print_timestamp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='print_timestamp'>
<parameter type-id='3502e3ff' name='timestamp_fmt'/>
<return type-id='48b5725f'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='thread_pool.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='384' id='36d7f119'>
<subrange length='48' type-id='7359adad' id='8f6d2a81'/>
</array-type-def>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='448' id='6093ff7c'>
<subrange length='56' type-id='7359adad' id='f8137894'/>
</array-type-def>
<type-decl name='long long unsigned int' size-in-bits='64' id='3a47d82b'/>
<array-type-def dimensions='1' type-id='f0981eeb' size-in-bits='64' id='0d532ec1'>
<subrange length='2' type-id='7359adad' id='52efc4ef'/>
</array-type-def>
<typedef-decl name='tpool_t' type-id='88d1b7f9' id='b1bbf10d'/>
<typedef-decl name='tpool_job_t' type-id='3b8579e5' id='66a0afc9'/>
<class-decl name='tpool_job' size-in-bits='192' is-struct='yes' visibility='default' id='3b8579e5'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='tpj_next' type-id='f32b30e4' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='tpj_func' type-id='b7f9d8e6' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='tpj_arg' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='tpool_active_t' type-id='c8d086f4' id='6fcda10e'/>
<class-decl name='tpool_active' size-in-bits='128' is-struct='yes' visibility='default' id='c8d086f4'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='tpa_next' type-id='ad33e5e7' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='tpa_tid' type-id='4051f5e7' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='tpool' size-in-bits='2496' is-struct='yes' visibility='default' id='88d1b7f9'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='tp_forw' type-id='9cf59a50' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='tp_back' type-id='9cf59a50' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='tp_mutex' type-id='7a6844eb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='tp_busycv' type-id='62fab762' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='832'>
<var-decl name='tp_workcv' type-id='62fab762' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1216'>
<var-decl name='tp_waitcv' type-id='62fab762' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1600'>
<var-decl name='tp_active' type-id='ad33e5e7' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1664'>
<var-decl name='tp_head' type-id='f32b30e4' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1728'>
<var-decl name='tp_tail' type-id='f32b30e4' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1792'>
<var-decl name='tp_attr' type-id='7d8569fd' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2240'>
<var-decl name='tp_flags' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2272'>
<var-decl name='tp_linger' type-id='3502e3ff' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2304'>
<var-decl name='tp_njobs' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2336'>
<var-decl name='tp_minimum' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2368'>
<var-decl name='tp_maximum' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2400'>
<var-decl name='tp_current' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2432'>
<var-decl name='tp_idle' type-id='95e97e5e' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='pthread_t' type-id='7359adad' id='4051f5e7'/>
<union-decl name='pthread_attr_t' size-in-bits='448' visibility='default' id='b63afacd'>
<data-member access='public'>
<var-decl name='__size' type-id='6093ff7c' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
</data-member>
</union-decl>
<typedef-decl name='pthread_attr_t' type-id='b63afacd' id='7d8569fd'/>
<union-decl name='pthread_cond_t' size-in-bits='384' naming-typedef-id='62fab762' visibility='default' id='cbb12c12'>
<data-member access='public'>
<var-decl name='__data' type-id='c987b47c' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='__size' type-id='36d7f119' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='__align' type-id='1eb56b1e' visibility='default'/>
</data-member>
</union-decl>
<typedef-decl name='pthread_cond_t' type-id='cbb12c12' id='62fab762'/>
<class-decl name='__pthread_cond_s' size-in-bits='384' is-struct='yes' visibility='default' id='c987b47c'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='' type-id='ac5ab595' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='' type-id='ac5ab596' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='__g_refs' type-id='0d532ec1' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='__g_size' type-id='0d532ec1' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='__g1_orig_size' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='288'>
<var-decl name='__wrefs' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='__g_signals' type-id='0d532ec1' visibility='default'/>
</data-member>
</class-decl>
<union-decl name='__anonymous_union__1' size-in-bits='64' is-anonymous='yes' visibility='default' id='ac5ab595'>
<data-member access='public'>
<var-decl name='__wseq' type-id='3a47d82b' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='__wseq32' type-id='e7f43f72' visibility='default'/>
</data-member>
</union-decl>
<class-decl name='__anonymous_struct__' size-in-bits='64' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f72'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='__low' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='__high' type-id='f0981eeb' visibility='default'/>
</data-member>
</class-decl>
<union-decl name='__anonymous_union__2' size-in-bits='64' is-anonymous='yes' visibility='default' id='ac5ab596'>
<data-member access='public'>
<var-decl name='__g1_start' type-id='3a47d82b' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='__g1_start32' type-id='e7f43f72' visibility='default'/>
</data-member>
</union-decl>
<pointer-type-def type-id='7d8569fd' size-in-bits='64' id='7347a39e'/>
<pointer-type-def type-id='6fcda10e' size-in-bits='64' id='ad33e5e7'/>
<pointer-type-def type-id='66a0afc9' size-in-bits='64' id='f32b30e4'/>
<pointer-type-def type-id='b1bbf10d' size-in-bits='64' id='9cf59a50'/>
<pointer-type-def type-id='c5c76c9c' size-in-bits='64' id='b7f9d8e6'/>
<function-decl name='tpool_create' mangled-name='tpool_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_create'>
<parameter type-id='3502e3ff' name='min_threads'/>
<parameter type-id='3502e3ff' name='max_threads'/>
<parameter type-id='3502e3ff' name='linger'/>
<parameter type-id='7347a39e' name='attr'/>
<return type-id='9cf59a50'/>
</function-decl>
<function-decl name='tpool_dispatch' mangled-name='tpool_dispatch' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_dispatch'>
<parameter type-id='9cf59a50' name='tpool'/>
<parameter type-id='b7f9d8e6' name='func'/>
<parameter type-id='eaa32e2f' name='arg'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='tpool_destroy' mangled-name='tpool_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_destroy'>
<parameter type-id='9cf59a50' name='tpool'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='tpool_abandon' mangled-name='tpool_abandon' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_abandon'>
<parameter type-id='9cf59a50' name='tpool'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='tpool_wait' mangled-name='tpool_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_wait'>
<parameter type-id='9cf59a50' name='tpool'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='tpool_suspend' mangled-name='tpool_suspend' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_suspend'>
<parameter type-id='9cf59a50' name='tpool'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='tpool_suspended' mangled-name='tpool_suspended' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_suspended'>
<parameter type-id='9cf59a50' name='tpool'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='tpool_resume' mangled-name='tpool_resume' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_resume'>
<parameter type-id='9cf59a50' name='tpool'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='tpool_member' mangled-name='tpool_member' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_member'>
<parameter type-id='9cf59a50' name='tpool'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-type size-in-bits='64' id='c5c76c9c'>
<parameter type-id='eaa32e2f'/>
<return type-id='48b5725f'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/cityhash.c' language='LANG_C99'>
<function-decl name='cityhash4' mangled-name='cityhash4' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='cityhash4'>
<parameter type-id='9c313c2d' name='w1'/>
<parameter type-id='9c313c2d' name='w2'/>
<parameter type-id='9c313c2d' name='w3'/>
<parameter type-id='9c313c2d' name='w4'/>
<return type-id='9c313c2d'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfeature_common.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='83f29ca2' size-in-bits='15232' id='d96379d0'>
<subrange length='34' type-id='7359adad' id='6a6a7e00'/>
</array-type-def>
<enum-decl name='spa_feature' id='33ecb627'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='SPA_FEATURE_NONE' value='-1'/>
<enumerator name='SPA_FEATURE_ASYNC_DESTROY' value='0'/>
<enumerator name='SPA_FEATURE_EMPTY_BPOBJ' value='1'/>
<enumerator name='SPA_FEATURE_LZ4_COMPRESS' value='2'/>
<enumerator name='SPA_FEATURE_MULTI_VDEV_CRASH_DUMP' value='3'/>
<enumerator name='SPA_FEATURE_SPACEMAP_HISTOGRAM' value='4'/>
<enumerator name='SPA_FEATURE_ENABLED_TXG' value='5'/>
<enumerator name='SPA_FEATURE_HOLE_BIRTH' value='6'/>
<enumerator name='SPA_FEATURE_EXTENSIBLE_DATASET' value='7'/>
<enumerator name='SPA_FEATURE_EMBEDDED_DATA' value='8'/>
<enumerator name='SPA_FEATURE_BOOKMARKS' value='9'/>
<enumerator name='SPA_FEATURE_FS_SS_LIMIT' value='10'/>
<enumerator name='SPA_FEATURE_LARGE_BLOCKS' value='11'/>
<enumerator name='SPA_FEATURE_LARGE_DNODE' value='12'/>
<enumerator name='SPA_FEATURE_SHA512' value='13'/>
<enumerator name='SPA_FEATURE_SKEIN' value='14'/>
<enumerator name='SPA_FEATURE_EDONR' value='15'/>
<enumerator name='SPA_FEATURE_USEROBJ_ACCOUNTING' value='16'/>
<enumerator name='SPA_FEATURE_ENCRYPTION' value='17'/>
<enumerator name='SPA_FEATURE_PROJECT_QUOTA' value='18'/>
<enumerator name='SPA_FEATURE_DEVICE_REMOVAL' value='19'/>
<enumerator name='SPA_FEATURE_OBSOLETE_COUNTS' value='20'/>
<enumerator name='SPA_FEATURE_POOL_CHECKPOINT' value='21'/>
<enumerator name='SPA_FEATURE_SPACEMAP_V2' value='22'/>
<enumerator name='SPA_FEATURE_ALLOCATION_CLASSES' value='23'/>
<enumerator name='SPA_FEATURE_RESILVER_DEFER' value='24'/>
<enumerator name='SPA_FEATURE_BOOKMARK_V2' value='25'/>
<enumerator name='SPA_FEATURE_REDACTION_BOOKMARKS' value='26'/>
<enumerator name='SPA_FEATURE_REDACTED_DATASETS' value='27'/>
<enumerator name='SPA_FEATURE_BOOKMARK_WRITTEN' value='28'/>
<enumerator name='SPA_FEATURE_LOG_SPACEMAP' value='29'/>
<enumerator name='SPA_FEATURE_LIVELIST' value='30'/>
<enumerator name='SPA_FEATURE_DEVICE_REBUILD' value='31'/>
<enumerator name='SPA_FEATURE_ZSTD_COMPRESS' value='32'/>
<enumerator name='SPA_FEATURE_DRAID' value='33'/>
<enumerator name='SPA_FEATURES' value='34'/>
</enum-decl>
<typedef-decl name='spa_feature_t' type-id='33ecb627' id='d6618c78'/>
<enum-decl name='zfeature_flags' id='6db816a4'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZFEATURE_FLAG_READONLY_COMPAT' value='1'/>
<enumerator name='ZFEATURE_FLAG_MOS' value='2'/>
<enumerator name='ZFEATURE_FLAG_ACTIVATE_ON_ENABLE' value='4'/>
<enumerator name='ZFEATURE_FLAG_PER_DATASET' value='8'/>
</enum-decl>
<typedef-decl name='zfeature_flags_t' type-id='6db816a4' id='fc329033'/>
<enum-decl name='zfeature_type' id='c4fa2355'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZFEATURE_TYPE_BOOLEAN' value='0'/>
<enumerator name='ZFEATURE_TYPE_UINT64_ARRAY' value='1'/>
<enumerator name='ZFEATURE_NUM_TYPES' value='2'/>
</enum-decl>
<typedef-decl name='zfeature_type_t' type-id='c4fa2355' id='732d2bb2'/>
<class-decl name='zfeature_info' size-in-bits='448' is-struct='yes' visibility='default' id='1178d146'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='fi_feature' type-id='d6618c78' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='fi_uname' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='fi_guid' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='fi_desc' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='fi_flags' type-id='fc329033' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='288'>
<var-decl name='fi_zfs_mod_supported' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='fi_type' type-id='732d2bb2' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='fi_depends' type-id='1acff326' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zfeature_info_t' type-id='1178d146' id='83f29ca2'/>
<class-decl name='zfs_mod_supported_features' size-in-bits='128' is-struct='yes' visibility='default' id='3eee3342'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='tree' type-id='eaa32e2f' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='all_features' type-id='c19b74c3' visibility='default'/>
</data-member>
</class-decl>
<qualified-type-def type-id='d6618c78' const='yes' id='81a65028'/>
<pointer-type-def type-id='81a65028' size-in-bits='64' id='1acff326'/>
<qualified-type-def type-id='3eee3342' const='yes' id='0c1d5bbb'/>
<pointer-type-def type-id='0c1d5bbb' size-in-bits='64' id='a3372543'/>
<pointer-type-def type-id='d6618c78' size-in-bits='64' id='a8425263'/>
<var-decl name='spa_feature_table' type-id='d96379d0' mangled-name='spa_feature_table' visibility='default' elf-symbol-id='spa_feature_table'/>
<var-decl name='zfeature_checks_disable' type-id='c19b74c3' mangled-name='zfeature_checks_disable' visibility='default' elf-symbol-id='zfeature_checks_disable'/>
<function-decl name='zfeature_is_valid_guid' mangled-name='zfeature_is_valid_guid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_is_valid_guid'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfeature_is_supported' mangled-name='zfeature_is_supported' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_is_supported'>
<parameter type-id='80f4b756' name='guid'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfeature_lookup_guid' mangled-name='zfeature_lookup_guid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_lookup_guid'>
<parameter type-id='80f4b756' name='guid'/>
<parameter type-id='a8425263' name='res'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfeature_lookup_name' mangled-name='zfeature_lookup_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_lookup_name'>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='a8425263' name='res'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfeature_depends_on' mangled-name='zfeature_depends_on' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_depends_on'>
<parameter type-id='d6618c78' name='fid'/>
<parameter type-id='d6618c78' name='check'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_mod_supported' mangled-name='zfs_mod_supported' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mod_supported'>
<parameter type-id='80f4b756' name='scope'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='a3372543' name='sfeatures'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zpool_feature_init' mangled-name='zpool_feature_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_feature_init'>
<return type-id='48b5725f'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_comutil.c' language='LANG_C99'>
- <array-type-def dimensions='1' type-id='80f4b756' size-in-bits='2624' id='ef31fedf'>
+ <array-type-def dimensions='1' type-id='b99c00c9' size-in-bits='2624' id='5ce15418'>
<subrange length='41' type-id='7359adad' id='cb834f44'/>
</array-type-def>
<class-decl name='zpool_load_policy' size-in-bits='256' is-struct='yes' visibility='default' id='2f65b36f'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='zlp_rewind' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='zlp_maxmeta' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='zlp_maxdata' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='zlp_txg' type-id='9c313c2d' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zpool_load_policy_t' type-id='2f65b36f' id='d11b7617'/>
+ <qualified-type-def type-id='80f4b756' const='yes' id='b99c00c9'/>
<pointer-type-def type-id='d11b7617' size-in-bits='64' id='23432aaa'/>
- <var-decl name='zfs_history_event_names' type-id='ef31fedf' mangled-name='zfs_history_event_names' visibility='default' elf-symbol-id='zfs_history_event_names'/>
+ <var-decl name='zfs_history_event_names' type-id='5ce15418' mangled-name='zfs_history_event_names' visibility='default' elf-symbol-id='zfs_history_event_names'/>
<function-decl name='zfs_allocatable_devs' mangled-name='zfs_allocatable_devs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_allocatable_devs'>
<parameter type-id='5ce45b60' name='nv'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_special_devs' mangled-name='zfs_special_devs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_special_devs'>
<parameter type-id='5ce45b60' name='nv'/>
<parameter type-id='26a90f95' name='type'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zpool_get_load_policy' mangled-name='zpool_get_load_policy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_load_policy'>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='23432aaa' name='zlpp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_zpl_version_map' mangled-name='zfs_zpl_version_map' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_zpl_version_map'>
<parameter type-id='95e97e5e' name='spa_version'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_spa_version_map' mangled-name='zfs_spa_version_map' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_spa_version_map'>
<parameter type-id='95e97e5e' name='zpl_version'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_dataset_name_hidden' mangled-name='zfs_dataset_name_hidden' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dataset_name_hidden'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_deleg.c' language='LANG_C99'>
- <array-type-def dimensions='1' type-id='f3f851ad' size-in-bits='infinite' id='bc4e5d90'>
+ <array-type-def dimensions='1' type-id='fa1870fd' size-in-bits='infinite' id='7c00e69d'>
<subrange length='infinite' id='031f2035'/>
</array-type-def>
<enum-decl name='zfs_deleg_who_type_t' naming-typedef-id='36d4bd5a' id='b5fa5816'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZFS_DELEG_WHO_UNKNOWN' value='0'/>
<enumerator name='ZFS_DELEG_USER' value='117'/>
<enumerator name='ZFS_DELEG_USER_SETS' value='85'/>
<enumerator name='ZFS_DELEG_GROUP' value='103'/>
<enumerator name='ZFS_DELEG_GROUP_SETS' value='71'/>
<enumerator name='ZFS_DELEG_EVERYONE' value='101'/>
<enumerator name='ZFS_DELEG_EVERYONE_SETS' value='69'/>
<enumerator name='ZFS_DELEG_CREATE' value='99'/>
<enumerator name='ZFS_DELEG_CREATE_SETS' value='67'/>
<enumerator name='ZFS_DELEG_NAMED_SET' value='115'/>
<enumerator name='ZFS_DELEG_NAMED_SET_SETS' value='83'/>
</enum-decl>
<typedef-decl name='zfs_deleg_who_type_t' type-id='b5fa5816' id='36d4bd5a'/>
<enum-decl name='zfs_deleg_note_t' naming-typedef-id='4613c173' id='729d4547'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZFS_DELEG_NOTE_CREATE' value='0'/>
<enumerator name='ZFS_DELEG_NOTE_DESTROY' value='1'/>
<enumerator name='ZFS_DELEG_NOTE_SNAPSHOT' value='2'/>
<enumerator name='ZFS_DELEG_NOTE_ROLLBACK' value='3'/>
<enumerator name='ZFS_DELEG_NOTE_CLONE' value='4'/>
<enumerator name='ZFS_DELEG_NOTE_PROMOTE' value='5'/>
<enumerator name='ZFS_DELEG_NOTE_RENAME' value='6'/>
<enumerator name='ZFS_DELEG_NOTE_SEND' value='7'/>
<enumerator name='ZFS_DELEG_NOTE_RECEIVE' value='8'/>
<enumerator name='ZFS_DELEG_NOTE_ALLOW' value='9'/>
<enumerator name='ZFS_DELEG_NOTE_USERPROP' value='10'/>
<enumerator name='ZFS_DELEG_NOTE_MOUNT' value='11'/>
<enumerator name='ZFS_DELEG_NOTE_SHARE' value='12'/>
<enumerator name='ZFS_DELEG_NOTE_USERQUOTA' value='13'/>
<enumerator name='ZFS_DELEG_NOTE_GROUPQUOTA' value='14'/>
<enumerator name='ZFS_DELEG_NOTE_USERUSED' value='15'/>
<enumerator name='ZFS_DELEG_NOTE_GROUPUSED' value='16'/>
<enumerator name='ZFS_DELEG_NOTE_USEROBJQUOTA' value='17'/>
<enumerator name='ZFS_DELEG_NOTE_GROUPOBJQUOTA' value='18'/>
<enumerator name='ZFS_DELEG_NOTE_USEROBJUSED' value='19'/>
<enumerator name='ZFS_DELEG_NOTE_GROUPOBJUSED' value='20'/>
<enumerator name='ZFS_DELEG_NOTE_HOLD' value='21'/>
<enumerator name='ZFS_DELEG_NOTE_RELEASE' value='22'/>
<enumerator name='ZFS_DELEG_NOTE_DIFF' value='23'/>
<enumerator name='ZFS_DELEG_NOTE_BOOKMARK' value='24'/>
<enumerator name='ZFS_DELEG_NOTE_LOAD_KEY' value='25'/>
<enumerator name='ZFS_DELEG_NOTE_CHANGE_KEY' value='26'/>
<enumerator name='ZFS_DELEG_NOTE_PROJECTUSED' value='27'/>
<enumerator name='ZFS_DELEG_NOTE_PROJECTQUOTA' value='28'/>
<enumerator name='ZFS_DELEG_NOTE_PROJECTOBJUSED' value='29'/>
<enumerator name='ZFS_DELEG_NOTE_PROJECTOBJQUOTA' value='30'/>
<enumerator name='ZFS_DELEG_NOTE_NONE' value='31'/>
</enum-decl>
<typedef-decl name='zfs_deleg_note_t' type-id='729d4547' id='4613c173'/>
<class-decl name='zfs_deleg_perm_tab' size-in-bits='128' is-struct='yes' visibility='default' id='5aa05c1f'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='z_perm' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='z_note' type-id='4613c173' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zfs_deleg_perm_tab_t' type-id='5aa05c1f' id='f3f851ad'/>
- <var-decl name='zfs_deleg_perm_tab' type-id='bc4e5d90' mangled-name='zfs_deleg_perm_tab' visibility='default' elf-symbol-id='zfs_deleg_perm_tab'/>
+ <qualified-type-def type-id='f3f851ad' const='yes' id='fa1870fd'/>
+ <var-decl name='zfs_deleg_perm_tab' type-id='7c00e69d' mangled-name='zfs_deleg_perm_tab' visibility='default' elf-symbol-id='zfs_deleg_perm_tab'/>
<function-decl name='zfs_deleg_canonicalize_perm' mangled-name='zfs_deleg_canonicalize_perm' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_deleg_canonicalize_perm'>
<parameter type-id='80f4b756' name='perm'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_deleg_verify_nvlist' mangled-name='zfs_deleg_verify_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_deleg_verify_nvlist'>
<parameter type-id='5ce45b60' name='nvp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_deleg_whokey' mangled-name='zfs_deleg_whokey' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_deleg_whokey'>
<parameter type-id='26a90f95' name='attr'/>
<parameter type-id='36d4bd5a' name='type'/>
<parameter type-id='a84c031d' name='inheritchr'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='48b5725f'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_fletcher.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='256' id='85c64d26'>
<subrange length='4' type-id='7359adad' id='16fe7105'/>
</array-type-def>
<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='512' id='c5d13f42'>
<subrange length='8' type-id='7359adad' id='56e0c0b1'/>
</array-type-def>
<array-type-def dimensions='1' type-id='90dbb6d6' size-in-bits='2048' id='16582e69'>
<subrange length='4' type-id='7359adad' id='16fe7105'/>
</array-type-def>
<array-type-def dimensions='1' type-id='8240361c' size-in-bits='1024' id='481f90b1'>
<subrange length='4' type-id='7359adad' id='16fe7105'/>
</array-type-def>
<array-type-def dimensions='1' type-id='7c1ab40c' size-in-bits='512' id='cbd91ec1'>
<subrange length='4' type-id='7359adad' id='16fe7105'/>
</array-type-def>
<array-type-def dimensions='1' type-id='6d059eaa' size-in-bits='1024' id='729b6ebb'>
<subrange length='4' type-id='7359adad' id='16fe7105'/>
</array-type-def>
<class-decl name='zio_cksum' size-in-bits='256' is-struct='yes' visibility='default' id='1d53e28b'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='zc_word' type-id='85c64d26' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zio_cksum_t' type-id='1d53e28b' id='39730d0b'/>
<enum-decl name='zio_byteorder_t' naming-typedef-id='595a65ec' id='fc861be0'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZIO_CHECKSUM_NATIVE' value='0'/>
<enumerator name='ZIO_CHECKSUM_BYTESWAP' value='1'/>
</enum-decl>
<typedef-decl name='zio_byteorder_t' type-id='fc861be0' id='595a65ec'/>
<class-decl name='zio_abd_checksum_data' size-in-bits='256' is-struct='yes' visibility='default' id='4bf4b004'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='acd_byteorder' type-id='595a65ec' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='acd_ctx' type-id='0f7df99e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='acd_zcp' type-id='c24fc2ee' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='acd_private' type-id='eaa32e2f' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zio_abd_checksum_data_t' type-id='4bf4b004' id='74e39470'/>
<typedef-decl name='zio_abd_checksum_init_t' type-id='a5444274' id='029a8ebe'/>
<typedef-decl name='zio_abd_checksum_fini_t' type-id='a5444274' id='d6fd5c6c'/>
<typedef-decl name='zio_abd_checksum_iter_t' type-id='f4a1892e' id='cefa0f4a'/>
<class-decl name='zio_abd_checksum_func' size-in-bits='192' is-struct='yes' visibility='default' id='aa14691a'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='acf_init' type-id='0bcca125' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='acf_fini' type-id='bfe36153' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='acf_iter' type-id='1e276399' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zio_abd_checksum_func_t' type-id='3f8e8d11' id='c2eb138a'/>
<class-decl name='zfs_fletcher_superscalar' size-in-bits='256' is-struct='yes' visibility='default' id='28efb250'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='v' type-id='85c64d26' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zfs_fletcher_superscalar_t' type-id='28efb250' id='6d059eaa'/>
<class-decl name='zfs_fletcher_sse' size-in-bits='128' is-struct='yes' visibility='default' id='acd4019a'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='v' type-id='c1c22e6c' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zfs_fletcher_sse_t' type-id='acd4019a' id='7c1ab40c'/>
<class-decl name='zfs_fletcher_avx' size-in-bits='256' is-struct='yes' visibility='default' id='8c208dfa'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='v' type-id='85c64d26' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zfs_fletcher_avx_t' type-id='8c208dfa' id='8240361c'/>
<class-decl name='zfs_fletcher_avx512' size-in-bits='512' is-struct='yes' visibility='default' id='c6d0c382'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='v' type-id='c5d13f42' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zfs_fletcher_avx512_t' type-id='c6d0c382' id='90dbb6d6'/>
<union-decl name='fletcher_4_ctx' size-in-bits='2048' visibility='default' id='1f951ade'>
<data-member access='public'>
<var-decl name='scalar' type-id='39730d0b' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='superscalar' type-id='729b6ebb' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='sse' type-id='cbd91ec1' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='avx' type-id='481f90b1' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='avx512' type-id='16582e69' visibility='default'/>
</data-member>
</union-decl>
<typedef-decl name='fletcher_4_ctx_t' type-id='1f951ade' id='4b675395'/>
<qualified-type-def type-id='aa14691a' const='yes' id='3f8e8d11'/>
<pointer-type-def type-id='4b675395' size-in-bits='64' id='0f7df99e'/>
<pointer-type-def type-id='74e39470' size-in-bits='64' id='eefe7427'/>
<pointer-type-def type-id='d6fd5c6c' size-in-bits='64' id='bfe36153'/>
<pointer-type-def type-id='029a8ebe' size-in-bits='64' id='0bcca125'/>
<pointer-type-def type-id='cefa0f4a' size-in-bits='64' id='1e276399'/>
<pointer-type-def type-id='39730d0b' size-in-bits='64' id='c24fc2ee'/>
<var-decl name='fletcher_4_abd_ops' type-id='c2eb138a' mangled-name='fletcher_4_abd_ops' visibility='default' elf-symbol-id='fletcher_4_abd_ops'/>
<function-decl name='fletcher_init' mangled-name='fletcher_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_init'>
<parameter type-id='c24fc2ee' name='zcp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fletcher_2_incremental_native' mangled-name='fletcher_2_incremental_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_incremental_native'>
<parameter type-id='eaa32e2f' name='buf'/>
<parameter type-id='b59d7dce' name='size'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='fletcher_2_native' mangled-name='fletcher_2_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_native'>
<parameter type-id='eaa32e2f' name='buf'/>
<parameter type-id='9c313c2d' name='size'/>
<parameter type-id='eaa32e2f' name='ctx_template'/>
<parameter type-id='c24fc2ee' name='zcp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fletcher_2_incremental_byteswap' mangled-name='fletcher_2_incremental_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_incremental_byteswap'>
<parameter type-id='eaa32e2f' name='buf'/>
<parameter type-id='b59d7dce' name='size'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='fletcher_2_byteswap' mangled-name='fletcher_2_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_byteswap'>
<parameter type-id='eaa32e2f' name='buf'/>
<parameter type-id='9c313c2d' name='size'/>
<parameter type-id='eaa32e2f' name='ctx_template'/>
<parameter type-id='c24fc2ee' name='zcp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fletcher_4_impl_set' mangled-name='fletcher_4_impl_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_impl_set'>
<parameter type-id='80f4b756' name='val'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='fletcher_4_native' mangled-name='fletcher_4_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_native'>
<parameter type-id='eaa32e2f' name='buf'/>
<parameter type-id='9c313c2d' name='size'/>
<parameter type-id='eaa32e2f' name='ctx_template'/>
<parameter type-id='c24fc2ee' name='zcp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fletcher_4_native_varsize' mangled-name='fletcher_4_native_varsize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_native_varsize'>
<parameter type-id='eaa32e2f' name='buf'/>
<parameter type-id='9c313c2d' name='size'/>
<parameter type-id='c24fc2ee' name='zcp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fletcher_4_byteswap' mangled-name='fletcher_4_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_byteswap'>
<parameter type-id='eaa32e2f' name='buf'/>
<parameter type-id='9c313c2d' name='size'/>
<parameter type-id='eaa32e2f' name='ctx_template'/>
<parameter type-id='c24fc2ee' name='zcp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fletcher_4_incremental_native' mangled-name='fletcher_4_incremental_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_incremental_native'>
<parameter type-id='eaa32e2f' name='buf'/>
<parameter type-id='b59d7dce' name='size'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='fletcher_4_incremental_byteswap' mangled-name='fletcher_4_incremental_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_incremental_byteswap'>
<parameter type-id='eaa32e2f' name='buf'/>
<parameter type-id='b59d7dce' name='size'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='fletcher_4_init' mangled-name='fletcher_4_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_init'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='fletcher_4_fini' mangled-name='fletcher_4_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_fini'>
<return type-id='48b5725f'/>
</function-decl>
<function-type size-in-bits='64' id='f4a1892e'>
<parameter type-id='eaa32e2f'/>
<parameter type-id='b59d7dce'/>
<parameter type-id='eaa32e2f'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='a5444274'>
<parameter type-id='eefe7427'/>
<return type-id='48b5725f'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_fletcher_avx512.c' language='LANG_C99'>
<typedef-decl name='fletcher_4_init_f' type-id='173aa527' id='b9ae1656'/>
<typedef-decl name='fletcher_4_fini_f' type-id='0ad5b8a8' id='c4c1f4fc'/>
<typedef-decl name='fletcher_4_compute_f' type-id='38147eff' id='ad1dc4cb'/>
<class-decl name='fletcher_4_func' size-in-bits='512' is-struct='yes' visibility='default' id='57f479a0'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='init_native' type-id='b9ae1656' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='fini_native' type-id='c4c1f4fc' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='compute_native' type-id='ad1dc4cb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='init_byteswap' type-id='b9ae1656' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='fini_byteswap' type-id='c4c1f4fc' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='compute_byteswap' type-id='ad1dc4cb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='valid' type-id='297d38bc' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='name' type-id='80f4b756' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='fletcher_4_ops_t' type-id='57f479a0' id='eba91718'/>
<qualified-type-def type-id='eba91718' const='yes' id='9eeabdc8'/>
<pointer-type-def type-id='e9e61702' size-in-bits='64' id='297d38bc'/>
<pointer-type-def type-id='fe40251b' size-in-bits='64' id='173aa527'/>
<pointer-type-def type-id='17fb1f83' size-in-bits='64' id='38147eff'/>
<pointer-type-def type-id='fb39e25e' size-in-bits='64' id='0ad5b8a8'/>
<var-decl name='fletcher_4_avx512f_ops' type-id='9eeabdc8' mangled-name='fletcher_4_avx512f_ops' visibility='default' elf-symbol-id='fletcher_4_avx512f_ops'/>
<var-decl name='fletcher_4_avx512bw_ops' type-id='9eeabdc8' mangled-name='fletcher_4_avx512bw_ops' visibility='default' elf-symbol-id='fletcher_4_avx512bw_ops'/>
<function-type size-in-bits='64' id='e9e61702'>
<return type-id='c19b74c3'/>
</function-type>
<function-type size-in-bits='64' id='fe40251b'>
<parameter type-id='0f7df99e'/>
<return type-id='48b5725f'/>
</function-type>
<function-type size-in-bits='64' id='17fb1f83'>
<parameter type-id='0f7df99e'/>
<parameter type-id='eaa32e2f'/>
<parameter type-id='9c313c2d'/>
<return type-id='48b5725f'/>
</function-type>
<function-type size-in-bits='64' id='fb39e25e'>
<parameter type-id='0f7df99e'/>
<parameter type-id='c24fc2ee'/>
<return type-id='48b5725f'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_fletcher_intel.c' language='LANG_C99'>
<var-decl name='fletcher_4_avx2_ops' type-id='9eeabdc8' mangled-name='fletcher_4_avx2_ops' visibility='default' elf-symbol-id='fletcher_4_avx2_ops'/>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_fletcher_sse.c' language='LANG_C99'>
<var-decl name='fletcher_4_sse2_ops' type-id='9eeabdc8' mangled-name='fletcher_4_sse2_ops' visibility='default' elf-symbol-id='fletcher_4_sse2_ops'/>
<var-decl name='fletcher_4_ssse3_ops' type-id='9eeabdc8' mangled-name='fletcher_4_ssse3_ops' visibility='default' elf-symbol-id='fletcher_4_ssse3_ops'/>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_fletcher_superscalar.c' language='LANG_C99'>
<var-decl name='fletcher_4_superscalar_ops' type-id='9eeabdc8' mangled-name='fletcher_4_superscalar_ops' visibility='default' elf-symbol-id='fletcher_4_superscalar_ops'/>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_fletcher_superscalar4.c' language='LANG_C99'>
<var-decl name='fletcher_4_superscalar4_ops' type-id='9eeabdc8' mangled-name='fletcher_4_superscalar4_ops' visibility='default' elf-symbol-id='fletcher_4_superscalar4_ops'/>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_namecheck.c' language='LANG_C99'>
<enum-decl name='namecheck_err_t' naming-typedef-id='8e0af06e' id='f43bbcda'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='NAME_ERR_LEADING_SLASH' value='0'/>
<enumerator name='NAME_ERR_EMPTY_COMPONENT' value='1'/>
<enumerator name='NAME_ERR_TRAILING_SLASH' value='2'/>
<enumerator name='NAME_ERR_INVALCHAR' value='3'/>
<enumerator name='NAME_ERR_MULTIPLE_DELIMITERS' value='4'/>
<enumerator name='NAME_ERR_NOLETTER' value='5'/>
<enumerator name='NAME_ERR_RESERVED' value='6'/>
<enumerator name='NAME_ERR_DISKLIKE' value='7'/>
<enumerator name='NAME_ERR_TOOLONG' value='8'/>
<enumerator name='NAME_ERR_SELF_REF' value='9'/>
<enumerator name='NAME_ERR_PARENT_REF' value='10'/>
<enumerator name='NAME_ERR_NO_AT' value='11'/>
<enumerator name='NAME_ERR_NO_POUND' value='12'/>
</enum-decl>
<typedef-decl name='namecheck_err_t' type-id='f43bbcda' id='8e0af06e'/>
<pointer-type-def type-id='8e0af06e' size-in-bits='64' id='053457bd'/>
<var-decl name='zfs_max_dataset_nesting' type-id='95e97e5e' mangled-name='zfs_max_dataset_nesting' visibility='default' elf-symbol-id='zfs_max_dataset_nesting'/>
<function-decl name='get_dataset_depth' mangled-name='get_dataset_depth' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='get_dataset_depth'>
<parameter type-id='80f4b756' name='path'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_component_namecheck' mangled-name='zfs_component_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_component_namecheck'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='053457bd' name='why'/>
<parameter type-id='26a90f95' name='what'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='permset_namecheck' mangled-name='permset_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='permset_namecheck'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='053457bd' name='why'/>
<parameter type-id='26a90f95' name='what'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='dataset_nestcheck' mangled-name='dataset_nestcheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='dataset_nestcheck'>
<parameter type-id='80f4b756' name='path'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='entity_namecheck' mangled-name='entity_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='entity_namecheck'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='053457bd' name='why'/>
<parameter type-id='26a90f95' name='what'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='dataset_namecheck' mangled-name='dataset_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='dataset_namecheck'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='053457bd' name='why'/>
<parameter type-id='26a90f95' name='what'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='bookmark_namecheck' mangled-name='bookmark_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='bookmark_namecheck'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='053457bd' name='why'/>
<parameter type-id='26a90f95' name='what'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='snapshot_namecheck' mangled-name='snapshot_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='snapshot_namecheck'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='053457bd' name='why'/>
<parameter type-id='26a90f95' name='what'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='mountpoint_namecheck' mangled-name='mountpoint_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='mountpoint_namecheck'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='053457bd' name='why'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='pool_namecheck' mangled-name='pool_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='pool_namecheck'>
<parameter type-id='80f4b756' name='pool'/>
<parameter type-id='053457bd' name='why'/>
<parameter type-id='26a90f95' name='what'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zfs_prop.c' language='LANG_C99'>
- <array-type-def dimensions='1' type-id='80f4b756' size-in-bits='768' id='35e4b367'>
+ <array-type-def dimensions='1' type-id='b99c00c9' size-in-bits='768' id='bcc77e38'>
<subrange length='12' type-id='7359adad' id='84827bdc'/>
</array-type-def>
<enum-decl name='zprop_type_t' naming-typedef-id='31429eff' id='87676253'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='PROP_TYPE_NUMBER' value='0'/>
<enumerator name='PROP_TYPE_STRING' value='1'/>
<enumerator name='PROP_TYPE_INDEX' value='2'/>
</enum-decl>
<typedef-decl name='zprop_type_t' type-id='87676253' id='31429eff'/>
<enum-decl name='zprop_attr_t' naming-typedef-id='999701cc' id='77d05200'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='PROP_DEFAULT' value='0'/>
<enumerator name='PROP_READONLY' value='1'/>
<enumerator name='PROP_INHERIT' value='2'/>
<enumerator name='PROP_ONETIME' value='3'/>
<enumerator name='PROP_ONETIME_DEFAULT' value='4'/>
</enum-decl>
<typedef-decl name='zprop_attr_t' type-id='77d05200' id='999701cc'/>
<class-decl name='zfs_index' size-in-bits='128' is-struct='yes' visibility='default' id='87957af9'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='pi_name' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='pi_value' type-id='9c313c2d' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zprop_index_t' type-id='87957af9' id='64636ce3'/>
<class-decl name='zprop_desc_t' size-in-bits='704' is-struct='yes' naming-typedef-id='ffa52b96' visibility='default' id='bbff5e4b'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='pd_name' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='pd_propnum' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='96'>
<var-decl name='pd_proptype' type-id='31429eff' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='pd_strdefault' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='pd_numdefault' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='pd_attr' type-id='999701cc' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='288'>
<var-decl name='pd_types' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='pd_values' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='pd_colname' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='pd_rightalign' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='480'>
<var-decl name='pd_visible' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='512'>
<var-decl name='pd_zfs_mod_supported' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='576'>
<var-decl name='pd_table' type-id='c8bc397b' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='640'>
<var-decl name='pd_table_size' type-id='b59d7dce' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zprop_desc_t' type-id='bbff5e4b' id='ffa52b96'/>
<pointer-type-def type-id='80f4b756' size-in-bits='64' id='7d3cd834'/>
<qualified-type-def type-id='64636ce3' const='yes' id='072f7953'/>
<pointer-type-def type-id='072f7953' size-in-bits='64' id='c8bc397b'/>
<pointer-type-def type-id='ffa52b96' size-in-bits='64' id='76c8174b'/>
- <var-decl name='zfs_userquota_prop_prefixes' type-id='35e4b367' mangled-name='zfs_userquota_prop_prefixes' visibility='default' elf-symbol-id='zfs_userquota_prop_prefixes'/>
+ <var-decl name='zfs_userquota_prop_prefixes' type-id='bcc77e38' mangled-name='zfs_userquota_prop_prefixes' visibility='default' elf-symbol-id='zfs_userquota_prop_prefixes'/>
<function-decl name='zfs_prop_get_table' mangled-name='zfs_prop_get_table' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_table'>
<return type-id='76c8174b'/>
</function-decl>
<function-decl name='zfs_prop_init' mangled-name='zfs_prop_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_init'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_prop_delegatable' mangled-name='zfs_prop_delegatable' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_delegatable'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_name_to_prop' mangled-name='zfs_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_name_to_prop'>
<parameter type-id='80f4b756' name='propname'/>
<return type-id='58603c44'/>
</function-decl>
<function-decl name='zfs_prop_user' mangled-name='zfs_prop_user' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_user'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_userquota' mangled-name='zfs_prop_userquota' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_userquota'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_written' mangled-name='zfs_prop_written' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_written'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_string_to_index' mangled-name='zfs_prop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_string_to_index'>
<parameter type-id='58603c44' name='prop'/>
<parameter type-id='80f4b756' name='string'/>
<parameter type-id='5d6479ae' name='index'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_index_to_string' mangled-name='zfs_prop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_index_to_string'>
<parameter type-id='58603c44' name='prop'/>
<parameter type-id='9c313c2d' name='index'/>
<parameter type-id='7d3cd834' name='string'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_random_value' mangled-name='zfs_prop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_random_value'>
<parameter type-id='58603c44' name='prop'/>
<parameter type-id='9c313c2d' name='seed'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zfs_prop_valid_for_type' mangled-name='zfs_prop_valid_for_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_valid_for_type'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='2e45de5d' name='types'/>
<parameter type-id='c19b74c3' name='headcheck'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_get_type' mangled-name='zfs_prop_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_type'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='31429eff'/>
</function-decl>
<function-decl name='zfs_prop_readonly' mangled-name='zfs_prop_readonly' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_readonly'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_visible' mangled-name='zfs_prop_visible' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_visible'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_setonce' mangled-name='zfs_prop_setonce' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_setonce'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_default_string' mangled-name='zfs_prop_default_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_default_string'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_prop_default_numeric' mangled-name='zfs_prop_default_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_default_numeric'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zfs_prop_to_name' mangled-name='zfs_prop_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_to_name'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_prop_inheritable' mangled-name='zfs_prop_inheritable' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_inheritable'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_encryption_key_param' mangled-name='zfs_prop_encryption_key_param' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_encryption_key_param'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_valid_keylocation' mangled-name='zfs_prop_valid_keylocation' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_valid_keylocation'>
<parameter type-id='80f4b756' name='str'/>
<parameter type-id='c19b74c3' name='encrypted'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_prop_values' mangled-name='zfs_prop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_values'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_prop_is_string' mangled-name='zfs_prop_is_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_is_string'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_column_name' mangled-name='zfs_prop_column_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_column_name'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_prop_align_right' mangled-name='zfs_prop_align_right' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_align_right'>
<parameter type-id='58603c44' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zpool_prop.c' language='LANG_C99'>
<function-decl name='zpool_prop_get_table' mangled-name='zpool_prop_get_table' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_get_table'>
<return type-id='76c8174b'/>
</function-decl>
<function-decl name='zpool_prop_init' mangled-name='zpool_prop_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_init'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_name_to_prop' mangled-name='zpool_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_name_to_prop'>
<parameter type-id='80f4b756' name='propname'/>
<return type-id='5d0c23fb'/>
</function-decl>
<function-decl name='zpool_prop_to_name' mangled-name='zpool_prop_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_to_name'>
<parameter type-id='5d0c23fb' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zpool_prop_get_type' mangled-name='zpool_prop_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_get_type'>
<parameter type-id='5d0c23fb' name='prop'/>
<return type-id='31429eff'/>
</function-decl>
<function-decl name='zpool_prop_readonly' mangled-name='zpool_prop_readonly' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_readonly'>
<parameter type-id='5d0c23fb' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zpool_prop_setonce' mangled-name='zpool_prop_setonce' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_setonce'>
<parameter type-id='5d0c23fb' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zpool_prop_default_string' mangled-name='zpool_prop_default_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_default_string'>
<parameter type-id='5d0c23fb' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zpool_prop_default_numeric' mangled-name='zpool_prop_default_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_default_numeric'>
<parameter type-id='5d0c23fb' name='prop'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zpool_prop_feature' mangled-name='zpool_prop_feature' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_feature'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zpool_prop_unsupported' mangled-name='zpool_prop_unsupported' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_unsupported'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zpool_prop_string_to_index' mangled-name='zpool_prop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_string_to_index'>
<parameter type-id='5d0c23fb' name='prop'/>
<parameter type-id='80f4b756' name='string'/>
<parameter type-id='5d6479ae' name='index'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_prop_index_to_string' mangled-name='zpool_prop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_index_to_string'>
<parameter type-id='5d0c23fb' name='prop'/>
<parameter type-id='9c313c2d' name='index'/>
<parameter type-id='7d3cd834' name='string'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_prop_random_value' mangled-name='zpool_prop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_random_value'>
<parameter type-id='5d0c23fb' name='prop'/>
<parameter type-id='9c313c2d' name='seed'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zpool_prop_values' mangled-name='zpool_prop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_values'>
<parameter type-id='5d0c23fb' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zpool_prop_column_name' mangled-name='zpool_prop_column_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_column_name'>
<parameter type-id='5d0c23fb' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zpool_prop_align_right' mangled-name='zpool_prop_align_right' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_align_right'>
<parameter type-id='5d0c23fb' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='vdev_prop_get_table' mangled-name='vdev_prop_get_table' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_get_table'>
<return type-id='76c8174b'/>
</function-decl>
<function-decl name='vdev_prop_init' mangled-name='vdev_prop_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_init'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='vdev_name_to_prop' mangled-name='vdev_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_name_to_prop'>
<parameter type-id='80f4b756' name='propname'/>
<return type-id='5aa5c90c'/>
</function-decl>
<function-decl name='vdev_prop_user' mangled-name='vdev_prop_user' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_user'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='vdev_prop_to_name' mangled-name='vdev_prop_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_to_name'>
<parameter type-id='5aa5c90c' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='vdev_prop_get_type' mangled-name='vdev_prop_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_get_type'>
<parameter type-id='5aa5c90c' name='prop'/>
<return type-id='31429eff'/>
</function-decl>
<function-decl name='vdev_prop_readonly' mangled-name='vdev_prop_readonly' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_readonly'>
<parameter type-id='5aa5c90c' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='vdev_prop_default_string' mangled-name='vdev_prop_default_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_default_string'>
<parameter type-id='5aa5c90c' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='vdev_prop_default_numeric' mangled-name='vdev_prop_default_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_default_numeric'>
<parameter type-id='5aa5c90c' name='prop'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='vdev_prop_string_to_index' mangled-name='vdev_prop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_string_to_index'>
<parameter type-id='5aa5c90c' name='prop'/>
<parameter type-id='80f4b756' name='string'/>
<parameter type-id='5d6479ae' name='index'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='vdev_prop_index_to_string' mangled-name='vdev_prop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_index_to_string'>
<parameter type-id='5aa5c90c' name='prop'/>
<parameter type-id='9c313c2d' name='index'/>
<parameter type-id='7d3cd834' name='string'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_prop_vdev' mangled-name='zpool_prop_vdev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_vdev'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='vdev_prop_random_value' mangled-name='vdev_prop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_random_value'>
<parameter type-id='5aa5c90c' name='prop'/>
<parameter type-id='9c313c2d' name='seed'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='vdev_prop_values' mangled-name='vdev_prop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_values'>
<parameter type-id='5aa5c90c' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='vdev_prop_column_name' mangled-name='vdev_prop_column_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_column_name'>
<parameter type-id='5aa5c90c' name='prop'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='vdev_prop_align_right' mangled-name='vdev_prop_align_right' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_align_right'>
<parameter type-id='5aa5c90c' name='prop'/>
<return type-id='c19b74c3'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='../../module/zcommon/zprop_common.c' language='LANG_C99'>
<function-decl name='zprop_register_impl' mangled-name='zprop_register_impl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_impl'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='31429eff' name='type'/>
<parameter type-id='9c313c2d' name='numdefault'/>
<parameter type-id='80f4b756' name='strdefault'/>
<parameter type-id='999701cc' name='attr'/>
<parameter type-id='95e97e5e' name='objset_types'/>
<parameter type-id='80f4b756' name='values'/>
<parameter type-id='80f4b756' name='colname'/>
<parameter type-id='c19b74c3' name='rightalign'/>
<parameter type-id='c19b74c3' name='visible'/>
<parameter type-id='c8bc397b' name='idx_tbl'/>
<parameter type-id='a3372543' name='sfeatures'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zprop_register_string' mangled-name='zprop_register_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_string'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='80f4b756' name='def'/>
<parameter type-id='999701cc' name='attr'/>
<parameter type-id='95e97e5e' name='objset_types'/>
<parameter type-id='80f4b756' name='values'/>
<parameter type-id='80f4b756' name='colname'/>
<parameter type-id='a3372543' name='sfeatures'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zprop_register_number' mangled-name='zprop_register_number' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_number'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9c313c2d' name='def'/>
<parameter type-id='999701cc' name='attr'/>
<parameter type-id='95e97e5e' name='objset_types'/>
<parameter type-id='80f4b756' name='values'/>
<parameter type-id='80f4b756' name='colname'/>
<parameter type-id='a3372543' name='sfeatures'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zprop_register_index' mangled-name='zprop_register_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_index'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='9c313c2d' name='def'/>
<parameter type-id='999701cc' name='attr'/>
<parameter type-id='95e97e5e' name='objset_types'/>
<parameter type-id='80f4b756' name='values'/>
<parameter type-id='80f4b756' name='colname'/>
<parameter type-id='c8bc397b' name='idx_tbl'/>
<parameter type-id='a3372543' name='sfeatures'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zprop_register_hidden' mangled-name='zprop_register_hidden' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_hidden'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='31429eff' name='type'/>
<parameter type-id='999701cc' name='attr'/>
<parameter type-id='95e97e5e' name='objset_types'/>
<parameter type-id='80f4b756' name='colname'/>
<parameter type-id='a3372543' name='sfeatures'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zprop_iter_common' mangled-name='zprop_iter_common' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_iter_common'>
<parameter type-id='1ec3747a' name='func'/>
<parameter type-id='eaa32e2f' name='cb'/>
<parameter type-id='c19b74c3' name='show_all'/>
<parameter type-id='c19b74c3' name='ordered'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zprop_name_to_prop' mangled-name='zprop_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_name_to_prop'>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zprop_string_to_index' mangled-name='zprop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_string_to_index'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='80f4b756' name='string'/>
<parameter type-id='5d6479ae' name='index'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zprop_index_to_string' mangled-name='zprop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_index_to_string'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='9c313c2d' name='index'/>
<parameter type-id='7d3cd834' name='string'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zprop_random_value' mangled-name='zprop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_random_value'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='9c313c2d' name='seed'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zprop_values' mangled-name='zprop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_values'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zprop_valid_for_type' mangled-name='zprop_valid_for_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_valid_for_type'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='2e45de5d' name='type'/>
<parameter type-id='c19b74c3' name='headcheck'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zprop_valid_char' mangled-name='zprop_valid_char' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_valid_char'>
<parameter type-id='a84c031d' name='c'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zprop_width' mangled-name='zprop_width' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_width'>
<parameter type-id='95e97e5e' name='prop'/>
<parameter type-id='37e3bd22' name='fixed'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='b59d7dce'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='libzfs_changelist.c' language='LANG_C99'>
<type-decl name='void' id='48b5725f'/>
</abi-instr>
<abi-instr address-size='64' path='libzfs_config.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='bf311473' size-in-bits='128' id='f0f65199'>
<subrange length='2' type-id='7359adad' id='52efc4ef'/>
</array-type-def>
<type-decl name='char' size-in-bits='8' id='a84c031d'/>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8192' id='b54ce520'>
<subrange length='1024' type-id='7359adad' id='c60446f8'/>
</array-type-def>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='2048' id='d1617432'>
<subrange length='256' type-id='7359adad' id='36e5b9fa'/>
</array-type-def>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='320' id='36c46961'>
<subrange length='40' type-id='7359adad' id='8f80b239'/>
</array-type-def>
+ <class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
+ <class-decl name='uu_avl' is-struct='yes' visibility='default' is-declaration-only='yes' id='4af029d1'/>
+ <class-decl name='uu_avl_pool' is-struct='yes' visibility='default' is-declaration-only='yes' id='12a530a8'/>
<type-decl name='int' size-in-bits='32' id='95e97e5e'/>
<type-decl name='long int' size-in-bits='64' id='bd54fe1a'/>
<type-decl name='long long int' size-in-bits='64' id='1eb56b1e'/>
<type-decl name='short int' size-in-bits='16' id='a2185560'/>
<type-decl name='unnamed-enum-underlying-type-32' is-anonymous='yes' size-in-bits='32' alignment-in-bits='32' id='9cac1fee'/>
<type-decl name='unsigned char' size-in-bits='8' id='002ac4a6'/>
<type-decl name='unsigned int' size-in-bits='32' id='f0981eeb'/>
<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
<typedef-decl name='uu_avl_pool_t' type-id='12a530a8' id='7f84e390'/>
<typedef-decl name='uu_avl_t' type-id='4af029d1' id='bb7f0973'/>
<typedef-decl name='zfs_handle_t' type-id='f6ee4445' id='775509eb'/>
<typedef-decl name='zpool_handle_t' type-id='67002a8a' id='b1efc708'/>
<typedef-decl name='libzfs_handle_t' type-id='c8a9d9d8' id='95942d0c'/>
<typedef-decl name='zpool_iter_f' type-id='3aebb66f' id='fa476e62'/>
<typedef-decl name='zfs_iter_f' type-id='5571cde4' id='d8e49ab9'/>
<typedef-decl name='avl_tree_t' type-id='b351119f' id='f20fbd51'/>
<class-decl name='avl_node' size-in-bits='192' is-struct='yes' visibility='default' id='428b67b3'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='avl_child' type-id='f0f65199' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='avl_pcb' type-id='e475ab95' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='avl_tree' size-in-bits='320' is-struct='yes' visibility='default' id='b351119f'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='avl_root' type-id='bf311473' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='avl_compar' type-id='585e1de9' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='avl_offset' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='avl_numnodes' type-id='ee1f298e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='avl_pad' type-id='b59d7dce' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='dmu_objset_stats' size-in-bits='2304' is-struct='yes' visibility='default' id='098f0221'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='dds_num_clones' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='dds_creation_txg' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='dds_guid' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='dds_type' type-id='230f1e16' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='224'>
<var-decl name='dds_is_snapshot' type-id='b96825af' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='232'>
<var-decl name='dds_inconsistent' type-id='b96825af' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='240'>
<var-decl name='dds_redacted' type-id='b96825af' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='248'>
<var-decl name='dds_origin' type-id='d1617432' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='dmu_objset_stats_t' type-id='098f0221' id='b2c14f17'/>
<enum-decl name='zfs_type_t' naming-typedef-id='2e45de5d' id='5d8f7321'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZFS_TYPE_FILESYSTEM' value='1'/>
<enumerator name='ZFS_TYPE_SNAPSHOT' value='2'/>
<enumerator name='ZFS_TYPE_VOLUME' value='4'/>
<enumerator name='ZFS_TYPE_POOL' value='8'/>
<enumerator name='ZFS_TYPE_BOOKMARK' value='16'/>
<enumerator name='ZFS_TYPE_VDEV' value='32'/>
</enum-decl>
<typedef-decl name='zfs_type_t' type-id='5d8f7321' id='2e45de5d'/>
<enum-decl name='dmu_objset_type' id='6b1b19f9'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='DMU_OST_NONE' value='0'/>
<enumerator name='DMU_OST_META' value='1'/>
<enumerator name='DMU_OST_ZFS' value='2'/>
<enumerator name='DMU_OST_ZVOL' value='3'/>
<enumerator name='DMU_OST_OTHER' value='4'/>
<enumerator name='DMU_OST_ANY' value='5'/>
<enumerator name='DMU_OST_NUMTYPES' value='6'/>
</enum-decl>
<typedef-decl name='dmu_objset_type_t' type-id='6b1b19f9' id='230f1e16'/>
<class-decl name='nvlist' size-in-bits='192' is-struct='yes' visibility='default' id='ac266fd9'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='nvl_version' type-id='3ff5601b' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='nvl_nvflag' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='nvl_priv' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='nvl_flag' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='160'>
<var-decl name='nvl_pad' type-id='3ff5601b' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='nvlist_t' type-id='ac266fd9' id='8e8d4be3'/>
<enum-decl name='boolean_t' naming-typedef-id='c19b74c3' id='f58c8277'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='B_FALSE' value='0'/>
<enumerator name='B_TRUE' value='1'/>
</enum-decl>
<typedef-decl name='boolean_t' type-id='f58c8277' id='c19b74c3'/>
<typedef-decl name='ulong_t' type-id='7359adad' id='ee1f298e'/>
<typedef-decl name='longlong_t' type-id='1eb56b1e' id='9b3ff54f'/>
<typedef-decl name='diskaddr_t' type-id='9b3ff54f' id='804dc465'/>
<class-decl name='libzfs_handle' size-in-bits='18240' is-struct='yes' visibility='default' id='c8a9d9d8'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='libzfs_error' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='libzfs_fd' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='libzfs_pool_handles' type-id='4c81de99' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='libzfs_ns_avlpool' type-id='de82c773' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='libzfs_ns_avl' type-id='a5c21a38' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='libzfs_ns_gen' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='libzfs_desc_active' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='352'>
<var-decl name='libzfs_action' type-id='b54ce520' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='8544'>
<var-decl name='libzfs_desc' type-id='b54ce520' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='16736'>
<var-decl name='libzfs_printerr' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='16768'>
<var-decl name='libzfs_mnttab_enable' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='16832'>
<var-decl name='libzfs_mnttab_cache_lock' type-id='7a6844eb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='17152'>
<var-decl name='libzfs_mnttab_cache' type-id='f20fbd51' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='17472'>
<var-decl name='libzfs_pool_iter' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='17504'>
<var-decl name='libzfs_prop_debug' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='17536'>
<var-decl name='libzfs_urire' type-id='aca3bac8' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='18048'>
<var-decl name='libzfs_max_nvlist' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='18112'>
<var-decl name='libfetch' type-id='eaa32e2f' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='18176'>
<var-decl name='libfetch_load_error' type-id='26a90f95' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='zfs_handle' size-in-bits='4928' is-struct='yes' visibility='default' id='f6ee4445'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='zfs_hdl' type-id='b0382bb3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='zpool_hdl' type-id='4c81de99' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='zfs_name' type-id='d1617432' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2176'>
<var-decl name='zfs_type' type-id='2e45de5d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2208'>
<var-decl name='zfs_head_type' type-id='2e45de5d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2240'>
<var-decl name='zfs_dmustats' type-id='b2c14f17' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='4544'>
<var-decl name='zfs_props' type-id='5ce45b60' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='4608'>
<var-decl name='zfs_user_props' type-id='5ce45b60' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='4672'>
<var-decl name='zfs_recvd_props' type-id='5ce45b60' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='4736'>
<var-decl name='zfs_mntcheck' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='4800'>
<var-decl name='zfs_mntopts' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='4864'>
<var-decl name='zfs_props_table' type-id='ae3e8ca6' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='zpool_handle' size-in-bits='2560' is-struct='yes' visibility='default' id='67002a8a'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='zpool_hdl' type-id='b0382bb3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='zpool_next' type-id='4c81de99' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='zpool_name' type-id='d1617432' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2176'>
<var-decl name='zpool_state' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2240'>
<var-decl name='zpool_config_size' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2304'>
<var-decl name='zpool_config' type-id='5ce45b60' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2368'>
<var-decl name='zpool_old_config' type-id='5ce45b60' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2432'>
<var-decl name='zpool_props' type-id='5ce45b60' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2496'>
<var-decl name='zpool_start_block' type-id='804dc465' visibility='default'/>
</data-member>
</class-decl>
+ <typedef-decl name='__re_long_size_t' type-id='7359adad' id='ba516949'/>
<typedef-decl name='reg_syntax_t' type-id='7359adad' id='1b72c3b3'/>
<class-decl name='re_pattern_buffer' size-in-bits='512' is-struct='yes' visibility='default' id='19fc9a8c'>
<data-member access='public' layout-offset-in-bits='0'>
- <var-decl name='buffer' type-id='cf536864' visibility='default'/>
+ <var-decl name='buffer' type-id='33976309' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
- <var-decl name='allocated' type-id='7359adad' visibility='default'/>
+ <var-decl name='allocated' type-id='ba516949' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
- <var-decl name='used' type-id='7359adad' visibility='default'/>
+ <var-decl name='used' type-id='ba516949' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='syntax' type-id='1b72c3b3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='fastmap' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='translate' type-id='cf536864' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='re_nsub' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='can_be_null' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='449'>
<var-decl name='regs_allocated' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='451'>
<var-decl name='fastmap_accurate' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='452'>
<var-decl name='no_sub' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='453'>
<var-decl name='not_bol' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='454'>
<var-decl name='not_eol' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='455'>
<var-decl name='newline_anchor' type-id='f0981eeb' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='regex_t' type-id='19fc9a8c' id='aca3bac8'/>
<typedef-decl name='uintptr_t' type-id='7359adad' id='e475ab95'/>
<union-decl name='pthread_mutex_t' size-in-bits='320' naming-typedef-id='7a6844eb' visibility='default' id='70681f9b'>
<data-member access='public'>
<var-decl name='__data' type-id='4c734837' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='__size' type-id='36c46961' visibility='default'/>
</data-member>
<data-member access='public'>
<var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
</data-member>
</union-decl>
<typedef-decl name='pthread_mutex_t' type-id='70681f9b' id='7a6844eb'/>
<typedef-decl name='int32_t' type-id='33f57a65' id='3ff5601b'/>
<typedef-decl name='uint8_t' type-id='c51d6389' id='b96825af'/>
<typedef-decl name='uint32_t' type-id='62f1140c' id='8f92235e'/>
<typedef-decl name='uint64_t' type-id='8910171f' id='9c313c2d'/>
- <class-decl name='__pthread_internal_list' size-in-bits='128' is-struct='yes' visibility='default' id='0e01899c'>
- <data-member access='public' layout-offset-in-bits='0'>
- <var-decl name='__prev' type-id='4d98cd5a' visibility='default'/>
- </data-member>
- <data-member access='public' layout-offset-in-bits='64'>
- <var-decl name='__next' type-id='4d98cd5a' visibility='default'/>
- </data-member>
- </class-decl>
- <typedef-decl name='__pthread_list_t' type-id='0e01899c' id='518fb49c'/>
<class-decl name='__pthread_mutex_s' size-in-bits='320' is-struct='yes' visibility='default' id='4c734837'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='__lock' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='__count' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='__owner' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='96'>
<var-decl name='__nusers' type-id='f0981eeb' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='__kind' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='160'>
<var-decl name='__spins' type-id='a2185560' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='176'>
<var-decl name='__elision' type-id='a2185560' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='__list' type-id='518fb49c' visibility='default'/>
</data-member>
</class-decl>
+ <class-decl name='__pthread_internal_list' size-in-bits='128' is-struct='yes' visibility='default' id='0e01899c'>
+ <data-member access='public' layout-offset-in-bits='0'>
+ <var-decl name='__prev' type-id='4d98cd5a' visibility='default'/>
+ </data-member>
+ <data-member access='public' layout-offset-in-bits='64'>
+ <var-decl name='__next' type-id='4d98cd5a' visibility='default'/>
+ </data-member>
+ </class-decl>
+ <typedef-decl name='__pthread_list_t' type-id='0e01899c' id='518fb49c'/>
<typedef-decl name='__uint8_t' type-id='002ac4a6' id='c51d6389'/>
<typedef-decl name='__int32_t' type-id='95e97e5e' id='33f57a65'/>
<typedef-decl name='__uint32_t' type-id='f0981eeb' id='62f1140c'/>
<typedef-decl name='__uint64_t' type-id='7359adad' id='8910171f'/>
<typedef-decl name='size_t' type-id='7359adad' id='b59d7dce'/>
<pointer-type-def type-id='0e01899c' size-in-bits='64' id='4d98cd5a'/>
<pointer-type-def type-id='428b67b3' size-in-bits='64' id='bf311473'/>
<pointer-type-def type-id='c19b74c3' size-in-bits='64' id='37e3bd22'/>
<pointer-type-def type-id='a84c031d' size-in-bits='64' id='26a90f95'/>
<qualified-type-def type-id='a84c031d' const='yes' id='9b45d938'/>
<pointer-type-def type-id='9b45d938' size-in-bits='64' id='80f4b756'/>
<pointer-type-def type-id='96ee24a5' size-in-bits='64' id='585e1de9'/>
<pointer-type-def type-id='cb9628fa' size-in-bits='64' id='5571cde4'/>
<pointer-type-def type-id='2bce87e3' size-in-bits='64' id='3aebb66f'/>
<pointer-type-def type-id='95942d0c' size-in-bits='64' id='b0382bb3'/>
<pointer-type-def type-id='8e8d4be3' size-in-bits='64' id='5ce45b60'/>
<pointer-type-def type-id='5ce45b60' size-in-bits='64' id='857bb57e'/>
+ <pointer-type-def type-id='b48d2441' size-in-bits='64' id='33976309'/>
<pointer-type-def type-id='b96825af' size-in-bits='64' id='ae3e8ca6'/>
<pointer-type-def type-id='002ac4a6' size-in-bits='64' id='cf536864'/>
<pointer-type-def type-id='7f84e390' size-in-bits='64' id='de82c773'/>
<pointer-type-def type-id='bb7f0973' size-in-bits='64' id='a5c21a38'/>
<pointer-type-def type-id='48b5725f' size-in-bits='64' id='eaa32e2f'/>
<pointer-type-def type-id='775509eb' size-in-bits='64' id='9200a744'/>
<pointer-type-def type-id='b1efc708' size-in-bits='64' id='4c81de99'/>
+ <class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
<class-decl name='uu_avl' is-struct='yes' visibility='default' is-declaration-only='yes' id='4af029d1'/>
<class-decl name='uu_avl_pool' is-struct='yes' visibility='default' is-declaration-only='yes' id='12a530a8'/>
<function-decl name='zpool_get_config' mangled-name='zpool_get_config' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_config'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='857bb57e' name='oldconfig'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zpool_get_features' mangled-name='zpool_get_features' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_features'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zpool_refresh_stats' mangled-name='zpool_refresh_stats' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_refresh_stats'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='37e3bd22' name='missing'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_skip_pool' mangled-name='zpool_skip_pool' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_skip_pool'>
<parameter type-id='80f4b756' name='poolname'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zpool_iter' mangled-name='zpool_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_iter'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='fa476e62' name='func'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_iter_root' mangled-name='zfs_iter_root' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_root'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='d8e49ab9' name='func'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-type size-in-bits='64' id='cb9628fa'>
<parameter type-id='9200a744'/>
<parameter type-id='eaa32e2f'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='2bce87e3'>
<parameter type-id='4c81de99'/>
<parameter type-id='eaa32e2f'/>
<return type-id='95e97e5e'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='libzfs_crypto.c' language='LANG_C99'>
<typedef-decl name='uint_t' type-id='f0981eeb' id='3502e3ff'/>
<pointer-type-def type-id='ae3e8ca6' size-in-bits='64' id='d8774064'/>
<pointer-type-def type-id='3502e3ff' size-in-bits='64' id='4dd26a40'/>
<function-decl name='zfs_crypto_get_encryption_root' mangled-name='zfs_crypto_get_encryption_root' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_get_encryption_root'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='37e3bd22' name='is_encroot'/>
<parameter type-id='26a90f95' name='buf'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_crypto_create' mangled-name='zfs_crypto_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_create'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='26a90f95' name='parent_name'/>
<parameter type-id='5ce45b60' name='props'/>
<parameter type-id='5ce45b60' name='pool_props'/>
<parameter type-id='c19b74c3' name='stdin_available'/>
<parameter type-id='d8774064' name='wkeydata_out'/>
<parameter type-id='4dd26a40' name='wkeylen_out'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_crypto_clone_check' mangled-name='zfs_crypto_clone_check' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_clone_check'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='9200a744' name='origin_zhp'/>
<parameter type-id='26a90f95' name='parent_name'/>
<parameter type-id='5ce45b60' name='props'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_crypto_attempt_load_keys' mangled-name='zfs_crypto_attempt_load_keys' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_attempt_load_keys'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='26a90f95' name='fsname'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_crypto_load_key' mangled-name='zfs_crypto_load_key' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_load_key'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='c19b74c3' name='noop'/>
<parameter type-id='26a90f95' name='alt_keylocation'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_crypto_unload_key' mangled-name='zfs_crypto_unload_key' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_unload_key'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_crypto_rewrap' mangled-name='zfs_crypto_rewrap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_rewrap'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='5ce45b60' name='raw_props'/>
<parameter type-id='c19b74c3' name='inheritkey'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='libzfs_dataset.c' language='LANG_C99'>
<class-decl name='zprop_list' size-in-bits='448' is-struct='yes' visibility='default' id='bd9b4291'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='pl_prop' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='pl_user_prop' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='pl_next' type-id='9f1a1109' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='pl_all' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='pl_width' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='pl_recvd_width' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='pl_fixed' type-id='c19b74c3' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zprop_list_t' type-id='bd9b4291' id='bdb8ac4f'/>
<class-decl name='renameflags' size-in-bits='32' is-struct='yes' visibility='default' id='7aee5792'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='recursive' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1'>
<var-decl name='nounmount' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2'>
<var-decl name='forceunmount' type-id='95e97e5e' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='renameflags_t' type-id='7aee5792' id='067170c2'/>
<typedef-decl name='zfs_userspace_cb_t' type-id='ca64ff60' id='16c5f410'/>
<enum-decl name='zfs_prop_t' naming-typedef-id='58603c44' id='4b000d60'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZPROP_CONT' value='-2'/>
<enumerator name='ZPROP_INVAL' value='-1'/>
<enumerator name='ZFS_PROP_TYPE' value='0'/>
<enumerator name='ZFS_PROP_CREATION' value='1'/>
<enumerator name='ZFS_PROP_USED' value='2'/>
<enumerator name='ZFS_PROP_AVAILABLE' value='3'/>
<enumerator name='ZFS_PROP_REFERENCED' value='4'/>
<enumerator name='ZFS_PROP_COMPRESSRATIO' value='5'/>
<enumerator name='ZFS_PROP_MOUNTED' value='6'/>
<enumerator name='ZFS_PROP_ORIGIN' value='7'/>
<enumerator name='ZFS_PROP_QUOTA' value='8'/>
<enumerator name='ZFS_PROP_RESERVATION' value='9'/>
<enumerator name='ZFS_PROP_VOLSIZE' value='10'/>
<enumerator name='ZFS_PROP_VOLBLOCKSIZE' value='11'/>
<enumerator name='ZFS_PROP_RECORDSIZE' value='12'/>
<enumerator name='ZFS_PROP_MOUNTPOINT' value='13'/>
<enumerator name='ZFS_PROP_SHARENFS' value='14'/>
<enumerator name='ZFS_PROP_CHECKSUM' value='15'/>
<enumerator name='ZFS_PROP_COMPRESSION' value='16'/>
<enumerator name='ZFS_PROP_ATIME' value='17'/>
<enumerator name='ZFS_PROP_DEVICES' value='18'/>
<enumerator name='ZFS_PROP_EXEC' value='19'/>
<enumerator name='ZFS_PROP_SETUID' value='20'/>
<enumerator name='ZFS_PROP_READONLY' value='21'/>
<enumerator name='ZFS_PROP_ZONED' value='22'/>
<enumerator name='ZFS_PROP_SNAPDIR' value='23'/>
<enumerator name='ZFS_PROP_ACLMODE' value='24'/>
<enumerator name='ZFS_PROP_ACLINHERIT' value='25'/>
<enumerator name='ZFS_PROP_CREATETXG' value='26'/>
<enumerator name='ZFS_PROP_NAME' value='27'/>
<enumerator name='ZFS_PROP_CANMOUNT' value='28'/>
<enumerator name='ZFS_PROP_ISCSIOPTIONS' value='29'/>
<enumerator name='ZFS_PROP_XATTR' value='30'/>
<enumerator name='ZFS_PROP_NUMCLONES' value='31'/>
<enumerator name='ZFS_PROP_COPIES' value='32'/>
<enumerator name='ZFS_PROP_VERSION' value='33'/>
<enumerator name='ZFS_PROP_UTF8ONLY' value='34'/>
<enumerator name='ZFS_PROP_NORMALIZE' value='35'/>
<enumerator name='ZFS_PROP_CASE' value='36'/>
<enumerator name='ZFS_PROP_VSCAN' value='37'/>
<enumerator name='ZFS_PROP_NBMAND' value='38'/>
<enumerator name='ZFS_PROP_SHARESMB' value='39'/>
<enumerator name='ZFS_PROP_REFQUOTA' value='40'/>
<enumerator name='ZFS_PROP_REFRESERVATION' value='41'/>
<enumerator name='ZFS_PROP_GUID' value='42'/>
<enumerator name='ZFS_PROP_PRIMARYCACHE' value='43'/>
<enumerator name='ZFS_PROP_SECONDARYCACHE' value='44'/>
<enumerator name='ZFS_PROP_USEDSNAP' value='45'/>
<enumerator name='ZFS_PROP_USEDDS' value='46'/>
<enumerator name='ZFS_PROP_USEDCHILD' value='47'/>
<enumerator name='ZFS_PROP_USEDREFRESERV' value='48'/>
<enumerator name='ZFS_PROP_USERACCOUNTING' value='49'/>
<enumerator name='ZFS_PROP_STMF_SHAREINFO' value='50'/>
<enumerator name='ZFS_PROP_DEFER_DESTROY' value='51'/>
<enumerator name='ZFS_PROP_USERREFS' value='52'/>
<enumerator name='ZFS_PROP_LOGBIAS' value='53'/>
<enumerator name='ZFS_PROP_UNIQUE' value='54'/>
<enumerator name='ZFS_PROP_OBJSETID' value='55'/>
<enumerator name='ZFS_PROP_DEDUP' value='56'/>
<enumerator name='ZFS_PROP_MLSLABEL' value='57'/>
<enumerator name='ZFS_PROP_SYNC' value='58'/>
<enumerator name='ZFS_PROP_DNODESIZE' value='59'/>
<enumerator name='ZFS_PROP_REFRATIO' value='60'/>
<enumerator name='ZFS_PROP_WRITTEN' value='61'/>
<enumerator name='ZFS_PROP_CLONES' value='62'/>
<enumerator name='ZFS_PROP_LOGICALUSED' value='63'/>
<enumerator name='ZFS_PROP_LOGICALREFERENCED' value='64'/>
<enumerator name='ZFS_PROP_INCONSISTENT' value='65'/>
<enumerator name='ZFS_PROP_VOLMODE' value='66'/>
<enumerator name='ZFS_PROP_FILESYSTEM_LIMIT' value='67'/>
<enumerator name='ZFS_PROP_SNAPSHOT_LIMIT' value='68'/>
<enumerator name='ZFS_PROP_FILESYSTEM_COUNT' value='69'/>
<enumerator name='ZFS_PROP_SNAPSHOT_COUNT' value='70'/>
<enumerator name='ZFS_PROP_SNAPDEV' value='71'/>
<enumerator name='ZFS_PROP_ACLTYPE' value='72'/>
<enumerator name='ZFS_PROP_SELINUX_CONTEXT' value='73'/>
<enumerator name='ZFS_PROP_SELINUX_FSCONTEXT' value='74'/>
<enumerator name='ZFS_PROP_SELINUX_DEFCONTEXT' value='75'/>
<enumerator name='ZFS_PROP_SELINUX_ROOTCONTEXT' value='76'/>
<enumerator name='ZFS_PROP_RELATIME' value='77'/>
<enumerator name='ZFS_PROP_REDUNDANT_METADATA' value='78'/>
<enumerator name='ZFS_PROP_OVERLAY' value='79'/>
<enumerator name='ZFS_PROP_PREV_SNAP' value='80'/>
<enumerator name='ZFS_PROP_RECEIVE_RESUME_TOKEN' value='81'/>
<enumerator name='ZFS_PROP_ENCRYPTION' value='82'/>
<enumerator name='ZFS_PROP_KEYLOCATION' value='83'/>
<enumerator name='ZFS_PROP_KEYFORMAT' value='84'/>
<enumerator name='ZFS_PROP_PBKDF2_SALT' value='85'/>
<enumerator name='ZFS_PROP_PBKDF2_ITERS' value='86'/>
<enumerator name='ZFS_PROP_ENCRYPTION_ROOT' value='87'/>
<enumerator name='ZFS_PROP_KEY_GUID' value='88'/>
<enumerator name='ZFS_PROP_KEYSTATUS' value='89'/>
<enumerator name='ZFS_PROP_REMAPTXG' value='90'/>
<enumerator name='ZFS_PROP_SPECIAL_SMALL_BLOCKS' value='91'/>
<enumerator name='ZFS_PROP_IVSET_GUID' value='92'/>
<enumerator name='ZFS_PROP_REDACTED' value='93'/>
<enumerator name='ZFS_PROP_REDACT_SNAPS' value='94'/>
<enumerator name='ZFS_NUM_PROPS' value='95'/>
</enum-decl>
<typedef-decl name='zfs_prop_t' type-id='4b000d60' id='58603c44'/>
<enum-decl name='zfs_userquota_prop_t' naming-typedef-id='279fde6a' id='5258d2f6'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZFS_PROP_USERUSED' value='0'/>
<enumerator name='ZFS_PROP_USERQUOTA' value='1'/>
<enumerator name='ZFS_PROP_GROUPUSED' value='2'/>
<enumerator name='ZFS_PROP_GROUPQUOTA' value='3'/>
<enumerator name='ZFS_PROP_USEROBJUSED' value='4'/>
<enumerator name='ZFS_PROP_USEROBJQUOTA' value='5'/>
<enumerator name='ZFS_PROP_GROUPOBJUSED' value='6'/>
<enumerator name='ZFS_PROP_GROUPOBJQUOTA' value='7'/>
<enumerator name='ZFS_PROP_PROJECTUSED' value='8'/>
<enumerator name='ZFS_PROP_PROJECTQUOTA' value='9'/>
<enumerator name='ZFS_PROP_PROJECTOBJUSED' value='10'/>
<enumerator name='ZFS_PROP_PROJECTOBJQUOTA' value='11'/>
<enumerator name='ZFS_NUM_USERQUOTA_PROPS' value='12'/>
</enum-decl>
<typedef-decl name='zfs_userquota_prop_t' type-id='5258d2f6' id='279fde6a'/>
<enum-decl name='zprop_source_t' naming-typedef-id='a2256d42' id='5903f80e'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZPROP_SRC_NONE' value='1'/>
<enumerator name='ZPROP_SRC_DEFAULT' value='2'/>
<enumerator name='ZPROP_SRC_TEMPORARY' value='4'/>
<enumerator name='ZPROP_SRC_LOCAL' value='8'/>
<enumerator name='ZPROP_SRC_INHERITED' value='16'/>
<enumerator name='ZPROP_SRC_RECEIVED' value='32'/>
</enum-decl>
<typedef-decl name='zprop_source_t' type-id='5903f80e' id='a2256d42'/>
<enum-decl name='zfs_wait_activity_t' naming-typedef-id='3024501a' id='527d5dc6'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZFS_WAIT_DELETEQ' value='0'/>
<enumerator name='ZFS_WAIT_NUM_ACTIVITIES' value='1'/>
</enum-decl>
<typedef-decl name='zfs_wait_activity_t' type-id='527d5dc6' id='3024501a'/>
<class-decl name='mnttab' size-in-bits='256' is-struct='yes' visibility='default' id='1b055409'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='__uid_t' type-id='f0981eeb' id='cc5fcceb'/>
<typedef-decl name='uid_t' type-id='cc5fcceb' id='354978ed'/>
<pointer-type-def type-id='26a90f95' size-in-bits='64' id='9b23c9ad'/>
<qualified-type-def type-id='775509eb' const='yes' id='5eadf2db'/>
<pointer-type-def type-id='5eadf2db' size-in-bits='64' id='fcd57163'/>
<pointer-type-def type-id='7e291ce6' size-in-bits='64' id='ca64ff60'/>
<pointer-type-def type-id='95e97e5e' size-in-bits='64' id='7292109c'/>
<pointer-type-def type-id='1b055409' size-in-bits='64' id='9d424d31'/>
<pointer-type-def type-id='9c313c2d' size-in-bits='64' id='5d6479ae'/>
<pointer-type-def type-id='bd9b4291' size-in-bits='64' id='9f1a1109'/>
<pointer-type-def type-id='bdb8ac4f' size-in-bits='64' id='3a9b2288'/>
<pointer-type-def type-id='3a9b2288' size-in-bits='64' id='e4378506'/>
<pointer-type-def type-id='a2256d42' size-in-bits='64' id='debc6aa3'/>
<function-decl name='zfs_type_to_name' mangled-name='zfs_type_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_type_to_name'>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_name_valid' mangled-name='zfs_name_valid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_name_valid'>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_free_handles' mangled-name='zpool_free_handles' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_free_handles'>
<parameter type-id='b0382bb3' name='hdl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_refresh_properties' mangled-name='zfs_refresh_properties' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_refresh_properties'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_handle_dup' mangled-name='zfs_handle_dup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_handle_dup'>
<parameter type-id='9200a744' name='zhp_orig'/>
<return type-id='9200a744'/>
</function-decl>
<function-decl name='zfs_bookmark_exists' mangled-name='zfs_bookmark_exists' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_bookmark_exists'>
<parameter type-id='80f4b756' name='path'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_open' mangled-name='zfs_open' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_open'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='95e97e5e' name='types'/>
<return type-id='9200a744'/>
</function-decl>
<function-decl name='zfs_close' mangled-name='zfs_close' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_close'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='libzfs_mnttab_init' mangled-name='libzfs_mnttab_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_init'>
<parameter type-id='b0382bb3' name='hdl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='libzfs_mnttab_fini' mangled-name='libzfs_mnttab_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_fini'>
<parameter type-id='b0382bb3' name='hdl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='libzfs_mnttab_cache' mangled-name='libzfs_mnttab_cache' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_cache'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='c19b74c3' name='enable'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='libzfs_mnttab_find' mangled-name='libzfs_mnttab_find' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_find'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='fsname'/>
<parameter type-id='9d424d31' name='entry'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='libzfs_mnttab_add' mangled-name='libzfs_mnttab_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_add'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='special'/>
<parameter type-id='80f4b756' name='mountp'/>
<parameter type-id='80f4b756' name='mntopts'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='libzfs_mnttab_remove' mangled-name='libzfs_mnttab_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_remove'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='fsname'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_spa_version' mangled-name='zfs_spa_version' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_spa_version'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='7292109c' name='spa_version'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_valid_proplist' mangled-name='zfs_valid_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_valid_proplist'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='2e45de5d' name='type'/>
<parameter type-id='5ce45b60' name='nvl'/>
<parameter type-id='9c313c2d' name='zoned'/>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='4c81de99' name='zpool_hdl'/>
<parameter type-id='c19b74c3' name='key_params_ok'/>
<parameter type-id='80f4b756' name='errbuf'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zfs_prop_set' mangled-name='zfs_prop_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_set'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='80f4b756' name='propval'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_set_list' mangled-name='zfs_prop_set_list' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_set_list'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='5ce45b60' name='props'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_inherit' mangled-name='zfs_prop_inherit' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_inherit'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='c19b74c3' name='received'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='getprop_uint64' mangled-name='getprop_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getprop_uint64'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='58603c44' name='prop'/>
<parameter type-id='9b23c9ad' name='source'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zfs_prop_get_recvd' mangled-name='zfs_prop_get_recvd' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_recvd'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='26a90f95' name='propbuf'/>
<parameter type-id='b59d7dce' name='proplen'/>
<parameter type-id='c19b74c3' name='literal'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_get_clones_nvl' mangled-name='zfs_get_clones_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_clones_nvl'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zfs_prop_get' mangled-name='zfs_prop_get' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='58603c44' name='prop'/>
<parameter type-id='26a90f95' name='propbuf'/>
<parameter type-id='b59d7dce' name='proplen'/>
<parameter type-id='debc6aa3' name='src'/>
<parameter type-id='26a90f95' name='statbuf'/>
<parameter type-id='b59d7dce' name='statlen'/>
<parameter type-id='c19b74c3' name='literal'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_get_int' mangled-name='zfs_prop_get_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_int'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='58603c44' name='prop'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zfs_prop_get_numeric' mangled-name='zfs_prop_get_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_numeric'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='58603c44' name='prop'/>
<parameter type-id='5d6479ae' name='value'/>
<parameter type-id='debc6aa3' name='src'/>
<parameter type-id='26a90f95' name='statbuf'/>
<parameter type-id='b59d7dce' name='statlen'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_get_userquota_int' mangled-name='zfs_prop_get_userquota_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_userquota_int'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='5d6479ae' name='propvalue'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_get_userquota' mangled-name='zfs_prop_get_userquota' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_userquota'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='26a90f95' name='propbuf'/>
<parameter type-id='95e97e5e' name='proplen'/>
<parameter type-id='c19b74c3' name='literal'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_get_written_int' mangled-name='zfs_prop_get_written_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_written_int'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='5d6479ae' name='propvalue'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prop_get_written' mangled-name='zfs_prop_get_written' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_written'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='26a90f95' name='propbuf'/>
<parameter type-id='95e97e5e' name='proplen'/>
<parameter type-id='c19b74c3' name='literal'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_get_name' mangled-name='zfs_get_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_name'>
<parameter type-id='fcd57163' name='zhp'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_get_pool_name' mangled-name='zfs_get_pool_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_pool_name'>
<parameter type-id='fcd57163' name='zhp'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_get_type' mangled-name='zfs_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_type'>
<parameter type-id='fcd57163' name='zhp'/>
<return type-id='2e45de5d'/>
</function-decl>
<function-decl name='zfs_get_underlying_type' mangled-name='zfs_get_underlying_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_underlying_type'>
<parameter type-id='fcd57163' name='zhp'/>
<return type-id='2e45de5d'/>
</function-decl>
<function-decl name='zfs_parent_name' mangled-name='zfs_parent_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_parent_name'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='buflen'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_dataset_exists' mangled-name='zfs_dataset_exists' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dataset_exists'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='2e45de5d' name='types'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_create_ancestors' mangled-name='zfs_create_ancestors' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_create_ancestors'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='path'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_create' mangled-name='zfs_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_create'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='2e45de5d' name='type'/>
<parameter type-id='5ce45b60' name='props'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_destroy' mangled-name='zfs_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='c19b74c3' name='defer'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_destroy_snaps' mangled-name='zfs_destroy_snaps' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy_snaps'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='26a90f95' name='snapname'/>
<parameter type-id='c19b74c3' name='defer'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_destroy_snaps_nvl' mangled-name='zfs_destroy_snaps_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy_snaps_nvl'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='5ce45b60' name='snaps'/>
<parameter type-id='c19b74c3' name='defer'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_clone' mangled-name='zfs_clone' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_clone'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='target'/>
<parameter type-id='5ce45b60' name='props'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_promote' mangled-name='zfs_promote' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_promote'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_snapshot_nvl' mangled-name='zfs_snapshot_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_snapshot_nvl'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='5ce45b60' name='snaps'/>
<parameter type-id='5ce45b60' name='props'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_snapshot' mangled-name='zfs_snapshot' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_snapshot'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='c19b74c3' name='recursive'/>
<parameter type-id='5ce45b60' name='props'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_rollback' mangled-name='zfs_rollback' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_rollback'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='9200a744' name='snap'/>
<parameter type-id='c19b74c3' name='force'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_rename' mangled-name='zfs_rename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_rename'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='target'/>
<parameter type-id='067170c2' name='flags'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_get_all_props' mangled-name='zfs_get_all_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_all_props'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zfs_get_recvd_props' mangled-name='zfs_get_recvd_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_recvd_props'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zfs_get_user_props' mangled-name='zfs_get_user_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_user_props'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zfs_expand_proplist' mangled-name='zfs_expand_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_expand_proplist'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='e4378506' name='plp'/>
<parameter type-id='c19b74c3' name='received'/>
<parameter type-id='c19b74c3' name='literal'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_prune_proplist' mangled-name='zfs_prune_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prune_proplist'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='ae3e8ca6' name='props'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_smb_acl_add' mangled-name='zfs_smb_acl_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_add'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='26a90f95' name='dataset'/>
<parameter type-id='26a90f95' name='path'/>
<parameter type-id='26a90f95' name='resource'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_smb_acl_remove' mangled-name='zfs_smb_acl_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_remove'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='26a90f95' name='dataset'/>
<parameter type-id='26a90f95' name='path'/>
<parameter type-id='26a90f95' name='resource'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_smb_acl_purge' mangled-name='zfs_smb_acl_purge' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_purge'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='26a90f95' name='dataset'/>
<parameter type-id='26a90f95' name='path'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_smb_acl_rename' mangled-name='zfs_smb_acl_rename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_rename'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='26a90f95' name='dataset'/>
<parameter type-id='26a90f95' name='path'/>
<parameter type-id='26a90f95' name='oldname'/>
<parameter type-id='26a90f95' name='newname'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_userspace' mangled-name='zfs_userspace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_userspace'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='279fde6a' name='type'/>
<parameter type-id='16c5f410' name='func'/>
<parameter type-id='eaa32e2f' name='arg'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_hold' mangled-name='zfs_hold' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_hold'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='snapname'/>
<parameter type-id='80f4b756' name='tag'/>
<parameter type-id='c19b74c3' name='recursive'/>
<parameter type-id='95e97e5e' name='cleanup_fd'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_hold_nvl' mangled-name='zfs_hold_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_hold_nvl'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='95e97e5e' name='cleanup_fd'/>
<parameter type-id='5ce45b60' name='holds'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_release' mangled-name='zfs_release' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_release'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='snapname'/>
<parameter type-id='80f4b756' name='tag'/>
<parameter type-id='c19b74c3' name='recursive'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_get_fsacl' mangled-name='zfs_get_fsacl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_fsacl'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='857bb57e' name='nvl'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_set_fsacl' mangled-name='zfs_set_fsacl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_set_fsacl'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='c19b74c3' name='un'/>
<parameter type-id='5ce45b60' name='nvl'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_get_holds' mangled-name='zfs_get_holds' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_holds'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='857bb57e' name='nvl'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zvol_volsize_to_reservation' mangled-name='zvol_volsize_to_reservation' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zvol_volsize_to_reservation'>
<parameter type-id='4c81de99' name='zph'/>
<parameter type-id='9c313c2d' name='volsize'/>
<parameter type-id='5ce45b60' name='props'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zfs_wait_status' mangled-name='zfs_wait_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_wait_status'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='3024501a' name='activity'/>
<parameter type-id='37e3bd22' name='missing'/>
<parameter type-id='37e3bd22' name='waited'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-type size-in-bits='64' id='7e291ce6'>
<parameter type-id='eaa32e2f'/>
<parameter type-id='80f4b756'/>
<parameter type-id='354978ed'/>
<parameter type-id='9c313c2d'/>
<return type-id='95e97e5e'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='libzfs_diff.c' language='LANG_C99'>
<function-decl name='zfs_show_diffs' mangled-name='zfs_show_diffs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_show_diffs'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='95e97e5e' name='outfd'/>
<parameter type-id='80f4b756' name='fromsnap'/>
<parameter type-id='80f4b756' name='tosnap'/>
<parameter type-id='95e97e5e' name='flags'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='libzfs_import.c' language='LANG_C99'>
<typedef-decl name='refresh_config_func_t' type-id='29f040d2' id='b7c58eaa'/>
<typedef-decl name='pool_active_func_t' type-id='baa42fef' id='de5d1d8f'/>
<class-decl name='pool_config_ops' size-in-bits='128' is-struct='yes' visibility='default' id='8b092c69'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='pco_refresh_config' type-id='e7c00489' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='pco_pool_active' type-id='9eadf5e0' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='pool_config_ops_t' type-id='1a21babe' id='b1e62775'/>
<enum-decl name='pool_state' id='4871ac24'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='POOL_STATE_ACTIVE' value='0'/>
<enumerator name='POOL_STATE_EXPORTED' value='1'/>
<enumerator name='POOL_STATE_DESTROYED' value='2'/>
<enumerator name='POOL_STATE_SPARE' value='3'/>
<enumerator name='POOL_STATE_L2CACHE' value='4'/>
<enumerator name='POOL_STATE_UNINITIALIZED' value='5'/>
<enumerator name='POOL_STATE_UNAVAIL' value='6'/>
<enumerator name='POOL_STATE_POTENTIALLY_ACTIVE' value='7'/>
</enum-decl>
<typedef-decl name='pool_state_t' type-id='4871ac24' id='084a08a3'/>
<qualified-type-def type-id='8b092c69' const='yes' id='1a21babe'/>
<pointer-type-def type-id='de5d1d8f' size-in-bits='64' id='9eadf5e0'/>
<pointer-type-def type-id='084a08a3' size-in-bits='64' id='b9ea57b8'/>
<pointer-type-def type-id='b7c58eaa' size-in-bits='64' id='e7c00489'/>
<var-decl name='libzfs_config_ops' type-id='b1e62775' mangled-name='libzfs_config_ops' visibility='default' elf-symbol-id='libzfs_config_ops'/>
<function-decl name='zpool_clear_label' mangled-name='zpool_clear_label' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_clear_label'>
<parameter type-id='95e97e5e' name='fd'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_in_use' mangled-name='zpool_in_use' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_in_use'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='95e97e5e' name='fd'/>
<parameter type-id='b9ea57b8' name='state'/>
<parameter type-id='9b23c9ad' name='namestr'/>
<parameter type-id='37e3bd22' name='inuse'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-type size-in-bits='64' id='baa42fef'>
<parameter type-id='eaa32e2f'/>
<parameter type-id='80f4b756'/>
<parameter type-id='9c313c2d'/>
<parameter type-id='37e3bd22'/>
<return type-id='95e97e5e'/>
</function-type>
<function-type size-in-bits='64' id='29f040d2'>
<parameter type-id='eaa32e2f'/>
<parameter type-id='5ce45b60'/>
<return type-id='5ce45b60'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='libzfs_iter.c' language='LANG_C99'>
<function-decl name='zfs_iter_filesystems' mangled-name='zfs_iter_filesystems' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_filesystems'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='d8e49ab9' name='func'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_iter_snapshots' mangled-name='zfs_iter_snapshots' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapshots'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='c19b74c3' name='simple'/>
<parameter type-id='d8e49ab9' name='func'/>
<parameter type-id='eaa32e2f' name='data'/>
<parameter type-id='9c313c2d' name='min_txg'/>
<parameter type-id='9c313c2d' name='max_txg'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_iter_bookmarks' mangled-name='zfs_iter_bookmarks' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_bookmarks'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='d8e49ab9' name='func'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_iter_snapshots_sorted' mangled-name='zfs_iter_snapshots_sorted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapshots_sorted'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='d8e49ab9' name='callback'/>
<parameter type-id='eaa32e2f' name='data'/>
<parameter type-id='9c313c2d' name='min_txg'/>
<parameter type-id='9c313c2d' name='max_txg'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_iter_snapspec' mangled-name='zfs_iter_snapspec' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapspec'>
<parameter type-id='9200a744' name='fs_zhp'/>
<parameter type-id='80f4b756' name='spec_orig'/>
<parameter type-id='d8e49ab9' name='func'/>
<parameter type-id='eaa32e2f' name='arg'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_iter_children' mangled-name='zfs_iter_children' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_children'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='d8e49ab9' name='func'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_iter_dependents' mangled-name='zfs_iter_dependents' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_dependents'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='c19b74c3' name='allowrecursion'/>
<parameter type-id='d8e49ab9' name='func'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_iter_mounted' mangled-name='zfs_iter_mounted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_mounted'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='d8e49ab9' name='func'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='libzfs_mount.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='f1bd64e2' size-in-bits='384' id='b2c36c9f'>
<subrange length='2' type-id='7359adad' id='52efc4ef'/>
</array-type-def>
<class-decl name='get_all_cb' size-in-bits='192' is-struct='yes' visibility='default' id='803dac95'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='cb_handles' type-id='4507922a' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='cb_alloc' type-id='b59d7dce' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='cb_used' type-id='b59d7dce' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='get_all_cb_t' type-id='803dac95' id='9b293607'/>
<class-decl name='proto_table_t' size-in-bits='192' is-struct='yes' naming-typedef-id='f1bd64e2' visibility='default' id='f4c8e1ed'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='p_prop' type-id='58603c44' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='p_name' type-id='26a90f95' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='p_share_err' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='160'>
<var-decl name='p_unshare_err' type-id='95e97e5e' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='proto_table_t' type-id='f4c8e1ed' id='f1bd64e2'/>
<pointer-type-def type-id='9b293607' size-in-bits='64' id='77bf1784'/>
<pointer-type-def type-id='9200a744' size-in-bits='64' id='4507922a'/>
<var-decl name='proto_table' type-id='b2c36c9f' visibility='default'/>
<function-decl name='is_mounted' mangled-name='is_mounted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='is_mounted'>
<parameter type-id='b0382bb3' name='zfs_hdl'/>
<parameter type-id='80f4b756' name='special'/>
<parameter type-id='9b23c9ad' name='where'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_is_mounted' mangled-name='zfs_is_mounted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_is_mounted'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='9b23c9ad' name='where'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_mount' mangled-name='zfs_mount' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mount'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='options'/>
<parameter type-id='95e97e5e' name='flags'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_mount_at' mangled-name='zfs_mount_at' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mount_at'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='options'/>
<parameter type-id='95e97e5e' name='flags'/>
<parameter type-id='80f4b756' name='mountpoint'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_unmount' mangled-name='zfs_unmount' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unmount'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='mountpoint'/>
<parameter type-id='95e97e5e' name='flags'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_unmountall' mangled-name='zfs_unmountall' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unmountall'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='95e97e5e' name='flags'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_is_shared' mangled-name='zfs_is_shared' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_is_shared'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_share' mangled-name='zfs_share' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_share'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_unshare' mangled-name='zfs_unshare' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshare'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_is_shared_nfs' mangled-name='zfs_is_shared_nfs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_is_shared_nfs'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='9b23c9ad' name='where'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_is_shared_smb' mangled-name='zfs_is_shared_smb' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_is_shared_smb'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='9b23c9ad' name='where'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_commit_nfs_shares' mangled-name='zfs_commit_nfs_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_commit_nfs_shares'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_commit_smb_shares' mangled-name='zfs_commit_smb_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_commit_smb_shares'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_commit_all_shares' mangled-name='zfs_commit_all_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_commit_all_shares'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_commit_shares' mangled-name='zfs_commit_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_commit_shares'>
<parameter type-id='80f4b756' name='proto'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_share_nfs' mangled-name='zfs_share_nfs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_share_nfs'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_share_smb' mangled-name='zfs_share_smb' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_share_smb'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
- <function-decl name='zfs_shareall' mangled-name='zfs_shareall' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_shareall'>
- <parameter type-id='9200a744' name='zhp'/>
- <return type-id='95e97e5e'/>
- </function-decl>
<function-decl name='zfs_unshare_nfs' mangled-name='zfs_unshare_nfs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshare_nfs'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='mountpoint'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_unshare_smb' mangled-name='zfs_unshare_smb' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshare_smb'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='mountpoint'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_unshareall_nfs' mangled-name='zfs_unshareall_nfs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshareall_nfs'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_unshareall_smb' mangled-name='zfs_unshareall_smb' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshareall_smb'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_unshareall' mangled-name='zfs_unshareall' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshareall'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_unshareall_bypath' mangled-name='zfs_unshareall_bypath' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshareall_bypath'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='mountpoint'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_unshareall_bytype' mangled-name='zfs_unshareall_bytype' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshareall_bytype'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='mountpoint'/>
<parameter type-id='80f4b756' name='proto'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='libzfs_add_handle' mangled-name='libzfs_add_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_add_handle'>
<parameter type-id='77bf1784' name='cbp'/>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_foreach_mountpoint' mangled-name='zfs_foreach_mountpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_foreach_mountpoint'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='4507922a' name='handles'/>
<parameter type-id='b59d7dce' name='num_handles'/>
<parameter type-id='d8e49ab9' name='func'/>
<parameter type-id='eaa32e2f' name='data'/>
<parameter type-id='c19b74c3' name='parallel'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_enable_datasets' mangled-name='zpool_enable_datasets' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_enable_datasets'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='mntopts'/>
<parameter type-id='95e97e5e' name='flags'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_disable_datasets' mangled-name='zpool_disable_datasets' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disable_datasets'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='c19b74c3' name='force'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='libzfs_pool.c' language='LANG_C99'>
<class-decl name='splitflags' size-in-bits='64' is-struct='yes' visibility='default' id='dc01bf52'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='dryrun' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='1'>
<var-decl name='import' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='name_flags' type-id='95e97e5e' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='splitflags_t' type-id='dc01bf52' id='325c1e34'/>
<class-decl name='trimflags' size-in-bits='192' is-struct='yes' visibility='default' id='8ef58008'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='fullpool' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='secure' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='wait' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='rate' type-id='9c313c2d' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='trimflags_t' type-id='8ef58008' id='a093cbb8'/>
<enum-decl name='zpool_compat_status_t' naming-typedef-id='901b78d1' id='20676925'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZPOOL_COMPATIBILITY_OK' value='0'/>
<enumerator name='ZPOOL_COMPATIBILITY_WARNTOKEN' value='1'/>
<enumerator name='ZPOOL_COMPATIBILITY_BADTOKEN' value='2'/>
<enumerator name='ZPOOL_COMPATIBILITY_BADFILE' value='3'/>
<enumerator name='ZPOOL_COMPATIBILITY_NOFILES' value='4'/>
</enum-decl>
<typedef-decl name='zpool_compat_status_t' type-id='20676925' id='901b78d1'/>
<enum-decl name='zpool_prop_t' naming-typedef-id='5d0c23fb' id='af1ba157'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZPOOL_PROP_INVAL' value='-1'/>
<enumerator name='ZPOOL_PROP_NAME' value='0'/>
<enumerator name='ZPOOL_PROP_SIZE' value='1'/>
<enumerator name='ZPOOL_PROP_CAPACITY' value='2'/>
<enumerator name='ZPOOL_PROP_ALTROOT' value='3'/>
<enumerator name='ZPOOL_PROP_HEALTH' value='4'/>
<enumerator name='ZPOOL_PROP_GUID' value='5'/>
<enumerator name='ZPOOL_PROP_VERSION' value='6'/>
<enumerator name='ZPOOL_PROP_BOOTFS' value='7'/>
<enumerator name='ZPOOL_PROP_DELEGATION' value='8'/>
<enumerator name='ZPOOL_PROP_AUTOREPLACE' value='9'/>
<enumerator name='ZPOOL_PROP_CACHEFILE' value='10'/>
<enumerator name='ZPOOL_PROP_FAILUREMODE' value='11'/>
<enumerator name='ZPOOL_PROP_LISTSNAPS' value='12'/>
<enumerator name='ZPOOL_PROP_AUTOEXPAND' value='13'/>
<enumerator name='ZPOOL_PROP_DEDUPDITTO' value='14'/>
<enumerator name='ZPOOL_PROP_DEDUPRATIO' value='15'/>
<enumerator name='ZPOOL_PROP_FREE' value='16'/>
<enumerator name='ZPOOL_PROP_ALLOCATED' value='17'/>
<enumerator name='ZPOOL_PROP_READONLY' value='18'/>
<enumerator name='ZPOOL_PROP_ASHIFT' value='19'/>
<enumerator name='ZPOOL_PROP_COMMENT' value='20'/>
<enumerator name='ZPOOL_PROP_EXPANDSZ' value='21'/>
<enumerator name='ZPOOL_PROP_FREEING' value='22'/>
<enumerator name='ZPOOL_PROP_FRAGMENTATION' value='23'/>
<enumerator name='ZPOOL_PROP_LEAKED' value='24'/>
<enumerator name='ZPOOL_PROP_MAXBLOCKSIZE' value='25'/>
<enumerator name='ZPOOL_PROP_TNAME' value='26'/>
<enumerator name='ZPOOL_PROP_MAXDNODESIZE' value='27'/>
<enumerator name='ZPOOL_PROP_MULTIHOST' value='28'/>
<enumerator name='ZPOOL_PROP_CHECKPOINT' value='29'/>
<enumerator name='ZPOOL_PROP_LOAD_GUID' value='30'/>
<enumerator name='ZPOOL_PROP_AUTOTRIM' value='31'/>
<enumerator name='ZPOOL_PROP_COMPATIBILITY' value='32'/>
<enumerator name='ZPOOL_NUM_PROPS' value='33'/>
</enum-decl>
<typedef-decl name='zpool_prop_t' type-id='af1ba157' id='5d0c23fb'/>
<enum-decl name='vdev_prop_t' naming-typedef-id='5aa5c90c' id='1573bec8'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='VDEV_PROP_INVAL' value='-1'/>
<enumerator name='VDEV_PROP_NAME' value='0'/>
<enumerator name='VDEV_PROP_CAPACITY' value='1'/>
<enumerator name='VDEV_PROP_STATE' value='2'/>
<enumerator name='VDEV_PROP_GUID' value='3'/>
<enumerator name='VDEV_PROP_ASIZE' value='4'/>
<enumerator name='VDEV_PROP_PSIZE' value='5'/>
<enumerator name='VDEV_PROP_ASHIFT' value='6'/>
<enumerator name='VDEV_PROP_SIZE' value='7'/>
<enumerator name='VDEV_PROP_FREE' value='8'/>
<enumerator name='VDEV_PROP_ALLOCATED' value='9'/>
<enumerator name='VDEV_PROP_COMMENT' value='10'/>
<enumerator name='VDEV_PROP_EXPANDSZ' value='11'/>
<enumerator name='VDEV_PROP_FRAGMENTATION' value='12'/>
<enumerator name='VDEV_PROP_BOOTSIZE' value='13'/>
<enumerator name='VDEV_PROP_PARITY' value='14'/>
<enumerator name='VDEV_PROP_PATH' value='15'/>
<enumerator name='VDEV_PROP_DEVID' value='16'/>
<enumerator name='VDEV_PROP_PHYS_PATH' value='17'/>
<enumerator name='VDEV_PROP_ENC_PATH' value='18'/>
<enumerator name='VDEV_PROP_FRU' value='19'/>
<enumerator name='VDEV_PROP_PARENT' value='20'/>
<enumerator name='VDEV_PROP_CHILDREN' value='21'/>
<enumerator name='VDEV_PROP_NUMCHILDREN' value='22'/>
<enumerator name='VDEV_PROP_READ_ERRORS' value='23'/>
<enumerator name='VDEV_PROP_WRITE_ERRORS' value='24'/>
<enumerator name='VDEV_PROP_CHECKSUM_ERRORS' value='25'/>
<enumerator name='VDEV_PROP_INITIALIZE_ERRORS' value='26'/>
<enumerator name='VDEV_PROP_OPS_NULL' value='27'/>
<enumerator name='VDEV_PROP_OPS_READ' value='28'/>
<enumerator name='VDEV_PROP_OPS_WRITE' value='29'/>
<enumerator name='VDEV_PROP_OPS_FREE' value='30'/>
<enumerator name='VDEV_PROP_OPS_CLAIM' value='31'/>
<enumerator name='VDEV_PROP_OPS_TRIM' value='32'/>
<enumerator name='VDEV_PROP_BYTES_NULL' value='33'/>
<enumerator name='VDEV_PROP_BYTES_READ' value='34'/>
<enumerator name='VDEV_PROP_BYTES_WRITE' value='35'/>
<enumerator name='VDEV_PROP_BYTES_FREE' value='36'/>
<enumerator name='VDEV_PROP_BYTES_CLAIM' value='37'/>
<enumerator name='VDEV_PROP_BYTES_TRIM' value='38'/>
<enumerator name='VDEV_PROP_REMOVING' value='39'/>
<enumerator name='VDEV_PROP_ALLOCATING' value='40'/>
<enumerator name='VDEV_NUM_PROPS' value='41'/>
</enum-decl>
<typedef-decl name='vdev_prop_t' type-id='1573bec8' id='5aa5c90c'/>
<enum-decl name='vdev_state' id='21566197'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='VDEV_STATE_UNKNOWN' value='0'/>
<enumerator name='VDEV_STATE_CLOSED' value='1'/>
<enumerator name='VDEV_STATE_OFFLINE' value='2'/>
<enumerator name='VDEV_STATE_REMOVED' value='3'/>
<enumerator name='VDEV_STATE_CANT_OPEN' value='4'/>
<enumerator name='VDEV_STATE_FAULTED' value='5'/>
<enumerator name='VDEV_STATE_DEGRADED' value='6'/>
<enumerator name='VDEV_STATE_HEALTHY' value='7'/>
</enum-decl>
<typedef-decl name='vdev_state_t' type-id='21566197' id='35acf840'/>
<enum-decl name='vdev_aux' id='7f5bcca4'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='VDEV_AUX_NONE' value='0'/>
<enumerator name='VDEV_AUX_OPEN_FAILED' value='1'/>
<enumerator name='VDEV_AUX_CORRUPT_DATA' value='2'/>
<enumerator name='VDEV_AUX_NO_REPLICAS' value='3'/>
<enumerator name='VDEV_AUX_BAD_GUID_SUM' value='4'/>
<enumerator name='VDEV_AUX_TOO_SMALL' value='5'/>
<enumerator name='VDEV_AUX_BAD_LABEL' value='6'/>
<enumerator name='VDEV_AUX_VERSION_NEWER' value='7'/>
<enumerator name='VDEV_AUX_VERSION_OLDER' value='8'/>
<enumerator name='VDEV_AUX_UNSUP_FEAT' value='9'/>
<enumerator name='VDEV_AUX_SPARED' value='10'/>
<enumerator name='VDEV_AUX_ERR_EXCEEDED' value='11'/>
<enumerator name='VDEV_AUX_IO_FAILURE' value='12'/>
<enumerator name='VDEV_AUX_BAD_LOG' value='13'/>
<enumerator name='VDEV_AUX_EXTERNAL' value='14'/>
<enumerator name='VDEV_AUX_SPLIT_POOL' value='15'/>
<enumerator name='VDEV_AUX_BAD_ASHIFT' value='16'/>
<enumerator name='VDEV_AUX_EXTERNAL_PERSIST' value='17'/>
<enumerator name='VDEV_AUX_ACTIVE' value='18'/>
<enumerator name='VDEV_AUX_CHILDREN_OFFLINE' value='19'/>
<enumerator name='VDEV_AUX_ASHIFT_TOO_BIG' value='20'/>
</enum-decl>
<typedef-decl name='vdev_aux_t' type-id='7f5bcca4' id='9d774e0b'/>
<enum-decl name='pool_scan_func' id='1b092565'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='POOL_SCAN_NONE' value='0'/>
<enumerator name='POOL_SCAN_SCRUB' value='1'/>
<enumerator name='POOL_SCAN_RESILVER' value='2'/>
<enumerator name='POOL_SCAN_FUNCS' value='3'/>
</enum-decl>
<typedef-decl name='pool_scan_func_t' type-id='1b092565' id='7313fbe2'/>
<enum-decl name='pool_scrub_cmd' id='a1474cbd'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='POOL_SCRUB_NORMAL' value='0'/>
<enumerator name='POOL_SCRUB_PAUSE' value='1'/>
<enumerator name='POOL_SCRUB_FLAGS_END' value='2'/>
</enum-decl>
<typedef-decl name='pool_scrub_cmd_t' type-id='a1474cbd' id='b51cf3c2'/>
<enum-decl name='pool_initialize_func' id='5c246ad4'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='POOL_INITIALIZE_START' value='0'/>
<enumerator name='POOL_INITIALIZE_CANCEL' value='1'/>
<enumerator name='POOL_INITIALIZE_SUSPEND' value='2'/>
<enumerator name='POOL_INITIALIZE_FUNCS' value='3'/>
</enum-decl>
<typedef-decl name='pool_initialize_func_t' type-id='5c246ad4' id='7063e1ab'/>
<enum-decl name='pool_trim_func' id='54ed608a'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='POOL_TRIM_START' value='0'/>
<enumerator name='POOL_TRIM_CANCEL' value='1'/>
<enumerator name='POOL_TRIM_SUSPEND' value='2'/>
<enumerator name='POOL_TRIM_FUNCS' value='3'/>
</enum-decl>
<typedef-decl name='pool_trim_func_t' type-id='54ed608a' id='b1146b8d'/>
<enum-decl name='zpool_wait_activity_t' naming-typedef-id='73446457' id='849338e3'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZPOOL_WAIT_CKPT_DISCARD' value='0'/>
<enumerator name='ZPOOL_WAIT_FREE' value='1'/>
<enumerator name='ZPOOL_WAIT_INITIALIZE' value='2'/>
<enumerator name='ZPOOL_WAIT_REPLACE' value='3'/>
<enumerator name='ZPOOL_WAIT_REMOVE' value='4'/>
<enumerator name='ZPOOL_WAIT_RESILVER' value='5'/>
<enumerator name='ZPOOL_WAIT_SCRUB' value='6'/>
<enumerator name='ZPOOL_WAIT_TRIM' value='7'/>
<enumerator name='ZPOOL_WAIT_NUM_ACTIVITIES' value='8'/>
</enum-decl>
<typedef-decl name='zpool_wait_activity_t' type-id='849338e3' id='73446457'/>
<qualified-type-def type-id='8e8d4be3' const='yes' id='693c3853'/>
<pointer-type-def type-id='693c3853' size-in-bits='64' id='22cce67b'/>
<pointer-type-def type-id='a093cbb8' size-in-bits='64' id='b13f38c3'/>
<pointer-type-def type-id='35acf840' size-in-bits='64' id='17f3480d'/>
<function-decl name='zpool_props_refresh' mangled-name='zpool_props_refresh' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_props_refresh'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_get_prop_int' mangled-name='zpool_get_prop_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_prop_int'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='5d0c23fb' name='prop'/>
<parameter type-id='debc6aa3' name='src'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zpool_state_to_name' mangled-name='zpool_state_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_state_to_name'>
<parameter type-id='35acf840' name='state'/>
<parameter type-id='9d774e0b' name='aux'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zpool_pool_state_to_name' mangled-name='zpool_pool_state_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_pool_state_to_name'>
<parameter type-id='084a08a3' name='state'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zpool_get_state_str' mangled-name='zpool_get_state_str' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_state_str'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zpool_get_prop' mangled-name='zpool_get_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_prop'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='5d0c23fb' name='prop'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='len'/>
<parameter type-id='debc6aa3' name='srctype'/>
<parameter type-id='c19b74c3' name='literal'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_set_prop' mangled-name='zpool_set_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_set_prop'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='80f4b756' name='propval'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_expand_proplist' mangled-name='zpool_expand_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_expand_proplist'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='e4378506' name='plp'/>
<parameter type-id='2e45de5d' name='type'/>
<parameter type-id='c19b74c3' name='literal'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='vdev_expand_proplist' mangled-name='vdev_expand_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_expand_proplist'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='vdevname'/>
<parameter type-id='e4378506' name='plp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_prop_get_feature' mangled-name='zpool_prop_get_feature' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_get_feature'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='len'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_open_canfail' mangled-name='zpool_open_canfail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_open_canfail'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='pool'/>
<return type-id='4c81de99'/>
</function-decl>
<function-decl name='zpool_open' mangled-name='zpool_open' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_open'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='pool'/>
<return type-id='4c81de99'/>
</function-decl>
<function-decl name='zpool_close' mangled-name='zpool_close' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_close'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_get_name' mangled-name='zpool_get_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_name'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zpool_get_state' mangled-name='zpool_get_state' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_state'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_is_draid_spare' mangled-name='zpool_is_draid_spare' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_is_draid_spare'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zpool_create' mangled-name='zpool_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_create'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='pool'/>
<parameter type-id='5ce45b60' name='nvroot'/>
<parameter type-id='5ce45b60' name='props'/>
<parameter type-id='5ce45b60' name='fsprops'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_destroy' mangled-name='zpool_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_destroy'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='log_str'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_checkpoint' mangled-name='zpool_checkpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_checkpoint'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_discard_checkpoint' mangled-name='zpool_discard_checkpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_discard_checkpoint'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_add' mangled-name='zpool_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_add'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='5ce45b60' name='nvroot'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_export' mangled-name='zpool_export' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_export'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='c19b74c3' name='force'/>
<parameter type-id='80f4b756' name='log_str'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_export_force' mangled-name='zpool_export_force' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_export_force'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='log_str'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_explain_recover' mangled-name='zpool_explain_recover' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_explain_recover'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='95e97e5e' name='reason'/>
<parameter type-id='5ce45b60' name='config'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_import' mangled-name='zpool_import' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_import'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='5ce45b60' name='config'/>
<parameter type-id='80f4b756' name='newname'/>
<parameter type-id='26a90f95' name='altroot'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_print_unsup_feat' mangled-name='zpool_print_unsup_feat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_print_unsup_feat'>
<parameter type-id='5ce45b60' name='config'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_import_props' mangled-name='zpool_import_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_import_props'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='5ce45b60' name='config'/>
<parameter type-id='80f4b756' name='newname'/>
<parameter type-id='5ce45b60' name='props'/>
<parameter type-id='95e97e5e' name='flags'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_initialize' mangled-name='zpool_initialize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_initialize'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='7063e1ab' name='cmd_type'/>
<parameter type-id='5ce45b60' name='vds'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_initialize_wait' mangled-name='zpool_initialize_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_initialize_wait'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='7063e1ab' name='cmd_type'/>
<parameter type-id='5ce45b60' name='vds'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_trim' mangled-name='zpool_trim' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_trim'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='b1146b8d' name='cmd_type'/>
<parameter type-id='5ce45b60' name='vds'/>
<parameter type-id='b13f38c3' name='trim_flags'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_scan' mangled-name='zpool_scan' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_scan'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='7313fbe2' name='func'/>
<parameter type-id='b51cf3c2' name='cmd'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_find_vdev_by_physpath' mangled-name='zpool_find_vdev_by_physpath' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_find_vdev_by_physpath'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='ppath'/>
<parameter type-id='37e3bd22' name='avail_spare'/>
<parameter type-id='37e3bd22' name='l2cache'/>
<parameter type-id='37e3bd22' name='log'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zpool_find_vdev' mangled-name='zpool_find_vdev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_find_vdev'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='37e3bd22' name='avail_spare'/>
<parameter type-id='37e3bd22' name='l2cache'/>
<parameter type-id='37e3bd22' name='log'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zpool_get_physpath' mangled-name='zpool_get_physpath' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_physpath'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='26a90f95' name='physpath'/>
<parameter type-id='b59d7dce' name='phypath_size'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_path_to_guid' mangled-name='zpool_vdev_path_to_guid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_path_to_guid'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='path'/>
<return type-id='9c313c2d'/>
</function-decl>
<function-decl name='zpool_vdev_online' mangled-name='zpool_vdev_online' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_online'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='95e97e5e' name='flags'/>
<parameter type-id='17f3480d' name='newstate'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_offline' mangled-name='zpool_vdev_offline' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_offline'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='c19b74c3' name='istmp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_fault' mangled-name='zpool_vdev_fault' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_fault'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='9c313c2d' name='guid'/>
<parameter type-id='9d774e0b' name='aux'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_degrade' mangled-name='zpool_vdev_degrade' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_degrade'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='9c313c2d' name='guid'/>
<parameter type-id='9d774e0b' name='aux'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_attach' mangled-name='zpool_vdev_attach' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_attach'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='old_disk'/>
<parameter type-id='80f4b756' name='new_disk'/>
<parameter type-id='5ce45b60' name='nvroot'/>
<parameter type-id='95e97e5e' name='replacing'/>
<parameter type-id='c19b74c3' name='rebuild'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_detach' mangled-name='zpool_vdev_detach' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_detach'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='path'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_split' mangled-name='zpool_vdev_split' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_split'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='26a90f95' name='newname'/>
<parameter type-id='857bb57e' name='newroot'/>
<parameter type-id='5ce45b60' name='props'/>
<parameter type-id='325c1e34' name='flags'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_remove' mangled-name='zpool_vdev_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_remove'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='path'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_remove_cancel' mangled-name='zpool_vdev_remove_cancel' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_remove_cancel'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_indirect_size' mangled-name='zpool_vdev_indirect_size' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_indirect_size'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='5d6479ae' name='sizep'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_clear' mangled-name='zpool_clear' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_clear'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='5ce45b60' name='rewindnvl'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_clear' mangled-name='zpool_vdev_clear' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_clear'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='9c313c2d' name='guid'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_reguid' mangled-name='zpool_reguid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_reguid'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_reopen_one' mangled-name='zpool_reopen_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_reopen_one'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_sync_one' mangled-name='zpool_sync_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_sync_one'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='eaa32e2f' name='data'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_vdev_name' mangled-name='zpool_vdev_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_name'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='5ce45b60' name='nv'/>
<parameter type-id='95e97e5e' name='name_flags'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='zpool_get_errlog' mangled-name='zpool_get_errlog' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_errlog'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='857bb57e' name='nverrlistp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_upgrade' mangled-name='zpool_upgrade' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_upgrade'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='9c313c2d' name='new_version'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_save_arguments' mangled-name='zfs_save_arguments' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_save_arguments'>
<parameter type-id='95e97e5e' name='argc'/>
<parameter type-id='9b23c9ad' name='argv'/>
<parameter type-id='26a90f95' name='string'/>
<parameter type-id='95e97e5e' name='len'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_log_history' mangled-name='zpool_log_history' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_log_history'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='message'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_get_history' mangled-name='zpool_get_history' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_history'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='857bb57e' name='nvhisp'/>
<parameter type-id='5d6479ae' name='off'/>
<parameter type-id='37e3bd22' name='eof'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_events_next' mangled-name='zpool_events_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_events_next'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='857bb57e' name='nvp'/>
<parameter type-id='7292109c' name='dropped'/>
<parameter type-id='f0981eeb' name='flags'/>
<parameter type-id='95e97e5e' name='zevent_fd'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_events_clear' mangled-name='zpool_events_clear' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_events_clear'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='7292109c' name='count'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_events_seek' mangled-name='zpool_events_seek' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_events_seek'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='9c313c2d' name='eid'/>
<parameter type-id='95e97e5e' name='zevent_fd'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_obj_to_path' mangled-name='zpool_obj_to_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_obj_to_path'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='9c313c2d' name='dsobj'/>
<parameter type-id='9c313c2d' name='obj'/>
<parameter type-id='26a90f95' name='pathname'/>
<parameter type-id='b59d7dce' name='len'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_obj_to_path_ds' mangled-name='zpool_obj_to_path_ds' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_obj_to_path_ds'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='9c313c2d' name='dsobj'/>
<parameter type-id='9c313c2d' name='obj'/>
<parameter type-id='26a90f95' name='pathname'/>
<parameter type-id='b59d7dce' name='len'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_wait' mangled-name='zpool_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_wait'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='73446457' name='activity'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_wait_status' mangled-name='zpool_wait_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_wait_status'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='73446457' name='activity'/>
<parameter type-id='37e3bd22' name='missing'/>
<parameter type-id='37e3bd22' name='waited'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_set_bootenv' mangled-name='zpool_set_bootenv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_set_bootenv'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='22cce67b' name='envmap'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_get_bootenv' mangled-name='zpool_get_bootenv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_bootenv'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='857bb57e' name='nvlp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_load_compat' mangled-name='zpool_load_compat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_load_compat'>
<parameter type-id='80f4b756' name='compat'/>
<parameter type-id='37e3bd22' name='features'/>
<parameter type-id='26a90f95' name='report'/>
<parameter type-id='b59d7dce' name='rlen'/>
<return type-id='901b78d1'/>
</function-decl>
<function-decl name='zpool_get_vdev_prop_value' mangled-name='zpool_get_vdev_prop_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_vdev_prop_value'>
<parameter type-id='5ce45b60' name='nvprop'/>
<parameter type-id='5aa5c90c' name='prop'/>
<parameter type-id='26a90f95' name='prop_name'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='len'/>
<parameter type-id='debc6aa3' name='srctype'/>
<parameter type-id='c19b74c3' name='literal'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_get_vdev_prop' mangled-name='zpool_get_vdev_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_vdev_prop'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='vdevname'/>
<parameter type-id='5aa5c90c' name='prop'/>
<parameter type-id='26a90f95' name='prop_name'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='len'/>
<parameter type-id='debc6aa3' name='srctype'/>
<parameter type-id='c19b74c3' name='literal'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_get_all_vdev_props' mangled-name='zpool_get_all_vdev_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_all_vdev_props'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='vdevname'/>
<parameter type-id='857bb57e' name='outnvl'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_set_vdev_prop' mangled-name='zpool_set_vdev_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_set_vdev_prop'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='vdevname'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='80f4b756' name='propval'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='libzfs_sendrecv.c' language='LANG_C99'>
<class-decl name='sendflags' size-in-bits='544' is-struct='yes' visibility='default' id='f6aa15be'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='verbosity' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='replicate' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='skipmissing' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='96'>
<var-decl name='doall' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='fromorigin' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='160'>
<var-decl name='pad' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='props' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='224'>
<var-decl name='dryrun' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='parsable' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='288'>
<var-decl name='progress' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='largeblock' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='352'>
<var-decl name='embed_data' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='compress' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='416'>
<var-decl name='raw' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='backup' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='480'>
<var-decl name='holds' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='512'>
<var-decl name='saved' type-id='c19b74c3' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='sendflags_t' type-id='f6aa15be' id='945467e6'/>
<typedef-decl name='snapfilter_cb_t' type-id='d2a5e211' id='3d3ffb69'/>
<class-decl name='recvflags' size-in-bits='416' is-struct='yes' visibility='default' id='34a384dc'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='verbose' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='isprefix' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='istail' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='96'>
<var-decl name='dryrun' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='force' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='160'>
<var-decl name='canmountoff' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='resumable' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='224'>
<var-decl name='byteswap' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='nomount' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='288'>
<var-decl name='holds' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='skipholds' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='352'>
<var-decl name='domount' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='forceunmount' type-id='c19b74c3' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='recvflags_t' type-id='34a384dc' id='9e59d1d4'/>
<pointer-type-def type-id='f20fbd51' size-in-bits='64' id='a3681dea'/>
<pointer-type-def type-id='9e59d1d4' size-in-bits='64' id='4ea84b4f'/>
<pointer-type-def type-id='945467e6' size-in-bits='64' id='8def7735'/>
<pointer-type-def type-id='3d3ffb69' size-in-bits='64' id='72a26210'/>
<function-decl name='zfs_send_progress' mangled-name='zfs_send_progress' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_progress'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='95e97e5e' name='fd'/>
<parameter type-id='5d6479ae' name='bytes_written'/>
<parameter type-id='5d6479ae' name='blocks_visited'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_send_resume_token_to_nvlist' mangled-name='zfs_send_resume_token_to_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_resume_token_to_nvlist'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='token'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zfs_send_resume' mangled-name='zfs_send_resume' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_resume'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='8def7735' name='flags'/>
<parameter type-id='95e97e5e' name='outfd'/>
<parameter type-id='80f4b756' name='resume_token'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_send_saved' mangled-name='zfs_send_saved' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_saved'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='8def7735' name='flags'/>
<parameter type-id='95e97e5e' name='outfd'/>
<parameter type-id='80f4b756' name='resume_token'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_send' mangled-name='zfs_send' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='fromsnap'/>
<parameter type-id='80f4b756' name='tosnap'/>
<parameter type-id='8def7735' name='flags'/>
<parameter type-id='95e97e5e' name='outfd'/>
<parameter type-id='72a26210' name='filter_func'/>
<parameter type-id='eaa32e2f' name='cb_arg'/>
<parameter type-id='857bb57e' name='debugnvp'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_send_one' mangled-name='zfs_send_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_one'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='from'/>
<parameter type-id='95e97e5e' name='fd'/>
<parameter type-id='8def7735' name='flags'/>
<parameter type-id='80f4b756' name='redactbook'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_receive' mangled-name='zfs_receive' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_receive'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='tosnap'/>
<parameter type-id='5ce45b60' name='props'/>
<parameter type-id='4ea84b4f' name='flags'/>
<parameter type-id='95e97e5e' name='infd'/>
<parameter type-id='a3681dea' name='stream_avl'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-type size-in-bits='64' id='d2a5e211'>
<parameter type-id='9200a744'/>
<parameter type-id='eaa32e2f'/>
<return type-id='c19b74c3'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='libzfs_status.c' language='LANG_C99'>
<enum-decl name='zpool_status_t' naming-typedef-id='d3dd6294' id='5e770b40'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZPOOL_STATUS_CORRUPT_CACHE' value='0'/>
<enumerator name='ZPOOL_STATUS_MISSING_DEV_R' value='1'/>
<enumerator name='ZPOOL_STATUS_MISSING_DEV_NR' value='2'/>
<enumerator name='ZPOOL_STATUS_CORRUPT_LABEL_R' value='3'/>
<enumerator name='ZPOOL_STATUS_CORRUPT_LABEL_NR' value='4'/>
<enumerator name='ZPOOL_STATUS_BAD_GUID_SUM' value='5'/>
<enumerator name='ZPOOL_STATUS_CORRUPT_POOL' value='6'/>
<enumerator name='ZPOOL_STATUS_CORRUPT_DATA' value='7'/>
<enumerator name='ZPOOL_STATUS_FAILING_DEV' value='8'/>
<enumerator name='ZPOOL_STATUS_VERSION_NEWER' value='9'/>
<enumerator name='ZPOOL_STATUS_HOSTID_MISMATCH' value='10'/>
<enumerator name='ZPOOL_STATUS_HOSTID_ACTIVE' value='11'/>
<enumerator name='ZPOOL_STATUS_HOSTID_REQUIRED' value='12'/>
<enumerator name='ZPOOL_STATUS_IO_FAILURE_WAIT' value='13'/>
<enumerator name='ZPOOL_STATUS_IO_FAILURE_CONTINUE' value='14'/>
<enumerator name='ZPOOL_STATUS_IO_FAILURE_MMP' value='15'/>
<enumerator name='ZPOOL_STATUS_BAD_LOG' value='16'/>
<enumerator name='ZPOOL_STATUS_ERRATA' value='17'/>
<enumerator name='ZPOOL_STATUS_UNSUP_FEAT_READ' value='18'/>
<enumerator name='ZPOOL_STATUS_UNSUP_FEAT_WRITE' value='19'/>
<enumerator name='ZPOOL_STATUS_FAULTED_DEV_R' value='20'/>
<enumerator name='ZPOOL_STATUS_FAULTED_DEV_NR' value='21'/>
<enumerator name='ZPOOL_STATUS_VERSION_OLDER' value='22'/>
<enumerator name='ZPOOL_STATUS_FEAT_DISABLED' value='23'/>
<enumerator name='ZPOOL_STATUS_RESILVERING' value='24'/>
<enumerator name='ZPOOL_STATUS_OFFLINE_DEV' value='25'/>
<enumerator name='ZPOOL_STATUS_REMOVED_DEV' value='26'/>
<enumerator name='ZPOOL_STATUS_REBUILDING' value='27'/>
<enumerator name='ZPOOL_STATUS_REBUILD_SCRUB' value='28'/>
<enumerator name='ZPOOL_STATUS_NON_NATIVE_ASHIFT' value='29'/>
<enumerator name='ZPOOL_STATUS_COMPATIBILITY_ERR' value='30'/>
<enumerator name='ZPOOL_STATUS_INCOMPATIBLE_FEAT' value='31'/>
<enumerator name='ZPOOL_STATUS_OK' value='32'/>
</enum-decl>
<typedef-decl name='zpool_status_t' type-id='5e770b40' id='d3dd6294'/>
<enum-decl name='zpool_errata' id='d9abbf54'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZPOOL_ERRATA_NONE' value='0'/>
<enumerator name='ZPOOL_ERRATA_ZOL_2094_SCRUB' value='1'/>
<enumerator name='ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY' value='2'/>
<enumerator name='ZPOOL_ERRATA_ZOL_6845_ENCRYPTION' value='3'/>
<enumerator name='ZPOOL_ERRATA_ZOL_8308_ENCRYPTION' value='4'/>
</enum-decl>
<typedef-decl name='zpool_errata_t' type-id='d9abbf54' id='688c495b'/>
<pointer-type-def type-id='688c495b' size-in-bits='64' id='cec6f2e4'/>
<function-decl name='zpool_get_status' mangled-name='zpool_get_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_status'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='9b23c9ad' name='msgid'/>
<parameter type-id='cec6f2e4' name='errata'/>
<return type-id='d3dd6294'/>
</function-decl>
<function-decl name='zpool_import_status' mangled-name='zpool_import_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_import_status'>
<parameter type-id='5ce45b60' name='config'/>
<parameter type-id='9b23c9ad' name='msgid'/>
<parameter type-id='cec6f2e4' name='errata'/>
<return type-id='d3dd6294'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='libzfs_util.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='95e97e5e' size-in-bits='192' id='e41bdf22'>
<subrange length='6' type-id='7359adad' id='52fa524b'/>
</array-type-def>
<type-decl name='variadic parameter type' id='2c1145c5'/>
<array-type-def dimensions='1' type-id='19cefcee' size-in-bits='160' alignment-in-bits='32' id='3fcf57d2'>
<subrange length='5' type-id='7359adad' id='53010e10'/>
</array-type-def>
<enum-decl name='zfs_get_column_t' naming-typedef-id='19cefcee' id='223bdcaa'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='GET_COL_NONE' value='0'/>
<enumerator name='GET_COL_NAME' value='1'/>
<enumerator name='GET_COL_PROPERTY' value='2'/>
<enumerator name='GET_COL_VALUE' value='3'/>
<enumerator name='GET_COL_RECVD' value='4'/>
<enumerator name='GET_COL_SOURCE' value='5'/>
</enum-decl>
<typedef-decl name='zfs_get_column_t' type-id='223bdcaa' id='19cefcee'/>
<class-decl name='vdev_cbdata' size-in-bits='192' is-struct='yes' visibility='default' id='b8006be8'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='cb_name_flags' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='cb_names' type-id='9b23c9ad' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='cb_names_count' type-id='f0981eeb' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='vdev_cbdata_t' type-id='b8006be8' id='a9679c94'/>
<class-decl name='zprop_get_cbdata' size-in-bits='832' is-struct='yes' visibility='default' id='f3d3c319'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='cb_sources' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32'>
<var-decl name='cb_columns' type-id='3fcf57d2' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='cb_colwidths' type-id='e41bdf22' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='cb_scripted' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='416'>
<var-decl name='cb_literal' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='cb_first' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='512'>
<var-decl name='cb_proplist' type-id='3a9b2288' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='576'>
<var-decl name='cb_type' type-id='2e45de5d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='640'>
<var-decl name='cb_vdevs' type-id='a9679c94' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zprop_get_cbdata_t' type-id='f3d3c319' id='f3d87113'/>
<typedef-decl name='zprop_func' type-id='2e711a2a' id='1ec3747a'/>
<pointer-type-def type-id='9b23c9ad' size-in-bits='64' id='c0563f85'/>
<pointer-type-def type-id='c70fa2e8' size-in-bits='64' id='2e711a2a'/>
<pointer-type-def type-id='f3d87113' size-in-bits='64' id='0d2a0670'/>
<function-decl name='libzfs_errno' mangled-name='libzfs_errno' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_errno'>
<parameter type-id='b0382bb3' name='hdl'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='libzfs_error_action' mangled-name='libzfs_error_action' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_error_action'>
<parameter type-id='b0382bb3' name='hdl'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='libzfs_error_description' mangled-name='libzfs_error_description' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_error_description'>
<parameter type-id='b0382bb3' name='hdl'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_standard_error' mangled-name='zfs_standard_error' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_standard_error'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='95e97e5e' name='error'/>
<parameter type-id='80f4b756' name='msg'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='libzfs_print_on_error' mangled-name='libzfs_print_on_error' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_print_on_error'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='c19b74c3' name='printerr'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='libzfs_run_process' mangled-name='libzfs_run_process' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_run_process'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='9b23c9ad' name='argv'/>
<parameter type-id='95e97e5e' name='flags'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='libzfs_run_process_get_stdout' mangled-name='libzfs_run_process_get_stdout' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_run_process_get_stdout'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='9b23c9ad' name='argv'/>
<parameter type-id='9b23c9ad' name='env'/>
<parameter type-id='c0563f85' name='lines'/>
<parameter type-id='7292109c' name='lines_cnt'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='libzfs_run_process_get_stdout_nopath' mangled-name='libzfs_run_process_get_stdout_nopath' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_run_process_get_stdout_nopath'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='9b23c9ad' name='argv'/>
<parameter type-id='9b23c9ad' name='env'/>
<parameter type-id='c0563f85' name='lines'/>
<parameter type-id='7292109c' name='lines_cnt'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='libzfs_free_str_array' mangled-name='libzfs_free_str_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_free_str_array'>
<parameter type-id='9b23c9ad' name='strs'/>
<parameter type-id='95e97e5e' name='count'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='libzfs_envvar_is_set' mangled-name='libzfs_envvar_is_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_envvar_is_set'>
<parameter type-id='26a90f95' name='envvar'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='libzfs_init' mangled-name='libzfs_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_init'>
<return type-id='b0382bb3'/>
</function-decl>
<function-decl name='libzfs_fini' mangled-name='libzfs_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_fini'>
<parameter type-id='b0382bb3' name='hdl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_get_handle' mangled-name='zpool_get_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_handle'>
<parameter type-id='4c81de99' name='zhp'/>
<return type-id='b0382bb3'/>
</function-decl>
<function-decl name='zfs_get_handle' mangled-name='zfs_get_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_handle'>
<parameter type-id='9200a744' name='zhp'/>
<return type-id='b0382bb3'/>
</function-decl>
<function-decl name='zfs_get_pool_handle' mangled-name='zfs_get_pool_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_pool_handle'>
<parameter type-id='fcd57163' name='zhp'/>
<return type-id='4c81de99'/>
</function-decl>
<function-decl name='zfs_path_to_zhandle' mangled-name='zfs_path_to_zhandle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_path_to_zhandle'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='2e45de5d' name='argtype'/>
<return type-id='9200a744'/>
</function-decl>
<function-decl name='zprop_print_one_property' mangled-name='zprop_print_one_property' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_print_one_property'>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='0d2a0670' name='cbp'/>
<parameter type-id='80f4b756' name='propname'/>
<parameter type-id='80f4b756' name='value'/>
<parameter type-id='a2256d42' name='sourcetype'/>
<parameter type-id='80f4b756' name='source'/>
<parameter type-id='80f4b756' name='recvd_value'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_nicestrtonum' mangled-name='zfs_nicestrtonum' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicestrtonum'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='80f4b756' name='value'/>
<parameter type-id='5d6479ae' name='num'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zprop_get_list' mangled-name='zprop_get_list' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_get_list'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='26a90f95' name='props'/>
<parameter type-id='e4378506' name='listp'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zprop_free_list' mangled-name='zprop_free_list' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_free_list'>
<parameter type-id='3a9b2288' name='pl'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zprop_iter' mangled-name='zprop_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_iter'>
<parameter type-id='1ec3747a' name='func'/>
<parameter type-id='eaa32e2f' name='cb'/>
<parameter type-id='c19b74c3' name='show_all'/>
<parameter type-id='c19b74c3' name='ordered'/>
<parameter type-id='2e45de5d' name='type'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_version_userland' mangled-name='zfs_version_userland' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_version_userland'>
<parameter type-id='26a90f95' name='version'/>
<parameter type-id='95e97e5e' name='len'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_version_print' mangled-name='zfs_version_print' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_version_print'>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='color_start' mangled-name='color_start' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='color_start'>
<parameter type-id='26a90f95' name='color'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='color_end' mangled-name='color_end' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='color_end'>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='printf_color' mangled-name='printf_color' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='printf_color'>
<parameter type-id='26a90f95' name='color'/>
<parameter type-id='26a90f95' name='format'/>
<parameter is-variadic='yes'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-type size-in-bits='64' id='c70fa2e8'>
<parameter type-id='95e97e5e'/>
<parameter type-id='eaa32e2f'/>
<return type-id='95e97e5e'/>
</function-type>
</abi-instr>
<abi-instr address-size='64' path='os/linux/libzfs_mount_os.c' language='LANG_C99'>
<pointer-type-def type-id='7359adad' size-in-bits='64' id='1d2c2b85'/>
<function-decl name='zfs_parse_mount_options' mangled-name='zfs_parse_mount_options' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_parse_mount_options'>
<parameter type-id='26a90f95' name='mntopts'/>
<parameter type-id='1d2c2b85' name='mntflags'/>
<parameter type-id='1d2c2b85' name='zfsflags'/>
<parameter type-id='95e97e5e' name='sloppy'/>
<parameter type-id='26a90f95' name='badopt'/>
<parameter type-id='26a90f95' name='mtabopt'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_adjust_mount_options' mangled-name='zfs_adjust_mount_options' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_adjust_mount_options'>
<parameter type-id='9200a744' name='zhp'/>
<parameter type-id='80f4b756' name='mntpoint'/>
<parameter type-id='26a90f95' name='mntopts'/>
<parameter type-id='26a90f95' name='mtabopt'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_mount_delegation_check' mangled-name='zfs_mount_delegation_check' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mount_delegation_check'>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_disable_datasets_os' mangled-name='zpool_disable_datasets_os' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disable_datasets_os'>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='c19b74c3' name='force'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_disable_volume_os' mangled-name='zpool_disable_volume_os' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disable_volume_os'>
<parameter type-id='80f4b756' name='name'/>
<return type-id='48b5725f'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='os/linux/libzfs_pool_os.c' language='LANG_C99'>
<function-decl name='zpool_label_disk' mangled-name='zpool_label_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_label_disk'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='4c81de99' name='zhp'/>
<parameter type-id='80f4b756' name='name'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='os/linux/libzfs_util_os.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='32768' id='d16c6df4'>
<subrange length='4096' type-id='7359adad' id='bc1b5ddc'/>
</array-type-def>
<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='65536' id='163f6aa5'>
<subrange length='8192' type-id='7359adad' id='c88f397d'/>
</array-type-def>
<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='128' id='c1c22e6c'>
<subrange length='2' type-id='7359adad' id='52efc4ef'/>
</array-type-def>
<array-type-def dimensions='1' type-id='b96825af' size-in-bits='24' id='d3490169'>
<subrange length='3' type-id='7359adad' id='56f209d2'/>
</array-type-def>
<class-decl name='drr_begin' size-in-bits='2432' is-struct='yes' visibility='default' id='09fcdc01'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='drr_magic' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='drr_versioninfo' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='drr_creation_time' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='drr_type' type-id='230f1e16' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='224'>
<var-decl name='drr_flags' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='drr_fromguid' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='drr_toname' type-id='d1617432' visibility='default'/>
</data-member>
</class-decl>
<class-decl name='zinject_record' size-in-bits='2816' is-struct='yes' visibility='default' id='3216f820'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='zi_objset' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='zi_object' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='zi_start' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='zi_end' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='zi_guid' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='zi_level' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='352'>
<var-decl name='zi_error' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='zi_type' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='zi_freq' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='480'>
<var-decl name='zi_failfast' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='512'>
<var-decl name='zi_func' type-id='d1617432' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2560'>
<var-decl name='zi_iotype' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2592'>
<var-decl name='zi_duration' type-id='3ff5601b' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2624'>
<var-decl name='zi_timer' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2688'>
<var-decl name='zi_nlanes' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2752'>
<var-decl name='zi_cmd' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='2784'>
<var-decl name='zi_dvas' type-id='8f92235e' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zinject_record_t' type-id='3216f820' id='a4301ca6'/>
<class-decl name='zfs_share' size-in-bits='256' is-struct='yes' visibility='default' id='feb6f2da'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='z_exportdata' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='z_sharedata' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='z_sharetype' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='z_sharemax' type-id='9c313c2d' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zfs_share_t' type-id='feb6f2da' id='ee5cec36'/>
<class-decl name='zfs_cmd' size-in-bits='109952' is-struct='yes' visibility='default' id='3522cd69'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='zc_name' type-id='d16c6df4' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32768'>
<var-decl name='zc_nvlist_src' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32832'>
<var-decl name='zc_nvlist_src_size' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32896'>
<var-decl name='zc_nvlist_dst' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='32960'>
<var-decl name='zc_nvlist_dst_size' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='33024'>
<var-decl name='zc_nvlist_dst_filled' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='33056'>
<var-decl name='zc_pad2' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='33088'>
<var-decl name='zc_history' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='33152'>
<var-decl name='zc_value' type-id='163f6aa5' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='98688'>
<var-decl name='zc_string' type-id='d1617432' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='100736'>
<var-decl name='zc_guid' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='100800'>
<var-decl name='zc_nvlist_conf' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='100864'>
<var-decl name='zc_nvlist_conf_size' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='100928'>
<var-decl name='zc_cookie' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='100992'>
<var-decl name='zc_objset_type' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='101056'>
<var-decl name='zc_perm_action' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='101120'>
<var-decl name='zc_history_len' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='101184'>
<var-decl name='zc_history_offset' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='101248'>
<var-decl name='zc_obj' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='101312'>
<var-decl name='zc_iflags' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='101376'>
<var-decl name='zc_share' type-id='ee5cec36' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='101632'>
<var-decl name='zc_objset_stats' type-id='b2c14f17' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='103936'>
<var-decl name='zc_begin_record' type-id='09fcdc01' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='106368'>
<var-decl name='zc_inject_record' type-id='a4301ca6' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109184'>
<var-decl name='zc_defer_destroy' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109216'>
<var-decl name='zc_flags' type-id='8f92235e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109248'>
<var-decl name='zc_action_handle' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109312'>
<var-decl name='zc_cleanup_fd' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109344'>
<var-decl name='zc_simple' type-id='b96825af' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109352'>
<var-decl name='zc_pad' type-id='d3490169' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109376'>
<var-decl name='zc_sendobj' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109440'>
<var-decl name='zc_fromobj' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109504'>
<var-decl name='zc_createtxg' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109568'>
<var-decl name='zc_stat' type-id='0371a9c7' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='109888'>
<var-decl name='zc_zoneid' type-id='9c313c2d' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zfs_cmd_t' type-id='3522cd69' id='a5559cdd'/>
<class-decl name='zfs_stat' size-in-bits='320' is-struct='yes' visibility='default' id='6417f0b9'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='zs_gen' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='zs_mode' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='zs_links' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='zs_ctime' type-id='c1c22e6c' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='zfs_stat_t' type-id='6417f0b9' id='0371a9c7'/>
<pointer-type-def type-id='a5559cdd' size-in-bits='64' id='e4ec4540'/>
<function-decl name='zfs_ioctl' mangled-name='zfs_ioctl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_ioctl'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='95e97e5e' name='request'/>
<parameter type-id='e4ec4540' name='zc'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='libzfs_error_init' mangled-name='libzfs_error_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_error_init'>
<parameter type-id='95e97e5e' name='error'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_destroy_snaps_nvl_os' mangled-name='zfs_destroy_snaps_nvl_os' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy_snaps_nvl_os'>
<parameter type-id='b0382bb3' name='hdl'/>
<parameter type-id='5ce45b60' name='snaps'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_version_kernel' mangled-name='zfs_version_kernel' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_version_kernel'>
<parameter type-id='26a90f95' name='version'/>
<parameter type-id='95e97e5e' name='len'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='os/linux/zutil_device_path_os.c' language='LANG_C99'>
<function-decl name='zfs_append_partition' mangled-name='zfs_append_partition' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_append_partition'>
<parameter type-id='26a90f95' name='path'/>
<parameter type-id='b59d7dce' name='max_len'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_strip_partition' mangled-name='zfs_strip_partition' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_strip_partition'>
<parameter type-id='26a90f95' name='path'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='zfs_strip_path' mangled-name='zfs_strip_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_strip_path'>
<parameter type-id='26a90f95' name='path'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='zfs_get_enclosure_sysfs_path' mangled-name='zfs_get_enclosure_sysfs_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_enclosure_sysfs_path'>
<parameter type-id='80f4b756' name='dev_name'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='zfs_dev_is_dm' mangled-name='zfs_dev_is_dm' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dev_is_dm'>
<parameter type-id='80f4b756' name='dev_name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_dev_is_whole_disk' mangled-name='zfs_dev_is_whole_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dev_is_whole_disk'>
<parameter type-id='80f4b756' name='dev_name'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_get_underlying_path' mangled-name='zfs_get_underlying_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_underlying_path'>
<parameter type-id='80f4b756' name='dev_name'/>
<return type-id='26a90f95'/>
</function-decl>
<function-decl name='is_mpath_whole_disk' mangled-name='is_mpath_whole_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='is_mpath_whole_disk'>
<parameter type-id='80f4b756' name='path'/>
<return type-id='c19b74c3'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='os/linux/zutil_import_os.c' language='LANG_C99'>
- <qualified-type-def type-id='80f4b756' const='yes' id='b99c00c9'/>
+ <class-decl name='udev_device' is-struct='yes' visibility='default' is-declaration-only='yes' id='640b33ca'/>
<pointer-type-def type-id='b99c00c9' size-in-bits='64' id='13956559'/>
<pointer-type-def type-id='b59d7dce' size-in-bits='64' id='78c01427'/>
<pointer-type-def type-id='640b33ca' size-in-bits='64' id='b32bae08'/>
<class-decl name='udev_device' is-struct='yes' visibility='default' is-declaration-only='yes' id='640b33ca'/>
<function-decl name='zfs_dev_flush' mangled-name='zfs_dev_flush' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dev_flush'>
<parameter type-id='95e97e5e' name='fd'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_default_search_paths' mangled-name='zpool_default_search_paths' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_default_search_paths'>
<parameter type-id='78c01427' name='count'/>
<return type-id='13956559'/>
</function-decl>
<function-decl name='zfs_device_get_devid' mangled-name='zfs_device_get_devid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_device_get_devid'>
<parameter type-id='b32bae08' name='dev'/>
<parameter type-id='26a90f95' name='bufptr'/>
<parameter type-id='b59d7dce' name='buflen'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_device_get_physical' mangled-name='zfs_device_get_physical' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_device_get_physical'>
<parameter type-id='b32bae08' name='dev'/>
<parameter type-id='26a90f95' name='bufptr'/>
<parameter type-id='b59d7dce' name='buflen'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_label_disk_wait' mangled-name='zpool_label_disk_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_label_disk_wait'>
<parameter type-id='80f4b756' name='path'/>
<parameter type-id='95e97e5e' name='timeout_ms'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='update_vdev_config_dev_strs' mangled-name='update_vdev_config_dev_strs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='update_vdev_config_dev_strs'>
<parameter type-id='5ce45b60' name='nv'/>
<return type-id='48b5725f'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='zutil_device_path.c' language='LANG_C99'>
<typedef-decl name='ssize_t' type-id='41060289' id='79a0948f'/>
<typedef-decl name='__ssize_t' type-id='bd54fe1a' id='41060289'/>
<function-decl name='zfs_basename' mangled-name='zfs_basename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_basename'>
<parameter type-id='80f4b756' name='path'/>
<return type-id='80f4b756'/>
</function-decl>
<function-decl name='zfs_dirnamelen' mangled-name='zfs_dirnamelen' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dirnamelen'>
<parameter type-id='80f4b756' name='path'/>
<return type-id='79a0948f'/>
</function-decl>
<function-decl name='zfs_resolve_shortname' mangled-name='zfs_resolve_shortname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_resolve_shortname'>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='26a90f95' name='path'/>
<parameter type-id='b59d7dce' name='len'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zfs_strcmp_pathname' mangled-name='zfs_strcmp_pathname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_strcmp_pathname'>
<parameter type-id='80f4b756' name='name'/>
<parameter type-id='80f4b756' name='cmp'/>
<parameter type-id='95e97e5e' name='wholedisk'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='zutil_import.c' language='LANG_C99'>
<class-decl name='importargs' size-in-bits='448' is-struct='yes' visibility='default' id='7ac83801'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='path' type-id='9b23c9ad' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='paths' type-id='95e97e5e' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='poolname' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='guid' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='cachefile' type-id='80f4b756' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='can_be_active' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='352'>
<var-decl name='scan' type-id='c19b74c3' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='policy' type-id='5ce45b60' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='importargs_t' type-id='7ac83801' id='7a842a6b'/>
<pointer-type-def type-id='7a842a6b' size-in-bits='64' id='07ee4a58'/>
<pointer-type-def type-id='b1e62775' size-in-bits='64' id='f095e320'/>
<function-decl name='zpool_read_label' mangled-name='zpool_read_label' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_read_label'>
<parameter type-id='95e97e5e' name='fd'/>
<parameter type-id='857bb57e' name='config'/>
<parameter type-id='7292109c' name='num_labels'/>
<return type-id='95e97e5e'/>
</function-decl>
<function-decl name='zpool_search_import' mangled-name='zpool_search_import' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_search_import'>
<parameter type-id='eaa32e2f' name='hdl'/>
<parameter type-id='07ee4a58' name='import'/>
<parameter type-id='f095e320' name='pco'/>
<return type-id='5ce45b60'/>
</function-decl>
<function-decl name='zpool_find_config' mangled-name='zpool_find_config' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_find_config'>
<parameter type-id='eaa32e2f' name='hdl'/>
<parameter type-id='80f4b756' name='target'/>
<parameter type-id='857bb57e' name='configp'/>
<parameter type-id='07ee4a58' name='args'/>
<parameter type-id='f095e320' name='pco'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='zutil_nicenum.c' language='LANG_C99'>
<enum-decl name='zfs_nicenum_format' id='29cf1969'>
<underlying-type type-id='9cac1fee'/>
<enumerator name='ZFS_NICENUM_1024' value='0'/>
<enumerator name='ZFS_NICENUM_BYTES' value='1'/>
<enumerator name='ZFS_NICENUM_TIME' value='2'/>
<enumerator name='ZFS_NICENUM_RAW' value='3'/>
<enumerator name='ZFS_NICENUM_RAWTIME' value='4'/>
</enum-decl>
<function-decl name='zfs_isnumber' mangled-name='zfs_isnumber' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_isnumber'>
<parameter type-id='80f4b756' name='str'/>
<return type-id='c19b74c3'/>
</function-decl>
<function-decl name='zfs_nicenum_format' mangled-name='zfs_nicenum_format' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicenum_format'>
<parameter type-id='9c313c2d' name='num'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='buflen'/>
<parameter type-id='29cf1969' name='format'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_nicenum' mangled-name='zfs_nicenum' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicenum'>
<parameter type-id='9c313c2d' name='num'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='buflen'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_nicetime' mangled-name='zfs_nicetime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicetime'>
<parameter type-id='9c313c2d' name='num'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='buflen'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_niceraw' mangled-name='zfs_niceraw' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_niceraw'>
<parameter type-id='9c313c2d' name='num'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='buflen'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zfs_nicebytes' mangled-name='zfs_nicebytes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicebytes'>
<parameter type-id='9c313c2d' name='num'/>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='b59d7dce' name='buflen'/>
<return type-id='48b5725f'/>
</function-decl>
</abi-instr>
<abi-instr address-size='64' path='zutil_pool.c' language='LANG_C99'>
<array-type-def dimensions='1' type-id='853fd5dc' size-in-bits='32768' id='b505fc2f'>
<subrange length='64' type-id='7359adad' id='b10be967'/>
</array-type-def>
<class-decl name='ddt_stat' size-in-bits='512' is-struct='yes' visibility='default' id='65242dfe'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='dds_blocks' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='64'>
<var-decl name='dds_lsize' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='128'>
<var-decl name='dds_psize' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='192'>
<var-decl name='dds_dsize' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='256'>
<var-decl name='dds_ref_blocks' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='320'>
<var-decl name='dds_ref_lsize' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='384'>
<var-decl name='dds_ref_psize' type-id='9c313c2d' visibility='default'/>
</data-member>
<data-member access='public' layout-offset-in-bits='448'>
<var-decl name='dds_ref_dsize' type-id='9c313c2d' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='ddt_stat_t' type-id='65242dfe' id='853fd5dc'/>
<class-decl name='ddt_histogram' size-in-bits='32768' is-struct='yes' visibility='default' id='bc2b3086'>
<data-member access='public' layout-offset-in-bits='0'>
<var-decl name='ddh_stat' type-id='b505fc2f' visibility='default'/>
</data-member>
</class-decl>
<typedef-decl name='ddt_histogram_t' type-id='bc2b3086' id='2d7fe832'/>
<qualified-type-def type-id='2d7fe832' const='yes' id='ec92d602'/>
<pointer-type-def type-id='ec92d602' size-in-bits='64' id='932720f8'/>
<qualified-type-def type-id='853fd5dc' const='yes' id='764c298c'/>
<pointer-type-def type-id='764c298c' size-in-bits='64' id='dfe59052'/>
<pointer-type-def type-id='857bb57e' size-in-bits='64' id='75be733c'/>
<function-decl name='zpool_dump_ddt' mangled-name='zpool_dump_ddt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_dump_ddt'>
<parameter type-id='dfe59052' name='dds_total'/>
<parameter type-id='932720f8' name='ddh'/>
<return type-id='48b5725f'/>
</function-decl>
<function-decl name='zpool_history_unpack' mangled-name='zpool_history_unpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_history_unpack'>
<parameter type-id='26a90f95' name='buf'/>
<parameter type-id='9c313c2d' name='bytes_read'/>
<parameter type-id='5d6479ae' name='leftover'/>
<parameter type-id='75be733c' name='records'/>
<parameter type-id='4dd26a40' name='numrecords'/>
<return type-id='95e97e5e'/>
</function-decl>
</abi-instr>
</abi-corpus>
diff --git a/module/avl/avl.c b/module/avl/avl.c
index e53e2fa5dfb6..c86dc9835a54 100644
--- a/module/avl/avl.c
+++ b/module/avl/avl.c
@@ -1,1090 +1,1090 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2015 by Delphix. All rights reserved.
*/
/*
* AVL - generic AVL tree implementation for kernel use
*
* A complete description of AVL trees can be found in many CS textbooks.
*
* Here is a very brief overview. An AVL tree is a binary search tree that is
* almost perfectly balanced. By "almost" perfectly balanced, we mean that at
* any given node, the left and right subtrees are allowed to differ in height
* by at most 1 level.
*
* This relaxation from a perfectly balanced binary tree allows doing
* insertion and deletion relatively efficiently. Searching the tree is
* still a fast operation, roughly O(log(N)).
*
* The key to insertion and deletion is a set of tree manipulations called
* rotations, which bring unbalanced subtrees back into the semi-balanced state.
*
* This implementation of AVL trees has the following peculiarities:
*
* - The AVL specific data structures are physically embedded as fields
* in the "using" data structures. To maintain generality the code
* must constantly translate between "avl_node_t *" and containing
* data structure "void *"s by adding/subtracting the avl_offset.
*
* - Since the AVL data is always embedded in other structures, there is
* no locking or memory allocation in the AVL routines. This must be
* provided for by the enclosing data structure's semantics. Typically,
* avl_insert()/_add()/_remove()/avl_insert_here() require some kind of
* exclusive write lock. Other operations require a read lock.
*
* - The implementation uses iteration instead of explicit recursion,
* since it is intended to run on limited size kernel stacks. Since
* there is no recursion stack present to move "up" in the tree,
* there is an explicit "parent" link in the avl_node_t.
*
* - The left/right children pointers of a node are in an array.
* In the code, variables (instead of constants) are used to represent
* left and right indices. The implementation is written as if it only
* dealt with left handed manipulations. By changing the value assigned
* to "left", the code also works for right handed trees. The
* following variables/terms are frequently used:
*
* int left; // 0 when dealing with left children,
* // 1 for dealing with right children
*
* int left_heavy; // -1 when left subtree is taller at some node,
* // +1 when right subtree is taller
*
* int right; // will be the opposite of left (0 or 1)
* int right_heavy;// will be the opposite of left_heavy (-1 or 1)
*
* int direction; // 0 for "<" (ie. left child); 1 for ">" (right)
*
* Though it is a little more confusing to read the code, the approach
* allows using half as much code (and hence cache footprint) for tree
* manipulations and eliminates many conditional branches.
*
* - The avl_index_t is an opaque "cookie" used to find nodes at or
* adjacent to where a new value would be inserted in the tree. The value
* is a modified "avl_node_t *". The bottom bit (normally 0 for a
* pointer) is set to indicate if that the new node has a value greater
* than the value of the indicated "avl_node_t *".
*
* Note - in addition to userland (e.g. libavl and libutil) and the kernel
* (e.g. genunix), avl.c is compiled into ld.so and kmdb's genunix module,
* which each have their own compilation environments and subsequent
* requirements. Each of these environments must be considered when adding
* dependencies from avl.c.
*
* Link to Illumos.org for more information on avl function:
* [1] https://illumos.org/man/9f/avl
*/
#include <sys/types.h>
#include <sys/param.h>
#include <sys/debug.h>
#include <sys/avl.h>
#include <sys/cmn_err.h>
#include <sys/mod.h>
/*
* Small arrays to translate between balance (or diff) values and child indices.
*
* Code that deals with binary tree data structures will randomly use
* left and right children when examining a tree. C "if()" statements
* which evaluate randomly suffer from very poor hardware branch prediction.
* In this code we avoid some of the branch mispredictions by using the
* following translation arrays. They replace random branches with an
* additional memory reference. Since the translation arrays are both very
* small the data should remain efficiently in cache.
*/
-static const int avl_child2balance[2] = {-1, 1};
+static const int avl_child2balance[] = {-1, 1};
static const int avl_balance2child[] = {0, 0, 1};
/*
* Walk from one node to the previous valued node (ie. an infix walk
* towards the left). At any given node we do one of 2 things:
*
* - If there is a left child, go to it, then to it's rightmost descendant.
*
* - otherwise we return through parent nodes until we've come from a right
* child.
*
* Return Value:
* NULL - if at the end of the nodes
* otherwise next node
*/
void *
avl_walk(avl_tree_t *tree, void *oldnode, int left)
{
size_t off = tree->avl_offset;
avl_node_t *node = AVL_DATA2NODE(oldnode, off);
int right = 1 - left;
int was_child;
/*
* nowhere to walk to if tree is empty
*/
if (node == NULL)
return (NULL);
/*
* Visit the previous valued node. There are two possibilities:
*
* If this node has a left child, go down one left, then all
* the way right.
*/
if (node->avl_child[left] != NULL) {
for (node = node->avl_child[left];
node->avl_child[right] != NULL;
node = node->avl_child[right])
;
/*
* Otherwise, return through left children as far as we can.
*/
} else {
for (;;) {
was_child = AVL_XCHILD(node);
node = AVL_XPARENT(node);
if (node == NULL)
return (NULL);
if (was_child == right)
break;
}
}
return (AVL_NODE2DATA(node, off));
}
/*
* Return the lowest valued node in a tree or NULL.
* (leftmost child from root of tree)
*/
void *
avl_first(avl_tree_t *tree)
{
avl_node_t *node;
avl_node_t *prev = NULL;
size_t off = tree->avl_offset;
for (node = tree->avl_root; node != NULL; node = node->avl_child[0])
prev = node;
if (prev != NULL)
return (AVL_NODE2DATA(prev, off));
return (NULL);
}
/*
* Return the highest valued node in a tree or NULL.
* (rightmost child from root of tree)
*/
void *
avl_last(avl_tree_t *tree)
{
avl_node_t *node;
avl_node_t *prev = NULL;
size_t off = tree->avl_offset;
for (node = tree->avl_root; node != NULL; node = node->avl_child[1])
prev = node;
if (prev != NULL)
return (AVL_NODE2DATA(prev, off));
return (NULL);
}
/*
* Access the node immediately before or after an insertion point.
*
* "avl_index_t" is a (avl_node_t *) with the bottom bit indicating a child
*
* Return value:
* NULL: no node in the given direction
* "void *" of the found tree node
*/
void *
avl_nearest(avl_tree_t *tree, avl_index_t where, int direction)
{
int child = AVL_INDEX2CHILD(where);
avl_node_t *node = AVL_INDEX2NODE(where);
void *data;
size_t off = tree->avl_offset;
if (node == NULL) {
ASSERT(tree->avl_root == NULL);
return (NULL);
}
data = AVL_NODE2DATA(node, off);
if (child != direction)
return (data);
return (avl_walk(tree, data, direction));
}
/*
* Search for the node which contains "value". The algorithm is a
* simple binary tree search.
*
* return value:
* NULL: the value is not in the AVL tree
* *where (if not NULL) is set to indicate the insertion point
* "void *" of the found tree node
*/
void *
avl_find(avl_tree_t *tree, const void *value, avl_index_t *where)
{
avl_node_t *node;
avl_node_t *prev = NULL;
int child = 0;
int diff;
size_t off = tree->avl_offset;
for (node = tree->avl_root; node != NULL;
node = node->avl_child[child]) {
prev = node;
diff = tree->avl_compar(value, AVL_NODE2DATA(node, off));
ASSERT(-1 <= diff && diff <= 1);
if (diff == 0) {
#ifdef ZFS_DEBUG
if (where != NULL)
*where = 0;
#endif
return (AVL_NODE2DATA(node, off));
}
child = avl_balance2child[1 + diff];
}
if (where != NULL)
*where = AVL_MKINDEX(prev, child);
return (NULL);
}
/*
* Perform a rotation to restore balance at the subtree given by depth.
*
* This routine is used by both insertion and deletion. The return value
* indicates:
* 0 : subtree did not change height
* !0 : subtree was reduced in height
*
* The code is written as if handling left rotations, right rotations are
* symmetric and handled by swapping values of variables right/left[_heavy]
*
* On input balance is the "new" balance at "node". This value is either
* -2 or +2.
*/
static int
avl_rotation(avl_tree_t *tree, avl_node_t *node, int balance)
{
int left = !(balance < 0); /* when balance = -2, left will be 0 */
int right = 1 - left;
int left_heavy = balance >> 1;
int right_heavy = -left_heavy;
avl_node_t *parent = AVL_XPARENT(node);
avl_node_t *child = node->avl_child[left];
avl_node_t *cright;
avl_node_t *gchild;
avl_node_t *gright;
avl_node_t *gleft;
int which_child = AVL_XCHILD(node);
int child_bal = AVL_XBALANCE(child);
/* BEGIN CSTYLED */
/*
* case 1 : node is overly left heavy, the left child is balanced or
* also left heavy. This requires the following rotation.
*
* (node bal:-2)
* / \
* / \
* (child bal:0 or -1)
* / \
* / \
* cright
*
* becomes:
*
* (child bal:1 or 0)
* / \
* / \
* (node bal:-1 or 0)
* / \
* / \
* cright
*
* we detect this situation by noting that child's balance is not
* right_heavy.
*/
/* END CSTYLED */
if (child_bal != right_heavy) {
/*
* compute new balance of nodes
*
* If child used to be left heavy (now balanced) we reduced
* the height of this sub-tree -- used in "return...;" below
*/
child_bal += right_heavy; /* adjust towards right */
/*
* move "cright" to be node's left child
*/
cright = child->avl_child[right];
node->avl_child[left] = cright;
if (cright != NULL) {
AVL_SETPARENT(cright, node);
AVL_SETCHILD(cright, left);
}
/*
* move node to be child's right child
*/
child->avl_child[right] = node;
AVL_SETBALANCE(node, -child_bal);
AVL_SETCHILD(node, right);
AVL_SETPARENT(node, child);
/*
* update the pointer into this subtree
*/
AVL_SETBALANCE(child, child_bal);
AVL_SETCHILD(child, which_child);
AVL_SETPARENT(child, parent);
if (parent != NULL)
parent->avl_child[which_child] = child;
else
tree->avl_root = child;
return (child_bal == 0);
}
/* BEGIN CSTYLED */
/*
* case 2 : When node is left heavy, but child is right heavy we use
* a different rotation.
*
* (node b:-2)
* / \
* / \
* / \
* (child b:+1)
* / \
* / \
* (gchild b: != 0)
* / \
* / \
* gleft gright
*
* becomes:
*
* (gchild b:0)
* / \
* / \
* / \
* (child b:?) (node b:?)
* / \ / \
* / \ / \
* gleft gright
*
* computing the new balances is more complicated. As an example:
* if gchild was right_heavy, then child is now left heavy
* else it is balanced
*/
/* END CSTYLED */
gchild = child->avl_child[right];
gleft = gchild->avl_child[left];
gright = gchild->avl_child[right];
/*
* move gright to left child of node and
*
* move gleft to right child of node
*/
node->avl_child[left] = gright;
if (gright != NULL) {
AVL_SETPARENT(gright, node);
AVL_SETCHILD(gright, left);
}
child->avl_child[right] = gleft;
if (gleft != NULL) {
AVL_SETPARENT(gleft, child);
AVL_SETCHILD(gleft, right);
}
/*
* move child to left child of gchild and
*
* move node to right child of gchild and
*
* fixup parent of all this to point to gchild
*/
balance = AVL_XBALANCE(gchild);
gchild->avl_child[left] = child;
AVL_SETBALANCE(child, (balance == right_heavy ? left_heavy : 0));
AVL_SETPARENT(child, gchild);
AVL_SETCHILD(child, left);
gchild->avl_child[right] = node;
AVL_SETBALANCE(node, (balance == left_heavy ? right_heavy : 0));
AVL_SETPARENT(node, gchild);
AVL_SETCHILD(node, right);
AVL_SETBALANCE(gchild, 0);
AVL_SETPARENT(gchild, parent);
AVL_SETCHILD(gchild, which_child);
if (parent != NULL)
parent->avl_child[which_child] = gchild;
else
tree->avl_root = gchild;
return (1); /* the new tree is always shorter */
}
/*
* Insert a new node into an AVL tree at the specified (from avl_find()) place.
*
* Newly inserted nodes are always leaf nodes in the tree, since avl_find()
* searches out to the leaf positions. The avl_index_t indicates the node
* which will be the parent of the new node.
*
* After the node is inserted, a single rotation further up the tree may
* be necessary to maintain an acceptable AVL balance.
*/
void
avl_insert(avl_tree_t *tree, void *new_data, avl_index_t where)
{
avl_node_t *node;
avl_node_t *parent = AVL_INDEX2NODE(where);
int old_balance;
int new_balance;
int which_child = AVL_INDEX2CHILD(where);
size_t off = tree->avl_offset;
#ifdef _LP64
ASSERT(((uintptr_t)new_data & 0x7) == 0);
#endif
node = AVL_DATA2NODE(new_data, off);
/*
* First, add the node to the tree at the indicated position.
*/
++tree->avl_numnodes;
node->avl_child[0] = NULL;
node->avl_child[1] = NULL;
AVL_SETCHILD(node, which_child);
AVL_SETBALANCE(node, 0);
AVL_SETPARENT(node, parent);
if (parent != NULL) {
ASSERT(parent->avl_child[which_child] == NULL);
parent->avl_child[which_child] = node;
} else {
ASSERT(tree->avl_root == NULL);
tree->avl_root = node;
}
/*
* Now, back up the tree modifying the balance of all nodes above the
* insertion point. If we get to a highly unbalanced ancestor, we
* need to do a rotation. If we back out of the tree we are done.
* If we brought any subtree into perfect balance (0), we are also done.
*/
for (;;) {
node = parent;
if (node == NULL)
return;
/*
* Compute the new balance
*/
old_balance = AVL_XBALANCE(node);
new_balance = old_balance + avl_child2balance[which_child];
/*
* If we introduced equal balance, then we are done immediately
*/
if (new_balance == 0) {
AVL_SETBALANCE(node, 0);
return;
}
/*
* If both old and new are not zero we went
* from -1 to -2 balance, do a rotation.
*/
if (old_balance != 0)
break;
AVL_SETBALANCE(node, new_balance);
parent = AVL_XPARENT(node);
which_child = AVL_XCHILD(node);
}
/*
* perform a rotation to fix the tree and return
*/
(void) avl_rotation(tree, node, new_balance);
}
/*
* Insert "new_data" in "tree" in the given "direction" either after or
* before (AVL_AFTER, AVL_BEFORE) the data "here".
*
* Insertions can only be done at empty leaf points in the tree, therefore
* if the given child of the node is already present we move to either
* the AVL_PREV or AVL_NEXT and reverse the insertion direction. Since
* every other node in the tree is a leaf, this always works.
*
* To help developers using this interface, we assert that the new node
* is correctly ordered at every step of the way in DEBUG kernels.
*/
void
avl_insert_here(
avl_tree_t *tree,
void *new_data,
void *here,
int direction)
{
avl_node_t *node;
int child = direction; /* rely on AVL_BEFORE == 0, AVL_AFTER == 1 */
#ifdef ZFS_DEBUG
int diff;
#endif
ASSERT(tree != NULL);
ASSERT(new_data != NULL);
ASSERT(here != NULL);
ASSERT(direction == AVL_BEFORE || direction == AVL_AFTER);
/*
* If corresponding child of node is not NULL, go to the neighboring
* node and reverse the insertion direction.
*/
node = AVL_DATA2NODE(here, tree->avl_offset);
#ifdef ZFS_DEBUG
diff = tree->avl_compar(new_data, here);
ASSERT(-1 <= diff && diff <= 1);
ASSERT(diff != 0);
ASSERT(diff > 0 ? child == 1 : child == 0);
#endif
if (node->avl_child[child] != NULL) {
node = node->avl_child[child];
child = 1 - child;
while (node->avl_child[child] != NULL) {
#ifdef ZFS_DEBUG
diff = tree->avl_compar(new_data,
AVL_NODE2DATA(node, tree->avl_offset));
ASSERT(-1 <= diff && diff <= 1);
ASSERT(diff != 0);
ASSERT(diff > 0 ? child == 1 : child == 0);
#endif
node = node->avl_child[child];
}
#ifdef ZFS_DEBUG
diff = tree->avl_compar(new_data,
AVL_NODE2DATA(node, tree->avl_offset));
ASSERT(-1 <= diff && diff <= 1);
ASSERT(diff != 0);
ASSERT(diff > 0 ? child == 1 : child == 0);
#endif
}
ASSERT(node->avl_child[child] == NULL);
avl_insert(tree, new_data, AVL_MKINDEX(node, child));
}
/*
* Add a new node to an AVL tree. Strictly enforce that no duplicates can
* be added to the tree with a VERIFY which is enabled for non-DEBUG builds.
*/
void
avl_add(avl_tree_t *tree, void *new_node)
{
avl_index_t where = 0;
VERIFY(avl_find(tree, new_node, &where) == NULL);
avl_insert(tree, new_node, where);
}
/*
* Delete a node from the AVL tree. Deletion is similar to insertion, but
* with 2 complications.
*
* First, we may be deleting an interior node. Consider the following subtree:
*
* d c c
* / \ / \ / \
* b e b e b e
* / \ / \ /
* a c a a
*
* When we are deleting node (d), we find and bring up an adjacent valued leaf
* node, say (c), to take the interior node's place. In the code this is
* handled by temporarily swapping (d) and (c) in the tree and then using
* common code to delete (d) from the leaf position.
*
* Secondly, an interior deletion from a deep tree may require more than one
* rotation to fix the balance. This is handled by moving up the tree through
* parents and applying rotations as needed. The return value from
* avl_rotation() is used to detect when a subtree did not change overall
* height due to a rotation.
*/
void
avl_remove(avl_tree_t *tree, void *data)
{
avl_node_t *delete;
avl_node_t *parent;
avl_node_t *node;
avl_node_t tmp;
int old_balance;
int new_balance;
int left;
int right;
int which_child;
size_t off = tree->avl_offset;
delete = AVL_DATA2NODE(data, off);
/*
* Deletion is easiest with a node that has at most 1 child.
* We swap a node with 2 children with a sequentially valued
* neighbor node. That node will have at most 1 child. Note this
* has no effect on the ordering of the remaining nodes.
*
* As an optimization, we choose the greater neighbor if the tree
* is right heavy, otherwise the left neighbor. This reduces the
* number of rotations needed.
*/
if (delete->avl_child[0] != NULL && delete->avl_child[1] != NULL) {
/*
* choose node to swap from whichever side is taller
*/
old_balance = AVL_XBALANCE(delete);
left = avl_balance2child[old_balance + 1];
right = 1 - left;
/*
* get to the previous value'd node
* (down 1 left, as far as possible right)
*/
for (node = delete->avl_child[left];
node->avl_child[right] != NULL;
node = node->avl_child[right])
;
/*
* create a temp placeholder for 'node'
* move 'node' to delete's spot in the tree
*/
tmp = *node;
*node = *delete;
if (node->avl_child[left] == node)
node->avl_child[left] = &tmp;
parent = AVL_XPARENT(node);
if (parent != NULL)
parent->avl_child[AVL_XCHILD(node)] = node;
else
tree->avl_root = node;
AVL_SETPARENT(node->avl_child[left], node);
AVL_SETPARENT(node->avl_child[right], node);
/*
* Put tmp where node used to be (just temporary).
* It always has a parent and at most 1 child.
*/
delete = &tmp;
parent = AVL_XPARENT(delete);
parent->avl_child[AVL_XCHILD(delete)] = delete;
which_child = (delete->avl_child[1] != 0);
if (delete->avl_child[which_child] != NULL)
AVL_SETPARENT(delete->avl_child[which_child], delete);
}
/*
* Here we know "delete" is at least partially a leaf node. It can
* be easily removed from the tree.
*/
ASSERT(tree->avl_numnodes > 0);
--tree->avl_numnodes;
parent = AVL_XPARENT(delete);
which_child = AVL_XCHILD(delete);
if (delete->avl_child[0] != NULL)
node = delete->avl_child[0];
else
node = delete->avl_child[1];
/*
* Connect parent directly to node (leaving out delete).
*/
if (node != NULL) {
AVL_SETPARENT(node, parent);
AVL_SETCHILD(node, which_child);
}
if (parent == NULL) {
tree->avl_root = node;
return;
}
parent->avl_child[which_child] = node;
/*
* Since the subtree is now shorter, begin adjusting parent balances
* and performing any needed rotations.
*/
do {
/*
* Move up the tree and adjust the balance
*
* Capture the parent and which_child values for the next
* iteration before any rotations occur.
*/
node = parent;
old_balance = AVL_XBALANCE(node);
new_balance = old_balance - avl_child2balance[which_child];
parent = AVL_XPARENT(node);
which_child = AVL_XCHILD(node);
/*
* If a node was in perfect balance but isn't anymore then
* we can stop, since the height didn't change above this point
* due to a deletion.
*/
if (old_balance == 0) {
AVL_SETBALANCE(node, new_balance);
break;
}
/*
* If the new balance is zero, we don't need to rotate
* else
* need a rotation to fix the balance.
* If the rotation doesn't change the height
* of the sub-tree we have finished adjusting.
*/
if (new_balance == 0)
AVL_SETBALANCE(node, new_balance);
else if (!avl_rotation(tree, node, new_balance))
break;
} while (parent != NULL);
}
#define AVL_REINSERT(tree, obj) \
avl_remove((tree), (obj)); \
avl_add((tree), (obj))
boolean_t
avl_update_lt(avl_tree_t *t, void *obj)
{
void *neighbor;
ASSERT(((neighbor = AVL_NEXT(t, obj)) == NULL) ||
(t->avl_compar(obj, neighbor) <= 0));
neighbor = AVL_PREV(t, obj);
if ((neighbor != NULL) && (t->avl_compar(obj, neighbor) < 0)) {
AVL_REINSERT(t, obj);
return (B_TRUE);
}
return (B_FALSE);
}
boolean_t
avl_update_gt(avl_tree_t *t, void *obj)
{
void *neighbor;
ASSERT(((neighbor = AVL_PREV(t, obj)) == NULL) ||
(t->avl_compar(obj, neighbor) >= 0));
neighbor = AVL_NEXT(t, obj);
if ((neighbor != NULL) && (t->avl_compar(obj, neighbor) > 0)) {
AVL_REINSERT(t, obj);
return (B_TRUE);
}
return (B_FALSE);
}
boolean_t
avl_update(avl_tree_t *t, void *obj)
{
void *neighbor;
neighbor = AVL_PREV(t, obj);
if ((neighbor != NULL) && (t->avl_compar(obj, neighbor) < 0)) {
AVL_REINSERT(t, obj);
return (B_TRUE);
}
neighbor = AVL_NEXT(t, obj);
if ((neighbor != NULL) && (t->avl_compar(obj, neighbor) > 0)) {
AVL_REINSERT(t, obj);
return (B_TRUE);
}
return (B_FALSE);
}
void
avl_swap(avl_tree_t *tree1, avl_tree_t *tree2)
{
avl_node_t *temp_node;
ulong_t temp_numnodes;
ASSERT3P(tree1->avl_compar, ==, tree2->avl_compar);
ASSERT3U(tree1->avl_offset, ==, tree2->avl_offset);
temp_node = tree1->avl_root;
temp_numnodes = tree1->avl_numnodes;
tree1->avl_root = tree2->avl_root;
tree1->avl_numnodes = tree2->avl_numnodes;
tree2->avl_root = temp_node;
tree2->avl_numnodes = temp_numnodes;
}
/*
* initialize a new AVL tree
*/
void
avl_create(avl_tree_t *tree, int (*compar) (const void *, const void *),
size_t size, size_t offset)
{
ASSERT(tree);
ASSERT(compar);
ASSERT(size > 0);
ASSERT(size >= offset + sizeof (avl_node_t));
#ifdef _LP64
ASSERT((offset & 0x7) == 0);
#endif
tree->avl_compar = compar;
tree->avl_root = NULL;
tree->avl_numnodes = 0;
tree->avl_offset = offset;
}
/*
* Delete a tree.
*/
void
avl_destroy(avl_tree_t *tree)
{
ASSERT(tree);
ASSERT(tree->avl_numnodes == 0);
ASSERT(tree->avl_root == NULL);
}
/*
* Return the number of nodes in an AVL tree.
*/
ulong_t
avl_numnodes(avl_tree_t *tree)
{
ASSERT(tree);
return (tree->avl_numnodes);
}
boolean_t
avl_is_empty(avl_tree_t *tree)
{
ASSERT(tree);
return (tree->avl_numnodes == 0);
}
#define CHILDBIT (1L)
/*
* Post-order tree walk used to visit all tree nodes and destroy the tree
* in post order. This is used for removing all the nodes from a tree without
* paying any cost for rebalancing it.
*
* example:
*
* void *cookie = NULL;
* my_data_t *node;
*
* while ((node = avl_destroy_nodes(tree, &cookie)) != NULL)
* free(node);
* avl_destroy(tree);
*
* The cookie is really an avl_node_t to the current node's parent and
* an indication of which child you looked at last.
*
* On input, a cookie value of CHILDBIT indicates the tree is done.
*/
void *
avl_destroy_nodes(avl_tree_t *tree, void **cookie)
{
avl_node_t *node;
avl_node_t *parent;
int child;
void *first;
size_t off = tree->avl_offset;
/*
* Initial calls go to the first node or it's right descendant.
*/
if (*cookie == NULL) {
first = avl_first(tree);
/*
* deal with an empty tree
*/
if (first == NULL) {
*cookie = (void *)CHILDBIT;
return (NULL);
}
node = AVL_DATA2NODE(first, off);
parent = AVL_XPARENT(node);
goto check_right_side;
}
/*
* If there is no parent to return to we are done.
*/
parent = (avl_node_t *)((uintptr_t)(*cookie) & ~CHILDBIT);
if (parent == NULL) {
if (tree->avl_root != NULL) {
ASSERT(tree->avl_numnodes == 1);
tree->avl_root = NULL;
tree->avl_numnodes = 0;
}
return (NULL);
}
/*
* Remove the child pointer we just visited from the parent and tree.
*/
child = (uintptr_t)(*cookie) & CHILDBIT;
parent->avl_child[child] = NULL;
ASSERT(tree->avl_numnodes > 1);
--tree->avl_numnodes;
/*
* If we just removed a right child or there isn't one, go up to parent.
*/
if (child == 1 || parent->avl_child[1] == NULL) {
node = parent;
parent = AVL_XPARENT(parent);
goto done;
}
/*
* Do parent's right child, then leftmost descendent.
*/
node = parent->avl_child[1];
while (node->avl_child[0] != NULL) {
parent = node;
node = node->avl_child[0];
}
/*
* If here, we moved to a left child. It may have one
* child on the right (when balance == +1).
*/
check_right_side:
if (node->avl_child[1] != NULL) {
ASSERT(AVL_XBALANCE(node) == 1);
parent = node;
node = node->avl_child[1];
ASSERT(node->avl_child[0] == NULL &&
node->avl_child[1] == NULL);
} else {
ASSERT(AVL_XBALANCE(node) <= 0);
}
done:
if (parent == NULL) {
*cookie = (void *)CHILDBIT;
ASSERT(node == tree->avl_root);
} else {
*cookie = (void *)((uintptr_t)parent | AVL_XCHILD(node));
}
return (AVL_NODE2DATA(node, off));
}
#if defined(_KERNEL)
static int __init
avl_init(void)
{
return (0);
}
static void __exit
avl_fini(void)
{
}
module_init(avl_init);
module_exit(avl_fini);
#endif
ZFS_MODULE_DESCRIPTION("Generic AVL tree implementation");
ZFS_MODULE_AUTHOR(ZFS_META_AUTHOR);
ZFS_MODULE_LICENSE(ZFS_META_LICENSE);
ZFS_MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
EXPORT_SYMBOL(avl_create);
EXPORT_SYMBOL(avl_find);
EXPORT_SYMBOL(avl_insert);
EXPORT_SYMBOL(avl_insert_here);
EXPORT_SYMBOL(avl_walk);
EXPORT_SYMBOL(avl_first);
EXPORT_SYMBOL(avl_last);
EXPORT_SYMBOL(avl_nearest);
EXPORT_SYMBOL(avl_add);
EXPORT_SYMBOL(avl_swap);
EXPORT_SYMBOL(avl_is_empty);
EXPORT_SYMBOL(avl_remove);
EXPORT_SYMBOL(avl_numnodes);
EXPORT_SYMBOL(avl_destroy_nodes);
EXPORT_SYMBOL(avl_destroy);
EXPORT_SYMBOL(avl_update_lt);
EXPORT_SYMBOL(avl_update_gt);
EXPORT_SYMBOL(avl_update);
diff --git a/module/icp/algs/aes/aes_impl.c b/module/icp/algs/aes/aes_impl.c
index a5b88b8aab25..c238bee2170b 100644
--- a/module/icp/algs/aes/aes_impl.c
+++ b/module/icp/algs/aes/aes_impl.c
@@ -1,442 +1,442 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/crypto/icp.h>
#include <sys/crypto/spi.h>
#include <sys/simd.h>
#include <modes/modes.h>
#include <aes/aes_impl.h>
/*
* Initialize AES encryption and decryption key schedules.
*
* Parameters:
* cipherKey User key
* keyBits AES key size (128, 192, or 256 bits)
* keysched AES key schedule to be initialized, of type aes_key_t.
* Allocated by aes_alloc_keysched().
*/
void
aes_init_keysched(const uint8_t *cipherKey, uint_t keyBits, void *keysched)
{
const aes_impl_ops_t *ops = aes_impl_get_ops();
aes_key_t *newbie = keysched;
uint_t keysize, i, j;
union {
uint64_t ka64[4];
uint32_t ka32[8];
} keyarr;
switch (keyBits) {
case 128:
newbie->nr = 10;
break;
case 192:
newbie->nr = 12;
break;
case 256:
newbie->nr = 14;
break;
default:
/* should never get here */
return;
}
keysize = CRYPTO_BITS2BYTES(keyBits);
/*
* Generic C implementation requires byteswap for little endian
* machines, various accelerated implementations for various
* architectures may not.
*/
if (!ops->needs_byteswap) {
/* no byteswap needed */
if (IS_P2ALIGNED(cipherKey, sizeof (uint64_t))) {
for (i = 0, j = 0; j < keysize; i++, j += 8) {
/* LINTED: pointer alignment */
keyarr.ka64[i] = *((uint64_t *)&cipherKey[j]);
}
} else {
bcopy(cipherKey, keyarr.ka32, keysize);
}
} else {
/* byte swap */
for (i = 0, j = 0; j < keysize; i++, j += 4) {
keyarr.ka32[i] =
htonl(*(uint32_t *)(void *)&cipherKey[j]);
}
}
ops->generate(newbie, keyarr.ka32, keyBits);
newbie->ops = ops;
/*
* Note: if there are systems that need the AES_64BIT_KS type in the
* future, move setting key schedule type to individual implementations
*/
newbie->type = AES_32BIT_KS;
}
/*
* Encrypt one block using AES.
* Align if needed and (for x86 32-bit only) byte-swap.
*
* Parameters:
* ks Key schedule, of type aes_key_t
* pt Input block (plain text)
* ct Output block (crypto text). Can overlap with pt
*/
int
aes_encrypt_block(const void *ks, const uint8_t *pt, uint8_t *ct)
{
aes_key_t *ksch = (aes_key_t *)ks;
const aes_impl_ops_t *ops = ksch->ops;
if (IS_P2ALIGNED2(pt, ct, sizeof (uint32_t)) && !ops->needs_byteswap) {
/* LINTED: pointer alignment */
ops->encrypt(&ksch->encr_ks.ks32[0], ksch->nr,
/* LINTED: pointer alignment */
(uint32_t *)pt, (uint32_t *)ct);
} else {
uint32_t buffer[AES_BLOCK_LEN / sizeof (uint32_t)];
/* Copy input block into buffer */
if (ops->needs_byteswap) {
buffer[0] = htonl(*(uint32_t *)(void *)&pt[0]);
buffer[1] = htonl(*(uint32_t *)(void *)&pt[4]);
buffer[2] = htonl(*(uint32_t *)(void *)&pt[8]);
buffer[3] = htonl(*(uint32_t *)(void *)&pt[12]);
} else
bcopy(pt, &buffer, AES_BLOCK_LEN);
ops->encrypt(&ksch->encr_ks.ks32[0], ksch->nr, buffer, buffer);
/* Copy result from buffer to output block */
if (ops->needs_byteswap) {
*(uint32_t *)(void *)&ct[0] = htonl(buffer[0]);
*(uint32_t *)(void *)&ct[4] = htonl(buffer[1]);
*(uint32_t *)(void *)&ct[8] = htonl(buffer[2]);
*(uint32_t *)(void *)&ct[12] = htonl(buffer[3]);
} else
bcopy(&buffer, ct, AES_BLOCK_LEN);
}
return (CRYPTO_SUCCESS);
}
/*
* Decrypt one block using AES.
* Align and byte-swap if needed.
*
* Parameters:
* ks Key schedule, of type aes_key_t
* ct Input block (crypto text)
* pt Output block (plain text). Can overlap with pt
*/
int
aes_decrypt_block(const void *ks, const uint8_t *ct, uint8_t *pt)
{
aes_key_t *ksch = (aes_key_t *)ks;
const aes_impl_ops_t *ops = ksch->ops;
if (IS_P2ALIGNED2(ct, pt, sizeof (uint32_t)) && !ops->needs_byteswap) {
/* LINTED: pointer alignment */
ops->decrypt(&ksch->decr_ks.ks32[0], ksch->nr,
/* LINTED: pointer alignment */
(uint32_t *)ct, (uint32_t *)pt);
} else {
uint32_t buffer[AES_BLOCK_LEN / sizeof (uint32_t)];
/* Copy input block into buffer */
if (ops->needs_byteswap) {
buffer[0] = htonl(*(uint32_t *)(void *)&ct[0]);
buffer[1] = htonl(*(uint32_t *)(void *)&ct[4]);
buffer[2] = htonl(*(uint32_t *)(void *)&ct[8]);
buffer[3] = htonl(*(uint32_t *)(void *)&ct[12]);
} else
bcopy(ct, &buffer, AES_BLOCK_LEN);
ops->decrypt(&ksch->decr_ks.ks32[0], ksch->nr, buffer, buffer);
/* Copy result from buffer to output block */
if (ops->needs_byteswap) {
*(uint32_t *)(void *)&pt[0] = htonl(buffer[0]);
*(uint32_t *)(void *)&pt[4] = htonl(buffer[1]);
*(uint32_t *)(void *)&pt[8] = htonl(buffer[2]);
*(uint32_t *)(void *)&pt[12] = htonl(buffer[3]);
} else
bcopy(&buffer, pt, AES_BLOCK_LEN);
}
return (CRYPTO_SUCCESS);
}
/*
* Allocate key schedule for AES.
*
* Return the pointer and set size to the number of bytes allocated.
* Memory allocated must be freed by the caller when done.
*
* Parameters:
* size Size of key schedule allocated, in bytes
* kmflag Flag passed to kmem_alloc(9F); ignored in userland.
*/
void *
aes_alloc_keysched(size_t *size, int kmflag)
{
aes_key_t *keysched;
keysched = (aes_key_t *)kmem_alloc(sizeof (aes_key_t), kmflag);
if (keysched != NULL) {
*size = sizeof (aes_key_t);
return (keysched);
}
return (NULL);
}
/* AES implementation that contains the fastest methods */
static aes_impl_ops_t aes_fastest_impl = {
.name = "fastest"
};
/* All compiled in implementations */
-const aes_impl_ops_t *aes_all_impl[] = {
+static const aes_impl_ops_t *aes_all_impl[] = {
&aes_generic_impl,
#if defined(__x86_64)
&aes_x86_64_impl,
#endif
#if defined(__x86_64) && defined(HAVE_AES)
&aes_aesni_impl,
#endif
};
/* Indicate that benchmark has been completed */
static boolean_t aes_impl_initialized = B_FALSE;
/* Select aes implementation */
#define IMPL_FASTEST (UINT32_MAX)
#define IMPL_CYCLE (UINT32_MAX-1)
#define AES_IMPL_READ(i) (*(volatile uint32_t *) &(i))
static uint32_t icp_aes_impl = IMPL_FASTEST;
static uint32_t user_sel_impl = IMPL_FASTEST;
/* Hold all supported implementations */
static size_t aes_supp_impl_cnt = 0;
static aes_impl_ops_t *aes_supp_impl[ARRAY_SIZE(aes_all_impl)];
/*
* Returns the AES operations for encrypt/decrypt/key setup. When a
* SIMD implementation is not allowed in the current context, then
* fallback to the fastest generic implementation.
*/
const aes_impl_ops_t *
aes_impl_get_ops(void)
{
if (!kfpu_allowed())
return (&aes_generic_impl);
const aes_impl_ops_t *ops = NULL;
const uint32_t impl = AES_IMPL_READ(icp_aes_impl);
switch (impl) {
case IMPL_FASTEST:
ASSERT(aes_impl_initialized);
ops = &aes_fastest_impl;
break;
case IMPL_CYCLE:
/* Cycle through supported implementations */
ASSERT(aes_impl_initialized);
ASSERT3U(aes_supp_impl_cnt, >, 0);
static size_t cycle_impl_idx = 0;
size_t idx = (++cycle_impl_idx) % aes_supp_impl_cnt;
ops = aes_supp_impl[idx];
break;
default:
ASSERT3U(impl, <, aes_supp_impl_cnt);
ASSERT3U(aes_supp_impl_cnt, >, 0);
if (impl < ARRAY_SIZE(aes_all_impl))
ops = aes_supp_impl[impl];
break;
}
ASSERT3P(ops, !=, NULL);
return (ops);
}
/*
* Initialize all supported implementations.
*/
void
aes_impl_init(void)
{
aes_impl_ops_t *curr_impl;
int i, c;
/* Move supported implementations into aes_supp_impls */
for (i = 0, c = 0; i < ARRAY_SIZE(aes_all_impl); i++) {
curr_impl = (aes_impl_ops_t *)aes_all_impl[i];
if (curr_impl->is_supported())
aes_supp_impl[c++] = (aes_impl_ops_t *)curr_impl;
}
aes_supp_impl_cnt = c;
/*
* Set the fastest implementation given the assumption that the
* hardware accelerated version is the fastest.
*/
#if defined(__x86_64)
#if defined(HAVE_AES)
if (aes_aesni_impl.is_supported()) {
memcpy(&aes_fastest_impl, &aes_aesni_impl,
sizeof (aes_fastest_impl));
} else
#endif
{
memcpy(&aes_fastest_impl, &aes_x86_64_impl,
sizeof (aes_fastest_impl));
}
#else
memcpy(&aes_fastest_impl, &aes_generic_impl,
sizeof (aes_fastest_impl));
#endif
strlcpy(aes_fastest_impl.name, "fastest", AES_IMPL_NAME_MAX);
/* Finish initialization */
atomic_swap_32(&icp_aes_impl, user_sel_impl);
aes_impl_initialized = B_TRUE;
}
static const struct {
char *name;
uint32_t sel;
} aes_impl_opts[] = {
{ "cycle", IMPL_CYCLE },
{ "fastest", IMPL_FASTEST },
};
/*
* Function sets desired aes implementation.
*
* If we are called before init(), user preference will be saved in
* user_sel_impl, and applied in later init() call. This occurs when module
* parameter is specified on module load. Otherwise, directly update
* icp_aes_impl.
*
* @val Name of aes implementation to use
* @param Unused.
*/
int
aes_impl_set(const char *val)
{
int err = -EINVAL;
char req_name[AES_IMPL_NAME_MAX];
uint32_t impl = AES_IMPL_READ(user_sel_impl);
size_t i;
/* sanitize input */
i = strnlen(val, AES_IMPL_NAME_MAX);
if (i == 0 || i >= AES_IMPL_NAME_MAX)
return (err);
strlcpy(req_name, val, AES_IMPL_NAME_MAX);
while (i > 0 && isspace(req_name[i-1]))
i--;
req_name[i] = '\0';
/* Check mandatory options */
for (i = 0; i < ARRAY_SIZE(aes_impl_opts); i++) {
if (strcmp(req_name, aes_impl_opts[i].name) == 0) {
impl = aes_impl_opts[i].sel;
err = 0;
break;
}
}
/* check all supported impl if init() was already called */
if (err != 0 && aes_impl_initialized) {
/* check all supported implementations */
for (i = 0; i < aes_supp_impl_cnt; i++) {
if (strcmp(req_name, aes_supp_impl[i]->name) == 0) {
impl = i;
err = 0;
break;
}
}
}
if (err == 0) {
if (aes_impl_initialized)
atomic_swap_32(&icp_aes_impl, impl);
else
atomic_swap_32(&user_sel_impl, impl);
}
return (err);
}
#if defined(_KERNEL) && defined(__linux__)
static int
icp_aes_impl_set(const char *val, zfs_kernel_param_t *kp)
{
return (aes_impl_set(val));
}
static int
icp_aes_impl_get(char *buffer, zfs_kernel_param_t *kp)
{
int i, cnt = 0;
char *fmt;
const uint32_t impl = AES_IMPL_READ(icp_aes_impl);
ASSERT(aes_impl_initialized);
/* list mandatory options */
for (i = 0; i < ARRAY_SIZE(aes_impl_opts); i++) {
fmt = (impl == aes_impl_opts[i].sel) ? "[%s] " : "%s ";
cnt += sprintf(buffer + cnt, fmt, aes_impl_opts[i].name);
}
/* list all supported implementations */
for (i = 0; i < aes_supp_impl_cnt; i++) {
fmt = (i == impl) ? "[%s] " : "%s ";
cnt += sprintf(buffer + cnt, fmt, aes_supp_impl[i]->name);
}
return (cnt);
}
module_param_call(icp_aes_impl, icp_aes_impl_set, icp_aes_impl_get,
NULL, 0644);
MODULE_PARM_DESC(icp_aes_impl, "Select aes implementation.");
#endif
diff --git a/module/icp/algs/edonr/edonr_byteorder.h b/module/icp/algs/edonr/edonr_byteorder.h
index 2b5d48287f26..cd35e5e4c7c9 100644
--- a/module/icp/algs/edonr/edonr_byteorder.h
+++ b/module/icp/algs/edonr/edonr_byteorder.h
@@ -1,216 +1,216 @@
/*
* IDI,NTNU
*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://opensource.org/licenses/CDDL-1.0.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*
* Copyright (C) 2009, 2010, Jorn Amundsen <jorn.amundsen@ntnu.no>
*
* C header file to determine compile machine byte order. Take care when cross
* compiling.
*
* $Id: byteorder.h 517 2013-02-17 20:34:39Z joern $
*/
/*
* Portions copyright (c) 2013, Saso Kiselkov, All rights reserved
*/
#ifndef _CRYPTO_EDONR_BYTEORDER_H
#define _CRYPTO_EDONR_BYTEORDER_H
#include <sys/sysmacros.h>
#include <sys/param.h>
#if defined(__BYTE_ORDER)
#if (__BYTE_ORDER == __BIG_ENDIAN)
#define MACHINE_IS_BIG_ENDIAN
#elif (__BYTE_ORDER == __LITTLE_ENDIAN)
#define MACHINE_IS_LITTLE_ENDIAN
#endif
#elif defined(BYTE_ORDER)
#if (BYTE_ORDER == BIG_ENDIAN)
#define MACHINE_IS_BIG_ENDIAN
#elif (BYTE_ORDER == LITTLE_ENDIAN)
#define MACHINE_IS_LITTLE_ENDIAN
#endif
#endif /* __BYTE_ORDER || BYTE_ORDER */
#if !defined(MACHINE_IS_BIG_ENDIAN) && !defined(MACHINE_IS_LITTLE_ENDIAN)
#if defined(_ZFS_BIG_ENDIAN) || defined(_MIPSEB)
#define MACHINE_IS_BIG_ENDIAN
#endif
#if defined(_ZFS_LITTLE_ENDIAN) || defined(_MIPSEL)
#define MACHINE_IS_LITTLE_ENDIAN
#endif
#endif /* !MACHINE_IS_BIG_ENDIAN && !MACHINE_IS_LITTLE_ENDIAN */
#if !defined(MACHINE_IS_BIG_ENDIAN) && !defined(MACHINE_IS_LITTLE_ENDIAN)
-#error unknown machine byte sex
+#error unknown machine byte order
#endif
#define BYTEORDER_INCLUDED
#if defined(MACHINE_IS_BIG_ENDIAN)
/*
* Byte swapping macros for big endian architectures and compilers,
* add as appropriate for other architectures and/or compilers.
*
* ld_swap64(src,dst) : uint64_t dst = *(src)
* st_swap64(src,dst) : *(dst) = uint64_t src
*/
#if defined(__PPC__) || defined(_ARCH_PPC)
#if defined(__64BIT__)
#if defined(_ARCH_PWR7)
#define aix_ld_swap64(s64, d64)\
__asm__("ldbrx %0,0,%1" : "=r"(d64) : "r"(s64))
#define aix_st_swap64(s64, d64)\
__asm__ volatile("stdbrx %1,0,%0" : : "r"(d64), "r"(s64))
#else
#define aix_ld_swap64(s64, d64) \
{ \
uint64_t *s4 = 0, h; /* initialize to zero for gcc warning */ \
\
__asm__("addi %0,%3,4;lwbrx %1,0,%3;lwbrx %2,0,%0;rldimi %1,%2,32,0"\
: "+r"(s4), "=r"(d64), "=r"(h) : "b"(s64)); \
}
#define aix_st_swap64(s64, d64) \
{ \
uint64_t *s4 = 0, h; /* initialize to zero for gcc warning */ \
h = (s64) >> 32; \
__asm__ volatile("addi %0,%3,4;stwbrx %1,0,%3;stwbrx %2,0,%0" \
: "+r"(s4) : "r"(s64), "r"(h), "b"(d64)); \
}
#endif /* 64BIT && PWR7 */
#else
#define aix_ld_swap64(s64, d64) \
{ \
uint32_t *s4 = 0, h, l; /* initialize to zero for gcc warning */\
__asm__("addi %0,%3,4;lwbrx %1,0,%3;lwbrx %2,0,%0" \
: "+r"(s4), "=r"(l), "=r"(h) : "b"(s64)); \
d64 = ((uint64_t)h<<32) | l; \
}
#define aix_st_swap64(s64, d64) \
{ \
uint32_t *s4 = 0, h, l; /* initialize to zero for gcc warning */\
l = (s64) & 0xfffffffful, h = (s64) >> 32; \
__asm__ volatile("addi %0,%3,4;stwbrx %1,0,%3;stwbrx %2,0,%0" \
: "+r"(s4) : "r"(l), "r"(h), "b"(d64)); \
}
#endif /* __64BIT__ */
#define aix_ld_swap32(s32, d32)\
__asm__("lwbrx %0,0,%1" : "=r"(d32) : "r"(s32))
#define aix_st_swap32(s32, d32)\
__asm__ volatile("stwbrx %1,0,%0" : : "r"(d32), "r"(s32))
#define ld_swap32(s, d) aix_ld_swap32(s, d)
#define st_swap32(s, d) aix_st_swap32(s, d)
#define ld_swap64(s, d) aix_ld_swap64(s, d)
#define st_swap64(s, d) aix_st_swap64(s, d)
#endif /* __PPC__ || _ARCH_PPC */
#if defined(__sparc)
#if !defined(__arch64__) && !defined(__sparcv8) && defined(__sparcv9)
#define __arch64__
#endif
#if defined(__GNUC__) || (defined(__SUNPRO_C) && __SUNPRO_C > 0x590)
/* need Sun Studio C 5.10 and above for GNU inline assembly */
#if defined(__arch64__)
#define sparc_ld_swap64(s64, d64) \
__asm__("ldxa [%1]0x88,%0" : "=r"(d64) : "r"(s64))
#define sparc_st_swap64(s64, d64) \
__asm__ volatile("stxa %0,[%1]0x88" : : "r"(s64), "r"(d64))
#define st_swap64(s, d) sparc_st_swap64(s, d)
#else
#define sparc_ld_swap64(s64, d64) \
{ \
uint32_t *s4, h, l; \
__asm__("add %3,4,%0\n\tlda [%3]0x88,%1\n\tlda [%0]0x88,%2" \
: "+r"(s4), "=r"(l), "=r"(h) : "r"(s64)); \
d64 = ((uint64_t)h<<32) | l; \
}
#define sparc_st_swap64(s64, d64) \
{ \
uint32_t *s4, h, l; \
l = (s64) & 0xfffffffful, h = (s64) >> 32; \
__asm__ volatile("add %3,4,%0\n\tsta %1,[%3]0x88\n\tsta %2,[%0]0x88"\
: "+r"(s4) : "r"(l), "r"(h), "r"(d64)); \
}
#endif /* sparc64 */
#define sparc_ld_swap32(s32, d32)\
__asm__("lda [%1]0x88,%0" : "=r"(d32) : "r"(s32))
#define sparc_st_swap32(s32, d32)\
__asm__ volatile("sta %0,[%1]0x88" : : "r"(s32), "r"(d32))
#define ld_swap32(s, d) sparc_ld_swap32(s, d)
#define st_swap32(s, d) sparc_st_swap32(s, d)
#define ld_swap64(s, d) sparc_ld_swap64(s, d)
#define st_swap64(s, d) sparc_st_swap64(s, d)
#endif /* GCC || Sun Studio C > 5.9 */
#endif /* sparc */
/* GCC fallback */
#if ((__GNUC__ >= 4) || defined(__PGIC__)) && !defined(ld_swap32)
#define ld_swap32(s, d) (d = __builtin_bswap32(*(s)))
#define st_swap32(s, d) (*(d) = __builtin_bswap32(s))
#endif /* GCC4/PGIC && !swap32 */
#if ((__GNUC__ >= 4) || defined(__PGIC__)) && !defined(ld_swap64)
#define ld_swap64(s, d) (d = __builtin_bswap64(*(s)))
#define st_swap64(s, d) (*(d) = __builtin_bswap64(s))
#endif /* GCC4/PGIC && !swap64 */
/* generic fallback */
#if !defined(ld_swap32)
#define ld_swap32(s, d) \
(d = (*(s) >> 24) | (*(s) >> 8 & 0xff00) | \
(*(s) << 8 & 0xff0000) | (*(s) << 24))
#define st_swap32(s, d) \
(*(d) = ((s) >> 24) | ((s) >> 8 & 0xff00) | \
((s) << 8 & 0xff0000) | ((s) << 24))
#endif
#if !defined(ld_swap64)
#define ld_swap64(s, d) \
(d = (*(s) >> 56) | (*(s) >> 40 & 0xff00) | \
(*(s) >> 24 & 0xff0000) | (*(s) >> 8 & 0xff000000) | \
(*(s) & 0xff000000) << 8 | (*(s) & 0xff0000) << 24 | \
(*(s) & 0xff00) << 40 | *(s) << 56)
#define st_swap64(s, d) \
(*(d) = ((s) >> 56) | ((s) >> 40 & 0xff00) | \
((s) >> 24 & 0xff0000) | ((s) >> 8 & 0xff000000) | \
((s) & 0xff000000) << 8 | ((s) & 0xff0000) << 24 | \
((s) & 0xff00) << 40 | (s) << 56)
#endif
#endif /* MACHINE_IS_BIG_ENDIAN */
#if defined(MACHINE_IS_LITTLE_ENDIAN)
/* replace swaps with simple assignments on little endian systems */
#undef ld_swap32
#undef st_swap32
#define ld_swap32(s, d) (d = *(s))
#define st_swap32(s, d) (*(d) = s)
#undef ld_swap64
#undef st_swap64
#define ld_swap64(s, d) (d = *(s))
#define st_swap64(s, d) (*(d) = s)
#endif /* MACHINE_IS_LITTLE_ENDIAN */
#endif /* _CRYPTO_EDONR_BYTEORDER_H */
diff --git a/module/icp/algs/modes/gcm.c b/module/icp/algs/modes/gcm.c
index 3a1660d93ab7..d9796cd0ed49 100644
--- a/module/icp/algs/modes/gcm.c
+++ b/module/icp/algs/modes/gcm.c
@@ -1,1588 +1,1588 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <modes/modes.h>
#include <sys/crypto/common.h>
#include <sys/crypto/icp.h>
#include <sys/crypto/impl.h>
#include <sys/byteorder.h>
#include <sys/simd.h>
#include <modes/gcm_impl.h>
#ifdef CAN_USE_GCM_ASM
#include <aes/aes_impl.h>
#endif
#define GHASH(c, d, t, o) \
xor_block((uint8_t *)(d), (uint8_t *)(c)->gcm_ghash); \
(o)->mul((uint64_t *)(void *)(c)->gcm_ghash, (c)->gcm_H, \
(uint64_t *)(void *)(t));
/* Select GCM implementation */
#define IMPL_FASTEST (UINT32_MAX)
#define IMPL_CYCLE (UINT32_MAX-1)
#ifdef CAN_USE_GCM_ASM
#define IMPL_AVX (UINT32_MAX-2)
#endif
#define GCM_IMPL_READ(i) (*(volatile uint32_t *) &(i))
static uint32_t icp_gcm_impl = IMPL_FASTEST;
static uint32_t user_sel_impl = IMPL_FASTEST;
#ifdef CAN_USE_GCM_ASM
/* Does the architecture we run on support the MOVBE instruction? */
boolean_t gcm_avx_can_use_movbe = B_FALSE;
/*
* Whether to use the optimized openssl gcm and ghash implementations.
* Set to true if module parameter icp_gcm_impl == "avx".
*/
static boolean_t gcm_use_avx = B_FALSE;
#define GCM_IMPL_USE_AVX (*(volatile boolean_t *)&gcm_use_avx)
extern boolean_t atomic_toggle_boolean_nv(volatile boolean_t *);
static inline boolean_t gcm_avx_will_work(void);
static inline void gcm_set_avx(boolean_t);
static inline boolean_t gcm_toggle_avx(void);
static inline size_t gcm_simd_get_htab_size(boolean_t);
static int gcm_mode_encrypt_contiguous_blocks_avx(gcm_ctx_t *, char *, size_t,
crypto_data_t *, size_t);
static int gcm_encrypt_final_avx(gcm_ctx_t *, crypto_data_t *, size_t);
static int gcm_decrypt_final_avx(gcm_ctx_t *, crypto_data_t *, size_t);
static int gcm_init_avx(gcm_ctx_t *, unsigned char *, size_t, unsigned char *,
size_t, size_t);
#endif /* ifdef CAN_USE_GCM_ASM */
/*
* Encrypt multiple blocks of data in GCM mode. Decrypt for GCM mode
* is done in another function.
*/
int
gcm_mode_encrypt_contiguous_blocks(gcm_ctx_t *ctx, char *data, size_t length,
crypto_data_t *out, size_t block_size,
int (*encrypt_block)(const void *, const uint8_t *, uint8_t *),
void (*copy_block)(uint8_t *, uint8_t *),
void (*xor_block)(uint8_t *, uint8_t *))
{
#ifdef CAN_USE_GCM_ASM
if (ctx->gcm_use_avx == B_TRUE)
return (gcm_mode_encrypt_contiguous_blocks_avx(
ctx, data, length, out, block_size));
#endif
const gcm_impl_ops_t *gops;
size_t remainder = length;
size_t need = 0;
uint8_t *datap = (uint8_t *)data;
uint8_t *blockp;
uint8_t *lastp;
void *iov_or_mp;
offset_t offset;
uint8_t *out_data_1;
uint8_t *out_data_2;
size_t out_data_1_len;
uint64_t counter;
uint64_t counter_mask = ntohll(0x00000000ffffffffULL);
if (length + ctx->gcm_remainder_len < block_size) {
/* accumulate bytes here and return */
bcopy(datap,
(uint8_t *)ctx->gcm_remainder + ctx->gcm_remainder_len,
length);
ctx->gcm_remainder_len += length;
if (ctx->gcm_copy_to == NULL) {
ctx->gcm_copy_to = datap;
}
return (CRYPTO_SUCCESS);
}
lastp = (uint8_t *)ctx->gcm_cb;
crypto_init_ptrs(out, &iov_or_mp, &offset);
gops = gcm_impl_get_ops();
do {
/* Unprocessed data from last call. */
if (ctx->gcm_remainder_len > 0) {
need = block_size - ctx->gcm_remainder_len;
if (need > remainder)
return (CRYPTO_DATA_LEN_RANGE);
bcopy(datap, &((uint8_t *)ctx->gcm_remainder)
[ctx->gcm_remainder_len], need);
blockp = (uint8_t *)ctx->gcm_remainder;
} else {
blockp = datap;
}
/*
* Increment counter. Counter bits are confined
* to the bottom 32 bits of the counter block.
*/
counter = ntohll(ctx->gcm_cb[1] & counter_mask);
counter = htonll(counter + 1);
counter &= counter_mask;
ctx->gcm_cb[1] = (ctx->gcm_cb[1] & ~counter_mask) | counter;
encrypt_block(ctx->gcm_keysched, (uint8_t *)ctx->gcm_cb,
(uint8_t *)ctx->gcm_tmp);
xor_block(blockp, (uint8_t *)ctx->gcm_tmp);
lastp = (uint8_t *)ctx->gcm_tmp;
ctx->gcm_processed_data_len += block_size;
crypto_get_ptrs(out, &iov_or_mp, &offset, &out_data_1,
&out_data_1_len, &out_data_2, block_size);
/* copy block to where it belongs */
if (out_data_1_len == block_size) {
copy_block(lastp, out_data_1);
} else {
bcopy(lastp, out_data_1, out_data_1_len);
if (out_data_2 != NULL) {
bcopy(lastp + out_data_1_len,
out_data_2,
block_size - out_data_1_len);
}
}
/* update offset */
out->cd_offset += block_size;
/* add ciphertext to the hash */
GHASH(ctx, ctx->gcm_tmp, ctx->gcm_ghash, gops);
/* Update pointer to next block of data to be processed. */
if (ctx->gcm_remainder_len != 0) {
datap += need;
ctx->gcm_remainder_len = 0;
} else {
datap += block_size;
}
remainder = (size_t)&data[length] - (size_t)datap;
/* Incomplete last block. */
if (remainder > 0 && remainder < block_size) {
bcopy(datap, ctx->gcm_remainder, remainder);
ctx->gcm_remainder_len = remainder;
ctx->gcm_copy_to = datap;
goto out;
}
ctx->gcm_copy_to = NULL;
} while (remainder > 0);
out:
return (CRYPTO_SUCCESS);
}
int
gcm_encrypt_final(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size,
int (*encrypt_block)(const void *, const uint8_t *, uint8_t *),
void (*copy_block)(uint8_t *, uint8_t *),
void (*xor_block)(uint8_t *, uint8_t *))
{
(void) copy_block;
#ifdef CAN_USE_GCM_ASM
if (ctx->gcm_use_avx == B_TRUE)
return (gcm_encrypt_final_avx(ctx, out, block_size));
#endif
const gcm_impl_ops_t *gops;
uint64_t counter_mask = ntohll(0x00000000ffffffffULL);
uint8_t *ghash, *macp = NULL;
int i, rv;
if (out->cd_length <
(ctx->gcm_remainder_len + ctx->gcm_tag_len)) {
return (CRYPTO_DATA_LEN_RANGE);
}
gops = gcm_impl_get_ops();
ghash = (uint8_t *)ctx->gcm_ghash;
if (ctx->gcm_remainder_len > 0) {
uint64_t counter;
uint8_t *tmpp = (uint8_t *)ctx->gcm_tmp;
/*
* Here is where we deal with data that is not a
* multiple of the block size.
*/
/*
* Increment counter.
*/
counter = ntohll(ctx->gcm_cb[1] & counter_mask);
counter = htonll(counter + 1);
counter &= counter_mask;
ctx->gcm_cb[1] = (ctx->gcm_cb[1] & ~counter_mask) | counter;
encrypt_block(ctx->gcm_keysched, (uint8_t *)ctx->gcm_cb,
(uint8_t *)ctx->gcm_tmp);
macp = (uint8_t *)ctx->gcm_remainder;
bzero(macp + ctx->gcm_remainder_len,
block_size - ctx->gcm_remainder_len);
/* XOR with counter block */
for (i = 0; i < ctx->gcm_remainder_len; i++) {
macp[i] ^= tmpp[i];
}
/* add ciphertext to the hash */
GHASH(ctx, macp, ghash, gops);
ctx->gcm_processed_data_len += ctx->gcm_remainder_len;
}
ctx->gcm_len_a_len_c[1] =
htonll(CRYPTO_BYTES2BITS(ctx->gcm_processed_data_len));
GHASH(ctx, ctx->gcm_len_a_len_c, ghash, gops);
encrypt_block(ctx->gcm_keysched, (uint8_t *)ctx->gcm_J0,
(uint8_t *)ctx->gcm_J0);
xor_block((uint8_t *)ctx->gcm_J0, ghash);
if (ctx->gcm_remainder_len > 0) {
rv = crypto_put_output_data(macp, out, ctx->gcm_remainder_len);
if (rv != CRYPTO_SUCCESS)
return (rv);
}
out->cd_offset += ctx->gcm_remainder_len;
ctx->gcm_remainder_len = 0;
rv = crypto_put_output_data(ghash, out, ctx->gcm_tag_len);
if (rv != CRYPTO_SUCCESS)
return (rv);
out->cd_offset += ctx->gcm_tag_len;
return (CRYPTO_SUCCESS);
}
/*
* This will only deal with decrypting the last block of the input that
* might not be a multiple of block length.
*/
static void
gcm_decrypt_incomplete_block(gcm_ctx_t *ctx, size_t block_size, size_t index,
int (*encrypt_block)(const void *, const uint8_t *, uint8_t *),
void (*xor_block)(uint8_t *, uint8_t *))
{
uint8_t *datap, *outp, *counterp;
uint64_t counter;
uint64_t counter_mask = ntohll(0x00000000ffffffffULL);
int i;
/*
* Increment counter.
* Counter bits are confined to the bottom 32 bits
*/
counter = ntohll(ctx->gcm_cb[1] & counter_mask);
counter = htonll(counter + 1);
counter &= counter_mask;
ctx->gcm_cb[1] = (ctx->gcm_cb[1] & ~counter_mask) | counter;
datap = (uint8_t *)ctx->gcm_remainder;
outp = &((ctx->gcm_pt_buf)[index]);
counterp = (uint8_t *)ctx->gcm_tmp;
/* authentication tag */
bzero((uint8_t *)ctx->gcm_tmp, block_size);
bcopy(datap, (uint8_t *)ctx->gcm_tmp, ctx->gcm_remainder_len);
/* add ciphertext to the hash */
GHASH(ctx, ctx->gcm_tmp, ctx->gcm_ghash, gcm_impl_get_ops());
/* decrypt remaining ciphertext */
encrypt_block(ctx->gcm_keysched, (uint8_t *)ctx->gcm_cb, counterp);
/* XOR with counter block */
for (i = 0; i < ctx->gcm_remainder_len; i++) {
outp[i] = datap[i] ^ counterp[i];
}
}
int
gcm_mode_decrypt_contiguous_blocks(gcm_ctx_t *ctx, char *data, size_t length,
crypto_data_t *out, size_t block_size,
int (*encrypt_block)(const void *, const uint8_t *, uint8_t *),
void (*copy_block)(uint8_t *, uint8_t *),
void (*xor_block)(uint8_t *, uint8_t *))
{
(void) out, (void) block_size, (void) encrypt_block, (void) copy_block,
(void) xor_block;
size_t new_len;
uint8_t *new;
/*
* Copy contiguous ciphertext input blocks to plaintext buffer.
* Ciphertext will be decrypted in the final.
*/
if (length > 0) {
new_len = ctx->gcm_pt_buf_len + length;
new = vmem_alloc(new_len, ctx->gcm_kmflag);
if (new == NULL) {
vmem_free(ctx->gcm_pt_buf, ctx->gcm_pt_buf_len);
ctx->gcm_pt_buf = NULL;
return (CRYPTO_HOST_MEMORY);
}
bcopy(ctx->gcm_pt_buf, new, ctx->gcm_pt_buf_len);
vmem_free(ctx->gcm_pt_buf, ctx->gcm_pt_buf_len);
ctx->gcm_pt_buf = new;
ctx->gcm_pt_buf_len = new_len;
bcopy(data, &ctx->gcm_pt_buf[ctx->gcm_processed_data_len],
length);
ctx->gcm_processed_data_len += length;
}
ctx->gcm_remainder_len = 0;
return (CRYPTO_SUCCESS);
}
int
gcm_decrypt_final(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size,
int (*encrypt_block)(const void *, const uint8_t *, uint8_t *),
void (*xor_block)(uint8_t *, uint8_t *))
{
#ifdef CAN_USE_GCM_ASM
if (ctx->gcm_use_avx == B_TRUE)
return (gcm_decrypt_final_avx(ctx, out, block_size));
#endif
const gcm_impl_ops_t *gops;
size_t pt_len;
size_t remainder;
uint8_t *ghash;
uint8_t *blockp;
uint8_t *cbp;
uint64_t counter;
uint64_t counter_mask = ntohll(0x00000000ffffffffULL);
int processed = 0, rv;
ASSERT(ctx->gcm_processed_data_len == ctx->gcm_pt_buf_len);
gops = gcm_impl_get_ops();
pt_len = ctx->gcm_processed_data_len - ctx->gcm_tag_len;
ghash = (uint8_t *)ctx->gcm_ghash;
blockp = ctx->gcm_pt_buf;
remainder = pt_len;
while (remainder > 0) {
/* Incomplete last block */
if (remainder < block_size) {
bcopy(blockp, ctx->gcm_remainder, remainder);
ctx->gcm_remainder_len = remainder;
/*
* not expecting anymore ciphertext, just
* compute plaintext for the remaining input
*/
gcm_decrypt_incomplete_block(ctx, block_size,
processed, encrypt_block, xor_block);
ctx->gcm_remainder_len = 0;
goto out;
}
/* add ciphertext to the hash */
GHASH(ctx, blockp, ghash, gops);
/*
* Increment counter.
* Counter bits are confined to the bottom 32 bits
*/
counter = ntohll(ctx->gcm_cb[1] & counter_mask);
counter = htonll(counter + 1);
counter &= counter_mask;
ctx->gcm_cb[1] = (ctx->gcm_cb[1] & ~counter_mask) | counter;
cbp = (uint8_t *)ctx->gcm_tmp;
encrypt_block(ctx->gcm_keysched, (uint8_t *)ctx->gcm_cb, cbp);
/* XOR with ciphertext */
xor_block(cbp, blockp);
processed += block_size;
blockp += block_size;
remainder -= block_size;
}
out:
ctx->gcm_len_a_len_c[1] = htonll(CRYPTO_BYTES2BITS(pt_len));
GHASH(ctx, ctx->gcm_len_a_len_c, ghash, gops);
encrypt_block(ctx->gcm_keysched, (uint8_t *)ctx->gcm_J0,
(uint8_t *)ctx->gcm_J0);
xor_block((uint8_t *)ctx->gcm_J0, ghash);
/* compare the input authentication tag with what we calculated */
if (bcmp(&ctx->gcm_pt_buf[pt_len], ghash, ctx->gcm_tag_len)) {
/* They don't match */
return (CRYPTO_INVALID_MAC);
} else {
rv = crypto_put_output_data(ctx->gcm_pt_buf, out, pt_len);
if (rv != CRYPTO_SUCCESS)
return (rv);
out->cd_offset += pt_len;
}
return (CRYPTO_SUCCESS);
}
static int
gcm_validate_args(CK_AES_GCM_PARAMS *gcm_param)
{
size_t tag_len;
/*
* Check the length of the authentication tag (in bits).
*/
tag_len = gcm_param->ulTagBits;
switch (tag_len) {
case 32:
case 64:
case 96:
case 104:
case 112:
case 120:
case 128:
break;
default:
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
if (gcm_param->ulIvLen == 0)
return (CRYPTO_MECHANISM_PARAM_INVALID);
return (CRYPTO_SUCCESS);
}
static void
gcm_format_initial_blocks(uchar_t *iv, ulong_t iv_len,
gcm_ctx_t *ctx, size_t block_size,
void (*copy_block)(uint8_t *, uint8_t *),
void (*xor_block)(uint8_t *, uint8_t *))
{
const gcm_impl_ops_t *gops;
uint8_t *cb;
ulong_t remainder = iv_len;
ulong_t processed = 0;
uint8_t *datap, *ghash;
uint64_t len_a_len_c[2];
gops = gcm_impl_get_ops();
ghash = (uint8_t *)ctx->gcm_ghash;
cb = (uint8_t *)ctx->gcm_cb;
if (iv_len == 12) {
bcopy(iv, cb, 12);
cb[12] = 0;
cb[13] = 0;
cb[14] = 0;
cb[15] = 1;
/* J0 will be used again in the final */
copy_block(cb, (uint8_t *)ctx->gcm_J0);
} else {
/* GHASH the IV */
do {
if (remainder < block_size) {
bzero(cb, block_size);
bcopy(&(iv[processed]), cb, remainder);
datap = (uint8_t *)cb;
remainder = 0;
} else {
datap = (uint8_t *)(&(iv[processed]));
processed += block_size;
remainder -= block_size;
}
GHASH(ctx, datap, ghash, gops);
} while (remainder > 0);
len_a_len_c[0] = 0;
len_a_len_c[1] = htonll(CRYPTO_BYTES2BITS(iv_len));
GHASH(ctx, len_a_len_c, ctx->gcm_J0, gops);
/* J0 will be used again in the final */
copy_block((uint8_t *)ctx->gcm_J0, (uint8_t *)cb);
}
}
static int
gcm_init(gcm_ctx_t *ctx, unsigned char *iv, size_t iv_len,
unsigned char *auth_data, size_t auth_data_len, size_t block_size,
int (*encrypt_block)(const void *, const uint8_t *, uint8_t *),
void (*copy_block)(uint8_t *, uint8_t *),
void (*xor_block)(uint8_t *, uint8_t *))
{
const gcm_impl_ops_t *gops;
uint8_t *ghash, *datap, *authp;
size_t remainder, processed;
/* encrypt zero block to get subkey H */
bzero(ctx->gcm_H, sizeof (ctx->gcm_H));
encrypt_block(ctx->gcm_keysched, (uint8_t *)ctx->gcm_H,
(uint8_t *)ctx->gcm_H);
gcm_format_initial_blocks(iv, iv_len, ctx, block_size,
copy_block, xor_block);
gops = gcm_impl_get_ops();
authp = (uint8_t *)ctx->gcm_tmp;
ghash = (uint8_t *)ctx->gcm_ghash;
bzero(authp, block_size);
bzero(ghash, block_size);
processed = 0;
remainder = auth_data_len;
do {
if (remainder < block_size) {
/*
* There's not a block full of data, pad rest of
* buffer with zero
*/
bzero(authp, block_size);
bcopy(&(auth_data[processed]), authp, remainder);
datap = (uint8_t *)authp;
remainder = 0;
} else {
datap = (uint8_t *)(&(auth_data[processed]));
processed += block_size;
remainder -= block_size;
}
/* add auth data to the hash */
GHASH(ctx, datap, ghash, gops);
} while (remainder > 0);
return (CRYPTO_SUCCESS);
}
/*
* The following function is called at encrypt or decrypt init time
* for AES GCM mode.
*
* Init the GCM context struct. Handle the cycle and avx implementations here.
*/
int
gcm_init_ctx(gcm_ctx_t *gcm_ctx, char *param, size_t block_size,
int (*encrypt_block)(const void *, const uint8_t *, uint8_t *),
void (*copy_block)(uint8_t *, uint8_t *),
void (*xor_block)(uint8_t *, uint8_t *))
{
int rv;
CK_AES_GCM_PARAMS *gcm_param;
if (param != NULL) {
gcm_param = (CK_AES_GCM_PARAMS *)(void *)param;
if ((rv = gcm_validate_args(gcm_param)) != 0) {
return (rv);
}
gcm_ctx->gcm_tag_len = gcm_param->ulTagBits;
gcm_ctx->gcm_tag_len >>= 3;
gcm_ctx->gcm_processed_data_len = 0;
/* these values are in bits */
gcm_ctx->gcm_len_a_len_c[0]
= htonll(CRYPTO_BYTES2BITS(gcm_param->ulAADLen));
rv = CRYPTO_SUCCESS;
gcm_ctx->gcm_flags |= GCM_MODE;
} else {
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
#ifdef CAN_USE_GCM_ASM
if (GCM_IMPL_READ(icp_gcm_impl) != IMPL_CYCLE) {
gcm_ctx->gcm_use_avx = GCM_IMPL_USE_AVX;
} else {
/*
* Handle the "cycle" implementation by creating avx and
* non-avx contexts alternately.
*/
gcm_ctx->gcm_use_avx = gcm_toggle_avx();
/*
* We don't handle byte swapped key schedules in the avx
* code path.
*/
aes_key_t *ks = (aes_key_t *)gcm_ctx->gcm_keysched;
if (ks->ops->needs_byteswap == B_TRUE) {
gcm_ctx->gcm_use_avx = B_FALSE;
}
/* Use the MOVBE and the BSWAP variants alternately. */
if (gcm_ctx->gcm_use_avx == B_TRUE &&
zfs_movbe_available() == B_TRUE) {
(void) atomic_toggle_boolean_nv(
(volatile boolean_t *)&gcm_avx_can_use_movbe);
}
}
/* Allocate Htab memory as needed. */
if (gcm_ctx->gcm_use_avx == B_TRUE) {
size_t htab_len = gcm_simd_get_htab_size(gcm_ctx->gcm_use_avx);
if (htab_len == 0) {
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
gcm_ctx->gcm_htab_len = htab_len;
gcm_ctx->gcm_Htable =
(uint64_t *)kmem_alloc(htab_len, gcm_ctx->gcm_kmflag);
if (gcm_ctx->gcm_Htable == NULL) {
return (CRYPTO_HOST_MEMORY);
}
}
/* Avx and non avx context initialization differs from here on. */
if (gcm_ctx->gcm_use_avx == B_FALSE) {
#endif /* ifdef CAN_USE_GCM_ASM */
if (gcm_init(gcm_ctx, gcm_param->pIv, gcm_param->ulIvLen,
gcm_param->pAAD, gcm_param->ulAADLen, block_size,
encrypt_block, copy_block, xor_block) != 0) {
rv = CRYPTO_MECHANISM_PARAM_INVALID;
}
#ifdef CAN_USE_GCM_ASM
} else {
if (gcm_init_avx(gcm_ctx, gcm_param->pIv, gcm_param->ulIvLen,
gcm_param->pAAD, gcm_param->ulAADLen, block_size) != 0) {
rv = CRYPTO_MECHANISM_PARAM_INVALID;
}
}
#endif /* ifdef CAN_USE_GCM_ASM */
return (rv);
}
int
gmac_init_ctx(gcm_ctx_t *gcm_ctx, char *param, size_t block_size,
int (*encrypt_block)(const void *, const uint8_t *, uint8_t *),
void (*copy_block)(uint8_t *, uint8_t *),
void (*xor_block)(uint8_t *, uint8_t *))
{
int rv;
CK_AES_GMAC_PARAMS *gmac_param;
if (param != NULL) {
gmac_param = (CK_AES_GMAC_PARAMS *)(void *)param;
gcm_ctx->gcm_tag_len = CRYPTO_BITS2BYTES(AES_GMAC_TAG_BITS);
gcm_ctx->gcm_processed_data_len = 0;
/* these values are in bits */
gcm_ctx->gcm_len_a_len_c[0]
= htonll(CRYPTO_BYTES2BITS(gmac_param->ulAADLen));
rv = CRYPTO_SUCCESS;
gcm_ctx->gcm_flags |= GMAC_MODE;
} else {
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
#ifdef CAN_USE_GCM_ASM
/*
* Handle the "cycle" implementation by creating avx and non avx
* contexts alternately.
*/
if (GCM_IMPL_READ(icp_gcm_impl) != IMPL_CYCLE) {
gcm_ctx->gcm_use_avx = GCM_IMPL_USE_AVX;
} else {
gcm_ctx->gcm_use_avx = gcm_toggle_avx();
}
/* We don't handle byte swapped key schedules in the avx code path. */
aes_key_t *ks = (aes_key_t *)gcm_ctx->gcm_keysched;
if (ks->ops->needs_byteswap == B_TRUE) {
gcm_ctx->gcm_use_avx = B_FALSE;
}
/* Allocate Htab memory as needed. */
if (gcm_ctx->gcm_use_avx == B_TRUE) {
size_t htab_len = gcm_simd_get_htab_size(gcm_ctx->gcm_use_avx);
if (htab_len == 0) {
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
gcm_ctx->gcm_htab_len = htab_len;
gcm_ctx->gcm_Htable =
(uint64_t *)kmem_alloc(htab_len, gcm_ctx->gcm_kmflag);
if (gcm_ctx->gcm_Htable == NULL) {
return (CRYPTO_HOST_MEMORY);
}
}
/* Avx and non avx context initialization differs from here on. */
if (gcm_ctx->gcm_use_avx == B_FALSE) {
#endif /* ifdef CAN_USE_GCM_ASM */
if (gcm_init(gcm_ctx, gmac_param->pIv, AES_GMAC_IV_LEN,
gmac_param->pAAD, gmac_param->ulAADLen, block_size,
encrypt_block, copy_block, xor_block) != 0) {
rv = CRYPTO_MECHANISM_PARAM_INVALID;
}
#ifdef CAN_USE_GCM_ASM
} else {
if (gcm_init_avx(gcm_ctx, gmac_param->pIv, AES_GMAC_IV_LEN,
gmac_param->pAAD, gmac_param->ulAADLen, block_size) != 0) {
rv = CRYPTO_MECHANISM_PARAM_INVALID;
}
}
#endif /* ifdef CAN_USE_GCM_ASM */
return (rv);
}
void *
gcm_alloc_ctx(int kmflag)
{
gcm_ctx_t *gcm_ctx;
if ((gcm_ctx = kmem_zalloc(sizeof (gcm_ctx_t), kmflag)) == NULL)
return (NULL);
gcm_ctx->gcm_flags = GCM_MODE;
return (gcm_ctx);
}
void *
gmac_alloc_ctx(int kmflag)
{
gcm_ctx_t *gcm_ctx;
if ((gcm_ctx = kmem_zalloc(sizeof (gcm_ctx_t), kmflag)) == NULL)
return (NULL);
gcm_ctx->gcm_flags = GMAC_MODE;
return (gcm_ctx);
}
void
gcm_set_kmflag(gcm_ctx_t *ctx, int kmflag)
{
ctx->gcm_kmflag = kmflag;
}
/* GCM implementation that contains the fastest methods */
static gcm_impl_ops_t gcm_fastest_impl = {
.name = "fastest"
};
/* All compiled in implementations */
-const gcm_impl_ops_t *gcm_all_impl[] = {
+static const gcm_impl_ops_t *gcm_all_impl[] = {
&gcm_generic_impl,
#if defined(__x86_64) && defined(HAVE_PCLMULQDQ)
&gcm_pclmulqdq_impl,
#endif
};
/* Indicate that benchmark has been completed */
static boolean_t gcm_impl_initialized = B_FALSE;
/* Hold all supported implementations */
static size_t gcm_supp_impl_cnt = 0;
static gcm_impl_ops_t *gcm_supp_impl[ARRAY_SIZE(gcm_all_impl)];
/*
* Returns the GCM operations for encrypt/decrypt/key setup. When a
* SIMD implementation is not allowed in the current context, then
* fallback to the fastest generic implementation.
*/
const gcm_impl_ops_t *
gcm_impl_get_ops()
{
if (!kfpu_allowed())
return (&gcm_generic_impl);
const gcm_impl_ops_t *ops = NULL;
const uint32_t impl = GCM_IMPL_READ(icp_gcm_impl);
switch (impl) {
case IMPL_FASTEST:
ASSERT(gcm_impl_initialized);
ops = &gcm_fastest_impl;
break;
case IMPL_CYCLE:
/* Cycle through supported implementations */
ASSERT(gcm_impl_initialized);
ASSERT3U(gcm_supp_impl_cnt, >, 0);
static size_t cycle_impl_idx = 0;
size_t idx = (++cycle_impl_idx) % gcm_supp_impl_cnt;
ops = gcm_supp_impl[idx];
break;
#ifdef CAN_USE_GCM_ASM
case IMPL_AVX:
/*
* Make sure that we return a valid implementation while
* switching to the avx implementation since there still
* may be unfinished non-avx contexts around.
*/
ops = &gcm_generic_impl;
break;
#endif
default:
ASSERT3U(impl, <, gcm_supp_impl_cnt);
ASSERT3U(gcm_supp_impl_cnt, >, 0);
if (impl < ARRAY_SIZE(gcm_all_impl))
ops = gcm_supp_impl[impl];
break;
}
ASSERT3P(ops, !=, NULL);
return (ops);
}
/*
* Initialize all supported implementations.
*/
void
gcm_impl_init(void)
{
gcm_impl_ops_t *curr_impl;
int i, c;
/* Move supported implementations into gcm_supp_impls */
for (i = 0, c = 0; i < ARRAY_SIZE(gcm_all_impl); i++) {
curr_impl = (gcm_impl_ops_t *)gcm_all_impl[i];
if (curr_impl->is_supported())
gcm_supp_impl[c++] = (gcm_impl_ops_t *)curr_impl;
}
gcm_supp_impl_cnt = c;
/*
* Set the fastest implementation given the assumption that the
* hardware accelerated version is the fastest.
*/
#if defined(__x86_64) && defined(HAVE_PCLMULQDQ)
if (gcm_pclmulqdq_impl.is_supported()) {
memcpy(&gcm_fastest_impl, &gcm_pclmulqdq_impl,
sizeof (gcm_fastest_impl));
} else
#endif
{
memcpy(&gcm_fastest_impl, &gcm_generic_impl,
sizeof (gcm_fastest_impl));
}
strlcpy(gcm_fastest_impl.name, "fastest", GCM_IMPL_NAME_MAX);
#ifdef CAN_USE_GCM_ASM
/*
* Use the avx implementation if it's available and the implementation
* hasn't changed from its default value of fastest on module load.
*/
if (gcm_avx_will_work()) {
#ifdef HAVE_MOVBE
if (zfs_movbe_available() == B_TRUE) {
atomic_swap_32(&gcm_avx_can_use_movbe, B_TRUE);
}
#endif
if (GCM_IMPL_READ(user_sel_impl) == IMPL_FASTEST) {
gcm_set_avx(B_TRUE);
}
}
#endif
/* Finish initialization */
atomic_swap_32(&icp_gcm_impl, user_sel_impl);
gcm_impl_initialized = B_TRUE;
}
static const struct {
char *name;
uint32_t sel;
} gcm_impl_opts[] = {
{ "cycle", IMPL_CYCLE },
{ "fastest", IMPL_FASTEST },
#ifdef CAN_USE_GCM_ASM
{ "avx", IMPL_AVX },
#endif
};
/*
* Function sets desired gcm implementation.
*
* If we are called before init(), user preference will be saved in
* user_sel_impl, and applied in later init() call. This occurs when module
* parameter is specified on module load. Otherwise, directly update
* icp_gcm_impl.
*
* @val Name of gcm implementation to use
* @param Unused.
*/
int
gcm_impl_set(const char *val)
{
int err = -EINVAL;
char req_name[GCM_IMPL_NAME_MAX];
uint32_t impl = GCM_IMPL_READ(user_sel_impl);
size_t i;
/* sanitize input */
i = strnlen(val, GCM_IMPL_NAME_MAX);
if (i == 0 || i >= GCM_IMPL_NAME_MAX)
return (err);
strlcpy(req_name, val, GCM_IMPL_NAME_MAX);
while (i > 0 && isspace(req_name[i-1]))
i--;
req_name[i] = '\0';
/* Check mandatory options */
for (i = 0; i < ARRAY_SIZE(gcm_impl_opts); i++) {
#ifdef CAN_USE_GCM_ASM
/* Ignore avx implementation if it won't work. */
if (gcm_impl_opts[i].sel == IMPL_AVX && !gcm_avx_will_work()) {
continue;
}
#endif
if (strcmp(req_name, gcm_impl_opts[i].name) == 0) {
impl = gcm_impl_opts[i].sel;
err = 0;
break;
}
}
/* check all supported impl if init() was already called */
if (err != 0 && gcm_impl_initialized) {
/* check all supported implementations */
for (i = 0; i < gcm_supp_impl_cnt; i++) {
if (strcmp(req_name, gcm_supp_impl[i]->name) == 0) {
impl = i;
err = 0;
break;
}
}
}
#ifdef CAN_USE_GCM_ASM
/*
* Use the avx implementation if available and the requested one is
* avx or fastest.
*/
if (gcm_avx_will_work() == B_TRUE &&
(impl == IMPL_AVX || impl == IMPL_FASTEST)) {
gcm_set_avx(B_TRUE);
} else {
gcm_set_avx(B_FALSE);
}
#endif
if (err == 0) {
if (gcm_impl_initialized)
atomic_swap_32(&icp_gcm_impl, impl);
else
atomic_swap_32(&user_sel_impl, impl);
}
return (err);
}
#if defined(_KERNEL) && defined(__linux__)
static int
icp_gcm_impl_set(const char *val, zfs_kernel_param_t *kp)
{
return (gcm_impl_set(val));
}
static int
icp_gcm_impl_get(char *buffer, zfs_kernel_param_t *kp)
{
int i, cnt = 0;
char *fmt;
const uint32_t impl = GCM_IMPL_READ(icp_gcm_impl);
ASSERT(gcm_impl_initialized);
/* list mandatory options */
for (i = 0; i < ARRAY_SIZE(gcm_impl_opts); i++) {
#ifdef CAN_USE_GCM_ASM
/* Ignore avx implementation if it won't work. */
if (gcm_impl_opts[i].sel == IMPL_AVX && !gcm_avx_will_work()) {
continue;
}
#endif
fmt = (impl == gcm_impl_opts[i].sel) ? "[%s] " : "%s ";
cnt += sprintf(buffer + cnt, fmt, gcm_impl_opts[i].name);
}
/* list all supported implementations */
for (i = 0; i < gcm_supp_impl_cnt; i++) {
fmt = (i == impl) ? "[%s] " : "%s ";
cnt += sprintf(buffer + cnt, fmt, gcm_supp_impl[i]->name);
}
return (cnt);
}
module_param_call(icp_gcm_impl, icp_gcm_impl_set, icp_gcm_impl_get,
NULL, 0644);
MODULE_PARM_DESC(icp_gcm_impl, "Select gcm implementation.");
#endif /* defined(__KERNEL) */
#ifdef CAN_USE_GCM_ASM
#define GCM_BLOCK_LEN 16
/*
* The openssl asm routines are 6x aggregated and need that many bytes
* at minimum.
*/
#define GCM_AVX_MIN_DECRYPT_BYTES (GCM_BLOCK_LEN * 6)
#define GCM_AVX_MIN_ENCRYPT_BYTES (GCM_BLOCK_LEN * 6 * 3)
/*
* Ensure the chunk size is reasonable since we are allocating a
* GCM_AVX_MAX_CHUNK_SIZEd buffer and disabling preemption and interrupts.
*/
#define GCM_AVX_MAX_CHUNK_SIZE \
(((128*1024)/GCM_AVX_MIN_DECRYPT_BYTES) * GCM_AVX_MIN_DECRYPT_BYTES)
-/* Get the chunk size module parameter. */
-#define GCM_CHUNK_SIZE_READ *(volatile uint32_t *) &gcm_avx_chunk_size
-
/* Clear the FPU registers since they hold sensitive internal state. */
#define clear_fpu_regs() clear_fpu_regs_avx()
#define GHASH_AVX(ctx, in, len) \
gcm_ghash_avx((ctx)->gcm_ghash, (const uint64_t *)(ctx)->gcm_Htable, \
in, len)
#define gcm_incr_counter_block(ctx) gcm_incr_counter_block_by(ctx, 1)
+/* Get the chunk size module parameter. */
+#define GCM_CHUNK_SIZE_READ *(volatile uint32_t *) &gcm_avx_chunk_size
+
/*
* Module parameter: number of bytes to process at once while owning the FPU.
* Rounded down to the next GCM_AVX_MIN_DECRYPT_BYTES byte boundary and is
* ensured to be greater or equal than GCM_AVX_MIN_DECRYPT_BYTES.
*/
static uint32_t gcm_avx_chunk_size =
((32 * 1024) / GCM_AVX_MIN_DECRYPT_BYTES) * GCM_AVX_MIN_DECRYPT_BYTES;
extern void clear_fpu_regs_avx(void);
extern void gcm_xor_avx(const uint8_t *src, uint8_t *dst);
extern void aes_encrypt_intel(const uint32_t rk[], int nr,
const uint32_t pt[4], uint32_t ct[4]);
extern void gcm_init_htab_avx(uint64_t *Htable, const uint64_t H[2]);
extern void gcm_ghash_avx(uint64_t ghash[2], const uint64_t *Htable,
const uint8_t *in, size_t len);
extern size_t aesni_gcm_encrypt(const uint8_t *, uint8_t *, size_t,
const void *, uint64_t *, uint64_t *);
extern size_t aesni_gcm_decrypt(const uint8_t *, uint8_t *, size_t,
const void *, uint64_t *, uint64_t *);
static inline boolean_t
gcm_avx_will_work(void)
{
/* Avx should imply aes-ni and pclmulqdq, but make sure anyhow. */
return (kfpu_allowed() &&
zfs_avx_available() && zfs_aes_available() &&
zfs_pclmulqdq_available());
}
static inline void
gcm_set_avx(boolean_t val)
{
if (gcm_avx_will_work() == B_TRUE) {
atomic_swap_32(&gcm_use_avx, val);
}
}
static inline boolean_t
gcm_toggle_avx(void)
{
if (gcm_avx_will_work() == B_TRUE) {
return (atomic_toggle_boolean_nv(&GCM_IMPL_USE_AVX));
} else {
return (B_FALSE);
}
}
static inline size_t
gcm_simd_get_htab_size(boolean_t simd_mode)
{
switch (simd_mode) {
case B_TRUE:
return (2 * 6 * 2 * sizeof (uint64_t));
default:
return (0);
}
}
/*
* Clear sensitive data in the context.
*
* ctx->gcm_remainder may contain a plaintext remainder. ctx->gcm_H and
* ctx->gcm_Htable contain the hash sub key which protects authentication.
*
* Although extremely unlikely, ctx->gcm_J0 and ctx->gcm_tmp could be used for
* a known plaintext attack, they consists of the IV and the first and last
* counter respectively. If they should be cleared is debatable.
*/
static inline void
gcm_clear_ctx(gcm_ctx_t *ctx)
{
bzero(ctx->gcm_remainder, sizeof (ctx->gcm_remainder));
bzero(ctx->gcm_H, sizeof (ctx->gcm_H));
bzero(ctx->gcm_J0, sizeof (ctx->gcm_J0));
bzero(ctx->gcm_tmp, sizeof (ctx->gcm_tmp));
}
/* Increment the GCM counter block by n. */
static inline void
gcm_incr_counter_block_by(gcm_ctx_t *ctx, int n)
{
uint64_t counter_mask = ntohll(0x00000000ffffffffULL);
uint64_t counter = ntohll(ctx->gcm_cb[1] & counter_mask);
counter = htonll(counter + n);
counter &= counter_mask;
ctx->gcm_cb[1] = (ctx->gcm_cb[1] & ~counter_mask) | counter;
}
/*
* Encrypt multiple blocks of data in GCM mode.
* This is done in gcm_avx_chunk_size chunks, utilizing AVX assembler routines
* if possible. While processing a chunk the FPU is "locked".
*/
static int
gcm_mode_encrypt_contiguous_blocks_avx(gcm_ctx_t *ctx, char *data,
size_t length, crypto_data_t *out, size_t block_size)
{
size_t bleft = length;
size_t need = 0;
size_t done = 0;
uint8_t *datap = (uint8_t *)data;
size_t chunk_size = (size_t)GCM_CHUNK_SIZE_READ;
const aes_key_t *key = ((aes_key_t *)ctx->gcm_keysched);
uint64_t *ghash = ctx->gcm_ghash;
uint64_t *cb = ctx->gcm_cb;
uint8_t *ct_buf = NULL;
uint8_t *tmp = (uint8_t *)ctx->gcm_tmp;
int rv = CRYPTO_SUCCESS;
ASSERT(block_size == GCM_BLOCK_LEN);
/*
* If the last call left an incomplete block, try to fill
* it first.
*/
if (ctx->gcm_remainder_len > 0) {
need = block_size - ctx->gcm_remainder_len;
if (length < need) {
/* Accumulate bytes here and return. */
bcopy(datap, (uint8_t *)ctx->gcm_remainder +
ctx->gcm_remainder_len, length);
ctx->gcm_remainder_len += length;
if (ctx->gcm_copy_to == NULL) {
ctx->gcm_copy_to = datap;
}
return (CRYPTO_SUCCESS);
} else {
/* Complete incomplete block. */
bcopy(datap, (uint8_t *)ctx->gcm_remainder +
ctx->gcm_remainder_len, need);
ctx->gcm_copy_to = NULL;
}
}
/* Allocate a buffer to encrypt to if there is enough input. */
if (bleft >= GCM_AVX_MIN_ENCRYPT_BYTES) {
ct_buf = vmem_alloc(chunk_size, ctx->gcm_kmflag);
if (ct_buf == NULL) {
return (CRYPTO_HOST_MEMORY);
}
}
/* If we completed an incomplete block, encrypt and write it out. */
if (ctx->gcm_remainder_len > 0) {
kfpu_begin();
aes_encrypt_intel(key->encr_ks.ks32, key->nr,
(const uint32_t *)cb, (uint32_t *)tmp);
gcm_xor_avx((const uint8_t *) ctx->gcm_remainder, tmp);
GHASH_AVX(ctx, tmp, block_size);
clear_fpu_regs();
kfpu_end();
rv = crypto_put_output_data(tmp, out, block_size);
out->cd_offset += block_size;
gcm_incr_counter_block(ctx);
ctx->gcm_processed_data_len += block_size;
bleft -= need;
datap += need;
ctx->gcm_remainder_len = 0;
}
/* Do the bulk encryption in chunk_size blocks. */
for (; bleft >= chunk_size; bleft -= chunk_size) {
kfpu_begin();
done = aesni_gcm_encrypt(
datap, ct_buf, chunk_size, key, cb, ghash);
clear_fpu_regs();
kfpu_end();
if (done != chunk_size) {
rv = CRYPTO_FAILED;
goto out_nofpu;
}
rv = crypto_put_output_data(ct_buf, out, chunk_size);
if (rv != CRYPTO_SUCCESS) {
goto out_nofpu;
}
out->cd_offset += chunk_size;
datap += chunk_size;
ctx->gcm_processed_data_len += chunk_size;
}
/* Check if we are already done. */
if (bleft == 0) {
goto out_nofpu;
}
/* Bulk encrypt the remaining data. */
kfpu_begin();
if (bleft >= GCM_AVX_MIN_ENCRYPT_BYTES) {
done = aesni_gcm_encrypt(datap, ct_buf, bleft, key, cb, ghash);
if (done == 0) {
rv = CRYPTO_FAILED;
goto out;
}
rv = crypto_put_output_data(ct_buf, out, done);
if (rv != CRYPTO_SUCCESS) {
goto out;
}
out->cd_offset += done;
ctx->gcm_processed_data_len += done;
datap += done;
bleft -= done;
}
/* Less than GCM_AVX_MIN_ENCRYPT_BYTES remain, operate on blocks. */
while (bleft > 0) {
if (bleft < block_size) {
bcopy(datap, ctx->gcm_remainder, bleft);
ctx->gcm_remainder_len = bleft;
ctx->gcm_copy_to = datap;
goto out;
}
/* Encrypt, hash and write out. */
aes_encrypt_intel(key->encr_ks.ks32, key->nr,
(const uint32_t *)cb, (uint32_t *)tmp);
gcm_xor_avx(datap, tmp);
GHASH_AVX(ctx, tmp, block_size);
rv = crypto_put_output_data(tmp, out, block_size);
if (rv != CRYPTO_SUCCESS) {
goto out;
}
out->cd_offset += block_size;
gcm_incr_counter_block(ctx);
ctx->gcm_processed_data_len += block_size;
datap += block_size;
bleft -= block_size;
}
out:
clear_fpu_regs();
kfpu_end();
out_nofpu:
if (ct_buf != NULL) {
vmem_free(ct_buf, chunk_size);
}
return (rv);
}
/*
* Finalize the encryption: Zero fill, encrypt, hash and write out an eventual
* incomplete last block. Encrypt the ICB. Calculate the tag and write it out.
*/
static int
gcm_encrypt_final_avx(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size)
{
uint8_t *ghash = (uint8_t *)ctx->gcm_ghash;
uint32_t *J0 = (uint32_t *)ctx->gcm_J0;
uint8_t *remainder = (uint8_t *)ctx->gcm_remainder;
size_t rem_len = ctx->gcm_remainder_len;
const void *keysched = ((aes_key_t *)ctx->gcm_keysched)->encr_ks.ks32;
int aes_rounds = ((aes_key_t *)keysched)->nr;
int rv;
ASSERT(block_size == GCM_BLOCK_LEN);
if (out->cd_length < (rem_len + ctx->gcm_tag_len)) {
return (CRYPTO_DATA_LEN_RANGE);
}
kfpu_begin();
/* Pad last incomplete block with zeros, encrypt and hash. */
if (rem_len > 0) {
uint8_t *tmp = (uint8_t *)ctx->gcm_tmp;
const uint32_t *cb = (uint32_t *)ctx->gcm_cb;
aes_encrypt_intel(keysched, aes_rounds, cb, (uint32_t *)tmp);
bzero(remainder + rem_len, block_size - rem_len);
for (int i = 0; i < rem_len; i++) {
remainder[i] ^= tmp[i];
}
GHASH_AVX(ctx, remainder, block_size);
ctx->gcm_processed_data_len += rem_len;
/* No need to increment counter_block, it's the last block. */
}
/* Finish tag. */
ctx->gcm_len_a_len_c[1] =
htonll(CRYPTO_BYTES2BITS(ctx->gcm_processed_data_len));
GHASH_AVX(ctx, (const uint8_t *)ctx->gcm_len_a_len_c, block_size);
aes_encrypt_intel(keysched, aes_rounds, J0, J0);
gcm_xor_avx((uint8_t *)J0, ghash);
clear_fpu_regs();
kfpu_end();
/* Output remainder. */
if (rem_len > 0) {
rv = crypto_put_output_data(remainder, out, rem_len);
if (rv != CRYPTO_SUCCESS)
return (rv);
}
out->cd_offset += rem_len;
ctx->gcm_remainder_len = 0;
rv = crypto_put_output_data(ghash, out, ctx->gcm_tag_len);
if (rv != CRYPTO_SUCCESS)
return (rv);
out->cd_offset += ctx->gcm_tag_len;
/* Clear sensitive data in the context before returning. */
gcm_clear_ctx(ctx);
return (CRYPTO_SUCCESS);
}
/*
* Finalize decryption: We just have accumulated crypto text, so now we
* decrypt it here inplace.
*/
static int
gcm_decrypt_final_avx(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size)
{
ASSERT3U(ctx->gcm_processed_data_len, ==, ctx->gcm_pt_buf_len);
ASSERT3U(block_size, ==, 16);
size_t chunk_size = (size_t)GCM_CHUNK_SIZE_READ;
size_t pt_len = ctx->gcm_processed_data_len - ctx->gcm_tag_len;
uint8_t *datap = ctx->gcm_pt_buf;
const aes_key_t *key = ((aes_key_t *)ctx->gcm_keysched);
uint32_t *cb = (uint32_t *)ctx->gcm_cb;
uint64_t *ghash = ctx->gcm_ghash;
uint32_t *tmp = (uint32_t *)ctx->gcm_tmp;
int rv = CRYPTO_SUCCESS;
size_t bleft, done;
/*
* Decrypt in chunks of gcm_avx_chunk_size, which is asserted to be
* greater or equal than GCM_AVX_MIN_ENCRYPT_BYTES, and a multiple of
* GCM_AVX_MIN_DECRYPT_BYTES.
*/
for (bleft = pt_len; bleft >= chunk_size; bleft -= chunk_size) {
kfpu_begin();
done = aesni_gcm_decrypt(datap, datap, chunk_size,
(const void *)key, ctx->gcm_cb, ghash);
clear_fpu_regs();
kfpu_end();
if (done != chunk_size) {
return (CRYPTO_FAILED);
}
datap += done;
}
/* Decrypt remainder, which is less than chunk size, in one go. */
kfpu_begin();
if (bleft >= GCM_AVX_MIN_DECRYPT_BYTES) {
done = aesni_gcm_decrypt(datap, datap, bleft,
(const void *)key, ctx->gcm_cb, ghash);
if (done == 0) {
clear_fpu_regs();
kfpu_end();
return (CRYPTO_FAILED);
}
datap += done;
bleft -= done;
}
ASSERT(bleft < GCM_AVX_MIN_DECRYPT_BYTES);
/*
* Now less than GCM_AVX_MIN_DECRYPT_BYTES bytes remain,
* decrypt them block by block.
*/
while (bleft > 0) {
/* Incomplete last block. */
if (bleft < block_size) {
uint8_t *lastb = (uint8_t *)ctx->gcm_remainder;
bzero(lastb, block_size);
bcopy(datap, lastb, bleft);
/* The GCM processing. */
GHASH_AVX(ctx, lastb, block_size);
aes_encrypt_intel(key->encr_ks.ks32, key->nr, cb, tmp);
for (size_t i = 0; i < bleft; i++) {
datap[i] = lastb[i] ^ ((uint8_t *)tmp)[i];
}
break;
}
/* The GCM processing. */
GHASH_AVX(ctx, datap, block_size);
aes_encrypt_intel(key->encr_ks.ks32, key->nr, cb, tmp);
gcm_xor_avx((uint8_t *)tmp, datap);
gcm_incr_counter_block(ctx);
datap += block_size;
bleft -= block_size;
}
if (rv != CRYPTO_SUCCESS) {
clear_fpu_regs();
kfpu_end();
return (rv);
}
/* Decryption done, finish the tag. */
ctx->gcm_len_a_len_c[1] = htonll(CRYPTO_BYTES2BITS(pt_len));
GHASH_AVX(ctx, (uint8_t *)ctx->gcm_len_a_len_c, block_size);
aes_encrypt_intel(key->encr_ks.ks32, key->nr, (uint32_t *)ctx->gcm_J0,
(uint32_t *)ctx->gcm_J0);
gcm_xor_avx((uint8_t *)ctx->gcm_J0, (uint8_t *)ghash);
/* We are done with the FPU, restore its state. */
clear_fpu_regs();
kfpu_end();
/* Compare the input authentication tag with what we calculated. */
if (bcmp(&ctx->gcm_pt_buf[pt_len], ghash, ctx->gcm_tag_len)) {
/* They don't match. */
return (CRYPTO_INVALID_MAC);
}
rv = crypto_put_output_data(ctx->gcm_pt_buf, out, pt_len);
if (rv != CRYPTO_SUCCESS) {
return (rv);
}
out->cd_offset += pt_len;
gcm_clear_ctx(ctx);
return (CRYPTO_SUCCESS);
}
/*
* Initialize the GCM params H, Htabtle and the counter block. Save the
* initial counter block.
*/
static int
gcm_init_avx(gcm_ctx_t *ctx, unsigned char *iv, size_t iv_len,
unsigned char *auth_data, size_t auth_data_len, size_t block_size)
{
uint8_t *cb = (uint8_t *)ctx->gcm_cb;
uint64_t *H = ctx->gcm_H;
const void *keysched = ((aes_key_t *)ctx->gcm_keysched)->encr_ks.ks32;
int aes_rounds = ((aes_key_t *)ctx->gcm_keysched)->nr;
uint8_t *datap = auth_data;
size_t chunk_size = (size_t)GCM_CHUNK_SIZE_READ;
size_t bleft;
ASSERT(block_size == GCM_BLOCK_LEN);
/* Init H (encrypt zero block) and create the initial counter block. */
bzero(ctx->gcm_ghash, sizeof (ctx->gcm_ghash));
bzero(H, sizeof (ctx->gcm_H));
kfpu_begin();
aes_encrypt_intel(keysched, aes_rounds,
(const uint32_t *)H, (uint32_t *)H);
gcm_init_htab_avx(ctx->gcm_Htable, H);
if (iv_len == 12) {
bcopy(iv, cb, 12);
cb[12] = 0;
cb[13] = 0;
cb[14] = 0;
cb[15] = 1;
/* We need the ICB later. */
bcopy(cb, ctx->gcm_J0, sizeof (ctx->gcm_J0));
} else {
/*
* Most consumers use 12 byte IVs, so it's OK to use the
* original routines for other IV sizes, just avoid nesting
* kfpu_begin calls.
*/
clear_fpu_regs();
kfpu_end();
gcm_format_initial_blocks(iv, iv_len, ctx, block_size,
aes_copy_block, aes_xor_block);
kfpu_begin();
}
/* Openssl post increments the counter, adjust for that. */
gcm_incr_counter_block(ctx);
/* Ghash AAD in chunk_size blocks. */
for (bleft = auth_data_len; bleft >= chunk_size; bleft -= chunk_size) {
GHASH_AVX(ctx, datap, chunk_size);
datap += chunk_size;
clear_fpu_regs();
kfpu_end();
kfpu_begin();
}
/* Ghash the remainder and handle possible incomplete GCM block. */
if (bleft > 0) {
size_t incomp = bleft % block_size;
bleft -= incomp;
if (bleft > 0) {
GHASH_AVX(ctx, datap, bleft);
datap += bleft;
}
if (incomp > 0) {
/* Zero pad and hash incomplete last block. */
uint8_t *authp = (uint8_t *)ctx->gcm_tmp;
bzero(authp, block_size);
bcopy(datap, authp, incomp);
GHASH_AVX(ctx, authp, block_size);
}
}
clear_fpu_regs();
kfpu_end();
return (CRYPTO_SUCCESS);
}
#if defined(_KERNEL)
static int
icp_gcm_avx_set_chunk_size(const char *buf, zfs_kernel_param_t *kp)
{
unsigned long val;
char val_rounded[16];
int error = 0;
error = kstrtoul(buf, 0, &val);
if (error)
return (error);
val = (val / GCM_AVX_MIN_DECRYPT_BYTES) * GCM_AVX_MIN_DECRYPT_BYTES;
if (val < GCM_AVX_MIN_ENCRYPT_BYTES || val > GCM_AVX_MAX_CHUNK_SIZE)
return (-EINVAL);
snprintf(val_rounded, 16, "%u", (uint32_t)val);
error = param_set_uint(val_rounded, kp);
return (error);
}
module_param_call(icp_gcm_avx_chunk_size, icp_gcm_avx_set_chunk_size,
param_get_uint, &gcm_avx_chunk_size, 0644);
MODULE_PARM_DESC(icp_gcm_avx_chunk_size,
"How many bytes to process while owning the FPU");
#endif /* defined(__KERNEL) */
#endif /* ifdef CAN_USE_GCM_ASM */
diff --git a/module/icp/algs/sha2/sha2.c b/module/icp/algs/sha2/sha2.c
index 75f6a3c1af4b..6f1e9b7193d4 100644
--- a/module/icp/algs/sha2/sha2.c
+++ b/module/icp/algs/sha2/sha2.c
@@ -1,956 +1,956 @@
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright 2013 Saso Kiselkov. All rights reserved.
*/
/*
* The basic framework for this code came from the reference
* implementation for MD5. That implementation is Copyright (C)
* 1991-2, RSA Data Security, Inc. Created 1991. All rights reserved.
*
* License to copy and use this software is granted provided that it
* is identified as the "RSA Data Security, Inc. MD5 Message-Digest
* Algorithm" in all material mentioning or referencing this software
* or this function.
*
* License is also granted to make and use derivative works provided
* that such works are identified as "derived from the RSA Data
* Security, Inc. MD5 Message-Digest Algorithm" in all material
* mentioning or referencing the derived work.
*
* RSA Data Security, Inc. makes no representations concerning either
* the merchantability of this software or the suitability of this
* software for any particular purpose. It is provided "as is"
* without express or implied warranty of any kind.
*
* These notices must be retained in any copies of any part of this
* documentation and/or software.
*
* NOTE: Cleaned-up and optimized, version of SHA2, based on the FIPS 180-2
* standard, available at
* http://csrc.nist.gov/publications/fips/fips180-2/fips180-2.pdf
* Not as fast as one would like -- further optimizations are encouraged
* and appreciated.
*/
#include <sys/zfs_context.h>
#define _SHA2_IMPL
#include <sys/sha2.h>
#include <sha2/sha2_consts.h>
#define _RESTRICT_KYWD
#ifdef _ZFS_LITTLE_ENDIAN
#include <sys/byteorder.h>
#define HAVE_HTONL
#endif
#include <sys/isa_defs.h> /* for _ILP32 */
static void Encode(uint8_t *, uint32_t *, size_t);
static void Encode64(uint8_t *, uint64_t *, size_t);
/* userspace only supports the generic version */
#if defined(__amd64) && defined(_KERNEL)
#define SHA512Transform(ctx, in) SHA512TransformBlocks((ctx), (in), 1)
#define SHA256Transform(ctx, in) SHA256TransformBlocks((ctx), (in), 1)
void SHA512TransformBlocks(SHA2_CTX *ctx, const void *in, size_t num);
void SHA256TransformBlocks(SHA2_CTX *ctx, const void *in, size_t num);
#else
static void SHA256Transform(SHA2_CTX *, const uint8_t *);
static void SHA512Transform(SHA2_CTX *, const uint8_t *);
#endif /* __amd64 && _KERNEL */
-static uint8_t PADDING[128] = { 0x80, /* all zeros */ };
+static const uint8_t PADDING[128] = { 0x80, /* all zeros */ };
/*
* The low-level checksum routines use a lot of stack space. On systems where
* small stacks are enforced (like 32-bit kernel builds), insert compiler memory
* barriers to reduce stack frame size. This can reduce the SHA512Transform()
* stack frame usage from 3k to <1k on ARM32, for example.
*/
#if defined(_ILP32) || defined(__powerpc) /* small stack */
#define SMALL_STACK_MEMORY_BARRIER asm volatile("": : :"memory");
#else
#define SMALL_STACK_MEMORY_BARRIER
#endif
/* Ch and Maj are the basic SHA2 functions. */
#define Ch(b, c, d) (((b) & (c)) ^ ((~b) & (d)))
#define Maj(b, c, d) (((b) & (c)) ^ ((b) & (d)) ^ ((c) & (d)))
/* Rotates x right n bits. */
#define ROTR(x, n) \
(((x) >> (n)) | ((x) << ((sizeof (x) * NBBY)-(n))))
/* Shift x right n bits */
#define SHR(x, n) ((x) >> (n))
/* SHA256 Functions */
#define BIGSIGMA0_256(x) (ROTR((x), 2) ^ ROTR((x), 13) ^ ROTR((x), 22))
#define BIGSIGMA1_256(x) (ROTR((x), 6) ^ ROTR((x), 11) ^ ROTR((x), 25))
#define SIGMA0_256(x) (ROTR((x), 7) ^ ROTR((x), 18) ^ SHR((x), 3))
#define SIGMA1_256(x) (ROTR((x), 17) ^ ROTR((x), 19) ^ SHR((x), 10))
#define SHA256ROUND(a, b, c, d, e, f, g, h, i, w) \
T1 = h + BIGSIGMA1_256(e) + Ch(e, f, g) + SHA256_CONST(i) + w; \
d += T1; \
T2 = BIGSIGMA0_256(a) + Maj(a, b, c); \
h = T1 + T2
/* SHA384/512 Functions */
#define BIGSIGMA0(x) (ROTR((x), 28) ^ ROTR((x), 34) ^ ROTR((x), 39))
#define BIGSIGMA1(x) (ROTR((x), 14) ^ ROTR((x), 18) ^ ROTR((x), 41))
#define SIGMA0(x) (ROTR((x), 1) ^ ROTR((x), 8) ^ SHR((x), 7))
#define SIGMA1(x) (ROTR((x), 19) ^ ROTR((x), 61) ^ SHR((x), 6))
#define SHA512ROUND(a, b, c, d, e, f, g, h, i, w) \
T1 = h + BIGSIGMA1(e) + Ch(e, f, g) + SHA512_CONST(i) + w; \
d += T1; \
T2 = BIGSIGMA0(a) + Maj(a, b, c); \
h = T1 + T2; \
SMALL_STACK_MEMORY_BARRIER;
/*
* sparc optimization:
*
* on the sparc, we can load big endian 32-bit data easily. note that
* special care must be taken to ensure the address is 32-bit aligned.
* in the interest of speed, we don't check to make sure, since
* careful programming can guarantee this for us.
*/
#if defined(_ZFS_BIG_ENDIAN)
#define LOAD_BIG_32(addr) (*(uint32_t *)(addr))
#define LOAD_BIG_64(addr) (*(uint64_t *)(addr))
#elif defined(HAVE_HTONL)
#define LOAD_BIG_32(addr) htonl(*((uint32_t *)(addr)))
#define LOAD_BIG_64(addr) htonll(*((uint64_t *)(addr)))
#else
/* little endian -- will work on big endian, but slowly */
#define LOAD_BIG_32(addr) \
(((addr)[0] << 24) | ((addr)[1] << 16) | ((addr)[2] << 8) | (addr)[3])
#define LOAD_BIG_64(addr) \
(((uint64_t)(addr)[0] << 56) | ((uint64_t)(addr)[1] << 48) | \
((uint64_t)(addr)[2] << 40) | ((uint64_t)(addr)[3] << 32) | \
((uint64_t)(addr)[4] << 24) | ((uint64_t)(addr)[5] << 16) | \
((uint64_t)(addr)[6] << 8) | (uint64_t)(addr)[7])
#endif /* _BIG_ENDIAN */
#if !defined(__amd64) || !defined(_KERNEL)
/* SHA256 Transform */
static void
SHA256Transform(SHA2_CTX *ctx, const uint8_t *blk)
{
uint32_t a = ctx->state.s32[0];
uint32_t b = ctx->state.s32[1];
uint32_t c = ctx->state.s32[2];
uint32_t d = ctx->state.s32[3];
uint32_t e = ctx->state.s32[4];
uint32_t f = ctx->state.s32[5];
uint32_t g = ctx->state.s32[6];
uint32_t h = ctx->state.s32[7];
uint32_t w0, w1, w2, w3, w4, w5, w6, w7;
uint32_t w8, w9, w10, w11, w12, w13, w14, w15;
uint32_t T1, T2;
#if defined(__sparc)
static const uint32_t sha256_consts[] = {
SHA256_CONST_0, SHA256_CONST_1, SHA256_CONST_2,
SHA256_CONST_3, SHA256_CONST_4, SHA256_CONST_5,
SHA256_CONST_6, SHA256_CONST_7, SHA256_CONST_8,
SHA256_CONST_9, SHA256_CONST_10, SHA256_CONST_11,
SHA256_CONST_12, SHA256_CONST_13, SHA256_CONST_14,
SHA256_CONST_15, SHA256_CONST_16, SHA256_CONST_17,
SHA256_CONST_18, SHA256_CONST_19, SHA256_CONST_20,
SHA256_CONST_21, SHA256_CONST_22, SHA256_CONST_23,
SHA256_CONST_24, SHA256_CONST_25, SHA256_CONST_26,
SHA256_CONST_27, SHA256_CONST_28, SHA256_CONST_29,
SHA256_CONST_30, SHA256_CONST_31, SHA256_CONST_32,
SHA256_CONST_33, SHA256_CONST_34, SHA256_CONST_35,
SHA256_CONST_36, SHA256_CONST_37, SHA256_CONST_38,
SHA256_CONST_39, SHA256_CONST_40, SHA256_CONST_41,
SHA256_CONST_42, SHA256_CONST_43, SHA256_CONST_44,
SHA256_CONST_45, SHA256_CONST_46, SHA256_CONST_47,
SHA256_CONST_48, SHA256_CONST_49, SHA256_CONST_50,
SHA256_CONST_51, SHA256_CONST_52, SHA256_CONST_53,
SHA256_CONST_54, SHA256_CONST_55, SHA256_CONST_56,
SHA256_CONST_57, SHA256_CONST_58, SHA256_CONST_59,
SHA256_CONST_60, SHA256_CONST_61, SHA256_CONST_62,
SHA256_CONST_63
};
#endif /* __sparc */
if ((uintptr_t)blk & 0x3) { /* not 4-byte aligned? */
bcopy(blk, ctx->buf_un.buf32, sizeof (ctx->buf_un.buf32));
blk = (uint8_t *)ctx->buf_un.buf32;
}
/* LINTED E_BAD_PTR_CAST_ALIGN */
w0 = LOAD_BIG_32(blk + 4 * 0);
SHA256ROUND(a, b, c, d, e, f, g, h, 0, w0);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w1 = LOAD_BIG_32(blk + 4 * 1);
SHA256ROUND(h, a, b, c, d, e, f, g, 1, w1);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w2 = LOAD_BIG_32(blk + 4 * 2);
SHA256ROUND(g, h, a, b, c, d, e, f, 2, w2);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w3 = LOAD_BIG_32(blk + 4 * 3);
SHA256ROUND(f, g, h, a, b, c, d, e, 3, w3);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w4 = LOAD_BIG_32(blk + 4 * 4);
SHA256ROUND(e, f, g, h, a, b, c, d, 4, w4);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w5 = LOAD_BIG_32(blk + 4 * 5);
SHA256ROUND(d, e, f, g, h, a, b, c, 5, w5);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w6 = LOAD_BIG_32(blk + 4 * 6);
SHA256ROUND(c, d, e, f, g, h, a, b, 6, w6);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w7 = LOAD_BIG_32(blk + 4 * 7);
SHA256ROUND(b, c, d, e, f, g, h, a, 7, w7);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w8 = LOAD_BIG_32(blk + 4 * 8);
SHA256ROUND(a, b, c, d, e, f, g, h, 8, w8);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w9 = LOAD_BIG_32(blk + 4 * 9);
SHA256ROUND(h, a, b, c, d, e, f, g, 9, w9);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w10 = LOAD_BIG_32(blk + 4 * 10);
SHA256ROUND(g, h, a, b, c, d, e, f, 10, w10);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w11 = LOAD_BIG_32(blk + 4 * 11);
SHA256ROUND(f, g, h, a, b, c, d, e, 11, w11);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w12 = LOAD_BIG_32(blk + 4 * 12);
SHA256ROUND(e, f, g, h, a, b, c, d, 12, w12);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w13 = LOAD_BIG_32(blk + 4 * 13);
SHA256ROUND(d, e, f, g, h, a, b, c, 13, w13);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w14 = LOAD_BIG_32(blk + 4 * 14);
SHA256ROUND(c, d, e, f, g, h, a, b, 14, w14);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w15 = LOAD_BIG_32(blk + 4 * 15);
SHA256ROUND(b, c, d, e, f, g, h, a, 15, w15);
w0 = SIGMA1_256(w14) + w9 + SIGMA0_256(w1) + w0;
SHA256ROUND(a, b, c, d, e, f, g, h, 16, w0);
w1 = SIGMA1_256(w15) + w10 + SIGMA0_256(w2) + w1;
SHA256ROUND(h, a, b, c, d, e, f, g, 17, w1);
w2 = SIGMA1_256(w0) + w11 + SIGMA0_256(w3) + w2;
SHA256ROUND(g, h, a, b, c, d, e, f, 18, w2);
w3 = SIGMA1_256(w1) + w12 + SIGMA0_256(w4) + w3;
SHA256ROUND(f, g, h, a, b, c, d, e, 19, w3);
w4 = SIGMA1_256(w2) + w13 + SIGMA0_256(w5) + w4;
SHA256ROUND(e, f, g, h, a, b, c, d, 20, w4);
w5 = SIGMA1_256(w3) + w14 + SIGMA0_256(w6) + w5;
SHA256ROUND(d, e, f, g, h, a, b, c, 21, w5);
w6 = SIGMA1_256(w4) + w15 + SIGMA0_256(w7) + w6;
SHA256ROUND(c, d, e, f, g, h, a, b, 22, w6);
w7 = SIGMA1_256(w5) + w0 + SIGMA0_256(w8) + w7;
SHA256ROUND(b, c, d, e, f, g, h, a, 23, w7);
w8 = SIGMA1_256(w6) + w1 + SIGMA0_256(w9) + w8;
SHA256ROUND(a, b, c, d, e, f, g, h, 24, w8);
w9 = SIGMA1_256(w7) + w2 + SIGMA0_256(w10) + w9;
SHA256ROUND(h, a, b, c, d, e, f, g, 25, w9);
w10 = SIGMA1_256(w8) + w3 + SIGMA0_256(w11) + w10;
SHA256ROUND(g, h, a, b, c, d, e, f, 26, w10);
w11 = SIGMA1_256(w9) + w4 + SIGMA0_256(w12) + w11;
SHA256ROUND(f, g, h, a, b, c, d, e, 27, w11);
w12 = SIGMA1_256(w10) + w5 + SIGMA0_256(w13) + w12;
SHA256ROUND(e, f, g, h, a, b, c, d, 28, w12);
w13 = SIGMA1_256(w11) + w6 + SIGMA0_256(w14) + w13;
SHA256ROUND(d, e, f, g, h, a, b, c, 29, w13);
w14 = SIGMA1_256(w12) + w7 + SIGMA0_256(w15) + w14;
SHA256ROUND(c, d, e, f, g, h, a, b, 30, w14);
w15 = SIGMA1_256(w13) + w8 + SIGMA0_256(w0) + w15;
SHA256ROUND(b, c, d, e, f, g, h, a, 31, w15);
w0 = SIGMA1_256(w14) + w9 + SIGMA0_256(w1) + w0;
SHA256ROUND(a, b, c, d, e, f, g, h, 32, w0);
w1 = SIGMA1_256(w15) + w10 + SIGMA0_256(w2) + w1;
SHA256ROUND(h, a, b, c, d, e, f, g, 33, w1);
w2 = SIGMA1_256(w0) + w11 + SIGMA0_256(w3) + w2;
SHA256ROUND(g, h, a, b, c, d, e, f, 34, w2);
w3 = SIGMA1_256(w1) + w12 + SIGMA0_256(w4) + w3;
SHA256ROUND(f, g, h, a, b, c, d, e, 35, w3);
w4 = SIGMA1_256(w2) + w13 + SIGMA0_256(w5) + w4;
SHA256ROUND(e, f, g, h, a, b, c, d, 36, w4);
w5 = SIGMA1_256(w3) + w14 + SIGMA0_256(w6) + w5;
SHA256ROUND(d, e, f, g, h, a, b, c, 37, w5);
w6 = SIGMA1_256(w4) + w15 + SIGMA0_256(w7) + w6;
SHA256ROUND(c, d, e, f, g, h, a, b, 38, w6);
w7 = SIGMA1_256(w5) + w0 + SIGMA0_256(w8) + w7;
SHA256ROUND(b, c, d, e, f, g, h, a, 39, w7);
w8 = SIGMA1_256(w6) + w1 + SIGMA0_256(w9) + w8;
SHA256ROUND(a, b, c, d, e, f, g, h, 40, w8);
w9 = SIGMA1_256(w7) + w2 + SIGMA0_256(w10) + w9;
SHA256ROUND(h, a, b, c, d, e, f, g, 41, w9);
w10 = SIGMA1_256(w8) + w3 + SIGMA0_256(w11) + w10;
SHA256ROUND(g, h, a, b, c, d, e, f, 42, w10);
w11 = SIGMA1_256(w9) + w4 + SIGMA0_256(w12) + w11;
SHA256ROUND(f, g, h, a, b, c, d, e, 43, w11);
w12 = SIGMA1_256(w10) + w5 + SIGMA0_256(w13) + w12;
SHA256ROUND(e, f, g, h, a, b, c, d, 44, w12);
w13 = SIGMA1_256(w11) + w6 + SIGMA0_256(w14) + w13;
SHA256ROUND(d, e, f, g, h, a, b, c, 45, w13);
w14 = SIGMA1_256(w12) + w7 + SIGMA0_256(w15) + w14;
SHA256ROUND(c, d, e, f, g, h, a, b, 46, w14);
w15 = SIGMA1_256(w13) + w8 + SIGMA0_256(w0) + w15;
SHA256ROUND(b, c, d, e, f, g, h, a, 47, w15);
w0 = SIGMA1_256(w14) + w9 + SIGMA0_256(w1) + w0;
SHA256ROUND(a, b, c, d, e, f, g, h, 48, w0);
w1 = SIGMA1_256(w15) + w10 + SIGMA0_256(w2) + w1;
SHA256ROUND(h, a, b, c, d, e, f, g, 49, w1);
w2 = SIGMA1_256(w0) + w11 + SIGMA0_256(w3) + w2;
SHA256ROUND(g, h, a, b, c, d, e, f, 50, w2);
w3 = SIGMA1_256(w1) + w12 + SIGMA0_256(w4) + w3;
SHA256ROUND(f, g, h, a, b, c, d, e, 51, w3);
w4 = SIGMA1_256(w2) + w13 + SIGMA0_256(w5) + w4;
SHA256ROUND(e, f, g, h, a, b, c, d, 52, w4);
w5 = SIGMA1_256(w3) + w14 + SIGMA0_256(w6) + w5;
SHA256ROUND(d, e, f, g, h, a, b, c, 53, w5);
w6 = SIGMA1_256(w4) + w15 + SIGMA0_256(w7) + w6;
SHA256ROUND(c, d, e, f, g, h, a, b, 54, w6);
w7 = SIGMA1_256(w5) + w0 + SIGMA0_256(w8) + w7;
SHA256ROUND(b, c, d, e, f, g, h, a, 55, w7);
w8 = SIGMA1_256(w6) + w1 + SIGMA0_256(w9) + w8;
SHA256ROUND(a, b, c, d, e, f, g, h, 56, w8);
w9 = SIGMA1_256(w7) + w2 + SIGMA0_256(w10) + w9;
SHA256ROUND(h, a, b, c, d, e, f, g, 57, w9);
w10 = SIGMA1_256(w8) + w3 + SIGMA0_256(w11) + w10;
SHA256ROUND(g, h, a, b, c, d, e, f, 58, w10);
w11 = SIGMA1_256(w9) + w4 + SIGMA0_256(w12) + w11;
SHA256ROUND(f, g, h, a, b, c, d, e, 59, w11);
w12 = SIGMA1_256(w10) + w5 + SIGMA0_256(w13) + w12;
SHA256ROUND(e, f, g, h, a, b, c, d, 60, w12);
w13 = SIGMA1_256(w11) + w6 + SIGMA0_256(w14) + w13;
SHA256ROUND(d, e, f, g, h, a, b, c, 61, w13);
w14 = SIGMA1_256(w12) + w7 + SIGMA0_256(w15) + w14;
SHA256ROUND(c, d, e, f, g, h, a, b, 62, w14);
w15 = SIGMA1_256(w13) + w8 + SIGMA0_256(w0) + w15;
SHA256ROUND(b, c, d, e, f, g, h, a, 63, w15);
ctx->state.s32[0] += a;
ctx->state.s32[1] += b;
ctx->state.s32[2] += c;
ctx->state.s32[3] += d;
ctx->state.s32[4] += e;
ctx->state.s32[5] += f;
ctx->state.s32[6] += g;
ctx->state.s32[7] += h;
}
/* SHA384 and SHA512 Transform */
static void
SHA512Transform(SHA2_CTX *ctx, const uint8_t *blk)
{
uint64_t a = ctx->state.s64[0];
uint64_t b = ctx->state.s64[1];
uint64_t c = ctx->state.s64[2];
uint64_t d = ctx->state.s64[3];
uint64_t e = ctx->state.s64[4];
uint64_t f = ctx->state.s64[5];
uint64_t g = ctx->state.s64[6];
uint64_t h = ctx->state.s64[7];
uint64_t w0, w1, w2, w3, w4, w5, w6, w7;
uint64_t w8, w9, w10, w11, w12, w13, w14, w15;
uint64_t T1, T2;
#if defined(__sparc)
static const uint64_t sha512_consts[] = {
SHA512_CONST_0, SHA512_CONST_1, SHA512_CONST_2,
SHA512_CONST_3, SHA512_CONST_4, SHA512_CONST_5,
SHA512_CONST_6, SHA512_CONST_7, SHA512_CONST_8,
SHA512_CONST_9, SHA512_CONST_10, SHA512_CONST_11,
SHA512_CONST_12, SHA512_CONST_13, SHA512_CONST_14,
SHA512_CONST_15, SHA512_CONST_16, SHA512_CONST_17,
SHA512_CONST_18, SHA512_CONST_19, SHA512_CONST_20,
SHA512_CONST_21, SHA512_CONST_22, SHA512_CONST_23,
SHA512_CONST_24, SHA512_CONST_25, SHA512_CONST_26,
SHA512_CONST_27, SHA512_CONST_28, SHA512_CONST_29,
SHA512_CONST_30, SHA512_CONST_31, SHA512_CONST_32,
SHA512_CONST_33, SHA512_CONST_34, SHA512_CONST_35,
SHA512_CONST_36, SHA512_CONST_37, SHA512_CONST_38,
SHA512_CONST_39, SHA512_CONST_40, SHA512_CONST_41,
SHA512_CONST_42, SHA512_CONST_43, SHA512_CONST_44,
SHA512_CONST_45, SHA512_CONST_46, SHA512_CONST_47,
SHA512_CONST_48, SHA512_CONST_49, SHA512_CONST_50,
SHA512_CONST_51, SHA512_CONST_52, SHA512_CONST_53,
SHA512_CONST_54, SHA512_CONST_55, SHA512_CONST_56,
SHA512_CONST_57, SHA512_CONST_58, SHA512_CONST_59,
SHA512_CONST_60, SHA512_CONST_61, SHA512_CONST_62,
SHA512_CONST_63, SHA512_CONST_64, SHA512_CONST_65,
SHA512_CONST_66, SHA512_CONST_67, SHA512_CONST_68,
SHA512_CONST_69, SHA512_CONST_70, SHA512_CONST_71,
SHA512_CONST_72, SHA512_CONST_73, SHA512_CONST_74,
SHA512_CONST_75, SHA512_CONST_76, SHA512_CONST_77,
SHA512_CONST_78, SHA512_CONST_79
};
#endif /* __sparc */
if ((uintptr_t)blk & 0x7) { /* not 8-byte aligned? */
bcopy(blk, ctx->buf_un.buf64, sizeof (ctx->buf_un.buf64));
blk = (uint8_t *)ctx->buf_un.buf64;
}
/* LINTED E_BAD_PTR_CAST_ALIGN */
w0 = LOAD_BIG_64(blk + 8 * 0);
SHA512ROUND(a, b, c, d, e, f, g, h, 0, w0);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w1 = LOAD_BIG_64(blk + 8 * 1);
SHA512ROUND(h, a, b, c, d, e, f, g, 1, w1);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w2 = LOAD_BIG_64(blk + 8 * 2);
SHA512ROUND(g, h, a, b, c, d, e, f, 2, w2);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w3 = LOAD_BIG_64(blk + 8 * 3);
SHA512ROUND(f, g, h, a, b, c, d, e, 3, w3);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w4 = LOAD_BIG_64(blk + 8 * 4);
SHA512ROUND(e, f, g, h, a, b, c, d, 4, w4);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w5 = LOAD_BIG_64(blk + 8 * 5);
SHA512ROUND(d, e, f, g, h, a, b, c, 5, w5);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w6 = LOAD_BIG_64(blk + 8 * 6);
SHA512ROUND(c, d, e, f, g, h, a, b, 6, w6);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w7 = LOAD_BIG_64(blk + 8 * 7);
SHA512ROUND(b, c, d, e, f, g, h, a, 7, w7);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w8 = LOAD_BIG_64(blk + 8 * 8);
SHA512ROUND(a, b, c, d, e, f, g, h, 8, w8);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w9 = LOAD_BIG_64(blk + 8 * 9);
SHA512ROUND(h, a, b, c, d, e, f, g, 9, w9);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w10 = LOAD_BIG_64(blk + 8 * 10);
SHA512ROUND(g, h, a, b, c, d, e, f, 10, w10);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w11 = LOAD_BIG_64(blk + 8 * 11);
SHA512ROUND(f, g, h, a, b, c, d, e, 11, w11);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w12 = LOAD_BIG_64(blk + 8 * 12);
SHA512ROUND(e, f, g, h, a, b, c, d, 12, w12);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w13 = LOAD_BIG_64(blk + 8 * 13);
SHA512ROUND(d, e, f, g, h, a, b, c, 13, w13);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w14 = LOAD_BIG_64(blk + 8 * 14);
SHA512ROUND(c, d, e, f, g, h, a, b, 14, w14);
/* LINTED E_BAD_PTR_CAST_ALIGN */
w15 = LOAD_BIG_64(blk + 8 * 15);
SHA512ROUND(b, c, d, e, f, g, h, a, 15, w15);
w0 = SIGMA1(w14) + w9 + SIGMA0(w1) + w0;
SHA512ROUND(a, b, c, d, e, f, g, h, 16, w0);
w1 = SIGMA1(w15) + w10 + SIGMA0(w2) + w1;
SHA512ROUND(h, a, b, c, d, e, f, g, 17, w1);
w2 = SIGMA1(w0) + w11 + SIGMA0(w3) + w2;
SHA512ROUND(g, h, a, b, c, d, e, f, 18, w2);
w3 = SIGMA1(w1) + w12 + SIGMA0(w4) + w3;
SHA512ROUND(f, g, h, a, b, c, d, e, 19, w3);
w4 = SIGMA1(w2) + w13 + SIGMA0(w5) + w4;
SHA512ROUND(e, f, g, h, a, b, c, d, 20, w4);
w5 = SIGMA1(w3) + w14 + SIGMA0(w6) + w5;
SHA512ROUND(d, e, f, g, h, a, b, c, 21, w5);
w6 = SIGMA1(w4) + w15 + SIGMA0(w7) + w6;
SHA512ROUND(c, d, e, f, g, h, a, b, 22, w6);
w7 = SIGMA1(w5) + w0 + SIGMA0(w8) + w7;
SHA512ROUND(b, c, d, e, f, g, h, a, 23, w7);
w8 = SIGMA1(w6) + w1 + SIGMA0(w9) + w8;
SHA512ROUND(a, b, c, d, e, f, g, h, 24, w8);
w9 = SIGMA1(w7) + w2 + SIGMA0(w10) + w9;
SHA512ROUND(h, a, b, c, d, e, f, g, 25, w9);
w10 = SIGMA1(w8) + w3 + SIGMA0(w11) + w10;
SHA512ROUND(g, h, a, b, c, d, e, f, 26, w10);
w11 = SIGMA1(w9) + w4 + SIGMA0(w12) + w11;
SHA512ROUND(f, g, h, a, b, c, d, e, 27, w11);
w12 = SIGMA1(w10) + w5 + SIGMA0(w13) + w12;
SHA512ROUND(e, f, g, h, a, b, c, d, 28, w12);
w13 = SIGMA1(w11) + w6 + SIGMA0(w14) + w13;
SHA512ROUND(d, e, f, g, h, a, b, c, 29, w13);
w14 = SIGMA1(w12) + w7 + SIGMA0(w15) + w14;
SHA512ROUND(c, d, e, f, g, h, a, b, 30, w14);
w15 = SIGMA1(w13) + w8 + SIGMA0(w0) + w15;
SHA512ROUND(b, c, d, e, f, g, h, a, 31, w15);
w0 = SIGMA1(w14) + w9 + SIGMA0(w1) + w0;
SHA512ROUND(a, b, c, d, e, f, g, h, 32, w0);
w1 = SIGMA1(w15) + w10 + SIGMA0(w2) + w1;
SHA512ROUND(h, a, b, c, d, e, f, g, 33, w1);
w2 = SIGMA1(w0) + w11 + SIGMA0(w3) + w2;
SHA512ROUND(g, h, a, b, c, d, e, f, 34, w2);
w3 = SIGMA1(w1) + w12 + SIGMA0(w4) + w3;
SHA512ROUND(f, g, h, a, b, c, d, e, 35, w3);
w4 = SIGMA1(w2) + w13 + SIGMA0(w5) + w4;
SHA512ROUND(e, f, g, h, a, b, c, d, 36, w4);
w5 = SIGMA1(w3) + w14 + SIGMA0(w6) + w5;
SHA512ROUND(d, e, f, g, h, a, b, c, 37, w5);
w6 = SIGMA1(w4) + w15 + SIGMA0(w7) + w6;
SHA512ROUND(c, d, e, f, g, h, a, b, 38, w6);
w7 = SIGMA1(w5) + w0 + SIGMA0(w8) + w7;
SHA512ROUND(b, c, d, e, f, g, h, a, 39, w7);
w8 = SIGMA1(w6) + w1 + SIGMA0(w9) + w8;
SHA512ROUND(a, b, c, d, e, f, g, h, 40, w8);
w9 = SIGMA1(w7) + w2 + SIGMA0(w10) + w9;
SHA512ROUND(h, a, b, c, d, e, f, g, 41, w9);
w10 = SIGMA1(w8) + w3 + SIGMA0(w11) + w10;
SHA512ROUND(g, h, a, b, c, d, e, f, 42, w10);
w11 = SIGMA1(w9) + w4 + SIGMA0(w12) + w11;
SHA512ROUND(f, g, h, a, b, c, d, e, 43, w11);
w12 = SIGMA1(w10) + w5 + SIGMA0(w13) + w12;
SHA512ROUND(e, f, g, h, a, b, c, d, 44, w12);
w13 = SIGMA1(w11) + w6 + SIGMA0(w14) + w13;
SHA512ROUND(d, e, f, g, h, a, b, c, 45, w13);
w14 = SIGMA1(w12) + w7 + SIGMA0(w15) + w14;
SHA512ROUND(c, d, e, f, g, h, a, b, 46, w14);
w15 = SIGMA1(w13) + w8 + SIGMA0(w0) + w15;
SHA512ROUND(b, c, d, e, f, g, h, a, 47, w15);
w0 = SIGMA1(w14) + w9 + SIGMA0(w1) + w0;
SHA512ROUND(a, b, c, d, e, f, g, h, 48, w0);
w1 = SIGMA1(w15) + w10 + SIGMA0(w2) + w1;
SHA512ROUND(h, a, b, c, d, e, f, g, 49, w1);
w2 = SIGMA1(w0) + w11 + SIGMA0(w3) + w2;
SHA512ROUND(g, h, a, b, c, d, e, f, 50, w2);
w3 = SIGMA1(w1) + w12 + SIGMA0(w4) + w3;
SHA512ROUND(f, g, h, a, b, c, d, e, 51, w3);
w4 = SIGMA1(w2) + w13 + SIGMA0(w5) + w4;
SHA512ROUND(e, f, g, h, a, b, c, d, 52, w4);
w5 = SIGMA1(w3) + w14 + SIGMA0(w6) + w5;
SHA512ROUND(d, e, f, g, h, a, b, c, 53, w5);
w6 = SIGMA1(w4) + w15 + SIGMA0(w7) + w6;
SHA512ROUND(c, d, e, f, g, h, a, b, 54, w6);
w7 = SIGMA1(w5) + w0 + SIGMA0(w8) + w7;
SHA512ROUND(b, c, d, e, f, g, h, a, 55, w7);
w8 = SIGMA1(w6) + w1 + SIGMA0(w9) + w8;
SHA512ROUND(a, b, c, d, e, f, g, h, 56, w8);
w9 = SIGMA1(w7) + w2 + SIGMA0(w10) + w9;
SHA512ROUND(h, a, b, c, d, e, f, g, 57, w9);
w10 = SIGMA1(w8) + w3 + SIGMA0(w11) + w10;
SHA512ROUND(g, h, a, b, c, d, e, f, 58, w10);
w11 = SIGMA1(w9) + w4 + SIGMA0(w12) + w11;
SHA512ROUND(f, g, h, a, b, c, d, e, 59, w11);
w12 = SIGMA1(w10) + w5 + SIGMA0(w13) + w12;
SHA512ROUND(e, f, g, h, a, b, c, d, 60, w12);
w13 = SIGMA1(w11) + w6 + SIGMA0(w14) + w13;
SHA512ROUND(d, e, f, g, h, a, b, c, 61, w13);
w14 = SIGMA1(w12) + w7 + SIGMA0(w15) + w14;
SHA512ROUND(c, d, e, f, g, h, a, b, 62, w14);
w15 = SIGMA1(w13) + w8 + SIGMA0(w0) + w15;
SHA512ROUND(b, c, d, e, f, g, h, a, 63, w15);
w0 = SIGMA1(w14) + w9 + SIGMA0(w1) + w0;
SHA512ROUND(a, b, c, d, e, f, g, h, 64, w0);
w1 = SIGMA1(w15) + w10 + SIGMA0(w2) + w1;
SHA512ROUND(h, a, b, c, d, e, f, g, 65, w1);
w2 = SIGMA1(w0) + w11 + SIGMA0(w3) + w2;
SHA512ROUND(g, h, a, b, c, d, e, f, 66, w2);
w3 = SIGMA1(w1) + w12 + SIGMA0(w4) + w3;
SHA512ROUND(f, g, h, a, b, c, d, e, 67, w3);
w4 = SIGMA1(w2) + w13 + SIGMA0(w5) + w4;
SHA512ROUND(e, f, g, h, a, b, c, d, 68, w4);
w5 = SIGMA1(w3) + w14 + SIGMA0(w6) + w5;
SHA512ROUND(d, e, f, g, h, a, b, c, 69, w5);
w6 = SIGMA1(w4) + w15 + SIGMA0(w7) + w6;
SHA512ROUND(c, d, e, f, g, h, a, b, 70, w6);
w7 = SIGMA1(w5) + w0 + SIGMA0(w8) + w7;
SHA512ROUND(b, c, d, e, f, g, h, a, 71, w7);
w8 = SIGMA1(w6) + w1 + SIGMA0(w9) + w8;
SHA512ROUND(a, b, c, d, e, f, g, h, 72, w8);
w9 = SIGMA1(w7) + w2 + SIGMA0(w10) + w9;
SHA512ROUND(h, a, b, c, d, e, f, g, 73, w9);
w10 = SIGMA1(w8) + w3 + SIGMA0(w11) + w10;
SHA512ROUND(g, h, a, b, c, d, e, f, 74, w10);
w11 = SIGMA1(w9) + w4 + SIGMA0(w12) + w11;
SHA512ROUND(f, g, h, a, b, c, d, e, 75, w11);
w12 = SIGMA1(w10) + w5 + SIGMA0(w13) + w12;
SHA512ROUND(e, f, g, h, a, b, c, d, 76, w12);
w13 = SIGMA1(w11) + w6 + SIGMA0(w14) + w13;
SHA512ROUND(d, e, f, g, h, a, b, c, 77, w13);
w14 = SIGMA1(w12) + w7 + SIGMA0(w15) + w14;
SHA512ROUND(c, d, e, f, g, h, a, b, 78, w14);
w15 = SIGMA1(w13) + w8 + SIGMA0(w0) + w15;
SHA512ROUND(b, c, d, e, f, g, h, a, 79, w15);
ctx->state.s64[0] += a;
ctx->state.s64[1] += b;
ctx->state.s64[2] += c;
ctx->state.s64[3] += d;
ctx->state.s64[4] += e;
ctx->state.s64[5] += f;
ctx->state.s64[6] += g;
ctx->state.s64[7] += h;
}
#endif /* !__amd64 || !_KERNEL */
/*
* Encode()
*
* purpose: to convert a list of numbers from little endian to big endian
* input: uint8_t * : place to store the converted big endian numbers
* uint32_t * : place to get numbers to convert from
* size_t : the length of the input in bytes
* output: void
*/
static void
Encode(uint8_t *_RESTRICT_KYWD output, uint32_t *_RESTRICT_KYWD input,
size_t len)
{
size_t i, j;
#if defined(__sparc)
if (IS_P2ALIGNED(output, sizeof (uint32_t))) {
for (i = 0, j = 0; j < len; i++, j += 4) {
/* LINTED E_BAD_PTR_CAST_ALIGN */
*((uint32_t *)(output + j)) = input[i];
}
} else {
#endif /* little endian -- will work on big endian, but slowly */
for (i = 0, j = 0; j < len; i++, j += 4) {
output[j] = (input[i] >> 24) & 0xff;
output[j + 1] = (input[i] >> 16) & 0xff;
output[j + 2] = (input[i] >> 8) & 0xff;
output[j + 3] = input[i] & 0xff;
}
#if defined(__sparc)
}
#endif
}
static void
Encode64(uint8_t *_RESTRICT_KYWD output, uint64_t *_RESTRICT_KYWD input,
size_t len)
{
size_t i, j;
#if defined(__sparc)
if (IS_P2ALIGNED(output, sizeof (uint64_t))) {
for (i = 0, j = 0; j < len; i++, j += 8) {
/* LINTED E_BAD_PTR_CAST_ALIGN */
*((uint64_t *)(output + j)) = input[i];
}
} else {
#endif /* little endian -- will work on big endian, but slowly */
for (i = 0, j = 0; j < len; i++, j += 8) {
output[j] = (input[i] >> 56) & 0xff;
output[j + 1] = (input[i] >> 48) & 0xff;
output[j + 2] = (input[i] >> 40) & 0xff;
output[j + 3] = (input[i] >> 32) & 0xff;
output[j + 4] = (input[i] >> 24) & 0xff;
output[j + 5] = (input[i] >> 16) & 0xff;
output[j + 6] = (input[i] >> 8) & 0xff;
output[j + 7] = input[i] & 0xff;
}
#if defined(__sparc)
}
#endif
}
void
SHA2Init(uint64_t mech, SHA2_CTX *ctx)
{
switch (mech) {
case SHA256_MECH_INFO_TYPE:
case SHA256_HMAC_MECH_INFO_TYPE:
case SHA256_HMAC_GEN_MECH_INFO_TYPE:
ctx->state.s32[0] = 0x6a09e667U;
ctx->state.s32[1] = 0xbb67ae85U;
ctx->state.s32[2] = 0x3c6ef372U;
ctx->state.s32[3] = 0xa54ff53aU;
ctx->state.s32[4] = 0x510e527fU;
ctx->state.s32[5] = 0x9b05688cU;
ctx->state.s32[6] = 0x1f83d9abU;
ctx->state.s32[7] = 0x5be0cd19U;
break;
case SHA384_MECH_INFO_TYPE:
case SHA384_HMAC_MECH_INFO_TYPE:
case SHA384_HMAC_GEN_MECH_INFO_TYPE:
ctx->state.s64[0] = 0xcbbb9d5dc1059ed8ULL;
ctx->state.s64[1] = 0x629a292a367cd507ULL;
ctx->state.s64[2] = 0x9159015a3070dd17ULL;
ctx->state.s64[3] = 0x152fecd8f70e5939ULL;
ctx->state.s64[4] = 0x67332667ffc00b31ULL;
ctx->state.s64[5] = 0x8eb44a8768581511ULL;
ctx->state.s64[6] = 0xdb0c2e0d64f98fa7ULL;
ctx->state.s64[7] = 0x47b5481dbefa4fa4ULL;
break;
case SHA512_MECH_INFO_TYPE:
case SHA512_HMAC_MECH_INFO_TYPE:
case SHA512_HMAC_GEN_MECH_INFO_TYPE:
ctx->state.s64[0] = 0x6a09e667f3bcc908ULL;
ctx->state.s64[1] = 0xbb67ae8584caa73bULL;
ctx->state.s64[2] = 0x3c6ef372fe94f82bULL;
ctx->state.s64[3] = 0xa54ff53a5f1d36f1ULL;
ctx->state.s64[4] = 0x510e527fade682d1ULL;
ctx->state.s64[5] = 0x9b05688c2b3e6c1fULL;
ctx->state.s64[6] = 0x1f83d9abfb41bd6bULL;
ctx->state.s64[7] = 0x5be0cd19137e2179ULL;
break;
case SHA512_224_MECH_INFO_TYPE:
ctx->state.s64[0] = 0x8C3D37C819544DA2ULL;
ctx->state.s64[1] = 0x73E1996689DCD4D6ULL;
ctx->state.s64[2] = 0x1DFAB7AE32FF9C82ULL;
ctx->state.s64[3] = 0x679DD514582F9FCFULL;
ctx->state.s64[4] = 0x0F6D2B697BD44DA8ULL;
ctx->state.s64[5] = 0x77E36F7304C48942ULL;
ctx->state.s64[6] = 0x3F9D85A86A1D36C8ULL;
ctx->state.s64[7] = 0x1112E6AD91D692A1ULL;
break;
case SHA512_256_MECH_INFO_TYPE:
ctx->state.s64[0] = 0x22312194FC2BF72CULL;
ctx->state.s64[1] = 0x9F555FA3C84C64C2ULL;
ctx->state.s64[2] = 0x2393B86B6F53B151ULL;
ctx->state.s64[3] = 0x963877195940EABDULL;
ctx->state.s64[4] = 0x96283EE2A88EFFE3ULL;
ctx->state.s64[5] = 0xBE5E1E2553863992ULL;
ctx->state.s64[6] = 0x2B0199FC2C85B8AAULL;
ctx->state.s64[7] = 0x0EB72DDC81C52CA2ULL;
break;
#ifdef _KERNEL
default:
cmn_err(CE_PANIC,
"sha2_init: failed to find a supported algorithm: 0x%x",
(uint32_t)mech);
#endif /* _KERNEL */
}
ctx->algotype = (uint32_t)mech;
ctx->count.c64[0] = ctx->count.c64[1] = 0;
}
#ifndef _KERNEL
// #pragma inline(SHA256Init, SHA384Init, SHA512Init)
void
SHA256Init(SHA256_CTX *ctx)
{
SHA2Init(SHA256, ctx);
}
void
SHA384Init(SHA384_CTX *ctx)
{
SHA2Init(SHA384, ctx);
}
void
SHA512Init(SHA512_CTX *ctx)
{
SHA2Init(SHA512, ctx);
}
#endif /* _KERNEL */
/*
* SHA2Update()
*
* purpose: continues an sha2 digest operation, using the message block
* to update the context.
* input: SHA2_CTX * : the context to update
* void * : the message block
* size_t : the length of the message block, in bytes
* output: void
*/
void
SHA2Update(SHA2_CTX *ctx, const void *inptr, size_t input_len)
{
uint32_t i, buf_index, buf_len, buf_limit;
const uint8_t *input = inptr;
uint32_t algotype = ctx->algotype;
/* check for noop */
if (input_len == 0)
return;
if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
buf_limit = 64;
/* compute number of bytes mod 64 */
buf_index = (ctx->count.c32[1] >> 3) & 0x3F;
/* update number of bits */
if ((ctx->count.c32[1] += (input_len << 3)) < (input_len << 3))
ctx->count.c32[0]++;
ctx->count.c32[0] += (input_len >> 29);
} else {
buf_limit = 128;
/* compute number of bytes mod 128 */
buf_index = (ctx->count.c64[1] >> 3) & 0x7F;
/* update number of bits */
if ((ctx->count.c64[1] += (input_len << 3)) < (input_len << 3))
ctx->count.c64[0]++;
ctx->count.c64[0] += (input_len >> 29);
}
buf_len = buf_limit - buf_index;
/* transform as many times as possible */
i = 0;
if (input_len >= buf_len) {
/*
* general optimization:
*
* only do initial bcopy() and SHA2Transform() if
* buf_index != 0. if buf_index == 0, we're just
* wasting our time doing the bcopy() since there
* wasn't any data left over from a previous call to
* SHA2Update().
*/
if (buf_index) {
bcopy(input, &ctx->buf_un.buf8[buf_index], buf_len);
if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE)
SHA256Transform(ctx, ctx->buf_un.buf8);
else
SHA512Transform(ctx, ctx->buf_un.buf8);
i = buf_len;
}
#if !defined(__amd64) || !defined(_KERNEL)
if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
for (; i + buf_limit - 1 < input_len; i += buf_limit) {
SHA256Transform(ctx, &input[i]);
}
} else {
for (; i + buf_limit - 1 < input_len; i += buf_limit) {
SHA512Transform(ctx, &input[i]);
}
}
#else
uint32_t block_count;
if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
block_count = (input_len - i) >> 6;
if (block_count > 0) {
SHA256TransformBlocks(ctx, &input[i],
block_count);
i += block_count << 6;
}
} else {
block_count = (input_len - i) >> 7;
if (block_count > 0) {
SHA512TransformBlocks(ctx, &input[i],
block_count);
i += block_count << 7;
}
}
#endif /* !__amd64 || !_KERNEL */
/*
* general optimization:
*
* if i and input_len are the same, return now instead
* of calling bcopy(), since the bcopy() in this case
* will be an expensive noop.
*/
if (input_len == i)
return;
buf_index = 0;
}
/* buffer remaining input */
bcopy(&input[i], &ctx->buf_un.buf8[buf_index], input_len - i);
}
/*
* SHA2Final()
*
* purpose: ends an sha2 digest operation, finalizing the message digest and
* zeroing the context.
* input: uchar_t * : a buffer to store the digest
* : The function actually uses void* because many
* : callers pass things other than uchar_t here.
* SHA2_CTX * : the context to finalize, save, and zero
* output: void
*/
void
SHA2Final(void *digest, SHA2_CTX *ctx)
{
uint8_t bitcount_be[sizeof (ctx->count.c32)];
uint8_t bitcount_be64[sizeof (ctx->count.c64)];
uint32_t index;
uint32_t algotype = ctx->algotype;
if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
index = (ctx->count.c32[1] >> 3) & 0x3f;
Encode(bitcount_be, ctx->count.c32, sizeof (bitcount_be));
SHA2Update(ctx, PADDING, ((index < 56) ? 56 : 120) - index);
SHA2Update(ctx, bitcount_be, sizeof (bitcount_be));
Encode(digest, ctx->state.s32, sizeof (ctx->state.s32));
} else {
index = (ctx->count.c64[1] >> 3) & 0x7f;
Encode64(bitcount_be64, ctx->count.c64,
sizeof (bitcount_be64));
SHA2Update(ctx, PADDING, ((index < 112) ? 112 : 240) - index);
SHA2Update(ctx, bitcount_be64, sizeof (bitcount_be64));
if (algotype <= SHA384_HMAC_GEN_MECH_INFO_TYPE) {
ctx->state.s64[6] = ctx->state.s64[7] = 0;
Encode64(digest, ctx->state.s64,
sizeof (uint64_t) * 6);
} else if (algotype == SHA512_224_MECH_INFO_TYPE) {
uint8_t last[sizeof (uint64_t)];
/*
* Since SHA-512/224 doesn't align well to 64-bit
* boundaries, we must do the encoding in three steps:
* 1) encode the three 64-bit words that fit neatly
* 2) encode the last 64-bit word to a temp buffer
* 3) chop out the lower 32-bits from the temp buffer
* and append them to the digest
*/
Encode64(digest, ctx->state.s64, sizeof (uint64_t) * 3);
Encode64(last, &ctx->state.s64[3], sizeof (uint64_t));
bcopy(last, (uint8_t *)digest + 24, 4);
} else if (algotype == SHA512_256_MECH_INFO_TYPE) {
Encode64(digest, ctx->state.s64, sizeof (uint64_t) * 4);
} else {
Encode64(digest, ctx->state.s64,
sizeof (ctx->state.s64));
}
}
/* zeroize sensitive information */
bzero(ctx, sizeof (*ctx));
}
#ifdef _KERNEL
EXPORT_SYMBOL(SHA2Init);
EXPORT_SYMBOL(SHA2Update);
EXPORT_SYMBOL(SHA2Final);
#endif
diff --git a/module/icp/algs/skein/skein_impl.h b/module/icp/algs/skein/skein_impl.h
index 2f6307fa7b55..1fa249e95e4b 100644
--- a/module/icp/algs/skein/skein_impl.h
+++ b/module/icp/algs/skein/skein_impl.h
@@ -1,284 +1,282 @@
/*
* Internal definitions for Skein hashing.
* Source code author: Doug Whiting, 2008.
* This algorithm and source code is released to the public domain.
*
* The following compile-time switches may be defined to control some
* tradeoffs between speed, code size, error checking, and security.
*
* The "default" note explains what happens when the switch is not defined.
*
* SKEIN_DEBUG -- make callouts from inside Skein code
* to examine/display intermediate values.
* [default: no callouts (no overhead)]
*
* SKEIN_ERR_CHECK -- how error checking is handled inside Skein
* code. If not defined, most error checking
* is disabled (for performance). Otherwise,
* the switch value is interpreted as:
* 0: use assert() to flag errors
* 1: return SKEIN_FAIL to flag errors
*/
/* Copyright 2013 Doug Whiting. This code is released to the public domain. */
#ifndef _SKEIN_IMPL_H_
#define _SKEIN_IMPL_H_
#include <sys/skein.h>
#include <sys/strings.h>
#include "skein_impl.h"
#include "skein_port.h"
/*
* "Internal" Skein definitions
* -- not needed for sequential hashing API, but will be
* helpful for other uses of Skein (e.g., tree hash mode).
* -- included here so that they can be shared between
* reference and optimized code.
*/
/* tweak word T[1]: bit field starting positions */
/* offset 64 because it's the second word */
#define SKEIN_T1_BIT(BIT) ((BIT) - 64)
/* bits 112..118: level in hash tree */
#define SKEIN_T1_POS_TREE_LVL SKEIN_T1_BIT(112)
/* bit 119: partial final input byte */
#define SKEIN_T1_POS_BIT_PAD SKEIN_T1_BIT(119)
/* bits 120..125: type field */
#define SKEIN_T1_POS_BLK_TYPE SKEIN_T1_BIT(120)
/* bits 126: first block flag */
#define SKEIN_T1_POS_FIRST SKEIN_T1_BIT(126)
/* bit 127: final block flag */
#define SKEIN_T1_POS_FINAL SKEIN_T1_BIT(127)
/* tweak word T[1]: flag bit definition(s) */
#define SKEIN_T1_FLAG_FIRST (((uint64_t)1) << SKEIN_T1_POS_FIRST)
#define SKEIN_T1_FLAG_FINAL (((uint64_t)1) << SKEIN_T1_POS_FINAL)
#define SKEIN_T1_FLAG_BIT_PAD (((uint64_t)1) << SKEIN_T1_POS_BIT_PAD)
/* tweak word T[1]: tree level bit field mask */
#define SKEIN_T1_TREE_LVL_MASK (((uint64_t)0x7F) << SKEIN_T1_POS_TREE_LVL)
#define SKEIN_T1_TREE_LEVEL(n) (((uint64_t)(n)) << SKEIN_T1_POS_TREE_LVL)
/* tweak word T[1]: block type field */
#define SKEIN_BLK_TYPE_KEY (0) /* key, for MAC and KDF */
#define SKEIN_BLK_TYPE_CFG (4) /* configuration block */
#define SKEIN_BLK_TYPE_PERS (8) /* personalization string */
#define SKEIN_BLK_TYPE_PK (12) /* public key (for signature hashing) */
#define SKEIN_BLK_TYPE_KDF (16) /* key identifier for KDF */
#define SKEIN_BLK_TYPE_NONCE (20) /* nonce for PRNG */
#define SKEIN_BLK_TYPE_MSG (48) /* message processing */
#define SKEIN_BLK_TYPE_OUT (63) /* output stage */
#define SKEIN_BLK_TYPE_MASK (63) /* bit field mask */
#define SKEIN_T1_BLK_TYPE(T) \
(((uint64_t)(SKEIN_BLK_TYPE_##T)) << SKEIN_T1_POS_BLK_TYPE)
/* key, for MAC and KDF */
#define SKEIN_T1_BLK_TYPE_KEY SKEIN_T1_BLK_TYPE(KEY)
/* configuration block */
#define SKEIN_T1_BLK_TYPE_CFG SKEIN_T1_BLK_TYPE(CFG)
/* personalization string */
#define SKEIN_T1_BLK_TYPE_PERS SKEIN_T1_BLK_TYPE(PERS)
/* public key (for digital signature hashing) */
#define SKEIN_T1_BLK_TYPE_PK SKEIN_T1_BLK_TYPE(PK)
/* key identifier for KDF */
#define SKEIN_T1_BLK_TYPE_KDF SKEIN_T1_BLK_TYPE(KDF)
/* nonce for PRNG */
#define SKEIN_T1_BLK_TYPE_NONCE SKEIN_T1_BLK_TYPE(NONCE)
/* message processing */
#define SKEIN_T1_BLK_TYPE_MSG SKEIN_T1_BLK_TYPE(MSG)
/* output stage */
#define SKEIN_T1_BLK_TYPE_OUT SKEIN_T1_BLK_TYPE(OUT)
/* field bit mask */
#define SKEIN_T1_BLK_TYPE_MASK SKEIN_T1_BLK_TYPE(MASK)
#define SKEIN_T1_BLK_TYPE_CFG_FINAL \
(SKEIN_T1_BLK_TYPE_CFG | SKEIN_T1_FLAG_FINAL)
#define SKEIN_T1_BLK_TYPE_OUT_FINAL \
(SKEIN_T1_BLK_TYPE_OUT | SKEIN_T1_FLAG_FINAL)
#define SKEIN_VERSION (1)
#ifndef SKEIN_ID_STRING_LE /* allow compile-time personalization */
#define SKEIN_ID_STRING_LE (0x33414853) /* "SHA3" (little-endian) */
#endif
#define SKEIN_MK_64(hi32, lo32) ((lo32) + (((uint64_t)(hi32)) << 32))
#define SKEIN_SCHEMA_VER SKEIN_MK_64(SKEIN_VERSION, SKEIN_ID_STRING_LE)
#define SKEIN_KS_PARITY SKEIN_MK_64(0x1BD11BDA, 0xA9FC1A22)
#define SKEIN_CFG_STR_LEN (4*8)
/* bit field definitions in config block treeInfo word */
#define SKEIN_CFG_TREE_LEAF_SIZE_POS (0)
#define SKEIN_CFG_TREE_NODE_SIZE_POS (8)
#define SKEIN_CFG_TREE_MAX_LEVEL_POS (16)
#define SKEIN_CFG_TREE_LEAF_SIZE_MSK \
(((uint64_t)0xFF) << SKEIN_CFG_TREE_LEAF_SIZE_POS)
#define SKEIN_CFG_TREE_NODE_SIZE_MSK \
(((uint64_t)0xFF) << SKEIN_CFG_TREE_NODE_SIZE_POS)
#define SKEIN_CFG_TREE_MAX_LEVEL_MSK \
(((uint64_t)0xFF) << SKEIN_CFG_TREE_MAX_LEVEL_POS)
#define SKEIN_CFG_TREE_INFO(leaf, node, maxLvl) \
((((uint64_t)(leaf)) << SKEIN_CFG_TREE_LEAF_SIZE_POS) | \
(((uint64_t)(node)) << SKEIN_CFG_TREE_NODE_SIZE_POS) | \
(((uint64_t)(maxLvl)) << SKEIN_CFG_TREE_MAX_LEVEL_POS))
/* use as treeInfo in InitExt() call for sequential processing */
#define SKEIN_CFG_TREE_INFO_SEQUENTIAL SKEIN_CFG_TREE_INFO(0, 0, 0)
/*
* Skein macros for getting/setting tweak words, etc.
* These are useful for partial input bytes, hash tree init/update, etc.
*/
#define Skein_Get_Tweak(ctxPtr, TWK_NUM) ((ctxPtr)->h.T[TWK_NUM])
#define Skein_Set_Tweak(ctxPtr, TWK_NUM, tVal) \
do { \
(ctxPtr)->h.T[TWK_NUM] = (tVal); \
} while (0)
#define Skein_Get_T0(ctxPtr) Skein_Get_Tweak(ctxPtr, 0)
#define Skein_Get_T1(ctxPtr) Skein_Get_Tweak(ctxPtr, 1)
#define Skein_Set_T0(ctxPtr, T0) Skein_Set_Tweak(ctxPtr, 0, T0)
#define Skein_Set_T1(ctxPtr, T1) Skein_Set_Tweak(ctxPtr, 1, T1)
/* set both tweak words at once */
#define Skein_Set_T0_T1(ctxPtr, T0, T1) \
do { \
Skein_Set_T0(ctxPtr, (T0)); \
Skein_Set_T1(ctxPtr, (T1)); \
} while (0)
#define Skein_Set_Type(ctxPtr, BLK_TYPE) \
Skein_Set_T1(ctxPtr, SKEIN_T1_BLK_TYPE_##BLK_TYPE)
/*
* set up for starting with a new type: h.T[0]=0; h.T[1] = NEW_TYPE; h.bCnt=0;
*/
#define Skein_Start_New_Type(ctxPtr, BLK_TYPE) \
do { \
Skein_Set_T0_T1(ctxPtr, 0, SKEIN_T1_FLAG_FIRST | \
SKEIN_T1_BLK_TYPE_ ## BLK_TYPE); \
(ctxPtr)->h.bCnt = 0; \
} while (0)
#define Skein_Clear_First_Flag(hdr) \
do { \
(hdr).T[1] &= ~SKEIN_T1_FLAG_FIRST; \
} while (0)
#define Skein_Set_Bit_Pad_Flag(hdr) \
do { \
(hdr).T[1] |= SKEIN_T1_FLAG_BIT_PAD; \
} while (0)
#define Skein_Set_Tree_Level(hdr, height) \
do { \
(hdr).T[1] |= SKEIN_T1_TREE_LEVEL(height); \
} while (0)
/*
* "Internal" Skein definitions for debugging and error checking
* Note: in Illumos we always disable debugging features.
*/
#define Skein_Show_Block(bits, ctx, X, blkPtr, wPtr, ksEvenPtr, ksOddPtr)
#define Skein_Show_Round(bits, ctx, r, X)
#define Skein_Show_R_Ptr(bits, ctx, r, X_ptr)
#define Skein_Show_Final(bits, ctx, cnt, outPtr)
#define Skein_Show_Key(bits, ctx, key, keyBytes)
/* run-time checks (e.g., bad params, uninitialized context)? */
#ifndef SKEIN_ERR_CHECK
/* default: ignore all Asserts, for performance */
#define Skein_Assert(x, retCode)
#define Skein_assert(x)
#elif defined(SKEIN_ASSERT)
#include <sys/debug.h>
#define Skein_Assert(x, retCode) ASSERT(x)
#define Skein_assert(x) ASSERT(x)
#else
#include <sys/debug.h>
/* caller error */
#define Skein_Assert(x, retCode) \
do { \
if (!(x)) \
return (retCode); \
} while (0)
/* internal error */
#define Skein_assert(x) ASSERT(x)
#endif
/*
* Skein block function constants (shared across Ref and Opt code)
*/
enum {
/* Skein_256 round rotation constants */
R_256_0_0 = 14, R_256_0_1 = 16,
R_256_1_0 = 52, R_256_1_1 = 57,
R_256_2_0 = 23, R_256_2_1 = 40,
R_256_3_0 = 5, R_256_3_1 = 37,
R_256_4_0 = 25, R_256_4_1 = 33,
R_256_5_0 = 46, R_256_5_1 = 12,
R_256_6_0 = 58, R_256_6_1 = 22,
R_256_7_0 = 32, R_256_7_1 = 32,
/* Skein_512 round rotation constants */
R_512_0_0 = 46, R_512_0_1 = 36, R_512_0_2 = 19, R_512_0_3 = 37,
R_512_1_0 = 33, R_512_1_1 = 27, R_512_1_2 = 14, R_512_1_3 = 42,
R_512_2_0 = 17, R_512_2_1 = 49, R_512_2_2 = 36, R_512_2_3 = 39,
R_512_3_0 = 44, R_512_3_1 = 9, R_512_3_2 = 54, R_512_3_3 = 56,
R_512_4_0 = 39, R_512_4_1 = 30, R_512_4_2 = 34, R_512_4_3 = 24,
R_512_5_0 = 13, R_512_5_1 = 50, R_512_5_2 = 10, R_512_5_3 = 17,
R_512_6_0 = 25, R_512_6_1 = 29, R_512_6_2 = 39, R_512_6_3 = 43,
R_512_7_0 = 8, R_512_7_1 = 35, R_512_7_2 = 56, R_512_7_3 = 22,
/* Skein1024 round rotation constants */
R1024_0_0 = 24, R1024_0_1 = 13, R1024_0_2 = 8, R1024_0_3 =
47, R1024_0_4 = 8, R1024_0_5 = 17, R1024_0_6 = 22, R1024_0_7 = 37,
R1024_1_0 = 38, R1024_1_1 = 19, R1024_1_2 = 10, R1024_1_3 =
55, R1024_1_4 = 49, R1024_1_5 = 18, R1024_1_6 = 23, R1024_1_7 = 52,
R1024_2_0 = 33, R1024_2_1 = 4, R1024_2_2 = 51, R1024_2_3 =
13, R1024_2_4 = 34, R1024_2_5 = 41, R1024_2_6 = 59, R1024_2_7 = 17,
R1024_3_0 = 5, R1024_3_1 = 20, R1024_3_2 = 48, R1024_3_3 =
41, R1024_3_4 = 47, R1024_3_5 = 28, R1024_3_6 = 16, R1024_3_7 = 25,
R1024_4_0 = 41, R1024_4_1 = 9, R1024_4_2 = 37, R1024_4_3 =
31, R1024_4_4 = 12, R1024_4_5 = 47, R1024_4_6 = 44, R1024_4_7 = 30,
R1024_5_0 = 16, R1024_5_1 = 34, R1024_5_2 = 56, R1024_5_3 =
51, R1024_5_4 = 4, R1024_5_5 = 53, R1024_5_6 = 42, R1024_5_7 = 41,
R1024_6_0 = 31, R1024_6_1 = 44, R1024_6_2 = 47, R1024_6_3 =
46, R1024_6_4 = 19, R1024_6_5 = 42, R1024_6_6 = 44, R1024_6_7 = 25,
R1024_7_0 = 9, R1024_7_1 = 48, R1024_7_2 = 35, R1024_7_3 =
52, R1024_7_4 = 23, R1024_7_5 = 31, R1024_7_6 = 37, R1024_7_7 = 20
};
/* number of rounds for the different block sizes */
#define SKEIN_256_ROUNDS_TOTAL (72)
#define SKEIN_512_ROUNDS_TOTAL (72)
#define SKEIN1024_ROUNDS_TOTAL (80)
extern const uint64_t SKEIN_256_IV_128[];
extern const uint64_t SKEIN_256_IV_160[];
extern const uint64_t SKEIN_256_IV_224[];
extern const uint64_t SKEIN_256_IV_256[];
-extern const uint64_t SKEIN_512_IV_128[];
-extern const uint64_t SKEIN_512_IV_160[];
extern const uint64_t SKEIN_512_IV_224[];
extern const uint64_t SKEIN_512_IV_256[];
extern const uint64_t SKEIN_512_IV_384[];
extern const uint64_t SKEIN_512_IV_512[];
extern const uint64_t SKEIN1024_IV_384[];
extern const uint64_t SKEIN1024_IV_512[];
extern const uint64_t SKEIN1024_IV_1024[];
/* Functions to process blkCnt (nonzero) full block(s) of data. */
void Skein_256_Process_Block(Skein_256_Ctxt_t *ctx, const uint8_t *blkPtr,
size_t blkCnt, size_t byteCntAdd);
void Skein_512_Process_Block(Skein_512_Ctxt_t *ctx, const uint8_t *blkPtr,
size_t blkCnt, size_t byteCntAdd);
void Skein1024_Process_Block(Skein1024_Ctxt_t *ctx, const uint8_t *blkPtr,
size_t blkCnt, size_t byteCntAdd);
#endif /* _SKEIN_IMPL_H_ */
diff --git a/module/icp/algs/skein/skein_iv.c b/module/icp/algs/skein/skein_iv.c
index 140d38f76547..84cefe4546ca 100644
--- a/module/icp/algs/skein/skein_iv.c
+++ b/module/icp/algs/skein/skein_iv.c
@@ -1,185 +1,161 @@
/*
* Pre-computed Skein IVs
*
* NOTE: these values are not "magic" constants, but
* are generated using the Threefish block function.
* They are pre-computed here only for speed; i.e., to
* avoid the need for a Threefish call during Init().
*
* The IV for any fixed hash length may be pre-computed.
* Only the most common values are included here.
*/
/* Copyright 2013 Doug Whiting. This code is released to the public domain. */
/*
* Illumos implementation note: these constants are for Skein v1.3 as per:
* http://www.skein-hash.info/sites/default/files/skein1.3.pdf
*/
#include <sys/skein.h> /* get Skein macros and types */
#include "skein_impl.h" /* get internal definitions */
#define MK_64 SKEIN_MK_64
/* blkSize = 256 bits. hashSize = 128 bits */
const uint64_t SKEIN_256_IV_128[] = {
MK_64(0xE1111906, 0x964D7260),
MK_64(0x883DAAA7, 0x7C8D811C),
MK_64(0x10080DF4, 0x91960F7A),
MK_64(0xCCF7DDE5, 0xB45BC1C2)
};
/* blkSize = 256 bits. hashSize = 160 bits */
const uint64_t SKEIN_256_IV_160[] = {
MK_64(0x14202314, 0x72825E98),
MK_64(0x2AC4E9A2, 0x5A77E590),
MK_64(0xD47A5856, 0x8838D63E),
MK_64(0x2DD2E496, 0x8586AB7D)
};
/* blkSize = 256 bits. hashSize = 224 bits */
const uint64_t SKEIN_256_IV_224[] = {
MK_64(0xC6098A8C, 0x9AE5EA0B),
MK_64(0x876D5686, 0x08C5191C),
MK_64(0x99CB88D7, 0xD7F53884),
MK_64(0x384BDDB1, 0xAEDDB5DE)
};
/* blkSize = 256 bits. hashSize = 256 bits */
const uint64_t SKEIN_256_IV_256[] = {
MK_64(0xFC9DA860, 0xD048B449),
MK_64(0x2FCA6647, 0x9FA7D833),
MK_64(0xB33BC389, 0x6656840F),
MK_64(0x6A54E920, 0xFDE8DA69)
};
-/* blkSize = 512 bits. hashSize = 128 bits */
-const uint64_t SKEIN_512_IV_128[] = {
- MK_64(0xA8BC7BF3, 0x6FBF9F52),
- MK_64(0x1E9872CE, 0xBD1AF0AA),
- MK_64(0x309B1790, 0xB32190D3),
- MK_64(0xBCFBB854, 0x3F94805C),
- MK_64(0x0DA61BCD, 0x6E31B11B),
- MK_64(0x1A18EBEA, 0xD46A32E3),
- MK_64(0xA2CC5B18, 0xCE84AA82),
- MK_64(0x6982AB28, 0x9D46982D)
-};
-
-/* blkSize = 512 bits. hashSize = 160 bits */
-const uint64_t SKEIN_512_IV_160[] = {
- MK_64(0x28B81A2A, 0xE013BD91),
- MK_64(0xC2F11668, 0xB5BDF78F),
- MK_64(0x1760D8F3, 0xF6A56F12),
- MK_64(0x4FB74758, 0x8239904F),
- MK_64(0x21EDE07F, 0x7EAF5056),
- MK_64(0xD908922E, 0x63ED70B8),
- MK_64(0xB8EC76FF, 0xECCB52FA),
- MK_64(0x01A47BB8, 0xA3F27A6E)
-};
-
/* blkSize = 512 bits. hashSize = 224 bits */
const uint64_t SKEIN_512_IV_224[] = {
MK_64(0xCCD06162, 0x48677224),
MK_64(0xCBA65CF3, 0xA92339EF),
MK_64(0x8CCD69D6, 0x52FF4B64),
MK_64(0x398AED7B, 0x3AB890B4),
MK_64(0x0F59D1B1, 0x457D2BD0),
MK_64(0x6776FE65, 0x75D4EB3D),
MK_64(0x99FBC70E, 0x997413E9),
MK_64(0x9E2CFCCF, 0xE1C41EF7)
};
/* blkSize = 512 bits. hashSize = 256 bits */
const uint64_t SKEIN_512_IV_256[] = {
MK_64(0xCCD044A1, 0x2FDB3E13),
MK_64(0xE8359030, 0x1A79A9EB),
MK_64(0x55AEA061, 0x4F816E6F),
MK_64(0x2A2767A4, 0xAE9B94DB),
MK_64(0xEC06025E, 0x74DD7683),
MK_64(0xE7A436CD, 0xC4746251),
MK_64(0xC36FBAF9, 0x393AD185),
MK_64(0x3EEDBA18, 0x33EDFC13)
};
/* blkSize = 512 bits. hashSize = 384 bits */
const uint64_t SKEIN_512_IV_384[] = {
MK_64(0xA3F6C6BF, 0x3A75EF5F),
MK_64(0xB0FEF9CC, 0xFD84FAA4),
MK_64(0x9D77DD66, 0x3D770CFE),
MK_64(0xD798CBF3, 0xB468FDDA),
MK_64(0x1BC4A666, 0x8A0E4465),
MK_64(0x7ED7D434, 0xE5807407),
MK_64(0x548FC1AC, 0xD4EC44D6),
MK_64(0x266E1754, 0x6AA18FF8)
};
/* blkSize = 512 bits. hashSize = 512 bits */
const uint64_t SKEIN_512_IV_512[] = {
MK_64(0x4903ADFF, 0x749C51CE),
MK_64(0x0D95DE39, 0x9746DF03),
MK_64(0x8FD19341, 0x27C79BCE),
MK_64(0x9A255629, 0xFF352CB1),
MK_64(0x5DB62599, 0xDF6CA7B0),
MK_64(0xEABE394C, 0xA9D5C3F4),
MK_64(0x991112C7, 0x1A75B523),
MK_64(0xAE18A40B, 0x660FCC33)
};
/* blkSize = 1024 bits. hashSize = 384 bits */
const uint64_t SKEIN1024_IV_384[] = {
MK_64(0x5102B6B8, 0xC1894A35),
MK_64(0xFEEBC9E3, 0xFE8AF11A),
MK_64(0x0C807F06, 0xE32BED71),
MK_64(0x60C13A52, 0xB41A91F6),
MK_64(0x9716D35D, 0xD4917C38),
MK_64(0xE780DF12, 0x6FD31D3A),
MK_64(0x797846B6, 0xC898303A),
MK_64(0xB172C2A8, 0xB3572A3B),
MK_64(0xC9BC8203, 0xA6104A6C),
MK_64(0x65909338, 0xD75624F4),
MK_64(0x94BCC568, 0x4B3F81A0),
MK_64(0x3EBBF51E, 0x10ECFD46),
MK_64(0x2DF50F0B, 0xEEB08542),
MK_64(0x3B5A6530, 0x0DBC6516),
MK_64(0x484B9CD2, 0x167BBCE1),
MK_64(0x2D136947, 0xD4CBAFEA)
};
/* blkSize = 1024 bits. hashSize = 512 bits */
const uint64_t SKEIN1024_IV_512[] = {
MK_64(0xCAEC0E5D, 0x7C1B1B18),
MK_64(0xA01B0E04, 0x5F03E802),
MK_64(0x33840451, 0xED912885),
MK_64(0x374AFB04, 0xEAEC2E1C),
MK_64(0xDF25A0E2, 0x813581F7),
MK_64(0xE4004093, 0x8B12F9D2),
MK_64(0xA662D539, 0xC2ED39B6),
MK_64(0xFA8B85CF, 0x45D8C75A),
MK_64(0x8316ED8E, 0x29EDE796),
MK_64(0x053289C0, 0x2E9F91B8),
MK_64(0xC3F8EF1D, 0x6D518B73),
MK_64(0xBDCEC3C4, 0xD5EF332E),
MK_64(0x549A7E52, 0x22974487),
MK_64(0x67070872, 0x5B749816),
MK_64(0xB9CD28FB, 0xF0581BD1),
MK_64(0x0E2940B8, 0x15804974)
};
/* blkSize = 1024 bits. hashSize = 1024 bits */
const uint64_t SKEIN1024_IV_1024[] = {
MK_64(0xD593DA07, 0x41E72355),
MK_64(0x15B5E511, 0xAC73E00C),
MK_64(0x5180E5AE, 0xBAF2C4F0),
MK_64(0x03BD41D3, 0xFCBCAFAF),
MK_64(0x1CAEC6FD, 0x1983A898),
MK_64(0x6E510B8B, 0xCDD0589F),
MK_64(0x77E2BDFD, 0xC6394ADA),
MK_64(0xC11E1DB5, 0x24DCB0A3),
MK_64(0xD6D14AF9, 0xC6329AB5),
MK_64(0x6A9B0BFC, 0x6EB67E0D),
MK_64(0x9243C60D, 0xCCFF1332),
MK_64(0x1A1F1DDE, 0x743F02D4),
MK_64(0x0996753C, 0x10ED0BB8),
MK_64(0x6572DD22, 0xF2B4969A),
MK_64(0x61FD3062, 0xD00A579A),
MK_64(0x1DE0536E, 0x8682E539)
};
diff --git a/module/icp/api/kcf_miscapi.c b/module/icp/api/kcf_miscapi.c
index c0f415b264a7..5c0d60391f44 100644
--- a/module/icp/api/kcf_miscapi.c
+++ b/module/icp/api/kcf_miscapi.c
@@ -1,127 +1,127 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
#include <sys/crypto/api.h>
#include <sys/crypto/impl.h>
#include <sys/crypto/sched_impl.h>
/*
* All event subscribers are put on a list. kcf_notify_list_lock
* protects changes to this list.
*
* The following locking order is maintained in the code - The
* global kcf_notify_list_lock followed by the individual lock
* in a kcf_ntfy_elem structure (kn_lock).
*/
kmutex_t ntfy_list_lock;
kcondvar_t ntfy_list_cv; /* cv the service thread waits on */
static kcf_ntfy_elem_t *ntfy_list_head;
/*
* crypto_mech2id()
*
* Arguments:
* . mechname: A null-terminated string identifying the mechanism name.
*
* Description:
* Walks the mechanisms tables, looking for an entry that matches the
* mechname. Once it find it, it builds the 64-bit mech_type and returns
* it. If there are no hardware or software providers for the mechanism,
* but there is an unloaded software provider, this routine will attempt
* to load it.
*
* Context:
* Process and interruption.
*
* Returns:
* The unique mechanism identified by 'mechname', if found.
* CRYPTO_MECH_INVALID otherwise.
*/
crypto_mech_type_t
-crypto_mech2id(char *mechname)
+crypto_mech2id(const char *mechname)
{
return (crypto_mech2id_common(mechname, B_TRUE));
}
/*
* We walk the notification list and do the callbacks.
*/
void
kcf_walk_ntfylist(uint32_t event, void *event_arg)
{
kcf_ntfy_elem_t *nep;
int nelem = 0;
mutex_enter(&ntfy_list_lock);
/*
* Count how many clients are on the notification list. We need
* this count to ensure that clients which joined the list after we
* have started this walk, are not wrongly notified.
*/
for (nep = ntfy_list_head; nep != NULL; nep = nep->kn_next)
nelem++;
for (nep = ntfy_list_head; (nep != NULL && nelem); nep = nep->kn_next) {
nelem--;
/*
* Check if this client is interested in the
* event.
*/
if (!(nep->kn_event_mask & event))
continue;
mutex_enter(&nep->kn_lock);
nep->kn_state = NTFY_RUNNING;
mutex_exit(&nep->kn_lock);
mutex_exit(&ntfy_list_lock);
/*
* We invoke the callback routine with no locks held. Another
* client could have joined the list meanwhile. This is fine
* as we maintain nelem as stated above. The NULL check in the
* for loop guards against shrinkage. Also, any callers of
* crypto_unnotify_events() at this point cv_wait till kn_state
* changes to NTFY_WAITING. Hence, nep is assured to be valid.
*/
(*nep->kn_func)(event, event_arg);
mutex_enter(&nep->kn_lock);
nep->kn_state = NTFY_WAITING;
cv_broadcast(&nep->kn_cv);
mutex_exit(&nep->kn_lock);
mutex_enter(&ntfy_list_lock);
}
mutex_exit(&ntfy_list_lock);
}
#if defined(_KERNEL)
EXPORT_SYMBOL(crypto_mech2id);
#endif
diff --git a/module/icp/asm-x86_64/aes/aes_amd64.S b/module/icp/asm-x86_64/aes/aes_amd64.S
index 272720e517e3..d16cc9996e25 100644
--- a/module/icp/asm-x86_64/aes/aes_amd64.S
+++ b/module/icp/asm-x86_64/aes/aes_amd64.S
@@ -1,906 +1,906 @@
/*
* ---------------------------------------------------------------------------
* Copyright (c) 1998-2007, Brian Gladman, Worcester, UK. All rights reserved.
*
* LICENSE TERMS
*
* The free distribution and use of this software is allowed (with or without
* changes) provided that:
*
* 1. source code distributions include the above copyright notice, this
* list of conditions and the following disclaimer;
*
* 2. binary distributions include the above copyright notice, this list
* of conditions and the following disclaimer in their documentation;
*
* 3. the name of the copyright holder is not used to endorse products
* built using this software without specific written permission.
*
* DISCLAIMER
*
* This software is provided 'as is' with no explicit or implied warranties
* in respect of its properties, including, but not limited to, correctness
* and/or fitness for purpose.
* ---------------------------------------------------------------------------
* Issue 20/12/2007
*
* I am grateful to Dag Arne Osvik for many discussions of the techniques that
* can be used to optimise AES assembler code on AMD64/EM64T architectures.
* Some of the techniques used in this implementation are the result of
* suggestions made by him for which I am most grateful.
*
* An AES implementation for AMD64 processors using the YASM assembler. This
* implementation provides only encryption, decryption and hence requires key
* scheduling support in C. It uses 8k bytes of tables but its encryption and
* decryption performance is very close to that obtained using large tables.
* It can use either MS Windows or Gnu/Linux/OpenSolaris OS calling conventions,
* which are as follows:
* ms windows gnu/linux/opensolaris os
*
* in_blk rcx rdi
* out_blk rdx rsi
* context (cx) r8 rdx
*
* preserved rsi - + rbx, rbp, rsp, r12, r13, r14 & r15
* registers rdi - on both
*
* destroyed - rsi + rax, rcx, rdx, r8, r9, r10 & r11
* registers - rdi on both
*
* The convention used here is that for gnu/linux/opensolaris os.
*
* This code provides the standard AES block size (128 bits, 16 bytes) and the
* three standard AES key sizes (128, 192 and 256 bits). It has the same call
* interface as my C implementation. It uses the Microsoft C AMD64 calling
* conventions in which the three parameters are placed in rcx, rdx and r8
* respectively. The rbx, rsi, rdi, rbp and r12..r15 registers are preserved.
*
* OpenSolaris Note:
* Modified to use GNU/Linux/Solaris calling conventions.
* That is parameters are placed in rdi, rsi, rdx, and rcx, respectively.
*
* AES_RETURN aes_encrypt(const unsigned char in_blk[],
* unsigned char out_blk[], const aes_encrypt_ctx cx[1])/
*
* AES_RETURN aes_decrypt(const unsigned char in_blk[],
* unsigned char out_blk[], const aes_decrypt_ctx cx[1])/
*
* AES_RETURN aes_encrypt_key<NNN>(const unsigned char key[],
* const aes_encrypt_ctx cx[1])/
*
* AES_RETURN aes_decrypt_key<NNN>(const unsigned char key[],
* const aes_decrypt_ctx cx[1])/
*
* AES_RETURN aes_encrypt_key(const unsigned char key[],
* unsigned int len, const aes_decrypt_ctx cx[1])/
*
* AES_RETURN aes_decrypt_key(const unsigned char key[],
* unsigned int len, const aes_decrypt_ctx cx[1])/
*
* where <NNN> is 128, 102 or 256. In the last two calls the length can be in
* either bits or bytes.
*
* Comment in/out the following lines to obtain the desired subroutines. These
* selections MUST match those in the C header file aesopt.h
*/
#define AES_REV_DKS /* define if key decryption schedule is reversed */
#define LAST_ROUND_TABLES /* define for the faster version using extra tables */
/*
* The encryption key schedule has the following in memory layout where N is the
* number of rounds (10, 12 or 14):
*
* lo: | input key (round 0) | / each round is four 32-bit words
* | encryption round 1 |
* | encryption round 2 |
* ....
* | encryption round N-1 |
* hi: | encryption round N |
*
* The decryption key schedule is normally set up so that it has the same
* layout as above by actually reversing the order of the encryption key
* schedule in memory (this happens when AES_REV_DKS is set):
*
* lo: | decryption round 0 | = | encryption round N |
* | decryption round 1 | = INV_MIX_COL[ | encryption round N-1 | ]
* | decryption round 2 | = INV_MIX_COL[ | encryption round N-2 | ]
* .... ....
* | decryption round N-1 | = INV_MIX_COL[ | encryption round 1 | ]
* hi: | decryption round N | = | input key (round 0) |
*
* with rounds except the first and last modified using inv_mix_column()
* But if AES_REV_DKS is NOT set the order of keys is left as it is for
* encryption so that it has to be accessed in reverse when used for
* decryption (although the inverse mix column modifications are done)
*
* lo: | decryption round 0 | = | input key (round 0) |
* | decryption round 1 | = INV_MIX_COL[ | encryption round 1 | ]
* | decryption round 2 | = INV_MIX_COL[ | encryption round 2 | ]
* .... ....
* | decryption round N-1 | = INV_MIX_COL[ | encryption round N-1 | ]
* hi: | decryption round N | = | encryption round N |
*
* This layout is faster when the assembler key scheduling provided here
* is used.
*
* End of user defines
*/
/*
* ---------------------------------------------------------------------------
* OpenSolaris OS modifications
*
* This source originates from Brian Gladman file aes_amd64.asm
* in http://fp.gladman.plus.com/AES/aes-src-04-03-08.zip
* with these changes:
*
* 1. Removed MS Windows-specific code within DLL_EXPORT, _SEH_, and
* !__GNUC__ ifdefs. Also removed ENCRYPTION, DECRYPTION,
* AES_128, AES_192, AES_256, AES_VAR ifdefs.
*
* 2. Translate yasm/nasm %define and .macro definitions to cpp(1) #define
*
* 3. Translate yasm/nasm %ifdef/%ifndef to cpp(1) #ifdef
*
* 4. Translate Intel/yasm/nasm syntax to ATT/OpenSolaris as(1) syntax
* (operands reversed, literals prefixed with "$", registers prefixed with "%",
* and "[register+offset]", addressing changed to "offset(register)",
* parenthesis in constant expressions "()" changed to square brackets "[]",
* "." removed from local (numeric) labels, and other changes.
* Examples:
* Intel/yasm/nasm Syntax ATT/OpenSolaris Syntax
* mov rax,(4*20h) mov $[4*0x20],%rax
* mov rax,[ebx+20h] mov 0x20(%ebx),%rax
* lea rax,[ebx+ecx] lea (%ebx,%ecx),%rax
* sub rax,[ebx+ecx*4-20h] sub -0x20(%ebx,%ecx,4),%rax
*
* 5. Added OpenSolaris ENTRY_NP/SET_SIZE macros from
* /usr/include/sys/asm_linkage.h, lint(1B) guards, and dummy C function
* definitions for lint.
*
* 6. Renamed functions and reordered parameters to match OpenSolaris:
* Original Gladman interface:
* int aes_encrypt(const unsigned char *in,
* unsigned char *out, const aes_encrypt_ctx cx[1])/
* int aes_decrypt(const unsigned char *in,
* unsigned char *out, const aes_encrypt_ctx cx[1])/
* Note: aes_encrypt_ctx contains ks, a 60 element array of uint32_t,
* and a union type, inf., containing inf.l, a uint32_t and
* inf.b, a 4-element array of uint32_t. Only b[0] in the array (aka "l") is
* used and contains the key schedule length * 16 where key schedule length is
* 10, 12, or 14 bytes.
*
* OpenSolaris OS interface:
* void aes_encrypt_amd64(const aes_ks_t *ks, int Nr,
* const uint32_t pt[4], uint32_t ct[4])/
* void aes_decrypt_amd64(const aes_ks_t *ks, int Nr,
* const uint32_t pt[4], uint32_t ct[4])/
* typedef union {uint64_t ks64[(MAX_AES_NR + 1) * 4]/
* uint32_t ks32[(MAX_AES_NR + 1) * 4]/ } aes_ks_t/
* Note: ks is the AES key schedule, Nr is number of rounds, pt is plain text,
* ct is crypto text, and MAX_AES_NR is 14.
* For the x86 64-bit architecture, OpenSolaris OS uses ks32 instead of ks64.
*/
#if defined(lint) || defined(__lint)
#include <sys/types.h>
void
aes_encrypt_amd64(const uint32_t rk[], int Nr, const uint32_t pt[4],
uint32_t ct[4]) {
(void) rk, (void) Nr, (void) pt, (void) ct;
}
void
aes_decrypt_amd64(const uint32_t rk[], int Nr, const uint32_t ct[4],
uint32_t pt[4]) {
(void) rk, (void) Nr, (void) pt, (void) ct;
}
#else
#define _ASM
#include <sys/asm_linkage.h>
#define KS_LENGTH 60
#define raxd eax
#define rdxd edx
#define rcxd ecx
#define rbxd ebx
#define rsid esi
#define rdid edi
#define raxb al
#define rdxb dl
#define rcxb cl
#define rbxb bl
#define rsib sil
#define rdib dil
// finite field multiplies by {02}, {04} and {08}
#define f2(x) [[x<<1]^[[[x>>7]&1]*0x11b]]
#define f4(x) [[x<<2]^[[[x>>6]&1]*0x11b]^[[[x>>6]&2]*0x11b]]
#define f8(x) [[x<<3]^[[[x>>5]&1]*0x11b]^[[[x>>5]&2]*0x11b]^[[[x>>5]&4]*0x11b]]
// finite field multiplies required in table generation
#define f3(x) [[f2(x)] ^ [x]]
#define f9(x) [[f8(x)] ^ [x]]
#define fb(x) [[f8(x)] ^ [f2(x)] ^ [x]]
#define fd(x) [[f8(x)] ^ [f4(x)] ^ [x]]
#define fe(x) [[f8(x)] ^ [f4(x)] ^ [f2(x)]]
// macros for expanding S-box data
#define u8(x) [f2(x)], [x], [x], [f3(x)], [f2(x)], [x], [x], [f3(x)]
#define v8(x) [fe(x)], [f9(x)], [fd(x)], [fb(x)], [fe(x)], [f9(x)], [fd(x)], [x]
#define w8(x) [x], 0, 0, 0, [x], 0, 0, 0
#define enc_vals(x) \
.byte x(0x63),x(0x7c),x(0x77),x(0x7b),x(0xf2),x(0x6b),x(0x6f),x(0xc5); \
.byte x(0x30),x(0x01),x(0x67),x(0x2b),x(0xfe),x(0xd7),x(0xab),x(0x76); \
.byte x(0xca),x(0x82),x(0xc9),x(0x7d),x(0xfa),x(0x59),x(0x47),x(0xf0); \
.byte x(0xad),x(0xd4),x(0xa2),x(0xaf),x(0x9c),x(0xa4),x(0x72),x(0xc0); \
.byte x(0xb7),x(0xfd),x(0x93),x(0x26),x(0x36),x(0x3f),x(0xf7),x(0xcc); \
.byte x(0x34),x(0xa5),x(0xe5),x(0xf1),x(0x71),x(0xd8),x(0x31),x(0x15); \
.byte x(0x04),x(0xc7),x(0x23),x(0xc3),x(0x18),x(0x96),x(0x05),x(0x9a); \
.byte x(0x07),x(0x12),x(0x80),x(0xe2),x(0xeb),x(0x27),x(0xb2),x(0x75); \
.byte x(0x09),x(0x83),x(0x2c),x(0x1a),x(0x1b),x(0x6e),x(0x5a),x(0xa0); \
.byte x(0x52),x(0x3b),x(0xd6),x(0xb3),x(0x29),x(0xe3),x(0x2f),x(0x84); \
.byte x(0x53),x(0xd1),x(0x00),x(0xed),x(0x20),x(0xfc),x(0xb1),x(0x5b); \
.byte x(0x6a),x(0xcb),x(0xbe),x(0x39),x(0x4a),x(0x4c),x(0x58),x(0xcf); \
.byte x(0xd0),x(0xef),x(0xaa),x(0xfb),x(0x43),x(0x4d),x(0x33),x(0x85); \
.byte x(0x45),x(0xf9),x(0x02),x(0x7f),x(0x50),x(0x3c),x(0x9f),x(0xa8); \
.byte x(0x51),x(0xa3),x(0x40),x(0x8f),x(0x92),x(0x9d),x(0x38),x(0xf5); \
.byte x(0xbc),x(0xb6),x(0xda),x(0x21),x(0x10),x(0xff),x(0xf3),x(0xd2); \
.byte x(0xcd),x(0x0c),x(0x13),x(0xec),x(0x5f),x(0x97),x(0x44),x(0x17); \
.byte x(0xc4),x(0xa7),x(0x7e),x(0x3d),x(0x64),x(0x5d),x(0x19),x(0x73); \
.byte x(0x60),x(0x81),x(0x4f),x(0xdc),x(0x22),x(0x2a),x(0x90),x(0x88); \
.byte x(0x46),x(0xee),x(0xb8),x(0x14),x(0xde),x(0x5e),x(0x0b),x(0xdb); \
.byte x(0xe0),x(0x32),x(0x3a),x(0x0a),x(0x49),x(0x06),x(0x24),x(0x5c); \
.byte x(0xc2),x(0xd3),x(0xac),x(0x62),x(0x91),x(0x95),x(0xe4),x(0x79); \
.byte x(0xe7),x(0xc8),x(0x37),x(0x6d),x(0x8d),x(0xd5),x(0x4e),x(0xa9); \
.byte x(0x6c),x(0x56),x(0xf4),x(0xea),x(0x65),x(0x7a),x(0xae),x(0x08); \
.byte x(0xba),x(0x78),x(0x25),x(0x2e),x(0x1c),x(0xa6),x(0xb4),x(0xc6); \
.byte x(0xe8),x(0xdd),x(0x74),x(0x1f),x(0x4b),x(0xbd),x(0x8b),x(0x8a); \
.byte x(0x70),x(0x3e),x(0xb5),x(0x66),x(0x48),x(0x03),x(0xf6),x(0x0e); \
.byte x(0x61),x(0x35),x(0x57),x(0xb9),x(0x86),x(0xc1),x(0x1d),x(0x9e); \
.byte x(0xe1),x(0xf8),x(0x98),x(0x11),x(0x69),x(0xd9),x(0x8e),x(0x94); \
.byte x(0x9b),x(0x1e),x(0x87),x(0xe9),x(0xce),x(0x55),x(0x28),x(0xdf); \
.byte x(0x8c),x(0xa1),x(0x89),x(0x0d),x(0xbf),x(0xe6),x(0x42),x(0x68); \
.byte x(0x41),x(0x99),x(0x2d),x(0x0f),x(0xb0),x(0x54),x(0xbb),x(0x16)
#define dec_vals(x) \
.byte x(0x52),x(0x09),x(0x6a),x(0xd5),x(0x30),x(0x36),x(0xa5),x(0x38); \
.byte x(0xbf),x(0x40),x(0xa3),x(0x9e),x(0x81),x(0xf3),x(0xd7),x(0xfb); \
.byte x(0x7c),x(0xe3),x(0x39),x(0x82),x(0x9b),x(0x2f),x(0xff),x(0x87); \
.byte x(0x34),x(0x8e),x(0x43),x(0x44),x(0xc4),x(0xde),x(0xe9),x(0xcb); \
.byte x(0x54),x(0x7b),x(0x94),x(0x32),x(0xa6),x(0xc2),x(0x23),x(0x3d); \
.byte x(0xee),x(0x4c),x(0x95),x(0x0b),x(0x42),x(0xfa),x(0xc3),x(0x4e); \
.byte x(0x08),x(0x2e),x(0xa1),x(0x66),x(0x28),x(0xd9),x(0x24),x(0xb2); \
.byte x(0x76),x(0x5b),x(0xa2),x(0x49),x(0x6d),x(0x8b),x(0xd1),x(0x25); \
.byte x(0x72),x(0xf8),x(0xf6),x(0x64),x(0x86),x(0x68),x(0x98),x(0x16); \
.byte x(0xd4),x(0xa4),x(0x5c),x(0xcc),x(0x5d),x(0x65),x(0xb6),x(0x92); \
.byte x(0x6c),x(0x70),x(0x48),x(0x50),x(0xfd),x(0xed),x(0xb9),x(0xda); \
.byte x(0x5e),x(0x15),x(0x46),x(0x57),x(0xa7),x(0x8d),x(0x9d),x(0x84); \
.byte x(0x90),x(0xd8),x(0xab),x(0x00),x(0x8c),x(0xbc),x(0xd3),x(0x0a); \
.byte x(0xf7),x(0xe4),x(0x58),x(0x05),x(0xb8),x(0xb3),x(0x45),x(0x06); \
.byte x(0xd0),x(0x2c),x(0x1e),x(0x8f),x(0xca),x(0x3f),x(0x0f),x(0x02); \
.byte x(0xc1),x(0xaf),x(0xbd),x(0x03),x(0x01),x(0x13),x(0x8a),x(0x6b); \
.byte x(0x3a),x(0x91),x(0x11),x(0x41),x(0x4f),x(0x67),x(0xdc),x(0xea); \
.byte x(0x97),x(0xf2),x(0xcf),x(0xce),x(0xf0),x(0xb4),x(0xe6),x(0x73); \
.byte x(0x96),x(0xac),x(0x74),x(0x22),x(0xe7),x(0xad),x(0x35),x(0x85); \
.byte x(0xe2),x(0xf9),x(0x37),x(0xe8),x(0x1c),x(0x75),x(0xdf),x(0x6e); \
.byte x(0x47),x(0xf1),x(0x1a),x(0x71),x(0x1d),x(0x29),x(0xc5),x(0x89); \
.byte x(0x6f),x(0xb7),x(0x62),x(0x0e),x(0xaa),x(0x18),x(0xbe),x(0x1b); \
.byte x(0xfc),x(0x56),x(0x3e),x(0x4b),x(0xc6),x(0xd2),x(0x79),x(0x20); \
.byte x(0x9a),x(0xdb),x(0xc0),x(0xfe),x(0x78),x(0xcd),x(0x5a),x(0xf4); \
.byte x(0x1f),x(0xdd),x(0xa8),x(0x33),x(0x88),x(0x07),x(0xc7),x(0x31); \
.byte x(0xb1),x(0x12),x(0x10),x(0x59),x(0x27),x(0x80),x(0xec),x(0x5f); \
.byte x(0x60),x(0x51),x(0x7f),x(0xa9),x(0x19),x(0xb5),x(0x4a),x(0x0d); \
.byte x(0x2d),x(0xe5),x(0x7a),x(0x9f),x(0x93),x(0xc9),x(0x9c),x(0xef); \
.byte x(0xa0),x(0xe0),x(0x3b),x(0x4d),x(0xae),x(0x2a),x(0xf5),x(0xb0); \
.byte x(0xc8),x(0xeb),x(0xbb),x(0x3c),x(0x83),x(0x53),x(0x99),x(0x61); \
.byte x(0x17),x(0x2b),x(0x04),x(0x7e),x(0xba),x(0x77),x(0xd6),x(0x26); \
.byte x(0xe1),x(0x69),x(0x14),x(0x63),x(0x55),x(0x21),x(0x0c),x(0x7d)
#define tptr %rbp /* table pointer */
#define kptr %r8 /* key schedule pointer */
#define fofs 128 /* adjust offset in key schedule to keep |disp| < 128 */
#define fk_ref(x, y) -16*x+fofs+4*y(kptr)
#ifdef AES_REV_DKS
#define rofs 128
#define ik_ref(x, y) -16*x+rofs+4*y(kptr)
#else
#define rofs -128
#define ik_ref(x, y) 16*x+rofs+4*y(kptr)
#endif /* AES_REV_DKS */
#define tab_0(x) (tptr,x,8)
#define tab_1(x) 3(tptr,x,8)
#define tab_2(x) 2(tptr,x,8)
#define tab_3(x) 1(tptr,x,8)
#define tab_f(x) 1(tptr,x,8)
#define tab_i(x) 7(tptr,x,8)
#define ff_rnd(p1, p2, p3, p4, round) /* normal forward round */ \
mov fk_ref(round,0), p1; \
mov fk_ref(round,1), p2; \
mov fk_ref(round,2), p3; \
mov fk_ref(round,3), p4; \
\
movzx %al, %esi; \
movzx %ah, %edi; \
shr $16, %eax; \
xor tab_0(%rsi), p1; \
xor tab_1(%rdi), p4; \
movzx %al, %esi; \
movzx %ah, %edi; \
xor tab_2(%rsi), p3; \
xor tab_3(%rdi), p2; \
\
movzx %bl, %esi; \
movzx %bh, %edi; \
shr $16, %ebx; \
xor tab_0(%rsi), p2; \
xor tab_1(%rdi), p1; \
movzx %bl, %esi; \
movzx %bh, %edi; \
xor tab_2(%rsi), p4; \
xor tab_3(%rdi), p3; \
\
movzx %cl, %esi; \
movzx %ch, %edi; \
shr $16, %ecx; \
xor tab_0(%rsi), p3; \
xor tab_1(%rdi), p2; \
movzx %cl, %esi; \
movzx %ch, %edi; \
xor tab_2(%rsi), p1; \
xor tab_3(%rdi), p4; \
\
movzx %dl, %esi; \
movzx %dh, %edi; \
shr $16, %edx; \
xor tab_0(%rsi), p4; \
xor tab_1(%rdi), p3; \
movzx %dl, %esi; \
movzx %dh, %edi; \
xor tab_2(%rsi), p2; \
xor tab_3(%rdi), p1; \
\
mov p1, %eax; \
mov p2, %ebx; \
mov p3, %ecx; \
mov p4, %edx
#ifdef LAST_ROUND_TABLES
#define fl_rnd(p1, p2, p3, p4, round) /* last forward round */ \
add $2048, tptr; \
mov fk_ref(round,0), p1; \
mov fk_ref(round,1), p2; \
mov fk_ref(round,2), p3; \
mov fk_ref(round,3), p4; \
\
movzx %al, %esi; \
movzx %ah, %edi; \
shr $16, %eax; \
xor tab_0(%rsi), p1; \
xor tab_1(%rdi), p4; \
movzx %al, %esi; \
movzx %ah, %edi; \
xor tab_2(%rsi), p3; \
xor tab_3(%rdi), p2; \
\
movzx %bl, %esi; \
movzx %bh, %edi; \
shr $16, %ebx; \
xor tab_0(%rsi), p2; \
xor tab_1(%rdi), p1; \
movzx %bl, %esi; \
movzx %bh, %edi; \
xor tab_2(%rsi), p4; \
xor tab_3(%rdi), p3; \
\
movzx %cl, %esi; \
movzx %ch, %edi; \
shr $16, %ecx; \
xor tab_0(%rsi), p3; \
xor tab_1(%rdi), p2; \
movzx %cl, %esi; \
movzx %ch, %edi; \
xor tab_2(%rsi), p1; \
xor tab_3(%rdi), p4; \
\
movzx %dl, %esi; \
movzx %dh, %edi; \
shr $16, %edx; \
xor tab_0(%rsi), p4; \
xor tab_1(%rdi), p3; \
movzx %dl, %esi; \
movzx %dh, %edi; \
xor tab_2(%rsi), p2; \
xor tab_3(%rdi), p1
#else
#define fl_rnd(p1, p2, p3, p4, round) /* last forward round */ \
mov fk_ref(round,0), p1; \
mov fk_ref(round,1), p2; \
mov fk_ref(round,2), p3; \
mov fk_ref(round,3), p4; \
\
movzx %al, %esi; \
movzx %ah, %edi; \
shr $16, %eax; \
movzx tab_f(%rsi), %esi; \
movzx tab_f(%rdi), %edi; \
xor %esi, p1; \
rol $8, %edi; \
xor %edi, p4; \
movzx %al, %esi; \
movzx %ah, %edi; \
movzx tab_f(%rsi), %esi; \
movzx tab_f(%rdi), %edi; \
rol $16, %esi; \
rol $24, %edi; \
xor %esi, p3; \
xor %edi, p2; \
\
movzx %bl, %esi; \
movzx %bh, %edi; \
shr $16, %ebx; \
movzx tab_f(%rsi), %esi; \
movzx tab_f(%rdi), %edi; \
xor %esi, p2; \
rol $8, %edi; \
xor %edi, p1; \
movzx %bl, %esi; \
movzx %bh, %edi; \
movzx tab_f(%rsi), %esi; \
movzx tab_f(%rdi), %edi; \
rol $16, %esi; \
rol $24, %edi; \
xor %esi, p4; \
xor %edi, p3; \
\
movzx %cl, %esi; \
movzx %ch, %edi; \
movzx tab_f(%rsi), %esi; \
movzx tab_f(%rdi), %edi; \
shr $16, %ecx; \
xor %esi, p3; \
rol $8, %edi; \
xor %edi, p2; \
movzx %cl, %esi; \
movzx %ch, %edi; \
movzx tab_f(%rsi), %esi; \
movzx tab_f(%rdi), %edi; \
rol $16, %esi; \
rol $24, %edi; \
xor %esi, p1; \
xor %edi, p4; \
\
movzx %dl, %esi; \
movzx %dh, %edi; \
movzx tab_f(%rsi), %esi; \
movzx tab_f(%rdi), %edi; \
shr $16, %edx; \
xor %esi, p4; \
rol $8, %edi; \
xor %edi, p3; \
movzx %dl, %esi; \
movzx %dh, %edi; \
movzx tab_f(%rsi), %esi; \
movzx tab_f(%rdi), %edi; \
rol $16, %esi; \
rol $24, %edi; \
xor %esi, p2; \
xor %edi, p1
#endif /* LAST_ROUND_TABLES */
#define ii_rnd(p1, p2, p3, p4, round) /* normal inverse round */ \
mov ik_ref(round,0), p1; \
mov ik_ref(round,1), p2; \
mov ik_ref(round,2), p3; \
mov ik_ref(round,3), p4; \
\
movzx %al, %esi; \
movzx %ah, %edi; \
shr $16, %eax; \
xor tab_0(%rsi), p1; \
xor tab_1(%rdi), p2; \
movzx %al, %esi; \
movzx %ah, %edi; \
xor tab_2(%rsi), p3; \
xor tab_3(%rdi), p4; \
\
movzx %bl, %esi; \
movzx %bh, %edi; \
shr $16, %ebx; \
xor tab_0(%rsi), p2; \
xor tab_1(%rdi), p3; \
movzx %bl, %esi; \
movzx %bh, %edi; \
xor tab_2(%rsi), p4; \
xor tab_3(%rdi), p1; \
\
movzx %cl, %esi; \
movzx %ch, %edi; \
shr $16, %ecx; \
xor tab_0(%rsi), p3; \
xor tab_1(%rdi), p4; \
movzx %cl, %esi; \
movzx %ch, %edi; \
xor tab_2(%rsi), p1; \
xor tab_3(%rdi), p2; \
\
movzx %dl, %esi; \
movzx %dh, %edi; \
shr $16, %edx; \
xor tab_0(%rsi), p4; \
xor tab_1(%rdi), p1; \
movzx %dl, %esi; \
movzx %dh, %edi; \
xor tab_2(%rsi), p2; \
xor tab_3(%rdi), p3; \
\
mov p1, %eax; \
mov p2, %ebx; \
mov p3, %ecx; \
mov p4, %edx
#ifdef LAST_ROUND_TABLES
#define il_rnd(p1, p2, p3, p4, round) /* last inverse round */ \
add $2048, tptr; \
mov ik_ref(round,0), p1; \
mov ik_ref(round,1), p2; \
mov ik_ref(round,2), p3; \
mov ik_ref(round,3), p4; \
\
movzx %al, %esi; \
movzx %ah, %edi; \
shr $16, %eax; \
xor tab_0(%rsi), p1; \
xor tab_1(%rdi), p2; \
movzx %al, %esi; \
movzx %ah, %edi; \
xor tab_2(%rsi), p3; \
xor tab_3(%rdi), p4; \
\
movzx %bl, %esi; \
movzx %bh, %edi; \
shr $16, %ebx; \
xor tab_0(%rsi), p2; \
xor tab_1(%rdi), p3; \
movzx %bl, %esi; \
movzx %bh, %edi; \
xor tab_2(%rsi), p4; \
xor tab_3(%rdi), p1; \
\
movzx %cl, %esi; \
movzx %ch, %edi; \
shr $16, %ecx; \
xor tab_0(%rsi), p3; \
xor tab_1(%rdi), p4; \
movzx %cl, %esi; \
movzx %ch, %edi; \
xor tab_2(%rsi), p1; \
xor tab_3(%rdi), p2; \
\
movzx %dl, %esi; \
movzx %dh, %edi; \
shr $16, %edx; \
xor tab_0(%rsi), p4; \
xor tab_1(%rdi), p1; \
movzx %dl, %esi; \
movzx %dh, %edi; \
xor tab_2(%rsi), p2; \
xor tab_3(%rdi), p3
#else
#define il_rnd(p1, p2, p3, p4, round) /* last inverse round */ \
mov ik_ref(round,0), p1; \
mov ik_ref(round,1), p2; \
mov ik_ref(round,2), p3; \
mov ik_ref(round,3), p4; \
\
movzx %al, %esi; \
movzx %ah, %edi; \
movzx tab_i(%rsi), %esi; \
movzx tab_i(%rdi), %edi; \
shr $16, %eax; \
xor %esi, p1; \
rol $8, %edi; \
xor %edi, p2; \
movzx %al, %esi; \
movzx %ah, %edi; \
movzx tab_i(%rsi), %esi; \
movzx tab_i(%rdi), %edi; \
rol $16, %esi; \
rol $24, %edi; \
xor %esi, p3; \
xor %edi, p4; \
\
movzx %bl, %esi; \
movzx %bh, %edi; \
movzx tab_i(%rsi), %esi; \
movzx tab_i(%rdi), %edi; \
shr $16, %ebx; \
xor %esi, p2; \
rol $8, %edi; \
xor %edi, p3; \
movzx %bl, %esi; \
movzx %bh, %edi; \
movzx tab_i(%rsi), %esi; \
movzx tab_i(%rdi), %edi; \
rol $16, %esi; \
rol $24, %edi; \
xor %esi, p4; \
xor %edi, p1; \
\
movzx %cl, %esi; \
movzx %ch, %edi; \
movzx tab_i(%rsi), %esi; \
movzx tab_i(%rdi), %edi; \
shr $16, %ecx; \
xor %esi, p3; \
rol $8, %edi; \
xor %edi, p4; \
movzx %cl, %esi; \
movzx %ch, %edi; \
movzx tab_i(%rsi), %esi; \
movzx tab_i(%rdi), %edi; \
rol $16, %esi; \
rol $24, %edi; \
xor %esi, p1; \
xor %edi, p2; \
\
movzx %dl, %esi; \
movzx %dh, %edi; \
movzx tab_i(%rsi), %esi; \
movzx tab_i(%rdi), %edi; \
shr $16, %edx; \
xor %esi, p4; \
rol $8, %edi; \
xor %edi, p1; \
movzx %dl, %esi; \
movzx %dh, %edi; \
movzx tab_i(%rsi), %esi; \
movzx tab_i(%rdi), %edi; \
rol $16, %esi; \
rol $24, %edi; \
xor %esi, p2; \
xor %edi, p3
#endif /* LAST_ROUND_TABLES */
/*
* OpenSolaris OS:
* void aes_encrypt_amd64(const aes_ks_t *ks, int Nr,
* const uint32_t pt[4], uint32_t ct[4])/
*
* Original interface:
* int aes_encrypt(const unsigned char *in,
* unsigned char *out, const aes_encrypt_ctx cx[1])/
*/
-.data
+.section .rodata
.align 64
enc_tab:
enc_vals(u8)
#ifdef LAST_ROUND_TABLES
// Last Round Tables:
enc_vals(w8)
#endif
ENTRY_NP(aes_encrypt_amd64)
#ifdef GLADMAN_INTERFACE
// Original interface
sub $[4*8], %rsp // gnu/linux/opensolaris binary interface
mov %rsi, (%rsp) // output pointer (P2)
mov %rdx, %r8 // context (P3)
mov %rbx, 1*8(%rsp) // P1: input pointer in rdi
mov %rbp, 2*8(%rsp) // P2: output pointer in (rsp)
mov %r12, 3*8(%rsp) // P3: context in r8
movzx 4*KS_LENGTH(kptr), %esi // Get byte key length * 16
#else
// OpenSolaris OS interface
sub $[4*8], %rsp // Make room on stack to save registers
mov %rcx, (%rsp) // Save output pointer (P4) on stack
mov %rdi, %r8 // context (P1)
mov %rdx, %rdi // P3: save input pointer
shl $4, %esi // P2: esi byte key length * 16
mov %rbx, 1*8(%rsp) // Save registers
mov %rbp, 2*8(%rsp)
mov %r12, 3*8(%rsp)
// P1: context in r8
// P2: byte key length * 16 in esi
// P3: input pointer in rdi
// P4: output pointer in (rsp)
#endif /* GLADMAN_INTERFACE */
lea enc_tab(%rip), tptr
sub $fofs, kptr
// Load input block into registers
mov (%rdi), %eax
mov 1*4(%rdi), %ebx
mov 2*4(%rdi), %ecx
mov 3*4(%rdi), %edx
xor fofs(kptr), %eax
xor fofs+4(kptr), %ebx
xor fofs+8(kptr), %ecx
xor fofs+12(kptr), %edx
lea (kptr,%rsi), kptr
// Jump based on byte key length * 16:
cmp $[10*16], %esi
je 3f
cmp $[12*16], %esi
je 2f
cmp $[14*16], %esi
je 1f
mov $-1, %rax // error
jmp 4f
// Perform normal forward rounds
1: ff_rnd(%r9d, %r10d, %r11d, %r12d, 13)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 12)
2: ff_rnd(%r9d, %r10d, %r11d, %r12d, 11)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 10)
3: ff_rnd(%r9d, %r10d, %r11d, %r12d, 9)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 8)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 7)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 6)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 5)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 4)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 3)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 2)
ff_rnd(%r9d, %r10d, %r11d, %r12d, 1)
fl_rnd(%r9d, %r10d, %r11d, %r12d, 0)
// Copy results
mov (%rsp), %rbx
mov %r9d, (%rbx)
mov %r10d, 4(%rbx)
mov %r11d, 8(%rbx)
mov %r12d, 12(%rbx)
xor %rax, %rax
4: // Restore registers
mov 1*8(%rsp), %rbx
mov 2*8(%rsp), %rbp
mov 3*8(%rsp), %r12
add $[4*8], %rsp
ret
SET_SIZE(aes_encrypt_amd64)
/*
* OpenSolaris OS:
* void aes_decrypt_amd64(const aes_ks_t *ks, int Nr,
* const uint32_t pt[4], uint32_t ct[4])/
*
* Original interface:
* int aes_decrypt(const unsigned char *in,
* unsigned char *out, const aes_encrypt_ctx cx[1])/
*/
-.data
+.section .rodata
.align 64
dec_tab:
dec_vals(v8)
#ifdef LAST_ROUND_TABLES
// Last Round Tables:
dec_vals(w8)
#endif
ENTRY_NP(aes_decrypt_amd64)
#ifdef GLADMAN_INTERFACE
// Original interface
sub $[4*8], %rsp // gnu/linux/opensolaris binary interface
mov %rsi, (%rsp) // output pointer (P2)
mov %rdx, %r8 // context (P3)
mov %rbx, 1*8(%rsp) // P1: input pointer in rdi
mov %rbp, 2*8(%rsp) // P2: output pointer in (rsp)
mov %r12, 3*8(%rsp) // P3: context in r8
movzx 4*KS_LENGTH(kptr), %esi // Get byte key length * 16
#else
// OpenSolaris OS interface
sub $[4*8], %rsp // Make room on stack to save registers
mov %rcx, (%rsp) // Save output pointer (P4) on stack
mov %rdi, %r8 // context (P1)
mov %rdx, %rdi // P3: save input pointer
shl $4, %esi // P2: esi byte key length * 16
mov %rbx, 1*8(%rsp) // Save registers
mov %rbp, 2*8(%rsp)
mov %r12, 3*8(%rsp)
// P1: context in r8
// P2: byte key length * 16 in esi
// P3: input pointer in rdi
// P4: output pointer in (rsp)
#endif /* GLADMAN_INTERFACE */
lea dec_tab(%rip), tptr
sub $rofs, kptr
// Load input block into registers
mov (%rdi), %eax
mov 1*4(%rdi), %ebx
mov 2*4(%rdi), %ecx
mov 3*4(%rdi), %edx
#ifdef AES_REV_DKS
mov kptr, %rdi
lea (kptr,%rsi), kptr
#else
lea (kptr,%rsi), %rdi
#endif
xor rofs(%rdi), %eax
xor rofs+4(%rdi), %ebx
xor rofs+8(%rdi), %ecx
xor rofs+12(%rdi), %edx
// Jump based on byte key length * 16:
cmp $[10*16], %esi
je 3f
cmp $[12*16], %esi
je 2f
cmp $[14*16], %esi
je 1f
mov $-1, %rax // error
jmp 4f
// Perform normal inverse rounds
1: ii_rnd(%r9d, %r10d, %r11d, %r12d, 13)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 12)
2: ii_rnd(%r9d, %r10d, %r11d, %r12d, 11)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 10)
3: ii_rnd(%r9d, %r10d, %r11d, %r12d, 9)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 8)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 7)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 6)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 5)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 4)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 3)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 2)
ii_rnd(%r9d, %r10d, %r11d, %r12d, 1)
il_rnd(%r9d, %r10d, %r11d, %r12d, 0)
// Copy results
mov (%rsp), %rbx
mov %r9d, (%rbx)
mov %r10d, 4(%rbx)
mov %r11d, 8(%rbx)
mov %r12d, 12(%rbx)
xor %rax, %rax
4: // Restore registers
mov 1*8(%rsp), %rbx
mov 2*8(%rsp), %rbp
mov 3*8(%rsp), %r12
add $[4*8], %rsp
ret
SET_SIZE(aes_decrypt_amd64)
#endif /* lint || __lint */
#ifdef __ELF__
.section .note.GNU-stack,"",%progbits
#endif
diff --git a/module/icp/asm-x86_64/modes/gcm_pclmulqdq.S b/module/icp/asm-x86_64/modes/gcm_pclmulqdq.S
index 0e1e04b78c5d..74eacbbe6388 100644
--- a/module/icp/asm-x86_64/modes/gcm_pclmulqdq.S
+++ b/module/icp/asm-x86_64/modes/gcm_pclmulqdq.S
@@ -1,254 +1,254 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2009 Intel Corporation
* All Rights Reserved.
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Accelerated GHASH implementation with Intel PCLMULQDQ-NI
* instructions. This file contains an accelerated
* Galois Field Multiplication implementation.
*
* PCLMULQDQ is used to accelerate the most time-consuming part of GHASH,
* carry-less multiplication. More information about PCLMULQDQ can be
* found at:
* http://software.intel.com/en-us/articles/
* carry-less-multiplication-and-its-usage-for-computing-the-gcm-mode/
*
*/
/*
* ====================================================================
* OpenSolaris OS modifications
*
* This source originates as file galois_hash_asm.c from
* Intel Corporation dated September 21, 2009.
*
* This OpenSolaris version has these major changes from the original source:
*
* 1. Added OpenSolaris ENTRY_NP/SET_SIZE macros from
* /usr/include/sys/asm_linkage.h, lint(1B) guards, and a dummy C function
* definition for lint.
*
* 2. Formatted code, added comments, and added #includes and #defines.
*
* 3. If bit CR0.TS is set, clear and set the TS bit, after and before
* calling kpreempt_disable() and kpreempt_enable().
* If the TS bit is not set, Save and restore %xmm registers at the beginning
* and end of function calls (%xmm* registers are not saved and restored by
* during kernel thread preemption).
*
* 4. Removed code to perform hashing. This is already done with C macro
* GHASH in gcm.c. For better performance, this removed code should be
* reintegrated in the future to replace the C GHASH macro.
*
* 5. Added code to byte swap 16-byte input and output.
*
* 6. Folded in comments from the original C source with embedded assembly
* (SB_w_shift_xor.c)
*
* 7. Renamed function and reordered parameters to match OpenSolaris:
* Intel interface:
* void galois_hash_asm(unsigned char *hk, unsigned char *s,
* unsigned char *d, int length)
* OpenSolaris OS interface:
* void gcm_mul_pclmulqdq(uint64_t *x_in, uint64_t *y, uint64_t *res);
* ====================================================================
*/
#if defined(lint) || defined(__lint) /* lint */
#include <sys/types.h>
void
gcm_mul_pclmulqdq(uint64_t *x_in, uint64_t *y, uint64_t *res) {
(void) x_in, (void) y, (void) res;
}
#elif defined(HAVE_PCLMULQDQ) /* guard by instruction set */
#define _ASM
#include <sys/asm_linkage.h>
/*
* Use this mask to byte-swap a 16-byte integer with the pshufb instruction
*/
// static uint8_t byte_swap16_mask[] = {
// 15, 14, 13, 12, 11, 10, 9, 8, 7, 6 ,5, 4, 3, 2, 1, 0 };
-.data
+.section .rodata
.align XMM_ALIGN
.Lbyte_swap16_mask:
.byte 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0
/*
* void gcm_mul_pclmulqdq(uint64_t *x_in, uint64_t *y, uint64_t *res);
*
* Perform a carry-less multiplication (that is, use XOR instead of the
* multiply operator) on P1 and P2 and place the result in P3.
*
* Byte swap the input and the output.
*
* Note: x_in, y, and res all point to a block of 20-byte numbers
* (an array of two 64-bit integers).
*
* Note2: For kernel code, caller is responsible for ensuring
* kpreempt_disable() has been called. This is because %xmm registers are
* not saved/restored. Clear and set the CR0.TS bit on entry and exit,
* respectively, if TS is set on entry. Otherwise, if TS is not set,
* save and restore %xmm registers on the stack.
*
* Note3: Original Intel definition:
* void galois_hash_asm(unsigned char *hk, unsigned char *s,
* unsigned char *d, int length)
*
* Note4: Register/parameter mapping:
* Intel:
* Parameter 1: %rcx (copied to %xmm0) hk or x_in
* Parameter 2: %rdx (copied to %xmm1) s or y
* Parameter 3: %rdi (result) d or res
* OpenSolaris:
* Parameter 1: %rdi (copied to %xmm0) x_in
* Parameter 2: %rsi (copied to %xmm1) y
* Parameter 3: %rdx (result) res
*/
ENTRY_NP(gcm_mul_pclmulqdq)
//
// Copy Parameters
//
movdqu (%rdi), %xmm0 // P1
movdqu (%rsi), %xmm1 // P2
//
// Byte swap 16-byte input
//
lea .Lbyte_swap16_mask(%rip), %rax
movups (%rax), %xmm10
pshufb %xmm10, %xmm0
pshufb %xmm10, %xmm1
//
// Multiply with the hash key
//
movdqu %xmm0, %xmm3
pclmulqdq $0, %xmm1, %xmm3 // xmm3 holds a0*b0
movdqu %xmm0, %xmm4
pclmulqdq $16, %xmm1, %xmm4 // xmm4 holds a0*b1
movdqu %xmm0, %xmm5
pclmulqdq $1, %xmm1, %xmm5 // xmm5 holds a1*b0
movdqu %xmm0, %xmm6
pclmulqdq $17, %xmm1, %xmm6 // xmm6 holds a1*b1
pxor %xmm5, %xmm4 // xmm4 holds a0*b1 + a1*b0
movdqu %xmm4, %xmm5 // move the contents of xmm4 to xmm5
psrldq $8, %xmm4 // shift by xmm4 64 bits to the right
pslldq $8, %xmm5 // shift by xmm5 64 bits to the left
pxor %xmm5, %xmm3
pxor %xmm4, %xmm6 // Register pair <xmm6:xmm3> holds the result
// of the carry-less multiplication of
// xmm0 by xmm1.
// We shift the result of the multiplication by one bit position
// to the left to cope for the fact that the bits are reversed.
movdqu %xmm3, %xmm7
movdqu %xmm6, %xmm8
pslld $1, %xmm3
pslld $1, %xmm6
psrld $31, %xmm7
psrld $31, %xmm8
movdqu %xmm7, %xmm9
pslldq $4, %xmm8
pslldq $4, %xmm7
psrldq $12, %xmm9
por %xmm7, %xmm3
por %xmm8, %xmm6
por %xmm9, %xmm6
//
// First phase of the reduction
//
// Move xmm3 into xmm7, xmm8, xmm9 in order to perform the shifts
// independently.
movdqu %xmm3, %xmm7
movdqu %xmm3, %xmm8
movdqu %xmm3, %xmm9
pslld $31, %xmm7 // packed right shift shifting << 31
pslld $30, %xmm8 // packed right shift shifting << 30
pslld $25, %xmm9 // packed right shift shifting << 25
pxor %xmm8, %xmm7 // xor the shifted versions
pxor %xmm9, %xmm7
movdqu %xmm7, %xmm8
pslldq $12, %xmm7
psrldq $4, %xmm8
pxor %xmm7, %xmm3 // first phase of the reduction complete
//
// Second phase of the reduction
//
// Make 3 copies of xmm3 in xmm2, xmm4, xmm5 for doing these
// shift operations.
movdqu %xmm3, %xmm2
movdqu %xmm3, %xmm4 // packed left shifting >> 1
movdqu %xmm3, %xmm5
psrld $1, %xmm2
psrld $2, %xmm4 // packed left shifting >> 2
psrld $7, %xmm5 // packed left shifting >> 7
pxor %xmm4, %xmm2 // xor the shifted versions
pxor %xmm5, %xmm2
pxor %xmm8, %xmm2
pxor %xmm2, %xmm3
pxor %xmm3, %xmm6 // the result is in xmm6
//
// Byte swap 16-byte result
//
pshufb %xmm10, %xmm6 // %xmm10 has the swap mask
//
// Store the result
//
movdqu %xmm6, (%rdx) // P3
//
// Return
//
ret
SET_SIZE(gcm_mul_pclmulqdq)
#endif /* lint || __lint */
#ifdef __ELF__
.section .note.GNU-stack,"",%progbits
#endif
diff --git a/module/icp/asm-x86_64/sha2/sha256_impl.S b/module/icp/asm-x86_64/sha2/sha256_impl.S
index ccd4a3e6b3af..951297c72ff8 100644
--- a/module/icp/asm-x86_64/sha2/sha256_impl.S
+++ b/module/icp/asm-x86_64/sha2/sha256_impl.S
@@ -1,2089 +1,2089 @@
/*
* ====================================================================
* Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
* project. Rights for redistribution and usage in source and binary
* forms are granted according to the OpenSSL license.
* ====================================================================
*
* sha256/512_block procedure for x86_64.
*
* 40% improvement over compiler-generated code on Opteron. On EM64T
* sha256 was observed to run >80% faster and sha512 - >40%. No magical
* tricks, just straight implementation... I really wonder why gcc
* [being armed with inline assembler] fails to generate as fast code.
* The only thing which is cool about this module is that it's very
* same instruction sequence used for both SHA-256 and SHA-512. In
* former case the instructions operate on 32-bit operands, while in
* latter - on 64-bit ones. All I had to do is to get one flavor right,
* the other one passed the test right away:-)
*
* sha256_block runs in ~1005 cycles on Opteron, which gives you
* asymptotic performance of 64*1000/1005=63.7MBps times CPU clock
* frequency in GHz. sha512_block runs in ~1275 cycles, which results
* in 128*1000/1275=100MBps per GHz. Is there room for improvement?
* Well, if you compare it to IA-64 implementation, which maintains
* X[16] in register bank[!], tends to 4 instructions per CPU clock
* cycle and runs in 1003 cycles, 1275 is very good result for 3-way
* issue Opteron pipeline and X[16] maintained in memory. So that *if*
* there is a way to improve it, *then* the only way would be to try to
* offload X[16] updates to SSE unit, but that would require "deeper"
* loop unroll, which in turn would naturally cause size blow-up, not
* to mention increased complexity! And once again, only *if* it's
* actually possible to noticeably improve overall ILP, instruction
* level parallelism, on a given CPU implementation in this case.
*
* Special note on Intel EM64T. While Opteron CPU exhibits perfect
* performance ratio of 1.5 between 64- and 32-bit flavors [see above],
* [currently available] EM64T CPUs apparently are far from it. On the
* contrary, 64-bit version, sha512_block, is ~30% *slower* than 32-bit
* sha256_block:-( This is presumably because 64-bit shifts/rotates
* apparently are not atomic instructions, but implemented in microcode.
*/
/*
* OpenSolaris OS modifications
*
* Sun elects to use this software under the BSD license.
*
* This source originates from OpenSSL file sha512-x86_64.pl at
* ftp://ftp.openssl.org/snapshot/openssl-0.9.8-stable-SNAP-20080131.tar.gz
* (presumably for future OpenSSL release 0.9.8h), with these changes:
*
* 1. Added perl "use strict" and declared variables.
*
* 2. Added OpenSolaris ENTRY_NP/SET_SIZE macros from
* /usr/include/sys/asm_linkage.h, .ident keywords, and lint(1B) guards.
*
* 3. Removed x86_64-xlate.pl script (not needed for as(1) or gas(1)
* assemblers). Replaced the .picmeup macro with assembler code.
*
* 4. Added 8 to $ctx, as OpenSolaris OS has an extra 4-byte field, "algotype",
* at the beginning of SHA2_CTX (the next field is 8-byte aligned).
*/
/*
* This file was generated by a perl script (sha512-x86_64.pl) that were
* used to generate sha256 and sha512 variants from the same code base.
* The comments from the original file have been pasted above.
*/
#if defined(lint) || defined(__lint)
#include <sys/stdint.h>
#include <sha2/sha2.h>
void
SHA256TransformBlocks(SHA2_CTX *ctx, const void *in, size_t num)
{
(void) ctx, (void) in, (void) num;
}
#else
#define _ASM
#include <sys/asm_linkage.h>
ENTRY_NP(SHA256TransformBlocks)
.cfi_startproc
movq %rsp, %rax
.cfi_def_cfa_register %rax
push %rbx
.cfi_offset %rbx,-16
push %rbp
.cfi_offset %rbp,-24
push %r12
.cfi_offset %r12,-32
push %r13
.cfi_offset %r13,-40
push %r14
.cfi_offset %r14,-48
push %r15
.cfi_offset %r15,-56
mov %rsp,%rbp # copy %rsp
shl $4,%rdx # num*16
sub $16*4+4*8,%rsp
lea (%rsi,%rdx,4),%rdx # inp+num*16*4
and $-64,%rsp # align stack frame
add $8,%rdi # Skip OpenSolaris field, "algotype"
mov %rdi,16*4+0*8(%rsp) # save ctx, 1st arg
mov %rsi,16*4+1*8(%rsp) # save inp, 2nd arg
mov %rdx,16*4+2*8(%rsp) # save end pointer, "3rd" arg
mov %rbp,16*4+3*8(%rsp) # save copy of %rsp
# echo ".cfi_cfa_expression %rsp+88,deref,+56" |
# openssl/crypto/perlasm/x86_64-xlate.pl
.cfi_escape 0x0f,0x06,0x77,0xd8,0x00,0x06,0x23,0x38
#.picmeup %rbp
# The .picmeup pseudo-directive, from perlasm/x86_64_xlate.pl, puts
# the address of the "next" instruction into the target register
# (%rbp). This generates these 2 instructions:
lea .Llea(%rip),%rbp
#nop # .picmeup generates a nop for mod 8 alignment--not needed here
.Llea:
lea K256-.(%rbp),%rbp
mov 4*0(%rdi),%eax
mov 4*1(%rdi),%ebx
mov 4*2(%rdi),%ecx
mov 4*3(%rdi),%edx
mov 4*4(%rdi),%r8d
mov 4*5(%rdi),%r9d
mov 4*6(%rdi),%r10d
mov 4*7(%rdi),%r11d
jmp .Lloop
.align 16
.Lloop:
xor %rdi,%rdi
mov 4*0(%rsi),%r12d
bswap %r12d
mov %r8d,%r13d
mov %r8d,%r14d
mov %r9d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r10d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r8d,%r15d # (f^g)&e
mov %r12d,0(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r10d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r11d,%r12d # T1+=h
mov %eax,%r11d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %eax,%r13d
mov %eax,%r14d
ror $2,%r11d
ror $13,%r13d
mov %eax,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r11d
ror $9,%r13d
or %ecx,%r14d # a|c
xor %r13d,%r11d # h=Sigma0(a)
and %ecx,%r15d # a&c
add %r12d,%edx # d+=T1
and %ebx,%r14d # (a|c)&b
add %r12d,%r11d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r11d # h+=Maj(a,b,c)
mov 4*1(%rsi),%r12d
bswap %r12d
mov %edx,%r13d
mov %edx,%r14d
mov %r8d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r9d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %edx,%r15d # (f^g)&e
mov %r12d,4(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r9d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r10d,%r12d # T1+=h
mov %r11d,%r10d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r11d,%r13d
mov %r11d,%r14d
ror $2,%r10d
ror $13,%r13d
mov %r11d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r10d
ror $9,%r13d
or %ebx,%r14d # a|c
xor %r13d,%r10d # h=Sigma0(a)
and %ebx,%r15d # a&c
add %r12d,%ecx # d+=T1
and %eax,%r14d # (a|c)&b
add %r12d,%r10d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r10d # h+=Maj(a,b,c)
mov 4*2(%rsi),%r12d
bswap %r12d
mov %ecx,%r13d
mov %ecx,%r14d
mov %edx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r8d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %ecx,%r15d # (f^g)&e
mov %r12d,8(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r8d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r9d,%r12d # T1+=h
mov %r10d,%r9d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r10d,%r13d
mov %r10d,%r14d
ror $2,%r9d
ror $13,%r13d
mov %r10d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r9d
ror $9,%r13d
or %eax,%r14d # a|c
xor %r13d,%r9d # h=Sigma0(a)
and %eax,%r15d # a&c
add %r12d,%ebx # d+=T1
and %r11d,%r14d # (a|c)&b
add %r12d,%r9d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r9d # h+=Maj(a,b,c)
mov 4*3(%rsi),%r12d
bswap %r12d
mov %ebx,%r13d
mov %ebx,%r14d
mov %ecx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %edx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %ebx,%r15d # (f^g)&e
mov %r12d,12(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %edx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r8d,%r12d # T1+=h
mov %r9d,%r8d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r9d,%r13d
mov %r9d,%r14d
ror $2,%r8d
ror $13,%r13d
mov %r9d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r8d
ror $9,%r13d
or %r11d,%r14d # a|c
xor %r13d,%r8d # h=Sigma0(a)
and %r11d,%r15d # a&c
add %r12d,%eax # d+=T1
and %r10d,%r14d # (a|c)&b
add %r12d,%r8d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r8d # h+=Maj(a,b,c)
mov 4*4(%rsi),%r12d
bswap %r12d
mov %eax,%r13d
mov %eax,%r14d
mov %ebx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %ecx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %eax,%r15d # (f^g)&e
mov %r12d,16(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %ecx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %edx,%r12d # T1+=h
mov %r8d,%edx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r8d,%r13d
mov %r8d,%r14d
ror $2,%edx
ror $13,%r13d
mov %r8d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%edx
ror $9,%r13d
or %r10d,%r14d # a|c
xor %r13d,%edx # h=Sigma0(a)
and %r10d,%r15d # a&c
add %r12d,%r11d # d+=T1
and %r9d,%r14d # (a|c)&b
add %r12d,%edx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%edx # h+=Maj(a,b,c)
mov 4*5(%rsi),%r12d
bswap %r12d
mov %r11d,%r13d
mov %r11d,%r14d
mov %eax,%r15d
ror $6,%r13d
ror $11,%r14d
xor %ebx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r11d,%r15d # (f^g)&e
mov %r12d,20(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %ebx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %ecx,%r12d # T1+=h
mov %edx,%ecx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %edx,%r13d
mov %edx,%r14d
ror $2,%ecx
ror $13,%r13d
mov %edx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%ecx
ror $9,%r13d
or %r9d,%r14d # a|c
xor %r13d,%ecx # h=Sigma0(a)
and %r9d,%r15d # a&c
add %r12d,%r10d # d+=T1
and %r8d,%r14d # (a|c)&b
add %r12d,%ecx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%ecx # h+=Maj(a,b,c)
mov 4*6(%rsi),%r12d
bswap %r12d
mov %r10d,%r13d
mov %r10d,%r14d
mov %r11d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %eax,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r10d,%r15d # (f^g)&e
mov %r12d,24(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %eax,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %ebx,%r12d # T1+=h
mov %ecx,%ebx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %ecx,%r13d
mov %ecx,%r14d
ror $2,%ebx
ror $13,%r13d
mov %ecx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%ebx
ror $9,%r13d
or %r8d,%r14d # a|c
xor %r13d,%ebx # h=Sigma0(a)
and %r8d,%r15d # a&c
add %r12d,%r9d # d+=T1
and %edx,%r14d # (a|c)&b
add %r12d,%ebx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%ebx # h+=Maj(a,b,c)
mov 4*7(%rsi),%r12d
bswap %r12d
mov %r9d,%r13d
mov %r9d,%r14d
mov %r10d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r11d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r9d,%r15d # (f^g)&e
mov %r12d,28(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r11d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %eax,%r12d # T1+=h
mov %ebx,%eax
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %ebx,%r13d
mov %ebx,%r14d
ror $2,%eax
ror $13,%r13d
mov %ebx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%eax
ror $9,%r13d
or %edx,%r14d # a|c
xor %r13d,%eax # h=Sigma0(a)
and %edx,%r15d # a&c
add %r12d,%r8d # d+=T1
and %ecx,%r14d # (a|c)&b
add %r12d,%eax # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%eax # h+=Maj(a,b,c)
mov 4*8(%rsi),%r12d
bswap %r12d
mov %r8d,%r13d
mov %r8d,%r14d
mov %r9d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r10d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r8d,%r15d # (f^g)&e
mov %r12d,32(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r10d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r11d,%r12d # T1+=h
mov %eax,%r11d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %eax,%r13d
mov %eax,%r14d
ror $2,%r11d
ror $13,%r13d
mov %eax,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r11d
ror $9,%r13d
or %ecx,%r14d # a|c
xor %r13d,%r11d # h=Sigma0(a)
and %ecx,%r15d # a&c
add %r12d,%edx # d+=T1
and %ebx,%r14d # (a|c)&b
add %r12d,%r11d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r11d # h+=Maj(a,b,c)
mov 4*9(%rsi),%r12d
bswap %r12d
mov %edx,%r13d
mov %edx,%r14d
mov %r8d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r9d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %edx,%r15d # (f^g)&e
mov %r12d,36(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r9d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r10d,%r12d # T1+=h
mov %r11d,%r10d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r11d,%r13d
mov %r11d,%r14d
ror $2,%r10d
ror $13,%r13d
mov %r11d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r10d
ror $9,%r13d
or %ebx,%r14d # a|c
xor %r13d,%r10d # h=Sigma0(a)
and %ebx,%r15d # a&c
add %r12d,%ecx # d+=T1
and %eax,%r14d # (a|c)&b
add %r12d,%r10d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r10d # h+=Maj(a,b,c)
mov 4*10(%rsi),%r12d
bswap %r12d
mov %ecx,%r13d
mov %ecx,%r14d
mov %edx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r8d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %ecx,%r15d # (f^g)&e
mov %r12d,40(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r8d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r9d,%r12d # T1+=h
mov %r10d,%r9d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r10d,%r13d
mov %r10d,%r14d
ror $2,%r9d
ror $13,%r13d
mov %r10d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r9d
ror $9,%r13d
or %eax,%r14d # a|c
xor %r13d,%r9d # h=Sigma0(a)
and %eax,%r15d # a&c
add %r12d,%ebx # d+=T1
and %r11d,%r14d # (a|c)&b
add %r12d,%r9d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r9d # h+=Maj(a,b,c)
mov 4*11(%rsi),%r12d
bswap %r12d
mov %ebx,%r13d
mov %ebx,%r14d
mov %ecx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %edx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %ebx,%r15d # (f^g)&e
mov %r12d,44(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %edx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r8d,%r12d # T1+=h
mov %r9d,%r8d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r9d,%r13d
mov %r9d,%r14d
ror $2,%r8d
ror $13,%r13d
mov %r9d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r8d
ror $9,%r13d
or %r11d,%r14d # a|c
xor %r13d,%r8d # h=Sigma0(a)
and %r11d,%r15d # a&c
add %r12d,%eax # d+=T1
and %r10d,%r14d # (a|c)&b
add %r12d,%r8d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r8d # h+=Maj(a,b,c)
mov 4*12(%rsi),%r12d
bswap %r12d
mov %eax,%r13d
mov %eax,%r14d
mov %ebx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %ecx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %eax,%r15d # (f^g)&e
mov %r12d,48(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %ecx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %edx,%r12d # T1+=h
mov %r8d,%edx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r8d,%r13d
mov %r8d,%r14d
ror $2,%edx
ror $13,%r13d
mov %r8d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%edx
ror $9,%r13d
or %r10d,%r14d # a|c
xor %r13d,%edx # h=Sigma0(a)
and %r10d,%r15d # a&c
add %r12d,%r11d # d+=T1
and %r9d,%r14d # (a|c)&b
add %r12d,%edx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%edx # h+=Maj(a,b,c)
mov 4*13(%rsi),%r12d
bswap %r12d
mov %r11d,%r13d
mov %r11d,%r14d
mov %eax,%r15d
ror $6,%r13d
ror $11,%r14d
xor %ebx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r11d,%r15d # (f^g)&e
mov %r12d,52(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %ebx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %ecx,%r12d # T1+=h
mov %edx,%ecx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %edx,%r13d
mov %edx,%r14d
ror $2,%ecx
ror $13,%r13d
mov %edx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%ecx
ror $9,%r13d
or %r9d,%r14d # a|c
xor %r13d,%ecx # h=Sigma0(a)
and %r9d,%r15d # a&c
add %r12d,%r10d # d+=T1
and %r8d,%r14d # (a|c)&b
add %r12d,%ecx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%ecx # h+=Maj(a,b,c)
mov 4*14(%rsi),%r12d
bswap %r12d
mov %r10d,%r13d
mov %r10d,%r14d
mov %r11d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %eax,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r10d,%r15d # (f^g)&e
mov %r12d,56(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %eax,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %ebx,%r12d # T1+=h
mov %ecx,%ebx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %ecx,%r13d
mov %ecx,%r14d
ror $2,%ebx
ror $13,%r13d
mov %ecx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%ebx
ror $9,%r13d
or %r8d,%r14d # a|c
xor %r13d,%ebx # h=Sigma0(a)
and %r8d,%r15d # a&c
add %r12d,%r9d # d+=T1
and %edx,%r14d # (a|c)&b
add %r12d,%ebx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%ebx # h+=Maj(a,b,c)
mov 4*15(%rsi),%r12d
bswap %r12d
mov %r9d,%r13d
mov %r9d,%r14d
mov %r10d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r11d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r9d,%r15d # (f^g)&e
mov %r12d,60(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r11d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %eax,%r12d # T1+=h
mov %ebx,%eax
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %ebx,%r13d
mov %ebx,%r14d
ror $2,%eax
ror $13,%r13d
mov %ebx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%eax
ror $9,%r13d
or %edx,%r14d # a|c
xor %r13d,%eax # h=Sigma0(a)
and %edx,%r15d # a&c
add %r12d,%r8d # d+=T1
and %ecx,%r14d # (a|c)&b
add %r12d,%eax # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%eax # h+=Maj(a,b,c)
jmp .Lrounds_16_xx
.align 16
.Lrounds_16_xx:
mov 4(%rsp),%r13d
mov 56(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 36(%rsp),%r12d
add 0(%rsp),%r12d
mov %r8d,%r13d
mov %r8d,%r14d
mov %r9d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r10d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r8d,%r15d # (f^g)&e
mov %r12d,0(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r10d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r11d,%r12d # T1+=h
mov %eax,%r11d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %eax,%r13d
mov %eax,%r14d
ror $2,%r11d
ror $13,%r13d
mov %eax,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r11d
ror $9,%r13d
or %ecx,%r14d # a|c
xor %r13d,%r11d # h=Sigma0(a)
and %ecx,%r15d # a&c
add %r12d,%edx # d+=T1
and %ebx,%r14d # (a|c)&b
add %r12d,%r11d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r11d # h+=Maj(a,b,c)
mov 8(%rsp),%r13d
mov 60(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 40(%rsp),%r12d
add 4(%rsp),%r12d
mov %edx,%r13d
mov %edx,%r14d
mov %r8d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r9d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %edx,%r15d # (f^g)&e
mov %r12d,4(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r9d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r10d,%r12d # T1+=h
mov %r11d,%r10d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r11d,%r13d
mov %r11d,%r14d
ror $2,%r10d
ror $13,%r13d
mov %r11d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r10d
ror $9,%r13d
or %ebx,%r14d # a|c
xor %r13d,%r10d # h=Sigma0(a)
and %ebx,%r15d # a&c
add %r12d,%ecx # d+=T1
and %eax,%r14d # (a|c)&b
add %r12d,%r10d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r10d # h+=Maj(a,b,c)
mov 12(%rsp),%r13d
mov 0(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 44(%rsp),%r12d
add 8(%rsp),%r12d
mov %ecx,%r13d
mov %ecx,%r14d
mov %edx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r8d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %ecx,%r15d # (f^g)&e
mov %r12d,8(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r8d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r9d,%r12d # T1+=h
mov %r10d,%r9d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r10d,%r13d
mov %r10d,%r14d
ror $2,%r9d
ror $13,%r13d
mov %r10d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r9d
ror $9,%r13d
or %eax,%r14d # a|c
xor %r13d,%r9d # h=Sigma0(a)
and %eax,%r15d # a&c
add %r12d,%ebx # d+=T1
and %r11d,%r14d # (a|c)&b
add %r12d,%r9d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r9d # h+=Maj(a,b,c)
mov 16(%rsp),%r13d
mov 4(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 48(%rsp),%r12d
add 12(%rsp),%r12d
mov %ebx,%r13d
mov %ebx,%r14d
mov %ecx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %edx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %ebx,%r15d # (f^g)&e
mov %r12d,12(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %edx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r8d,%r12d # T1+=h
mov %r9d,%r8d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r9d,%r13d
mov %r9d,%r14d
ror $2,%r8d
ror $13,%r13d
mov %r9d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r8d
ror $9,%r13d
or %r11d,%r14d # a|c
xor %r13d,%r8d # h=Sigma0(a)
and %r11d,%r15d # a&c
add %r12d,%eax # d+=T1
and %r10d,%r14d # (a|c)&b
add %r12d,%r8d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r8d # h+=Maj(a,b,c)
mov 20(%rsp),%r13d
mov 8(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 52(%rsp),%r12d
add 16(%rsp),%r12d
mov %eax,%r13d
mov %eax,%r14d
mov %ebx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %ecx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %eax,%r15d # (f^g)&e
mov %r12d,16(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %ecx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %edx,%r12d # T1+=h
mov %r8d,%edx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r8d,%r13d
mov %r8d,%r14d
ror $2,%edx
ror $13,%r13d
mov %r8d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%edx
ror $9,%r13d
or %r10d,%r14d # a|c
xor %r13d,%edx # h=Sigma0(a)
and %r10d,%r15d # a&c
add %r12d,%r11d # d+=T1
and %r9d,%r14d # (a|c)&b
add %r12d,%edx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%edx # h+=Maj(a,b,c)
mov 24(%rsp),%r13d
mov 12(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 56(%rsp),%r12d
add 20(%rsp),%r12d
mov %r11d,%r13d
mov %r11d,%r14d
mov %eax,%r15d
ror $6,%r13d
ror $11,%r14d
xor %ebx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r11d,%r15d # (f^g)&e
mov %r12d,20(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %ebx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %ecx,%r12d # T1+=h
mov %edx,%ecx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %edx,%r13d
mov %edx,%r14d
ror $2,%ecx
ror $13,%r13d
mov %edx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%ecx
ror $9,%r13d
or %r9d,%r14d # a|c
xor %r13d,%ecx # h=Sigma0(a)
and %r9d,%r15d # a&c
add %r12d,%r10d # d+=T1
and %r8d,%r14d # (a|c)&b
add %r12d,%ecx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%ecx # h+=Maj(a,b,c)
mov 28(%rsp),%r13d
mov 16(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 60(%rsp),%r12d
add 24(%rsp),%r12d
mov %r10d,%r13d
mov %r10d,%r14d
mov %r11d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %eax,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r10d,%r15d # (f^g)&e
mov %r12d,24(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %eax,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %ebx,%r12d # T1+=h
mov %ecx,%ebx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %ecx,%r13d
mov %ecx,%r14d
ror $2,%ebx
ror $13,%r13d
mov %ecx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%ebx
ror $9,%r13d
or %r8d,%r14d # a|c
xor %r13d,%ebx # h=Sigma0(a)
and %r8d,%r15d # a&c
add %r12d,%r9d # d+=T1
and %edx,%r14d # (a|c)&b
add %r12d,%ebx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%ebx # h+=Maj(a,b,c)
mov 32(%rsp),%r13d
mov 20(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 0(%rsp),%r12d
add 28(%rsp),%r12d
mov %r9d,%r13d
mov %r9d,%r14d
mov %r10d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r11d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r9d,%r15d # (f^g)&e
mov %r12d,28(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r11d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %eax,%r12d # T1+=h
mov %ebx,%eax
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %ebx,%r13d
mov %ebx,%r14d
ror $2,%eax
ror $13,%r13d
mov %ebx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%eax
ror $9,%r13d
or %edx,%r14d # a|c
xor %r13d,%eax # h=Sigma0(a)
and %edx,%r15d # a&c
add %r12d,%r8d # d+=T1
and %ecx,%r14d # (a|c)&b
add %r12d,%eax # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%eax # h+=Maj(a,b,c)
mov 36(%rsp),%r13d
mov 24(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 4(%rsp),%r12d
add 32(%rsp),%r12d
mov %r8d,%r13d
mov %r8d,%r14d
mov %r9d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r10d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r8d,%r15d # (f^g)&e
mov %r12d,32(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r10d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r11d,%r12d # T1+=h
mov %eax,%r11d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %eax,%r13d
mov %eax,%r14d
ror $2,%r11d
ror $13,%r13d
mov %eax,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r11d
ror $9,%r13d
or %ecx,%r14d # a|c
xor %r13d,%r11d # h=Sigma0(a)
and %ecx,%r15d # a&c
add %r12d,%edx # d+=T1
and %ebx,%r14d # (a|c)&b
add %r12d,%r11d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r11d # h+=Maj(a,b,c)
mov 40(%rsp),%r13d
mov 28(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 8(%rsp),%r12d
add 36(%rsp),%r12d
mov %edx,%r13d
mov %edx,%r14d
mov %r8d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r9d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %edx,%r15d # (f^g)&e
mov %r12d,36(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r9d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r10d,%r12d # T1+=h
mov %r11d,%r10d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r11d,%r13d
mov %r11d,%r14d
ror $2,%r10d
ror $13,%r13d
mov %r11d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r10d
ror $9,%r13d
or %ebx,%r14d # a|c
xor %r13d,%r10d # h=Sigma0(a)
and %ebx,%r15d # a&c
add %r12d,%ecx # d+=T1
and %eax,%r14d # (a|c)&b
add %r12d,%r10d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r10d # h+=Maj(a,b,c)
mov 44(%rsp),%r13d
mov 32(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 12(%rsp),%r12d
add 40(%rsp),%r12d
mov %ecx,%r13d
mov %ecx,%r14d
mov %edx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r8d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %ecx,%r15d # (f^g)&e
mov %r12d,40(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r8d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r9d,%r12d # T1+=h
mov %r10d,%r9d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r10d,%r13d
mov %r10d,%r14d
ror $2,%r9d
ror $13,%r13d
mov %r10d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r9d
ror $9,%r13d
or %eax,%r14d # a|c
xor %r13d,%r9d # h=Sigma0(a)
and %eax,%r15d # a&c
add %r12d,%ebx # d+=T1
and %r11d,%r14d # (a|c)&b
add %r12d,%r9d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r9d # h+=Maj(a,b,c)
mov 48(%rsp),%r13d
mov 36(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 16(%rsp),%r12d
add 44(%rsp),%r12d
mov %ebx,%r13d
mov %ebx,%r14d
mov %ecx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %edx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %ebx,%r15d # (f^g)&e
mov %r12d,44(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %edx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %r8d,%r12d # T1+=h
mov %r9d,%r8d
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r9d,%r13d
mov %r9d,%r14d
ror $2,%r8d
ror $13,%r13d
mov %r9d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%r8d
ror $9,%r13d
or %r11d,%r14d # a|c
xor %r13d,%r8d # h=Sigma0(a)
and %r11d,%r15d # a&c
add %r12d,%eax # d+=T1
and %r10d,%r14d # (a|c)&b
add %r12d,%r8d # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%r8d # h+=Maj(a,b,c)
mov 52(%rsp),%r13d
mov 40(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 20(%rsp),%r12d
add 48(%rsp),%r12d
mov %eax,%r13d
mov %eax,%r14d
mov %ebx,%r15d
ror $6,%r13d
ror $11,%r14d
xor %ecx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %eax,%r15d # (f^g)&e
mov %r12d,48(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %ecx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %edx,%r12d # T1+=h
mov %r8d,%edx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %r8d,%r13d
mov %r8d,%r14d
ror $2,%edx
ror $13,%r13d
mov %r8d,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%edx
ror $9,%r13d
or %r10d,%r14d # a|c
xor %r13d,%edx # h=Sigma0(a)
and %r10d,%r15d # a&c
add %r12d,%r11d # d+=T1
and %r9d,%r14d # (a|c)&b
add %r12d,%edx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%edx # h+=Maj(a,b,c)
mov 56(%rsp),%r13d
mov 44(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 24(%rsp),%r12d
add 52(%rsp),%r12d
mov %r11d,%r13d
mov %r11d,%r14d
mov %eax,%r15d
ror $6,%r13d
ror $11,%r14d
xor %ebx,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r11d,%r15d # (f^g)&e
mov %r12d,52(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %ebx,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %ecx,%r12d # T1+=h
mov %edx,%ecx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %edx,%r13d
mov %edx,%r14d
ror $2,%ecx
ror $13,%r13d
mov %edx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%ecx
ror $9,%r13d
or %r9d,%r14d # a|c
xor %r13d,%ecx # h=Sigma0(a)
and %r9d,%r15d # a&c
add %r12d,%r10d # d+=T1
and %r8d,%r14d # (a|c)&b
add %r12d,%ecx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%ecx # h+=Maj(a,b,c)
mov 60(%rsp),%r13d
mov 48(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 28(%rsp),%r12d
add 56(%rsp),%r12d
mov %r10d,%r13d
mov %r10d,%r14d
mov %r11d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %eax,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r10d,%r15d # (f^g)&e
mov %r12d,56(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %eax,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %ebx,%r12d # T1+=h
mov %ecx,%ebx
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %ecx,%r13d
mov %ecx,%r14d
ror $2,%ebx
ror $13,%r13d
mov %ecx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%ebx
ror $9,%r13d
or %r8d,%r14d # a|c
xor %r13d,%ebx # h=Sigma0(a)
and %r8d,%r15d # a&c
add %r12d,%r9d # d+=T1
and %edx,%r14d # (a|c)&b
add %r12d,%ebx # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%ebx # h+=Maj(a,b,c)
mov 0(%rsp),%r13d
mov 52(%rsp),%r12d
mov %r13d,%r15d
shr $3,%r13d
ror $7,%r15d
xor %r15d,%r13d
ror $11,%r15d
xor %r15d,%r13d # sigma0(X[(i+1)&0xf])
mov %r12d,%r14d
shr $10,%r12d
ror $17,%r14d
xor %r14d,%r12d
ror $2,%r14d
xor %r14d,%r12d # sigma1(X[(i+14)&0xf])
add %r13d,%r12d
add 32(%rsp),%r12d
add 60(%rsp),%r12d
mov %r9d,%r13d
mov %r9d,%r14d
mov %r10d,%r15d
ror $6,%r13d
ror $11,%r14d
xor %r11d,%r15d # f^g
xor %r14d,%r13d
ror $14,%r14d
and %r9d,%r15d # (f^g)&e
mov %r12d,60(%rsp)
xor %r14d,%r13d # Sigma1(e)
xor %r11d,%r15d # Ch(e,f,g)=((f^g)&e)^g
add %eax,%r12d # T1+=h
mov %ebx,%eax
add %r13d,%r12d # T1+=Sigma1(e)
add %r15d,%r12d # T1+=Ch(e,f,g)
mov %ebx,%r13d
mov %ebx,%r14d
ror $2,%eax
ror $13,%r13d
mov %ebx,%r15d
add (%rbp,%rdi,4),%r12d # T1+=K[round]
xor %r13d,%eax
ror $9,%r13d
or %edx,%r14d # a|c
xor %r13d,%eax # h=Sigma0(a)
and %edx,%r15d # a&c
add %r12d,%r8d # d+=T1
and %ecx,%r14d # (a|c)&b
add %r12d,%eax # h+=T1
or %r15d,%r14d # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14d,%eax # h+=Maj(a,b,c)
cmp $64,%rdi
jb .Lrounds_16_xx
mov 16*4+0*8(%rsp),%rdi
lea 16*4(%rsi),%rsi
add 4*0(%rdi),%eax
add 4*1(%rdi),%ebx
add 4*2(%rdi),%ecx
add 4*3(%rdi),%edx
add 4*4(%rdi),%r8d
add 4*5(%rdi),%r9d
add 4*6(%rdi),%r10d
add 4*7(%rdi),%r11d
cmp 16*4+2*8(%rsp),%rsi
mov %eax,4*0(%rdi)
mov %ebx,4*1(%rdi)
mov %ecx,4*2(%rdi)
mov %edx,4*3(%rdi)
mov %r8d,4*4(%rdi)
mov %r9d,4*5(%rdi)
mov %r10d,4*6(%rdi)
mov %r11d,4*7(%rdi)
jb .Lloop
mov 16*4+3*8(%rsp),%rsp
.cfi_def_cfa %rsp,56
pop %r15
.cfi_adjust_cfa_offset -8
.cfi_restore %r15
pop %r14
.cfi_adjust_cfa_offset -8
.cfi_restore %r14
pop %r13
.cfi_adjust_cfa_offset -8
.cfi_restore %r13
pop %r12
.cfi_adjust_cfa_offset -8
.cfi_restore %r12
pop %rbp
.cfi_adjust_cfa_offset -8
.cfi_restore %rbp
pop %rbx
.cfi_adjust_cfa_offset -8
.cfi_restore %rbx
ret
.cfi_endproc
SET_SIZE(SHA256TransformBlocks)
-.data
+.section .rodata
.align 64
.type K256,@object
K256:
.long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
.long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
.long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3
.long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174
.long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc
.long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da
.long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7
.long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967
.long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13
.long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85
.long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3
.long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070
.long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5
.long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3
.long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
.long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
#endif /* !lint && !__lint */
#ifdef __ELF__
.section .note.GNU-stack,"",%progbits
#endif
diff --git a/module/icp/asm-x86_64/sha2/sha512_impl.S b/module/icp/asm-x86_64/sha2/sha512_impl.S
index c6e7efd86038..921d3d8cddae 100644
--- a/module/icp/asm-x86_64/sha2/sha512_impl.S
+++ b/module/icp/asm-x86_64/sha2/sha512_impl.S
@@ -1,2114 +1,2114 @@
/*
* ====================================================================
* Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
* project. Rights for redistribution and usage in source and binary
* forms are granted according to the OpenSSL license.
* ====================================================================
*
* sha256/512_block procedure for x86_64.
*
* 40% improvement over compiler-generated code on Opteron. On EM64T
* sha256 was observed to run >80% faster and sha512 - >40%. No magical
* tricks, just straight implementation... I really wonder why gcc
* [being armed with inline assembler] fails to generate as fast code.
* The only thing which is cool about this module is that it's very
* same instruction sequence used for both SHA-256 and SHA-512. In
* former case the instructions operate on 32-bit operands, while in
* latter - on 64-bit ones. All I had to do is to get one flavor right,
* the other one passed the test right away:-)
*
* sha256_block runs in ~1005 cycles on Opteron, which gives you
* asymptotic performance of 64*1000/1005=63.7MBps times CPU clock
* frequency in GHz. sha512_block runs in ~1275 cycles, which results
* in 128*1000/1275=100MBps per GHz. Is there room for improvement?
* Well, if you compare it to IA-64 implementation, which maintains
* X[16] in register bank[!], tends to 4 instructions per CPU clock
* cycle and runs in 1003 cycles, 1275 is very good result for 3-way
* issue Opteron pipeline and X[16] maintained in memory. So that *if*
* there is a way to improve it, *then* the only way would be to try to
* offload X[16] updates to SSE unit, but that would require "deeper"
* loop unroll, which in turn would naturally cause size blow-up, not
* to mention increased complexity! And once again, only *if* it's
* actually possible to noticeably improve overall ILP, instruction
* level parallelism, on a given CPU implementation in this case.
*
* Special note on Intel EM64T. While Opteron CPU exhibits perfect
* performance ratio of 1.5 between 64- and 32-bit flavors [see above],
* [currently available] EM64T CPUs apparently are far from it. On the
* contrary, 64-bit version, sha512_block, is ~30% *slower* than 32-bit
* sha256_block:-( This is presumably because 64-bit shifts/rotates
* apparently are not atomic instructions, but implemented in microcode.
*/
/*
* OpenSolaris OS modifications
*
* Sun elects to use this software under the BSD license.
*
* This source originates from OpenSSL file sha512-x86_64.pl at
* ftp://ftp.openssl.org/snapshot/openssl-0.9.8-stable-SNAP-20080131.tar.gz
* (presumably for future OpenSSL release 0.9.8h), with these changes:
*
* 1. Added perl "use strict" and declared variables.
*
* 2. Added OpenSolaris ENTRY_NP/SET_SIZE macros from
* /usr/include/sys/asm_linkage.h, .ident keywords, and lint(1B) guards.
*
* 3. Removed x86_64-xlate.pl script (not needed for as(1) or gas(1)
* assemblers). Replaced the .picmeup macro with assembler code.
*
* 4. Added 8 to $ctx, as OpenSolaris OS has an extra 4-byte field, "algotype",
* at the beginning of SHA2_CTX (the next field is 8-byte aligned).
*/
/*
* This file was generated by a perl script (sha512-x86_64.pl) that were
* used to generate sha256 and sha512 variants from the same code base.
* The comments from the original file have been pasted above.
*/
#if defined(lint) || defined(__lint)
#include <sys/stdint.h>
#include <sha2/sha2.h>
void
SHA512TransformBlocks(SHA2_CTX *ctx, const void *in, size_t num)
{
(void) ctx, (void) in, (void) num;
}
#else
#define _ASM
#include <sys/asm_linkage.h>
ENTRY_NP(SHA512TransformBlocks)
.cfi_startproc
movq %rsp, %rax
.cfi_def_cfa_register %rax
push %rbx
.cfi_offset %rbx,-16
push %rbp
.cfi_offset %rbp,-24
push %r12
.cfi_offset %r12,-32
push %r13
.cfi_offset %r13,-40
push %r14
.cfi_offset %r14,-48
push %r15
.cfi_offset %r15,-56
mov %rsp,%rbp # copy %rsp
shl $4,%rdx # num*16
sub $16*8+4*8,%rsp
lea (%rsi,%rdx,8),%rdx # inp+num*16*8
and $-64,%rsp # align stack frame
add $8,%rdi # Skip OpenSolaris field, "algotype"
mov %rdi,16*8+0*8(%rsp) # save ctx, 1st arg
mov %rsi,16*8+1*8(%rsp) # save inp, 2nd arg
mov %rdx,16*8+2*8(%rsp) # save end pointer, "3rd" arg
mov %rbp,16*8+3*8(%rsp) # save copy of %rsp
# echo ".cfi_cfa_expression %rsp+152,deref,+56" |
# openssl/crypto/perlasm/x86_64-xlate.pl
.cfi_escape 0x0f,0x06,0x77,0x98,0x01,0x06,0x23,0x38
#.picmeup %rbp
# The .picmeup pseudo-directive, from perlasm/x86_64_xlate.pl, puts
# the address of the "next" instruction into the target register
# (%rbp). This generates these 2 instructions:
lea .Llea(%rip),%rbp
#nop # .picmeup generates a nop for mod 8 alignment--not needed here
.Llea:
lea K512-.(%rbp),%rbp
mov 8*0(%rdi),%rax
mov 8*1(%rdi),%rbx
mov 8*2(%rdi),%rcx
mov 8*3(%rdi),%rdx
mov 8*4(%rdi),%r8
mov 8*5(%rdi),%r9
mov 8*6(%rdi),%r10
mov 8*7(%rdi),%r11
jmp .Lloop
.align 16
.Lloop:
xor %rdi,%rdi
mov 8*0(%rsi),%r12
bswap %r12
mov %r8,%r13
mov %r8,%r14
mov %r9,%r15
ror $14,%r13
ror $18,%r14
xor %r10,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r8,%r15 # (f^g)&e
mov %r12,0(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r10,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r11,%r12 # T1+=h
mov %rax,%r11
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rax,%r13
mov %rax,%r14
ror $28,%r11
ror $34,%r13
mov %rax,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r11
ror $5,%r13
or %rcx,%r14 # a|c
xor %r13,%r11 # h=Sigma0(a)
and %rcx,%r15 # a&c
add %r12,%rdx # d+=T1
and %rbx,%r14 # (a|c)&b
add %r12,%r11 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r11 # h+=Maj(a,b,c)
mov 8*1(%rsi),%r12
bswap %r12
mov %rdx,%r13
mov %rdx,%r14
mov %r8,%r15
ror $14,%r13
ror $18,%r14
xor %r9,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rdx,%r15 # (f^g)&e
mov %r12,8(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r9,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r10,%r12 # T1+=h
mov %r11,%r10
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r11,%r13
mov %r11,%r14
ror $28,%r10
ror $34,%r13
mov %r11,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r10
ror $5,%r13
or %rbx,%r14 # a|c
xor %r13,%r10 # h=Sigma0(a)
and %rbx,%r15 # a&c
add %r12,%rcx # d+=T1
and %rax,%r14 # (a|c)&b
add %r12,%r10 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r10 # h+=Maj(a,b,c)
mov 8*2(%rsi),%r12
bswap %r12
mov %rcx,%r13
mov %rcx,%r14
mov %rdx,%r15
ror $14,%r13
ror $18,%r14
xor %r8,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rcx,%r15 # (f^g)&e
mov %r12,16(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r8,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r9,%r12 # T1+=h
mov %r10,%r9
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r10,%r13
mov %r10,%r14
ror $28,%r9
ror $34,%r13
mov %r10,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r9
ror $5,%r13
or %rax,%r14 # a|c
xor %r13,%r9 # h=Sigma0(a)
and %rax,%r15 # a&c
add %r12,%rbx # d+=T1
and %r11,%r14 # (a|c)&b
add %r12,%r9 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r9 # h+=Maj(a,b,c)
mov 8*3(%rsi),%r12
bswap %r12
mov %rbx,%r13
mov %rbx,%r14
mov %rcx,%r15
ror $14,%r13
ror $18,%r14
xor %rdx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rbx,%r15 # (f^g)&e
mov %r12,24(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rdx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r8,%r12 # T1+=h
mov %r9,%r8
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r9,%r13
mov %r9,%r14
ror $28,%r8
ror $34,%r13
mov %r9,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r8
ror $5,%r13
or %r11,%r14 # a|c
xor %r13,%r8 # h=Sigma0(a)
and %r11,%r15 # a&c
add %r12,%rax # d+=T1
and %r10,%r14 # (a|c)&b
add %r12,%r8 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r8 # h+=Maj(a,b,c)
mov 8*4(%rsi),%r12
bswap %r12
mov %rax,%r13
mov %rax,%r14
mov %rbx,%r15
ror $14,%r13
ror $18,%r14
xor %rcx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rax,%r15 # (f^g)&e
mov %r12,32(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rcx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rdx,%r12 # T1+=h
mov %r8,%rdx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r8,%r13
mov %r8,%r14
ror $28,%rdx
ror $34,%r13
mov %r8,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rdx
ror $5,%r13
or %r10,%r14 # a|c
xor %r13,%rdx # h=Sigma0(a)
and %r10,%r15 # a&c
add %r12,%r11 # d+=T1
and %r9,%r14 # (a|c)&b
add %r12,%rdx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rdx # h+=Maj(a,b,c)
mov 8*5(%rsi),%r12
bswap %r12
mov %r11,%r13
mov %r11,%r14
mov %rax,%r15
ror $14,%r13
ror $18,%r14
xor %rbx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r11,%r15 # (f^g)&e
mov %r12,40(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rbx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rcx,%r12 # T1+=h
mov %rdx,%rcx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rdx,%r13
mov %rdx,%r14
ror $28,%rcx
ror $34,%r13
mov %rdx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rcx
ror $5,%r13
or %r9,%r14 # a|c
xor %r13,%rcx # h=Sigma0(a)
and %r9,%r15 # a&c
add %r12,%r10 # d+=T1
and %r8,%r14 # (a|c)&b
add %r12,%rcx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rcx # h+=Maj(a,b,c)
mov 8*6(%rsi),%r12
bswap %r12
mov %r10,%r13
mov %r10,%r14
mov %r11,%r15
ror $14,%r13
ror $18,%r14
xor %rax,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r10,%r15 # (f^g)&e
mov %r12,48(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rax,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rbx,%r12 # T1+=h
mov %rcx,%rbx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rcx,%r13
mov %rcx,%r14
ror $28,%rbx
ror $34,%r13
mov %rcx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rbx
ror $5,%r13
or %r8,%r14 # a|c
xor %r13,%rbx # h=Sigma0(a)
and %r8,%r15 # a&c
add %r12,%r9 # d+=T1
and %rdx,%r14 # (a|c)&b
add %r12,%rbx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rbx # h+=Maj(a,b,c)
mov 8*7(%rsi),%r12
bswap %r12
mov %r9,%r13
mov %r9,%r14
mov %r10,%r15
ror $14,%r13
ror $18,%r14
xor %r11,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r9,%r15 # (f^g)&e
mov %r12,56(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r11,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rax,%r12 # T1+=h
mov %rbx,%rax
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rbx,%r13
mov %rbx,%r14
ror $28,%rax
ror $34,%r13
mov %rbx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rax
ror $5,%r13
or %rdx,%r14 # a|c
xor %r13,%rax # h=Sigma0(a)
and %rdx,%r15 # a&c
add %r12,%r8 # d+=T1
and %rcx,%r14 # (a|c)&b
add %r12,%rax # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rax # h+=Maj(a,b,c)
mov 8*8(%rsi),%r12
bswap %r12
mov %r8,%r13
mov %r8,%r14
mov %r9,%r15
ror $14,%r13
ror $18,%r14
xor %r10,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r8,%r15 # (f^g)&e
mov %r12,64(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r10,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r11,%r12 # T1+=h
mov %rax,%r11
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rax,%r13
mov %rax,%r14
ror $28,%r11
ror $34,%r13
mov %rax,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r11
ror $5,%r13
or %rcx,%r14 # a|c
xor %r13,%r11 # h=Sigma0(a)
and %rcx,%r15 # a&c
add %r12,%rdx # d+=T1
and %rbx,%r14 # (a|c)&b
add %r12,%r11 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r11 # h+=Maj(a,b,c)
mov 8*9(%rsi),%r12
bswap %r12
mov %rdx,%r13
mov %rdx,%r14
mov %r8,%r15
ror $14,%r13
ror $18,%r14
xor %r9,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rdx,%r15 # (f^g)&e
mov %r12,72(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r9,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r10,%r12 # T1+=h
mov %r11,%r10
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r11,%r13
mov %r11,%r14
ror $28,%r10
ror $34,%r13
mov %r11,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r10
ror $5,%r13
or %rbx,%r14 # a|c
xor %r13,%r10 # h=Sigma0(a)
and %rbx,%r15 # a&c
add %r12,%rcx # d+=T1
and %rax,%r14 # (a|c)&b
add %r12,%r10 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r10 # h+=Maj(a,b,c)
mov 8*10(%rsi),%r12
bswap %r12
mov %rcx,%r13
mov %rcx,%r14
mov %rdx,%r15
ror $14,%r13
ror $18,%r14
xor %r8,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rcx,%r15 # (f^g)&e
mov %r12,80(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r8,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r9,%r12 # T1+=h
mov %r10,%r9
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r10,%r13
mov %r10,%r14
ror $28,%r9
ror $34,%r13
mov %r10,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r9
ror $5,%r13
or %rax,%r14 # a|c
xor %r13,%r9 # h=Sigma0(a)
and %rax,%r15 # a&c
add %r12,%rbx # d+=T1
and %r11,%r14 # (a|c)&b
add %r12,%r9 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r9 # h+=Maj(a,b,c)
mov 8*11(%rsi),%r12
bswap %r12
mov %rbx,%r13
mov %rbx,%r14
mov %rcx,%r15
ror $14,%r13
ror $18,%r14
xor %rdx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rbx,%r15 # (f^g)&e
mov %r12,88(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rdx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r8,%r12 # T1+=h
mov %r9,%r8
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r9,%r13
mov %r9,%r14
ror $28,%r8
ror $34,%r13
mov %r9,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r8
ror $5,%r13
or %r11,%r14 # a|c
xor %r13,%r8 # h=Sigma0(a)
and %r11,%r15 # a&c
add %r12,%rax # d+=T1
and %r10,%r14 # (a|c)&b
add %r12,%r8 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r8 # h+=Maj(a,b,c)
mov 8*12(%rsi),%r12
bswap %r12
mov %rax,%r13
mov %rax,%r14
mov %rbx,%r15
ror $14,%r13
ror $18,%r14
xor %rcx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rax,%r15 # (f^g)&e
mov %r12,96(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rcx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rdx,%r12 # T1+=h
mov %r8,%rdx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r8,%r13
mov %r8,%r14
ror $28,%rdx
ror $34,%r13
mov %r8,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rdx
ror $5,%r13
or %r10,%r14 # a|c
xor %r13,%rdx # h=Sigma0(a)
and %r10,%r15 # a&c
add %r12,%r11 # d+=T1
and %r9,%r14 # (a|c)&b
add %r12,%rdx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rdx # h+=Maj(a,b,c)
mov 8*13(%rsi),%r12
bswap %r12
mov %r11,%r13
mov %r11,%r14
mov %rax,%r15
ror $14,%r13
ror $18,%r14
xor %rbx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r11,%r15 # (f^g)&e
mov %r12,104(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rbx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rcx,%r12 # T1+=h
mov %rdx,%rcx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rdx,%r13
mov %rdx,%r14
ror $28,%rcx
ror $34,%r13
mov %rdx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rcx
ror $5,%r13
or %r9,%r14 # a|c
xor %r13,%rcx # h=Sigma0(a)
and %r9,%r15 # a&c
add %r12,%r10 # d+=T1
and %r8,%r14 # (a|c)&b
add %r12,%rcx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rcx # h+=Maj(a,b,c)
mov 8*14(%rsi),%r12
bswap %r12
mov %r10,%r13
mov %r10,%r14
mov %r11,%r15
ror $14,%r13
ror $18,%r14
xor %rax,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r10,%r15 # (f^g)&e
mov %r12,112(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rax,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rbx,%r12 # T1+=h
mov %rcx,%rbx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rcx,%r13
mov %rcx,%r14
ror $28,%rbx
ror $34,%r13
mov %rcx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rbx
ror $5,%r13
or %r8,%r14 # a|c
xor %r13,%rbx # h=Sigma0(a)
and %r8,%r15 # a&c
add %r12,%r9 # d+=T1
and %rdx,%r14 # (a|c)&b
add %r12,%rbx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rbx # h+=Maj(a,b,c)
mov 8*15(%rsi),%r12
bswap %r12
mov %r9,%r13
mov %r9,%r14
mov %r10,%r15
ror $14,%r13
ror $18,%r14
xor %r11,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r9,%r15 # (f^g)&e
mov %r12,120(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r11,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rax,%r12 # T1+=h
mov %rbx,%rax
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rbx,%r13
mov %rbx,%r14
ror $28,%rax
ror $34,%r13
mov %rbx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rax
ror $5,%r13
or %rdx,%r14 # a|c
xor %r13,%rax # h=Sigma0(a)
and %rdx,%r15 # a&c
add %r12,%r8 # d+=T1
and %rcx,%r14 # (a|c)&b
add %r12,%rax # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rax # h+=Maj(a,b,c)
jmp .Lrounds_16_xx
.align 16
.Lrounds_16_xx:
mov 8(%rsp),%r13
mov 112(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 72(%rsp),%r12
add 0(%rsp),%r12
mov %r8,%r13
mov %r8,%r14
mov %r9,%r15
ror $14,%r13
ror $18,%r14
xor %r10,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r8,%r15 # (f^g)&e
mov %r12,0(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r10,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r11,%r12 # T1+=h
mov %rax,%r11
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rax,%r13
mov %rax,%r14
ror $28,%r11
ror $34,%r13
mov %rax,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r11
ror $5,%r13
or %rcx,%r14 # a|c
xor %r13,%r11 # h=Sigma0(a)
and %rcx,%r15 # a&c
add %r12,%rdx # d+=T1
and %rbx,%r14 # (a|c)&b
add %r12,%r11 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r11 # h+=Maj(a,b,c)
mov 16(%rsp),%r13
mov 120(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 80(%rsp),%r12
add 8(%rsp),%r12
mov %rdx,%r13
mov %rdx,%r14
mov %r8,%r15
ror $14,%r13
ror $18,%r14
xor %r9,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rdx,%r15 # (f^g)&e
mov %r12,8(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r9,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r10,%r12 # T1+=h
mov %r11,%r10
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r11,%r13
mov %r11,%r14
ror $28,%r10
ror $34,%r13
mov %r11,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r10
ror $5,%r13
or %rbx,%r14 # a|c
xor %r13,%r10 # h=Sigma0(a)
and %rbx,%r15 # a&c
add %r12,%rcx # d+=T1
and %rax,%r14 # (a|c)&b
add %r12,%r10 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r10 # h+=Maj(a,b,c)
mov 24(%rsp),%r13
mov 0(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 88(%rsp),%r12
add 16(%rsp),%r12
mov %rcx,%r13
mov %rcx,%r14
mov %rdx,%r15
ror $14,%r13
ror $18,%r14
xor %r8,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rcx,%r15 # (f^g)&e
mov %r12,16(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r8,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r9,%r12 # T1+=h
mov %r10,%r9
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r10,%r13
mov %r10,%r14
ror $28,%r9
ror $34,%r13
mov %r10,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r9
ror $5,%r13
or %rax,%r14 # a|c
xor %r13,%r9 # h=Sigma0(a)
and %rax,%r15 # a&c
add %r12,%rbx # d+=T1
and %r11,%r14 # (a|c)&b
add %r12,%r9 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r9 # h+=Maj(a,b,c)
mov 32(%rsp),%r13
mov 8(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 96(%rsp),%r12
add 24(%rsp),%r12
mov %rbx,%r13
mov %rbx,%r14
mov %rcx,%r15
ror $14,%r13
ror $18,%r14
xor %rdx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rbx,%r15 # (f^g)&e
mov %r12,24(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rdx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r8,%r12 # T1+=h
mov %r9,%r8
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r9,%r13
mov %r9,%r14
ror $28,%r8
ror $34,%r13
mov %r9,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r8
ror $5,%r13
or %r11,%r14 # a|c
xor %r13,%r8 # h=Sigma0(a)
and %r11,%r15 # a&c
add %r12,%rax # d+=T1
and %r10,%r14 # (a|c)&b
add %r12,%r8 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r8 # h+=Maj(a,b,c)
mov 40(%rsp),%r13
mov 16(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 104(%rsp),%r12
add 32(%rsp),%r12
mov %rax,%r13
mov %rax,%r14
mov %rbx,%r15
ror $14,%r13
ror $18,%r14
xor %rcx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rax,%r15 # (f^g)&e
mov %r12,32(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rcx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rdx,%r12 # T1+=h
mov %r8,%rdx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r8,%r13
mov %r8,%r14
ror $28,%rdx
ror $34,%r13
mov %r8,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rdx
ror $5,%r13
or %r10,%r14 # a|c
xor %r13,%rdx # h=Sigma0(a)
and %r10,%r15 # a&c
add %r12,%r11 # d+=T1
and %r9,%r14 # (a|c)&b
add %r12,%rdx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rdx # h+=Maj(a,b,c)
mov 48(%rsp),%r13
mov 24(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 112(%rsp),%r12
add 40(%rsp),%r12
mov %r11,%r13
mov %r11,%r14
mov %rax,%r15
ror $14,%r13
ror $18,%r14
xor %rbx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r11,%r15 # (f^g)&e
mov %r12,40(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rbx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rcx,%r12 # T1+=h
mov %rdx,%rcx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rdx,%r13
mov %rdx,%r14
ror $28,%rcx
ror $34,%r13
mov %rdx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rcx
ror $5,%r13
or %r9,%r14 # a|c
xor %r13,%rcx # h=Sigma0(a)
and %r9,%r15 # a&c
add %r12,%r10 # d+=T1
and %r8,%r14 # (a|c)&b
add %r12,%rcx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rcx # h+=Maj(a,b,c)
mov 56(%rsp),%r13
mov 32(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 120(%rsp),%r12
add 48(%rsp),%r12
mov %r10,%r13
mov %r10,%r14
mov %r11,%r15
ror $14,%r13
ror $18,%r14
xor %rax,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r10,%r15 # (f^g)&e
mov %r12,48(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rax,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rbx,%r12 # T1+=h
mov %rcx,%rbx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rcx,%r13
mov %rcx,%r14
ror $28,%rbx
ror $34,%r13
mov %rcx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rbx
ror $5,%r13
or %r8,%r14 # a|c
xor %r13,%rbx # h=Sigma0(a)
and %r8,%r15 # a&c
add %r12,%r9 # d+=T1
and %rdx,%r14 # (a|c)&b
add %r12,%rbx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rbx # h+=Maj(a,b,c)
mov 64(%rsp),%r13
mov 40(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 0(%rsp),%r12
add 56(%rsp),%r12
mov %r9,%r13
mov %r9,%r14
mov %r10,%r15
ror $14,%r13
ror $18,%r14
xor %r11,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r9,%r15 # (f^g)&e
mov %r12,56(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r11,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rax,%r12 # T1+=h
mov %rbx,%rax
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rbx,%r13
mov %rbx,%r14
ror $28,%rax
ror $34,%r13
mov %rbx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rax
ror $5,%r13
or %rdx,%r14 # a|c
xor %r13,%rax # h=Sigma0(a)
and %rdx,%r15 # a&c
add %r12,%r8 # d+=T1
and %rcx,%r14 # (a|c)&b
add %r12,%rax # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rax # h+=Maj(a,b,c)
mov 72(%rsp),%r13
mov 48(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 8(%rsp),%r12
add 64(%rsp),%r12
mov %r8,%r13
mov %r8,%r14
mov %r9,%r15
ror $14,%r13
ror $18,%r14
xor %r10,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r8,%r15 # (f^g)&e
mov %r12,64(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r10,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r11,%r12 # T1+=h
mov %rax,%r11
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rax,%r13
mov %rax,%r14
ror $28,%r11
ror $34,%r13
mov %rax,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r11
ror $5,%r13
or %rcx,%r14 # a|c
xor %r13,%r11 # h=Sigma0(a)
and %rcx,%r15 # a&c
add %r12,%rdx # d+=T1
and %rbx,%r14 # (a|c)&b
add %r12,%r11 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r11 # h+=Maj(a,b,c)
mov 80(%rsp),%r13
mov 56(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 16(%rsp),%r12
add 72(%rsp),%r12
mov %rdx,%r13
mov %rdx,%r14
mov %r8,%r15
ror $14,%r13
ror $18,%r14
xor %r9,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rdx,%r15 # (f^g)&e
mov %r12,72(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r9,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r10,%r12 # T1+=h
mov %r11,%r10
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r11,%r13
mov %r11,%r14
ror $28,%r10
ror $34,%r13
mov %r11,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r10
ror $5,%r13
or %rbx,%r14 # a|c
xor %r13,%r10 # h=Sigma0(a)
and %rbx,%r15 # a&c
add %r12,%rcx # d+=T1
and %rax,%r14 # (a|c)&b
add %r12,%r10 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r10 # h+=Maj(a,b,c)
mov 88(%rsp),%r13
mov 64(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 24(%rsp),%r12
add 80(%rsp),%r12
mov %rcx,%r13
mov %rcx,%r14
mov %rdx,%r15
ror $14,%r13
ror $18,%r14
xor %r8,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rcx,%r15 # (f^g)&e
mov %r12,80(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r8,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r9,%r12 # T1+=h
mov %r10,%r9
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r10,%r13
mov %r10,%r14
ror $28,%r9
ror $34,%r13
mov %r10,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r9
ror $5,%r13
or %rax,%r14 # a|c
xor %r13,%r9 # h=Sigma0(a)
and %rax,%r15 # a&c
add %r12,%rbx # d+=T1
and %r11,%r14 # (a|c)&b
add %r12,%r9 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r9 # h+=Maj(a,b,c)
mov 96(%rsp),%r13
mov 72(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 32(%rsp),%r12
add 88(%rsp),%r12
mov %rbx,%r13
mov %rbx,%r14
mov %rcx,%r15
ror $14,%r13
ror $18,%r14
xor %rdx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rbx,%r15 # (f^g)&e
mov %r12,88(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rdx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %r8,%r12 # T1+=h
mov %r9,%r8
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r9,%r13
mov %r9,%r14
ror $28,%r8
ror $34,%r13
mov %r9,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%r8
ror $5,%r13
or %r11,%r14 # a|c
xor %r13,%r8 # h=Sigma0(a)
and %r11,%r15 # a&c
add %r12,%rax # d+=T1
and %r10,%r14 # (a|c)&b
add %r12,%r8 # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%r8 # h+=Maj(a,b,c)
mov 104(%rsp),%r13
mov 80(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 40(%rsp),%r12
add 96(%rsp),%r12
mov %rax,%r13
mov %rax,%r14
mov %rbx,%r15
ror $14,%r13
ror $18,%r14
xor %rcx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %rax,%r15 # (f^g)&e
mov %r12,96(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rcx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rdx,%r12 # T1+=h
mov %r8,%rdx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %r8,%r13
mov %r8,%r14
ror $28,%rdx
ror $34,%r13
mov %r8,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rdx
ror $5,%r13
or %r10,%r14 # a|c
xor %r13,%rdx # h=Sigma0(a)
and %r10,%r15 # a&c
add %r12,%r11 # d+=T1
and %r9,%r14 # (a|c)&b
add %r12,%rdx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rdx # h+=Maj(a,b,c)
mov 112(%rsp),%r13
mov 88(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 48(%rsp),%r12
add 104(%rsp),%r12
mov %r11,%r13
mov %r11,%r14
mov %rax,%r15
ror $14,%r13
ror $18,%r14
xor %rbx,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r11,%r15 # (f^g)&e
mov %r12,104(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rbx,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rcx,%r12 # T1+=h
mov %rdx,%rcx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rdx,%r13
mov %rdx,%r14
ror $28,%rcx
ror $34,%r13
mov %rdx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rcx
ror $5,%r13
or %r9,%r14 # a|c
xor %r13,%rcx # h=Sigma0(a)
and %r9,%r15 # a&c
add %r12,%r10 # d+=T1
and %r8,%r14 # (a|c)&b
add %r12,%rcx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rcx # h+=Maj(a,b,c)
mov 120(%rsp),%r13
mov 96(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 56(%rsp),%r12
add 112(%rsp),%r12
mov %r10,%r13
mov %r10,%r14
mov %r11,%r15
ror $14,%r13
ror $18,%r14
xor %rax,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r10,%r15 # (f^g)&e
mov %r12,112(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %rax,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rbx,%r12 # T1+=h
mov %rcx,%rbx
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rcx,%r13
mov %rcx,%r14
ror $28,%rbx
ror $34,%r13
mov %rcx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rbx
ror $5,%r13
or %r8,%r14 # a|c
xor %r13,%rbx # h=Sigma0(a)
and %r8,%r15 # a&c
add %r12,%r9 # d+=T1
and %rdx,%r14 # (a|c)&b
add %r12,%rbx # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rbx # h+=Maj(a,b,c)
mov 0(%rsp),%r13
mov 104(%rsp),%r12
mov %r13,%r15
shr $7,%r13
ror $1,%r15
xor %r15,%r13
ror $7,%r15
xor %r15,%r13 # sigma0(X[(i+1)&0xf])
mov %r12,%r14
shr $6,%r12
ror $19,%r14
xor %r14,%r12
ror $42,%r14
xor %r14,%r12 # sigma1(X[(i+14)&0xf])
add %r13,%r12
add 64(%rsp),%r12
add 120(%rsp),%r12
mov %r9,%r13
mov %r9,%r14
mov %r10,%r15
ror $14,%r13
ror $18,%r14
xor %r11,%r15 # f^g
xor %r14,%r13
ror $23,%r14
and %r9,%r15 # (f^g)&e
mov %r12,120(%rsp)
xor %r14,%r13 # Sigma1(e)
xor %r11,%r15 # Ch(e,f,g)=((f^g)&e)^g
add %rax,%r12 # T1+=h
mov %rbx,%rax
add %r13,%r12 # T1+=Sigma1(e)
add %r15,%r12 # T1+=Ch(e,f,g)
mov %rbx,%r13
mov %rbx,%r14
ror $28,%rax
ror $34,%r13
mov %rbx,%r15
add (%rbp,%rdi,8),%r12 # T1+=K[round]
xor %r13,%rax
ror $5,%r13
or %rdx,%r14 # a|c
xor %r13,%rax # h=Sigma0(a)
and %rdx,%r15 # a&c
add %r12,%r8 # d+=T1
and %rcx,%r14 # (a|c)&b
add %r12,%rax # h+=T1
or %r15,%r14 # Maj(a,b,c)=((a|c)&b)|(a&c)
lea 1(%rdi),%rdi # round++
add %r14,%rax # h+=Maj(a,b,c)
cmp $80,%rdi
jb .Lrounds_16_xx
mov 16*8+0*8(%rsp),%rdi
lea 16*8(%rsi),%rsi
add 8*0(%rdi),%rax
add 8*1(%rdi),%rbx
add 8*2(%rdi),%rcx
add 8*3(%rdi),%rdx
add 8*4(%rdi),%r8
add 8*5(%rdi),%r9
add 8*6(%rdi),%r10
add 8*7(%rdi),%r11
cmp 16*8+2*8(%rsp),%rsi
mov %rax,8*0(%rdi)
mov %rbx,8*1(%rdi)
mov %rcx,8*2(%rdi)
mov %rdx,8*3(%rdi)
mov %r8,8*4(%rdi)
mov %r9,8*5(%rdi)
mov %r10,8*6(%rdi)
mov %r11,8*7(%rdi)
jb .Lloop
mov 16*8+3*8(%rsp),%rsp
.cfi_def_cfa %rsp,56
pop %r15
.cfi_adjust_cfa_offset -8
.cfi_restore %r15
pop %r14
.cfi_adjust_cfa_offset -8
.cfi_restore %r14
pop %r13
.cfi_adjust_cfa_offset -8
.cfi_restore %r13
pop %r12
.cfi_adjust_cfa_offset -8
.cfi_restore %r12
pop %rbp
.cfi_adjust_cfa_offset -8
.cfi_restore %rbp
pop %rbx
.cfi_adjust_cfa_offset -8
.cfi_restore %rbx
ret
.cfi_endproc
SET_SIZE(SHA512TransformBlocks)
-.data
+.section .rodata
.align 64
.type K512,@object
K512:
.quad 0x428a2f98d728ae22,0x7137449123ef65cd
.quad 0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc
.quad 0x3956c25bf348b538,0x59f111f1b605d019
.quad 0x923f82a4af194f9b,0xab1c5ed5da6d8118
.quad 0xd807aa98a3030242,0x12835b0145706fbe
.quad 0x243185be4ee4b28c,0x550c7dc3d5ffb4e2
.quad 0x72be5d74f27b896f,0x80deb1fe3b1696b1
.quad 0x9bdc06a725c71235,0xc19bf174cf692694
.quad 0xe49b69c19ef14ad2,0xefbe4786384f25e3
.quad 0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65
.quad 0x2de92c6f592b0275,0x4a7484aa6ea6e483
.quad 0x5cb0a9dcbd41fbd4,0x76f988da831153b5
.quad 0x983e5152ee66dfab,0xa831c66d2db43210
.quad 0xb00327c898fb213f,0xbf597fc7beef0ee4
.quad 0xc6e00bf33da88fc2,0xd5a79147930aa725
.quad 0x06ca6351e003826f,0x142929670a0e6e70
.quad 0x27b70a8546d22ffc,0x2e1b21385c26c926
.quad 0x4d2c6dfc5ac42aed,0x53380d139d95b3df
.quad 0x650a73548baf63de,0x766a0abb3c77b2a8
.quad 0x81c2c92e47edaee6,0x92722c851482353b
.quad 0xa2bfe8a14cf10364,0xa81a664bbc423001
.quad 0xc24b8b70d0f89791,0xc76c51a30654be30
.quad 0xd192e819d6ef5218,0xd69906245565a910
.quad 0xf40e35855771202a,0x106aa07032bbd1b8
.quad 0x19a4c116b8d2d0c8,0x1e376c085141ab53
.quad 0x2748774cdf8eeb99,0x34b0bcb5e19b48a8
.quad 0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb
.quad 0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3
.quad 0x748f82ee5defb2fc,0x78a5636f43172f60
.quad 0x84c87814a1f0ab72,0x8cc702081a6439ec
.quad 0x90befffa23631e28,0xa4506cebde82bde9
.quad 0xbef9a3f7b2c67915,0xc67178f2e372532b
.quad 0xca273eceea26619c,0xd186b8c721c0c207
.quad 0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178
.quad 0x06f067aa72176fba,0x0a637dc5a2c898a6
.quad 0x113f9804bef90dae,0x1b710b35131c471b
.quad 0x28db77f523047d84,0x32caab7b40c72493
.quad 0x3c9ebe0a15c9bebc,0x431d67c49c100d4c
.quad 0x4cc5d4becb3e42b6,0x597f299cfc657e2a
.quad 0x5fcb6fab3ad6faec,0x6c44198c4a475817
#endif /* !lint && !__lint */
#ifdef __ELF__
.section .note.GNU-stack,"",%progbits
#endif
diff --git a/module/icp/core/kcf_callprov.c b/module/icp/core/kcf_callprov.c
index fd2f7e1aac3d..345014d0a1e4 100644
--- a/module/icp/core/kcf_callprov.c
+++ b/module/icp/core/kcf_callprov.c
@@ -1,1567 +1,1567 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#include <sys/crypto/common.h>
#include <sys/crypto/impl.h>
#include <sys/crypto/sched_impl.h>
static int kcf_emulate_dual(kcf_provider_desc_t *, crypto_ctx_t *,
kcf_req_params_t *);
void
kcf_free_triedlist(kcf_prov_tried_t *list)
{
kcf_prov_tried_t *l;
while ((l = list) != NULL) {
list = list->pt_next;
KCF_PROV_REFRELE(l->pt_pd);
kmem_free(l, sizeof (kcf_prov_tried_t));
}
}
kcf_prov_tried_t *
kcf_insert_triedlist(kcf_prov_tried_t **list, kcf_provider_desc_t *pd,
int kmflag)
{
kcf_prov_tried_t *l;
l = kmem_alloc(sizeof (kcf_prov_tried_t), kmflag);
if (l == NULL)
return (NULL);
l->pt_pd = pd;
l->pt_next = *list;
*list = l;
return (l);
}
static boolean_t
is_in_triedlist(kcf_provider_desc_t *pd, kcf_prov_tried_t *triedl)
{
while (triedl != NULL) {
if (triedl->pt_pd == pd)
return (B_TRUE);
triedl = triedl->pt_next;
};
return (B_FALSE);
}
/*
* Search a mech entry's hardware provider list for the specified
* provider. Return true if found.
*/
static boolean_t
is_valid_provider_for_mech(kcf_provider_desc_t *pd, kcf_mech_entry_t *me,
crypto_func_group_t fg)
{
kcf_prov_mech_desc_t *prov_chain;
prov_chain = me->me_hw_prov_chain;
if (prov_chain != NULL) {
ASSERT(me->me_num_hwprov > 0);
for (; prov_chain != NULL; prov_chain = prov_chain->pm_next) {
if (prov_chain->pm_prov_desc == pd &&
IS_FG_SUPPORTED(prov_chain, fg)) {
return (B_TRUE);
}
}
}
return (B_FALSE);
}
/*
* This routine, given a logical provider, returns the least loaded
* provider belonging to the logical provider. The provider must be
* able to do the specified mechanism, i.e. check that the mechanism
* hasn't been disabled. In addition, just in case providers are not
* entirely equivalent, the provider's entry point is checked for
* non-nullness. This is accomplished by having the caller pass, as
* arguments, the offset of the function group (offset_1), and the
* offset of the function within the function group (offset_2).
* Returns NULL if no provider can be found.
*/
int
kcf_get_hardware_provider(crypto_mech_type_t mech_type_1,
crypto_mech_type_t mech_type_2, boolean_t call_restrict,
kcf_provider_desc_t *old, kcf_provider_desc_t **new, crypto_func_group_t fg)
{
kcf_provider_desc_t *provider, *real_pd = old;
kcf_provider_desc_t *gpd = NULL; /* good provider */
kcf_provider_desc_t *bpd = NULL; /* busy provider */
kcf_provider_list_t *p;
kcf_ops_class_t class;
kcf_mech_entry_t *me;
- kcf_mech_entry_tab_t *me_tab;
+ const kcf_mech_entry_tab_t *me_tab;
int index, len, gqlen = INT_MAX, rv = CRYPTO_SUCCESS;
/* get the mech entry for the specified mechanism */
class = KCF_MECH2CLASS(mech_type_1);
if ((class < KCF_FIRST_OPSCLASS) || (class > KCF_LAST_OPSCLASS)) {
return (CRYPTO_MECHANISM_INVALID);
}
me_tab = &kcf_mech_tabs_tab[class];
index = KCF_MECH2INDEX(mech_type_1);
if ((index < 0) || (index >= me_tab->met_size)) {
return (CRYPTO_MECHANISM_INVALID);
}
me = &((me_tab->met_tab)[index]);
mutex_enter(&me->me_mutex);
/*
* We assume the provider descriptor will not go away because
* it is being held somewhere, i.e. its reference count has been
* incremented. In the case of the crypto module, the provider
* descriptor is held by the session structure.
*/
if (old->pd_prov_type == CRYPTO_LOGICAL_PROVIDER) {
if (old->pd_provider_list == NULL) {
real_pd = NULL;
rv = CRYPTO_DEVICE_ERROR;
goto out;
}
/*
* Find the least loaded real provider. KCF_PROV_LOAD gives
* the load (number of pending requests) of the provider.
*/
mutex_enter(&old->pd_lock);
p = old->pd_provider_list;
while (p != NULL) {
provider = p->pl_provider;
ASSERT(provider->pd_prov_type !=
CRYPTO_LOGICAL_PROVIDER);
if (call_restrict &&
(provider->pd_flags & KCF_PROV_RESTRICTED)) {
p = p->pl_next;
continue;
}
if (!is_valid_provider_for_mech(provider, me, fg)) {
p = p->pl_next;
continue;
}
/* provider does second mech */
if (mech_type_2 != CRYPTO_MECH_INVALID) {
int i;
i = KCF_TO_PROV_MECH_INDX(provider,
mech_type_2);
if (i == KCF_INVALID_INDX) {
p = p->pl_next;
continue;
}
}
if (provider->pd_state != KCF_PROV_READY) {
/* choose BUSY if no READY providers */
if (provider->pd_state == KCF_PROV_BUSY)
bpd = provider;
p = p->pl_next;
continue;
}
len = KCF_PROV_LOAD(provider);
if (len < gqlen) {
gqlen = len;
gpd = provider;
}
p = p->pl_next;
}
if (gpd != NULL) {
real_pd = gpd;
KCF_PROV_REFHOLD(real_pd);
} else if (bpd != NULL) {
real_pd = bpd;
KCF_PROV_REFHOLD(real_pd);
} else {
/* can't find provider */
real_pd = NULL;
rv = CRYPTO_MECHANISM_INVALID;
}
mutex_exit(&old->pd_lock);
} else {
if (!KCF_IS_PROV_USABLE(old) ||
(call_restrict && (old->pd_flags & KCF_PROV_RESTRICTED))) {
real_pd = NULL;
rv = CRYPTO_DEVICE_ERROR;
goto out;
}
if (!is_valid_provider_for_mech(old, me, fg)) {
real_pd = NULL;
rv = CRYPTO_MECHANISM_INVALID;
goto out;
}
KCF_PROV_REFHOLD(real_pd);
}
out:
mutex_exit(&me->me_mutex);
*new = real_pd;
return (rv);
}
/*
* Return the best provider for the specified mechanism. The provider
* is held and it is the caller's responsibility to release it when done.
* The fg input argument is used as a search criterion to pick a provider.
* A provider has to support this function group to be picked.
*
* Find the least loaded provider in the list of providers. We do a linear
* search to find one. This is fine as we assume there are only a few
* number of providers in this list. If this assumption ever changes,
* we should revisit this.
*
* call_restrict represents if the caller should not be allowed to
* use restricted providers.
*/
kcf_provider_desc_t *
kcf_get_mech_provider(crypto_mech_type_t mech_type, kcf_mech_entry_t **mepp,
int *error, kcf_prov_tried_t *triedl, crypto_func_group_t fg,
boolean_t call_restrict, size_t data_size)
{
kcf_provider_desc_t *pd = NULL, *gpd = NULL;
kcf_prov_mech_desc_t *prov_chain, *mdesc;
int len, gqlen = INT_MAX;
kcf_ops_class_t class;
int index;
kcf_mech_entry_t *me;
- kcf_mech_entry_tab_t *me_tab;
+ const kcf_mech_entry_tab_t *me_tab;
class = KCF_MECH2CLASS(mech_type);
if ((class < KCF_FIRST_OPSCLASS) || (class > KCF_LAST_OPSCLASS)) {
*error = CRYPTO_MECHANISM_INVALID;
return (NULL);
}
me_tab = &kcf_mech_tabs_tab[class];
index = KCF_MECH2INDEX(mech_type);
if ((index < 0) || (index >= me_tab->met_size)) {
*error = CRYPTO_MECHANISM_INVALID;
return (NULL);
}
me = &((me_tab->met_tab)[index]);
if (mepp != NULL)
*mepp = me;
mutex_enter(&me->me_mutex);
prov_chain = me->me_hw_prov_chain;
/*
* We check for the threshold for using a hardware provider for
* this amount of data. If there is no software provider available
* for the mechanism, then the threshold is ignored.
*/
if ((prov_chain != NULL) &&
((data_size == 0) || (me->me_threshold == 0) ||
(data_size >= me->me_threshold) ||
((mdesc = me->me_sw_prov) == NULL) ||
(!IS_FG_SUPPORTED(mdesc, fg)) ||
(!KCF_IS_PROV_USABLE(mdesc->pm_prov_desc)))) {
ASSERT(me->me_num_hwprov > 0);
/* there is at least one provider */
/*
* Find the least loaded real provider. KCF_PROV_LOAD gives
* the load (number of pending requests) of the provider.
*/
while (prov_chain != NULL) {
pd = prov_chain->pm_prov_desc;
if (!IS_FG_SUPPORTED(prov_chain, fg) ||
!KCF_IS_PROV_USABLE(pd) ||
IS_PROVIDER_TRIED(pd, triedl) ||
(call_restrict &&
(pd->pd_flags & KCF_PROV_RESTRICTED))) {
prov_chain = prov_chain->pm_next;
continue;
}
if ((len = KCF_PROV_LOAD(pd)) < gqlen) {
gqlen = len;
gpd = pd;
}
prov_chain = prov_chain->pm_next;
}
pd = gpd;
}
/* No HW provider for this mech, is there a SW provider? */
if (pd == NULL && (mdesc = me->me_sw_prov) != NULL) {
pd = mdesc->pm_prov_desc;
if (!IS_FG_SUPPORTED(mdesc, fg) ||
!KCF_IS_PROV_USABLE(pd) ||
IS_PROVIDER_TRIED(pd, triedl) ||
(call_restrict && (pd->pd_flags & KCF_PROV_RESTRICTED)))
pd = NULL;
}
if (pd == NULL) {
/*
* We do not want to report CRYPTO_MECH_NOT_SUPPORTED, when
* we are in the "fallback to the next provider" case. Rather
* we preserve the error, so that the client gets the right
* error code.
*/
if (triedl == NULL)
*error = CRYPTO_MECH_NOT_SUPPORTED;
} else
KCF_PROV_REFHOLD(pd);
mutex_exit(&me->me_mutex);
return (pd);
}
/*
* Very similar to kcf_get_mech_provider(). Finds the best provider capable of
* a dual operation with both me1 and me2.
* When no dual-ops capable providers are available, return the best provider
* for me1 only, and sets *prov_mt2 to CRYPTO_INVALID_MECHID;
* We assume/expect that a slower HW capable of the dual is still
* faster than the 2 fastest providers capable of the individual ops
* separately.
*/
kcf_provider_desc_t *
kcf_get_dual_provider(crypto_mechanism_t *mech1, crypto_mechanism_t *mech2,
kcf_mech_entry_t **mepp, crypto_mech_type_t *prov_mt1,
crypto_mech_type_t *prov_mt2, int *error, kcf_prov_tried_t *triedl,
crypto_func_group_t fg1, crypto_func_group_t fg2, boolean_t call_restrict,
size_t data_size)
{
kcf_provider_desc_t *pd = NULL, *pdm1 = NULL, *pdm1m2 = NULL;
kcf_prov_mech_desc_t *prov_chain, *mdesc;
int len, gqlen = INT_MAX, dgqlen = INT_MAX;
crypto_mech_info_list_t *mil;
crypto_mech_type_t m2id = mech2->cm_type;
kcf_mech_entry_t *me;
/* when mech is a valid mechanism, me will be its mech_entry */
if (kcf_get_mech_entry(mech1->cm_type, &me) != KCF_SUCCESS) {
*error = CRYPTO_MECHANISM_INVALID;
return (NULL);
}
*prov_mt2 = CRYPTO_MECH_INVALID;
if (mepp != NULL)
*mepp = me;
mutex_enter(&me->me_mutex);
prov_chain = me->me_hw_prov_chain;
/*
* We check the threshold for using a hardware provider for
* this amount of data. If there is no software provider available
* for the first mechanism, then the threshold is ignored.
*/
if ((prov_chain != NULL) &&
((data_size == 0) || (me->me_threshold == 0) ||
(data_size >= me->me_threshold) ||
((mdesc = me->me_sw_prov) == NULL) ||
(!IS_FG_SUPPORTED(mdesc, fg1)) ||
(!KCF_IS_PROV_USABLE(mdesc->pm_prov_desc)))) {
/* there is at least one provider */
ASSERT(me->me_num_hwprov > 0);
/*
* Find the least loaded provider capable of the combo
* me1 + me2, and save a pointer to the least loaded
* provider capable of me1 only.
*/
while (prov_chain != NULL) {
pd = prov_chain->pm_prov_desc;
len = KCF_PROV_LOAD(pd);
if (!IS_FG_SUPPORTED(prov_chain, fg1) ||
!KCF_IS_PROV_USABLE(pd) ||
IS_PROVIDER_TRIED(pd, triedl) ||
(call_restrict &&
(pd->pd_flags & KCF_PROV_RESTRICTED))) {
prov_chain = prov_chain->pm_next;
continue;
}
/* Save the best provider capable of m1 */
if (len < gqlen) {
*prov_mt1 =
prov_chain->pm_mech_info.cm_mech_number;
gqlen = len;
pdm1 = pd;
}
/* See if pd can do me2 too */
for (mil = prov_chain->pm_mi_list;
mil != NULL; mil = mil->ml_next) {
if ((mil->ml_mech_info.cm_func_group_mask &
fg2) == 0)
continue;
if ((mil->ml_kcf_mechid == m2id) &&
(len < dgqlen)) {
/* Bingo! */
dgqlen = len;
pdm1m2 = pd;
*prov_mt2 =
mil->ml_mech_info.cm_mech_number;
*prov_mt1 = prov_chain->
pm_mech_info.cm_mech_number;
break;
}
}
prov_chain = prov_chain->pm_next;
}
pd = (pdm1m2 != NULL) ? pdm1m2 : pdm1;
}
/* no HW provider for this mech, is there a SW provider? */
if (pd == NULL && (mdesc = me->me_sw_prov) != NULL) {
pd = mdesc->pm_prov_desc;
if (!IS_FG_SUPPORTED(mdesc, fg1) ||
!KCF_IS_PROV_USABLE(pd) ||
IS_PROVIDER_TRIED(pd, triedl) ||
(call_restrict && (pd->pd_flags & KCF_PROV_RESTRICTED)))
pd = NULL;
else {
/* See if pd can do me2 too */
for (mil = me->me_sw_prov->pm_mi_list;
mil != NULL; mil = mil->ml_next) {
if ((mil->ml_mech_info.cm_func_group_mask &
fg2) == 0)
continue;
if (mil->ml_kcf_mechid == m2id) {
/* Bingo! */
*prov_mt2 =
mil->ml_mech_info.cm_mech_number;
break;
}
}
*prov_mt1 = me->me_sw_prov->pm_mech_info.cm_mech_number;
}
}
if (pd == NULL)
*error = CRYPTO_MECH_NOT_SUPPORTED;
else
KCF_PROV_REFHOLD(pd);
mutex_exit(&me->me_mutex);
return (pd);
}
/*
* Do the actual work of calling the provider routines.
*
* pd - Provider structure
* ctx - Context for this operation
* params - Parameters for this operation
* rhndl - Request handle to use for notification
*
* The return values are the same as that of the respective SPI.
*/
int
common_submit_request(kcf_provider_desc_t *pd, crypto_ctx_t *ctx,
kcf_req_params_t *params, crypto_req_handle_t rhndl)
{
int err = CRYPTO_ARGUMENTS_BAD;
kcf_op_type_t optype;
optype = params->rp_optype;
switch (params->rp_opgrp) {
case KCF_OG_DIGEST: {
kcf_digest_ops_params_t *dops = &params->rp_u.digest_params;
switch (optype) {
case KCF_OP_INIT:
/*
* We should do this only here and not in KCF_WRAP_*
* macros. This is because we may want to try other
* providers, in case we recover from a failure.
*/
KCF_SET_PROVIDER_MECHNUM(dops->do_framework_mechtype,
pd, &dops->do_mech);
err = KCF_PROV_DIGEST_INIT(pd, ctx, &dops->do_mech,
rhndl);
break;
case KCF_OP_SINGLE:
err = KCF_PROV_DIGEST(pd, ctx, dops->do_data,
dops->do_digest, rhndl);
break;
case KCF_OP_UPDATE:
err = KCF_PROV_DIGEST_UPDATE(pd, ctx,
dops->do_data, rhndl);
break;
case KCF_OP_FINAL:
err = KCF_PROV_DIGEST_FINAL(pd, ctx,
dops->do_digest, rhndl);
break;
case KCF_OP_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(dops->do_framework_mechtype,
pd, &dops->do_mech);
err = KCF_PROV_DIGEST_ATOMIC(pd, dops->do_sid,
&dops->do_mech, dops->do_data, dops->do_digest,
rhndl);
break;
case KCF_OP_DIGEST_KEY:
err = KCF_PROV_DIGEST_KEY(pd, ctx, dops->do_digest_key,
rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_MAC: {
kcf_mac_ops_params_t *mops = &params->rp_u.mac_params;
switch (optype) {
case KCF_OP_INIT:
KCF_SET_PROVIDER_MECHNUM(mops->mo_framework_mechtype,
pd, &mops->mo_mech);
err = KCF_PROV_MAC_INIT(pd, ctx, &mops->mo_mech,
mops->mo_key, mops->mo_templ, rhndl);
break;
case KCF_OP_SINGLE:
err = KCF_PROV_MAC(pd, ctx, mops->mo_data,
mops->mo_mac, rhndl);
break;
case KCF_OP_UPDATE:
err = KCF_PROV_MAC_UPDATE(pd, ctx, mops->mo_data,
rhndl);
break;
case KCF_OP_FINAL:
err = KCF_PROV_MAC_FINAL(pd, ctx, mops->mo_mac, rhndl);
break;
case KCF_OP_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(mops->mo_framework_mechtype,
pd, &mops->mo_mech);
err = KCF_PROV_MAC_ATOMIC(pd, mops->mo_sid,
&mops->mo_mech, mops->mo_key, mops->mo_data,
mops->mo_mac, mops->mo_templ, rhndl);
break;
case KCF_OP_MAC_VERIFY_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(mops->mo_framework_mechtype,
pd, &mops->mo_mech);
err = KCF_PROV_MAC_VERIFY_ATOMIC(pd, mops->mo_sid,
&mops->mo_mech, mops->mo_key, mops->mo_data,
mops->mo_mac, mops->mo_templ, rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_ENCRYPT: {
kcf_encrypt_ops_params_t *eops = &params->rp_u.encrypt_params;
switch (optype) {
case KCF_OP_INIT:
KCF_SET_PROVIDER_MECHNUM(eops->eo_framework_mechtype,
pd, &eops->eo_mech);
err = KCF_PROV_ENCRYPT_INIT(pd, ctx, &eops->eo_mech,
eops->eo_key, eops->eo_templ, rhndl);
break;
case KCF_OP_SINGLE:
err = KCF_PROV_ENCRYPT(pd, ctx, eops->eo_plaintext,
eops->eo_ciphertext, rhndl);
break;
case KCF_OP_UPDATE:
err = KCF_PROV_ENCRYPT_UPDATE(pd, ctx,
eops->eo_plaintext, eops->eo_ciphertext, rhndl);
break;
case KCF_OP_FINAL:
err = KCF_PROV_ENCRYPT_FINAL(pd, ctx,
eops->eo_ciphertext, rhndl);
break;
case KCF_OP_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(eops->eo_framework_mechtype,
pd, &eops->eo_mech);
err = KCF_PROV_ENCRYPT_ATOMIC(pd, eops->eo_sid,
&eops->eo_mech, eops->eo_key, eops->eo_plaintext,
eops->eo_ciphertext, eops->eo_templ, rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_DECRYPT: {
kcf_decrypt_ops_params_t *dcrops = &params->rp_u.decrypt_params;
switch (optype) {
case KCF_OP_INIT:
KCF_SET_PROVIDER_MECHNUM(dcrops->dop_framework_mechtype,
pd, &dcrops->dop_mech);
err = KCF_PROV_DECRYPT_INIT(pd, ctx, &dcrops->dop_mech,
dcrops->dop_key, dcrops->dop_templ, rhndl);
break;
case KCF_OP_SINGLE:
err = KCF_PROV_DECRYPT(pd, ctx, dcrops->dop_ciphertext,
dcrops->dop_plaintext, rhndl);
break;
case KCF_OP_UPDATE:
err = KCF_PROV_DECRYPT_UPDATE(pd, ctx,
dcrops->dop_ciphertext, dcrops->dop_plaintext,
rhndl);
break;
case KCF_OP_FINAL:
err = KCF_PROV_DECRYPT_FINAL(pd, ctx,
dcrops->dop_plaintext, rhndl);
break;
case KCF_OP_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(dcrops->dop_framework_mechtype,
pd, &dcrops->dop_mech);
err = KCF_PROV_DECRYPT_ATOMIC(pd, dcrops->dop_sid,
&dcrops->dop_mech, dcrops->dop_key,
dcrops->dop_ciphertext, dcrops->dop_plaintext,
dcrops->dop_templ, rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_SIGN: {
kcf_sign_ops_params_t *sops = &params->rp_u.sign_params;
switch (optype) {
case KCF_OP_INIT:
KCF_SET_PROVIDER_MECHNUM(sops->so_framework_mechtype,
pd, &sops->so_mech);
err = KCF_PROV_SIGN_INIT(pd, ctx, &sops->so_mech,
sops->so_key, sops->so_templ, rhndl);
break;
case KCF_OP_SIGN_RECOVER_INIT:
KCF_SET_PROVIDER_MECHNUM(sops->so_framework_mechtype,
pd, &sops->so_mech);
err = KCF_PROV_SIGN_RECOVER_INIT(pd, ctx,
&sops->so_mech, sops->so_key, sops->so_templ,
rhndl);
break;
case KCF_OP_SINGLE:
err = KCF_PROV_SIGN(pd, ctx, sops->so_data,
sops->so_signature, rhndl);
break;
case KCF_OP_SIGN_RECOVER:
err = KCF_PROV_SIGN_RECOVER(pd, ctx,
sops->so_data, sops->so_signature, rhndl);
break;
case KCF_OP_UPDATE:
err = KCF_PROV_SIGN_UPDATE(pd, ctx, sops->so_data,
rhndl);
break;
case KCF_OP_FINAL:
err = KCF_PROV_SIGN_FINAL(pd, ctx, sops->so_signature,
rhndl);
break;
case KCF_OP_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(sops->so_framework_mechtype,
pd, &sops->so_mech);
err = KCF_PROV_SIGN_ATOMIC(pd, sops->so_sid,
&sops->so_mech, sops->so_key, sops->so_data,
sops->so_templ, sops->so_signature, rhndl);
break;
case KCF_OP_SIGN_RECOVER_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(sops->so_framework_mechtype,
pd, &sops->so_mech);
err = KCF_PROV_SIGN_RECOVER_ATOMIC(pd, sops->so_sid,
&sops->so_mech, sops->so_key, sops->so_data,
sops->so_templ, sops->so_signature, rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_VERIFY: {
kcf_verify_ops_params_t *vops = &params->rp_u.verify_params;
switch (optype) {
case KCF_OP_INIT:
KCF_SET_PROVIDER_MECHNUM(vops->vo_framework_mechtype,
pd, &vops->vo_mech);
err = KCF_PROV_VERIFY_INIT(pd, ctx, &vops->vo_mech,
vops->vo_key, vops->vo_templ, rhndl);
break;
case KCF_OP_VERIFY_RECOVER_INIT:
KCF_SET_PROVIDER_MECHNUM(vops->vo_framework_mechtype,
pd, &vops->vo_mech);
err = KCF_PROV_VERIFY_RECOVER_INIT(pd, ctx,
&vops->vo_mech, vops->vo_key, vops->vo_templ,
rhndl);
break;
case KCF_OP_SINGLE:
err = KCF_PROV_VERIFY(pd, ctx, vops->vo_data,
vops->vo_signature, rhndl);
break;
case KCF_OP_VERIFY_RECOVER:
err = KCF_PROV_VERIFY_RECOVER(pd, ctx,
vops->vo_signature, vops->vo_data, rhndl);
break;
case KCF_OP_UPDATE:
err = KCF_PROV_VERIFY_UPDATE(pd, ctx, vops->vo_data,
rhndl);
break;
case KCF_OP_FINAL:
err = KCF_PROV_VERIFY_FINAL(pd, ctx, vops->vo_signature,
rhndl);
break;
case KCF_OP_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(vops->vo_framework_mechtype,
pd, &vops->vo_mech);
err = KCF_PROV_VERIFY_ATOMIC(pd, vops->vo_sid,
&vops->vo_mech, vops->vo_key, vops->vo_data,
vops->vo_templ, vops->vo_signature, rhndl);
break;
case KCF_OP_VERIFY_RECOVER_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(vops->vo_framework_mechtype,
pd, &vops->vo_mech);
err = KCF_PROV_VERIFY_RECOVER_ATOMIC(pd, vops->vo_sid,
&vops->vo_mech, vops->vo_key, vops->vo_signature,
vops->vo_templ, vops->vo_data, rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_ENCRYPT_MAC: {
kcf_encrypt_mac_ops_params_t *eops =
&params->rp_u.encrypt_mac_params;
kcf_context_t *kcf_secondctx;
switch (optype) {
case KCF_OP_INIT:
kcf_secondctx = ((kcf_context_t *)
(ctx->cc_framework_private))->kc_secondctx;
if (kcf_secondctx != NULL) {
err = kcf_emulate_dual(pd, ctx, params);
break;
}
KCF_SET_PROVIDER_MECHNUM(
eops->em_framework_encr_mechtype,
pd, &eops->em_encr_mech);
KCF_SET_PROVIDER_MECHNUM(
eops->em_framework_mac_mechtype,
pd, &eops->em_mac_mech);
err = KCF_PROV_ENCRYPT_MAC_INIT(pd, ctx,
&eops->em_encr_mech, eops->em_encr_key,
&eops->em_mac_mech, eops->em_mac_key,
eops->em_encr_templ, eops->em_mac_templ,
rhndl);
break;
case KCF_OP_SINGLE:
err = KCF_PROV_ENCRYPT_MAC(pd, ctx,
eops->em_plaintext, eops->em_ciphertext,
eops->em_mac, rhndl);
break;
case KCF_OP_UPDATE:
kcf_secondctx = ((kcf_context_t *)
(ctx->cc_framework_private))->kc_secondctx;
if (kcf_secondctx != NULL) {
err = kcf_emulate_dual(pd, ctx, params);
break;
}
err = KCF_PROV_ENCRYPT_MAC_UPDATE(pd, ctx,
eops->em_plaintext, eops->em_ciphertext, rhndl);
break;
case KCF_OP_FINAL:
kcf_secondctx = ((kcf_context_t *)
(ctx->cc_framework_private))->kc_secondctx;
if (kcf_secondctx != NULL) {
err = kcf_emulate_dual(pd, ctx, params);
break;
}
err = KCF_PROV_ENCRYPT_MAC_FINAL(pd, ctx,
eops->em_ciphertext, eops->em_mac, rhndl);
break;
case KCF_OP_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(
eops->em_framework_encr_mechtype,
pd, &eops->em_encr_mech);
KCF_SET_PROVIDER_MECHNUM(
eops->em_framework_mac_mechtype,
pd, &eops->em_mac_mech);
err = KCF_PROV_ENCRYPT_MAC_ATOMIC(pd, eops->em_sid,
&eops->em_encr_mech, eops->em_encr_key,
&eops->em_mac_mech, eops->em_mac_key,
eops->em_plaintext, eops->em_ciphertext,
eops->em_mac,
eops->em_encr_templ, eops->em_mac_templ,
rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_MAC_DECRYPT: {
kcf_mac_decrypt_ops_params_t *dops =
&params->rp_u.mac_decrypt_params;
kcf_context_t *kcf_secondctx;
switch (optype) {
case KCF_OP_INIT:
kcf_secondctx = ((kcf_context_t *)
(ctx->cc_framework_private))->kc_secondctx;
if (kcf_secondctx != NULL) {
err = kcf_emulate_dual(pd, ctx, params);
break;
}
KCF_SET_PROVIDER_MECHNUM(
dops->md_framework_mac_mechtype,
pd, &dops->md_mac_mech);
KCF_SET_PROVIDER_MECHNUM(
dops->md_framework_decr_mechtype,
pd, &dops->md_decr_mech);
err = KCF_PROV_MAC_DECRYPT_INIT(pd, ctx,
&dops->md_mac_mech, dops->md_mac_key,
&dops->md_decr_mech, dops->md_decr_key,
dops->md_mac_templ, dops->md_decr_templ,
rhndl);
break;
case KCF_OP_SINGLE:
err = KCF_PROV_MAC_DECRYPT(pd, ctx,
dops->md_ciphertext, dops->md_mac,
dops->md_plaintext, rhndl);
break;
case KCF_OP_UPDATE:
kcf_secondctx = ((kcf_context_t *)
(ctx->cc_framework_private))->kc_secondctx;
if (kcf_secondctx != NULL) {
err = kcf_emulate_dual(pd, ctx, params);
break;
}
err = KCF_PROV_MAC_DECRYPT_UPDATE(pd, ctx,
dops->md_ciphertext, dops->md_plaintext, rhndl);
break;
case KCF_OP_FINAL:
kcf_secondctx = ((kcf_context_t *)
(ctx->cc_framework_private))->kc_secondctx;
if (kcf_secondctx != NULL) {
err = kcf_emulate_dual(pd, ctx, params);
break;
}
err = KCF_PROV_MAC_DECRYPT_FINAL(pd, ctx,
dops->md_mac, dops->md_plaintext, rhndl);
break;
case KCF_OP_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(
dops->md_framework_mac_mechtype,
pd, &dops->md_mac_mech);
KCF_SET_PROVIDER_MECHNUM(
dops->md_framework_decr_mechtype,
pd, &dops->md_decr_mech);
err = KCF_PROV_MAC_DECRYPT_ATOMIC(pd, dops->md_sid,
&dops->md_mac_mech, dops->md_mac_key,
&dops->md_decr_mech, dops->md_decr_key,
dops->md_ciphertext, dops->md_mac,
dops->md_plaintext,
dops->md_mac_templ, dops->md_decr_templ,
rhndl);
break;
case KCF_OP_MAC_VERIFY_DECRYPT_ATOMIC:
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(
dops->md_framework_mac_mechtype,
pd, &dops->md_mac_mech);
KCF_SET_PROVIDER_MECHNUM(
dops->md_framework_decr_mechtype,
pd, &dops->md_decr_mech);
err = KCF_PROV_MAC_VERIFY_DECRYPT_ATOMIC(pd,
dops->md_sid, &dops->md_mac_mech, dops->md_mac_key,
&dops->md_decr_mech, dops->md_decr_key,
dops->md_ciphertext, dops->md_mac,
dops->md_plaintext,
dops->md_mac_templ, dops->md_decr_templ,
rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_KEY: {
kcf_key_ops_params_t *kops = &params->rp_u.key_params;
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(kops->ko_framework_mechtype, pd,
&kops->ko_mech);
switch (optype) {
case KCF_OP_KEY_GENERATE:
err = KCF_PROV_KEY_GENERATE(pd, kops->ko_sid,
&kops->ko_mech,
kops->ko_key_template, kops->ko_key_attribute_count,
kops->ko_key_object_id_ptr, rhndl);
break;
case KCF_OP_KEY_GENERATE_PAIR:
err = KCF_PROV_KEY_GENERATE_PAIR(pd, kops->ko_sid,
&kops->ko_mech,
kops->ko_key_template, kops->ko_key_attribute_count,
kops->ko_private_key_template,
kops->ko_private_key_attribute_count,
kops->ko_key_object_id_ptr,
kops->ko_private_key_object_id_ptr, rhndl);
break;
case KCF_OP_KEY_WRAP:
err = KCF_PROV_KEY_WRAP(pd, kops->ko_sid,
&kops->ko_mech,
kops->ko_key, kops->ko_key_object_id_ptr,
kops->ko_wrapped_key, kops->ko_wrapped_key_len_ptr,
rhndl);
break;
case KCF_OP_KEY_UNWRAP:
err = KCF_PROV_KEY_UNWRAP(pd, kops->ko_sid,
&kops->ko_mech,
kops->ko_key, kops->ko_wrapped_key,
kops->ko_wrapped_key_len_ptr,
kops->ko_key_template, kops->ko_key_attribute_count,
kops->ko_key_object_id_ptr, rhndl);
break;
case KCF_OP_KEY_DERIVE:
err = KCF_PROV_KEY_DERIVE(pd, kops->ko_sid,
&kops->ko_mech,
kops->ko_key, kops->ko_key_template,
kops->ko_key_attribute_count,
kops->ko_key_object_id_ptr, rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_RANDOM: {
kcf_random_number_ops_params_t *rops =
&params->rp_u.random_number_params;
ASSERT(ctx == NULL);
switch (optype) {
case KCF_OP_RANDOM_SEED:
err = KCF_PROV_SEED_RANDOM(pd, rops->rn_sid,
rops->rn_buf, rops->rn_buflen, rops->rn_entropy_est,
rops->rn_flags, rhndl);
break;
case KCF_OP_RANDOM_GENERATE:
err = KCF_PROV_GENERATE_RANDOM(pd, rops->rn_sid,
rops->rn_buf, rops->rn_buflen, rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_SESSION: {
kcf_session_ops_params_t *sops = &params->rp_u.session_params;
ASSERT(ctx == NULL);
switch (optype) {
case KCF_OP_SESSION_OPEN:
/*
* so_pd may be a logical provider, in which case
* we need to check whether it has been removed.
*/
if (KCF_IS_PROV_REMOVED(sops->so_pd)) {
err = CRYPTO_DEVICE_ERROR;
break;
}
err = KCF_PROV_SESSION_OPEN(pd, sops->so_sid_ptr,
rhndl, sops->so_pd);
break;
case KCF_OP_SESSION_CLOSE:
/*
* so_pd may be a logical provider, in which case
* we need to check whether it has been removed.
*/
if (KCF_IS_PROV_REMOVED(sops->so_pd)) {
err = CRYPTO_DEVICE_ERROR;
break;
}
err = KCF_PROV_SESSION_CLOSE(pd, sops->so_sid,
rhndl, sops->so_pd);
break;
case KCF_OP_SESSION_LOGIN:
err = KCF_PROV_SESSION_LOGIN(pd, sops->so_sid,
sops->so_user_type, sops->so_pin,
sops->so_pin_len, rhndl);
break;
case KCF_OP_SESSION_LOGOUT:
err = KCF_PROV_SESSION_LOGOUT(pd, sops->so_sid, rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_OBJECT: {
kcf_object_ops_params_t *jops = &params->rp_u.object_params;
ASSERT(ctx == NULL);
switch (optype) {
case KCF_OP_OBJECT_CREATE:
err = KCF_PROV_OBJECT_CREATE(pd, jops->oo_sid,
jops->oo_template, jops->oo_attribute_count,
jops->oo_object_id_ptr, rhndl);
break;
case KCF_OP_OBJECT_COPY:
err = KCF_PROV_OBJECT_COPY(pd, jops->oo_sid,
jops->oo_object_id,
jops->oo_template, jops->oo_attribute_count,
jops->oo_object_id_ptr, rhndl);
break;
case KCF_OP_OBJECT_DESTROY:
err = KCF_PROV_OBJECT_DESTROY(pd, jops->oo_sid,
jops->oo_object_id, rhndl);
break;
case KCF_OP_OBJECT_GET_SIZE:
err = KCF_PROV_OBJECT_GET_SIZE(pd, jops->oo_sid,
jops->oo_object_id, jops->oo_object_size, rhndl);
break;
case KCF_OP_OBJECT_GET_ATTRIBUTE_VALUE:
err = KCF_PROV_OBJECT_GET_ATTRIBUTE_VALUE(pd,
jops->oo_sid, jops->oo_object_id,
jops->oo_template, jops->oo_attribute_count, rhndl);
break;
case KCF_OP_OBJECT_SET_ATTRIBUTE_VALUE:
err = KCF_PROV_OBJECT_SET_ATTRIBUTE_VALUE(pd,
jops->oo_sid, jops->oo_object_id,
jops->oo_template, jops->oo_attribute_count, rhndl);
break;
case KCF_OP_OBJECT_FIND_INIT:
err = KCF_PROV_OBJECT_FIND_INIT(pd, jops->oo_sid,
jops->oo_template, jops->oo_attribute_count,
jops->oo_find_init_pp_ptr, rhndl);
break;
case KCF_OP_OBJECT_FIND:
err = KCF_PROV_OBJECT_FIND(pd, jops->oo_find_pp,
jops->oo_object_id_ptr, jops->oo_max_object_count,
jops->oo_object_count_ptr, rhndl);
break;
case KCF_OP_OBJECT_FIND_FINAL:
err = KCF_PROV_OBJECT_FIND_FINAL(pd, jops->oo_find_pp,
rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_PROVMGMT: {
kcf_provmgmt_ops_params_t *pops = &params->rp_u.provmgmt_params;
ASSERT(ctx == NULL);
switch (optype) {
case KCF_OP_MGMT_EXTINFO:
/*
* po_pd may be a logical provider, in which case
* we need to check whether it has been removed.
*/
if (KCF_IS_PROV_REMOVED(pops->po_pd)) {
err = CRYPTO_DEVICE_ERROR;
break;
}
err = KCF_PROV_EXT_INFO(pd, pops->po_ext_info, rhndl,
pops->po_pd);
break;
case KCF_OP_MGMT_INITTOKEN:
err = KCF_PROV_INIT_TOKEN(pd, pops->po_pin,
pops->po_pin_len, pops->po_label, rhndl);
break;
case KCF_OP_MGMT_INITPIN:
err = KCF_PROV_INIT_PIN(pd, pops->po_sid, pops->po_pin,
pops->po_pin_len, rhndl);
break;
case KCF_OP_MGMT_SETPIN:
err = KCF_PROV_SET_PIN(pd, pops->po_sid,
pops->po_old_pin, pops->po_old_pin_len,
pops->po_pin, pops->po_pin_len, rhndl);
break;
default:
break;
}
break;
}
case KCF_OG_NOSTORE_KEY: {
kcf_key_ops_params_t *kops = &params->rp_u.key_params;
ASSERT(ctx == NULL);
KCF_SET_PROVIDER_MECHNUM(kops->ko_framework_mechtype, pd,
&kops->ko_mech);
switch (optype) {
case KCF_OP_KEY_GENERATE:
err = KCF_PROV_NOSTORE_KEY_GENERATE(pd, kops->ko_sid,
&kops->ko_mech, kops->ko_key_template,
kops->ko_key_attribute_count,
kops->ko_out_template1,
kops->ko_out_attribute_count1, rhndl);
break;
case KCF_OP_KEY_GENERATE_PAIR:
err = KCF_PROV_NOSTORE_KEY_GENERATE_PAIR(pd,
kops->ko_sid, &kops->ko_mech,
kops->ko_key_template, kops->ko_key_attribute_count,
kops->ko_private_key_template,
kops->ko_private_key_attribute_count,
kops->ko_out_template1,
kops->ko_out_attribute_count1,
kops->ko_out_template2,
kops->ko_out_attribute_count2,
rhndl);
break;
case KCF_OP_KEY_DERIVE:
err = KCF_PROV_NOSTORE_KEY_DERIVE(pd, kops->ko_sid,
&kops->ko_mech, kops->ko_key,
kops->ko_key_template,
kops->ko_key_attribute_count,
kops->ko_out_template1,
kops->ko_out_attribute_count1, rhndl);
break;
default:
break;
}
break;
}
default:
break;
} /* end of switch(params->rp_opgrp) */
KCF_PROV_INCRSTATS(pd, err);
return (err);
}
/*
* Emulate the call for a multipart dual ops with 2 single steps.
* This routine is always called in the context of a working thread
* running kcf_svc_do_run().
* The single steps are submitted in a pure synchronous way (blocking).
* When this routine returns, kcf_svc_do_run() will call kcf_aop_done()
* so the originating consumer's callback gets invoked. kcf_aop_done()
* takes care of freeing the operation context. So, this routine does
* not free the operation context.
*
* The provider descriptor is assumed held by the callers.
*/
static int
kcf_emulate_dual(kcf_provider_desc_t *pd, crypto_ctx_t *ctx,
kcf_req_params_t *params)
{
int err = CRYPTO_ARGUMENTS_BAD;
kcf_op_type_t optype;
size_t save_len;
off_t save_offset;
optype = params->rp_optype;
switch (params->rp_opgrp) {
case KCF_OG_ENCRYPT_MAC: {
kcf_encrypt_mac_ops_params_t *cmops =
&params->rp_u.encrypt_mac_params;
kcf_context_t *encr_kcf_ctx;
crypto_ctx_t *mac_ctx;
kcf_req_params_t encr_params;
encr_kcf_ctx = (kcf_context_t *)(ctx->cc_framework_private);
switch (optype) {
case KCF_OP_INIT: {
encr_kcf_ctx->kc_secondctx = NULL;
KCF_WRAP_ENCRYPT_OPS_PARAMS(&encr_params, KCF_OP_INIT,
pd->pd_sid, &cmops->em_encr_mech,
cmops->em_encr_key, NULL, NULL,
cmops->em_encr_templ);
err = kcf_submit_request(pd, ctx, NULL, &encr_params,
B_FALSE);
/* It can't be CRYPTO_QUEUED */
if (err != CRYPTO_SUCCESS) {
break;
}
err = crypto_mac_init(&cmops->em_mac_mech,
cmops->em_mac_key, cmops->em_mac_templ,
(crypto_context_t *)&mac_ctx, NULL);
if (err == CRYPTO_SUCCESS) {
encr_kcf_ctx->kc_secondctx = (kcf_context_t *)
mac_ctx->cc_framework_private;
KCF_CONTEXT_REFHOLD((kcf_context_t *)
mac_ctx->cc_framework_private);
}
break;
}
case KCF_OP_UPDATE: {
crypto_dual_data_t *ct = cmops->em_ciphertext;
crypto_data_t *pt = cmops->em_plaintext;
kcf_context_t *mac_kcf_ctx = encr_kcf_ctx->kc_secondctx;
crypto_ctx_t *mac_ctx = &mac_kcf_ctx->kc_glbl_ctx;
KCF_WRAP_ENCRYPT_OPS_PARAMS(&encr_params, KCF_OP_UPDATE,
pd->pd_sid, NULL, NULL, pt, (crypto_data_t *)ct,
NULL);
err = kcf_submit_request(pd, ctx, NULL, &encr_params,
B_FALSE);
/* It can't be CRYPTO_QUEUED */
if (err != CRYPTO_SUCCESS) {
break;
}
save_offset = ct->dd_offset1;
save_len = ct->dd_len1;
if (ct->dd_len2 == 0) {
/*
* The previous encrypt step was an
* accumulation only and didn't produce any
* partial output
*/
if (ct->dd_len1 == 0)
break;
} else {
ct->dd_offset1 = ct->dd_offset2;
ct->dd_len1 = ct->dd_len2;
}
err = crypto_mac_update((crypto_context_t)mac_ctx,
(crypto_data_t *)ct, NULL);
ct->dd_offset1 = save_offset;
ct->dd_len1 = save_len;
break;
}
case KCF_OP_FINAL: {
crypto_dual_data_t *ct = cmops->em_ciphertext;
crypto_data_t *mac = cmops->em_mac;
kcf_context_t *mac_kcf_ctx = encr_kcf_ctx->kc_secondctx;
crypto_ctx_t *mac_ctx = &mac_kcf_ctx->kc_glbl_ctx;
crypto_context_t mac_context = mac_ctx;
KCF_WRAP_ENCRYPT_OPS_PARAMS(&encr_params, KCF_OP_FINAL,
pd->pd_sid, NULL, NULL, NULL, (crypto_data_t *)ct,
NULL);
err = kcf_submit_request(pd, ctx, NULL, &encr_params,
B_FALSE);
/* It can't be CRYPTO_QUEUED */
if (err != CRYPTO_SUCCESS) {
crypto_cancel_ctx(mac_context);
break;
}
if (ct->dd_len2 > 0) {
save_offset = ct->dd_offset1;
save_len = ct->dd_len1;
ct->dd_offset1 = ct->dd_offset2;
ct->dd_len1 = ct->dd_len2;
err = crypto_mac_update(mac_context,
(crypto_data_t *)ct, NULL);
ct->dd_offset1 = save_offset;
ct->dd_len1 = save_len;
if (err != CRYPTO_SUCCESS) {
crypto_cancel_ctx(mac_context);
return (err);
}
}
/* and finally, collect the MAC */
err = crypto_mac_final(mac_context, mac, NULL);
break;
}
default:
break;
}
KCF_PROV_INCRSTATS(pd, err);
break;
}
case KCF_OG_MAC_DECRYPT: {
kcf_mac_decrypt_ops_params_t *mdops =
&params->rp_u.mac_decrypt_params;
kcf_context_t *decr_kcf_ctx;
crypto_ctx_t *mac_ctx;
kcf_req_params_t decr_params;
decr_kcf_ctx = (kcf_context_t *)(ctx->cc_framework_private);
switch (optype) {
case KCF_OP_INIT: {
decr_kcf_ctx->kc_secondctx = NULL;
err = crypto_mac_init(&mdops->md_mac_mech,
mdops->md_mac_key, mdops->md_mac_templ,
(crypto_context_t *)&mac_ctx, NULL);
/* It can't be CRYPTO_QUEUED */
if (err != CRYPTO_SUCCESS) {
break;
}
KCF_WRAP_DECRYPT_OPS_PARAMS(&decr_params, KCF_OP_INIT,
pd->pd_sid, &mdops->md_decr_mech,
mdops->md_decr_key, NULL, NULL,
mdops->md_decr_templ);
err = kcf_submit_request(pd, ctx, NULL, &decr_params,
B_FALSE);
/* It can't be CRYPTO_QUEUED */
if (err != CRYPTO_SUCCESS) {
crypto_cancel_ctx((crypto_context_t)mac_ctx);
break;
}
decr_kcf_ctx->kc_secondctx = (kcf_context_t *)
mac_ctx->cc_framework_private;
KCF_CONTEXT_REFHOLD((kcf_context_t *)
mac_ctx->cc_framework_private);
break;
default:
break;
}
case KCF_OP_UPDATE: {
crypto_dual_data_t *ct = mdops->md_ciphertext;
crypto_data_t *pt = mdops->md_plaintext;
kcf_context_t *mac_kcf_ctx = decr_kcf_ctx->kc_secondctx;
crypto_ctx_t *mac_ctx = &mac_kcf_ctx->kc_glbl_ctx;
err = crypto_mac_update((crypto_context_t)mac_ctx,
(crypto_data_t *)ct, NULL);
if (err != CRYPTO_SUCCESS)
break;
save_offset = ct->dd_offset1;
save_len = ct->dd_len1;
/* zero ct->dd_len2 means decrypt everything */
if (ct->dd_len2 > 0) {
ct->dd_offset1 = ct->dd_offset2;
ct->dd_len1 = ct->dd_len2;
}
err = crypto_decrypt_update((crypto_context_t)ctx,
(crypto_data_t *)ct, pt, NULL);
ct->dd_offset1 = save_offset;
ct->dd_len1 = save_len;
break;
}
case KCF_OP_FINAL: {
crypto_data_t *pt = mdops->md_plaintext;
crypto_data_t *mac = mdops->md_mac;
kcf_context_t *mac_kcf_ctx = decr_kcf_ctx->kc_secondctx;
crypto_ctx_t *mac_ctx = &mac_kcf_ctx->kc_glbl_ctx;
err = crypto_mac_final((crypto_context_t)mac_ctx,
mac, NULL);
if (err != CRYPTO_SUCCESS) {
crypto_cancel_ctx(ctx);
break;
}
/* Get the last chunk of plaintext */
KCF_CONTEXT_REFHOLD(decr_kcf_ctx);
err = crypto_decrypt_final((crypto_context_t)ctx, pt,
NULL);
break;
}
}
break;
}
default:
break;
} /* end of switch(params->rp_opgrp) */
return (err);
}
diff --git a/module/icp/core/kcf_mech_tabs.c b/module/icp/core/kcf_mech_tabs.c
index 461b7946c2c8..60055e78af68 100644
--- a/module/icp/core/kcf_mech_tabs.c
+++ b/module/icp/core/kcf_mech_tabs.c
@@ -1,785 +1,784 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
#include <sys/crypto/api.h>
#include <sys/crypto/impl.h>
#include <sys/modhash.h>
/* Cryptographic mechanisms tables and their access functions */
/*
* Internal numbers assigned to mechanisms are coded as follows:
*
* +----------------+----------------+
* | mech. class | mech. index |
* <--- 32-bits --->+<--- 32-bits --->
*
* the mech_class identifies the table the mechanism belongs to.
* mech_index is the index for that mechanism in the table.
* A mechanism belongs to exactly 1 table.
* The tables are:
* . digest_mechs_tab[] for the msg digest mechs.
* . cipher_mechs_tab[] for encrypt/decrypt and wrap/unwrap mechs.
* . mac_mechs_tab[] for MAC mechs.
* . sign_mechs_tab[] for sign & verify mechs.
* . keyops_mechs_tab[] for key/key pair generation, and key derivation.
* . misc_mechs_tab[] for mechs that don't belong to any of the above.
*
* There are no holes in the tables.
*/
/*
* Locking conventions:
* --------------------
* A global mutex, kcf_mech_tabs_lock, serializes writes to the
* mechanism table via kcf_create_mech_entry().
*
* A mutex is associated with every entry of the tables.
* The mutex is acquired whenever the entry is accessed for
* 1) retrieving the mech_id (comparing the mech name)
* 2) finding a provider for an xxx_init() or atomic operation.
* 3) altering the mechs entry to add or remove a provider.
*
* In 2), after a provider is chosen, its prov_desc is held and the
* entry's mutex must be dropped. The provider's working function (SPI) is
* called outside the mech_entry's mutex.
*
* The number of providers for a particular mechanism is not expected to be
* long enough to justify the cost of using rwlocks, so the per-mechanism
* entry mutex won't be very *hot*.
*
* When both kcf_mech_tabs_lock and a mech_entry mutex need to be held,
* kcf_mech_tabs_lock must always be acquired first.
*
*/
/* Mechanisms tables */
/* RFE 4687834 Will deal with the extensibility of these tables later */
-kcf_mech_entry_t kcf_digest_mechs_tab[KCF_MAXDIGEST];
-kcf_mech_entry_t kcf_cipher_mechs_tab[KCF_MAXCIPHER];
-kcf_mech_entry_t kcf_mac_mechs_tab[KCF_MAXMAC];
-kcf_mech_entry_t kcf_sign_mechs_tab[KCF_MAXSIGN];
-kcf_mech_entry_t kcf_keyops_mechs_tab[KCF_MAXKEYOPS];
-kcf_mech_entry_t kcf_misc_mechs_tab[KCF_MAXMISC];
-
-kcf_mech_entry_tab_t kcf_mech_tabs_tab[KCF_LAST_OPSCLASS + 1] = {
+static kcf_mech_entry_t kcf_digest_mechs_tab[KCF_MAXDIGEST];
+static kcf_mech_entry_t kcf_cipher_mechs_tab[KCF_MAXCIPHER];
+static kcf_mech_entry_t kcf_mac_mechs_tab[KCF_MAXMAC];
+static kcf_mech_entry_t kcf_sign_mechs_tab[KCF_MAXSIGN];
+static kcf_mech_entry_t kcf_keyops_mechs_tab[KCF_MAXKEYOPS];
+static kcf_mech_entry_t kcf_misc_mechs_tab[KCF_MAXMISC];
+
+const kcf_mech_entry_tab_t kcf_mech_tabs_tab[KCF_LAST_OPSCLASS + 1] = {
{0, NULL}, /* No class zero */
{KCF_MAXDIGEST, kcf_digest_mechs_tab},
{KCF_MAXCIPHER, kcf_cipher_mechs_tab},
{KCF_MAXMAC, kcf_mac_mechs_tab},
{KCF_MAXSIGN, kcf_sign_mechs_tab},
{KCF_MAXKEYOPS, kcf_keyops_mechs_tab},
{KCF_MAXMISC, kcf_misc_mechs_tab}
};
/*
* Per-algorithm internal thresholds for the minimum input size of before
* offloading to hardware provider.
* Dispatching a crypto operation to a hardware provider entails paying the
* cost of an additional context switch. Measurements with Sun Accelerator 4000
* shows that 512-byte jobs or smaller are better handled in software.
* There is room for refinement here.
*
*/
-int kcf_md5_threshold = 512;
-int kcf_sha1_threshold = 512;
-int kcf_des_threshold = 512;
-int kcf_des3_threshold = 512;
-int kcf_aes_threshold = 512;
-int kcf_bf_threshold = 512;
-int kcf_rc4_threshold = 512;
-
-kmutex_t kcf_mech_tabs_lock;
+static const int kcf_md5_threshold = 512;
+static const int kcf_sha1_threshold = 512;
+static const int kcf_des_threshold = 512;
+static const int kcf_des3_threshold = 512;
+static const int kcf_aes_threshold = 512;
+static const int kcf_bf_threshold = 512;
+static const int kcf_rc4_threshold = 512;
+
+static kmutex_t kcf_mech_tabs_lock;
static uint32_t kcf_gen_swprov = 0;
-int kcf_mech_hash_size = 256;
-mod_hash_t *kcf_mech_hash; /* mech name to id hash */
+static const int kcf_mech_hash_size = 256;
+static mod_hash_t *kcf_mech_hash; /* mech name to id hash */
static crypto_mech_type_t
-kcf_mech_hash_find(char *mechname)
+kcf_mech_hash_find(const char *mechname)
{
mod_hash_val_t hv;
crypto_mech_type_t mt;
mt = CRYPTO_MECH_INVALID;
if (mod_hash_find(kcf_mech_hash, (mod_hash_key_t)mechname, &hv) == 0) {
mt = *(crypto_mech_type_t *)hv;
ASSERT(mt != CRYPTO_MECH_INVALID);
}
return (mt);
}
void
kcf_destroy_mech_tabs(void)
{
int i, max;
kcf_ops_class_t class;
kcf_mech_entry_t *me_tab;
if (kcf_mech_hash)
mod_hash_destroy_hash(kcf_mech_hash);
mutex_destroy(&kcf_mech_tabs_lock);
for (class = KCF_FIRST_OPSCLASS; class <= KCF_LAST_OPSCLASS; class++) {
max = kcf_mech_tabs_tab[class].met_size;
me_tab = kcf_mech_tabs_tab[class].met_tab;
for (i = 0; i < max; i++)
mutex_destroy(&(me_tab[i].me_mutex));
}
}
/*
* kcf_init_mech_tabs()
*
* Called by the misc/kcf's _init() routine to initialize the tables
* of mech_entry's.
*/
void
kcf_init_mech_tabs(void)
{
- int i, max;
kcf_ops_class_t class;
kcf_mech_entry_t *me_tab;
/* Initializes the mutex locks. */
mutex_init(&kcf_mech_tabs_lock, NULL, MUTEX_DEFAULT, NULL);
/* Then the pre-defined mechanism entries */
/* Two digests */
(void) strncpy(kcf_digest_mechs_tab[0].me_name, SUN_CKM_MD5,
CRYPTO_MAX_MECH_NAME);
kcf_digest_mechs_tab[0].me_threshold = kcf_md5_threshold;
(void) strncpy(kcf_digest_mechs_tab[1].me_name, SUN_CKM_SHA1,
CRYPTO_MAX_MECH_NAME);
kcf_digest_mechs_tab[1].me_threshold = kcf_sha1_threshold;
/* The symmetric ciphers in various modes */
(void) strncpy(kcf_cipher_mechs_tab[0].me_name, SUN_CKM_DES_CBC,
CRYPTO_MAX_MECH_NAME);
kcf_cipher_mechs_tab[0].me_threshold = kcf_des_threshold;
(void) strncpy(kcf_cipher_mechs_tab[1].me_name, SUN_CKM_DES3_CBC,
CRYPTO_MAX_MECH_NAME);
kcf_cipher_mechs_tab[1].me_threshold = kcf_des3_threshold;
(void) strncpy(kcf_cipher_mechs_tab[2].me_name, SUN_CKM_DES_ECB,
CRYPTO_MAX_MECH_NAME);
kcf_cipher_mechs_tab[2].me_threshold = kcf_des_threshold;
(void) strncpy(kcf_cipher_mechs_tab[3].me_name, SUN_CKM_DES3_ECB,
CRYPTO_MAX_MECH_NAME);
kcf_cipher_mechs_tab[3].me_threshold = kcf_des3_threshold;
(void) strncpy(kcf_cipher_mechs_tab[4].me_name, SUN_CKM_BLOWFISH_CBC,
CRYPTO_MAX_MECH_NAME);
kcf_cipher_mechs_tab[4].me_threshold = kcf_bf_threshold;
(void) strncpy(kcf_cipher_mechs_tab[5].me_name, SUN_CKM_BLOWFISH_ECB,
CRYPTO_MAX_MECH_NAME);
kcf_cipher_mechs_tab[5].me_threshold = kcf_bf_threshold;
(void) strncpy(kcf_cipher_mechs_tab[6].me_name, SUN_CKM_AES_CBC,
CRYPTO_MAX_MECH_NAME);
kcf_cipher_mechs_tab[6].me_threshold = kcf_aes_threshold;
(void) strncpy(kcf_cipher_mechs_tab[7].me_name, SUN_CKM_AES_ECB,
CRYPTO_MAX_MECH_NAME);
kcf_cipher_mechs_tab[7].me_threshold = kcf_aes_threshold;
(void) strncpy(kcf_cipher_mechs_tab[8].me_name, SUN_CKM_RC4,
CRYPTO_MAX_MECH_NAME);
kcf_cipher_mechs_tab[8].me_threshold = kcf_rc4_threshold;
/* 4 HMACs */
(void) strncpy(kcf_mac_mechs_tab[0].me_name, SUN_CKM_MD5_HMAC,
CRYPTO_MAX_MECH_NAME);
kcf_mac_mechs_tab[0].me_threshold = kcf_md5_threshold;
(void) strncpy(kcf_mac_mechs_tab[1].me_name, SUN_CKM_MD5_HMAC_GENERAL,
CRYPTO_MAX_MECH_NAME);
kcf_mac_mechs_tab[1].me_threshold = kcf_md5_threshold;
(void) strncpy(kcf_mac_mechs_tab[2].me_name, SUN_CKM_SHA1_HMAC,
CRYPTO_MAX_MECH_NAME);
kcf_mac_mechs_tab[2].me_threshold = kcf_sha1_threshold;
(void) strncpy(kcf_mac_mechs_tab[3].me_name, SUN_CKM_SHA1_HMAC_GENERAL,
CRYPTO_MAX_MECH_NAME);
kcf_mac_mechs_tab[3].me_threshold = kcf_sha1_threshold;
/* 1 random number generation pseudo mechanism */
(void) strncpy(kcf_misc_mechs_tab[0].me_name, SUN_RANDOM,
CRYPTO_MAX_MECH_NAME);
kcf_mech_hash = mod_hash_create_strhash_nodtr("kcf mech2id hash",
kcf_mech_hash_size, mod_hash_null_valdtor);
for (class = KCF_FIRST_OPSCLASS; class <= KCF_LAST_OPSCLASS; class++) {
- max = kcf_mech_tabs_tab[class].met_size;
+ int max = kcf_mech_tabs_tab[class].met_size;
me_tab = kcf_mech_tabs_tab[class].met_tab;
- for (i = 0; i < max; i++) {
+ for (int i = 0; i < max; i++) {
mutex_init(&(me_tab[i].me_mutex), NULL,
MUTEX_DEFAULT, NULL);
if (me_tab[i].me_name[0] != 0) {
me_tab[i].me_mechid = KCF_MECHID(class, i);
(void) mod_hash_insert(kcf_mech_hash,
(mod_hash_key_t)me_tab[i].me_name,
(mod_hash_val_t)&(me_tab[i].me_mechid));
}
}
}
}
/*
* kcf_create_mech_entry()
*
* Arguments:
* . The class of mechanism.
* . the name of the new mechanism.
*
* Description:
* Creates a new mech_entry for a mechanism not yet known to the
* framework.
* This routine is called by kcf_add_mech_provider, which is
* in turn invoked for each mechanism supported by a provider.
* The'class' argument depends on the crypto_func_group_t bitmask
* in the registering provider's mech_info struct for this mechanism.
* When there is ambiguity in the mapping between the crypto_func_group_t
* and a class (dual ops, ...) the KCF_MISC_CLASS should be used.
*
* Context:
* User context only.
*
* Returns:
* KCF_INVALID_MECH_CLASS or KCF_INVALID_MECH_NAME if the class or
* the mechname is bogus.
* KCF_MECH_TAB_FULL when there is no room left in the mech. tabs.
* KCF_SUCCESS otherwise.
*/
static int
kcf_create_mech_entry(kcf_ops_class_t class, char *mechname)
{
crypto_mech_type_t mt;
kcf_mech_entry_t *me_tab;
int i = 0, size;
if ((class < KCF_FIRST_OPSCLASS) || (class > KCF_LAST_OPSCLASS))
return (KCF_INVALID_MECH_CLASS);
if ((mechname == NULL) || (mechname[0] == 0))
return (KCF_INVALID_MECH_NAME);
/*
* First check if the mechanism is already in one of the tables.
* The mech_entry could be in another class.
*/
mutex_enter(&kcf_mech_tabs_lock);
mt = kcf_mech_hash_find(mechname);
if (mt != CRYPTO_MECH_INVALID) {
/* Nothing to do, regardless the suggested class. */
mutex_exit(&kcf_mech_tabs_lock);
return (KCF_SUCCESS);
}
/* Now take the next unused mech entry in the class's tab */
me_tab = kcf_mech_tabs_tab[class].met_tab;
size = kcf_mech_tabs_tab[class].met_size;
while (i < size) {
mutex_enter(&(me_tab[i].me_mutex));
if (me_tab[i].me_name[0] == 0) {
/* Found an empty spot */
(void) strlcpy(me_tab[i].me_name, mechname,
CRYPTO_MAX_MECH_NAME);
me_tab[i].me_name[CRYPTO_MAX_MECH_NAME-1] = '\0';
me_tab[i].me_mechid = KCF_MECHID(class, i);
/*
* No a-priori information about the new mechanism, so
* the threshold is set to zero.
*/
me_tab[i].me_threshold = 0;
mutex_exit(&(me_tab[i].me_mutex));
/* Add the new mechanism to the hash table */
(void) mod_hash_insert(kcf_mech_hash,
(mod_hash_key_t)me_tab[i].me_name,
(mod_hash_val_t)&(me_tab[i].me_mechid));
break;
}
mutex_exit(&(me_tab[i].me_mutex));
i++;
}
mutex_exit(&kcf_mech_tabs_lock);
if (i == size) {
return (KCF_MECH_TAB_FULL);
}
return (KCF_SUCCESS);
}
/*
* kcf_add_mech_provider()
*
* Arguments:
* . An index in to the provider mechanism array
* . A pointer to the provider descriptor
* . A storage for the kcf_prov_mech_desc_t the entry was added at.
*
* Description:
* Adds a new provider of a mechanism to the mechanism's mech_entry
* chain.
*
* Context:
* User context only.
*
* Returns
* KCF_SUCCESS on success
* KCF_MECH_TAB_FULL otherwise.
*/
int
kcf_add_mech_provider(short mech_indx,
kcf_provider_desc_t *prov_desc, kcf_prov_mech_desc_t **pmdpp)
{
int error;
kcf_mech_entry_t *mech_entry = NULL;
crypto_mech_info_t *mech_info;
crypto_mech_type_t kcf_mech_type, mt;
kcf_prov_mech_desc_t *prov_mech, *prov_mech2;
crypto_func_group_t simple_fg_mask, dual_fg_mask;
crypto_mech_info_t *dmi;
crypto_mech_info_list_t *mil, *mil2;
kcf_mech_entry_t *me;
int i;
ASSERT(prov_desc->pd_prov_type != CRYPTO_LOGICAL_PROVIDER);
mech_info = &prov_desc->pd_mechanisms[mech_indx];
/*
* A mechanism belongs to exactly one mechanism table.
* Find the class corresponding to the function group flag of
* the mechanism.
*/
kcf_mech_type = kcf_mech_hash_find(mech_info->cm_mech_name);
if (kcf_mech_type == CRYPTO_MECH_INVALID) {
crypto_func_group_t fg = mech_info->cm_func_group_mask;
kcf_ops_class_t class;
if (fg & CRYPTO_FG_DIGEST || fg & CRYPTO_FG_DIGEST_ATOMIC)
class = KCF_DIGEST_CLASS;
else if (fg & CRYPTO_FG_ENCRYPT || fg & CRYPTO_FG_DECRYPT ||
fg & CRYPTO_FG_ENCRYPT_ATOMIC ||
fg & CRYPTO_FG_DECRYPT_ATOMIC)
class = KCF_CIPHER_CLASS;
else if (fg & CRYPTO_FG_MAC || fg & CRYPTO_FG_MAC_ATOMIC)
class = KCF_MAC_CLASS;
else if (fg & CRYPTO_FG_SIGN || fg & CRYPTO_FG_VERIFY ||
fg & CRYPTO_FG_SIGN_ATOMIC ||
fg & CRYPTO_FG_VERIFY_ATOMIC ||
fg & CRYPTO_FG_SIGN_RECOVER ||
fg & CRYPTO_FG_VERIFY_RECOVER)
class = KCF_SIGN_CLASS;
else if (fg & CRYPTO_FG_GENERATE ||
fg & CRYPTO_FG_GENERATE_KEY_PAIR ||
fg & CRYPTO_FG_WRAP || fg & CRYPTO_FG_UNWRAP ||
fg & CRYPTO_FG_DERIVE)
class = KCF_KEYOPS_CLASS;
else
class = KCF_MISC_CLASS;
/*
* Attempt to create a new mech_entry for the specified
* mechanism. kcf_create_mech_entry() can handle the case
* where such an entry already exists.
*/
if ((error = kcf_create_mech_entry(class,
mech_info->cm_mech_name)) != KCF_SUCCESS) {
return (error);
}
/* get the KCF mech type that was assigned to the mechanism */
kcf_mech_type = kcf_mech_hash_find(mech_info->cm_mech_name);
ASSERT(kcf_mech_type != CRYPTO_MECH_INVALID);
}
error = kcf_get_mech_entry(kcf_mech_type, &mech_entry);
ASSERT(error == KCF_SUCCESS);
/* allocate and initialize new kcf_prov_mech_desc */
prov_mech = kmem_zalloc(sizeof (kcf_prov_mech_desc_t), KM_SLEEP);
bcopy(mech_info, &prov_mech->pm_mech_info, sizeof (crypto_mech_info_t));
prov_mech->pm_prov_desc = prov_desc;
prov_desc->pd_mech_indx[KCF_MECH2CLASS(kcf_mech_type)]
[KCF_MECH2INDEX(kcf_mech_type)] = mech_indx;
KCF_PROV_REFHOLD(prov_desc);
KCF_PROV_IREFHOLD(prov_desc);
dual_fg_mask = mech_info->cm_func_group_mask & CRYPTO_FG_DUAL_MASK;
if (dual_fg_mask == ((crypto_func_group_t)0))
goto add_entry;
simple_fg_mask = (mech_info->cm_func_group_mask &
CRYPTO_FG_SIMPLEOP_MASK) | CRYPTO_FG_RANDOM;
for (i = 0; i < prov_desc->pd_mech_list_count; i++) {
dmi = &prov_desc->pd_mechanisms[i];
/* skip self */
if (dmi->cm_mech_number == mech_info->cm_mech_number)
continue;
/* skip if not a dual operation mechanism */
if (!(dmi->cm_func_group_mask & dual_fg_mask) ||
(dmi->cm_func_group_mask & simple_fg_mask))
continue;
mt = kcf_mech_hash_find(dmi->cm_mech_name);
if (mt == CRYPTO_MECH_INVALID)
continue;
if (kcf_get_mech_entry(mt, &me) != KCF_SUCCESS)
continue;
mil = kmem_zalloc(sizeof (*mil), KM_SLEEP);
mil2 = kmem_zalloc(sizeof (*mil2), KM_SLEEP);
/*
* Ignore hard-coded entries in the mech table
* if the provider hasn't registered.
*/
mutex_enter(&me->me_mutex);
if (me->me_hw_prov_chain == NULL && me->me_sw_prov == NULL) {
mutex_exit(&me->me_mutex);
kmem_free(mil, sizeof (*mil));
kmem_free(mil2, sizeof (*mil2));
continue;
}
/*
* Add other dual mechanisms that have registered
* with the framework to this mechanism's
* cross-reference list.
*/
mil->ml_mech_info = *dmi; /* struct assignment */
mil->ml_kcf_mechid = mt;
/* add to head of list */
mil->ml_next = prov_mech->pm_mi_list;
prov_mech->pm_mi_list = mil;
if (prov_desc->pd_prov_type == CRYPTO_HW_PROVIDER)
prov_mech2 = me->me_hw_prov_chain;
else
prov_mech2 = me->me_sw_prov;
if (prov_mech2 == NULL) {
kmem_free(mil2, sizeof (*mil2));
mutex_exit(&me->me_mutex);
continue;
}
/*
* Update all other cross-reference lists by
* adding this new mechanism.
*/
while (prov_mech2 != NULL) {
if (prov_mech2->pm_prov_desc == prov_desc) {
/* struct assignment */
mil2->ml_mech_info = *mech_info;
mil2->ml_kcf_mechid = kcf_mech_type;
/* add to head of list */
mil2->ml_next = prov_mech2->pm_mi_list;
prov_mech2->pm_mi_list = mil2;
break;
}
prov_mech2 = prov_mech2->pm_next;
}
if (prov_mech2 == NULL)
kmem_free(mil2, sizeof (*mil2));
mutex_exit(&me->me_mutex);
}
add_entry:
/*
* Add new kcf_prov_mech_desc at the front of HW providers
* chain.
*/
switch (prov_desc->pd_prov_type) {
case CRYPTO_HW_PROVIDER:
mutex_enter(&mech_entry->me_mutex);
prov_mech->pm_me = mech_entry;
prov_mech->pm_next = mech_entry->me_hw_prov_chain;
mech_entry->me_hw_prov_chain = prov_mech;
mech_entry->me_num_hwprov++;
mutex_exit(&mech_entry->me_mutex);
break;
case CRYPTO_SW_PROVIDER:
mutex_enter(&mech_entry->me_mutex);
if (mech_entry->me_sw_prov != NULL) {
/*
* There is already a SW provider for this mechanism.
* Since we allow only one SW provider per mechanism,
* report this condition.
*/
cmn_err(CE_WARN, "The cryptographic software provider "
"\"%s\" will not be used for %s. The provider "
"\"%s\" will be used for this mechanism "
"instead.", prov_desc->pd_description,
mech_info->cm_mech_name,
mech_entry->me_sw_prov->pm_prov_desc->
pd_description);
KCF_PROV_REFRELE(prov_desc);
kmem_free(prov_mech, sizeof (kcf_prov_mech_desc_t));
prov_mech = NULL;
} else {
/*
* Set the provider as the software provider for
* this mechanism.
*/
mech_entry->me_sw_prov = prov_mech;
/* We'll wrap around after 4 billion registrations! */
mech_entry->me_gen_swprov = kcf_gen_swprov++;
}
mutex_exit(&mech_entry->me_mutex);
break;
default:
break;
}
*pmdpp = prov_mech;
return (KCF_SUCCESS);
}
/*
* kcf_remove_mech_provider()
*
* Arguments:
* . mech_name: the name of the mechanism.
* . prov_desc: The provider descriptor
*
* Description:
* Removes a provider from chain of provider descriptors.
* The provider is made unavailable to kernel consumers for the specified
* mechanism.
*
* Context:
* User context only.
*/
void
kcf_remove_mech_provider(char *mech_name, kcf_provider_desc_t *prov_desc)
{
crypto_mech_type_t mech_type;
kcf_prov_mech_desc_t *prov_mech = NULL, *prov_chain;
kcf_prov_mech_desc_t **prev_entry_next;
kcf_mech_entry_t *mech_entry;
crypto_mech_info_list_t *mil, *mil2, *next, **prev_next;
ASSERT(prov_desc->pd_prov_type != CRYPTO_LOGICAL_PROVIDER);
/* get the KCF mech type that was assigned to the mechanism */
if ((mech_type = kcf_mech_hash_find(mech_name)) ==
CRYPTO_MECH_INVALID) {
/*
* Provider was not allowed for this mech due to policy or
* configuration.
*/
return;
}
/* get a ptr to the mech_entry that was created */
if (kcf_get_mech_entry(mech_type, &mech_entry) != KCF_SUCCESS) {
/*
* Provider was not allowed for this mech due to policy or
* configuration.
*/
return;
}
mutex_enter(&mech_entry->me_mutex);
switch (prov_desc->pd_prov_type) {
case CRYPTO_HW_PROVIDER:
/* find the provider in the mech_entry chain */
prev_entry_next = &mech_entry->me_hw_prov_chain;
prov_mech = mech_entry->me_hw_prov_chain;
while (prov_mech != NULL &&
prov_mech->pm_prov_desc != prov_desc) {
prev_entry_next = &prov_mech->pm_next;
prov_mech = prov_mech->pm_next;
}
if (prov_mech == NULL) {
/* entry not found, simply return */
mutex_exit(&mech_entry->me_mutex);
return;
}
/* remove provider entry from mech_entry chain */
*prev_entry_next = prov_mech->pm_next;
ASSERT(mech_entry->me_num_hwprov > 0);
mech_entry->me_num_hwprov--;
break;
case CRYPTO_SW_PROVIDER:
if (mech_entry->me_sw_prov == NULL ||
mech_entry->me_sw_prov->pm_prov_desc != prov_desc) {
/* not the software provider for this mechanism */
mutex_exit(&mech_entry->me_mutex);
return;
}
prov_mech = mech_entry->me_sw_prov;
mech_entry->me_sw_prov = NULL;
break;
default:
/* unexpected crypto_provider_type_t */
mutex_exit(&mech_entry->me_mutex);
return;
}
mutex_exit(&mech_entry->me_mutex);
/* Free the dual ops cross-reference lists */
mil = prov_mech->pm_mi_list;
while (mil != NULL) {
next = mil->ml_next;
if (kcf_get_mech_entry(mil->ml_kcf_mechid,
&mech_entry) != KCF_SUCCESS) {
mil = next;
continue;
}
mutex_enter(&mech_entry->me_mutex);
if (prov_desc->pd_prov_type == CRYPTO_HW_PROVIDER)
prov_chain = mech_entry->me_hw_prov_chain;
else
prov_chain = mech_entry->me_sw_prov;
while (prov_chain != NULL) {
if (prov_chain->pm_prov_desc == prov_desc) {
prev_next = &prov_chain->pm_mi_list;
mil2 = prov_chain->pm_mi_list;
while (mil2 != NULL &&
mil2->ml_kcf_mechid != mech_type) {
prev_next = &mil2->ml_next;
mil2 = mil2->ml_next;
}
if (mil2 != NULL) {
*prev_next = mil2->ml_next;
kmem_free(mil2, sizeof (*mil2));
}
break;
}
prov_chain = prov_chain->pm_next;
}
mutex_exit(&mech_entry->me_mutex);
kmem_free(mil, sizeof (crypto_mech_info_list_t));
mil = next;
}
/* free entry */
KCF_PROV_REFRELE(prov_mech->pm_prov_desc);
KCF_PROV_IREFRELE(prov_mech->pm_prov_desc);
kmem_free(prov_mech, sizeof (kcf_prov_mech_desc_t));
}
/*
* kcf_get_mech_entry()
*
* Arguments:
* . The framework mechanism type
* . Storage for the mechanism entry
*
* Description:
* Retrieves the mechanism entry for the mech.
*
* Context:
* User and interrupt contexts.
*
* Returns:
* KCF_MECHANISM_XXX appropriate error code.
* KCF_SUCCESS otherwise.
*/
int
kcf_get_mech_entry(crypto_mech_type_t mech_type, kcf_mech_entry_t **mep)
{
kcf_ops_class_t class;
int index;
- kcf_mech_entry_tab_t *me_tab;
+ const kcf_mech_entry_tab_t *me_tab;
ASSERT(mep != NULL);
class = KCF_MECH2CLASS(mech_type);
if ((class < KCF_FIRST_OPSCLASS) || (class > KCF_LAST_OPSCLASS)) {
/* the caller won't need to know it's an invalid class */
return (KCF_INVALID_MECH_NUMBER);
}
me_tab = &kcf_mech_tabs_tab[class];
index = KCF_MECH2INDEX(mech_type);
if ((index < 0) || (index >= me_tab->met_size)) {
return (KCF_INVALID_MECH_NUMBER);
}
*mep = &((me_tab->met_tab)[index]);
return (KCF_SUCCESS);
}
/* CURRENTLY UNSUPPORTED: attempting to load the module if it isn't found */
/*
* Lookup the hash table for an entry that matches the mechname.
* If there are no hardware or software providers for the mechanism,
* but there is an unloaded software provider, this routine will attempt
* to load it.
*/
crypto_mech_type_t
-crypto_mech2id_common(char *mechname, boolean_t load_module)
+crypto_mech2id_common(const char *mechname, boolean_t load_module)
{
(void) load_module;
return (kcf_mech_hash_find(mechname));
}
diff --git a/module/icp/core/kcf_prov_tabs.c b/module/icp/core/kcf_prov_tabs.c
index 9d303d022517..664e96da9c28 100644
--- a/module/icp/core/kcf_prov_tabs.c
+++ b/module/icp/core/kcf_prov_tabs.c
@@ -1,645 +1,646 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* This file is part of the core Kernel Cryptographic Framework.
* It implements the management of tables of Providers. Entries to
* added and removed when cryptographic providers register with
* and unregister from the framework, respectively. The KCF scheduler
* and ioctl pseudo driver call this function to obtain the list
* of available providers.
*
* The provider table is indexed by crypto_provider_id_t. Each
* element of the table contains a pointer to a provider descriptor,
* or NULL if the entry is free.
*
* This file also implements helper functions to allocate and free
* provider descriptors.
*/
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
#include <sys/crypto/impl.h>
#include <sys/crypto/sched_impl.h>
#include <sys/crypto/spi.h>
#define KCF_MAX_PROVIDERS 512 /* max number of providers */
/*
* Prov_tab is an array of providers which is updated when
* a crypto provider registers with kcf. The provider calls the
* SPI routine, crypto_register_provider(), which in turn calls
* kcf_prov_tab_add_provider().
*
* A provider unregisters by calling crypto_unregister_provider()
* which triggers the removal of the prov_tab entry.
* It also calls kcf_remove_mech_provider().
*
* prov_tab entries are not updated from kcf.conf or by cryptoadm(1M).
*/
static kcf_provider_desc_t **prov_tab = NULL;
static kmutex_t prov_tab_mutex; /* ensure exclusive access to the table */
static uint_t prov_tab_num = 0; /* number of providers in table */
static uint_t prov_tab_max = KCF_MAX_PROVIDERS;
void
kcf_prov_tab_destroy(void)
{
mutex_destroy(&prov_tab_mutex);
if (prov_tab)
kmem_free(prov_tab, prov_tab_max *
sizeof (kcf_provider_desc_t *));
}
/*
* Initialize a mutex and the KCF providers table, prov_tab.
* The providers table is dynamically allocated with prov_tab_max entries.
* Called from kcf module _init().
*/
void
kcf_prov_tab_init(void)
{
mutex_init(&prov_tab_mutex, NULL, MUTEX_DEFAULT, NULL);
prov_tab = kmem_zalloc(prov_tab_max * sizeof (kcf_provider_desc_t *),
KM_SLEEP);
}
/*
* Add a provider to the provider table. If no free entry can be found
* for the new provider, returns CRYPTO_HOST_MEMORY. Otherwise, add
* the provider to the table, initialize the pd_prov_id field
* of the specified provider descriptor to the index in that table,
* and return CRYPTO_SUCCESS. Note that a REFHOLD is done on the
* provider when pointed to by a table entry.
*/
int
kcf_prov_tab_add_provider(kcf_provider_desc_t *prov_desc)
{
uint_t i;
ASSERT(prov_tab != NULL);
mutex_enter(&prov_tab_mutex);
/* find free slot in providers table */
for (i = 1; i < KCF_MAX_PROVIDERS && prov_tab[i] != NULL; i++)
;
if (i == KCF_MAX_PROVIDERS) {
/* ran out of providers entries */
mutex_exit(&prov_tab_mutex);
cmn_err(CE_WARN, "out of providers entries");
return (CRYPTO_HOST_MEMORY);
}
/* initialize entry */
prov_tab[i] = prov_desc;
KCF_PROV_REFHOLD(prov_desc);
KCF_PROV_IREFHOLD(prov_desc);
prov_tab_num++;
mutex_exit(&prov_tab_mutex);
/* update provider descriptor */
prov_desc->pd_prov_id = i;
/*
* The KCF-private provider handle is defined as the internal
* provider id.
*/
prov_desc->pd_kcf_prov_handle =
(crypto_kcf_provider_handle_t)prov_desc->pd_prov_id;
return (CRYPTO_SUCCESS);
}
/*
* Remove the provider specified by its id. A REFRELE is done on the
* corresponding provider descriptor before this function returns.
* Returns CRYPTO_UNKNOWN_PROVIDER if the provider id is not valid.
*/
int
kcf_prov_tab_rem_provider(crypto_provider_id_t prov_id)
{
kcf_provider_desc_t *prov_desc;
ASSERT(prov_tab != NULL);
ASSERT(prov_tab_num >= 0);
/*
* Validate provider id, since it can be specified by a 3rd-party
* provider.
*/
mutex_enter(&prov_tab_mutex);
if (prov_id >= KCF_MAX_PROVIDERS ||
((prov_desc = prov_tab[prov_id]) == NULL)) {
mutex_exit(&prov_tab_mutex);
return (CRYPTO_INVALID_PROVIDER_ID);
}
mutex_exit(&prov_tab_mutex);
/*
* The provider id must remain valid until the associated provider
* descriptor is freed. For this reason, we simply release our
* reference to the descriptor here. When the reference count
* reaches zero, kcf_free_provider_desc() will be invoked and
* the associated entry in the providers table will be released
* at that time.
*/
KCF_PROV_REFRELE(prov_desc);
KCF_PROV_IREFRELE(prov_desc);
return (CRYPTO_SUCCESS);
}
/*
* Returns the provider descriptor corresponding to the specified
* provider id. A REFHOLD is done on the descriptor before it is
* returned to the caller. It is the responsibility of the caller
* to do a REFRELE once it is done with the provider descriptor.
*/
kcf_provider_desc_t *
kcf_prov_tab_lookup(crypto_provider_id_t prov_id)
{
kcf_provider_desc_t *prov_desc;
mutex_enter(&prov_tab_mutex);
prov_desc = prov_tab[prov_id];
if (prov_desc == NULL) {
mutex_exit(&prov_tab_mutex);
return (NULL);
}
KCF_PROV_REFHOLD(prov_desc);
mutex_exit(&prov_tab_mutex);
return (prov_desc);
}
static void
-allocate_ops_v1(crypto_ops_t *src, crypto_ops_t *dst, uint_t *mech_list_count)
+allocate_ops_v1(const crypto_ops_t *src, crypto_ops_t *dst,
+ uint_t *mech_list_count)
{
if (src->co_control_ops != NULL)
dst->co_control_ops = kmem_alloc(sizeof (crypto_control_ops_t),
KM_SLEEP);
if (src->co_digest_ops != NULL)
dst->co_digest_ops = kmem_alloc(sizeof (crypto_digest_ops_t),
KM_SLEEP);
if (src->co_cipher_ops != NULL)
dst->co_cipher_ops = kmem_alloc(sizeof (crypto_cipher_ops_t),
KM_SLEEP);
if (src->co_mac_ops != NULL)
dst->co_mac_ops = kmem_alloc(sizeof (crypto_mac_ops_t),
KM_SLEEP);
if (src->co_sign_ops != NULL)
dst->co_sign_ops = kmem_alloc(sizeof (crypto_sign_ops_t),
KM_SLEEP);
if (src->co_verify_ops != NULL)
dst->co_verify_ops = kmem_alloc(sizeof (crypto_verify_ops_t),
KM_SLEEP);
if (src->co_dual_ops != NULL)
dst->co_dual_ops = kmem_alloc(sizeof (crypto_dual_ops_t),
KM_SLEEP);
if (src->co_dual_cipher_mac_ops != NULL)
dst->co_dual_cipher_mac_ops = kmem_alloc(
sizeof (crypto_dual_cipher_mac_ops_t), KM_SLEEP);
if (src->co_random_ops != NULL) {
dst->co_random_ops = kmem_alloc(
sizeof (crypto_random_number_ops_t), KM_SLEEP);
/*
* Allocate storage to store the array of supported mechanisms
* specified by provider. We allocate extra mechanism storage
* if the provider has random_ops since we keep an internal
* mechanism, SUN_RANDOM, in this case.
*/
(*mech_list_count)++;
}
if (src->co_session_ops != NULL)
dst->co_session_ops = kmem_alloc(sizeof (crypto_session_ops_t),
KM_SLEEP);
if (src->co_object_ops != NULL)
dst->co_object_ops = kmem_alloc(sizeof (crypto_object_ops_t),
KM_SLEEP);
if (src->co_key_ops != NULL)
dst->co_key_ops = kmem_alloc(sizeof (crypto_key_ops_t),
KM_SLEEP);
if (src->co_provider_ops != NULL)
dst->co_provider_ops = kmem_alloc(
sizeof (crypto_provider_management_ops_t), KM_SLEEP);
if (src->co_ctx_ops != NULL)
dst->co_ctx_ops = kmem_alloc(sizeof (crypto_ctx_ops_t),
KM_SLEEP);
}
static void
-allocate_ops_v2(crypto_ops_t *src, crypto_ops_t *dst)
+allocate_ops_v2(const crypto_ops_t *src, crypto_ops_t *dst)
{
if (src->co_mech_ops != NULL)
dst->co_mech_ops = kmem_alloc(sizeof (crypto_mech_ops_t),
KM_SLEEP);
}
static void
-allocate_ops_v3(crypto_ops_t *src, crypto_ops_t *dst)
+allocate_ops_v3(const crypto_ops_t *src, crypto_ops_t *dst)
{
if (src->co_nostore_key_ops != NULL)
dst->co_nostore_key_ops =
kmem_alloc(sizeof (crypto_nostore_key_ops_t), KM_SLEEP);
}
/*
* Allocate a provider descriptor. mech_list_count specifies the
* number of mechanisms supported by the providers, and is used
* to allocate storage for the mechanism table.
* This function may sleep while allocating memory, which is OK
* since it is invoked from user context during provider registration.
*/
kcf_provider_desc_t *
-kcf_alloc_provider_desc(crypto_provider_info_t *info)
+kcf_alloc_provider_desc(const crypto_provider_info_t *info)
{
- int i, j;
kcf_provider_desc_t *desc;
uint_t mech_list_count = info->pi_mech_list_count;
- crypto_ops_t *src_ops = info->pi_ops_vector;
+ const crypto_ops_t *src_ops = info->pi_ops_vector;
desc = kmem_zalloc(sizeof (kcf_provider_desc_t), KM_SLEEP);
/*
* pd_description serves two purposes
* - Appears as a blank padded PKCS#11 style string, that will be
* returned to applications in CK_SLOT_INFO.slotDescription.
* This means that we should not have a null character in the
* first CRYPTO_PROVIDER_DESCR_MAX_LEN bytes.
* - Appears as a null-terminated string that can be used by
* other kcf routines.
*
* So, we allocate enough room for one extra null terminator
* which keeps every one happy.
*/
desc->pd_description = kmem_alloc(CRYPTO_PROVIDER_DESCR_MAX_LEN + 1,
KM_SLEEP);
(void) memset(desc->pd_description, ' ',
CRYPTO_PROVIDER_DESCR_MAX_LEN);
desc->pd_description[CRYPTO_PROVIDER_DESCR_MAX_LEN] = '\0';
/*
* Since the framework does not require the ops vector specified
* by the providers during registration to be persistent,
* KCF needs to allocate storage where copies of the ops
* vectors are copied.
*/
- desc->pd_ops_vector = kmem_zalloc(sizeof (crypto_ops_t), KM_SLEEP);
+ crypto_ops_t *opvec = kmem_zalloc(sizeof (crypto_ops_t), KM_SLEEP);
if (info->pi_provider_type != CRYPTO_LOGICAL_PROVIDER) {
- allocate_ops_v1(src_ops, desc->pd_ops_vector, &mech_list_count);
+ allocate_ops_v1(src_ops, opvec, &mech_list_count);
if (info->pi_interface_version >= CRYPTO_SPI_VERSION_2)
- allocate_ops_v2(src_ops, desc->pd_ops_vector);
+ allocate_ops_v2(src_ops, opvec);
if (info->pi_interface_version == CRYPTO_SPI_VERSION_3)
- allocate_ops_v3(src_ops, desc->pd_ops_vector);
+ allocate_ops_v3(src_ops, opvec);
}
+ desc->pd_ops_vector = opvec;
desc->pd_mech_list_count = mech_list_count;
desc->pd_mechanisms = kmem_zalloc(sizeof (crypto_mech_info_t) *
mech_list_count, KM_SLEEP);
- for (i = 0; i < KCF_OPS_CLASSSIZE; i++)
- for (j = 0; j < KCF_MAXMECHTAB; j++)
+ for (int i = 0; i < KCF_OPS_CLASSSIZE; i++)
+ for (int j = 0; j < KCF_MAXMECHTAB; j++)
desc->pd_mech_indx[i][j] = KCF_INVALID_INDX;
desc->pd_prov_id = KCF_PROVID_INVALID;
desc->pd_state = KCF_PROV_ALLOCATED;
mutex_init(&desc->pd_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&desc->pd_resume_cv, NULL, CV_DEFAULT, NULL);
cv_init(&desc->pd_remove_cv, NULL, CV_DEFAULT, NULL);
return (desc);
}
/*
* Called by KCF_PROV_REFRELE when a provider's reference count drops
* to zero. We free the descriptor when the last reference is released.
* However, for software providers, we do not free it when there is an
* unregister thread waiting. We signal that thread in this case and
* that thread is responsible for freeing the descriptor.
*/
void
kcf_provider_zero_refcnt(kcf_provider_desc_t *desc)
{
mutex_enter(&desc->pd_lock);
switch (desc->pd_prov_type) {
case CRYPTO_SW_PROVIDER:
if (desc->pd_state == KCF_PROV_REMOVED ||
desc->pd_state == KCF_PROV_DISABLED) {
desc->pd_state = KCF_PROV_FREED;
cv_broadcast(&desc->pd_remove_cv);
mutex_exit(&desc->pd_lock);
break;
}
fallthrough;
case CRYPTO_HW_PROVIDER:
case CRYPTO_LOGICAL_PROVIDER:
mutex_exit(&desc->pd_lock);
kcf_free_provider_desc(desc);
}
}
/*
* Free a provider descriptor.
*/
void
kcf_free_provider_desc(kcf_provider_desc_t *desc)
{
if (desc == NULL)
return;
mutex_enter(&prov_tab_mutex);
if (desc->pd_prov_id != KCF_PROVID_INVALID) {
/* release the associated providers table entry */
ASSERT(prov_tab[desc->pd_prov_id] != NULL);
prov_tab[desc->pd_prov_id] = NULL;
prov_tab_num--;
}
mutex_exit(&prov_tab_mutex);
/* free the kernel memory associated with the provider descriptor */
if (desc->pd_description != NULL)
kmem_free(desc->pd_description,
CRYPTO_PROVIDER_DESCR_MAX_LEN + 1);
if (desc->pd_ops_vector != NULL) {
if (desc->pd_ops_vector->co_control_ops != NULL)
kmem_free(desc->pd_ops_vector->co_control_ops,
sizeof (crypto_control_ops_t));
if (desc->pd_ops_vector->co_digest_ops != NULL)
kmem_free(desc->pd_ops_vector->co_digest_ops,
sizeof (crypto_digest_ops_t));
if (desc->pd_ops_vector->co_cipher_ops != NULL)
kmem_free(desc->pd_ops_vector->co_cipher_ops,
sizeof (crypto_cipher_ops_t));
if (desc->pd_ops_vector->co_mac_ops != NULL)
kmem_free(desc->pd_ops_vector->co_mac_ops,
sizeof (crypto_mac_ops_t));
if (desc->pd_ops_vector->co_sign_ops != NULL)
kmem_free(desc->pd_ops_vector->co_sign_ops,
sizeof (crypto_sign_ops_t));
if (desc->pd_ops_vector->co_verify_ops != NULL)
kmem_free(desc->pd_ops_vector->co_verify_ops,
sizeof (crypto_verify_ops_t));
if (desc->pd_ops_vector->co_dual_ops != NULL)
kmem_free(desc->pd_ops_vector->co_dual_ops,
sizeof (crypto_dual_ops_t));
if (desc->pd_ops_vector->co_dual_cipher_mac_ops != NULL)
kmem_free(desc->pd_ops_vector->co_dual_cipher_mac_ops,
sizeof (crypto_dual_cipher_mac_ops_t));
if (desc->pd_ops_vector->co_random_ops != NULL)
kmem_free(desc->pd_ops_vector->co_random_ops,
sizeof (crypto_random_number_ops_t));
if (desc->pd_ops_vector->co_session_ops != NULL)
kmem_free(desc->pd_ops_vector->co_session_ops,
sizeof (crypto_session_ops_t));
if (desc->pd_ops_vector->co_object_ops != NULL)
kmem_free(desc->pd_ops_vector->co_object_ops,
sizeof (crypto_object_ops_t));
if (desc->pd_ops_vector->co_key_ops != NULL)
kmem_free(desc->pd_ops_vector->co_key_ops,
sizeof (crypto_key_ops_t));
if (desc->pd_ops_vector->co_provider_ops != NULL)
kmem_free(desc->pd_ops_vector->co_provider_ops,
sizeof (crypto_provider_management_ops_t));
if (desc->pd_ops_vector->co_ctx_ops != NULL)
kmem_free(desc->pd_ops_vector->co_ctx_ops,
sizeof (crypto_ctx_ops_t));
if (desc->pd_ops_vector->co_mech_ops != NULL)
kmem_free(desc->pd_ops_vector->co_mech_ops,
sizeof (crypto_mech_ops_t));
if (desc->pd_ops_vector->co_nostore_key_ops != NULL)
kmem_free(desc->pd_ops_vector->co_nostore_key_ops,
sizeof (crypto_nostore_key_ops_t));
kmem_free(desc->pd_ops_vector, sizeof (crypto_ops_t));
}
if (desc->pd_mechanisms != NULL)
/* free the memory associated with the mechanism info's */
kmem_free(desc->pd_mechanisms, sizeof (crypto_mech_info_t) *
desc->pd_mech_list_count);
if (desc->pd_sched_info.ks_taskq != NULL)
taskq_destroy(desc->pd_sched_info.ks_taskq);
mutex_destroy(&desc->pd_lock);
cv_destroy(&desc->pd_resume_cv);
cv_destroy(&desc->pd_remove_cv);
kmem_free(desc, sizeof (kcf_provider_desc_t));
}
/*
* Returns an array of hardware and logical provider descriptors,
* a.k.a the PKCS#11 slot list. A REFHOLD is done on each descriptor
* before the array is returned. The entire table can be freed by
* calling kcf_free_provider_tab().
*/
int
kcf_get_slot_list(uint_t *count, kcf_provider_desc_t ***array,
boolean_t unverified)
{
kcf_provider_desc_t *prov_desc;
kcf_provider_desc_t **p = NULL;
char *last;
uint_t cnt = 0;
uint_t i, j;
int rval = CRYPTO_SUCCESS;
size_t n, final_size;
/* count the providers */
mutex_enter(&prov_tab_mutex);
for (i = 0; i < KCF_MAX_PROVIDERS; i++) {
if ((prov_desc = prov_tab[i]) != NULL &&
((prov_desc->pd_prov_type == CRYPTO_HW_PROVIDER &&
(prov_desc->pd_flags & CRYPTO_HIDE_PROVIDER) == 0) ||
prov_desc->pd_prov_type == CRYPTO_LOGICAL_PROVIDER)) {
if (KCF_IS_PROV_USABLE(prov_desc) ||
(unverified && KCF_IS_PROV_UNVERIFIED(prov_desc))) {
cnt++;
}
}
}
mutex_exit(&prov_tab_mutex);
if (cnt == 0)
goto out;
n = cnt * sizeof (kcf_provider_desc_t *);
again:
p = kmem_zalloc(n, KM_SLEEP);
/* pointer to last entry in the array */
last = (char *)&p[cnt-1];
mutex_enter(&prov_tab_mutex);
/* fill the slot list */
for (i = 0, j = 0; i < KCF_MAX_PROVIDERS; i++) {
if ((prov_desc = prov_tab[i]) != NULL &&
((prov_desc->pd_prov_type == CRYPTO_HW_PROVIDER &&
(prov_desc->pd_flags & CRYPTO_HIDE_PROVIDER) == 0) ||
prov_desc->pd_prov_type == CRYPTO_LOGICAL_PROVIDER)) {
if (KCF_IS_PROV_USABLE(prov_desc) ||
(unverified && KCF_IS_PROV_UNVERIFIED(prov_desc))) {
if ((char *)&p[j] > last) {
mutex_exit(&prov_tab_mutex);
kcf_free_provider_tab(cnt, p);
n = n << 1;
cnt = cnt << 1;
goto again;
}
p[j++] = prov_desc;
KCF_PROV_REFHOLD(prov_desc);
}
}
}
mutex_exit(&prov_tab_mutex);
final_size = j * sizeof (kcf_provider_desc_t *);
cnt = j;
ASSERT(final_size <= n);
/* check if buffer we allocated is too large */
if (final_size < n) {
char *final_buffer = NULL;
if (final_size > 0) {
final_buffer = kmem_alloc(final_size, KM_SLEEP);
bcopy(p, final_buffer, final_size);
}
kmem_free(p, n);
p = (kcf_provider_desc_t **)final_buffer;
}
out:
*count = cnt;
*array = p;
return (rval);
}
/*
* Free an array of hardware provider descriptors. A REFRELE
* is done on each descriptor before the table is freed.
*/
void
kcf_free_provider_tab(uint_t count, kcf_provider_desc_t **array)
{
kcf_provider_desc_t *prov_desc;
int i;
for (i = 0; i < count; i++) {
if ((prov_desc = array[i]) != NULL) {
KCF_PROV_REFRELE(prov_desc);
}
}
kmem_free(array, count * sizeof (kcf_provider_desc_t *));
}
/*
* Returns in the location pointed to by pd a pointer to the descriptor
* for the software provider for the specified mechanism.
* The provider descriptor is returned held and it is the caller's
* responsibility to release it when done. The mechanism entry
* is returned if the optional argument mep is non NULL.
*
* Returns one of the CRYPTO_ * error codes on failure, and
* CRYPTO_SUCCESS on success.
*/
int
kcf_get_sw_prov(crypto_mech_type_t mech_type, kcf_provider_desc_t **pd,
kcf_mech_entry_t **mep, boolean_t log_warn)
{
kcf_mech_entry_t *me;
/* get the mechanism entry for this mechanism */
if (kcf_get_mech_entry(mech_type, &me) != KCF_SUCCESS)
return (CRYPTO_MECHANISM_INVALID);
/*
* Get the software provider for this mechanism.
* Lock the mech_entry until we grab the 'pd'.
*/
mutex_enter(&me->me_mutex);
if (me->me_sw_prov == NULL ||
(*pd = me->me_sw_prov->pm_prov_desc) == NULL) {
/* no SW provider for this mechanism */
if (log_warn)
cmn_err(CE_WARN, "no SW provider for \"%s\"\n",
me->me_name);
mutex_exit(&me->me_mutex);
return (CRYPTO_MECH_NOT_SUPPORTED);
}
KCF_PROV_REFHOLD(*pd);
mutex_exit(&me->me_mutex);
if (mep != NULL)
*mep = me;
return (CRYPTO_SUCCESS);
}
diff --git a/module/icp/core/kcf_sched.c b/module/icp/core/kcf_sched.c
index e4ccdbde9fb4..062e96059f20 100644
--- a/module/icp/core/kcf_sched.c
+++ b/module/icp/core/kcf_sched.c
@@ -1,1780 +1,1775 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* This file contains the core framework routines for the
* kernel cryptographic framework. These routines are at the
* layer, between the kernel API/ioctls and the SPI.
*/
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
#include <sys/crypto/impl.h>
#include <sys/crypto/sched_impl.h>
#include <sys/crypto/api.h>
-kcf_global_swq_t *gswq; /* Global software queue */
+static kcf_global_swq_t *gswq; /* Global software queue */
/* Thread pool related variables */
static kcf_pool_t *kcfpool; /* Thread pool of kcfd LWPs */
-int kcf_maxthreads = 2;
-int kcf_minthreads = 1;
-int kcf_thr_multiple = 2; /* Boot-time tunable for experimentation */
-static ulong_t kcf_idlethr_timeout;
-#define KCF_DEFAULT_THRTIMEOUT 60000000 /* 60 seconds */
+static const int kcf_maxthreads = 2;
+static const int kcf_minthreads = 1;
/* kmem caches used by the scheduler */
static kmem_cache_t *kcf_sreq_cache;
static kmem_cache_t *kcf_areq_cache;
static kmem_cache_t *kcf_context_cache;
/* Global request ID table */
static kcf_reqid_table_t *kcf_reqid_table[REQID_TABLES];
/* KCF stats. Not protected. */
static kcf_stats_t kcf_ksdata = {
{ "total threads in pool", KSTAT_DATA_UINT32},
{ "idle threads in pool", KSTAT_DATA_UINT32},
{ "min threads in pool", KSTAT_DATA_UINT32},
{ "max threads in pool", KSTAT_DATA_UINT32},
{ "requests in gswq", KSTAT_DATA_UINT32},
{ "max requests in gswq", KSTAT_DATA_UINT32},
{ "threads for HW taskq", KSTAT_DATA_UINT32},
{ "minalloc for HW taskq", KSTAT_DATA_UINT32},
{ "maxalloc for HW taskq", KSTAT_DATA_UINT32}
};
static kstat_t *kcf_misc_kstat = NULL;
ulong_t kcf_swprov_hndl = 0;
static kcf_areq_node_t *kcf_areqnode_alloc(kcf_provider_desc_t *,
kcf_context_t *, crypto_call_req_t *, kcf_req_params_t *, boolean_t);
static int kcf_disp_sw_request(kcf_areq_node_t *);
static void process_req_hwp(void *);
static int kcf_enqueue(kcf_areq_node_t *);
static void kcfpool_alloc(void);
static void kcf_reqid_delete(kcf_areq_node_t *areq);
static crypto_req_id_t kcf_reqid_insert(kcf_areq_node_t *areq);
static int kcf_misc_kstat_update(kstat_t *ksp, int rw);
/*
* Create a new context.
*/
crypto_ctx_t *
kcf_new_ctx(crypto_call_req_t *crq, kcf_provider_desc_t *pd,
crypto_session_id_t sid)
{
crypto_ctx_t *ctx;
kcf_context_t *kcf_ctx;
kcf_ctx = kmem_cache_alloc(kcf_context_cache,
(crq == NULL) ? KM_SLEEP : KM_NOSLEEP);
if (kcf_ctx == NULL)
return (NULL);
/* initialize the context for the consumer */
kcf_ctx->kc_refcnt = 1;
kcf_ctx->kc_req_chain_first = NULL;
kcf_ctx->kc_req_chain_last = NULL;
kcf_ctx->kc_secondctx = NULL;
KCF_PROV_REFHOLD(pd);
kcf_ctx->kc_prov_desc = pd;
kcf_ctx->kc_sw_prov_desc = NULL;
kcf_ctx->kc_mech = NULL;
ctx = &kcf_ctx->kc_glbl_ctx;
ctx->cc_provider = pd->pd_prov_handle;
ctx->cc_session = sid;
ctx->cc_provider_private = NULL;
ctx->cc_framework_private = (void *)kcf_ctx;
ctx->cc_flags = 0;
ctx->cc_opstate = NULL;
return (ctx);
}
/*
* Allocate a new async request node.
*
* ictx - Framework private context pointer
* crq - Has callback function and argument. Should be non NULL.
* req - The parameters to pass to the SPI
*/
static kcf_areq_node_t *
kcf_areqnode_alloc(kcf_provider_desc_t *pd, kcf_context_t *ictx,
crypto_call_req_t *crq, kcf_req_params_t *req, boolean_t isdual)
{
kcf_areq_node_t *arptr, *areq;
ASSERT(crq != NULL);
arptr = kmem_cache_alloc(kcf_areq_cache, KM_NOSLEEP);
if (arptr == NULL)
return (NULL);
arptr->an_state = REQ_ALLOCATED;
arptr->an_reqarg = *crq;
arptr->an_params = *req;
arptr->an_context = ictx;
arptr->an_isdual = isdual;
arptr->an_next = arptr->an_prev = NULL;
KCF_PROV_REFHOLD(pd);
arptr->an_provider = pd;
arptr->an_tried_plist = NULL;
arptr->an_refcnt = 1;
arptr->an_idnext = arptr->an_idprev = NULL;
/*
* Requests for context-less operations do not use the
* fields - an_is_my_turn, and an_ctxchain_next.
*/
if (ictx == NULL)
return (arptr);
KCF_CONTEXT_REFHOLD(ictx);
/*
* Chain this request to the context.
*/
mutex_enter(&ictx->kc_in_use_lock);
arptr->an_ctxchain_next = NULL;
if ((areq = ictx->kc_req_chain_last) == NULL) {
arptr->an_is_my_turn = B_TRUE;
ictx->kc_req_chain_last =
ictx->kc_req_chain_first = arptr;
} else {
ASSERT(ictx->kc_req_chain_first != NULL);
arptr->an_is_my_turn = B_FALSE;
/* Insert the new request to the end of the chain. */
areq->an_ctxchain_next = arptr;
ictx->kc_req_chain_last = arptr;
}
mutex_exit(&ictx->kc_in_use_lock);
return (arptr);
}
/*
* Queue the request node and do one of the following:
* - If there is an idle thread signal it to run.
* - If there is no idle thread and max running threads is not
* reached, signal the creator thread for more threads.
*
* If the two conditions above are not met, we don't need to do
* anything. The request will be picked up by one of the
* worker threads when it becomes available.
*/
static int
kcf_disp_sw_request(kcf_areq_node_t *areq)
{
int err;
int cnt = 0;
if ((err = kcf_enqueue(areq)) != 0)
return (err);
if (kcfpool->kp_idlethreads > 0) {
/* Signal an idle thread to run */
mutex_enter(&gswq->gs_lock);
cv_signal(&gswq->gs_cv);
mutex_exit(&gswq->gs_lock);
return (CRYPTO_QUEUED);
}
/*
* We keep the number of running threads to be at
* kcf_minthreads to reduce gs_lock contention.
*/
cnt = kcf_minthreads -
(kcfpool->kp_threads - kcfpool->kp_blockedthreads);
if (cnt > 0) {
/*
* The following ensures the number of threads in pool
* does not exceed kcf_maxthreads.
*/
cnt = MIN(cnt, kcf_maxthreads - (int)kcfpool->kp_threads);
if (cnt > 0) {
/* Signal the creator thread for more threads */
mutex_enter(&kcfpool->kp_user_lock);
if (!kcfpool->kp_signal_create_thread) {
kcfpool->kp_signal_create_thread = B_TRUE;
kcfpool->kp_nthrs = cnt;
cv_signal(&kcfpool->kp_user_cv);
}
mutex_exit(&kcfpool->kp_user_lock);
}
}
return (CRYPTO_QUEUED);
}
/*
* This routine is called by the taskq associated with
* each hardware provider. We notify the kernel consumer
* via the callback routine in case of CRYPTO_SUCCESS or
* a failure.
*
* A request can be of type kcf_areq_node_t or of type
* kcf_sreq_node_t.
*/
static void
process_req_hwp(void *ireq)
{
int error = 0;
crypto_ctx_t *ctx;
kcf_call_type_t ctype;
kcf_provider_desc_t *pd;
kcf_areq_node_t *areq = (kcf_areq_node_t *)ireq;
kcf_sreq_node_t *sreq = (kcf_sreq_node_t *)ireq;
pd = ((ctype = GET_REQ_TYPE(ireq)) == CRYPTO_SYNCH) ?
sreq->sn_provider : areq->an_provider;
/*
* Wait if flow control is in effect for the provider. A
* CRYPTO_PROVIDER_READY or CRYPTO_PROVIDER_FAILED
* notification will signal us. We also get signaled if
* the provider is unregistering.
*/
if (pd->pd_state == KCF_PROV_BUSY) {
mutex_enter(&pd->pd_lock);
while (pd->pd_state == KCF_PROV_BUSY)
cv_wait(&pd->pd_resume_cv, &pd->pd_lock);
mutex_exit(&pd->pd_lock);
}
/*
* Bump the internal reference count while the request is being
* processed. This is how we know when it's safe to unregister
* a provider. This step must precede the pd_state check below.
*/
KCF_PROV_IREFHOLD(pd);
/*
* Fail the request if the provider has failed. We return a
* recoverable error and the notified clients attempt any
* recovery. For async clients this is done in kcf_aop_done()
* and for sync clients it is done in the k-api routines.
*/
if (pd->pd_state >= KCF_PROV_FAILED) {
error = CRYPTO_DEVICE_ERROR;
goto bail;
}
if (ctype == CRYPTO_SYNCH) {
mutex_enter(&sreq->sn_lock);
sreq->sn_state = REQ_INPROGRESS;
mutex_exit(&sreq->sn_lock);
ctx = sreq->sn_context ? &sreq->sn_context->kc_glbl_ctx : NULL;
error = common_submit_request(sreq->sn_provider, ctx,
sreq->sn_params, sreq);
} else {
kcf_context_t *ictx;
ASSERT(ctype == CRYPTO_ASYNCH);
/*
* We are in the per-hardware provider thread context and
* hence can sleep. Note that the caller would have done
* a taskq_dispatch(..., TQ_NOSLEEP) and would have returned.
*/
ctx = (ictx = areq->an_context) ? &ictx->kc_glbl_ctx : NULL;
mutex_enter(&areq->an_lock);
/*
* We need to maintain ordering for multi-part requests.
* an_is_my_turn is set to B_TRUE initially for a request
* when it is enqueued and there are no other requests
* for that context. It is set later from kcf_aop_done() when
* the request before us in the chain of requests for the
* context completes. We get signaled at that point.
*/
if (ictx != NULL) {
ASSERT(ictx->kc_prov_desc == areq->an_provider);
while (areq->an_is_my_turn == B_FALSE) {
cv_wait(&areq->an_turn_cv, &areq->an_lock);
}
}
areq->an_state = REQ_INPROGRESS;
mutex_exit(&areq->an_lock);
error = common_submit_request(areq->an_provider, ctx,
&areq->an_params, areq);
}
bail:
if (error == CRYPTO_QUEUED) {
/*
* The request is queued by the provider and we should
* get a crypto_op_notification() from the provider later.
* We notify the consumer at that time.
*/
return;
} else { /* CRYPTO_SUCCESS or other failure */
KCF_PROV_IREFRELE(pd);
if (ctype == CRYPTO_SYNCH)
kcf_sop_done(sreq, error);
else
kcf_aop_done(areq, error);
}
}
/*
* This routine checks if a request can be retried on another
* provider. If true, mech1 is initialized to point to the mechanism
* structure. mech2 is also initialized in case of a dual operation. fg
* is initialized to the correct crypto_func_group_t bit flag. They are
* initialized by this routine, so that the caller can pass them to a
* kcf_get_mech_provider() or kcf_get_dual_provider() with no further change.
*
* We check that the request is for a init or atomic routine and that
* it is for one of the operation groups used from k-api .
*/
static boolean_t
can_resubmit(kcf_areq_node_t *areq, crypto_mechanism_t **mech1,
crypto_mechanism_t **mech2, crypto_func_group_t *fg)
{
kcf_req_params_t *params;
kcf_op_type_t optype;
params = &areq->an_params;
optype = params->rp_optype;
if (!(IS_INIT_OP(optype) || IS_ATOMIC_OP(optype)))
return (B_FALSE);
switch (params->rp_opgrp) {
case KCF_OG_DIGEST: {
kcf_digest_ops_params_t *dops = &params->rp_u.digest_params;
dops->do_mech.cm_type = dops->do_framework_mechtype;
*mech1 = &dops->do_mech;
*fg = (optype == KCF_OP_INIT) ? CRYPTO_FG_DIGEST :
CRYPTO_FG_DIGEST_ATOMIC;
break;
}
case KCF_OG_MAC: {
kcf_mac_ops_params_t *mops = &params->rp_u.mac_params;
mops->mo_mech.cm_type = mops->mo_framework_mechtype;
*mech1 = &mops->mo_mech;
*fg = (optype == KCF_OP_INIT) ? CRYPTO_FG_MAC :
CRYPTO_FG_MAC_ATOMIC;
break;
}
case KCF_OG_SIGN: {
kcf_sign_ops_params_t *sops = &params->rp_u.sign_params;
sops->so_mech.cm_type = sops->so_framework_mechtype;
*mech1 = &sops->so_mech;
switch (optype) {
case KCF_OP_INIT:
*fg = CRYPTO_FG_SIGN;
break;
case KCF_OP_ATOMIC:
*fg = CRYPTO_FG_SIGN_ATOMIC;
break;
default:
ASSERT(optype == KCF_OP_SIGN_RECOVER_ATOMIC);
*fg = CRYPTO_FG_SIGN_RECOVER_ATOMIC;
}
break;
}
case KCF_OG_VERIFY: {
kcf_verify_ops_params_t *vops = &params->rp_u.verify_params;
vops->vo_mech.cm_type = vops->vo_framework_mechtype;
*mech1 = &vops->vo_mech;
switch (optype) {
case KCF_OP_INIT:
*fg = CRYPTO_FG_VERIFY;
break;
case KCF_OP_ATOMIC:
*fg = CRYPTO_FG_VERIFY_ATOMIC;
break;
default:
ASSERT(optype == KCF_OP_VERIFY_RECOVER_ATOMIC);
*fg = CRYPTO_FG_VERIFY_RECOVER_ATOMIC;
}
break;
}
case KCF_OG_ENCRYPT: {
kcf_encrypt_ops_params_t *eops = &params->rp_u.encrypt_params;
eops->eo_mech.cm_type = eops->eo_framework_mechtype;
*mech1 = &eops->eo_mech;
*fg = (optype == KCF_OP_INIT) ? CRYPTO_FG_ENCRYPT :
CRYPTO_FG_ENCRYPT_ATOMIC;
break;
}
case KCF_OG_DECRYPT: {
kcf_decrypt_ops_params_t *dcrops = &params->rp_u.decrypt_params;
dcrops->dop_mech.cm_type = dcrops->dop_framework_mechtype;
*mech1 = &dcrops->dop_mech;
*fg = (optype == KCF_OP_INIT) ? CRYPTO_FG_DECRYPT :
CRYPTO_FG_DECRYPT_ATOMIC;
break;
}
case KCF_OG_ENCRYPT_MAC: {
kcf_encrypt_mac_ops_params_t *eops =
&params->rp_u.encrypt_mac_params;
eops->em_encr_mech.cm_type = eops->em_framework_encr_mechtype;
*mech1 = &eops->em_encr_mech;
eops->em_mac_mech.cm_type = eops->em_framework_mac_mechtype;
*mech2 = &eops->em_mac_mech;
*fg = (optype == KCF_OP_INIT) ? CRYPTO_FG_ENCRYPT_MAC :
CRYPTO_FG_ENCRYPT_MAC_ATOMIC;
break;
}
case KCF_OG_MAC_DECRYPT: {
kcf_mac_decrypt_ops_params_t *dops =
&params->rp_u.mac_decrypt_params;
dops->md_mac_mech.cm_type = dops->md_framework_mac_mechtype;
*mech1 = &dops->md_mac_mech;
dops->md_decr_mech.cm_type = dops->md_framework_decr_mechtype;
*mech2 = &dops->md_decr_mech;
*fg = (optype == KCF_OP_INIT) ? CRYPTO_FG_MAC_DECRYPT :
CRYPTO_FG_MAC_DECRYPT_ATOMIC;
break;
}
default:
return (B_FALSE);
}
return (B_TRUE);
}
/*
* This routine is called when a request to a provider has failed
* with a recoverable error. This routine tries to find another provider
* and dispatches the request to the new provider, if one is available.
* We reuse the request structure.
*
* A return value of NULL from kcf_get_mech_provider() indicates
* we have tried the last provider.
*/
static int
kcf_resubmit_request(kcf_areq_node_t *areq)
{
int error = CRYPTO_FAILED;
kcf_context_t *ictx;
kcf_provider_desc_t *old_pd;
kcf_provider_desc_t *new_pd;
crypto_mechanism_t *mech1 = NULL, *mech2 = NULL;
crypto_mech_type_t prov_mt1, prov_mt2;
crypto_func_group_t fg = 0;
if (!can_resubmit(areq, &mech1, &mech2, &fg))
return (error);
old_pd = areq->an_provider;
/*
* Add old_pd to the list of providers already tried. We release
* the hold on old_pd (from the earlier kcf_get_mech_provider()) in
* kcf_free_triedlist().
*/
if (kcf_insert_triedlist(&areq->an_tried_plist, old_pd,
KM_NOSLEEP) == NULL)
return (error);
if (mech1 && !mech2) {
new_pd = kcf_get_mech_provider(mech1->cm_type, NULL, &error,
areq->an_tried_plist, fg,
(areq->an_reqarg.cr_flag & CRYPTO_RESTRICTED), 0);
} else {
ASSERT(mech1 != NULL && mech2 != NULL);
new_pd = kcf_get_dual_provider(mech1, mech2, NULL, &prov_mt1,
&prov_mt2, &error, areq->an_tried_plist, fg, fg,
(areq->an_reqarg.cr_flag & CRYPTO_RESTRICTED), 0);
}
if (new_pd == NULL)
return (error);
/*
* We reuse the old context by resetting provider specific
* fields in it.
*/
if ((ictx = areq->an_context) != NULL) {
crypto_ctx_t *ctx;
ASSERT(old_pd == ictx->kc_prov_desc);
KCF_PROV_REFRELE(ictx->kc_prov_desc);
KCF_PROV_REFHOLD(new_pd);
ictx->kc_prov_desc = new_pd;
ctx = &ictx->kc_glbl_ctx;
ctx->cc_provider = new_pd->pd_prov_handle;
ctx->cc_session = new_pd->pd_sid;
ctx->cc_provider_private = NULL;
}
/* We reuse areq. by resetting the provider and context fields. */
KCF_PROV_REFRELE(old_pd);
KCF_PROV_REFHOLD(new_pd);
areq->an_provider = new_pd;
mutex_enter(&areq->an_lock);
areq->an_state = REQ_WAITING;
mutex_exit(&areq->an_lock);
switch (new_pd->pd_prov_type) {
case CRYPTO_SW_PROVIDER:
error = kcf_disp_sw_request(areq);
break;
case CRYPTO_HW_PROVIDER: {
taskq_t *taskq = new_pd->pd_sched_info.ks_taskq;
if (taskq_dispatch(taskq, process_req_hwp, areq, TQ_NOSLEEP) ==
TASKQID_INVALID) {
error = CRYPTO_HOST_MEMORY;
} else {
error = CRYPTO_QUEUED;
}
break;
default:
break;
}
}
return (error);
}
static inline int EMPTY_TASKQ(taskq_t *tq)
{
#ifdef _KERNEL
return (tq->tq_lowest_id == tq->tq_next_id);
#else
return (tq->tq_task.tqent_next == &tq->tq_task || tq->tq_active == 0);
#endif
}
/*
* Routine called by both ioctl and k-api. The consumer should
* bundle the parameters into a kcf_req_params_t structure. A bunch
* of macros are available in ops_impl.h for this bundling. They are:
*
* KCF_WRAP_DIGEST_OPS_PARAMS()
* KCF_WRAP_MAC_OPS_PARAMS()
* KCF_WRAP_ENCRYPT_OPS_PARAMS()
* KCF_WRAP_DECRYPT_OPS_PARAMS() ... etc.
*
* It is the caller's responsibility to free the ctx argument when
* appropriate. See the KCF_CONTEXT_COND_RELEASE macro for details.
*/
int
kcf_submit_request(kcf_provider_desc_t *pd, crypto_ctx_t *ctx,
crypto_call_req_t *crq, kcf_req_params_t *params, boolean_t cont)
{
int error = CRYPTO_SUCCESS;
kcf_areq_node_t *areq;
kcf_sreq_node_t *sreq;
kcf_context_t *kcf_ctx;
taskq_t *taskq = pd->pd_sched_info.ks_taskq;
kcf_ctx = ctx ? (kcf_context_t *)ctx->cc_framework_private : NULL;
/* Synchronous cases */
if (crq == NULL) {
switch (pd->pd_prov_type) {
case CRYPTO_SW_PROVIDER:
error = common_submit_request(pd, ctx, params,
KCF_RHNDL(KM_SLEEP));
break;
case CRYPTO_HW_PROVIDER:
/*
* Special case for CRYPTO_SYNCHRONOUS providers that
* never return a CRYPTO_QUEUED error. We skip any
* request allocation and call the SPI directly.
*/
if ((pd->pd_flags & CRYPTO_SYNCHRONOUS) &&
EMPTY_TASKQ(taskq)) {
KCF_PROV_IREFHOLD(pd);
if (pd->pd_state == KCF_PROV_READY) {
error = common_submit_request(pd, ctx,
params, KCF_RHNDL(KM_SLEEP));
KCF_PROV_IREFRELE(pd);
ASSERT(error != CRYPTO_QUEUED);
break;
}
KCF_PROV_IREFRELE(pd);
}
sreq = kmem_cache_alloc(kcf_sreq_cache, KM_SLEEP);
sreq->sn_state = REQ_ALLOCATED;
sreq->sn_rv = CRYPTO_FAILED;
sreq->sn_params = params;
/*
* Note that we do not need to hold the context
* for synchronous case as the context will never
* become invalid underneath us. We do not need to hold
* the provider here either as the caller has a hold.
*/
sreq->sn_context = kcf_ctx;
ASSERT(KCF_PROV_REFHELD(pd));
sreq->sn_provider = pd;
ASSERT(taskq != NULL);
/*
* Call the SPI directly if the taskq is empty and the
* provider is not busy, else dispatch to the taskq.
* Calling directly is fine as this is the synchronous
* case. This is unlike the asynchronous case where we
* must always dispatch to the taskq.
*/
if (EMPTY_TASKQ(taskq) &&
pd->pd_state == KCF_PROV_READY) {
process_req_hwp(sreq);
} else {
/*
* We can not tell from taskq_dispatch() return
* value if we exceeded maxalloc. Hence the
* check here. Since we are allowed to wait in
* the synchronous case, we wait for the taskq
* to become empty.
*/
if (taskq->tq_nalloc >= crypto_taskq_maxalloc) {
taskq_wait(taskq);
}
(void) taskq_dispatch(taskq, process_req_hwp,
sreq, TQ_SLEEP);
}
/*
* Wait for the notification to arrive,
* if the operation is not done yet.
* Bug# 4722589 will make the wait a cv_wait_sig().
*/
mutex_enter(&sreq->sn_lock);
while (sreq->sn_state < REQ_DONE)
cv_wait(&sreq->sn_cv, &sreq->sn_lock);
mutex_exit(&sreq->sn_lock);
error = sreq->sn_rv;
kmem_cache_free(kcf_sreq_cache, sreq);
break;
default:
error = CRYPTO_FAILED;
break;
}
} else { /* Asynchronous cases */
switch (pd->pd_prov_type) {
case CRYPTO_SW_PROVIDER:
if (!(crq->cr_flag & CRYPTO_ALWAYS_QUEUE)) {
/*
* This case has less overhead since there is
* no switching of context.
*/
error = common_submit_request(pd, ctx, params,
KCF_RHNDL(KM_NOSLEEP));
} else {
/*
* CRYPTO_ALWAYS_QUEUE is set. We need to
* queue the request and return.
*/
areq = kcf_areqnode_alloc(pd, kcf_ctx, crq,
params, cont);
if (areq == NULL)
error = CRYPTO_HOST_MEMORY;
else {
if (!(crq->cr_flag
& CRYPTO_SKIP_REQID)) {
/*
* Set the request handle. This handle
* is used for any crypto_cancel_req(9f)
* calls from the consumer. We have to
* do this before dispatching the
* request.
*/
crq->cr_reqid = kcf_reqid_insert(areq);
}
error = kcf_disp_sw_request(areq);
/*
* There is an error processing this
* request. Remove the handle and
* release the request structure.
*/
if (error != CRYPTO_QUEUED) {
if (!(crq->cr_flag
& CRYPTO_SKIP_REQID))
kcf_reqid_delete(areq);
KCF_AREQ_REFRELE(areq);
}
}
}
break;
case CRYPTO_HW_PROVIDER:
/*
* We need to queue the request and return.
*/
areq = kcf_areqnode_alloc(pd, kcf_ctx, crq, params,
cont);
if (areq == NULL) {
error = CRYPTO_HOST_MEMORY;
goto done;
}
ASSERT(taskq != NULL);
/*
* We can not tell from taskq_dispatch() return
* value if we exceeded maxalloc. Hence the check
* here.
*/
if (taskq->tq_nalloc >= crypto_taskq_maxalloc) {
error = CRYPTO_BUSY;
KCF_AREQ_REFRELE(areq);
goto done;
}
if (!(crq->cr_flag & CRYPTO_SKIP_REQID)) {
/*
* Set the request handle. This handle is used
* for any crypto_cancel_req(9f) calls from the
* consumer. We have to do this before dispatching
* the request.
*/
crq->cr_reqid = kcf_reqid_insert(areq);
}
if (taskq_dispatch(taskq,
process_req_hwp, areq, TQ_NOSLEEP) ==
TASKQID_INVALID) {
error = CRYPTO_HOST_MEMORY;
if (!(crq->cr_flag & CRYPTO_SKIP_REQID))
kcf_reqid_delete(areq);
KCF_AREQ_REFRELE(areq);
} else {
error = CRYPTO_QUEUED;
}
break;
default:
error = CRYPTO_FAILED;
break;
}
}
done:
return (error);
}
/*
* We're done with this framework context, so free it. Note that freeing
* framework context (kcf_context) frees the global context (crypto_ctx).
*
* The provider is responsible for freeing provider private context after a
* final or single operation and resetting the cc_provider_private field
* to NULL. It should do this before it notifies the framework of the
* completion. We still need to call KCF_PROV_FREE_CONTEXT to handle cases
* like crypto_cancel_ctx(9f).
*/
void
kcf_free_context(kcf_context_t *kcf_ctx)
{
kcf_provider_desc_t *pd = kcf_ctx->kc_prov_desc;
crypto_ctx_t *gctx = &kcf_ctx->kc_glbl_ctx;
kcf_context_t *kcf_secondctx = kcf_ctx->kc_secondctx;
/* Release the second context, if any */
if (kcf_secondctx != NULL)
KCF_CONTEXT_REFRELE(kcf_secondctx);
if (gctx->cc_provider_private != NULL) {
mutex_enter(&pd->pd_lock);
if (!KCF_IS_PROV_REMOVED(pd)) {
/*
* Increment the provider's internal refcnt so it
* doesn't unregister from the framework while
* we're calling the entry point.
*/
KCF_PROV_IREFHOLD(pd);
mutex_exit(&pd->pd_lock);
(void) KCF_PROV_FREE_CONTEXT(pd, gctx);
KCF_PROV_IREFRELE(pd);
} else {
mutex_exit(&pd->pd_lock);
}
}
/* kcf_ctx->kc_prov_desc has a hold on pd */
KCF_PROV_REFRELE(kcf_ctx->kc_prov_desc);
/* check if this context is shared with a software provider */
if ((gctx->cc_flags & CRYPTO_INIT_OPSTATE) &&
kcf_ctx->kc_sw_prov_desc != NULL) {
KCF_PROV_REFRELE(kcf_ctx->kc_sw_prov_desc);
}
kmem_cache_free(kcf_context_cache, kcf_ctx);
}
/*
* Free the request after releasing all the holds.
*/
void
kcf_free_req(kcf_areq_node_t *areq)
{
KCF_PROV_REFRELE(areq->an_provider);
if (areq->an_context != NULL)
KCF_CONTEXT_REFRELE(areq->an_context);
if (areq->an_tried_plist != NULL)
kcf_free_triedlist(areq->an_tried_plist);
kmem_cache_free(kcf_areq_cache, areq);
}
/*
* Utility routine to remove a request from the chain of requests
* hanging off a context.
*/
static void
kcf_removereq_in_ctxchain(kcf_context_t *ictx, kcf_areq_node_t *areq)
{
kcf_areq_node_t *cur, *prev;
/*
* Get context lock, search for areq in the chain and remove it.
*/
ASSERT(ictx != NULL);
mutex_enter(&ictx->kc_in_use_lock);
prev = cur = ictx->kc_req_chain_first;
while (cur != NULL) {
if (cur == areq) {
if (prev == cur) {
if ((ictx->kc_req_chain_first =
cur->an_ctxchain_next) == NULL)
ictx->kc_req_chain_last = NULL;
} else {
if (cur == ictx->kc_req_chain_last)
ictx->kc_req_chain_last = prev;
prev->an_ctxchain_next = cur->an_ctxchain_next;
}
break;
}
prev = cur;
cur = cur->an_ctxchain_next;
}
mutex_exit(&ictx->kc_in_use_lock);
}
/*
* Remove the specified node from the global software queue.
*
* The caller must hold the queue lock and request lock (an_lock).
*/
static void
kcf_remove_node(kcf_areq_node_t *node)
{
kcf_areq_node_t *nextp = node->an_next;
kcf_areq_node_t *prevp = node->an_prev;
if (nextp != NULL)
nextp->an_prev = prevp;
else
gswq->gs_last = prevp;
if (prevp != NULL)
prevp->an_next = nextp;
else
gswq->gs_first = nextp;
node->an_state = REQ_CANCELED;
}
/*
* Add the request node to the end of the global software queue.
*
* The caller should not hold the queue lock. Returns 0 if the
* request is successfully queued. Returns CRYPTO_BUSY if the limit
* on the number of jobs is exceeded.
*/
static int
kcf_enqueue(kcf_areq_node_t *node)
{
kcf_areq_node_t *tnode;
mutex_enter(&gswq->gs_lock);
if (gswq->gs_njobs >= gswq->gs_maxjobs) {
mutex_exit(&gswq->gs_lock);
return (CRYPTO_BUSY);
}
if (gswq->gs_last == NULL) {
gswq->gs_first = gswq->gs_last = node;
} else {
ASSERT(gswq->gs_last->an_next == NULL);
tnode = gswq->gs_last;
tnode->an_next = node;
gswq->gs_last = node;
node->an_prev = tnode;
}
gswq->gs_njobs++;
/* an_lock not needed here as we hold gs_lock */
node->an_state = REQ_WAITING;
mutex_exit(&gswq->gs_lock);
return (0);
}
/*
* kmem_cache_alloc constructor for sync request structure.
*/
static int
kcf_sreq_cache_constructor(void *buf, void *cdrarg, int kmflags)
{
(void) cdrarg, (void) kmflags;
kcf_sreq_node_t *sreq = (kcf_sreq_node_t *)buf;
sreq->sn_type = CRYPTO_SYNCH;
cv_init(&sreq->sn_cv, NULL, CV_DEFAULT, NULL);
mutex_init(&sreq->sn_lock, NULL, MUTEX_DEFAULT, NULL);
return (0);
}
static void
kcf_sreq_cache_destructor(void *buf, void *cdrarg)
{
(void) cdrarg;
kcf_sreq_node_t *sreq = (kcf_sreq_node_t *)buf;
mutex_destroy(&sreq->sn_lock);
cv_destroy(&sreq->sn_cv);
}
/*
* kmem_cache_alloc constructor for async request structure.
*/
static int
kcf_areq_cache_constructor(void *buf, void *cdrarg, int kmflags)
{
(void) cdrarg, (void) kmflags;
kcf_areq_node_t *areq = (kcf_areq_node_t *)buf;
areq->an_type = CRYPTO_ASYNCH;
areq->an_refcnt = 0;
mutex_init(&areq->an_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&areq->an_done, NULL, CV_DEFAULT, NULL);
cv_init(&areq->an_turn_cv, NULL, CV_DEFAULT, NULL);
return (0);
}
static void
kcf_areq_cache_destructor(void *buf, void *cdrarg)
{
(void) cdrarg;
kcf_areq_node_t *areq = (kcf_areq_node_t *)buf;
ASSERT(areq->an_refcnt == 0);
mutex_destroy(&areq->an_lock);
cv_destroy(&areq->an_done);
cv_destroy(&areq->an_turn_cv);
}
/*
* kmem_cache_alloc constructor for kcf_context structure.
*/
static int
kcf_context_cache_constructor(void *buf, void *cdrarg, int kmflags)
{
(void) cdrarg, (void) kmflags;
kcf_context_t *kctx = (kcf_context_t *)buf;
kctx->kc_refcnt = 0;
mutex_init(&kctx->kc_in_use_lock, NULL, MUTEX_DEFAULT, NULL);
return (0);
}
static void
kcf_context_cache_destructor(void *buf, void *cdrarg)
{
(void) cdrarg;
kcf_context_t *kctx = (kcf_context_t *)buf;
ASSERT(kctx->kc_refcnt == 0);
mutex_destroy(&kctx->kc_in_use_lock);
}
void
kcf_sched_destroy(void)
{
int i;
if (kcf_misc_kstat)
kstat_delete(kcf_misc_kstat);
if (kcfpool) {
mutex_destroy(&kcfpool->kp_thread_lock);
cv_destroy(&kcfpool->kp_nothr_cv);
mutex_destroy(&kcfpool->kp_user_lock);
cv_destroy(&kcfpool->kp_user_cv);
kmem_free(kcfpool, sizeof (kcf_pool_t));
}
for (i = 0; i < REQID_TABLES; i++) {
if (kcf_reqid_table[i]) {
mutex_destroy(&(kcf_reqid_table[i]->rt_lock));
kmem_free(kcf_reqid_table[i],
sizeof (kcf_reqid_table_t));
}
}
if (gswq) {
mutex_destroy(&gswq->gs_lock);
cv_destroy(&gswq->gs_cv);
kmem_free(gswq, sizeof (kcf_global_swq_t));
}
if (kcf_context_cache)
kmem_cache_destroy(kcf_context_cache);
if (kcf_areq_cache)
kmem_cache_destroy(kcf_areq_cache);
if (kcf_sreq_cache)
kmem_cache_destroy(kcf_sreq_cache);
mutex_destroy(&ntfy_list_lock);
cv_destroy(&ntfy_list_cv);
}
/*
* Creates and initializes all the structures needed by the framework.
*/
void
kcf_sched_init(void)
{
int i;
kcf_reqid_table_t *rt;
/*
* Create all the kmem caches needed by the framework. We set the
* align argument to 64, to get a slab aligned to 64-byte as well as
* have the objects (cache_chunksize) to be a 64-byte multiple.
* This helps to avoid false sharing as this is the size of the
* CPU cache line.
*/
kcf_sreq_cache = kmem_cache_create("kcf_sreq_cache",
sizeof (struct kcf_sreq_node), 64, kcf_sreq_cache_constructor,
kcf_sreq_cache_destructor, NULL, NULL, NULL, 0);
kcf_areq_cache = kmem_cache_create("kcf_areq_cache",
sizeof (struct kcf_areq_node), 64, kcf_areq_cache_constructor,
kcf_areq_cache_destructor, NULL, NULL, NULL, 0);
kcf_context_cache = kmem_cache_create("kcf_context_cache",
sizeof (struct kcf_context), 64, kcf_context_cache_constructor,
kcf_context_cache_destructor, NULL, NULL, NULL, 0);
gswq = kmem_alloc(sizeof (kcf_global_swq_t), KM_SLEEP);
mutex_init(&gswq->gs_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&gswq->gs_cv, NULL, CV_DEFAULT, NULL);
gswq->gs_njobs = 0;
gswq->gs_maxjobs = kcf_maxthreads * crypto_taskq_maxalloc;
gswq->gs_first = gswq->gs_last = NULL;
/* Initialize the global reqid table */
for (i = 0; i < REQID_TABLES; i++) {
rt = kmem_zalloc(sizeof (kcf_reqid_table_t), KM_SLEEP);
kcf_reqid_table[i] = rt;
mutex_init(&rt->rt_lock, NULL, MUTEX_DEFAULT, NULL);
rt->rt_curid = i;
}
/* Allocate and initialize the thread pool */
kcfpool_alloc();
/* Initialize the event notification list variables */
mutex_init(&ntfy_list_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&ntfy_list_cv, NULL, CV_DEFAULT, NULL);
/* Create the kcf kstat */
kcf_misc_kstat = kstat_create("kcf", 0, "framework_stats", "crypto",
KSTAT_TYPE_NAMED, sizeof (kcf_stats_t) / sizeof (kstat_named_t),
KSTAT_FLAG_VIRTUAL);
if (kcf_misc_kstat != NULL) {
kcf_misc_kstat->ks_data = &kcf_ksdata;
kcf_misc_kstat->ks_update = kcf_misc_kstat_update;
kstat_install(kcf_misc_kstat);
}
}
/*
* Signal the waiting sync client.
*/
void
kcf_sop_done(kcf_sreq_node_t *sreq, int error)
{
mutex_enter(&sreq->sn_lock);
sreq->sn_state = REQ_DONE;
sreq->sn_rv = error;
cv_signal(&sreq->sn_cv);
mutex_exit(&sreq->sn_lock);
}
/*
* Callback the async client with the operation status.
* We free the async request node and possibly the context.
* We also handle any chain of requests hanging off of
* the context.
*/
void
kcf_aop_done(kcf_areq_node_t *areq, int error)
{
kcf_op_type_t optype;
boolean_t skip_notify = B_FALSE;
kcf_context_t *ictx;
kcf_areq_node_t *nextreq;
/*
* Handle recoverable errors. This has to be done first
* before doing anything else in this routine so that
* we do not change the state of the request.
*/
if (error != CRYPTO_SUCCESS && IS_RECOVERABLE(error)) {
/*
* We try another provider, if one is available. Else
* we continue with the failure notification to the
* client.
*/
if (kcf_resubmit_request(areq) == CRYPTO_QUEUED)
return;
}
mutex_enter(&areq->an_lock);
areq->an_state = REQ_DONE;
mutex_exit(&areq->an_lock);
optype = (&areq->an_params)->rp_optype;
if ((ictx = areq->an_context) != NULL) {
/*
* A request after it is removed from the request
* queue, still stays on a chain of requests hanging
* of its context structure. It needs to be removed
* from this chain at this point.
*/
mutex_enter(&ictx->kc_in_use_lock);
nextreq = areq->an_ctxchain_next;
if (nextreq != NULL) {
mutex_enter(&nextreq->an_lock);
nextreq->an_is_my_turn = B_TRUE;
cv_signal(&nextreq->an_turn_cv);
mutex_exit(&nextreq->an_lock);
}
ictx->kc_req_chain_first = nextreq;
if (nextreq == NULL)
ictx->kc_req_chain_last = NULL;
mutex_exit(&ictx->kc_in_use_lock);
if (IS_SINGLE_OP(optype) || IS_FINAL_OP(optype)) {
ASSERT(nextreq == NULL);
KCF_CONTEXT_REFRELE(ictx);
} else if (error != CRYPTO_SUCCESS && IS_INIT_OP(optype)) {
/*
* NOTE - We do not release the context in case of update
* operations. We require the consumer to free it explicitly,
* in case it wants to abandon an update operation. This is done
* as there may be mechanisms in ECB mode that can continue
* even if an operation on a block fails.
*/
KCF_CONTEXT_REFRELE(ictx);
}
}
/* Deal with the internal continuation to this request first */
if (areq->an_isdual) {
kcf_dual_req_t *next_arg;
next_arg = (kcf_dual_req_t *)areq->an_reqarg.cr_callback_arg;
next_arg->kr_areq = areq;
KCF_AREQ_REFHOLD(areq);
areq->an_isdual = B_FALSE;
NOTIFY_CLIENT(areq, error);
return;
}
/*
* If CRYPTO_NOTIFY_OPDONE flag is set, we should notify
* always. If this flag is clear, we skip the notification
* provided there are no errors. We check this flag for only
* init or update operations. It is ignored for single, final or
* atomic operations.
*/
skip_notify = (IS_UPDATE_OP(optype) || IS_INIT_OP(optype)) &&
(!(areq->an_reqarg.cr_flag & CRYPTO_NOTIFY_OPDONE)) &&
(error == CRYPTO_SUCCESS);
if (!skip_notify) {
NOTIFY_CLIENT(areq, error);
}
if (!(areq->an_reqarg.cr_flag & CRYPTO_SKIP_REQID))
kcf_reqid_delete(areq);
KCF_AREQ_REFRELE(areq);
}
/*
* Allocate the thread pool and initialize all the fields.
*/
static void
kcfpool_alloc()
{
kcfpool = kmem_alloc(sizeof (kcf_pool_t), KM_SLEEP);
kcfpool->kp_threads = kcfpool->kp_idlethreads = 0;
kcfpool->kp_blockedthreads = 0;
kcfpool->kp_signal_create_thread = B_FALSE;
kcfpool->kp_nthrs = 0;
kcfpool->kp_user_waiting = B_FALSE;
mutex_init(&kcfpool->kp_thread_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&kcfpool->kp_nothr_cv, NULL, CV_DEFAULT, NULL);
mutex_init(&kcfpool->kp_user_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&kcfpool->kp_user_cv, NULL, CV_DEFAULT, NULL);
-
- kcf_idlethr_timeout = KCF_DEFAULT_THRTIMEOUT;
}
/*
* Insert the async request in the hash table after assigning it
* an ID. Returns the ID.
*
* The ID is used by the caller to pass as an argument to a
* cancel_req() routine later.
*/
static crypto_req_id_t
kcf_reqid_insert(kcf_areq_node_t *areq)
{
int indx;
crypto_req_id_t id;
kcf_areq_node_t *headp;
kcf_reqid_table_t *rt;
rt = kcf_reqid_table[CPU_SEQID_UNSTABLE & REQID_TABLE_MASK];
mutex_enter(&rt->rt_lock);
rt->rt_curid = id =
(rt->rt_curid - REQID_COUNTER_LOW) | REQID_COUNTER_HIGH;
SET_REQID(areq, id);
indx = REQID_HASH(id);
headp = areq->an_idnext = rt->rt_idhash[indx];
areq->an_idprev = NULL;
if (headp != NULL)
headp->an_idprev = areq;
rt->rt_idhash[indx] = areq;
mutex_exit(&rt->rt_lock);
return (id);
}
/*
* Delete the async request from the hash table.
*/
static void
kcf_reqid_delete(kcf_areq_node_t *areq)
{
int indx;
kcf_areq_node_t *nextp, *prevp;
crypto_req_id_t id = GET_REQID(areq);
kcf_reqid_table_t *rt;
rt = kcf_reqid_table[id & REQID_TABLE_MASK];
indx = REQID_HASH(id);
mutex_enter(&rt->rt_lock);
nextp = areq->an_idnext;
prevp = areq->an_idprev;
if (nextp != NULL)
nextp->an_idprev = prevp;
if (prevp != NULL)
prevp->an_idnext = nextp;
else
rt->rt_idhash[indx] = nextp;
SET_REQID(areq, 0);
cv_broadcast(&areq->an_done);
mutex_exit(&rt->rt_lock);
}
/*
* Cancel a single asynchronous request.
*
* We guarantee that no problems will result from calling
* crypto_cancel_req() for a request which is either running, or
* has already completed. We remove the request from any queues
* if it is possible. We wait for request completion if the
* request is dispatched to a provider.
*
* Calling context:
* Can be called from user context only.
*
* NOTE: We acquire the following locks in this routine (in order):
* - rt_lock (kcf_reqid_table_t)
* - gswq->gs_lock
* - areq->an_lock
* - ictx->kc_in_use_lock (from kcf_removereq_in_ctxchain())
*
* This locking order MUST be maintained in code every where else.
*/
void
crypto_cancel_req(crypto_req_id_t id)
{
int indx;
kcf_areq_node_t *areq;
kcf_provider_desc_t *pd;
kcf_context_t *ictx;
kcf_reqid_table_t *rt;
rt = kcf_reqid_table[id & REQID_TABLE_MASK];
indx = REQID_HASH(id);
mutex_enter(&rt->rt_lock);
for (areq = rt->rt_idhash[indx]; areq; areq = areq->an_idnext) {
if (GET_REQID(areq) == id) {
/*
* We found the request. It is either still waiting
* in the framework queues or running at the provider.
*/
pd = areq->an_provider;
ASSERT(pd != NULL);
switch (pd->pd_prov_type) {
case CRYPTO_SW_PROVIDER:
mutex_enter(&gswq->gs_lock);
mutex_enter(&areq->an_lock);
/* This request can be safely canceled. */
if (areq->an_state <= REQ_WAITING) {
/* Remove from gswq, global software queue. */
kcf_remove_node(areq);
if ((ictx = areq->an_context) != NULL)
kcf_removereq_in_ctxchain(ictx, areq);
mutex_exit(&areq->an_lock);
mutex_exit(&gswq->gs_lock);
mutex_exit(&rt->rt_lock);
/* Remove areq from hash table and free it. */
kcf_reqid_delete(areq);
KCF_AREQ_REFRELE(areq);
return;
}
mutex_exit(&areq->an_lock);
mutex_exit(&gswq->gs_lock);
break;
case CRYPTO_HW_PROVIDER:
/*
* There is no interface to remove an entry
* once it is on the taskq. So, we do not do
* anything for a hardware provider.
*/
break;
default:
break;
}
/*
* The request is running. Wait for the request completion
* to notify us.
*/
KCF_AREQ_REFHOLD(areq);
while (GET_REQID(areq) == id)
cv_wait(&areq->an_done, &rt->rt_lock);
KCF_AREQ_REFRELE(areq);
break;
}
}
mutex_exit(&rt->rt_lock);
}
/*
* Cancel all asynchronous requests associated with the
* passed in crypto context and free it.
*
* A client SHOULD NOT call this routine after calling a crypto_*_final
* routine. This routine is called only during intermediate operations.
* The client should not use the crypto context after this function returns
* since we destroy it.
*
* Calling context:
* Can be called from user context only.
*/
void
crypto_cancel_ctx(crypto_context_t ctx)
{
kcf_context_t *ictx;
kcf_areq_node_t *areq;
if (ctx == NULL)
return;
ictx = (kcf_context_t *)((crypto_ctx_t *)ctx)->cc_framework_private;
mutex_enter(&ictx->kc_in_use_lock);
/* Walk the chain and cancel each request */
while ((areq = ictx->kc_req_chain_first) != NULL) {
/*
* We have to drop the lock here as we may have
* to wait for request completion. We hold the
* request before dropping the lock though, so that it
* won't be freed underneath us.
*/
KCF_AREQ_REFHOLD(areq);
mutex_exit(&ictx->kc_in_use_lock);
crypto_cancel_req(GET_REQID(areq));
KCF_AREQ_REFRELE(areq);
mutex_enter(&ictx->kc_in_use_lock);
}
mutex_exit(&ictx->kc_in_use_lock);
KCF_CONTEXT_REFRELE(ictx);
}
/*
* Update kstats.
*/
static int
kcf_misc_kstat_update(kstat_t *ksp, int rw)
{
uint_t tcnt;
kcf_stats_t *ks_data;
if (rw == KSTAT_WRITE)
return (EACCES);
ks_data = ksp->ks_data;
ks_data->ks_thrs_in_pool.value.ui32 = kcfpool->kp_threads;
/*
* The failover thread is counted in kp_idlethreads in
* some corner cases. This is done to avoid doing more checks
* when submitting a request. We account for those cases below.
*/
if ((tcnt = kcfpool->kp_idlethreads) == (kcfpool->kp_threads + 1))
tcnt--;
ks_data->ks_idle_thrs.value.ui32 = tcnt;
ks_data->ks_minthrs.value.ui32 = kcf_minthreads;
ks_data->ks_maxthrs.value.ui32 = kcf_maxthreads;
ks_data->ks_swq_njobs.value.ui32 = gswq->gs_njobs;
ks_data->ks_swq_maxjobs.value.ui32 = gswq->gs_maxjobs;
ks_data->ks_taskq_threads.value.ui32 = crypto_taskq_threads;
ks_data->ks_taskq_minalloc.value.ui32 = crypto_taskq_minalloc;
ks_data->ks_taskq_maxalloc.value.ui32 = crypto_taskq_maxalloc;
return (0);
}
/*
* Allocate and initialize a kcf_dual_req, used for saving the arguments of
* a dual operation or an atomic operation that has to be internally
* simulated with multiple single steps.
* crq determines the memory allocation flags.
*/
kcf_dual_req_t *
kcf_alloc_req(crypto_call_req_t *crq)
{
kcf_dual_req_t *kcr;
kcr = kmem_alloc(sizeof (kcf_dual_req_t), KCF_KMFLAG(crq));
if (kcr == NULL)
return (NULL);
/* Copy the whole crypto_call_req struct, as it isn't persistent */
if (crq != NULL)
kcr->kr_callreq = *crq;
else
bzero(&(kcr->kr_callreq), sizeof (crypto_call_req_t));
kcr->kr_areq = NULL;
kcr->kr_saveoffset = 0;
kcr->kr_savelen = 0;
return (kcr);
}
/*
* Callback routine for the next part of a simulated dual part.
* Schedules the next step.
*
* This routine can be called from interrupt context.
*/
void
kcf_next_req(void *next_req_arg, int status)
{
kcf_dual_req_t *next_req = (kcf_dual_req_t *)next_req_arg;
kcf_req_params_t *params = &(next_req->kr_params);
kcf_areq_node_t *areq = next_req->kr_areq;
int error = status;
kcf_provider_desc_t *pd = NULL;
crypto_dual_data_t *ct = NULL;
/* Stop the processing if an error occurred at this step */
if (error != CRYPTO_SUCCESS) {
out:
areq->an_reqarg = next_req->kr_callreq;
KCF_AREQ_REFRELE(areq);
kmem_free(next_req, sizeof (kcf_dual_req_t));
areq->an_isdual = B_FALSE;
kcf_aop_done(areq, error);
return;
}
switch (params->rp_opgrp) {
case KCF_OG_MAC: {
/*
* The next req is submitted with the same reqid as the
* first part. The consumer only got back that reqid, and
* should still be able to cancel the operation during its
* second step.
*/
kcf_mac_ops_params_t *mops = &(params->rp_u.mac_params);
crypto_ctx_template_t mac_tmpl;
kcf_mech_entry_t *me;
ct = (crypto_dual_data_t *)mops->mo_data;
mac_tmpl = (crypto_ctx_template_t)mops->mo_templ;
/* No expected recoverable failures, so no retry list */
pd = kcf_get_mech_provider(mops->mo_framework_mechtype,
&me, &error, NULL, CRYPTO_FG_MAC_ATOMIC,
(areq->an_reqarg.cr_flag & CRYPTO_RESTRICTED), ct->dd_len2);
if (pd == NULL) {
error = CRYPTO_MECH_NOT_SUPPORTED;
goto out;
}
/* Validate the MAC context template here */
if ((pd->pd_prov_type == CRYPTO_SW_PROVIDER) &&
(mac_tmpl != NULL)) {
kcf_ctx_template_t *ctx_mac_tmpl;
ctx_mac_tmpl = (kcf_ctx_template_t *)mac_tmpl;
if (ctx_mac_tmpl->ct_generation != me->me_gen_swprov) {
KCF_PROV_REFRELE(pd);
error = CRYPTO_OLD_CTX_TEMPLATE;
goto out;
}
mops->mo_templ = ctx_mac_tmpl->ct_prov_tmpl;
}
break;
}
case KCF_OG_DECRYPT: {
kcf_decrypt_ops_params_t *dcrops =
&(params->rp_u.decrypt_params);
ct = (crypto_dual_data_t *)dcrops->dop_ciphertext;
/* No expected recoverable failures, so no retry list */
pd = kcf_get_mech_provider(dcrops->dop_framework_mechtype,
NULL, &error, NULL, CRYPTO_FG_DECRYPT_ATOMIC,
(areq->an_reqarg.cr_flag & CRYPTO_RESTRICTED), ct->dd_len1);
if (pd == NULL) {
error = CRYPTO_MECH_NOT_SUPPORTED;
goto out;
}
break;
}
default:
break;
}
/* The second step uses len2 and offset2 of the dual_data */
next_req->kr_saveoffset = ct->dd_offset1;
next_req->kr_savelen = ct->dd_len1;
ct->dd_offset1 = ct->dd_offset2;
ct->dd_len1 = ct->dd_len2;
/* preserve if the caller is restricted */
if (areq->an_reqarg.cr_flag & CRYPTO_RESTRICTED) {
areq->an_reqarg.cr_flag = CRYPTO_RESTRICTED;
} else {
areq->an_reqarg.cr_flag = 0;
}
areq->an_reqarg.cr_callback_func = kcf_last_req;
areq->an_reqarg.cr_callback_arg = next_req;
areq->an_isdual = B_TRUE;
/*
* We would like to call kcf_submit_request() here. But,
* that is not possible as that routine allocates a new
* kcf_areq_node_t request structure, while we need to
* reuse the existing request structure.
*/
switch (pd->pd_prov_type) {
case CRYPTO_SW_PROVIDER:
error = common_submit_request(pd, NULL, params,
KCF_RHNDL(KM_NOSLEEP));
break;
case CRYPTO_HW_PROVIDER: {
kcf_provider_desc_t *old_pd;
taskq_t *taskq = pd->pd_sched_info.ks_taskq;
/*
* Set the params for the second step in the
* dual-ops.
*/
areq->an_params = *params;
old_pd = areq->an_provider;
KCF_PROV_REFRELE(old_pd);
KCF_PROV_REFHOLD(pd);
areq->an_provider = pd;
/*
* Note that we have to do a taskq_dispatch()
* here as we may be in interrupt context.
*/
if (taskq_dispatch(taskq, process_req_hwp, areq,
TQ_NOSLEEP) == (taskqid_t)0) {
error = CRYPTO_HOST_MEMORY;
} else {
error = CRYPTO_QUEUED;
}
break;
}
default:
break;
}
/*
* We have to release the holds on the request and the provider
* in all cases.
*/
KCF_AREQ_REFRELE(areq);
KCF_PROV_REFRELE(pd);
if (error != CRYPTO_QUEUED) {
/* restore, clean up, and invoke the client's callback */
ct->dd_offset1 = next_req->kr_saveoffset;
ct->dd_len1 = next_req->kr_savelen;
areq->an_reqarg = next_req->kr_callreq;
kmem_free(next_req, sizeof (kcf_dual_req_t));
areq->an_isdual = B_FALSE;
kcf_aop_done(areq, error);
}
}
/*
* Last part of an emulated dual operation.
* Clean up and restore ...
*/
void
kcf_last_req(void *last_req_arg, int status)
{
kcf_dual_req_t *last_req = (kcf_dual_req_t *)last_req_arg;
kcf_req_params_t *params = &(last_req->kr_params);
kcf_areq_node_t *areq = last_req->kr_areq;
crypto_dual_data_t *ct = NULL;
switch (params->rp_opgrp) {
case KCF_OG_MAC: {
kcf_mac_ops_params_t *mops = &(params->rp_u.mac_params);
ct = (crypto_dual_data_t *)mops->mo_data;
break;
}
case KCF_OG_DECRYPT: {
kcf_decrypt_ops_params_t *dcrops =
&(params->rp_u.decrypt_params);
ct = (crypto_dual_data_t *)dcrops->dop_ciphertext;
break;
}
default: {
panic("invalid kcf_op_group_t %d", (int)params->rp_opgrp);
return;
}
}
ct->dd_offset1 = last_req->kr_saveoffset;
ct->dd_len1 = last_req->kr_savelen;
/* The submitter used kcf_last_req as its callback */
if (areq == NULL) {
crypto_call_req_t *cr = &last_req->kr_callreq;
(*(cr->cr_callback_func))(cr->cr_callback_arg, status);
kmem_free(last_req, sizeof (kcf_dual_req_t));
return;
}
areq->an_reqarg = last_req->kr_callreq;
KCF_AREQ_REFRELE(areq);
kmem_free(last_req, sizeof (kcf_dual_req_t));
areq->an_isdual = B_FALSE;
kcf_aop_done(areq, status);
}
diff --git a/module/icp/include/sys/crypto/impl.h b/module/icp/include/sys/crypto/impl.h
index 8ebc4be483ba..4906549b5687 100644
--- a/module/icp/include/sys/crypto/impl.h
+++ b/module/icp/include/sys/crypto/impl.h
@@ -1,1359 +1,1347 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#ifndef _SYS_CRYPTO_IMPL_H
#define _SYS_CRYPTO_IMPL_H
/*
* Kernel Cryptographic Framework private implementation definitions.
*/
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
#include <sys/crypto/api.h>
#include <sys/crypto/spi.h>
#ifdef __cplusplus
extern "C" {
#endif
#define KCF_MODULE "kcf"
/*
* Prefixes convention: structures internal to the kernel cryptographic
* framework start with 'kcf_'. Exposed structure start with 'crypto_'.
*/
/* Provider stats. Not protected. */
typedef struct kcf_prov_stats {
kstat_named_t ps_ops_total;
kstat_named_t ps_ops_passed;
kstat_named_t ps_ops_failed;
kstat_named_t ps_ops_busy_rval;
} kcf_prov_stats_t;
/* Various kcf stats. Not protected. */
typedef struct kcf_stats {
kstat_named_t ks_thrs_in_pool;
kstat_named_t ks_idle_thrs;
kstat_named_t ks_minthrs;
kstat_named_t ks_maxthrs;
kstat_named_t ks_swq_njobs;
kstat_named_t ks_swq_maxjobs;
kstat_named_t ks_taskq_threads;
kstat_named_t ks_taskq_minalloc;
kstat_named_t ks_taskq_maxalloc;
} kcf_stats_t;
/*
* Keep all the information needed by the scheduler from
* this provider.
*/
typedef struct kcf_sched_info {
/* The number of operations dispatched. */
uint64_t ks_ndispatches;
/* The number of operations that failed. */
uint64_t ks_nfails;
/* The number of operations that returned CRYPTO_BUSY. */
uint64_t ks_nbusy_rval;
/* taskq used to dispatch crypto requests */
taskq_t *ks_taskq;
} kcf_sched_info_t;
/*
* pd_irefcnt approximates the number of inflight requests to the
* provider. Though we increment this counter during registration for
* other purposes, that base value is mostly same across all providers.
* So, it is a good measure of the load on a provider when it is not
* in a busy state. Once a provider notifies it is busy, requests
* backup in the taskq. So, we use tq_nalloc in that case which gives
* the number of task entries in the task queue. Note that we do not
* acquire any locks here as it is not critical to get the exact number
* and the lock contention may be too costly for this code path.
*/
#define KCF_PROV_LOAD(pd) ((pd)->pd_state != KCF_PROV_BUSY ? \
(pd)->pd_irefcnt : (pd)->pd_sched_info.ks_taskq->tq_nalloc)
#define KCF_PROV_INCRSTATS(pd, error) { \
(pd)->pd_sched_info.ks_ndispatches++; \
if (error == CRYPTO_BUSY) \
(pd)->pd_sched_info.ks_nbusy_rval++; \
else if (error != CRYPTO_SUCCESS && error != CRYPTO_QUEUED) \
(pd)->pd_sched_info.ks_nfails++; \
}
/*
* The following two macros should be
* #define KCF_OPS_CLASSSIZE (KCF_LAST_OPSCLASS - KCF_FIRST_OPSCLASS + 2)
* #define KCF_MAXMECHTAB KCF_MAXCIPHER
*
* However, doing that would involve reorganizing the header file a bit.
* When impl.h is broken up (bug# 4703218), this will be done. For now,
* we hardcode these values.
*/
#define KCF_OPS_CLASSSIZE 8
#define KCF_MAXMECHTAB 32
/*
* Valid values for the state of a provider. The order of
* the elements is important.
*
* Routines which get a provider or the list of providers
* should pick only those that are either in KCF_PROV_READY state
* or in KCF_PROV_BUSY state.
*/
typedef enum {
KCF_PROV_ALLOCATED = 1,
KCF_PROV_UNVERIFIED,
KCF_PROV_VERIFICATION_FAILED,
/*
* state < KCF_PROV_READY means the provider can not
* be used at all.
*/
KCF_PROV_READY,
KCF_PROV_BUSY,
/*
* state > KCF_PROV_BUSY means the provider can not
* be used for new requests.
*/
KCF_PROV_FAILED,
/*
* Threads setting the following two states should do so only
* if the current state < KCF_PROV_DISABLED.
*/
KCF_PROV_DISABLED,
KCF_PROV_REMOVED,
KCF_PROV_FREED
} kcf_prov_state_t;
#define KCF_IS_PROV_UNVERIFIED(pd) ((pd)->pd_state == KCF_PROV_UNVERIFIED)
#define KCF_IS_PROV_USABLE(pd) ((pd)->pd_state == KCF_PROV_READY || \
(pd)->pd_state == KCF_PROV_BUSY)
#define KCF_IS_PROV_REMOVED(pd) ((pd)->pd_state >= KCF_PROV_REMOVED)
/* Internal flags valid for pd_flags field */
#define KCF_PROV_RESTRICTED 0x40000000
#define KCF_LPROV_MEMBER 0x80000000 /* is member of a logical provider */
/*
* A provider descriptor structure. There is one such structure per
* provider. It is allocated and initialized at registration time and
* freed when the provider unregisters.
*
* pd_prov_type: Provider type, hardware or software
* pd_sid: Session ID of the provider used by kernel clients.
* This is valid only for session-oriented providers.
* pd_refcnt: Reference counter to this provider descriptor
* pd_irefcnt: References held by the framework internal structs
* pd_lock: lock protects pd_state and pd_provider_list
* pd_state: State value of the provider
* pd_provider_list: Used to cross-reference logical providers and their
* members. Not used for software providers.
* pd_resume_cv: cv to wait for state to change from KCF_PROV_BUSY
* pd_prov_handle: Provider handle specified by provider
* pd_ops_vector: The ops vector specified by Provider
* pd_mech_indx: Lookup table which maps a core framework mechanism
* number to an index in pd_mechanisms array
* pd_mechanisms: Array of mechanisms supported by the provider, specified
* by the provider during registration
* pd_sched_info: Scheduling information associated with the provider
* pd_mech_list_count: The number of entries in pi_mechanisms, specified
* by the provider during registration
* pd_name: Device name or module name
* pd_instance: Device instance
* pd_module_id: Module ID returned by modload
* pd_mctlp: Pointer to modctl structure for this provider
* pd_remove_cv: cv to wait on while the provider queue drains
* pd_description: Provider description string
* pd_flags bitwise OR of pi_flags from crypto_provider_info_t
* and other internal flags defined above.
* pd_hash_limit Maximum data size that hash mechanisms of this provider
* can support.
* pd_kcf_prov_handle: KCF-private handle assigned by KCF
* pd_prov_id: Identification # assigned by KCF to provider
* pd_kstat: kstat associated with the provider
* pd_ks_data: kstat data
*/
typedef struct kcf_provider_desc {
crypto_provider_type_t pd_prov_type;
crypto_session_id_t pd_sid;
uint_t pd_refcnt;
uint_t pd_irefcnt;
kmutex_t pd_lock;
kcf_prov_state_t pd_state;
struct kcf_provider_list *pd_provider_list;
kcondvar_t pd_resume_cv;
crypto_provider_handle_t pd_prov_handle;
- crypto_ops_t *pd_ops_vector;
+ const crypto_ops_t *pd_ops_vector;
ushort_t pd_mech_indx[KCF_OPS_CLASSSIZE]\
[KCF_MAXMECHTAB];
crypto_mech_info_t *pd_mechanisms;
kcf_sched_info_t pd_sched_info;
uint_t pd_mech_list_count;
// char *pd_name;
// uint_t pd_instance;
// int pd_module_id;
// struct modctl *pd_mctlp;
kcondvar_t pd_remove_cv;
char *pd_description;
uint_t pd_flags;
uint_t pd_hash_limit;
crypto_kcf_provider_handle_t pd_kcf_prov_handle;
crypto_provider_id_t pd_prov_id;
kstat_t *pd_kstat;
kcf_prov_stats_t pd_ks_data;
} kcf_provider_desc_t;
/* useful for making a list of providers */
typedef struct kcf_provider_list {
struct kcf_provider_list *pl_next;
struct kcf_provider_desc *pl_provider;
} kcf_provider_list_t;
/* atomic operations in linux implicitly form a memory barrier */
#define membar_exit()
/*
* If a component has a reference to a kcf_provider_desc_t,
* it REFHOLD()s. A new provider descriptor which is referenced only
* by the providers table has a reference counter of one.
*/
#define KCF_PROV_REFHOLD(desc) { \
atomic_add_32(&(desc)->pd_refcnt, 1); \
ASSERT((desc)->pd_refcnt != 0); \
}
#define KCF_PROV_IREFHOLD(desc) { \
atomic_add_32(&(desc)->pd_irefcnt, 1); \
ASSERT((desc)->pd_irefcnt != 0); \
}
#define KCF_PROV_IREFRELE(desc) { \
ASSERT((desc)->pd_irefcnt != 0); \
membar_exit(); \
if (atomic_add_32_nv(&(desc)->pd_irefcnt, -1) == 0) { \
cv_broadcast(&(desc)->pd_remove_cv); \
} \
}
#define KCF_PROV_REFHELD(desc) ((desc)->pd_refcnt >= 1)
#define KCF_PROV_REFRELE(desc) { \
ASSERT((desc)->pd_refcnt != 0); \
membar_exit(); \
if (atomic_add_32_nv(&(desc)->pd_refcnt, -1) == 0) { \
kcf_provider_zero_refcnt((desc)); \
} \
}
/* list of crypto_mech_info_t valid as the second mech in a dual operation */
typedef struct crypto_mech_info_list {
struct crypto_mech_info_list *ml_next;
crypto_mech_type_t ml_kcf_mechid; /* KCF's id */
crypto_mech_info_t ml_mech_info;
} crypto_mech_info_list_t;
/*
* An element in a mechanism provider descriptors chain.
* The kcf_prov_mech_desc_t is duplicated in every chain the provider belongs
* to. This is a small tradeoff memory vs mutex spinning time to access the
* common provider field.
*/
typedef struct kcf_prov_mech_desc {
struct kcf_mech_entry *pm_me; /* Back to the head */
struct kcf_prov_mech_desc *pm_next; /* Next in the chain */
crypto_mech_info_t pm_mech_info; /* Provider mech info */
crypto_mech_info_list_t *pm_mi_list; /* list for duals */
kcf_provider_desc_t *pm_prov_desc; /* Common desc. */
} kcf_prov_mech_desc_t;
/* and the notation shortcuts ... */
#define pm_provider_type pm_prov_desc.pd_provider_type
#define pm_provider_handle pm_prov_desc.pd_provider_handle
#define pm_ops_vector pm_prov_desc.pd_ops_vector
/*
* A mechanism entry in an xxx_mech_tab[]. me_pad was deemed
* to be unnecessary and removed.
*/
typedef struct kcf_mech_entry {
crypto_mech_name_t me_name; /* mechanism name */
crypto_mech_type_t me_mechid; /* Internal id for mechanism */
kmutex_t me_mutex; /* access protection */
kcf_prov_mech_desc_t *me_hw_prov_chain; /* list of HW providers */
kcf_prov_mech_desc_t *me_sw_prov; /* SW provider */
/*
* Number of HW providers in the chain. There is only one
* SW provider. So, we need only a count of HW providers.
*/
int me_num_hwprov;
/*
* When a SW provider is present, this is the generation number that
* ensures no objects from old SW providers are used in the new one
*/
uint32_t me_gen_swprov;
/*
* threshold for using hardware providers for this mech
*/
size_t me_threshold;
} kcf_mech_entry_t;
/*
* A policy descriptor structure. It is allocated and initialized
* when administrative ioctls load disabled mechanisms.
*
* pd_prov_type: Provider type, hardware or software
* pd_name: Device name or module name.
* pd_instance: Device instance.
* pd_refcnt: Reference counter for this policy descriptor
* pd_mutex: Protects array and count of disabled mechanisms.
* pd_disabled_count: Count of disabled mechanisms.
* pd_disabled_mechs: Array of disabled mechanisms.
*/
typedef struct kcf_policy_desc {
crypto_provider_type_t pd_prov_type;
char *pd_name;
uint_t pd_instance;
uint_t pd_refcnt;
kmutex_t pd_mutex;
uint_t pd_disabled_count;
crypto_mech_name_t *pd_disabled_mechs;
} kcf_policy_desc_t;
/*
* If a component has a reference to a kcf_policy_desc_t,
* it REFHOLD()s. A new policy descriptor which is referenced only
* by the policy table has a reference count of one.
*/
#define KCF_POLICY_REFHOLD(desc) { \
atomic_add_32(&(desc)->pd_refcnt, 1); \
ASSERT((desc)->pd_refcnt != 0); \
}
/*
* Releases a reference to a policy descriptor. When the last
* reference is released, the descriptor is freed.
*/
#define KCF_POLICY_REFRELE(desc) { \
ASSERT((desc)->pd_refcnt != 0); \
membar_exit(); \
if (atomic_add_32_nv(&(desc)->pd_refcnt, -1) == 0) \
kcf_policy_free_desc(desc); \
}
/*
* This entry stores the name of a software module and its
* mechanisms. The mechanisms are 'hints' that are used to
* trigger loading of the module.
*/
typedef struct kcf_soft_conf_entry {
struct kcf_soft_conf_entry *ce_next;
char *ce_name;
crypto_mech_name_t *ce_mechs;
uint_t ce_count;
} kcf_soft_conf_entry_t;
extern kmutex_t soft_config_mutex;
extern kcf_soft_conf_entry_t *soft_config_list;
/*
* Global tables. The sizes are from the predefined PKCS#11 v2.20 mechanisms,
* with a margin of few extra empty entry points
*/
#define KCF_MAXDIGEST 16 /* Digests */
#define KCF_MAXCIPHER 64 /* Ciphers */
#define KCF_MAXMAC 40 /* Message authentication codes */
#define KCF_MAXSIGN 24 /* Sign/Verify */
#define KCF_MAXKEYOPS 116 /* Key generation and derivation */
#define KCF_MAXMISC 16 /* Others ... */
-#define KCF_MAXMECHS KCF_MAXDIGEST + KCF_MAXCIPHER + KCF_MAXMAC + \
- KCF_MAXSIGN + KCF_MAXKEYOPS + \
- KCF_MAXMISC
-
-extern kcf_mech_entry_t kcf_digest_mechs_tab[];
-extern kcf_mech_entry_t kcf_cipher_mechs_tab[];
-extern kcf_mech_entry_t kcf_mac_mechs_tab[];
-extern kcf_mech_entry_t kcf_sign_mechs_tab[];
-extern kcf_mech_entry_t kcf_keyops_mechs_tab[];
-extern kcf_mech_entry_t kcf_misc_mechs_tab[];
-
-extern kmutex_t kcf_mech_tabs_lock;
-
typedef enum {
KCF_DIGEST_CLASS = 1,
KCF_CIPHER_CLASS,
KCF_MAC_CLASS,
KCF_SIGN_CLASS,
KCF_KEYOPS_CLASS,
KCF_MISC_CLASS
} kcf_ops_class_t;
#define KCF_FIRST_OPSCLASS KCF_DIGEST_CLASS
#define KCF_LAST_OPSCLASS KCF_MISC_CLASS
/* The table of all the kcf_xxx_mech_tab[]s, indexed by kcf_ops_class */
typedef struct kcf_mech_entry_tab {
int met_size; /* Size of the met_tab[] */
kcf_mech_entry_t *met_tab; /* the table */
} kcf_mech_entry_tab_t;
-extern kcf_mech_entry_tab_t kcf_mech_tabs_tab[];
+extern const kcf_mech_entry_tab_t kcf_mech_tabs_tab[];
#define KCF_MECHID(class, index) \
(((crypto_mech_type_t)(class) << 32) | (crypto_mech_type_t)(index))
#define KCF_MECH2CLASS(mech_type) ((kcf_ops_class_t)((mech_type) >> 32))
#define KCF_MECH2INDEX(mech_type) ((int)(mech_type))
#define KCF_TO_PROV_MECH_INDX(pd, mech_type) \
((pd)->pd_mech_indx[KCF_MECH2CLASS(mech_type)] \
[KCF_MECH2INDEX(mech_type)])
#define KCF_TO_PROV_MECHINFO(pd, mech_type) \
((pd)->pd_mechanisms[KCF_TO_PROV_MECH_INDX(pd, mech_type)])
#define KCF_TO_PROV_MECHNUM(pd, mech_type) \
(KCF_TO_PROV_MECHINFO(pd, mech_type).cm_mech_number)
#define KCF_CAN_SHARE_OPSTATE(pd, mech_type) \
((KCF_TO_PROV_MECHINFO(pd, mech_type).cm_mech_flags) & \
CRYPTO_CAN_SHARE_OPSTATE)
/* ps_refcnt is protected by cm_lock in the crypto_minor structure */
typedef struct crypto_provider_session {
struct crypto_provider_session *ps_next;
crypto_session_id_t ps_session;
kcf_provider_desc_t *ps_provider;
kcf_provider_desc_t *ps_real_provider;
uint_t ps_refcnt;
} crypto_provider_session_t;
typedef struct crypto_session_data {
kmutex_t sd_lock;
kcondvar_t sd_cv;
uint32_t sd_flags;
int sd_pre_approved_amount;
crypto_ctx_t *sd_digest_ctx;
crypto_ctx_t *sd_encr_ctx;
crypto_ctx_t *sd_decr_ctx;
crypto_ctx_t *sd_sign_ctx;
crypto_ctx_t *sd_verify_ctx;
crypto_ctx_t *sd_sign_recover_ctx;
crypto_ctx_t *sd_verify_recover_ctx;
kcf_provider_desc_t *sd_provider;
void *sd_find_init_cookie;
crypto_provider_session_t *sd_provider_session;
} crypto_session_data_t;
#define CRYPTO_SESSION_IN_USE 0x00000001
#define CRYPTO_SESSION_IS_BUSY 0x00000002
#define CRYPTO_SESSION_IS_CLOSED 0x00000004
#define KCF_MAX_PIN_LEN 1024
/*
* Per-minor info.
*
* cm_lock protects everything in this structure except for cm_refcnt.
*/
typedef struct crypto_minor {
uint_t cm_refcnt;
kmutex_t cm_lock;
kcondvar_t cm_cv;
crypto_session_data_t **cm_session_table;
uint_t cm_session_table_count;
kcf_provider_desc_t **cm_provider_array;
uint_t cm_provider_count;
crypto_provider_session_t *cm_provider_session;
} crypto_minor_t;
/*
* Return codes for internal functions
*/
#define KCF_SUCCESS 0x0 /* Successful call */
#define KCF_INVALID_MECH_NUMBER 0x1 /* invalid mechanism number */
#define KCF_INVALID_MECH_NAME 0x2 /* invalid mechanism name */
#define KCF_INVALID_MECH_CLASS 0x3 /* invalid mechanism class */
#define KCF_MECH_TAB_FULL 0x4 /* Need more room in the mech tabs. */
#define KCF_INVALID_INDX ((ushort_t)-1)
/*
* kCF internal mechanism and function group for tracking RNG providers.
*/
#define SUN_RANDOM "random"
#define CRYPTO_FG_RANDOM 0x80000000 /* generate_random() */
/*
* Wrappers for ops vectors. In the wrapper definitions below, the pd
* argument always corresponds to a pointer to a provider descriptor
* of type kcf_prov_desc_t.
*/
#define KCF_PROV_CONTROL_OPS(pd) ((pd)->pd_ops_vector->co_control_ops)
#define KCF_PROV_CTX_OPS(pd) ((pd)->pd_ops_vector->co_ctx_ops)
#define KCF_PROV_DIGEST_OPS(pd) ((pd)->pd_ops_vector->co_digest_ops)
#define KCF_PROV_CIPHER_OPS(pd) ((pd)->pd_ops_vector->co_cipher_ops)
#define KCF_PROV_MAC_OPS(pd) ((pd)->pd_ops_vector->co_mac_ops)
#define KCF_PROV_SIGN_OPS(pd) ((pd)->pd_ops_vector->co_sign_ops)
#define KCF_PROV_VERIFY_OPS(pd) ((pd)->pd_ops_vector->co_verify_ops)
#define KCF_PROV_DUAL_OPS(pd) ((pd)->pd_ops_vector->co_dual_ops)
#define KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) \
((pd)->pd_ops_vector->co_dual_cipher_mac_ops)
#define KCF_PROV_RANDOM_OPS(pd) ((pd)->pd_ops_vector->co_random_ops)
#define KCF_PROV_SESSION_OPS(pd) ((pd)->pd_ops_vector->co_session_ops)
#define KCF_PROV_OBJECT_OPS(pd) ((pd)->pd_ops_vector->co_object_ops)
#define KCF_PROV_KEY_OPS(pd) ((pd)->pd_ops_vector->co_key_ops)
#define KCF_PROV_PROVIDER_OPS(pd) ((pd)->pd_ops_vector->co_provider_ops)
#define KCF_PROV_MECH_OPS(pd) ((pd)->pd_ops_vector->co_mech_ops)
#define KCF_PROV_NOSTORE_KEY_OPS(pd) \
((pd)->pd_ops_vector->co_nostore_key_ops)
/*
* Wrappers for crypto_control_ops(9S) entry points.
*/
#define KCF_PROV_STATUS(pd, status) ( \
(KCF_PROV_CONTROL_OPS(pd) && \
KCF_PROV_CONTROL_OPS(pd)->provider_status) ? \
KCF_PROV_CONTROL_OPS(pd)->provider_status( \
(pd)->pd_prov_handle, status) : \
CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_ctx_ops(9S) entry points.
*/
#define KCF_PROV_CREATE_CTX_TEMPLATE(pd, mech, key, template, size, req) ( \
(KCF_PROV_CTX_OPS(pd) && KCF_PROV_CTX_OPS(pd)->create_ctx_template) ? \
KCF_PROV_CTX_OPS(pd)->create_ctx_template( \
(pd)->pd_prov_handle, mech, key, template, size, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_FREE_CONTEXT(pd, ctx) ( \
(KCF_PROV_CTX_OPS(pd) && KCF_PROV_CTX_OPS(pd)->free_context) ? \
KCF_PROV_CTX_OPS(pd)->free_context(ctx) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_COPYIN_MECH(pd, umech, kmech, errorp, mode) ( \
(KCF_PROV_MECH_OPS(pd) && KCF_PROV_MECH_OPS(pd)->copyin_mechanism) ? \
KCF_PROV_MECH_OPS(pd)->copyin_mechanism( \
(pd)->pd_prov_handle, umech, kmech, errorp, mode) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_COPYOUT_MECH(pd, kmech, umech, errorp, mode) ( \
(KCF_PROV_MECH_OPS(pd) && KCF_PROV_MECH_OPS(pd)->copyout_mechanism) ? \
KCF_PROV_MECH_OPS(pd)->copyout_mechanism( \
(pd)->pd_prov_handle, kmech, umech, errorp, mode) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_FREE_MECH(pd, prov_mech) ( \
(KCF_PROV_MECH_OPS(pd) && KCF_PROV_MECH_OPS(pd)->free_mechanism) ? \
KCF_PROV_MECH_OPS(pd)->free_mechanism( \
(pd)->pd_prov_handle, prov_mech) : CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_digest_ops(9S) entry points.
*/
#define KCF_PROV_DIGEST_INIT(pd, ctx, mech, req) ( \
(KCF_PROV_DIGEST_OPS(pd) && KCF_PROV_DIGEST_OPS(pd)->digest_init) ? \
KCF_PROV_DIGEST_OPS(pd)->digest_init(ctx, mech, req) : \
CRYPTO_NOT_SUPPORTED)
/*
* The _ (underscore) in _digest is needed to avoid replacing the
* function digest().
*/
#define KCF_PROV_DIGEST(pd, ctx, data, _digest, req) ( \
(KCF_PROV_DIGEST_OPS(pd) && KCF_PROV_DIGEST_OPS(pd)->digest) ? \
KCF_PROV_DIGEST_OPS(pd)->digest(ctx, data, _digest, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DIGEST_UPDATE(pd, ctx, data, req) ( \
(KCF_PROV_DIGEST_OPS(pd) && KCF_PROV_DIGEST_OPS(pd)->digest_update) ? \
KCF_PROV_DIGEST_OPS(pd)->digest_update(ctx, data, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DIGEST_KEY(pd, ctx, key, req) ( \
(KCF_PROV_DIGEST_OPS(pd) && KCF_PROV_DIGEST_OPS(pd)->digest_key) ? \
KCF_PROV_DIGEST_OPS(pd)->digest_key(ctx, key, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DIGEST_FINAL(pd, ctx, digest, req) ( \
(KCF_PROV_DIGEST_OPS(pd) && KCF_PROV_DIGEST_OPS(pd)->digest_final) ? \
KCF_PROV_DIGEST_OPS(pd)->digest_final(ctx, digest, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DIGEST_ATOMIC(pd, session, mech, data, digest, req) ( \
(KCF_PROV_DIGEST_OPS(pd) && KCF_PROV_DIGEST_OPS(pd)->digest_atomic) ? \
KCF_PROV_DIGEST_OPS(pd)->digest_atomic( \
(pd)->pd_prov_handle, session, mech, data, digest, req) : \
CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_cipher_ops(9S) entry points.
*/
#define KCF_PROV_ENCRYPT_INIT(pd, ctx, mech, key, template, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->encrypt_init) ? \
KCF_PROV_CIPHER_OPS(pd)->encrypt_init(ctx, mech, key, template, \
req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_ENCRYPT(pd, ctx, plaintext, ciphertext, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->encrypt) ? \
KCF_PROV_CIPHER_OPS(pd)->encrypt(ctx, plaintext, ciphertext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_ENCRYPT_UPDATE(pd, ctx, plaintext, ciphertext, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->encrypt_update) ? \
KCF_PROV_CIPHER_OPS(pd)->encrypt_update(ctx, plaintext, \
ciphertext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_ENCRYPT_FINAL(pd, ctx, ciphertext, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->encrypt_final) ? \
KCF_PROV_CIPHER_OPS(pd)->encrypt_final(ctx, ciphertext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_ENCRYPT_ATOMIC(pd, session, mech, key, plaintext, ciphertext, \
template, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->encrypt_atomic) ? \
KCF_PROV_CIPHER_OPS(pd)->encrypt_atomic( \
(pd)->pd_prov_handle, session, mech, key, plaintext, ciphertext, \
template, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DECRYPT_INIT(pd, ctx, mech, key, template, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->decrypt_init) ? \
KCF_PROV_CIPHER_OPS(pd)->decrypt_init(ctx, mech, key, template, \
req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DECRYPT(pd, ctx, ciphertext, plaintext, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->decrypt) ? \
KCF_PROV_CIPHER_OPS(pd)->decrypt(ctx, ciphertext, plaintext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DECRYPT_UPDATE(pd, ctx, ciphertext, plaintext, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->decrypt_update) ? \
KCF_PROV_CIPHER_OPS(pd)->decrypt_update(ctx, ciphertext, \
plaintext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DECRYPT_FINAL(pd, ctx, plaintext, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->decrypt_final) ? \
KCF_PROV_CIPHER_OPS(pd)->decrypt_final(ctx, plaintext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DECRYPT_ATOMIC(pd, session, mech, key, ciphertext, plaintext, \
template, req) ( \
(KCF_PROV_CIPHER_OPS(pd) && KCF_PROV_CIPHER_OPS(pd)->decrypt_atomic) ? \
KCF_PROV_CIPHER_OPS(pd)->decrypt_atomic( \
(pd)->pd_prov_handle, session, mech, key, ciphertext, plaintext, \
template, req) : \
CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_mac_ops(9S) entry points.
*/
#define KCF_PROV_MAC_INIT(pd, ctx, mech, key, template, req) ( \
(KCF_PROV_MAC_OPS(pd) && KCF_PROV_MAC_OPS(pd)->mac_init) ? \
KCF_PROV_MAC_OPS(pd)->mac_init(ctx, mech, key, template, req) \
: CRYPTO_NOT_SUPPORTED)
/*
* The _ (underscore) in _mac is needed to avoid replacing the
* function mac().
*/
#define KCF_PROV_MAC(pd, ctx, data, _mac, req) ( \
(KCF_PROV_MAC_OPS(pd) && KCF_PROV_MAC_OPS(pd)->mac) ? \
KCF_PROV_MAC_OPS(pd)->mac(ctx, data, _mac, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_UPDATE(pd, ctx, data, req) ( \
(KCF_PROV_MAC_OPS(pd) && KCF_PROV_MAC_OPS(pd)->mac_update) ? \
KCF_PROV_MAC_OPS(pd)->mac_update(ctx, data, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_FINAL(pd, ctx, mac, req) ( \
(KCF_PROV_MAC_OPS(pd) && KCF_PROV_MAC_OPS(pd)->mac_final) ? \
KCF_PROV_MAC_OPS(pd)->mac_final(ctx, mac, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_ATOMIC(pd, session, mech, key, data, mac, template, \
req) ( \
(KCF_PROV_MAC_OPS(pd) && KCF_PROV_MAC_OPS(pd)->mac_atomic) ? \
KCF_PROV_MAC_OPS(pd)->mac_atomic( \
(pd)->pd_prov_handle, session, mech, key, data, mac, template, \
req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_VERIFY_ATOMIC(pd, session, mech, key, data, mac, \
template, req) ( \
(KCF_PROV_MAC_OPS(pd) && KCF_PROV_MAC_OPS(pd)->mac_verify_atomic) ? \
KCF_PROV_MAC_OPS(pd)->mac_verify_atomic( \
(pd)->pd_prov_handle, session, mech, key, data, mac, template, \
req) : \
CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_sign_ops(9S) entry points.
*/
#define KCF_PROV_SIGN_INIT(pd, ctx, mech, key, template, req) ( \
(KCF_PROV_SIGN_OPS(pd) && KCF_PROV_SIGN_OPS(pd)->sign_init) ? \
KCF_PROV_SIGN_OPS(pd)->sign_init( \
ctx, mech, key, template, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SIGN(pd, ctx, data, sig, req) ( \
(KCF_PROV_SIGN_OPS(pd) && KCF_PROV_SIGN_OPS(pd)->sign) ? \
KCF_PROV_SIGN_OPS(pd)->sign(ctx, data, sig, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SIGN_UPDATE(pd, ctx, data, req) ( \
(KCF_PROV_SIGN_OPS(pd) && KCF_PROV_SIGN_OPS(pd)->sign_update) ? \
KCF_PROV_SIGN_OPS(pd)->sign_update(ctx, data, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SIGN_FINAL(pd, ctx, sig, req) ( \
(KCF_PROV_SIGN_OPS(pd) && KCF_PROV_SIGN_OPS(pd)->sign_final) ? \
KCF_PROV_SIGN_OPS(pd)->sign_final(ctx, sig, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SIGN_ATOMIC(pd, session, mech, key, data, template, \
sig, req) ( \
(KCF_PROV_SIGN_OPS(pd) && KCF_PROV_SIGN_OPS(pd)->sign_atomic) ? \
KCF_PROV_SIGN_OPS(pd)->sign_atomic( \
(pd)->pd_prov_handle, session, mech, key, data, sig, template, \
req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SIGN_RECOVER_INIT(pd, ctx, mech, key, template, \
req) ( \
(KCF_PROV_SIGN_OPS(pd) && KCF_PROV_SIGN_OPS(pd)->sign_recover_init) ? \
KCF_PROV_SIGN_OPS(pd)->sign_recover_init(ctx, mech, key, template, \
req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SIGN_RECOVER(pd, ctx, data, sig, req) ( \
(KCF_PROV_SIGN_OPS(pd) && KCF_PROV_SIGN_OPS(pd)->sign_recover) ? \
KCF_PROV_SIGN_OPS(pd)->sign_recover(ctx, data, sig, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SIGN_RECOVER_ATOMIC(pd, session, mech, key, data, template, \
sig, req) ( \
(KCF_PROV_SIGN_OPS(pd) && \
KCF_PROV_SIGN_OPS(pd)->sign_recover_atomic) ? \
KCF_PROV_SIGN_OPS(pd)->sign_recover_atomic( \
(pd)->pd_prov_handle, session, mech, key, data, sig, template, \
req) : CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_verify_ops(9S) entry points.
*/
#define KCF_PROV_VERIFY_INIT(pd, ctx, mech, key, template, req) ( \
(KCF_PROV_VERIFY_OPS(pd) && KCF_PROV_VERIFY_OPS(pd)->verify_init) ? \
KCF_PROV_VERIFY_OPS(pd)->verify_init(ctx, mech, key, template, \
req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_VERIFY(pd, ctx, data, sig, req) ( \
(KCF_PROV_VERIFY_OPS(pd) && KCF_PROV_VERIFY_OPS(pd)->do_verify) ? \
KCF_PROV_VERIFY_OPS(pd)->do_verify(ctx, data, sig, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_VERIFY_UPDATE(pd, ctx, data, req) ( \
(KCF_PROV_VERIFY_OPS(pd) && KCF_PROV_VERIFY_OPS(pd)->verify_update) ? \
KCF_PROV_VERIFY_OPS(pd)->verify_update(ctx, data, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_VERIFY_FINAL(pd, ctx, sig, req) ( \
(KCF_PROV_VERIFY_OPS(pd) && KCF_PROV_VERIFY_OPS(pd)->verify_final) ? \
KCF_PROV_VERIFY_OPS(pd)->verify_final(ctx, sig, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_VERIFY_ATOMIC(pd, session, mech, key, data, template, sig, \
req) ( \
(KCF_PROV_VERIFY_OPS(pd) && KCF_PROV_VERIFY_OPS(pd)->verify_atomic) ? \
KCF_PROV_VERIFY_OPS(pd)->verify_atomic( \
(pd)->pd_prov_handle, session, mech, key, data, sig, template, \
req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_VERIFY_RECOVER_INIT(pd, ctx, mech, key, template, \
req) ( \
(KCF_PROV_VERIFY_OPS(pd) && \
KCF_PROV_VERIFY_OPS(pd)->verify_recover_init) ? \
KCF_PROV_VERIFY_OPS(pd)->verify_recover_init(ctx, mech, key, \
template, req) : CRYPTO_NOT_SUPPORTED)
/* verify_recover() CSPI routine has different argument order than verify() */
#define KCF_PROV_VERIFY_RECOVER(pd, ctx, sig, data, req) ( \
(KCF_PROV_VERIFY_OPS(pd) && KCF_PROV_VERIFY_OPS(pd)->verify_recover) ? \
KCF_PROV_VERIFY_OPS(pd)->verify_recover(ctx, sig, data, req) : \
CRYPTO_NOT_SUPPORTED)
/*
* verify_recover_atomic() CSPI routine has different argument order
* than verify_atomic().
*/
#define KCF_PROV_VERIFY_RECOVER_ATOMIC(pd, session, mech, key, sig, \
template, data, req) ( \
(KCF_PROV_VERIFY_OPS(pd) && \
KCF_PROV_VERIFY_OPS(pd)->verify_recover_atomic) ? \
KCF_PROV_VERIFY_OPS(pd)->verify_recover_atomic( \
(pd)->pd_prov_handle, session, mech, key, sig, data, template, \
req) : CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_dual_ops(9S) entry points.
*/
#define KCF_PROV_DIGEST_ENCRYPT_UPDATE(digest_ctx, encrypt_ctx, plaintext, \
ciphertext, req) ( \
(KCF_PROV_DUAL_OPS(pd) && \
KCF_PROV_DUAL_OPS(pd)->digest_encrypt_update) ? \
KCF_PROV_DUAL_OPS(pd)->digest_encrypt_update( \
digest_ctx, encrypt_ctx, plaintext, ciphertext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DECRYPT_DIGEST_UPDATE(decrypt_ctx, digest_ctx, ciphertext, \
plaintext, req) ( \
(KCF_PROV_DUAL_OPS(pd) && \
KCF_PROV_DUAL_OPS(pd)->decrypt_digest_update) ? \
KCF_PROV_DUAL_OPS(pd)->decrypt_digest_update( \
decrypt_ctx, digest_ctx, ciphertext, plaintext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SIGN_ENCRYPT_UPDATE(sign_ctx, encrypt_ctx, plaintext, \
ciphertext, req) ( \
(KCF_PROV_DUAL_OPS(pd) && \
KCF_PROV_DUAL_OPS(pd)->sign_encrypt_update) ? \
KCF_PROV_DUAL_OPS(pd)->sign_encrypt_update( \
sign_ctx, encrypt_ctx, plaintext, ciphertext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_DECRYPT_VERIFY_UPDATE(decrypt_ctx, verify_ctx, ciphertext, \
plaintext, req) ( \
(KCF_PROV_DUAL_OPS(pd) && \
KCF_PROV_DUAL_OPS(pd)->decrypt_verify_update) ? \
KCF_PROV_DUAL_OPS(pd)->decrypt_verify_update( \
decrypt_ctx, verify_ctx, ciphertext, plaintext, req) : \
CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_dual_cipher_mac_ops(9S) entry points.
*/
#define KCF_PROV_ENCRYPT_MAC_INIT(pd, ctx, encr_mech, encr_key, mac_mech, \
mac_key, encr_ctx_template, mac_ctx_template, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac_init) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac_init( \
ctx, encr_mech, encr_key, mac_mech, mac_key, encr_ctx_template, \
mac_ctx_template, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_ENCRYPT_MAC(pd, ctx, plaintext, ciphertext, mac, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac( \
ctx, plaintext, ciphertext, mac, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_ENCRYPT_MAC_UPDATE(pd, ctx, plaintext, ciphertext, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac_update) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac_update( \
ctx, plaintext, ciphertext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_ENCRYPT_MAC_FINAL(pd, ctx, ciphertext, mac, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac_final) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac_final( \
ctx, ciphertext, mac, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_ENCRYPT_MAC_ATOMIC(pd, session, encr_mech, encr_key, \
mac_mech, mac_key, plaintext, ciphertext, mac, \
encr_ctx_template, mac_ctx_template, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac_atomic) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->encrypt_mac_atomic( \
(pd)->pd_prov_handle, session, encr_mech, encr_key, \
mac_mech, mac_key, plaintext, ciphertext, mac, \
encr_ctx_template, mac_ctx_template, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_DECRYPT_INIT(pd, ctx, mac_mech, mac_key, decr_mech, \
decr_key, mac_ctx_template, decr_ctx_template, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt_init) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt_init( \
ctx, mac_mech, mac_key, decr_mech, decr_key, mac_ctx_template, \
decr_ctx_template, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_DECRYPT(pd, ctx, ciphertext, mac, plaintext, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt( \
ctx, ciphertext, mac, plaintext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_DECRYPT_UPDATE(pd, ctx, ciphertext, plaintext, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt_update) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt_update( \
ctx, ciphertext, plaintext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_DECRYPT_FINAL(pd, ctx, mac, plaintext, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt_final) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt_final( \
ctx, mac, plaintext, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_DECRYPT_ATOMIC(pd, session, mac_mech, mac_key, \
decr_mech, decr_key, ciphertext, mac, plaintext, \
mac_ctx_template, decr_ctx_template, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt_atomic) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_decrypt_atomic( \
(pd)->pd_prov_handle, session, mac_mech, mac_key, \
decr_mech, decr_key, ciphertext, mac, plaintext, \
mac_ctx_template, decr_ctx_template, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_MAC_VERIFY_DECRYPT_ATOMIC(pd, session, mac_mech, mac_key, \
decr_mech, decr_key, ciphertext, mac, plaintext, \
mac_ctx_template, decr_ctx_template, req) ( \
(KCF_PROV_DUAL_CIPHER_MAC_OPS(pd) && \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_verify_decrypt_atomic \
!= NULL) ? \
KCF_PROV_DUAL_CIPHER_MAC_OPS(pd)->mac_verify_decrypt_atomic( \
(pd)->pd_prov_handle, session, mac_mech, mac_key, \
decr_mech, decr_key, ciphertext, mac, plaintext, \
mac_ctx_template, decr_ctx_template, req) : \
CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_random_number_ops(9S) entry points.
*/
#define KCF_PROV_SEED_RANDOM(pd, session, buf, len, est, flags, req) ( \
(KCF_PROV_RANDOM_OPS(pd) && KCF_PROV_RANDOM_OPS(pd)->seed_random) ? \
KCF_PROV_RANDOM_OPS(pd)->seed_random((pd)->pd_prov_handle, \
session, buf, len, est, flags, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_GENERATE_RANDOM(pd, session, buf, len, req) ( \
(KCF_PROV_RANDOM_OPS(pd) && \
KCF_PROV_RANDOM_OPS(pd)->generate_random) ? \
KCF_PROV_RANDOM_OPS(pd)->generate_random((pd)->pd_prov_handle, \
session, buf, len, req) : CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_session_ops(9S) entry points.
*
* ops_pd is the provider descriptor that supplies the ops_vector.
* pd is the descriptor that supplies the provider handle.
* Only session open/close needs two handles.
*/
#define KCF_PROV_SESSION_OPEN(ops_pd, session, req, pd) ( \
(KCF_PROV_SESSION_OPS(ops_pd) && \
KCF_PROV_SESSION_OPS(ops_pd)->session_open) ? \
KCF_PROV_SESSION_OPS(ops_pd)->session_open((pd)->pd_prov_handle, \
session, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SESSION_CLOSE(ops_pd, session, req, pd) ( \
(KCF_PROV_SESSION_OPS(ops_pd) && \
KCF_PROV_SESSION_OPS(ops_pd)->session_close) ? \
KCF_PROV_SESSION_OPS(ops_pd)->session_close((pd)->pd_prov_handle, \
session, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SESSION_LOGIN(pd, session, user_type, pin, len, req) ( \
(KCF_PROV_SESSION_OPS(pd) && \
KCF_PROV_SESSION_OPS(pd)->session_login) ? \
KCF_PROV_SESSION_OPS(pd)->session_login((pd)->pd_prov_handle, \
session, user_type, pin, len, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SESSION_LOGOUT(pd, session, req) ( \
(KCF_PROV_SESSION_OPS(pd) && \
KCF_PROV_SESSION_OPS(pd)->session_logout) ? \
KCF_PROV_SESSION_OPS(pd)->session_logout((pd)->pd_prov_handle, \
session, req) : CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_object_ops(9S) entry points.
*/
#define KCF_PROV_OBJECT_CREATE(pd, session, template, count, object, req) ( \
(KCF_PROV_OBJECT_OPS(pd) && KCF_PROV_OBJECT_OPS(pd)->object_create) ? \
KCF_PROV_OBJECT_OPS(pd)->object_create((pd)->pd_prov_handle, \
session, template, count, object, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_OBJECT_COPY(pd, session, object, template, count, \
new_object, req) ( \
(KCF_PROV_OBJECT_OPS(pd) && KCF_PROV_OBJECT_OPS(pd)->object_copy) ? \
KCF_PROV_OBJECT_OPS(pd)->object_copy((pd)->pd_prov_handle, \
session, object, template, count, new_object, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_OBJECT_DESTROY(pd, session, object, req) ( \
(KCF_PROV_OBJECT_OPS(pd) && KCF_PROV_OBJECT_OPS(pd)->object_destroy) ? \
KCF_PROV_OBJECT_OPS(pd)->object_destroy((pd)->pd_prov_handle, \
session, object, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_OBJECT_GET_SIZE(pd, session, object, size, req) ( \
(KCF_PROV_OBJECT_OPS(pd) && \
KCF_PROV_OBJECT_OPS(pd)->object_get_size) ? \
KCF_PROV_OBJECT_OPS(pd)->object_get_size((pd)->pd_prov_handle, \
session, object, size, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_OBJECT_GET_ATTRIBUTE_VALUE(pd, session, object, template, \
count, req) ( \
(KCF_PROV_OBJECT_OPS(pd) && \
KCF_PROV_OBJECT_OPS(pd)->object_get_attribute_value) ? \
KCF_PROV_OBJECT_OPS(pd)->object_get_attribute_value( \
(pd)->pd_prov_handle, session, object, template, count, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_OBJECT_SET_ATTRIBUTE_VALUE(pd, session, object, template, \
count, req) ( \
(KCF_PROV_OBJECT_OPS(pd) && \
KCF_PROV_OBJECT_OPS(pd)->object_set_attribute_value) ? \
KCF_PROV_OBJECT_OPS(pd)->object_set_attribute_value( \
(pd)->pd_prov_handle, session, object, template, count, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_OBJECT_FIND_INIT(pd, session, template, count, ppriv, \
req) ( \
(KCF_PROV_OBJECT_OPS(pd) && \
KCF_PROV_OBJECT_OPS(pd)->object_find_init) ? \
KCF_PROV_OBJECT_OPS(pd)->object_find_init((pd)->pd_prov_handle, \
session, template, count, ppriv, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_OBJECT_FIND(pd, ppriv, objects, max_objects, object_count, \
req) ( \
(KCF_PROV_OBJECT_OPS(pd) && KCF_PROV_OBJECT_OPS(pd)->object_find) ? \
KCF_PROV_OBJECT_OPS(pd)->object_find( \
(pd)->pd_prov_handle, ppriv, objects, max_objects, object_count, \
req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_OBJECT_FIND_FINAL(pd, ppriv, req) ( \
(KCF_PROV_OBJECT_OPS(pd) && \
KCF_PROV_OBJECT_OPS(pd)->object_find_final) ? \
KCF_PROV_OBJECT_OPS(pd)->object_find_final( \
(pd)->pd_prov_handle, ppriv, req) : CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_key_ops(9S) entry points.
*/
#define KCF_PROV_KEY_GENERATE(pd, session, mech, template, count, object, \
req) ( \
(KCF_PROV_KEY_OPS(pd) && KCF_PROV_KEY_OPS(pd)->key_generate) ? \
KCF_PROV_KEY_OPS(pd)->key_generate((pd)->pd_prov_handle, \
session, mech, template, count, object, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_KEY_GENERATE_PAIR(pd, session, mech, pub_template, \
pub_count, priv_template, priv_count, pub_key, priv_key, req) ( \
(KCF_PROV_KEY_OPS(pd) && KCF_PROV_KEY_OPS(pd)->key_generate_pair) ? \
KCF_PROV_KEY_OPS(pd)->key_generate_pair((pd)->pd_prov_handle, \
session, mech, pub_template, pub_count, priv_template, \
priv_count, pub_key, priv_key, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_KEY_WRAP(pd, session, mech, wrapping_key, key, wrapped_key, \
wrapped_key_len, req) ( \
(KCF_PROV_KEY_OPS(pd) && KCF_PROV_KEY_OPS(pd)->key_wrap) ? \
KCF_PROV_KEY_OPS(pd)->key_wrap((pd)->pd_prov_handle, \
session, mech, wrapping_key, key, wrapped_key, wrapped_key_len, \
req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_KEY_UNWRAP(pd, session, mech, unwrapping_key, wrapped_key, \
wrapped_key_len, template, count, key, req) ( \
(KCF_PROV_KEY_OPS(pd) && KCF_PROV_KEY_OPS(pd)->key_unwrap) ? \
KCF_PROV_KEY_OPS(pd)->key_unwrap((pd)->pd_prov_handle, \
session, mech, unwrapping_key, wrapped_key, wrapped_key_len, \
template, count, key, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_KEY_DERIVE(pd, session, mech, base_key, template, count, \
key, req) ( \
(KCF_PROV_KEY_OPS(pd) && KCF_PROV_KEY_OPS(pd)->key_derive) ? \
KCF_PROV_KEY_OPS(pd)->key_derive((pd)->pd_prov_handle, \
session, mech, base_key, template, count, key, req) : \
CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_KEY_CHECK(pd, mech, key) ( \
(KCF_PROV_KEY_OPS(pd) && KCF_PROV_KEY_OPS(pd)->key_check) ? \
KCF_PROV_KEY_OPS(pd)->key_check((pd)->pd_prov_handle, mech, key) : \
CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_provider_management_ops(9S) entry points.
*
* ops_pd is the provider descriptor that supplies the ops_vector.
* pd is the descriptor that supplies the provider handle.
* Only ext_info needs two handles.
*/
#define KCF_PROV_EXT_INFO(ops_pd, provext_info, req, pd) ( \
(KCF_PROV_PROVIDER_OPS(ops_pd) && \
KCF_PROV_PROVIDER_OPS(ops_pd)->ext_info) ? \
KCF_PROV_PROVIDER_OPS(ops_pd)->ext_info((pd)->pd_prov_handle, \
provext_info, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_INIT_TOKEN(pd, pin, pin_len, label, req) ( \
(KCF_PROV_PROVIDER_OPS(pd) && KCF_PROV_PROVIDER_OPS(pd)->init_token) ? \
KCF_PROV_PROVIDER_OPS(pd)->init_token((pd)->pd_prov_handle, \
pin, pin_len, label, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_INIT_PIN(pd, session, pin, pin_len, req) ( \
(KCF_PROV_PROVIDER_OPS(pd) && KCF_PROV_PROVIDER_OPS(pd)->init_pin) ? \
KCF_PROV_PROVIDER_OPS(pd)->init_pin((pd)->pd_prov_handle, \
session, pin, pin_len, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_SET_PIN(pd, session, old_pin, old_len, new_pin, new_len, \
req) ( \
(KCF_PROV_PROVIDER_OPS(pd) && KCF_PROV_PROVIDER_OPS(pd)->set_pin) ? \
KCF_PROV_PROVIDER_OPS(pd)->set_pin((pd)->pd_prov_handle, \
session, old_pin, old_len, new_pin, new_len, req) : \
CRYPTO_NOT_SUPPORTED)
/*
* Wrappers for crypto_nostore_key_ops(9S) entry points.
*/
#define KCF_PROV_NOSTORE_KEY_GENERATE(pd, session, mech, template, count, \
out_template, out_count, req) ( \
(KCF_PROV_NOSTORE_KEY_OPS(pd) && \
KCF_PROV_NOSTORE_KEY_OPS(pd)->nostore_key_generate) ? \
KCF_PROV_NOSTORE_KEY_OPS(pd)->nostore_key_generate( \
(pd)->pd_prov_handle, session, mech, template, count, \
out_template, out_count, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_NOSTORE_KEY_GENERATE_PAIR(pd, session, mech, pub_template, \
pub_count, priv_template, priv_count, out_pub_template, \
out_pub_count, out_priv_template, out_priv_count, req) ( \
(KCF_PROV_NOSTORE_KEY_OPS(pd) && \
KCF_PROV_NOSTORE_KEY_OPS(pd)->nostore_key_generate_pair) ? \
KCF_PROV_NOSTORE_KEY_OPS(pd)->nostore_key_generate_pair( \
(pd)->pd_prov_handle, session, mech, pub_template, pub_count, \
priv_template, priv_count, out_pub_template, out_pub_count, \
out_priv_template, out_priv_count, req) : CRYPTO_NOT_SUPPORTED)
#define KCF_PROV_NOSTORE_KEY_DERIVE(pd, session, mech, base_key, template, \
count, out_template, out_count, req) ( \
(KCF_PROV_NOSTORE_KEY_OPS(pd) && \
KCF_PROV_NOSTORE_KEY_OPS(pd)->nostore_key_derive) ? \
KCF_PROV_NOSTORE_KEY_OPS(pd)->nostore_key_derive( \
(pd)->pd_prov_handle, session, mech, base_key, template, count, \
out_template, out_count, req) : CRYPTO_NOT_SUPPORTED)
/*
* The following routines are exported by the kcf module (/kernel/misc/kcf)
* to the crypto and cryptoadmin modules.
*/
/* Digest/mac/cipher entry points that take a provider descriptor and session */
extern int crypto_digest_single(crypto_context_t, crypto_data_t *,
crypto_data_t *, crypto_call_req_t *);
extern int crypto_mac_single(crypto_context_t, crypto_data_t *,
crypto_data_t *, crypto_call_req_t *);
extern int crypto_encrypt_single(crypto_context_t, crypto_data_t *,
crypto_data_t *, crypto_call_req_t *);
extern int crypto_decrypt_single(crypto_context_t, crypto_data_t *,
crypto_data_t *, crypto_call_req_t *);
/* Other private digest/mac/cipher entry points not exported through k-API */
extern int crypto_digest_key_prov(crypto_context_t, crypto_key_t *,
crypto_call_req_t *);
/* Private sign entry points exported by KCF */
extern int crypto_sign_single(crypto_context_t, crypto_data_t *,
crypto_data_t *, crypto_call_req_t *);
extern int crypto_sign_recover_single(crypto_context_t, crypto_data_t *,
crypto_data_t *, crypto_call_req_t *);
/* Private verify entry points exported by KCF */
extern int crypto_verify_single(crypto_context_t, crypto_data_t *,
crypto_data_t *, crypto_call_req_t *);
extern int crypto_verify_recover_single(crypto_context_t, crypto_data_t *,
crypto_data_t *, crypto_call_req_t *);
/* Private dual operations entry points exported by KCF */
extern int crypto_digest_encrypt_update(crypto_context_t, crypto_context_t,
crypto_data_t *, crypto_data_t *, crypto_call_req_t *);
extern int crypto_decrypt_digest_update(crypto_context_t, crypto_context_t,
crypto_data_t *, crypto_data_t *, crypto_call_req_t *);
extern int crypto_sign_encrypt_update(crypto_context_t, crypto_context_t,
crypto_data_t *, crypto_data_t *, crypto_call_req_t *);
extern int crypto_decrypt_verify_update(crypto_context_t, crypto_context_t,
crypto_data_t *, crypto_data_t *, crypto_call_req_t *);
/* Random Number Generation */
int crypto_seed_random(crypto_provider_handle_t provider, uchar_t *buf,
size_t len, crypto_call_req_t *req);
int crypto_generate_random(crypto_provider_handle_t provider, uchar_t *buf,
size_t len, crypto_call_req_t *req);
/* Provider Management */
int crypto_get_provider_info(crypto_provider_id_t id,
crypto_provider_info_t **info, crypto_call_req_t *req);
int crypto_get_provider_mechanisms(crypto_minor_t *, crypto_provider_id_t id,
uint_t *count, crypto_mech_name_t **list);
int crypto_init_token(crypto_provider_handle_t provider, char *pin,
size_t pin_len, char *label, crypto_call_req_t *);
int crypto_init_pin(crypto_provider_handle_t provider, char *pin,
size_t pin_len, crypto_call_req_t *req);
int crypto_set_pin(crypto_provider_handle_t provider, char *old_pin,
size_t old_len, char *new_pin, size_t new_len, crypto_call_req_t *req);
void crypto_free_provider_list(crypto_provider_entry_t *list, uint_t count);
void crypto_free_provider_info(crypto_provider_info_t *info);
/* Administrative */
int crypto_get_dev_list(uint_t *count, crypto_dev_list_entry_t **list);
int crypto_get_soft_list(uint_t *count, char **list, size_t *len);
int crypto_get_dev_info(char *name, uint_t instance, uint_t *count,
crypto_mech_name_t **list);
int crypto_get_soft_info(caddr_t name, uint_t *count,
crypto_mech_name_t **list);
int crypto_load_dev_disabled(char *name, uint_t instance, uint_t count,
crypto_mech_name_t *list);
int crypto_load_soft_disabled(caddr_t name, uint_t count,
crypto_mech_name_t *list);
int crypto_unload_soft_module(caddr_t path);
int crypto_load_soft_config(caddr_t name, uint_t count,
crypto_mech_name_t *list);
int crypto_load_door(uint_t did);
void crypto_free_mech_list(crypto_mech_name_t *list, uint_t count);
void crypto_free_dev_list(crypto_dev_list_entry_t *list, uint_t count);
/* Miscellaneous */
int crypto_get_mechanism_number(caddr_t name, crypto_mech_type_t *number);
int crypto_build_permitted_mech_names(kcf_provider_desc_t *,
crypto_mech_name_t **, uint_t *, int);
extern void kcf_destroy_mech_tabs(void);
extern void kcf_init_mech_tabs(void);
extern int kcf_add_mech_provider(short, kcf_provider_desc_t *,
kcf_prov_mech_desc_t **);
extern void kcf_remove_mech_provider(char *, kcf_provider_desc_t *);
extern int kcf_get_mech_entry(crypto_mech_type_t, kcf_mech_entry_t **);
-extern kcf_provider_desc_t *kcf_alloc_provider_desc(crypto_provider_info_t *);
+extern kcf_provider_desc_t *kcf_alloc_provider_desc(
+ const crypto_provider_info_t *);
extern void kcf_provider_zero_refcnt(kcf_provider_desc_t *);
extern void kcf_free_provider_desc(kcf_provider_desc_t *);
extern void kcf_soft_config_init(void);
extern int get_sw_provider_for_mech(crypto_mech_name_t, char **);
-extern crypto_mech_type_t crypto_mech2id_common(char *, boolean_t);
+extern crypto_mech_type_t crypto_mech2id_common(const char *, boolean_t);
extern void undo_register_provider(kcf_provider_desc_t *, boolean_t);
extern void redo_register_provider(kcf_provider_desc_t *);
extern void kcf_rnd_init(void);
extern boolean_t kcf_rngprov_check(void);
extern int kcf_rnd_get_pseudo_bytes(uint8_t *, size_t);
extern int kcf_rnd_get_bytes(uint8_t *, size_t, boolean_t, boolean_t);
extern int random_add_pseudo_entropy(uint8_t *, size_t, uint_t);
extern void kcf_rnd_schedule_timeout(boolean_t);
extern int crypto_uio_data(crypto_data_t *, uchar_t *, int, cmd_type_t,
void *, void (*update)(void));
extern int crypto_mblk_data(crypto_data_t *, uchar_t *, int, cmd_type_t,
void *, void (*update)(void));
extern int crypto_put_output_data(uchar_t *, crypto_data_t *, int);
extern int crypto_get_input_data(crypto_data_t *, uchar_t **, uchar_t *);
extern int crypto_copy_key_to_ctx(crypto_key_t *, crypto_key_t **, size_t *,
int kmflag);
extern int crypto_digest_data(crypto_data_t *, void *, uchar_t *,
void (*update)(void), void (*final)(void), uchar_t);
extern int crypto_update_iov(void *, crypto_data_t *, crypto_data_t *,
int (*cipher)(void *, caddr_t, size_t, crypto_data_t *),
void (*copy_block)(uint8_t *, uint64_t *));
extern int crypto_update_uio(void *, crypto_data_t *, crypto_data_t *,
int (*cipher)(void *, caddr_t, size_t, crypto_data_t *),
void (*copy_block)(uint8_t *, uint64_t *));
extern int crypto_update_mp(void *, crypto_data_t *, crypto_data_t *,
int (*cipher)(void *, caddr_t, size_t, crypto_data_t *),
void (*copy_block)(uint8_t *, uint64_t *));
extern int crypto_get_key_attr(crypto_key_t *, crypto_attr_type_t, uchar_t **,
ssize_t *);
/* Access to the provider's table */
extern void kcf_prov_tab_destroy(void);
extern void kcf_prov_tab_init(void);
extern int kcf_prov_tab_add_provider(kcf_provider_desc_t *);
extern int kcf_prov_tab_rem_provider(crypto_provider_id_t);
extern kcf_provider_desc_t *kcf_prov_tab_lookup_by_name(char *);
extern kcf_provider_desc_t *kcf_prov_tab_lookup_by_dev(char *, uint_t);
extern int kcf_get_hw_prov_tab(uint_t *, kcf_provider_desc_t ***, int,
char *, uint_t, boolean_t);
extern int kcf_get_slot_list(uint_t *, kcf_provider_desc_t ***, boolean_t);
extern void kcf_free_provider_tab(uint_t, kcf_provider_desc_t **);
extern kcf_provider_desc_t *kcf_prov_tab_lookup(crypto_provider_id_t);
extern int kcf_get_sw_prov(crypto_mech_type_t, kcf_provider_desc_t **,
kcf_mech_entry_t **, boolean_t);
/* Access to the policy table */
extern boolean_t is_mech_disabled(kcf_provider_desc_t *, crypto_mech_name_t);
extern boolean_t is_mech_disabled_byname(crypto_provider_type_t, char *,
uint_t, crypto_mech_name_t);
extern void kcf_policy_tab_init(void);
extern void kcf_policy_free_desc(kcf_policy_desc_t *);
extern void kcf_policy_remove_by_name(char *, uint_t *, crypto_mech_name_t **);
extern void kcf_policy_remove_by_dev(char *, uint_t, uint_t *,
crypto_mech_name_t **);
extern kcf_policy_desc_t *kcf_policy_lookup_by_name(char *);
extern kcf_policy_desc_t *kcf_policy_lookup_by_dev(char *, uint_t);
extern int kcf_policy_load_soft_disabled(char *, uint_t, crypto_mech_name_t *,
uint_t *, crypto_mech_name_t **);
extern int kcf_policy_load_dev_disabled(char *, uint_t, uint_t,
crypto_mech_name_t *, uint_t *, crypto_mech_name_t **);
extern boolean_t in_soft_config_list(char *);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_CRYPTO_IMPL_H */
diff --git a/module/icp/include/sys/crypto/sched_impl.h b/module/icp/include/sys/crypto/sched_impl.h
index 85ea0ba1d092..29ef8021f0fc 100644
--- a/module/icp/include/sys/crypto/sched_impl.h
+++ b/module/icp/include/sys/crypto/sched_impl.h
@@ -1,531 +1,528 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#ifndef _SYS_CRYPTO_SCHED_IMPL_H
#define _SYS_CRYPTO_SCHED_IMPL_H
/*
* Scheduler internal structures.
*/
#ifdef __cplusplus
extern "C" {
#endif
#include <sys/zfs_context.h>
#include <sys/crypto/api.h>
#include <sys/crypto/spi.h>
#include <sys/crypto/impl.h>
#include <sys/crypto/common.h>
#include <sys/crypto/ops_impl.h>
typedef void (kcf_func_t)(void *, int);
typedef enum kcf_req_status {
REQ_ALLOCATED = 1,
REQ_WAITING, /* At the framework level */
REQ_INPROGRESS, /* At the provider level */
REQ_DONE,
REQ_CANCELED
} kcf_req_status_t;
typedef enum kcf_call_type {
CRYPTO_SYNCH = 1,
CRYPTO_ASYNCH
} kcf_call_type_t;
#define CHECK_RESTRICT(crq) (crq != NULL && \
((crq)->cr_flag & CRYPTO_RESTRICTED))
#define CHECK_RESTRICT_FALSE B_FALSE
#define CHECK_FASTPATH(crq, pd) ((crq) == NULL || \
!((crq)->cr_flag & CRYPTO_ALWAYS_QUEUE)) && \
(pd)->pd_prov_type == CRYPTO_SW_PROVIDER
#define KCF_KMFLAG(crq) (((crq) == NULL) ? KM_SLEEP : KM_NOSLEEP)
/*
* The framework keeps an internal handle to use in the adaptive
* asynchronous case. This is the case when a client has the
* CRYPTO_ALWAYS_QUEUE bit clear and a software provider is used for
* the request. The request is completed in the context of the calling
* thread and kernel memory must be allocated with KM_NOSLEEP.
*
* The framework passes a pointer to the handle in crypto_req_handle_t
* argument when it calls the SPI of the software provider. The macros
* KCF_RHNDL() and KCF_SWFP_RHNDL() are used to do this.
*
* When a provider asks the framework for kmflag value via
* crypto_kmflag(9S) we use REQHNDL2_KMFLAG() macro.
*/
extern ulong_t kcf_swprov_hndl;
#define KCF_RHNDL(kmflag) (((kmflag) == KM_SLEEP) ? NULL : &kcf_swprov_hndl)
#define KCF_SWFP_RHNDL(crq) (((crq) == NULL) ? NULL : &kcf_swprov_hndl)
#define REQHNDL2_KMFLAG(rhndl) \
((rhndl == &kcf_swprov_hndl) ? KM_NOSLEEP : KM_SLEEP)
/* Internal call_req flags. They start after the public ones in api.h */
#define CRYPTO_SETDUAL 0x00001000 /* Set the 'cont' boolean before */
/* submitting the request */
#define KCF_ISDUALREQ(crq) \
(((crq) == NULL) ? B_FALSE : (crq->cr_flag & CRYPTO_SETDUAL))
typedef struct kcf_prov_tried {
kcf_provider_desc_t *pt_pd;
struct kcf_prov_tried *pt_next;
} kcf_prov_tried_t;
#define IS_FG_SUPPORTED(mdesc, fg) \
(((mdesc)->pm_mech_info.cm_func_group_mask & (fg)) != 0)
#define IS_PROVIDER_TRIED(pd, tlist) \
(tlist != NULL && is_in_triedlist(pd, tlist))
#define IS_RECOVERABLE(error) \
(error == CRYPTO_BUFFER_TOO_BIG || \
error == CRYPTO_BUSY || \
error == CRYPTO_DEVICE_ERROR || \
error == CRYPTO_DEVICE_MEMORY || \
error == CRYPTO_KEY_SIZE_RANGE || \
error == CRYPTO_NO_PERMISSION)
#define KCF_ATOMIC_INCR(x) atomic_add_32(&(x), 1)
#define KCF_ATOMIC_DECR(x) atomic_add_32(&(x), -1)
/*
* Node structure for synchronous requests.
*/
typedef struct kcf_sreq_node {
/* Should always be the first field in this structure */
kcf_call_type_t sn_type;
/*
* sn_cv and sr_lock are used to wait for the
* operation to complete. sn_lock also protects
* the sn_state field.
*/
kcondvar_t sn_cv;
kmutex_t sn_lock;
kcf_req_status_t sn_state;
/*
* Return value from the operation. This will be
* one of the CRYPTO_* errors defined in common.h.
*/
int sn_rv;
/*
* parameters to call the SPI with. This can be
* a pointer as we know the caller context/stack stays.
*/
struct kcf_req_params *sn_params;
/* Internal context for this request */
struct kcf_context *sn_context;
/* Provider handling this request */
kcf_provider_desc_t *sn_provider;
} kcf_sreq_node_t;
/*
* Node structure for asynchronous requests. A node can be on
* on a chain of requests hanging of the internal context
* structure and can be in the global software provider queue.
*/
typedef struct kcf_areq_node {
/* Should always be the first field in this structure */
kcf_call_type_t an_type;
/* an_lock protects the field an_state */
kmutex_t an_lock;
kcf_req_status_t an_state;
crypto_call_req_t an_reqarg;
/*
* parameters to call the SPI with. We need to
* save the params since the caller stack can go away.
*/
struct kcf_req_params an_params;
/*
* The next two fields should be NULL for operations that
* don't need a context.
*/
/* Internal context for this request */
struct kcf_context *an_context;
/* next in chain of requests for context */
struct kcf_areq_node *an_ctxchain_next;
kcondvar_t an_turn_cv;
boolean_t an_is_my_turn;
boolean_t an_isdual; /* for internal reuse */
/*
* Next and previous nodes in the global software
* queue. These fields are NULL for a hardware
* provider since we use a taskq there.
*/
struct kcf_areq_node *an_next;
struct kcf_areq_node *an_prev;
/* Provider handling this request */
kcf_provider_desc_t *an_provider;
kcf_prov_tried_t *an_tried_plist;
struct kcf_areq_node *an_idnext; /* Next in ID hash */
struct kcf_areq_node *an_idprev; /* Prev in ID hash */
kcondvar_t an_done; /* Signal request completion */
uint_t an_refcnt;
} kcf_areq_node_t;
#define KCF_AREQ_REFHOLD(areq) { \
atomic_add_32(&(areq)->an_refcnt, 1); \
ASSERT((areq)->an_refcnt != 0); \
}
#define KCF_AREQ_REFRELE(areq) { \
ASSERT((areq)->an_refcnt != 0); \
membar_exit(); \
if (atomic_add_32_nv(&(areq)->an_refcnt, -1) == 0) \
kcf_free_req(areq); \
}
#define GET_REQ_TYPE(arg) *((kcf_call_type_t *)(arg))
#define NOTIFY_CLIENT(areq, err) (*(areq)->an_reqarg.cr_callback_func)(\
(areq)->an_reqarg.cr_callback_arg, err);
/* For internally generated call requests for dual operations */
typedef struct kcf_call_req {
crypto_call_req_t kr_callreq; /* external client call req */
kcf_req_params_t kr_params; /* Params saved for next call */
kcf_areq_node_t *kr_areq; /* Use this areq */
off_t kr_saveoffset;
size_t kr_savelen;
} kcf_dual_req_t;
/*
* The following are some what similar to macros in callo.h, which implement
* callout tables.
*
* The lower four bits of the ID are used to encode the table ID to
* index in to. The REQID_COUNTER_HIGH bit is used to avoid any check for
* wrap around when generating ID. We assume that there won't be a request
* which takes more time than 2^^(sizeof (long) - 5) other requests submitted
* after it. This ensures there won't be any ID collision.
*/
#define REQID_COUNTER_HIGH (1UL << (8 * sizeof (long) - 1))
#define REQID_COUNTER_SHIFT 4
#define REQID_COUNTER_LOW (1 << REQID_COUNTER_SHIFT)
#define REQID_TABLES 16
#define REQID_TABLE_MASK (REQID_TABLES - 1)
#define REQID_BUCKETS 512
#define REQID_BUCKET_MASK (REQID_BUCKETS - 1)
#define REQID_HASH(id) (((id) >> REQID_COUNTER_SHIFT) & REQID_BUCKET_MASK)
#define GET_REQID(areq) (areq)->an_reqarg.cr_reqid
#define SET_REQID(areq, val) GET_REQID(areq) = val
/*
* Hash table for async requests.
*/
typedef struct kcf_reqid_table {
kmutex_t rt_lock;
crypto_req_id_t rt_curid;
kcf_areq_node_t *rt_idhash[REQID_BUCKETS];
} kcf_reqid_table_t;
/*
* Global software provider queue structure. Requests to be
* handled by a SW provider and have the ALWAYS_QUEUE flag set
* get queued here.
*/
typedef struct kcf_global_swq {
/*
* gs_cv and gs_lock are used to wait for new requests.
* gs_lock protects the changes to the queue.
*/
kcondvar_t gs_cv;
kmutex_t gs_lock;
uint_t gs_njobs;
uint_t gs_maxjobs;
kcf_areq_node_t *gs_first;
kcf_areq_node_t *gs_last;
} kcf_global_swq_t;
/*
* Internal representation of a canonical context. We contain crypto_ctx_t
* structure in order to have just one memory allocation. The SPI
* ((crypto_ctx_t *)ctx)->cc_framework_private maps to this structure.
*/
typedef struct kcf_context {
crypto_ctx_t kc_glbl_ctx;
uint_t kc_refcnt;
kmutex_t kc_in_use_lock;
/*
* kc_req_chain_first and kc_req_chain_last are used to chain
* multiple async requests using the same context. They should be
* NULL for sync requests.
*/
kcf_areq_node_t *kc_req_chain_first;
kcf_areq_node_t *kc_req_chain_last;
kcf_provider_desc_t *kc_prov_desc; /* Prov. descriptor */
kcf_provider_desc_t *kc_sw_prov_desc; /* Prov. descriptor */
kcf_mech_entry_t *kc_mech;
struct kcf_context *kc_secondctx; /* for dual contexts */
} kcf_context_t;
/*
* Bump up the reference count on the framework private context. A
* global context or a request that references this structure should
* do a hold.
*/
#define KCF_CONTEXT_REFHOLD(ictx) { \
atomic_add_32(&(ictx)->kc_refcnt, 1); \
ASSERT((ictx)->kc_refcnt != 0); \
}
/*
* Decrement the reference count on the framework private context.
* When the last reference is released, the framework private
* context structure is freed along with the global context.
*/
#define KCF_CONTEXT_REFRELE(ictx) { \
ASSERT((ictx)->kc_refcnt != 0); \
membar_exit(); \
if (atomic_add_32_nv(&(ictx)->kc_refcnt, -1) == 0) \
kcf_free_context(ictx); \
}
/*
* Check if we can release the context now. In case of CRYPTO_QUEUED
* we do not release it as we can do it only after the provider notified
* us. In case of CRYPTO_BUSY, the client can retry the request using
* the context, so we do not release the context.
*
* This macro should be called only from the final routine in
* an init/update/final sequence. We do not release the context in case
* of update operations. We require the consumer to free it
* explicitly, in case it wants to abandon the operation. This is done
* as there may be mechanisms in ECB mode that can continue even if
* an operation on a block fails.
*/
#define KCF_CONTEXT_COND_RELEASE(rv, kcf_ctx) { \
if (KCF_CONTEXT_DONE(rv)) \
KCF_CONTEXT_REFRELE(kcf_ctx); \
}
/*
* This macro determines whether we're done with a context.
*/
#define KCF_CONTEXT_DONE(rv) \
((rv) != CRYPTO_QUEUED && (rv) != CRYPTO_BUSY && \
(rv) != CRYPTO_BUFFER_TOO_SMALL)
/*
* A crypto_ctx_template_t is internally a pointer to this struct
*/
typedef struct kcf_ctx_template {
crypto_kcf_provider_handle_t ct_prov_handle; /* provider handle */
uint_t ct_generation; /* generation # */
size_t ct_size; /* for freeing */
crypto_spi_ctx_template_t ct_prov_tmpl; /* context template */
/* from the SW prov */
} kcf_ctx_template_t;
/*
* Structure for pool of threads working on global software queue.
*/
typedef struct kcf_pool {
uint32_t kp_threads; /* Number of threads in pool */
uint32_t kp_idlethreads; /* Idle threads in pool */
uint32_t kp_blockedthreads; /* Blocked threads in pool */
/*
* cv & lock to monitor the condition when no threads
* are around. In this case the failover thread kicks in.
*/
kcondvar_t kp_nothr_cv;
kmutex_t kp_thread_lock;
/* Userspace thread creator variables. */
boolean_t kp_signal_create_thread; /* Create requested flag */
int kp_nthrs; /* # of threads to create */
boolean_t kp_user_waiting; /* Thread waiting for work */
/*
* cv & lock for the condition where more threads need to be
* created. kp_user_lock also protects the three fields above.
*/
kcondvar_t kp_user_cv; /* Creator cond. variable */
kmutex_t kp_user_lock; /* Creator lock */
} kcf_pool_t;
/*
* State of a crypto bufcall element.
*/
typedef enum cbuf_state {
CBUF_FREE = 1,
CBUF_WAITING,
CBUF_RUNNING
} cbuf_state_t;
/*
* Structure of a crypto bufcall element.
*/
typedef struct kcf_cbuf_elem {
/*
* lock and cv to wait for CBUF_RUNNING to be done
* kc_lock also protects kc_state.
*/
kmutex_t kc_lock;
kcondvar_t kc_cv;
cbuf_state_t kc_state;
struct kcf_cbuf_elem *kc_next;
struct kcf_cbuf_elem *kc_prev;
void (*kc_func)(void *arg);
void *kc_arg;
} kcf_cbuf_elem_t;
/*
* State of a notify element.
*/
typedef enum ntfy_elem_state {
NTFY_WAITING = 1,
NTFY_RUNNING
} ntfy_elem_state_t;
/*
* Structure of a notify list element.
*/
typedef struct kcf_ntfy_elem {
/*
* lock and cv to wait for NTFY_RUNNING to be done.
* kn_lock also protects kn_state.
*/
kmutex_t kn_lock;
kcondvar_t kn_cv;
ntfy_elem_state_t kn_state;
struct kcf_ntfy_elem *kn_next;
struct kcf_ntfy_elem *kn_prev;
crypto_notify_callback_t kn_func;
uint32_t kn_event_mask;
} kcf_ntfy_elem_t;
/*
* The following values are based on the assumption that it would
* take around eight cpus to load a hardware provider (This is true for
* at least one product) and a kernel client may come from different
* low-priority interrupt levels. We will have CRYPTO_TASKQ_MIN number
* of cached taskq entries. The CRYPTO_TASKQ_MAX number is based on
* a throughput of 1GB/s using 512-byte buffers. These are just
* reasonable estimates and might need to change in future.
*/
#define CRYPTO_TASKQ_THREADS 8
#define CRYPTO_TASKQ_MIN 64
#define CRYPTO_TASKQ_MAX 2 * 1024 * 1024
-extern int crypto_taskq_threads;
-extern int crypto_taskq_minalloc;
-extern int crypto_taskq_maxalloc;
-extern kcf_global_swq_t *gswq;
-extern int kcf_maxthreads;
-extern int kcf_minthreads;
+extern const int crypto_taskq_threads;
+extern const int crypto_taskq_minalloc;
+extern const int crypto_taskq_maxalloc;
/*
* All pending crypto bufcalls are put on a list. cbuf_list_lock
* protects changes to this list.
*/
extern kmutex_t cbuf_list_lock;
extern kcondvar_t cbuf_list_cv;
/*
* All event subscribers are put on a list. kcf_notify_list_lock
* protects changes to this list.
*/
extern kmutex_t ntfy_list_lock;
extern kcondvar_t ntfy_list_cv;
boolean_t kcf_get_next_logical_provider_member(kcf_provider_desc_t *,
kcf_provider_desc_t *, kcf_provider_desc_t **);
extern int kcf_get_hardware_provider(crypto_mech_type_t, crypto_mech_type_t,
boolean_t, kcf_provider_desc_t *, kcf_provider_desc_t **,
crypto_func_group_t);
extern int kcf_get_hardware_provider_nomech(offset_t, offset_t,
boolean_t, kcf_provider_desc_t *, kcf_provider_desc_t **);
extern void kcf_free_triedlist(kcf_prov_tried_t *);
extern kcf_prov_tried_t *kcf_insert_triedlist(kcf_prov_tried_t **,
kcf_provider_desc_t *, int);
extern kcf_provider_desc_t *kcf_get_mech_provider(crypto_mech_type_t,
kcf_mech_entry_t **, int *, kcf_prov_tried_t *, crypto_func_group_t,
boolean_t, size_t);
extern kcf_provider_desc_t *kcf_get_dual_provider(crypto_mechanism_t *,
crypto_mechanism_t *, kcf_mech_entry_t **, crypto_mech_type_t *,
crypto_mech_type_t *, int *, kcf_prov_tried_t *,
crypto_func_group_t, crypto_func_group_t, boolean_t, size_t);
extern crypto_ctx_t *kcf_new_ctx(crypto_call_req_t *, kcf_provider_desc_t *,
crypto_session_id_t);
extern int kcf_submit_request(kcf_provider_desc_t *, crypto_ctx_t *,
crypto_call_req_t *, kcf_req_params_t *, boolean_t);
extern void kcf_sched_destroy(void);
extern void kcf_sched_init(void);
extern void kcf_sched_start(void);
extern void kcf_sop_done(kcf_sreq_node_t *, int);
extern void kcf_aop_done(kcf_areq_node_t *, int);
extern int common_submit_request(kcf_provider_desc_t *,
crypto_ctx_t *, kcf_req_params_t *, crypto_req_handle_t);
extern void kcf_free_context(kcf_context_t *);
extern int kcf_svc_wait(int *);
extern int kcf_svc_do_run(void);
extern int kcf_need_signature_verification(kcf_provider_desc_t *);
extern void kcf_verify_signature(void *);
extern struct modctl *kcf_get_modctl(crypto_provider_info_t *);
extern void verify_unverified_providers(void);
extern void kcf_free_req(kcf_areq_node_t *areq);
extern void crypto_bufcall_service(void);
extern void kcf_walk_ntfylist(uint32_t, void *);
extern void kcf_do_notify(kcf_provider_desc_t *, boolean_t);
extern kcf_dual_req_t *kcf_alloc_req(crypto_call_req_t *);
extern void kcf_next_req(void *, int);
extern void kcf_last_req(void *, int);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_CRYPTO_SCHED_IMPL_H */
diff --git a/module/icp/include/sys/crypto/spi.h b/module/icp/include/sys/crypto/spi.h
index 2c62b5706651..0f1b455c808e 100644
--- a/module/icp/include/sys/crypto/spi.h
+++ b/module/icp/include/sys/crypto/spi.h
@@ -1,726 +1,726 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#ifndef _SYS_CRYPTO_SPI_H
#define _SYS_CRYPTO_SPI_H
/*
* CSPI: Cryptographic Service Provider Interface.
*/
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
#ifdef __cplusplus
extern "C" {
#endif
#ifdef CONSTIFY_PLUGIN
#define __no_const __attribute__((no_const))
#else
#define __no_const
#endif /* CONSTIFY_PLUGIN */
#define CRYPTO_SPI_VERSION_1 1
#define CRYPTO_SPI_VERSION_2 2
#define CRYPTO_SPI_VERSION_3 3
/*
* Provider-private handle. This handle is specified by a provider
* when it registers by means of the pi_provider_handle field of
* the crypto_provider_info structure, and passed to the provider
* when its entry points are invoked.
*/
typedef void *crypto_provider_handle_t;
/*
* Context templates can be used to by software providers to pre-process
* keying material, such as key schedules. They are allocated by
* a software provider create_ctx_template(9E) entry point, and passed
* as argument to initialization and atomic provider entry points.
*/
typedef void *crypto_spi_ctx_template_t;
/*
* Request handles are used by the kernel to identify an asynchronous
* request being processed by a provider. It is passed by the kernel
* to a hardware provider when submitting a request, and must be
* specified by a provider when calling crypto_op_notification(9F)
*/
typedef void *crypto_req_handle_t;
/* Values for cc_flags field */
#define CRYPTO_INIT_OPSTATE 0x00000001 /* allocate and init cc_opstate */
#define CRYPTO_USE_OPSTATE 0x00000002 /* .. start using it as context */
/*
* The context structure is passed from the kernel to a provider.
* It contains the information needed to process a multi-part or
* single part operation. The context structure is not used
* by atomic operations.
*
* Parameters needed to perform a cryptographic operation, such
* as keys, mechanisms, input and output buffers, are passed
* as separate arguments to Provider routines.
*/
typedef struct crypto_ctx {
crypto_provider_handle_t cc_provider;
crypto_session_id_t cc_session;
void *cc_provider_private; /* owned by provider */
void *cc_framework_private; /* owned by framework */
uint32_t cc_flags; /* flags */
void *cc_opstate; /* state */
} crypto_ctx_t;
/*
* Extended provider information.
*/
/*
* valid values for ei_flags field of extended info structure
* They match the RSA Security, Inc PKCS#11 tokenInfo flags.
*/
#define CRYPTO_EXTF_RNG 0x00000001
#define CRYPTO_EXTF_WRITE_PROTECTED 0x00000002
#define CRYPTO_EXTF_LOGIN_REQUIRED 0x00000004
#define CRYPTO_EXTF_USER_PIN_INITIALIZED 0x00000008
#define CRYPTO_EXTF_CLOCK_ON_TOKEN 0x00000040
#define CRYPTO_EXTF_PROTECTED_AUTHENTICATION_PATH 0x00000100
#define CRYPTO_EXTF_DUAL_CRYPTO_OPERATIONS 0x00000200
#define CRYPTO_EXTF_TOKEN_INITIALIZED 0x00000400
#define CRYPTO_EXTF_USER_PIN_COUNT_LOW 0x00010000
#define CRYPTO_EXTF_USER_PIN_FINAL_TRY 0x00020000
#define CRYPTO_EXTF_USER_PIN_LOCKED 0x00040000
#define CRYPTO_EXTF_USER_PIN_TO_BE_CHANGED 0x00080000
#define CRYPTO_EXTF_SO_PIN_COUNT_LOW 0x00100000
#define CRYPTO_EXTF_SO_PIN_FINAL_TRY 0x00200000
#define CRYPTO_EXTF_SO_PIN_LOCKED 0x00400000
#define CRYPTO_EXTF_SO_PIN_TO_BE_CHANGED 0x00800000
/*
* The crypto_control_ops structure contains pointers to control
* operations for cryptographic providers. It is passed through
* the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_control_ops {
void (*provider_status)(crypto_provider_handle_t, uint_t *);
} __no_const crypto_control_ops_t;
/*
* The crypto_ctx_ops structure contains points to context and context
* templates management operations for cryptographic providers. It is
* passed through the crypto_ops(9S) structure when providers register
* with the kernel using crypto_register_provider(9F).
*/
typedef struct crypto_ctx_ops {
int (*create_ctx_template)(crypto_provider_handle_t,
crypto_mechanism_t *, crypto_key_t *,
crypto_spi_ctx_template_t *, size_t *, crypto_req_handle_t);
int (*free_context)(crypto_ctx_t *);
} __no_const crypto_ctx_ops_t;
/*
* The crypto_digest_ops structure contains pointers to digest
* operations for cryptographic providers. It is passed through
* the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_digest_ops {
int (*digest_init)(crypto_ctx_t *, crypto_mechanism_t *,
crypto_req_handle_t);
int (*digest)(crypto_ctx_t *, crypto_data_t *, crypto_data_t *,
crypto_req_handle_t);
int (*digest_update)(crypto_ctx_t *, crypto_data_t *,
crypto_req_handle_t);
int (*digest_key)(crypto_ctx_t *, crypto_key_t *, crypto_req_handle_t);
int (*digest_final)(crypto_ctx_t *, crypto_data_t *,
crypto_req_handle_t);
int (*digest_atomic)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_data_t *,
crypto_data_t *, crypto_req_handle_t);
} __no_const crypto_digest_ops_t;
/*
* The crypto_cipher_ops structure contains pointers to encryption
* and decryption operations for cryptographic providers. It is
* passed through the crypto_ops(9S) structure when providers register
* with the kernel using crypto_register_provider(9F).
*/
typedef struct crypto_cipher_ops {
int (*encrypt_init)(crypto_ctx_t *,
crypto_mechanism_t *, crypto_key_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
int (*encrypt)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*encrypt_update)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*encrypt_final)(crypto_ctx_t *,
crypto_data_t *, crypto_req_handle_t);
int (*encrypt_atomic)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_data_t *, crypto_spi_ctx_template_t, crypto_req_handle_t);
int (*decrypt_init)(crypto_ctx_t *,
crypto_mechanism_t *, crypto_key_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
int (*decrypt)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*decrypt_update)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*decrypt_final)(crypto_ctx_t *,
crypto_data_t *, crypto_req_handle_t);
int (*decrypt_atomic)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_data_t *, crypto_spi_ctx_template_t, crypto_req_handle_t);
} __no_const crypto_cipher_ops_t;
/*
* The crypto_mac_ops structure contains pointers to MAC
* operations for cryptographic providers. It is passed through
* the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_mac_ops {
int (*mac_init)(crypto_ctx_t *,
crypto_mechanism_t *, crypto_key_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
int (*mac)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*mac_update)(crypto_ctx_t *,
crypto_data_t *, crypto_req_handle_t);
int (*mac_final)(crypto_ctx_t *,
crypto_data_t *, crypto_req_handle_t);
int (*mac_atomic)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_data_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
int (*mac_verify_atomic)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_data_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
} __no_const crypto_mac_ops_t;
/*
* The crypto_sign_ops structure contains pointers to signing
* operations for cryptographic providers. It is passed through
* the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_sign_ops {
int (*sign_init)(crypto_ctx_t *,
crypto_mechanism_t *, crypto_key_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
int (*sign)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*sign_update)(crypto_ctx_t *,
crypto_data_t *, crypto_req_handle_t);
int (*sign_final)(crypto_ctx_t *,
crypto_data_t *, crypto_req_handle_t);
int (*sign_atomic)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_data_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
int (*sign_recover_init)(crypto_ctx_t *, crypto_mechanism_t *,
crypto_key_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
int (*sign_recover)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*sign_recover_atomic)(crypto_provider_handle_t,
crypto_session_id_t, crypto_mechanism_t *, crypto_key_t *,
crypto_data_t *, crypto_data_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
} __no_const crypto_sign_ops_t;
/*
* The crypto_verify_ops structure contains pointers to verify
* operations for cryptographic providers. It is passed through
* the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_verify_ops {
int (*verify_init)(crypto_ctx_t *,
crypto_mechanism_t *, crypto_key_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
int (*do_verify)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*verify_update)(crypto_ctx_t *,
crypto_data_t *, crypto_req_handle_t);
int (*verify_final)(crypto_ctx_t *,
crypto_data_t *, crypto_req_handle_t);
int (*verify_atomic)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_data_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
int (*verify_recover_init)(crypto_ctx_t *, crypto_mechanism_t *,
crypto_key_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
int (*verify_recover)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*verify_recover_atomic)(crypto_provider_handle_t,
crypto_session_id_t, crypto_mechanism_t *, crypto_key_t *,
crypto_data_t *, crypto_data_t *, crypto_spi_ctx_template_t,
crypto_req_handle_t);
} __no_const crypto_verify_ops_t;
/*
* The crypto_dual_ops structure contains pointers to dual
* cipher and sign/verify operations for cryptographic providers.
* It is passed through the crypto_ops(9S) structure when
* providers register with the kernel using
* crypto_register_provider(9F).
*/
typedef struct crypto_dual_ops {
int (*digest_encrypt_update)(
crypto_ctx_t *, crypto_ctx_t *, crypto_data_t *,
crypto_data_t *, crypto_req_handle_t);
int (*decrypt_digest_update)(
crypto_ctx_t *, crypto_ctx_t *, crypto_data_t *,
crypto_data_t *, crypto_req_handle_t);
int (*sign_encrypt_update)(
crypto_ctx_t *, crypto_ctx_t *, crypto_data_t *,
crypto_data_t *, crypto_req_handle_t);
int (*decrypt_verify_update)(
crypto_ctx_t *, crypto_ctx_t *, crypto_data_t *,
crypto_data_t *, crypto_req_handle_t);
} __no_const crypto_dual_ops_t;
/*
* The crypto_dual_cipher_mac_ops structure contains pointers to dual
* cipher and MAC operations for cryptographic providers.
* It is passed through the crypto_ops(9S) structure when
* providers register with the kernel using
* crypto_register_provider(9F).
*/
typedef struct crypto_dual_cipher_mac_ops {
int (*encrypt_mac_init)(crypto_ctx_t *,
crypto_mechanism_t *, crypto_key_t *, crypto_mechanism_t *,
crypto_key_t *, crypto_spi_ctx_template_t,
crypto_spi_ctx_template_t, crypto_req_handle_t);
int (*encrypt_mac)(crypto_ctx_t *,
crypto_data_t *, crypto_dual_data_t *, crypto_data_t *,
crypto_req_handle_t);
int (*encrypt_mac_update)(crypto_ctx_t *,
crypto_data_t *, crypto_dual_data_t *, crypto_req_handle_t);
int (*encrypt_mac_final)(crypto_ctx_t *,
crypto_dual_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*encrypt_mac_atomic)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_mechanism_t *,
crypto_key_t *, crypto_data_t *, crypto_dual_data_t *,
crypto_data_t *, crypto_spi_ctx_template_t,
crypto_spi_ctx_template_t, crypto_req_handle_t);
int (*mac_decrypt_init)(crypto_ctx_t *,
crypto_mechanism_t *, crypto_key_t *, crypto_mechanism_t *,
crypto_key_t *, crypto_spi_ctx_template_t,
crypto_spi_ctx_template_t, crypto_req_handle_t);
int (*mac_decrypt)(crypto_ctx_t *,
crypto_dual_data_t *, crypto_data_t *, crypto_data_t *,
crypto_req_handle_t);
int (*mac_decrypt_update)(crypto_ctx_t *,
crypto_dual_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*mac_decrypt_final)(crypto_ctx_t *,
crypto_data_t *, crypto_data_t *, crypto_req_handle_t);
int (*mac_decrypt_atomic)(crypto_provider_handle_t,
crypto_session_id_t, crypto_mechanism_t *, crypto_key_t *,
crypto_mechanism_t *, crypto_key_t *, crypto_dual_data_t *,
crypto_data_t *, crypto_data_t *, crypto_spi_ctx_template_t,
crypto_spi_ctx_template_t, crypto_req_handle_t);
int (*mac_verify_decrypt_atomic)(crypto_provider_handle_t,
crypto_session_id_t, crypto_mechanism_t *, crypto_key_t *,
crypto_mechanism_t *, crypto_key_t *, crypto_dual_data_t *,
crypto_data_t *, crypto_data_t *, crypto_spi_ctx_template_t,
crypto_spi_ctx_template_t, crypto_req_handle_t);
} __no_const crypto_dual_cipher_mac_ops_t;
/*
* The crypto_random_number_ops structure contains pointers to random
* number operations for cryptographic providers. It is passed through
* the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_random_number_ops {
int (*seed_random)(crypto_provider_handle_t, crypto_session_id_t,
uchar_t *, size_t, uint_t, uint32_t, crypto_req_handle_t);
int (*generate_random)(crypto_provider_handle_t, crypto_session_id_t,
uchar_t *, size_t, crypto_req_handle_t);
} __no_const crypto_random_number_ops_t;
/*
* Flag values for seed_random.
*/
#define CRYPTO_SEED_NOW 0x00000001
/*
* The crypto_session_ops structure contains pointers to session
* operations for cryptographic providers. It is passed through
* the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_session_ops {
int (*session_open)(crypto_provider_handle_t, crypto_session_id_t *,
crypto_req_handle_t);
int (*session_close)(crypto_provider_handle_t, crypto_session_id_t,
crypto_req_handle_t);
int (*session_login)(crypto_provider_handle_t, crypto_session_id_t,
crypto_user_type_t, char *, size_t, crypto_req_handle_t);
int (*session_logout)(crypto_provider_handle_t, crypto_session_id_t,
crypto_req_handle_t);
} __no_const crypto_session_ops_t;
/*
* The crypto_object_ops structure contains pointers to object
* operations for cryptographic providers. It is passed through
* the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_object_ops {
int (*object_create)(crypto_provider_handle_t, crypto_session_id_t,
crypto_object_attribute_t *, uint_t, crypto_object_id_t *,
crypto_req_handle_t);
int (*object_copy)(crypto_provider_handle_t, crypto_session_id_t,
crypto_object_id_t, crypto_object_attribute_t *, uint_t,
crypto_object_id_t *, crypto_req_handle_t);
int (*object_destroy)(crypto_provider_handle_t, crypto_session_id_t,
crypto_object_id_t, crypto_req_handle_t);
int (*object_get_size)(crypto_provider_handle_t, crypto_session_id_t,
crypto_object_id_t, size_t *, crypto_req_handle_t);
int (*object_get_attribute_value)(crypto_provider_handle_t,
crypto_session_id_t, crypto_object_id_t,
crypto_object_attribute_t *, uint_t, crypto_req_handle_t);
int (*object_set_attribute_value)(crypto_provider_handle_t,
crypto_session_id_t, crypto_object_id_t,
crypto_object_attribute_t *, uint_t, crypto_req_handle_t);
int (*object_find_init)(crypto_provider_handle_t, crypto_session_id_t,
crypto_object_attribute_t *, uint_t, void **,
crypto_req_handle_t);
int (*object_find)(crypto_provider_handle_t, void *,
crypto_object_id_t *, uint_t, uint_t *, crypto_req_handle_t);
int (*object_find_final)(crypto_provider_handle_t, void *,
crypto_req_handle_t);
} __no_const crypto_object_ops_t;
/*
* The crypto_key_ops structure contains pointers to key
* operations for cryptographic providers. It is passed through
* the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_key_ops {
int (*key_generate)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_object_attribute_t *, uint_t,
crypto_object_id_t *, crypto_req_handle_t);
int (*key_generate_pair)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_object_attribute_t *, uint_t,
crypto_object_attribute_t *, uint_t, crypto_object_id_t *,
crypto_object_id_t *, crypto_req_handle_t);
int (*key_wrap)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_object_id_t *,
uchar_t *, size_t *, crypto_req_handle_t);
int (*key_unwrap)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, uchar_t *, size_t *,
crypto_object_attribute_t *, uint_t,
crypto_object_id_t *, crypto_req_handle_t);
int (*key_derive)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_object_attribute_t *,
uint_t, crypto_object_id_t *, crypto_req_handle_t);
int (*key_check)(crypto_provider_handle_t, crypto_mechanism_t *,
crypto_key_t *);
} __no_const crypto_key_ops_t;
/*
* The crypto_provider_management_ops structure contains pointers
* to management operations for cryptographic providers. It is passed
* through the crypto_ops(9S) structure when providers register with the
* kernel using crypto_register_provider(9F).
*/
typedef struct crypto_provider_management_ops {
int (*ext_info)(crypto_provider_handle_t,
crypto_provider_ext_info_t *, crypto_req_handle_t);
int (*init_token)(crypto_provider_handle_t, char *, size_t,
char *, crypto_req_handle_t);
int (*init_pin)(crypto_provider_handle_t, crypto_session_id_t,
char *, size_t, crypto_req_handle_t);
int (*set_pin)(crypto_provider_handle_t, crypto_session_id_t,
char *, size_t, char *, size_t, crypto_req_handle_t);
} __no_const crypto_provider_management_ops_t;
typedef struct crypto_mech_ops {
int (*copyin_mechanism)(crypto_provider_handle_t,
crypto_mechanism_t *, crypto_mechanism_t *, int *, int);
int (*copyout_mechanism)(crypto_provider_handle_t,
crypto_mechanism_t *, crypto_mechanism_t *, int *, int);
int (*free_mechanism)(crypto_provider_handle_t, crypto_mechanism_t *);
} __no_const crypto_mech_ops_t;
typedef struct crypto_nostore_key_ops {
int (*nostore_key_generate)(crypto_provider_handle_t,
crypto_session_id_t, crypto_mechanism_t *,
crypto_object_attribute_t *, uint_t, crypto_object_attribute_t *,
uint_t, crypto_req_handle_t);
int (*nostore_key_generate_pair)(crypto_provider_handle_t,
crypto_session_id_t, crypto_mechanism_t *,
crypto_object_attribute_t *, uint_t, crypto_object_attribute_t *,
uint_t, crypto_object_attribute_t *, uint_t,
crypto_object_attribute_t *, uint_t, crypto_req_handle_t);
int (*nostore_key_derive)(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_object_attribute_t *,
uint_t, crypto_object_attribute_t *, uint_t, crypto_req_handle_t);
} __no_const crypto_nostore_key_ops_t;
/*
* The crypto_ops(9S) structure contains the structures containing
* the pointers to functions implemented by cryptographic providers.
* It is specified as part of the crypto_provider_info(9S)
* supplied by a provider when it registers with the kernel
* by calling crypto_register_provider(9F).
*/
typedef struct crypto_ops_v1 {
- crypto_control_ops_t *co_control_ops;
- crypto_digest_ops_t *co_digest_ops;
- crypto_cipher_ops_t *co_cipher_ops;
- crypto_mac_ops_t *co_mac_ops;
+ const crypto_control_ops_t *co_control_ops;
+ const crypto_digest_ops_t *co_digest_ops;
+ const crypto_cipher_ops_t *co_cipher_ops;
+ const crypto_mac_ops_t *co_mac_ops;
crypto_sign_ops_t *co_sign_ops;
crypto_verify_ops_t *co_verify_ops;
crypto_dual_ops_t *co_dual_ops;
crypto_dual_cipher_mac_ops_t *co_dual_cipher_mac_ops;
crypto_random_number_ops_t *co_random_ops;
crypto_session_ops_t *co_session_ops;
crypto_object_ops_t *co_object_ops;
crypto_key_ops_t *co_key_ops;
crypto_provider_management_ops_t *co_provider_ops;
- crypto_ctx_ops_t *co_ctx_ops;
+ const crypto_ctx_ops_t *co_ctx_ops;
} crypto_ops_v1_t;
typedef struct crypto_ops_v2 {
crypto_ops_v1_t v1_ops;
crypto_mech_ops_t *co_mech_ops;
} crypto_ops_v2_t;
typedef struct crypto_ops_v3 {
crypto_ops_v2_t v2_ops;
crypto_nostore_key_ops_t *co_nostore_key_ops;
} crypto_ops_v3_t;
typedef struct crypto_ops {
union {
crypto_ops_v3_t cou_v3;
crypto_ops_v2_t cou_v2;
crypto_ops_v1_t cou_v1;
} cou;
} crypto_ops_t;
#define co_control_ops cou.cou_v1.co_control_ops
#define co_digest_ops cou.cou_v1.co_digest_ops
#define co_cipher_ops cou.cou_v1.co_cipher_ops
#define co_mac_ops cou.cou_v1.co_mac_ops
#define co_sign_ops cou.cou_v1.co_sign_ops
#define co_verify_ops cou.cou_v1.co_verify_ops
#define co_dual_ops cou.cou_v1.co_dual_ops
#define co_dual_cipher_mac_ops cou.cou_v1.co_dual_cipher_mac_ops
#define co_random_ops cou.cou_v1.co_random_ops
#define co_session_ops cou.cou_v1.co_session_ops
#define co_object_ops cou.cou_v1.co_object_ops
#define co_key_ops cou.cou_v1.co_key_ops
#define co_provider_ops cou.cou_v1.co_provider_ops
#define co_ctx_ops cou.cou_v1.co_ctx_ops
#define co_mech_ops cou.cou_v2.co_mech_ops
#define co_nostore_key_ops cou.cou_v3.co_nostore_key_ops
/*
* The mechanism info structure crypto_mech_info_t contains a function group
* bit mask cm_func_group_mask. This field, of type crypto_func_group_t,
* specifies the provider entry point that can be used a particular
* mechanism. The function group mask is a combination of the following values.
*/
typedef uint32_t crypto_func_group_t;
#define CRYPTO_FG_ENCRYPT 0x00000001 /* encrypt_init() */
#define CRYPTO_FG_DECRYPT 0x00000002 /* decrypt_init() */
#define CRYPTO_FG_DIGEST 0x00000004 /* digest_init() */
#define CRYPTO_FG_SIGN 0x00000008 /* sign_init() */
#define CRYPTO_FG_SIGN_RECOVER 0x00000010 /* sign_recover_init() */
#define CRYPTO_FG_VERIFY 0x00000020 /* verify_init() */
#define CRYPTO_FG_VERIFY_RECOVER 0x00000040 /* verify_recover_init() */
#define CRYPTO_FG_GENERATE 0x00000080 /* key_generate() */
#define CRYPTO_FG_GENERATE_KEY_PAIR 0x00000100 /* key_generate_pair() */
#define CRYPTO_FG_WRAP 0x00000200 /* key_wrap() */
#define CRYPTO_FG_UNWRAP 0x00000400 /* key_unwrap() */
#define CRYPTO_FG_DERIVE 0x00000800 /* key_derive() */
#define CRYPTO_FG_MAC 0x00001000 /* mac_init() */
#define CRYPTO_FG_ENCRYPT_MAC 0x00002000 /* encrypt_mac_init() */
#define CRYPTO_FG_MAC_DECRYPT 0x00004000 /* decrypt_mac_init() */
#define CRYPTO_FG_ENCRYPT_ATOMIC 0x00008000 /* encrypt_atomic() */
#define CRYPTO_FG_DECRYPT_ATOMIC 0x00010000 /* decrypt_atomic() */
#define CRYPTO_FG_MAC_ATOMIC 0x00020000 /* mac_atomic() */
#define CRYPTO_FG_DIGEST_ATOMIC 0x00040000 /* digest_atomic() */
#define CRYPTO_FG_SIGN_ATOMIC 0x00080000 /* sign_atomic() */
#define CRYPTO_FG_SIGN_RECOVER_ATOMIC 0x00100000 /* sign_recover_atomic() */
#define CRYPTO_FG_VERIFY_ATOMIC 0x00200000 /* verify_atomic() */
#define CRYPTO_FG_VERIFY_RECOVER_ATOMIC 0x00400000 /* verify_recover_atomic() */
#define CRYPTO_FG_ENCRYPT_MAC_ATOMIC 0x00800000 /* encrypt_mac_atomic() */
#define CRYPTO_FG_MAC_DECRYPT_ATOMIC 0x01000000 /* mac_decrypt_atomic() */
#define CRYPTO_FG_RESERVED 0x80000000
/*
* Maximum length of the pi_provider_description field of the
* crypto_provider_info structure.
*/
#define CRYPTO_PROVIDER_DESCR_MAX_LEN 64
/* Bit mask for all the simple operations */
#define CRYPTO_FG_SIMPLEOP_MASK (CRYPTO_FG_ENCRYPT | CRYPTO_FG_DECRYPT | \
CRYPTO_FG_DIGEST | CRYPTO_FG_SIGN | CRYPTO_FG_VERIFY | CRYPTO_FG_MAC | \
CRYPTO_FG_ENCRYPT_ATOMIC | CRYPTO_FG_DECRYPT_ATOMIC | \
CRYPTO_FG_MAC_ATOMIC | CRYPTO_FG_DIGEST_ATOMIC | CRYPTO_FG_SIGN_ATOMIC | \
CRYPTO_FG_VERIFY_ATOMIC)
/* Bit mask for all the dual operations */
#define CRYPTO_FG_MAC_CIPHER_MASK (CRYPTO_FG_ENCRYPT_MAC | \
CRYPTO_FG_MAC_DECRYPT | CRYPTO_FG_ENCRYPT_MAC_ATOMIC | \
CRYPTO_FG_MAC_DECRYPT_ATOMIC)
/* Add other combos to CRYPTO_FG_DUAL_MASK */
#define CRYPTO_FG_DUAL_MASK CRYPTO_FG_MAC_CIPHER_MASK
/*
* The crypto_mech_info structure specifies one of the mechanisms
* supported by a cryptographic provider. The pi_mechanisms field of
* the crypto_provider_info structure contains a pointer to an array
* of crypto_mech_info's.
*/
typedef struct crypto_mech_info {
crypto_mech_name_t cm_mech_name;
crypto_mech_type_t cm_mech_number;
crypto_func_group_t cm_func_group_mask;
ssize_t cm_min_key_length;
ssize_t cm_max_key_length;
uint32_t cm_mech_flags;
} crypto_mech_info_t;
/* Alias the old name to the new name for compatibility. */
#define cm_keysize_unit cm_mech_flags
/*
* The following is used by a provider that sets
* CRYPTO_HASH_NO_UPDATE. It needs to specify the maximum
* input data size it can digest in this field.
*/
#define cm_max_input_length cm_max_key_length
/*
* crypto_kcf_provider_handle_t is a handle allocated by the kernel.
* It is returned after the provider registers with
* crypto_register_provider(), and must be specified by the provider
* when calling crypto_unregister_provider(), and
* crypto_provider_notification().
*/
typedef uint_t crypto_kcf_provider_handle_t;
/*
* Provider information. Passed as argument to crypto_register_provider(9F).
* Describes the provider and its capabilities. Multiple providers can
* register for the same device instance. In this case, the same
* pi_provider_dev must be specified with a different pi_provider_handle.
*/
typedef struct crypto_provider_info_v1 {
uint_t pi_interface_version;
char *pi_provider_description;
crypto_provider_type_t pi_provider_type;
crypto_provider_handle_t pi_provider_handle;
- crypto_ops_t *pi_ops_vector;
+ const crypto_ops_t *pi_ops_vector;
uint_t pi_mech_list_count;
- crypto_mech_info_t *pi_mechanisms;
+ const crypto_mech_info_t *pi_mechanisms;
uint_t pi_logical_provider_count;
crypto_kcf_provider_handle_t *pi_logical_providers;
} crypto_provider_info_v1_t;
typedef struct crypto_provider_info_v2 {
crypto_provider_info_v1_t v1_info;
uint_t pi_flags;
} crypto_provider_info_v2_t;
typedef struct crypto_provider_info {
union {
crypto_provider_info_v2_t piu_v2;
crypto_provider_info_v1_t piu_v1;
} piu;
} crypto_provider_info_t;
#define pi_interface_version piu.piu_v1.pi_interface_version
#define pi_provider_description piu.piu_v1.pi_provider_description
#define pi_provider_type piu.piu_v1.pi_provider_type
#define pi_provider_handle piu.piu_v1.pi_provider_handle
#define pi_ops_vector piu.piu_v1.pi_ops_vector
#define pi_mech_list_count piu.piu_v1.pi_mech_list_count
#define pi_mechanisms piu.piu_v1.pi_mechanisms
#define pi_logical_provider_count piu.piu_v1.pi_logical_provider_count
#define pi_logical_providers piu.piu_v1.pi_logical_providers
#define pi_flags piu.piu_v2.pi_flags
/* hidden providers can only be accessed via a logical provider */
#define CRYPTO_HIDE_PROVIDER 0x00000001
/*
* provider can not do multi-part digest (updates) and has a limit
* on maximum input data that it can digest.
*/
#define CRYPTO_HASH_NO_UPDATE 0x00000002
/* provider can handle the request without returning a CRYPTO_QUEUED */
#define CRYPTO_SYNCHRONOUS 0x00000004
#define CRYPTO_PIFLAGS_RESERVED2 0x40000000
#define CRYPTO_PIFLAGS_RESERVED1 0x80000000
/*
* Provider status passed by a provider to crypto_provider_notification(9F)
* and returned by the provider_status(9E) entry point.
*/
#define CRYPTO_PROVIDER_READY 0
#define CRYPTO_PROVIDER_BUSY 1
#define CRYPTO_PROVIDER_FAILED 2
/*
* Functions exported by Solaris to cryptographic providers. Providers
* call these functions to register and unregister, notify the kernel
* of state changes, and notify the kernel when a asynchronous request
* completed.
*/
-extern int crypto_register_provider(crypto_provider_info_t *,
+extern int crypto_register_provider(const crypto_provider_info_t *,
crypto_kcf_provider_handle_t *);
extern int crypto_unregister_provider(crypto_kcf_provider_handle_t);
extern void crypto_provider_notification(crypto_kcf_provider_handle_t, uint_t);
extern void crypto_op_notification(crypto_req_handle_t, int);
extern int crypto_kmflag(crypto_req_handle_t);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_CRYPTO_SPI_H */
diff --git a/module/icp/io/aes.c b/module/icp/io/aes.c
index 2f24739a4bda..be1736864c0b 100644
--- a/module/icp/io/aes.c
+++ b/module/icp/io/aes.c
@@ -1,1433 +1,1433 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
*/
/*
* AES provider for the Kernel Cryptographic Framework (KCF)
*/
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
#include <sys/crypto/impl.h>
#include <sys/crypto/spi.h>
#include <sys/crypto/icp.h>
#include <modes/modes.h>
#define _AES_IMPL
#include <aes/aes_impl.h>
#include <modes/gcm_impl.h>
/*
* Mechanism info structure passed to KCF during registration.
*/
-static crypto_mech_info_t aes_mech_info_tab[] = {
+static const crypto_mech_info_t aes_mech_info_tab[] = {
/* AES_ECB */
{SUN_CKM_AES_ECB, AES_ECB_MECH_INFO_TYPE,
CRYPTO_FG_ENCRYPT | CRYPTO_FG_ENCRYPT_ATOMIC |
CRYPTO_FG_DECRYPT | CRYPTO_FG_DECRYPT_ATOMIC,
AES_MIN_KEY_BYTES, AES_MAX_KEY_BYTES, CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* AES_CBC */
{SUN_CKM_AES_CBC, AES_CBC_MECH_INFO_TYPE,
CRYPTO_FG_ENCRYPT | CRYPTO_FG_ENCRYPT_ATOMIC |
CRYPTO_FG_DECRYPT | CRYPTO_FG_DECRYPT_ATOMIC,
AES_MIN_KEY_BYTES, AES_MAX_KEY_BYTES, CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* AES_CTR */
{SUN_CKM_AES_CTR, AES_CTR_MECH_INFO_TYPE,
CRYPTO_FG_ENCRYPT | CRYPTO_FG_ENCRYPT_ATOMIC |
CRYPTO_FG_DECRYPT | CRYPTO_FG_DECRYPT_ATOMIC,
AES_MIN_KEY_BYTES, AES_MAX_KEY_BYTES, CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* AES_CCM */
{SUN_CKM_AES_CCM, AES_CCM_MECH_INFO_TYPE,
CRYPTO_FG_ENCRYPT | CRYPTO_FG_ENCRYPT_ATOMIC |
CRYPTO_FG_DECRYPT | CRYPTO_FG_DECRYPT_ATOMIC,
AES_MIN_KEY_BYTES, AES_MAX_KEY_BYTES, CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* AES_GCM */
{SUN_CKM_AES_GCM, AES_GCM_MECH_INFO_TYPE,
CRYPTO_FG_ENCRYPT | CRYPTO_FG_ENCRYPT_ATOMIC |
CRYPTO_FG_DECRYPT | CRYPTO_FG_DECRYPT_ATOMIC,
AES_MIN_KEY_BYTES, AES_MAX_KEY_BYTES, CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* AES_GMAC */
{SUN_CKM_AES_GMAC, AES_GMAC_MECH_INFO_TYPE,
CRYPTO_FG_ENCRYPT | CRYPTO_FG_ENCRYPT_ATOMIC |
CRYPTO_FG_DECRYPT | CRYPTO_FG_DECRYPT_ATOMIC |
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC |
CRYPTO_FG_SIGN | CRYPTO_FG_SIGN_ATOMIC |
CRYPTO_FG_VERIFY | CRYPTO_FG_VERIFY_ATOMIC,
AES_MIN_KEY_BYTES, AES_MAX_KEY_BYTES, CRYPTO_KEYSIZE_UNIT_IN_BYTES}
};
static void aes_provider_status(crypto_provider_handle_t, uint_t *);
-static crypto_control_ops_t aes_control_ops = {
+static const crypto_control_ops_t aes_control_ops = {
aes_provider_status
};
static int aes_encrypt_init(crypto_ctx_t *, crypto_mechanism_t *,
crypto_key_t *, crypto_spi_ctx_template_t, crypto_req_handle_t);
static int aes_decrypt_init(crypto_ctx_t *, crypto_mechanism_t *,
crypto_key_t *, crypto_spi_ctx_template_t, crypto_req_handle_t);
static int aes_common_init(crypto_ctx_t *, crypto_mechanism_t *,
crypto_key_t *, crypto_spi_ctx_template_t, crypto_req_handle_t, boolean_t);
static int aes_common_init_ctx(aes_ctx_t *, crypto_spi_ctx_template_t *,
crypto_mechanism_t *, crypto_key_t *, int, boolean_t);
static int aes_encrypt_final(crypto_ctx_t *, crypto_data_t *,
crypto_req_handle_t);
static int aes_decrypt_final(crypto_ctx_t *, crypto_data_t *,
crypto_req_handle_t);
static int aes_encrypt(crypto_ctx_t *, crypto_data_t *, crypto_data_t *,
crypto_req_handle_t);
static int aes_encrypt_update(crypto_ctx_t *, crypto_data_t *,
crypto_data_t *, crypto_req_handle_t);
static int aes_encrypt_atomic(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_data_t *, crypto_spi_ctx_template_t, crypto_req_handle_t);
static int aes_decrypt(crypto_ctx_t *, crypto_data_t *, crypto_data_t *,
crypto_req_handle_t);
static int aes_decrypt_update(crypto_ctx_t *, crypto_data_t *,
crypto_data_t *, crypto_req_handle_t);
static int aes_decrypt_atomic(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *,
crypto_data_t *, crypto_spi_ctx_template_t, crypto_req_handle_t);
-static crypto_cipher_ops_t aes_cipher_ops = {
+static const crypto_cipher_ops_t aes_cipher_ops = {
.encrypt_init = aes_encrypt_init,
.encrypt = aes_encrypt,
.encrypt_update = aes_encrypt_update,
.encrypt_final = aes_encrypt_final,
.encrypt_atomic = aes_encrypt_atomic,
.decrypt_init = aes_decrypt_init,
.decrypt = aes_decrypt,
.decrypt_update = aes_decrypt_update,
.decrypt_final = aes_decrypt_final,
.decrypt_atomic = aes_decrypt_atomic
};
static int aes_mac_atomic(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *, crypto_data_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
static int aes_mac_verify_atomic(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *, crypto_data_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
-static crypto_mac_ops_t aes_mac_ops = {
+static const crypto_mac_ops_t aes_mac_ops = {
.mac_init = NULL,
.mac = NULL,
.mac_update = NULL,
.mac_final = NULL,
.mac_atomic = aes_mac_atomic,
.mac_verify_atomic = aes_mac_verify_atomic
};
static int aes_create_ctx_template(crypto_provider_handle_t,
crypto_mechanism_t *, crypto_key_t *, crypto_spi_ctx_template_t *,
size_t *, crypto_req_handle_t);
static int aes_free_context(crypto_ctx_t *);
-static crypto_ctx_ops_t aes_ctx_ops = {
+static const crypto_ctx_ops_t aes_ctx_ops = {
.create_ctx_template = aes_create_ctx_template,
.free_context = aes_free_context
};
-static crypto_ops_t aes_crypto_ops = {{{{{
+static const crypto_ops_t aes_crypto_ops = {{{{{
&aes_control_ops,
NULL,
&aes_cipher_ops,
&aes_mac_ops,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
&aes_ctx_ops
}}}}};
-static crypto_provider_info_t aes_prov_info = {{{{
+static const crypto_provider_info_t aes_prov_info = {{{{
CRYPTO_SPI_VERSION_1,
"AES Software Provider",
CRYPTO_SW_PROVIDER,
NULL,
&aes_crypto_ops,
- sizeof (aes_mech_info_tab)/sizeof (crypto_mech_info_t),
+ sizeof (aes_mech_info_tab) / sizeof (crypto_mech_info_t),
aes_mech_info_tab
}}}};
static crypto_kcf_provider_handle_t aes_prov_handle = 0;
static crypto_data_t null_crypto_data = { CRYPTO_DATA_RAW };
int
aes_mod_init(void)
{
/* Determine the fastest available implementation. */
aes_impl_init();
gcm_impl_init();
/* Register with KCF. If the registration fails, remove the module. */
if (crypto_register_provider(&aes_prov_info, &aes_prov_handle))
return (EACCES);
return (0);
}
int
aes_mod_fini(void)
{
/* Unregister from KCF if module is registered */
if (aes_prov_handle != 0) {
if (crypto_unregister_provider(aes_prov_handle))
return (EBUSY);
aes_prov_handle = 0;
}
return (0);
}
static int
aes_check_mech_param(crypto_mechanism_t *mechanism, aes_ctx_t **ctx, int kmflag)
{
void *p = NULL;
boolean_t param_required = B_TRUE;
size_t param_len;
void *(*alloc_fun)(int);
int rv = CRYPTO_SUCCESS;
switch (mechanism->cm_type) {
case AES_ECB_MECH_INFO_TYPE:
param_required = B_FALSE;
alloc_fun = ecb_alloc_ctx;
break;
case AES_CBC_MECH_INFO_TYPE:
param_len = AES_BLOCK_LEN;
alloc_fun = cbc_alloc_ctx;
break;
case AES_CTR_MECH_INFO_TYPE:
param_len = sizeof (CK_AES_CTR_PARAMS);
alloc_fun = ctr_alloc_ctx;
break;
case AES_CCM_MECH_INFO_TYPE:
param_len = sizeof (CK_AES_CCM_PARAMS);
alloc_fun = ccm_alloc_ctx;
break;
case AES_GCM_MECH_INFO_TYPE:
param_len = sizeof (CK_AES_GCM_PARAMS);
alloc_fun = gcm_alloc_ctx;
break;
case AES_GMAC_MECH_INFO_TYPE:
param_len = sizeof (CK_AES_GMAC_PARAMS);
alloc_fun = gmac_alloc_ctx;
break;
default:
rv = CRYPTO_MECHANISM_INVALID;
return (rv);
}
if (param_required && mechanism->cm_param != NULL &&
mechanism->cm_param_len != param_len) {
rv = CRYPTO_MECHANISM_PARAM_INVALID;
}
if (ctx != NULL) {
p = (alloc_fun)(kmflag);
*ctx = p;
}
return (rv);
}
/*
* Initialize key schedules for AES
*/
static int
init_keysched(crypto_key_t *key, void *newbie)
{
/*
* Only keys by value are supported by this module.
*/
switch (key->ck_format) {
case CRYPTO_KEY_RAW:
if (key->ck_length < AES_MINBITS ||
key->ck_length > AES_MAXBITS) {
return (CRYPTO_KEY_SIZE_RANGE);
}
/* key length must be either 128, 192, or 256 */
if ((key->ck_length & 63) != 0)
return (CRYPTO_KEY_SIZE_RANGE);
break;
default:
return (CRYPTO_KEY_TYPE_INCONSISTENT);
}
aes_init_keysched(key->ck_data, key->ck_length, newbie);
return (CRYPTO_SUCCESS);
}
/*
* KCF software provider control entry points.
*/
static void
aes_provider_status(crypto_provider_handle_t provider, uint_t *status)
{
(void) provider;
*status = CRYPTO_PROVIDER_READY;
}
static int
aes_encrypt_init(crypto_ctx_t *ctx, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_spi_ctx_template_t template,
crypto_req_handle_t req)
{
return (aes_common_init(ctx, mechanism, key, template, req, B_TRUE));
}
static int
aes_decrypt_init(crypto_ctx_t *ctx, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_spi_ctx_template_t template,
crypto_req_handle_t req)
{
return (aes_common_init(ctx, mechanism, key, template, req, B_FALSE));
}
/*
* KCF software provider encrypt entry points.
*/
static int
aes_common_init(crypto_ctx_t *ctx, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_spi_ctx_template_t template,
crypto_req_handle_t req, boolean_t is_encrypt_init)
{
aes_ctx_t *aes_ctx;
int rv;
int kmflag;
/*
* Only keys by value are supported by this module.
*/
if (key->ck_format != CRYPTO_KEY_RAW) {
return (CRYPTO_KEY_TYPE_INCONSISTENT);
}
kmflag = crypto_kmflag(req);
if ((rv = aes_check_mech_param(mechanism, &aes_ctx, kmflag))
!= CRYPTO_SUCCESS)
return (rv);
rv = aes_common_init_ctx(aes_ctx, template, mechanism, key, kmflag,
is_encrypt_init);
if (rv != CRYPTO_SUCCESS) {
crypto_free_mode_ctx(aes_ctx);
return (rv);
}
ctx->cc_provider_private = aes_ctx;
return (CRYPTO_SUCCESS);
}
static void
aes_copy_block64(uint8_t *in, uint64_t *out)
{
if (IS_P2ALIGNED(in, sizeof (uint64_t))) {
/* LINTED: pointer alignment */
out[0] = *(uint64_t *)&in[0];
/* LINTED: pointer alignment */
out[1] = *(uint64_t *)&in[8];
} else {
uint8_t *iv8 = (uint8_t *)&out[0];
AES_COPY_BLOCK(in, iv8);
}
}
static int
aes_encrypt(crypto_ctx_t *ctx, crypto_data_t *plaintext,
crypto_data_t *ciphertext, crypto_req_handle_t req)
{
int ret = CRYPTO_FAILED;
aes_ctx_t *aes_ctx;
size_t saved_length, saved_offset, length_needed;
ASSERT(ctx->cc_provider_private != NULL);
aes_ctx = ctx->cc_provider_private;
/*
* For block ciphers, plaintext must be a multiple of AES block size.
* This test is only valid for ciphers whose blocksize is a power of 2.
*/
if (((aes_ctx->ac_flags & (CTR_MODE|CCM_MODE|GCM_MODE|GMAC_MODE))
== 0) && (plaintext->cd_length & (AES_BLOCK_LEN - 1)) != 0)
return (CRYPTO_DATA_LEN_RANGE);
ASSERT(ciphertext != NULL);
/*
* We need to just return the length needed to store the output.
* We should not destroy the context for the following case.
*/
switch (aes_ctx->ac_flags & (CCM_MODE|GCM_MODE|GMAC_MODE)) {
case CCM_MODE:
length_needed = plaintext->cd_length + aes_ctx->ac_mac_len;
break;
case GCM_MODE:
length_needed = plaintext->cd_length + aes_ctx->ac_tag_len;
break;
case GMAC_MODE:
if (plaintext->cd_length != 0)
return (CRYPTO_ARGUMENTS_BAD);
length_needed = aes_ctx->ac_tag_len;
break;
default:
length_needed = plaintext->cd_length;
}
if (ciphertext->cd_length < length_needed) {
ciphertext->cd_length = length_needed;
return (CRYPTO_BUFFER_TOO_SMALL);
}
saved_length = ciphertext->cd_length;
saved_offset = ciphertext->cd_offset;
/*
* Do an update on the specified input data.
*/
ret = aes_encrypt_update(ctx, plaintext, ciphertext, req);
if (ret != CRYPTO_SUCCESS) {
return (ret);
}
/*
* For CCM mode, aes_ccm_encrypt_final() will take care of any
* left-over unprocessed data, and compute the MAC
*/
if (aes_ctx->ac_flags & CCM_MODE) {
/*
* ccm_encrypt_final() will compute the MAC and append
* it to existing ciphertext. So, need to adjust the left over
* length value accordingly
*/
/* order of following 2 lines MUST not be reversed */
ciphertext->cd_offset = ciphertext->cd_length;
ciphertext->cd_length = saved_length - ciphertext->cd_length;
ret = ccm_encrypt_final((ccm_ctx_t *)aes_ctx, ciphertext,
AES_BLOCK_LEN, aes_encrypt_block, aes_xor_block);
if (ret != CRYPTO_SUCCESS) {
return (ret);
}
if (plaintext != ciphertext) {
ciphertext->cd_length =
ciphertext->cd_offset - saved_offset;
}
ciphertext->cd_offset = saved_offset;
} else if (aes_ctx->ac_flags & (GCM_MODE|GMAC_MODE)) {
/*
* gcm_encrypt_final() will compute the MAC and append
* it to existing ciphertext. So, need to adjust the left over
* length value accordingly
*/
/* order of following 2 lines MUST not be reversed */
ciphertext->cd_offset = ciphertext->cd_length;
ciphertext->cd_length = saved_length - ciphertext->cd_length;
ret = gcm_encrypt_final((gcm_ctx_t *)aes_ctx, ciphertext,
AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
aes_xor_block);
if (ret != CRYPTO_SUCCESS) {
return (ret);
}
if (plaintext != ciphertext) {
ciphertext->cd_length =
ciphertext->cd_offset - saved_offset;
}
ciphertext->cd_offset = saved_offset;
}
ASSERT(aes_ctx->ac_remainder_len == 0);
(void) aes_free_context(ctx);
return (ret);
}
static int
aes_decrypt(crypto_ctx_t *ctx, crypto_data_t *ciphertext,
crypto_data_t *plaintext, crypto_req_handle_t req)
{
int ret = CRYPTO_FAILED;
aes_ctx_t *aes_ctx;
off_t saved_offset;
size_t saved_length, length_needed;
ASSERT(ctx->cc_provider_private != NULL);
aes_ctx = ctx->cc_provider_private;
/*
* For block ciphers, plaintext must be a multiple of AES block size.
* This test is only valid for ciphers whose blocksize is a power of 2.
*/
if (((aes_ctx->ac_flags & (CTR_MODE|CCM_MODE|GCM_MODE|GMAC_MODE))
== 0) && (ciphertext->cd_length & (AES_BLOCK_LEN - 1)) != 0) {
return (CRYPTO_ENCRYPTED_DATA_LEN_RANGE);
}
ASSERT(plaintext != NULL);
/*
* Return length needed to store the output.
* Do not destroy context when plaintext buffer is too small.
*
* CCM: plaintext is MAC len smaller than cipher text
* GCM: plaintext is TAG len smaller than cipher text
* GMAC: plaintext length must be zero
*/
switch (aes_ctx->ac_flags & (CCM_MODE|GCM_MODE|GMAC_MODE)) {
case CCM_MODE:
length_needed = aes_ctx->ac_processed_data_len;
break;
case GCM_MODE:
length_needed = ciphertext->cd_length - aes_ctx->ac_tag_len;
break;
case GMAC_MODE:
if (plaintext->cd_length != 0)
return (CRYPTO_ARGUMENTS_BAD);
length_needed = 0;
break;
default:
length_needed = ciphertext->cd_length;
}
if (plaintext->cd_length < length_needed) {
plaintext->cd_length = length_needed;
return (CRYPTO_BUFFER_TOO_SMALL);
}
saved_offset = plaintext->cd_offset;
saved_length = plaintext->cd_length;
/*
* Do an update on the specified input data.
*/
ret = aes_decrypt_update(ctx, ciphertext, plaintext, req);
if (ret != CRYPTO_SUCCESS) {
goto cleanup;
}
if (aes_ctx->ac_flags & CCM_MODE) {
ASSERT(aes_ctx->ac_processed_data_len == aes_ctx->ac_data_len);
ASSERT(aes_ctx->ac_processed_mac_len == aes_ctx->ac_mac_len);
/* order of following 2 lines MUST not be reversed */
plaintext->cd_offset = plaintext->cd_length;
plaintext->cd_length = saved_length - plaintext->cd_length;
ret = ccm_decrypt_final((ccm_ctx_t *)aes_ctx, plaintext,
AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
aes_xor_block);
if (ret == CRYPTO_SUCCESS) {
if (plaintext != ciphertext) {
plaintext->cd_length =
plaintext->cd_offset - saved_offset;
}
} else {
plaintext->cd_length = saved_length;
}
plaintext->cd_offset = saved_offset;
} else if (aes_ctx->ac_flags & (GCM_MODE|GMAC_MODE)) {
/* order of following 2 lines MUST not be reversed */
plaintext->cd_offset = plaintext->cd_length;
plaintext->cd_length = saved_length - plaintext->cd_length;
ret = gcm_decrypt_final((gcm_ctx_t *)aes_ctx, plaintext,
AES_BLOCK_LEN, aes_encrypt_block, aes_xor_block);
if (ret == CRYPTO_SUCCESS) {
if (plaintext != ciphertext) {
plaintext->cd_length =
plaintext->cd_offset - saved_offset;
}
} else {
plaintext->cd_length = saved_length;
}
plaintext->cd_offset = saved_offset;
}
ASSERT(aes_ctx->ac_remainder_len == 0);
cleanup:
(void) aes_free_context(ctx);
return (ret);
}
static int
aes_encrypt_update(crypto_ctx_t *ctx, crypto_data_t *plaintext,
crypto_data_t *ciphertext, crypto_req_handle_t req)
{
(void) req;
off_t saved_offset;
size_t saved_length, out_len;
int ret = CRYPTO_SUCCESS;
aes_ctx_t *aes_ctx;
ASSERT(ctx->cc_provider_private != NULL);
aes_ctx = ctx->cc_provider_private;
ASSERT(ciphertext != NULL);
/* compute number of bytes that will hold the ciphertext */
out_len = aes_ctx->ac_remainder_len;
out_len += plaintext->cd_length;
out_len &= ~(AES_BLOCK_LEN - 1);
/* return length needed to store the output */
if (ciphertext->cd_length < out_len) {
ciphertext->cd_length = out_len;
return (CRYPTO_BUFFER_TOO_SMALL);
}
saved_offset = ciphertext->cd_offset;
saved_length = ciphertext->cd_length;
/*
* Do the AES update on the specified input data.
*/
switch (plaintext->cd_format) {
case CRYPTO_DATA_RAW:
ret = crypto_update_iov(ctx->cc_provider_private,
plaintext, ciphertext, aes_encrypt_contiguous_blocks,
aes_copy_block64);
break;
case CRYPTO_DATA_UIO:
ret = crypto_update_uio(ctx->cc_provider_private,
plaintext, ciphertext, aes_encrypt_contiguous_blocks,
aes_copy_block64);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
/*
* Since AES counter mode is a stream cipher, we call
* ctr_mode_final() to pick up any remaining bytes.
* It is an internal function that does not destroy
* the context like *normal* final routines.
*/
if ((aes_ctx->ac_flags & CTR_MODE) && (aes_ctx->ac_remainder_len > 0)) {
ret = ctr_mode_final((ctr_ctx_t *)aes_ctx,
ciphertext, aes_encrypt_block);
}
if (ret == CRYPTO_SUCCESS) {
if (plaintext != ciphertext)
ciphertext->cd_length =
ciphertext->cd_offset - saved_offset;
} else {
ciphertext->cd_length = saved_length;
}
ciphertext->cd_offset = saved_offset;
return (ret);
}
static int
aes_decrypt_update(crypto_ctx_t *ctx, crypto_data_t *ciphertext,
crypto_data_t *plaintext, crypto_req_handle_t req)
{
off_t saved_offset;
size_t saved_length, out_len;
int ret = CRYPTO_SUCCESS;
aes_ctx_t *aes_ctx;
ASSERT(ctx->cc_provider_private != NULL);
aes_ctx = ctx->cc_provider_private;
ASSERT(plaintext != NULL);
/*
* Compute number of bytes that will hold the plaintext.
* This is not necessary for CCM, GCM, and GMAC since these
* mechanisms never return plaintext for update operations.
*/
if ((aes_ctx->ac_flags & (CCM_MODE|GCM_MODE|GMAC_MODE)) == 0) {
out_len = aes_ctx->ac_remainder_len;
out_len += ciphertext->cd_length;
out_len &= ~(AES_BLOCK_LEN - 1);
/* return length needed to store the output */
if (plaintext->cd_length < out_len) {
plaintext->cd_length = out_len;
return (CRYPTO_BUFFER_TOO_SMALL);
}
}
saved_offset = plaintext->cd_offset;
saved_length = plaintext->cd_length;
if (aes_ctx->ac_flags & (GCM_MODE|GMAC_MODE))
gcm_set_kmflag((gcm_ctx_t *)aes_ctx, crypto_kmflag(req));
/*
* Do the AES update on the specified input data.
*/
switch (ciphertext->cd_format) {
case CRYPTO_DATA_RAW:
ret = crypto_update_iov(ctx->cc_provider_private,
ciphertext, plaintext, aes_decrypt_contiguous_blocks,
aes_copy_block64);
break;
case CRYPTO_DATA_UIO:
ret = crypto_update_uio(ctx->cc_provider_private,
ciphertext, plaintext, aes_decrypt_contiguous_blocks,
aes_copy_block64);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
/*
* Since AES counter mode is a stream cipher, we call
* ctr_mode_final() to pick up any remaining bytes.
* It is an internal function that does not destroy
* the context like *normal* final routines.
*/
if ((aes_ctx->ac_flags & CTR_MODE) && (aes_ctx->ac_remainder_len > 0)) {
ret = ctr_mode_final((ctr_ctx_t *)aes_ctx, plaintext,
aes_encrypt_block);
if (ret == CRYPTO_DATA_LEN_RANGE)
ret = CRYPTO_ENCRYPTED_DATA_LEN_RANGE;
}
if (ret == CRYPTO_SUCCESS) {
if (ciphertext != plaintext)
plaintext->cd_length =
plaintext->cd_offset - saved_offset;
} else {
plaintext->cd_length = saved_length;
}
plaintext->cd_offset = saved_offset;
return (ret);
}
static int
aes_encrypt_final(crypto_ctx_t *ctx, crypto_data_t *data,
crypto_req_handle_t req)
{
(void) req;
aes_ctx_t *aes_ctx;
int ret;
ASSERT(ctx->cc_provider_private != NULL);
aes_ctx = ctx->cc_provider_private;
if (data->cd_format != CRYPTO_DATA_RAW &&
data->cd_format != CRYPTO_DATA_UIO) {
return (CRYPTO_ARGUMENTS_BAD);
}
if (aes_ctx->ac_flags & CTR_MODE) {
if (aes_ctx->ac_remainder_len > 0) {
ret = ctr_mode_final((ctr_ctx_t *)aes_ctx, data,
aes_encrypt_block);
if (ret != CRYPTO_SUCCESS)
return (ret);
}
} else if (aes_ctx->ac_flags & CCM_MODE) {
ret = ccm_encrypt_final((ccm_ctx_t *)aes_ctx, data,
AES_BLOCK_LEN, aes_encrypt_block, aes_xor_block);
if (ret != CRYPTO_SUCCESS) {
return (ret);
}
} else if (aes_ctx->ac_flags & (GCM_MODE|GMAC_MODE)) {
size_t saved_offset = data->cd_offset;
ret = gcm_encrypt_final((gcm_ctx_t *)aes_ctx, data,
AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
aes_xor_block);
if (ret != CRYPTO_SUCCESS) {
return (ret);
}
data->cd_length = data->cd_offset - saved_offset;
data->cd_offset = saved_offset;
} else {
/*
* There must be no unprocessed plaintext.
* This happens if the length of the last data is
* not a multiple of the AES block length.
*/
if (aes_ctx->ac_remainder_len > 0) {
return (CRYPTO_DATA_LEN_RANGE);
}
data->cd_length = 0;
}
(void) aes_free_context(ctx);
return (CRYPTO_SUCCESS);
}
static int
aes_decrypt_final(crypto_ctx_t *ctx, crypto_data_t *data,
crypto_req_handle_t req)
{
(void) req;
aes_ctx_t *aes_ctx;
int ret;
off_t saved_offset;
size_t saved_length;
ASSERT(ctx->cc_provider_private != NULL);
aes_ctx = ctx->cc_provider_private;
if (data->cd_format != CRYPTO_DATA_RAW &&
data->cd_format != CRYPTO_DATA_UIO) {
return (CRYPTO_ARGUMENTS_BAD);
}
/*
* There must be no unprocessed ciphertext.
* This happens if the length of the last ciphertext is
* not a multiple of the AES block length.
*/
if (aes_ctx->ac_remainder_len > 0) {
if ((aes_ctx->ac_flags & CTR_MODE) == 0)
return (CRYPTO_ENCRYPTED_DATA_LEN_RANGE);
else {
ret = ctr_mode_final((ctr_ctx_t *)aes_ctx, data,
aes_encrypt_block);
if (ret == CRYPTO_DATA_LEN_RANGE)
ret = CRYPTO_ENCRYPTED_DATA_LEN_RANGE;
if (ret != CRYPTO_SUCCESS)
return (ret);
}
}
if (aes_ctx->ac_flags & CCM_MODE) {
/*
* This is where all the plaintext is returned, make sure
* the plaintext buffer is big enough
*/
size_t pt_len = aes_ctx->ac_data_len;
if (data->cd_length < pt_len) {
data->cd_length = pt_len;
return (CRYPTO_BUFFER_TOO_SMALL);
}
ASSERT(aes_ctx->ac_processed_data_len == pt_len);
ASSERT(aes_ctx->ac_processed_mac_len == aes_ctx->ac_mac_len);
saved_offset = data->cd_offset;
saved_length = data->cd_length;
ret = ccm_decrypt_final((ccm_ctx_t *)aes_ctx, data,
AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
aes_xor_block);
if (ret == CRYPTO_SUCCESS) {
data->cd_length = data->cd_offset - saved_offset;
} else {
data->cd_length = saved_length;
}
data->cd_offset = saved_offset;
if (ret != CRYPTO_SUCCESS) {
return (ret);
}
} else if (aes_ctx->ac_flags & (GCM_MODE|GMAC_MODE)) {
/*
* This is where all the plaintext is returned, make sure
* the plaintext buffer is big enough
*/
gcm_ctx_t *ctx = (gcm_ctx_t *)aes_ctx;
size_t pt_len = ctx->gcm_processed_data_len - ctx->gcm_tag_len;
if (data->cd_length < pt_len) {
data->cd_length = pt_len;
return (CRYPTO_BUFFER_TOO_SMALL);
}
saved_offset = data->cd_offset;
saved_length = data->cd_length;
ret = gcm_decrypt_final((gcm_ctx_t *)aes_ctx, data,
AES_BLOCK_LEN, aes_encrypt_block, aes_xor_block);
if (ret == CRYPTO_SUCCESS) {
data->cd_length = data->cd_offset - saved_offset;
} else {
data->cd_length = saved_length;
}
data->cd_offset = saved_offset;
if (ret != CRYPTO_SUCCESS) {
return (ret);
}
}
if ((aes_ctx->ac_flags & (CTR_MODE|CCM_MODE|GCM_MODE|GMAC_MODE)) == 0) {
data->cd_length = 0;
}
(void) aes_free_context(ctx);
return (CRYPTO_SUCCESS);
}
static int
aes_encrypt_atomic(crypto_provider_handle_t provider,
crypto_session_id_t session_id, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_data_t *plaintext, crypto_data_t *ciphertext,
crypto_spi_ctx_template_t template, crypto_req_handle_t req)
{
(void) provider, (void) session_id;
aes_ctx_t aes_ctx; /* on the stack */
off_t saved_offset;
size_t saved_length;
size_t length_needed;
int ret;
ASSERT(ciphertext != NULL);
/*
* CTR, CCM, GCM, and GMAC modes do not require that plaintext
* be a multiple of AES block size.
*/
switch (mechanism->cm_type) {
case AES_CTR_MECH_INFO_TYPE:
case AES_CCM_MECH_INFO_TYPE:
case AES_GCM_MECH_INFO_TYPE:
case AES_GMAC_MECH_INFO_TYPE:
break;
default:
if ((plaintext->cd_length & (AES_BLOCK_LEN - 1)) != 0)
return (CRYPTO_DATA_LEN_RANGE);
}
if ((ret = aes_check_mech_param(mechanism, NULL, 0)) != CRYPTO_SUCCESS)
return (ret);
bzero(&aes_ctx, sizeof (aes_ctx_t));
ret = aes_common_init_ctx(&aes_ctx, template, mechanism, key,
crypto_kmflag(req), B_TRUE);
if (ret != CRYPTO_SUCCESS)
return (ret);
switch (mechanism->cm_type) {
case AES_CCM_MECH_INFO_TYPE:
length_needed = plaintext->cd_length + aes_ctx.ac_mac_len;
break;
case AES_GMAC_MECH_INFO_TYPE:
if (plaintext->cd_length != 0)
return (CRYPTO_ARGUMENTS_BAD);
fallthrough;
case AES_GCM_MECH_INFO_TYPE:
length_needed = plaintext->cd_length + aes_ctx.ac_tag_len;
break;
default:
length_needed = plaintext->cd_length;
}
/* return size of buffer needed to store output */
if (ciphertext->cd_length < length_needed) {
ciphertext->cd_length = length_needed;
ret = CRYPTO_BUFFER_TOO_SMALL;
goto out;
}
saved_offset = ciphertext->cd_offset;
saved_length = ciphertext->cd_length;
/*
* Do an update on the specified input data.
*/
switch (plaintext->cd_format) {
case CRYPTO_DATA_RAW:
ret = crypto_update_iov(&aes_ctx, plaintext, ciphertext,
aes_encrypt_contiguous_blocks, aes_copy_block64);
break;
case CRYPTO_DATA_UIO:
ret = crypto_update_uio(&aes_ctx, plaintext, ciphertext,
aes_encrypt_contiguous_blocks, aes_copy_block64);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
if (ret == CRYPTO_SUCCESS) {
if (mechanism->cm_type == AES_CCM_MECH_INFO_TYPE) {
ret = ccm_encrypt_final((ccm_ctx_t *)&aes_ctx,
ciphertext, AES_BLOCK_LEN, aes_encrypt_block,
aes_xor_block);
if (ret != CRYPTO_SUCCESS)
goto out;
ASSERT(aes_ctx.ac_remainder_len == 0);
} else if (mechanism->cm_type == AES_GCM_MECH_INFO_TYPE ||
mechanism->cm_type == AES_GMAC_MECH_INFO_TYPE) {
ret = gcm_encrypt_final((gcm_ctx_t *)&aes_ctx,
ciphertext, AES_BLOCK_LEN, aes_encrypt_block,
aes_copy_block, aes_xor_block);
if (ret != CRYPTO_SUCCESS)
goto out;
ASSERT(aes_ctx.ac_remainder_len == 0);
} else if (mechanism->cm_type == AES_CTR_MECH_INFO_TYPE) {
if (aes_ctx.ac_remainder_len > 0) {
ret = ctr_mode_final((ctr_ctx_t *)&aes_ctx,
ciphertext, aes_encrypt_block);
if (ret != CRYPTO_SUCCESS)
goto out;
}
} else {
ASSERT(aes_ctx.ac_remainder_len == 0);
}
if (plaintext != ciphertext) {
ciphertext->cd_length =
ciphertext->cd_offset - saved_offset;
}
} else {
ciphertext->cd_length = saved_length;
}
ciphertext->cd_offset = saved_offset;
out:
if (aes_ctx.ac_flags & PROVIDER_OWNS_KEY_SCHEDULE) {
bzero(aes_ctx.ac_keysched, aes_ctx.ac_keysched_len);
kmem_free(aes_ctx.ac_keysched, aes_ctx.ac_keysched_len);
}
#ifdef CAN_USE_GCM_ASM
if (aes_ctx.ac_flags & (GCM_MODE|GMAC_MODE) &&
((gcm_ctx_t *)&aes_ctx)->gcm_Htable != NULL) {
gcm_ctx_t *ctx = (gcm_ctx_t *)&aes_ctx;
bzero(ctx->gcm_Htable, ctx->gcm_htab_len);
kmem_free(ctx->gcm_Htable, ctx->gcm_htab_len);
}
#endif
return (ret);
}
static int
aes_decrypt_atomic(crypto_provider_handle_t provider,
crypto_session_id_t session_id, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_data_t *ciphertext, crypto_data_t *plaintext,
crypto_spi_ctx_template_t template, crypto_req_handle_t req)
{
(void) provider, (void) session_id;
aes_ctx_t aes_ctx; /* on the stack */
off_t saved_offset;
size_t saved_length;
size_t length_needed;
int ret;
ASSERT(plaintext != NULL);
/*
* CCM, GCM, CTR, and GMAC modes do not require that ciphertext
* be a multiple of AES block size.
*/
switch (mechanism->cm_type) {
case AES_CTR_MECH_INFO_TYPE:
case AES_CCM_MECH_INFO_TYPE:
case AES_GCM_MECH_INFO_TYPE:
case AES_GMAC_MECH_INFO_TYPE:
break;
default:
if ((ciphertext->cd_length & (AES_BLOCK_LEN - 1)) != 0)
return (CRYPTO_ENCRYPTED_DATA_LEN_RANGE);
}
if ((ret = aes_check_mech_param(mechanism, NULL, 0)) != CRYPTO_SUCCESS)
return (ret);
bzero(&aes_ctx, sizeof (aes_ctx_t));
ret = aes_common_init_ctx(&aes_ctx, template, mechanism, key,
crypto_kmflag(req), B_FALSE);
if (ret != CRYPTO_SUCCESS)
return (ret);
switch (mechanism->cm_type) {
case AES_CCM_MECH_INFO_TYPE:
length_needed = aes_ctx.ac_data_len;
break;
case AES_GCM_MECH_INFO_TYPE:
length_needed = ciphertext->cd_length - aes_ctx.ac_tag_len;
break;
case AES_GMAC_MECH_INFO_TYPE:
if (plaintext->cd_length != 0)
return (CRYPTO_ARGUMENTS_BAD);
length_needed = 0;
break;
default:
length_needed = ciphertext->cd_length;
}
/* return size of buffer needed to store output */
if (plaintext->cd_length < length_needed) {
plaintext->cd_length = length_needed;
ret = CRYPTO_BUFFER_TOO_SMALL;
goto out;
}
saved_offset = plaintext->cd_offset;
saved_length = plaintext->cd_length;
if (mechanism->cm_type == AES_GCM_MECH_INFO_TYPE ||
mechanism->cm_type == AES_GMAC_MECH_INFO_TYPE)
gcm_set_kmflag((gcm_ctx_t *)&aes_ctx, crypto_kmflag(req));
/*
* Do an update on the specified input data.
*/
switch (ciphertext->cd_format) {
case CRYPTO_DATA_RAW:
ret = crypto_update_iov(&aes_ctx, ciphertext, plaintext,
aes_decrypt_contiguous_blocks, aes_copy_block64);
break;
case CRYPTO_DATA_UIO:
ret = crypto_update_uio(&aes_ctx, ciphertext, plaintext,
aes_decrypt_contiguous_blocks, aes_copy_block64);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
if (ret == CRYPTO_SUCCESS) {
if (mechanism->cm_type == AES_CCM_MECH_INFO_TYPE) {
ASSERT(aes_ctx.ac_processed_data_len
== aes_ctx.ac_data_len);
ASSERT(aes_ctx.ac_processed_mac_len
== aes_ctx.ac_mac_len);
ret = ccm_decrypt_final((ccm_ctx_t *)&aes_ctx,
plaintext, AES_BLOCK_LEN, aes_encrypt_block,
aes_copy_block, aes_xor_block);
ASSERT(aes_ctx.ac_remainder_len == 0);
if ((ret == CRYPTO_SUCCESS) &&
(ciphertext != plaintext)) {
plaintext->cd_length =
plaintext->cd_offset - saved_offset;
} else {
plaintext->cd_length = saved_length;
}
} else if (mechanism->cm_type == AES_GCM_MECH_INFO_TYPE ||
mechanism->cm_type == AES_GMAC_MECH_INFO_TYPE) {
ret = gcm_decrypt_final((gcm_ctx_t *)&aes_ctx,
plaintext, AES_BLOCK_LEN, aes_encrypt_block,
aes_xor_block);
ASSERT(aes_ctx.ac_remainder_len == 0);
if ((ret == CRYPTO_SUCCESS) &&
(ciphertext != plaintext)) {
plaintext->cd_length =
plaintext->cd_offset - saved_offset;
} else {
plaintext->cd_length = saved_length;
}
} else if (mechanism->cm_type != AES_CTR_MECH_INFO_TYPE) {
ASSERT(aes_ctx.ac_remainder_len == 0);
if (ciphertext != plaintext)
plaintext->cd_length =
plaintext->cd_offset - saved_offset;
} else {
if (aes_ctx.ac_remainder_len > 0) {
ret = ctr_mode_final((ctr_ctx_t *)&aes_ctx,
plaintext, aes_encrypt_block);
if (ret == CRYPTO_DATA_LEN_RANGE)
ret = CRYPTO_ENCRYPTED_DATA_LEN_RANGE;
if (ret != CRYPTO_SUCCESS)
goto out;
}
if (ciphertext != plaintext)
plaintext->cd_length =
plaintext->cd_offset - saved_offset;
}
} else {
plaintext->cd_length = saved_length;
}
plaintext->cd_offset = saved_offset;
out:
if (aes_ctx.ac_flags & PROVIDER_OWNS_KEY_SCHEDULE) {
bzero(aes_ctx.ac_keysched, aes_ctx.ac_keysched_len);
kmem_free(aes_ctx.ac_keysched, aes_ctx.ac_keysched_len);
}
if (aes_ctx.ac_flags & CCM_MODE) {
if (aes_ctx.ac_pt_buf != NULL) {
vmem_free(aes_ctx.ac_pt_buf, aes_ctx.ac_data_len);
}
} else if (aes_ctx.ac_flags & (GCM_MODE|GMAC_MODE)) {
if (((gcm_ctx_t *)&aes_ctx)->gcm_pt_buf != NULL) {
vmem_free(((gcm_ctx_t *)&aes_ctx)->gcm_pt_buf,
((gcm_ctx_t *)&aes_ctx)->gcm_pt_buf_len);
}
#ifdef CAN_USE_GCM_ASM
if (((gcm_ctx_t *)&aes_ctx)->gcm_Htable != NULL) {
gcm_ctx_t *ctx = (gcm_ctx_t *)&aes_ctx;
bzero(ctx->gcm_Htable, ctx->gcm_htab_len);
kmem_free(ctx->gcm_Htable, ctx->gcm_htab_len);
}
#endif
}
return (ret);
}
/*
* KCF software provider context template entry points.
*/
static int
aes_create_ctx_template(crypto_provider_handle_t provider,
crypto_mechanism_t *mechanism, crypto_key_t *key,
crypto_spi_ctx_template_t *tmpl, size_t *tmpl_size, crypto_req_handle_t req)
{
(void) provider;
void *keysched;
size_t size;
int rv;
if (mechanism->cm_type != AES_ECB_MECH_INFO_TYPE &&
mechanism->cm_type != AES_CBC_MECH_INFO_TYPE &&
mechanism->cm_type != AES_CTR_MECH_INFO_TYPE &&
mechanism->cm_type != AES_CCM_MECH_INFO_TYPE &&
mechanism->cm_type != AES_GCM_MECH_INFO_TYPE &&
mechanism->cm_type != AES_GMAC_MECH_INFO_TYPE)
return (CRYPTO_MECHANISM_INVALID);
if ((keysched = aes_alloc_keysched(&size,
crypto_kmflag(req))) == NULL) {
return (CRYPTO_HOST_MEMORY);
}
/*
* Initialize key schedule. Key length information is stored
* in the key.
*/
if ((rv = init_keysched(key, keysched)) != CRYPTO_SUCCESS) {
bzero(keysched, size);
kmem_free(keysched, size);
return (rv);
}
*tmpl = keysched;
*tmpl_size = size;
return (CRYPTO_SUCCESS);
}
static int
aes_free_context(crypto_ctx_t *ctx)
{
aes_ctx_t *aes_ctx = ctx->cc_provider_private;
if (aes_ctx != NULL) {
if (aes_ctx->ac_flags & PROVIDER_OWNS_KEY_SCHEDULE) {
ASSERT(aes_ctx->ac_keysched_len != 0);
bzero(aes_ctx->ac_keysched, aes_ctx->ac_keysched_len);
kmem_free(aes_ctx->ac_keysched,
aes_ctx->ac_keysched_len);
}
crypto_free_mode_ctx(aes_ctx);
ctx->cc_provider_private = NULL;
}
return (CRYPTO_SUCCESS);
}
static int
aes_common_init_ctx(aes_ctx_t *aes_ctx, crypto_spi_ctx_template_t *template,
crypto_mechanism_t *mechanism, crypto_key_t *key, int kmflag,
boolean_t is_encrypt_init)
{
int rv = CRYPTO_SUCCESS;
void *keysched;
size_t size = 0;
if (template == NULL) {
if ((keysched = aes_alloc_keysched(&size, kmflag)) == NULL)
return (CRYPTO_HOST_MEMORY);
/*
* Initialize key schedule.
* Key length is stored in the key.
*/
if ((rv = init_keysched(key, keysched)) != CRYPTO_SUCCESS) {
kmem_free(keysched, size);
return (rv);
}
aes_ctx->ac_flags |= PROVIDER_OWNS_KEY_SCHEDULE;
aes_ctx->ac_keysched_len = size;
} else {
keysched = template;
}
aes_ctx->ac_keysched = keysched;
switch (mechanism->cm_type) {
case AES_CBC_MECH_INFO_TYPE:
rv = cbc_init_ctx((cbc_ctx_t *)aes_ctx, mechanism->cm_param,
mechanism->cm_param_len, AES_BLOCK_LEN, aes_copy_block64);
break;
case AES_CTR_MECH_INFO_TYPE: {
CK_AES_CTR_PARAMS *pp;
if (mechanism->cm_param == NULL ||
mechanism->cm_param_len != sizeof (CK_AES_CTR_PARAMS)) {
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
pp = (CK_AES_CTR_PARAMS *)(void *)mechanism->cm_param;
rv = ctr_init_ctx((ctr_ctx_t *)aes_ctx, pp->ulCounterBits,
pp->cb, aes_copy_block);
break;
}
case AES_CCM_MECH_INFO_TYPE:
if (mechanism->cm_param == NULL ||
mechanism->cm_param_len != sizeof (CK_AES_CCM_PARAMS)) {
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
rv = ccm_init_ctx((ccm_ctx_t *)aes_ctx, mechanism->cm_param,
kmflag, is_encrypt_init, AES_BLOCK_LEN, aes_encrypt_block,
aes_xor_block);
break;
case AES_GCM_MECH_INFO_TYPE:
if (mechanism->cm_param == NULL ||
mechanism->cm_param_len != sizeof (CK_AES_GCM_PARAMS)) {
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
rv = gcm_init_ctx((gcm_ctx_t *)aes_ctx, mechanism->cm_param,
AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
aes_xor_block);
break;
case AES_GMAC_MECH_INFO_TYPE:
if (mechanism->cm_param == NULL ||
mechanism->cm_param_len != sizeof (CK_AES_GMAC_PARAMS)) {
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
rv = gmac_init_ctx((gcm_ctx_t *)aes_ctx, mechanism->cm_param,
AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
aes_xor_block);
break;
case AES_ECB_MECH_INFO_TYPE:
aes_ctx->ac_flags |= ECB_MODE;
}
if (rv != CRYPTO_SUCCESS) {
if (aes_ctx->ac_flags & PROVIDER_OWNS_KEY_SCHEDULE) {
bzero(keysched, size);
kmem_free(keysched, size);
}
}
return (rv);
}
static int
process_gmac_mech(crypto_mechanism_t *mech, crypto_data_t *data,
CK_AES_GCM_PARAMS *gcm_params)
{
/* LINTED: pointer alignment */
CK_AES_GMAC_PARAMS *params = (CK_AES_GMAC_PARAMS *)mech->cm_param;
if (mech->cm_type != AES_GMAC_MECH_INFO_TYPE)
return (CRYPTO_MECHANISM_INVALID);
if (mech->cm_param_len != sizeof (CK_AES_GMAC_PARAMS))
return (CRYPTO_MECHANISM_PARAM_INVALID);
if (params->pIv == NULL)
return (CRYPTO_MECHANISM_PARAM_INVALID);
gcm_params->pIv = params->pIv;
gcm_params->ulIvLen = AES_GMAC_IV_LEN;
gcm_params->ulTagBits = AES_GMAC_TAG_BITS;
if (data == NULL)
return (CRYPTO_SUCCESS);
if (data->cd_format != CRYPTO_DATA_RAW)
return (CRYPTO_ARGUMENTS_BAD);
gcm_params->pAAD = (uchar_t *)data->cd_raw.iov_base;
gcm_params->ulAADLen = data->cd_length;
return (CRYPTO_SUCCESS);
}
static int
aes_mac_atomic(crypto_provider_handle_t provider,
crypto_session_id_t session_id, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_data_t *data, crypto_data_t *mac,
crypto_spi_ctx_template_t template, crypto_req_handle_t req)
{
CK_AES_GCM_PARAMS gcm_params;
crypto_mechanism_t gcm_mech;
int rv;
if ((rv = process_gmac_mech(mechanism, data, &gcm_params))
!= CRYPTO_SUCCESS)
return (rv);
gcm_mech.cm_type = AES_GCM_MECH_INFO_TYPE;
gcm_mech.cm_param_len = sizeof (CK_AES_GCM_PARAMS);
gcm_mech.cm_param = (char *)&gcm_params;
return (aes_encrypt_atomic(provider, session_id, &gcm_mech,
key, &null_crypto_data, mac, template, req));
}
static int
aes_mac_verify_atomic(crypto_provider_handle_t provider,
crypto_session_id_t session_id, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_data_t *data, crypto_data_t *mac,
crypto_spi_ctx_template_t template, crypto_req_handle_t req)
{
CK_AES_GCM_PARAMS gcm_params;
crypto_mechanism_t gcm_mech;
int rv;
if ((rv = process_gmac_mech(mechanism, data, &gcm_params))
!= CRYPTO_SUCCESS)
return (rv);
gcm_mech.cm_type = AES_GCM_MECH_INFO_TYPE;
gcm_mech.cm_param_len = sizeof (CK_AES_GCM_PARAMS);
gcm_mech.cm_param = (char *)&gcm_params;
return (aes_decrypt_atomic(provider, session_id, &gcm_mech,
key, mac, &null_crypto_data, template, req));
}
diff --git a/module/icp/io/sha2_mod.c b/module/icp/io/sha2_mod.c
index c096a3ce1f45..a43c7c5b7b8e 100644
--- a/module/icp/io/sha2_mod.c
+++ b/module/icp/io/sha2_mod.c
@@ -1,1377 +1,1377 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2010 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
#include <sys/crypto/spi.h>
#include <sys/crypto/icp.h>
#define _SHA2_IMPL
#include <sys/sha2.h>
#include <sha2/sha2_impl.h>
/*
* Macros to access the SHA2 or SHA2-HMAC contexts from a context passed
* by KCF to one of the entry points.
*/
#define PROV_SHA2_CTX(ctx) ((sha2_ctx_t *)(ctx)->cc_provider_private)
#define PROV_SHA2_HMAC_CTX(ctx) ((sha2_hmac_ctx_t *)(ctx)->cc_provider_private)
/* to extract the digest length passed as mechanism parameter */
#define PROV_SHA2_GET_DIGEST_LEN(m, len) { \
if (IS_P2ALIGNED((m)->cm_param, sizeof (ulong_t))) \
(len) = (uint32_t)*((ulong_t *)(m)->cm_param); \
else { \
ulong_t tmp_ulong; \
bcopy((m)->cm_param, &tmp_ulong, sizeof (ulong_t)); \
(len) = (uint32_t)tmp_ulong; \
} \
}
#define PROV_SHA2_DIGEST_KEY(mech, ctx, key, len, digest) { \
SHA2Init(mech, ctx); \
SHA2Update(ctx, key, len); \
SHA2Final(digest, ctx); \
}
/*
* Mechanism info structure passed to KCF during registration.
*/
-static crypto_mech_info_t sha2_mech_info_tab[] = {
+static const crypto_mech_info_t sha2_mech_info_tab[] = {
/* SHA256 */
{SUN_CKM_SHA256, SHA256_MECH_INFO_TYPE,
CRYPTO_FG_DIGEST | CRYPTO_FG_DIGEST_ATOMIC,
0, 0, CRYPTO_KEYSIZE_UNIT_IN_BITS},
/* SHA256-HMAC */
{SUN_CKM_SHA256_HMAC, SHA256_HMAC_MECH_INFO_TYPE,
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC,
SHA2_HMAC_MIN_KEY_LEN, SHA2_HMAC_MAX_KEY_LEN,
CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* SHA256-HMAC GENERAL */
{SUN_CKM_SHA256_HMAC_GENERAL, SHA256_HMAC_GEN_MECH_INFO_TYPE,
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC,
SHA2_HMAC_MIN_KEY_LEN, SHA2_HMAC_MAX_KEY_LEN,
CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* SHA384 */
{SUN_CKM_SHA384, SHA384_MECH_INFO_TYPE,
CRYPTO_FG_DIGEST | CRYPTO_FG_DIGEST_ATOMIC,
0, 0, CRYPTO_KEYSIZE_UNIT_IN_BITS},
/* SHA384-HMAC */
{SUN_CKM_SHA384_HMAC, SHA384_HMAC_MECH_INFO_TYPE,
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC,
SHA2_HMAC_MIN_KEY_LEN, SHA2_HMAC_MAX_KEY_LEN,
CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* SHA384-HMAC GENERAL */
{SUN_CKM_SHA384_HMAC_GENERAL, SHA384_HMAC_GEN_MECH_INFO_TYPE,
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC,
SHA2_HMAC_MIN_KEY_LEN, SHA2_HMAC_MAX_KEY_LEN,
CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* SHA512 */
{SUN_CKM_SHA512, SHA512_MECH_INFO_TYPE,
CRYPTO_FG_DIGEST | CRYPTO_FG_DIGEST_ATOMIC,
0, 0, CRYPTO_KEYSIZE_UNIT_IN_BITS},
/* SHA512-HMAC */
{SUN_CKM_SHA512_HMAC, SHA512_HMAC_MECH_INFO_TYPE,
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC,
SHA2_HMAC_MIN_KEY_LEN, SHA2_HMAC_MAX_KEY_LEN,
CRYPTO_KEYSIZE_UNIT_IN_BYTES},
/* SHA512-HMAC GENERAL */
{SUN_CKM_SHA512_HMAC_GENERAL, SHA512_HMAC_GEN_MECH_INFO_TYPE,
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC,
SHA2_HMAC_MIN_KEY_LEN, SHA2_HMAC_MAX_KEY_LEN,
CRYPTO_KEYSIZE_UNIT_IN_BYTES}
};
static void sha2_provider_status(crypto_provider_handle_t, uint_t *);
-static crypto_control_ops_t sha2_control_ops = {
+static const crypto_control_ops_t sha2_control_ops = {
sha2_provider_status
};
static int sha2_digest_init(crypto_ctx_t *, crypto_mechanism_t *,
crypto_req_handle_t);
static int sha2_digest(crypto_ctx_t *, crypto_data_t *, crypto_data_t *,
crypto_req_handle_t);
static int sha2_digest_update(crypto_ctx_t *, crypto_data_t *,
crypto_req_handle_t);
static int sha2_digest_final(crypto_ctx_t *, crypto_data_t *,
crypto_req_handle_t);
static int sha2_digest_atomic(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_data_t *, crypto_data_t *,
crypto_req_handle_t);
-static crypto_digest_ops_t sha2_digest_ops = {
+static const crypto_digest_ops_t sha2_digest_ops = {
.digest_init = sha2_digest_init,
.digest = sha2_digest,
.digest_update = sha2_digest_update,
.digest_key = NULL,
.digest_final = sha2_digest_final,
.digest_atomic = sha2_digest_atomic
};
static int sha2_mac_init(crypto_ctx_t *, crypto_mechanism_t *, crypto_key_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
static int sha2_mac_update(crypto_ctx_t *, crypto_data_t *,
crypto_req_handle_t);
static int sha2_mac_final(crypto_ctx_t *, crypto_data_t *, crypto_req_handle_t);
static int sha2_mac_atomic(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *, crypto_data_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
static int sha2_mac_verify_atomic(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *, crypto_data_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
-static crypto_mac_ops_t sha2_mac_ops = {
+static const crypto_mac_ops_t sha2_mac_ops = {
.mac_init = sha2_mac_init,
.mac = NULL,
.mac_update = sha2_mac_update,
.mac_final = sha2_mac_final,
.mac_atomic = sha2_mac_atomic,
.mac_verify_atomic = sha2_mac_verify_atomic
};
static int sha2_create_ctx_template(crypto_provider_handle_t,
crypto_mechanism_t *, crypto_key_t *, crypto_spi_ctx_template_t *,
size_t *, crypto_req_handle_t);
static int sha2_free_context(crypto_ctx_t *);
-static crypto_ctx_ops_t sha2_ctx_ops = {
+static const crypto_ctx_ops_t sha2_ctx_ops = {
.create_ctx_template = sha2_create_ctx_template,
.free_context = sha2_free_context
};
-static crypto_ops_t sha2_crypto_ops = {{{{{
+static const crypto_ops_t sha2_crypto_ops = {{{{{
&sha2_control_ops,
&sha2_digest_ops,
NULL,
&sha2_mac_ops,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
&sha2_ctx_ops
}}}}};
-static crypto_provider_info_t sha2_prov_info = {{{{
+static const crypto_provider_info_t sha2_prov_info = {{{{
CRYPTO_SPI_VERSION_1,
"SHA2 Software Provider",
CRYPTO_SW_PROVIDER,
NULL,
&sha2_crypto_ops,
- sizeof (sha2_mech_info_tab)/sizeof (crypto_mech_info_t),
+ sizeof (sha2_mech_info_tab) / sizeof (crypto_mech_info_t),
sha2_mech_info_tab
}}}};
static crypto_kcf_provider_handle_t sha2_prov_handle = 0;
int
sha2_mod_init(void)
{
int ret;
/*
* Register with KCF. If the registration fails, log an
* error but do not uninstall the module, since the functionality
* provided by misc/sha2 should still be available.
*/
if ((ret = crypto_register_provider(&sha2_prov_info,
&sha2_prov_handle)) != CRYPTO_SUCCESS)
cmn_err(CE_WARN, "sha2 _init: "
"crypto_register_provider() failed (0x%x)", ret);
return (0);
}
int
sha2_mod_fini(void)
{
int ret = 0;
if (sha2_prov_handle != 0) {
if ((ret = crypto_unregister_provider(sha2_prov_handle)) !=
CRYPTO_SUCCESS) {
cmn_err(CE_WARN,
"sha2 _fini: crypto_unregister_provider() "
"failed (0x%x)", ret);
return (EBUSY);
}
sha2_prov_handle = 0;
}
return (ret);
}
/*
* KCF software provider control entry points.
*/
static void
sha2_provider_status(crypto_provider_handle_t provider, uint_t *status)
{
(void) provider;
*status = CRYPTO_PROVIDER_READY;
}
/*
* KCF software provider digest entry points.
*/
static int
sha2_digest_init(crypto_ctx_t *ctx, crypto_mechanism_t *mechanism,
crypto_req_handle_t req)
{
/*
* Allocate and initialize SHA2 context.
*/
ctx->cc_provider_private = kmem_alloc(sizeof (sha2_ctx_t),
crypto_kmflag(req));
if (ctx->cc_provider_private == NULL)
return (CRYPTO_HOST_MEMORY);
PROV_SHA2_CTX(ctx)->sc_mech_type = mechanism->cm_type;
SHA2Init(mechanism->cm_type, &PROV_SHA2_CTX(ctx)->sc_sha2_ctx);
return (CRYPTO_SUCCESS);
}
/*
* Helper SHA2 digest update function for uio data.
*/
static int
sha2_digest_update_uio(SHA2_CTX *sha2_ctx, crypto_data_t *data)
{
off_t offset = data->cd_offset;
size_t length = data->cd_length;
uint_t vec_idx = 0;
size_t cur_len;
/* we support only kernel buffer */
if (zfs_uio_segflg(data->cd_uio) != UIO_SYSSPACE)
return (CRYPTO_ARGUMENTS_BAD);
/*
* Jump to the first iovec containing data to be
* digested.
*/
offset = zfs_uio_index_at_offset(data->cd_uio, offset, &vec_idx);
if (vec_idx == zfs_uio_iovcnt(data->cd_uio)) {
/*
* The caller specified an offset that is larger than the
* total size of the buffers it provided.
*/
return (CRYPTO_DATA_LEN_RANGE);
}
/*
* Now do the digesting on the iovecs.
*/
while (vec_idx < zfs_uio_iovcnt(data->cd_uio) && length > 0) {
cur_len = MIN(zfs_uio_iovlen(data->cd_uio, vec_idx) -
offset, length);
SHA2Update(sha2_ctx, (uint8_t *)zfs_uio_iovbase(data->cd_uio,
vec_idx) + offset, cur_len);
length -= cur_len;
vec_idx++;
offset = 0;
}
if (vec_idx == zfs_uio_iovcnt(data->cd_uio) && length > 0) {
/*
* The end of the specified iovec's was reached but
* the length requested could not be processed, i.e.
* The caller requested to digest more data than it provided.
*/
return (CRYPTO_DATA_LEN_RANGE);
}
return (CRYPTO_SUCCESS);
}
/*
* Helper SHA2 digest final function for uio data.
* digest_len is the length of the desired digest. If digest_len
* is smaller than the default SHA2 digest length, the caller
* must pass a scratch buffer, digest_scratch, which must
* be at least the algorithm's digest length bytes.
*/
static int
sha2_digest_final_uio(SHA2_CTX *sha2_ctx, crypto_data_t *digest,
ulong_t digest_len, uchar_t *digest_scratch)
{
off_t offset = digest->cd_offset;
uint_t vec_idx = 0;
/* we support only kernel buffer */
if (zfs_uio_segflg(digest->cd_uio) != UIO_SYSSPACE)
return (CRYPTO_ARGUMENTS_BAD);
/*
* Jump to the first iovec containing ptr to the digest to
* be returned.
*/
offset = zfs_uio_index_at_offset(digest->cd_uio, offset, &vec_idx);
if (vec_idx == zfs_uio_iovcnt(digest->cd_uio)) {
/*
* The caller specified an offset that is
* larger than the total size of the buffers
* it provided.
*/
return (CRYPTO_DATA_LEN_RANGE);
}
if (offset + digest_len <=
zfs_uio_iovlen(digest->cd_uio, vec_idx)) {
/*
* The computed SHA2 digest will fit in the current
* iovec.
*/
if (((sha2_ctx->algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) &&
(digest_len != SHA256_DIGEST_LENGTH)) ||
((sha2_ctx->algotype > SHA256_HMAC_GEN_MECH_INFO_TYPE) &&
(digest_len != SHA512_DIGEST_LENGTH))) {
/*
* The caller requested a short digest. Digest
* into a scratch buffer and return to
* the user only what was requested.
*/
SHA2Final(digest_scratch, sha2_ctx);
bcopy(digest_scratch, (uchar_t *)
zfs_uio_iovbase(digest->cd_uio, vec_idx) + offset,
digest_len);
} else {
SHA2Final((uchar_t *)zfs_uio_iovbase(digest->
cd_uio, vec_idx) + offset,
sha2_ctx);
}
} else {
/*
* The computed digest will be crossing one or more iovec's.
* This is bad performance-wise but we need to support it.
* Allocate a small scratch buffer on the stack and
* copy it piece meal to the specified digest iovec's.
*/
uchar_t digest_tmp[SHA512_DIGEST_LENGTH];
off_t scratch_offset = 0;
size_t length = digest_len;
size_t cur_len;
SHA2Final(digest_tmp, sha2_ctx);
while (vec_idx < zfs_uio_iovcnt(digest->cd_uio) && length > 0) {
cur_len =
MIN(zfs_uio_iovlen(digest->cd_uio, vec_idx) -
offset, length);
bcopy(digest_tmp + scratch_offset,
zfs_uio_iovbase(digest->cd_uio, vec_idx) + offset,
cur_len);
length -= cur_len;
vec_idx++;
scratch_offset += cur_len;
offset = 0;
}
if (vec_idx == zfs_uio_iovcnt(digest->cd_uio) && length > 0) {
/*
* The end of the specified iovec's was reached but
* the length requested could not be processed, i.e.
* The caller requested to digest more data than it
* provided.
*/
return (CRYPTO_DATA_LEN_RANGE);
}
}
return (CRYPTO_SUCCESS);
}
static int
sha2_digest(crypto_ctx_t *ctx, crypto_data_t *data, crypto_data_t *digest,
crypto_req_handle_t req)
{
(void) req;
int ret = CRYPTO_SUCCESS;
uint_t sha_digest_len;
ASSERT(ctx->cc_provider_private != NULL);
switch (PROV_SHA2_CTX(ctx)->sc_mech_type) {
case SHA256_MECH_INFO_TYPE:
sha_digest_len = SHA256_DIGEST_LENGTH;
break;
case SHA384_MECH_INFO_TYPE:
sha_digest_len = SHA384_DIGEST_LENGTH;
break;
case SHA512_MECH_INFO_TYPE:
sha_digest_len = SHA512_DIGEST_LENGTH;
break;
default:
return (CRYPTO_MECHANISM_INVALID);
}
/*
* We need to just return the length needed to store the output.
* We should not destroy the context for the following cases.
*/
if ((digest->cd_length == 0) ||
(digest->cd_length < sha_digest_len)) {
digest->cd_length = sha_digest_len;
return (CRYPTO_BUFFER_TOO_SMALL);
}
/*
* Do the SHA2 update on the specified input data.
*/
switch (data->cd_format) {
case CRYPTO_DATA_RAW:
SHA2Update(&PROV_SHA2_CTX(ctx)->sc_sha2_ctx,
(uint8_t *)data->cd_raw.iov_base + data->cd_offset,
data->cd_length);
break;
case CRYPTO_DATA_UIO:
ret = sha2_digest_update_uio(&PROV_SHA2_CTX(ctx)->sc_sha2_ctx,
data);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
if (ret != CRYPTO_SUCCESS) {
/* the update failed, free context and bail */
kmem_free(ctx->cc_provider_private, sizeof (sha2_ctx_t));
ctx->cc_provider_private = NULL;
digest->cd_length = 0;
return (ret);
}
/*
* Do a SHA2 final, must be done separately since the digest
* type can be different than the input data type.
*/
switch (digest->cd_format) {
case CRYPTO_DATA_RAW:
SHA2Final((unsigned char *)digest->cd_raw.iov_base +
digest->cd_offset, &PROV_SHA2_CTX(ctx)->sc_sha2_ctx);
break;
case CRYPTO_DATA_UIO:
ret = sha2_digest_final_uio(&PROV_SHA2_CTX(ctx)->sc_sha2_ctx,
digest, sha_digest_len, NULL);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
/* all done, free context and return */
if (ret == CRYPTO_SUCCESS)
digest->cd_length = sha_digest_len;
else
digest->cd_length = 0;
kmem_free(ctx->cc_provider_private, sizeof (sha2_ctx_t));
ctx->cc_provider_private = NULL;
return (ret);
}
static int
sha2_digest_update(crypto_ctx_t *ctx, crypto_data_t *data,
crypto_req_handle_t req)
{
(void) req;
int ret = CRYPTO_SUCCESS;
ASSERT(ctx->cc_provider_private != NULL);
/*
* Do the SHA2 update on the specified input data.
*/
switch (data->cd_format) {
case CRYPTO_DATA_RAW:
SHA2Update(&PROV_SHA2_CTX(ctx)->sc_sha2_ctx,
(uint8_t *)data->cd_raw.iov_base + data->cd_offset,
data->cd_length);
break;
case CRYPTO_DATA_UIO:
ret = sha2_digest_update_uio(&PROV_SHA2_CTX(ctx)->sc_sha2_ctx,
data);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
return (ret);
}
static int
sha2_digest_final(crypto_ctx_t *ctx, crypto_data_t *digest,
crypto_req_handle_t req)
{
(void) req;
int ret = CRYPTO_SUCCESS;
uint_t sha_digest_len;
ASSERT(ctx->cc_provider_private != NULL);
switch (PROV_SHA2_CTX(ctx)->sc_mech_type) {
case SHA256_MECH_INFO_TYPE:
sha_digest_len = SHA256_DIGEST_LENGTH;
break;
case SHA384_MECH_INFO_TYPE:
sha_digest_len = SHA384_DIGEST_LENGTH;
break;
case SHA512_MECH_INFO_TYPE:
sha_digest_len = SHA512_DIGEST_LENGTH;
break;
default:
return (CRYPTO_MECHANISM_INVALID);
}
/*
* We need to just return the length needed to store the output.
* We should not destroy the context for the following cases.
*/
if ((digest->cd_length == 0) ||
(digest->cd_length < sha_digest_len)) {
digest->cd_length = sha_digest_len;
return (CRYPTO_BUFFER_TOO_SMALL);
}
/*
* Do a SHA2 final.
*/
switch (digest->cd_format) {
case CRYPTO_DATA_RAW:
SHA2Final((unsigned char *)digest->cd_raw.iov_base +
digest->cd_offset, &PROV_SHA2_CTX(ctx)->sc_sha2_ctx);
break;
case CRYPTO_DATA_UIO:
ret = sha2_digest_final_uio(&PROV_SHA2_CTX(ctx)->sc_sha2_ctx,
digest, sha_digest_len, NULL);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
/* all done, free context and return */
if (ret == CRYPTO_SUCCESS)
digest->cd_length = sha_digest_len;
else
digest->cd_length = 0;
kmem_free(ctx->cc_provider_private, sizeof (sha2_ctx_t));
ctx->cc_provider_private = NULL;
return (ret);
}
static int
sha2_digest_atomic(crypto_provider_handle_t provider,
crypto_session_id_t session_id, crypto_mechanism_t *mechanism,
crypto_data_t *data, crypto_data_t *digest,
crypto_req_handle_t req)
{
(void) provider, (void) session_id, (void) req;
int ret = CRYPTO_SUCCESS;
SHA2_CTX sha2_ctx;
uint32_t sha_digest_len;
/*
* Do the SHA inits.
*/
SHA2Init(mechanism->cm_type, &sha2_ctx);
switch (data->cd_format) {
case CRYPTO_DATA_RAW:
SHA2Update(&sha2_ctx, (uint8_t *)data->
cd_raw.iov_base + data->cd_offset, data->cd_length);
break;
case CRYPTO_DATA_UIO:
ret = sha2_digest_update_uio(&sha2_ctx, data);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
/*
* Do the SHA updates on the specified input data.
*/
if (ret != CRYPTO_SUCCESS) {
/* the update failed, bail */
digest->cd_length = 0;
return (ret);
}
if (mechanism->cm_type <= SHA256_HMAC_GEN_MECH_INFO_TYPE)
sha_digest_len = SHA256_DIGEST_LENGTH;
else
sha_digest_len = SHA512_DIGEST_LENGTH;
/*
* Do a SHA2 final, must be done separately since the digest
* type can be different than the input data type.
*/
switch (digest->cd_format) {
case CRYPTO_DATA_RAW:
SHA2Final((unsigned char *)digest->cd_raw.iov_base +
digest->cd_offset, &sha2_ctx);
break;
case CRYPTO_DATA_UIO:
ret = sha2_digest_final_uio(&sha2_ctx, digest,
sha_digest_len, NULL);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
if (ret == CRYPTO_SUCCESS)
digest->cd_length = sha_digest_len;
else
digest->cd_length = 0;
return (ret);
}
/*
* KCF software provider mac entry points.
*
* SHA2 HMAC is: SHA2(key XOR opad, SHA2(key XOR ipad, text))
*
* Init:
* The initialization routine initializes what we denote
* as the inner and outer contexts by doing
* - for inner context: SHA2(key XOR ipad)
* - for outer context: SHA2(key XOR opad)
*
* Update:
* Each subsequent SHA2 HMAC update will result in an
* update of the inner context with the specified data.
*
* Final:
* The SHA2 HMAC final will do a SHA2 final operation on the
* inner context, and the resulting digest will be used
* as the data for an update on the outer context. Last
* but not least, a SHA2 final on the outer context will
* be performed to obtain the SHA2 HMAC digest to return
* to the user.
*/
/*
* Initialize a SHA2-HMAC context.
*/
static void
sha2_mac_init_ctx(sha2_hmac_ctx_t *ctx, void *keyval, uint_t length_in_bytes)
{
uint64_t ipad[SHA512_HMAC_BLOCK_SIZE / sizeof (uint64_t)];
uint64_t opad[SHA512_HMAC_BLOCK_SIZE / sizeof (uint64_t)];
int i, block_size, blocks_per_int64;
/* Determine the block size */
if (ctx->hc_mech_type <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
block_size = SHA256_HMAC_BLOCK_SIZE;
blocks_per_int64 = SHA256_HMAC_BLOCK_SIZE / sizeof (uint64_t);
} else {
block_size = SHA512_HMAC_BLOCK_SIZE;
blocks_per_int64 = SHA512_HMAC_BLOCK_SIZE / sizeof (uint64_t);
}
(void) bzero(ipad, block_size);
(void) bzero(opad, block_size);
(void) bcopy(keyval, ipad, length_in_bytes);
(void) bcopy(keyval, opad, length_in_bytes);
/* XOR key with ipad (0x36) and opad (0x5c) */
for (i = 0; i < blocks_per_int64; i ++) {
ipad[i] ^= 0x3636363636363636;
opad[i] ^= 0x5c5c5c5c5c5c5c5c;
}
/* perform SHA2 on ipad */
SHA2Init(ctx->hc_mech_type, &ctx->hc_icontext);
SHA2Update(&ctx->hc_icontext, (uint8_t *)ipad, block_size);
/* perform SHA2 on opad */
SHA2Init(ctx->hc_mech_type, &ctx->hc_ocontext);
SHA2Update(&ctx->hc_ocontext, (uint8_t *)opad, block_size);
}
/*
*/
static int
sha2_mac_init(crypto_ctx_t *ctx, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_spi_ctx_template_t ctx_template,
crypto_req_handle_t req)
{
int ret = CRYPTO_SUCCESS;
uint_t keylen_in_bytes = CRYPTO_BITS2BYTES(key->ck_length);
uint_t sha_digest_len, sha_hmac_block_size;
/*
* Set the digest length and block size to values appropriate to the
* mechanism
*/
switch (mechanism->cm_type) {
case SHA256_HMAC_MECH_INFO_TYPE:
case SHA256_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = SHA256_DIGEST_LENGTH;
sha_hmac_block_size = SHA256_HMAC_BLOCK_SIZE;
break;
case SHA384_HMAC_MECH_INFO_TYPE:
case SHA384_HMAC_GEN_MECH_INFO_TYPE:
case SHA512_HMAC_MECH_INFO_TYPE:
case SHA512_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = SHA512_DIGEST_LENGTH;
sha_hmac_block_size = SHA512_HMAC_BLOCK_SIZE;
break;
default:
return (CRYPTO_MECHANISM_INVALID);
}
if (key->ck_format != CRYPTO_KEY_RAW)
return (CRYPTO_ARGUMENTS_BAD);
ctx->cc_provider_private = kmem_alloc(sizeof (sha2_hmac_ctx_t),
crypto_kmflag(req));
if (ctx->cc_provider_private == NULL)
return (CRYPTO_HOST_MEMORY);
PROV_SHA2_HMAC_CTX(ctx)->hc_mech_type = mechanism->cm_type;
if (ctx_template != NULL) {
/* reuse context template */
bcopy(ctx_template, PROV_SHA2_HMAC_CTX(ctx),
sizeof (sha2_hmac_ctx_t));
} else {
/* no context template, compute context */
if (keylen_in_bytes > sha_hmac_block_size) {
uchar_t digested_key[SHA512_DIGEST_LENGTH];
sha2_hmac_ctx_t *hmac_ctx = ctx->cc_provider_private;
/*
* Hash the passed-in key to get a smaller key.
* The inner context is used since it hasn't been
* initialized yet.
*/
PROV_SHA2_DIGEST_KEY(mechanism->cm_type / 3,
&hmac_ctx->hc_icontext,
key->ck_data, keylen_in_bytes, digested_key);
sha2_mac_init_ctx(PROV_SHA2_HMAC_CTX(ctx),
digested_key, sha_digest_len);
} else {
sha2_mac_init_ctx(PROV_SHA2_HMAC_CTX(ctx),
key->ck_data, keylen_in_bytes);
}
}
/*
* Get the mechanism parameters, if applicable.
*/
if (mechanism->cm_type % 3 == 2) {
if (mechanism->cm_param == NULL ||
mechanism->cm_param_len != sizeof (ulong_t))
ret = CRYPTO_MECHANISM_PARAM_INVALID;
PROV_SHA2_GET_DIGEST_LEN(mechanism,
PROV_SHA2_HMAC_CTX(ctx)->hc_digest_len);
if (PROV_SHA2_HMAC_CTX(ctx)->hc_digest_len > sha_digest_len)
ret = CRYPTO_MECHANISM_PARAM_INVALID;
}
if (ret != CRYPTO_SUCCESS) {
bzero(ctx->cc_provider_private, sizeof (sha2_hmac_ctx_t));
kmem_free(ctx->cc_provider_private, sizeof (sha2_hmac_ctx_t));
ctx->cc_provider_private = NULL;
}
return (ret);
}
static int
sha2_mac_update(crypto_ctx_t *ctx, crypto_data_t *data,
crypto_req_handle_t req)
{
(void) req;
int ret = CRYPTO_SUCCESS;
ASSERT(ctx->cc_provider_private != NULL);
/*
* Do a SHA2 update of the inner context using the specified
* data.
*/
switch (data->cd_format) {
case CRYPTO_DATA_RAW:
SHA2Update(&PROV_SHA2_HMAC_CTX(ctx)->hc_icontext,
(uint8_t *)data->cd_raw.iov_base + data->cd_offset,
data->cd_length);
break;
case CRYPTO_DATA_UIO:
ret = sha2_digest_update_uio(
&PROV_SHA2_HMAC_CTX(ctx)->hc_icontext, data);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
return (ret);
}
static int
sha2_mac_final(crypto_ctx_t *ctx, crypto_data_t *mac, crypto_req_handle_t req)
{
(void) req;
int ret = CRYPTO_SUCCESS;
uchar_t digest[SHA512_DIGEST_LENGTH];
uint32_t digest_len, sha_digest_len;
ASSERT(ctx->cc_provider_private != NULL);
/* Set the digest lengths to values appropriate to the mechanism */
switch (PROV_SHA2_HMAC_CTX(ctx)->hc_mech_type) {
case SHA256_HMAC_MECH_INFO_TYPE:
sha_digest_len = digest_len = SHA256_DIGEST_LENGTH;
break;
case SHA384_HMAC_MECH_INFO_TYPE:
sha_digest_len = digest_len = SHA384_DIGEST_LENGTH;
break;
case SHA512_HMAC_MECH_INFO_TYPE:
sha_digest_len = digest_len = SHA512_DIGEST_LENGTH;
break;
case SHA256_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = SHA256_DIGEST_LENGTH;
digest_len = PROV_SHA2_HMAC_CTX(ctx)->hc_digest_len;
break;
case SHA384_HMAC_GEN_MECH_INFO_TYPE:
case SHA512_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = SHA512_DIGEST_LENGTH;
digest_len = PROV_SHA2_HMAC_CTX(ctx)->hc_digest_len;
break;
default:
return (CRYPTO_ARGUMENTS_BAD);
}
/*
* We need to just return the length needed to store the output.
* We should not destroy the context for the following cases.
*/
if ((mac->cd_length == 0) || (mac->cd_length < digest_len)) {
mac->cd_length = digest_len;
return (CRYPTO_BUFFER_TOO_SMALL);
}
/*
* Do a SHA2 final on the inner context.
*/
SHA2Final(digest, &PROV_SHA2_HMAC_CTX(ctx)->hc_icontext);
/*
* Do a SHA2 update on the outer context, feeding the inner
* digest as data.
*/
SHA2Update(&PROV_SHA2_HMAC_CTX(ctx)->hc_ocontext, digest,
sha_digest_len);
/*
* Do a SHA2 final on the outer context, storing the computing
* digest in the users buffer.
*/
switch (mac->cd_format) {
case CRYPTO_DATA_RAW:
if (digest_len != sha_digest_len) {
/*
* The caller requested a short digest. Digest
* into a scratch buffer and return to
* the user only what was requested.
*/
SHA2Final(digest,
&PROV_SHA2_HMAC_CTX(ctx)->hc_ocontext);
bcopy(digest, (unsigned char *)mac->cd_raw.iov_base +
mac->cd_offset, digest_len);
} else {
SHA2Final((unsigned char *)mac->cd_raw.iov_base +
mac->cd_offset,
&PROV_SHA2_HMAC_CTX(ctx)->hc_ocontext);
}
break;
case CRYPTO_DATA_UIO:
ret = sha2_digest_final_uio(
&PROV_SHA2_HMAC_CTX(ctx)->hc_ocontext, mac,
digest_len, digest);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
if (ret == CRYPTO_SUCCESS)
mac->cd_length = digest_len;
else
mac->cd_length = 0;
bzero(ctx->cc_provider_private, sizeof (sha2_hmac_ctx_t));
kmem_free(ctx->cc_provider_private, sizeof (sha2_hmac_ctx_t));
ctx->cc_provider_private = NULL;
return (ret);
}
#define SHA2_MAC_UPDATE(data, ctx, ret) { \
switch (data->cd_format) { \
case CRYPTO_DATA_RAW: \
SHA2Update(&(ctx).hc_icontext, \
(uint8_t *)data->cd_raw.iov_base + \
data->cd_offset, data->cd_length); \
break; \
case CRYPTO_DATA_UIO: \
ret = sha2_digest_update_uio(&(ctx).hc_icontext, data); \
break; \
default: \
ret = CRYPTO_ARGUMENTS_BAD; \
} \
}
static int
sha2_mac_atomic(crypto_provider_handle_t provider,
crypto_session_id_t session_id, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_data_t *data, crypto_data_t *mac,
crypto_spi_ctx_template_t ctx_template, crypto_req_handle_t req)
{
(void) provider, (void) session_id, (void) req;
int ret = CRYPTO_SUCCESS;
uchar_t digest[SHA512_DIGEST_LENGTH];
sha2_hmac_ctx_t sha2_hmac_ctx;
uint32_t sha_digest_len, digest_len, sha_hmac_block_size;
uint_t keylen_in_bytes = CRYPTO_BITS2BYTES(key->ck_length);
/*
* Set the digest length and block size to values appropriate to the
* mechanism
*/
switch (mechanism->cm_type) {
case SHA256_HMAC_MECH_INFO_TYPE:
case SHA256_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = digest_len = SHA256_DIGEST_LENGTH;
sha_hmac_block_size = SHA256_HMAC_BLOCK_SIZE;
break;
case SHA384_HMAC_MECH_INFO_TYPE:
case SHA384_HMAC_GEN_MECH_INFO_TYPE:
case SHA512_HMAC_MECH_INFO_TYPE:
case SHA512_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = digest_len = SHA512_DIGEST_LENGTH;
sha_hmac_block_size = SHA512_HMAC_BLOCK_SIZE;
break;
default:
return (CRYPTO_MECHANISM_INVALID);
}
/* Add support for key by attributes (RFE 4706552) */
if (key->ck_format != CRYPTO_KEY_RAW)
return (CRYPTO_ARGUMENTS_BAD);
if (ctx_template != NULL) {
/* reuse context template */
bcopy(ctx_template, &sha2_hmac_ctx, sizeof (sha2_hmac_ctx_t));
} else {
sha2_hmac_ctx.hc_mech_type = mechanism->cm_type;
/* no context template, initialize context */
if (keylen_in_bytes > sha_hmac_block_size) {
/*
* Hash the passed-in key to get a smaller key.
* The inner context is used since it hasn't been
* initialized yet.
*/
PROV_SHA2_DIGEST_KEY(mechanism->cm_type / 3,
&sha2_hmac_ctx.hc_icontext,
key->ck_data, keylen_in_bytes, digest);
sha2_mac_init_ctx(&sha2_hmac_ctx, digest,
sha_digest_len);
} else {
sha2_mac_init_ctx(&sha2_hmac_ctx, key->ck_data,
keylen_in_bytes);
}
}
/* get the mechanism parameters, if applicable */
if ((mechanism->cm_type % 3) == 2) {
if (mechanism->cm_param == NULL ||
mechanism->cm_param_len != sizeof (ulong_t)) {
ret = CRYPTO_MECHANISM_PARAM_INVALID;
goto bail;
}
PROV_SHA2_GET_DIGEST_LEN(mechanism, digest_len);
if (digest_len > sha_digest_len) {
ret = CRYPTO_MECHANISM_PARAM_INVALID;
goto bail;
}
}
/* do a SHA2 update of the inner context using the specified data */
SHA2_MAC_UPDATE(data, sha2_hmac_ctx, ret);
if (ret != CRYPTO_SUCCESS)
/* the update failed, free context and bail */
goto bail;
/*
* Do a SHA2 final on the inner context.
*/
SHA2Final(digest, &sha2_hmac_ctx.hc_icontext);
/*
* Do an SHA2 update on the outer context, feeding the inner
* digest as data.
*
* HMAC-SHA384 needs special handling as the outer hash needs only 48
* bytes of the inner hash value.
*/
if (mechanism->cm_type == SHA384_HMAC_MECH_INFO_TYPE ||
mechanism->cm_type == SHA384_HMAC_GEN_MECH_INFO_TYPE)
SHA2Update(&sha2_hmac_ctx.hc_ocontext, digest,
SHA384_DIGEST_LENGTH);
else
SHA2Update(&sha2_hmac_ctx.hc_ocontext, digest, sha_digest_len);
/*
* Do a SHA2 final on the outer context, storing the computed
* digest in the users buffer.
*/
switch (mac->cd_format) {
case CRYPTO_DATA_RAW:
if (digest_len != sha_digest_len) {
/*
* The caller requested a short digest. Digest
* into a scratch buffer and return to
* the user only what was requested.
*/
SHA2Final(digest, &sha2_hmac_ctx.hc_ocontext);
bcopy(digest, (unsigned char *)mac->cd_raw.iov_base +
mac->cd_offset, digest_len);
} else {
SHA2Final((unsigned char *)mac->cd_raw.iov_base +
mac->cd_offset, &sha2_hmac_ctx.hc_ocontext);
}
break;
case CRYPTO_DATA_UIO:
ret = sha2_digest_final_uio(&sha2_hmac_ctx.hc_ocontext, mac,
digest_len, digest);
break;
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
if (ret == CRYPTO_SUCCESS) {
mac->cd_length = digest_len;
return (CRYPTO_SUCCESS);
}
bail:
bzero(&sha2_hmac_ctx, sizeof (sha2_hmac_ctx_t));
mac->cd_length = 0;
return (ret);
}
static int
sha2_mac_verify_atomic(crypto_provider_handle_t provider,
crypto_session_id_t session_id, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_data_t *data, crypto_data_t *mac,
crypto_spi_ctx_template_t ctx_template, crypto_req_handle_t req)
{
(void) provider, (void) session_id, (void) req;
int ret = CRYPTO_SUCCESS;
uchar_t digest[SHA512_DIGEST_LENGTH];
sha2_hmac_ctx_t sha2_hmac_ctx;
uint32_t sha_digest_len, digest_len, sha_hmac_block_size;
uint_t keylen_in_bytes = CRYPTO_BITS2BYTES(key->ck_length);
/*
* Set the digest length and block size to values appropriate to the
* mechanism
*/
switch (mechanism->cm_type) {
case SHA256_HMAC_MECH_INFO_TYPE:
case SHA256_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = digest_len = SHA256_DIGEST_LENGTH;
sha_hmac_block_size = SHA256_HMAC_BLOCK_SIZE;
break;
case SHA384_HMAC_MECH_INFO_TYPE:
case SHA384_HMAC_GEN_MECH_INFO_TYPE:
case SHA512_HMAC_MECH_INFO_TYPE:
case SHA512_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = digest_len = SHA512_DIGEST_LENGTH;
sha_hmac_block_size = SHA512_HMAC_BLOCK_SIZE;
break;
default:
return (CRYPTO_MECHANISM_INVALID);
}
/* Add support for key by attributes (RFE 4706552) */
if (key->ck_format != CRYPTO_KEY_RAW)
return (CRYPTO_ARGUMENTS_BAD);
if (ctx_template != NULL) {
/* reuse context template */
bcopy(ctx_template, &sha2_hmac_ctx, sizeof (sha2_hmac_ctx_t));
} else {
sha2_hmac_ctx.hc_mech_type = mechanism->cm_type;
/* no context template, initialize context */
if (keylen_in_bytes > sha_hmac_block_size) {
/*
* Hash the passed-in key to get a smaller key.
* The inner context is used since it hasn't been
* initialized yet.
*/
PROV_SHA2_DIGEST_KEY(mechanism->cm_type / 3,
&sha2_hmac_ctx.hc_icontext,
key->ck_data, keylen_in_bytes, digest);
sha2_mac_init_ctx(&sha2_hmac_ctx, digest,
sha_digest_len);
} else {
sha2_mac_init_ctx(&sha2_hmac_ctx, key->ck_data,
keylen_in_bytes);
}
}
/* get the mechanism parameters, if applicable */
if (mechanism->cm_type % 3 == 2) {
if (mechanism->cm_param == NULL ||
mechanism->cm_param_len != sizeof (ulong_t)) {
ret = CRYPTO_MECHANISM_PARAM_INVALID;
goto bail;
}
PROV_SHA2_GET_DIGEST_LEN(mechanism, digest_len);
if (digest_len > sha_digest_len) {
ret = CRYPTO_MECHANISM_PARAM_INVALID;
goto bail;
}
}
if (mac->cd_length != digest_len) {
ret = CRYPTO_INVALID_MAC;
goto bail;
}
/* do a SHA2 update of the inner context using the specified data */
SHA2_MAC_UPDATE(data, sha2_hmac_ctx, ret);
if (ret != CRYPTO_SUCCESS)
/* the update failed, free context and bail */
goto bail;
/* do a SHA2 final on the inner context */
SHA2Final(digest, &sha2_hmac_ctx.hc_icontext);
/*
* Do an SHA2 update on the outer context, feeding the inner
* digest as data.
*
* HMAC-SHA384 needs special handling as the outer hash needs only 48
* bytes of the inner hash value.
*/
if (mechanism->cm_type == SHA384_HMAC_MECH_INFO_TYPE ||
mechanism->cm_type == SHA384_HMAC_GEN_MECH_INFO_TYPE)
SHA2Update(&sha2_hmac_ctx.hc_ocontext, digest,
SHA384_DIGEST_LENGTH);
else
SHA2Update(&sha2_hmac_ctx.hc_ocontext, digest, sha_digest_len);
/*
* Do a SHA2 final on the outer context, storing the computed
* digest in the users buffer.
*/
SHA2Final(digest, &sha2_hmac_ctx.hc_ocontext);
/*
* Compare the computed digest against the expected digest passed
* as argument.
*/
switch (mac->cd_format) {
case CRYPTO_DATA_RAW:
if (bcmp(digest, (unsigned char *)mac->cd_raw.iov_base +
mac->cd_offset, digest_len) != 0)
ret = CRYPTO_INVALID_MAC;
break;
case CRYPTO_DATA_UIO: {
off_t offset = mac->cd_offset;
uint_t vec_idx = 0;
off_t scratch_offset = 0;
size_t length = digest_len;
size_t cur_len;
/* we support only kernel buffer */
if (zfs_uio_segflg(mac->cd_uio) != UIO_SYSSPACE)
return (CRYPTO_ARGUMENTS_BAD);
/* jump to the first iovec containing the expected digest */
offset = zfs_uio_index_at_offset(mac->cd_uio, offset, &vec_idx);
if (vec_idx == zfs_uio_iovcnt(mac->cd_uio)) {
/*
* The caller specified an offset that is
* larger than the total size of the buffers
* it provided.
*/
ret = CRYPTO_DATA_LEN_RANGE;
break;
}
/* do the comparison of computed digest vs specified one */
while (vec_idx < zfs_uio_iovcnt(mac->cd_uio) && length > 0) {
cur_len = MIN(zfs_uio_iovlen(mac->cd_uio, vec_idx) -
offset, length);
if (bcmp(digest + scratch_offset,
zfs_uio_iovbase(mac->cd_uio, vec_idx) + offset,
cur_len) != 0) {
ret = CRYPTO_INVALID_MAC;
break;
}
length -= cur_len;
vec_idx++;
scratch_offset += cur_len;
offset = 0;
}
break;
}
default:
ret = CRYPTO_ARGUMENTS_BAD;
}
return (ret);
bail:
bzero(&sha2_hmac_ctx, sizeof (sha2_hmac_ctx_t));
mac->cd_length = 0;
return (ret);
}
/*
* KCF software provider context management entry points.
*/
static int
sha2_create_ctx_template(crypto_provider_handle_t provider,
crypto_mechanism_t *mechanism, crypto_key_t *key,
crypto_spi_ctx_template_t *ctx_template, size_t *ctx_template_size,
crypto_req_handle_t req)
{
(void) provider;
sha2_hmac_ctx_t *sha2_hmac_ctx_tmpl;
uint_t keylen_in_bytes = CRYPTO_BITS2BYTES(key->ck_length);
uint32_t sha_digest_len, sha_hmac_block_size;
/*
* Set the digest length and block size to values appropriate to the
* mechanism
*/
switch (mechanism->cm_type) {
case SHA256_HMAC_MECH_INFO_TYPE:
case SHA256_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = SHA256_DIGEST_LENGTH;
sha_hmac_block_size = SHA256_HMAC_BLOCK_SIZE;
break;
case SHA384_HMAC_MECH_INFO_TYPE:
case SHA384_HMAC_GEN_MECH_INFO_TYPE:
case SHA512_HMAC_MECH_INFO_TYPE:
case SHA512_HMAC_GEN_MECH_INFO_TYPE:
sha_digest_len = SHA512_DIGEST_LENGTH;
sha_hmac_block_size = SHA512_HMAC_BLOCK_SIZE;
break;
default:
return (CRYPTO_MECHANISM_INVALID);
}
/* Add support for key by attributes (RFE 4706552) */
if (key->ck_format != CRYPTO_KEY_RAW)
return (CRYPTO_ARGUMENTS_BAD);
/*
* Allocate and initialize SHA2 context.
*/
sha2_hmac_ctx_tmpl = kmem_alloc(sizeof (sha2_hmac_ctx_t),
crypto_kmflag(req));
if (sha2_hmac_ctx_tmpl == NULL)
return (CRYPTO_HOST_MEMORY);
sha2_hmac_ctx_tmpl->hc_mech_type = mechanism->cm_type;
if (keylen_in_bytes > sha_hmac_block_size) {
uchar_t digested_key[SHA512_DIGEST_LENGTH];
/*
* Hash the passed-in key to get a smaller key.
* The inner context is used since it hasn't been
* initialized yet.
*/
PROV_SHA2_DIGEST_KEY(mechanism->cm_type / 3,
&sha2_hmac_ctx_tmpl->hc_icontext,
key->ck_data, keylen_in_bytes, digested_key);
sha2_mac_init_ctx(sha2_hmac_ctx_tmpl, digested_key,
sha_digest_len);
} else {
sha2_mac_init_ctx(sha2_hmac_ctx_tmpl, key->ck_data,
keylen_in_bytes);
}
*ctx_template = (crypto_spi_ctx_template_t)sha2_hmac_ctx_tmpl;
*ctx_template_size = sizeof (sha2_hmac_ctx_t);
return (CRYPTO_SUCCESS);
}
static int
sha2_free_context(crypto_ctx_t *ctx)
{
uint_t ctx_len;
if (ctx->cc_provider_private == NULL)
return (CRYPTO_SUCCESS);
/*
* We have to free either SHA2 or SHA2-HMAC contexts, which
* have different lengths.
*
* Note: Below is dependent on the mechanism ordering.
*/
if (PROV_SHA2_CTX(ctx)->sc_mech_type % 3 == 0)
ctx_len = sizeof (sha2_ctx_t);
else
ctx_len = sizeof (sha2_hmac_ctx_t);
bzero(ctx->cc_provider_private, ctx_len);
kmem_free(ctx->cc_provider_private, ctx_len);
ctx->cc_provider_private = NULL;
return (CRYPTO_SUCCESS);
}
diff --git a/module/icp/io/skein_mod.c b/module/icp/io/skein_mod.c
index 250c0001a6f4..d0917e71b12e 100644
--- a/module/icp/io/skein_mod.c
+++ b/module/icp/io/skein_mod.c
@@ -1,702 +1,702 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://opensource.org/licenses/CDDL-1.0.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2013 Saso Kiselkov. All rights reserved.
*/
#include <sys/crypto/common.h>
#include <sys/crypto/icp.h>
#include <sys/crypto/spi.h>
#include <sys/sysmacros.h>
#define SKEIN_MODULE_IMPL
#include <sys/skein.h>
-static crypto_mech_info_t skein_mech_info_tab[] = {
+static const crypto_mech_info_t skein_mech_info_tab[] = {
{CKM_SKEIN_256, SKEIN_256_MECH_INFO_TYPE,
CRYPTO_FG_DIGEST | CRYPTO_FG_DIGEST_ATOMIC,
0, 0, CRYPTO_KEYSIZE_UNIT_IN_BITS},
{CKM_SKEIN_256_MAC, SKEIN_256_MAC_MECH_INFO_TYPE,
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC, 1, INT_MAX,
CRYPTO_KEYSIZE_UNIT_IN_BYTES},
{CKM_SKEIN_512, SKEIN_512_MECH_INFO_TYPE,
CRYPTO_FG_DIGEST | CRYPTO_FG_DIGEST_ATOMIC,
0, 0, CRYPTO_KEYSIZE_UNIT_IN_BITS},
{CKM_SKEIN_512_MAC, SKEIN_512_MAC_MECH_INFO_TYPE,
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC, 1, INT_MAX,
CRYPTO_KEYSIZE_UNIT_IN_BYTES},
{CKM_SKEIN1024, SKEIN1024_MECH_INFO_TYPE,
CRYPTO_FG_DIGEST | CRYPTO_FG_DIGEST_ATOMIC,
0, 0, CRYPTO_KEYSIZE_UNIT_IN_BITS},
{CKM_SKEIN1024_MAC, SKEIN1024_MAC_MECH_INFO_TYPE,
CRYPTO_FG_MAC | CRYPTO_FG_MAC_ATOMIC, 1, INT_MAX,
CRYPTO_KEYSIZE_UNIT_IN_BYTES}
};
static void skein_provider_status(crypto_provider_handle_t, uint_t *);
-static crypto_control_ops_t skein_control_ops = {
+static const crypto_control_ops_t skein_control_ops = {
skein_provider_status
};
static int skein_digest_init(crypto_ctx_t *, crypto_mechanism_t *,
crypto_req_handle_t);
static int skein_digest(crypto_ctx_t *, crypto_data_t *, crypto_data_t *,
crypto_req_handle_t);
static int skein_update(crypto_ctx_t *, crypto_data_t *, crypto_req_handle_t);
static int skein_final(crypto_ctx_t *, crypto_data_t *, crypto_req_handle_t);
static int skein_digest_atomic(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_data_t *, crypto_data_t *,
crypto_req_handle_t);
-static crypto_digest_ops_t skein_digest_ops = {
+static const crypto_digest_ops_t skein_digest_ops = {
.digest_init = skein_digest_init,
.digest = skein_digest,
.digest_update = skein_update,
.digest_key = NULL,
.digest_final = skein_final,
.digest_atomic = skein_digest_atomic
};
static int skein_mac_init(crypto_ctx_t *, crypto_mechanism_t *, crypto_key_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
static int skein_mac_atomic(crypto_provider_handle_t, crypto_session_id_t,
crypto_mechanism_t *, crypto_key_t *, crypto_data_t *, crypto_data_t *,
crypto_spi_ctx_template_t, crypto_req_handle_t);
-static crypto_mac_ops_t skein_mac_ops = {
+static const crypto_mac_ops_t skein_mac_ops = {
.mac_init = skein_mac_init,
.mac = NULL,
.mac_update = skein_update, /* using regular digest update is OK here */
.mac_final = skein_final, /* using regular digest final is OK here */
.mac_atomic = skein_mac_atomic,
.mac_verify_atomic = NULL
};
static int skein_create_ctx_template(crypto_provider_handle_t,
crypto_mechanism_t *, crypto_key_t *, crypto_spi_ctx_template_t *,
size_t *, crypto_req_handle_t);
static int skein_free_context(crypto_ctx_t *);
-static crypto_ctx_ops_t skein_ctx_ops = {
+static const crypto_ctx_ops_t skein_ctx_ops = {
.create_ctx_template = skein_create_ctx_template,
.free_context = skein_free_context
};
-static crypto_ops_t skein_crypto_ops = {{{{{
+static const crypto_ops_t skein_crypto_ops = {{{{{
&skein_control_ops,
&skein_digest_ops,
NULL,
&skein_mac_ops,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
&skein_ctx_ops,
}}}}};
-static crypto_provider_info_t skein_prov_info = {{{{
+static const crypto_provider_info_t skein_prov_info = {{{{
CRYPTO_SPI_VERSION_1,
"Skein Software Provider",
CRYPTO_SW_PROVIDER,
NULL,
&skein_crypto_ops,
sizeof (skein_mech_info_tab) / sizeof (crypto_mech_info_t),
skein_mech_info_tab
}}}};
static crypto_kcf_provider_handle_t skein_prov_handle = 0;
typedef struct skein_ctx {
skein_mech_type_t sc_mech_type;
size_t sc_digest_bitlen;
/*LINTED(E_ANONYMOUS_UNION_DECL)*/
union {
Skein_256_Ctxt_t sc_256;
Skein_512_Ctxt_t sc_512;
Skein1024_Ctxt_t sc_1024;
};
} skein_ctx_t;
#define SKEIN_CTX(_ctx_) ((skein_ctx_t *)((_ctx_)->cc_provider_private))
#define SKEIN_CTX_LVALUE(_ctx_) (_ctx_)->cc_provider_private
#define SKEIN_OP(_skein_ctx, _op, ...) \
do { \
skein_ctx_t *sc = (_skein_ctx); \
switch (sc->sc_mech_type) { \
case SKEIN_256_MECH_INFO_TYPE: \
case SKEIN_256_MAC_MECH_INFO_TYPE: \
(void) Skein_256_ ## _op(&sc->sc_256, __VA_ARGS__);\
break; \
case SKEIN_512_MECH_INFO_TYPE: \
case SKEIN_512_MAC_MECH_INFO_TYPE: \
(void) Skein_512_ ## _op(&sc->sc_512, __VA_ARGS__);\
break; \
case SKEIN1024_MECH_INFO_TYPE: \
case SKEIN1024_MAC_MECH_INFO_TYPE: \
(void) Skein1024_ ## _op(&sc->sc_1024, __VA_ARGS__);\
break; \
} \
} while (0)
static int
skein_get_digest_bitlen(const crypto_mechanism_t *mechanism, size_t *result)
{
if (mechanism->cm_param != NULL) {
/*LINTED(E_BAD_PTR_CAST_ALIGN)*/
skein_param_t *param = (skein_param_t *)mechanism->cm_param;
if (mechanism->cm_param_len != sizeof (*param) ||
param->sp_digest_bitlen == 0) {
return (CRYPTO_MECHANISM_PARAM_INVALID);
}
*result = param->sp_digest_bitlen;
} else {
switch (mechanism->cm_type) {
case SKEIN_256_MECH_INFO_TYPE:
*result = 256;
break;
case SKEIN_512_MECH_INFO_TYPE:
*result = 512;
break;
case SKEIN1024_MECH_INFO_TYPE:
*result = 1024;
break;
default:
return (CRYPTO_MECHANISM_INVALID);
}
}
return (CRYPTO_SUCCESS);
}
int
skein_mod_init(void)
{
/*
* Try to register with KCF - failure shouldn't unload us, since we
* still may want to continue providing misc/skein functionality.
*/
(void) crypto_register_provider(&skein_prov_info, &skein_prov_handle);
return (0);
}
int
skein_mod_fini(void)
{
int ret = 0;
if (skein_prov_handle != 0) {
if ((ret = crypto_unregister_provider(skein_prov_handle)) !=
CRYPTO_SUCCESS) {
cmn_err(CE_WARN,
"skein _fini: crypto_unregister_provider() "
"failed (0x%x)", ret);
return (EBUSY);
}
skein_prov_handle = 0;
}
return (0);
}
/*
* KCF software provider control entry points.
*/
static void
skein_provider_status(crypto_provider_handle_t provider, uint_t *status)
{
(void) provider;
*status = CRYPTO_PROVIDER_READY;
}
/*
* General Skein hashing helper functions.
*/
/*
* Performs an Update on a context with uio input data.
*/
static int
skein_digest_update_uio(skein_ctx_t *ctx, const crypto_data_t *data)
{
off_t offset = data->cd_offset;
size_t length = data->cd_length;
uint_t vec_idx = 0;
size_t cur_len;
zfs_uio_t *uio = data->cd_uio;
/* we support only kernel buffer */
if (zfs_uio_segflg(uio) != UIO_SYSSPACE)
return (CRYPTO_ARGUMENTS_BAD);
/*
* Jump to the first iovec containing data to be
* digested.
*/
offset = zfs_uio_index_at_offset(uio, offset, &vec_idx);
if (vec_idx == zfs_uio_iovcnt(uio)) {
/*
* The caller specified an offset that is larger than the
* total size of the buffers it provided.
*/
return (CRYPTO_DATA_LEN_RANGE);
}
/*
* Now do the digesting on the iovecs.
*/
while (vec_idx < zfs_uio_iovcnt(uio) && length > 0) {
cur_len = MIN(zfs_uio_iovlen(uio, vec_idx) - offset, length);
SKEIN_OP(ctx, Update, (uint8_t *)zfs_uio_iovbase(uio, vec_idx)
+ offset, cur_len);
length -= cur_len;
vec_idx++;
offset = 0;
}
if (vec_idx == zfs_uio_iovcnt(uio) && length > 0) {
/*
* The end of the specified iovec's was reached but
* the length requested could not be processed, i.e.
* The caller requested to digest more data than it provided.
*/
return (CRYPTO_DATA_LEN_RANGE);
}
return (CRYPTO_SUCCESS);
}
/*
* Performs a Final on a context and writes to a uio digest output.
*/
static int
skein_digest_final_uio(skein_ctx_t *ctx, crypto_data_t *digest,
crypto_req_handle_t req)
{
off_t offset = digest->cd_offset;
uint_t vec_idx = 0;
zfs_uio_t *uio = digest->cd_uio;
/* we support only kernel buffer */
if (zfs_uio_segflg(uio) != UIO_SYSSPACE)
return (CRYPTO_ARGUMENTS_BAD);
/*
* Jump to the first iovec containing ptr to the digest to be returned.
*/
offset = zfs_uio_index_at_offset(uio, offset, &vec_idx);
if (vec_idx == zfs_uio_iovcnt(uio)) {
/*
* The caller specified an offset that is larger than the
* total size of the buffers it provided.
*/
return (CRYPTO_DATA_LEN_RANGE);
}
if (offset + CRYPTO_BITS2BYTES(ctx->sc_digest_bitlen) <=
zfs_uio_iovlen(uio, vec_idx)) {
/* The computed digest will fit in the current iovec. */
SKEIN_OP(ctx, Final,
(uchar_t *)zfs_uio_iovbase(uio, vec_idx) + offset);
} else {
uint8_t *digest_tmp;
off_t scratch_offset = 0;
size_t length = CRYPTO_BITS2BYTES(ctx->sc_digest_bitlen);
size_t cur_len;
digest_tmp = kmem_alloc(CRYPTO_BITS2BYTES(
ctx->sc_digest_bitlen), crypto_kmflag(req));
if (digest_tmp == NULL)
return (CRYPTO_HOST_MEMORY);
SKEIN_OP(ctx, Final, digest_tmp);
while (vec_idx < zfs_uio_iovcnt(uio) && length > 0) {
cur_len = MIN(zfs_uio_iovlen(uio, vec_idx) - offset,
length);
bcopy(digest_tmp + scratch_offset,
zfs_uio_iovbase(uio, vec_idx) + offset, cur_len);
length -= cur_len;
vec_idx++;
scratch_offset += cur_len;
offset = 0;
}
kmem_free(digest_tmp, CRYPTO_BITS2BYTES(ctx->sc_digest_bitlen));
if (vec_idx == zfs_uio_iovcnt(uio) && length > 0) {
/*
* The end of the specified iovec's was reached but
* the length requested could not be processed, i.e.
* The caller requested to digest more data than it
* provided.
*/
return (CRYPTO_DATA_LEN_RANGE);
}
}
return (CRYPTO_SUCCESS);
}
/*
* KCF software provider digest entry points.
*/
/*
* Initializes a skein digest context to the configuration in `mechanism'.
* The mechanism cm_type must be one of SKEIN_*_MECH_INFO_TYPE. The cm_param
* field may contain a skein_param_t structure indicating the length of the
* digest the algorithm should produce. Otherwise the default output lengths
* are applied (32 bytes for Skein-256, 64 bytes for Skein-512 and 128 bytes
* for Skein-1024).
*/
static int
skein_digest_init(crypto_ctx_t *ctx, crypto_mechanism_t *mechanism,
crypto_req_handle_t req)
{
int error = CRYPTO_SUCCESS;
if (!VALID_SKEIN_DIGEST_MECH(mechanism->cm_type))
return (CRYPTO_MECHANISM_INVALID);
SKEIN_CTX_LVALUE(ctx) = kmem_alloc(sizeof (*SKEIN_CTX(ctx)),
crypto_kmflag(req));
if (SKEIN_CTX(ctx) == NULL)
return (CRYPTO_HOST_MEMORY);
SKEIN_CTX(ctx)->sc_mech_type = mechanism->cm_type;
error = skein_get_digest_bitlen(mechanism,
&SKEIN_CTX(ctx)->sc_digest_bitlen);
if (error != CRYPTO_SUCCESS)
goto errout;
SKEIN_OP(SKEIN_CTX(ctx), Init, SKEIN_CTX(ctx)->sc_digest_bitlen);
return (CRYPTO_SUCCESS);
errout:
bzero(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
kmem_free(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
SKEIN_CTX_LVALUE(ctx) = NULL;
return (error);
}
/*
* Executes a skein_update and skein_digest on a pre-initialized crypto
* context in a single step. See the documentation to these functions to
* see what to pass here.
*/
static int
skein_digest(crypto_ctx_t *ctx, crypto_data_t *data, crypto_data_t *digest,
crypto_req_handle_t req)
{
int error = CRYPTO_SUCCESS;
ASSERT(SKEIN_CTX(ctx) != NULL);
if (digest->cd_length <
CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen)) {
digest->cd_length =
CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen);
return (CRYPTO_BUFFER_TOO_SMALL);
}
error = skein_update(ctx, data, req);
if (error != CRYPTO_SUCCESS) {
bzero(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
kmem_free(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
SKEIN_CTX_LVALUE(ctx) = NULL;
digest->cd_length = 0;
return (error);
}
error = skein_final(ctx, digest, req);
return (error);
}
/*
* Performs a skein Update with the input message in `data' (successive calls
* can push more data). This is used both for digest and MAC operation.
* Supported input data formats are raw, uio and mblk.
*/
static int
skein_update(crypto_ctx_t *ctx, crypto_data_t *data, crypto_req_handle_t req)
{
(void) req;
int error = CRYPTO_SUCCESS;
ASSERT(SKEIN_CTX(ctx) != NULL);
switch (data->cd_format) {
case CRYPTO_DATA_RAW:
SKEIN_OP(SKEIN_CTX(ctx), Update,
(uint8_t *)data->cd_raw.iov_base + data->cd_offset,
data->cd_length);
break;
case CRYPTO_DATA_UIO:
error = skein_digest_update_uio(SKEIN_CTX(ctx), data);
break;
default:
error = CRYPTO_ARGUMENTS_BAD;
}
return (error);
}
/*
* Performs a skein Final, writing the output to `digest'. This is used both
* for digest and MAC operation.
* Supported output digest formats are raw, uio and mblk.
*/
static int
skein_final(crypto_ctx_t *ctx, crypto_data_t *digest, crypto_req_handle_t req)
{
int error = CRYPTO_SUCCESS;
ASSERT(SKEIN_CTX(ctx) != NULL);
if (digest->cd_length <
CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen)) {
digest->cd_length =
CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen);
return (CRYPTO_BUFFER_TOO_SMALL);
}
switch (digest->cd_format) {
case CRYPTO_DATA_RAW:
SKEIN_OP(SKEIN_CTX(ctx), Final,
(uint8_t *)digest->cd_raw.iov_base + digest->cd_offset);
break;
case CRYPTO_DATA_UIO:
error = skein_digest_final_uio(SKEIN_CTX(ctx), digest, req);
break;
default:
error = CRYPTO_ARGUMENTS_BAD;
}
if (error == CRYPTO_SUCCESS)
digest->cd_length =
CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen);
else
digest->cd_length = 0;
bzero(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
kmem_free(SKEIN_CTX(ctx), sizeof (*(SKEIN_CTX(ctx))));
SKEIN_CTX_LVALUE(ctx) = NULL;
return (error);
}
/*
* Performs a full skein digest computation in a single call, configuring the
* algorithm according to `mechanism', reading the input to be digested from
* `data' and writing the output to `digest'.
* Supported input/output formats are raw, uio and mblk.
*/
static int
skein_digest_atomic(crypto_provider_handle_t provider,
crypto_session_id_t session_id, crypto_mechanism_t *mechanism,
crypto_data_t *data, crypto_data_t *digest, crypto_req_handle_t req)
{
(void) provider, (void) session_id, (void) req;
int error;
skein_ctx_t skein_ctx;
crypto_ctx_t ctx;
SKEIN_CTX_LVALUE(&ctx) = &skein_ctx;
/* Init */
if (!VALID_SKEIN_DIGEST_MECH(mechanism->cm_type))
return (CRYPTO_MECHANISM_INVALID);
skein_ctx.sc_mech_type = mechanism->cm_type;
error = skein_get_digest_bitlen(mechanism, &skein_ctx.sc_digest_bitlen);
if (error != CRYPTO_SUCCESS)
goto out;
SKEIN_OP(&skein_ctx, Init, skein_ctx.sc_digest_bitlen);
if ((error = skein_update(&ctx, data, digest)) != CRYPTO_SUCCESS)
goto out;
if ((error = skein_final(&ctx, data, digest)) != CRYPTO_SUCCESS)
goto out;
out:
if (error == CRYPTO_SUCCESS)
digest->cd_length =
CRYPTO_BITS2BYTES(skein_ctx.sc_digest_bitlen);
else
digest->cd_length = 0;
bzero(&skein_ctx, sizeof (skein_ctx));
return (error);
}
/*
* Helper function that builds a Skein MAC context from the provided
* mechanism and key.
*/
static int
skein_mac_ctx_build(skein_ctx_t *ctx, crypto_mechanism_t *mechanism,
crypto_key_t *key)
{
int error;
if (!VALID_SKEIN_MAC_MECH(mechanism->cm_type))
return (CRYPTO_MECHANISM_INVALID);
if (key->ck_format != CRYPTO_KEY_RAW)
return (CRYPTO_ARGUMENTS_BAD);
ctx->sc_mech_type = mechanism->cm_type;
error = skein_get_digest_bitlen(mechanism, &ctx->sc_digest_bitlen);
if (error != CRYPTO_SUCCESS)
return (error);
SKEIN_OP(ctx, InitExt, ctx->sc_digest_bitlen, 0, key->ck_data,
CRYPTO_BITS2BYTES(key->ck_length));
return (CRYPTO_SUCCESS);
}
/*
* KCF software provide mac entry points.
*/
/*
* Initializes a skein MAC context. You may pass a ctx_template, in which
* case the template will be reused to make initialization more efficient.
* Otherwise a new context will be constructed. The mechanism cm_type must
* be one of SKEIN_*_MAC_MECH_INFO_TYPE. Same as in skein_digest_init, you
* may pass a skein_param_t in cm_param to configure the length of the
* digest. The key must be in raw format.
*/
static int
skein_mac_init(crypto_ctx_t *ctx, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_spi_ctx_template_t ctx_template,
crypto_req_handle_t req)
{
int error;
SKEIN_CTX_LVALUE(ctx) = kmem_alloc(sizeof (*SKEIN_CTX(ctx)),
crypto_kmflag(req));
if (SKEIN_CTX(ctx) == NULL)
return (CRYPTO_HOST_MEMORY);
if (ctx_template != NULL) {
bcopy(ctx_template, SKEIN_CTX(ctx),
sizeof (*SKEIN_CTX(ctx)));
} else {
error = skein_mac_ctx_build(SKEIN_CTX(ctx), mechanism, key);
if (error != CRYPTO_SUCCESS)
goto errout;
}
return (CRYPTO_SUCCESS);
errout:
bzero(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
kmem_free(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
return (error);
}
/*
* The MAC update and final calls are reused from the regular digest code.
*/
/*
* Same as skein_digest_atomic, performs an atomic Skein MAC operation in
* one step. All the same properties apply to the arguments of this
* function as to those of the partial operations above.
*/
static int
skein_mac_atomic(crypto_provider_handle_t provider,
crypto_session_id_t session_id, crypto_mechanism_t *mechanism,
crypto_key_t *key, crypto_data_t *data, crypto_data_t *mac,
crypto_spi_ctx_template_t ctx_template, crypto_req_handle_t req)
{
/* faux crypto context just for skein_digest_{update,final} */
(void) provider, (void) session_id;
int error;
crypto_ctx_t ctx;
skein_ctx_t skein_ctx;
SKEIN_CTX_LVALUE(&ctx) = &skein_ctx;
if (ctx_template != NULL) {
bcopy(ctx_template, &skein_ctx, sizeof (skein_ctx));
} else {
error = skein_mac_ctx_build(&skein_ctx, mechanism, key);
if (error != CRYPTO_SUCCESS)
goto errout;
}
if ((error = skein_update(&ctx, data, req)) != CRYPTO_SUCCESS)
goto errout;
if ((error = skein_final(&ctx, mac, req)) != CRYPTO_SUCCESS)
goto errout;
return (CRYPTO_SUCCESS);
errout:
bzero(&skein_ctx, sizeof (skein_ctx));
return (error);
}
/*
* KCF software provider context management entry points.
*/
/*
* Constructs a context template for the Skein MAC algorithm. The same
* properties apply to the arguments of this function as to those of
* skein_mac_init.
*/
static int
skein_create_ctx_template(crypto_provider_handle_t provider,
crypto_mechanism_t *mechanism, crypto_key_t *key,
crypto_spi_ctx_template_t *ctx_template, size_t *ctx_template_size,
crypto_req_handle_t req)
{
(void) provider;
int error;
skein_ctx_t *ctx_tmpl;
ctx_tmpl = kmem_alloc(sizeof (*ctx_tmpl), crypto_kmflag(req));
if (ctx_tmpl == NULL)
return (CRYPTO_HOST_MEMORY);
error = skein_mac_ctx_build(ctx_tmpl, mechanism, key);
if (error != CRYPTO_SUCCESS)
goto errout;
*ctx_template = ctx_tmpl;
*ctx_template_size = sizeof (*ctx_tmpl);
return (CRYPTO_SUCCESS);
errout:
bzero(ctx_tmpl, sizeof (*ctx_tmpl));
kmem_free(ctx_tmpl, sizeof (*ctx_tmpl));
return (error);
}
/*
* Frees a skein context in a parent crypto context.
*/
static int
skein_free_context(crypto_ctx_t *ctx)
{
if (SKEIN_CTX(ctx) != NULL) {
bzero(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
kmem_free(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
SKEIN_CTX_LVALUE(ctx) = NULL;
}
return (CRYPTO_SUCCESS);
}
diff --git a/module/icp/spi/kcf_spi.c b/module/icp/spi/kcf_spi.c
index 34b36b81c0ab..25fe9b5b66be 100644
--- a/module/icp/spi/kcf_spi.c
+++ b/module/icp/spi/kcf_spi.c
@@ -1,925 +1,925 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* This file is part of the core Kernel Cryptographic Framework.
* It implements the SPI functions exported to cryptographic
* providers.
*/
#include <sys/zfs_context.h>
#include <sys/crypto/common.h>
#include <sys/crypto/impl.h>
#include <sys/crypto/sched_impl.h>
#include <sys/crypto/spi.h>
/*
* minalloc and maxalloc values to be used for taskq_create().
*/
-int crypto_taskq_threads = CRYPTO_TASKQ_THREADS;
-int crypto_taskq_minalloc = CRYPTO_TASKQ_MIN;
-int crypto_taskq_maxalloc = CRYPTO_TASKQ_MAX;
+const int crypto_taskq_threads = CRYPTO_TASKQ_THREADS;
+const int crypto_taskq_minalloc = CRYPTO_TASKQ_MIN;
+const int crypto_taskq_maxalloc = CRYPTO_TASKQ_MAX;
static void remove_provider(kcf_provider_desc_t *);
-static void process_logical_providers(crypto_provider_info_t *,
+static void process_logical_providers(const crypto_provider_info_t *,
+ kcf_provider_desc_t *);
+static int init_prov_mechs(const crypto_provider_info_t *,
kcf_provider_desc_t *);
-static int init_prov_mechs(crypto_provider_info_t *, kcf_provider_desc_t *);
static int kcf_prov_kstat_update(kstat_t *, int);
static void delete_kstat(kcf_provider_desc_t *);
-static kcf_prov_stats_t kcf_stats_ks_data_template = {
+static const kcf_prov_stats_t kcf_stats_ks_data_template = {
{ "kcf_ops_total", KSTAT_DATA_UINT64 },
{ "kcf_ops_passed", KSTAT_DATA_UINT64 },
{ "kcf_ops_failed", KSTAT_DATA_UINT64 },
{ "kcf_ops_returned_busy", KSTAT_DATA_UINT64 }
};
#define KCF_SPI_COPY_OPS(src, dst, ops) if ((src)->ops != NULL) \
- *((dst)->ops) = *((src)->ops);
+ memcpy((void *) (dst)->ops, (src)->ops, sizeof (*(src)->ops));
/*
* Copy an ops vector from src to dst. Used during provider registration
* to copy the ops vector from the provider info structure to the
* provider descriptor maintained by KCF.
* Copying the ops vector specified by the provider is needed since the
* framework does not require the provider info structure to be
* persistent.
*/
static void
-copy_ops_vector_v1(crypto_ops_t *src_ops, crypto_ops_t *dst_ops)
+copy_ops_vector_v1(const crypto_ops_t *src_ops, crypto_ops_t *dst_ops)
{
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_control_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_digest_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_cipher_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_mac_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_sign_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_verify_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_dual_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_dual_cipher_mac_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_random_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_session_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_object_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_key_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_provider_ops);
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_ctx_ops);
}
static void
-copy_ops_vector_v2(crypto_ops_t *src_ops, crypto_ops_t *dst_ops)
+copy_ops_vector_v2(const crypto_ops_t *src_ops, crypto_ops_t *dst_ops)
{
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_mech_ops);
}
static void
-copy_ops_vector_v3(crypto_ops_t *src_ops, crypto_ops_t *dst_ops)
+copy_ops_vector_v3(const crypto_ops_t *src_ops, crypto_ops_t *dst_ops)
{
KCF_SPI_COPY_OPS(src_ops, dst_ops, co_nostore_key_ops);
}
/*
* This routine is used to add cryptographic providers to the KEF framework.
* Providers pass a crypto_provider_info structure to crypto_register_provider()
* and get back a handle. The crypto_provider_info structure contains a
* list of mechanisms supported by the provider and an ops vector containing
* provider entry points. Hardware providers call this routine in their attach
* routines. Software providers call this routine in their _init() routine.
*/
int
-crypto_register_provider(crypto_provider_info_t *info,
+crypto_register_provider(const crypto_provider_info_t *info,
crypto_kcf_provider_handle_t *handle)
{
char *ks_name;
kcf_provider_desc_t *prov_desc = NULL;
int ret = CRYPTO_ARGUMENTS_BAD;
if (info->pi_interface_version > CRYPTO_SPI_VERSION_3)
return (CRYPTO_VERSION_MISMATCH);
/*
* Check provider type, must be software, hardware, or logical.
*/
if (info->pi_provider_type != CRYPTO_HW_PROVIDER &&
info->pi_provider_type != CRYPTO_SW_PROVIDER &&
info->pi_provider_type != CRYPTO_LOGICAL_PROVIDER)
return (CRYPTO_ARGUMENTS_BAD);
/*
* Allocate and initialize a new provider descriptor. We also
* hold it and release it when done.
*/
prov_desc = kcf_alloc_provider_desc(info);
KCF_PROV_REFHOLD(prov_desc);
prov_desc->pd_prov_type = info->pi_provider_type;
/* provider-private handle, opaque to KCF */
prov_desc->pd_prov_handle = info->pi_provider_handle;
/* copy provider description string */
if (info->pi_provider_description != NULL) {
/*
* pi_provider_descriptor is a string that can contain
* up to CRYPTO_PROVIDER_DESCR_MAX_LEN + 1 characters
* INCLUDING the terminating null character. A bcopy()
* is necessary here as pd_description should not have
* a null character. See comments in kcf_alloc_provider_desc()
* for details on pd_description field.
*/
bcopy(info->pi_provider_description, prov_desc->pd_description,
MIN(strlen(info->pi_provider_description),
(size_t)CRYPTO_PROVIDER_DESCR_MAX_LEN));
}
if (info->pi_provider_type != CRYPTO_LOGICAL_PROVIDER) {
if (info->pi_ops_vector == NULL) {
goto bail;
}
- copy_ops_vector_v1(info->pi_ops_vector,
- prov_desc->pd_ops_vector);
+ crypto_ops_t *pvec = (crypto_ops_t *)prov_desc->pd_ops_vector;
+ copy_ops_vector_v1(info->pi_ops_vector, pvec);
if (info->pi_interface_version >= CRYPTO_SPI_VERSION_2) {
- copy_ops_vector_v2(info->pi_ops_vector,
- prov_desc->pd_ops_vector);
+ copy_ops_vector_v2(info->pi_ops_vector, pvec);
prov_desc->pd_flags = info->pi_flags;
}
if (info->pi_interface_version == CRYPTO_SPI_VERSION_3) {
- copy_ops_vector_v3(info->pi_ops_vector,
- prov_desc->pd_ops_vector);
+ copy_ops_vector_v3(info->pi_ops_vector, pvec);
}
}
/* object_ops and nostore_key_ops are mutually exclusive */
if (prov_desc->pd_ops_vector->co_object_ops &&
prov_desc->pd_ops_vector->co_nostore_key_ops) {
goto bail;
}
/* process the mechanisms supported by the provider */
if ((ret = init_prov_mechs(info, prov_desc)) != CRYPTO_SUCCESS)
goto bail;
/*
* Add provider to providers tables, also sets the descriptor
* pd_prov_id field.
*/
if ((ret = kcf_prov_tab_add_provider(prov_desc)) != CRYPTO_SUCCESS) {
undo_register_provider(prov_desc, B_FALSE);
goto bail;
}
/*
* We create a taskq only for a hardware provider. The global
* software queue is used for software providers. We handle ordering
* of multi-part requests in the taskq routine. So, it is safe to
* have multiple threads for the taskq. We pass TASKQ_PREPOPULATE flag
* to keep some entries cached to improve performance.
*/
if (prov_desc->pd_prov_type == CRYPTO_HW_PROVIDER)
prov_desc->pd_sched_info.ks_taskq = taskq_create("kcf_taskq",
- crypto_taskq_threads, minclsyspri,
- crypto_taskq_minalloc, crypto_taskq_maxalloc,
+ CRYPTO_TASKQ_THREADS, minclsyspri,
+ CRYPTO_TASKQ_MIN, CRYPTO_TASKQ_MAX,
TASKQ_PREPOPULATE);
else
prov_desc->pd_sched_info.ks_taskq = NULL;
/* no kernel session to logical providers */
if (prov_desc->pd_prov_type != CRYPTO_LOGICAL_PROVIDER) {
/*
* Open a session for session-oriented providers. This session
* is used for all kernel consumers. This is fine as a provider
* is required to support multiple thread access to a session.
* We can do this only after the taskq has been created as we
* do a kcf_submit_request() to open the session.
*/
if (KCF_PROV_SESSION_OPS(prov_desc) != NULL) {
kcf_req_params_t params;
KCF_WRAP_SESSION_OPS_PARAMS(&params,
KCF_OP_SESSION_OPEN, &prov_desc->pd_sid, 0,
CRYPTO_USER, NULL, 0, prov_desc);
ret = kcf_submit_request(prov_desc, NULL, NULL, &params,
B_FALSE);
if (ret != CRYPTO_SUCCESS) {
undo_register_provider(prov_desc, B_TRUE);
ret = CRYPTO_FAILED;
goto bail;
}
}
}
if (prov_desc->pd_prov_type != CRYPTO_LOGICAL_PROVIDER) {
/*
* Create the kstat for this provider. There is a kstat
* installed for each successfully registered provider.
* This kstat is deleted, when the provider unregisters.
*/
if (prov_desc->pd_prov_type == CRYPTO_SW_PROVIDER) {
ks_name = kmem_asprintf("%s_%s",
"NONAME", "provider_stats");
} else {
ks_name = kmem_asprintf("%s_%d_%u_%s",
"NONAME", 0, prov_desc->pd_prov_id,
"provider_stats");
}
prov_desc->pd_kstat = kstat_create("kcf", 0, ks_name, "crypto",
KSTAT_TYPE_NAMED, sizeof (kcf_prov_stats_t) /
sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
if (prov_desc->pd_kstat != NULL) {
bcopy(&kcf_stats_ks_data_template,
&prov_desc->pd_ks_data,
sizeof (kcf_stats_ks_data_template));
prov_desc->pd_kstat->ks_data = &prov_desc->pd_ks_data;
KCF_PROV_REFHOLD(prov_desc);
KCF_PROV_IREFHOLD(prov_desc);
prov_desc->pd_kstat->ks_private = prov_desc;
prov_desc->pd_kstat->ks_update = kcf_prov_kstat_update;
kstat_install(prov_desc->pd_kstat);
}
kmem_strfree(ks_name);
}
if (prov_desc->pd_prov_type == CRYPTO_HW_PROVIDER)
process_logical_providers(info, prov_desc);
mutex_enter(&prov_desc->pd_lock);
prov_desc->pd_state = KCF_PROV_READY;
mutex_exit(&prov_desc->pd_lock);
kcf_do_notify(prov_desc, B_TRUE);
*handle = prov_desc->pd_kcf_prov_handle;
ret = CRYPTO_SUCCESS;
bail:
KCF_PROV_REFRELE(prov_desc);
return (ret);
}
/*
* This routine is used to notify the framework when a provider is being
* removed. Hardware providers call this routine in their detach routines.
* Software providers call this routine in their _fini() routine.
*/
int
crypto_unregister_provider(crypto_kcf_provider_handle_t handle)
{
uint_t mech_idx;
kcf_provider_desc_t *desc;
kcf_prov_state_t saved_state;
/* lookup provider descriptor */
if ((desc = kcf_prov_tab_lookup((crypto_provider_id_t)handle)) == NULL)
return (CRYPTO_UNKNOWN_PROVIDER);
mutex_enter(&desc->pd_lock);
/*
* Check if any other thread is disabling or removing
* this provider. We return if this is the case.
*/
if (desc->pd_state >= KCF_PROV_DISABLED) {
mutex_exit(&desc->pd_lock);
/* Release reference held by kcf_prov_tab_lookup(). */
KCF_PROV_REFRELE(desc);
return (CRYPTO_BUSY);
}
saved_state = desc->pd_state;
desc->pd_state = KCF_PROV_REMOVED;
if (saved_state == KCF_PROV_BUSY) {
/*
* The per-provider taskq threads may be waiting. We
* signal them so that they can start failing requests.
*/
cv_broadcast(&desc->pd_resume_cv);
}
if (desc->pd_prov_type == CRYPTO_SW_PROVIDER) {
/*
* Check if this provider is currently being used.
* pd_irefcnt is the number of holds from the internal
* structures. We add one to account for the above lookup.
*/
if (desc->pd_refcnt > desc->pd_irefcnt + 1) {
desc->pd_state = saved_state;
mutex_exit(&desc->pd_lock);
/* Release reference held by kcf_prov_tab_lookup(). */
KCF_PROV_REFRELE(desc);
/*
* The administrator presumably will stop the clients
* thus removing the holds, when they get the busy
* return value. Any retry will succeed then.
*/
return (CRYPTO_BUSY);
}
}
mutex_exit(&desc->pd_lock);
if (desc->pd_prov_type != CRYPTO_SW_PROVIDER) {
remove_provider(desc);
}
if (desc->pd_prov_type != CRYPTO_LOGICAL_PROVIDER) {
/* remove the provider from the mechanisms tables */
for (mech_idx = 0; mech_idx < desc->pd_mech_list_count;
mech_idx++) {
kcf_remove_mech_provider(
desc->pd_mechanisms[mech_idx].cm_mech_name, desc);
}
}
/* remove provider from providers table */
if (kcf_prov_tab_rem_provider((crypto_provider_id_t)handle) !=
CRYPTO_SUCCESS) {
/* Release reference held by kcf_prov_tab_lookup(). */
KCF_PROV_REFRELE(desc);
return (CRYPTO_UNKNOWN_PROVIDER);
}
delete_kstat(desc);
if (desc->pd_prov_type == CRYPTO_SW_PROVIDER) {
/* Release reference held by kcf_prov_tab_lookup(). */
KCF_PROV_REFRELE(desc);
/*
* Wait till the existing requests complete.
*/
mutex_enter(&desc->pd_lock);
while (desc->pd_state != KCF_PROV_FREED)
cv_wait(&desc->pd_remove_cv, &desc->pd_lock);
mutex_exit(&desc->pd_lock);
} else {
/*
* Wait until requests that have been sent to the provider
* complete.
*/
mutex_enter(&desc->pd_lock);
while (desc->pd_irefcnt > 0)
cv_wait(&desc->pd_remove_cv, &desc->pd_lock);
mutex_exit(&desc->pd_lock);
}
kcf_do_notify(desc, B_FALSE);
if (desc->pd_prov_type == CRYPTO_SW_PROVIDER) {
/*
* This is the only place where kcf_free_provider_desc()
* is called directly. KCF_PROV_REFRELE() should free the
* structure in all other places.
*/
ASSERT(desc->pd_state == KCF_PROV_FREED &&
desc->pd_refcnt == 0);
kcf_free_provider_desc(desc);
} else {
KCF_PROV_REFRELE(desc);
}
return (CRYPTO_SUCCESS);
}
/*
* This routine is used to notify the framework that the state of
* a cryptographic provider has changed. Valid state codes are:
*
* CRYPTO_PROVIDER_READY
* The provider indicates that it can process more requests. A provider
* will notify with this event if it previously has notified us with a
* CRYPTO_PROVIDER_BUSY.
*
* CRYPTO_PROVIDER_BUSY
* The provider can not take more requests.
*
* CRYPTO_PROVIDER_FAILED
* The provider encountered an internal error. The framework will not
* be sending any more requests to the provider. The provider may notify
* with a CRYPTO_PROVIDER_READY, if it is able to recover from the error.
*
* This routine can be called from user or interrupt context.
*/
void
crypto_provider_notification(crypto_kcf_provider_handle_t handle, uint_t state)
{
kcf_provider_desc_t *pd;
/* lookup the provider from the given handle */
if ((pd = kcf_prov_tab_lookup((crypto_provider_id_t)handle)) == NULL)
return;
mutex_enter(&pd->pd_lock);
if (pd->pd_state <= KCF_PROV_VERIFICATION_FAILED)
goto out;
if (pd->pd_prov_type == CRYPTO_LOGICAL_PROVIDER) {
cmn_err(CE_WARN, "crypto_provider_notification: "
"logical provider (%x) ignored\n", handle);
goto out;
}
switch (state) {
case CRYPTO_PROVIDER_READY:
switch (pd->pd_state) {
case KCF_PROV_BUSY:
pd->pd_state = KCF_PROV_READY;
/*
* Signal the per-provider taskq threads that they
* can start submitting requests.
*/
cv_broadcast(&pd->pd_resume_cv);
break;
case KCF_PROV_FAILED:
/*
* The provider recovered from the error. Let us
* use it now.
*/
pd->pd_state = KCF_PROV_READY;
break;
default:
break;
}
break;
case CRYPTO_PROVIDER_BUSY:
switch (pd->pd_state) {
case KCF_PROV_READY:
pd->pd_state = KCF_PROV_BUSY;
break;
default:
break;
}
break;
case CRYPTO_PROVIDER_FAILED:
/*
* We note the failure and return. The per-provider taskq
* threads check this flag and start failing the
* requests, if it is set. See process_req_hwp() for details.
*/
switch (pd->pd_state) {
case KCF_PROV_READY:
pd->pd_state = KCF_PROV_FAILED;
break;
case KCF_PROV_BUSY:
pd->pd_state = KCF_PROV_FAILED;
/*
* The per-provider taskq threads may be waiting. We
* signal them so that they can start failing requests.
*/
cv_broadcast(&pd->pd_resume_cv);
break;
default:
break;
}
break;
default:
break;
}
out:
mutex_exit(&pd->pd_lock);
KCF_PROV_REFRELE(pd);
}
/*
* This routine is used to notify the framework the result of
* an asynchronous request handled by a provider. Valid error
* codes are the same as the CRYPTO_* errors defined in common.h.
*
* This routine can be called from user or interrupt context.
*/
void
crypto_op_notification(crypto_req_handle_t handle, int error)
{
kcf_call_type_t ctype;
if (handle == NULL)
return;
if ((ctype = GET_REQ_TYPE(handle)) == CRYPTO_SYNCH) {
kcf_sreq_node_t *sreq = (kcf_sreq_node_t *)handle;
if (error != CRYPTO_SUCCESS)
sreq->sn_provider->pd_sched_info.ks_nfails++;
KCF_PROV_IREFRELE(sreq->sn_provider);
kcf_sop_done(sreq, error);
} else {
kcf_areq_node_t *areq = (kcf_areq_node_t *)handle;
ASSERT(ctype == CRYPTO_ASYNCH);
if (error != CRYPTO_SUCCESS)
areq->an_provider->pd_sched_info.ks_nfails++;
KCF_PROV_IREFRELE(areq->an_provider);
kcf_aop_done(areq, error);
}
}
/*
* This routine is used by software providers to determine
* whether to use KM_SLEEP or KM_NOSLEEP during memory allocation.
* Note that hardware providers can always use KM_SLEEP. So,
* they do not need to call this routine.
*
* This routine can be called from user or interrupt context.
*/
int
crypto_kmflag(crypto_req_handle_t handle)
{
return (REQHNDL2_KMFLAG(handle));
}
/*
* Process the mechanism info structures specified by the provider
* during registration. A NULL crypto_provider_info_t indicates
* an already initialized provider descriptor.
*
* Mechanisms are not added to the kernel's mechanism table if the
* provider is a logical provider.
*
* Returns CRYPTO_SUCCESS on success, CRYPTO_ARGUMENTS if one
* of the specified mechanisms was malformed, or CRYPTO_HOST_MEMORY
* if the table of mechanisms is full.
*/
static int
-init_prov_mechs(crypto_provider_info_t *info, kcf_provider_desc_t *desc)
+init_prov_mechs(const crypto_provider_info_t *info, kcf_provider_desc_t *desc)
{
uint_t mech_idx;
uint_t cleanup_idx;
int err = CRYPTO_SUCCESS;
kcf_prov_mech_desc_t *pmd;
int desc_use_count = 0;
int mcount = desc->pd_mech_list_count;
if (desc->pd_prov_type == CRYPTO_LOGICAL_PROVIDER) {
if (info != NULL) {
ASSERT(info->pi_mechanisms != NULL);
bcopy(info->pi_mechanisms, desc->pd_mechanisms,
sizeof (crypto_mech_info_t) * mcount);
}
return (CRYPTO_SUCCESS);
}
/*
* Copy the mechanism list from the provider info to the provider
* descriptor. desc->pd_mechanisms has an extra crypto_mech_info_t
* element if the provider has random_ops since we keep an internal
* mechanism, SUN_RANDOM, in this case.
*/
if (info != NULL) {
if (info->pi_ops_vector->co_random_ops != NULL) {
crypto_mech_info_t *rand_mi;
/*
* Need the following check as it is possible to have
* a provider that implements just random_ops and has
* pi_mechanisms == NULL.
*/
if (info->pi_mechanisms != NULL) {
bcopy(info->pi_mechanisms, desc->pd_mechanisms,
sizeof (crypto_mech_info_t) * (mcount - 1));
}
rand_mi = &desc->pd_mechanisms[mcount - 1];
bzero(rand_mi, sizeof (crypto_mech_info_t));
(void) strncpy(rand_mi->cm_mech_name, SUN_RANDOM,
CRYPTO_MAX_MECH_NAME);
rand_mi->cm_func_group_mask = CRYPTO_FG_RANDOM;
} else {
ASSERT(info->pi_mechanisms != NULL);
bcopy(info->pi_mechanisms, desc->pd_mechanisms,
sizeof (crypto_mech_info_t) * mcount);
}
}
/*
* For each mechanism support by the provider, add the provider
* to the corresponding KCF mechanism mech_entry chain.
*/
for (mech_idx = 0; mech_idx < desc->pd_mech_list_count; mech_idx++) {
crypto_mech_info_t *mi = &desc->pd_mechanisms[mech_idx];
if ((mi->cm_mech_flags & CRYPTO_KEYSIZE_UNIT_IN_BITS) &&
(mi->cm_mech_flags & CRYPTO_KEYSIZE_UNIT_IN_BYTES)) {
err = CRYPTO_ARGUMENTS_BAD;
break;
}
if (desc->pd_flags & CRYPTO_HASH_NO_UPDATE &&
mi->cm_func_group_mask & CRYPTO_FG_DIGEST) {
/*
* We ask the provider to specify the limit
* per hash mechanism. But, in practice, a
* hardware limitation means all hash mechanisms
* will have the same maximum size allowed for
* input data. So, we make it a per provider
* limit to keep it simple.
*/
if (mi->cm_max_input_length == 0) {
err = CRYPTO_ARGUMENTS_BAD;
break;
} else {
desc->pd_hash_limit = mi->cm_max_input_length;
}
}
if ((err = kcf_add_mech_provider(mech_idx, desc, &pmd)) !=
KCF_SUCCESS)
break;
if (pmd == NULL)
continue;
/* The provider will be used for this mechanism */
desc_use_count++;
}
/*
* Don't allow multiple software providers with disabled mechanisms
* to register. Subsequent enabling of mechanisms will result in
* an unsupported configuration, i.e. multiple software providers
* per mechanism.
*/
if (desc_use_count == 0 && desc->pd_prov_type == CRYPTO_SW_PROVIDER)
return (CRYPTO_ARGUMENTS_BAD);
if (err == KCF_SUCCESS)
return (CRYPTO_SUCCESS);
/*
* An error occurred while adding the mechanism, cleanup
* and bail.
*/
for (cleanup_idx = 0; cleanup_idx < mech_idx; cleanup_idx++) {
kcf_remove_mech_provider(
desc->pd_mechanisms[cleanup_idx].cm_mech_name, desc);
}
if (err == KCF_MECH_TAB_FULL)
return (CRYPTO_HOST_MEMORY);
return (CRYPTO_ARGUMENTS_BAD);
}
/*
* Update routine for kstat. Only privileged users are allowed to
* access this information, since this information is sensitive.
* There are some cryptographic attacks (e.g. traffic analysis)
* which can use this information.
*/
static int
kcf_prov_kstat_update(kstat_t *ksp, int rw)
{
kcf_prov_stats_t *ks_data;
kcf_provider_desc_t *pd = (kcf_provider_desc_t *)ksp->ks_private;
if (rw == KSTAT_WRITE)
return (EACCES);
ks_data = ksp->ks_data;
ks_data->ps_ops_total.value.ui64 = pd->pd_sched_info.ks_ndispatches;
ks_data->ps_ops_failed.value.ui64 = pd->pd_sched_info.ks_nfails;
ks_data->ps_ops_busy_rval.value.ui64 = pd->pd_sched_info.ks_nbusy_rval;
ks_data->ps_ops_passed.value.ui64 =
pd->pd_sched_info.ks_ndispatches -
pd->pd_sched_info.ks_nfails -
pd->pd_sched_info.ks_nbusy_rval;
return (0);
}
/*
* Utility routine called from failure paths in crypto_register_provider()
* and from crypto_load_soft_disabled().
*/
void
undo_register_provider(kcf_provider_desc_t *desc, boolean_t remove_prov)
{
uint_t mech_idx;
/* remove the provider from the mechanisms tables */
for (mech_idx = 0; mech_idx < desc->pd_mech_list_count;
mech_idx++) {
kcf_remove_mech_provider(
desc->pd_mechanisms[mech_idx].cm_mech_name, desc);
}
/* remove provider from providers table */
if (remove_prov)
(void) kcf_prov_tab_rem_provider(desc->pd_prov_id);
}
/*
* Utility routine called from crypto_load_soft_disabled(). Callers
* should have done a prior undo_register_provider().
*/
void
redo_register_provider(kcf_provider_desc_t *pd)
{
/* process the mechanisms supported by the provider */
(void) init_prov_mechs(NULL, pd);
/*
* Hold provider in providers table. We should not call
* kcf_prov_tab_add_provider() here as the provider descriptor
* is still valid which means it has an entry in the provider
* table.
*/
KCF_PROV_REFHOLD(pd);
KCF_PROV_IREFHOLD(pd);
}
/*
* Add provider (p1) to another provider's array of providers (p2).
* Hardware and logical providers use this array to cross-reference
* each other.
*/
static void
add_provider_to_array(kcf_provider_desc_t *p1, kcf_provider_desc_t *p2)
{
kcf_provider_list_t *new;
new = kmem_alloc(sizeof (kcf_provider_list_t), KM_SLEEP);
mutex_enter(&p2->pd_lock);
new->pl_next = p2->pd_provider_list;
p2->pd_provider_list = new;
KCF_PROV_IREFHOLD(p1);
new->pl_provider = p1;
mutex_exit(&p2->pd_lock);
}
/*
* Remove provider (p1) from another provider's array of providers (p2).
* Hardware and logical providers use this array to cross-reference
* each other.
*/
static void
remove_provider_from_array(kcf_provider_desc_t *p1, kcf_provider_desc_t *p2)
{
kcf_provider_list_t *pl = NULL, **prev;
mutex_enter(&p2->pd_lock);
for (pl = p2->pd_provider_list, prev = &p2->pd_provider_list;
pl != NULL; prev = &pl->pl_next, pl = pl->pl_next) {
if (pl->pl_provider == p1) {
break;
}
}
if (p1 == NULL) {
mutex_exit(&p2->pd_lock);
return;
}
/* detach and free kcf_provider_list structure */
KCF_PROV_IREFRELE(p1);
*prev = pl->pl_next;
kmem_free(pl, sizeof (*pl));
mutex_exit(&p2->pd_lock);
}
/*
* Convert an array of logical provider handles (crypto_provider_id)
* stored in a crypto_provider_info structure into an array of provider
* descriptors (kcf_provider_desc_t) attached to a logical provider.
*/
static void
-process_logical_providers(crypto_provider_info_t *info, kcf_provider_desc_t *hp)
+process_logical_providers(const crypto_provider_info_t *info,
+ kcf_provider_desc_t *hp)
{
kcf_provider_desc_t *lp;
crypto_provider_id_t handle;
int count = info->pi_logical_provider_count;
int i;
/* add hardware provider to each logical provider */
for (i = 0; i < count; i++) {
handle = info->pi_logical_providers[i];
lp = kcf_prov_tab_lookup((crypto_provider_id_t)handle);
if (lp == NULL) {
continue;
}
add_provider_to_array(hp, lp);
hp->pd_flags |= KCF_LPROV_MEMBER;
/*
* A hardware provider has to have the provider descriptor of
* every logical provider it belongs to, so it can be removed
* from the logical provider if the hardware provider
* unregisters from the framework.
*/
add_provider_to_array(lp, hp);
KCF_PROV_REFRELE(lp);
}
}
/*
* This routine removes a provider from all of the logical or
* hardware providers it belongs to, and frees the provider's
* array of pointers to providers.
*/
static void
remove_provider(kcf_provider_desc_t *pp)
{
kcf_provider_desc_t *p;
kcf_provider_list_t *e, *next;
mutex_enter(&pp->pd_lock);
for (e = pp->pd_provider_list; e != NULL; e = next) {
p = e->pl_provider;
remove_provider_from_array(pp, p);
if (p->pd_prov_type == CRYPTO_HW_PROVIDER &&
p->pd_provider_list == NULL)
p->pd_flags &= ~KCF_LPROV_MEMBER;
KCF_PROV_IREFRELE(p);
next = e->pl_next;
kmem_free(e, sizeof (*e));
}
pp->pd_provider_list = NULL;
mutex_exit(&pp->pd_lock);
}
/*
* Dispatch events as needed for a provider. is_added flag tells
* whether the provider is registering or unregistering.
*/
void
kcf_do_notify(kcf_provider_desc_t *prov_desc, boolean_t is_added)
{
int i;
crypto_notify_event_change_t ec;
ASSERT(prov_desc->pd_state > KCF_PROV_VERIFICATION_FAILED);
/*
* Inform interested clients of the mechanisms becoming
* available/unavailable. We skip this for logical providers
* as they do not affect mechanisms.
*/
if (prov_desc->pd_prov_type != CRYPTO_LOGICAL_PROVIDER) {
ec.ec_provider_type = prov_desc->pd_prov_type;
ec.ec_change = is_added ? CRYPTO_MECH_ADDED :
CRYPTO_MECH_REMOVED;
for (i = 0; i < prov_desc->pd_mech_list_count; i++) {
(void) strlcpy(ec.ec_mech_name,
prov_desc->pd_mechanisms[i].cm_mech_name,
CRYPTO_MAX_MECH_NAME);
kcf_walk_ntfylist(CRYPTO_EVENT_MECHS_CHANGED, &ec);
}
}
/*
* Inform interested clients about the new or departing provider.
* In case of a logical provider, we need to notify the event only
* for the logical provider and not for the underlying
* providers which are known by the KCF_LPROV_MEMBER bit.
*/
if (prov_desc->pd_prov_type == CRYPTO_LOGICAL_PROVIDER ||
(prov_desc->pd_flags & KCF_LPROV_MEMBER) == 0) {
kcf_walk_ntfylist(is_added ? CRYPTO_EVENT_PROVIDER_REGISTERED :
CRYPTO_EVENT_PROVIDER_UNREGISTERED, prov_desc);
}
}
static void
delete_kstat(kcf_provider_desc_t *desc)
{
/* destroy the kstat created for this provider */
if (desc->pd_kstat != NULL) {
kcf_provider_desc_t *kspd = desc->pd_kstat->ks_private;
/* release reference held by desc->pd_kstat->ks_private */
ASSERT(desc == kspd);
kstat_delete(kspd->pd_kstat);
desc->pd_kstat = NULL;
KCF_PROV_REFRELE(kspd);
KCF_PROV_IREFRELE(kspd);
}
}
diff --git a/module/nvpair/nvpair.c b/module/nvpair/nvpair.c
index 8230cca208bf..b4463dd7308f 100644
--- a/module/nvpair/nvpair.c
+++ b/module/nvpair/nvpair.c
@@ -1,3814 +1,3814 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2000, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2015, 2017 by Delphix. All rights reserved.
* Copyright 2018 RackTop Systems.
*/
/*
* Links to Illumos.org for more information on Interface Libraries:
* [1] https://illumos.org/man/3lib/libnvpair
* [2] https://illumos.org/man/3nvpair/nvlist_alloc
* [3] https://illumos.org/man/9f/nvlist_alloc
* [4] https://illumos.org/man/9f/nvlist_next_nvpair
* [5] https://illumos.org/man/9f/nvpair_value_byte
*/
#include <sys/debug.h>
#include <sys/isa_defs.h>
#include <sys/nvpair.h>
#include <sys/nvpair_impl.h>
#include <sys/types.h>
#include <sys/param.h>
#include <sys/strings.h>
#include <rpc/xdr.h>
#include <sys/mod.h>
#if defined(_KERNEL)
#include <sys/sunddi.h>
#include <sys/sysmacros.h>
#else
#include <stdarg.h>
#include <stdlib.h>
#include <stddef.h>
#endif
#define skip_whitespace(p) while ((*(p) == ' ') || (*(p) == '\t')) p++
/*
* nvpair.c - Provides kernel & userland interfaces for manipulating
* name-value pairs.
*
* Overview Diagram
*
* +--------------+
* | nvlist_t |
* |--------------|
* | nvl_version |
* | nvl_nvflag |
* | nvl_priv -+-+
* | nvl_flag | |
* | nvl_pad | |
* +--------------+ |
* V
* +--------------+ last i_nvp in list
* | nvpriv_t | +--------------------->
* |--------------| |
* +--+- nvp_list | | +------------+
* | | nvp_last -+--+ + nv_alloc_t |
* | | nvp_curr | |------------|
* | | nvp_nva -+----> | nva_ops |
* | | nvp_stat | | nva_arg |
* | +--------------+ +------------+
* |
* +-------+
* V
* +---------------------+ +-------------------+
* | i_nvp_t | +-->| i_nvp_t | +-->
* |---------------------| | |-------------------| |
* | nvi_next -+--+ | nvi_next -+--+
* | nvi_prev (NULL) | <----+ nvi_prev |
* | . . . . . . . . . . | | . . . . . . . . . |
* | nvp (nvpair_t) | | nvp (nvpair_t) |
* | - nvp_size | | - nvp_size |
* | - nvp_name_sz | | - nvp_name_sz |
* | - nvp_value_elem | | - nvp_value_elem |
* | - nvp_type | | - nvp_type |
* | - data ... | | - data ... |
* +---------------------+ +-------------------+
*
*
*
* +---------------------+ +---------------------+
* | i_nvp_t | +--> +-->| i_nvp_t (last) |
* |---------------------| | | |---------------------|
* | nvi_next -+--+ ... --+ | nvi_next (NULL) |
* <-+- nvi_prev |<-- ... <----+ nvi_prev |
* | . . . . . . . . . | | . . . . . . . . . |
* | nvp (nvpair_t) | | nvp (nvpair_t) |
* | - nvp_size | | - nvp_size |
* | - nvp_name_sz | | - nvp_name_sz |
* | - nvp_value_elem | | - nvp_value_elem |
* | - DATA_TYPE_NVLIST | | - nvp_type |
* | - data (embedded) | | - data ... |
* | nvlist name | +---------------------+
* | +--------------+ |
* | | nvlist_t | |
* | |--------------| |
* | | nvl_version | |
* | | nvl_nvflag | |
* | | nvl_priv --+---+---->
* | | nvl_flag | |
* | | nvl_pad | |
* | +--------------+ |
* +---------------------+
*
*
* N.B. nvpair_t may be aligned on 4 byte boundary, so +4 will
* allow value to be aligned on 8 byte boundary
*
* name_len is the length of the name string including the null terminator
* so it must be >= 1
*/
#define NVP_SIZE_CALC(name_len, data_len) \
(NV_ALIGN((sizeof (nvpair_t)) + name_len) + NV_ALIGN(data_len))
static int i_get_value_size(data_type_t type, const void *data, uint_t nelem);
static int nvlist_add_common(nvlist_t *nvl, const char *name, data_type_t type,
uint_t nelem, const void *data);
#define NV_STAT_EMBEDDED 0x1
#define EMBEDDED_NVL(nvp) ((nvlist_t *)(void *)NVP_VALUE(nvp))
#define EMBEDDED_NVL_ARRAY(nvp) ((nvlist_t **)(void *)NVP_VALUE(nvp))
#define NVP_VALOFF(nvp) (NV_ALIGN(sizeof (nvpair_t) + (nvp)->nvp_name_sz))
#define NVPAIR2I_NVP(nvp) \
((i_nvp_t *)((size_t)(nvp) - offsetof(i_nvp_t, nvi_nvp)))
#ifdef _KERNEL
-int nvpair_max_recursion = 20;
+static const int nvpair_max_recursion = 20;
#else
-int nvpair_max_recursion = 100;
+static const int nvpair_max_recursion = 100;
#endif
-uint64_t nvlist_hashtable_init_size = (1 << 4);
+static const uint64_t nvlist_hashtable_init_size = (1 << 4);
int
nv_alloc_init(nv_alloc_t *nva, const nv_alloc_ops_t *nvo, /* args */ ...)
{
va_list valist;
int err = 0;
nva->nva_ops = nvo;
nva->nva_arg = NULL;
va_start(valist, nvo);
if (nva->nva_ops->nv_ao_init != NULL)
err = nva->nva_ops->nv_ao_init(nva, valist);
va_end(valist);
return (err);
}
void
nv_alloc_reset(nv_alloc_t *nva)
{
if (nva->nva_ops->nv_ao_reset != NULL)
nva->nva_ops->nv_ao_reset(nva);
}
void
nv_alloc_fini(nv_alloc_t *nva)
{
if (nva->nva_ops->nv_ao_fini != NULL)
nva->nva_ops->nv_ao_fini(nva);
}
nv_alloc_t *
nvlist_lookup_nv_alloc(nvlist_t *nvl)
{
nvpriv_t *priv;
if (nvl == NULL ||
(priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv) == NULL)
return (NULL);
return (priv->nvp_nva);
}
static void *
nv_mem_zalloc(nvpriv_t *nvp, size_t size)
{
nv_alloc_t *nva = nvp->nvp_nva;
void *buf;
if ((buf = nva->nva_ops->nv_ao_alloc(nva, size)) != NULL)
bzero(buf, size);
return (buf);
}
static void
nv_mem_free(nvpriv_t *nvp, void *buf, size_t size)
{
nv_alloc_t *nva = nvp->nvp_nva;
nva->nva_ops->nv_ao_free(nva, buf, size);
}
static void
nv_priv_init(nvpriv_t *priv, nv_alloc_t *nva, uint32_t stat)
{
bzero(priv, sizeof (nvpriv_t));
priv->nvp_nva = nva;
priv->nvp_stat = stat;
}
static nvpriv_t *
nv_priv_alloc(nv_alloc_t *nva)
{
nvpriv_t *priv;
/*
* nv_mem_alloc() cannot called here because it needs the priv
* argument.
*/
if ((priv = nva->nva_ops->nv_ao_alloc(nva, sizeof (nvpriv_t))) == NULL)
return (NULL);
nv_priv_init(priv, nva, 0);
return (priv);
}
/*
* Embedded lists need their own nvpriv_t's. We create a new
* nvpriv_t using the parameters and allocator from the parent
* list's nvpriv_t.
*/
static nvpriv_t *
nv_priv_alloc_embedded(nvpriv_t *priv)
{
nvpriv_t *emb_priv;
if ((emb_priv = nv_mem_zalloc(priv, sizeof (nvpriv_t))) == NULL)
return (NULL);
nv_priv_init(emb_priv, priv->nvp_nva, NV_STAT_EMBEDDED);
return (emb_priv);
}
static int
nvt_tab_alloc(nvpriv_t *priv, uint64_t buckets)
{
ASSERT3P(priv->nvp_hashtable, ==, NULL);
ASSERT0(priv->nvp_nbuckets);
ASSERT0(priv->nvp_nentries);
i_nvp_t **tab = nv_mem_zalloc(priv, buckets * sizeof (i_nvp_t *));
if (tab == NULL)
return (ENOMEM);
priv->nvp_hashtable = tab;
priv->nvp_nbuckets = buckets;
return (0);
}
static void
nvt_tab_free(nvpriv_t *priv)
{
i_nvp_t **tab = priv->nvp_hashtable;
if (tab == NULL) {
ASSERT0(priv->nvp_nbuckets);
ASSERT0(priv->nvp_nentries);
return;
}
nv_mem_free(priv, tab, priv->nvp_nbuckets * sizeof (i_nvp_t *));
priv->nvp_hashtable = NULL;
priv->nvp_nbuckets = 0;
priv->nvp_nentries = 0;
}
static uint32_t
nvt_hash(const char *p)
{
uint32_t g, hval = 0;
while (*p) {
hval = (hval << 4) + *p++;
if ((g = (hval & 0xf0000000)) != 0)
hval ^= g >> 24;
hval &= ~g;
}
return (hval);
}
static boolean_t
nvt_nvpair_match(const nvpair_t *nvp1, const nvpair_t *nvp2, uint32_t nvflag)
{
boolean_t match = B_FALSE;
if (nvflag & NV_UNIQUE_NAME_TYPE) {
if (strcmp(NVP_NAME(nvp1), NVP_NAME(nvp2)) == 0 &&
NVP_TYPE(nvp1) == NVP_TYPE(nvp2))
match = B_TRUE;
} else {
ASSERT(nvflag == 0 || nvflag & NV_UNIQUE_NAME);
if (strcmp(NVP_NAME(nvp1), NVP_NAME(nvp2)) == 0)
match = B_TRUE;
}
return (match);
}
static nvpair_t *
nvt_lookup_name_type(const nvlist_t *nvl, const char *name, data_type_t type)
{
const nvpriv_t *priv = (const nvpriv_t *)(uintptr_t)nvl->nvl_priv;
ASSERT(priv != NULL);
i_nvp_t **tab = priv->nvp_hashtable;
if (tab == NULL) {
ASSERT3P(priv->nvp_list, ==, NULL);
ASSERT0(priv->nvp_nbuckets);
ASSERT0(priv->nvp_nentries);
return (NULL);
} else {
ASSERT(priv->nvp_nbuckets != 0);
}
uint64_t hash = nvt_hash(name);
uint64_t index = hash & (priv->nvp_nbuckets - 1);
ASSERT3U(index, <, priv->nvp_nbuckets);
i_nvp_t *entry = tab[index];
for (i_nvp_t *e = entry; e != NULL; e = e->nvi_hashtable_next) {
if (strcmp(NVP_NAME(&e->nvi_nvp), name) == 0 &&
(type == DATA_TYPE_DONTCARE ||
NVP_TYPE(&e->nvi_nvp) == type))
return (&e->nvi_nvp);
}
return (NULL);
}
static nvpair_t *
nvt_lookup_name(const nvlist_t *nvl, const char *name)
{
return (nvt_lookup_name_type(nvl, name, DATA_TYPE_DONTCARE));
}
static int
nvt_resize(nvpriv_t *priv, uint32_t new_size)
{
i_nvp_t **tab = priv->nvp_hashtable;
/*
* Migrate all the entries from the current table
* to a newly-allocated table with the new size by
* re-adjusting the pointers of their entries.
*/
uint32_t size = priv->nvp_nbuckets;
uint32_t new_mask = new_size - 1;
ASSERT(ISP2(new_size));
i_nvp_t **new_tab = nv_mem_zalloc(priv, new_size * sizeof (i_nvp_t *));
if (new_tab == NULL)
return (ENOMEM);
uint32_t nentries = 0;
for (uint32_t i = 0; i < size; i++) {
i_nvp_t *next, *e = tab[i];
while (e != NULL) {
next = e->nvi_hashtable_next;
uint32_t hash = nvt_hash(NVP_NAME(&e->nvi_nvp));
uint32_t index = hash & new_mask;
e->nvi_hashtable_next = new_tab[index];
new_tab[index] = e;
nentries++;
e = next;
}
tab[i] = NULL;
}
ASSERT3U(nentries, ==, priv->nvp_nentries);
nvt_tab_free(priv);
priv->nvp_hashtable = new_tab;
priv->nvp_nbuckets = new_size;
priv->nvp_nentries = nentries;
return (0);
}
static boolean_t
nvt_needs_togrow(nvpriv_t *priv)
{
/*
* Grow only when we have more elements than buckets
* and the # of buckets doesn't overflow.
*/
return (priv->nvp_nentries > priv->nvp_nbuckets &&
(UINT32_MAX >> 1) >= priv->nvp_nbuckets);
}
/*
* Allocate a new table that's twice the size of the old one,
* and migrate all the entries from the old one to the new
* one by re-adjusting their pointers.
*/
static int
nvt_grow(nvpriv_t *priv)
{
uint32_t current_size = priv->nvp_nbuckets;
/* ensure we won't overflow */
ASSERT3U(UINT32_MAX >> 1, >=, current_size);
return (nvt_resize(priv, current_size << 1));
}
static boolean_t
nvt_needs_toshrink(nvpriv_t *priv)
{
/*
* Shrink only when the # of elements is less than or
* equal to 1/4 the # of buckets. Never shrink less than
* nvlist_hashtable_init_size.
*/
ASSERT3U(priv->nvp_nbuckets, >=, nvlist_hashtable_init_size);
if (priv->nvp_nbuckets == nvlist_hashtable_init_size)
return (B_FALSE);
return (priv->nvp_nentries <= (priv->nvp_nbuckets >> 2));
}
/*
* Allocate a new table that's half the size of the old one,
* and migrate all the entries from the old one to the new
* one by re-adjusting their pointers.
*/
static int
nvt_shrink(nvpriv_t *priv)
{
uint32_t current_size = priv->nvp_nbuckets;
/* ensure we won't overflow */
ASSERT3U(current_size, >=, nvlist_hashtable_init_size);
return (nvt_resize(priv, current_size >> 1));
}
static int
nvt_remove_nvpair(nvlist_t *nvl, const nvpair_t *nvp)
{
nvpriv_t *priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv;
if (nvt_needs_toshrink(priv)) {
int err = nvt_shrink(priv);
if (err != 0)
return (err);
}
i_nvp_t **tab = priv->nvp_hashtable;
char *name = NVP_NAME(nvp);
uint64_t hash = nvt_hash(name);
uint64_t index = hash & (priv->nvp_nbuckets - 1);
ASSERT3U(index, <, priv->nvp_nbuckets);
i_nvp_t *bucket = tab[index];
for (i_nvp_t *prev = NULL, *e = bucket;
e != NULL; prev = e, e = e->nvi_hashtable_next) {
if (nvt_nvpair_match(&e->nvi_nvp, nvp, nvl->nvl_nvflag)) {
if (prev != NULL) {
prev->nvi_hashtable_next =
e->nvi_hashtable_next;
} else {
ASSERT3P(e, ==, bucket);
tab[index] = e->nvi_hashtable_next;
}
e->nvi_hashtable_next = NULL;
priv->nvp_nentries--;
break;
}
}
return (0);
}
static int
nvt_add_nvpair(nvlist_t *nvl, nvpair_t *nvp)
{
nvpriv_t *priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv;
/* initialize nvpair table now if it doesn't exist. */
if (priv->nvp_hashtable == NULL) {
int err = nvt_tab_alloc(priv, nvlist_hashtable_init_size);
if (err != 0)
return (err);
}
/*
* if we don't allow duplicate entries, make sure to
* unlink any existing entries from the table.
*/
if (nvl->nvl_nvflag != 0) {
int err = nvt_remove_nvpair(nvl, nvp);
if (err != 0)
return (err);
}
if (nvt_needs_togrow(priv)) {
int err = nvt_grow(priv);
if (err != 0)
return (err);
}
i_nvp_t **tab = priv->nvp_hashtable;
char *name = NVP_NAME(nvp);
uint64_t hash = nvt_hash(name);
uint64_t index = hash & (priv->nvp_nbuckets - 1);
ASSERT3U(index, <, priv->nvp_nbuckets);
// cppcheck-suppress nullPointerRedundantCheck
i_nvp_t *bucket = tab[index];
/* insert link at the beginning of the bucket */
i_nvp_t *new_entry = NVPAIR2I_NVP(nvp);
ASSERT3P(new_entry->nvi_hashtable_next, ==, NULL);
new_entry->nvi_hashtable_next = bucket;
// cppcheck-suppress nullPointerRedundantCheck
tab[index] = new_entry;
priv->nvp_nentries++;
return (0);
}
static void
nvlist_init(nvlist_t *nvl, uint32_t nvflag, nvpriv_t *priv)
{
nvl->nvl_version = NV_VERSION;
nvl->nvl_nvflag = nvflag & (NV_UNIQUE_NAME|NV_UNIQUE_NAME_TYPE);
nvl->nvl_priv = (uint64_t)(uintptr_t)priv;
nvl->nvl_flag = 0;
nvl->nvl_pad = 0;
}
uint_t
nvlist_nvflag(nvlist_t *nvl)
{
return (nvl->nvl_nvflag);
}
static nv_alloc_t *
nvlist_nv_alloc(int kmflag)
{
#if defined(_KERNEL)
switch (kmflag) {
case KM_SLEEP:
return (nv_alloc_sleep);
case KM_NOSLEEP:
return (nv_alloc_nosleep);
default:
return (nv_alloc_pushpage);
}
#else
(void) kmflag;
return (nv_alloc_nosleep);
#endif /* _KERNEL */
}
/*
* nvlist_alloc - Allocate nvlist.
*/
int
nvlist_alloc(nvlist_t **nvlp, uint_t nvflag, int kmflag)
{
return (nvlist_xalloc(nvlp, nvflag, nvlist_nv_alloc(kmflag)));
}
int
nvlist_xalloc(nvlist_t **nvlp, uint_t nvflag, nv_alloc_t *nva)
{
nvpriv_t *priv;
if (nvlp == NULL || nva == NULL)
return (EINVAL);
if ((priv = nv_priv_alloc(nva)) == NULL)
return (ENOMEM);
if ((*nvlp = nv_mem_zalloc(priv,
NV_ALIGN(sizeof (nvlist_t)))) == NULL) {
nv_mem_free(priv, priv, sizeof (nvpriv_t));
return (ENOMEM);
}
nvlist_init(*nvlp, nvflag, priv);
return (0);
}
/*
* nvp_buf_alloc - Allocate i_nvp_t for storing a new nv pair.
*/
static nvpair_t *
nvp_buf_alloc(nvlist_t *nvl, size_t len)
{
nvpriv_t *priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv;
i_nvp_t *buf;
nvpair_t *nvp;
size_t nvsize;
/*
* Allocate the buffer
*/
nvsize = len + offsetof(i_nvp_t, nvi_nvp);
if ((buf = nv_mem_zalloc(priv, nvsize)) == NULL)
return (NULL);
nvp = &buf->nvi_nvp;
nvp->nvp_size = len;
return (nvp);
}
/*
* nvp_buf_free - de-Allocate an i_nvp_t.
*/
static void
nvp_buf_free(nvlist_t *nvl, nvpair_t *nvp)
{
nvpriv_t *priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv;
size_t nvsize = nvp->nvp_size + offsetof(i_nvp_t, nvi_nvp);
nv_mem_free(priv, NVPAIR2I_NVP(nvp), nvsize);
}
/*
* nvp_buf_link - link a new nv pair into the nvlist.
*/
static void
nvp_buf_link(nvlist_t *nvl, nvpair_t *nvp)
{
nvpriv_t *priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv;
i_nvp_t *curr = NVPAIR2I_NVP(nvp);
/* Put element at end of nvlist */
if (priv->nvp_list == NULL) {
priv->nvp_list = priv->nvp_last = curr;
} else {
curr->nvi_prev = priv->nvp_last;
priv->nvp_last->nvi_next = curr;
priv->nvp_last = curr;
}
}
/*
* nvp_buf_unlink - unlink an removed nvpair out of the nvlist.
*/
static void
nvp_buf_unlink(nvlist_t *nvl, nvpair_t *nvp)
{
nvpriv_t *priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv;
i_nvp_t *curr = NVPAIR2I_NVP(nvp);
/*
* protect nvlist_next_nvpair() against walking on freed memory.
*/
if (priv->nvp_curr == curr)
priv->nvp_curr = curr->nvi_next;
if (curr == priv->nvp_list)
priv->nvp_list = curr->nvi_next;
else
curr->nvi_prev->nvi_next = curr->nvi_next;
if (curr == priv->nvp_last)
priv->nvp_last = curr->nvi_prev;
else
curr->nvi_next->nvi_prev = curr->nvi_prev;
}
/*
* take a nvpair type and number of elements and make sure the are valid
*/
static int
i_validate_type_nelem(data_type_t type, uint_t nelem)
{
switch (type) {
case DATA_TYPE_BOOLEAN:
if (nelem != 0)
return (EINVAL);
break;
case DATA_TYPE_BOOLEAN_VALUE:
case DATA_TYPE_BYTE:
case DATA_TYPE_INT8:
case DATA_TYPE_UINT8:
case DATA_TYPE_INT16:
case DATA_TYPE_UINT16:
case DATA_TYPE_INT32:
case DATA_TYPE_UINT32:
case DATA_TYPE_INT64:
case DATA_TYPE_UINT64:
case DATA_TYPE_STRING:
case DATA_TYPE_HRTIME:
case DATA_TYPE_NVLIST:
#if !defined(_KERNEL)
case DATA_TYPE_DOUBLE:
#endif
if (nelem != 1)
return (EINVAL);
break;
case DATA_TYPE_BOOLEAN_ARRAY:
case DATA_TYPE_BYTE_ARRAY:
case DATA_TYPE_INT8_ARRAY:
case DATA_TYPE_UINT8_ARRAY:
case DATA_TYPE_INT16_ARRAY:
case DATA_TYPE_UINT16_ARRAY:
case DATA_TYPE_INT32_ARRAY:
case DATA_TYPE_UINT32_ARRAY:
case DATA_TYPE_INT64_ARRAY:
case DATA_TYPE_UINT64_ARRAY:
case DATA_TYPE_STRING_ARRAY:
case DATA_TYPE_NVLIST_ARRAY:
/* we allow arrays with 0 elements */
break;
default:
return (EINVAL);
}
return (0);
}
/*
* Verify nvp_name_sz and check the name string length.
*/
static int
i_validate_nvpair_name(nvpair_t *nvp)
{
if ((nvp->nvp_name_sz <= 0) ||
(nvp->nvp_size < NVP_SIZE_CALC(nvp->nvp_name_sz, 0)))
return (EFAULT);
/* verify the name string, make sure its terminated */
if (NVP_NAME(nvp)[nvp->nvp_name_sz - 1] != '\0')
return (EFAULT);
return (strlen(NVP_NAME(nvp)) == nvp->nvp_name_sz - 1 ? 0 : EFAULT);
}
static int
i_validate_nvpair_value(data_type_t type, uint_t nelem, const void *data)
{
switch (type) {
case DATA_TYPE_BOOLEAN_VALUE:
if (*(boolean_t *)data != B_TRUE &&
*(boolean_t *)data != B_FALSE)
return (EINVAL);
break;
case DATA_TYPE_BOOLEAN_ARRAY: {
int i;
for (i = 0; i < nelem; i++)
if (((boolean_t *)data)[i] != B_TRUE &&
((boolean_t *)data)[i] != B_FALSE)
return (EINVAL);
break;
}
default:
break;
}
return (0);
}
/*
* This function takes a pointer to what should be a nvpair and it's size
* and then verifies that all the nvpair fields make sense and can be
* trusted. This function is used when decoding packed nvpairs.
*/
static int
i_validate_nvpair(nvpair_t *nvp)
{
data_type_t type = NVP_TYPE(nvp);
int size1, size2;
/* verify nvp_name_sz, check the name string length */
if (i_validate_nvpair_name(nvp) != 0)
return (EFAULT);
if (i_validate_nvpair_value(type, NVP_NELEM(nvp), NVP_VALUE(nvp)) != 0)
return (EFAULT);
/*
* verify nvp_type, nvp_value_elem, and also possibly
* verify string values and get the value size.
*/
size2 = i_get_value_size(type, NVP_VALUE(nvp), NVP_NELEM(nvp));
size1 = nvp->nvp_size - NVP_VALOFF(nvp);
if (size2 < 0 || size1 != NV_ALIGN(size2))
return (EFAULT);
return (0);
}
static int
nvlist_copy_pairs(const nvlist_t *snvl, nvlist_t *dnvl)
{
const nvpriv_t *priv;
const i_nvp_t *curr;
if ((priv = (const nvpriv_t *)(uintptr_t)snvl->nvl_priv) == NULL)
return (EINVAL);
for (curr = priv->nvp_list; curr != NULL; curr = curr->nvi_next) {
const nvpair_t *nvp = &curr->nvi_nvp;
int err;
if ((err = nvlist_add_common(dnvl, NVP_NAME(nvp), NVP_TYPE(nvp),
NVP_NELEM(nvp), NVP_VALUE(nvp))) != 0)
return (err);
}
return (0);
}
/*
* Frees all memory allocated for an nvpair (like embedded lists) with
* the exception of the nvpair buffer itself.
*/
static void
nvpair_free(nvpair_t *nvp)
{
switch (NVP_TYPE(nvp)) {
case DATA_TYPE_NVLIST:
nvlist_free(EMBEDDED_NVL(nvp));
break;
case DATA_TYPE_NVLIST_ARRAY: {
nvlist_t **nvlp = EMBEDDED_NVL_ARRAY(nvp);
int i;
for (i = 0; i < NVP_NELEM(nvp); i++)
if (nvlp[i] != NULL)
nvlist_free(nvlp[i]);
break;
}
default:
break;
}
}
/*
* nvlist_free - free an unpacked nvlist
*/
void
nvlist_free(nvlist_t *nvl)
{
nvpriv_t *priv;
i_nvp_t *curr;
if (nvl == NULL ||
(priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv) == NULL)
return;
/*
* Unpacked nvlist are linked through i_nvp_t
*/
curr = priv->nvp_list;
while (curr != NULL) {
nvpair_t *nvp = &curr->nvi_nvp;
curr = curr->nvi_next;
nvpair_free(nvp);
nvp_buf_free(nvl, nvp);
}
if (!(priv->nvp_stat & NV_STAT_EMBEDDED))
nv_mem_free(priv, nvl, NV_ALIGN(sizeof (nvlist_t)));
else
nvl->nvl_priv = 0;
nvt_tab_free(priv);
nv_mem_free(priv, priv, sizeof (nvpriv_t));
}
static int
nvlist_contains_nvp(const nvlist_t *nvl, const nvpair_t *nvp)
{
const nvpriv_t *priv = (const nvpriv_t *)(uintptr_t)nvl->nvl_priv;
const i_nvp_t *curr;
if (nvp == NULL)
return (0);
for (curr = priv->nvp_list; curr != NULL; curr = curr->nvi_next)
if (&curr->nvi_nvp == nvp)
return (1);
return (0);
}
/*
* Make a copy of nvlist
*/
int
nvlist_dup(const nvlist_t *nvl, nvlist_t **nvlp, int kmflag)
{
return (nvlist_xdup(nvl, nvlp, nvlist_nv_alloc(kmflag)));
}
int
nvlist_xdup(const nvlist_t *nvl, nvlist_t **nvlp, nv_alloc_t *nva)
{
int err;
nvlist_t *ret;
if (nvl == NULL || nvlp == NULL)
return (EINVAL);
if ((err = nvlist_xalloc(&ret, nvl->nvl_nvflag, nva)) != 0)
return (err);
if ((err = nvlist_copy_pairs(nvl, ret)) != 0)
nvlist_free(ret);
else
*nvlp = ret;
return (err);
}
/*
* Remove all with matching name
*/
int
nvlist_remove_all(nvlist_t *nvl, const char *name)
{
int error = ENOENT;
if (nvl == NULL || name == NULL || nvl->nvl_priv == 0)
return (EINVAL);
nvpair_t *nvp;
while ((nvp = nvt_lookup_name(nvl, name)) != NULL) {
VERIFY0(nvlist_remove_nvpair(nvl, nvp));
error = 0;
}
return (error);
}
/*
* Remove first one with matching name and type
*/
int
nvlist_remove(nvlist_t *nvl, const char *name, data_type_t type)
{
if (nvl == NULL || name == NULL || nvl->nvl_priv == 0)
return (EINVAL);
nvpair_t *nvp = nvt_lookup_name_type(nvl, name, type);
if (nvp == NULL)
return (ENOENT);
return (nvlist_remove_nvpair(nvl, nvp));
}
int
nvlist_remove_nvpair(nvlist_t *nvl, nvpair_t *nvp)
{
if (nvl == NULL || nvp == NULL)
return (EINVAL);
int err = nvt_remove_nvpair(nvl, nvp);
if (err != 0)
return (err);
nvp_buf_unlink(nvl, nvp);
nvpair_free(nvp);
nvp_buf_free(nvl, nvp);
return (0);
}
/*
* This function calculates the size of an nvpair value.
*
* The data argument controls the behavior in case of the data types
* DATA_TYPE_STRING and
* DATA_TYPE_STRING_ARRAY
* Is data == NULL then the size of the string(s) is excluded.
*/
static int
i_get_value_size(data_type_t type, const void *data, uint_t nelem)
{
uint64_t value_sz;
if (i_validate_type_nelem(type, nelem) != 0)
return (-1);
/* Calculate required size for holding value */
switch (type) {
case DATA_TYPE_BOOLEAN:
value_sz = 0;
break;
case DATA_TYPE_BOOLEAN_VALUE:
value_sz = sizeof (boolean_t);
break;
case DATA_TYPE_BYTE:
value_sz = sizeof (uchar_t);
break;
case DATA_TYPE_INT8:
value_sz = sizeof (int8_t);
break;
case DATA_TYPE_UINT8:
value_sz = sizeof (uint8_t);
break;
case DATA_TYPE_INT16:
value_sz = sizeof (int16_t);
break;
case DATA_TYPE_UINT16:
value_sz = sizeof (uint16_t);
break;
case DATA_TYPE_INT32:
value_sz = sizeof (int32_t);
break;
case DATA_TYPE_UINT32:
value_sz = sizeof (uint32_t);
break;
case DATA_TYPE_INT64:
value_sz = sizeof (int64_t);
break;
case DATA_TYPE_UINT64:
value_sz = sizeof (uint64_t);
break;
#if !defined(_KERNEL)
case DATA_TYPE_DOUBLE:
value_sz = sizeof (double);
break;
#endif
case DATA_TYPE_STRING:
if (data == NULL)
value_sz = 0;
else
value_sz = strlen(data) + 1;
break;
case DATA_TYPE_BOOLEAN_ARRAY:
value_sz = (uint64_t)nelem * sizeof (boolean_t);
break;
case DATA_TYPE_BYTE_ARRAY:
value_sz = (uint64_t)nelem * sizeof (uchar_t);
break;
case DATA_TYPE_INT8_ARRAY:
value_sz = (uint64_t)nelem * sizeof (int8_t);
break;
case DATA_TYPE_UINT8_ARRAY:
value_sz = (uint64_t)nelem * sizeof (uint8_t);
break;
case DATA_TYPE_INT16_ARRAY:
value_sz = (uint64_t)nelem * sizeof (int16_t);
break;
case DATA_TYPE_UINT16_ARRAY:
value_sz = (uint64_t)nelem * sizeof (uint16_t);
break;
case DATA_TYPE_INT32_ARRAY:
value_sz = (uint64_t)nelem * sizeof (int32_t);
break;
case DATA_TYPE_UINT32_ARRAY:
value_sz = (uint64_t)nelem * sizeof (uint32_t);
break;
case DATA_TYPE_INT64_ARRAY:
value_sz = (uint64_t)nelem * sizeof (int64_t);
break;
case DATA_TYPE_UINT64_ARRAY:
value_sz = (uint64_t)nelem * sizeof (uint64_t);
break;
case DATA_TYPE_STRING_ARRAY:
value_sz = (uint64_t)nelem * sizeof (uint64_t);
if (data != NULL) {
char *const *strs = data;
uint_t i;
/* no alignment requirement for strings */
for (i = 0; i < nelem; i++) {
if (strs[i] == NULL)
return (-1);
value_sz += strlen(strs[i]) + 1;
}
}
break;
case DATA_TYPE_HRTIME:
value_sz = sizeof (hrtime_t);
break;
case DATA_TYPE_NVLIST:
value_sz = NV_ALIGN(sizeof (nvlist_t));
break;
case DATA_TYPE_NVLIST_ARRAY:
value_sz = (uint64_t)nelem * sizeof (uint64_t) +
(uint64_t)nelem * NV_ALIGN(sizeof (nvlist_t));
break;
default:
return (-1);
}
return (value_sz > INT32_MAX ? -1 : (int)value_sz);
}
static int
nvlist_copy_embedded(nvlist_t *nvl, nvlist_t *onvl, nvlist_t *emb_nvl)
{
nvpriv_t *priv;
int err;
if ((priv = nv_priv_alloc_embedded((nvpriv_t *)(uintptr_t)
nvl->nvl_priv)) == NULL)
return (ENOMEM);
nvlist_init(emb_nvl, onvl->nvl_nvflag, priv);
if ((err = nvlist_copy_pairs(onvl, emb_nvl)) != 0) {
nvlist_free(emb_nvl);
emb_nvl->nvl_priv = 0;
}
return (err);
}
/*
* nvlist_add_common - Add new <name,value> pair to nvlist
*/
static int
nvlist_add_common(nvlist_t *nvl, const char *name,
data_type_t type, uint_t nelem, const void *data)
{
nvpair_t *nvp;
uint_t i;
int nvp_sz, name_sz, value_sz;
int err = 0;
if (name == NULL || nvl == NULL || nvl->nvl_priv == 0)
return (EINVAL);
if (nelem != 0 && data == NULL)
return (EINVAL);
/*
* Verify type and nelem and get the value size.
* In case of data types DATA_TYPE_STRING and DATA_TYPE_STRING_ARRAY
* is the size of the string(s) included.
*/
if ((value_sz = i_get_value_size(type, data, nelem)) < 0)
return (EINVAL);
if (i_validate_nvpair_value(type, nelem, data) != 0)
return (EINVAL);
/*
* If we're adding an nvlist or nvlist array, ensure that we are not
* adding the input nvlist to itself, which would cause recursion,
* and ensure that no NULL nvlist pointers are present.
*/
switch (type) {
case DATA_TYPE_NVLIST:
if (data == nvl || data == NULL)
return (EINVAL);
break;
case DATA_TYPE_NVLIST_ARRAY: {
nvlist_t **onvlp = (nvlist_t **)data;
for (i = 0; i < nelem; i++) {
if (onvlp[i] == nvl || onvlp[i] == NULL)
return (EINVAL);
}
break;
}
default:
break;
}
/* calculate sizes of the nvpair elements and the nvpair itself */
name_sz = strlen(name) + 1;
if (name_sz >= 1ULL << (sizeof (nvp->nvp_name_sz) * NBBY - 1))
return (EINVAL);
nvp_sz = NVP_SIZE_CALC(name_sz, value_sz);
if ((nvp = nvp_buf_alloc(nvl, nvp_sz)) == NULL)
return (ENOMEM);
ASSERT(nvp->nvp_size == nvp_sz);
nvp->nvp_name_sz = name_sz;
nvp->nvp_value_elem = nelem;
nvp->nvp_type = type;
bcopy(name, NVP_NAME(nvp), name_sz);
switch (type) {
case DATA_TYPE_BOOLEAN:
break;
case DATA_TYPE_STRING_ARRAY: {
char *const *strs = data;
char *buf = NVP_VALUE(nvp);
char **cstrs = (void *)buf;
/* skip pre-allocated space for pointer array */
buf += nelem * sizeof (uint64_t);
for (i = 0; i < nelem; i++) {
int slen = strlen(strs[i]) + 1;
bcopy(strs[i], buf, slen);
cstrs[i] = buf;
buf += slen;
}
break;
}
case DATA_TYPE_NVLIST: {
nvlist_t *nnvl = EMBEDDED_NVL(nvp);
nvlist_t *onvl = (nvlist_t *)data;
if ((err = nvlist_copy_embedded(nvl, onvl, nnvl)) != 0) {
nvp_buf_free(nvl, nvp);
return (err);
}
break;
}
case DATA_TYPE_NVLIST_ARRAY: {
nvlist_t **onvlp = (nvlist_t **)data;
nvlist_t **nvlp = EMBEDDED_NVL_ARRAY(nvp);
nvlist_t *embedded = (nvlist_t *)
((uintptr_t)nvlp + nelem * sizeof (uint64_t));
for (i = 0; i < nelem; i++) {
if ((err = nvlist_copy_embedded(nvl,
onvlp[i], embedded)) != 0) {
/*
* Free any successfully created lists
*/
nvpair_free(nvp);
nvp_buf_free(nvl, nvp);
return (err);
}
nvlp[i] = embedded++;
}
break;
}
default:
bcopy(data, NVP_VALUE(nvp), value_sz);
}
/* if unique name, remove before add */
if (nvl->nvl_nvflag & NV_UNIQUE_NAME)
(void) nvlist_remove_all(nvl, name);
else if (nvl->nvl_nvflag & NV_UNIQUE_NAME_TYPE)
(void) nvlist_remove(nvl, name, type);
err = nvt_add_nvpair(nvl, nvp);
if (err != 0) {
nvpair_free(nvp);
nvp_buf_free(nvl, nvp);
return (err);
}
nvp_buf_link(nvl, nvp);
return (0);
}
int
nvlist_add_boolean(nvlist_t *nvl, const char *name)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_BOOLEAN, 0, NULL));
}
int
nvlist_add_boolean_value(nvlist_t *nvl, const char *name, boolean_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_BOOLEAN_VALUE, 1, &val));
}
int
nvlist_add_byte(nvlist_t *nvl, const char *name, uchar_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_BYTE, 1, &val));
}
int
nvlist_add_int8(nvlist_t *nvl, const char *name, int8_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_INT8, 1, &val));
}
int
nvlist_add_uint8(nvlist_t *nvl, const char *name, uint8_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_UINT8, 1, &val));
}
int
nvlist_add_int16(nvlist_t *nvl, const char *name, int16_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_INT16, 1, &val));
}
int
nvlist_add_uint16(nvlist_t *nvl, const char *name, uint16_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_UINT16, 1, &val));
}
int
nvlist_add_int32(nvlist_t *nvl, const char *name, int32_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_INT32, 1, &val));
}
int
nvlist_add_uint32(nvlist_t *nvl, const char *name, uint32_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_UINT32, 1, &val));
}
int
nvlist_add_int64(nvlist_t *nvl, const char *name, int64_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_INT64, 1, &val));
}
int
nvlist_add_uint64(nvlist_t *nvl, const char *name, uint64_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_UINT64, 1, &val));
}
#if !defined(_KERNEL)
int
nvlist_add_double(nvlist_t *nvl, const char *name, double val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_DOUBLE, 1, &val));
}
#endif
int
nvlist_add_string(nvlist_t *nvl, const char *name, const char *val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_STRING, 1, (void *)val));
}
int
nvlist_add_boolean_array(nvlist_t *nvl, const char *name,
const boolean_t *a, uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_BOOLEAN_ARRAY, n, a));
}
int
nvlist_add_byte_array(nvlist_t *nvl, const char *name, const uchar_t *a,
uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_BYTE_ARRAY, n, a));
}
int
nvlist_add_int8_array(nvlist_t *nvl, const char *name, const int8_t *a,
uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_INT8_ARRAY, n, a));
}
int
nvlist_add_uint8_array(nvlist_t *nvl, const char *name, const uint8_t *a,
uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_UINT8_ARRAY, n, a));
}
int
nvlist_add_int16_array(nvlist_t *nvl, const char *name, const int16_t *a,
uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_INT16_ARRAY, n, a));
}
int
nvlist_add_uint16_array(nvlist_t *nvl, const char *name, const uint16_t *a,
uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_UINT16_ARRAY, n, a));
}
int
nvlist_add_int32_array(nvlist_t *nvl, const char *name, const int32_t *a,
uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_INT32_ARRAY, n, a));
}
int
nvlist_add_uint32_array(nvlist_t *nvl, const char *name, const uint32_t *a,
uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_UINT32_ARRAY, n, a));
}
int
nvlist_add_int64_array(nvlist_t *nvl, const char *name, const int64_t *a,
uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_INT64_ARRAY, n, a));
}
int
nvlist_add_uint64_array(nvlist_t *nvl, const char *name, const uint64_t *a,
uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_UINT64_ARRAY, n, a));
}
int
nvlist_add_string_array(nvlist_t *nvl, const char *name,
const char *const *a, uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_STRING_ARRAY, n, a));
}
int
nvlist_add_hrtime(nvlist_t *nvl, const char *name, hrtime_t val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_HRTIME, 1, &val));
}
int
nvlist_add_nvlist(nvlist_t *nvl, const char *name, const nvlist_t *val)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_NVLIST, 1, val));
}
int
nvlist_add_nvlist_array(nvlist_t *nvl, const char *name,
const nvlist_t * const *a, uint_t n)
{
return (nvlist_add_common(nvl, name, DATA_TYPE_NVLIST_ARRAY, n, a));
}
/* reading name-value pairs */
nvpair_t *
nvlist_next_nvpair(nvlist_t *nvl, const nvpair_t *nvp)
{
nvpriv_t *priv;
i_nvp_t *curr;
if (nvl == NULL ||
(priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv) == NULL)
return (NULL);
curr = NVPAIR2I_NVP(nvp);
/*
* Ensure that nvp is a valid nvpair on this nvlist.
* NB: nvp_curr is used only as a hint so that we don't always
* have to walk the list to determine if nvp is still on the list.
*/
if (nvp == NULL)
curr = priv->nvp_list;
else if (priv->nvp_curr == curr || nvlist_contains_nvp(nvl, nvp))
curr = curr->nvi_next;
else
curr = NULL;
priv->nvp_curr = curr;
return (curr != NULL ? &curr->nvi_nvp : NULL);
}
nvpair_t *
nvlist_prev_nvpair(nvlist_t *nvl, const nvpair_t *nvp)
{
nvpriv_t *priv;
i_nvp_t *curr;
if (nvl == NULL ||
(priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv) == NULL)
return (NULL);
curr = NVPAIR2I_NVP(nvp);
if (nvp == NULL)
curr = priv->nvp_last;
else if (priv->nvp_curr == curr || nvlist_contains_nvp(nvl, nvp))
curr = curr->nvi_prev;
else
curr = NULL;
priv->nvp_curr = curr;
return (curr != NULL ? &curr->nvi_nvp : NULL);
}
boolean_t
nvlist_empty(const nvlist_t *nvl)
{
const nvpriv_t *priv;
if (nvl == NULL ||
(priv = (const nvpriv_t *)(uintptr_t)nvl->nvl_priv) == NULL)
return (B_TRUE);
return (priv->nvp_list == NULL);
}
char *
nvpair_name(const nvpair_t *nvp)
{
return (NVP_NAME(nvp));
}
data_type_t
nvpair_type(const nvpair_t *nvp)
{
return (NVP_TYPE(nvp));
}
int
nvpair_type_is_array(const nvpair_t *nvp)
{
data_type_t type = NVP_TYPE(nvp);
if ((type == DATA_TYPE_BYTE_ARRAY) ||
(type == DATA_TYPE_INT8_ARRAY) ||
(type == DATA_TYPE_UINT8_ARRAY) ||
(type == DATA_TYPE_INT16_ARRAY) ||
(type == DATA_TYPE_UINT16_ARRAY) ||
(type == DATA_TYPE_INT32_ARRAY) ||
(type == DATA_TYPE_UINT32_ARRAY) ||
(type == DATA_TYPE_INT64_ARRAY) ||
(type == DATA_TYPE_UINT64_ARRAY) ||
(type == DATA_TYPE_BOOLEAN_ARRAY) ||
(type == DATA_TYPE_STRING_ARRAY) ||
(type == DATA_TYPE_NVLIST_ARRAY))
return (1);
return (0);
}
static int
nvpair_value_common(const nvpair_t *nvp, data_type_t type, uint_t *nelem,
void *data)
{
int value_sz;
if (nvp == NULL || nvpair_type(nvp) != type)
return (EINVAL);
/*
* For non-array types, we copy the data.
* For array types (including string), we set a pointer.
*/
switch (type) {
case DATA_TYPE_BOOLEAN:
if (nelem != NULL)
*nelem = 0;
break;
case DATA_TYPE_BOOLEAN_VALUE:
case DATA_TYPE_BYTE:
case DATA_TYPE_INT8:
case DATA_TYPE_UINT8:
case DATA_TYPE_INT16:
case DATA_TYPE_UINT16:
case DATA_TYPE_INT32:
case DATA_TYPE_UINT32:
case DATA_TYPE_INT64:
case DATA_TYPE_UINT64:
case DATA_TYPE_HRTIME:
#if !defined(_KERNEL)
case DATA_TYPE_DOUBLE:
#endif
if (data == NULL)
return (EINVAL);
if ((value_sz = i_get_value_size(type, NULL, 1)) < 0)
return (EINVAL);
bcopy(NVP_VALUE(nvp), data, (size_t)value_sz);
if (nelem != NULL)
*nelem = 1;
break;
case DATA_TYPE_NVLIST:
case DATA_TYPE_STRING:
if (data == NULL)
return (EINVAL);
/*
* This discards the const from nvp, so all callers for these
* types must not accept const nvpairs.
*/
*(void **)data = (void *)NVP_VALUE(nvp);
if (nelem != NULL)
*nelem = 1;
break;
case DATA_TYPE_BOOLEAN_ARRAY:
case DATA_TYPE_BYTE_ARRAY:
case DATA_TYPE_INT8_ARRAY:
case DATA_TYPE_UINT8_ARRAY:
case DATA_TYPE_INT16_ARRAY:
case DATA_TYPE_UINT16_ARRAY:
case DATA_TYPE_INT32_ARRAY:
case DATA_TYPE_UINT32_ARRAY:
case DATA_TYPE_INT64_ARRAY:
case DATA_TYPE_UINT64_ARRAY:
case DATA_TYPE_STRING_ARRAY:
case DATA_TYPE_NVLIST_ARRAY:
if (nelem == NULL || data == NULL)
return (EINVAL);
/*
* This discards the const from nvp, so all callers for these
* types must not accept const nvpairs.
*/
if ((*nelem = NVP_NELEM(nvp)) != 0)
*(void **)data = (void *)NVP_VALUE(nvp);
else
*(void **)data = NULL;
break;
default:
return (ENOTSUP);
}
return (0);
}
static int
nvlist_lookup_common(const nvlist_t *nvl, const char *name, data_type_t type,
uint_t *nelem, void *data)
{
if (name == NULL || nvl == NULL || nvl->nvl_priv == 0)
return (EINVAL);
if (!(nvl->nvl_nvflag & (NV_UNIQUE_NAME | NV_UNIQUE_NAME_TYPE)))
return (ENOTSUP);
nvpair_t *nvp = nvt_lookup_name_type(nvl, name, type);
if (nvp == NULL)
return (ENOENT);
return (nvpair_value_common(nvp, type, nelem, data));
}
int
nvlist_lookup_boolean(const nvlist_t *nvl, const char *name)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_BOOLEAN, NULL, NULL));
}
int
nvlist_lookup_boolean_value(const nvlist_t *nvl, const char *name,
boolean_t *val)
{
return (nvlist_lookup_common(nvl, name,
DATA_TYPE_BOOLEAN_VALUE, NULL, val));
}
int
nvlist_lookup_byte(const nvlist_t *nvl, const char *name, uchar_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_BYTE, NULL, val));
}
int
nvlist_lookup_int8(const nvlist_t *nvl, const char *name, int8_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_INT8, NULL, val));
}
int
nvlist_lookup_uint8(const nvlist_t *nvl, const char *name, uint8_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_UINT8, NULL, val));
}
int
nvlist_lookup_int16(const nvlist_t *nvl, const char *name, int16_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_INT16, NULL, val));
}
int
nvlist_lookup_uint16(const nvlist_t *nvl, const char *name, uint16_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_UINT16, NULL, val));
}
int
nvlist_lookup_int32(const nvlist_t *nvl, const char *name, int32_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_INT32, NULL, val));
}
int
nvlist_lookup_uint32(const nvlist_t *nvl, const char *name, uint32_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_UINT32, NULL, val));
}
int
nvlist_lookup_int64(const nvlist_t *nvl, const char *name, int64_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_INT64, NULL, val));
}
int
nvlist_lookup_uint64(const nvlist_t *nvl, const char *name, uint64_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_UINT64, NULL, val));
}
#if !defined(_KERNEL)
int
nvlist_lookup_double(const nvlist_t *nvl, const char *name, double *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_DOUBLE, NULL, val));
}
#endif
int
nvlist_lookup_string(nvlist_t *nvl, const char *name, char **val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_STRING, NULL, val));
}
int
nvlist_lookup_nvlist(nvlist_t *nvl, const char *name, nvlist_t **val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_NVLIST, NULL, val));
}
int
nvlist_lookup_boolean_array(nvlist_t *nvl, const char *name,
boolean_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name,
DATA_TYPE_BOOLEAN_ARRAY, n, a));
}
int
nvlist_lookup_byte_array(nvlist_t *nvl, const char *name,
uchar_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_BYTE_ARRAY, n, a));
}
int
nvlist_lookup_int8_array(nvlist_t *nvl, const char *name, int8_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_INT8_ARRAY, n, a));
}
int
nvlist_lookup_uint8_array(nvlist_t *nvl, const char *name,
uint8_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_UINT8_ARRAY, n, a));
}
int
nvlist_lookup_int16_array(nvlist_t *nvl, const char *name,
int16_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_INT16_ARRAY, n, a));
}
int
nvlist_lookup_uint16_array(nvlist_t *nvl, const char *name,
uint16_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_UINT16_ARRAY, n, a));
}
int
nvlist_lookup_int32_array(nvlist_t *nvl, const char *name,
int32_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_INT32_ARRAY, n, a));
}
int
nvlist_lookup_uint32_array(nvlist_t *nvl, const char *name,
uint32_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_UINT32_ARRAY, n, a));
}
int
nvlist_lookup_int64_array(nvlist_t *nvl, const char *name,
int64_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_INT64_ARRAY, n, a));
}
int
nvlist_lookup_uint64_array(nvlist_t *nvl, const char *name,
uint64_t **a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_UINT64_ARRAY, n, a));
}
int
nvlist_lookup_string_array(nvlist_t *nvl, const char *name,
char ***a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_STRING_ARRAY, n, a));
}
int
nvlist_lookup_nvlist_array(nvlist_t *nvl, const char *name,
nvlist_t ***a, uint_t *n)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_NVLIST_ARRAY, n, a));
}
int
nvlist_lookup_hrtime(nvlist_t *nvl, const char *name, hrtime_t *val)
{
return (nvlist_lookup_common(nvl, name, DATA_TYPE_HRTIME, NULL, val));
}
int
nvlist_lookup_pairs(nvlist_t *nvl, int flag, ...)
{
va_list ap;
char *name;
int noentok = (flag & NV_FLAG_NOENTOK ? 1 : 0);
int ret = 0;
va_start(ap, flag);
while (ret == 0 && (name = va_arg(ap, char *)) != NULL) {
data_type_t type;
void *val;
uint_t *nelem;
switch (type = va_arg(ap, data_type_t)) {
case DATA_TYPE_BOOLEAN:
ret = nvlist_lookup_common(nvl, name, type, NULL, NULL);
break;
case DATA_TYPE_BOOLEAN_VALUE:
case DATA_TYPE_BYTE:
case DATA_TYPE_INT8:
case DATA_TYPE_UINT8:
case DATA_TYPE_INT16:
case DATA_TYPE_UINT16:
case DATA_TYPE_INT32:
case DATA_TYPE_UINT32:
case DATA_TYPE_INT64:
case DATA_TYPE_UINT64:
case DATA_TYPE_HRTIME:
case DATA_TYPE_STRING:
case DATA_TYPE_NVLIST:
#if !defined(_KERNEL)
case DATA_TYPE_DOUBLE:
#endif
val = va_arg(ap, void *);
ret = nvlist_lookup_common(nvl, name, type, NULL, val);
break;
case DATA_TYPE_BYTE_ARRAY:
case DATA_TYPE_BOOLEAN_ARRAY:
case DATA_TYPE_INT8_ARRAY:
case DATA_TYPE_UINT8_ARRAY:
case DATA_TYPE_INT16_ARRAY:
case DATA_TYPE_UINT16_ARRAY:
case DATA_TYPE_INT32_ARRAY:
case DATA_TYPE_UINT32_ARRAY:
case DATA_TYPE_INT64_ARRAY:
case DATA_TYPE_UINT64_ARRAY:
case DATA_TYPE_STRING_ARRAY:
case DATA_TYPE_NVLIST_ARRAY:
val = va_arg(ap, void *);
nelem = va_arg(ap, uint_t *);
ret = nvlist_lookup_common(nvl, name, type, nelem, val);
break;
default:
ret = EINVAL;
}
if (ret == ENOENT && noentok)
ret = 0;
}
va_end(ap);
return (ret);
}
/*
* Find the 'name'ed nvpair in the nvlist 'nvl'. If 'name' found, the function
* returns zero and a pointer to the matching nvpair is returned in '*ret'
* (given 'ret' is non-NULL). If 'sep' is specified then 'name' will penitrate
* multiple levels of embedded nvlists, with 'sep' as the separator. As an
* example, if sep is '.', name might look like: "a" or "a.b" or "a.c[3]" or
* "a.d[3].e[1]". This matches the C syntax for array embed (for convenience,
* code also supports "a.d[3]e[1]" syntax).
*
* If 'ip' is non-NULL and the last name component is an array, return the
* value of the "...[index]" array index in *ip. For an array reference that
* is not indexed, *ip will be returned as -1. If there is a syntax error in
* 'name', and 'ep' is non-NULL then *ep will be set to point to the location
* inside the 'name' string where the syntax error was detected.
*/
static int
nvlist_lookup_nvpair_ei_sep(nvlist_t *nvl, const char *name, const char sep,
nvpair_t **ret, int *ip, char **ep)
{
nvpair_t *nvp;
const char *np;
char *sepp = NULL;
char *idxp, *idxep;
nvlist_t **nva;
long idx = 0;
int n;
if (ip)
*ip = -1; /* not indexed */
if (ep)
*ep = NULL;
if ((nvl == NULL) || (name == NULL))
return (EINVAL);
sepp = NULL;
idx = 0;
/* step through components of name */
for (np = name; np && *np; np = sepp) {
/* ensure unique names */
if (!(nvl->nvl_nvflag & NV_UNIQUE_NAME))
return (ENOTSUP);
/* skip white space */
skip_whitespace(np);
if (*np == 0)
break;
/* set 'sepp' to end of current component 'np' */
if (sep)
sepp = strchr(np, sep);
else
sepp = NULL;
/* find start of next "[ index ]..." */
idxp = strchr(np, '[');
/* if sepp comes first, set idxp to NULL */
if (sepp && idxp && (sepp < idxp))
idxp = NULL;
/*
* At this point 'idxp' is set if there is an index
* expected for the current component.
*/
if (idxp) {
/* set 'n' to length of current 'np' name component */
n = idxp++ - np;
/* keep sepp up to date for *ep use as we advance */
skip_whitespace(idxp);
sepp = idxp;
/* determine the index value */
#if defined(_KERNEL)
if (ddi_strtol(idxp, &idxep, 0, &idx))
goto fail;
#else
idx = strtol(idxp, &idxep, 0);
#endif
if (idxep == idxp)
goto fail;
/* keep sepp up to date for *ep use as we advance */
sepp = idxep;
/* skip white space index value and check for ']' */
skip_whitespace(sepp);
if (*sepp++ != ']')
goto fail;
/* for embedded arrays, support C syntax: "a[1].b" */
skip_whitespace(sepp);
if (sep && (*sepp == sep))
sepp++;
} else if (sepp) {
n = sepp++ - np;
} else {
n = strlen(np);
}
/* trim trailing whitespace by reducing length of 'np' */
if (n == 0)
goto fail;
for (n--; (np[n] == ' ') || (np[n] == '\t'); n--)
;
n++;
/* skip whitespace, and set sepp to NULL if complete */
if (sepp) {
skip_whitespace(sepp);
if (*sepp == 0)
sepp = NULL;
}
/*
* At this point:
* o 'n' is the length of current 'np' component.
* o 'idxp' is set if there was an index, and value 'idx'.
* o 'sepp' is set to the beginning of the next component,
* and set to NULL if we have no more components.
*
* Search for nvpair with matching component name.
*/
for (nvp = nvlist_next_nvpair(nvl, NULL); nvp != NULL;
nvp = nvlist_next_nvpair(nvl, nvp)) {
/* continue if no match on name */
if (strncmp(np, nvpair_name(nvp), n) ||
(strlen(nvpair_name(nvp)) != n))
continue;
/* if indexed, verify type is array oriented */
if (idxp && !nvpair_type_is_array(nvp))
goto fail;
/*
* Full match found, return nvp and idx if this
* was the last component.
*/
if (sepp == NULL) {
if (ret)
*ret = nvp;
if (ip && idxp)
*ip = (int)idx; /* return index */
return (0); /* found */
}
/*
* More components: current match must be
* of DATA_TYPE_NVLIST or DATA_TYPE_NVLIST_ARRAY
* to support going deeper.
*/
if (nvpair_type(nvp) == DATA_TYPE_NVLIST) {
nvl = EMBEDDED_NVL(nvp);
break;
} else if (nvpair_type(nvp) == DATA_TYPE_NVLIST_ARRAY) {
(void) nvpair_value_nvlist_array(nvp,
&nva, (uint_t *)&n);
if ((n < 0) || (idx >= n))
goto fail;
nvl = nva[idx];
break;
}
/* type does not support more levels */
goto fail;
}
if (nvp == NULL)
goto fail; /* 'name' not found */
/* search for match of next component in embedded 'nvl' list */
}
fail: if (ep && sepp)
*ep = sepp;
return (EINVAL);
}
/*
* Return pointer to nvpair with specified 'name'.
*/
int
nvlist_lookup_nvpair(nvlist_t *nvl, const char *name, nvpair_t **ret)
{
return (nvlist_lookup_nvpair_ei_sep(nvl, name, 0, ret, NULL, NULL));
}
/*
* Determine if named nvpair exists in nvlist (use embedded separator of '.'
* and return array index). See nvlist_lookup_nvpair_ei_sep for more detailed
* description.
*/
int nvlist_lookup_nvpair_embedded_index(nvlist_t *nvl,
const char *name, nvpair_t **ret, int *ip, char **ep)
{
return (nvlist_lookup_nvpair_ei_sep(nvl, name, '.', ret, ip, ep));
}
boolean_t
nvlist_exists(const nvlist_t *nvl, const char *name)
{
nvpriv_t *priv;
nvpair_t *nvp;
i_nvp_t *curr;
if (name == NULL || nvl == NULL ||
(priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv) == NULL)
return (B_FALSE);
for (curr = priv->nvp_list; curr != NULL; curr = curr->nvi_next) {
nvp = &curr->nvi_nvp;
if (strcmp(name, NVP_NAME(nvp)) == 0)
return (B_TRUE);
}
return (B_FALSE);
}
int
nvpair_value_boolean_value(const nvpair_t *nvp, boolean_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_BOOLEAN_VALUE, NULL, val));
}
int
nvpair_value_byte(const nvpair_t *nvp, uchar_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_BYTE, NULL, val));
}
int
nvpair_value_int8(const nvpair_t *nvp, int8_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_INT8, NULL, val));
}
int
nvpair_value_uint8(const nvpair_t *nvp, uint8_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_UINT8, NULL, val));
}
int
nvpair_value_int16(const nvpair_t *nvp, int16_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_INT16, NULL, val));
}
int
nvpair_value_uint16(const nvpair_t *nvp, uint16_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_UINT16, NULL, val));
}
int
nvpair_value_int32(const nvpair_t *nvp, int32_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_INT32, NULL, val));
}
int
nvpair_value_uint32(const nvpair_t *nvp, uint32_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_UINT32, NULL, val));
}
int
nvpair_value_int64(const nvpair_t *nvp, int64_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_INT64, NULL, val));
}
int
nvpair_value_uint64(const nvpair_t *nvp, uint64_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_UINT64, NULL, val));
}
#if !defined(_KERNEL)
int
nvpair_value_double(const nvpair_t *nvp, double *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_DOUBLE, NULL, val));
}
#endif
int
nvpair_value_string(nvpair_t *nvp, char **val)
{
return (nvpair_value_common(nvp, DATA_TYPE_STRING, NULL, val));
}
int
nvpair_value_nvlist(nvpair_t *nvp, nvlist_t **val)
{
return (nvpair_value_common(nvp, DATA_TYPE_NVLIST, NULL, val));
}
int
nvpair_value_boolean_array(nvpair_t *nvp, boolean_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_BOOLEAN_ARRAY, nelem, val));
}
int
nvpair_value_byte_array(nvpair_t *nvp, uchar_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_BYTE_ARRAY, nelem, val));
}
int
nvpair_value_int8_array(nvpair_t *nvp, int8_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_INT8_ARRAY, nelem, val));
}
int
nvpair_value_uint8_array(nvpair_t *nvp, uint8_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_UINT8_ARRAY, nelem, val));
}
int
nvpair_value_int16_array(nvpair_t *nvp, int16_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_INT16_ARRAY, nelem, val));
}
int
nvpair_value_uint16_array(nvpair_t *nvp, uint16_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_UINT16_ARRAY, nelem, val));
}
int
nvpair_value_int32_array(nvpair_t *nvp, int32_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_INT32_ARRAY, nelem, val));
}
int
nvpair_value_uint32_array(nvpair_t *nvp, uint32_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_UINT32_ARRAY, nelem, val));
}
int
nvpair_value_int64_array(nvpair_t *nvp, int64_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_INT64_ARRAY, nelem, val));
}
int
nvpair_value_uint64_array(nvpair_t *nvp, uint64_t **val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_UINT64_ARRAY, nelem, val));
}
int
nvpair_value_string_array(nvpair_t *nvp, char ***val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_STRING_ARRAY, nelem, val));
}
int
nvpair_value_nvlist_array(nvpair_t *nvp, nvlist_t ***val, uint_t *nelem)
{
return (nvpair_value_common(nvp, DATA_TYPE_NVLIST_ARRAY, nelem, val));
}
int
nvpair_value_hrtime(nvpair_t *nvp, hrtime_t *val)
{
return (nvpair_value_common(nvp, DATA_TYPE_HRTIME, NULL, val));
}
/*
* Add specified pair to the list.
*/
int
nvlist_add_nvpair(nvlist_t *nvl, nvpair_t *nvp)
{
if (nvl == NULL || nvp == NULL)
return (EINVAL);
return (nvlist_add_common(nvl, NVP_NAME(nvp), NVP_TYPE(nvp),
NVP_NELEM(nvp), NVP_VALUE(nvp)));
}
/*
* Merge the supplied nvlists and put the result in dst.
* The merged list will contain all names specified in both lists,
* the values are taken from nvl in the case of duplicates.
* Return 0 on success.
*/
int
nvlist_merge(nvlist_t *dst, nvlist_t *nvl, int flag)
{
(void) flag;
if (nvl == NULL || dst == NULL)
return (EINVAL);
if (dst != nvl)
return (nvlist_copy_pairs(nvl, dst));
return (0);
}
/*
* Encoding related routines
*/
#define NVS_OP_ENCODE 0
#define NVS_OP_DECODE 1
#define NVS_OP_GETSIZE 2
typedef struct nvs_ops nvs_ops_t;
typedef struct {
int nvs_op;
const nvs_ops_t *nvs_ops;
void *nvs_private;
nvpriv_t *nvs_priv;
int nvs_recursion;
} nvstream_t;
/*
* nvs operations are:
* - nvs_nvlist
* encoding / decoding of an nvlist header (nvlist_t)
* calculates the size used for header and end detection
*
* - nvs_nvpair
* responsible for the first part of encoding / decoding of an nvpair
* calculates the decoded size of an nvpair
*
* - nvs_nvp_op
* second part of encoding / decoding of an nvpair
*
* - nvs_nvp_size
* calculates the encoding size of an nvpair
*
* - nvs_nvl_fini
* encodes the end detection mark (zeros).
*/
struct nvs_ops {
int (*nvs_nvlist)(nvstream_t *, nvlist_t *, size_t *);
int (*nvs_nvpair)(nvstream_t *, nvpair_t *, size_t *);
int (*nvs_nvp_op)(nvstream_t *, nvpair_t *);
int (*nvs_nvp_size)(nvstream_t *, nvpair_t *, size_t *);
int (*nvs_nvl_fini)(nvstream_t *);
};
typedef struct {
char nvh_encoding; /* nvs encoding method */
char nvh_endian; /* nvs endian */
char nvh_reserved1; /* reserved for future use */
char nvh_reserved2; /* reserved for future use */
} nvs_header_t;
static int
nvs_encode_pairs(nvstream_t *nvs, nvlist_t *nvl)
{
nvpriv_t *priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv;
i_nvp_t *curr;
/*
* Walk nvpair in list and encode each nvpair
*/
for (curr = priv->nvp_list; curr != NULL; curr = curr->nvi_next)
if (nvs->nvs_ops->nvs_nvpair(nvs, &curr->nvi_nvp, NULL) != 0)
return (EFAULT);
return (nvs->nvs_ops->nvs_nvl_fini(nvs));
}
static int
nvs_decode_pairs(nvstream_t *nvs, nvlist_t *nvl)
{
nvpair_t *nvp;
size_t nvsize;
int err;
/*
* Get decoded size of next pair in stream, alloc
* memory for nvpair_t, then decode the nvpair
*/
while ((err = nvs->nvs_ops->nvs_nvpair(nvs, NULL, &nvsize)) == 0) {
if (nvsize == 0) /* end of list */
break;
/* make sure len makes sense */
if (nvsize < NVP_SIZE_CALC(1, 0))
return (EFAULT);
if ((nvp = nvp_buf_alloc(nvl, nvsize)) == NULL)
return (ENOMEM);
if ((err = nvs->nvs_ops->nvs_nvp_op(nvs, nvp)) != 0) {
nvp_buf_free(nvl, nvp);
return (err);
}
if (i_validate_nvpair(nvp) != 0) {
nvpair_free(nvp);
nvp_buf_free(nvl, nvp);
return (EFAULT);
}
err = nvt_add_nvpair(nvl, nvp);
if (err != 0) {
nvpair_free(nvp);
nvp_buf_free(nvl, nvp);
return (err);
}
nvp_buf_link(nvl, nvp);
}
return (err);
}
static int
nvs_getsize_pairs(nvstream_t *nvs, nvlist_t *nvl, size_t *buflen)
{
nvpriv_t *priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv;
i_nvp_t *curr;
uint64_t nvsize = *buflen;
size_t size;
/*
* Get encoded size of nvpairs in nvlist
*/
for (curr = priv->nvp_list; curr != NULL; curr = curr->nvi_next) {
if (nvs->nvs_ops->nvs_nvp_size(nvs, &curr->nvi_nvp, &size) != 0)
return (EINVAL);
if ((nvsize += size) > INT32_MAX)
return (EINVAL);
}
*buflen = nvsize;
return (0);
}
static int
nvs_operation(nvstream_t *nvs, nvlist_t *nvl, size_t *buflen)
{
int err;
if (nvl->nvl_priv == 0)
return (EFAULT);
/*
* Perform the operation, starting with header, then each nvpair
*/
if ((err = nvs->nvs_ops->nvs_nvlist(nvs, nvl, buflen)) != 0)
return (err);
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
err = nvs_encode_pairs(nvs, nvl);
break;
case NVS_OP_DECODE:
err = nvs_decode_pairs(nvs, nvl);
break;
case NVS_OP_GETSIZE:
err = nvs_getsize_pairs(nvs, nvl, buflen);
break;
default:
err = EINVAL;
}
return (err);
}
static int
nvs_embedded(nvstream_t *nvs, nvlist_t *embedded)
{
switch (nvs->nvs_op) {
case NVS_OP_ENCODE: {
int err;
if (nvs->nvs_recursion >= nvpair_max_recursion)
return (EINVAL);
nvs->nvs_recursion++;
err = nvs_operation(nvs, embedded, NULL);
nvs->nvs_recursion--;
return (err);
}
case NVS_OP_DECODE: {
nvpriv_t *priv;
int err;
if (embedded->nvl_version != NV_VERSION)
return (ENOTSUP);
if ((priv = nv_priv_alloc_embedded(nvs->nvs_priv)) == NULL)
return (ENOMEM);
nvlist_init(embedded, embedded->nvl_nvflag, priv);
if (nvs->nvs_recursion >= nvpair_max_recursion) {
nvlist_free(embedded);
return (EINVAL);
}
nvs->nvs_recursion++;
if ((err = nvs_operation(nvs, embedded, NULL)) != 0)
nvlist_free(embedded);
nvs->nvs_recursion--;
return (err);
}
default:
break;
}
return (EINVAL);
}
static int
nvs_embedded_nvl_array(nvstream_t *nvs, nvpair_t *nvp, size_t *size)
{
size_t nelem = NVP_NELEM(nvp);
nvlist_t **nvlp = EMBEDDED_NVL_ARRAY(nvp);
int i;
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
for (i = 0; i < nelem; i++)
if (nvs_embedded(nvs, nvlp[i]) != 0)
return (EFAULT);
break;
case NVS_OP_DECODE: {
size_t len = nelem * sizeof (uint64_t);
nvlist_t *embedded = (nvlist_t *)((uintptr_t)nvlp + len);
bzero(nvlp, len); /* don't trust packed data */
for (i = 0; i < nelem; i++) {
if (nvs_embedded(nvs, embedded) != 0) {
nvpair_free(nvp);
return (EFAULT);
}
nvlp[i] = embedded++;
}
break;
}
case NVS_OP_GETSIZE: {
uint64_t nvsize = 0;
for (i = 0; i < nelem; i++) {
size_t nvp_sz = 0;
if (nvs_operation(nvs, nvlp[i], &nvp_sz) != 0)
return (EINVAL);
if ((nvsize += nvp_sz) > INT32_MAX)
return (EINVAL);
}
*size = nvsize;
break;
}
default:
return (EINVAL);
}
return (0);
}
static int nvs_native(nvstream_t *, nvlist_t *, char *, size_t *);
static int nvs_xdr(nvstream_t *, nvlist_t *, char *, size_t *);
/*
* Common routine for nvlist operations:
* encode, decode, getsize (encoded size).
*/
static int
nvlist_common(nvlist_t *nvl, char *buf, size_t *buflen, int encoding,
int nvs_op)
{
int err = 0;
nvstream_t nvs;
int nvl_endian;
#if defined(_ZFS_LITTLE_ENDIAN)
int host_endian = 1;
#elif defined(_ZFS_BIG_ENDIAN)
int host_endian = 0;
#else
#error "No endian defined!"
#endif /* _ZFS_LITTLE_ENDIAN */
nvs_header_t *nvh;
if (buflen == NULL || nvl == NULL ||
(nvs.nvs_priv = (nvpriv_t *)(uintptr_t)nvl->nvl_priv) == NULL)
return (EINVAL);
nvs.nvs_op = nvs_op;
nvs.nvs_recursion = 0;
/*
* For NVS_OP_ENCODE and NVS_OP_DECODE make sure an nvlist and
* a buffer is allocated. The first 4 bytes in the buffer are
* used for encoding method and host endian.
*/
switch (nvs_op) {
case NVS_OP_ENCODE:
if (buf == NULL || *buflen < sizeof (nvs_header_t))
return (EINVAL);
nvh = (void *)buf;
nvh->nvh_encoding = encoding;
nvh->nvh_endian = nvl_endian = host_endian;
nvh->nvh_reserved1 = 0;
nvh->nvh_reserved2 = 0;
break;
case NVS_OP_DECODE:
if (buf == NULL || *buflen < sizeof (nvs_header_t))
return (EINVAL);
/* get method of encoding from first byte */
nvh = (void *)buf;
encoding = nvh->nvh_encoding;
nvl_endian = nvh->nvh_endian;
break;
case NVS_OP_GETSIZE:
nvl_endian = host_endian;
/*
* add the size for encoding
*/
*buflen = sizeof (nvs_header_t);
break;
default:
return (ENOTSUP);
}
/*
* Create an nvstream with proper encoding method
*/
switch (encoding) {
case NV_ENCODE_NATIVE:
/*
* check endianness, in case we are unpacking
* from a file
*/
if (nvl_endian != host_endian)
return (ENOTSUP);
err = nvs_native(&nvs, nvl, buf, buflen);
break;
case NV_ENCODE_XDR:
err = nvs_xdr(&nvs, nvl, buf, buflen);
break;
default:
err = ENOTSUP;
break;
}
return (err);
}
int
nvlist_size(nvlist_t *nvl, size_t *size, int encoding)
{
return (nvlist_common(nvl, NULL, size, encoding, NVS_OP_GETSIZE));
}
/*
* Pack nvlist into contiguous memory
*/
int
nvlist_pack(nvlist_t *nvl, char **bufp, size_t *buflen, int encoding,
int kmflag)
{
return (nvlist_xpack(nvl, bufp, buflen, encoding,
nvlist_nv_alloc(kmflag)));
}
int
nvlist_xpack(nvlist_t *nvl, char **bufp, size_t *buflen, int encoding,
nv_alloc_t *nva)
{
nvpriv_t nvpriv;
size_t alloc_size;
char *buf;
int err;
if (nva == NULL || nvl == NULL || bufp == NULL || buflen == NULL)
return (EINVAL);
if (*bufp != NULL)
return (nvlist_common(nvl, *bufp, buflen, encoding,
NVS_OP_ENCODE));
/*
* Here is a difficult situation:
* 1. The nvlist has fixed allocator properties.
* All other nvlist routines (like nvlist_add_*, ...) use
* these properties.
* 2. When using nvlist_pack() the user can specify their own
* allocator properties (e.g. by using KM_NOSLEEP).
*
* We use the user specified properties (2). A clearer solution
* will be to remove the kmflag from nvlist_pack(), but we will
* not change the interface.
*/
nv_priv_init(&nvpriv, nva, 0);
if ((err = nvlist_size(nvl, &alloc_size, encoding)))
return (err);
if ((buf = nv_mem_zalloc(&nvpriv, alloc_size)) == NULL)
return (ENOMEM);
if ((err = nvlist_common(nvl, buf, &alloc_size, encoding,
NVS_OP_ENCODE)) != 0) {
nv_mem_free(&nvpriv, buf, alloc_size);
} else {
*buflen = alloc_size;
*bufp = buf;
}
return (err);
}
/*
* Unpack buf into an nvlist_t
*/
int
nvlist_unpack(char *buf, size_t buflen, nvlist_t **nvlp, int kmflag)
{
return (nvlist_xunpack(buf, buflen, nvlp, nvlist_nv_alloc(kmflag)));
}
int
nvlist_xunpack(char *buf, size_t buflen, nvlist_t **nvlp, nv_alloc_t *nva)
{
nvlist_t *nvl;
int err;
if (nvlp == NULL)
return (EINVAL);
if ((err = nvlist_xalloc(&nvl, 0, nva)) != 0)
return (err);
if ((err = nvlist_common(nvl, buf, &buflen, NV_ENCODE_NATIVE,
NVS_OP_DECODE)) != 0)
nvlist_free(nvl);
else
*nvlp = nvl;
return (err);
}
/*
* Native encoding functions
*/
typedef struct {
/*
* This structure is used when decoding a packed nvpair in
* the native format. n_base points to a buffer containing the
* packed nvpair. n_end is a pointer to the end of the buffer.
* (n_end actually points to the first byte past the end of the
* buffer.) n_curr is a pointer that lies between n_base and n_end.
* It points to the current data that we are decoding.
* The amount of data left in the buffer is equal to n_end - n_curr.
* n_flag is used to recognize a packed embedded list.
*/
caddr_t n_base;
caddr_t n_end;
caddr_t n_curr;
uint_t n_flag;
} nvs_native_t;
static int
nvs_native_create(nvstream_t *nvs, nvs_native_t *native, char *buf,
size_t buflen)
{
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
case NVS_OP_DECODE:
nvs->nvs_private = native;
native->n_curr = native->n_base = buf;
native->n_end = buf + buflen;
native->n_flag = 0;
return (0);
case NVS_OP_GETSIZE:
nvs->nvs_private = native;
native->n_curr = native->n_base = native->n_end = NULL;
native->n_flag = 0;
return (0);
default:
return (EINVAL);
}
}
static void
nvs_native_destroy(nvstream_t *nvs)
{
(void) nvs;
}
static int
native_cp(nvstream_t *nvs, void *buf, size_t size)
{
nvs_native_t *native = (nvs_native_t *)nvs->nvs_private;
if (native->n_curr + size > native->n_end)
return (EFAULT);
/*
* The bcopy() below eliminates alignment requirement
* on the buffer (stream) and is preferred over direct access.
*/
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
bcopy(buf, native->n_curr, size);
break;
case NVS_OP_DECODE:
bcopy(native->n_curr, buf, size);
break;
default:
return (EINVAL);
}
native->n_curr += size;
return (0);
}
/*
* operate on nvlist_t header
*/
static int
nvs_native_nvlist(nvstream_t *nvs, nvlist_t *nvl, size_t *size)
{
nvs_native_t *native = nvs->nvs_private;
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
case NVS_OP_DECODE:
if (native->n_flag)
return (0); /* packed embedded list */
native->n_flag = 1;
/* copy version and nvflag of the nvlist_t */
if (native_cp(nvs, &nvl->nvl_version, sizeof (int32_t)) != 0 ||
native_cp(nvs, &nvl->nvl_nvflag, sizeof (int32_t)) != 0)
return (EFAULT);
return (0);
case NVS_OP_GETSIZE:
/*
* if calculate for packed embedded list
* 4 for end of the embedded list
* else
* 2 * sizeof (int32_t) for nvl_version and nvl_nvflag
* and 4 for end of the entire list
*/
if (native->n_flag) {
*size += 4;
} else {
native->n_flag = 1;
*size += 2 * sizeof (int32_t) + 4;
}
return (0);
default:
return (EINVAL);
}
}
static int
nvs_native_nvl_fini(nvstream_t *nvs)
{
if (nvs->nvs_op == NVS_OP_ENCODE) {
nvs_native_t *native = (nvs_native_t *)nvs->nvs_private;
/*
* Add 4 zero bytes at end of nvlist. They are used
* for end detection by the decode routine.
*/
if (native->n_curr + sizeof (int) > native->n_end)
return (EFAULT);
bzero(native->n_curr, sizeof (int));
native->n_curr += sizeof (int);
}
return (0);
}
static int
nvpair_native_embedded(nvstream_t *nvs, nvpair_t *nvp)
{
if (nvs->nvs_op == NVS_OP_ENCODE) {
nvs_native_t *native = (nvs_native_t *)nvs->nvs_private;
nvlist_t *packed = (void *)
(native->n_curr - nvp->nvp_size + NVP_VALOFF(nvp));
/*
* Null out the pointer that is meaningless in the packed
* structure. The address may not be aligned, so we have
* to use bzero.
*/
bzero((char *)packed + offsetof(nvlist_t, nvl_priv),
sizeof (uint64_t));
}
return (nvs_embedded(nvs, EMBEDDED_NVL(nvp)));
}
static int
nvpair_native_embedded_array(nvstream_t *nvs, nvpair_t *nvp)
{
if (nvs->nvs_op == NVS_OP_ENCODE) {
nvs_native_t *native = (nvs_native_t *)nvs->nvs_private;
char *value = native->n_curr - nvp->nvp_size + NVP_VALOFF(nvp);
size_t len = NVP_NELEM(nvp) * sizeof (uint64_t);
nvlist_t *packed = (nvlist_t *)((uintptr_t)value + len);
int i;
/*
* Null out pointers that are meaningless in the packed
* structure. The addresses may not be aligned, so we have
* to use bzero.
*/
bzero(value, len);
for (i = 0; i < NVP_NELEM(nvp); i++, packed++)
/*
* Null out the pointer that is meaningless in the
* packed structure. The address may not be aligned,
* so we have to use bzero.
*/
bzero((char *)packed + offsetof(nvlist_t, nvl_priv),
sizeof (uint64_t));
}
return (nvs_embedded_nvl_array(nvs, nvp, NULL));
}
static void
nvpair_native_string_array(nvstream_t *nvs, nvpair_t *nvp)
{
switch (nvs->nvs_op) {
case NVS_OP_ENCODE: {
nvs_native_t *native = (nvs_native_t *)nvs->nvs_private;
uint64_t *strp = (void *)
(native->n_curr - nvp->nvp_size + NVP_VALOFF(nvp));
/*
* Null out pointers that are meaningless in the packed
* structure. The addresses may not be aligned, so we have
* to use bzero.
*/
bzero(strp, NVP_NELEM(nvp) * sizeof (uint64_t));
break;
}
case NVS_OP_DECODE: {
char **strp = (void *)NVP_VALUE(nvp);
char *buf = ((char *)strp + NVP_NELEM(nvp) * sizeof (uint64_t));
int i;
for (i = 0; i < NVP_NELEM(nvp); i++) {
strp[i] = buf;
buf += strlen(buf) + 1;
}
break;
}
}
}
static int
nvs_native_nvp_op(nvstream_t *nvs, nvpair_t *nvp)
{
data_type_t type;
int value_sz;
int ret = 0;
/*
* We do the initial bcopy of the data before we look at
* the nvpair type, because when we're decoding, we won't
* have the correct values for the pair until we do the bcopy.
*/
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
case NVS_OP_DECODE:
if (native_cp(nvs, nvp, nvp->nvp_size) != 0)
return (EFAULT);
break;
default:
return (EINVAL);
}
/* verify nvp_name_sz, check the name string length */
if (i_validate_nvpair_name(nvp) != 0)
return (EFAULT);
type = NVP_TYPE(nvp);
/*
* Verify type and nelem and get the value size.
* In case of data types DATA_TYPE_STRING and DATA_TYPE_STRING_ARRAY
* is the size of the string(s) excluded.
*/
if ((value_sz = i_get_value_size(type, NULL, NVP_NELEM(nvp))) < 0)
return (EFAULT);
if (NVP_SIZE_CALC(nvp->nvp_name_sz, value_sz) > nvp->nvp_size)
return (EFAULT);
switch (type) {
case DATA_TYPE_NVLIST:
ret = nvpair_native_embedded(nvs, nvp);
break;
case DATA_TYPE_NVLIST_ARRAY:
ret = nvpair_native_embedded_array(nvs, nvp);
break;
case DATA_TYPE_STRING_ARRAY:
nvpair_native_string_array(nvs, nvp);
break;
default:
break;
}
return (ret);
}
static int
nvs_native_nvp_size(nvstream_t *nvs, nvpair_t *nvp, size_t *size)
{
uint64_t nvp_sz = nvp->nvp_size;
switch (NVP_TYPE(nvp)) {
case DATA_TYPE_NVLIST: {
size_t nvsize = 0;
if (nvs_operation(nvs, EMBEDDED_NVL(nvp), &nvsize) != 0)
return (EINVAL);
nvp_sz += nvsize;
break;
}
case DATA_TYPE_NVLIST_ARRAY: {
size_t nvsize;
if (nvs_embedded_nvl_array(nvs, nvp, &nvsize) != 0)
return (EINVAL);
nvp_sz += nvsize;
break;
}
default:
break;
}
if (nvp_sz > INT32_MAX)
return (EINVAL);
*size = nvp_sz;
return (0);
}
static int
nvs_native_nvpair(nvstream_t *nvs, nvpair_t *nvp, size_t *size)
{
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
return (nvs_native_nvp_op(nvs, nvp));
case NVS_OP_DECODE: {
nvs_native_t *native = (nvs_native_t *)nvs->nvs_private;
int32_t decode_len;
/* try to read the size value from the stream */
if (native->n_curr + sizeof (int32_t) > native->n_end)
return (EFAULT);
bcopy(native->n_curr, &decode_len, sizeof (int32_t));
/* sanity check the size value */
if (decode_len < 0 ||
decode_len > native->n_end - native->n_curr)
return (EFAULT);
*size = decode_len;
/*
* If at the end of the stream then move the cursor
* forward, otherwise nvpair_native_op() will read
* the entire nvpair at the same cursor position.
*/
if (*size == 0)
native->n_curr += sizeof (int32_t);
break;
}
default:
return (EINVAL);
}
return (0);
}
static const nvs_ops_t nvs_native_ops = {
.nvs_nvlist = nvs_native_nvlist,
.nvs_nvpair = nvs_native_nvpair,
.nvs_nvp_op = nvs_native_nvp_op,
.nvs_nvp_size = nvs_native_nvp_size,
.nvs_nvl_fini = nvs_native_nvl_fini
};
static int
nvs_native(nvstream_t *nvs, nvlist_t *nvl, char *buf, size_t *buflen)
{
nvs_native_t native;
int err;
nvs->nvs_ops = &nvs_native_ops;
if ((err = nvs_native_create(nvs, &native, buf + sizeof (nvs_header_t),
*buflen - sizeof (nvs_header_t))) != 0)
return (err);
err = nvs_operation(nvs, nvl, buflen);
nvs_native_destroy(nvs);
return (err);
}
/*
* XDR encoding functions
*
* An xdr packed nvlist is encoded as:
*
* - encoding method and host endian (4 bytes)
* - nvl_version (4 bytes)
* - nvl_nvflag (4 bytes)
*
* - encoded nvpairs, the format of one xdr encoded nvpair is:
* - encoded size of the nvpair (4 bytes)
* - decoded size of the nvpair (4 bytes)
* - name string, (4 + sizeof(NV_ALIGN4(string))
* a string is coded as size (4 bytes) and data
* - data type (4 bytes)
* - number of elements in the nvpair (4 bytes)
* - data
*
* - 2 zero's for end of the entire list (8 bytes)
*/
static int
nvs_xdr_create(nvstream_t *nvs, XDR *xdr, char *buf, size_t buflen)
{
/* xdr data must be 4 byte aligned */
if ((ulong_t)buf % 4 != 0)
return (EFAULT);
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
xdrmem_create(xdr, buf, (uint_t)buflen, XDR_ENCODE);
nvs->nvs_private = xdr;
return (0);
case NVS_OP_DECODE:
xdrmem_create(xdr, buf, (uint_t)buflen, XDR_DECODE);
nvs->nvs_private = xdr;
return (0);
case NVS_OP_GETSIZE:
nvs->nvs_private = NULL;
return (0);
default:
return (EINVAL);
}
}
static void
nvs_xdr_destroy(nvstream_t *nvs)
{
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
case NVS_OP_DECODE:
xdr_destroy((XDR *)nvs->nvs_private);
break;
default:
break;
}
}
static int
nvs_xdr_nvlist(nvstream_t *nvs, nvlist_t *nvl, size_t *size)
{
switch (nvs->nvs_op) {
case NVS_OP_ENCODE:
case NVS_OP_DECODE: {
XDR *xdr = nvs->nvs_private;
if (!xdr_int(xdr, &nvl->nvl_version) ||
!xdr_u_int(xdr, &nvl->nvl_nvflag))
return (EFAULT);
break;
}
case NVS_OP_GETSIZE: {
/*
* 2 * 4 for nvl_version + nvl_nvflag
* and 8 for end of the entire list
*/
*size += 2 * 4 + 8;
break;
}
default:
return (EINVAL);
}
return (0);
}
static int
nvs_xdr_nvl_fini(nvstream_t *nvs)
{
if (nvs->nvs_op == NVS_OP_ENCODE) {
XDR *xdr = nvs->nvs_private;
int zero = 0;
if (!xdr_int(xdr, &zero) || !xdr_int(xdr, &zero))
return (EFAULT);
}
return (0);
}
/*
* xdrproc_t-compatible callbacks for xdr_array()
*/
#if defined(_KERNEL) && defined(__linux__) /* Linux kernel */
#define NVS_BUILD_XDRPROC_T(type) \
static bool_t \
nvs_xdr_nvp_##type(XDR *xdrs, void *ptr) \
{ \
return (xdr_##type(xdrs, ptr)); \
}
#elif !defined(_KERNEL) && defined(XDR_CONTROL) /* tirpc */
#define NVS_BUILD_XDRPROC_T(type) \
static bool_t \
nvs_xdr_nvp_##type(XDR *xdrs, ...) \
{ \
va_list args; \
void *ptr; \
\
va_start(args, xdrs); \
ptr = va_arg(args, void *); \
va_end(args); \
\
return (xdr_##type(xdrs, ptr)); \
}
#else /* FreeBSD, sunrpc */
#define NVS_BUILD_XDRPROC_T(type) \
static bool_t \
nvs_xdr_nvp_##type(XDR *xdrs, void *ptr, ...) \
{ \
return (xdr_##type(xdrs, ptr)); \
}
#endif
/* BEGIN CSTYLED */
NVS_BUILD_XDRPROC_T(char);
NVS_BUILD_XDRPROC_T(short);
NVS_BUILD_XDRPROC_T(u_short);
NVS_BUILD_XDRPROC_T(int);
NVS_BUILD_XDRPROC_T(u_int);
NVS_BUILD_XDRPROC_T(longlong_t);
NVS_BUILD_XDRPROC_T(u_longlong_t);
/* END CSTYLED */
/*
* The format of xdr encoded nvpair is:
* encode_size, decode_size, name string, data type, nelem, data
*/
static int
nvs_xdr_nvp_op(nvstream_t *nvs, nvpair_t *nvp)
{
ASSERT(nvs != NULL && nvp != NULL);
data_type_t type;
char *buf;
char *buf_end = (char *)nvp + nvp->nvp_size;
int value_sz;
uint_t nelem, buflen;
bool_t ret = FALSE;
XDR *xdr = nvs->nvs_private;
ASSERT(xdr != NULL);
/* name string */
if ((buf = NVP_NAME(nvp)) >= buf_end)
return (EFAULT);
buflen = buf_end - buf;
if (!xdr_string(xdr, &buf, buflen - 1))
return (EFAULT);
nvp->nvp_name_sz = strlen(buf) + 1;
/* type and nelem */
if (!xdr_int(xdr, (int *)&nvp->nvp_type) ||
!xdr_int(xdr, &nvp->nvp_value_elem))
return (EFAULT);
type = NVP_TYPE(nvp);
nelem = nvp->nvp_value_elem;
/*
* Verify type and nelem and get the value size.
* In case of data types DATA_TYPE_STRING and DATA_TYPE_STRING_ARRAY
* is the size of the string(s) excluded.
*/
if ((value_sz = i_get_value_size(type, NULL, nelem)) < 0)
return (EFAULT);
/* if there is no data to extract then return */
if (nelem == 0)
return (0);
/* value */
if ((buf = NVP_VALUE(nvp)) >= buf_end)
return (EFAULT);
buflen = buf_end - buf;
if (buflen < value_sz)
return (EFAULT);
switch (type) {
case DATA_TYPE_NVLIST:
if (nvs_embedded(nvs, (void *)buf) == 0)
return (0);
break;
case DATA_TYPE_NVLIST_ARRAY:
if (nvs_embedded_nvl_array(nvs, nvp, NULL) == 0)
return (0);
break;
case DATA_TYPE_BOOLEAN:
ret = TRUE;
break;
case DATA_TYPE_BYTE:
case DATA_TYPE_INT8:
case DATA_TYPE_UINT8:
ret = xdr_char(xdr, buf);
break;
case DATA_TYPE_INT16:
ret = xdr_short(xdr, (void *)buf);
break;
case DATA_TYPE_UINT16:
ret = xdr_u_short(xdr, (void *)buf);
break;
case DATA_TYPE_BOOLEAN_VALUE:
case DATA_TYPE_INT32:
ret = xdr_int(xdr, (void *)buf);
break;
case DATA_TYPE_UINT32:
ret = xdr_u_int(xdr, (void *)buf);
break;
case DATA_TYPE_INT64:
ret = xdr_longlong_t(xdr, (void *)buf);
break;
case DATA_TYPE_UINT64:
ret = xdr_u_longlong_t(xdr, (void *)buf);
break;
case DATA_TYPE_HRTIME:
/*
* NOTE: must expose the definition of hrtime_t here
*/
ret = xdr_longlong_t(xdr, (void *)buf);
break;
#if !defined(_KERNEL)
case DATA_TYPE_DOUBLE:
ret = xdr_double(xdr, (void *)buf);
break;
#endif
case DATA_TYPE_STRING:
ret = xdr_string(xdr, &buf, buflen - 1);
break;
case DATA_TYPE_BYTE_ARRAY:
ret = xdr_opaque(xdr, buf, nelem);
break;
case DATA_TYPE_INT8_ARRAY:
case DATA_TYPE_UINT8_ARRAY:
ret = xdr_array(xdr, &buf, &nelem, buflen, sizeof (int8_t),
nvs_xdr_nvp_char);
break;
case DATA_TYPE_INT16_ARRAY:
ret = xdr_array(xdr, &buf, &nelem, buflen / sizeof (int16_t),
sizeof (int16_t), nvs_xdr_nvp_short);
break;
case DATA_TYPE_UINT16_ARRAY:
ret = xdr_array(xdr, &buf, &nelem, buflen / sizeof (uint16_t),
sizeof (uint16_t), nvs_xdr_nvp_u_short);
break;
case DATA_TYPE_BOOLEAN_ARRAY:
case DATA_TYPE_INT32_ARRAY:
ret = xdr_array(xdr, &buf, &nelem, buflen / sizeof (int32_t),
sizeof (int32_t), nvs_xdr_nvp_int);
break;
case DATA_TYPE_UINT32_ARRAY:
ret = xdr_array(xdr, &buf, &nelem, buflen / sizeof (uint32_t),
sizeof (uint32_t), nvs_xdr_nvp_u_int);
break;
case DATA_TYPE_INT64_ARRAY:
ret = xdr_array(xdr, &buf, &nelem, buflen / sizeof (int64_t),
sizeof (int64_t), nvs_xdr_nvp_longlong_t);
break;
case DATA_TYPE_UINT64_ARRAY:
ret = xdr_array(xdr, &buf, &nelem, buflen / sizeof (uint64_t),
sizeof (uint64_t), nvs_xdr_nvp_u_longlong_t);
break;
case DATA_TYPE_STRING_ARRAY: {
size_t len = nelem * sizeof (uint64_t);
char **strp = (void *)buf;
int i;
if (nvs->nvs_op == NVS_OP_DECODE)
bzero(buf, len); /* don't trust packed data */
for (i = 0; i < nelem; i++) {
if (buflen <= len)
return (EFAULT);
buf += len;
buflen -= len;
if (xdr_string(xdr, &buf, buflen - 1) != TRUE)
return (EFAULT);
if (nvs->nvs_op == NVS_OP_DECODE)
strp[i] = buf;
len = strlen(buf) + 1;
}
ret = TRUE;
break;
}
default:
break;
}
return (ret == TRUE ? 0 : EFAULT);
}
static int
nvs_xdr_nvp_size(nvstream_t *nvs, nvpair_t *nvp, size_t *size)
{
data_type_t type = NVP_TYPE(nvp);
/*
* encode_size + decode_size + name string size + data type + nelem
* where name string size = 4 + NV_ALIGN4(strlen(NVP_NAME(nvp)))
*/
uint64_t nvp_sz = 4 + 4 + 4 + NV_ALIGN4(strlen(NVP_NAME(nvp))) + 4 + 4;
switch (type) {
case DATA_TYPE_BOOLEAN:
break;
case DATA_TYPE_BOOLEAN_VALUE:
case DATA_TYPE_BYTE:
case DATA_TYPE_INT8:
case DATA_TYPE_UINT8:
case DATA_TYPE_INT16:
case DATA_TYPE_UINT16:
case DATA_TYPE_INT32:
case DATA_TYPE_UINT32:
nvp_sz += 4; /* 4 is the minimum xdr unit */
break;
case DATA_TYPE_INT64:
case DATA_TYPE_UINT64:
case DATA_TYPE_HRTIME:
#if !defined(_KERNEL)
case DATA_TYPE_DOUBLE:
#endif
nvp_sz += 8;
break;
case DATA_TYPE_STRING:
nvp_sz += 4 + NV_ALIGN4(strlen((char *)NVP_VALUE(nvp)));
break;
case DATA_TYPE_BYTE_ARRAY:
nvp_sz += NV_ALIGN4(NVP_NELEM(nvp));
break;
case DATA_TYPE_BOOLEAN_ARRAY:
case DATA_TYPE_INT8_ARRAY:
case DATA_TYPE_UINT8_ARRAY:
case DATA_TYPE_INT16_ARRAY:
case DATA_TYPE_UINT16_ARRAY:
case DATA_TYPE_INT32_ARRAY:
case DATA_TYPE_UINT32_ARRAY:
nvp_sz += 4 + 4 * (uint64_t)NVP_NELEM(nvp);
break;
case DATA_TYPE_INT64_ARRAY:
case DATA_TYPE_UINT64_ARRAY:
nvp_sz += 4 + 8 * (uint64_t)NVP_NELEM(nvp);
break;
case DATA_TYPE_STRING_ARRAY: {
int i;
char **strs = (void *)NVP_VALUE(nvp);
for (i = 0; i < NVP_NELEM(nvp); i++)
nvp_sz += 4 + NV_ALIGN4(strlen(strs[i]));
break;
}
case DATA_TYPE_NVLIST:
case DATA_TYPE_NVLIST_ARRAY: {
size_t nvsize = 0;
int old_nvs_op = nvs->nvs_op;
int err;
nvs->nvs_op = NVS_OP_GETSIZE;
if (type == DATA_TYPE_NVLIST)
err = nvs_operation(nvs, EMBEDDED_NVL(nvp), &nvsize);
else
err = nvs_embedded_nvl_array(nvs, nvp, &nvsize);
nvs->nvs_op = old_nvs_op;
if (err != 0)
return (EINVAL);
nvp_sz += nvsize;
break;
}
default:
return (EINVAL);
}
if (nvp_sz > INT32_MAX)
return (EINVAL);
*size = nvp_sz;
return (0);
}
/*
* The NVS_XDR_MAX_LEN macro takes a packed xdr buffer of size x and estimates
* the largest nvpair that could be encoded in the buffer.
*
* See comments above nvpair_xdr_op() for the format of xdr encoding.
* The size of a xdr packed nvpair without any data is 5 words.
*
* Using the size of the data directly as an estimate would be ok
* in all cases except one. If the data type is of DATA_TYPE_STRING_ARRAY
* then the actual nvpair has space for an array of pointers to index
* the strings. These pointers are not encoded into the packed xdr buffer.
*
* If the data is of type DATA_TYPE_STRING_ARRAY and all the strings are
* of length 0, then each string is encoded in xdr format as a single word.
* Therefore when expanded to an nvpair there will be 2.25 word used for
* each string. (a int64_t allocated for pointer usage, and a single char
* for the null termination.)
*
* This is the calculation performed by the NVS_XDR_MAX_LEN macro.
*/
#define NVS_XDR_HDR_LEN ((size_t)(5 * 4))
#define NVS_XDR_DATA_LEN(y) (((size_t)(y) <= NVS_XDR_HDR_LEN) ? \
0 : ((size_t)(y) - NVS_XDR_HDR_LEN))
#define NVS_XDR_MAX_LEN(x) (NVP_SIZE_CALC(1, 0) + \
(NVS_XDR_DATA_LEN(x) * 2) + \
NV_ALIGN4((NVS_XDR_DATA_LEN(x) / 4)))
static int
nvs_xdr_nvpair(nvstream_t *nvs, nvpair_t *nvp, size_t *size)
{
XDR *xdr = nvs->nvs_private;
int32_t encode_len, decode_len;
switch (nvs->nvs_op) {
case NVS_OP_ENCODE: {
size_t nvsize;
if (nvs_xdr_nvp_size(nvs, nvp, &nvsize) != 0)
return (EFAULT);
decode_len = nvp->nvp_size;
encode_len = nvsize;
if (!xdr_int(xdr, &encode_len) || !xdr_int(xdr, &decode_len))
return (EFAULT);
return (nvs_xdr_nvp_op(nvs, nvp));
}
case NVS_OP_DECODE: {
struct xdr_bytesrec bytesrec;
/* get the encode and decode size */
if (!xdr_int(xdr, &encode_len) || !xdr_int(xdr, &decode_len))
return (EFAULT);
*size = decode_len;
/* are we at the end of the stream? */
if (*size == 0)
return (0);
/* sanity check the size parameter */
if (!xdr_control(xdr, XDR_GET_BYTES_AVAIL, &bytesrec))
return (EFAULT);
if (*size > NVS_XDR_MAX_LEN(bytesrec.xc_num_avail))
return (EFAULT);
break;
}
default:
return (EINVAL);
}
return (0);
}
static const struct nvs_ops nvs_xdr_ops = {
.nvs_nvlist = nvs_xdr_nvlist,
.nvs_nvpair = nvs_xdr_nvpair,
.nvs_nvp_op = nvs_xdr_nvp_op,
.nvs_nvp_size = nvs_xdr_nvp_size,
.nvs_nvl_fini = nvs_xdr_nvl_fini
};
static int
nvs_xdr(nvstream_t *nvs, nvlist_t *nvl, char *buf, size_t *buflen)
{
XDR xdr;
int err;
nvs->nvs_ops = &nvs_xdr_ops;
if ((err = nvs_xdr_create(nvs, &xdr, buf + sizeof (nvs_header_t),
*buflen - sizeof (nvs_header_t))) != 0)
return (err);
err = nvs_operation(nvs, nvl, buflen);
nvs_xdr_destroy(nvs);
return (err);
}
#if defined(_KERNEL)
static int __init
nvpair_init(void)
{
return (0);
}
static void __exit
nvpair_fini(void)
{
}
module_init(nvpair_init);
module_exit(nvpair_fini);
#endif
ZFS_MODULE_DESCRIPTION("Generic name/value pair implementation");
ZFS_MODULE_AUTHOR(ZFS_META_AUTHOR);
ZFS_MODULE_LICENSE(ZFS_META_LICENSE);
ZFS_MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
EXPORT_SYMBOL(nv_alloc_init);
EXPORT_SYMBOL(nv_alloc_reset);
EXPORT_SYMBOL(nv_alloc_fini);
/* list management */
EXPORT_SYMBOL(nvlist_alloc);
EXPORT_SYMBOL(nvlist_free);
EXPORT_SYMBOL(nvlist_size);
EXPORT_SYMBOL(nvlist_pack);
EXPORT_SYMBOL(nvlist_unpack);
EXPORT_SYMBOL(nvlist_dup);
EXPORT_SYMBOL(nvlist_merge);
EXPORT_SYMBOL(nvlist_xalloc);
EXPORT_SYMBOL(nvlist_xpack);
EXPORT_SYMBOL(nvlist_xunpack);
EXPORT_SYMBOL(nvlist_xdup);
EXPORT_SYMBOL(nvlist_lookup_nv_alloc);
EXPORT_SYMBOL(nvlist_add_nvpair);
EXPORT_SYMBOL(nvlist_add_boolean);
EXPORT_SYMBOL(nvlist_add_boolean_value);
EXPORT_SYMBOL(nvlist_add_byte);
EXPORT_SYMBOL(nvlist_add_int8);
EXPORT_SYMBOL(nvlist_add_uint8);
EXPORT_SYMBOL(nvlist_add_int16);
EXPORT_SYMBOL(nvlist_add_uint16);
EXPORT_SYMBOL(nvlist_add_int32);
EXPORT_SYMBOL(nvlist_add_uint32);
EXPORT_SYMBOL(nvlist_add_int64);
EXPORT_SYMBOL(nvlist_add_uint64);
EXPORT_SYMBOL(nvlist_add_string);
EXPORT_SYMBOL(nvlist_add_nvlist);
EXPORT_SYMBOL(nvlist_add_boolean_array);
EXPORT_SYMBOL(nvlist_add_byte_array);
EXPORT_SYMBOL(nvlist_add_int8_array);
EXPORT_SYMBOL(nvlist_add_uint8_array);
EXPORT_SYMBOL(nvlist_add_int16_array);
EXPORT_SYMBOL(nvlist_add_uint16_array);
EXPORT_SYMBOL(nvlist_add_int32_array);
EXPORT_SYMBOL(nvlist_add_uint32_array);
EXPORT_SYMBOL(nvlist_add_int64_array);
EXPORT_SYMBOL(nvlist_add_uint64_array);
EXPORT_SYMBOL(nvlist_add_string_array);
EXPORT_SYMBOL(nvlist_add_nvlist_array);
EXPORT_SYMBOL(nvlist_next_nvpair);
EXPORT_SYMBOL(nvlist_prev_nvpair);
EXPORT_SYMBOL(nvlist_empty);
EXPORT_SYMBOL(nvlist_add_hrtime);
EXPORT_SYMBOL(nvlist_remove);
EXPORT_SYMBOL(nvlist_remove_nvpair);
EXPORT_SYMBOL(nvlist_remove_all);
EXPORT_SYMBOL(nvlist_lookup_boolean);
EXPORT_SYMBOL(nvlist_lookup_boolean_value);
EXPORT_SYMBOL(nvlist_lookup_byte);
EXPORT_SYMBOL(nvlist_lookup_int8);
EXPORT_SYMBOL(nvlist_lookup_uint8);
EXPORT_SYMBOL(nvlist_lookup_int16);
EXPORT_SYMBOL(nvlist_lookup_uint16);
EXPORT_SYMBOL(nvlist_lookup_int32);
EXPORT_SYMBOL(nvlist_lookup_uint32);
EXPORT_SYMBOL(nvlist_lookup_int64);
EXPORT_SYMBOL(nvlist_lookup_uint64);
EXPORT_SYMBOL(nvlist_lookup_string);
EXPORT_SYMBOL(nvlist_lookup_nvlist);
EXPORT_SYMBOL(nvlist_lookup_boolean_array);
EXPORT_SYMBOL(nvlist_lookup_byte_array);
EXPORT_SYMBOL(nvlist_lookup_int8_array);
EXPORT_SYMBOL(nvlist_lookup_uint8_array);
EXPORT_SYMBOL(nvlist_lookup_int16_array);
EXPORT_SYMBOL(nvlist_lookup_uint16_array);
EXPORT_SYMBOL(nvlist_lookup_int32_array);
EXPORT_SYMBOL(nvlist_lookup_uint32_array);
EXPORT_SYMBOL(nvlist_lookup_int64_array);
EXPORT_SYMBOL(nvlist_lookup_uint64_array);
EXPORT_SYMBOL(nvlist_lookup_string_array);
EXPORT_SYMBOL(nvlist_lookup_nvlist_array);
EXPORT_SYMBOL(nvlist_lookup_hrtime);
EXPORT_SYMBOL(nvlist_lookup_pairs);
EXPORT_SYMBOL(nvlist_lookup_nvpair);
EXPORT_SYMBOL(nvlist_exists);
/* processing nvpair */
EXPORT_SYMBOL(nvpair_name);
EXPORT_SYMBOL(nvpair_type);
EXPORT_SYMBOL(nvpair_value_boolean_value);
EXPORT_SYMBOL(nvpair_value_byte);
EXPORT_SYMBOL(nvpair_value_int8);
EXPORT_SYMBOL(nvpair_value_uint8);
EXPORT_SYMBOL(nvpair_value_int16);
EXPORT_SYMBOL(nvpair_value_uint16);
EXPORT_SYMBOL(nvpair_value_int32);
EXPORT_SYMBOL(nvpair_value_uint32);
EXPORT_SYMBOL(nvpair_value_int64);
EXPORT_SYMBOL(nvpair_value_uint64);
EXPORT_SYMBOL(nvpair_value_string);
EXPORT_SYMBOL(nvpair_value_nvlist);
EXPORT_SYMBOL(nvpair_value_boolean_array);
EXPORT_SYMBOL(nvpair_value_byte_array);
EXPORT_SYMBOL(nvpair_value_int8_array);
EXPORT_SYMBOL(nvpair_value_uint8_array);
EXPORT_SYMBOL(nvpair_value_int16_array);
EXPORT_SYMBOL(nvpair_value_uint16_array);
EXPORT_SYMBOL(nvpair_value_int32_array);
EXPORT_SYMBOL(nvpair_value_uint32_array);
EXPORT_SYMBOL(nvpair_value_int64_array);
EXPORT_SYMBOL(nvpair_value_uint64_array);
EXPORT_SYMBOL(nvpair_value_string_array);
EXPORT_SYMBOL(nvpair_value_nvlist_array);
EXPORT_SYMBOL(nvpair_value_hrtime);
diff --git a/module/nvpair/nvpair_alloc_fixed.c b/module/nvpair/nvpair_alloc_fixed.c
index ca3f68676100..d7d3e7afd520 100644
--- a/module/nvpair/nvpair_alloc_fixed.c
+++ b/module/nvpair/nvpair_alloc_fixed.c
@@ -1,115 +1,115 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#include <sys/isa_defs.h>
#include <sys/nvpair.h>
#include <sys/sysmacros.h>
/*
* This allocator is very simple.
* - it uses a pre-allocated buffer for memory allocations.
* - it does _not_ free memory in the pre-allocated buffer.
*
* The reason for the selected implementation is simplicity.
* This allocator is designed for the usage in interrupt context when
* the caller may not wait for free memory.
*/
/* pre-allocated buffer for memory allocations */
typedef struct nvbuf {
uintptr_t nvb_buf; /* address of pre-allocated buffer */
uintptr_t nvb_lim; /* limit address in the buffer */
uintptr_t nvb_cur; /* current address in the buffer */
} nvbuf_t;
/*
* Initialize the pre-allocated buffer allocator. The caller needs to supply
*
* buf address of pre-allocated buffer
* bufsz size of pre-allocated buffer
*
* nv_fixed_init() calculates the remaining members of nvbuf_t.
*/
static int
nv_fixed_init(nv_alloc_t *nva, va_list valist)
{
uintptr_t base = va_arg(valist, uintptr_t);
uintptr_t lim = base + va_arg(valist, size_t);
nvbuf_t *nvb = (nvbuf_t *)P2ROUNDUP(base, sizeof (uintptr_t));
if (base == 0 || (uintptr_t)&nvb[1] > lim)
return (EINVAL);
nvb->nvb_buf = (uintptr_t)&nvb[0];
nvb->nvb_cur = (uintptr_t)&nvb[1];
nvb->nvb_lim = lim;
nva->nva_arg = nvb;
return (0);
}
static void *
nv_fixed_alloc(nv_alloc_t *nva, size_t size)
{
nvbuf_t *nvb = nva->nva_arg;
uintptr_t new = nvb->nvb_cur;
if (size == 0 || new + size > nvb->nvb_lim)
return (NULL);
nvb->nvb_cur = P2ROUNDUP(new + size, sizeof (uintptr_t));
return ((void *)new);
}
static void
nv_fixed_free(nv_alloc_t *nva, void *buf, size_t size)
{
/* don't free memory in the pre-allocated buffer */
(void) nva, (void) buf, (void) size;
}
static void
nv_fixed_reset(nv_alloc_t *nva)
{
nvbuf_t *nvb = nva->nva_arg;
nvb->nvb_cur = (uintptr_t)&nvb[1];
}
-const nv_alloc_ops_t nv_fixed_ops_def = {
+static const nv_alloc_ops_t nv_fixed_ops_def = {
.nv_ao_init = nv_fixed_init,
.nv_ao_fini = NULL,
.nv_ao_alloc = nv_fixed_alloc,
.nv_ao_free = nv_fixed_free,
.nv_ao_reset = nv_fixed_reset
};
-const nv_alloc_ops_t *nv_fixed_ops = &nv_fixed_ops_def;
+const nv_alloc_ops_t *const nv_fixed_ops = &nv_fixed_ops_def;
#if defined(_KERNEL)
EXPORT_SYMBOL(nv_fixed_ops);
#endif
diff --git a/module/nvpair/nvpair_alloc_spl.c b/module/nvpair/nvpair_alloc_spl.c
index ed8fa4d09402..aa344b6423e8 100644
--- a/module/nvpair/nvpair_alloc_spl.c
+++ b/module/nvpair/nvpair_alloc_spl.c
@@ -1,96 +1,96 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License, Version 1.0 only
* (the "License"). You may not use this file except in compliance
* with the License.
*
* You can obtain a copy of the license at * usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2004 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#include <sys/nvpair.h>
#include <sys/kmem.h>
#include <sys/vmem.h>
static void *
nv_alloc_sleep_spl(nv_alloc_t *nva, size_t size)
{
return (vmem_alloc(size, KM_SLEEP));
}
static void *
nv_alloc_pushpage_spl(nv_alloc_t *nva, size_t size)
{
return (vmem_alloc(size, KM_PUSHPAGE));
}
static void *
nv_alloc_nosleep_spl(nv_alloc_t *nva, size_t size)
{
return (kmem_alloc(size, KM_NOSLEEP));
}
static void
nv_free_spl(nv_alloc_t *nva, void *buf, size_t size)
{
kmem_free(buf, size);
}
-const nv_alloc_ops_t spl_sleep_ops_def = {
+static const nv_alloc_ops_t spl_sleep_ops_def = {
.nv_ao_init = NULL,
.nv_ao_fini = NULL,
.nv_ao_alloc = nv_alloc_sleep_spl,
.nv_ao_free = nv_free_spl,
.nv_ao_reset = NULL
};
-const nv_alloc_ops_t spl_pushpage_ops_def = {
+static const nv_alloc_ops_t spl_pushpage_ops_def = {
.nv_ao_init = NULL,
.nv_ao_fini = NULL,
.nv_ao_alloc = nv_alloc_pushpage_spl,
.nv_ao_free = nv_free_spl,
.nv_ao_reset = NULL
};
-const nv_alloc_ops_t spl_nosleep_ops_def = {
+static const nv_alloc_ops_t spl_nosleep_ops_def = {
.nv_ao_init = NULL,
.nv_ao_fini = NULL,
.nv_ao_alloc = nv_alloc_nosleep_spl,
.nv_ao_free = nv_free_spl,
.nv_ao_reset = NULL
};
-nv_alloc_t nv_alloc_sleep_def = {
+static nv_alloc_t nv_alloc_sleep_def = {
&spl_sleep_ops_def,
NULL
};
-nv_alloc_t nv_alloc_pushpage_def = {
+static nv_alloc_t nv_alloc_pushpage_def = {
&spl_pushpage_ops_def,
NULL
};
-nv_alloc_t nv_alloc_nosleep_def = {
+static nv_alloc_t nv_alloc_nosleep_def = {
&spl_nosleep_ops_def,
NULL
};
-nv_alloc_t *nv_alloc_sleep = &nv_alloc_sleep_def;
-nv_alloc_t *nv_alloc_pushpage = &nv_alloc_pushpage_def;
-nv_alloc_t *nv_alloc_nosleep = &nv_alloc_nosleep_def;
+nv_alloc_t *const nv_alloc_sleep = &nv_alloc_sleep_def;
+nv_alloc_t *const nv_alloc_pushpage = &nv_alloc_pushpage_def;
+nv_alloc_t *const nv_alloc_nosleep = &nv_alloc_nosleep_def;
diff --git a/module/os/freebsd/zfs/abd_os.c b/module/os/freebsd/zfs/abd_os.c
index fa1034ff88bc..722a8898cde8 100644
--- a/module/os/freebsd/zfs/abd_os.c
+++ b/module/os/freebsd/zfs/abd_os.c
@@ -1,509 +1,509 @@
/*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*/
/*
* Copyright (c) 2014 by Chunwei Chen. All rights reserved.
* Copyright (c) 2016 by Delphix. All rights reserved.
*/
/*
* See abd.c for a general overview of the arc buffered data (ABD).
*
* Using a large proportion of scattered ABDs decreases ARC fragmentation since
* when we are at the limit of allocatable space, using equal-size chunks will
* allow us to quickly reclaim enough space for a new large allocation (assuming
* it is also scattered).
*
* ABDs are allocated scattered by default unless the caller uses
* abd_alloc_linear() or zfs_abd_scatter_enabled is disabled.
*/
#include <sys/abd_impl.h>
#include <sys/param.h>
#include <sys/types.h>
#include <sys/zio.h>
#include <sys/zfs_context.h>
#include <sys/zfs_znode.h>
typedef struct abd_stats {
kstat_named_t abdstat_struct_size;
kstat_named_t abdstat_scatter_cnt;
kstat_named_t abdstat_scatter_data_size;
kstat_named_t abdstat_scatter_chunk_waste;
kstat_named_t abdstat_linear_cnt;
kstat_named_t abdstat_linear_data_size;
} abd_stats_t;
static abd_stats_t abd_stats = {
/* Amount of memory occupied by all of the abd_t struct allocations */
{ "struct_size", KSTAT_DATA_UINT64 },
/*
* The number of scatter ABDs which are currently allocated, excluding
* ABDs which don't own their data (for instance the ones which were
* allocated through abd_get_offset()).
*/
{ "scatter_cnt", KSTAT_DATA_UINT64 },
/* Amount of data stored in all scatter ABDs tracked by scatter_cnt */
{ "scatter_data_size", KSTAT_DATA_UINT64 },
/*
* The amount of space wasted at the end of the last chunk across all
* scatter ABDs tracked by scatter_cnt.
*/
{ "scatter_chunk_waste", KSTAT_DATA_UINT64 },
/*
* The number of linear ABDs which are currently allocated, excluding
* ABDs which don't own their data (for instance the ones which were
* allocated through abd_get_offset() and abd_get_from_buf()). If an
* ABD takes ownership of its buf then it will become tracked.
*/
{ "linear_cnt", KSTAT_DATA_UINT64 },
/* Amount of data stored in all linear ABDs tracked by linear_cnt */
{ "linear_data_size", KSTAT_DATA_UINT64 },
};
struct {
wmsum_t abdstat_struct_size;
wmsum_t abdstat_scatter_cnt;
wmsum_t abdstat_scatter_data_size;
wmsum_t abdstat_scatter_chunk_waste;
wmsum_t abdstat_linear_cnt;
wmsum_t abdstat_linear_data_size;
} abd_sums;
/*
* zfs_abd_scatter_min_size is the minimum allocation size to use scatter
* ABD's for. Smaller allocations will use linear ABD's which use
* zio_[data_]buf_alloc().
*
* Scatter ABD's use at least one page each, so sub-page allocations waste
* some space when allocated as scatter (e.g. 2KB scatter allocation wastes
* half of each page). Using linear ABD's for small allocations means that
* they will be put on slabs which contain many allocations.
*
* Linear ABDs for multi-page allocations are easier to use, and in some cases
* it allows to avoid buffer copying. But allocation and especially free
* of multi-page linear ABDs are expensive operations due to KVA mapping and
* unmapping, and with time they cause KVA fragmentations.
*/
-size_t zfs_abd_scatter_min_size = PAGE_SIZE + 1;
+static size_t zfs_abd_scatter_min_size = PAGE_SIZE + 1;
#if defined(_KERNEL)
SYSCTL_DECL(_vfs_zfs);
SYSCTL_INT(_vfs_zfs, OID_AUTO, abd_scatter_enabled, CTLFLAG_RWTUN,
&zfs_abd_scatter_enabled, 0, "Enable scattered ARC data buffers");
SYSCTL_ULONG(_vfs_zfs, OID_AUTO, abd_scatter_min_size, CTLFLAG_RWTUN,
&zfs_abd_scatter_min_size, 0, "Minimum size of scatter allocations.");
#endif
kmem_cache_t *abd_chunk_cache;
static kstat_t *abd_ksp;
/*
* We use a scattered SPA_MAXBLOCKSIZE sized ABD whose chunks are
* just a single zero'd page-sized buffer. This allows us to conserve
* memory by only using a single zero buffer for the scatter chunks.
*/
abd_t *abd_zero_scatter = NULL;
static char *abd_zero_buf = NULL;
static uint_t
abd_chunkcnt_for_bytes(size_t size)
{
return ((size + PAGE_MASK) >> PAGE_SHIFT);
}
static inline uint_t
abd_scatter_chunkcnt(abd_t *abd)
{
ASSERT(!abd_is_linear(abd));
return (abd_chunkcnt_for_bytes(
ABD_SCATTER(abd).abd_offset + abd->abd_size));
}
boolean_t
abd_size_alloc_linear(size_t size)
{
return (!zfs_abd_scatter_enabled || size < zfs_abd_scatter_min_size);
}
void
abd_update_scatter_stats(abd_t *abd, abd_stats_op_t op)
{
uint_t n = abd_scatter_chunkcnt(abd);
ASSERT(op == ABDSTAT_INCR || op == ABDSTAT_DECR);
int waste = (n << PAGE_SHIFT) - abd->abd_size;
if (op == ABDSTAT_INCR) {
ABDSTAT_BUMP(abdstat_scatter_cnt);
ABDSTAT_INCR(abdstat_scatter_data_size, abd->abd_size);
ABDSTAT_INCR(abdstat_scatter_chunk_waste, waste);
arc_space_consume(waste, ARC_SPACE_ABD_CHUNK_WASTE);
} else {
ABDSTAT_BUMPDOWN(abdstat_scatter_cnt);
ABDSTAT_INCR(abdstat_scatter_data_size, -(int)abd->abd_size);
ABDSTAT_INCR(abdstat_scatter_chunk_waste, -waste);
arc_space_return(waste, ARC_SPACE_ABD_CHUNK_WASTE);
}
}
void
abd_update_linear_stats(abd_t *abd, abd_stats_op_t op)
{
ASSERT(op == ABDSTAT_INCR || op == ABDSTAT_DECR);
if (op == ABDSTAT_INCR) {
ABDSTAT_BUMP(abdstat_linear_cnt);
ABDSTAT_INCR(abdstat_linear_data_size, abd->abd_size);
} else {
ABDSTAT_BUMPDOWN(abdstat_linear_cnt);
ABDSTAT_INCR(abdstat_linear_data_size, -(int)abd->abd_size);
}
}
void
abd_verify_scatter(abd_t *abd)
{
uint_t i, n;
/*
* There is no scatter linear pages in FreeBSD so there is
* an error if the ABD has been marked as a linear page.
*/
ASSERT(!abd_is_linear_page(abd));
ASSERT3U(ABD_SCATTER(abd).abd_offset, <, PAGE_SIZE);
n = abd_scatter_chunkcnt(abd);
for (i = 0; i < n; i++) {
ASSERT3P(ABD_SCATTER(abd).abd_chunks[i], !=, NULL);
}
}
void
abd_alloc_chunks(abd_t *abd, size_t size)
{
uint_t i, n;
n = abd_chunkcnt_for_bytes(size);
for (i = 0; i < n; i++) {
ABD_SCATTER(abd).abd_chunks[i] =
kmem_cache_alloc(abd_chunk_cache, KM_PUSHPAGE);
}
}
void
abd_free_chunks(abd_t *abd)
{
uint_t i, n;
n = abd_scatter_chunkcnt(abd);
for (i = 0; i < n; i++) {
kmem_cache_free(abd_chunk_cache,
ABD_SCATTER(abd).abd_chunks[i]);
}
}
abd_t *
abd_alloc_struct_impl(size_t size)
{
uint_t chunkcnt = abd_chunkcnt_for_bytes(size);
/*
* In the event we are allocating a gang ABD, the size passed in
* will be 0. We must make sure to set abd_size to the size of an
* ABD struct as opposed to an ABD scatter with 0 chunks. The gang
* ABD struct allocation accounts for an additional 24 bytes over
* a scatter ABD with 0 chunks.
*/
size_t abd_size = MAX(sizeof (abd_t),
offsetof(abd_t, abd_u.abd_scatter.abd_chunks[chunkcnt]));
abd_t *abd = kmem_alloc(abd_size, KM_PUSHPAGE);
ASSERT3P(abd, !=, NULL);
ABDSTAT_INCR(abdstat_struct_size, abd_size);
return (abd);
}
void
abd_free_struct_impl(abd_t *abd)
{
uint_t chunkcnt = abd_is_linear(abd) || abd_is_gang(abd) ? 0 :
abd_scatter_chunkcnt(abd);
ssize_t size = MAX(sizeof (abd_t),
offsetof(abd_t, abd_u.abd_scatter.abd_chunks[chunkcnt]));
kmem_free(abd, size);
ABDSTAT_INCR(abdstat_struct_size, -size);
}
/*
* Allocate scatter ABD of size SPA_MAXBLOCKSIZE, where
* each chunk in the scatterlist will be set to abd_zero_buf.
*/
static void
abd_alloc_zero_scatter(void)
{
uint_t i, n;
n = abd_chunkcnt_for_bytes(SPA_MAXBLOCKSIZE);
abd_zero_buf = kmem_cache_alloc(abd_chunk_cache, KM_PUSHPAGE);
bzero(abd_zero_buf, PAGE_SIZE);
abd_zero_scatter = abd_alloc_struct(SPA_MAXBLOCKSIZE);
abd_zero_scatter->abd_flags |= ABD_FLAG_OWNER | ABD_FLAG_ZEROS;
abd_zero_scatter->abd_size = SPA_MAXBLOCKSIZE;
ABD_SCATTER(abd_zero_scatter).abd_offset = 0;
for (i = 0; i < n; i++) {
ABD_SCATTER(abd_zero_scatter).abd_chunks[i] =
abd_zero_buf;
}
ABDSTAT_BUMP(abdstat_scatter_cnt);
ABDSTAT_INCR(abdstat_scatter_data_size, PAGE_SIZE);
}
static void
abd_free_zero_scatter(void)
{
ABDSTAT_BUMPDOWN(abdstat_scatter_cnt);
ABDSTAT_INCR(abdstat_scatter_data_size, -(int)PAGE_SIZE);
abd_free_struct(abd_zero_scatter);
abd_zero_scatter = NULL;
kmem_cache_free(abd_chunk_cache, abd_zero_buf);
}
static int
abd_kstats_update(kstat_t *ksp, int rw)
{
abd_stats_t *as = ksp->ks_data;
if (rw == KSTAT_WRITE)
return (EACCES);
as->abdstat_struct_size.value.ui64 =
wmsum_value(&abd_sums.abdstat_struct_size);
as->abdstat_scatter_cnt.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_cnt);
as->abdstat_scatter_data_size.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_data_size);
as->abdstat_scatter_chunk_waste.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_chunk_waste);
as->abdstat_linear_cnt.value.ui64 =
wmsum_value(&abd_sums.abdstat_linear_cnt);
as->abdstat_linear_data_size.value.ui64 =
wmsum_value(&abd_sums.abdstat_linear_data_size);
return (0);
}
void
abd_init(void)
{
abd_chunk_cache = kmem_cache_create("abd_chunk", PAGE_SIZE, 0,
NULL, NULL, NULL, NULL, 0, KMC_NODEBUG);
wmsum_init(&abd_sums.abdstat_struct_size, 0);
wmsum_init(&abd_sums.abdstat_scatter_cnt, 0);
wmsum_init(&abd_sums.abdstat_scatter_data_size, 0);
wmsum_init(&abd_sums.abdstat_scatter_chunk_waste, 0);
wmsum_init(&abd_sums.abdstat_linear_cnt, 0);
wmsum_init(&abd_sums.abdstat_linear_data_size, 0);
abd_ksp = kstat_create("zfs", 0, "abdstats", "misc", KSTAT_TYPE_NAMED,
sizeof (abd_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
if (abd_ksp != NULL) {
abd_ksp->ks_data = &abd_stats;
abd_ksp->ks_update = abd_kstats_update;
kstat_install(abd_ksp);
}
abd_alloc_zero_scatter();
}
void
abd_fini(void)
{
abd_free_zero_scatter();
if (abd_ksp != NULL) {
kstat_delete(abd_ksp);
abd_ksp = NULL;
}
wmsum_fini(&abd_sums.abdstat_struct_size);
wmsum_fini(&abd_sums.abdstat_scatter_cnt);
wmsum_fini(&abd_sums.abdstat_scatter_data_size);
wmsum_fini(&abd_sums.abdstat_scatter_chunk_waste);
wmsum_fini(&abd_sums.abdstat_linear_cnt);
wmsum_fini(&abd_sums.abdstat_linear_data_size);
kmem_cache_destroy(abd_chunk_cache);
abd_chunk_cache = NULL;
}
void
abd_free_linear_page(abd_t *abd)
{
/*
* FreeBSD does not have scatter linear pages
* so there is an error.
*/
VERIFY(0);
}
/*
* If we're going to use this ABD for doing I/O using the block layer, the
* consumer of the ABD data doesn't care if it's scattered or not, and we don't
* plan to store this ABD in memory for a long period of time, we should
* allocate the ABD type that requires the least data copying to do the I/O.
*
* Currently this is linear ABDs, however if ldi_strategy() can ever issue I/Os
* using a scatter/gather list we should switch to that and replace this call
* with vanilla abd_alloc().
*/
abd_t *
abd_alloc_for_io(size_t size, boolean_t is_metadata)
{
return (abd_alloc_linear(size, is_metadata));
}
abd_t *
abd_get_offset_scatter(abd_t *abd, abd_t *sabd, size_t off,
size_t size)
{
abd_verify(sabd);
ASSERT3U(off, <=, sabd->abd_size);
size_t new_offset = ABD_SCATTER(sabd).abd_offset + off;
size_t chunkcnt = abd_chunkcnt_for_bytes(
(new_offset & PAGE_MASK) + size);
ASSERT3U(chunkcnt, <=, abd_scatter_chunkcnt(sabd));
/*
* If an abd struct is provided, it is only the minimum size. If we
* need additional chunks, we need to allocate a new struct.
*/
if (abd != NULL &&
offsetof(abd_t, abd_u.abd_scatter.abd_chunks[chunkcnt]) >
sizeof (abd_t)) {
abd = NULL;
}
if (abd == NULL)
abd = abd_alloc_struct(chunkcnt << PAGE_SHIFT);
/*
* Even if this buf is filesystem metadata, we only track that
* if we own the underlying data buffer, which is not true in
* this case. Therefore, we don't ever use ABD_FLAG_META here.
*/
ABD_SCATTER(abd).abd_offset = new_offset & PAGE_MASK;
/* Copy the scatterlist starting at the correct offset */
(void) memcpy(&ABD_SCATTER(abd).abd_chunks,
&ABD_SCATTER(sabd).abd_chunks[new_offset >> PAGE_SHIFT],
chunkcnt * sizeof (void *));
return (abd);
}
/*
* Initialize the abd_iter.
*/
void
abd_iter_init(struct abd_iter *aiter, abd_t *abd)
{
ASSERT(!abd_is_gang(abd));
abd_verify(abd);
aiter->iter_abd = abd;
aiter->iter_pos = 0;
aiter->iter_mapaddr = NULL;
aiter->iter_mapsize = 0;
}
/*
* This is just a helper function to see if we have exhausted the
* abd_iter and reached the end.
*/
boolean_t
abd_iter_at_end(struct abd_iter *aiter)
{
return (aiter->iter_pos == aiter->iter_abd->abd_size);
}
/*
* Advance the iterator by a certain amount. Cannot be called when a chunk is
* in use. This can be safely called when the aiter has already exhausted, in
* which case this does nothing.
*/
void
abd_iter_advance(struct abd_iter *aiter, size_t amount)
{
ASSERT3P(aiter->iter_mapaddr, ==, NULL);
ASSERT0(aiter->iter_mapsize);
/* There's nothing left to advance to, so do nothing */
if (abd_iter_at_end(aiter))
return;
aiter->iter_pos += amount;
}
/*
* Map the current chunk into aiter. This can be safely called when the aiter
* has already exhausted, in which case this does nothing.
*/
void
abd_iter_map(struct abd_iter *aiter)
{
void *paddr;
ASSERT3P(aiter->iter_mapaddr, ==, NULL);
ASSERT0(aiter->iter_mapsize);
/* There's nothing left to iterate over, so do nothing */
if (abd_iter_at_end(aiter))
return;
abd_t *abd = aiter->iter_abd;
size_t offset = aiter->iter_pos;
if (abd_is_linear(abd)) {
aiter->iter_mapsize = abd->abd_size - offset;
paddr = ABD_LINEAR_BUF(abd);
} else {
offset += ABD_SCATTER(abd).abd_offset;
paddr = ABD_SCATTER(abd).abd_chunks[offset >> PAGE_SHIFT];
offset &= PAGE_MASK;
aiter->iter_mapsize = MIN(PAGE_SIZE - offset,
abd->abd_size - aiter->iter_pos);
}
aiter->iter_mapaddr = (char *)paddr + offset;
}
/*
* Unmap the current chunk from aiter. This can be safely called when the aiter
* has already exhausted, in which case this does nothing.
*/
void
abd_iter_unmap(struct abd_iter *aiter)
{
if (!abd_iter_at_end(aiter)) {
ASSERT3P(aiter->iter_mapaddr, !=, NULL);
ASSERT3U(aiter->iter_mapsize, >, 0);
}
aiter->iter_mapaddr = NULL;
aiter->iter_mapsize = 0;
}
void
abd_cache_reap_now(void)
{
kmem_cache_reap_soon(abd_chunk_cache);
}
diff --git a/module/os/freebsd/zfs/crypto_os.c b/module/os/freebsd/zfs/crypto_os.c
index 6a67dbc9f616..f971b62bd124 100644
--- a/module/os/freebsd/zfs/crypto_os.c
+++ b/module/os/freebsd/zfs/crypto_os.c
@@ -1,624 +1,624 @@
/*
* Copyright (c) 2005-2010 Pawel Jakub Dawidek <pjd@FreeBSD.org>
* Copyright (c) 2018 Sean Eric Fagan <sef@ixsystems.com>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* Portions of this file are derived from sys/geom/eli/g_eli_hmac.c
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/types.h>
#include <sys/errno.h>
#ifdef _KERNEL
#include <sys/libkern.h>
#include <sys/malloc.h>
#include <sys/sysctl.h>
#include <opencrypto/cryptodev.h>
#include <opencrypto/xform.h>
#else
#include <strings.h>
#endif
#include <sys/zio_crypt.h>
#include <sys/fs/zfs.h>
#include <sys/zio.h>
#include <sys/freebsd_crypto.h>
#define SHA512_HMAC_BLOCK_SIZE 128
static int crypt_sessions = 0;
SYSCTL_DECL(_vfs_zfs);
SYSCTL_INT(_vfs_zfs, OID_AUTO, crypt_sessions, CTLFLAG_RD,
&crypt_sessions, 0, "Number of cryptographic sessions created");
void
crypto_mac_init(struct hmac_ctx *ctx, const crypto_key_t *c_key)
{
uint8_t k_ipad[SHA512_HMAC_BLOCK_SIZE],
k_opad[SHA512_HMAC_BLOCK_SIZE],
key[SHA512_HMAC_BLOCK_SIZE];
SHA512_CTX lctx;
int i;
size_t cl_bytes = CRYPTO_BITS2BYTES(c_key->ck_length);
/*
* This code is based on the similar code in geom/eli/g_eli_hmac.c
*/
explicit_bzero(key, sizeof (key));
if (c_key->ck_length == 0)
/* do nothing */;
else if (cl_bytes <= SHA512_HMAC_BLOCK_SIZE)
bcopy(c_key->ck_data, key, cl_bytes);
else {
/*
* If key is longer than 128 bytes reset it to
* key = SHA512(key).
*/
SHA512_Init(&lctx);
SHA512_Update(&lctx, c_key->ck_data, cl_bytes);
SHA512_Final(key, &lctx);
}
/* XOR key with ipad and opad values. */
for (i = 0; i < sizeof (key); i++) {
k_ipad[i] = key[i] ^ 0x36;
k_opad[i] = key[i] ^ 0x5c;
}
explicit_bzero(key, sizeof (key));
/* Start inner SHA512. */
SHA512_Init(&ctx->innerctx);
SHA512_Update(&ctx->innerctx, k_ipad, sizeof (k_ipad));
explicit_bzero(k_ipad, sizeof (k_ipad));
/* Start outer SHA512. */
SHA512_Init(&ctx->outerctx);
SHA512_Update(&ctx->outerctx, k_opad, sizeof (k_opad));
explicit_bzero(k_opad, sizeof (k_opad));
}
void
crypto_mac_update(struct hmac_ctx *ctx, const void *data, size_t datasize)
{
SHA512_Update(&ctx->innerctx, data, datasize);
}
void
crypto_mac_final(struct hmac_ctx *ctx, void *md, size_t mdsize)
{
uint8_t digest[SHA512_DIGEST_LENGTH];
/* Complete inner hash */
SHA512_Final(digest, &ctx->innerctx);
/* Complete outer hash */
SHA512_Update(&ctx->outerctx, digest, sizeof (digest));
SHA512_Final(digest, &ctx->outerctx);
explicit_bzero(ctx, sizeof (*ctx));
/* mdsize == 0 means "Give me the whole hash!" */
if (mdsize == 0)
mdsize = SHA512_DIGEST_LENGTH;
bcopy(digest, md, mdsize);
explicit_bzero(digest, sizeof (digest));
}
void
crypto_mac(const crypto_key_t *key, const void *in_data, size_t in_data_size,
void *out_data, size_t out_data_size)
{
struct hmac_ctx ctx;
crypto_mac_init(&ctx, key);
crypto_mac_update(&ctx, in_data, in_data_size);
crypto_mac_final(&ctx, out_data, out_data_size);
}
static int
freebsd_zfs_crypt_done(struct cryptop *crp)
{
freebsd_crypt_session_t *ses;
ses = crp->crp_opaque;
mtx_lock(&ses->fs_lock);
ses->fs_done = true;
mtx_unlock(&ses->fs_lock);
wakeup(crp);
return (0);
}
void
freebsd_crypt_freesession(freebsd_crypt_session_t *sess)
{
mtx_destroy(&sess->fs_lock);
crypto_freesession(sess->fs_sid);
explicit_bzero(sess, sizeof (*sess));
}
static int
zfs_crypto_dispatch(freebsd_crypt_session_t *session, struct cryptop *crp)
{
int error;
crp->crp_opaque = session;
crp->crp_callback = freebsd_zfs_crypt_done;
for (;;) {
error = crypto_dispatch(crp);
if (error)
break;
mtx_lock(&session->fs_lock);
while (session->fs_done == false)
msleep(crp, &session->fs_lock, 0,
"zfs_crypto", 0);
mtx_unlock(&session->fs_lock);
if (crp->crp_etype == ENOMEM) {
pause("zcrnomem", 1);
} else if (crp->crp_etype != EAGAIN) {
error = crp->crp_etype;
break;
}
crp->crp_etype = 0;
crp->crp_flags &= ~CRYPTO_F_DONE;
session->fs_done = false;
#if __FreeBSD_version < 1300087
/*
* Session ID changed, so we should record that,
* and try again
*/
session->fs_sid = crp->crp_session;
#endif
}
return (error);
}
static void
freebsd_crypt_uio_debug_log(boolean_t encrypt,
freebsd_crypt_session_t *input_sessionp,
- struct zio_crypt_info *c_info,
+ const struct zio_crypt_info *c_info,
zfs_uio_t *data_uio,
crypto_key_t *key,
uint8_t *ivbuf,
size_t datalen,
size_t auth_len)
{
#ifdef FCRYPTO_DEBUG
struct cryptodesc *crd;
uint8_t *p = NULL;
size_t total = 0;
printf("%s(%s, %p, { %s, %d, %d, %s }, %p, { %d, %p, %u }, "
"%p, %u, %u)\n",
__FUNCTION__, encrypt ? "encrypt" : "decrypt", input_sessionp,
c_info->ci_algname, c_info->ci_crypt_type,
(unsigned int)c_info->ci_keylen, c_info->ci_name,
data_uio, key->ck_format, key->ck_data,
(unsigned int)key->ck_length,
ivbuf, (unsigned int)datalen, (unsigned int)auth_len);
printf("\tkey = { ");
for (int i = 0; i < key->ck_length / 8; i++) {
uint8_t *b = (uint8_t *)key->ck_data;
printf("%02x ", b[i]);
}
printf("}\n");
for (int i = 0; i < zfs_uio_iovcnt(data_uio); i++) {
printf("\tiovec #%d: <%p, %u>\n", i,
zfs_uio_iovbase(data_uio, i),
(unsigned int)zfs_uio_iovlen(data_uio, i));
total += zfs_uio_iovlen(data_uio, i);
}
zfs_uio_resid(data_uio) = total;
#endif
}
/*
* Create a new cryptographic session. This should
* happen every time the key changes (including when
* it's first loaded).
*/
#if __FreeBSD_version >= 1300087
int
freebsd_crypt_newsession(freebsd_crypt_session_t *sessp,
- struct zio_crypt_info *c_info, crypto_key_t *key)
+ const struct zio_crypt_info *c_info, crypto_key_t *key)
{
struct crypto_session_params csp;
int error = 0;
#ifdef FCRYPTO_DEBUG
printf("%s(%p, { %s, %d, %d, %s }, { %d, %p, %u })\n",
__FUNCTION__, sessp,
c_info->ci_algname, c_info->ci_crypt_type,
(unsigned int)c_info->ci_keylen, c_info->ci_name,
key->ck_format, key->ck_data, (unsigned int)key->ck_length);
printf("\tkey = { ");
for (int i = 0; i < key->ck_length / 8; i++) {
uint8_t *b = (uint8_t *)key->ck_data;
printf("%02x ", b[i]);
}
printf("}\n");
#endif
bzero(&csp, sizeof (csp));
csp.csp_mode = CSP_MODE_AEAD;
csp.csp_cipher_key = key->ck_data;
csp.csp_cipher_klen = key->ck_length / 8;
switch (c_info->ci_crypt_type) {
case ZC_TYPE_GCM:
csp.csp_cipher_alg = CRYPTO_AES_NIST_GCM_16;
csp.csp_ivlen = AES_GCM_IV_LEN;
switch (key->ck_length/8) {
case AES_128_GMAC_KEY_LEN:
case AES_192_GMAC_KEY_LEN:
case AES_256_GMAC_KEY_LEN:
break;
default:
error = EINVAL;
goto bad;
}
break;
case ZC_TYPE_CCM:
csp.csp_cipher_alg = CRYPTO_AES_CCM_16;
csp.csp_ivlen = AES_CCM_IV_LEN;
switch (key->ck_length/8) {
case AES_128_CBC_MAC_KEY_LEN:
case AES_192_CBC_MAC_KEY_LEN:
case AES_256_CBC_MAC_KEY_LEN:
break;
default:
error = EINVAL;
goto bad;
break;
}
break;
default:
error = ENOTSUP;
goto bad;
}
/*
* Disable the use of hardware drivers on FreeBSD 13 and later since
* common crypto offload drivers impose constraints on AES-GCM AAD
* lengths that make them unusable for ZFS, and we currently do not have
* a mechanism to fall back to a software driver for requests not
* handled by a hardware driver.
*
* On 12 we continue to permit the use of hardware drivers since
* CPU-accelerated drivers such as aesni(4) register themselves as
* hardware drivers.
*/
error = crypto_newsession(&sessp->fs_sid, &csp, CRYPTOCAP_F_SOFTWARE);
mtx_init(&sessp->fs_lock, "FreeBSD Cryptographic Session Lock",
NULL, MTX_DEF);
crypt_sessions++;
bad:
#ifdef FCRYPTO_DEBUG
if (error)
printf("%s: returning error %d\n", __FUNCTION__, error);
#endif
return (error);
}
int
freebsd_crypt_uio(boolean_t encrypt,
freebsd_crypt_session_t *input_sessionp,
- struct zio_crypt_info *c_info,
+ const struct zio_crypt_info *c_info,
zfs_uio_t *data_uio,
crypto_key_t *key,
uint8_t *ivbuf,
size_t datalen,
size_t auth_len)
{
struct cryptop *crp;
freebsd_crypt_session_t *session = NULL;
int error = 0;
size_t total = 0;
freebsd_crypt_uio_debug_log(encrypt, input_sessionp, c_info, data_uio,
key, ivbuf, datalen, auth_len);
for (int i = 0; i < zfs_uio_iovcnt(data_uio); i++)
total += zfs_uio_iovlen(data_uio, i);
zfs_uio_resid(data_uio) = total;
if (input_sessionp == NULL) {
session = kmem_zalloc(sizeof (*session), KM_SLEEP);
error = freebsd_crypt_newsession(session, c_info, key);
if (error)
goto out;
} else
session = input_sessionp;
crp = crypto_getreq(session->fs_sid, M_WAITOK);
if (encrypt) {
crp->crp_op = CRYPTO_OP_ENCRYPT |
CRYPTO_OP_COMPUTE_DIGEST;
} else {
crp->crp_op = CRYPTO_OP_DECRYPT |
CRYPTO_OP_VERIFY_DIGEST;
}
crp->crp_flags = CRYPTO_F_CBIFSYNC | CRYPTO_F_IV_SEPARATE;
crypto_use_uio(crp, GET_UIO_STRUCT(data_uio));
crp->crp_aad_start = 0;
crp->crp_aad_length = auth_len;
crp->crp_payload_start = auth_len;
crp->crp_payload_length = datalen;
crp->crp_digest_start = auth_len + datalen;
bcopy(ivbuf, crp->crp_iv, ZIO_DATA_IV_LEN);
error = zfs_crypto_dispatch(session, crp);
crypto_freereq(crp);
out:
#ifdef FCRYPTO_DEBUG
if (error)
printf("%s: returning error %d\n", __FUNCTION__, error);
#endif
if (input_sessionp == NULL) {
freebsd_crypt_freesession(session);
kmem_free(session, sizeof (*session));
}
return (error);
}
#else
int
freebsd_crypt_newsession(freebsd_crypt_session_t *sessp,
- struct zio_crypt_info *c_info, crypto_key_t *key)
+ const struct zio_crypt_info *c_info, crypto_key_t *key)
{
struct cryptoini cria, crie, *crip;
struct enc_xform *xform;
struct auth_hash *xauth;
int error = 0;
crypto_session_t sid;
#ifdef FCRYPTO_DEBUG
printf("%s(%p, { %s, %d, %d, %s }, { %d, %p, %u })\n",
__FUNCTION__, sessp,
c_info->ci_algname, c_info->ci_crypt_type,
(unsigned int)c_info->ci_keylen, c_info->ci_name,
key->ck_format, key->ck_data, (unsigned int)key->ck_length);
printf("\tkey = { ");
for (int i = 0; i < key->ck_length / 8; i++) {
uint8_t *b = (uint8_t *)key->ck_data;
printf("%02x ", b[i]);
}
printf("}\n");
#endif
switch (c_info->ci_crypt_type) {
case ZC_TYPE_GCM:
xform = &enc_xform_aes_nist_gcm;
switch (key->ck_length/8) {
case AES_128_GMAC_KEY_LEN:
xauth = &auth_hash_nist_gmac_aes_128;
break;
case AES_192_GMAC_KEY_LEN:
xauth = &auth_hash_nist_gmac_aes_192;
break;
case AES_256_GMAC_KEY_LEN:
xauth = &auth_hash_nist_gmac_aes_256;
break;
default:
error = EINVAL;
goto bad;
}
break;
case ZC_TYPE_CCM:
xform = &enc_xform_ccm;
switch (key->ck_length/8) {
case AES_128_CBC_MAC_KEY_LEN:
xauth = &auth_hash_ccm_cbc_mac_128;
break;
case AES_192_CBC_MAC_KEY_LEN:
xauth = &auth_hash_ccm_cbc_mac_192;
break;
case AES_256_CBC_MAC_KEY_LEN:
xauth = &auth_hash_ccm_cbc_mac_256;
break;
default:
error = EINVAL;
goto bad;
break;
}
break;
default:
error = ENOTSUP;
goto bad;
}
#ifdef FCRYPTO_DEBUG
printf("%s(%d): Using crypt %s (key length %u [%u bytes]), "
"auth %s (key length %d)\n",
__FUNCTION__, __LINE__,
xform->name, (unsigned int)key->ck_length,
(unsigned int)key->ck_length/8,
xauth->name, xauth->keysize);
#endif
bzero(&crie, sizeof (crie));
bzero(&cria, sizeof (cria));
crie.cri_alg = xform->type;
crie.cri_key = key->ck_data;
crie.cri_klen = key->ck_length;
cria.cri_alg = xauth->type;
cria.cri_key = key->ck_data;
cria.cri_klen = key->ck_length;
cria.cri_next = &crie;
crie.cri_next = NULL;
crip = &cria;
// Everything else is bzero'd
error = crypto_newsession(&sid, crip,
CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE);
if (error != 0) {
printf("%s(%d): crypto_newsession failed with %d\n",
__FUNCTION__, __LINE__, error);
goto bad;
}
sessp->fs_sid = sid;
mtx_init(&sessp->fs_lock, "FreeBSD Cryptographic Session Lock",
NULL, MTX_DEF);
crypt_sessions++;
bad:
return (error);
}
/*
* The meat of encryption/decryption.
* If sessp is NULL, then it will create a
* temporary cryptographic session, and release
* it when done.
*/
int
freebsd_crypt_uio(boolean_t encrypt,
freebsd_crypt_session_t *input_sessionp,
- struct zio_crypt_info *c_info,
+ const struct zio_crypt_info *c_info,
zfs_uio_t *data_uio,
crypto_key_t *key,
uint8_t *ivbuf,
size_t datalen,
size_t auth_len)
{
struct cryptop *crp;
struct cryptodesc *enc_desc, *auth_desc;
struct enc_xform *xform;
struct auth_hash *xauth;
freebsd_crypt_session_t *session = NULL;
int error;
freebsd_crypt_uio_debug_log(encrypt, input_sessionp, c_info, data_uio,
key, ivbuf, datalen, auth_len);
switch (c_info->ci_crypt_type) {
case ZC_TYPE_GCM:
xform = &enc_xform_aes_nist_gcm;
switch (key->ck_length/8) {
case AES_128_GMAC_KEY_LEN:
xauth = &auth_hash_nist_gmac_aes_128;
break;
case AES_192_GMAC_KEY_LEN:
xauth = &auth_hash_nist_gmac_aes_192;
break;
case AES_256_GMAC_KEY_LEN:
xauth = &auth_hash_nist_gmac_aes_256;
break;
default:
error = EINVAL;
goto bad;
}
break;
case ZC_TYPE_CCM:
xform = &enc_xform_ccm;
switch (key->ck_length/8) {
case AES_128_CBC_MAC_KEY_LEN:
xauth = &auth_hash_ccm_cbc_mac_128;
break;
case AES_192_CBC_MAC_KEY_LEN:
xauth = &auth_hash_ccm_cbc_mac_192;
break;
case AES_256_CBC_MAC_KEY_LEN:
xauth = &auth_hash_ccm_cbc_mac_256;
break;
default:
error = EINVAL;
goto bad;
break;
}
break;
default:
error = ENOTSUP;
goto bad;
}
#ifdef FCRYPTO_DEBUG
printf("%s(%d): Using crypt %s (key length %u [%u bytes]), "
"auth %s (key length %d)\n",
__FUNCTION__, __LINE__,
xform->name, (unsigned int)key->ck_length,
(unsigned int)key->ck_length/8,
xauth->name, xauth->keysize);
#endif
if (input_sessionp == NULL) {
session = kmem_zalloc(sizeof (*session), KM_SLEEP);
error = freebsd_crypt_newsession(session, c_info, key);
if (error)
goto out;
} else
session = input_sessionp;
crp = crypto_getreq(2);
if (crp == NULL) {
error = ENOMEM;
goto bad;
}
auth_desc = crp->crp_desc;
enc_desc = auth_desc->crd_next;
crp->crp_session = session->fs_sid;
crp->crp_ilen = auth_len + datalen;
crp->crp_buf = (void*)GET_UIO_STRUCT(data_uio);
crp->crp_flags = CRYPTO_F_IOV | CRYPTO_F_CBIFSYNC;
auth_desc->crd_skip = 0;
auth_desc->crd_len = auth_len;
auth_desc->crd_inject = auth_len + datalen;
auth_desc->crd_alg = xauth->type;
#ifdef FCRYPTO_DEBUG
printf("%s: auth: skip = %u, len = %u, inject = %u\n",
__FUNCTION__, auth_desc->crd_skip, auth_desc->crd_len,
auth_desc->crd_inject);
#endif
enc_desc->crd_skip = auth_len;
enc_desc->crd_len = datalen;
enc_desc->crd_inject = auth_len;
enc_desc->crd_alg = xform->type;
enc_desc->crd_flags = CRD_F_IV_EXPLICIT | CRD_F_IV_PRESENT;
bcopy(ivbuf, enc_desc->crd_iv, ZIO_DATA_IV_LEN);
enc_desc->crd_next = NULL;
#ifdef FCRYPTO_DEBUG
printf("%s: enc: skip = %u, len = %u, inject = %u\n",
__FUNCTION__, enc_desc->crd_skip, enc_desc->crd_len,
enc_desc->crd_inject);
#endif
if (encrypt)
enc_desc->crd_flags |= CRD_F_ENCRYPT;
error = zfs_crypto_dispatch(session, crp);
crypto_freereq(crp);
out:
if (input_sessionp == NULL) {
freebsd_crypt_freesession(session);
kmem_free(session, sizeof (*session));
}
bad:
#ifdef FCRYPTO_DEBUG
if (error)
printf("%s: returning error %d\n", __FUNCTION__, error);
#endif
return (error);
}
#endif
diff --git a/module/os/freebsd/zfs/vdev_file.c b/module/os/freebsd/zfs/vdev_file.c
index 2d926813637e..ef87d6610977 100644
--- a/module/os/freebsd/zfs/vdev_file.c
+++ b/module/os/freebsd/zfs/vdev_file.c
@@ -1,356 +1,356 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/file.h>
#include <sys/vdev_file.h>
#include <sys/vdev_impl.h>
#include <sys/zio.h>
#include <sys/fs/zfs.h>
#include <sys/fm/fs/zfs.h>
#include <sys/abd.h>
#include <sys/stat.h>
/*
* Virtual device vector for files.
*/
static taskq_t *vdev_file_taskq;
-unsigned long vdev_file_logical_ashift = SPA_MINBLOCKSHIFT;
-unsigned long vdev_file_physical_ashift = SPA_MINBLOCKSHIFT;
+static unsigned long vdev_file_logical_ashift = SPA_MINBLOCKSHIFT;
+static unsigned long vdev_file_physical_ashift = SPA_MINBLOCKSHIFT;
void
vdev_file_init(void)
{
vdev_file_taskq = taskq_create("z_vdev_file", MAX(max_ncpus, 16),
minclsyspri, max_ncpus, INT_MAX, 0);
}
void
vdev_file_fini(void)
{
taskq_destroy(vdev_file_taskq);
}
static void
vdev_file_hold(vdev_t *vd)
{
ASSERT3P(vd->vdev_path, !=, NULL);
}
static void
vdev_file_rele(vdev_t *vd)
{
ASSERT3P(vd->vdev_path, !=, NULL);
}
static mode_t
vdev_file_open_mode(spa_mode_t spa_mode)
{
mode_t mode = 0;
if ((spa_mode & SPA_MODE_READ) && (spa_mode & SPA_MODE_WRITE)) {
mode = O_RDWR;
} else if (spa_mode & SPA_MODE_READ) {
mode = O_RDONLY;
} else if (spa_mode & SPA_MODE_WRITE) {
mode = O_WRONLY;
}
return (mode | O_LARGEFILE);
}
static int
vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
uint64_t *logical_ashift, uint64_t *physical_ashift)
{
vdev_file_t *vf;
zfs_file_t *fp;
zfs_file_attr_t zfa;
int error;
/*
* Rotational optimizations only make sense on block devices.
*/
vd->vdev_nonrot = B_TRUE;
/*
* Allow TRIM on file based vdevs. This may not always be supported,
* since it depends on your kernel version and underlying filesystem
* type but it is always safe to attempt.
*/
vd->vdev_has_trim = B_TRUE;
/*
* Disable secure TRIM on file based vdevs. There is no way to
* request this behavior from the underlying filesystem.
*/
vd->vdev_has_securetrim = B_FALSE;
/*
* We must have a pathname, and it must be absolute.
*/
if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
/*
* Reopen the device if it's not currently open. Otherwise,
* just update the physical size of the device.
*/
if (vd->vdev_tsd != NULL) {
ASSERT(vd->vdev_reopening);
vf = vd->vdev_tsd;
goto skip_open;
}
vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
/*
* We always open the files from the root of the global zone, even if
* we're in a local zone. If the user has gotten to this point, the
* administrator has already decided that the pool should be available
* to local zone users, so the underlying devices should be as well.
*/
ASSERT3P(vd->vdev_path, !=, NULL);
ASSERT(vd->vdev_path[0] == '/');
error = zfs_file_open(vd->vdev_path,
vdev_file_open_mode(spa_mode(vd->vdev_spa)), 0, &fp);
if (error) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
return (error);
}
vf->vf_file = fp;
#ifdef _KERNEL
/*
* Make sure it's a regular file.
*/
if (zfs_file_getattr(fp, &zfa)) {
return (SET_ERROR(ENODEV));
}
if (!S_ISREG(zfa.zfa_mode)) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
return (SET_ERROR(ENODEV));
}
#endif
skip_open:
error = zfs_file_getattr(vf->vf_file, &zfa);
if (error) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
return (error);
}
*max_psize = *psize = zfa.zfa_size;
*logical_ashift = vdev_file_logical_ashift;
*physical_ashift = vdev_file_physical_ashift;
return (0);
}
static void
vdev_file_close(vdev_t *vd)
{
vdev_file_t *vf = vd->vdev_tsd;
if (vd->vdev_reopening || vf == NULL)
return;
if (vf->vf_file != NULL) {
zfs_file_close(vf->vf_file);
}
vd->vdev_delayed_close = B_FALSE;
kmem_free(vf, sizeof (vdev_file_t));
vd->vdev_tsd = NULL;
}
/*
* Implements the interrupt side for file vdev types. This routine will be
* called when the I/O completes allowing us to transfer the I/O to the
* interrupt taskqs. For consistency, the code structure mimics disk vdev
* types.
*/
static void
vdev_file_io_intr(zio_t *zio)
{
zio_delay_interrupt(zio);
}
static void
vdev_file_io_strategy(void *arg)
{
zio_t *zio = arg;
vdev_t *vd = zio->io_vd;
vdev_file_t *vf;
void *buf;
ssize_t resid;
loff_t off;
ssize_t size;
int err;
off = zio->io_offset;
size = zio->io_size;
resid = 0;
vf = vd->vdev_tsd;
ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
if (zio->io_type == ZIO_TYPE_READ) {
buf = abd_borrow_buf(zio->io_abd, zio->io_size);
err = zfs_file_pread(vf->vf_file, buf, size, off, &resid);
abd_return_buf_copy(zio->io_abd, buf, size);
} else {
buf = abd_borrow_buf_copy(zio->io_abd, zio->io_size);
err = zfs_file_pwrite(vf->vf_file, buf, size, off, &resid);
abd_return_buf(zio->io_abd, buf, size);
}
zio->io_error = err;
if (resid != 0 && zio->io_error == 0)
zio->io_error = ENOSPC;
vdev_file_io_intr(zio);
}
static void
vdev_file_io_start(zio_t *zio)
{
vdev_t *vd = zio->io_vd;
vdev_file_t *vf = vd->vdev_tsd;
if (zio->io_type == ZIO_TYPE_IOCTL) {
/* XXPOLICY */
if (!vdev_readable(vd)) {
zio->io_error = SET_ERROR(ENXIO);
zio_interrupt(zio);
return;
}
switch (zio->io_cmd) {
case DKIOCFLUSHWRITECACHE:
zio->io_error = zfs_file_fsync(vf->vf_file,
O_SYNC|O_DSYNC);
break;
default:
zio->io_error = SET_ERROR(ENOTSUP);
}
zio_execute(zio);
return;
} else if (zio->io_type == ZIO_TYPE_TRIM) {
#ifdef notyet
int mode = 0;
ASSERT3U(zio->io_size, !=, 0);
/* XXX FreeBSD has no fallocate routine in file ops */
zio->io_error = zfs_file_fallocate(vf->vf_file,
mode, zio->io_offset, zio->io_size);
#endif
zio->io_error = SET_ERROR(ENOTSUP);
zio_execute(zio);
return;
}
ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
zio->io_target_timestamp = zio_handle_io_delay(zio);
VERIFY3U(taskq_dispatch(vdev_file_taskq, vdev_file_io_strategy, zio,
TQ_SLEEP), !=, 0);
}
static void
vdev_file_io_done(zio_t *zio)
{
(void) zio;
}
vdev_ops_t vdev_file_ops = {
.vdev_op_init = NULL,
.vdev_op_fini = NULL,
.vdev_op_open = vdev_file_open,
.vdev_op_close = vdev_file_close,
.vdev_op_asize = vdev_default_asize,
.vdev_op_min_asize = vdev_default_min_asize,
.vdev_op_min_alloc = NULL,
.vdev_op_io_start = vdev_file_io_start,
.vdev_op_io_done = vdev_file_io_done,
.vdev_op_state_change = NULL,
.vdev_op_need_resilver = NULL,
.vdev_op_hold = vdev_file_hold,
.vdev_op_rele = vdev_file_rele,
.vdev_op_remap = NULL,
.vdev_op_xlate = vdev_default_xlate,
.vdev_op_rebuild_asize = NULL,
.vdev_op_metaslab_init = NULL,
.vdev_op_config_generate = NULL,
.vdev_op_nparity = NULL,
.vdev_op_ndisks = NULL,
.vdev_op_type = VDEV_TYPE_FILE, /* name of this vdev type */
.vdev_op_leaf = B_TRUE /* leaf vdev */
};
/*
* From userland we access disks just like files.
*/
#ifndef _KERNEL
vdev_ops_t vdev_disk_ops = {
.vdev_op_init = NULL,
.vdev_op_fini = NULL,
.vdev_op_open = vdev_file_open,
.vdev_op_close = vdev_file_close,
.vdev_op_asize = vdev_default_asize,
.vdev_op_min_asize = vdev_default_min_asize,
.vdev_op_min_alloc = NULL,
.vdev_op_io_start = vdev_file_io_start,
.vdev_op_io_done = vdev_file_io_done,
.vdev_op_state_change = NULL,
.vdev_op_need_resilver = NULL,
.vdev_op_hold = vdev_file_hold,
.vdev_op_rele = vdev_file_rele,
.vdev_op_remap = NULL,
.vdev_op_xlate = vdev_default_xlate,
.vdev_op_rebuild_asize = NULL,
.vdev_op_metaslab_init = NULL,
.vdev_op_config_generate = NULL,
.vdev_op_nparity = NULL,
.vdev_op_ndisks = NULL,
.vdev_op_type = VDEV_TYPE_DISK, /* name of this vdev type */
.vdev_op_leaf = B_TRUE /* leaf vdev */
};
#endif
ZFS_MODULE_PARAM(zfs_vdev_file, vdev_file_, logical_ashift, ULONG, ZMOD_RW,
"Logical ashift for file-based devices");
ZFS_MODULE_PARAM(zfs_vdev_file, vdev_file_, physical_ashift, ULONG, ZMOD_RW,
"Physical ashift for file-based devices");
diff --git a/module/os/freebsd/zfs/zfs_acl.c b/module/os/freebsd/zfs/zfs_acl.c
index ae758bcefe21..bd22cda416bd 100644
--- a/module/os/freebsd/zfs/zfs_acl.c
+++ b/module/os/freebsd/zfs/zfs_acl.c
@@ -1,2672 +1,2672 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright 2017 Nexenta Systems, Inc. All rights reserved.
*/
#include <sys/types.h>
#include <sys/param.h>
#include <sys/time.h>
#include <sys/systm.h>
#include <sys/sysmacros.h>
#include <sys/resource.h>
#include <sys/vfs.h>
#include <sys/vnode.h>
#include <sys/file.h>
#include <sys/stat.h>
#include <sys/kmem.h>
#include <sys/cmn_err.h>
#include <sys/errno.h>
#include <sys/unistd.h>
#include <sys/sdt.h>
#include <sys/fs/zfs.h>
#include <sys/policy.h>
#include <sys/zfs_znode.h>
#include <sys/zfs_fuid.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_dir.h>
#include <sys/zfs_quota.h>
#include <sys/zfs_vfsops.h>
#include <sys/dmu.h>
#include <sys/dnode.h>
#include <sys/zap.h>
#include <sys/sa.h>
#include <acl/acl_common.h>
#define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
#define DENY ACE_ACCESS_DENIED_ACE_TYPE
#define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
#define MIN_ACE_TYPE ALLOW
#define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
#define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
#define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
#define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
#define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
#define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
#define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
ACE_DELETE|ACE_DELETE_CHILD)
#define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
#define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
#define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
#define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
#define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
#define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
ZFS_ACL_PROTECTED)
#define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
ZFS_ACL_OBJ_ACE)
#define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
static uint16_t
zfs_ace_v0_get_type(void *acep)
{
return (((zfs_oldace_t *)acep)->z_type);
}
static uint16_t
zfs_ace_v0_get_flags(void *acep)
{
return (((zfs_oldace_t *)acep)->z_flags);
}
static uint32_t
zfs_ace_v0_get_mask(void *acep)
{
return (((zfs_oldace_t *)acep)->z_access_mask);
}
static uint64_t
zfs_ace_v0_get_who(void *acep)
{
return (((zfs_oldace_t *)acep)->z_fuid);
}
static void
zfs_ace_v0_set_type(void *acep, uint16_t type)
{
((zfs_oldace_t *)acep)->z_type = type;
}
static void
zfs_ace_v0_set_flags(void *acep, uint16_t flags)
{
((zfs_oldace_t *)acep)->z_flags = flags;
}
static void
zfs_ace_v0_set_mask(void *acep, uint32_t mask)
{
((zfs_oldace_t *)acep)->z_access_mask = mask;
}
static void
zfs_ace_v0_set_who(void *acep, uint64_t who)
{
((zfs_oldace_t *)acep)->z_fuid = who;
}
/*ARGSUSED*/
static size_t
zfs_ace_v0_size(void *acep)
{
return (sizeof (zfs_oldace_t));
}
static size_t
zfs_ace_v0_abstract_size(void)
{
return (sizeof (zfs_oldace_t));
}
static int
zfs_ace_v0_mask_off(void)
{
return (offsetof(zfs_oldace_t, z_access_mask));
}
/*ARGSUSED*/
static int
zfs_ace_v0_data(void *acep, void **datap)
{
*datap = NULL;
return (0);
}
-static acl_ops_t zfs_acl_v0_ops = {
+static const acl_ops_t zfs_acl_v0_ops = {
zfs_ace_v0_get_mask,
zfs_ace_v0_set_mask,
zfs_ace_v0_get_flags,
zfs_ace_v0_set_flags,
zfs_ace_v0_get_type,
zfs_ace_v0_set_type,
zfs_ace_v0_get_who,
zfs_ace_v0_set_who,
zfs_ace_v0_size,
zfs_ace_v0_abstract_size,
zfs_ace_v0_mask_off,
zfs_ace_v0_data
};
static uint16_t
zfs_ace_fuid_get_type(void *acep)
{
return (((zfs_ace_hdr_t *)acep)->z_type);
}
static uint16_t
zfs_ace_fuid_get_flags(void *acep)
{
return (((zfs_ace_hdr_t *)acep)->z_flags);
}
static uint32_t
zfs_ace_fuid_get_mask(void *acep)
{
return (((zfs_ace_hdr_t *)acep)->z_access_mask);
}
static uint64_t
zfs_ace_fuid_get_who(void *args)
{
uint16_t entry_type;
zfs_ace_t *acep = args;
entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE)
return (-1);
return (((zfs_ace_t *)acep)->z_fuid);
}
static void
zfs_ace_fuid_set_type(void *acep, uint16_t type)
{
((zfs_ace_hdr_t *)acep)->z_type = type;
}
static void
zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
{
((zfs_ace_hdr_t *)acep)->z_flags = flags;
}
static void
zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
{
((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
}
static void
zfs_ace_fuid_set_who(void *arg, uint64_t who)
{
zfs_ace_t *acep = arg;
uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE)
return;
acep->z_fuid = who;
}
static size_t
zfs_ace_fuid_size(void *acep)
{
zfs_ace_hdr_t *zacep = acep;
uint16_t entry_type;
switch (zacep->z_type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
return (sizeof (zfs_object_ace_t));
case ALLOW:
case DENY:
entry_type =
(((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
if (entry_type == ACE_OWNER ||
entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE)
return (sizeof (zfs_ace_hdr_t));
fallthrough;
default:
return (sizeof (zfs_ace_t));
}
}
static size_t
zfs_ace_fuid_abstract_size(void)
{
return (sizeof (zfs_ace_hdr_t));
}
static int
zfs_ace_fuid_mask_off(void)
{
return (offsetof(zfs_ace_hdr_t, z_access_mask));
}
static int
zfs_ace_fuid_data(void *acep, void **datap)
{
zfs_ace_t *zacep = acep;
zfs_object_ace_t *zobjp;
switch (zacep->z_hdr.z_type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
zobjp = acep;
*datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
default:
*datap = NULL;
return (0);
}
}
-static acl_ops_t zfs_acl_fuid_ops = {
+static const acl_ops_t zfs_acl_fuid_ops = {
zfs_ace_fuid_get_mask,
zfs_ace_fuid_set_mask,
zfs_ace_fuid_get_flags,
zfs_ace_fuid_set_flags,
zfs_ace_fuid_get_type,
zfs_ace_fuid_set_type,
zfs_ace_fuid_get_who,
zfs_ace_fuid_set_who,
zfs_ace_fuid_size,
zfs_ace_fuid_abstract_size,
zfs_ace_fuid_mask_off,
zfs_ace_fuid_data
};
/*
* The following three functions are provided for compatibility with
* older ZPL version in order to determine if the file use to have
* an external ACL and what version of ACL previously existed on the
* file. Would really be nice to not need this, sigh.
*/
uint64_t
zfs_external_acl(znode_t *zp)
{
zfs_acl_phys_t acl_phys;
int error;
if (zp->z_is_sa)
return (0);
/*
* Need to deal with a potential
* race where zfs_sa_upgrade could cause
* z_isa_sa to change.
*
* If the lookup fails then the state of z_is_sa should have
* changed.
*/
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
&acl_phys, sizeof (acl_phys))) == 0)
return (acl_phys.z_acl_extern_obj);
else {
/*
* after upgrade the SA_ZPL_ZNODE_ACL should have been
* removed
*/
VERIFY(zp->z_is_sa);
VERIFY3S(error, ==, ENOENT);
return (0);
}
}
/*
* Determine size of ACL in bytes
*
* This is more complicated than it should be since we have to deal
* with old external ACLs.
*/
static int
zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
zfs_acl_phys_t *aclphys)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
uint64_t acl_count;
int size;
int error;
ASSERT(MUTEX_HELD(&zp->z_acl_lock));
if (zp->z_is_sa) {
if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zfsvfs),
&size)) != 0)
return (error);
*aclsize = size;
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zfsvfs),
&acl_count, sizeof (acl_count))) != 0)
return (error);
*aclcount = acl_count;
} else {
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
aclphys, sizeof (*aclphys))) != 0)
return (error);
if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
*aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
*aclcount = aclphys->z_acl_size;
} else {
*aclsize = aclphys->z_acl_size;
*aclcount = aclphys->z_acl_count;
}
}
return (0);
}
int
zfs_znode_acl_version(znode_t *zp)
{
zfs_acl_phys_t acl_phys;
if (zp->z_is_sa)
return (ZFS_ACL_VERSION_FUID);
else {
int error;
/*
* Need to deal with a potential
* race where zfs_sa_upgrade could cause
* z_isa_sa to change.
*
* If the lookup fails then the state of z_is_sa should have
* changed.
*/
if ((error = sa_lookup(zp->z_sa_hdl,
SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
&acl_phys, sizeof (acl_phys))) == 0)
return (acl_phys.z_acl_version);
else {
/*
* After upgrade SA_ZPL_ZNODE_ACL should have
* been removed.
*/
VERIFY(zp->z_is_sa);
VERIFY3S(error, ==, ENOENT);
return (ZFS_ACL_VERSION_FUID);
}
}
}
static int
zfs_acl_version(int version)
{
if (version < ZPL_VERSION_FUID)
return (ZFS_ACL_VERSION_INITIAL);
else
return (ZFS_ACL_VERSION_FUID);
}
static int
zfs_acl_version_zp(znode_t *zp)
{
return (zfs_acl_version(zp->z_zfsvfs->z_version));
}
zfs_acl_t *
zfs_acl_alloc(int vers)
{
zfs_acl_t *aclp;
aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
offsetof(zfs_acl_node_t, z_next));
aclp->z_version = vers;
if (vers == ZFS_ACL_VERSION_FUID)
aclp->z_ops = &zfs_acl_fuid_ops;
else
aclp->z_ops = &zfs_acl_v0_ops;
return (aclp);
}
zfs_acl_node_t *
zfs_acl_node_alloc(size_t bytes)
{
zfs_acl_node_t *aclnode;
aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
if (bytes) {
aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
aclnode->z_allocdata = aclnode->z_acldata;
aclnode->z_allocsize = bytes;
aclnode->z_size = bytes;
}
return (aclnode);
}
static void
zfs_acl_node_free(zfs_acl_node_t *aclnode)
{
if (aclnode->z_allocsize)
kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
kmem_free(aclnode, sizeof (zfs_acl_node_t));
}
static void
zfs_acl_release_nodes(zfs_acl_t *aclp)
{
zfs_acl_node_t *aclnode;
while ((aclnode = list_head(&aclp->z_acl))) {
list_remove(&aclp->z_acl, aclnode);
zfs_acl_node_free(aclnode);
}
aclp->z_acl_count = 0;
aclp->z_acl_bytes = 0;
}
void
zfs_acl_free(zfs_acl_t *aclp)
{
zfs_acl_release_nodes(aclp);
list_destroy(&aclp->z_acl);
kmem_free(aclp, sizeof (zfs_acl_t));
}
static boolean_t
zfs_acl_valid_ace_type(uint_t type, uint_t flags)
{
uint16_t entry_type;
switch (type) {
case ALLOW:
case DENY:
case ACE_SYSTEM_AUDIT_ACE_TYPE:
case ACE_SYSTEM_ALARM_ACE_TYPE:
entry_type = flags & ACE_TYPE_FLAGS;
return (entry_type == ACE_OWNER ||
entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE || entry_type == 0 ||
entry_type == ACE_IDENTIFIER_GROUP);
default:
if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
return (B_TRUE);
}
return (B_FALSE);
}
static boolean_t
zfs_ace_valid(vtype_t obj_type, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
{
/*
* first check type of entry
*/
if (!zfs_acl_valid_ace_type(type, iflags))
return (B_FALSE);
switch (type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
if (aclp->z_version < ZFS_ACL_VERSION_FUID)
return (B_FALSE);
aclp->z_hints |= ZFS_ACL_OBJ_ACE;
}
/*
* next check inheritance level flags
*/
if (obj_type == VDIR &&
(iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
aclp->z_hints |= ZFS_INHERIT_ACE;
if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
if ((iflags & (ACE_FILE_INHERIT_ACE|
ACE_DIRECTORY_INHERIT_ACE)) == 0) {
return (B_FALSE);
}
}
return (B_TRUE);
}
static void *
zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
{
zfs_acl_node_t *aclnode;
ASSERT3P(aclp, !=, NULL);
if (start == NULL) {
aclnode = list_head(&aclp->z_acl);
if (aclnode == NULL)
return (NULL);
aclp->z_next_ace = aclnode->z_acldata;
aclp->z_curr_node = aclnode;
aclnode->z_ace_idx = 0;
}
aclnode = aclp->z_curr_node;
if (aclnode == NULL)
return (NULL);
if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
aclnode = list_next(&aclp->z_acl, aclnode);
if (aclnode == NULL)
return (NULL);
else {
aclp->z_curr_node = aclnode;
aclnode->z_ace_idx = 0;
aclp->z_next_ace = aclnode->z_acldata;
}
}
if (aclnode->z_ace_idx < aclnode->z_ace_count) {
void *acep = aclp->z_next_ace;
size_t ace_size;
/*
* Make sure we don't overstep our bounds
*/
ace_size = aclp->z_ops->ace_size(acep);
if (((caddr_t)acep + ace_size) >
((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
return (NULL);
}
*iflags = aclp->z_ops->ace_flags_get(acep);
*type = aclp->z_ops->ace_type_get(acep);
*access_mask = aclp->z_ops->ace_mask_get(acep);
*who = aclp->z_ops->ace_who_get(acep);
aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
aclnode->z_ace_idx++;
return ((void *)acep);
}
return (NULL);
}
/*ARGSUSED*/
static uint64_t
zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
uint16_t *flags, uint16_t *type, uint32_t *mask)
{
zfs_acl_t *aclp = datap;
zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
uint64_t who;
acep = zfs_acl_next_ace(aclp, acep, &who, mask,
flags, type);
return ((uint64_t)(uintptr_t)acep);
}
/*
* Copy ACE to internal ZFS format.
* While processing the ACL each ACE will be validated for correctness.
* ACE FUIDs will be created later.
*/
static int
zfs_copy_ace_2_fuid(zfsvfs_t *zfsvfs, vtype_t obj_type, zfs_acl_t *aclp,
void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
zfs_fuid_info_t **fuidp, cred_t *cr)
{
int i;
uint16_t entry_type;
zfs_ace_t *aceptr = z_acl;
ace_t *acep = datap;
zfs_object_ace_t *zobjacep;
ace_object_t *aceobjp;
for (i = 0; i != aclcnt; i++) {
aceptr->z_hdr.z_access_mask = acep->a_access_mask;
aceptr->z_hdr.z_flags = acep->a_flags;
aceptr->z_hdr.z_type = acep->a_type;
entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
entry_type != ACE_EVERYONE) {
aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
cr, (entry_type == 0) ?
ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
}
/*
* Make sure ACE is valid
*/
if (zfs_ace_valid(obj_type, aclp, aceptr->z_hdr.z_type,
aceptr->z_hdr.z_flags) != B_TRUE)
return (SET_ERROR(EINVAL));
switch (acep->a_type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
zobjacep = (zfs_object_ace_t *)aceptr;
aceobjp = (ace_object_t *)acep;
bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
sizeof (aceobjp->a_obj_type));
bcopy(aceobjp->a_inherit_obj_type,
zobjacep->z_inherit_type,
sizeof (aceobjp->a_inherit_obj_type));
acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
break;
default:
acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
}
aceptr = (zfs_ace_t *)((caddr_t)aceptr +
aclp->z_ops->ace_size(aceptr));
}
*size = (caddr_t)aceptr - (caddr_t)z_acl;
return (0);
}
/*
* Copy ZFS ACEs to fixed size ace_t layout
*/
static void
zfs_copy_fuid_2_ace(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, cred_t *cr,
void *datap, int filter)
{
uint64_t who;
uint32_t access_mask;
uint16_t iflags, type;
zfs_ace_hdr_t *zacep = NULL;
ace_t *acep = datap;
ace_object_t *objacep;
zfs_object_ace_t *zobjacep;
size_t ace_size;
uint16_t entry_type;
while ((zacep = zfs_acl_next_ace(aclp, zacep,
&who, &access_mask, &iflags, &type))) {
switch (type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
if (filter) {
continue;
}
zobjacep = (zfs_object_ace_t *)zacep;
objacep = (ace_object_t *)acep;
bcopy(zobjacep->z_object_type,
objacep->a_obj_type,
sizeof (zobjacep->z_object_type));
bcopy(zobjacep->z_inherit_type,
objacep->a_inherit_obj_type,
sizeof (zobjacep->z_inherit_type));
ace_size = sizeof (ace_object_t);
break;
default:
ace_size = sizeof (ace_t);
break;
}
entry_type = (iflags & ACE_TYPE_FLAGS);
if ((entry_type != ACE_OWNER &&
entry_type != OWNING_GROUP &&
entry_type != ACE_EVERYONE)) {
acep->a_who = zfs_fuid_map_id(zfsvfs, who,
cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
ZFS_ACE_GROUP : ZFS_ACE_USER);
} else {
acep->a_who = (uid_t)(int64_t)who;
}
acep->a_access_mask = access_mask;
acep->a_flags = iflags;
acep->a_type = type;
acep = (ace_t *)((caddr_t)acep + ace_size);
}
}
static int
zfs_copy_ace_2_oldace(vtype_t obj_type, zfs_acl_t *aclp, ace_t *acep,
zfs_oldace_t *z_acl, int aclcnt, size_t *size)
{
int i;
zfs_oldace_t *aceptr = z_acl;
for (i = 0; i != aclcnt; i++, aceptr++) {
aceptr->z_access_mask = acep[i].a_access_mask;
aceptr->z_type = acep[i].a_type;
aceptr->z_flags = acep[i].a_flags;
aceptr->z_fuid = acep[i].a_who;
/*
* Make sure ACE is valid
*/
if (zfs_ace_valid(obj_type, aclp, aceptr->z_type,
aceptr->z_flags) != B_TRUE)
return (SET_ERROR(EINVAL));
}
*size = (caddr_t)aceptr - (caddr_t)z_acl;
return (0);
}
/*
* convert old ACL format to new
*/
void
zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
{
zfs_oldace_t *oldaclp;
int i;
uint16_t type, iflags;
uint32_t access_mask;
uint64_t who;
void *cookie = NULL;
zfs_acl_node_t *newaclnode;
ASSERT3U(aclp->z_version, ==, ZFS_ACL_VERSION_INITIAL);
/*
* First create the ACE in a contiguous piece of memory
* for zfs_copy_ace_2_fuid().
*
* We only convert an ACL once, so this won't happen
* everytime.
*/
oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
KM_SLEEP);
i = 0;
while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
&access_mask, &iflags, &type))) {
oldaclp[i].z_flags = iflags;
oldaclp[i].z_type = type;
oldaclp[i].z_fuid = who;
oldaclp[i++].z_access_mask = access_mask;
}
newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
sizeof (zfs_object_ace_t));
aclp->z_ops = &zfs_acl_fuid_ops;
VERIFY0(zfs_copy_ace_2_fuid(zp->z_zfsvfs, ZTOV(zp)->v_type, aclp,
oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
&newaclnode->z_size, NULL, cr));
newaclnode->z_ace_count = aclp->z_acl_count;
aclp->z_version = ZFS_ACL_VERSION;
kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
/*
* Release all previous ACL nodes
*/
zfs_acl_release_nodes(aclp);
list_insert_head(&aclp->z_acl, newaclnode);
aclp->z_acl_bytes = newaclnode->z_size;
aclp->z_acl_count = newaclnode->z_ace_count;
}
/*
* Convert unix access mask to v4 access mask
*/
static uint32_t
zfs_unix_to_v4(uint32_t access_mask)
{
uint32_t new_mask = 0;
if (access_mask & S_IXOTH)
new_mask |= ACE_EXECUTE;
if (access_mask & S_IWOTH)
new_mask |= ACE_WRITE_DATA;
if (access_mask & S_IROTH)
new_mask |= ACE_READ_DATA;
return (new_mask);
}
static void
zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
uint16_t access_type, uint64_t fuid, uint16_t entry_type)
{
uint16_t type = entry_type & ACE_TYPE_FLAGS;
aclp->z_ops->ace_mask_set(acep, access_mask);
aclp->z_ops->ace_type_set(acep, access_type);
aclp->z_ops->ace_flags_set(acep, entry_type);
if ((type != ACE_OWNER && type != OWNING_GROUP &&
type != ACE_EVERYONE))
aclp->z_ops->ace_who_set(acep, fuid);
}
/*
* Determine mode of file based on ACL.
*/
uint64_t
zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
uint64_t *pflags, uint64_t fuid, uint64_t fgid)
{
int entry_type;
mode_t mode;
mode_t seen = 0;
zfs_ace_hdr_t *acep = NULL;
uint64_t who;
uint16_t iflags, type;
uint32_t access_mask;
boolean_t an_exec_denied = B_FALSE;
mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
while ((acep = zfs_acl_next_ace(aclp, acep, &who,
&access_mask, &iflags, &type))) {
if (!zfs_acl_valid_ace_type(type, iflags))
continue;
entry_type = (iflags & ACE_TYPE_FLAGS);
/*
* Skip over any inherit_only ACEs
*/
if (iflags & ACE_INHERIT_ONLY_ACE)
continue;
if (entry_type == ACE_OWNER || (entry_type == 0 &&
who == fuid)) {
if ((access_mask & ACE_READ_DATA) &&
(!(seen & S_IRUSR))) {
seen |= S_IRUSR;
if (type == ALLOW) {
mode |= S_IRUSR;
}
}
if ((access_mask & ACE_WRITE_DATA) &&
(!(seen & S_IWUSR))) {
seen |= S_IWUSR;
if (type == ALLOW) {
mode |= S_IWUSR;
}
}
if ((access_mask & ACE_EXECUTE) &&
(!(seen & S_IXUSR))) {
seen |= S_IXUSR;
if (type == ALLOW) {
mode |= S_IXUSR;
}
}
} else if (entry_type == OWNING_GROUP ||
(entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
if ((access_mask & ACE_READ_DATA) &&
(!(seen & S_IRGRP))) {
seen |= S_IRGRP;
if (type == ALLOW) {
mode |= S_IRGRP;
}
}
if ((access_mask & ACE_WRITE_DATA) &&
(!(seen & S_IWGRP))) {
seen |= S_IWGRP;
if (type == ALLOW) {
mode |= S_IWGRP;
}
}
if ((access_mask & ACE_EXECUTE) &&
(!(seen & S_IXGRP))) {
seen |= S_IXGRP;
if (type == ALLOW) {
mode |= S_IXGRP;
}
}
} else if (entry_type == ACE_EVERYONE) {
if ((access_mask & ACE_READ_DATA)) {
if (!(seen & S_IRUSR)) {
seen |= S_IRUSR;
if (type == ALLOW) {
mode |= S_IRUSR;
}
}
if (!(seen & S_IRGRP)) {
seen |= S_IRGRP;
if (type == ALLOW) {
mode |= S_IRGRP;
}
}
if (!(seen & S_IROTH)) {
seen |= S_IROTH;
if (type == ALLOW) {
mode |= S_IROTH;
}
}
}
if ((access_mask & ACE_WRITE_DATA)) {
if (!(seen & S_IWUSR)) {
seen |= S_IWUSR;
if (type == ALLOW) {
mode |= S_IWUSR;
}
}
if (!(seen & S_IWGRP)) {
seen |= S_IWGRP;
if (type == ALLOW) {
mode |= S_IWGRP;
}
}
if (!(seen & S_IWOTH)) {
seen |= S_IWOTH;
if (type == ALLOW) {
mode |= S_IWOTH;
}
}
}
if ((access_mask & ACE_EXECUTE)) {
if (!(seen & S_IXUSR)) {
seen |= S_IXUSR;
if (type == ALLOW) {
mode |= S_IXUSR;
}
}
if (!(seen & S_IXGRP)) {
seen |= S_IXGRP;
if (type == ALLOW) {
mode |= S_IXGRP;
}
}
if (!(seen & S_IXOTH)) {
seen |= S_IXOTH;
if (type == ALLOW) {
mode |= S_IXOTH;
}
}
}
} else {
/*
* Only care if this IDENTIFIER_GROUP or
* USER ACE denies execute access to someone,
* mode is not affected
*/
if ((access_mask & ACE_EXECUTE) && type == DENY)
an_exec_denied = B_TRUE;
}
}
/*
* Failure to allow is effectively a deny, so execute permission
* is denied if it was never mentioned or if we explicitly
* weren't allowed it.
*/
if (!an_exec_denied &&
((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
(mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
an_exec_denied = B_TRUE;
if (an_exec_denied)
*pflags &= ~ZFS_NO_EXECS_DENIED;
else
*pflags |= ZFS_NO_EXECS_DENIED;
return (mode);
}
/*
* Read an external acl object. If the intent is to modify, always
* create a new acl and leave any cached acl in place.
*/
int
zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp,
boolean_t will_modify)
{
zfs_acl_t *aclp;
int aclsize;
int acl_count;
zfs_acl_node_t *aclnode;
zfs_acl_phys_t znode_acl;
int version;
int error;
ASSERT(MUTEX_HELD(&zp->z_acl_lock));
if (zp->z_zfsvfs->z_replay == B_FALSE)
ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
if (zp->z_acl_cached && !will_modify) {
*aclpp = zp->z_acl_cached;
return (0);
}
version = zfs_znode_acl_version(zp);
if ((error = zfs_acl_znode_info(zp, &aclsize,
&acl_count, &znode_acl)) != 0) {
goto done;
}
aclp = zfs_acl_alloc(version);
aclp->z_acl_count = acl_count;
aclp->z_acl_bytes = aclsize;
aclnode = zfs_acl_node_alloc(aclsize);
aclnode->z_ace_count = aclp->z_acl_count;
aclnode->z_size = aclsize;
if (!zp->z_is_sa) {
if (znode_acl.z_acl_extern_obj) {
error = dmu_read(zp->z_zfsvfs->z_os,
znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
aclnode->z_acldata, DMU_READ_PREFETCH);
} else {
bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
aclnode->z_size);
}
} else {
error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zp->z_zfsvfs),
aclnode->z_acldata, aclnode->z_size);
}
if (error != 0) {
zfs_acl_free(aclp);
zfs_acl_node_free(aclnode);
/* convert checksum errors into IO errors */
if (error == ECKSUM)
error = SET_ERROR(EIO);
goto done;
}
list_insert_head(&aclp->z_acl, aclnode);
*aclpp = aclp;
if (!will_modify)
zp->z_acl_cached = aclp;
done:
return (error);
}
/*ARGSUSED*/
void
zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
boolean_t start, void *userdata)
{
zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
if (start) {
cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
} else {
cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
cb->cb_acl_node);
}
*dataptr = cb->cb_acl_node->z_acldata;
*length = cb->cb_acl_node->z_size;
}
int
zfs_acl_chown_setattr(znode_t *zp)
{
int error;
zfs_acl_t *aclp;
if (zp->z_zfsvfs->z_replay == B_FALSE) {
ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
ASSERT_VOP_IN_SEQC(ZTOV(zp));
}
ASSERT(MUTEX_HELD(&zp->z_acl_lock));
if ((error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE)) == 0)
zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
&zp->z_pflags, zp->z_uid, zp->z_gid);
return (error);
}
/*
* common code for setting ACLs.
*
* This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
* zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
* already checked the acl and knows whether to inherit.
*/
int
zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
{
int error;
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
dmu_object_type_t otype;
zfs_acl_locator_cb_t locate = { 0 };
uint64_t mode;
sa_bulk_attr_t bulk[5];
uint64_t ctime[2];
int count = 0;
zfs_acl_phys_t acl_phys;
if (zp->z_zfsvfs->z_replay == B_FALSE) {
ASSERT_VOP_IN_SEQC(ZTOV(zp));
}
mode = zp->z_mode;
mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
zp->z_uid, zp->z_gid);
zp->z_mode = mode;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
&mode, sizeof (mode));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
&ctime, sizeof (ctime));
if (zp->z_acl_cached) {
zfs_acl_free(zp->z_acl_cached);
zp->z_acl_cached = NULL;
}
/*
* Upgrade needed?
*/
if (!zfsvfs->z_use_fuids) {
otype = DMU_OT_OLDACL;
} else {
if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
(zfsvfs->z_version >= ZPL_VERSION_FUID))
zfs_acl_xform(zp, aclp, cr);
ASSERT3U(aclp->z_version, >=, ZFS_ACL_VERSION_FUID);
otype = DMU_OT_ACL;
}
/*
* Arrgh, we have to handle old on disk format
* as well as newer (preferred) SA format.
*/
if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
locate.cb_aclp = aclp;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
NULL, &aclp->z_acl_count, sizeof (uint64_t));
} else { /* Painful legacy way */
zfs_acl_node_t *aclnode;
uint64_t off = 0;
uint64_t aoid;
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
&acl_phys, sizeof (acl_phys))) != 0)
return (error);
aoid = acl_phys.z_acl_extern_obj;
if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
/*
* If ACL was previously external and we are now
* converting to new ACL format then release old
* ACL object and create a new one.
*/
if (aoid &&
aclp->z_version != acl_phys.z_acl_version) {
error = dmu_object_free(zfsvfs->z_os, aoid, tx);
if (error)
return (error);
aoid = 0;
}
if (aoid == 0) {
aoid = dmu_object_alloc(zfsvfs->z_os,
otype, aclp->z_acl_bytes,
otype == DMU_OT_ACL ?
DMU_OT_SYSACL : DMU_OT_NONE,
otype == DMU_OT_ACL ?
DN_OLD_MAX_BONUSLEN : 0, tx);
} else {
(void) dmu_object_set_blocksize(zfsvfs->z_os,
aoid, aclp->z_acl_bytes, 0, tx);
}
acl_phys.z_acl_extern_obj = aoid;
for (aclnode = list_head(&aclp->z_acl); aclnode;
aclnode = list_next(&aclp->z_acl, aclnode)) {
if (aclnode->z_ace_count == 0)
continue;
dmu_write(zfsvfs->z_os, aoid, off,
aclnode->z_size, aclnode->z_acldata, tx);
off += aclnode->z_size;
}
} else {
void *start = acl_phys.z_ace_data;
/*
* Migrating back embedded?
*/
if (acl_phys.z_acl_extern_obj) {
error = dmu_object_free(zfsvfs->z_os,
acl_phys.z_acl_extern_obj, tx);
if (error)
return (error);
acl_phys.z_acl_extern_obj = 0;
}
for (aclnode = list_head(&aclp->z_acl); aclnode;
aclnode = list_next(&aclp->z_acl, aclnode)) {
if (aclnode->z_ace_count == 0)
continue;
bcopy(aclnode->z_acldata, start,
aclnode->z_size);
start = (caddr_t)start + aclnode->z_size;
}
}
/*
* If Old version then swap count/bytes to match old
* layout of znode_acl_phys_t.
*/
if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
acl_phys.z_acl_size = aclp->z_acl_count;
acl_phys.z_acl_count = aclp->z_acl_bytes;
} else {
acl_phys.z_acl_size = aclp->z_acl_bytes;
acl_phys.z_acl_count = aclp->z_acl_count;
}
acl_phys.z_acl_version = aclp->z_version;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
&acl_phys, sizeof (acl_phys));
}
/*
* Replace ACL wide bits, but first clear them.
*/
zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
zp->z_pflags |= aclp->z_hints;
if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
zp->z_pflags |= ZFS_ACL_TRIVIAL;
zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
}
static void
zfs_acl_chmod(vtype_t vtype, uint64_t mode, boolean_t split, boolean_t trim,
zfs_acl_t *aclp)
{
void *acep = NULL;
uint64_t who;
int new_count, new_bytes;
int ace_size;
int entry_type;
uint16_t iflags, type;
uint32_t access_mask;
zfs_acl_node_t *newnode;
size_t abstract_size = aclp->z_ops->ace_abstract_size();
void *zacep;
boolean_t isdir;
trivial_acl_t masks;
new_count = new_bytes = 0;
isdir = (vtype == VDIR);
acl_trivial_access_masks((mode_t)mode, isdir, &masks);
newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
zacep = newnode->z_acldata;
if (masks.allow0) {
zfs_set_ace(aclp, zacep, masks.allow0, ALLOW, -1, ACE_OWNER);
zacep = (void *)((uintptr_t)zacep + abstract_size);
new_count++;
new_bytes += abstract_size;
}
if (masks.deny1) {
zfs_set_ace(aclp, zacep, masks.deny1, DENY, -1, ACE_OWNER);
zacep = (void *)((uintptr_t)zacep + abstract_size);
new_count++;
new_bytes += abstract_size;
}
if (masks.deny2) {
zfs_set_ace(aclp, zacep, masks.deny2, DENY, -1, OWNING_GROUP);
zacep = (void *)((uintptr_t)zacep + abstract_size);
new_count++;
new_bytes += abstract_size;
}
while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
&iflags, &type))) {
entry_type = (iflags & ACE_TYPE_FLAGS);
/*
* ACEs used to represent the file mode may be divided
* into an equivalent pair of inherit-only and regular
* ACEs, if they are inheritable.
* Skip regular ACEs, which are replaced by the new mode.
*/
if (split && (entry_type == ACE_OWNER ||
entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE)) {
if (!isdir || !(iflags &
(ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
continue;
/*
* We preserve owner@, group@, or @everyone
* permissions, if they are inheritable, by
* copying them to inherit_only ACEs. This
* prevents inheritable permissions from being
* altered along with the file mode.
*/
iflags |= ACE_INHERIT_ONLY_ACE;
}
/*
* If this ACL has any inheritable ACEs, mark that in
* the hints (which are later masked into the pflags)
* so create knows to do inheritance.
*/
if (isdir && (iflags &
(ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
aclp->z_hints |= ZFS_INHERIT_ACE;
if ((type != ALLOW && type != DENY) ||
(iflags & ACE_INHERIT_ONLY_ACE)) {
switch (type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
aclp->z_hints |= ZFS_ACL_OBJ_ACE;
break;
}
} else {
/*
* Limit permissions granted by ACEs to be no greater
* than permissions of the requested group mode.
* Applies when the "aclmode" property is set to
* "groupmask".
*/
if ((type == ALLOW) && trim)
access_mask &= masks.group;
}
zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
ace_size = aclp->z_ops->ace_size(acep);
zacep = (void *)((uintptr_t)zacep + ace_size);
new_count++;
new_bytes += ace_size;
}
zfs_set_ace(aclp, zacep, masks.owner, ALLOW, -1, ACE_OWNER);
zacep = (void *)((uintptr_t)zacep + abstract_size);
zfs_set_ace(aclp, zacep, masks.group, ALLOW, -1, OWNING_GROUP);
zacep = (void *)((uintptr_t)zacep + abstract_size);
zfs_set_ace(aclp, zacep, masks.everyone, ALLOW, -1, ACE_EVERYONE);
new_count += 3;
new_bytes += abstract_size * 3;
zfs_acl_release_nodes(aclp);
aclp->z_acl_count = new_count;
aclp->z_acl_bytes = new_bytes;
newnode->z_ace_count = new_count;
newnode->z_size = new_bytes;
list_insert_tail(&aclp->z_acl, newnode);
}
int
zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
{
int error = 0;
mutex_enter(&zp->z_acl_lock);
if (zp->z_zfsvfs->z_replay == B_FALSE)
ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_DISCARD)
*aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
else
error = zfs_acl_node_read(zp, B_TRUE, aclp, B_TRUE);
if (error == 0) {
(*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
zfs_acl_chmod(ZTOV(zp)->v_type, mode, B_TRUE,
(zp->z_zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK), *aclp);
}
mutex_exit(&zp->z_acl_lock);
return (error);
}
/*
* Should ACE be inherited?
*/
static int
zfs_ace_can_use(vtype_t vtype, uint16_t acep_flags)
{
int iflags = (acep_flags & 0xf);
if ((vtype == VDIR) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
return (1);
else if (iflags & ACE_FILE_INHERIT_ACE)
return (!((vtype == VDIR) &&
(iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
return (0);
}
/*
* inherit inheritable ACEs from parent
*/
static zfs_acl_t *
zfs_acl_inherit(zfsvfs_t *zfsvfs, vtype_t vtype, zfs_acl_t *paclp,
uint64_t mode, boolean_t *need_chmod)
{
void *pacep = NULL;
void *acep;
zfs_acl_node_t *aclnode;
zfs_acl_t *aclp = NULL;
uint64_t who;
uint32_t access_mask;
uint16_t iflags, newflags, type;
size_t ace_size;
void *data1, *data2;
size_t data1sz, data2sz;
uint_t aclinherit;
boolean_t isdir = (vtype == VDIR);
boolean_t isreg = (vtype == VREG);
*need_chmod = B_TRUE;
aclp = zfs_acl_alloc(paclp->z_version);
aclinherit = zfsvfs->z_acl_inherit;
if (aclinherit == ZFS_ACL_DISCARD || vtype == VLNK)
return (aclp);
while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
&access_mask, &iflags, &type))) {
/*
* don't inherit bogus ACEs
*/
if (!zfs_acl_valid_ace_type(type, iflags))
continue;
/*
* Check if ACE is inheritable by this vnode
*/
if ((aclinherit == ZFS_ACL_NOALLOW && type == ALLOW) ||
!zfs_ace_can_use(vtype, iflags))
continue;
/*
* If owner@, group@, or everyone@ inheritable
* then zfs_acl_chmod() isn't needed.
*/
if ((aclinherit == ZFS_ACL_PASSTHROUGH ||
aclinherit == ZFS_ACL_PASSTHROUGH_X) &&
((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
((iflags & OWNING_GROUP) == OWNING_GROUP)) &&
(isreg || (isdir && (iflags & ACE_DIRECTORY_INHERIT_ACE))))
*need_chmod = B_FALSE;
/*
* Strip inherited execute permission from file if
* not in mode
*/
if (aclinherit == ZFS_ACL_PASSTHROUGH_X && type == ALLOW &&
!isdir && ((mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0)) {
access_mask &= ~ACE_EXECUTE;
}
/*
* Strip write_acl and write_owner from permissions
* when inheriting an ACE
*/
if (aclinherit == ZFS_ACL_RESTRICTED && type == ALLOW) {
access_mask &= ~RESTRICTED_CLEAR;
}
ace_size = aclp->z_ops->ace_size(pacep);
aclnode = zfs_acl_node_alloc(ace_size);
list_insert_tail(&aclp->z_acl, aclnode);
acep = aclnode->z_acldata;
zfs_set_ace(aclp, acep, access_mask, type,
who, iflags|ACE_INHERITED_ACE);
/*
* Copy special opaque data if any
*/
if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
data2sz = aclp->z_ops->ace_data(acep, &data2);
VERIFY3U(data2sz, ==, data1sz);
bcopy(data1, data2, data2sz);
}
aclp->z_acl_count++;
aclnode->z_ace_count++;
aclp->z_acl_bytes += aclnode->z_size;
newflags = aclp->z_ops->ace_flags_get(acep);
/*
* If ACE is not to be inherited further, or if the vnode is
* not a directory, remove all inheritance flags
*/
if (!isdir || (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)) {
newflags &= ~ALL_INHERIT;
aclp->z_ops->ace_flags_set(acep,
newflags|ACE_INHERITED_ACE);
continue;
}
/*
* This directory has an inheritable ACE
*/
aclp->z_hints |= ZFS_INHERIT_ACE;
/*
* If only FILE_INHERIT is set then turn on
* inherit_only
*/
if ((iflags & (ACE_FILE_INHERIT_ACE |
ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
newflags |= ACE_INHERIT_ONLY_ACE;
aclp->z_ops->ace_flags_set(acep,
newflags|ACE_INHERITED_ACE);
} else {
newflags &= ~ACE_INHERIT_ONLY_ACE;
aclp->z_ops->ace_flags_set(acep,
newflags|ACE_INHERITED_ACE);
}
}
if (zfsvfs->z_acl_mode == ZFS_ACL_RESTRICTED &&
aclp->z_acl_count != 0) {
*need_chmod = B_FALSE;
}
return (aclp);
}
/*
* Create file system object initial permissions
* including inheritable ACEs.
* Also, create FUIDs for owner and group.
*/
int
zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
{
int error;
zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
zfs_acl_t *paclp;
gid_t gid;
boolean_t need_chmod = B_TRUE;
boolean_t trim = B_FALSE;
boolean_t inherited = B_FALSE;
if ((flag & IS_ROOT_NODE) == 0) {
if (zfsvfs->z_replay == B_FALSE)
ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
} else
ASSERT3P(dzp->z_vnode, ==, NULL);
bzero(acl_ids, sizeof (zfs_acl_ids_t));
acl_ids->z_mode = MAKEIMODE(vap->va_type, vap->va_mode);
if (vsecp)
if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, cr,
&acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
return (error);
/*
* Determine uid and gid.
*/
if ((flag & IS_ROOT_NODE) || zfsvfs->z_replay ||
((flag & IS_XATTR) && (vap->va_type == VDIR))) {
acl_ids->z_fuid = zfs_fuid_create(zfsvfs,
(uint64_t)vap->va_uid, cr,
ZFS_OWNER, &acl_ids->z_fuidp);
acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
(uint64_t)vap->va_gid, cr,
ZFS_GROUP, &acl_ids->z_fuidp);
gid = vap->va_gid;
} else {
acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER,
cr, &acl_ids->z_fuidp);
acl_ids->z_fgid = 0;
if (vap->va_mask & AT_GID) {
acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
(uint64_t)vap->va_gid,
cr, ZFS_GROUP, &acl_ids->z_fuidp);
gid = vap->va_gid;
if (acl_ids->z_fgid != dzp->z_gid &&
!groupmember(vap->va_gid, cr) &&
secpolicy_vnode_create_gid(cr) != 0)
acl_ids->z_fgid = 0;
}
if (acl_ids->z_fgid == 0) {
char *domain;
uint32_t rid;
acl_ids->z_fgid = dzp->z_gid;
gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
cr, ZFS_GROUP);
if (zfsvfs->z_use_fuids &&
IS_EPHEMERAL(acl_ids->z_fgid)) {
domain =
zfs_fuid_idx_domain(&zfsvfs->z_fuid_idx,
FUID_INDEX(acl_ids->z_fgid));
rid = FUID_RID(acl_ids->z_fgid);
zfs_fuid_node_add(&acl_ids->z_fuidp,
domain, rid, FUID_INDEX(acl_ids->z_fgid),
acl_ids->z_fgid, ZFS_GROUP);
}
}
}
/*
* If we're creating a directory, and the parent directory has the
* set-GID bit set, set in on the new directory.
* Otherwise, if the user is neither privileged nor a member of the
* file's new group, clear the file's set-GID bit.
*/
if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
(vap->va_type == VDIR)) {
acl_ids->z_mode |= S_ISGID;
} else {
if ((acl_ids->z_mode & S_ISGID) &&
secpolicy_vnode_setids_setgids(ZTOV(dzp), cr, gid) != 0)
acl_ids->z_mode &= ~S_ISGID;
}
if (acl_ids->z_aclp == NULL) {
mutex_enter(&dzp->z_acl_lock);
if (!(flag & IS_ROOT_NODE) &&
(dzp->z_pflags & ZFS_INHERIT_ACE) &&
!(dzp->z_pflags & ZFS_XATTR)) {
VERIFY0(zfs_acl_node_read(dzp, B_TRUE,
&paclp, B_FALSE));
acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
vap->va_type, paclp, acl_ids->z_mode, &need_chmod);
inherited = B_TRUE;
} else {
acl_ids->z_aclp =
zfs_acl_alloc(zfs_acl_version_zp(dzp));
acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
}
mutex_exit(&dzp->z_acl_lock);
if (need_chmod) {
if (vap->va_type == VDIR)
acl_ids->z_aclp->z_hints |=
ZFS_ACL_AUTO_INHERIT;
if (zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK &&
zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH &&
zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH_X)
trim = B_TRUE;
zfs_acl_chmod(vap->va_type, acl_ids->z_mode, B_FALSE,
trim, acl_ids->z_aclp);
}
}
if (inherited || vsecp) {
acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
acl_ids->z_fuid, acl_ids->z_fgid);
if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
}
return (0);
}
/*
* Free ACL and fuid_infop, but not the acl_ids structure
*/
void
zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
{
if (acl_ids->z_aclp)
zfs_acl_free(acl_ids->z_aclp);
if (acl_ids->z_fuidp)
zfs_fuid_info_free(acl_ids->z_fuidp);
acl_ids->z_aclp = NULL;
acl_ids->z_fuidp = NULL;
}
boolean_t
zfs_acl_ids_overquota(zfsvfs_t *zv, zfs_acl_ids_t *acl_ids, uint64_t projid)
{
return (zfs_id_overquota(zv, DMU_USERUSED_OBJECT, acl_ids->z_fuid) ||
zfs_id_overquota(zv, DMU_GROUPUSED_OBJECT, acl_ids->z_fgid) ||
(projid != ZFS_DEFAULT_PROJID && projid != ZFS_INVALID_PROJID &&
zfs_id_overquota(zv, DMU_PROJECTUSED_OBJECT, projid)));
}
/*
* Retrieve a file's ACL
*/
int
zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
{
zfs_acl_t *aclp;
ulong_t mask;
int error;
int count = 0;
int largeace = 0;
mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
if (mask == 0)
return (SET_ERROR(ENOSYS));
if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
return (error);
mutex_enter(&zp->z_acl_lock);
if (zp->z_zfsvfs->z_replay == B_FALSE)
ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
if (error != 0) {
mutex_exit(&zp->z_acl_lock);
return (error);
}
/*
* Scan ACL to determine number of ACEs
*/
if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
void *zacep = NULL;
uint64_t who;
uint32_t access_mask;
uint16_t type, iflags;
while ((zacep = zfs_acl_next_ace(aclp, zacep,
&who, &access_mask, &iflags, &type))) {
switch (type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
largeace++;
continue;
default:
count++;
}
}
vsecp->vsa_aclcnt = count;
} else
count = (int)aclp->z_acl_count;
if (mask & VSA_ACECNT) {
vsecp->vsa_aclcnt = count;
}
if (mask & VSA_ACE) {
size_t aclsz;
aclsz = count * sizeof (ace_t) +
sizeof (ace_object_t) * largeace;
vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
vsecp->vsa_aclentsz = aclsz;
if (aclp->z_version == ZFS_ACL_VERSION_FUID)
zfs_copy_fuid_2_ace(zp->z_zfsvfs, aclp, cr,
vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
else {
zfs_acl_node_t *aclnode;
void *start = vsecp->vsa_aclentp;
for (aclnode = list_head(&aclp->z_acl); aclnode;
aclnode = list_next(&aclp->z_acl, aclnode)) {
bcopy(aclnode->z_acldata, start,
aclnode->z_size);
start = (caddr_t)start + aclnode->z_size;
}
ASSERT3U((caddr_t)start - (caddr_t)vsecp->vsa_aclentp,
==, aclp->z_acl_bytes);
}
}
if (mask & VSA_ACE_ACLFLAGS) {
vsecp->vsa_aclflags = 0;
if (zp->z_pflags & ZFS_ACL_DEFAULTED)
vsecp->vsa_aclflags |= ACL_DEFAULTED;
if (zp->z_pflags & ZFS_ACL_PROTECTED)
vsecp->vsa_aclflags |= ACL_PROTECTED;
if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
}
mutex_exit(&zp->z_acl_lock);
return (0);
}
int
zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, umode_t obj_type,
vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
{
zfs_acl_t *aclp;
zfs_acl_node_t *aclnode;
int aclcnt = vsecp->vsa_aclcnt;
int error;
if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
return (SET_ERROR(EINVAL));
aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
aclp->z_hints = 0;
aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
if ((error = zfs_copy_ace_2_oldace(obj_type, aclp,
(ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
aclcnt, &aclnode->z_size)) != 0) {
zfs_acl_free(aclp);
zfs_acl_node_free(aclnode);
return (error);
}
} else {
if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_type, aclp,
vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
&aclnode->z_size, fuidp, cr)) != 0) {
zfs_acl_free(aclp);
zfs_acl_node_free(aclnode);
return (error);
}
}
aclp->z_acl_bytes = aclnode->z_size;
aclnode->z_ace_count = aclcnt;
aclp->z_acl_count = aclcnt;
list_insert_head(&aclp->z_acl, aclnode);
/*
* If flags are being set then add them to z_hints
*/
if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
if (vsecp->vsa_aclflags & ACL_PROTECTED)
aclp->z_hints |= ZFS_ACL_PROTECTED;
if (vsecp->vsa_aclflags & ACL_DEFAULTED)
aclp->z_hints |= ZFS_ACL_DEFAULTED;
if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
}
*zaclp = aclp;
return (0);
}
/*
* Set a file's ACL
*/
int
zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
zilog_t *zilog = zfsvfs->z_log;
ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
dmu_tx_t *tx;
int error;
zfs_acl_t *aclp;
zfs_fuid_info_t *fuidp = NULL;
boolean_t fuid_dirtied;
uint64_t acl_obj;
if (zp->z_zfsvfs->z_replay == B_FALSE)
ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
if (mask == 0)
return (SET_ERROR(ENOSYS));
if (zp->z_pflags & ZFS_IMMUTABLE)
return (SET_ERROR(EPERM));
if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
return (error);
error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, cr, &fuidp,
&aclp);
if (error)
return (error);
/*
* If ACL wide flags aren't being set then preserve any
* existing flags.
*/
if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
aclp->z_hints |=
(zp->z_pflags & V4_ACL_WIDE_FLAGS);
}
top:
mutex_enter(&zp->z_acl_lock);
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
fuid_dirtied = zfsvfs->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
/*
* If old version and ACL won't fit in bonus and we aren't
* upgrading then take out necessary DMU holds
*/
if ((acl_obj = zfs_external_acl(zp)) != 0) {
if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
dmu_tx_hold_free(tx, acl_obj, 0,
DMU_OBJECT_END);
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
aclp->z_acl_bytes);
} else {
dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
}
} else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
}
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_NOWAIT);
if (error) {
mutex_exit(&zp->z_acl_lock);
if (error == ERESTART) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
zfs_acl_free(aclp);
return (error);
}
error = zfs_aclset_common(zp, aclp, cr, tx);
ASSERT0(error);
ASSERT3P(zp->z_acl_cached, ==, NULL);
zp->z_acl_cached = aclp;
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
if (fuidp)
zfs_fuid_info_free(fuidp);
dmu_tx_commit(tx);
mutex_exit(&zp->z_acl_lock);
return (error);
}
/*
* Check accesses of interest (AoI) against attributes of the dataset
* such as read-only. Returns zero if no AoI conflict with dataset
* attributes, otherwise an appropriate errno is returned.
*/
static int
zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
{
if ((v4_mode & WRITE_MASK) &&
(zp->z_zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
(!IS_DEVVP(ZTOV(zp)) ||
(IS_DEVVP(ZTOV(zp)) && (v4_mode & WRITE_MASK_ATTRS)))) {
return (SET_ERROR(EROFS));
}
/*
* Intentionally allow ZFS_READONLY through here.
* See zfs_zaccess_common().
*/
if ((v4_mode & WRITE_MASK_DATA) &&
(zp->z_pflags & ZFS_IMMUTABLE)) {
return (SET_ERROR(EPERM));
}
/*
* In FreeBSD we allow to modify directory's content is ZFS_NOUNLINK
* (sunlnk) is set. We just don't allow directory removal, which is
* handled in zfs_zaccess_delete().
*/
if ((v4_mode & ACE_DELETE) &&
(zp->z_pflags & ZFS_NOUNLINK)) {
return (EPERM);
}
if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
(zp->z_pflags & ZFS_AV_QUARANTINED))) {
return (SET_ERROR(EACCES));
}
return (0);
}
/*
* The primary usage of this function is to loop through all of the
* ACEs in the znode, determining what accesses of interest (AoI) to
* the caller are allowed or denied. The AoI are expressed as bits in
* the working_mode parameter. As each ACE is processed, bits covered
* by that ACE are removed from the working_mode. This removal
* facilitates two things. The first is that when the working mode is
* empty (= 0), we know we've looked at all the AoI. The second is
* that the ACE interpretation rules don't allow a later ACE to undo
* something granted or denied by an earlier ACE. Removing the
* discovered access or denial enforces this rule. At the end of
* processing the ACEs, all AoI that were found to be denied are
* placed into the working_mode, giving the caller a mask of denied
* accesses. Returns:
* 0 if all AoI granted
* EACCESS if the denied mask is non-zero
* other error if abnormal failure (e.g., IO error)
*
* A secondary usage of the function is to determine if any of the
* AoI are granted. If an ACE grants any access in
* the working_mode, we immediately short circuit out of the function.
* This mode is chosen by setting anyaccess to B_TRUE. The
* working_mode is not a denied access mask upon exit if the function
* is used in this manner.
*/
static int
zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
boolean_t anyaccess, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
zfs_acl_t *aclp;
int error;
uid_t uid = crgetuid(cr);
uint64_t who;
uint16_t type, iflags;
uint16_t entry_type;
uint32_t access_mask;
uint32_t deny_mask = 0;
zfs_ace_hdr_t *acep = NULL;
boolean_t checkit;
uid_t gowner;
uid_t fowner;
zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
mutex_enter(&zp->z_acl_lock);
if (zp->z_zfsvfs->z_replay == B_FALSE)
ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
if (error != 0) {
mutex_exit(&zp->z_acl_lock);
return (error);
}
ASSERT3P(zp->z_acl_cached, !=, NULL);
while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
&iflags, &type))) {
uint32_t mask_matched;
if (!zfs_acl_valid_ace_type(type, iflags))
continue;
if (ZTOV(zp)->v_type == VDIR && (iflags & ACE_INHERIT_ONLY_ACE))
continue;
/* Skip ACE if it does not affect any AoI */
mask_matched = (access_mask & *working_mode);
if (!mask_matched)
continue;
entry_type = (iflags & ACE_TYPE_FLAGS);
checkit = B_FALSE;
switch (entry_type) {
case ACE_OWNER:
if (uid == fowner)
checkit = B_TRUE;
break;
case OWNING_GROUP:
who = gowner;
fallthrough;
case ACE_IDENTIFIER_GROUP:
checkit = zfs_groupmember(zfsvfs, who, cr);
break;
case ACE_EVERYONE:
checkit = B_TRUE;
break;
/* USER Entry */
default:
if (entry_type == 0) {
uid_t newid;
newid = zfs_fuid_map_id(zfsvfs, who, cr,
ZFS_ACE_USER);
if (newid != UID_NOBODY &&
uid == newid)
checkit = B_TRUE;
break;
} else {
mutex_exit(&zp->z_acl_lock);
return (SET_ERROR(EIO));
}
}
if (checkit) {
if (type == DENY) {
DTRACE_PROBE3(zfs__ace__denies,
znode_t *, zp,
zfs_ace_hdr_t *, acep,
uint32_t, mask_matched);
deny_mask |= mask_matched;
} else {
DTRACE_PROBE3(zfs__ace__allows,
znode_t *, zp,
zfs_ace_hdr_t *, acep,
uint32_t, mask_matched);
if (anyaccess) {
mutex_exit(&zp->z_acl_lock);
return (0);
}
}
*working_mode &= ~mask_matched;
}
/* Are we done? */
if (*working_mode == 0)
break;
}
mutex_exit(&zp->z_acl_lock);
/* Put the found 'denies' back on the working mode */
if (deny_mask) {
*working_mode |= deny_mask;
return (SET_ERROR(EACCES));
} else if (*working_mode) {
return (-1);
}
return (0);
}
/*
* Return true if any access whatsoever granted, we don't actually
* care what access is granted.
*/
boolean_t
zfs_has_access(znode_t *zp, cred_t *cr)
{
uint32_t have = ACE_ALL_PERMS;
if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
uid_t owner;
owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
return (secpolicy_vnode_any_access(cr, ZTOV(zp), owner) == 0);
}
return (B_TRUE);
}
static int
zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
int err;
*working_mode = v4_mode;
*check_privs = B_TRUE;
/*
* Short circuit empty requests
*/
if (v4_mode == 0 || zfsvfs->z_replay) {
*working_mode = 0;
return (0);
}
if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
*check_privs = B_FALSE;
return (err);
}
/*
* The caller requested that the ACL check be skipped. This
* would only happen if the caller checked VOP_ACCESS() with a
* 32 bit ACE mask and already had the appropriate permissions.
*/
if (skipaclchk) {
*working_mode = 0;
return (0);
}
/*
* Note: ZFS_READONLY represents the "DOS R/O" attribute.
* When that flag is set, we should behave as if write access
* were not granted by anything in the ACL. In particular:
* We _must_ allow writes after opening the file r/w, then
* setting the DOS R/O attribute, and writing some more.
* (Similar to how you can write after fchmod(fd, 0444).)
*
* Therefore ZFS_READONLY is ignored in the dataset check
* above, and checked here as if part of the ACL check.
* Also note: DOS R/O is ignored for directories.
*/
if ((v4_mode & WRITE_MASK_DATA) &&
(ZTOV(zp)->v_type != VDIR) &&
(zp->z_pflags & ZFS_READONLY)) {
return (SET_ERROR(EPERM));
}
return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
}
static int
zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
cred_t *cr)
{
if (*working_mode != ACE_WRITE_DATA)
return (SET_ERROR(EACCES));
return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
check_privs, B_FALSE, cr));
}
/*
* Check if VEXEC is allowed.
*
* This routine is based on zfs_fastaccesschk_execute which has slowpath
* calling zfs_zaccess. This would be incorrect on FreeBSD (see
* zfs_freebsd_access for the difference). Thus this variant let's the
* caller handle the slowpath (if necessary).
*
* On top of that we perform a lockless check for ZFS_NO_EXECS_DENIED.
*
* Safe access to znode_t is provided by the vnode lock.
*/
int
zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
{
boolean_t is_attr;
if (zdp->z_pflags & ZFS_AV_QUARANTINED)
return (1);
is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
(ZTOV(zdp)->v_type == VDIR));
if (is_attr)
return (1);
if (zdp->z_pflags & ZFS_NO_EXECS_DENIED)
return (0);
return (1);
}
/*
* Determine whether Access should be granted/denied.
*
* The least priv subsystem is always consulted as a basic privilege
* can define any form of access.
*/
int
zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
{
uint32_t working_mode;
int error;
int is_attr;
boolean_t check_privs;
znode_t *xzp = NULL;
znode_t *check_zp = zp;
mode_t needed_bits;
uid_t owner;
is_attr = ((zp->z_pflags & ZFS_XATTR) && (ZTOV(zp)->v_type == VDIR));
/*
* In FreeBSD, we don't care about permissions of individual ADS.
* Note that not checking them is not just an optimization - without
* this shortcut, EA operations may bogusly fail with EACCES.
*/
if (zp->z_pflags & ZFS_XATTR)
return (0);
owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
/*
* Map the bits required to the standard vnode flags VREAD|VWRITE|VEXEC
* in needed_bits. Map the bits mapped by working_mode (currently
* missing) in missing_bits.
* Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
* needed_bits.
*/
needed_bits = 0;
working_mode = mode;
if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
owner == crgetuid(cr))
working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
needed_bits |= VREAD;
if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
needed_bits |= VWRITE;
if (working_mode & ACE_EXECUTE)
needed_bits |= VEXEC;
if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
&check_privs, skipaclchk, cr)) == 0) {
if (is_attr)
VN_RELE(ZTOV(xzp));
return (secpolicy_vnode_access2(cr, ZTOV(zp), owner,
needed_bits, needed_bits));
}
if (error && !check_privs) {
if (is_attr)
VN_RELE(ZTOV(xzp));
return (error);
}
if (error && (flags & V_APPEND)) {
error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
}
if (error && check_privs) {
mode_t checkmode = 0;
vnode_t *check_vp = ZTOV(check_zp);
/*
* First check for implicit owner permission on
* read_acl/read_attributes
*/
error = 0;
ASSERT3U(working_mode, !=, 0);
if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
owner == crgetuid(cr)))
working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
checkmode |= VREAD;
if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
checkmode |= VWRITE;
if (working_mode & ACE_EXECUTE)
checkmode |= VEXEC;
error = secpolicy_vnode_access2(cr, check_vp, owner,
needed_bits & ~checkmode, needed_bits);
if (error == 0 && (working_mode & ACE_WRITE_OWNER))
error = secpolicy_vnode_chown(check_vp, cr, owner);
if (error == 0 && (working_mode & ACE_WRITE_ACL))
error = secpolicy_vnode_setdac(check_vp, cr, owner);
if (error == 0 && (working_mode &
(ACE_DELETE|ACE_DELETE_CHILD)))
error = secpolicy_vnode_remove(check_vp, cr);
if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
error = secpolicy_vnode_chown(check_vp, cr, owner);
}
if (error == 0) {
/*
* See if any bits other than those already checked
* for are still present. If so then return EACCES
*/
if (working_mode & ~(ZFS_CHECKED_MASKS)) {
error = SET_ERROR(EACCES);
}
}
} else if (error == 0) {
error = secpolicy_vnode_access2(cr, ZTOV(zp), owner,
needed_bits, needed_bits);
}
if (is_attr)
VN_RELE(ZTOV(xzp));
return (error);
}
/*
* Translate traditional unix VREAD/VWRITE/VEXEC mode into
* NFSv4-style ZFS ACL format and call zfs_zaccess()
*/
int
zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
{
return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
}
/*
* Access function for secpolicy_vnode_setattr
*/
int
zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
{
int v4_mode = zfs_unix_to_v4(mode >> 6);
return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
}
static int
zfs_delete_final_check(znode_t *zp, znode_t *dzp,
mode_t available_perms, cred_t *cr)
{
int error;
uid_t downer;
downer = zfs_fuid_map_id(dzp->z_zfsvfs, dzp->z_uid, cr, ZFS_OWNER);
error = secpolicy_vnode_access2(cr, ZTOV(dzp),
downer, available_perms, VWRITE|VEXEC);
if (error == 0)
error = zfs_sticky_remove_access(dzp, zp, cr);
return (error);
}
/*
* Determine whether Access should be granted/deny, without
* consulting least priv subsystem.
*
* The following chart is the recommended NFSv4 enforcement for
* ability to delete an object.
*
* -------------------------------------------------------
* | Parent Dir | Target Object Permissions |
* | permissions | |
* -------------------------------------------------------
* | | ACL Allows | ACL Denies| Delete |
* | | Delete | Delete | unspecified|
* -------------------------------------------------------
* | ACL Allows | Permit | Permit | Permit |
* | DELETE_CHILD | |
* -------------------------------------------------------
* | ACL Denies | Permit | Deny | Deny |
* | DELETE_CHILD | | | |
* -------------------------------------------------------
* | ACL specifies | | | |
* | only allow | Permit | Permit | Permit |
* | write and | | | |
* | execute | | | |
* -------------------------------------------------------
* | ACL denies | | | |
* | write and | Permit | Deny | Deny |
* | execute | | | |
* -------------------------------------------------------
* ^
* |
* No search privilege, can't even look up file?
*
*/
int
zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
{
uint32_t dzp_working_mode = 0;
uint32_t zp_working_mode = 0;
int dzp_error, zp_error;
mode_t available_perms;
boolean_t dzpcheck_privs = B_TRUE;
boolean_t zpcheck_privs = B_TRUE;
/*
* We want specific DELETE permissions to
* take precedence over WRITE/EXECUTE. We don't
* want an ACL such as this to mess us up.
* user:joe:write_data:deny,user:joe:delete:allow
*
* However, deny permissions may ultimately be overridden
* by secpolicy_vnode_access().
*
* We will ask for all of the necessary permissions and then
* look at the working modes from the directory and target object
* to determine what was found.
*/
if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
return (SET_ERROR(EPERM));
/*
* First row
* If the directory permissions allow the delete, we are done.
*/
if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
&dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
return (0);
/*
* If target object has delete permission then we are done
*/
if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
&zpcheck_privs, B_FALSE, cr)) == 0)
return (0);
ASSERT(dzp_error);
ASSERT(zp_error);
if (!dzpcheck_privs)
return (dzp_error);
if (!zpcheck_privs)
return (zp_error);
/*
* Second row
*
* If directory returns EACCES then delete_child was denied
* due to deny delete_child. In this case send the request through
* secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
* since that *could* allow the delete based on write/execute permission
* and we want delete permissions to override write/execute.
*/
if (dzp_error == EACCES) {
/* XXXPJD: s/dzp/zp/ ? */
return (secpolicy_vnode_remove(ZTOV(dzp), cr));
}
/*
* Third Row
* only need to see if we have write/execute on directory.
*/
dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
&dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
if (dzp_error != 0 && !dzpcheck_privs)
return (dzp_error);
/*
* Fourth row
*/
available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : VWRITE;
available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : VEXEC;
return (zfs_delete_final_check(zp, dzp, available_perms, cr));
}
int
zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
znode_t *tzp, cred_t *cr)
{
int add_perm;
int error;
if (szp->z_pflags & ZFS_AV_QUARANTINED)
return (SET_ERROR(EACCES));
add_perm = (ZTOV(szp)->v_type == VDIR) ?
ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
/*
* Rename permissions are combination of delete permission +
* add file/subdir permission.
*
* BSD operating systems also require write permission
* on the directory being moved from one parent directory
* to another.
*/
if (ZTOV(szp)->v_type == VDIR && ZTOV(sdzp) != ZTOV(tdzp)) {
if ((error = zfs_zaccess(szp, ACE_WRITE_DATA, 0, B_FALSE, cr)))
return (error);
}
/*
* first make sure we do the delete portion.
*
* If that succeeds then check for add_file/add_subdir permissions
*/
if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
return (error);
/*
* If we have a tzp, see if we can delete it?
*/
if (tzp && (error = zfs_zaccess_delete(tdzp, tzp, cr)))
return (error);
/*
* Now check for add permissions
*/
error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
return (error);
}
diff --git a/module/os/freebsd/zfs/zfs_debug.c b/module/os/freebsd/zfs/zfs_debug.c
index dad342b06fc1..0ff22cfe79ec 100644
--- a/module/os/freebsd/zfs/zfs_debug.c
+++ b/module/os/freebsd/zfs/zfs_debug.c
@@ -1,254 +1,254 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/kstat.h>
typedef struct zfs_dbgmsg {
list_node_t zdm_node;
time_t zdm_timestamp;
int zdm_size;
char zdm_msg[1]; /* variable length allocation */
} zfs_dbgmsg_t;
-list_t zfs_dbgmsgs;
-int zfs_dbgmsg_size = 0;
-kmutex_t zfs_dbgmsgs_lock;
+static list_t zfs_dbgmsgs;
+static int zfs_dbgmsg_size = 0;
+static kmutex_t zfs_dbgmsgs_lock;
int zfs_dbgmsg_maxsize = 4<<20; /* 4MB */
-kstat_t *zfs_dbgmsg_kstat;
+static kstat_t *zfs_dbgmsg_kstat;
/*
* Internal ZFS debug messages are enabled by default.
*
* # Print debug messages as they're logged
* dtrace -n 'zfs-dbgmsg { print(stringof(arg0)); }'
*
* # Print all logged dbgmsg entries
* sysctl kstat.zfs.misc.dbgmsg
*
* # Disable the kernel debug message log.
* sysctl vfs.zfs.dbgmsg_enable=0
*/
-int zfs_dbgmsg_enable = 1;
+int zfs_dbgmsg_enable = B_TRUE;
static int
zfs_dbgmsg_headers(char *buf, size_t size)
{
(void) snprintf(buf, size, "%-12s %-8s\n", "timestamp", "message");
return (0);
}
static int
zfs_dbgmsg_data(char *buf, size_t size, void *data)
{
zfs_dbgmsg_t *zdm = (zfs_dbgmsg_t *)data;
(void) snprintf(buf, size, "%-12llu %-s\n",
(u_longlong_t)zdm->zdm_timestamp, zdm->zdm_msg);
return (0);
}
static void *
zfs_dbgmsg_addr(kstat_t *ksp, loff_t n)
{
zfs_dbgmsg_t *zdm = (zfs_dbgmsg_t *)ksp->ks_private;
ASSERT(MUTEX_HELD(&zfs_dbgmsgs_lock));
if (n == 0)
ksp->ks_private = list_head(&zfs_dbgmsgs);
else if (zdm)
ksp->ks_private = list_next(&zfs_dbgmsgs, zdm);
return (ksp->ks_private);
}
static void
zfs_dbgmsg_purge(int max_size)
{
zfs_dbgmsg_t *zdm;
int size;
ASSERT(MUTEX_HELD(&zfs_dbgmsgs_lock));
while (zfs_dbgmsg_size > max_size) {
zdm = list_remove_head(&zfs_dbgmsgs);
if (zdm == NULL)
return;
size = zdm->zdm_size;
kmem_free(zdm, size);
zfs_dbgmsg_size -= size;
}
}
static int
zfs_dbgmsg_update(kstat_t *ksp, int rw)
{
if (rw == KSTAT_WRITE)
zfs_dbgmsg_purge(0);
return (0);
}
void
zfs_dbgmsg_init(void)
{
list_create(&zfs_dbgmsgs, sizeof (zfs_dbgmsg_t),
offsetof(zfs_dbgmsg_t, zdm_node));
mutex_init(&zfs_dbgmsgs_lock, NULL, MUTEX_DEFAULT, NULL);
zfs_dbgmsg_kstat = kstat_create("zfs", 0, "dbgmsg", "misc",
KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);
if (zfs_dbgmsg_kstat) {
zfs_dbgmsg_kstat->ks_lock = &zfs_dbgmsgs_lock;
zfs_dbgmsg_kstat->ks_ndata = UINT32_MAX;
zfs_dbgmsg_kstat->ks_private = NULL;
zfs_dbgmsg_kstat->ks_update = zfs_dbgmsg_update;
kstat_set_raw_ops(zfs_dbgmsg_kstat, zfs_dbgmsg_headers,
zfs_dbgmsg_data, zfs_dbgmsg_addr);
kstat_install(zfs_dbgmsg_kstat);
}
}
void
zfs_dbgmsg_fini(void)
{
if (zfs_dbgmsg_kstat)
kstat_delete(zfs_dbgmsg_kstat);
/*
* TODO - decide how to make this permanent
*/
#ifdef _KERNEL
mutex_enter(&zfs_dbgmsgs_lock);
zfs_dbgmsg_purge(0);
mutex_exit(&zfs_dbgmsgs_lock);
mutex_destroy(&zfs_dbgmsgs_lock);
#endif
}
void
__zfs_dbgmsg(char *buf)
{
zfs_dbgmsg_t *zdm;
int size;
DTRACE_PROBE1(zfs__dbgmsg, char *, buf);
size = sizeof (zfs_dbgmsg_t) + strlen(buf);
zdm = kmem_zalloc(size, KM_SLEEP);
zdm->zdm_size = size;
zdm->zdm_timestamp = gethrestime_sec();
strcpy(zdm->zdm_msg, buf);
mutex_enter(&zfs_dbgmsgs_lock);
list_insert_tail(&zfs_dbgmsgs, zdm);
zfs_dbgmsg_size += size;
zfs_dbgmsg_purge(MAX(zfs_dbgmsg_maxsize, 0));
mutex_exit(&zfs_dbgmsgs_lock);
}
void
__set_error(const char *file, const char *func, int line, int err)
{
/*
* To enable this:
*
* $ echo 512 >/sys/module/zfs/parameters/zfs_flags
*/
if (zfs_flags & ZFS_DEBUG_SET_ERROR)
__dprintf(B_FALSE, file, func, line, "error %lu", (ulong_t)err);
}
#ifdef _KERNEL
void
__dprintf(boolean_t dprint, const char *file, const char *func,
int line, const char *fmt, ...)
{
const char *newfile;
va_list adx;
size_t size;
char *buf;
char *nl;
int i;
size = 1024;
buf = kmem_alloc(size, KM_SLEEP);
/*
* Get rid of annoying prefix to filename.
*/
newfile = strrchr(file, '/');
if (newfile != NULL) {
newfile = newfile + 1; /* Get rid of leading / */
} else {
newfile = file;
}
i = snprintf(buf, size, "%s:%d:%s(): ", newfile, line, func);
if (i < size) {
va_start(adx, fmt);
(void) vsnprintf(buf + i, size - i, fmt, adx);
va_end(adx);
}
/*
* Get rid of trailing newline.
*/
nl = strrchr(buf, '\n');
if (nl != NULL)
*nl = '\0';
__zfs_dbgmsg(buf);
kmem_free(buf, size);
}
#else
void
zfs_dbgmsg_print(const char *tag)
{
zfs_dbgmsg_t *zdm;
(void) printf("ZFS_DBGMSG(%s):\n", tag);
mutex_enter(&zfs_dbgmsgs_lock);
for (zdm = list_head(&zfs_dbgmsgs); zdm;
zdm = list_next(&zfs_dbgmsgs, zdm))
(void) printf("%s\n", zdm->zdm_msg);
mutex_exit(&zfs_dbgmsgs_lock);
}
#endif /* _KERNEL */
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, dbgmsg_enable, INT, ZMOD_RW,
"Enable ZFS debug message log");
ZFS_MODULE_PARAM(zfs, zfs_, dbgmsg_maxsize, INT, ZMOD_RW,
"Maximum ZFS debug log size");
/* END CSTYLED */
diff --git a/module/os/freebsd/zfs/zio_crypt.c b/module/os/freebsd/zfs/zio_crypt.c
index 038f2250e775..c55c1ac25117 100644
--- a/module/os/freebsd/zfs/zio_crypt.c
+++ b/module/os/freebsd/zfs/zio_crypt.c
@@ -1,1823 +1,1814 @@
/*
* CDDL HEADER START
*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2017, Datto, Inc. All rights reserved.
*/
#include <sys/zio_crypt.h>
#include <sys/dmu.h>
#include <sys/dmu_objset.h>
#include <sys/dnode.h>
#include <sys/fs/zfs.h>
#include <sys/zio.h>
#include <sys/zil.h>
#include <sys/sha2.h>
#include <sys/hkdf.h>
/*
* This file is responsible for handling all of the details of generating
* encryption parameters and performing encryption and authentication.
*
* BLOCK ENCRYPTION PARAMETERS:
* Encryption /Authentication Algorithm Suite (crypt):
* The encryption algorithm, mode, and key length we are going to use. We
* currently support AES in either GCM or CCM modes with 128, 192, and 256 bit
* keys. All authentication is currently done with SHA512-HMAC.
*
* Plaintext:
* The unencrypted data that we want to encrypt.
*
* Initialization Vector (IV):
* An initialization vector for the encryption algorithms. This is used to
* "tweak" the encryption algorithms so that two blocks of the same data are
* encrypted into different ciphertext outputs, thus obfuscating block patterns.
* The supported encryption modes (AES-GCM and AES-CCM) require that an IV is
* never reused with the same encryption key. This value is stored unencrypted
* and must simply be provided to the decryption function. We use a 96 bit IV
* (as recommended by NIST) for all block encryption. For non-dedup blocks we
* derive the IV randomly. The first 64 bits of the IV are stored in the second
* word of DVA[2] and the remaining 32 bits are stored in the upper 32 bits of
* blk_fill. This is safe because encrypted blocks can't use the upper 32 bits
* of blk_fill. We only encrypt level 0 blocks, which normally have a fill count
* of 1. The only exception is for DMU_OT_DNODE objects, where the fill count of
* level 0 blocks is the number of allocated dnodes in that block. The on-disk
* format supports at most 2^15 slots per L0 dnode block, because the maximum
* block size is 16MB (2^24). In either case, for level 0 blocks this number
* will still be smaller than UINT32_MAX so it is safe to store the IV in the
* top 32 bits of blk_fill, while leaving the bottom 32 bits of the fill count
* for the dnode code.
*
* Master key:
* This is the most important secret data of an encrypted dataset. It is used
* along with the salt to generate that actual encryption keys via HKDF. We
* do not use the master key to directly encrypt any data because there are
* theoretical limits on how much data can actually be safely encrypted with
* any encryption mode. The master key is stored encrypted on disk with the
* user's wrapping key. Its length is determined by the encryption algorithm.
* For details on how this is stored see the block comment in dsl_crypt.c
*
* Salt:
* Used as an input to the HKDF function, along with the master key. We use a
* 64 bit salt, stored unencrypted in the first word of DVA[2]. Any given salt
* can be used for encrypting many blocks, so we cache the current salt and the
* associated derived key in zio_crypt_t so we do not need to derive it again
* needlessly.
*
* Encryption Key:
* A secret binary key, generated from an HKDF function used to encrypt and
* decrypt data.
*
* Message Authentication Code (MAC)
* The MAC is an output of authenticated encryption modes such as AES-GCM and
* AES-CCM. Its purpose is to ensure that an attacker cannot modify encrypted
* data on disk and return garbage to the application. Effectively, it is a
* checksum that can not be reproduced by an attacker. We store the MAC in the
* second 128 bits of blk_cksum, leaving the first 128 bits for a truncated
* regular checksum of the ciphertext which can be used for scrubbing.
*
* OBJECT AUTHENTICATION:
* Some object types, such as DMU_OT_MASTER_NODE cannot be encrypted because
* they contain some info that always needs to be readable. To prevent this
* data from being altered, we authenticate this data using SHA512-HMAC. This
* will produce a MAC (similar to the one produced via encryption) which can
* be used to verify the object was not modified. HMACs do not require key
* rotation or IVs, so we can keep up to the full 3 copies of authenticated
* data.
*
* ZIL ENCRYPTION:
* ZIL blocks have their bp written to disk ahead of the associated data, so we
* cannot store the MAC there as we normally do. For these blocks the MAC is
* stored in the embedded checksum within the zil_chain_t header. The salt and
* IV are generated for the block on bp allocation instead of at encryption
* time. In addition, ZIL blocks have some pieces that must be left in plaintext
* for claiming even though all of the sensitive user data still needs to be
* encrypted. The function zio_crypt_init_uios_zil() handles parsing which
* pieces of the block need to be encrypted. All data that is not encrypted is
* authenticated using the AAD mechanisms that the supported encryption modes
* provide for. In order to preserve the semantics of the ZIL for encrypted
* datasets, the ZIL is not protected at the objset level as described below.
*
* DNODE ENCRYPTION:
* Similarly to ZIL blocks, the core part of each dnode_phys_t needs to be left
* in plaintext for scrubbing and claiming, but the bonus buffers might contain
* sensitive user data. The function zio_crypt_init_uios_dnode() handles parsing
* which pieces of the block need to be encrypted. For more details about
* dnode authentication and encryption, see zio_crypt_init_uios_dnode().
*
* OBJECT SET AUTHENTICATION:
* Up to this point, everything we have encrypted and authenticated has been
* at level 0 (or -2 for the ZIL). If we did not do any further work the
* on-disk format would be susceptible to attacks that deleted or rearranged
* the order of level 0 blocks. Ideally, the cleanest solution would be to
* maintain a tree of authentication MACs going up the bp tree. However, this
* presents a problem for raw sends. Send files do not send information about
* indirect blocks so there would be no convenient way to transfer the MACs and
* they cannot be recalculated on the receive side without the master key which
* would defeat one of the purposes of raw sends in the first place. Instead,
* for the indirect levels of the bp tree, we use a regular SHA512 of the MACs
* from the level below. We also include some portable fields from blk_prop such
* as the lsize and compression algorithm to prevent the data from being
* misinterpreted.
*
* At the objset level, we maintain 2 separate 256 bit MACs in the
* objset_phys_t. The first one is "portable" and is the logical root of the
* MAC tree maintained in the metadnode's bps. The second, is "local" and is
* used as the root MAC for the user accounting objects, which are also not
* transferred via "zfs send". The portable MAC is sent in the DRR_BEGIN payload
* of the send file. The useraccounting code ensures that the useraccounting
* info is not present upon a receive, so the local MAC can simply be cleared
* out at that time. For more info about objset_phys_t authentication, see
* zio_crypt_do_objset_hmacs().
*
* CONSIDERATIONS FOR DEDUP:
* In order for dedup to work, blocks that we want to dedup with one another
* need to use the same IV and encryption key, so that they will have the same
* ciphertext. Normally, one should never reuse an IV with the same encryption
* key or else AES-GCM and AES-CCM can both actually leak the plaintext of both
* blocks. In this case, however, since we are using the same plaintext as
* well all that we end up with is a duplicate of the original ciphertext we
* already had. As a result, an attacker with read access to the raw disk will
* be able to tell which blocks are the same but this information is given away
* by dedup anyway. In order to get the same IVs and encryption keys for
* equivalent blocks of data we use an HMAC of the plaintext. We use an HMAC
* here so that a reproducible checksum of the plaintext is never available to
* the attacker. The HMAC key is kept alongside the master key, encrypted on
* disk. The first 64 bits of the HMAC are used in place of the random salt, and
* the next 96 bits are used as the IV. As a result of this mechanism, dedup
* will only work within a clone family since encrypted dedup requires use of
* the same master and HMAC keys.
*/
/*
* After encrypting many blocks with the same key we may start to run up
* against the theoretical limits of how much data can securely be encrypted
* with a single key using the supported encryption modes. The most obvious
* limitation is that our risk of generating 2 equivalent 96 bit IVs increases
* the more IVs we generate (which both GCM and CCM modes strictly forbid).
* This risk actually grows surprisingly quickly over time according to the
* Birthday Problem. With a total IV space of 2^(96 bits), and assuming we have
* generated n IVs with a cryptographically secure RNG, the approximate
* probability p(n) of a collision is given as:
*
* p(n) ~= e^(-n*(n-1)/(2*(2^96)))
*
* [http://www.math.cornell.edu/~mec/2008-2009/TianyiZheng/Birthday.html]
*
* Assuming that we want to ensure that p(n) never goes over 1 / 1 trillion
* we must not write more than 398,065,730 blocks with the same encryption key.
* Therefore, we rotate our keys after 400,000,000 blocks have been written by
* generating a new random 64 bit salt for our HKDF encryption key generation
* function.
*/
#define ZFS_KEY_MAX_SALT_USES_DEFAULT 400000000
#define ZFS_CURRENT_MAX_SALT_USES \
(MIN(zfs_key_max_salt_uses, ZFS_KEY_MAX_SALT_USES_DEFAULT))
-unsigned long zfs_key_max_salt_uses = ZFS_KEY_MAX_SALT_USES_DEFAULT;
-
-/*
- * Set to a nonzero value to cause zio_do_crypt_uio() to fail 1/this many
- * calls, to test decryption error handling code paths.
- */
-uint64_t zio_decrypt_fail_fraction = 0;
+static unsigned long zfs_key_max_salt_uses = ZFS_KEY_MAX_SALT_USES_DEFAULT;
typedef struct blkptr_auth_buf {
uint64_t bab_prop; /* blk_prop - portable mask */
uint8_t bab_mac[ZIO_DATA_MAC_LEN]; /* MAC from blk_cksum */
uint64_t bab_pad; /* reserved for future use */
} blkptr_auth_buf_t;
-zio_crypt_info_t zio_crypt_table[ZIO_CRYPT_FUNCTIONS] = {
+const zio_crypt_info_t zio_crypt_table[ZIO_CRYPT_FUNCTIONS] = {
{"", ZC_TYPE_NONE, 0, "inherit"},
{"", ZC_TYPE_NONE, 0, "on"},
{"", ZC_TYPE_NONE, 0, "off"},
{SUN_CKM_AES_CCM, ZC_TYPE_CCM, 16, "aes-128-ccm"},
{SUN_CKM_AES_CCM, ZC_TYPE_CCM, 24, "aes-192-ccm"},
{SUN_CKM_AES_CCM, ZC_TYPE_CCM, 32, "aes-256-ccm"},
{SUN_CKM_AES_GCM, ZC_TYPE_GCM, 16, "aes-128-gcm"},
{SUN_CKM_AES_GCM, ZC_TYPE_GCM, 24, "aes-192-gcm"},
{SUN_CKM_AES_GCM, ZC_TYPE_GCM, 32, "aes-256-gcm"}
};
static void
zio_crypt_key_destroy_early(zio_crypt_key_t *key)
{
rw_destroy(&key->zk_salt_lock);
/* free crypto templates */
bzero(&key->zk_session, sizeof (key->zk_session));
/* zero out sensitive data */
bzero(key, sizeof (zio_crypt_key_t));
}
void
zio_crypt_key_destroy(zio_crypt_key_t *key)
{
freebsd_crypt_freesession(&key->zk_session);
zio_crypt_key_destroy_early(key);
}
int
zio_crypt_key_init(uint64_t crypt, zio_crypt_key_t *key)
{
int ret;
crypto_mechanism_t mech __unused;
uint_t keydata_len;
- zio_crypt_info_t *ci = NULL;
+ const zio_crypt_info_t *ci = NULL;
ASSERT3P(key, !=, NULL);
ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);
ci = &zio_crypt_table[crypt];
if (ci->ci_crypt_type != ZC_TYPE_GCM &&
ci->ci_crypt_type != ZC_TYPE_CCM)
return (ENOTSUP);
keydata_len = zio_crypt_table[crypt].ci_keylen;
bzero(key, sizeof (zio_crypt_key_t));
rw_init(&key->zk_salt_lock, NULL, RW_DEFAULT, NULL);
/* fill keydata buffers and salt with random data */
ret = random_get_bytes((uint8_t *)&key->zk_guid, sizeof (uint64_t));
if (ret != 0)
goto error;
ret = random_get_bytes(key->zk_master_keydata, keydata_len);
if (ret != 0)
goto error;
ret = random_get_bytes(key->zk_hmac_keydata, SHA512_HMAC_KEYLEN);
if (ret != 0)
goto error;
ret = random_get_bytes(key->zk_salt, ZIO_DATA_SALT_LEN);
if (ret != 0)
goto error;
/* derive the current key from the master key */
ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
key->zk_salt, ZIO_DATA_SALT_LEN, key->zk_current_keydata,
keydata_len);
if (ret != 0)
goto error;
/* initialize keys for the ICP */
key->zk_current_key.ck_format = CRYPTO_KEY_RAW;
key->zk_current_key.ck_data = key->zk_current_keydata;
key->zk_current_key.ck_length = CRYPTO_BYTES2BITS(keydata_len);
key->zk_hmac_key.ck_format = CRYPTO_KEY_RAW;
key->zk_hmac_key.ck_data = &key->zk_hmac_key;
key->zk_hmac_key.ck_length = CRYPTO_BYTES2BITS(SHA512_HMAC_KEYLEN);
ci = &zio_crypt_table[crypt];
if (ci->ci_crypt_type != ZC_TYPE_GCM &&
ci->ci_crypt_type != ZC_TYPE_CCM)
return (ENOTSUP);
ret = freebsd_crypt_newsession(&key->zk_session, ci,
&key->zk_current_key);
if (ret)
goto error;
key->zk_crypt = crypt;
key->zk_version = ZIO_CRYPT_KEY_CURRENT_VERSION;
key->zk_salt_count = 0;
return (0);
error:
zio_crypt_key_destroy_early(key);
return (ret);
}
static int
zio_crypt_key_change_salt(zio_crypt_key_t *key)
{
int ret = 0;
uint8_t salt[ZIO_DATA_SALT_LEN];
crypto_mechanism_t mech __unused;
uint_t keydata_len = zio_crypt_table[key->zk_crypt].ci_keylen;
/* generate a new salt */
ret = random_get_bytes(salt, ZIO_DATA_SALT_LEN);
if (ret != 0)
goto error;
rw_enter(&key->zk_salt_lock, RW_WRITER);
/* someone beat us to the salt rotation, just unlock and return */
if (key->zk_salt_count < ZFS_CURRENT_MAX_SALT_USES)
goto out_unlock;
/* derive the current key from the master key and the new salt */
ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
salt, ZIO_DATA_SALT_LEN, key->zk_current_keydata, keydata_len);
if (ret != 0)
goto out_unlock;
/* assign the salt and reset the usage count */
bcopy(salt, key->zk_salt, ZIO_DATA_SALT_LEN);
key->zk_salt_count = 0;
freebsd_crypt_freesession(&key->zk_session);
ret = freebsd_crypt_newsession(&key->zk_session,
&zio_crypt_table[key->zk_crypt], &key->zk_current_key);
if (ret != 0)
goto out_unlock;
rw_exit(&key->zk_salt_lock);
return (0);
out_unlock:
rw_exit(&key->zk_salt_lock);
error:
return (ret);
}
/* See comment above zfs_key_max_salt_uses definition for details */
int
zio_crypt_key_get_salt(zio_crypt_key_t *key, uint8_t *salt)
{
int ret;
boolean_t salt_change;
rw_enter(&key->zk_salt_lock, RW_READER);
bcopy(key->zk_salt, salt, ZIO_DATA_SALT_LEN);
salt_change = (atomic_inc_64_nv(&key->zk_salt_count) >=
ZFS_CURRENT_MAX_SALT_USES);
rw_exit(&key->zk_salt_lock);
if (salt_change) {
ret = zio_crypt_key_change_salt(key);
if (ret != 0)
goto error;
}
return (0);
error:
return (ret);
}
void *failed_decrypt_buf;
int failed_decrypt_size;
/*
* This function handles all encryption and decryption in zfs. When
* encrypting it expects puio to reference the plaintext and cuio to
* reference the ciphertext. cuio must have enough space for the
* ciphertext + room for a MAC. datalen should be the length of the
* plaintext / ciphertext alone.
*/
/*
* The implementation for FreeBSD's OpenCrypto.
*
* The big difference between ICP and FOC is that FOC uses a single
* buffer for input and output. This means that (for AES-GCM, the
* only one supported right now) the source must be copied into the
* destination, and the destination must have the AAD, and the tag/MAC,
* already associated with it. (Both implementations can use a uio.)
*
* Since the auth data is part of the iovec array, all we need to know
* is the length: 0 means there's no AAD.
*
*/
static int
zio_do_crypt_uio_opencrypto(boolean_t encrypt, freebsd_crypt_session_t *sess,
uint64_t crypt, crypto_key_t *key, uint8_t *ivbuf, uint_t datalen,
zfs_uio_t *uio, uint_t auth_len)
{
- zio_crypt_info_t *ci;
- int ret;
-
- ci = &zio_crypt_table[crypt];
+ const zio_crypt_info_t *ci = &zio_crypt_table[crypt];
if (ci->ci_crypt_type != ZC_TYPE_GCM &&
ci->ci_crypt_type != ZC_TYPE_CCM)
return (ENOTSUP);
- ret = freebsd_crypt_uio(encrypt, sess, ci, uio, key, ivbuf,
+ int ret = freebsd_crypt_uio(encrypt, sess, ci, uio, key, ivbuf,
datalen, auth_len);
if (ret != 0) {
#ifdef FCRYPTO_DEBUG
printf("%s(%d): Returning error %s\n",
__FUNCTION__, __LINE__, encrypt ? "EIO" : "ECKSUM");
#endif
ret = SET_ERROR(encrypt ? EIO : ECKSUM);
}
return (ret);
}
int
zio_crypt_key_wrap(crypto_key_t *cwkey, zio_crypt_key_t *key, uint8_t *iv,
uint8_t *mac, uint8_t *keydata_out, uint8_t *hmac_keydata_out)
{
int ret;
uint64_t aad[3];
/*
* With OpenCrypto in FreeBSD, the same buffer is used for
* input and output. Also, the AAD (for AES-GMC at least)
* needs to logically go in front.
*/
zfs_uio_t cuio;
struct uio cuio_s;
iovec_t iovecs[4];
uint64_t crypt = key->zk_crypt;
uint_t enc_len, keydata_len, aad_len;
ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);
ASSERT3U(cwkey->ck_format, ==, CRYPTO_KEY_RAW);
zfs_uio_init(&cuio, &cuio_s);
keydata_len = zio_crypt_table[crypt].ci_keylen;
/* generate iv for wrapping the master and hmac key */
ret = random_get_pseudo_bytes(iv, WRAPPING_IV_LEN);
if (ret != 0)
goto error;
/*
* Since we only support one buffer, we need to copy
* the plain text (source) to the cipher buffer (dest).
* We set iovecs[0] -- the authentication data -- below.
*/
bcopy((void*)key->zk_master_keydata, keydata_out, keydata_len);
bcopy((void*)key->zk_hmac_keydata, hmac_keydata_out,
SHA512_HMAC_KEYLEN);
iovecs[1].iov_base = keydata_out;
iovecs[1].iov_len = keydata_len;
iovecs[2].iov_base = hmac_keydata_out;
iovecs[2].iov_len = SHA512_HMAC_KEYLEN;
iovecs[3].iov_base = mac;
iovecs[3].iov_len = WRAPPING_MAC_LEN;
/*
* Although we don't support writing to the old format, we do
* support rewrapping the key so that the user can move and
* quarantine datasets on the old format.
*/
if (key->zk_version == 0) {
aad_len = sizeof (uint64_t);
aad[0] = LE_64(key->zk_guid);
} else {
ASSERT3U(key->zk_version, ==, ZIO_CRYPT_KEY_CURRENT_VERSION);
aad_len = sizeof (uint64_t) * 3;
aad[0] = LE_64(key->zk_guid);
aad[1] = LE_64(crypt);
aad[2] = LE_64(key->zk_version);
}
iovecs[0].iov_base = aad;
iovecs[0].iov_len = aad_len;
enc_len = zio_crypt_table[crypt].ci_keylen + SHA512_HMAC_KEYLEN;
GET_UIO_STRUCT(&cuio)->uio_iov = iovecs;
zfs_uio_iovcnt(&cuio) = 4;
zfs_uio_segflg(&cuio) = UIO_SYSSPACE;
/* encrypt the keys and store the resulting ciphertext and mac */
ret = zio_do_crypt_uio_opencrypto(B_TRUE, NULL, crypt, cwkey,
iv, enc_len, &cuio, aad_len);
if (ret != 0)
goto error;
return (0);
error:
return (ret);
}
int
zio_crypt_key_unwrap(crypto_key_t *cwkey, uint64_t crypt, uint64_t version,
uint64_t guid, uint8_t *keydata, uint8_t *hmac_keydata, uint8_t *iv,
uint8_t *mac, zio_crypt_key_t *key)
{
int ret;
uint64_t aad[3];
/*
* With OpenCrypto in FreeBSD, the same buffer is used for
* input and output. Also, the AAD (for AES-GMC at least)
* needs to logically go in front.
*/
zfs_uio_t cuio;
struct uio cuio_s;
iovec_t iovecs[4];
void *src, *dst;
uint_t enc_len, keydata_len, aad_len;
ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);
ASSERT3U(cwkey->ck_format, ==, CRYPTO_KEY_RAW);
keydata_len = zio_crypt_table[crypt].ci_keylen;
rw_init(&key->zk_salt_lock, NULL, RW_DEFAULT, NULL);
zfs_uio_init(&cuio, &cuio_s);
/*
* Since we only support one buffer, we need to copy
* the encrypted buffer (source) to the plain buffer
* (dest). We set iovecs[0] -- the authentication data --
* below.
*/
dst = key->zk_master_keydata;
src = keydata;
bcopy(src, dst, keydata_len);
dst = key->zk_hmac_keydata;
src = hmac_keydata;
bcopy(src, dst, SHA512_HMAC_KEYLEN);
iovecs[1].iov_base = key->zk_master_keydata;
iovecs[1].iov_len = keydata_len;
iovecs[2].iov_base = key->zk_hmac_keydata;
iovecs[2].iov_len = SHA512_HMAC_KEYLEN;
iovecs[3].iov_base = mac;
iovecs[3].iov_len = WRAPPING_MAC_LEN;
if (version == 0) {
aad_len = sizeof (uint64_t);
aad[0] = LE_64(guid);
} else {
ASSERT3U(version, ==, ZIO_CRYPT_KEY_CURRENT_VERSION);
aad_len = sizeof (uint64_t) * 3;
aad[0] = LE_64(guid);
aad[1] = LE_64(crypt);
aad[2] = LE_64(version);
}
enc_len = keydata_len + SHA512_HMAC_KEYLEN;
iovecs[0].iov_base = aad;
iovecs[0].iov_len = aad_len;
GET_UIO_STRUCT(&cuio)->uio_iov = iovecs;
zfs_uio_iovcnt(&cuio) = 4;
zfs_uio_segflg(&cuio) = UIO_SYSSPACE;
/* decrypt the keys and store the result in the output buffers */
ret = zio_do_crypt_uio_opencrypto(B_FALSE, NULL, crypt, cwkey,
iv, enc_len, &cuio, aad_len);
if (ret != 0)
goto error;
/* generate a fresh salt */
ret = random_get_bytes(key->zk_salt, ZIO_DATA_SALT_LEN);
if (ret != 0)
goto error;
/* derive the current key from the master key */
ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
key->zk_salt, ZIO_DATA_SALT_LEN, key->zk_current_keydata,
keydata_len);
if (ret != 0)
goto error;
/* initialize keys for ICP */
key->zk_current_key.ck_format = CRYPTO_KEY_RAW;
key->zk_current_key.ck_data = key->zk_current_keydata;
key->zk_current_key.ck_length = CRYPTO_BYTES2BITS(keydata_len);
key->zk_hmac_key.ck_format = CRYPTO_KEY_RAW;
key->zk_hmac_key.ck_data = key->zk_hmac_keydata;
key->zk_hmac_key.ck_length = CRYPTO_BYTES2BITS(SHA512_HMAC_KEYLEN);
ret = freebsd_crypt_newsession(&key->zk_session,
&zio_crypt_table[crypt], &key->zk_current_key);
if (ret != 0)
goto error;
key->zk_crypt = crypt;
key->zk_version = version;
key->zk_guid = guid;
key->zk_salt_count = 0;
return (0);
error:
zio_crypt_key_destroy_early(key);
return (ret);
}
int
zio_crypt_generate_iv(uint8_t *ivbuf)
{
int ret;
/* randomly generate the IV */
ret = random_get_pseudo_bytes(ivbuf, ZIO_DATA_IV_LEN);
if (ret != 0)
goto error;
return (0);
error:
bzero(ivbuf, ZIO_DATA_IV_LEN);
return (ret);
}
int
zio_crypt_do_hmac(zio_crypt_key_t *key, uint8_t *data, uint_t datalen,
uint8_t *digestbuf, uint_t digestlen)
{
uint8_t raw_digestbuf[SHA512_DIGEST_LENGTH];
ASSERT3U(digestlen, <=, SHA512_DIGEST_LENGTH);
crypto_mac(&key->zk_hmac_key, data, datalen,
raw_digestbuf, SHA512_DIGEST_LENGTH);
bcopy(raw_digestbuf, digestbuf, digestlen);
return (0);
}
int
zio_crypt_generate_iv_salt_dedup(zio_crypt_key_t *key, uint8_t *data,
uint_t datalen, uint8_t *ivbuf, uint8_t *salt)
{
int ret;
uint8_t digestbuf[SHA512_DIGEST_LENGTH];
ret = zio_crypt_do_hmac(key, data, datalen,
digestbuf, SHA512_DIGEST_LENGTH);
if (ret != 0)
return (ret);
bcopy(digestbuf, salt, ZIO_DATA_SALT_LEN);
bcopy(digestbuf + ZIO_DATA_SALT_LEN, ivbuf, ZIO_DATA_IV_LEN);
return (0);
}
/*
* The following functions are used to encode and decode encryption parameters
* into blkptr_t and zil_header_t. The ICP wants to use these parameters as
* byte strings, which normally means that these strings would not need to deal
* with byteswapping at all. However, both blkptr_t and zil_header_t may be
* byteswapped by lower layers and so we must "undo" that byteswap here upon
* decoding and encoding in a non-native byteorder. These functions require
* that the byteorder bit is correct before being called.
*/
void
zio_crypt_encode_params_bp(blkptr_t *bp, uint8_t *salt, uint8_t *iv)
{
uint64_t val64;
uint32_t val32;
ASSERT(BP_IS_ENCRYPTED(bp));
if (!BP_SHOULD_BYTESWAP(bp)) {
bcopy(salt, &bp->blk_dva[2].dva_word[0], sizeof (uint64_t));
bcopy(iv, &bp->blk_dva[2].dva_word[1], sizeof (uint64_t));
bcopy(iv + sizeof (uint64_t), &val32, sizeof (uint32_t));
BP_SET_IV2(bp, val32);
} else {
bcopy(salt, &val64, sizeof (uint64_t));
bp->blk_dva[2].dva_word[0] = BSWAP_64(val64);
bcopy(iv, &val64, sizeof (uint64_t));
bp->blk_dva[2].dva_word[1] = BSWAP_64(val64);
bcopy(iv + sizeof (uint64_t), &val32, sizeof (uint32_t));
BP_SET_IV2(bp, BSWAP_32(val32));
}
}
void
zio_crypt_decode_params_bp(const blkptr_t *bp, uint8_t *salt, uint8_t *iv)
{
uint64_t val64;
uint32_t val32;
ASSERT(BP_IS_PROTECTED(bp));
/* for convenience, so callers don't need to check */
if (BP_IS_AUTHENTICATED(bp)) {
bzero(salt, ZIO_DATA_SALT_LEN);
bzero(iv, ZIO_DATA_IV_LEN);
return;
}
if (!BP_SHOULD_BYTESWAP(bp)) {
bcopy(&bp->blk_dva[2].dva_word[0], salt, sizeof (uint64_t));
bcopy(&bp->blk_dva[2].dva_word[1], iv, sizeof (uint64_t));
val32 = (uint32_t)BP_GET_IV2(bp);
bcopy(&val32, iv + sizeof (uint64_t), sizeof (uint32_t));
} else {
val64 = BSWAP_64(bp->blk_dva[2].dva_word[0]);
bcopy(&val64, salt, sizeof (uint64_t));
val64 = BSWAP_64(bp->blk_dva[2].dva_word[1]);
bcopy(&val64, iv, sizeof (uint64_t));
val32 = BSWAP_32((uint32_t)BP_GET_IV2(bp));
bcopy(&val32, iv + sizeof (uint64_t), sizeof (uint32_t));
}
}
void
zio_crypt_encode_mac_bp(blkptr_t *bp, uint8_t *mac)
{
uint64_t val64;
ASSERT(BP_USES_CRYPT(bp));
ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_OBJSET);
if (!BP_SHOULD_BYTESWAP(bp)) {
bcopy(mac, &bp->blk_cksum.zc_word[2], sizeof (uint64_t));
bcopy(mac + sizeof (uint64_t), &bp->blk_cksum.zc_word[3],
sizeof (uint64_t));
} else {
bcopy(mac, &val64, sizeof (uint64_t));
bp->blk_cksum.zc_word[2] = BSWAP_64(val64);
bcopy(mac + sizeof (uint64_t), &val64, sizeof (uint64_t));
bp->blk_cksum.zc_word[3] = BSWAP_64(val64);
}
}
void
zio_crypt_decode_mac_bp(const blkptr_t *bp, uint8_t *mac)
{
uint64_t val64;
ASSERT(BP_USES_CRYPT(bp) || BP_IS_HOLE(bp));
/* for convenience, so callers don't need to check */
if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
bzero(mac, ZIO_DATA_MAC_LEN);
return;
}
if (!BP_SHOULD_BYTESWAP(bp)) {
bcopy(&bp->blk_cksum.zc_word[2], mac, sizeof (uint64_t));
bcopy(&bp->blk_cksum.zc_word[3], mac + sizeof (uint64_t),
sizeof (uint64_t));
} else {
val64 = BSWAP_64(bp->blk_cksum.zc_word[2]);
bcopy(&val64, mac, sizeof (uint64_t));
val64 = BSWAP_64(bp->blk_cksum.zc_word[3]);
bcopy(&val64, mac + sizeof (uint64_t), sizeof (uint64_t));
}
}
void
zio_crypt_encode_mac_zil(void *data, uint8_t *mac)
{
zil_chain_t *zilc = data;
bcopy(mac, &zilc->zc_eck.zec_cksum.zc_word[2], sizeof (uint64_t));
bcopy(mac + sizeof (uint64_t), &zilc->zc_eck.zec_cksum.zc_word[3],
sizeof (uint64_t));
}
void
zio_crypt_decode_mac_zil(const void *data, uint8_t *mac)
{
/*
* The ZIL MAC is embedded in the block it protects, which will
* not have been byteswapped by the time this function has been called.
* As a result, we don't need to worry about byteswapping the MAC.
*/
const zil_chain_t *zilc = data;
bcopy(&zilc->zc_eck.zec_cksum.zc_word[2], mac, sizeof (uint64_t));
bcopy(&zilc->zc_eck.zec_cksum.zc_word[3], mac + sizeof (uint64_t),
sizeof (uint64_t));
}
/*
* This routine takes a block of dnodes (src_abd) and copies only the bonus
* buffers to the same offsets in the dst buffer. datalen should be the size
* of both the src_abd and the dst buffer (not just the length of the bonus
* buffers).
*/
void
zio_crypt_copy_dnode_bonus(abd_t *src_abd, uint8_t *dst, uint_t datalen)
{
uint_t i, max_dnp = datalen >> DNODE_SHIFT;
uint8_t *src;
dnode_phys_t *dnp, *sdnp, *ddnp;
src = abd_borrow_buf_copy(src_abd, datalen);
sdnp = (dnode_phys_t *)src;
ddnp = (dnode_phys_t *)dst;
for (i = 0; i < max_dnp; i += sdnp[i].dn_extra_slots + 1) {
dnp = &sdnp[i];
if (dnp->dn_type != DMU_OT_NONE &&
DMU_OT_IS_ENCRYPTED(dnp->dn_bonustype) &&
dnp->dn_bonuslen != 0) {
bcopy(DN_BONUS(dnp), DN_BONUS(&ddnp[i]),
DN_MAX_BONUS_LEN(dnp));
}
}
abd_return_buf(src_abd, src, datalen);
}
/*
* This function decides what fields from blk_prop are included in
* the on-disk various MAC algorithms.
*/
static void
zio_crypt_bp_zero_nonportable_blkprop(blkptr_t *bp, uint64_t version)
{
int avoidlint = SPA_MINBLOCKSIZE;
/*
* Version 0 did not properly zero out all non-portable fields
* as it should have done. We maintain this code so that we can
* do read-only imports of pools on this version.
*/
if (version == 0) {
BP_SET_DEDUP(bp, 0);
BP_SET_CHECKSUM(bp, 0);
BP_SET_PSIZE(bp, avoidlint);
return;
}
ASSERT3U(version, ==, ZIO_CRYPT_KEY_CURRENT_VERSION);
/*
* The hole_birth feature might set these fields even if this bp
* is a hole. We zero them out here to guarantee that raw sends
* will function with or without the feature.
*/
if (BP_IS_HOLE(bp)) {
bp->blk_prop = 0ULL;
return;
}
/*
* At L0 we want to verify these fields to ensure that data blocks
* can not be reinterpreted. For instance, we do not want an attacker
* to trick us into returning raw lz4 compressed data to the user
* by modifying the compression bits. At higher levels, we cannot
* enforce this policy since raw sends do not convey any information
* about indirect blocks, so these values might be different on the
* receive side. Fortunately, this does not open any new attack
* vectors, since any alterations that can be made to a higher level
* bp must still verify the correct order of the layer below it.
*/
if (BP_GET_LEVEL(bp) != 0) {
BP_SET_BYTEORDER(bp, 0);
BP_SET_COMPRESS(bp, 0);
/*
* psize cannot be set to zero or it will trigger
* asserts, but the value doesn't really matter as
* long as it is constant.
*/
BP_SET_PSIZE(bp, avoidlint);
}
BP_SET_DEDUP(bp, 0);
BP_SET_CHECKSUM(bp, 0);
}
static void
zio_crypt_bp_auth_init(uint64_t version, boolean_t should_bswap, blkptr_t *bp,
blkptr_auth_buf_t *bab, uint_t *bab_len)
{
blkptr_t tmpbp = *bp;
if (should_bswap)
byteswap_uint64_array(&tmpbp, sizeof (blkptr_t));
ASSERT(BP_USES_CRYPT(&tmpbp) || BP_IS_HOLE(&tmpbp));
ASSERT0(BP_IS_EMBEDDED(&tmpbp));
zio_crypt_decode_mac_bp(&tmpbp, bab->bab_mac);
/*
* We always MAC blk_prop in LE to ensure portability. This
* must be done after decoding the mac, since the endianness
* will get zero'd out here.
*/
zio_crypt_bp_zero_nonportable_blkprop(&tmpbp, version);
bab->bab_prop = LE_64(tmpbp.blk_prop);
bab->bab_pad = 0ULL;
/* version 0 did not include the padding */
*bab_len = sizeof (blkptr_auth_buf_t);
if (version == 0)
*bab_len -= sizeof (uint64_t);
}
static int
zio_crypt_bp_do_hmac_updates(crypto_context_t ctx, uint64_t version,
boolean_t should_bswap, blkptr_t *bp)
{
uint_t bab_len;
blkptr_auth_buf_t bab;
zio_crypt_bp_auth_init(version, should_bswap, bp, &bab, &bab_len);
crypto_mac_update(ctx, &bab, bab_len);
return (0);
}
static void
zio_crypt_bp_do_indrect_checksum_updates(SHA2_CTX *ctx, uint64_t version,
boolean_t should_bswap, blkptr_t *bp)
{
uint_t bab_len;
blkptr_auth_buf_t bab;
zio_crypt_bp_auth_init(version, should_bswap, bp, &bab, &bab_len);
SHA2Update(ctx, &bab, bab_len);
}
static void
zio_crypt_bp_do_aad_updates(uint8_t **aadp, uint_t *aad_len, uint64_t version,
boolean_t should_bswap, blkptr_t *bp)
{
uint_t bab_len;
blkptr_auth_buf_t bab;
zio_crypt_bp_auth_init(version, should_bswap, bp, &bab, &bab_len);
bcopy(&bab, *aadp, bab_len);
*aadp += bab_len;
*aad_len += bab_len;
}
static int
zio_crypt_do_dnode_hmac_updates(crypto_context_t ctx, uint64_t version,
boolean_t should_bswap, dnode_phys_t *dnp)
{
int ret, i;
dnode_phys_t *adnp;
boolean_t le_bswap = (should_bswap == ZFS_HOST_BYTEORDER);
uint8_t tmp_dncore[offsetof(dnode_phys_t, dn_blkptr)];
/* authenticate the core dnode (masking out non-portable bits) */
bcopy(dnp, tmp_dncore, sizeof (tmp_dncore));
adnp = (dnode_phys_t *)tmp_dncore;
if (le_bswap) {
adnp->dn_datablkszsec = BSWAP_16(adnp->dn_datablkszsec);
adnp->dn_bonuslen = BSWAP_16(adnp->dn_bonuslen);
adnp->dn_maxblkid = BSWAP_64(adnp->dn_maxblkid);
adnp->dn_used = BSWAP_64(adnp->dn_used);
}
adnp->dn_flags &= DNODE_CRYPT_PORTABLE_FLAGS_MASK;
adnp->dn_used = 0;
crypto_mac_update(ctx, adnp, sizeof (tmp_dncore));
for (i = 0; i < dnp->dn_nblkptr; i++) {
ret = zio_crypt_bp_do_hmac_updates(ctx, version,
should_bswap, &dnp->dn_blkptr[i]);
if (ret != 0)
goto error;
}
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
ret = zio_crypt_bp_do_hmac_updates(ctx, version,
should_bswap, DN_SPILL_BLKPTR(dnp));
if (ret != 0)
goto error;
}
return (0);
error:
return (ret);
}
/*
* objset_phys_t blocks introduce a number of exceptions to the normal
* authentication process. objset_phys_t's contain 2 separate HMACS for
* protecting the integrity of their data. The portable_mac protects the
* metadnode. This MAC can be sent with a raw send and protects against
* reordering of data within the metadnode. The local_mac protects the user
* accounting objects which are not sent from one system to another.
*
* In addition, objset blocks are the only blocks that can be modified and
* written to disk without the key loaded under certain circumstances. During
* zil_claim() we need to be able to update the zil_header_t to complete
* claiming log blocks and during raw receives we need to write out the
* portable_mac from the send file. Both of these actions are possible
* because these fields are not protected by either MAC so neither one will
* need to modify the MACs without the key. However, when the modified blocks
* are written out they will be byteswapped into the host machine's native
* endianness which will modify fields protected by the MAC. As a result, MAC
* calculation for objset blocks works slightly differently from other block
* types. Where other block types MAC the data in whatever endianness is
* written to disk, objset blocks always MAC little endian version of their
* values. In the code, should_bswap is the value from BP_SHOULD_BYTESWAP()
* and le_bswap indicates whether a byteswap is needed to get this block
* into little endian format.
*/
int
zio_crypt_do_objset_hmacs(zio_crypt_key_t *key, void *data, uint_t datalen,
boolean_t should_bswap, uint8_t *portable_mac, uint8_t *local_mac)
{
int ret;
struct hmac_ctx hash_ctx;
struct hmac_ctx *ctx = &hash_ctx;
objset_phys_t *osp = data;
uint64_t intval;
boolean_t le_bswap = (should_bswap == ZFS_HOST_BYTEORDER);
uint8_t raw_portable_mac[SHA512_DIGEST_LENGTH];
uint8_t raw_local_mac[SHA512_DIGEST_LENGTH];
/* calculate the portable MAC from the portable fields and metadnode */
crypto_mac_init(ctx, &key->zk_hmac_key);
/* add in the os_type */
intval = (le_bswap) ? osp->os_type : BSWAP_64(osp->os_type);
crypto_mac_update(ctx, &intval, sizeof (uint64_t));
/* add in the portable os_flags */
intval = osp->os_flags;
if (should_bswap)
intval = BSWAP_64(intval);
intval &= OBJSET_CRYPT_PORTABLE_FLAGS_MASK;
if (!ZFS_HOST_BYTEORDER)
intval = BSWAP_64(intval);
crypto_mac_update(ctx, &intval, sizeof (uint64_t));
/* add in fields from the metadnode */
ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
should_bswap, &osp->os_meta_dnode);
if (ret)
goto error;
crypto_mac_final(ctx, raw_portable_mac, SHA512_DIGEST_LENGTH);
bcopy(raw_portable_mac, portable_mac, ZIO_OBJSET_MAC_LEN);
/*
* The local MAC protects the user, group and project accounting.
* If these objects are not present, the local MAC is zeroed out.
*/
if ((datalen >= OBJSET_PHYS_SIZE_V3 &&
osp->os_userused_dnode.dn_type == DMU_OT_NONE &&
osp->os_groupused_dnode.dn_type == DMU_OT_NONE &&
osp->os_projectused_dnode.dn_type == DMU_OT_NONE) ||
(datalen >= OBJSET_PHYS_SIZE_V2 &&
osp->os_userused_dnode.dn_type == DMU_OT_NONE &&
osp->os_groupused_dnode.dn_type == DMU_OT_NONE) ||
(datalen <= OBJSET_PHYS_SIZE_V1)) {
bzero(local_mac, ZIO_OBJSET_MAC_LEN);
return (0);
}
/* calculate the local MAC from the userused and groupused dnodes */
crypto_mac_init(ctx, &key->zk_hmac_key);
/* add in the non-portable os_flags */
intval = osp->os_flags;
if (should_bswap)
intval = BSWAP_64(intval);
intval &= ~OBJSET_CRYPT_PORTABLE_FLAGS_MASK;
if (!ZFS_HOST_BYTEORDER)
intval = BSWAP_64(intval);
crypto_mac_update(ctx, &intval, sizeof (uint64_t));
/* XXX check dnode type ... */
/* add in fields from the user accounting dnodes */
if (osp->os_userused_dnode.dn_type != DMU_OT_NONE) {
ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
should_bswap, &osp->os_userused_dnode);
if (ret)
goto error;
}
if (osp->os_groupused_dnode.dn_type != DMU_OT_NONE) {
ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
should_bswap, &osp->os_groupused_dnode);
if (ret)
goto error;
}
if (osp->os_projectused_dnode.dn_type != DMU_OT_NONE &&
datalen >= OBJSET_PHYS_SIZE_V3) {
ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
should_bswap, &osp->os_projectused_dnode);
if (ret)
goto error;
}
crypto_mac_final(ctx, raw_local_mac, SHA512_DIGEST_LENGTH);
bcopy(raw_local_mac, local_mac, ZIO_OBJSET_MAC_LEN);
return (0);
error:
bzero(portable_mac, ZIO_OBJSET_MAC_LEN);
bzero(local_mac, ZIO_OBJSET_MAC_LEN);
return (ret);
}
static void
zio_crypt_destroy_uio(zfs_uio_t *uio)
{
if (GET_UIO_STRUCT(uio)->uio_iov)
kmem_free(GET_UIO_STRUCT(uio)->uio_iov,
zfs_uio_iovcnt(uio) * sizeof (iovec_t));
}
/*
* This function parses an uncompressed indirect block and returns a checksum
* of all the portable fields from all of the contained bps. The portable
* fields are the MAC and all of the fields from blk_prop except for the dedup,
* checksum, and psize bits. For an explanation of the purpose of this, see
* the comment block on object set authentication.
*/
static int
zio_crypt_do_indirect_mac_checksum_impl(boolean_t generate, void *buf,
uint_t datalen, uint64_t version, boolean_t byteswap, uint8_t *cksum)
{
blkptr_t *bp;
int i, epb = datalen >> SPA_BLKPTRSHIFT;
SHA2_CTX ctx;
uint8_t digestbuf[SHA512_DIGEST_LENGTH];
/* checksum all of the MACs from the layer below */
SHA2Init(SHA512, &ctx);
for (i = 0, bp = buf; i < epb; i++, bp++) {
zio_crypt_bp_do_indrect_checksum_updates(&ctx, version,
byteswap, bp);
}
SHA2Final(digestbuf, &ctx);
if (generate) {
bcopy(digestbuf, cksum, ZIO_DATA_MAC_LEN);
return (0);
}
if (bcmp(digestbuf, cksum, ZIO_DATA_MAC_LEN) != 0) {
#ifdef FCRYPTO_DEBUG
printf("%s(%d): Setting ECKSUM\n", __FUNCTION__, __LINE__);
#endif
return (SET_ERROR(ECKSUM));
}
return (0);
}
int
zio_crypt_do_indirect_mac_checksum(boolean_t generate, void *buf,
uint_t datalen, boolean_t byteswap, uint8_t *cksum)
{
int ret;
/*
* Unfortunately, callers of this function will not always have
* easy access to the on-disk format version. This info is
* normally found in the DSL Crypto Key, but the checksum-of-MACs
* is expected to be verifiable even when the key isn't loaded.
* Here, instead of doing a ZAP lookup for the version for each
* zio, we simply try both existing formats.
*/
ret = zio_crypt_do_indirect_mac_checksum_impl(generate, buf,
datalen, ZIO_CRYPT_KEY_CURRENT_VERSION, byteswap, cksum);
if (ret == ECKSUM) {
ASSERT(!generate);
ret = zio_crypt_do_indirect_mac_checksum_impl(generate,
buf, datalen, 0, byteswap, cksum);
}
return (ret);
}
int
zio_crypt_do_indirect_mac_checksum_abd(boolean_t generate, abd_t *abd,
uint_t datalen, boolean_t byteswap, uint8_t *cksum)
{
int ret;
void *buf;
buf = abd_borrow_buf_copy(abd, datalen);
ret = zio_crypt_do_indirect_mac_checksum(generate, buf, datalen,
byteswap, cksum);
abd_return_buf(abd, buf, datalen);
return (ret);
}
/*
* Special case handling routine for encrypting / decrypting ZIL blocks.
* We do not check for the older ZIL chain because the encryption feature
* was not available before the newer ZIL chain was introduced. The goal
* here is to encrypt everything except the blkptr_t of a lr_write_t and
* the zil_chain_t header. Everything that is not encrypted is authenticated.
*/
/*
* The OpenCrypto used in FreeBSD does not use separate source and
* destination buffers; instead, the same buffer is used. Further, to
* accommodate some of the drivers, the authbuf needs to be logically before
* the data. This means that we need to copy the source to the destination,
* and set up an extra iovec_t at the beginning to handle the authbuf.
* It also means we'll only return one zfs_uio_t.
*/
static int
zio_crypt_init_uios_zil(boolean_t encrypt, uint8_t *plainbuf,
uint8_t *cipherbuf, uint_t datalen, boolean_t byteswap, zfs_uio_t *puio,
zfs_uio_t *out_uio, uint_t *enc_len, uint8_t **authbuf, uint_t *auth_len,
boolean_t *no_crypt)
{
(void) puio;
uint8_t *aadbuf = zio_buf_alloc(datalen);
uint8_t *src, *dst, *slrp, *dlrp, *blkend, *aadp;
iovec_t *dst_iovecs;
zil_chain_t *zilc;
lr_t *lr;
uint64_t txtype, lr_len;
uint_t crypt_len, nr_iovecs, vec;
uint_t aad_len = 0, total_len = 0;
if (encrypt) {
src = plainbuf;
dst = cipherbuf;
} else {
src = cipherbuf;
dst = plainbuf;
}
bcopy(src, dst, datalen);
/* Find the start and end record of the log block. */
zilc = (zil_chain_t *)src;
slrp = src + sizeof (zil_chain_t);
aadp = aadbuf;
blkend = src + ((byteswap) ? BSWAP_64(zilc->zc_nused) : zilc->zc_nused);
/*
* Calculate the number of encrypted iovecs we will need.
*/
/* We need at least two iovecs -- one for the AAD, one for the MAC. */
nr_iovecs = 2;
for (; slrp < blkend; slrp += lr_len) {
lr = (lr_t *)slrp;
if (byteswap) {
txtype = BSWAP_64(lr->lrc_txtype);
lr_len = BSWAP_64(lr->lrc_reclen);
} else {
txtype = lr->lrc_txtype;
lr_len = lr->lrc_reclen;
}
nr_iovecs++;
if (txtype == TX_WRITE && lr_len != sizeof (lr_write_t))
nr_iovecs++;
}
dst_iovecs = kmem_alloc(nr_iovecs * sizeof (iovec_t), KM_SLEEP);
/*
* Copy the plain zil header over and authenticate everything except
* the checksum that will store our MAC. If we are writing the data
* the embedded checksum will not have been calculated yet, so we don't
* authenticate that.
*/
bcopy(src, aadp, sizeof (zil_chain_t) - sizeof (zio_eck_t));
aadp += sizeof (zil_chain_t) - sizeof (zio_eck_t);
aad_len += sizeof (zil_chain_t) - sizeof (zio_eck_t);
slrp = src + sizeof (zil_chain_t);
dlrp = dst + sizeof (zil_chain_t);
/*
* Loop over records again, filling in iovecs.
*/
/* The first iovec will contain the authbuf. */
vec = 1;
for (; slrp < blkend; slrp += lr_len, dlrp += lr_len) {
lr = (lr_t *)slrp;
if (!byteswap) {
txtype = lr->lrc_txtype;
lr_len = lr->lrc_reclen;
} else {
txtype = BSWAP_64(lr->lrc_txtype);
lr_len = BSWAP_64(lr->lrc_reclen);
}
/* copy the common lr_t */
bcopy(slrp, dlrp, sizeof (lr_t));
bcopy(slrp, aadp, sizeof (lr_t));
aadp += sizeof (lr_t);
aad_len += sizeof (lr_t);
/*
* If this is a TX_WRITE record we want to encrypt everything
* except the bp if exists. If the bp does exist we want to
* authenticate it.
*/
if (txtype == TX_WRITE) {
crypt_len = sizeof (lr_write_t) -
sizeof (lr_t) - sizeof (blkptr_t);
dst_iovecs[vec].iov_base = (char *)dlrp +
sizeof (lr_t);
dst_iovecs[vec].iov_len = crypt_len;
/* copy the bp now since it will not be encrypted */
bcopy(slrp + sizeof (lr_write_t) - sizeof (blkptr_t),
dlrp + sizeof (lr_write_t) - sizeof (blkptr_t),
sizeof (blkptr_t));
bcopy(slrp + sizeof (lr_write_t) - sizeof (blkptr_t),
aadp, sizeof (blkptr_t));
aadp += sizeof (blkptr_t);
aad_len += sizeof (blkptr_t);
vec++;
total_len += crypt_len;
if (lr_len != sizeof (lr_write_t)) {
crypt_len = lr_len - sizeof (lr_write_t);
dst_iovecs[vec].iov_base = (char *)
dlrp + sizeof (lr_write_t);
dst_iovecs[vec].iov_len = crypt_len;
vec++;
total_len += crypt_len;
}
} else {
crypt_len = lr_len - sizeof (lr_t);
dst_iovecs[vec].iov_base = (char *)dlrp +
sizeof (lr_t);
dst_iovecs[vec].iov_len = crypt_len;
vec++;
total_len += crypt_len;
}
}
/* The last iovec will contain the MAC. */
ASSERT3U(vec, ==, nr_iovecs - 1);
/* AAD */
dst_iovecs[0].iov_base = aadbuf;
dst_iovecs[0].iov_len = aad_len;
/* MAC */
dst_iovecs[vec].iov_base = 0;
dst_iovecs[vec].iov_len = 0;
*no_crypt = (vec == 1);
*enc_len = total_len;
*authbuf = aadbuf;
*auth_len = aad_len;
GET_UIO_STRUCT(out_uio)->uio_iov = dst_iovecs;
zfs_uio_iovcnt(out_uio) = nr_iovecs;
return (0);
}
/*
* Special case handling routine for encrypting / decrypting dnode blocks.
*/
static int
zio_crypt_init_uios_dnode(boolean_t encrypt, uint64_t version,
uint8_t *plainbuf, uint8_t *cipherbuf, uint_t datalen, boolean_t byteswap,
zfs_uio_t *puio, zfs_uio_t *out_uio, uint_t *enc_len, uint8_t **authbuf,
uint_t *auth_len, boolean_t *no_crypt)
{
uint8_t *aadbuf = zio_buf_alloc(datalen);
uint8_t *src, *dst, *aadp;
dnode_phys_t *dnp, *adnp, *sdnp, *ddnp;
iovec_t *dst_iovecs;
uint_t nr_iovecs, crypt_len, vec;
uint_t aad_len = 0, total_len = 0;
uint_t i, j, max_dnp = datalen >> DNODE_SHIFT;
if (encrypt) {
src = plainbuf;
dst = cipherbuf;
} else {
src = cipherbuf;
dst = plainbuf;
}
bcopy(src, dst, datalen);
sdnp = (dnode_phys_t *)src;
ddnp = (dnode_phys_t *)dst;
aadp = aadbuf;
/*
* Count the number of iovecs we will need to do the encryption by
* counting the number of bonus buffers that need to be encrypted.
*/
/* We need at least two iovecs -- one for the AAD, one for the MAC. */
nr_iovecs = 2;
for (i = 0; i < max_dnp; i += sdnp[i].dn_extra_slots + 1) {
/*
* This block may still be byteswapped. However, all of the
* values we use are either uint8_t's (for which byteswapping
* is a noop) or a * != 0 check, which will work regardless
* of whether or not we byteswap.
*/
if (sdnp[i].dn_type != DMU_OT_NONE &&
DMU_OT_IS_ENCRYPTED(sdnp[i].dn_bonustype) &&
sdnp[i].dn_bonuslen != 0) {
nr_iovecs++;
}
}
dst_iovecs = kmem_alloc(nr_iovecs * sizeof (iovec_t), KM_SLEEP);
/*
* Iterate through the dnodes again, this time filling in the uios
* we allocated earlier. We also concatenate any data we want to
* authenticate onto aadbuf.
*/
/* The first iovec will contain the authbuf. */
vec = 1;
for (i = 0; i < max_dnp; i += sdnp[i].dn_extra_slots + 1) {
dnp = &sdnp[i];
/* copy over the core fields and blkptrs (kept as plaintext) */
bcopy(dnp, &ddnp[i], (uint8_t *)DN_BONUS(dnp) - (uint8_t *)dnp);
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
bcopy(DN_SPILL_BLKPTR(dnp), DN_SPILL_BLKPTR(&ddnp[i]),
sizeof (blkptr_t));
}
/*
* Handle authenticated data. We authenticate everything in
* the dnode that can be brought over when we do a raw send.
* This includes all of the core fields as well as the MACs
* stored in the bp checksums and all of the portable bits
* from blk_prop. We include the dnode padding here in case it
* ever gets used in the future. Some dn_flags and dn_used are
* not portable so we mask those out values out of the
* authenticated data.
*/
crypt_len = offsetof(dnode_phys_t, dn_blkptr);
bcopy(dnp, aadp, crypt_len);
adnp = (dnode_phys_t *)aadp;
adnp->dn_flags &= DNODE_CRYPT_PORTABLE_FLAGS_MASK;
adnp->dn_used = 0;
aadp += crypt_len;
aad_len += crypt_len;
for (j = 0; j < dnp->dn_nblkptr; j++) {
zio_crypt_bp_do_aad_updates(&aadp, &aad_len,
version, byteswap, &dnp->dn_blkptr[j]);
}
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
zio_crypt_bp_do_aad_updates(&aadp, &aad_len,
version, byteswap, DN_SPILL_BLKPTR(dnp));
}
/*
* If this bonus buffer needs to be encrypted, we prepare an
* iovec_t. The encryption / decryption functions will fill
* this in for us with the encrypted or decrypted data.
* Otherwise we add the bonus buffer to the authenticated
* data buffer and copy it over to the destination. The
* encrypted iovec extends to DN_MAX_BONUS_LEN(dnp) so that
* we can guarantee alignment with the AES block size
* (128 bits).
*/
crypt_len = DN_MAX_BONUS_LEN(dnp);
if (dnp->dn_type != DMU_OT_NONE &&
DMU_OT_IS_ENCRYPTED(dnp->dn_bonustype) &&
dnp->dn_bonuslen != 0) {
dst_iovecs[vec].iov_base = DN_BONUS(&ddnp[i]);
dst_iovecs[vec].iov_len = crypt_len;
vec++;
total_len += crypt_len;
} else {
bcopy(DN_BONUS(dnp), DN_BONUS(&ddnp[i]), crypt_len);
bcopy(DN_BONUS(dnp), aadp, crypt_len);
aadp += crypt_len;
aad_len += crypt_len;
}
}
/* The last iovec will contain the MAC. */
ASSERT3U(vec, ==, nr_iovecs - 1);
/* AAD */
dst_iovecs[0].iov_base = aadbuf;
dst_iovecs[0].iov_len = aad_len;
/* MAC */
dst_iovecs[vec].iov_base = 0;
dst_iovecs[vec].iov_len = 0;
*no_crypt = (vec == 1);
*enc_len = total_len;
*authbuf = aadbuf;
*auth_len = aad_len;
GET_UIO_STRUCT(out_uio)->uio_iov = dst_iovecs;
zfs_uio_iovcnt(out_uio) = nr_iovecs;
return (0);
}
static int
zio_crypt_init_uios_normal(boolean_t encrypt, uint8_t *plainbuf,
uint8_t *cipherbuf, uint_t datalen, zfs_uio_t *puio, zfs_uio_t *out_uio,
uint_t *enc_len)
{
(void) puio;
int ret;
uint_t nr_plain = 1, nr_cipher = 2;
iovec_t *plain_iovecs = NULL, *cipher_iovecs = NULL;
void *src, *dst;
cipher_iovecs = kmem_alloc(nr_cipher * sizeof (iovec_t),
KM_SLEEP);
if (!cipher_iovecs) {
ret = SET_ERROR(ENOMEM);
goto error;
}
bzero(cipher_iovecs, nr_cipher * sizeof (iovec_t));
if (encrypt) {
src = plainbuf;
dst = cipherbuf;
} else {
src = cipherbuf;
dst = plainbuf;
}
bcopy(src, dst, datalen);
cipher_iovecs[0].iov_base = dst;
cipher_iovecs[0].iov_len = datalen;
*enc_len = datalen;
GET_UIO_STRUCT(out_uio)->uio_iov = cipher_iovecs;
zfs_uio_iovcnt(out_uio) = nr_cipher;
return (0);
error:
if (plain_iovecs != NULL)
kmem_free(plain_iovecs, nr_plain * sizeof (iovec_t));
if (cipher_iovecs != NULL)
kmem_free(cipher_iovecs, nr_cipher * sizeof (iovec_t));
*enc_len = 0;
GET_UIO_STRUCT(out_uio)->uio_iov = NULL;
zfs_uio_iovcnt(out_uio) = 0;
return (ret);
}
/*
* This function builds up the plaintext (puio) and ciphertext (cuio) uios so
* that they can be used for encryption and decryption by zio_do_crypt_uio().
* Most blocks will use zio_crypt_init_uios_normal(), with ZIL and dnode blocks
* requiring special handling to parse out pieces that are to be encrypted. The
* authbuf is used by these special cases to store additional authenticated
* data (AAD) for the encryption modes.
*/
static int
zio_crypt_init_uios(boolean_t encrypt, uint64_t version, dmu_object_type_t ot,
uint8_t *plainbuf, uint8_t *cipherbuf, uint_t datalen, boolean_t byteswap,
uint8_t *mac, zfs_uio_t *puio, zfs_uio_t *cuio, uint_t *enc_len,
uint8_t **authbuf, uint_t *auth_len, boolean_t *no_crypt)
{
int ret;
iovec_t *mac_iov;
ASSERT(DMU_OT_IS_ENCRYPTED(ot) || ot == DMU_OT_NONE);
/* route to handler */
switch (ot) {
case DMU_OT_INTENT_LOG:
ret = zio_crypt_init_uios_zil(encrypt, plainbuf, cipherbuf,
datalen, byteswap, puio, cuio, enc_len, authbuf, auth_len,
no_crypt);
break;
case DMU_OT_DNODE:
ret = zio_crypt_init_uios_dnode(encrypt, version, plainbuf,
cipherbuf, datalen, byteswap, puio, cuio, enc_len, authbuf,
auth_len, no_crypt);
break;
default:
ret = zio_crypt_init_uios_normal(encrypt, plainbuf, cipherbuf,
datalen, puio, cuio, enc_len);
*authbuf = NULL;
*auth_len = 0;
*no_crypt = B_FALSE;
break;
}
if (ret != 0)
goto error;
/* populate the uios */
zfs_uio_segflg(cuio) = UIO_SYSSPACE;
mac_iov =
((iovec_t *)&(GET_UIO_STRUCT(cuio)->
uio_iov[zfs_uio_iovcnt(cuio) - 1]));
mac_iov->iov_base = (void *)mac;
mac_iov->iov_len = ZIO_DATA_MAC_LEN;
return (0);
error:
return (ret);
}
void *failed_decrypt_buf;
int faile_decrypt_size;
/*
* Primary encryption / decryption entrypoint for zio data.
*/
int
zio_do_crypt_data(boolean_t encrypt, zio_crypt_key_t *key,
dmu_object_type_t ot, boolean_t byteswap, uint8_t *salt, uint8_t *iv,
uint8_t *mac, uint_t datalen, uint8_t *plainbuf, uint8_t *cipherbuf,
boolean_t *no_crypt)
{
int ret;
boolean_t locked = B_FALSE;
uint64_t crypt = key->zk_crypt;
uint_t keydata_len = zio_crypt_table[crypt].ci_keylen;
uint_t enc_len, auth_len;
zfs_uio_t puio, cuio;
struct uio puio_s, cuio_s;
uint8_t enc_keydata[MASTER_KEY_MAX_LEN];
crypto_key_t tmp_ckey, *ckey = NULL;
freebsd_crypt_session_t *tmpl = NULL;
uint8_t *authbuf = NULL;
zfs_uio_init(&puio, &puio_s);
zfs_uio_init(&cuio, &cuio_s);
bzero(GET_UIO_STRUCT(&puio), sizeof (struct uio));
bzero(GET_UIO_STRUCT(&cuio), sizeof (struct uio));
#ifdef FCRYPTO_DEBUG
printf("%s(%s, %p, %p, %d, %p, %p, %u, %s, %p, %p, %p)\n",
__FUNCTION__,
encrypt ? "encrypt" : "decrypt",
key, salt, ot, iv, mac, datalen,
byteswap ? "byteswap" : "native_endian", plainbuf,
cipherbuf, no_crypt);
printf("\tkey = {");
for (int i = 0; i < key->zk_current_key.ck_length/8; i++)
printf("%02x ", ((uint8_t *)key->zk_current_key.ck_data)[i]);
printf("}\n");
#endif
/* create uios for encryption */
ret = zio_crypt_init_uios(encrypt, key->zk_version, ot, plainbuf,
cipherbuf, datalen, byteswap, mac, &puio, &cuio, &enc_len,
&authbuf, &auth_len, no_crypt);
if (ret != 0)
return (ret);
/*
* If the needed key is the current one, just use it. Otherwise we
* need to generate a temporary one from the given salt + master key.
* If we are encrypting, we must return a copy of the current salt
* so that it can be stored in the blkptr_t.
*/
rw_enter(&key->zk_salt_lock, RW_READER);
locked = B_TRUE;
if (bcmp(salt, key->zk_salt, ZIO_DATA_SALT_LEN) == 0) {
ckey = &key->zk_current_key;
tmpl = &key->zk_session;
} else {
rw_exit(&key->zk_salt_lock);
locked = B_FALSE;
ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
salt, ZIO_DATA_SALT_LEN, enc_keydata, keydata_len);
if (ret != 0)
goto error;
tmp_ckey.ck_format = CRYPTO_KEY_RAW;
tmp_ckey.ck_data = enc_keydata;
tmp_ckey.ck_length = CRYPTO_BYTES2BITS(keydata_len);
ckey = &tmp_ckey;
tmpl = NULL;
}
/* perform the encryption / decryption */
ret = zio_do_crypt_uio_opencrypto(encrypt, tmpl, key->zk_crypt,
ckey, iv, enc_len, &cuio, auth_len);
if (ret != 0)
goto error;
if (locked) {
rw_exit(&key->zk_salt_lock);
locked = B_FALSE;
}
if (authbuf != NULL)
zio_buf_free(authbuf, datalen);
if (ckey == &tmp_ckey)
bzero(enc_keydata, keydata_len);
zio_crypt_destroy_uio(&puio);
zio_crypt_destroy_uio(&cuio);
return (0);
error:
if (!encrypt) {
if (failed_decrypt_buf != NULL)
kmem_free(failed_decrypt_buf, failed_decrypt_size);
failed_decrypt_buf = kmem_alloc(datalen, KM_SLEEP);
failed_decrypt_size = datalen;
bcopy(cipherbuf, failed_decrypt_buf, datalen);
}
if (locked)
rw_exit(&key->zk_salt_lock);
if (authbuf != NULL)
zio_buf_free(authbuf, datalen);
if (ckey == &tmp_ckey)
bzero(enc_keydata, keydata_len);
zio_crypt_destroy_uio(&puio);
zio_crypt_destroy_uio(&cuio);
return (SET_ERROR(ret));
}
/*
* Simple wrapper around zio_do_crypt_data() to work with abd's instead of
* linear buffers.
*/
int
zio_do_crypt_abd(boolean_t encrypt, zio_crypt_key_t *key, dmu_object_type_t ot,
boolean_t byteswap, uint8_t *salt, uint8_t *iv, uint8_t *mac,
uint_t datalen, abd_t *pabd, abd_t *cabd, boolean_t *no_crypt)
{
int ret;
void *ptmp, *ctmp;
if (encrypt) {
ptmp = abd_borrow_buf_copy(pabd, datalen);
ctmp = abd_borrow_buf(cabd, datalen);
} else {
ptmp = abd_borrow_buf(pabd, datalen);
ctmp = abd_borrow_buf_copy(cabd, datalen);
}
ret = zio_do_crypt_data(encrypt, key, ot, byteswap, salt, iv, mac,
datalen, ptmp, ctmp, no_crypt);
if (ret != 0)
goto error;
if (encrypt) {
abd_return_buf(pabd, ptmp, datalen);
abd_return_buf_copy(cabd, ctmp, datalen);
} else {
abd_return_buf_copy(pabd, ptmp, datalen);
abd_return_buf(cabd, ctmp, datalen);
}
return (0);
error:
if (encrypt) {
abd_return_buf(pabd, ptmp, datalen);
abd_return_buf_copy(cabd, ctmp, datalen);
} else {
abd_return_buf_copy(pabd, ptmp, datalen);
abd_return_buf(cabd, ctmp, datalen);
}
return (SET_ERROR(ret));
}
#if defined(_KERNEL) && defined(HAVE_SPL)
/* BEGIN CSTYLED */
module_param(zfs_key_max_salt_uses, ulong, 0644);
MODULE_PARM_DESC(zfs_key_max_salt_uses, "Max number of times a salt value "
"can be used for generating encryption keys before it is rotated");
/* END CSTYLED */
#endif
diff --git a/module/os/linux/spl/spl-generic.c b/module/os/linux/spl/spl-generic.c
index 5ea4fc635165..a7239385953e 100644
--- a/module/os/linux/spl/spl-generic.c
+++ b/module/os/linux/spl/spl-generic.c
@@ -1,844 +1,841 @@
/*
* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
* Copyright (C) 2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
* UCRL-CODE-235197
*
* This file is part of the SPL, Solaris Porting Layer.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*
* Solaris Porting Layer (SPL) Generic Implementation.
*/
#include <sys/sysmacros.h>
#include <sys/systeminfo.h>
#include <sys/vmsystm.h>
#include <sys/kmem.h>
#include <sys/kmem_cache.h>
#include <sys/vmem.h>
#include <sys/mutex.h>
#include <sys/rwlock.h>
#include <sys/taskq.h>
#include <sys/tsd.h>
#include <sys/zmod.h>
#include <sys/debug.h>
#include <sys/proc.h>
#include <sys/kstat.h>
#include <sys/file.h>
#include <sys/sunddi.h>
#include <linux/ctype.h>
#include <sys/disp.h>
#include <sys/random.h>
#include <sys/strings.h>
#include <linux/kmod.h>
-#include "zfs_gitrev.h"
#include <linux/mod_compat.h>
#include <sys/cred.h>
#include <sys/vnode.h>
-char spl_gitrev[64] = ZFS_META_GITREV;
-
/* BEGIN CSTYLED */
unsigned long spl_hostid = 0;
EXPORT_SYMBOL(spl_hostid);
-/* BEGIN CSTYLED */
+
module_param(spl_hostid, ulong, 0644);
MODULE_PARM_DESC(spl_hostid, "The system hostid.");
/* END CSTYLED */
proc_t p0;
EXPORT_SYMBOL(p0);
/*
* Xorshift Pseudo Random Number Generator based on work by Sebastiano Vigna
*
* "Further scramblings of Marsaglia's xorshift generators"
* http://vigna.di.unimi.it/ftp/papers/xorshiftplus.pdf
*
* random_get_pseudo_bytes() is an API function on Illumos whose sole purpose
* is to provide bytes containing random numbers. It is mapped to /dev/urandom
* on Illumos, which uses a "FIPS 186-2 algorithm". No user of the SPL's
* random_get_pseudo_bytes() needs bytes that are of cryptographic quality, so
* we can implement it using a fast PRNG that we seed using Linux' actual
* equivalent to random_get_pseudo_bytes(). We do this by providing each CPU
* with an independent seed so that all calls to random_get_pseudo_bytes() are
* free of atomic instructions.
*
* A consequence of using a fast PRNG is that using random_get_pseudo_bytes()
* to generate words larger than 128 bits will paradoxically be limited to
* `2^128 - 1` possibilities. This is because we have a sequence of `2^128 - 1`
* 128-bit words and selecting the first will implicitly select the second. If
* a caller finds this behavior undesirable, random_get_bytes() should be used
* instead.
*
* XXX: Linux interrupt handlers that trigger within the critical section
* formed by `s[1] = xp[1];` and `xp[0] = s[0];` and call this function will
* see the same numbers. Nothing in the code currently calls this in an
* interrupt handler, so this is considered to be okay. If that becomes a
* problem, we could create a set of per-cpu variables for interrupt handlers
* and use them when in_interrupt() from linux/preempt_mask.h evaluates to
* true.
*/
void __percpu *spl_pseudo_entropy;
/*
* spl_rand_next()/spl_rand_jump() are copied from the following CC-0 licensed
* file:
*
* http://xorshift.di.unimi.it/xorshift128plus.c
*/
static inline uint64_t
spl_rand_next(uint64_t *s)
{
uint64_t s1 = s[0];
const uint64_t s0 = s[1];
s[0] = s0;
s1 ^= s1 << 23; // a
s[1] = s1 ^ s0 ^ (s1 >> 18) ^ (s0 >> 5); // b, c
return (s[1] + s0);
}
static inline void
spl_rand_jump(uint64_t *s)
{
static const uint64_t JUMP[] =
{ 0x8a5cd789635d2dff, 0x121fd2155c472f96 };
uint64_t s0 = 0;
uint64_t s1 = 0;
int i, b;
for (i = 0; i < sizeof (JUMP) / sizeof (*JUMP); i++)
for (b = 0; b < 64; b++) {
if (JUMP[i] & 1ULL << b) {
s0 ^= s[0];
s1 ^= s[1];
}
(void) spl_rand_next(s);
}
s[0] = s0;
s[1] = s1;
}
int
random_get_pseudo_bytes(uint8_t *ptr, size_t len)
{
uint64_t *xp, s[2];
ASSERT(ptr);
xp = get_cpu_ptr(spl_pseudo_entropy);
s[0] = xp[0];
s[1] = xp[1];
while (len) {
union {
uint64_t ui64;
uint8_t byte[sizeof (uint64_t)];
}entropy;
int i = MIN(len, sizeof (uint64_t));
len -= i;
entropy.ui64 = spl_rand_next(s);
while (i--)
*ptr++ = entropy.byte[i];
}
xp[0] = s[0];
xp[1] = s[1];
put_cpu_ptr(spl_pseudo_entropy);
return (0);
}
EXPORT_SYMBOL(random_get_pseudo_bytes);
#if BITS_PER_LONG == 32
/*
* Support 64/64 => 64 division on a 32-bit platform. While the kernel
* provides a div64_u64() function for this we do not use it because the
* implementation is flawed. There are cases which return incorrect
* results as late as linux-2.6.35. Until this is fixed upstream the
* spl must provide its own implementation.
*
* This implementation is a slightly modified version of the algorithm
* proposed by the book 'Hacker's Delight'. The original source can be
* found here and is available for use without restriction.
*
* http://www.hackersdelight.org/HDcode/newCode/divDouble.c
*/
/*
* Calculate number of leading of zeros for a 64-bit value.
*/
static int
nlz64(uint64_t x)
{
register int n = 0;
if (x == 0)
return (64);
if (x <= 0x00000000FFFFFFFFULL) { n = n + 32; x = x << 32; }
if (x <= 0x0000FFFFFFFFFFFFULL) { n = n + 16; x = x << 16; }
if (x <= 0x00FFFFFFFFFFFFFFULL) { n = n + 8; x = x << 8; }
if (x <= 0x0FFFFFFFFFFFFFFFULL) { n = n + 4; x = x << 4; }
if (x <= 0x3FFFFFFFFFFFFFFFULL) { n = n + 2; x = x << 2; }
if (x <= 0x7FFFFFFFFFFFFFFFULL) { n = n + 1; }
return (n);
}
/*
* Newer kernels have a div_u64() function but we define our own
* to simplify portability between kernel versions.
*/
static inline uint64_t
__div_u64(uint64_t u, uint32_t v)
{
(void) do_div(u, v);
return (u);
}
/*
* Turn off missing prototypes warning for these functions. They are
* replacements for libgcc-provided functions and will never be called
* directly.
*/
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wmissing-prototypes"
/*
* Implementation of 64-bit unsigned division for 32-bit machines.
*
* First the procedure takes care of the case in which the divisor is a
* 32-bit quantity. There are two subcases: (1) If the left half of the
* dividend is less than the divisor, one execution of do_div() is all that
* is required (overflow is not possible). (2) Otherwise it does two
* divisions, using the grade school method.
*/
uint64_t
__udivdi3(uint64_t u, uint64_t v)
{
uint64_t u0, u1, v1, q0, q1, k;
int n;
if (v >> 32 == 0) { // If v < 2**32:
if (u >> 32 < v) { // If u/v cannot overflow,
return (__div_u64(u, v)); // just do one division.
} else { // If u/v would overflow:
u1 = u >> 32; // Break u into two halves.
u0 = u & 0xFFFFFFFF;
q1 = __div_u64(u1, v); // First quotient digit.
k = u1 - q1 * v; // First remainder, < v.
u0 += (k << 32);
q0 = __div_u64(u0, v); // Seconds quotient digit.
return ((q1 << 32) + q0);
}
} else { // If v >= 2**32:
n = nlz64(v); // 0 <= n <= 31.
v1 = (v << n) >> 32; // Normalize divisor, MSB is 1.
u1 = u >> 1; // To ensure no overflow.
q1 = __div_u64(u1, v1); // Get quotient from
q0 = (q1 << n) >> 31; // Undo normalization and
// division of u by 2.
if (q0 != 0) // Make q0 correct or
q0 = q0 - 1; // too small by 1.
if ((u - q0 * v) >= v)
q0 = q0 + 1; // Now q0 is correct.
return (q0);
}
}
EXPORT_SYMBOL(__udivdi3);
/* BEGIN CSTYLED */
#ifndef abs64
#define abs64(x) ({ uint64_t t = (x) >> 63; ((x) ^ t) - t; })
#endif
/* END CSTYLED */
/*
* Implementation of 64-bit signed division for 32-bit machines.
*/
int64_t
__divdi3(int64_t u, int64_t v)
{
int64_t q, t;
q = __udivdi3(abs64(u), abs64(v));
t = (u ^ v) >> 63; // If u, v have different
return ((q ^ t) - t); // signs, negate q.
}
EXPORT_SYMBOL(__divdi3);
/*
* Implementation of 64-bit unsigned modulo for 32-bit machines.
*/
uint64_t
__umoddi3(uint64_t dividend, uint64_t divisor)
{
return (dividend - (divisor * __udivdi3(dividend, divisor)));
}
EXPORT_SYMBOL(__umoddi3);
/* 64-bit signed modulo for 32-bit machines. */
int64_t
__moddi3(int64_t n, int64_t d)
{
int64_t q;
boolean_t nn = B_FALSE;
if (n < 0) {
nn = B_TRUE;
n = -n;
}
if (d < 0)
d = -d;
q = __umoddi3(n, d);
return (nn ? -q : q);
}
EXPORT_SYMBOL(__moddi3);
/*
* Implementation of 64-bit unsigned division/modulo for 32-bit machines.
*/
uint64_t
__udivmoddi4(uint64_t n, uint64_t d, uint64_t *r)
{
uint64_t q = __udivdi3(n, d);
if (r)
*r = n - d * q;
return (q);
}
EXPORT_SYMBOL(__udivmoddi4);
/*
* Implementation of 64-bit signed division/modulo for 32-bit machines.
*/
int64_t
__divmoddi4(int64_t n, int64_t d, int64_t *r)
{
int64_t q, rr;
boolean_t nn = B_FALSE;
boolean_t nd = B_FALSE;
if (n < 0) {
nn = B_TRUE;
n = -n;
}
if (d < 0) {
nd = B_TRUE;
d = -d;
}
q = __udivmoddi4(n, d, (uint64_t *)&rr);
if (nn != nd)
q = -q;
if (nn)
rr = -rr;
if (r)
*r = rr;
return (q);
}
EXPORT_SYMBOL(__divmoddi4);
#if defined(__arm) || defined(__arm__)
/*
* Implementation of 64-bit (un)signed division for 32-bit arm machines.
*
* Run-time ABI for the ARM Architecture (page 20). A pair of (unsigned)
* long longs is returned in {{r0, r1}, {r2,r3}}, the quotient in {r0, r1},
* and the remainder in {r2, r3}. The return type is specifically left
* set to 'void' to ensure the compiler does not overwrite these registers
* during the return. All results are in registers as per ABI
*/
void
__aeabi_uldivmod(uint64_t u, uint64_t v)
{
uint64_t res;
uint64_t mod;
res = __udivdi3(u, v);
mod = __umoddi3(u, v);
{
register uint32_t r0 asm("r0") = (res & 0xFFFFFFFF);
register uint32_t r1 asm("r1") = (res >> 32);
register uint32_t r2 asm("r2") = (mod & 0xFFFFFFFF);
register uint32_t r3 asm("r3") = (mod >> 32);
/* BEGIN CSTYLED */
asm volatile(""
: "+r"(r0), "+r"(r1), "+r"(r2),"+r"(r3) /* output */
: "r"(r0), "r"(r1), "r"(r2), "r"(r3)); /* input */
/* END CSTYLED */
return; /* r0; */
}
}
EXPORT_SYMBOL(__aeabi_uldivmod);
void
__aeabi_ldivmod(int64_t u, int64_t v)
{
int64_t res;
uint64_t mod;
res = __divdi3(u, v);
mod = __umoddi3(u, v);
{
register uint32_t r0 asm("r0") = (res & 0xFFFFFFFF);
register uint32_t r1 asm("r1") = (res >> 32);
register uint32_t r2 asm("r2") = (mod & 0xFFFFFFFF);
register uint32_t r3 asm("r3") = (mod >> 32);
/* BEGIN CSTYLED */
asm volatile(""
: "+r"(r0), "+r"(r1), "+r"(r2),"+r"(r3) /* output */
: "r"(r0), "r"(r1), "r"(r2), "r"(r3)); /* input */
/* END CSTYLED */
return; /* r0; */
}
}
EXPORT_SYMBOL(__aeabi_ldivmod);
#endif /* __arm || __arm__ */
#pragma GCC diagnostic pop
#endif /* BITS_PER_LONG */
/*
* NOTE: The strtoxx behavior is solely based on my reading of the Solaris
* ddi_strtol(9F) man page. I have not verified the behavior of these
* functions against their Solaris counterparts. It is possible that I
* may have misinterpreted the man page or the man page is incorrect.
*/
int ddi_strtoul(const char *, char **, int, unsigned long *);
int ddi_strtol(const char *, char **, int, long *);
int ddi_strtoull(const char *, char **, int, unsigned long long *);
int ddi_strtoll(const char *, char **, int, long long *);
#define define_ddi_strtoux(type, valtype) \
int ddi_strtou##type(const char *str, char **endptr, \
int base, valtype *result) \
{ \
valtype last_value, value = 0; \
char *ptr = (char *)str; \
int flag = 1, digit; \
\
if (strlen(ptr) == 0) \
return (EINVAL); \
\
/* Auto-detect base based on prefix */ \
if (!base) { \
if (str[0] == '0') { \
if (tolower(str[1]) == 'x' && isxdigit(str[2])) { \
base = 16; /* hex */ \
ptr += 2; \
} else if (str[1] >= '0' && str[1] < 8) { \
base = 8; /* octal */ \
ptr += 1; \
} else { \
return (EINVAL); \
} \
} else { \
base = 10; /* decimal */ \
} \
} \
\
while (1) { \
if (isdigit(*ptr)) \
digit = *ptr - '0'; \
else if (isalpha(*ptr)) \
digit = tolower(*ptr) - 'a' + 10; \
else \
break; \
\
if (digit >= base) \
break; \
\
last_value = value; \
value = value * base + digit; \
if (last_value > value) /* Overflow */ \
return (ERANGE); \
\
flag = 1; \
ptr++; \
} \
\
if (flag) \
*result = value; \
\
if (endptr) \
*endptr = (char *)(flag ? ptr : str); \
\
return (0); \
} \
#define define_ddi_strtox(type, valtype) \
int ddi_strto##type(const char *str, char **endptr, \
int base, valtype *result) \
{ \
int rc; \
\
if (*str == '-') { \
rc = ddi_strtou##type(str + 1, endptr, base, result); \
if (!rc) { \
if (*endptr == str + 1) \
*endptr = (char *)str; \
else \
*result = -*result; \
} \
} else { \
rc = ddi_strtou##type(str, endptr, base, result); \
} \
\
return (rc); \
}
define_ddi_strtoux(l, unsigned long)
define_ddi_strtox(l, long)
define_ddi_strtoux(ll, unsigned long long)
define_ddi_strtox(ll, long long)
EXPORT_SYMBOL(ddi_strtoul);
EXPORT_SYMBOL(ddi_strtol);
EXPORT_SYMBOL(ddi_strtoll);
EXPORT_SYMBOL(ddi_strtoull);
int
ddi_copyin(const void *from, void *to, size_t len, int flags)
{
/* Fake ioctl() issued by kernel, 'from' is a kernel address */
if (flags & FKIOCTL) {
memcpy(to, from, len);
return (0);
}
return (copyin(from, to, len));
}
EXPORT_SYMBOL(ddi_copyin);
int
ddi_copyout(const void *from, void *to, size_t len, int flags)
{
/* Fake ioctl() issued by kernel, 'from' is a kernel address */
if (flags & FKIOCTL) {
memcpy(to, from, len);
return (0);
}
return (copyout(from, to, len));
}
EXPORT_SYMBOL(ddi_copyout);
static ssize_t
spl_kernel_read(struct file *file, void *buf, size_t count, loff_t *pos)
{
#if defined(HAVE_KERNEL_READ_PPOS)
return (kernel_read(file, buf, count, pos));
#else
mm_segment_t saved_fs;
ssize_t ret;
saved_fs = get_fs();
set_fs(KERNEL_DS);
ret = vfs_read(file, (void __user *)buf, count, pos);
set_fs(saved_fs);
return (ret);
#endif
}
static int
spl_getattr(struct file *filp, struct kstat *stat)
{
int rc;
ASSERT(filp);
ASSERT(stat);
#if defined(HAVE_4ARGS_VFS_GETATTR)
rc = vfs_getattr(&filp->f_path, stat, STATX_BASIC_STATS,
AT_STATX_SYNC_AS_STAT);
#elif defined(HAVE_2ARGS_VFS_GETATTR)
rc = vfs_getattr(&filp->f_path, stat);
#elif defined(HAVE_3ARGS_VFS_GETATTR)
rc = vfs_getattr(filp->f_path.mnt, filp->f_dentry, stat);
#else
#error "No available vfs_getattr()"
#endif
if (rc)
return (-rc);
return (0);
}
/*
* Read the unique system identifier from the /etc/hostid file.
*
* The behavior of /usr/bin/hostid on Linux systems with the
* regular eglibc and coreutils is:
*
* 1. Generate the value if the /etc/hostid file does not exist
* or if the /etc/hostid file is less than four bytes in size.
*
* 2. If the /etc/hostid file is at least 4 bytes, then return
* the first four bytes [0..3] in native endian order.
*
* 3. Always ignore bytes [4..] if they exist in the file.
*
* Only the first four bytes are significant, even on systems that
* have a 64-bit word size.
*
* See:
*
* eglibc: sysdeps/unix/sysv/linux/gethostid.c
* coreutils: src/hostid.c
*
* Notes:
*
* The /etc/hostid file on Solaris is a text file that often reads:
*
* # DO NOT EDIT
* "0123456789"
*
* Directly copying this file to Linux results in a constant
* hostid of 4f442023 because the default comment constitutes
* the first four bytes of the file.
*
*/
-char *spl_hostid_path = HW_HOSTID_PATH;
+static char *spl_hostid_path = HW_HOSTID_PATH;
module_param(spl_hostid_path, charp, 0444);
MODULE_PARM_DESC(spl_hostid_path, "The system hostid file (/etc/hostid)");
static int
hostid_read(uint32_t *hostid)
{
uint64_t size;
uint32_t value = 0;
int error;
loff_t off;
struct file *filp;
struct kstat stat;
filp = filp_open(spl_hostid_path, 0, 0);
if (IS_ERR(filp))
return (ENOENT);
error = spl_getattr(filp, &stat);
if (error) {
filp_close(filp, 0);
return (error);
}
size = stat.size;
// cppcheck-suppress sizeofwithnumericparameter
if (size < sizeof (HW_HOSTID_MASK)) {
filp_close(filp, 0);
return (EINVAL);
}
off = 0;
/*
* Read directly into the variable like eglibc does.
* Short reads are okay; native behavior is preserved.
*/
error = spl_kernel_read(filp, &value, sizeof (value), &off);
if (error < 0) {
filp_close(filp, 0);
return (EIO);
}
/* Mask down to 32 bits like coreutils does. */
*hostid = (value & HW_HOSTID_MASK);
filp_close(filp, 0);
return (0);
}
/*
* Return the system hostid. Preferentially use the spl_hostid module option
* when set, otherwise use the value in the /etc/hostid file.
*/
uint32_t
zone_get_hostid(void *zone)
{
uint32_t hostid;
ASSERT3P(zone, ==, NULL);
if (spl_hostid != 0)
return ((uint32_t)(spl_hostid & HW_HOSTID_MASK));
if (hostid_read(&hostid) == 0)
return (hostid);
return (0);
}
EXPORT_SYMBOL(zone_get_hostid);
static int
spl_kvmem_init(void)
{
int rc = 0;
rc = spl_kmem_init();
if (rc)
return (rc);
rc = spl_vmem_init();
if (rc) {
spl_kmem_fini();
return (rc);
}
return (rc);
}
/*
* We initialize the random number generator with 128 bits of entropy from the
* system random number generator. In the improbable case that we have a zero
* seed, we fallback to the system jiffies, unless it is also zero, in which
* situation we use a preprogrammed seed. We step forward by 2^64 iterations to
* initialize each of the per-cpu seeds so that the sequences generated on each
* CPU are guaranteed to never overlap in practice.
*/
static void __init
spl_random_init(void)
{
uint64_t s[2];
int i = 0;
spl_pseudo_entropy = __alloc_percpu(2 * sizeof (uint64_t),
sizeof (uint64_t));
get_random_bytes(s, sizeof (s));
if (s[0] == 0 && s[1] == 0) {
if (jiffies != 0) {
s[0] = jiffies;
s[1] = ~0 - jiffies;
} else {
(void) memcpy(s, "improbable seed", sizeof (s));
}
printk("SPL: get_random_bytes() returned 0 "
"when generating random seed. Setting initial seed to "
"0x%016llx%016llx.\n", cpu_to_be64(s[0]),
cpu_to_be64(s[1]));
}
for_each_possible_cpu(i) {
uint64_t *wordp = per_cpu_ptr(spl_pseudo_entropy, i);
spl_rand_jump(s);
wordp[0] = s[0];
wordp[1] = s[1];
}
}
static void
spl_random_fini(void)
{
free_percpu(spl_pseudo_entropy);
}
static void
spl_kvmem_fini(void)
{
spl_vmem_fini();
spl_kmem_fini();
}
static int __init
spl_init(void)
{
int rc = 0;
bzero(&p0, sizeof (proc_t));
spl_random_init();
if ((rc = spl_kvmem_init()))
goto out1;
if ((rc = spl_tsd_init()))
goto out2;
if ((rc = spl_taskq_init()))
goto out3;
if ((rc = spl_kmem_cache_init()))
goto out4;
if ((rc = spl_proc_init()))
goto out5;
if ((rc = spl_kstat_init()))
goto out6;
if ((rc = spl_zlib_init()))
goto out7;
return (rc);
out7:
spl_kstat_fini();
out6:
spl_proc_fini();
out5:
spl_kmem_cache_fini();
out4:
spl_taskq_fini();
out3:
spl_tsd_fini();
out2:
spl_kvmem_fini();
out1:
return (rc);
}
static void __exit
spl_fini(void)
{
spl_zlib_fini();
spl_kstat_fini();
spl_proc_fini();
spl_kmem_cache_fini();
spl_taskq_fini();
spl_tsd_fini();
spl_kvmem_fini();
spl_random_fini();
}
module_init(spl_init);
module_exit(spl_fini);
ZFS_MODULE_DESCRIPTION("Solaris Porting Layer");
ZFS_MODULE_AUTHOR(ZFS_META_AUTHOR);
ZFS_MODULE_LICENSE("GPL");
ZFS_MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
diff --git a/module/os/linux/spl/spl-kmem-cache.c b/module/os/linux/spl/spl-kmem-cache.c
index 2151ef008fd6..3d926173674d 100644
--- a/module/os/linux/spl/spl-kmem-cache.c
+++ b/module/os/linux/spl/spl-kmem-cache.c
@@ -1,1463 +1,1463 @@
/*
* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
* Copyright (C) 2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
* UCRL-CODE-235197
*
* This file is part of the SPL, Solaris Porting Layer.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*/
#include <linux/percpu_compat.h>
#include <sys/kmem.h>
#include <sys/kmem_cache.h>
#include <sys/taskq.h>
#include <sys/timer.h>
#include <sys/vmem.h>
#include <sys/wait.h>
#include <linux/slab.h>
#include <linux/swap.h>
#include <linux/prefetch.h>
/*
* Within the scope of spl-kmem.c file the kmem_cache_* definitions
* are removed to allow access to the real Linux slab allocator.
*/
#undef kmem_cache_destroy
#undef kmem_cache_create
#undef kmem_cache_alloc
#undef kmem_cache_free
/*
* Linux 3.16 replaced smp_mb__{before,after}_{atomic,clear}_{dec,inc,bit}()
* with smp_mb__{before,after}_atomic() because they were redundant. This is
* only used inside our SLAB allocator, so we implement an internal wrapper
* here to give us smp_mb__{before,after}_atomic() on older kernels.
*/
#ifndef smp_mb__before_atomic
#define smp_mb__before_atomic(x) smp_mb__before_clear_bit(x)
#endif
#ifndef smp_mb__after_atomic
#define smp_mb__after_atomic(x) smp_mb__after_clear_bit(x)
#endif
/* BEGIN CSTYLED */
/*
* Cache magazines are an optimization designed to minimize the cost of
* allocating memory. They do this by keeping a per-cpu cache of recently
* freed objects, which can then be reallocated without taking a lock. This
* can improve performance on highly contended caches. However, because
* objects in magazines will prevent otherwise empty slabs from being
* immediately released this may not be ideal for low memory machines.
*
* For this reason spl_kmem_cache_magazine_size can be used to set a maximum
* magazine size. When this value is set to 0 the magazine size will be
* automatically determined based on the object size. Otherwise magazines
* will be limited to 2-256 objects per magazine (i.e per cpu). Magazines
* may never be entirely disabled in this implementation.
*/
-unsigned int spl_kmem_cache_magazine_size = 0;
+static unsigned int spl_kmem_cache_magazine_size = 0;
module_param(spl_kmem_cache_magazine_size, uint, 0444);
MODULE_PARM_DESC(spl_kmem_cache_magazine_size,
"Default magazine size (2-256), set automatically (0)");
/*
* The default behavior is to report the number of objects remaining in the
* cache. This allows the Linux VM to repeatedly reclaim objects from the
* cache when memory is low satisfy other memory allocations. Alternately,
* setting this value to KMC_RECLAIM_ONCE limits how aggressively the cache
* is reclaimed. This may increase the likelihood of out of memory events.
*/
-unsigned int spl_kmem_cache_reclaim = 0 /* KMC_RECLAIM_ONCE */;
+static unsigned int spl_kmem_cache_reclaim = 0 /* KMC_RECLAIM_ONCE */;
module_param(spl_kmem_cache_reclaim, uint, 0644);
MODULE_PARM_DESC(spl_kmem_cache_reclaim, "Single reclaim pass (0x1)");
-unsigned int spl_kmem_cache_obj_per_slab = SPL_KMEM_CACHE_OBJ_PER_SLAB;
+static unsigned int spl_kmem_cache_obj_per_slab = SPL_KMEM_CACHE_OBJ_PER_SLAB;
module_param(spl_kmem_cache_obj_per_slab, uint, 0644);
MODULE_PARM_DESC(spl_kmem_cache_obj_per_slab, "Number of objects per slab");
-unsigned int spl_kmem_cache_max_size = SPL_KMEM_CACHE_MAX_SIZE;
+static unsigned int spl_kmem_cache_max_size = SPL_KMEM_CACHE_MAX_SIZE;
module_param(spl_kmem_cache_max_size, uint, 0644);
MODULE_PARM_DESC(spl_kmem_cache_max_size, "Maximum size of slab in MB");
/*
* For small objects the Linux slab allocator should be used to make the most
* efficient use of the memory. However, large objects are not supported by
* the Linux slab and therefore the SPL implementation is preferred. A cutoff
* of 16K was determined to be optimal for architectures using 4K pages and
* to also work well on architecutres using larger 64K page sizes.
*/
-unsigned int spl_kmem_cache_slab_limit = 16384;
+static unsigned int spl_kmem_cache_slab_limit = 16384;
module_param(spl_kmem_cache_slab_limit, uint, 0644);
MODULE_PARM_DESC(spl_kmem_cache_slab_limit,
"Objects less than N bytes use the Linux slab");
/*
* The number of threads available to allocate new slabs for caches. This
* should not need to be tuned but it is available for performance analysis.
*/
-unsigned int spl_kmem_cache_kmem_threads = 4;
+static unsigned int spl_kmem_cache_kmem_threads = 4;
module_param(spl_kmem_cache_kmem_threads, uint, 0444);
MODULE_PARM_DESC(spl_kmem_cache_kmem_threads,
"Number of spl_kmem_cache threads");
/* END CSTYLED */
/*
* Slab allocation interfaces
*
* While the Linux slab implementation was inspired by the Solaris
* implementation I cannot use it to emulate the Solaris APIs. I
* require two features which are not provided by the Linux slab.
*
* 1) Constructors AND destructors. Recent versions of the Linux
* kernel have removed support for destructors. This is a deal
* breaker for the SPL which contains particularly expensive
* initializers for mutex's, condition variables, etc. We also
* require a minimal level of cleanup for these data types unlike
* many Linux data types which do need to be explicitly destroyed.
*
* 2) Virtual address space backed slab. Callers of the Solaris slab
* expect it to work well for both small are very large allocations.
* Because of memory fragmentation the Linux slab which is backed
* by kmalloc'ed memory performs very badly when confronted with
* large numbers of large allocations. Basing the slab on the
* virtual address space removes the need for contiguous pages
* and greatly improve performance for large allocations.
*
* For these reasons, the SPL has its own slab implementation with
* the needed features. It is not as highly optimized as either the
* Solaris or Linux slabs, but it should get me most of what is
* needed until it can be optimized or obsoleted by another approach.
*
* One serious concern I do have about this method is the relatively
* small virtual address space on 32bit arches. This will seriously
* constrain the size of the slab caches and their performance.
*/
struct list_head spl_kmem_cache_list; /* List of caches */
struct rw_semaphore spl_kmem_cache_sem; /* Cache list lock */
taskq_t *spl_kmem_cache_taskq; /* Task queue for aging / reclaim */
static void spl_cache_shrink(spl_kmem_cache_t *skc, void *obj);
static void *
kv_alloc(spl_kmem_cache_t *skc, int size, int flags)
{
gfp_t lflags = kmem_flags_convert(flags);
void *ptr;
ptr = spl_vmalloc(size, lflags | __GFP_HIGHMEM);
/* Resulting allocated memory will be page aligned */
ASSERT(IS_P2ALIGNED(ptr, PAGE_SIZE));
return (ptr);
}
static void
kv_free(spl_kmem_cache_t *skc, void *ptr, int size)
{
ASSERT(IS_P2ALIGNED(ptr, PAGE_SIZE));
/*
* The Linux direct reclaim path uses this out of band value to
* determine if forward progress is being made. Normally this is
* incremented by kmem_freepages() which is part of the various
* Linux slab implementations. However, since we are using none
* of that infrastructure we are responsible for incrementing it.
*/
if (current->reclaim_state)
current->reclaim_state->reclaimed_slab += size >> PAGE_SHIFT;
vfree(ptr);
}
/*
* Required space for each aligned sks.
*/
static inline uint32_t
spl_sks_size(spl_kmem_cache_t *skc)
{
return (P2ROUNDUP_TYPED(sizeof (spl_kmem_slab_t),
skc->skc_obj_align, uint32_t));
}
/*
* Required space for each aligned object.
*/
static inline uint32_t
spl_obj_size(spl_kmem_cache_t *skc)
{
uint32_t align = skc->skc_obj_align;
return (P2ROUNDUP_TYPED(skc->skc_obj_size, align, uint32_t) +
P2ROUNDUP_TYPED(sizeof (spl_kmem_obj_t), align, uint32_t));
}
uint64_t
spl_kmem_cache_inuse(kmem_cache_t *cache)
{
return (cache->skc_obj_total);
}
EXPORT_SYMBOL(spl_kmem_cache_inuse);
uint64_t
spl_kmem_cache_entry_size(kmem_cache_t *cache)
{
return (cache->skc_obj_size);
}
EXPORT_SYMBOL(spl_kmem_cache_entry_size);
/*
* Lookup the spl_kmem_object_t for an object given that object.
*/
static inline spl_kmem_obj_t *
spl_sko_from_obj(spl_kmem_cache_t *skc, void *obj)
{
return (obj + P2ROUNDUP_TYPED(skc->skc_obj_size,
skc->skc_obj_align, uint32_t));
}
/*
* It's important that we pack the spl_kmem_obj_t structure and the
* actual objects in to one large address space to minimize the number
* of calls to the allocator. It is far better to do a few large
* allocations and then subdivide it ourselves. Now which allocator
* we use requires balancing a few trade offs.
*
* For small objects we use kmem_alloc() because as long as you are
* only requesting a small number of pages (ideally just one) its cheap.
* However, when you start requesting multiple pages with kmem_alloc()
* it gets increasingly expensive since it requires contiguous pages.
* For this reason we shift to vmem_alloc() for slabs of large objects
* which removes the need for contiguous pages. We do not use
* vmem_alloc() in all cases because there is significant locking
* overhead in __get_vm_area_node(). This function takes a single
* global lock when acquiring an available virtual address range which
* serializes all vmem_alloc()'s for all slab caches. Using slightly
* different allocation functions for small and large objects should
* give us the best of both worlds.
*
* +------------------------+
* | spl_kmem_slab_t --+-+ |
* | skc_obj_size <-+ | |
* | spl_kmem_obj_t | |
* | skc_obj_size <---+ |
* | spl_kmem_obj_t | |
* | ... v |
* +------------------------+
*/
static spl_kmem_slab_t *
spl_slab_alloc(spl_kmem_cache_t *skc, int flags)
{
spl_kmem_slab_t *sks;
void *base;
uint32_t obj_size;
base = kv_alloc(skc, skc->skc_slab_size, flags);
if (base == NULL)
return (NULL);
sks = (spl_kmem_slab_t *)base;
sks->sks_magic = SKS_MAGIC;
sks->sks_objs = skc->skc_slab_objs;
sks->sks_age = jiffies;
sks->sks_cache = skc;
INIT_LIST_HEAD(&sks->sks_list);
INIT_LIST_HEAD(&sks->sks_free_list);
sks->sks_ref = 0;
obj_size = spl_obj_size(skc);
for (int i = 0; i < sks->sks_objs; i++) {
void *obj = base + spl_sks_size(skc) + (i * obj_size);
ASSERT(IS_P2ALIGNED(obj, skc->skc_obj_align));
spl_kmem_obj_t *sko = spl_sko_from_obj(skc, obj);
sko->sko_addr = obj;
sko->sko_magic = SKO_MAGIC;
sko->sko_slab = sks;
INIT_LIST_HEAD(&sko->sko_list);
list_add_tail(&sko->sko_list, &sks->sks_free_list);
}
return (sks);
}
/*
* Remove a slab from complete or partial list, it must be called with
* the 'skc->skc_lock' held but the actual free must be performed
* outside the lock to prevent deadlocking on vmem addresses.
*/
static void
spl_slab_free(spl_kmem_slab_t *sks,
struct list_head *sks_list, struct list_head *sko_list)
{
spl_kmem_cache_t *skc;
ASSERT(sks->sks_magic == SKS_MAGIC);
ASSERT(sks->sks_ref == 0);
skc = sks->sks_cache;
ASSERT(skc->skc_magic == SKC_MAGIC);
/*
* Update slab/objects counters in the cache, then remove the
* slab from the skc->skc_partial_list. Finally add the slab
* and all its objects in to the private work lists where the
* destructors will be called and the memory freed to the system.
*/
skc->skc_obj_total -= sks->sks_objs;
skc->skc_slab_total--;
list_del(&sks->sks_list);
list_add(&sks->sks_list, sks_list);
list_splice_init(&sks->sks_free_list, sko_list);
}
/*
* Reclaim empty slabs at the end of the partial list.
*/
static void
spl_slab_reclaim(spl_kmem_cache_t *skc)
{
spl_kmem_slab_t *sks = NULL, *m = NULL;
spl_kmem_obj_t *sko = NULL, *n = NULL;
LIST_HEAD(sks_list);
LIST_HEAD(sko_list);
/*
* Empty slabs and objects must be moved to a private list so they
* can be safely freed outside the spin lock. All empty slabs are
* at the end of skc->skc_partial_list, therefore once a non-empty
* slab is found we can stop scanning.
*/
spin_lock(&skc->skc_lock);
list_for_each_entry_safe_reverse(sks, m,
&skc->skc_partial_list, sks_list) {
if (sks->sks_ref > 0)
break;
spl_slab_free(sks, &sks_list, &sko_list);
}
spin_unlock(&skc->skc_lock);
/*
* The following two loops ensure all the object destructors are run,
* and the slabs themselves are freed. This is all done outside the
* skc->skc_lock since this allows the destructor to sleep, and
* allows us to perform a conditional reschedule when a freeing a
* large number of objects and slabs back to the system.
*/
list_for_each_entry_safe(sko, n, &sko_list, sko_list) {
ASSERT(sko->sko_magic == SKO_MAGIC);
}
list_for_each_entry_safe(sks, m, &sks_list, sks_list) {
ASSERT(sks->sks_magic == SKS_MAGIC);
kv_free(skc, sks, skc->skc_slab_size);
}
}
static spl_kmem_emergency_t *
spl_emergency_search(struct rb_root *root, void *obj)
{
struct rb_node *node = root->rb_node;
spl_kmem_emergency_t *ske;
unsigned long address = (unsigned long)obj;
while (node) {
ske = container_of(node, spl_kmem_emergency_t, ske_node);
if (address < ske->ske_obj)
node = node->rb_left;
else if (address > ske->ske_obj)
node = node->rb_right;
else
return (ske);
}
return (NULL);
}
static int
spl_emergency_insert(struct rb_root *root, spl_kmem_emergency_t *ske)
{
struct rb_node **new = &(root->rb_node), *parent = NULL;
spl_kmem_emergency_t *ske_tmp;
unsigned long address = ske->ske_obj;
while (*new) {
ske_tmp = container_of(*new, spl_kmem_emergency_t, ske_node);
parent = *new;
if (address < ske_tmp->ske_obj)
new = &((*new)->rb_left);
else if (address > ske_tmp->ske_obj)
new = &((*new)->rb_right);
else
return (0);
}
rb_link_node(&ske->ske_node, parent, new);
rb_insert_color(&ske->ske_node, root);
return (1);
}
/*
* Allocate a single emergency object and track it in a red black tree.
*/
static int
spl_emergency_alloc(spl_kmem_cache_t *skc, int flags, void **obj)
{
gfp_t lflags = kmem_flags_convert(flags);
spl_kmem_emergency_t *ske;
int order = get_order(skc->skc_obj_size);
int empty;
/* Last chance use a partial slab if one now exists */
spin_lock(&skc->skc_lock);
empty = list_empty(&skc->skc_partial_list);
spin_unlock(&skc->skc_lock);
if (!empty)
return (-EEXIST);
ske = kmalloc(sizeof (*ske), lflags);
if (ske == NULL)
return (-ENOMEM);
ske->ske_obj = __get_free_pages(lflags, order);
if (ske->ske_obj == 0) {
kfree(ske);
return (-ENOMEM);
}
spin_lock(&skc->skc_lock);
empty = spl_emergency_insert(&skc->skc_emergency_tree, ske);
if (likely(empty)) {
skc->skc_obj_total++;
skc->skc_obj_emergency++;
if (skc->skc_obj_emergency > skc->skc_obj_emergency_max)
skc->skc_obj_emergency_max = skc->skc_obj_emergency;
}
spin_unlock(&skc->skc_lock);
if (unlikely(!empty)) {
free_pages(ske->ske_obj, order);
kfree(ske);
return (-EINVAL);
}
*obj = (void *)ske->ske_obj;
return (0);
}
/*
* Locate the passed object in the red black tree and free it.
*/
static int
spl_emergency_free(spl_kmem_cache_t *skc, void *obj)
{
spl_kmem_emergency_t *ske;
int order = get_order(skc->skc_obj_size);
spin_lock(&skc->skc_lock);
ske = spl_emergency_search(&skc->skc_emergency_tree, obj);
if (ske) {
rb_erase(&ske->ske_node, &skc->skc_emergency_tree);
skc->skc_obj_emergency--;
skc->skc_obj_total--;
}
spin_unlock(&skc->skc_lock);
if (ske == NULL)
return (-ENOENT);
free_pages(ske->ske_obj, order);
kfree(ske);
return (0);
}
/*
* Release objects from the per-cpu magazine back to their slab. The flush
* argument contains the max number of entries to remove from the magazine.
*/
static void
spl_cache_flush(spl_kmem_cache_t *skc, spl_kmem_magazine_t *skm, int flush)
{
spin_lock(&skc->skc_lock);
ASSERT(skc->skc_magic == SKC_MAGIC);
ASSERT(skm->skm_magic == SKM_MAGIC);
int count = MIN(flush, skm->skm_avail);
for (int i = 0; i < count; i++)
spl_cache_shrink(skc, skm->skm_objs[i]);
skm->skm_avail -= count;
memmove(skm->skm_objs, &(skm->skm_objs[count]),
sizeof (void *) * skm->skm_avail);
spin_unlock(&skc->skc_lock);
}
/*
* Size a slab based on the size of each aligned object plus spl_kmem_obj_t.
* When on-slab we want to target spl_kmem_cache_obj_per_slab. However,
* for very small objects we may end up with more than this so as not
* to waste space in the minimal allocation of a single page.
*/
static int
spl_slab_size(spl_kmem_cache_t *skc, uint32_t *objs, uint32_t *size)
{
uint32_t sks_size, obj_size, max_size, tgt_size, tgt_objs;
sks_size = spl_sks_size(skc);
obj_size = spl_obj_size(skc);
max_size = (spl_kmem_cache_max_size * 1024 * 1024);
tgt_size = (spl_kmem_cache_obj_per_slab * obj_size + sks_size);
if (tgt_size <= max_size) {
tgt_objs = (tgt_size - sks_size) / obj_size;
} else {
tgt_objs = (max_size - sks_size) / obj_size;
tgt_size = (tgt_objs * obj_size) + sks_size;
}
if (tgt_objs == 0)
return (-ENOSPC);
*objs = tgt_objs;
*size = tgt_size;
return (0);
}
/*
* Make a guess at reasonable per-cpu magazine size based on the size of
* each object and the cost of caching N of them in each magazine. Long
* term this should really adapt based on an observed usage heuristic.
*/
static int
spl_magazine_size(spl_kmem_cache_t *skc)
{
uint32_t obj_size = spl_obj_size(skc);
int size;
if (spl_kmem_cache_magazine_size > 0)
return (MAX(MIN(spl_kmem_cache_magazine_size, 256), 2));
/* Per-magazine sizes below assume a 4Kib page size */
if (obj_size > (PAGE_SIZE * 256))
size = 4; /* Minimum 4Mib per-magazine */
else if (obj_size > (PAGE_SIZE * 32))
size = 16; /* Minimum 2Mib per-magazine */
else if (obj_size > (PAGE_SIZE))
size = 64; /* Minimum 256Kib per-magazine */
else if (obj_size > (PAGE_SIZE / 4))
size = 128; /* Minimum 128Kib per-magazine */
else
size = 256;
return (size);
}
/*
* Allocate a per-cpu magazine to associate with a specific core.
*/
static spl_kmem_magazine_t *
spl_magazine_alloc(spl_kmem_cache_t *skc, int cpu)
{
spl_kmem_magazine_t *skm;
int size = sizeof (spl_kmem_magazine_t) +
sizeof (void *) * skc->skc_mag_size;
skm = kmalloc_node(size, GFP_KERNEL, cpu_to_node(cpu));
if (skm) {
skm->skm_magic = SKM_MAGIC;
skm->skm_avail = 0;
skm->skm_size = skc->skc_mag_size;
skm->skm_refill = skc->skc_mag_refill;
skm->skm_cache = skc;
skm->skm_cpu = cpu;
}
return (skm);
}
/*
* Free a per-cpu magazine associated with a specific core.
*/
static void
spl_magazine_free(spl_kmem_magazine_t *skm)
{
ASSERT(skm->skm_magic == SKM_MAGIC);
ASSERT(skm->skm_avail == 0);
kfree(skm);
}
/*
* Create all pre-cpu magazines of reasonable sizes.
*/
static int
spl_magazine_create(spl_kmem_cache_t *skc)
{
int i = 0;
ASSERT((skc->skc_flags & KMC_SLAB) == 0);
skc->skc_mag = kzalloc(sizeof (spl_kmem_magazine_t *) *
num_possible_cpus(), kmem_flags_convert(KM_SLEEP));
skc->skc_mag_size = spl_magazine_size(skc);
skc->skc_mag_refill = (skc->skc_mag_size + 1) / 2;
for_each_possible_cpu(i) {
skc->skc_mag[i] = spl_magazine_alloc(skc, i);
if (!skc->skc_mag[i]) {
for (i--; i >= 0; i--)
spl_magazine_free(skc->skc_mag[i]);
kfree(skc->skc_mag);
return (-ENOMEM);
}
}
return (0);
}
/*
* Destroy all pre-cpu magazines.
*/
static void
spl_magazine_destroy(spl_kmem_cache_t *skc)
{
spl_kmem_magazine_t *skm;
int i = 0;
ASSERT((skc->skc_flags & KMC_SLAB) == 0);
for_each_possible_cpu(i) {
skm = skc->skc_mag[i];
spl_cache_flush(skc, skm, skm->skm_avail);
spl_magazine_free(skm);
}
kfree(skc->skc_mag);
}
/*
* Create a object cache based on the following arguments:
* name cache name
* size cache object size
* align cache object alignment
* ctor cache object constructor
* dtor cache object destructor
* reclaim cache object reclaim
* priv cache private data for ctor/dtor/reclaim
* vmp unused must be NULL
* flags
* KMC_KVMEM Force kvmem backed SPL cache
* KMC_SLAB Force Linux slab backed cache
* KMC_NODEBUG Disable debugging (unsupported)
*/
spl_kmem_cache_t *
spl_kmem_cache_create(char *name, size_t size, size_t align,
spl_kmem_ctor_t ctor, spl_kmem_dtor_t dtor, void *reclaim,
void *priv, void *vmp, int flags)
{
gfp_t lflags = kmem_flags_convert(KM_SLEEP);
spl_kmem_cache_t *skc;
int rc;
/*
* Unsupported flags
*/
ASSERT(vmp == NULL);
ASSERT(reclaim == NULL);
might_sleep();
skc = kzalloc(sizeof (*skc), lflags);
if (skc == NULL)
return (NULL);
skc->skc_magic = SKC_MAGIC;
skc->skc_name_size = strlen(name) + 1;
skc->skc_name = (char *)kmalloc(skc->skc_name_size, lflags);
if (skc->skc_name == NULL) {
kfree(skc);
return (NULL);
}
strncpy(skc->skc_name, name, skc->skc_name_size);
skc->skc_ctor = ctor;
skc->skc_dtor = dtor;
skc->skc_private = priv;
skc->skc_vmp = vmp;
skc->skc_linux_cache = NULL;
skc->skc_flags = flags;
skc->skc_obj_size = size;
skc->skc_obj_align = SPL_KMEM_CACHE_ALIGN;
atomic_set(&skc->skc_ref, 0);
INIT_LIST_HEAD(&skc->skc_list);
INIT_LIST_HEAD(&skc->skc_complete_list);
INIT_LIST_HEAD(&skc->skc_partial_list);
skc->skc_emergency_tree = RB_ROOT;
spin_lock_init(&skc->skc_lock);
init_waitqueue_head(&skc->skc_waitq);
skc->skc_slab_fail = 0;
skc->skc_slab_create = 0;
skc->skc_slab_destroy = 0;
skc->skc_slab_total = 0;
skc->skc_slab_alloc = 0;
skc->skc_slab_max = 0;
skc->skc_obj_total = 0;
skc->skc_obj_alloc = 0;
skc->skc_obj_max = 0;
skc->skc_obj_deadlock = 0;
skc->skc_obj_emergency = 0;
skc->skc_obj_emergency_max = 0;
rc = percpu_counter_init_common(&skc->skc_linux_alloc, 0,
GFP_KERNEL);
if (rc != 0) {
kfree(skc);
return (NULL);
}
/*
* Verify the requested alignment restriction is sane.
*/
if (align) {
VERIFY(ISP2(align));
VERIFY3U(align, >=, SPL_KMEM_CACHE_ALIGN);
VERIFY3U(align, <=, PAGE_SIZE);
skc->skc_obj_align = align;
}
/*
* When no specific type of slab is requested (kmem, vmem, or
* linuxslab) then select a cache type based on the object size
* and default tunables.
*/
if (!(skc->skc_flags & (KMC_SLAB | KMC_KVMEM))) {
if (spl_kmem_cache_slab_limit &&
size <= (size_t)spl_kmem_cache_slab_limit) {
/*
* Objects smaller than spl_kmem_cache_slab_limit can
* use the Linux slab for better space-efficiency.
*/
skc->skc_flags |= KMC_SLAB;
} else {
/*
* All other objects are considered large and are
* placed on kvmem backed slabs.
*/
skc->skc_flags |= KMC_KVMEM;
}
}
/*
* Given the type of slab allocate the required resources.
*/
if (skc->skc_flags & KMC_KVMEM) {
rc = spl_slab_size(skc,
&skc->skc_slab_objs, &skc->skc_slab_size);
if (rc)
goto out;
rc = spl_magazine_create(skc);
if (rc)
goto out;
} else {
unsigned long slabflags = 0;
if (size > (SPL_MAX_KMEM_ORDER_NR_PAGES * PAGE_SIZE)) {
rc = EINVAL;
goto out;
}
#if defined(SLAB_USERCOPY)
/*
* Required for PAX-enabled kernels if the slab is to be
* used for copying between user and kernel space.
*/
slabflags |= SLAB_USERCOPY;
#endif
#if defined(HAVE_KMEM_CACHE_CREATE_USERCOPY)
/*
* Newer grsec patchset uses kmem_cache_create_usercopy()
* instead of SLAB_USERCOPY flag
*/
skc->skc_linux_cache = kmem_cache_create_usercopy(
skc->skc_name, size, align, slabflags, 0, size, NULL);
#else
skc->skc_linux_cache = kmem_cache_create(
skc->skc_name, size, align, slabflags, NULL);
#endif
if (skc->skc_linux_cache == NULL) {
rc = ENOMEM;
goto out;
}
}
down_write(&spl_kmem_cache_sem);
list_add_tail(&skc->skc_list, &spl_kmem_cache_list);
up_write(&spl_kmem_cache_sem);
return (skc);
out:
kfree(skc->skc_name);
percpu_counter_destroy(&skc->skc_linux_alloc);
kfree(skc);
return (NULL);
}
EXPORT_SYMBOL(spl_kmem_cache_create);
/*
* Register a move callback for cache defragmentation.
* XXX: Unimplemented but harmless to stub out for now.
*/
void
spl_kmem_cache_set_move(spl_kmem_cache_t *skc,
kmem_cbrc_t (move)(void *, void *, size_t, void *))
{
ASSERT(move != NULL);
}
EXPORT_SYMBOL(spl_kmem_cache_set_move);
/*
* Destroy a cache and all objects associated with the cache.
*/
void
spl_kmem_cache_destroy(spl_kmem_cache_t *skc)
{
DECLARE_WAIT_QUEUE_HEAD(wq);
taskqid_t id;
ASSERT(skc->skc_magic == SKC_MAGIC);
ASSERT(skc->skc_flags & (KMC_KVMEM | KMC_SLAB));
down_write(&spl_kmem_cache_sem);
list_del_init(&skc->skc_list);
up_write(&spl_kmem_cache_sem);
/* Cancel any and wait for any pending delayed tasks */
VERIFY(!test_and_set_bit(KMC_BIT_DESTROY, &skc->skc_flags));
spin_lock(&skc->skc_lock);
id = skc->skc_taskqid;
spin_unlock(&skc->skc_lock);
taskq_cancel_id(spl_kmem_cache_taskq, id);
/*
* Wait until all current callers complete, this is mainly
* to catch the case where a low memory situation triggers a
* cache reaping action which races with this destroy.
*/
wait_event(wq, atomic_read(&skc->skc_ref) == 0);
if (skc->skc_flags & KMC_KVMEM) {
spl_magazine_destroy(skc);
spl_slab_reclaim(skc);
} else {
ASSERT(skc->skc_flags & KMC_SLAB);
kmem_cache_destroy(skc->skc_linux_cache);
}
spin_lock(&skc->skc_lock);
/*
* Validate there are no objects in use and free all the
* spl_kmem_slab_t, spl_kmem_obj_t, and object buffers.
*/
ASSERT3U(skc->skc_slab_alloc, ==, 0);
ASSERT3U(skc->skc_obj_alloc, ==, 0);
ASSERT3U(skc->skc_slab_total, ==, 0);
ASSERT3U(skc->skc_obj_total, ==, 0);
ASSERT3U(skc->skc_obj_emergency, ==, 0);
ASSERT(list_empty(&skc->skc_complete_list));
ASSERT3U(percpu_counter_sum(&skc->skc_linux_alloc), ==, 0);
percpu_counter_destroy(&skc->skc_linux_alloc);
spin_unlock(&skc->skc_lock);
kfree(skc->skc_name);
kfree(skc);
}
EXPORT_SYMBOL(spl_kmem_cache_destroy);
/*
* Allocate an object from a slab attached to the cache. This is used to
* repopulate the per-cpu magazine caches in batches when they run low.
*/
static void *
spl_cache_obj(spl_kmem_cache_t *skc, spl_kmem_slab_t *sks)
{
spl_kmem_obj_t *sko;
ASSERT(skc->skc_magic == SKC_MAGIC);
ASSERT(sks->sks_magic == SKS_MAGIC);
sko = list_entry(sks->sks_free_list.next, spl_kmem_obj_t, sko_list);
ASSERT(sko->sko_magic == SKO_MAGIC);
ASSERT(sko->sko_addr != NULL);
/* Remove from sks_free_list */
list_del_init(&sko->sko_list);
sks->sks_age = jiffies;
sks->sks_ref++;
skc->skc_obj_alloc++;
/* Track max obj usage statistics */
if (skc->skc_obj_alloc > skc->skc_obj_max)
skc->skc_obj_max = skc->skc_obj_alloc;
/* Track max slab usage statistics */
if (sks->sks_ref == 1) {
skc->skc_slab_alloc++;
if (skc->skc_slab_alloc > skc->skc_slab_max)
skc->skc_slab_max = skc->skc_slab_alloc;
}
return (sko->sko_addr);
}
/*
* Generic slab allocation function to run by the global work queues.
* It is responsible for allocating a new slab, linking it in to the list
* of partial slabs, and then waking any waiters.
*/
static int
__spl_cache_grow(spl_kmem_cache_t *skc, int flags)
{
spl_kmem_slab_t *sks;
fstrans_cookie_t cookie = spl_fstrans_mark();
sks = spl_slab_alloc(skc, flags);
spl_fstrans_unmark(cookie);
spin_lock(&skc->skc_lock);
if (sks) {
skc->skc_slab_total++;
skc->skc_obj_total += sks->sks_objs;
list_add_tail(&sks->sks_list, &skc->skc_partial_list);
smp_mb__before_atomic();
clear_bit(KMC_BIT_DEADLOCKED, &skc->skc_flags);
smp_mb__after_atomic();
}
spin_unlock(&skc->skc_lock);
return (sks == NULL ? -ENOMEM : 0);
}
static void
spl_cache_grow_work(void *data)
{
spl_kmem_alloc_t *ska = (spl_kmem_alloc_t *)data;
spl_kmem_cache_t *skc = ska->ska_cache;
int error = __spl_cache_grow(skc, ska->ska_flags);
atomic_dec(&skc->skc_ref);
smp_mb__before_atomic();
clear_bit(KMC_BIT_GROWING, &skc->skc_flags);
smp_mb__after_atomic();
if (error == 0)
wake_up_all(&skc->skc_waitq);
kfree(ska);
}
/*
* Returns non-zero when a new slab should be available.
*/
static int
spl_cache_grow_wait(spl_kmem_cache_t *skc)
{
return (!test_bit(KMC_BIT_GROWING, &skc->skc_flags));
}
/*
* No available objects on any slabs, create a new slab. Note that this
* functionality is disabled for KMC_SLAB caches which are backed by the
* Linux slab.
*/
static int
spl_cache_grow(spl_kmem_cache_t *skc, int flags, void **obj)
{
int remaining, rc = 0;
ASSERT0(flags & ~KM_PUBLIC_MASK);
ASSERT(skc->skc_magic == SKC_MAGIC);
ASSERT((skc->skc_flags & KMC_SLAB) == 0);
might_sleep();
*obj = NULL;
/*
* Before allocating a new slab wait for any reaping to complete and
* then return so the local magazine can be rechecked for new objects.
*/
if (test_bit(KMC_BIT_REAPING, &skc->skc_flags)) {
rc = spl_wait_on_bit(&skc->skc_flags, KMC_BIT_REAPING,
TASK_UNINTERRUPTIBLE);
return (rc ? rc : -EAGAIN);
}
/*
* Note: It would be nice to reduce the overhead of context switch
* and improve NUMA locality, by trying to allocate a new slab in the
* current process context with KM_NOSLEEP flag.
*
* However, this can't be applied to vmem/kvmem due to a bug that
* spl_vmalloc() doesn't honor gfp flags in page table allocation.
*/
/*
* This is handled by dispatching a work request to the global work
* queue. This allows us to asynchronously allocate a new slab while
* retaining the ability to safely fall back to a smaller synchronous
* allocations to ensure forward progress is always maintained.
*/
if (test_and_set_bit(KMC_BIT_GROWING, &skc->skc_flags) == 0) {
spl_kmem_alloc_t *ska;
ska = kmalloc(sizeof (*ska), kmem_flags_convert(flags));
if (ska == NULL) {
clear_bit_unlock(KMC_BIT_GROWING, &skc->skc_flags);
smp_mb__after_atomic();
wake_up_all(&skc->skc_waitq);
return (-ENOMEM);
}
atomic_inc(&skc->skc_ref);
ska->ska_cache = skc;
ska->ska_flags = flags;
taskq_init_ent(&ska->ska_tqe);
taskq_dispatch_ent(spl_kmem_cache_taskq,
spl_cache_grow_work, ska, 0, &ska->ska_tqe);
}
/*
* The goal here is to only detect the rare case where a virtual slab
* allocation has deadlocked. We must be careful to minimize the use
* of emergency objects which are more expensive to track. Therefore,
* we set a very long timeout for the asynchronous allocation and if
* the timeout is reached the cache is flagged as deadlocked. From
* this point only new emergency objects will be allocated until the
* asynchronous allocation completes and clears the deadlocked flag.
*/
if (test_bit(KMC_BIT_DEADLOCKED, &skc->skc_flags)) {
rc = spl_emergency_alloc(skc, flags, obj);
} else {
remaining = wait_event_timeout(skc->skc_waitq,
spl_cache_grow_wait(skc), HZ / 10);
if (!remaining) {
spin_lock(&skc->skc_lock);
if (test_bit(KMC_BIT_GROWING, &skc->skc_flags)) {
set_bit(KMC_BIT_DEADLOCKED, &skc->skc_flags);
skc->skc_obj_deadlock++;
}
spin_unlock(&skc->skc_lock);
}
rc = -ENOMEM;
}
return (rc);
}
/*
* Refill a per-cpu magazine with objects from the slabs for this cache.
* Ideally the magazine can be repopulated using existing objects which have
* been released, however if we are unable to locate enough free objects new
* slabs of objects will be created. On success NULL is returned, otherwise
* the address of a single emergency object is returned for use by the caller.
*/
static void *
spl_cache_refill(spl_kmem_cache_t *skc, spl_kmem_magazine_t *skm, int flags)
{
spl_kmem_slab_t *sks;
int count = 0, rc, refill;
void *obj = NULL;
ASSERT(skc->skc_magic == SKC_MAGIC);
ASSERT(skm->skm_magic == SKM_MAGIC);
refill = MIN(skm->skm_refill, skm->skm_size - skm->skm_avail);
spin_lock(&skc->skc_lock);
while (refill > 0) {
/* No slabs available we may need to grow the cache */
if (list_empty(&skc->skc_partial_list)) {
spin_unlock(&skc->skc_lock);
local_irq_enable();
rc = spl_cache_grow(skc, flags, &obj);
local_irq_disable();
/* Emergency object for immediate use by caller */
if (rc == 0 && obj != NULL)
return (obj);
if (rc)
goto out;
/* Rescheduled to different CPU skm is not local */
if (skm != skc->skc_mag[smp_processor_id()])
goto out;
/*
* Potentially rescheduled to the same CPU but
* allocations may have occurred from this CPU while
* we were sleeping so recalculate max refill.
*/
refill = MIN(refill, skm->skm_size - skm->skm_avail);
spin_lock(&skc->skc_lock);
continue;
}
/* Grab the next available slab */
sks = list_entry((&skc->skc_partial_list)->next,
spl_kmem_slab_t, sks_list);
ASSERT(sks->sks_magic == SKS_MAGIC);
ASSERT(sks->sks_ref < sks->sks_objs);
ASSERT(!list_empty(&sks->sks_free_list));
/*
* Consume as many objects as needed to refill the requested
* cache. We must also be careful not to overfill it.
*/
while (sks->sks_ref < sks->sks_objs && refill-- > 0 &&
++count) {
ASSERT(skm->skm_avail < skm->skm_size);
ASSERT(count < skm->skm_size);
skm->skm_objs[skm->skm_avail++] =
spl_cache_obj(skc, sks);
}
/* Move slab to skc_complete_list when full */
if (sks->sks_ref == sks->sks_objs) {
list_del(&sks->sks_list);
list_add(&sks->sks_list, &skc->skc_complete_list);
}
}
spin_unlock(&skc->skc_lock);
out:
return (NULL);
}
/*
* Release an object back to the slab from which it came.
*/
static void
spl_cache_shrink(spl_kmem_cache_t *skc, void *obj)
{
spl_kmem_slab_t *sks = NULL;
spl_kmem_obj_t *sko = NULL;
ASSERT(skc->skc_magic == SKC_MAGIC);
sko = spl_sko_from_obj(skc, obj);
ASSERT(sko->sko_magic == SKO_MAGIC);
sks = sko->sko_slab;
ASSERT(sks->sks_magic == SKS_MAGIC);
ASSERT(sks->sks_cache == skc);
list_add(&sko->sko_list, &sks->sks_free_list);
sks->sks_age = jiffies;
sks->sks_ref--;
skc->skc_obj_alloc--;
/*
* Move slab to skc_partial_list when no longer full. Slabs
* are added to the head to keep the partial list is quasi-full
* sorted order. Fuller at the head, emptier at the tail.
*/
if (sks->sks_ref == (sks->sks_objs - 1)) {
list_del(&sks->sks_list);
list_add(&sks->sks_list, &skc->skc_partial_list);
}
/*
* Move empty slabs to the end of the partial list so
* they can be easily found and freed during reclamation.
*/
if (sks->sks_ref == 0) {
list_del(&sks->sks_list);
list_add_tail(&sks->sks_list, &skc->skc_partial_list);
skc->skc_slab_alloc--;
}
}
/*
* Allocate an object from the per-cpu magazine, or if the magazine
* is empty directly allocate from a slab and repopulate the magazine.
*/
void *
spl_kmem_cache_alloc(spl_kmem_cache_t *skc, int flags)
{
spl_kmem_magazine_t *skm;
void *obj = NULL;
ASSERT0(flags & ~KM_PUBLIC_MASK);
ASSERT(skc->skc_magic == SKC_MAGIC);
ASSERT(!test_bit(KMC_BIT_DESTROY, &skc->skc_flags));
/*
* Allocate directly from a Linux slab. All optimizations are left
* to the underlying cache we only need to guarantee that KM_SLEEP
* callers will never fail.
*/
if (skc->skc_flags & KMC_SLAB) {
struct kmem_cache *slc = skc->skc_linux_cache;
do {
obj = kmem_cache_alloc(slc, kmem_flags_convert(flags));
} while ((obj == NULL) && !(flags & KM_NOSLEEP));
if (obj != NULL) {
/*
* Even though we leave everything up to the
* underlying cache we still keep track of
* how many objects we've allocated in it for
* better debuggability.
*/
percpu_counter_inc(&skc->skc_linux_alloc);
}
goto ret;
}
local_irq_disable();
restart:
/*
* Safe to update per-cpu structure without lock, but
* in the restart case we must be careful to reacquire
* the local magazine since this may have changed
* when we need to grow the cache.
*/
skm = skc->skc_mag[smp_processor_id()];
ASSERT(skm->skm_magic == SKM_MAGIC);
if (likely(skm->skm_avail)) {
/* Object available in CPU cache, use it */
obj = skm->skm_objs[--skm->skm_avail];
} else {
obj = spl_cache_refill(skc, skm, flags);
if ((obj == NULL) && !(flags & KM_NOSLEEP))
goto restart;
local_irq_enable();
goto ret;
}
local_irq_enable();
ASSERT(obj);
ASSERT(IS_P2ALIGNED(obj, skc->skc_obj_align));
ret:
/* Pre-emptively migrate object to CPU L1 cache */
if (obj) {
if (obj && skc->skc_ctor)
skc->skc_ctor(obj, skc->skc_private, flags);
else
prefetchw(obj);
}
return (obj);
}
EXPORT_SYMBOL(spl_kmem_cache_alloc);
/*
* Free an object back to the local per-cpu magazine, there is no
* guarantee that this is the same magazine the object was originally
* allocated from. We may need to flush entire from the magazine
* back to the slabs to make space.
*/
void
spl_kmem_cache_free(spl_kmem_cache_t *skc, void *obj)
{
spl_kmem_magazine_t *skm;
unsigned long flags;
int do_reclaim = 0;
int do_emergency = 0;
ASSERT(skc->skc_magic == SKC_MAGIC);
ASSERT(!test_bit(KMC_BIT_DESTROY, &skc->skc_flags));
/*
* Run the destructor
*/
if (skc->skc_dtor)
skc->skc_dtor(obj, skc->skc_private);
/*
* Free the object from the Linux underlying Linux slab.
*/
if (skc->skc_flags & KMC_SLAB) {
kmem_cache_free(skc->skc_linux_cache, obj);
percpu_counter_dec(&skc->skc_linux_alloc);
return;
}
/*
* While a cache has outstanding emergency objects all freed objects
* must be checked. However, since emergency objects will never use
* a virtual address these objects can be safely excluded as an
* optimization.
*/
if (!is_vmalloc_addr(obj)) {
spin_lock(&skc->skc_lock);
do_emergency = (skc->skc_obj_emergency > 0);
spin_unlock(&skc->skc_lock);
if (do_emergency && (spl_emergency_free(skc, obj) == 0))
return;
}
local_irq_save(flags);
/*
* Safe to update per-cpu structure without lock, but
* no remote memory allocation tracking is being performed
* it is entirely possible to allocate an object from one
* CPU cache and return it to another.
*/
skm = skc->skc_mag[smp_processor_id()];
ASSERT(skm->skm_magic == SKM_MAGIC);
/*
* Per-CPU cache full, flush it to make space for this object,
* this may result in an empty slab which can be reclaimed once
* interrupts are re-enabled.
*/
if (unlikely(skm->skm_avail >= skm->skm_size)) {
spl_cache_flush(skc, skm, skm->skm_refill);
do_reclaim = 1;
}
/* Available space in cache, use it */
skm->skm_objs[skm->skm_avail++] = obj;
local_irq_restore(flags);
if (do_reclaim)
spl_slab_reclaim(skc);
}
EXPORT_SYMBOL(spl_kmem_cache_free);
/*
* Depending on how many and which objects are released it may simply
* repopulate the local magazine which will then need to age-out. Objects
* which cannot fit in the magazine will be released back to their slabs
* which will also need to age out before being released. This is all just
* best effort and we do not want to thrash creating and destroying slabs.
*/
void
spl_kmem_cache_reap_now(spl_kmem_cache_t *skc)
{
ASSERT(skc->skc_magic == SKC_MAGIC);
ASSERT(!test_bit(KMC_BIT_DESTROY, &skc->skc_flags));
if (skc->skc_flags & KMC_SLAB)
return;
atomic_inc(&skc->skc_ref);
/*
* Prevent concurrent cache reaping when contended.
*/
if (test_and_set_bit(KMC_BIT_REAPING, &skc->skc_flags))
goto out;
/* Reclaim from the magazine and free all now empty slabs. */
unsigned long irq_flags;
local_irq_save(irq_flags);
spl_kmem_magazine_t *skm = skc->skc_mag[smp_processor_id()];
spl_cache_flush(skc, skm, skm->skm_avail);
local_irq_restore(irq_flags);
spl_slab_reclaim(skc);
clear_bit_unlock(KMC_BIT_REAPING, &skc->skc_flags);
smp_mb__after_atomic();
wake_up_bit(&skc->skc_flags, KMC_BIT_REAPING);
out:
atomic_dec(&skc->skc_ref);
}
EXPORT_SYMBOL(spl_kmem_cache_reap_now);
/*
* This is stubbed out for code consistency with other platforms. There
* is existing logic to prevent concurrent reaping so while this is ugly
* it should do no harm.
*/
int
spl_kmem_cache_reap_active()
{
return (0);
}
EXPORT_SYMBOL(spl_kmem_cache_reap_active);
/*
* Reap all free slabs from all registered caches.
*/
void
spl_kmem_reap(void)
{
spl_kmem_cache_t *skc = NULL;
down_read(&spl_kmem_cache_sem);
list_for_each_entry(skc, &spl_kmem_cache_list, skc_list) {
spl_kmem_cache_reap_now(skc);
}
up_read(&spl_kmem_cache_sem);
}
EXPORT_SYMBOL(spl_kmem_reap);
int
spl_kmem_cache_init(void)
{
init_rwsem(&spl_kmem_cache_sem);
INIT_LIST_HEAD(&spl_kmem_cache_list);
spl_kmem_cache_taskq = taskq_create("spl_kmem_cache",
spl_kmem_cache_kmem_threads, maxclsyspri,
spl_kmem_cache_kmem_threads * 8, INT_MAX,
TASKQ_PREPOPULATE | TASKQ_DYNAMIC);
return (0);
}
void
spl_kmem_cache_fini(void)
{
taskq_destroy(spl_kmem_cache_taskq);
}
diff --git a/module/os/linux/spl/spl-kstat.c b/module/os/linux/spl/spl-kstat.c
index 0c46708326d8..a417d4d7c358 100644
--- a/module/os/linux/spl/spl-kstat.c
+++ b/module/os/linux/spl/spl-kstat.c
@@ -1,715 +1,715 @@
/*
* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
* Copyright (C) 2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
* UCRL-CODE-235197
*
* This file is part of the SPL, Solaris Porting Layer.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*
* Solaris Porting Layer (SPL) Kstat Implementation.
*
* Links to Illumos.org for more information on kstat function:
* [1] https://illumos.org/man/1M/kstat
* [2] https://illumos.org/man/9f/kstat_create
*/
#include <linux/seq_file.h>
#include <sys/kstat.h>
#include <sys/vmem.h>
#include <sys/cmn_err.h>
#include <sys/sysmacros.h>
static kmutex_t kstat_module_lock;
static struct list_head kstat_module_list;
static kid_t kstat_id;
static int
kstat_resize_raw(kstat_t *ksp)
{
if (ksp->ks_raw_bufsize == KSTAT_RAW_MAX)
return (ENOMEM);
vmem_free(ksp->ks_raw_buf, ksp->ks_raw_bufsize);
ksp->ks_raw_bufsize = MIN(ksp->ks_raw_bufsize * 2, KSTAT_RAW_MAX);
ksp->ks_raw_buf = vmem_alloc(ksp->ks_raw_bufsize, KM_SLEEP);
return (0);
}
static int
kstat_seq_show_headers(struct seq_file *f)
{
kstat_t *ksp = (kstat_t *)f->private;
int rc = 0;
ASSERT(ksp->ks_magic == KS_MAGIC);
seq_printf(f, "%d %d 0x%02x %d %d %lld %lld\n",
ksp->ks_kid, ksp->ks_type, ksp->ks_flags,
ksp->ks_ndata, (int)ksp->ks_data_size,
ksp->ks_crtime, ksp->ks_snaptime);
switch (ksp->ks_type) {
case KSTAT_TYPE_RAW:
restart:
if (ksp->ks_raw_ops.headers) {
rc = ksp->ks_raw_ops.headers(
ksp->ks_raw_buf, ksp->ks_raw_bufsize);
if (rc == ENOMEM && !kstat_resize_raw(ksp))
goto restart;
if (!rc)
seq_puts(f, ksp->ks_raw_buf);
} else {
seq_printf(f, "raw data\n");
}
break;
case KSTAT_TYPE_NAMED:
seq_printf(f, "%-31s %-4s %s\n",
"name", "type", "data");
break;
case KSTAT_TYPE_INTR:
seq_printf(f, "%-8s %-8s %-8s %-8s %-8s\n",
"hard", "soft", "watchdog",
"spurious", "multsvc");
break;
case KSTAT_TYPE_IO:
seq_printf(f,
"%-8s %-8s %-8s %-8s %-8s %-8s "
"%-8s %-8s %-8s %-8s %-8s %-8s\n",
"nread", "nwritten", "reads", "writes",
"wtime", "wlentime", "wupdate",
"rtime", "rlentime", "rupdate",
"wcnt", "rcnt");
break;
case KSTAT_TYPE_TIMER:
seq_printf(f,
"%-31s %-8s "
"%-8s %-8s %-8s %-8s %-8s\n",
"name", "events", "elapsed",
"min", "max", "start", "stop");
break;
default:
PANIC("Undefined kstat type %d\n", ksp->ks_type);
}
return (-rc);
}
static int
kstat_seq_show_raw(struct seq_file *f, unsigned char *p, int l)
{
int i, j;
for (i = 0; ; i++) {
seq_printf(f, "%03x:", i);
for (j = 0; j < 16; j++) {
if (i * 16 + j >= l) {
seq_printf(f, "\n");
goto out;
}
seq_printf(f, " %02x", (unsigned char)p[i * 16 + j]);
}
seq_printf(f, "\n");
}
out:
return (0);
}
static int
kstat_seq_show_named(struct seq_file *f, kstat_named_t *knp)
{
seq_printf(f, "%-31s %-4d ", knp->name, knp->data_type);
switch (knp->data_type) {
case KSTAT_DATA_CHAR:
knp->value.c[15] = '\0'; /* NULL terminate */
seq_printf(f, "%-16s", knp->value.c);
break;
/*
* NOTE - We need to be more careful able what tokens are
* used for each arch, for now this is correct for x86_64.
*/
case KSTAT_DATA_INT32:
seq_printf(f, "%d", knp->value.i32);
break;
case KSTAT_DATA_UINT32:
seq_printf(f, "%u", knp->value.ui32);
break;
case KSTAT_DATA_INT64:
seq_printf(f, "%lld", (signed long long)knp->value.i64);
break;
case KSTAT_DATA_UINT64:
seq_printf(f, "%llu",
(unsigned long long)knp->value.ui64);
break;
case KSTAT_DATA_LONG:
seq_printf(f, "%ld", knp->value.l);
break;
case KSTAT_DATA_ULONG:
seq_printf(f, "%lu", knp->value.ul);
break;
case KSTAT_DATA_STRING:
KSTAT_NAMED_STR_PTR(knp)
[KSTAT_NAMED_STR_BUFLEN(knp)-1] = '\0';
seq_printf(f, "%s", KSTAT_NAMED_STR_PTR(knp));
break;
default:
PANIC("Undefined kstat data type %d\n", knp->data_type);
}
seq_printf(f, "\n");
return (0);
}
static int
kstat_seq_show_intr(struct seq_file *f, kstat_intr_t *kip)
{
seq_printf(f, "%-8u %-8u %-8u %-8u %-8u\n",
kip->intrs[KSTAT_INTR_HARD],
kip->intrs[KSTAT_INTR_SOFT],
kip->intrs[KSTAT_INTR_WATCHDOG],
kip->intrs[KSTAT_INTR_SPURIOUS],
kip->intrs[KSTAT_INTR_MULTSVC]);
return (0);
}
static int
kstat_seq_show_io(struct seq_file *f, kstat_io_t *kip)
{
/* though wlentime & friends are signed, they will never be negative */
seq_printf(f,
"%-8llu %-8llu %-8u %-8u %-8llu %-8llu "
"%-8llu %-8llu %-8llu %-8llu %-8u %-8u\n",
kip->nread, kip->nwritten,
kip->reads, kip->writes,
kip->wtime, kip->wlentime, kip->wlastupdate,
kip->rtime, kip->rlentime, kip->rlastupdate,
kip->wcnt, kip->rcnt);
return (0);
}
static int
kstat_seq_show_timer(struct seq_file *f, kstat_timer_t *ktp)
{
seq_printf(f,
"%-31s %-8llu %-8llu %-8llu %-8llu %-8llu %-8llu\n",
ktp->name, ktp->num_events, ktp->elapsed_time,
ktp->min_time, ktp->max_time,
ktp->start_time, ktp->stop_time);
return (0);
}
static int
kstat_seq_show(struct seq_file *f, void *p)
{
kstat_t *ksp = (kstat_t *)f->private;
int rc = 0;
ASSERT(ksp->ks_magic == KS_MAGIC);
switch (ksp->ks_type) {
case KSTAT_TYPE_RAW:
restart:
if (ksp->ks_raw_ops.data) {
rc = ksp->ks_raw_ops.data(
ksp->ks_raw_buf, ksp->ks_raw_bufsize, p);
if (rc == ENOMEM && !kstat_resize_raw(ksp))
goto restart;
if (!rc)
seq_puts(f, ksp->ks_raw_buf);
} else {
ASSERT(ksp->ks_ndata == 1);
rc = kstat_seq_show_raw(f, ksp->ks_data,
ksp->ks_data_size);
}
break;
case KSTAT_TYPE_NAMED:
rc = kstat_seq_show_named(f, (kstat_named_t *)p);
break;
case KSTAT_TYPE_INTR:
rc = kstat_seq_show_intr(f, (kstat_intr_t *)p);
break;
case KSTAT_TYPE_IO:
rc = kstat_seq_show_io(f, (kstat_io_t *)p);
break;
case KSTAT_TYPE_TIMER:
rc = kstat_seq_show_timer(f, (kstat_timer_t *)p);
break;
default:
PANIC("Undefined kstat type %d\n", ksp->ks_type);
}
return (-rc);
}
static int
kstat_default_update(kstat_t *ksp, int rw)
{
ASSERT(ksp != NULL);
if (rw == KSTAT_WRITE)
return (EACCES);
return (0);
}
static void *
kstat_seq_data_addr(kstat_t *ksp, loff_t n)
{
void *rc = NULL;
switch (ksp->ks_type) {
case KSTAT_TYPE_RAW:
if (ksp->ks_raw_ops.addr)
rc = ksp->ks_raw_ops.addr(ksp, n);
else
rc = ksp->ks_data;
break;
case KSTAT_TYPE_NAMED:
rc = ksp->ks_data + n * sizeof (kstat_named_t);
break;
case KSTAT_TYPE_INTR:
rc = ksp->ks_data + n * sizeof (kstat_intr_t);
break;
case KSTAT_TYPE_IO:
rc = ksp->ks_data + n * sizeof (kstat_io_t);
break;
case KSTAT_TYPE_TIMER:
rc = ksp->ks_data + n * sizeof (kstat_timer_t);
break;
default:
PANIC("Undefined kstat type %d\n", ksp->ks_type);
}
return (rc);
}
static void *
kstat_seq_start(struct seq_file *f, loff_t *pos)
{
loff_t n = *pos;
kstat_t *ksp = (kstat_t *)f->private;
ASSERT(ksp->ks_magic == KS_MAGIC);
mutex_enter(ksp->ks_lock);
if (ksp->ks_type == KSTAT_TYPE_RAW) {
ksp->ks_raw_bufsize = PAGE_SIZE;
ksp->ks_raw_buf = vmem_alloc(ksp->ks_raw_bufsize, KM_SLEEP);
}
/* Dynamically update kstat, on error existing kstats are used */
(void) ksp->ks_update(ksp, KSTAT_READ);
ksp->ks_snaptime = gethrtime();
if (!(ksp->ks_flags & KSTAT_FLAG_NO_HEADERS) && !n &&
kstat_seq_show_headers(f))
return (NULL);
if (n >= ksp->ks_ndata)
return (NULL);
return (kstat_seq_data_addr(ksp, n));
}
static void *
kstat_seq_next(struct seq_file *f, void *p, loff_t *pos)
{
kstat_t *ksp = (kstat_t *)f->private;
ASSERT(ksp->ks_magic == KS_MAGIC);
++*pos;
if (*pos >= ksp->ks_ndata)
return (NULL);
return (kstat_seq_data_addr(ksp, *pos));
}
static void
kstat_seq_stop(struct seq_file *f, void *v)
{
kstat_t *ksp = (kstat_t *)f->private;
ASSERT(ksp->ks_magic == KS_MAGIC);
if (ksp->ks_type == KSTAT_TYPE_RAW)
vmem_free(ksp->ks_raw_buf, ksp->ks_raw_bufsize);
mutex_exit(ksp->ks_lock);
}
-static struct seq_operations kstat_seq_ops = {
+static const struct seq_operations kstat_seq_ops = {
.show = kstat_seq_show,
.start = kstat_seq_start,
.next = kstat_seq_next,
.stop = kstat_seq_stop,
};
static kstat_module_t *
kstat_find_module(char *name)
{
kstat_module_t *module = NULL;
list_for_each_entry(module, &kstat_module_list, ksm_module_list) {
if (strncmp(name, module->ksm_name, KSTAT_STRLEN) == 0)
return (module);
}
return (NULL);
}
static kstat_module_t *
kstat_create_module(char *name)
{
kstat_module_t *module;
struct proc_dir_entry *pde;
pde = proc_mkdir(name, proc_spl_kstat);
if (pde == NULL)
return (NULL);
module = kmem_alloc(sizeof (kstat_module_t), KM_SLEEP);
module->ksm_proc = pde;
strlcpy(module->ksm_name, name, KSTAT_STRLEN+1);
INIT_LIST_HEAD(&module->ksm_kstat_list);
list_add_tail(&module->ksm_module_list, &kstat_module_list);
return (module);
}
static void
kstat_delete_module(kstat_module_t *module)
{
ASSERT(list_empty(&module->ksm_kstat_list));
remove_proc_entry(module->ksm_name, proc_spl_kstat);
list_del(&module->ksm_module_list);
kmem_free(module, sizeof (kstat_module_t));
}
static int
proc_kstat_open(struct inode *inode, struct file *filp)
{
struct seq_file *f;
int rc;
rc = seq_open(filp, &kstat_seq_ops);
if (rc)
return (rc);
f = filp->private_data;
f->private = PDE_DATA(inode);
return (0);
}
static ssize_t
proc_kstat_write(struct file *filp, const char __user *buf, size_t len,
loff_t *ppos)
{
struct seq_file *f = filp->private_data;
kstat_t *ksp = f->private;
int rc;
ASSERT(ksp->ks_magic == KS_MAGIC);
mutex_enter(ksp->ks_lock);
rc = ksp->ks_update(ksp, KSTAT_WRITE);
mutex_exit(ksp->ks_lock);
if (rc)
return (-rc);
*ppos += len;
return (len);
}
static const kstat_proc_op_t proc_kstat_operations = {
#ifdef HAVE_PROC_OPS_STRUCT
.proc_open = proc_kstat_open,
.proc_write = proc_kstat_write,
.proc_read = seq_read,
.proc_lseek = seq_lseek,
.proc_release = seq_release,
#else
.open = proc_kstat_open,
.write = proc_kstat_write,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release,
#endif
};
void
__kstat_set_raw_ops(kstat_t *ksp,
int (*headers)(char *buf, size_t size),
int (*data)(char *buf, size_t size, void *data),
void *(*addr)(kstat_t *ksp, loff_t index))
{
ksp->ks_raw_ops.headers = headers;
ksp->ks_raw_ops.data = data;
ksp->ks_raw_ops.addr = addr;
}
EXPORT_SYMBOL(__kstat_set_raw_ops);
void
kstat_proc_entry_init(kstat_proc_entry_t *kpep, const char *module,
const char *name)
{
kpep->kpe_owner = NULL;
kpep->kpe_proc = NULL;
INIT_LIST_HEAD(&kpep->kpe_list);
strncpy(kpep->kpe_module, module, KSTAT_STRLEN);
strncpy(kpep->kpe_name, name, KSTAT_STRLEN);
}
EXPORT_SYMBOL(kstat_proc_entry_init);
kstat_t *
__kstat_create(const char *ks_module, int ks_instance, const char *ks_name,
const char *ks_class, uchar_t ks_type, uint_t ks_ndata,
uchar_t ks_flags)
{
kstat_t *ksp;
ASSERT(ks_module);
ASSERT(ks_instance == 0);
ASSERT(ks_name);
if ((ks_type == KSTAT_TYPE_INTR) || (ks_type == KSTAT_TYPE_IO))
ASSERT(ks_ndata == 1);
ksp = kmem_zalloc(sizeof (*ksp), KM_SLEEP);
if (ksp == NULL)
return (ksp);
mutex_enter(&kstat_module_lock);
ksp->ks_kid = kstat_id;
kstat_id++;
mutex_exit(&kstat_module_lock);
ksp->ks_magic = KS_MAGIC;
mutex_init(&ksp->ks_private_lock, NULL, MUTEX_DEFAULT, NULL);
ksp->ks_lock = &ksp->ks_private_lock;
ksp->ks_crtime = gethrtime();
ksp->ks_snaptime = ksp->ks_crtime;
ksp->ks_instance = ks_instance;
strncpy(ksp->ks_class, ks_class, KSTAT_STRLEN);
ksp->ks_type = ks_type;
ksp->ks_flags = ks_flags;
ksp->ks_update = kstat_default_update;
ksp->ks_private = NULL;
ksp->ks_raw_ops.headers = NULL;
ksp->ks_raw_ops.data = NULL;
ksp->ks_raw_ops.addr = NULL;
ksp->ks_raw_buf = NULL;
ksp->ks_raw_bufsize = 0;
kstat_proc_entry_init(&ksp->ks_proc, ks_module, ks_name);
switch (ksp->ks_type) {
case KSTAT_TYPE_RAW:
ksp->ks_ndata = 1;
ksp->ks_data_size = ks_ndata;
break;
case KSTAT_TYPE_NAMED:
ksp->ks_ndata = ks_ndata;
ksp->ks_data_size = ks_ndata * sizeof (kstat_named_t);
break;
case KSTAT_TYPE_INTR:
ksp->ks_ndata = ks_ndata;
ksp->ks_data_size = ks_ndata * sizeof (kstat_intr_t);
break;
case KSTAT_TYPE_IO:
ksp->ks_ndata = ks_ndata;
ksp->ks_data_size = ks_ndata * sizeof (kstat_io_t);
break;
case KSTAT_TYPE_TIMER:
ksp->ks_ndata = ks_ndata;
ksp->ks_data_size = ks_ndata * sizeof (kstat_timer_t);
break;
default:
PANIC("Undefined kstat type %d\n", ksp->ks_type);
}
if (ksp->ks_flags & KSTAT_FLAG_VIRTUAL) {
ksp->ks_data = NULL;
} else {
ksp->ks_data = kmem_zalloc(ksp->ks_data_size, KM_SLEEP);
if (ksp->ks_data == NULL) {
kmem_free(ksp, sizeof (*ksp));
ksp = NULL;
}
}
return (ksp);
}
EXPORT_SYMBOL(__kstat_create);
static int
kstat_detect_collision(kstat_proc_entry_t *kpep)
{
kstat_module_t *module;
kstat_proc_entry_t *tmp = NULL;
char *parent;
char *cp;
parent = kmem_asprintf("%s", kpep->kpe_module);
if ((cp = strrchr(parent, '/')) == NULL) {
kmem_strfree(parent);
return (0);
}
cp[0] = '\0';
if ((module = kstat_find_module(parent)) != NULL) {
list_for_each_entry(tmp, &module->ksm_kstat_list, kpe_list) {
if (strncmp(tmp->kpe_name, cp+1, KSTAT_STRLEN) == 0) {
kmem_strfree(parent);
return (EEXIST);
}
}
}
kmem_strfree(parent);
return (0);
}
/*
* Add a file to the proc filesystem under the kstat namespace (i.e.
* /proc/spl/kstat/). The file need not necessarily be implemented as a
* kstat.
*/
void
kstat_proc_entry_install(kstat_proc_entry_t *kpep, mode_t mode,
const kstat_proc_op_t *proc_ops, void *data)
{
kstat_module_t *module;
kstat_proc_entry_t *tmp = NULL;
ASSERT(kpep);
mutex_enter(&kstat_module_lock);
module = kstat_find_module(kpep->kpe_module);
if (module == NULL) {
if (kstat_detect_collision(kpep) != 0) {
cmn_err(CE_WARN, "kstat_create('%s', '%s'): namespace" \
" collision", kpep->kpe_module, kpep->kpe_name);
goto out;
}
module = kstat_create_module(kpep->kpe_module);
if (module == NULL)
goto out;
}
/*
* Only one entry by this name per-module, on failure the module
* shouldn't be deleted because we know it has at least one entry.
*/
list_for_each_entry(tmp, &module->ksm_kstat_list, kpe_list) {
if (strncmp(tmp->kpe_name, kpep->kpe_name, KSTAT_STRLEN) == 0)
goto out;
}
list_add_tail(&kpep->kpe_list, &module->ksm_kstat_list);
kpep->kpe_owner = module;
kpep->kpe_proc = proc_create_data(kpep->kpe_name, mode,
module->ksm_proc, proc_ops, data);
if (kpep->kpe_proc == NULL) {
list_del_init(&kpep->kpe_list);
if (list_empty(&module->ksm_kstat_list))
kstat_delete_module(module);
}
out:
mutex_exit(&kstat_module_lock);
}
EXPORT_SYMBOL(kstat_proc_entry_install);
void
__kstat_install(kstat_t *ksp)
{
ASSERT(ksp);
mode_t mode;
/* Specify permission modes for different kstats */
if (strncmp(ksp->ks_proc.kpe_name, "dbufs", KSTAT_STRLEN) == 0) {
mode = 0600;
} else {
mode = 0644;
}
kstat_proc_entry_install(
&ksp->ks_proc, mode, &proc_kstat_operations, ksp);
}
EXPORT_SYMBOL(__kstat_install);
void
kstat_proc_entry_delete(kstat_proc_entry_t *kpep)
{
kstat_module_t *module = kpep->kpe_owner;
if (kpep->kpe_proc)
remove_proc_entry(kpep->kpe_name, module->ksm_proc);
mutex_enter(&kstat_module_lock);
list_del_init(&kpep->kpe_list);
/*
* Remove top level module directory if it wasn't empty before, but now
* is.
*/
if (kpep->kpe_proc && list_empty(&module->ksm_kstat_list))
kstat_delete_module(module);
mutex_exit(&kstat_module_lock);
}
EXPORT_SYMBOL(kstat_proc_entry_delete);
void
__kstat_delete(kstat_t *ksp)
{
kstat_proc_entry_delete(&ksp->ks_proc);
if (!(ksp->ks_flags & KSTAT_FLAG_VIRTUAL))
kmem_free(ksp->ks_data, ksp->ks_data_size);
ksp->ks_lock = NULL;
mutex_destroy(&ksp->ks_private_lock);
kmem_free(ksp, sizeof (*ksp));
}
EXPORT_SYMBOL(__kstat_delete);
int
spl_kstat_init(void)
{
mutex_init(&kstat_module_lock, NULL, MUTEX_DEFAULT, NULL);
INIT_LIST_HEAD(&kstat_module_list);
kstat_id = 0;
return (0);
}
void
spl_kstat_fini(void)
{
ASSERT(list_empty(&kstat_module_list));
mutex_destroy(&kstat_module_lock);
}
diff --git a/module/os/linux/spl/spl-proc.c b/module/os/linux/spl/spl-proc.c
index c4af27a7fcd7..f500492eafdc 100644
--- a/module/os/linux/spl/spl-proc.c
+++ b/module/os/linux/spl/spl-proc.c
@@ -1,731 +1,732 @@
/*
* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
* Copyright (C) 2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
* UCRL-CODE-235197
*
* This file is part of the SPL, Solaris Porting Layer.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*
* Solaris Porting Layer (SPL) Proc Implementation.
*/
#include <sys/systeminfo.h>
#include <sys/kstat.h>
#include <sys/kmem.h>
#include <sys/kmem_cache.h>
#include <sys/vmem.h>
#include <sys/taskq.h>
#include <sys/proc.h>
#include <linux/ctype.h>
#include <linux/kmod.h>
#include <linux/seq_file.h>
#include <linux/uaccess.h>
#include <linux/version.h>
+#include "zfs_gitrev.h"
#if defined(CONSTIFY_PLUGIN) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 8, 0)
typedef struct ctl_table __no_const spl_ctl_table;
#else
typedef struct ctl_table spl_ctl_table;
#endif
static unsigned long table_min = 0;
static unsigned long table_max = ~0;
static struct ctl_table_header *spl_header = NULL;
static struct proc_dir_entry *proc_spl = NULL;
static struct proc_dir_entry *proc_spl_kmem = NULL;
static struct proc_dir_entry *proc_spl_kmem_slab = NULL;
static struct proc_dir_entry *proc_spl_taskq_all = NULL;
static struct proc_dir_entry *proc_spl_taskq = NULL;
struct proc_dir_entry *proc_spl_kstat = NULL;
#ifdef DEBUG_KMEM
static int
proc_domemused(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp, loff_t *ppos)
{
int rc = 0;
unsigned long val;
spl_ctl_table dummy = *table;
dummy.data = &val;
dummy.proc_handler = &proc_dointvec;
dummy.extra1 = &table_min;
dummy.extra2 = &table_max;
if (write) {
*ppos += *lenp;
} else {
#ifdef HAVE_ATOMIC64_T
val = atomic64_read((atomic64_t *)table->data);
#else
val = atomic_read((atomic_t *)table->data);
#endif /* HAVE_ATOMIC64_T */
rc = proc_doulongvec_minmax(&dummy, write, buffer, lenp, ppos);
}
return (rc);
}
#endif /* DEBUG_KMEM */
static int
proc_doslab(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp, loff_t *ppos)
{
int rc = 0;
unsigned long val = 0, mask;
spl_ctl_table dummy = *table;
spl_kmem_cache_t *skc = NULL;
dummy.data = &val;
dummy.proc_handler = &proc_dointvec;
dummy.extra1 = &table_min;
dummy.extra2 = &table_max;
if (write) {
*ppos += *lenp;
} else {
down_read(&spl_kmem_cache_sem);
mask = (unsigned long)table->data;
list_for_each_entry(skc, &spl_kmem_cache_list, skc_list) {
/* Only use slabs of the correct kmem/vmem type */
if (!(skc->skc_flags & mask))
continue;
/* Sum the specified field for selected slabs */
switch (mask & (KMC_TOTAL | KMC_ALLOC | KMC_MAX)) {
case KMC_TOTAL:
val += skc->skc_slab_size * skc->skc_slab_total;
break;
case KMC_ALLOC:
val += skc->skc_obj_size * skc->skc_obj_alloc;
break;
case KMC_MAX:
val += skc->skc_obj_size * skc->skc_obj_max;
break;
}
}
up_read(&spl_kmem_cache_sem);
rc = proc_doulongvec_minmax(&dummy, write, buffer, lenp, ppos);
}
return (rc);
}
static int
proc_dohostid(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp, loff_t *ppos)
{
char *end, str[32];
unsigned long hid;
spl_ctl_table dummy = *table;
dummy.data = str;
dummy.maxlen = sizeof (str) - 1;
if (!write)
snprintf(str, sizeof (str), "%lx",
(unsigned long) zone_get_hostid(NULL));
/* always returns 0 */
proc_dostring(&dummy, write, buffer, lenp, ppos);
if (write) {
/*
* We can't use proc_doulongvec_minmax() in the write
* case here because hostid, while a hex value, has no
* leading 0x, which confuses the helper function.
*/
hid = simple_strtoul(str, &end, 16);
if (str == end)
return (-EINVAL);
spl_hostid = hid;
}
return (0);
}
static void
taskq_seq_show_headers(struct seq_file *f)
{
seq_printf(f, "%-25s %5s %5s %5s %5s %5s %5s %12s %5s %10s\n",
"taskq", "act", "nthr", "spwn", "maxt", "pri",
"mina", "maxa", "cura", "flags");
}
/* indices into the lheads array below */
#define LHEAD_PEND 0
#define LHEAD_PRIO 1
#define LHEAD_DELAY 2
#define LHEAD_WAIT 3
#define LHEAD_ACTIVE 4
#define LHEAD_SIZE 5
/* BEGIN CSTYLED */
static unsigned int spl_max_show_tasks = 512;
module_param(spl_max_show_tasks, uint, 0644);
MODULE_PARM_DESC(spl_max_show_tasks, "Max number of tasks shown in taskq proc");
/* END CSTYLED */
static int
taskq_seq_show_impl(struct seq_file *f, void *p, boolean_t allflag)
{
taskq_t *tq = p;
taskq_thread_t *tqt = NULL;
spl_wait_queue_entry_t *wq;
struct task_struct *tsk;
taskq_ent_t *tqe;
char name[100];
struct list_head *lheads[LHEAD_SIZE], *lh;
static char *list_names[LHEAD_SIZE] =
{"pend", "prio", "delay", "wait", "active" };
int i, j, have_lheads = 0;
unsigned long wflags, flags;
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
spin_lock_irqsave(&tq->tq_wait_waitq.lock, wflags);
/* get the various lists and check whether they're empty */
lheads[LHEAD_PEND] = &tq->tq_pend_list;
lheads[LHEAD_PRIO] = &tq->tq_prio_list;
lheads[LHEAD_DELAY] = &tq->tq_delay_list;
#ifdef HAVE_WAIT_QUEUE_HEAD_ENTRY
lheads[LHEAD_WAIT] = &tq->tq_wait_waitq.head;
#else
lheads[LHEAD_WAIT] = &tq->tq_wait_waitq.task_list;
#endif
lheads[LHEAD_ACTIVE] = &tq->tq_active_list;
for (i = 0; i < LHEAD_SIZE; ++i) {
if (list_empty(lheads[i]))
lheads[i] = NULL;
else
++have_lheads;
}
/* early return in non-"all" mode if lists are all empty */
if (!allflag && !have_lheads) {
spin_unlock_irqrestore(&tq->tq_wait_waitq.lock, wflags);
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (0);
}
/* unlock the waitq quickly */
if (!lheads[LHEAD_WAIT])
spin_unlock_irqrestore(&tq->tq_wait_waitq.lock, wflags);
/* show the base taskq contents */
snprintf(name, sizeof (name), "%s/%d", tq->tq_name, tq->tq_instance);
seq_printf(f, "%-25s ", name);
seq_printf(f, "%5d %5d %5d %5d %5d %5d %12d %5d %10x\n",
tq->tq_nactive, tq->tq_nthreads, tq->tq_nspawn,
tq->tq_maxthreads, tq->tq_pri, tq->tq_minalloc, tq->tq_maxalloc,
tq->tq_nalloc, tq->tq_flags);
/* show the active list */
if (lheads[LHEAD_ACTIVE]) {
j = 0;
list_for_each_entry(tqt, &tq->tq_active_list, tqt_active_list) {
if (j == 0)
seq_printf(f, "\t%s:",
list_names[LHEAD_ACTIVE]);
else if (j == 2) {
seq_printf(f, "\n\t ");
j = 0;
}
seq_printf(f, " [%d]%pf(%ps)",
tqt->tqt_thread->pid,
tqt->tqt_task->tqent_func,
tqt->tqt_task->tqent_arg);
++j;
}
seq_printf(f, "\n");
}
for (i = LHEAD_PEND; i <= LHEAD_WAIT; ++i)
if (lheads[i]) {
j = 0;
list_for_each(lh, lheads[i]) {
if (spl_max_show_tasks != 0 &&
j >= spl_max_show_tasks) {
seq_printf(f, "\n\t(truncated)");
break;
}
/* show the wait waitq list */
if (i == LHEAD_WAIT) {
#ifdef HAVE_WAIT_QUEUE_HEAD_ENTRY
wq = list_entry(lh,
spl_wait_queue_entry_t, entry);
#else
wq = list_entry(lh,
spl_wait_queue_entry_t, task_list);
#endif
if (j == 0)
seq_printf(f, "\t%s:",
list_names[i]);
else if (j % 8 == 0)
seq_printf(f, "\n\t ");
tsk = wq->private;
seq_printf(f, " %d", tsk->pid);
/* pend, prio and delay lists */
} else {
tqe = list_entry(lh, taskq_ent_t,
tqent_list);
if (j == 0)
seq_printf(f, "\t%s:",
list_names[i]);
else if (j % 2 == 0)
seq_printf(f, "\n\t ");
seq_printf(f, " %pf(%ps)",
tqe->tqent_func,
tqe->tqent_arg);
}
++j;
}
seq_printf(f, "\n");
}
if (lheads[LHEAD_WAIT])
spin_unlock_irqrestore(&tq->tq_wait_waitq.lock, wflags);
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (0);
}
static int
taskq_all_seq_show(struct seq_file *f, void *p)
{
return (taskq_seq_show_impl(f, p, B_TRUE));
}
static int
taskq_seq_show(struct seq_file *f, void *p)
{
return (taskq_seq_show_impl(f, p, B_FALSE));
}
static void *
taskq_seq_start(struct seq_file *f, loff_t *pos)
{
struct list_head *p;
loff_t n = *pos;
down_read(&tq_list_sem);
if (!n)
taskq_seq_show_headers(f);
p = tq_list.next;
while (n--) {
p = p->next;
if (p == &tq_list)
return (NULL);
}
return (list_entry(p, taskq_t, tq_taskqs));
}
static void *
taskq_seq_next(struct seq_file *f, void *p, loff_t *pos)
{
taskq_t *tq = p;
++*pos;
return ((tq->tq_taskqs.next == &tq_list) ?
NULL : list_entry(tq->tq_taskqs.next, taskq_t, tq_taskqs));
}
static void
slab_seq_show_headers(struct seq_file *f)
{
seq_printf(f,
"--------------------- cache ----------"
"--------------------------------------------- "
"----- slab ------ "
"---- object ----- "
"--- emergency ---\n");
seq_printf(f,
"name "
" flags size alloc slabsize objsize "
"total alloc max "
"total alloc max "
"dlock alloc max\n");
}
static int
slab_seq_show(struct seq_file *f, void *p)
{
spl_kmem_cache_t *skc = p;
ASSERT(skc->skc_magic == SKC_MAGIC);
if (skc->skc_flags & KMC_SLAB) {
/*
* This cache is backed by a generic Linux kmem cache which
* has its own accounting. For these caches we only track
* the number of active allocated objects that exist within
* the underlying Linux slabs. For the overall statistics of
* the underlying Linux cache please refer to /proc/slabinfo.
*/
spin_lock(&skc->skc_lock);
uint64_t objs_allocated =
percpu_counter_sum(&skc->skc_linux_alloc);
seq_printf(f, "%-36s ", skc->skc_name);
seq_printf(f, "0x%05lx %9s %9lu %8s %8u "
"%5s %5s %5s %5s %5lu %5s %5s %5s %5s\n",
(long unsigned)skc->skc_flags,
"-",
(long unsigned)(skc->skc_obj_size * objs_allocated),
"-",
(unsigned)skc->skc_obj_size,
"-", "-", "-", "-",
(long unsigned)objs_allocated,
"-", "-", "-", "-");
spin_unlock(&skc->skc_lock);
return (0);
}
spin_lock(&skc->skc_lock);
seq_printf(f, "%-36s ", skc->skc_name);
seq_printf(f, "0x%05lx %9lu %9lu %8u %8u "
"%5lu %5lu %5lu %5lu %5lu %5lu %5lu %5lu %5lu\n",
(long unsigned)skc->skc_flags,
(long unsigned)(skc->skc_slab_size * skc->skc_slab_total),
(long unsigned)(skc->skc_obj_size * skc->skc_obj_alloc),
(unsigned)skc->skc_slab_size,
(unsigned)skc->skc_obj_size,
(long unsigned)skc->skc_slab_total,
(long unsigned)skc->skc_slab_alloc,
(long unsigned)skc->skc_slab_max,
(long unsigned)skc->skc_obj_total,
(long unsigned)skc->skc_obj_alloc,
(long unsigned)skc->skc_obj_max,
(long unsigned)skc->skc_obj_deadlock,
(long unsigned)skc->skc_obj_emergency,
(long unsigned)skc->skc_obj_emergency_max);
spin_unlock(&skc->skc_lock);
return (0);
}
static void *
slab_seq_start(struct seq_file *f, loff_t *pos)
{
struct list_head *p;
loff_t n = *pos;
down_read(&spl_kmem_cache_sem);
if (!n)
slab_seq_show_headers(f);
p = spl_kmem_cache_list.next;
while (n--) {
p = p->next;
if (p == &spl_kmem_cache_list)
return (NULL);
}
return (list_entry(p, spl_kmem_cache_t, skc_list));
}
static void *
slab_seq_next(struct seq_file *f, void *p, loff_t *pos)
{
spl_kmem_cache_t *skc = p;
++*pos;
return ((skc->skc_list.next == &spl_kmem_cache_list) ?
NULL : list_entry(skc->skc_list.next, spl_kmem_cache_t, skc_list));
}
static void
slab_seq_stop(struct seq_file *f, void *v)
{
up_read(&spl_kmem_cache_sem);
}
-static struct seq_operations slab_seq_ops = {
+static const struct seq_operations slab_seq_ops = {
.show = slab_seq_show,
.start = slab_seq_start,
.next = slab_seq_next,
.stop = slab_seq_stop,
};
static int
proc_slab_open(struct inode *inode, struct file *filp)
{
return (seq_open(filp, &slab_seq_ops));
}
static const kstat_proc_op_t proc_slab_operations = {
#ifdef HAVE_PROC_OPS_STRUCT
.proc_open = proc_slab_open,
.proc_read = seq_read,
.proc_lseek = seq_lseek,
.proc_release = seq_release,
#else
.open = proc_slab_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release,
#endif
};
static void
taskq_seq_stop(struct seq_file *f, void *v)
{
up_read(&tq_list_sem);
}
-static struct seq_operations taskq_all_seq_ops = {
+static const struct seq_operations taskq_all_seq_ops = {
.show = taskq_all_seq_show,
.start = taskq_seq_start,
.next = taskq_seq_next,
.stop = taskq_seq_stop,
};
-static struct seq_operations taskq_seq_ops = {
+static const struct seq_operations taskq_seq_ops = {
.show = taskq_seq_show,
.start = taskq_seq_start,
.next = taskq_seq_next,
.stop = taskq_seq_stop,
};
static int
proc_taskq_all_open(struct inode *inode, struct file *filp)
{
return (seq_open(filp, &taskq_all_seq_ops));
}
static int
proc_taskq_open(struct inode *inode, struct file *filp)
{
return (seq_open(filp, &taskq_seq_ops));
}
static const kstat_proc_op_t proc_taskq_all_operations = {
#ifdef HAVE_PROC_OPS_STRUCT
.proc_open = proc_taskq_all_open,
.proc_read = seq_read,
.proc_lseek = seq_lseek,
.proc_release = seq_release,
#else
.open = proc_taskq_all_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release,
#endif
};
static const kstat_proc_op_t proc_taskq_operations = {
#ifdef HAVE_PROC_OPS_STRUCT
.proc_open = proc_taskq_open,
.proc_read = seq_read,
.proc_lseek = seq_lseek,
.proc_release = seq_release,
#else
.open = proc_taskq_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release,
#endif
};
static struct ctl_table spl_kmem_table[] = {
#ifdef DEBUG_KMEM
{
.procname = "kmem_used",
.data = &kmem_alloc_used,
#ifdef HAVE_ATOMIC64_T
.maxlen = sizeof (atomic64_t),
#else
.maxlen = sizeof (atomic_t),
#endif /* HAVE_ATOMIC64_T */
.mode = 0444,
.proc_handler = &proc_domemused,
},
{
.procname = "kmem_max",
.data = &kmem_alloc_max,
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doulongvec_minmax,
},
#endif /* DEBUG_KMEM */
{
.procname = "slab_kvmem_total",
.data = (void *)(KMC_KVMEM | KMC_TOTAL),
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doslab,
},
{
.procname = "slab_kvmem_alloc",
.data = (void *)(KMC_KVMEM | KMC_ALLOC),
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doslab,
},
{
.procname = "slab_kvmem_max",
.data = (void *)(KMC_KVMEM | KMC_MAX),
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doslab,
},
{},
};
static struct ctl_table spl_kstat_table[] = {
{},
};
static struct ctl_table spl_table[] = {
/*
* NB No .strategy entries have been provided since
* sysctl(8) prefers to go via /proc for portability.
*/
{
.procname = "gitrev",
- .data = spl_gitrev,
- .maxlen = sizeof (spl_gitrev),
+ .data = (char *)ZFS_META_GITREV,
+ .maxlen = sizeof (ZFS_META_GITREV),
.mode = 0444,
.proc_handler = &proc_dostring,
},
{
.procname = "hostid",
.data = &spl_hostid,
.maxlen = sizeof (unsigned long),
.mode = 0644,
.proc_handler = &proc_dohostid,
},
{
.procname = "kmem",
.mode = 0555,
.child = spl_kmem_table,
},
{
.procname = "kstat",
.mode = 0555,
.child = spl_kstat_table,
},
{},
};
static struct ctl_table spl_dir[] = {
{
.procname = "spl",
.mode = 0555,
.child = spl_table,
},
{}
};
static struct ctl_table spl_root[] = {
{
.procname = "kernel",
.mode = 0555,
.child = spl_dir,
},
{}
};
int
spl_proc_init(void)
{
int rc = 0;
spl_header = register_sysctl_table(spl_root);
if (spl_header == NULL)
return (-EUNATCH);
proc_spl = proc_mkdir("spl", NULL);
if (proc_spl == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_taskq_all = proc_create_data("taskq-all", 0444, proc_spl,
&proc_taskq_all_operations, NULL);
if (proc_spl_taskq_all == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_taskq = proc_create_data("taskq", 0444, proc_spl,
&proc_taskq_operations, NULL);
if (proc_spl_taskq == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_kmem = proc_mkdir("kmem", proc_spl);
if (proc_spl_kmem == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_kmem_slab = proc_create_data("slab", 0444, proc_spl_kmem,
&proc_slab_operations, NULL);
if (proc_spl_kmem_slab == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_kstat = proc_mkdir("kstat", proc_spl);
if (proc_spl_kstat == NULL) {
rc = -EUNATCH;
goto out;
}
out:
if (rc) {
remove_proc_entry("kstat", proc_spl);
remove_proc_entry("slab", proc_spl_kmem);
remove_proc_entry("kmem", proc_spl);
remove_proc_entry("taskq-all", proc_spl);
remove_proc_entry("taskq", proc_spl);
remove_proc_entry("spl", NULL);
unregister_sysctl_table(spl_header);
}
return (rc);
}
void
spl_proc_fini(void)
{
remove_proc_entry("kstat", proc_spl);
remove_proc_entry("slab", proc_spl_kmem);
remove_proc_entry("kmem", proc_spl);
remove_proc_entry("taskq-all", proc_spl);
remove_proc_entry("taskq", proc_spl);
remove_proc_entry("spl", NULL);
ASSERT(spl_header != NULL);
unregister_sysctl_table(spl_header);
}
diff --git a/module/os/linux/spl/spl-procfs-list.c b/module/os/linux/spl/spl-procfs-list.c
index cae13228c62c..1922825c94a8 100644
--- a/module/os/linux/spl/spl-procfs-list.c
+++ b/module/os/linux/spl/spl-procfs-list.c
@@ -1,284 +1,284 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2018 by Delphix. All rights reserved.
*/
#include <sys/list.h>
#include <sys/mutex.h>
#include <sys/procfs_list.h>
#include <linux/proc_fs.h>
/*
* A procfs_list is a wrapper around a linked list which implements the seq_file
* interface, allowing the contents of the list to be exposed through procfs.
* The kernel already has some utilities to help implement the seq_file
* interface for linked lists (seq_list_*), but they aren't appropriate for use
* with lists that have many entries, because seq_list_start walks the list at
* the start of each read syscall to find where it left off, so reading a file
* ends up being quadratic in the number of entries in the list.
*
* This implementation avoids this penalty by maintaining a separate cursor into
* the list per instance of the file that is open. It also maintains some extra
* information in each node of the list to prevent reads of entries that have
* been dropped from the list.
*
* Callers should only add elements to the list using procfs_list_add, which
* adds an element to the tail of the list. Other operations can be performed
* directly on the wrapped list using the normal list manipulation functions,
* but elements should only be removed from the head of the list.
*/
#define NODE_ID(procfs_list, obj) \
(((procfs_list_node_t *)(((char *)obj) + \
(procfs_list)->pl_node_offset))->pln_id)
typedef struct procfs_list_cursor {
procfs_list_t *procfs_list; /* List into which this cursor points */
void *cached_node; /* Most recently accessed node */
loff_t cached_pos; /* Position of cached_node */
} procfs_list_cursor_t;
static int
procfs_list_seq_show(struct seq_file *f, void *p)
{
procfs_list_cursor_t *cursor = f->private;
procfs_list_t *procfs_list = cursor->procfs_list;
ASSERT(MUTEX_HELD(&procfs_list->pl_lock));
if (p == SEQ_START_TOKEN) {
if (procfs_list->pl_show_header != NULL)
return (procfs_list->pl_show_header(f));
else
return (0);
}
return (procfs_list->pl_show(f, p));
}
static void *
procfs_list_next_node(procfs_list_cursor_t *cursor, loff_t *pos)
{
void *next_node;
procfs_list_t *procfs_list = cursor->procfs_list;
if (cursor->cached_node == SEQ_START_TOKEN)
next_node = list_head(&procfs_list->pl_list);
else
next_node = list_next(&procfs_list->pl_list,
cursor->cached_node);
if (next_node != NULL) {
cursor->cached_node = next_node;
cursor->cached_pos = NODE_ID(procfs_list, cursor->cached_node);
*pos = cursor->cached_pos;
} else {
/*
* seq_read() expects ->next() to update the position even
* when there are no more entries. Advance the position to
* prevent a warning from being logged.
*/
cursor->cached_node = NULL;
cursor->cached_pos++;
*pos = cursor->cached_pos;
}
return (next_node);
}
static void *
procfs_list_seq_start(struct seq_file *f, loff_t *pos)
{
procfs_list_cursor_t *cursor = f->private;
procfs_list_t *procfs_list = cursor->procfs_list;
mutex_enter(&procfs_list->pl_lock);
if (*pos == 0) {
cursor->cached_node = SEQ_START_TOKEN;
cursor->cached_pos = 0;
return (SEQ_START_TOKEN);
} else if (cursor->cached_node == NULL) {
return (NULL);
}
/*
* Check if our cached pointer has become stale, which happens if the
* the message where we left off has been dropped from the list since
* the last read syscall completed.
*/
void *oldest_node = list_head(&procfs_list->pl_list);
if (cursor->cached_node != SEQ_START_TOKEN && (oldest_node == NULL ||
NODE_ID(procfs_list, oldest_node) > cursor->cached_pos))
return (ERR_PTR(-EIO));
/*
* If it isn't starting from the beginning of the file, the seq_file
* code will either pick up at the same position it visited last or the
* following one.
*/
if (*pos == cursor->cached_pos) {
return (cursor->cached_node);
} else {
ASSERT3U(*pos, ==, cursor->cached_pos + 1);
return (procfs_list_next_node(cursor, pos));
}
}
static void *
procfs_list_seq_next(struct seq_file *f, void *p, loff_t *pos)
{
procfs_list_cursor_t *cursor = f->private;
ASSERT(MUTEX_HELD(&cursor->procfs_list->pl_lock));
return (procfs_list_next_node(cursor, pos));
}
static void
procfs_list_seq_stop(struct seq_file *f, void *p)
{
procfs_list_cursor_t *cursor = f->private;
procfs_list_t *procfs_list = cursor->procfs_list;
mutex_exit(&procfs_list->pl_lock);
}
-static struct seq_operations procfs_list_seq_ops = {
+static const struct seq_operations procfs_list_seq_ops = {
.show = procfs_list_seq_show,
.start = procfs_list_seq_start,
.next = procfs_list_seq_next,
.stop = procfs_list_seq_stop,
};
static int
procfs_list_open(struct inode *inode, struct file *filp)
{
int rc = seq_open_private(filp, &procfs_list_seq_ops,
sizeof (procfs_list_cursor_t));
if (rc != 0)
return (rc);
struct seq_file *f = filp->private_data;
procfs_list_cursor_t *cursor = f->private;
cursor->procfs_list = PDE_DATA(inode);
cursor->cached_node = NULL;
cursor->cached_pos = 0;
return (0);
}
static ssize_t
procfs_list_write(struct file *filp, const char __user *buf, size_t len,
loff_t *ppos)
{
struct seq_file *f = filp->private_data;
procfs_list_cursor_t *cursor = f->private;
procfs_list_t *procfs_list = cursor->procfs_list;
int rc;
if (procfs_list->pl_clear != NULL &&
(rc = procfs_list->pl_clear(procfs_list)) != 0)
return (-rc);
return (len);
}
static const kstat_proc_op_t procfs_list_operations = {
#ifdef HAVE_PROC_OPS_STRUCT
.proc_open = procfs_list_open,
.proc_write = procfs_list_write,
.proc_read = seq_read,
.proc_lseek = seq_lseek,
.proc_release = seq_release_private,
#else
.open = procfs_list_open,
.write = procfs_list_write,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release_private,
#endif
};
/*
* Initialize a procfs_list and create a file for it in the proc filesystem
* under the kstat namespace.
*/
void
procfs_list_install(const char *module,
const char *submodule,
const char *name,
mode_t mode,
procfs_list_t *procfs_list,
int (*show)(struct seq_file *f, void *p),
int (*show_header)(struct seq_file *f),
int (*clear)(procfs_list_t *procfs_list),
size_t procfs_list_node_off)
{
char *modulestr;
if (submodule != NULL)
modulestr = kmem_asprintf("%s/%s", module, submodule);
else
modulestr = kmem_asprintf("%s", module);
mutex_init(&procfs_list->pl_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&procfs_list->pl_list,
procfs_list_node_off + sizeof (procfs_list_node_t),
procfs_list_node_off + offsetof(procfs_list_node_t, pln_link));
procfs_list->pl_next_id = 1; /* Save id 0 for SEQ_START_TOKEN */
procfs_list->pl_show = show;
procfs_list->pl_show_header = show_header;
procfs_list->pl_clear = clear;
procfs_list->pl_node_offset = procfs_list_node_off;
kstat_proc_entry_init(&procfs_list->pl_kstat_entry, modulestr, name);
kstat_proc_entry_install(&procfs_list->pl_kstat_entry, mode,
&procfs_list_operations, procfs_list);
kmem_strfree(modulestr);
}
EXPORT_SYMBOL(procfs_list_install);
/* Remove the proc filesystem file corresponding to the given list */
void
procfs_list_uninstall(procfs_list_t *procfs_list)
{
kstat_proc_entry_delete(&procfs_list->pl_kstat_entry);
}
EXPORT_SYMBOL(procfs_list_uninstall);
void
procfs_list_destroy(procfs_list_t *procfs_list)
{
ASSERT(list_is_empty(&procfs_list->pl_list));
list_destroy(&procfs_list->pl_list);
mutex_destroy(&procfs_list->pl_lock);
}
EXPORT_SYMBOL(procfs_list_destroy);
/*
* Add a new node to the tail of the list. While the standard list manipulation
* functions can be use for all other operation, adding elements to the list
* should only be done using this helper so that the id of the new node is set
* correctly.
*/
void
procfs_list_add(procfs_list_t *procfs_list, void *p)
{
ASSERT(MUTEX_HELD(&procfs_list->pl_lock));
NODE_ID(procfs_list, p) = procfs_list->pl_next_id++;
list_insert_tail(&procfs_list->pl_list, p);
}
EXPORT_SYMBOL(procfs_list_add);
diff --git a/module/os/linux/spl/spl-taskq.c b/module/os/linux/spl/spl-taskq.c
index a879c2856302..0aab148975aa 100644
--- a/module/os/linux/spl/spl-taskq.c
+++ b/module/os/linux/spl/spl-taskq.c
@@ -1,1433 +1,1433 @@
/*
* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
* Copyright (C) 2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
* UCRL-CODE-235197
*
* This file is part of the SPL, Solaris Porting Layer.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*
* Solaris Porting Layer (SPL) Task Queue Implementation.
*/
#include <sys/timer.h>
#include <sys/taskq.h>
#include <sys/kmem.h>
#include <sys/tsd.h>
#include <sys/trace_spl.h>
#ifdef HAVE_CPU_HOTPLUG
#include <linux/cpuhotplug.h>
#endif
-int spl_taskq_thread_bind = 0;
+static int spl_taskq_thread_bind = 0;
module_param(spl_taskq_thread_bind, int, 0644);
MODULE_PARM_DESC(spl_taskq_thread_bind, "Bind taskq thread to CPU by default");
-int spl_taskq_thread_dynamic = 1;
+static int spl_taskq_thread_dynamic = 1;
module_param(spl_taskq_thread_dynamic, int, 0444);
MODULE_PARM_DESC(spl_taskq_thread_dynamic, "Allow dynamic taskq threads");
-int spl_taskq_thread_priority = 1;
+static int spl_taskq_thread_priority = 1;
module_param(spl_taskq_thread_priority, int, 0644);
MODULE_PARM_DESC(spl_taskq_thread_priority,
"Allow non-default priority for taskq threads");
-int spl_taskq_thread_sequential = 4;
+static int spl_taskq_thread_sequential = 4;
module_param(spl_taskq_thread_sequential, int, 0644);
MODULE_PARM_DESC(spl_taskq_thread_sequential,
"Create new taskq threads after N sequential tasks");
/*
* Global system-wide dynamic task queue available for all consumers. This
* taskq is not intended for long-running tasks; instead, a dedicated taskq
* should be created.
*/
taskq_t *system_taskq;
EXPORT_SYMBOL(system_taskq);
/* Global dynamic task queue for long delay */
taskq_t *system_delay_taskq;
EXPORT_SYMBOL(system_delay_taskq);
/* Private dedicated taskq for creating new taskq threads on demand. */
static taskq_t *dynamic_taskq;
static taskq_thread_t *taskq_thread_create(taskq_t *);
#ifdef HAVE_CPU_HOTPLUG
/* Multi-callback id for cpu hotplugging. */
static int spl_taskq_cpuhp_state;
#endif
/* List of all taskqs */
LIST_HEAD(tq_list);
struct rw_semaphore tq_list_sem;
static uint_t taskq_tsd;
static int
task_km_flags(uint_t flags)
{
if (flags & TQ_NOSLEEP)
return (KM_NOSLEEP);
if (flags & TQ_PUSHPAGE)
return (KM_PUSHPAGE);
return (KM_SLEEP);
}
/*
* taskq_find_by_name - Find the largest instance number of a named taskq.
*/
static int
taskq_find_by_name(const char *name)
{
struct list_head *tql = NULL;
taskq_t *tq;
list_for_each_prev(tql, &tq_list) {
tq = list_entry(tql, taskq_t, tq_taskqs);
if (strcmp(name, tq->tq_name) == 0)
return (tq->tq_instance);
}
return (-1);
}
/*
* NOTE: Must be called with tq->tq_lock held, returns a list_t which
* is not attached to the free, work, or pending taskq lists.
*/
static taskq_ent_t *
task_alloc(taskq_t *tq, uint_t flags, unsigned long *irqflags)
{
taskq_ent_t *t;
int count = 0;
ASSERT(tq);
retry:
/* Acquire taskq_ent_t's from free list if available */
if (!list_empty(&tq->tq_free_list) && !(flags & TQ_NEW)) {
t = list_entry(tq->tq_free_list.next, taskq_ent_t, tqent_list);
ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));
ASSERT(!(t->tqent_flags & TQENT_FLAG_CANCEL));
ASSERT(!timer_pending(&t->tqent_timer));
list_del_init(&t->tqent_list);
return (t);
}
/* Free list is empty and memory allocations are prohibited */
if (flags & TQ_NOALLOC)
return (NULL);
/* Hit maximum taskq_ent_t pool size */
if (tq->tq_nalloc >= tq->tq_maxalloc) {
if (flags & TQ_NOSLEEP)
return (NULL);
/*
* Sleep periodically polling the free list for an available
* taskq_ent_t. Dispatching with TQ_SLEEP should always succeed
* but we cannot block forever waiting for an taskq_ent_t to
* show up in the free list, otherwise a deadlock can happen.
*
* Therefore, we need to allocate a new task even if the number
* of allocated tasks is above tq->tq_maxalloc, but we still
* end up delaying the task allocation by one second, thereby
* throttling the task dispatch rate.
*/
spin_unlock_irqrestore(&tq->tq_lock, *irqflags);
schedule_timeout(HZ / 100);
spin_lock_irqsave_nested(&tq->tq_lock, *irqflags,
tq->tq_lock_class);
if (count < 100) {
count++;
goto retry;
}
}
spin_unlock_irqrestore(&tq->tq_lock, *irqflags);
t = kmem_alloc(sizeof (taskq_ent_t), task_km_flags(flags));
spin_lock_irqsave_nested(&tq->tq_lock, *irqflags, tq->tq_lock_class);
if (t) {
taskq_init_ent(t);
tq->tq_nalloc++;
}
return (t);
}
/*
* NOTE: Must be called with tq->tq_lock held, expects the taskq_ent_t
* to already be removed from the free, work, or pending taskq lists.
*/
static void
task_free(taskq_t *tq, taskq_ent_t *t)
{
ASSERT(tq);
ASSERT(t);
ASSERT(list_empty(&t->tqent_list));
ASSERT(!timer_pending(&t->tqent_timer));
kmem_free(t, sizeof (taskq_ent_t));
tq->tq_nalloc--;
}
/*
* NOTE: Must be called with tq->tq_lock held, either destroys the
* taskq_ent_t if too many exist or moves it to the free list for later use.
*/
static void
task_done(taskq_t *tq, taskq_ent_t *t)
{
ASSERT(tq);
ASSERT(t);
/* Wake tasks blocked in taskq_wait_id() */
wake_up_all(&t->tqent_waitq);
list_del_init(&t->tqent_list);
if (tq->tq_nalloc <= tq->tq_minalloc) {
t->tqent_id = TASKQID_INVALID;
t->tqent_func = NULL;
t->tqent_arg = NULL;
t->tqent_flags = 0;
list_add_tail(&t->tqent_list, &tq->tq_free_list);
} else {
task_free(tq, t);
}
}
/*
* When a delayed task timer expires remove it from the delay list and
* add it to the priority list in order for immediate processing.
*/
static void
task_expire_impl(taskq_ent_t *t)
{
taskq_ent_t *w;
taskq_t *tq = t->tqent_taskq;
struct list_head *l = NULL;
unsigned long flags;
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
if (t->tqent_flags & TQENT_FLAG_CANCEL) {
ASSERT(list_empty(&t->tqent_list));
spin_unlock_irqrestore(&tq->tq_lock, flags);
return;
}
t->tqent_birth = jiffies;
DTRACE_PROBE1(taskq_ent__birth, taskq_ent_t *, t);
/*
* The priority list must be maintained in strict task id order
* from lowest to highest for lowest_id to be easily calculable.
*/
list_del(&t->tqent_list);
list_for_each_prev(l, &tq->tq_prio_list) {
w = list_entry(l, taskq_ent_t, tqent_list);
if (w->tqent_id < t->tqent_id) {
list_add(&t->tqent_list, l);
break;
}
}
if (l == &tq->tq_prio_list)
list_add(&t->tqent_list, &tq->tq_prio_list);
spin_unlock_irqrestore(&tq->tq_lock, flags);
wake_up(&tq->tq_work_waitq);
}
static void
task_expire(spl_timer_list_t tl)
{
struct timer_list *tmr = (struct timer_list *)tl;
taskq_ent_t *t = from_timer(t, tmr, tqent_timer);
task_expire_impl(t);
}
/*
* Returns the lowest incomplete taskqid_t. The taskqid_t may
* be queued on the pending list, on the priority list, on the
* delay list, or on the work list currently being handled, but
* it is not 100% complete yet.
*/
static taskqid_t
taskq_lowest_id(taskq_t *tq)
{
taskqid_t lowest_id = tq->tq_next_id;
taskq_ent_t *t;
taskq_thread_t *tqt;
if (!list_empty(&tq->tq_pend_list)) {
t = list_entry(tq->tq_pend_list.next, taskq_ent_t, tqent_list);
lowest_id = MIN(lowest_id, t->tqent_id);
}
if (!list_empty(&tq->tq_prio_list)) {
t = list_entry(tq->tq_prio_list.next, taskq_ent_t, tqent_list);
lowest_id = MIN(lowest_id, t->tqent_id);
}
if (!list_empty(&tq->tq_delay_list)) {
t = list_entry(tq->tq_delay_list.next, taskq_ent_t, tqent_list);
lowest_id = MIN(lowest_id, t->tqent_id);
}
if (!list_empty(&tq->tq_active_list)) {
tqt = list_entry(tq->tq_active_list.next, taskq_thread_t,
tqt_active_list);
ASSERT(tqt->tqt_id != TASKQID_INVALID);
lowest_id = MIN(lowest_id, tqt->tqt_id);
}
return (lowest_id);
}
/*
* Insert a task into a list keeping the list sorted by increasing taskqid.
*/
static void
taskq_insert_in_order(taskq_t *tq, taskq_thread_t *tqt)
{
taskq_thread_t *w;
struct list_head *l = NULL;
ASSERT(tq);
ASSERT(tqt);
list_for_each_prev(l, &tq->tq_active_list) {
w = list_entry(l, taskq_thread_t, tqt_active_list);
if (w->tqt_id < tqt->tqt_id) {
list_add(&tqt->tqt_active_list, l);
break;
}
}
if (l == &tq->tq_active_list)
list_add(&tqt->tqt_active_list, &tq->tq_active_list);
}
/*
* Find and return a task from the given list if it exists. The list
* must be in lowest to highest task id order.
*/
static taskq_ent_t *
taskq_find_list(taskq_t *tq, struct list_head *lh, taskqid_t id)
{
struct list_head *l = NULL;
taskq_ent_t *t;
list_for_each(l, lh) {
t = list_entry(l, taskq_ent_t, tqent_list);
if (t->tqent_id == id)
return (t);
if (t->tqent_id > id)
break;
}
return (NULL);
}
/*
* Find an already dispatched task given the task id regardless of what
* state it is in. If a task is still pending it will be returned.
* If a task is executing, then -EBUSY will be returned instead.
* If the task has already been run then NULL is returned.
*/
static taskq_ent_t *
taskq_find(taskq_t *tq, taskqid_t id)
{
taskq_thread_t *tqt;
struct list_head *l = NULL;
taskq_ent_t *t;
t = taskq_find_list(tq, &tq->tq_delay_list, id);
if (t)
return (t);
t = taskq_find_list(tq, &tq->tq_prio_list, id);
if (t)
return (t);
t = taskq_find_list(tq, &tq->tq_pend_list, id);
if (t)
return (t);
list_for_each(l, &tq->tq_active_list) {
tqt = list_entry(l, taskq_thread_t, tqt_active_list);
if (tqt->tqt_id == id) {
/*
* Instead of returning tqt_task, we just return a non
* NULL value to prevent misuse, since tqt_task only
* has two valid fields.
*/
return (ERR_PTR(-EBUSY));
}
}
return (NULL);
}
/*
* Theory for the taskq_wait_id(), taskq_wait_outstanding(), and
* taskq_wait() functions below.
*
* Taskq waiting is accomplished by tracking the lowest outstanding task
* id and the next available task id. As tasks are dispatched they are
* added to the tail of the pending, priority, or delay lists. As worker
* threads become available the tasks are removed from the heads of these
* lists and linked to the worker threads. This ensures the lists are
* kept sorted by lowest to highest task id.
*
* Therefore the lowest outstanding task id can be quickly determined by
* checking the head item from all of these lists. This value is stored
* with the taskq as the lowest id. It only needs to be recalculated when
* either the task with the current lowest id completes or is canceled.
*
* By blocking until the lowest task id exceeds the passed task id the
* taskq_wait_outstanding() function can be easily implemented. Similarly,
* by blocking until the lowest task id matches the next task id taskq_wait()
* can be implemented.
*
* Callers should be aware that when there are multiple worked threads it
* is possible for larger task ids to complete before smaller ones. Also
* when the taskq contains delay tasks with small task ids callers may
* block for a considerable length of time waiting for them to expire and
* execute.
*/
static int
taskq_wait_id_check(taskq_t *tq, taskqid_t id)
{
int rc;
unsigned long flags;
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
rc = (taskq_find(tq, id) == NULL);
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (rc);
}
/*
* The taskq_wait_id() function blocks until the passed task id completes.
* This does not guarantee that all lower task ids have completed.
*/
void
taskq_wait_id(taskq_t *tq, taskqid_t id)
{
wait_event(tq->tq_wait_waitq, taskq_wait_id_check(tq, id));
}
EXPORT_SYMBOL(taskq_wait_id);
static int
taskq_wait_outstanding_check(taskq_t *tq, taskqid_t id)
{
int rc;
unsigned long flags;
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
rc = (id < tq->tq_lowest_id);
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (rc);
}
/*
* The taskq_wait_outstanding() function will block until all tasks with a
* lower taskqid than the passed 'id' have been completed. Note that all
* task id's are assigned monotonically at dispatch time. Zero may be
* passed for the id to indicate all tasks dispatch up to this point,
* but not after, should be waited for.
*/
void
taskq_wait_outstanding(taskq_t *tq, taskqid_t id)
{
id = id ? id : tq->tq_next_id - 1;
wait_event(tq->tq_wait_waitq, taskq_wait_outstanding_check(tq, id));
}
EXPORT_SYMBOL(taskq_wait_outstanding);
static int
taskq_wait_check(taskq_t *tq)
{
int rc;
unsigned long flags;
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
rc = (tq->tq_lowest_id == tq->tq_next_id);
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (rc);
}
/*
* The taskq_wait() function will block until the taskq is empty.
* This means that if a taskq re-dispatches work to itself taskq_wait()
* callers will block indefinitely.
*/
void
taskq_wait(taskq_t *tq)
{
wait_event(tq->tq_wait_waitq, taskq_wait_check(tq));
}
EXPORT_SYMBOL(taskq_wait);
int
taskq_member(taskq_t *tq, kthread_t *t)
{
return (tq == (taskq_t *)tsd_get_by_thread(taskq_tsd, t));
}
EXPORT_SYMBOL(taskq_member);
taskq_t *
taskq_of_curthread(void)
{
return (tsd_get(taskq_tsd));
}
EXPORT_SYMBOL(taskq_of_curthread);
/*
* Cancel an already dispatched task given the task id. Still pending tasks
* will be immediately canceled, and if the task is active the function will
* block until it completes. Preallocated tasks which are canceled must be
* freed by the caller.
*/
int
taskq_cancel_id(taskq_t *tq, taskqid_t id)
{
taskq_ent_t *t;
int rc = ENOENT;
unsigned long flags;
ASSERT(tq);
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
t = taskq_find(tq, id);
if (t && t != ERR_PTR(-EBUSY)) {
list_del_init(&t->tqent_list);
t->tqent_flags |= TQENT_FLAG_CANCEL;
/*
* When canceling the lowest outstanding task id we
* must recalculate the new lowest outstanding id.
*/
if (tq->tq_lowest_id == t->tqent_id) {
tq->tq_lowest_id = taskq_lowest_id(tq);
ASSERT3S(tq->tq_lowest_id, >, t->tqent_id);
}
/*
* The task_expire() function takes the tq->tq_lock so drop
* drop the lock before synchronously cancelling the timer.
*/
if (timer_pending(&t->tqent_timer)) {
spin_unlock_irqrestore(&tq->tq_lock, flags);
del_timer_sync(&t->tqent_timer);
spin_lock_irqsave_nested(&tq->tq_lock, flags,
tq->tq_lock_class);
}
if (!(t->tqent_flags & TQENT_FLAG_PREALLOC))
task_done(tq, t);
rc = 0;
}
spin_unlock_irqrestore(&tq->tq_lock, flags);
if (t == ERR_PTR(-EBUSY)) {
taskq_wait_id(tq, id);
rc = EBUSY;
}
return (rc);
}
EXPORT_SYMBOL(taskq_cancel_id);
static int taskq_thread_spawn(taskq_t *tq);
taskqid_t
taskq_dispatch(taskq_t *tq, task_func_t func, void *arg, uint_t flags)
{
taskq_ent_t *t;
taskqid_t rc = TASKQID_INVALID;
unsigned long irqflags;
ASSERT(tq);
ASSERT(func);
spin_lock_irqsave_nested(&tq->tq_lock, irqflags, tq->tq_lock_class);
/* Taskq being destroyed and all tasks drained */
if (!(tq->tq_flags & TASKQ_ACTIVE))
goto out;
/* Do not queue the task unless there is idle thread for it */
ASSERT(tq->tq_nactive <= tq->tq_nthreads);
if ((flags & TQ_NOQUEUE) && (tq->tq_nactive == tq->tq_nthreads)) {
/* Dynamic taskq may be able to spawn another thread */
if (!(tq->tq_flags & TASKQ_DYNAMIC) ||
taskq_thread_spawn(tq) == 0)
goto out;
}
if ((t = task_alloc(tq, flags, &irqflags)) == NULL)
goto out;
spin_lock(&t->tqent_lock);
/* Queue to the front of the list to enforce TQ_NOQUEUE semantics */
if (flags & TQ_NOQUEUE)
list_add(&t->tqent_list, &tq->tq_prio_list);
/* Queue to the priority list instead of the pending list */
else if (flags & TQ_FRONT)
list_add_tail(&t->tqent_list, &tq->tq_prio_list);
else
list_add_tail(&t->tqent_list, &tq->tq_pend_list);
t->tqent_id = rc = tq->tq_next_id;
tq->tq_next_id++;
t->tqent_func = func;
t->tqent_arg = arg;
t->tqent_taskq = tq;
t->tqent_timer.function = NULL;
t->tqent_timer.expires = 0;
t->tqent_birth = jiffies;
DTRACE_PROBE1(taskq_ent__birth, taskq_ent_t *, t);
ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));
spin_unlock(&t->tqent_lock);
wake_up(&tq->tq_work_waitq);
out:
/* Spawn additional taskq threads if required. */
if (!(flags & TQ_NOQUEUE) && tq->tq_nactive == tq->tq_nthreads)
(void) taskq_thread_spawn(tq);
spin_unlock_irqrestore(&tq->tq_lock, irqflags);
return (rc);
}
EXPORT_SYMBOL(taskq_dispatch);
taskqid_t
taskq_dispatch_delay(taskq_t *tq, task_func_t func, void *arg,
uint_t flags, clock_t expire_time)
{
taskqid_t rc = TASKQID_INVALID;
taskq_ent_t *t;
unsigned long irqflags;
ASSERT(tq);
ASSERT(func);
spin_lock_irqsave_nested(&tq->tq_lock, irqflags, tq->tq_lock_class);
/* Taskq being destroyed and all tasks drained */
if (!(tq->tq_flags & TASKQ_ACTIVE))
goto out;
if ((t = task_alloc(tq, flags, &irqflags)) == NULL)
goto out;
spin_lock(&t->tqent_lock);
/* Queue to the delay list for subsequent execution */
list_add_tail(&t->tqent_list, &tq->tq_delay_list);
t->tqent_id = rc = tq->tq_next_id;
tq->tq_next_id++;
t->tqent_func = func;
t->tqent_arg = arg;
t->tqent_taskq = tq;
t->tqent_timer.function = task_expire;
t->tqent_timer.expires = (unsigned long)expire_time;
add_timer(&t->tqent_timer);
ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));
spin_unlock(&t->tqent_lock);
out:
/* Spawn additional taskq threads if required. */
if (tq->tq_nactive == tq->tq_nthreads)
(void) taskq_thread_spawn(tq);
spin_unlock_irqrestore(&tq->tq_lock, irqflags);
return (rc);
}
EXPORT_SYMBOL(taskq_dispatch_delay);
void
taskq_dispatch_ent(taskq_t *tq, task_func_t func, void *arg, uint_t flags,
taskq_ent_t *t)
{
unsigned long irqflags;
ASSERT(tq);
ASSERT(func);
spin_lock_irqsave_nested(&tq->tq_lock, irqflags,
tq->tq_lock_class);
/* Taskq being destroyed and all tasks drained */
if (!(tq->tq_flags & TASKQ_ACTIVE)) {
t->tqent_id = TASKQID_INVALID;
goto out;
}
if ((flags & TQ_NOQUEUE) && (tq->tq_nactive == tq->tq_nthreads)) {
/* Dynamic taskq may be able to spawn another thread */
if (!(tq->tq_flags & TASKQ_DYNAMIC) ||
taskq_thread_spawn(tq) == 0)
goto out2;
flags |= TQ_FRONT;
}
spin_lock(&t->tqent_lock);
/*
* Make sure the entry is not on some other taskq; it is important to
* ASSERT() under lock
*/
ASSERT(taskq_empty_ent(t));
/*
* Mark it as a prealloc'd task. This is important
* to ensure that we don't free it later.
*/
t->tqent_flags |= TQENT_FLAG_PREALLOC;
/* Queue to the priority list instead of the pending list */
if (flags & TQ_FRONT)
list_add_tail(&t->tqent_list, &tq->tq_prio_list);
else
list_add_tail(&t->tqent_list, &tq->tq_pend_list);
t->tqent_id = tq->tq_next_id;
tq->tq_next_id++;
t->tqent_func = func;
t->tqent_arg = arg;
t->tqent_taskq = tq;
t->tqent_birth = jiffies;
DTRACE_PROBE1(taskq_ent__birth, taskq_ent_t *, t);
spin_unlock(&t->tqent_lock);
wake_up(&tq->tq_work_waitq);
out:
/* Spawn additional taskq threads if required. */
if (tq->tq_nactive == tq->tq_nthreads)
(void) taskq_thread_spawn(tq);
out2:
spin_unlock_irqrestore(&tq->tq_lock, irqflags);
}
EXPORT_SYMBOL(taskq_dispatch_ent);
int
taskq_empty_ent(taskq_ent_t *t)
{
return (list_empty(&t->tqent_list));
}
EXPORT_SYMBOL(taskq_empty_ent);
void
taskq_init_ent(taskq_ent_t *t)
{
spin_lock_init(&t->tqent_lock);
init_waitqueue_head(&t->tqent_waitq);
timer_setup(&t->tqent_timer, NULL, 0);
INIT_LIST_HEAD(&t->tqent_list);
t->tqent_id = 0;
t->tqent_func = NULL;
t->tqent_arg = NULL;
t->tqent_flags = 0;
t->tqent_taskq = NULL;
}
EXPORT_SYMBOL(taskq_init_ent);
/*
* Return the next pending task, preference is given to tasks on the
* priority list which were dispatched with TQ_FRONT.
*/
static taskq_ent_t *
taskq_next_ent(taskq_t *tq)
{
struct list_head *list;
if (!list_empty(&tq->tq_prio_list))
list = &tq->tq_prio_list;
else if (!list_empty(&tq->tq_pend_list))
list = &tq->tq_pend_list;
else
return (NULL);
return (list_entry(list->next, taskq_ent_t, tqent_list));
}
/*
* Spawns a new thread for the specified taskq.
*/
static void
taskq_thread_spawn_task(void *arg)
{
taskq_t *tq = (taskq_t *)arg;
unsigned long flags;
if (taskq_thread_create(tq) == NULL) {
/* restore spawning count if failed */
spin_lock_irqsave_nested(&tq->tq_lock, flags,
tq->tq_lock_class);
tq->tq_nspawn--;
spin_unlock_irqrestore(&tq->tq_lock, flags);
}
}
/*
* Spawn addition threads for dynamic taskqs (TASKQ_DYNAMIC) the current
* number of threads is insufficient to handle the pending tasks. These
* new threads must be created by the dedicated dynamic_taskq to avoid
* deadlocks between thread creation and memory reclaim. The system_taskq
* which is also a dynamic taskq cannot be safely used for this.
*/
static int
taskq_thread_spawn(taskq_t *tq)
{
int spawning = 0;
if (!(tq->tq_flags & TASKQ_DYNAMIC))
return (0);
if ((tq->tq_nthreads + tq->tq_nspawn < tq->tq_maxthreads) &&
(tq->tq_flags & TASKQ_ACTIVE)) {
spawning = (++tq->tq_nspawn);
taskq_dispatch(dynamic_taskq, taskq_thread_spawn_task,
tq, TQ_NOSLEEP);
}
return (spawning);
}
/*
* Threads in a dynamic taskq should only exit once it has been completely
* drained and no other threads are actively servicing tasks. This prevents
* threads from being created and destroyed more than is required.
*
* The first thread is the thread list is treated as the primary thread.
* There is nothing special about the primary thread but in order to avoid
* all the taskq pids from changing we opt to make it long running.
*/
static int
taskq_thread_should_stop(taskq_t *tq, taskq_thread_t *tqt)
{
if (!(tq->tq_flags & TASKQ_DYNAMIC))
return (0);
if (list_first_entry(&(tq->tq_thread_list), taskq_thread_t,
tqt_thread_list) == tqt)
return (0);
return
((tq->tq_nspawn == 0) && /* No threads are being spawned */
(tq->tq_nactive == 0) && /* No threads are handling tasks */
(tq->tq_nthreads > 1) && /* More than 1 thread is running */
(!taskq_next_ent(tq)) && /* There are no pending tasks */
(spl_taskq_thread_dynamic)); /* Dynamic taskqs are allowed */
}
static int
taskq_thread(void *args)
{
DECLARE_WAITQUEUE(wait, current);
sigset_t blocked;
taskq_thread_t *tqt = args;
taskq_t *tq;
taskq_ent_t *t;
int seq_tasks = 0;
unsigned long flags;
taskq_ent_t dup_task = {};
ASSERT(tqt);
ASSERT(tqt->tqt_tq);
tq = tqt->tqt_tq;
current->flags |= PF_NOFREEZE;
(void) spl_fstrans_mark();
sigfillset(&blocked);
sigprocmask(SIG_BLOCK, &blocked, NULL);
flush_signals(current);
tsd_set(taskq_tsd, tq);
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
/*
* If we are dynamically spawned, decrease spawning count. Note that
* we could be created during taskq_create, in which case we shouldn't
* do the decrement. But it's fine because taskq_create will reset
* tq_nspawn later.
*/
if (tq->tq_flags & TASKQ_DYNAMIC)
tq->tq_nspawn--;
/* Immediately exit if more threads than allowed were created. */
if (tq->tq_nthreads >= tq->tq_maxthreads)
goto error;
tq->tq_nthreads++;
list_add_tail(&tqt->tqt_thread_list, &tq->tq_thread_list);
wake_up(&tq->tq_wait_waitq);
set_current_state(TASK_INTERRUPTIBLE);
while (!kthread_should_stop()) {
if (list_empty(&tq->tq_pend_list) &&
list_empty(&tq->tq_prio_list)) {
if (taskq_thread_should_stop(tq, tqt)) {
wake_up_all(&tq->tq_wait_waitq);
break;
}
add_wait_queue_exclusive(&tq->tq_work_waitq, &wait);
spin_unlock_irqrestore(&tq->tq_lock, flags);
schedule();
seq_tasks = 0;
spin_lock_irqsave_nested(&tq->tq_lock, flags,
tq->tq_lock_class);
remove_wait_queue(&tq->tq_work_waitq, &wait);
} else {
__set_current_state(TASK_RUNNING);
}
if ((t = taskq_next_ent(tq)) != NULL) {
list_del_init(&t->tqent_list);
/*
* A TQENT_FLAG_PREALLOC task may be reused or freed
* during the task function call. Store tqent_id and
* tqent_flags here.
*
* Also use an on stack taskq_ent_t for tqt_task
* assignment in this case; we want to make sure
* to duplicate all fields, so the values are
* correct when it's accessed via DTRACE_PROBE*.
*/
tqt->tqt_id = t->tqent_id;
tqt->tqt_flags = t->tqent_flags;
if (t->tqent_flags & TQENT_FLAG_PREALLOC) {
dup_task = *t;
t = &dup_task;
}
tqt->tqt_task = t;
taskq_insert_in_order(tq, tqt);
tq->tq_nactive++;
spin_unlock_irqrestore(&tq->tq_lock, flags);
DTRACE_PROBE1(taskq_ent__start, taskq_ent_t *, t);
/* Perform the requested task */
t->tqent_func(t->tqent_arg);
DTRACE_PROBE1(taskq_ent__finish, taskq_ent_t *, t);
spin_lock_irqsave_nested(&tq->tq_lock, flags,
tq->tq_lock_class);
tq->tq_nactive--;
list_del_init(&tqt->tqt_active_list);
tqt->tqt_task = NULL;
/* For prealloc'd tasks, we don't free anything. */
if (!(tqt->tqt_flags & TQENT_FLAG_PREALLOC))
task_done(tq, t);
/*
* When the current lowest outstanding taskqid is
* done calculate the new lowest outstanding id
*/
if (tq->tq_lowest_id == tqt->tqt_id) {
tq->tq_lowest_id = taskq_lowest_id(tq);
ASSERT3S(tq->tq_lowest_id, >, tqt->tqt_id);
}
/* Spawn additional taskq threads if required. */
if ((++seq_tasks) > spl_taskq_thread_sequential &&
taskq_thread_spawn(tq))
seq_tasks = 0;
tqt->tqt_id = TASKQID_INVALID;
tqt->tqt_flags = 0;
wake_up_all(&tq->tq_wait_waitq);
} else {
if (taskq_thread_should_stop(tq, tqt))
break;
}
set_current_state(TASK_INTERRUPTIBLE);
}
__set_current_state(TASK_RUNNING);
tq->tq_nthreads--;
list_del_init(&tqt->tqt_thread_list);
error:
kmem_free(tqt, sizeof (taskq_thread_t));
spin_unlock_irqrestore(&tq->tq_lock, flags);
tsd_set(taskq_tsd, NULL);
thread_exit();
return (0);
}
static taskq_thread_t *
taskq_thread_create(taskq_t *tq)
{
static int last_used_cpu = 0;
taskq_thread_t *tqt;
tqt = kmem_alloc(sizeof (*tqt), KM_PUSHPAGE);
INIT_LIST_HEAD(&tqt->tqt_thread_list);
INIT_LIST_HEAD(&tqt->tqt_active_list);
tqt->tqt_tq = tq;
tqt->tqt_id = TASKQID_INVALID;
tqt->tqt_thread = spl_kthread_create(taskq_thread, tqt,
"%s", tq->tq_name);
if (tqt->tqt_thread == NULL) {
kmem_free(tqt, sizeof (taskq_thread_t));
return (NULL);
}
if (spl_taskq_thread_bind) {
last_used_cpu = (last_used_cpu + 1) % num_online_cpus();
kthread_bind(tqt->tqt_thread, last_used_cpu);
}
if (spl_taskq_thread_priority)
set_user_nice(tqt->tqt_thread, PRIO_TO_NICE(tq->tq_pri));
wake_up_process(tqt->tqt_thread);
return (tqt);
}
taskq_t *
taskq_create(const char *name, int threads_arg, pri_t pri,
int minalloc, int maxalloc, uint_t flags)
{
taskq_t *tq;
taskq_thread_t *tqt;
int count = 0, rc = 0, i;
unsigned long irqflags;
int nthreads = threads_arg;
ASSERT(name != NULL);
ASSERT(minalloc >= 0);
ASSERT(maxalloc <= INT_MAX);
ASSERT(!(flags & (TASKQ_CPR_SAFE))); /* Unsupported */
/* Scale the number of threads using nthreads as a percentage */
if (flags & TASKQ_THREADS_CPU_PCT) {
ASSERT(nthreads <= 100);
ASSERT(nthreads >= 0);
nthreads = MIN(threads_arg, 100);
nthreads = MAX(nthreads, 0);
nthreads = MAX((num_online_cpus() * nthreads) /100, 1);
}
tq = kmem_alloc(sizeof (*tq), KM_PUSHPAGE);
if (tq == NULL)
return (NULL);
tq->tq_hp_support = B_FALSE;
#ifdef HAVE_CPU_HOTPLUG
if (flags & TASKQ_THREADS_CPU_PCT) {
tq->tq_hp_support = B_TRUE;
if (cpuhp_state_add_instance_nocalls(spl_taskq_cpuhp_state,
&tq->tq_hp_cb_node) != 0) {
kmem_free(tq, sizeof (*tq));
return (NULL);
}
}
#endif
spin_lock_init(&tq->tq_lock);
INIT_LIST_HEAD(&tq->tq_thread_list);
INIT_LIST_HEAD(&tq->tq_active_list);
tq->tq_name = kmem_strdup(name);
tq->tq_nactive = 0;
tq->tq_nthreads = 0;
tq->tq_nspawn = 0;
tq->tq_maxthreads = nthreads;
tq->tq_cpu_pct = threads_arg;
tq->tq_pri = pri;
tq->tq_minalloc = minalloc;
tq->tq_maxalloc = maxalloc;
tq->tq_nalloc = 0;
tq->tq_flags = (flags | TASKQ_ACTIVE);
tq->tq_next_id = TASKQID_INITIAL;
tq->tq_lowest_id = TASKQID_INITIAL;
INIT_LIST_HEAD(&tq->tq_free_list);
INIT_LIST_HEAD(&tq->tq_pend_list);
INIT_LIST_HEAD(&tq->tq_prio_list);
INIT_LIST_HEAD(&tq->tq_delay_list);
init_waitqueue_head(&tq->tq_work_waitq);
init_waitqueue_head(&tq->tq_wait_waitq);
tq->tq_lock_class = TQ_LOCK_GENERAL;
INIT_LIST_HEAD(&tq->tq_taskqs);
if (flags & TASKQ_PREPOPULATE) {
spin_lock_irqsave_nested(&tq->tq_lock, irqflags,
tq->tq_lock_class);
for (i = 0; i < minalloc; i++)
task_done(tq, task_alloc(tq, TQ_PUSHPAGE | TQ_NEW,
&irqflags));
spin_unlock_irqrestore(&tq->tq_lock, irqflags);
}
if ((flags & TASKQ_DYNAMIC) && spl_taskq_thread_dynamic)
nthreads = 1;
for (i = 0; i < nthreads; i++) {
tqt = taskq_thread_create(tq);
if (tqt == NULL)
rc = 1;
else
count++;
}
/* Wait for all threads to be started before potential destroy */
wait_event(tq->tq_wait_waitq, tq->tq_nthreads == count);
/*
* taskq_thread might have touched nspawn, but we don't want them to
* because they're not dynamically spawned. So we reset it to 0
*/
tq->tq_nspawn = 0;
if (rc) {
taskq_destroy(tq);
tq = NULL;
} else {
down_write(&tq_list_sem);
tq->tq_instance = taskq_find_by_name(name) + 1;
list_add_tail(&tq->tq_taskqs, &tq_list);
up_write(&tq_list_sem);
}
return (tq);
}
EXPORT_SYMBOL(taskq_create);
void
taskq_destroy(taskq_t *tq)
{
struct task_struct *thread;
taskq_thread_t *tqt;
taskq_ent_t *t;
unsigned long flags;
ASSERT(tq);
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
tq->tq_flags &= ~TASKQ_ACTIVE;
spin_unlock_irqrestore(&tq->tq_lock, flags);
#ifdef HAVE_CPU_HOTPLUG
if (tq->tq_hp_support) {
VERIFY0(cpuhp_state_remove_instance_nocalls(
spl_taskq_cpuhp_state, &tq->tq_hp_cb_node));
}
#endif
/*
* When TASKQ_ACTIVE is clear new tasks may not be added nor may
* new worker threads be spawned for dynamic taskq.
*/
if (dynamic_taskq != NULL)
taskq_wait_outstanding(dynamic_taskq, 0);
taskq_wait(tq);
/* remove taskq from global list used by the kstats */
down_write(&tq_list_sem);
list_del(&tq->tq_taskqs);
up_write(&tq_list_sem);
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
/* wait for spawning threads to insert themselves to the list */
while (tq->tq_nspawn) {
spin_unlock_irqrestore(&tq->tq_lock, flags);
schedule_timeout_interruptible(1);
spin_lock_irqsave_nested(&tq->tq_lock, flags,
tq->tq_lock_class);
}
/*
* Signal each thread to exit and block until it does. Each thread
* is responsible for removing itself from the list and freeing its
* taskq_thread_t. This allows for idle threads to opt to remove
* themselves from the taskq. They can be recreated as needed.
*/
while (!list_empty(&tq->tq_thread_list)) {
tqt = list_entry(tq->tq_thread_list.next,
taskq_thread_t, tqt_thread_list);
thread = tqt->tqt_thread;
spin_unlock_irqrestore(&tq->tq_lock, flags);
kthread_stop(thread);
spin_lock_irqsave_nested(&tq->tq_lock, flags,
tq->tq_lock_class);
}
while (!list_empty(&tq->tq_free_list)) {
t = list_entry(tq->tq_free_list.next, taskq_ent_t, tqent_list);
ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));
list_del_init(&t->tqent_list);
task_free(tq, t);
}
ASSERT0(tq->tq_nthreads);
ASSERT0(tq->tq_nalloc);
ASSERT0(tq->tq_nspawn);
ASSERT(list_empty(&tq->tq_thread_list));
ASSERT(list_empty(&tq->tq_active_list));
ASSERT(list_empty(&tq->tq_free_list));
ASSERT(list_empty(&tq->tq_pend_list));
ASSERT(list_empty(&tq->tq_prio_list));
ASSERT(list_empty(&tq->tq_delay_list));
spin_unlock_irqrestore(&tq->tq_lock, flags);
kmem_strfree(tq->tq_name);
kmem_free(tq, sizeof (taskq_t));
}
EXPORT_SYMBOL(taskq_destroy);
static unsigned int spl_taskq_kick = 0;
/*
* 2.6.36 API Change
* module_param_cb is introduced to take kernel_param_ops and
* module_param_call is marked as obsolete. Also set and get operations
* were changed to take a 'const struct kernel_param *'.
*/
static int
#ifdef module_param_cb
param_set_taskq_kick(const char *val, const struct kernel_param *kp)
#else
param_set_taskq_kick(const char *val, struct kernel_param *kp)
#endif
{
int ret;
taskq_t *tq = NULL;
taskq_ent_t *t;
unsigned long flags;
ret = param_set_uint(val, kp);
if (ret < 0 || !spl_taskq_kick)
return (ret);
/* reset value */
spl_taskq_kick = 0;
down_read(&tq_list_sem);
list_for_each_entry(tq, &tq_list, tq_taskqs) {
spin_lock_irqsave_nested(&tq->tq_lock, flags,
tq->tq_lock_class);
/* Check if the first pending is older than 5 seconds */
t = taskq_next_ent(tq);
if (t && time_after(jiffies, t->tqent_birth + 5*HZ)) {
(void) taskq_thread_spawn(tq);
printk(KERN_INFO "spl: Kicked taskq %s/%d\n",
tq->tq_name, tq->tq_instance);
}
spin_unlock_irqrestore(&tq->tq_lock, flags);
}
up_read(&tq_list_sem);
return (ret);
}
#ifdef module_param_cb
static const struct kernel_param_ops param_ops_taskq_kick = {
.set = param_set_taskq_kick,
.get = param_get_uint,
};
module_param_cb(spl_taskq_kick, &param_ops_taskq_kick, &spl_taskq_kick, 0644);
#else
module_param_call(spl_taskq_kick, param_set_taskq_kick, param_get_uint,
&spl_taskq_kick, 0644);
#endif
MODULE_PARM_DESC(spl_taskq_kick,
"Write nonzero to kick stuck taskqs to spawn more threads");
#ifdef HAVE_CPU_HOTPLUG
/*
* This callback will be called exactly once for each core that comes online,
* for each dynamic taskq. We attempt to expand taskqs that have
* TASKQ_THREADS_CPU_PCT set. We need to redo the percentage calculation every
* time, to correctly determine whether or not to add a thread.
*/
static int
spl_taskq_expand(unsigned int cpu, struct hlist_node *node)
{
taskq_t *tq = list_entry(node, taskq_t, tq_hp_cb_node);
unsigned long flags;
int err = 0;
ASSERT(tq);
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
if (!(tq->tq_flags & TASKQ_ACTIVE)) {
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (err);
}
ASSERT(tq->tq_flags & TASKQ_THREADS_CPU_PCT);
int nthreads = MIN(tq->tq_cpu_pct, 100);
nthreads = MAX(((num_online_cpus() + 1) * nthreads) / 100, 1);
tq->tq_maxthreads = nthreads;
if (!((tq->tq_flags & TASKQ_DYNAMIC) && spl_taskq_thread_dynamic) &&
tq->tq_maxthreads > tq->tq_nthreads) {
spin_unlock_irqrestore(&tq->tq_lock, flags);
taskq_thread_t *tqt = taskq_thread_create(tq);
if (tqt == NULL)
err = -1;
return (err);
}
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (err);
}
/*
* While we don't support offlining CPUs, it is possible that CPUs will fail
* to online successfully. We do need to be able to handle this case
* gracefully.
*/
static int
spl_taskq_prepare_down(unsigned int cpu, struct hlist_node *node)
{
taskq_t *tq = list_entry(node, taskq_t, tq_hp_cb_node);
unsigned long flags;
ASSERT(tq);
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
if (!(tq->tq_flags & TASKQ_ACTIVE))
goto out;
ASSERT(tq->tq_flags & TASKQ_THREADS_CPU_PCT);
int nthreads = MIN(tq->tq_cpu_pct, 100);
nthreads = MAX(((num_online_cpus()) * nthreads) / 100, 1);
tq->tq_maxthreads = nthreads;
if (!((tq->tq_flags & TASKQ_DYNAMIC) && spl_taskq_thread_dynamic) &&
tq->tq_maxthreads < tq->tq_nthreads) {
ASSERT3U(tq->tq_maxthreads, ==, tq->tq_nthreads - 1);
taskq_thread_t *tqt = list_entry(tq->tq_thread_list.next,
taskq_thread_t, tqt_thread_list);
struct task_struct *thread = tqt->tqt_thread;
spin_unlock_irqrestore(&tq->tq_lock, flags);
kthread_stop(thread);
return (0);
}
out:
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (0);
}
#endif
int
spl_taskq_init(void)
{
init_rwsem(&tq_list_sem);
tsd_create(&taskq_tsd, NULL);
#ifdef HAVE_CPU_HOTPLUG
spl_taskq_cpuhp_state = cpuhp_setup_state_multi(CPUHP_AP_ONLINE_DYN,
"fs/spl_taskq:online", spl_taskq_expand, spl_taskq_prepare_down);
#endif
system_taskq = taskq_create("spl_system_taskq", MAX(boot_ncpus, 64),
maxclsyspri, boot_ncpus, INT_MAX, TASKQ_PREPOPULATE|TASKQ_DYNAMIC);
if (system_taskq == NULL)
return (1);
system_delay_taskq = taskq_create("spl_delay_taskq", MAX(boot_ncpus, 4),
maxclsyspri, boot_ncpus, INT_MAX, TASKQ_PREPOPULATE|TASKQ_DYNAMIC);
if (system_delay_taskq == NULL) {
#ifdef HAVE_CPU_HOTPLUG
cpuhp_remove_multi_state(spl_taskq_cpuhp_state);
#endif
taskq_destroy(system_taskq);
return (1);
}
dynamic_taskq = taskq_create("spl_dynamic_taskq", 1,
maxclsyspri, boot_ncpus, INT_MAX, TASKQ_PREPOPULATE);
if (dynamic_taskq == NULL) {
#ifdef HAVE_CPU_HOTPLUG
cpuhp_remove_multi_state(spl_taskq_cpuhp_state);
#endif
taskq_destroy(system_taskq);
taskq_destroy(system_delay_taskq);
return (1);
}
/*
* This is used to annotate tq_lock, so
* taskq_dispatch -> taskq_thread_spawn -> taskq_dispatch
* does not trigger a lockdep warning re: possible recursive locking
*/
dynamic_taskq->tq_lock_class = TQ_LOCK_DYNAMIC;
return (0);
}
void
spl_taskq_fini(void)
{
taskq_destroy(dynamic_taskq);
dynamic_taskq = NULL;
taskq_destroy(system_delay_taskq);
system_delay_taskq = NULL;
taskq_destroy(system_taskq);
system_taskq = NULL;
tsd_destroy(&taskq_tsd);
#ifdef HAVE_CPU_HOTPLUG
cpuhp_remove_multi_state(spl_taskq_cpuhp_state);
spl_taskq_cpuhp_state = 0;
#endif
}
diff --git a/module/os/linux/spl/spl-xdr.c b/module/os/linux/spl/spl-xdr.c
index 5e763c25606f..6b77524181db 100644
--- a/module/os/linux/spl/spl-xdr.c
+++ b/module/os/linux/spl/spl-xdr.c
@@ -1,512 +1,512 @@
/*
* Copyright (c) 2008-2010 Sun Microsystems, Inc.
* Written by Ricardo Correia <Ricardo.M.Correia@Sun.COM>
*
* This file is part of the SPL, Solaris Porting Layer.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*
* Solaris Porting Layer (SPL) XDR Implementation.
*/
#include <linux/string.h>
#include <sys/kmem.h>
#include <sys/debug.h>
#include <sys/types.h>
#include <sys/sysmacros.h>
#include <rpc/xdr.h>
/*
* SPL's XDR mem implementation.
*
* This is used by libnvpair to serialize/deserialize the name-value pair data
* structures into byte arrays in a well-defined and portable manner.
*
* These data structures are used by the DMU/ZFS to flexibly manipulate various
* information in memory and later serialize it/deserialize it to disk.
* Examples of usages include the pool configuration, lists of pool and dataset
* properties, etc.
*
* Reference documentation for the XDR representation and XDR operations can be
* found in RFC 1832 and xdr(3), respectively.
*
* === Implementation shortcomings ===
*
* It is assumed that the following C types have the following sizes:
*
* char/unsigned char: 1 byte
* short/unsigned short: 2 bytes
* int/unsigned int: 4 bytes
* longlong_t/u_longlong_t: 8 bytes
*
* The C standard allows these types to be larger (and in the case of ints,
* shorter), so if that is the case on some compiler/architecture, the build
* will fail (on purpose).
*
* If someone wants to fix the code to work properly on such environments, then:
*
* 1) Preconditions should be added to xdrmem_enc functions to make sure the
* caller doesn't pass arguments which exceed the expected range.
* 2) Functions which take signed integers should be changed to properly do
* sign extension.
* 3) For ints with less than 32 bits, well.. I suspect you'll have bigger
* problems than this implementation.
*
* It is also assumed that:
*
* 1) Chars have 8 bits.
* 2) We can always do 32-bit-aligned int memory accesses and byte-aligned
* memcpy, memset and memcmp.
* 3) Arrays passed to xdr_array() are packed and the compiler/architecture
* supports element-sized-aligned memory accesses.
* 4) Negative integers are natively stored in two's complement binary
* representation.
*
* No checks are done for the 4 assumptions above, though.
*
* === Caller expectations ===
*
* Existing documentation does not describe the semantics of XDR operations very
* well. Therefore, some assumptions about failure semantics will be made and
* will be described below:
*
* 1) If any encoding operation fails (e.g., due to lack of buffer space), the
* the stream should be considered valid only up to the encoding operation
* previous to the one that first failed. However, the stream size as returned
* by xdr_control() cannot be considered to be strictly correct (it may be
* bigger).
*
* Putting it another way, if there is an encoding failure it's undefined
* whether anything is added to the stream in that operation and therefore
* neither xdr_control() nor future encoding operations on the same stream can
* be relied upon to produce correct results.
*
* 2) If a decoding operation fails, it's undefined whether anything will be
* decoded into passed buffers/pointers during that operation, or what the
* values on those buffers will look like.
*
* Future decoding operations on the same stream will also have similar
* undefined behavior.
*
* 3) When the first decoding operation fails it is OK to trust the results of
* previous decoding operations on the same stream, as long as the caller
* expects a failure to be possible (e.g. due to end-of-stream).
*
* However, this is highly discouraged because the caller should know the
* stream size and should be coded to expect any decoding failure to be data
* corruption due to hardware, accidental or even malicious causes, which should
* be handled gracefully in all cases.
*
* In very rare situations where there are strong reasons to believe the data
* can be trusted to be valid and non-tampered with, then the caller may assume
* a decoding failure to be a bug (e.g. due to mismatched data types) and may
* fail non-gracefully.
*
* 4) Non-zero padding bytes will cause the decoding operation to fail.
*
* 5) Zero bytes on string types will also cause the decoding operation to fail.
*
* 6) It is assumed that either the pointer to the stream buffer given by the
* caller is 32-bit aligned or the architecture supports non-32-bit-aligned int
* memory accesses.
*
* 7) The stream buffer and encoding/decoding buffers/ptrs should not overlap.
*
* 8) If a caller passes pointers to non-kernel memory (e.g., pointers to user
* space or MMIO space), the computer may explode.
*/
-static struct xdr_ops xdrmem_encode_ops;
-static struct xdr_ops xdrmem_decode_ops;
+static const struct xdr_ops xdrmem_encode_ops;
+static const struct xdr_ops xdrmem_decode_ops;
void
xdrmem_create(XDR *xdrs, const caddr_t addr, const uint_t size,
const enum xdr_op op)
{
switch (op) {
case XDR_ENCODE:
xdrs->x_ops = &xdrmem_encode_ops;
break;
case XDR_DECODE:
xdrs->x_ops = &xdrmem_decode_ops;
break;
default:
xdrs->x_ops = NULL; /* Let the caller know we failed */
return;
}
xdrs->x_op = op;
xdrs->x_addr = addr;
xdrs->x_addr_end = addr + size;
if (xdrs->x_addr_end < xdrs->x_addr) {
xdrs->x_ops = NULL;
}
}
EXPORT_SYMBOL(xdrmem_create);
static bool_t
xdrmem_control(XDR *xdrs, int req, void *info)
{
struct xdr_bytesrec *rec = (struct xdr_bytesrec *)info;
if (req != XDR_GET_BYTES_AVAIL)
return (FALSE);
rec->xc_is_last_record = TRUE; /* always TRUE in xdrmem streams */
rec->xc_num_avail = xdrs->x_addr_end - xdrs->x_addr;
return (TRUE);
}
static bool_t
xdrmem_enc_bytes(XDR *xdrs, caddr_t cp, const uint_t cnt)
{
uint_t size = roundup(cnt, 4);
uint_t pad;
if (size < cnt)
return (FALSE); /* Integer overflow */
if (xdrs->x_addr > xdrs->x_addr_end)
return (FALSE);
if (xdrs->x_addr_end - xdrs->x_addr < size)
return (FALSE);
memcpy(xdrs->x_addr, cp, cnt);
xdrs->x_addr += cnt;
pad = size - cnt;
if (pad > 0) {
memset(xdrs->x_addr, 0, pad);
xdrs->x_addr += pad;
}
return (TRUE);
}
static bool_t
xdrmem_dec_bytes(XDR *xdrs, caddr_t cp, const uint_t cnt)
{
static uint32_t zero = 0;
uint_t size = roundup(cnt, 4);
uint_t pad;
if (size < cnt)
return (FALSE); /* Integer overflow */
if (xdrs->x_addr > xdrs->x_addr_end)
return (FALSE);
if (xdrs->x_addr_end - xdrs->x_addr < size)
return (FALSE);
memcpy(cp, xdrs->x_addr, cnt);
xdrs->x_addr += cnt;
pad = size - cnt;
if (pad > 0) {
/* An inverted memchr() would be useful here... */
if (memcmp(&zero, xdrs->x_addr, pad) != 0)
return (FALSE);
xdrs->x_addr += pad;
}
return (TRUE);
}
static bool_t
xdrmem_enc_uint32(XDR *xdrs, uint32_t val)
{
if (xdrs->x_addr + sizeof (uint32_t) > xdrs->x_addr_end)
return (FALSE);
*((uint32_t *)xdrs->x_addr) = cpu_to_be32(val);
xdrs->x_addr += sizeof (uint32_t);
return (TRUE);
}
static bool_t
xdrmem_dec_uint32(XDR *xdrs, uint32_t *val)
{
if (xdrs->x_addr + sizeof (uint32_t) > xdrs->x_addr_end)
return (FALSE);
*val = be32_to_cpu(*((uint32_t *)xdrs->x_addr));
xdrs->x_addr += sizeof (uint32_t);
return (TRUE);
}
static bool_t
xdrmem_enc_char(XDR *xdrs, char *cp)
{
uint32_t val;
BUILD_BUG_ON(sizeof (char) != 1);
val = *((unsigned char *) cp);
return (xdrmem_enc_uint32(xdrs, val));
}
static bool_t
xdrmem_dec_char(XDR *xdrs, char *cp)
{
uint32_t val;
BUILD_BUG_ON(sizeof (char) != 1);
if (!xdrmem_dec_uint32(xdrs, &val))
return (FALSE);
/*
* If any of the 3 other bytes are non-zero then val will be greater
* than 0xff and we fail because according to the RFC, this block does
* not have a char encoded in it.
*/
if (val > 0xff)
return (FALSE);
*((unsigned char *) cp) = val;
return (TRUE);
}
static bool_t
xdrmem_enc_ushort(XDR *xdrs, unsigned short *usp)
{
BUILD_BUG_ON(sizeof (unsigned short) != 2);
return (xdrmem_enc_uint32(xdrs, *usp));
}
static bool_t
xdrmem_dec_ushort(XDR *xdrs, unsigned short *usp)
{
uint32_t val;
BUILD_BUG_ON(sizeof (unsigned short) != 2);
if (!xdrmem_dec_uint32(xdrs, &val))
return (FALSE);
/*
* Short ints are not in the RFC, but we assume similar logic as in
* xdrmem_dec_char().
*/
if (val > 0xffff)
return (FALSE);
*usp = val;
return (TRUE);
}
static bool_t
xdrmem_enc_uint(XDR *xdrs, unsigned *up)
{
BUILD_BUG_ON(sizeof (unsigned) != 4);
return (xdrmem_enc_uint32(xdrs, *up));
}
static bool_t
xdrmem_dec_uint(XDR *xdrs, unsigned *up)
{
BUILD_BUG_ON(sizeof (unsigned) != 4);
return (xdrmem_dec_uint32(xdrs, (uint32_t *)up));
}
static bool_t
xdrmem_enc_ulonglong(XDR *xdrs, u_longlong_t *ullp)
{
BUILD_BUG_ON(sizeof (u_longlong_t) != 8);
if (!xdrmem_enc_uint32(xdrs, *ullp >> 32))
return (FALSE);
return (xdrmem_enc_uint32(xdrs, *ullp & 0xffffffff));
}
static bool_t
xdrmem_dec_ulonglong(XDR *xdrs, u_longlong_t *ullp)
{
uint32_t low, high;
BUILD_BUG_ON(sizeof (u_longlong_t) != 8);
if (!xdrmem_dec_uint32(xdrs, &high))
return (FALSE);
if (!xdrmem_dec_uint32(xdrs, &low))
return (FALSE);
*ullp = ((u_longlong_t)high << 32) | low;
return (TRUE);
}
static bool_t
xdr_enc_array(XDR *xdrs, caddr_t *arrp, uint_t *sizep, const uint_t maxsize,
const uint_t elsize, const xdrproc_t elproc)
{
uint_t i;
caddr_t addr = *arrp;
if (*sizep > maxsize || *sizep > UINT_MAX / elsize)
return (FALSE);
if (!xdrmem_enc_uint(xdrs, sizep))
return (FALSE);
for (i = 0; i < *sizep; i++) {
if (!elproc(xdrs, addr))
return (FALSE);
addr += elsize;
}
return (TRUE);
}
static bool_t
xdr_dec_array(XDR *xdrs, caddr_t *arrp, uint_t *sizep, const uint_t maxsize,
const uint_t elsize, const xdrproc_t elproc)
{
uint_t i, size;
bool_t alloc = FALSE;
caddr_t addr;
if (!xdrmem_dec_uint(xdrs, sizep))
return (FALSE);
size = *sizep;
if (size > maxsize || size > UINT_MAX / elsize)
return (FALSE);
/*
* The Solaris man page says: "If *arrp is NULL when decoding,
* xdr_array() allocates memory and *arrp points to it".
*/
if (*arrp == NULL) {
BUILD_BUG_ON(sizeof (uint_t) > sizeof (size_t));
*arrp = kmem_alloc(size * elsize, KM_NOSLEEP);
if (*arrp == NULL)
return (FALSE);
alloc = TRUE;
}
addr = *arrp;
for (i = 0; i < size; i++) {
if (!elproc(xdrs, addr)) {
if (alloc)
kmem_free(*arrp, size * elsize);
return (FALSE);
}
addr += elsize;
}
return (TRUE);
}
static bool_t
xdr_enc_string(XDR *xdrs, char **sp, const uint_t maxsize)
{
size_t slen = strlen(*sp);
uint_t len;
if (slen > maxsize)
return (FALSE);
len = slen;
if (!xdrmem_enc_uint(xdrs, &len))
return (FALSE);
return (xdrmem_enc_bytes(xdrs, *sp, len));
}
static bool_t
xdr_dec_string(XDR *xdrs, char **sp, const uint_t maxsize)
{
uint_t size;
bool_t alloc = FALSE;
if (!xdrmem_dec_uint(xdrs, &size))
return (FALSE);
if (size > maxsize || size > UINT_MAX - 1)
return (FALSE);
/*
* Solaris man page: "If *sp is NULL when decoding, xdr_string()
* allocates memory and *sp points to it".
*/
if (*sp == NULL) {
BUILD_BUG_ON(sizeof (uint_t) > sizeof (size_t));
*sp = kmem_alloc(size + 1, KM_NOSLEEP);
if (*sp == NULL)
return (FALSE);
alloc = TRUE;
}
if (!xdrmem_dec_bytes(xdrs, *sp, size))
goto fail;
if (memchr(*sp, 0, size) != NULL)
goto fail;
(*sp)[size] = '\0';
return (TRUE);
fail:
if (alloc)
kmem_free(*sp, size + 1);
return (FALSE);
}
-static struct xdr_ops xdrmem_encode_ops = {
+static const struct xdr_ops xdrmem_encode_ops = {
.xdr_control = xdrmem_control,
.xdr_char = xdrmem_enc_char,
.xdr_u_short = xdrmem_enc_ushort,
.xdr_u_int = xdrmem_enc_uint,
.xdr_u_longlong_t = xdrmem_enc_ulonglong,
.xdr_opaque = xdrmem_enc_bytes,
.xdr_string = xdr_enc_string,
.xdr_array = xdr_enc_array
};
-static struct xdr_ops xdrmem_decode_ops = {
+static const struct xdr_ops xdrmem_decode_ops = {
.xdr_control = xdrmem_control,
.xdr_char = xdrmem_dec_char,
.xdr_u_short = xdrmem_dec_ushort,
.xdr_u_int = xdrmem_dec_uint,
.xdr_u_longlong_t = xdrmem_dec_ulonglong,
.xdr_opaque = xdrmem_dec_bytes,
.xdr_string = xdr_dec_string,
.xdr_array = xdr_dec_array
};
diff --git a/module/os/linux/zfs/abd_os.c b/module/os/linux/zfs/abd_os.c
index 372c1dadc3b7..113aee58599f 100644
--- a/module/os/linux/zfs/abd_os.c
+++ b/module/os/linux/zfs/abd_os.c
@@ -1,1149 +1,1149 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2014 by Chunwei Chen. All rights reserved.
* Copyright (c) 2019 by Delphix. All rights reserved.
*/
/*
* See abd.c for a general overview of the arc buffered data (ABD).
*
* Linear buffers act exactly like normal buffers and are always mapped into the
* kernel's virtual memory space, while scattered ABD data chunks are allocated
* as physical pages and then mapped in only while they are actually being
* accessed through one of the abd_* library functions. Using scattered ABDs
* provides several benefits:
*
* (1) They avoid use of kmem_*, preventing performance problems where running
* kmem_reap on very large memory systems never finishes and causes
* constant TLB shootdowns.
*
* (2) Fragmentation is less of an issue since when we are at the limit of
* allocatable space, we won't have to search around for a long free
* hole in the VA space for large ARC allocations. Each chunk is mapped in
* individually, so even if we are using HIGHMEM (see next point) we
* wouldn't need to worry about finding a contiguous address range.
*
* (3) If we are not using HIGHMEM, then all physical memory is always
* mapped into the kernel's address space, so we also avoid the map /
* unmap costs on each ABD access.
*
* If we are not using HIGHMEM, scattered buffers which have only one chunk
* can be treated as linear buffers, because they are contiguous in the
* kernel's virtual address space. See abd_alloc_chunks() for details.
*/
#include <sys/abd_impl.h>
#include <sys/param.h>
#include <sys/zio.h>
#include <sys/arc.h>
#include <sys/zfs_context.h>
#include <sys/zfs_znode.h>
#ifdef _KERNEL
#include <linux/kmap_compat.h>
#include <linux/scatterlist.h>
#else
#define MAX_ORDER 1
#endif
typedef struct abd_stats {
kstat_named_t abdstat_struct_size;
kstat_named_t abdstat_linear_cnt;
kstat_named_t abdstat_linear_data_size;
kstat_named_t abdstat_scatter_cnt;
kstat_named_t abdstat_scatter_data_size;
kstat_named_t abdstat_scatter_chunk_waste;
kstat_named_t abdstat_scatter_orders[MAX_ORDER];
kstat_named_t abdstat_scatter_page_multi_chunk;
kstat_named_t abdstat_scatter_page_multi_zone;
kstat_named_t abdstat_scatter_page_alloc_retry;
kstat_named_t abdstat_scatter_sg_table_retry;
} abd_stats_t;
static abd_stats_t abd_stats = {
/* Amount of memory occupied by all of the abd_t struct allocations */
{ "struct_size", KSTAT_DATA_UINT64 },
/*
* The number of linear ABDs which are currently allocated, excluding
* ABDs which don't own their data (for instance the ones which were
* allocated through abd_get_offset() and abd_get_from_buf()). If an
* ABD takes ownership of its buf then it will become tracked.
*/
{ "linear_cnt", KSTAT_DATA_UINT64 },
/* Amount of data stored in all linear ABDs tracked by linear_cnt */
{ "linear_data_size", KSTAT_DATA_UINT64 },
/*
* The number of scatter ABDs which are currently allocated, excluding
* ABDs which don't own their data (for instance the ones which were
* allocated through abd_get_offset()).
*/
{ "scatter_cnt", KSTAT_DATA_UINT64 },
/* Amount of data stored in all scatter ABDs tracked by scatter_cnt */
{ "scatter_data_size", KSTAT_DATA_UINT64 },
/*
* The amount of space wasted at the end of the last chunk across all
* scatter ABDs tracked by scatter_cnt.
*/
{ "scatter_chunk_waste", KSTAT_DATA_UINT64 },
/*
* The number of compound allocations of a given order. These
* allocations are spread over all currently allocated ABDs, and
* act as a measure of memory fragmentation.
*/
{ { "scatter_order_N", KSTAT_DATA_UINT64 } },
/*
* The number of scatter ABDs which contain multiple chunks.
* ABDs are preferentially allocated from the minimum number of
* contiguous multi-page chunks, a single chunk is optimal.
*/
{ "scatter_page_multi_chunk", KSTAT_DATA_UINT64 },
/*
* The number of scatter ABDs which are split across memory zones.
* ABDs are preferentially allocated using pages from a single zone.
*/
{ "scatter_page_multi_zone", KSTAT_DATA_UINT64 },
/*
* The total number of retries encountered when attempting to
* allocate the pages to populate the scatter ABD.
*/
{ "scatter_page_alloc_retry", KSTAT_DATA_UINT64 },
/*
* The total number of retries encountered when attempting to
* allocate the sg table for an ABD.
*/
{ "scatter_sg_table_retry", KSTAT_DATA_UINT64 },
};
struct {
wmsum_t abdstat_struct_size;
wmsum_t abdstat_linear_cnt;
wmsum_t abdstat_linear_data_size;
wmsum_t abdstat_scatter_cnt;
wmsum_t abdstat_scatter_data_size;
wmsum_t abdstat_scatter_chunk_waste;
wmsum_t abdstat_scatter_orders[MAX_ORDER];
wmsum_t abdstat_scatter_page_multi_chunk;
wmsum_t abdstat_scatter_page_multi_zone;
wmsum_t abdstat_scatter_page_alloc_retry;
wmsum_t abdstat_scatter_sg_table_retry;
} abd_sums;
#define abd_for_each_sg(abd, sg, n, i) \
for_each_sg(ABD_SCATTER(abd).abd_sgl, sg, n, i)
-unsigned zfs_abd_scatter_max_order = MAX_ORDER - 1;
-
/*
* zfs_abd_scatter_min_size is the minimum allocation size to use scatter
* ABD's. Smaller allocations will use linear ABD's which uses
* zio_[data_]buf_alloc().
*
* Scatter ABD's use at least one page each, so sub-page allocations waste
* some space when allocated as scatter (e.g. 2KB scatter allocation wastes
* half of each page). Using linear ABD's for small allocations means that
* they will be put on slabs which contain many allocations. This can
* improve memory efficiency, but it also makes it much harder for ARC
* evictions to actually free pages, because all the buffers on one slab need
* to be freed in order for the slab (and underlying pages) to be freed.
* Typically, 512B and 1KB kmem caches have 16 buffers per slab, so it's
* possible for them to actually waste more memory than scatter (one page per
* buf = wasting 3/4 or 7/8th; one buf per slab = wasting 15/16th).
*
* Spill blocks are typically 512B and are heavily used on systems running
* selinux with the default dnode size and the `xattr=sa` property set.
*
* By default we use linear allocations for 512B and 1KB, and scatter
* allocations for larger (1.5KB and up).
*/
-int zfs_abd_scatter_min_size = 512 * 3;
+static int zfs_abd_scatter_min_size = 512 * 3;
/*
* We use a scattered SPA_MAXBLOCKSIZE sized ABD whose pages are
* just a single zero'd page. This allows us to conserve memory by
* only using a single zero page for the scatterlist.
*/
abd_t *abd_zero_scatter = NULL;
struct page;
/*
* abd_zero_page we will be an allocated zero'd PAGESIZE buffer, which is
* assigned to set each of the pages of abd_zero_scatter.
*/
static struct page *abd_zero_page = NULL;
static kmem_cache_t *abd_cache = NULL;
static kstat_t *abd_ksp;
static uint_t
abd_chunkcnt_for_bytes(size_t size)
{
return (P2ROUNDUP(size, PAGESIZE) / PAGESIZE);
}
abd_t *
abd_alloc_struct_impl(size_t size)
{
/*
* In Linux we do not use the size passed in during ABD
* allocation, so we just ignore it.
*/
(void) size;
abd_t *abd = kmem_cache_alloc(abd_cache, KM_PUSHPAGE);
ASSERT3P(abd, !=, NULL);
ABDSTAT_INCR(abdstat_struct_size, sizeof (abd_t));
return (abd);
}
void
abd_free_struct_impl(abd_t *abd)
{
kmem_cache_free(abd_cache, abd);
ABDSTAT_INCR(abdstat_struct_size, -(int)sizeof (abd_t));
}
#ifdef _KERNEL
+static unsigned zfs_abd_scatter_max_order = MAX_ORDER - 1;
+
/*
* Mark zfs data pages so they can be excluded from kernel crash dumps
*/
#ifdef _LP64
#define ABD_FILE_CACHE_PAGE 0x2F5ABDF11ECAC4E
static inline void
abd_mark_zfs_page(struct page *page)
{
get_page(page);
SetPagePrivate(page);
set_page_private(page, ABD_FILE_CACHE_PAGE);
}
static inline void
abd_unmark_zfs_page(struct page *page)
{
set_page_private(page, 0UL);
ClearPagePrivate(page);
put_page(page);
}
#else
#define abd_mark_zfs_page(page)
#define abd_unmark_zfs_page(page)
#endif /* _LP64 */
#ifndef CONFIG_HIGHMEM
#ifndef __GFP_RECLAIM
#define __GFP_RECLAIM __GFP_WAIT
#endif
/*
* The goal is to minimize fragmentation by preferentially populating ABDs
* with higher order compound pages from a single zone. Allocation size is
* progressively decreased until it can be satisfied without performing
* reclaim or compaction. When necessary this function will degenerate to
* allocating individual pages and allowing reclaim to satisfy allocations.
*/
void
abd_alloc_chunks(abd_t *abd, size_t size)
{
struct list_head pages;
struct sg_table table;
struct scatterlist *sg;
struct page *page, *tmp_page = NULL;
gfp_t gfp = __GFP_NOWARN | GFP_NOIO;
gfp_t gfp_comp = (gfp | __GFP_NORETRY | __GFP_COMP) & ~__GFP_RECLAIM;
int max_order = MIN(zfs_abd_scatter_max_order, MAX_ORDER - 1);
int nr_pages = abd_chunkcnt_for_bytes(size);
int chunks = 0, zones = 0;
size_t remaining_size;
int nid = NUMA_NO_NODE;
int alloc_pages = 0;
INIT_LIST_HEAD(&pages);
while (alloc_pages < nr_pages) {
unsigned chunk_pages;
int order;
order = MIN(highbit64(nr_pages - alloc_pages) - 1, max_order);
chunk_pages = (1U << order);
page = alloc_pages_node(nid, order ? gfp_comp : gfp, order);
if (page == NULL) {
if (order == 0) {
ABDSTAT_BUMP(abdstat_scatter_page_alloc_retry);
schedule_timeout_interruptible(1);
} else {
max_order = MAX(0, order - 1);
}
continue;
}
list_add_tail(&page->lru, &pages);
if ((nid != NUMA_NO_NODE) && (page_to_nid(page) != nid))
zones++;
nid = page_to_nid(page);
ABDSTAT_BUMP(abdstat_scatter_orders[order]);
chunks++;
alloc_pages += chunk_pages;
}
ASSERT3S(alloc_pages, ==, nr_pages);
while (sg_alloc_table(&table, chunks, gfp)) {
ABDSTAT_BUMP(abdstat_scatter_sg_table_retry);
schedule_timeout_interruptible(1);
}
sg = table.sgl;
remaining_size = size;
list_for_each_entry_safe(page, tmp_page, &pages, lru) {
size_t sg_size = MIN(PAGESIZE << compound_order(page),
remaining_size);
sg_set_page(sg, page, sg_size, 0);
abd_mark_zfs_page(page);
remaining_size -= sg_size;
sg = sg_next(sg);
list_del(&page->lru);
}
/*
* These conditions ensure that a possible transformation to a linear
* ABD would be valid.
*/
ASSERT(!PageHighMem(sg_page(table.sgl)));
ASSERT0(ABD_SCATTER(abd).abd_offset);
if (table.nents == 1) {
/*
* Since there is only one entry, this ABD can be represented
* as a linear buffer. All single-page (4K) ABD's can be
* represented this way. Some multi-page ABD's can also be
* represented this way, if we were able to allocate a single
* "chunk" (higher-order "page" which represents a power-of-2
* series of physically-contiguous pages). This is often the
* case for 2-page (8K) ABD's.
*
* Representing a single-entry scatter ABD as a linear ABD
* has the performance advantage of avoiding the copy (and
* allocation) in abd_borrow_buf_copy / abd_return_buf_copy.
* A performance increase of around 5% has been observed for
* ARC-cached reads (of small blocks which can take advantage
* of this).
*
* Note that this optimization is only possible because the
* pages are always mapped into the kernel's address space.
* This is not the case for highmem pages, so the
* optimization can not be made there.
*/
abd->abd_flags |= ABD_FLAG_LINEAR;
abd->abd_flags |= ABD_FLAG_LINEAR_PAGE;
abd->abd_u.abd_linear.abd_sgl = table.sgl;
ABD_LINEAR_BUF(abd) = page_address(sg_page(table.sgl));
} else if (table.nents > 1) {
ABDSTAT_BUMP(abdstat_scatter_page_multi_chunk);
abd->abd_flags |= ABD_FLAG_MULTI_CHUNK;
if (zones) {
ABDSTAT_BUMP(abdstat_scatter_page_multi_zone);
abd->abd_flags |= ABD_FLAG_MULTI_ZONE;
}
ABD_SCATTER(abd).abd_sgl = table.sgl;
ABD_SCATTER(abd).abd_nents = table.nents;
}
}
#else
/*
* Allocate N individual pages to construct a scatter ABD. This function
* makes no attempt to request contiguous pages and requires the minimal
* number of kernel interfaces. It's designed for maximum compatibility.
*/
void
abd_alloc_chunks(abd_t *abd, size_t size)
{
struct scatterlist *sg = NULL;
struct sg_table table;
struct page *page;
gfp_t gfp = __GFP_NOWARN | GFP_NOIO;
int nr_pages = abd_chunkcnt_for_bytes(size);
int i = 0;
while (sg_alloc_table(&table, nr_pages, gfp)) {
ABDSTAT_BUMP(abdstat_scatter_sg_table_retry);
schedule_timeout_interruptible(1);
}
ASSERT3U(table.nents, ==, nr_pages);
ABD_SCATTER(abd).abd_sgl = table.sgl;
ABD_SCATTER(abd).abd_nents = nr_pages;
abd_for_each_sg(abd, sg, nr_pages, i) {
while ((page = __page_cache_alloc(gfp)) == NULL) {
ABDSTAT_BUMP(abdstat_scatter_page_alloc_retry);
schedule_timeout_interruptible(1);
}
ABDSTAT_BUMP(abdstat_scatter_orders[0]);
sg_set_page(sg, page, PAGESIZE, 0);
abd_mark_zfs_page(page);
}
if (nr_pages > 1) {
ABDSTAT_BUMP(abdstat_scatter_page_multi_chunk);
abd->abd_flags |= ABD_FLAG_MULTI_CHUNK;
}
}
#endif /* !CONFIG_HIGHMEM */
/*
* This must be called if any of the sg_table allocation functions
* are called.
*/
static void
abd_free_sg_table(abd_t *abd)
{
struct sg_table table;
table.sgl = ABD_SCATTER(abd).abd_sgl;
table.nents = table.orig_nents = ABD_SCATTER(abd).abd_nents;
sg_free_table(&table);
}
void
abd_free_chunks(abd_t *abd)
{
struct scatterlist *sg = NULL;
struct page *page;
int nr_pages = ABD_SCATTER(abd).abd_nents;
int order, i = 0;
if (abd->abd_flags & ABD_FLAG_MULTI_ZONE)
ABDSTAT_BUMPDOWN(abdstat_scatter_page_multi_zone);
if (abd->abd_flags & ABD_FLAG_MULTI_CHUNK)
ABDSTAT_BUMPDOWN(abdstat_scatter_page_multi_chunk);
abd_for_each_sg(abd, sg, nr_pages, i) {
page = sg_page(sg);
abd_unmark_zfs_page(page);
order = compound_order(page);
__free_pages(page, order);
ASSERT3U(sg->length, <=, PAGE_SIZE << order);
ABDSTAT_BUMPDOWN(abdstat_scatter_orders[order]);
}
abd_free_sg_table(abd);
}
/*
* Allocate scatter ABD of size SPA_MAXBLOCKSIZE, where each page in
* the scatterlist will be set to the zero'd out buffer abd_zero_page.
*/
static void
abd_alloc_zero_scatter(void)
{
struct scatterlist *sg = NULL;
struct sg_table table;
gfp_t gfp = __GFP_NOWARN | GFP_NOIO;
gfp_t gfp_zero_page = gfp | __GFP_ZERO;
int nr_pages = abd_chunkcnt_for_bytes(SPA_MAXBLOCKSIZE);
int i = 0;
while ((abd_zero_page = __page_cache_alloc(gfp_zero_page)) == NULL) {
ABDSTAT_BUMP(abdstat_scatter_page_alloc_retry);
schedule_timeout_interruptible(1);
}
abd_mark_zfs_page(abd_zero_page);
while (sg_alloc_table(&table, nr_pages, gfp)) {
ABDSTAT_BUMP(abdstat_scatter_sg_table_retry);
schedule_timeout_interruptible(1);
}
ASSERT3U(table.nents, ==, nr_pages);
abd_zero_scatter = abd_alloc_struct(SPA_MAXBLOCKSIZE);
abd_zero_scatter->abd_flags |= ABD_FLAG_OWNER;
ABD_SCATTER(abd_zero_scatter).abd_offset = 0;
ABD_SCATTER(abd_zero_scatter).abd_sgl = table.sgl;
ABD_SCATTER(abd_zero_scatter).abd_nents = nr_pages;
abd_zero_scatter->abd_size = SPA_MAXBLOCKSIZE;
abd_zero_scatter->abd_flags |= ABD_FLAG_MULTI_CHUNK | ABD_FLAG_ZEROS;
abd_for_each_sg(abd_zero_scatter, sg, nr_pages, i) {
sg_set_page(sg, abd_zero_page, PAGESIZE, 0);
}
ABDSTAT_BUMP(abdstat_scatter_cnt);
ABDSTAT_INCR(abdstat_scatter_data_size, PAGESIZE);
ABDSTAT_BUMP(abdstat_scatter_page_multi_chunk);
}
#else /* _KERNEL */
#ifndef PAGE_SHIFT
#define PAGE_SHIFT (highbit64(PAGESIZE)-1)
#endif
#define zfs_kmap_atomic(chunk) ((void *)chunk)
#define zfs_kunmap_atomic(addr) do { (void)(addr); } while (0)
#define local_irq_save(flags) do { (void)(flags); } while (0)
#define local_irq_restore(flags) do { (void)(flags); } while (0)
#define nth_page(pg, i) \
((struct page *)((void *)(pg) + (i) * PAGESIZE))
struct scatterlist {
struct page *page;
int length;
int end;
};
static void
sg_init_table(struct scatterlist *sg, int nr)
{
memset(sg, 0, nr * sizeof (struct scatterlist));
sg[nr - 1].end = 1;
}
/*
* This must be called if any of the sg_table allocation functions
* are called.
*/
static void
abd_free_sg_table(abd_t *abd)
{
int nents = ABD_SCATTER(abd).abd_nents;
vmem_free(ABD_SCATTER(abd).abd_sgl,
nents * sizeof (struct scatterlist));
}
#define for_each_sg(sgl, sg, nr, i) \
for ((i) = 0, (sg) = (sgl); (i) < (nr); (i)++, (sg) = sg_next(sg))
static inline void
sg_set_page(struct scatterlist *sg, struct page *page, unsigned int len,
unsigned int offset)
{
/* currently we don't use offset */
ASSERT(offset == 0);
sg->page = page;
sg->length = len;
}
static inline struct page *
sg_page(struct scatterlist *sg)
{
return (sg->page);
}
static inline struct scatterlist *
sg_next(struct scatterlist *sg)
{
if (sg->end)
return (NULL);
return (sg + 1);
}
void
abd_alloc_chunks(abd_t *abd, size_t size)
{
unsigned nr_pages = abd_chunkcnt_for_bytes(size);
struct scatterlist *sg;
int i;
ABD_SCATTER(abd).abd_sgl = vmem_alloc(nr_pages *
sizeof (struct scatterlist), KM_SLEEP);
sg_init_table(ABD_SCATTER(abd).abd_sgl, nr_pages);
abd_for_each_sg(abd, sg, nr_pages, i) {
struct page *p = umem_alloc_aligned(PAGESIZE, 64, KM_SLEEP);
sg_set_page(sg, p, PAGESIZE, 0);
}
ABD_SCATTER(abd).abd_nents = nr_pages;
}
void
abd_free_chunks(abd_t *abd)
{
int i, n = ABD_SCATTER(abd).abd_nents;
struct scatterlist *sg;
abd_for_each_sg(abd, sg, n, i) {
for (int j = 0; j < sg->length; j += PAGESIZE) {
struct page *p = nth_page(sg_page(sg), j >> PAGE_SHIFT);
umem_free(p, PAGESIZE);
}
}
abd_free_sg_table(abd);
}
static void
abd_alloc_zero_scatter(void)
{
unsigned nr_pages = abd_chunkcnt_for_bytes(SPA_MAXBLOCKSIZE);
struct scatterlist *sg;
int i;
abd_zero_page = umem_alloc_aligned(PAGESIZE, 64, KM_SLEEP);
memset(abd_zero_page, 0, PAGESIZE);
abd_zero_scatter = abd_alloc_struct(SPA_MAXBLOCKSIZE);
abd_zero_scatter->abd_flags |= ABD_FLAG_OWNER;
abd_zero_scatter->abd_flags |= ABD_FLAG_MULTI_CHUNK | ABD_FLAG_ZEROS;
ABD_SCATTER(abd_zero_scatter).abd_offset = 0;
ABD_SCATTER(abd_zero_scatter).abd_nents = nr_pages;
abd_zero_scatter->abd_size = SPA_MAXBLOCKSIZE;
zfs_refcount_create(&abd_zero_scatter->abd_children);
ABD_SCATTER(abd_zero_scatter).abd_sgl = vmem_alloc(nr_pages *
sizeof (struct scatterlist), KM_SLEEP);
sg_init_table(ABD_SCATTER(abd_zero_scatter).abd_sgl, nr_pages);
abd_for_each_sg(abd_zero_scatter, sg, nr_pages, i) {
sg_set_page(sg, abd_zero_page, PAGESIZE, 0);
}
ABDSTAT_BUMP(abdstat_scatter_cnt);
ABDSTAT_INCR(abdstat_scatter_data_size, PAGESIZE);
ABDSTAT_BUMP(abdstat_scatter_page_multi_chunk);
}
#endif /* _KERNEL */
boolean_t
abd_size_alloc_linear(size_t size)
{
return (!zfs_abd_scatter_enabled || size < zfs_abd_scatter_min_size);
}
void
abd_update_scatter_stats(abd_t *abd, abd_stats_op_t op)
{
ASSERT(op == ABDSTAT_INCR || op == ABDSTAT_DECR);
int waste = P2ROUNDUP(abd->abd_size, PAGESIZE) - abd->abd_size;
if (op == ABDSTAT_INCR) {
ABDSTAT_BUMP(abdstat_scatter_cnt);
ABDSTAT_INCR(abdstat_scatter_data_size, abd->abd_size);
ABDSTAT_INCR(abdstat_scatter_chunk_waste, waste);
arc_space_consume(waste, ARC_SPACE_ABD_CHUNK_WASTE);
} else {
ABDSTAT_BUMPDOWN(abdstat_scatter_cnt);
ABDSTAT_INCR(abdstat_scatter_data_size, -(int)abd->abd_size);
ABDSTAT_INCR(abdstat_scatter_chunk_waste, -waste);
arc_space_return(waste, ARC_SPACE_ABD_CHUNK_WASTE);
}
}
void
abd_update_linear_stats(abd_t *abd, abd_stats_op_t op)
{
ASSERT(op == ABDSTAT_INCR || op == ABDSTAT_DECR);
if (op == ABDSTAT_INCR) {
ABDSTAT_BUMP(abdstat_linear_cnt);
ABDSTAT_INCR(abdstat_linear_data_size, abd->abd_size);
} else {
ABDSTAT_BUMPDOWN(abdstat_linear_cnt);
ABDSTAT_INCR(abdstat_linear_data_size, -(int)abd->abd_size);
}
}
void
abd_verify_scatter(abd_t *abd)
{
size_t n;
int i = 0;
struct scatterlist *sg = NULL;
ASSERT3U(ABD_SCATTER(abd).abd_nents, >, 0);
ASSERT3U(ABD_SCATTER(abd).abd_offset, <,
ABD_SCATTER(abd).abd_sgl->length);
n = ABD_SCATTER(abd).abd_nents;
abd_for_each_sg(abd, sg, n, i) {
ASSERT3P(sg_page(sg), !=, NULL);
}
}
static void
abd_free_zero_scatter(void)
{
ABDSTAT_BUMPDOWN(abdstat_scatter_cnt);
ABDSTAT_INCR(abdstat_scatter_data_size, -(int)PAGESIZE);
ABDSTAT_BUMPDOWN(abdstat_scatter_page_multi_chunk);
abd_free_sg_table(abd_zero_scatter);
abd_free_struct(abd_zero_scatter);
abd_zero_scatter = NULL;
ASSERT3P(abd_zero_page, !=, NULL);
#if defined(_KERNEL)
abd_unmark_zfs_page(abd_zero_page);
__free_page(abd_zero_page);
#else
umem_free(abd_zero_page, PAGESIZE);
#endif /* _KERNEL */
}
static int
abd_kstats_update(kstat_t *ksp, int rw)
{
abd_stats_t *as = ksp->ks_data;
if (rw == KSTAT_WRITE)
return (EACCES);
as->abdstat_struct_size.value.ui64 =
wmsum_value(&abd_sums.abdstat_struct_size);
as->abdstat_linear_cnt.value.ui64 =
wmsum_value(&abd_sums.abdstat_linear_cnt);
as->abdstat_linear_data_size.value.ui64 =
wmsum_value(&abd_sums.abdstat_linear_data_size);
as->abdstat_scatter_cnt.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_cnt);
as->abdstat_scatter_data_size.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_data_size);
as->abdstat_scatter_chunk_waste.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_chunk_waste);
for (int i = 0; i < MAX_ORDER; i++) {
as->abdstat_scatter_orders[i].value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_orders[i]);
}
as->abdstat_scatter_page_multi_chunk.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_page_multi_chunk);
as->abdstat_scatter_page_multi_zone.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_page_multi_zone);
as->abdstat_scatter_page_alloc_retry.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_page_alloc_retry);
as->abdstat_scatter_sg_table_retry.value.ui64 =
wmsum_value(&abd_sums.abdstat_scatter_sg_table_retry);
return (0);
}
void
abd_init(void)
{
int i;
abd_cache = kmem_cache_create("abd_t", sizeof (abd_t),
0, NULL, NULL, NULL, NULL, NULL, 0);
wmsum_init(&abd_sums.abdstat_struct_size, 0);
wmsum_init(&abd_sums.abdstat_linear_cnt, 0);
wmsum_init(&abd_sums.abdstat_linear_data_size, 0);
wmsum_init(&abd_sums.abdstat_scatter_cnt, 0);
wmsum_init(&abd_sums.abdstat_scatter_data_size, 0);
wmsum_init(&abd_sums.abdstat_scatter_chunk_waste, 0);
for (i = 0; i < MAX_ORDER; i++)
wmsum_init(&abd_sums.abdstat_scatter_orders[i], 0);
wmsum_init(&abd_sums.abdstat_scatter_page_multi_chunk, 0);
wmsum_init(&abd_sums.abdstat_scatter_page_multi_zone, 0);
wmsum_init(&abd_sums.abdstat_scatter_page_alloc_retry, 0);
wmsum_init(&abd_sums.abdstat_scatter_sg_table_retry, 0);
abd_ksp = kstat_create("zfs", 0, "abdstats", "misc", KSTAT_TYPE_NAMED,
sizeof (abd_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
if (abd_ksp != NULL) {
for (i = 0; i < MAX_ORDER; i++) {
snprintf(abd_stats.abdstat_scatter_orders[i].name,
KSTAT_STRLEN, "scatter_order_%d", i);
abd_stats.abdstat_scatter_orders[i].data_type =
KSTAT_DATA_UINT64;
}
abd_ksp->ks_data = &abd_stats;
abd_ksp->ks_update = abd_kstats_update;
kstat_install(abd_ksp);
}
abd_alloc_zero_scatter();
}
void
abd_fini(void)
{
abd_free_zero_scatter();
if (abd_ksp != NULL) {
kstat_delete(abd_ksp);
abd_ksp = NULL;
}
wmsum_fini(&abd_sums.abdstat_struct_size);
wmsum_fini(&abd_sums.abdstat_linear_cnt);
wmsum_fini(&abd_sums.abdstat_linear_data_size);
wmsum_fini(&abd_sums.abdstat_scatter_cnt);
wmsum_fini(&abd_sums.abdstat_scatter_data_size);
wmsum_fini(&abd_sums.abdstat_scatter_chunk_waste);
for (int i = 0; i < MAX_ORDER; i++)
wmsum_fini(&abd_sums.abdstat_scatter_orders[i]);
wmsum_fini(&abd_sums.abdstat_scatter_page_multi_chunk);
wmsum_fini(&abd_sums.abdstat_scatter_page_multi_zone);
wmsum_fini(&abd_sums.abdstat_scatter_page_alloc_retry);
wmsum_fini(&abd_sums.abdstat_scatter_sg_table_retry);
if (abd_cache) {
kmem_cache_destroy(abd_cache);
abd_cache = NULL;
}
}
void
abd_free_linear_page(abd_t *abd)
{
/* Transform it back into a scatter ABD for freeing */
struct scatterlist *sg = abd->abd_u.abd_linear.abd_sgl;
abd->abd_flags &= ~ABD_FLAG_LINEAR;
abd->abd_flags &= ~ABD_FLAG_LINEAR_PAGE;
ABD_SCATTER(abd).abd_nents = 1;
ABD_SCATTER(abd).abd_offset = 0;
ABD_SCATTER(abd).abd_sgl = sg;
abd_free_chunks(abd);
abd_update_scatter_stats(abd, ABDSTAT_DECR);
}
/*
* If we're going to use this ABD for doing I/O using the block layer, the
* consumer of the ABD data doesn't care if it's scattered or not, and we don't
* plan to store this ABD in memory for a long period of time, we should
* allocate the ABD type that requires the least data copying to do the I/O.
*
* On Linux the optimal thing to do would be to use abd_get_offset() and
* construct a new ABD which shares the original pages thereby eliminating
* the copy. But for the moment a new linear ABD is allocated until this
* performance optimization can be implemented.
*/
abd_t *
abd_alloc_for_io(size_t size, boolean_t is_metadata)
{
return (abd_alloc(size, is_metadata));
}
abd_t *
abd_get_offset_scatter(abd_t *abd, abd_t *sabd, size_t off,
size_t size)
{
(void) size;
int i = 0;
struct scatterlist *sg = NULL;
abd_verify(sabd);
ASSERT3U(off, <=, sabd->abd_size);
size_t new_offset = ABD_SCATTER(sabd).abd_offset + off;
if (abd == NULL)
abd = abd_alloc_struct(0);
/*
* Even if this buf is filesystem metadata, we only track that
* if we own the underlying data buffer, which is not true in
* this case. Therefore, we don't ever use ABD_FLAG_META here.
*/
abd_for_each_sg(sabd, sg, ABD_SCATTER(sabd).abd_nents, i) {
if (new_offset < sg->length)
break;
new_offset -= sg->length;
}
ABD_SCATTER(abd).abd_sgl = sg;
ABD_SCATTER(abd).abd_offset = new_offset;
ABD_SCATTER(abd).abd_nents = ABD_SCATTER(sabd).abd_nents - i;
return (abd);
}
/*
* Initialize the abd_iter.
*/
void
abd_iter_init(struct abd_iter *aiter, abd_t *abd)
{
ASSERT(!abd_is_gang(abd));
abd_verify(abd);
aiter->iter_abd = abd;
aiter->iter_mapaddr = NULL;
aiter->iter_mapsize = 0;
aiter->iter_pos = 0;
if (abd_is_linear(abd)) {
aiter->iter_offset = 0;
aiter->iter_sg = NULL;
} else {
aiter->iter_offset = ABD_SCATTER(abd).abd_offset;
aiter->iter_sg = ABD_SCATTER(abd).abd_sgl;
}
}
/*
* This is just a helper function to see if we have exhausted the
* abd_iter and reached the end.
*/
boolean_t
abd_iter_at_end(struct abd_iter *aiter)
{
return (aiter->iter_pos == aiter->iter_abd->abd_size);
}
/*
* Advance the iterator by a certain amount. Cannot be called when a chunk is
* in use. This can be safely called when the aiter has already exhausted, in
* which case this does nothing.
*/
void
abd_iter_advance(struct abd_iter *aiter, size_t amount)
{
ASSERT3P(aiter->iter_mapaddr, ==, NULL);
ASSERT0(aiter->iter_mapsize);
/* There's nothing left to advance to, so do nothing */
if (abd_iter_at_end(aiter))
return;
aiter->iter_pos += amount;
aiter->iter_offset += amount;
if (!abd_is_linear(aiter->iter_abd)) {
while (aiter->iter_offset >= aiter->iter_sg->length) {
aiter->iter_offset -= aiter->iter_sg->length;
aiter->iter_sg = sg_next(aiter->iter_sg);
if (aiter->iter_sg == NULL) {
ASSERT0(aiter->iter_offset);
break;
}
}
}
}
/*
* Map the current chunk into aiter. This can be safely called when the aiter
* has already exhausted, in which case this does nothing.
*/
void
abd_iter_map(struct abd_iter *aiter)
{
void *paddr;
size_t offset = 0;
ASSERT3P(aiter->iter_mapaddr, ==, NULL);
ASSERT0(aiter->iter_mapsize);
/* There's nothing left to iterate over, so do nothing */
if (abd_iter_at_end(aiter))
return;
if (abd_is_linear(aiter->iter_abd)) {
ASSERT3U(aiter->iter_pos, ==, aiter->iter_offset);
offset = aiter->iter_offset;
aiter->iter_mapsize = aiter->iter_abd->abd_size - offset;
paddr = ABD_LINEAR_BUF(aiter->iter_abd);
} else {
offset = aiter->iter_offset;
aiter->iter_mapsize = MIN(aiter->iter_sg->length - offset,
aiter->iter_abd->abd_size - aiter->iter_pos);
paddr = zfs_kmap_atomic(sg_page(aiter->iter_sg));
}
aiter->iter_mapaddr = (char *)paddr + offset;
}
/*
* Unmap the current chunk from aiter. This can be safely called when the aiter
* has already exhausted, in which case this does nothing.
*/
void
abd_iter_unmap(struct abd_iter *aiter)
{
/* There's nothing left to unmap, so do nothing */
if (abd_iter_at_end(aiter))
return;
if (!abd_is_linear(aiter->iter_abd)) {
/* LINTED E_FUNC_SET_NOT_USED */
zfs_kunmap_atomic(aiter->iter_mapaddr - aiter->iter_offset);
}
ASSERT3P(aiter->iter_mapaddr, !=, NULL);
ASSERT3U(aiter->iter_mapsize, >, 0);
aiter->iter_mapaddr = NULL;
aiter->iter_mapsize = 0;
}
void
abd_cache_reap_now(void)
{
}
#if defined(_KERNEL)
/*
* bio_nr_pages for ABD.
* @off is the offset in @abd
*/
unsigned long
abd_nr_pages_off(abd_t *abd, unsigned int size, size_t off)
{
unsigned long pos;
if (abd_is_gang(abd)) {
unsigned long count = 0;
for (abd_t *cabd = abd_gang_get_offset(abd, &off);
cabd != NULL && size != 0;
cabd = list_next(&ABD_GANG(abd).abd_gang_chain, cabd)) {
ASSERT3U(off, <, cabd->abd_size);
int mysize = MIN(size, cabd->abd_size - off);
count += abd_nr_pages_off(cabd, mysize, off);
size -= mysize;
off = 0;
}
return (count);
}
if (abd_is_linear(abd))
pos = (unsigned long)abd_to_buf(abd) + off;
else
pos = ABD_SCATTER(abd).abd_offset + off;
return (((pos + size + PAGESIZE - 1) >> PAGE_SHIFT) -
(pos >> PAGE_SHIFT));
}
static unsigned int
bio_map(struct bio *bio, void *buf_ptr, unsigned int bio_size)
{
unsigned int offset, size, i;
struct page *page;
offset = offset_in_page(buf_ptr);
for (i = 0; i < bio->bi_max_vecs; i++) {
size = PAGE_SIZE - offset;
if (bio_size <= 0)
break;
if (size > bio_size)
size = bio_size;
if (is_vmalloc_addr(buf_ptr))
page = vmalloc_to_page(buf_ptr);
else
page = virt_to_page(buf_ptr);
/*
* Some network related block device uses tcp_sendpage, which
* doesn't behave well when using 0-count page, this is a
* safety net to catch them.
*/
ASSERT3S(page_count(page), >, 0);
if (bio_add_page(bio, page, size, offset) != size)
break;
buf_ptr += size;
bio_size -= size;
offset = 0;
}
return (bio_size);
}
/*
* bio_map for gang ABD.
*/
static unsigned int
abd_gang_bio_map_off(struct bio *bio, abd_t *abd,
unsigned int io_size, size_t off)
{
ASSERT(abd_is_gang(abd));
for (abd_t *cabd = abd_gang_get_offset(abd, &off);
cabd != NULL;
cabd = list_next(&ABD_GANG(abd).abd_gang_chain, cabd)) {
ASSERT3U(off, <, cabd->abd_size);
int size = MIN(io_size, cabd->abd_size - off);
int remainder = abd_bio_map_off(bio, cabd, size, off);
io_size -= (size - remainder);
if (io_size == 0 || remainder > 0)
return (io_size);
off = 0;
}
ASSERT0(io_size);
return (io_size);
}
/*
* bio_map for ABD.
* @off is the offset in @abd
* Remaining IO size is returned
*/
unsigned int
abd_bio_map_off(struct bio *bio, abd_t *abd,
unsigned int io_size, size_t off)
{
struct abd_iter aiter;
ASSERT3U(io_size, <=, abd->abd_size - off);
if (abd_is_linear(abd))
return (bio_map(bio, ((char *)abd_to_buf(abd)) + off, io_size));
ASSERT(!abd_is_linear(abd));
if (abd_is_gang(abd))
return (abd_gang_bio_map_off(bio, abd, io_size, off));
abd_iter_init(&aiter, abd);
abd_iter_advance(&aiter, off);
for (int i = 0; i < bio->bi_max_vecs; i++) {
struct page *pg;
size_t len, sgoff, pgoff;
struct scatterlist *sg;
if (io_size <= 0)
break;
sg = aiter.iter_sg;
sgoff = aiter.iter_offset;
pgoff = sgoff & (PAGESIZE - 1);
len = MIN(io_size, PAGESIZE - pgoff);
ASSERT(len > 0);
pg = nth_page(sg_page(sg), sgoff >> PAGE_SHIFT);
if (bio_add_page(bio, pg, len, pgoff) != len)
break;
io_size -= len;
abd_iter_advance(&aiter, len);
}
return (io_size);
}
/* Tunable Parameters */
module_param(zfs_abd_scatter_enabled, int, 0644);
MODULE_PARM_DESC(zfs_abd_scatter_enabled,
"Toggle whether ABD allocations must be linear.");
module_param(zfs_abd_scatter_min_size, int, 0644);
MODULE_PARM_DESC(zfs_abd_scatter_min_size,
"Minimum size of scatter allocations.");
/* CSTYLED */
module_param(zfs_abd_scatter_max_order, uint, 0644);
MODULE_PARM_DESC(zfs_abd_scatter_max_order,
"Maximum order allocation used for a scatter ABD.");
#endif
diff --git a/module/os/linux/zfs/vdev_disk.c b/module/os/linux/zfs/vdev_disk.c
index 46b459f5c80b..c06ba613b16d 100644
--- a/module/os/linux/zfs/vdev_disk.c
+++ b/module/os/linux/zfs/vdev_disk.c
@@ -1,965 +1,965 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (C) 2008-2010 Lawrence Livermore National Security, LLC.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Rewritten for Linux by Brian Behlendorf <behlendorf1@llnl.gov>.
* LLNL-CODE-403049.
* Copyright (c) 2012, 2019 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa_impl.h>
#include <sys/vdev_disk.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_trim.h>
#include <sys/abd.h>
#include <sys/fs/zfs.h>
#include <sys/zio.h>
#include <linux/blkpg.h>
#include <linux/msdos_fs.h>
#include <linux/vfs_compat.h>
#ifdef HAVE_LINUX_BLK_CGROUP_HEADER
#include <linux/blk-cgroup.h>
#endif
typedef struct vdev_disk {
struct block_device *vd_bdev;
krwlock_t vd_lock;
} vdev_disk_t;
/*
* Unique identifier for the exclusive vdev holder.
*/
static void *zfs_vdev_holder = VDEV_HOLDER;
/*
* Wait up to zfs_vdev_open_timeout_ms milliseconds before determining the
* device is missing. The missing path may be transient since the links
* can be briefly removed and recreated in response to udev events.
*/
static unsigned zfs_vdev_open_timeout_ms = 1000;
/*
* Size of the "reserved" partition, in blocks.
*/
#define EFI_MIN_RESV_SIZE (16 * 1024)
/*
* Virtual device vector for disks.
*/
typedef struct dio_request {
zio_t *dr_zio; /* Parent ZIO */
atomic_t dr_ref; /* References */
int dr_error; /* Bio error */
int dr_bio_count; /* Count of bio's */
struct bio *dr_bio[0]; /* Attached bio's */
} dio_request_t;
static fmode_t
vdev_bdev_mode(spa_mode_t spa_mode)
{
fmode_t mode = 0;
if (spa_mode & SPA_MODE_READ)
mode |= FMODE_READ;
if (spa_mode & SPA_MODE_WRITE)
mode |= FMODE_WRITE;
return (mode);
}
/*
* Returns the usable capacity (in bytes) for the partition or disk.
*/
static uint64_t
bdev_capacity(struct block_device *bdev)
{
return (i_size_read(bdev->bd_inode));
}
#if !defined(HAVE_BDEV_WHOLE)
static inline struct block_device *
bdev_whole(struct block_device *bdev)
{
return (bdev->bd_contains);
}
#endif
/*
* Returns the maximum expansion capacity of the block device (in bytes).
*
* It is possible to expand a vdev when it has been created as a wholedisk
* and the containing block device has increased in capacity. Or when the
* partition containing the pool has been manually increased in size.
*
* This function is only responsible for calculating the potential expansion
* size so it can be reported by 'zpool list'. The efi_use_whole_disk() is
* responsible for verifying the expected partition layout in the wholedisk
* case, and updating the partition table if appropriate. Once the partition
* size has been increased the additional capacity will be visible using
* bdev_capacity().
*
* The returned maximum expansion capacity is always expected to be larger, or
* at the very least equal, to its usable capacity to prevent overestimating
* the pool expandsize.
*/
static uint64_t
bdev_max_capacity(struct block_device *bdev, uint64_t wholedisk)
{
uint64_t psize;
int64_t available;
if (wholedisk && bdev != bdev_whole(bdev)) {
/*
* When reporting maximum expansion capacity for a wholedisk
* deduct any capacity which is expected to be lost due to
* alignment restrictions. Over reporting this value isn't
* harmful and would only result in slightly less capacity
* than expected post expansion.
* The estimated available space may be slightly smaller than
* bdev_capacity() for devices where the number of sectors is
* not a multiple of the alignment size and the partition layout
* is keeping less than PARTITION_END_ALIGNMENT bytes after the
* "reserved" EFI partition: in such cases return the device
* usable capacity.
*/
available = i_size_read(bdev_whole(bdev)->bd_inode) -
((EFI_MIN_RESV_SIZE + NEW_START_BLOCK +
PARTITION_END_ALIGNMENT) << SECTOR_BITS);
psize = MAX(available, bdev_capacity(bdev));
} else {
psize = bdev_capacity(bdev);
}
return (psize);
}
static void
vdev_disk_error(zio_t *zio)
{
/*
* This function can be called in interrupt context, for instance while
* handling IRQs coming from a misbehaving disk device; use printk()
* which is safe from any context.
*/
printk(KERN_WARNING "zio pool=%s vdev=%s error=%d type=%d "
"offset=%llu size=%llu flags=%x\n", spa_name(zio->io_spa),
zio->io_vd->vdev_path, zio->io_error, zio->io_type,
(u_longlong_t)zio->io_offset, (u_longlong_t)zio->io_size,
zio->io_flags);
}
static int
vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *max_psize,
uint64_t *logical_ashift, uint64_t *physical_ashift)
{
struct block_device *bdev;
fmode_t mode = vdev_bdev_mode(spa_mode(v->vdev_spa));
hrtime_t timeout = MSEC2NSEC(zfs_vdev_open_timeout_ms);
vdev_disk_t *vd;
/* Must have a pathname and it must be absolute. */
if (v->vdev_path == NULL || v->vdev_path[0] != '/') {
v->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
vdev_dbgmsg(v, "invalid vdev_path");
return (SET_ERROR(EINVAL));
}
/*
* Reopen the device if it is currently open. When expanding a
* partition force re-scanning the partition table if userland
* did not take care of this already. We need to do this while closed
* in order to get an accurate updated block device size. Then
* since udev may need to recreate the device links increase the
* open retry timeout before reporting the device as unavailable.
*/
vd = v->vdev_tsd;
if (vd) {
char disk_name[BDEVNAME_SIZE + 6] = "/dev/";
boolean_t reread_part = B_FALSE;
rw_enter(&vd->vd_lock, RW_WRITER);
bdev = vd->vd_bdev;
vd->vd_bdev = NULL;
if (bdev) {
if (v->vdev_expanding && bdev != bdev_whole(bdev)) {
bdevname(bdev_whole(bdev), disk_name + 5);
/*
* If userland has BLKPG_RESIZE_PARTITION,
* then it should have updated the partition
* table already. We can detect this by
* comparing our current physical size
* with that of the device. If they are
* the same, then we must not have
* BLKPG_RESIZE_PARTITION or it failed to
* update the partition table online. We
* fallback to rescanning the partition
* table from the kernel below. However,
* if the capacity already reflects the
* updated partition, then we skip
* rescanning the partition table here.
*/
if (v->vdev_psize == bdev_capacity(bdev))
reread_part = B_TRUE;
}
blkdev_put(bdev, mode | FMODE_EXCL);
}
if (reread_part) {
bdev = blkdev_get_by_path(disk_name, mode | FMODE_EXCL,
zfs_vdev_holder);
if (!IS_ERR(bdev)) {
int error = vdev_bdev_reread_part(bdev);
blkdev_put(bdev, mode | FMODE_EXCL);
if (error == 0) {
timeout = MSEC2NSEC(
zfs_vdev_open_timeout_ms * 2);
}
}
}
} else {
vd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP);
rw_init(&vd->vd_lock, NULL, RW_DEFAULT, NULL);
rw_enter(&vd->vd_lock, RW_WRITER);
}
/*
* Devices are always opened by the path provided at configuration
* time. This means that if the provided path is a udev by-id path
* then drives may be re-cabled without an issue. If the provided
* path is a udev by-path path, then the physical location information
* will be preserved. This can be critical for more complicated
* configurations where drives are located in specific physical
* locations to maximize the systems tolerance to component failure.
*
* Alternatively, you can provide your own udev rule to flexibly map
* the drives as you see fit. It is not advised that you use the
* /dev/[hd]d devices which may be reordered due to probing order.
* Devices in the wrong locations will be detected by the higher
* level vdev validation.
*
* The specified paths may be briefly removed and recreated in
* response to udev events. This should be exceptionally unlikely
* because the zpool command makes every effort to verify these paths
* have already settled prior to reaching this point. Therefore,
* a ENOENT failure at this point is highly likely to be transient
* and it is reasonable to sleep and retry before giving up. In
* practice delays have been observed to be on the order of 100ms.
*
* When ERESTARTSYS is returned it indicates the block device is
* a zvol which could not be opened due to the deadlock detection
* logic in zvol_open(). Extend the timeout and retry the open
* subsequent attempts are expected to eventually succeed.
*/
hrtime_t start = gethrtime();
bdev = ERR_PTR(-ENXIO);
while (IS_ERR(bdev) && ((gethrtime() - start) < timeout)) {
bdev = blkdev_get_by_path(v->vdev_path, mode | FMODE_EXCL,
zfs_vdev_holder);
if (unlikely(PTR_ERR(bdev) == -ENOENT)) {
schedule_timeout(MSEC_TO_TICK(10));
} else if (unlikely(PTR_ERR(bdev) == -ERESTARTSYS)) {
timeout = MSEC2NSEC(zfs_vdev_open_timeout_ms * 10);
continue;
} else if (IS_ERR(bdev)) {
break;
}
}
if (IS_ERR(bdev)) {
int error = -PTR_ERR(bdev);
vdev_dbgmsg(v, "open error=%d timeout=%llu/%llu", error,
(u_longlong_t)(gethrtime() - start),
(u_longlong_t)timeout);
vd->vd_bdev = NULL;
v->vdev_tsd = vd;
rw_exit(&vd->vd_lock);
return (SET_ERROR(error));
} else {
vd->vd_bdev = bdev;
v->vdev_tsd = vd;
rw_exit(&vd->vd_lock);
}
struct request_queue *q = bdev_get_queue(vd->vd_bdev);
/* Determine the physical block size */
int physical_block_size = bdev_physical_block_size(vd->vd_bdev);
/* Determine the logical block size */
int logical_block_size = bdev_logical_block_size(vd->vd_bdev);
/* Clear the nowritecache bit, causes vdev_reopen() to try again. */
v->vdev_nowritecache = B_FALSE;
/* Set when device reports it supports TRIM. */
v->vdev_has_trim = !!blk_queue_discard(q);
/* Set when device reports it supports secure TRIM. */
v->vdev_has_securetrim = !!blk_queue_discard_secure(q);
/* Inform the ZIO pipeline that we are non-rotational */
v->vdev_nonrot = blk_queue_nonrot(q);
/* Physical volume size in bytes for the partition */
*psize = bdev_capacity(vd->vd_bdev);
/* Physical volume size in bytes including possible expansion space */
*max_psize = bdev_max_capacity(vd->vd_bdev, v->vdev_wholedisk);
/* Based on the minimum sector size set the block size */
*physical_ashift = highbit64(MAX(physical_block_size,
SPA_MINBLOCKSIZE)) - 1;
*logical_ashift = highbit64(MAX(logical_block_size,
SPA_MINBLOCKSIZE)) - 1;
return (0);
}
static void
vdev_disk_close(vdev_t *v)
{
vdev_disk_t *vd = v->vdev_tsd;
if (v->vdev_reopening || vd == NULL)
return;
if (vd->vd_bdev != NULL) {
blkdev_put(vd->vd_bdev,
vdev_bdev_mode(spa_mode(v->vdev_spa)) | FMODE_EXCL);
}
rw_destroy(&vd->vd_lock);
kmem_free(vd, sizeof (vdev_disk_t));
v->vdev_tsd = NULL;
}
static dio_request_t *
vdev_disk_dio_alloc(int bio_count)
{
dio_request_t *dr = kmem_zalloc(sizeof (dio_request_t) +
sizeof (struct bio *) * bio_count, KM_SLEEP);
atomic_set(&dr->dr_ref, 0);
dr->dr_bio_count = bio_count;
dr->dr_error = 0;
for (int i = 0; i < dr->dr_bio_count; i++)
dr->dr_bio[i] = NULL;
return (dr);
}
static void
vdev_disk_dio_free(dio_request_t *dr)
{
int i;
for (i = 0; i < dr->dr_bio_count; i++)
if (dr->dr_bio[i])
bio_put(dr->dr_bio[i]);
kmem_free(dr, sizeof (dio_request_t) +
sizeof (struct bio *) * dr->dr_bio_count);
}
static void
vdev_disk_dio_get(dio_request_t *dr)
{
atomic_inc(&dr->dr_ref);
}
static int
vdev_disk_dio_put(dio_request_t *dr)
{
int rc = atomic_dec_return(&dr->dr_ref);
/*
* Free the dio_request when the last reference is dropped and
* ensure zio_interpret is called only once with the correct zio
*/
if (rc == 0) {
zio_t *zio = dr->dr_zio;
int error = dr->dr_error;
vdev_disk_dio_free(dr);
if (zio) {
zio->io_error = error;
ASSERT3S(zio->io_error, >=, 0);
if (zio->io_error)
vdev_disk_error(zio);
zio_delay_interrupt(zio);
}
}
return (rc);
}
BIO_END_IO_PROTO(vdev_disk_physio_completion, bio, error)
{
dio_request_t *dr = bio->bi_private;
int rc;
if (dr->dr_error == 0) {
#ifdef HAVE_1ARG_BIO_END_IO_T
dr->dr_error = BIO_END_IO_ERROR(bio);
#else
if (error)
dr->dr_error = -(error);
else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
dr->dr_error = EIO;
#endif
}
/* Drop reference acquired by __vdev_disk_physio */
rc = vdev_disk_dio_put(dr);
}
static inline void
vdev_submit_bio_impl(struct bio *bio)
{
#ifdef HAVE_1ARG_SUBMIT_BIO
(void) submit_bio(bio);
#else
(void) submit_bio(0, bio);
#endif
}
/*
* preempt_schedule_notrace is GPL-only which breaks the ZFS build, so
* replace it with preempt_schedule under the following condition:
*/
#if defined(CONFIG_ARM64) && \
defined(CONFIG_PREEMPTION) && \
defined(CONFIG_BLK_CGROUP)
#define preempt_schedule_notrace(x) preempt_schedule(x)
#endif
#ifdef HAVE_BIO_SET_DEV
#if defined(CONFIG_BLK_CGROUP) && defined(HAVE_BIO_SET_DEV_GPL_ONLY)
/*
* The Linux 5.5 kernel updated percpu_ref_tryget() which is inlined by
* blkg_tryget() to use rcu_read_lock() instead of rcu_read_lock_sched().
* As a side effect the function was converted to GPL-only. Define our
* own version when needed which uses rcu_read_lock_sched().
*/
#if defined(HAVE_BLKG_TRYGET_GPL_ONLY)
static inline bool
vdev_blkg_tryget(struct blkcg_gq *blkg)
{
struct percpu_ref *ref = &blkg->refcnt;
unsigned long __percpu *count;
bool rc;
rcu_read_lock_sched();
if (__ref_is_percpu(ref, &count)) {
this_cpu_inc(*count);
rc = true;
} else {
#ifdef ZFS_PERCPU_REF_COUNT_IN_DATA
rc = atomic_long_inc_not_zero(&ref->data->count);
#else
rc = atomic_long_inc_not_zero(&ref->count);
#endif
}
rcu_read_unlock_sched();
return (rc);
}
#elif defined(HAVE_BLKG_TRYGET)
#define vdev_blkg_tryget(bg) blkg_tryget(bg)
#endif
#ifdef HAVE_BIO_SET_DEV_MACRO
/*
* The Linux 5.0 kernel updated the bio_set_dev() macro so it calls the
* GPL-only bio_associate_blkg() symbol thus inadvertently converting
* the entire macro. Provide a minimal version which always assigns the
* request queue's root_blkg to the bio.
*/
static inline void
vdev_bio_associate_blkg(struct bio *bio)
{
#if defined(HAVE_BIO_BDEV_DISK)
struct request_queue *q = bio->bi_bdev->bd_disk->queue;
#else
struct request_queue *q = bio->bi_disk->queue;
#endif
ASSERT3P(q, !=, NULL);
ASSERT3P(bio->bi_blkg, ==, NULL);
if (q->root_blkg && vdev_blkg_tryget(q->root_blkg))
bio->bi_blkg = q->root_blkg;
}
#define bio_associate_blkg vdev_bio_associate_blkg
#else
static inline void
vdev_bio_set_dev(struct bio *bio, struct block_device *bdev)
{
#if defined(HAVE_BIO_BDEV_DISK)
struct request_queue *q = bdev->bd_disk->queue;
#else
struct request_queue *q = bio->bi_disk->queue;
#endif
bio_clear_flag(bio, BIO_REMAPPED);
if (bio->bi_bdev != bdev)
bio_clear_flag(bio, BIO_THROTTLED);
bio->bi_bdev = bdev;
ASSERT3P(q, !=, NULL);
ASSERT3P(bio->bi_blkg, ==, NULL);
if (q->root_blkg && vdev_blkg_tryget(q->root_blkg))
bio->bi_blkg = q->root_blkg;
}
#define bio_set_dev vdev_bio_set_dev
#endif
#endif
#else
/*
* Provide a bio_set_dev() helper macro for pre-Linux 4.14 kernels.
*/
static inline void
bio_set_dev(struct bio *bio, struct block_device *bdev)
{
bio->bi_bdev = bdev;
}
#endif /* HAVE_BIO_SET_DEV */
static inline void
vdev_submit_bio(struct bio *bio)
{
struct bio_list *bio_list = current->bio_list;
current->bio_list = NULL;
vdev_submit_bio_impl(bio);
current->bio_list = bio_list;
}
static int
__vdev_disk_physio(struct block_device *bdev, zio_t *zio,
size_t io_size, uint64_t io_offset, int rw, int flags)
{
dio_request_t *dr;
uint64_t abd_offset;
uint64_t bio_offset;
int bio_size;
int bio_count = 16;
int error = 0;
struct blk_plug plug;
/*
* Accessing outside the block device is never allowed.
*/
if (io_offset + io_size > bdev->bd_inode->i_size) {
vdev_dbgmsg(zio->io_vd,
"Illegal access %llu size %llu, device size %llu",
(u_longlong_t)io_offset,
(u_longlong_t)io_size,
(u_longlong_t)i_size_read(bdev->bd_inode));
return (SET_ERROR(EIO));
}
retry:
dr = vdev_disk_dio_alloc(bio_count);
if (zio && !(zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD)))
bio_set_flags_failfast(bdev, &flags);
dr->dr_zio = zio;
/*
* Since bio's can have up to BIO_MAX_PAGES=256 iovec's, each of which
* is at least 512 bytes and at most PAGESIZE (typically 4K), one bio
* can cover at least 128KB and at most 1MB. When the required number
* of iovec's exceeds this, we are forced to break the IO in multiple
* bio's and wait for them all to complete. This is likely if the
* recordsize property is increased beyond 1MB. The default
* bio_count=16 should typically accommodate the maximum-size zio of
* 16MB.
*/
abd_offset = 0;
bio_offset = io_offset;
bio_size = io_size;
for (int i = 0; i <= dr->dr_bio_count; i++) {
/* Finished constructing bio's for given buffer */
if (bio_size <= 0)
break;
/*
* If additional bio's are required, we have to retry, but
* this should be rare - see the comment above.
*/
if (dr->dr_bio_count == i) {
vdev_disk_dio_free(dr);
bio_count *= 2;
goto retry;
}
/* bio_alloc() with __GFP_WAIT never returns NULL */
#ifdef HAVE_BIO_MAX_SEGS
dr->dr_bio[i] = bio_alloc(GFP_NOIO, bio_max_segs(
abd_nr_pages_off(zio->io_abd, bio_size, abd_offset)));
#else
dr->dr_bio[i] = bio_alloc(GFP_NOIO,
MIN(abd_nr_pages_off(zio->io_abd, bio_size, abd_offset),
BIO_MAX_PAGES));
#endif
if (unlikely(dr->dr_bio[i] == NULL)) {
vdev_disk_dio_free(dr);
return (SET_ERROR(ENOMEM));
}
/* Matching put called by vdev_disk_physio_completion */
vdev_disk_dio_get(dr);
bio_set_dev(dr->dr_bio[i], bdev);
BIO_BI_SECTOR(dr->dr_bio[i]) = bio_offset >> 9;
dr->dr_bio[i]->bi_end_io = vdev_disk_physio_completion;
dr->dr_bio[i]->bi_private = dr;
bio_set_op_attrs(dr->dr_bio[i], rw, flags);
/* Remaining size is returned to become the new size */
bio_size = abd_bio_map_off(dr->dr_bio[i], zio->io_abd,
bio_size, abd_offset);
/* Advance in buffer and construct another bio if needed */
abd_offset += BIO_BI_SIZE(dr->dr_bio[i]);
bio_offset += BIO_BI_SIZE(dr->dr_bio[i]);
}
/* Extra reference to protect dio_request during vdev_submit_bio */
vdev_disk_dio_get(dr);
if (dr->dr_bio_count > 1)
blk_start_plug(&plug);
/* Submit all bio's associated with this dio */
for (int i = 0; i < dr->dr_bio_count; i++) {
if (dr->dr_bio[i])
vdev_submit_bio(dr->dr_bio[i]);
}
if (dr->dr_bio_count > 1)
blk_finish_plug(&plug);
(void) vdev_disk_dio_put(dr);
return (error);
}
BIO_END_IO_PROTO(vdev_disk_io_flush_completion, bio, error)
{
zio_t *zio = bio->bi_private;
#ifdef HAVE_1ARG_BIO_END_IO_T
zio->io_error = BIO_END_IO_ERROR(bio);
#else
zio->io_error = -error;
#endif
if (zio->io_error && (zio->io_error == EOPNOTSUPP))
zio->io_vd->vdev_nowritecache = B_TRUE;
bio_put(bio);
ASSERT3S(zio->io_error, >=, 0);
if (zio->io_error)
vdev_disk_error(zio);
zio_interrupt(zio);
}
static int
vdev_disk_io_flush(struct block_device *bdev, zio_t *zio)
{
struct request_queue *q;
struct bio *bio;
q = bdev_get_queue(bdev);
if (!q)
return (SET_ERROR(ENXIO));
bio = bio_alloc(GFP_NOIO, 0);
/* bio_alloc() with __GFP_WAIT never returns NULL */
if (unlikely(bio == NULL))
return (SET_ERROR(ENOMEM));
bio->bi_end_io = vdev_disk_io_flush_completion;
bio->bi_private = zio;
bio_set_dev(bio, bdev);
bio_set_flush(bio);
vdev_submit_bio(bio);
invalidate_bdev(bdev);
return (0);
}
static void
vdev_disk_io_start(zio_t *zio)
{
vdev_t *v = zio->io_vd;
vdev_disk_t *vd = v->vdev_tsd;
unsigned long trim_flags = 0;
int rw, error;
/*
* If the vdev is closed, it's likely in the REMOVED or FAULTED state.
* Nothing to be done here but return failure.
*/
if (vd == NULL) {
zio->io_error = ENXIO;
zio_interrupt(zio);
return;
}
rw_enter(&vd->vd_lock, RW_READER);
/*
* If the vdev is closed, it's likely due to a failed reopen and is
* in the UNAVAIL state. Nothing to be done here but return failure.
*/
if (vd->vd_bdev == NULL) {
rw_exit(&vd->vd_lock);
zio->io_error = ENXIO;
zio_interrupt(zio);
return;
}
switch (zio->io_type) {
case ZIO_TYPE_IOCTL:
if (!vdev_readable(v)) {
rw_exit(&vd->vd_lock);
zio->io_error = SET_ERROR(ENXIO);
zio_interrupt(zio);
return;
}
switch (zio->io_cmd) {
case DKIOCFLUSHWRITECACHE:
if (zfs_nocacheflush)
break;
if (v->vdev_nowritecache) {
zio->io_error = SET_ERROR(ENOTSUP);
break;
}
error = vdev_disk_io_flush(vd->vd_bdev, zio);
if (error == 0) {
rw_exit(&vd->vd_lock);
return;
}
zio->io_error = error;
break;
default:
zio->io_error = SET_ERROR(ENOTSUP);
}
rw_exit(&vd->vd_lock);
zio_execute(zio);
return;
case ZIO_TYPE_WRITE:
rw = WRITE;
break;
case ZIO_TYPE_READ:
rw = READ;
break;
case ZIO_TYPE_TRIM:
#if defined(BLKDEV_DISCARD_SECURE)
if (zio->io_trim_flags & ZIO_TRIM_SECURE)
trim_flags |= BLKDEV_DISCARD_SECURE;
#endif
zio->io_error = -blkdev_issue_discard(vd->vd_bdev,
zio->io_offset >> 9, zio->io_size >> 9, GFP_NOFS,
trim_flags);
rw_exit(&vd->vd_lock);
zio_interrupt(zio);
return;
default:
rw_exit(&vd->vd_lock);
zio->io_error = SET_ERROR(ENOTSUP);
zio_interrupt(zio);
return;
}
zio->io_target_timestamp = zio_handle_io_delay(zio);
error = __vdev_disk_physio(vd->vd_bdev, zio,
zio->io_size, zio->io_offset, rw, 0);
rw_exit(&vd->vd_lock);
if (error) {
zio->io_error = error;
zio_interrupt(zio);
return;
}
}
static void
vdev_disk_io_done(zio_t *zio)
{
/*
* If the device returned EIO, we revalidate the media. If it is
* determined the media has changed this triggers the asynchronous
* removal of the device from the configuration.
*/
if (zio->io_error == EIO) {
vdev_t *v = zio->io_vd;
vdev_disk_t *vd = v->vdev_tsd;
if (zfs_check_media_change(vd->vd_bdev)) {
invalidate_bdev(vd->vd_bdev);
v->vdev_remove_wanted = B_TRUE;
spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE);
}
}
}
static void
vdev_disk_hold(vdev_t *vd)
{
ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER));
/* We must have a pathname, and it must be absolute. */
if (vd->vdev_path == NULL || vd->vdev_path[0] != '/')
return;
/*
* Only prefetch path and devid info if the device has
* never been opened.
*/
if (vd->vdev_tsd != NULL)
return;
}
static void
vdev_disk_rele(vdev_t *vd)
{
ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER));
/* XXX: Implement me as a vnode rele for the device */
}
vdev_ops_t vdev_disk_ops = {
.vdev_op_init = NULL,
.vdev_op_fini = NULL,
.vdev_op_open = vdev_disk_open,
.vdev_op_close = vdev_disk_close,
.vdev_op_asize = vdev_default_asize,
.vdev_op_min_asize = vdev_default_min_asize,
.vdev_op_min_alloc = NULL,
.vdev_op_io_start = vdev_disk_io_start,
.vdev_op_io_done = vdev_disk_io_done,
.vdev_op_state_change = NULL,
.vdev_op_need_resilver = NULL,
.vdev_op_hold = vdev_disk_hold,
.vdev_op_rele = vdev_disk_rele,
.vdev_op_remap = NULL,
.vdev_op_xlate = vdev_default_xlate,
.vdev_op_rebuild_asize = NULL,
.vdev_op_metaslab_init = NULL,
.vdev_op_config_generate = NULL,
.vdev_op_nparity = NULL,
.vdev_op_ndisks = NULL,
.vdev_op_type = VDEV_TYPE_DISK, /* name of this vdev type */
.vdev_op_leaf = B_TRUE /* leaf vdev */
};
/*
* The zfs_vdev_scheduler module option has been deprecated. Setting this
* value no longer has any effect. It has not yet been entirely removed
* to allow the module to be loaded if this option is specified in the
* /etc/modprobe.d/zfs.conf file. The following warning will be logged.
*/
static int
param_set_vdev_scheduler(const char *val, zfs_kernel_param_t *kp)
{
int error = param_set_charp(val, kp);
if (error == 0) {
printk(KERN_INFO "The 'zfs_vdev_scheduler' module option "
"is not supported.\n");
}
return (error);
}
-char *zfs_vdev_scheduler = "unused";
+static const char *zfs_vdev_scheduler = "unused";
module_param_call(zfs_vdev_scheduler, param_set_vdev_scheduler,
param_get_charp, &zfs_vdev_scheduler, 0644);
MODULE_PARM_DESC(zfs_vdev_scheduler, "I/O scheduler");
int
param_set_min_auto_ashift(const char *buf, zfs_kernel_param_t *kp)
{
uint64_t val;
int error;
error = kstrtoull(buf, 0, &val);
if (error < 0)
return (SET_ERROR(error));
if (val < ASHIFT_MIN || val > zfs_vdev_max_auto_ashift)
return (SET_ERROR(-EINVAL));
error = param_set_ulong(buf, kp);
if (error < 0)
return (SET_ERROR(error));
return (0);
}
int
param_set_max_auto_ashift(const char *buf, zfs_kernel_param_t *kp)
{
uint64_t val;
int error;
error = kstrtoull(buf, 0, &val);
if (error < 0)
return (SET_ERROR(error));
if (val > ASHIFT_MAX || val < zfs_vdev_min_auto_ashift)
return (SET_ERROR(-EINVAL));
error = param_set_ulong(buf, kp);
if (error < 0)
return (SET_ERROR(error));
return (0);
}
diff --git a/module/os/linux/zfs/vdev_file.c b/module/os/linux/zfs/vdev_file.c
index 98338e604fa8..f073145326e3 100644
--- a/module/os/linux/zfs/vdev_file.c
+++ b/module/os/linux/zfs/vdev_file.c
@@ -1,382 +1,382 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/vdev_file.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_trim.h>
#include <sys/zio.h>
#include <sys/fs/zfs.h>
#include <sys/fm/fs/zfs.h>
#include <sys/abd.h>
#include <sys/fcntl.h>
#include <sys/vnode.h>
#include <sys/zfs_file.h>
#ifdef _KERNEL
#include <linux/falloc.h>
#endif
/*
* Virtual device vector for files.
*/
static taskq_t *vdev_file_taskq;
/*
* By default, the logical/physical ashift for file vdevs is set to
* SPA_MINBLOCKSHIFT (9). This allows all file vdevs to use 512B (1 << 9)
* blocksizes. Users may opt to change one or both of these for testing
* or performance reasons. Care should be taken as these values will
* impact the vdev_ashift setting which can only be set at vdev creation
* time.
*/
-unsigned long vdev_file_logical_ashift = SPA_MINBLOCKSHIFT;
-unsigned long vdev_file_physical_ashift = SPA_MINBLOCKSHIFT;
+static unsigned long vdev_file_logical_ashift = SPA_MINBLOCKSHIFT;
+static unsigned long vdev_file_physical_ashift = SPA_MINBLOCKSHIFT;
static void
vdev_file_hold(vdev_t *vd)
{
ASSERT(vd->vdev_path != NULL);
}
static void
vdev_file_rele(vdev_t *vd)
{
ASSERT(vd->vdev_path != NULL);
}
static mode_t
vdev_file_open_mode(spa_mode_t spa_mode)
{
mode_t mode = 0;
if ((spa_mode & SPA_MODE_READ) && (spa_mode & SPA_MODE_WRITE)) {
mode = O_RDWR;
} else if (spa_mode & SPA_MODE_READ) {
mode = O_RDONLY;
} else if (spa_mode & SPA_MODE_WRITE) {
mode = O_WRONLY;
}
return (mode | O_LARGEFILE);
}
static int
vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
uint64_t *logical_ashift, uint64_t *physical_ashift)
{
vdev_file_t *vf;
zfs_file_t *fp;
zfs_file_attr_t zfa;
int error;
/*
* Rotational optimizations only make sense on block devices.
*/
vd->vdev_nonrot = B_TRUE;
/*
* Allow TRIM on file based vdevs. This may not always be supported,
* since it depends on your kernel version and underlying filesystem
* type but it is always safe to attempt.
*/
vd->vdev_has_trim = B_TRUE;
/*
* Disable secure TRIM on file based vdevs. There is no way to
* request this behavior from the underlying filesystem.
*/
vd->vdev_has_securetrim = B_FALSE;
/*
* We must have a pathname, and it must be absolute.
*/
if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
/*
* Reopen the device if it's not currently open. Otherwise,
* just update the physical size of the device.
*/
if (vd->vdev_tsd != NULL) {
ASSERT(vd->vdev_reopening);
vf = vd->vdev_tsd;
goto skip_open;
}
vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
/*
* We always open the files from the root of the global zone, even if
* we're in a local zone. If the user has gotten to this point, the
* administrator has already decided that the pool should be available
* to local zone users, so the underlying devices should be as well.
*/
ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
error = zfs_file_open(vd->vdev_path,
vdev_file_open_mode(spa_mode(vd->vdev_spa)), 0, &fp);
if (error) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
return (error);
}
vf->vf_file = fp;
#ifdef _KERNEL
/*
* Make sure it's a regular file.
*/
if (zfs_file_getattr(fp, &zfa)) {
return (SET_ERROR(ENODEV));
}
if (!S_ISREG(zfa.zfa_mode)) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
return (SET_ERROR(ENODEV));
}
#endif
skip_open:
error = zfs_file_getattr(vf->vf_file, &zfa);
if (error) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
return (error);
}
*max_psize = *psize = zfa.zfa_size;
*logical_ashift = vdev_file_logical_ashift;
*physical_ashift = vdev_file_physical_ashift;
return (0);
}
static void
vdev_file_close(vdev_t *vd)
{
vdev_file_t *vf = vd->vdev_tsd;
if (vd->vdev_reopening || vf == NULL)
return;
if (vf->vf_file != NULL) {
(void) zfs_file_close(vf->vf_file);
}
vd->vdev_delayed_close = B_FALSE;
kmem_free(vf, sizeof (vdev_file_t));
vd->vdev_tsd = NULL;
}
static void
vdev_file_io_strategy(void *arg)
{
zio_t *zio = (zio_t *)arg;
vdev_t *vd = zio->io_vd;
vdev_file_t *vf = vd->vdev_tsd;
ssize_t resid;
void *buf;
loff_t off;
ssize_t size;
int err;
off = zio->io_offset;
size = zio->io_size;
resid = 0;
if (zio->io_type == ZIO_TYPE_READ) {
buf = abd_borrow_buf(zio->io_abd, zio->io_size);
err = zfs_file_pread(vf->vf_file, buf, size, off, &resid);
abd_return_buf_copy(zio->io_abd, buf, size);
} else {
buf = abd_borrow_buf_copy(zio->io_abd, zio->io_size);
err = zfs_file_pwrite(vf->vf_file, buf, size, off, &resid);
abd_return_buf(zio->io_abd, buf, size);
}
zio->io_error = err;
if (resid != 0 && zio->io_error == 0)
zio->io_error = SET_ERROR(ENOSPC);
zio_delay_interrupt(zio);
}
static void
vdev_file_io_fsync(void *arg)
{
zio_t *zio = (zio_t *)arg;
vdev_file_t *vf = zio->io_vd->vdev_tsd;
zio->io_error = zfs_file_fsync(vf->vf_file, O_SYNC | O_DSYNC);
zio_interrupt(zio);
}
static void
vdev_file_io_start(zio_t *zio)
{
vdev_t *vd = zio->io_vd;
vdev_file_t *vf = vd->vdev_tsd;
if (zio->io_type == ZIO_TYPE_IOCTL) {
/* XXPOLICY */
if (!vdev_readable(vd)) {
zio->io_error = SET_ERROR(ENXIO);
zio_interrupt(zio);
return;
}
switch (zio->io_cmd) {
case DKIOCFLUSHWRITECACHE:
if (zfs_nocacheflush)
break;
/*
* We cannot safely call vfs_fsync() when PF_FSTRANS
* is set in the current context. Filesystems like
* XFS include sanity checks to verify it is not
* already set, see xfs_vm_writepage(). Therefore
* the sync must be dispatched to a different context.
*/
if (__spl_pf_fstrans_check()) {
VERIFY3U(taskq_dispatch(vdev_file_taskq,
vdev_file_io_fsync, zio, TQ_SLEEP), !=,
TASKQID_INVALID);
return;
}
zio->io_error = zfs_file_fsync(vf->vf_file,
O_SYNC | O_DSYNC);
break;
default:
zio->io_error = SET_ERROR(ENOTSUP);
}
zio_execute(zio);
return;
} else if (zio->io_type == ZIO_TYPE_TRIM) {
int mode = 0;
ASSERT3U(zio->io_size, !=, 0);
#ifdef __linux__
mode = FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE;
#endif
zio->io_error = zfs_file_fallocate(vf->vf_file,
mode, zio->io_offset, zio->io_size);
zio_execute(zio);
return;
}
zio->io_target_timestamp = zio_handle_io_delay(zio);
VERIFY3U(taskq_dispatch(vdev_file_taskq, vdev_file_io_strategy, zio,
TQ_SLEEP), !=, TASKQID_INVALID);
}
static void
vdev_file_io_done(zio_t *zio)
{
(void) zio;
}
vdev_ops_t vdev_file_ops = {
.vdev_op_init = NULL,
.vdev_op_fini = NULL,
.vdev_op_open = vdev_file_open,
.vdev_op_close = vdev_file_close,
.vdev_op_asize = vdev_default_asize,
.vdev_op_min_asize = vdev_default_min_asize,
.vdev_op_min_alloc = NULL,
.vdev_op_io_start = vdev_file_io_start,
.vdev_op_io_done = vdev_file_io_done,
.vdev_op_state_change = NULL,
.vdev_op_need_resilver = NULL,
.vdev_op_hold = vdev_file_hold,
.vdev_op_rele = vdev_file_rele,
.vdev_op_remap = NULL,
.vdev_op_xlate = vdev_default_xlate,
.vdev_op_rebuild_asize = NULL,
.vdev_op_metaslab_init = NULL,
.vdev_op_config_generate = NULL,
.vdev_op_nparity = NULL,
.vdev_op_ndisks = NULL,
.vdev_op_type = VDEV_TYPE_FILE, /* name of this vdev type */
.vdev_op_leaf = B_TRUE /* leaf vdev */
};
void
vdev_file_init(void)
{
vdev_file_taskq = taskq_create("z_vdev_file", MAX(boot_ncpus, 16),
minclsyspri, boot_ncpus, INT_MAX, TASKQ_DYNAMIC);
VERIFY(vdev_file_taskq);
}
void
vdev_file_fini(void)
{
taskq_destroy(vdev_file_taskq);
}
/*
* From userland we access disks just like files.
*/
#ifndef _KERNEL
vdev_ops_t vdev_disk_ops = {
.vdev_op_init = NULL,
.vdev_op_fini = NULL,
.vdev_op_open = vdev_file_open,
.vdev_op_close = vdev_file_close,
.vdev_op_asize = vdev_default_asize,
.vdev_op_min_asize = vdev_default_min_asize,
.vdev_op_min_alloc = NULL,
.vdev_op_io_start = vdev_file_io_start,
.vdev_op_io_done = vdev_file_io_done,
.vdev_op_state_change = NULL,
.vdev_op_need_resilver = NULL,
.vdev_op_hold = vdev_file_hold,
.vdev_op_rele = vdev_file_rele,
.vdev_op_remap = NULL,
.vdev_op_xlate = vdev_default_xlate,
.vdev_op_rebuild_asize = NULL,
.vdev_op_metaslab_init = NULL,
.vdev_op_config_generate = NULL,
.vdev_op_nparity = NULL,
.vdev_op_ndisks = NULL,
.vdev_op_type = VDEV_TYPE_DISK, /* name of this vdev type */
.vdev_op_leaf = B_TRUE /* leaf vdev */
};
#endif
ZFS_MODULE_PARAM(zfs_vdev_file, vdev_file_, logical_ashift, ULONG, ZMOD_RW,
"Logical ashift for file-based devices");
ZFS_MODULE_PARAM(zfs_vdev_file, vdev_file_, physical_ashift, ULONG, ZMOD_RW,
"Physical ashift for file-based devices");
diff --git a/module/os/linux/zfs/zfs_acl.c b/module/os/linux/zfs/zfs_acl.c
index cf37aecf8a22..94b20dd6e582 100644
--- a/module/os/linux/zfs/zfs_acl.c
+++ b/module/os/linux/zfs/zfs_acl.c
@@ -1,2948 +1,2948 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright 2014 Nexenta Systems, Inc. All rights reserved.
*/
#include <sys/types.h>
#include <sys/param.h>
#include <sys/time.h>
#include <sys/sysmacros.h>
#include <sys/vfs.h>
#include <sys/vnode.h>
#include <sys/sid.h>
#include <sys/file.h>
#include <sys/stat.h>
#include <sys/kmem.h>
#include <sys/cmn_err.h>
#include <sys/errno.h>
#include <sys/fs/zfs.h>
#include <sys/policy.h>
#include <sys/zfs_znode.h>
#include <sys/zfs_fuid.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_dir.h>
#include <sys/zfs_quota.h>
#include <sys/zfs_vfsops.h>
#include <sys/dmu.h>
#include <sys/dnode.h>
#include <sys/zap.h>
#include <sys/sa.h>
#include <sys/trace_acl.h>
#include <sys/zpl.h>
#define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
#define DENY ACE_ACCESS_DENIED_ACE_TYPE
#define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
#define MIN_ACE_TYPE ALLOW
#define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
#define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
#define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
#define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
#define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
#define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
#define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
ACE_DELETE|ACE_DELETE_CHILD)
#define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
#define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
#define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
#define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
#define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
#define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
ZFS_ACL_PROTECTED)
#define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
ZFS_ACL_OBJ_ACE)
#define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
#define IDMAP_WK_CREATOR_OWNER_UID 2147483648U
static uint16_t
zfs_ace_v0_get_type(void *acep)
{
return (((zfs_oldace_t *)acep)->z_type);
}
static uint16_t
zfs_ace_v0_get_flags(void *acep)
{
return (((zfs_oldace_t *)acep)->z_flags);
}
static uint32_t
zfs_ace_v0_get_mask(void *acep)
{
return (((zfs_oldace_t *)acep)->z_access_mask);
}
static uint64_t
zfs_ace_v0_get_who(void *acep)
{
return (((zfs_oldace_t *)acep)->z_fuid);
}
static void
zfs_ace_v0_set_type(void *acep, uint16_t type)
{
((zfs_oldace_t *)acep)->z_type = type;
}
static void
zfs_ace_v0_set_flags(void *acep, uint16_t flags)
{
((zfs_oldace_t *)acep)->z_flags = flags;
}
static void
zfs_ace_v0_set_mask(void *acep, uint32_t mask)
{
((zfs_oldace_t *)acep)->z_access_mask = mask;
}
static void
zfs_ace_v0_set_who(void *acep, uint64_t who)
{
((zfs_oldace_t *)acep)->z_fuid = who;
}
/*ARGSUSED*/
static size_t
zfs_ace_v0_size(void *acep)
{
return (sizeof (zfs_oldace_t));
}
static size_t
zfs_ace_v0_abstract_size(void)
{
return (sizeof (zfs_oldace_t));
}
static int
zfs_ace_v0_mask_off(void)
{
return (offsetof(zfs_oldace_t, z_access_mask));
}
/*ARGSUSED*/
static int
zfs_ace_v0_data(void *acep, void **datap)
{
*datap = NULL;
return (0);
}
-static acl_ops_t zfs_acl_v0_ops = {
+static const acl_ops_t zfs_acl_v0_ops = {
.ace_mask_get = zfs_ace_v0_get_mask,
.ace_mask_set = zfs_ace_v0_set_mask,
.ace_flags_get = zfs_ace_v0_get_flags,
.ace_flags_set = zfs_ace_v0_set_flags,
.ace_type_get = zfs_ace_v0_get_type,
.ace_type_set = zfs_ace_v0_set_type,
.ace_who_get = zfs_ace_v0_get_who,
.ace_who_set = zfs_ace_v0_set_who,
.ace_size = zfs_ace_v0_size,
.ace_abstract_size = zfs_ace_v0_abstract_size,
.ace_mask_off = zfs_ace_v0_mask_off,
.ace_data = zfs_ace_v0_data
};
static uint16_t
zfs_ace_fuid_get_type(void *acep)
{
return (((zfs_ace_hdr_t *)acep)->z_type);
}
static uint16_t
zfs_ace_fuid_get_flags(void *acep)
{
return (((zfs_ace_hdr_t *)acep)->z_flags);
}
static uint32_t
zfs_ace_fuid_get_mask(void *acep)
{
return (((zfs_ace_hdr_t *)acep)->z_access_mask);
}
static uint64_t
zfs_ace_fuid_get_who(void *args)
{
uint16_t entry_type;
zfs_ace_t *acep = args;
entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE)
return (-1);
return (((zfs_ace_t *)acep)->z_fuid);
}
static void
zfs_ace_fuid_set_type(void *acep, uint16_t type)
{
((zfs_ace_hdr_t *)acep)->z_type = type;
}
static void
zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
{
((zfs_ace_hdr_t *)acep)->z_flags = flags;
}
static void
zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
{
((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
}
static void
zfs_ace_fuid_set_who(void *arg, uint64_t who)
{
zfs_ace_t *acep = arg;
uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE)
return;
acep->z_fuid = who;
}
static size_t
zfs_ace_fuid_size(void *acep)
{
zfs_ace_hdr_t *zacep = acep;
uint16_t entry_type;
switch (zacep->z_type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
return (sizeof (zfs_object_ace_t));
case ALLOW:
case DENY:
entry_type =
(((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
if (entry_type == ACE_OWNER ||
entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE)
return (sizeof (zfs_ace_hdr_t));
fallthrough;
default:
return (sizeof (zfs_ace_t));
}
}
static size_t
zfs_ace_fuid_abstract_size(void)
{
return (sizeof (zfs_ace_hdr_t));
}
static int
zfs_ace_fuid_mask_off(void)
{
return (offsetof(zfs_ace_hdr_t, z_access_mask));
}
static int
zfs_ace_fuid_data(void *acep, void **datap)
{
zfs_ace_t *zacep = acep;
zfs_object_ace_t *zobjp;
switch (zacep->z_hdr.z_type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
zobjp = acep;
*datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
default:
*datap = NULL;
return (0);
}
}
-static acl_ops_t zfs_acl_fuid_ops = {
+static const acl_ops_t zfs_acl_fuid_ops = {
.ace_mask_get = zfs_ace_fuid_get_mask,
.ace_mask_set = zfs_ace_fuid_set_mask,
.ace_flags_get = zfs_ace_fuid_get_flags,
.ace_flags_set = zfs_ace_fuid_set_flags,
.ace_type_get = zfs_ace_fuid_get_type,
.ace_type_set = zfs_ace_fuid_set_type,
.ace_who_get = zfs_ace_fuid_get_who,
.ace_who_set = zfs_ace_fuid_set_who,
.ace_size = zfs_ace_fuid_size,
.ace_abstract_size = zfs_ace_fuid_abstract_size,
.ace_mask_off = zfs_ace_fuid_mask_off,
.ace_data = zfs_ace_fuid_data
};
/*
* The following three functions are provided for compatibility with
* older ZPL version in order to determine if the file use to have
* an external ACL and what version of ACL previously existed on the
* file. Would really be nice to not need this, sigh.
*/
uint64_t
zfs_external_acl(znode_t *zp)
{
zfs_acl_phys_t acl_phys;
int error;
if (zp->z_is_sa)
return (0);
/*
* Need to deal with a potential
* race where zfs_sa_upgrade could cause
* z_isa_sa to change.
*
* If the lookup fails then the state of z_is_sa should have
* changed.
*/
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
&acl_phys, sizeof (acl_phys))) == 0)
return (acl_phys.z_acl_extern_obj);
else {
/*
* after upgrade the SA_ZPL_ZNODE_ACL should have been
* removed
*/
VERIFY(zp->z_is_sa && error == ENOENT);
return (0);
}
}
/*
* Determine size of ACL in bytes
*
* This is more complicated than it should be since we have to deal
* with old external ACLs.
*/
static int
zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
zfs_acl_phys_t *aclphys)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
uint64_t acl_count;
int size;
int error;
ASSERT(MUTEX_HELD(&zp->z_acl_lock));
if (zp->z_is_sa) {
if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zfsvfs),
&size)) != 0)
return (error);
*aclsize = size;
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zfsvfs),
&acl_count, sizeof (acl_count))) != 0)
return (error);
*aclcount = acl_count;
} else {
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
aclphys, sizeof (*aclphys))) != 0)
return (error);
if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
*aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
*aclcount = aclphys->z_acl_size;
} else {
*aclsize = aclphys->z_acl_size;
*aclcount = aclphys->z_acl_count;
}
}
return (0);
}
int
zfs_znode_acl_version(znode_t *zp)
{
zfs_acl_phys_t acl_phys;
if (zp->z_is_sa)
return (ZFS_ACL_VERSION_FUID);
else {
int error;
/*
* Need to deal with a potential
* race where zfs_sa_upgrade could cause
* z_isa_sa to change.
*
* If the lookup fails then the state of z_is_sa should have
* changed.
*/
if ((error = sa_lookup(zp->z_sa_hdl,
SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
&acl_phys, sizeof (acl_phys))) == 0)
return (acl_phys.z_acl_version);
else {
/*
* After upgrade SA_ZPL_ZNODE_ACL should have
* been removed.
*/
VERIFY(zp->z_is_sa && error == ENOENT);
return (ZFS_ACL_VERSION_FUID);
}
}
}
static int
zfs_acl_version(int version)
{
if (version < ZPL_VERSION_FUID)
return (ZFS_ACL_VERSION_INITIAL);
else
return (ZFS_ACL_VERSION_FUID);
}
static int
zfs_acl_version_zp(znode_t *zp)
{
return (zfs_acl_version(ZTOZSB(zp)->z_version));
}
zfs_acl_t *
zfs_acl_alloc(int vers)
{
zfs_acl_t *aclp;
aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
offsetof(zfs_acl_node_t, z_next));
aclp->z_version = vers;
if (vers == ZFS_ACL_VERSION_FUID)
aclp->z_ops = &zfs_acl_fuid_ops;
else
aclp->z_ops = &zfs_acl_v0_ops;
return (aclp);
}
zfs_acl_node_t *
zfs_acl_node_alloc(size_t bytes)
{
zfs_acl_node_t *aclnode;
aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
if (bytes) {
aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
aclnode->z_allocdata = aclnode->z_acldata;
aclnode->z_allocsize = bytes;
aclnode->z_size = bytes;
}
return (aclnode);
}
static void
zfs_acl_node_free(zfs_acl_node_t *aclnode)
{
if (aclnode->z_allocsize)
kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
kmem_free(aclnode, sizeof (zfs_acl_node_t));
}
static void
zfs_acl_release_nodes(zfs_acl_t *aclp)
{
zfs_acl_node_t *aclnode;
while ((aclnode = list_head(&aclp->z_acl))) {
list_remove(&aclp->z_acl, aclnode);
zfs_acl_node_free(aclnode);
}
aclp->z_acl_count = 0;
aclp->z_acl_bytes = 0;
}
void
zfs_acl_free(zfs_acl_t *aclp)
{
zfs_acl_release_nodes(aclp);
list_destroy(&aclp->z_acl);
kmem_free(aclp, sizeof (zfs_acl_t));
}
static boolean_t
zfs_acl_valid_ace_type(uint_t type, uint_t flags)
{
uint16_t entry_type;
switch (type) {
case ALLOW:
case DENY:
case ACE_SYSTEM_AUDIT_ACE_TYPE:
case ACE_SYSTEM_ALARM_ACE_TYPE:
entry_type = flags & ACE_TYPE_FLAGS;
return (entry_type == ACE_OWNER ||
entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE || entry_type == 0 ||
entry_type == ACE_IDENTIFIER_GROUP);
default:
if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
return (B_TRUE);
}
return (B_FALSE);
}
static boolean_t
zfs_ace_valid(umode_t obj_mode, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
{
/*
* first check type of entry
*/
if (!zfs_acl_valid_ace_type(type, iflags))
return (B_FALSE);
switch (type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
if (aclp->z_version < ZFS_ACL_VERSION_FUID)
return (B_FALSE);
aclp->z_hints |= ZFS_ACL_OBJ_ACE;
}
/*
* next check inheritance level flags
*/
if (S_ISDIR(obj_mode) &&
(iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
aclp->z_hints |= ZFS_INHERIT_ACE;
if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
if ((iflags & (ACE_FILE_INHERIT_ACE|
ACE_DIRECTORY_INHERIT_ACE)) == 0) {
return (B_FALSE);
}
}
return (B_TRUE);
}
static void *
zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
{
zfs_acl_node_t *aclnode;
ASSERT(aclp);
if (start == NULL) {
aclnode = list_head(&aclp->z_acl);
if (aclnode == NULL)
return (NULL);
aclp->z_next_ace = aclnode->z_acldata;
aclp->z_curr_node = aclnode;
aclnode->z_ace_idx = 0;
}
aclnode = aclp->z_curr_node;
if (aclnode == NULL)
return (NULL);
if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
aclnode = list_next(&aclp->z_acl, aclnode);
if (aclnode == NULL)
return (NULL);
else {
aclp->z_curr_node = aclnode;
aclnode->z_ace_idx = 0;
aclp->z_next_ace = aclnode->z_acldata;
}
}
if (aclnode->z_ace_idx < aclnode->z_ace_count) {
void *acep = aclp->z_next_ace;
size_t ace_size;
/*
* Make sure we don't overstep our bounds
*/
ace_size = aclp->z_ops->ace_size(acep);
if (((caddr_t)acep + ace_size) >
((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
return (NULL);
}
*iflags = aclp->z_ops->ace_flags_get(acep);
*type = aclp->z_ops->ace_type_get(acep);
*access_mask = aclp->z_ops->ace_mask_get(acep);
*who = aclp->z_ops->ace_who_get(acep);
aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
aclnode->z_ace_idx++;
return ((void *)acep);
}
return (NULL);
}
/*ARGSUSED*/
static uint64_t
zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
uint16_t *flags, uint16_t *type, uint32_t *mask)
{
zfs_acl_t *aclp = datap;
zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
uint64_t who;
acep = zfs_acl_next_ace(aclp, acep, &who, mask,
flags, type);
return ((uint64_t)(uintptr_t)acep);
}
/*
* Copy ACE to internal ZFS format.
* While processing the ACL each ACE will be validated for correctness.
* ACE FUIDs will be created later.
*/
static int
zfs_copy_ace_2_fuid(zfsvfs_t *zfsvfs, umode_t obj_mode, zfs_acl_t *aclp,
void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
zfs_fuid_info_t **fuidp, cred_t *cr)
{
int i;
uint16_t entry_type;
zfs_ace_t *aceptr = z_acl;
ace_t *acep = datap;
zfs_object_ace_t *zobjacep;
ace_object_t *aceobjp;
for (i = 0; i != aclcnt; i++) {
aceptr->z_hdr.z_access_mask = acep->a_access_mask;
aceptr->z_hdr.z_flags = acep->a_flags;
aceptr->z_hdr.z_type = acep->a_type;
entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
entry_type != ACE_EVERYONE) {
aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
cr, (entry_type == 0) ?
ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
}
/*
* Make sure ACE is valid
*/
if (zfs_ace_valid(obj_mode, aclp, aceptr->z_hdr.z_type,
aceptr->z_hdr.z_flags) != B_TRUE)
return (SET_ERROR(EINVAL));
switch (acep->a_type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
zobjacep = (zfs_object_ace_t *)aceptr;
aceobjp = (ace_object_t *)acep;
bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
sizeof (aceobjp->a_obj_type));
bcopy(aceobjp->a_inherit_obj_type,
zobjacep->z_inherit_type,
sizeof (aceobjp->a_inherit_obj_type));
acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
break;
default:
acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
}
aceptr = (zfs_ace_t *)((caddr_t)aceptr +
aclp->z_ops->ace_size(aceptr));
}
*size = (caddr_t)aceptr - (caddr_t)z_acl;
return (0);
}
/*
* Copy ZFS ACEs to fixed size ace_t layout
*/
static void
zfs_copy_fuid_2_ace(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, cred_t *cr,
void *datap, int filter)
{
uint64_t who;
uint32_t access_mask;
uint16_t iflags, type;
zfs_ace_hdr_t *zacep = NULL;
ace_t *acep = datap;
ace_object_t *objacep;
zfs_object_ace_t *zobjacep;
size_t ace_size;
uint16_t entry_type;
while ((zacep = zfs_acl_next_ace(aclp, zacep,
&who, &access_mask, &iflags, &type))) {
switch (type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
if (filter) {
continue;
}
zobjacep = (zfs_object_ace_t *)zacep;
objacep = (ace_object_t *)acep;
bcopy(zobjacep->z_object_type,
objacep->a_obj_type,
sizeof (zobjacep->z_object_type));
bcopy(zobjacep->z_inherit_type,
objacep->a_inherit_obj_type,
sizeof (zobjacep->z_inherit_type));
ace_size = sizeof (ace_object_t);
break;
default:
ace_size = sizeof (ace_t);
break;
}
entry_type = (iflags & ACE_TYPE_FLAGS);
if ((entry_type != ACE_OWNER &&
entry_type != OWNING_GROUP &&
entry_type != ACE_EVERYONE)) {
acep->a_who = zfs_fuid_map_id(zfsvfs, who,
cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
ZFS_ACE_GROUP : ZFS_ACE_USER);
} else {
acep->a_who = (uid_t)(int64_t)who;
}
acep->a_access_mask = access_mask;
acep->a_flags = iflags;
acep->a_type = type;
acep = (ace_t *)((caddr_t)acep + ace_size);
}
}
static int
zfs_copy_ace_2_oldace(umode_t obj_mode, zfs_acl_t *aclp, ace_t *acep,
zfs_oldace_t *z_acl, int aclcnt, size_t *size)
{
int i;
zfs_oldace_t *aceptr = z_acl;
for (i = 0; i != aclcnt; i++, aceptr++) {
aceptr->z_access_mask = acep[i].a_access_mask;
aceptr->z_type = acep[i].a_type;
aceptr->z_flags = acep[i].a_flags;
aceptr->z_fuid = acep[i].a_who;
/*
* Make sure ACE is valid
*/
if (zfs_ace_valid(obj_mode, aclp, aceptr->z_type,
aceptr->z_flags) != B_TRUE)
return (SET_ERROR(EINVAL));
}
*size = (caddr_t)aceptr - (caddr_t)z_acl;
return (0);
}
/*
* convert old ACL format to new
*/
void
zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
{
zfs_oldace_t *oldaclp;
int i;
uint16_t type, iflags;
uint32_t access_mask;
uint64_t who;
void *cookie = NULL;
zfs_acl_node_t *newaclnode;
ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
/*
* First create the ACE in a contiguous piece of memory
* for zfs_copy_ace_2_fuid().
*
* We only convert an ACL once, so this won't happen
* every time.
*/
oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
KM_SLEEP);
i = 0;
while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
&access_mask, &iflags, &type))) {
oldaclp[i].z_flags = iflags;
oldaclp[i].z_type = type;
oldaclp[i].z_fuid = who;
oldaclp[i++].z_access_mask = access_mask;
}
newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
sizeof (zfs_object_ace_t));
aclp->z_ops = &zfs_acl_fuid_ops;
VERIFY(zfs_copy_ace_2_fuid(ZTOZSB(zp), ZTOI(zp)->i_mode,
aclp, oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
&newaclnode->z_size, NULL, cr) == 0);
newaclnode->z_ace_count = aclp->z_acl_count;
aclp->z_version = ZFS_ACL_VERSION;
kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
/*
* Release all previous ACL nodes
*/
zfs_acl_release_nodes(aclp);
list_insert_head(&aclp->z_acl, newaclnode);
aclp->z_acl_bytes = newaclnode->z_size;
aclp->z_acl_count = newaclnode->z_ace_count;
}
/*
* Convert unix access mask to v4 access mask
*/
static uint32_t
zfs_unix_to_v4(uint32_t access_mask)
{
uint32_t new_mask = 0;
if (access_mask & S_IXOTH)
new_mask |= ACE_EXECUTE;
if (access_mask & S_IWOTH)
new_mask |= ACE_WRITE_DATA;
if (access_mask & S_IROTH)
new_mask |= ACE_READ_DATA;
return (new_mask);
}
static void
zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
uint16_t access_type, uint64_t fuid, uint16_t entry_type)
{
uint16_t type = entry_type & ACE_TYPE_FLAGS;
aclp->z_ops->ace_mask_set(acep, access_mask);
aclp->z_ops->ace_type_set(acep, access_type);
aclp->z_ops->ace_flags_set(acep, entry_type);
if ((type != ACE_OWNER && type != OWNING_GROUP &&
type != ACE_EVERYONE))
aclp->z_ops->ace_who_set(acep, fuid);
}
/*
* Determine mode of file based on ACL.
*/
uint64_t
zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
uint64_t *pflags, uint64_t fuid, uint64_t fgid)
{
int entry_type;
mode_t mode;
mode_t seen = 0;
zfs_ace_hdr_t *acep = NULL;
uint64_t who;
uint16_t iflags, type;
uint32_t access_mask;
boolean_t an_exec_denied = B_FALSE;
mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
while ((acep = zfs_acl_next_ace(aclp, acep, &who,
&access_mask, &iflags, &type))) {
if (!zfs_acl_valid_ace_type(type, iflags))
continue;
entry_type = (iflags & ACE_TYPE_FLAGS);
/*
* Skip over any inherit_only ACEs
*/
if (iflags & ACE_INHERIT_ONLY_ACE)
continue;
if (entry_type == ACE_OWNER || (entry_type == 0 &&
who == fuid)) {
if ((access_mask & ACE_READ_DATA) &&
(!(seen & S_IRUSR))) {
seen |= S_IRUSR;
if (type == ALLOW) {
mode |= S_IRUSR;
}
}
if ((access_mask & ACE_WRITE_DATA) &&
(!(seen & S_IWUSR))) {
seen |= S_IWUSR;
if (type == ALLOW) {
mode |= S_IWUSR;
}
}
if ((access_mask & ACE_EXECUTE) &&
(!(seen & S_IXUSR))) {
seen |= S_IXUSR;
if (type == ALLOW) {
mode |= S_IXUSR;
}
}
} else if (entry_type == OWNING_GROUP ||
(entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
if ((access_mask & ACE_READ_DATA) &&
(!(seen & S_IRGRP))) {
seen |= S_IRGRP;
if (type == ALLOW) {
mode |= S_IRGRP;
}
}
if ((access_mask & ACE_WRITE_DATA) &&
(!(seen & S_IWGRP))) {
seen |= S_IWGRP;
if (type == ALLOW) {
mode |= S_IWGRP;
}
}
if ((access_mask & ACE_EXECUTE) &&
(!(seen & S_IXGRP))) {
seen |= S_IXGRP;
if (type == ALLOW) {
mode |= S_IXGRP;
}
}
} else if (entry_type == ACE_EVERYONE) {
if ((access_mask & ACE_READ_DATA)) {
if (!(seen & S_IRUSR)) {
seen |= S_IRUSR;
if (type == ALLOW) {
mode |= S_IRUSR;
}
}
if (!(seen & S_IRGRP)) {
seen |= S_IRGRP;
if (type == ALLOW) {
mode |= S_IRGRP;
}
}
if (!(seen & S_IROTH)) {
seen |= S_IROTH;
if (type == ALLOW) {
mode |= S_IROTH;
}
}
}
if ((access_mask & ACE_WRITE_DATA)) {
if (!(seen & S_IWUSR)) {
seen |= S_IWUSR;
if (type == ALLOW) {
mode |= S_IWUSR;
}
}
if (!(seen & S_IWGRP)) {
seen |= S_IWGRP;
if (type == ALLOW) {
mode |= S_IWGRP;
}
}
if (!(seen & S_IWOTH)) {
seen |= S_IWOTH;
if (type == ALLOW) {
mode |= S_IWOTH;
}
}
}
if ((access_mask & ACE_EXECUTE)) {
if (!(seen & S_IXUSR)) {
seen |= S_IXUSR;
if (type == ALLOW) {
mode |= S_IXUSR;
}
}
if (!(seen & S_IXGRP)) {
seen |= S_IXGRP;
if (type == ALLOW) {
mode |= S_IXGRP;
}
}
if (!(seen & S_IXOTH)) {
seen |= S_IXOTH;
if (type == ALLOW) {
mode |= S_IXOTH;
}
}
}
} else {
/*
* Only care if this IDENTIFIER_GROUP or
* USER ACE denies execute access to someone,
* mode is not affected
*/
if ((access_mask & ACE_EXECUTE) && type == DENY)
an_exec_denied = B_TRUE;
}
}
/*
* Failure to allow is effectively a deny, so execute permission
* is denied if it was never mentioned or if we explicitly
* weren't allowed it.
*/
if (!an_exec_denied &&
((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
(mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
an_exec_denied = B_TRUE;
if (an_exec_denied)
*pflags &= ~ZFS_NO_EXECS_DENIED;
else
*pflags |= ZFS_NO_EXECS_DENIED;
return (mode);
}
/*
* Read an external acl object. If the intent is to modify, always
* create a new acl and leave any cached acl in place.
*/
int
zfs_acl_node_read(struct znode *zp, boolean_t have_lock, zfs_acl_t **aclpp,
boolean_t will_modify)
{
zfs_acl_t *aclp;
int aclsize = 0;
int acl_count = 0;
zfs_acl_node_t *aclnode;
zfs_acl_phys_t znode_acl;
int version;
int error;
boolean_t drop_lock = B_FALSE;
ASSERT(MUTEX_HELD(&zp->z_acl_lock));
if (zp->z_acl_cached && !will_modify) {
*aclpp = zp->z_acl_cached;
return (0);
}
/*
* close race where znode could be upgrade while trying to
* read the znode attributes.
*
* But this could only happen if the file isn't already an SA
* znode
*/
if (!zp->z_is_sa && !have_lock) {
mutex_enter(&zp->z_lock);
drop_lock = B_TRUE;
}
version = zfs_znode_acl_version(zp);
if ((error = zfs_acl_znode_info(zp, &aclsize,
&acl_count, &znode_acl)) != 0) {
goto done;
}
aclp = zfs_acl_alloc(version);
aclp->z_acl_count = acl_count;
aclp->z_acl_bytes = aclsize;
aclnode = zfs_acl_node_alloc(aclsize);
aclnode->z_ace_count = aclp->z_acl_count;
aclnode->z_size = aclsize;
if (!zp->z_is_sa) {
if (znode_acl.z_acl_extern_obj) {
error = dmu_read(ZTOZSB(zp)->z_os,
znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
aclnode->z_acldata, DMU_READ_PREFETCH);
} else {
bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
aclnode->z_size);
}
} else {
error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(ZTOZSB(zp)),
aclnode->z_acldata, aclnode->z_size);
}
if (error != 0) {
zfs_acl_free(aclp);
zfs_acl_node_free(aclnode);
/* convert checksum errors into IO errors */
if (error == ECKSUM)
error = SET_ERROR(EIO);
goto done;
}
list_insert_head(&aclp->z_acl, aclnode);
*aclpp = aclp;
if (!will_modify)
zp->z_acl_cached = aclp;
done:
if (drop_lock)
mutex_exit(&zp->z_lock);
return (error);
}
/*ARGSUSED*/
void
zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
boolean_t start, void *userdata)
{
zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
if (start) {
cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
} else {
cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
cb->cb_acl_node);
}
*dataptr = cb->cb_acl_node->z_acldata;
*length = cb->cb_acl_node->z_size;
}
int
zfs_acl_chown_setattr(znode_t *zp)
{
int error;
zfs_acl_t *aclp;
if (ZTOZSB(zp)->z_acl_type == ZFS_ACLTYPE_POSIX)
return (0);
ASSERT(MUTEX_HELD(&zp->z_lock));
ASSERT(MUTEX_HELD(&zp->z_acl_lock));
error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
if (error == 0 && aclp->z_acl_count > 0)
zp->z_mode = ZTOI(zp)->i_mode =
zfs_mode_compute(zp->z_mode, aclp,
&zp->z_pflags, KUID_TO_SUID(ZTOI(zp)->i_uid),
KGID_TO_SGID(ZTOI(zp)->i_gid));
/*
* Some ZFS implementations (ZEVO) create neither a ZNODE_ACL
* nor a DACL_ACES SA in which case ENOENT is returned from
* zfs_acl_node_read() when the SA can't be located.
* Allow chown/chgrp to succeed in these cases rather than
* returning an error that makes no sense in the context of
* the caller.
*/
if (error == ENOENT)
return (0);
return (error);
}
typedef struct trivial_acl {
uint32_t allow0; /* allow mask for bits only in owner */
uint32_t deny1; /* deny mask for bits not in owner */
uint32_t deny2; /* deny mask for bits not in group */
uint32_t owner; /* allow mask matching mode */
uint32_t group; /* allow mask matching mode */
uint32_t everyone; /* allow mask matching mode */
} trivial_acl_t;
static void
acl_trivial_access_masks(mode_t mode, boolean_t isdir, trivial_acl_t *masks)
{
uint32_t read_mask = ACE_READ_DATA;
uint32_t write_mask = ACE_WRITE_DATA|ACE_APPEND_DATA;
uint32_t execute_mask = ACE_EXECUTE;
if (isdir)
write_mask |= ACE_DELETE_CHILD;
masks->deny1 = 0;
if (!(mode & S_IRUSR) && (mode & (S_IRGRP|S_IROTH)))
masks->deny1 |= read_mask;
if (!(mode & S_IWUSR) && (mode & (S_IWGRP|S_IWOTH)))
masks->deny1 |= write_mask;
if (!(mode & S_IXUSR) && (mode & (S_IXGRP|S_IXOTH)))
masks->deny1 |= execute_mask;
masks->deny2 = 0;
if (!(mode & S_IRGRP) && (mode & S_IROTH))
masks->deny2 |= read_mask;
if (!(mode & S_IWGRP) && (mode & S_IWOTH))
masks->deny2 |= write_mask;
if (!(mode & S_IXGRP) && (mode & S_IXOTH))
masks->deny2 |= execute_mask;
masks->allow0 = 0;
if ((mode & S_IRUSR) && (!(mode & S_IRGRP) && (mode & S_IROTH)))
masks->allow0 |= read_mask;
if ((mode & S_IWUSR) && (!(mode & S_IWGRP) && (mode & S_IWOTH)))
masks->allow0 |= write_mask;
if ((mode & S_IXUSR) && (!(mode & S_IXGRP) && (mode & S_IXOTH)))
masks->allow0 |= execute_mask;
masks->owner = ACE_WRITE_ATTRIBUTES|ACE_WRITE_OWNER|ACE_WRITE_ACL|
ACE_WRITE_NAMED_ATTRS|ACE_READ_ACL|ACE_READ_ATTRIBUTES|
ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE;
if (mode & S_IRUSR)
masks->owner |= read_mask;
if (mode & S_IWUSR)
masks->owner |= write_mask;
if (mode & S_IXUSR)
masks->owner |= execute_mask;
masks->group = ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_NAMED_ATTRS|
ACE_SYNCHRONIZE;
if (mode & S_IRGRP)
masks->group |= read_mask;
if (mode & S_IWGRP)
masks->group |= write_mask;
if (mode & S_IXGRP)
masks->group |= execute_mask;
masks->everyone = ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_NAMED_ATTRS|
ACE_SYNCHRONIZE;
if (mode & S_IROTH)
masks->everyone |= read_mask;
if (mode & S_IWOTH)
masks->everyone |= write_mask;
if (mode & S_IXOTH)
masks->everyone |= execute_mask;
}
/*
* ace_trivial:
* determine whether an ace_t acl is trivial
*
* Trivialness implies that the acl is composed of only
* owner, group, everyone entries. ACL can't
* have read_acl denied, and write_owner/write_acl/write_attributes
* can only be owner@ entry.
*/
static int
ace_trivial_common(void *acep, int aclcnt,
uint64_t (*walk)(void *, uint64_t, int aclcnt,
uint16_t *, uint16_t *, uint32_t *))
{
uint16_t flags;
uint32_t mask;
uint16_t type;
uint64_t cookie = 0;
while ((cookie = walk(acep, cookie, aclcnt, &flags, &type, &mask))) {
switch (flags & ACE_TYPE_FLAGS) {
case ACE_OWNER:
case ACE_GROUP|ACE_IDENTIFIER_GROUP:
case ACE_EVERYONE:
break;
default:
return (1);
}
if (flags & (ACE_FILE_INHERIT_ACE|
ACE_DIRECTORY_INHERIT_ACE|ACE_NO_PROPAGATE_INHERIT_ACE|
ACE_INHERIT_ONLY_ACE))
return (1);
/*
* Special check for some special bits
*
* Don't allow anybody to deny reading basic
* attributes or a files ACL.
*/
if ((mask & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
(type == ACE_ACCESS_DENIED_ACE_TYPE))
return (1);
/*
* Delete permission is never set by default
*/
if (mask & ACE_DELETE)
return (1);
/*
* Child delete permission should be accompanied by write
*/
if ((mask & ACE_DELETE_CHILD) && !(mask & ACE_WRITE_DATA))
return (1);
/*
* only allow owner@ to have
* write_acl/write_owner/write_attributes/write_xattr/
*/
if (type == ACE_ACCESS_ALLOWED_ACE_TYPE &&
(!(flags & ACE_OWNER) && (mask &
(ACE_WRITE_OWNER|ACE_WRITE_ACL| ACE_WRITE_ATTRIBUTES|
ACE_WRITE_NAMED_ATTRS))))
return (1);
}
return (0);
}
/*
* common code for setting ACLs.
*
* This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
* zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
* already checked the acl and knows whether to inherit.
*/
int
zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
{
int error;
zfsvfs_t *zfsvfs = ZTOZSB(zp);
dmu_object_type_t otype;
zfs_acl_locator_cb_t locate = { 0 };
uint64_t mode;
sa_bulk_attr_t bulk[5];
uint64_t ctime[2];
int count = 0;
zfs_acl_phys_t acl_phys;
mode = zp->z_mode;
mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
KUID_TO_SUID(ZTOI(zp)->i_uid), KGID_TO_SGID(ZTOI(zp)->i_gid));
zp->z_mode = ZTOI(zp)->i_mode = mode;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
&mode, sizeof (mode));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
&ctime, sizeof (ctime));
if (zp->z_acl_cached) {
zfs_acl_free(zp->z_acl_cached);
zp->z_acl_cached = NULL;
}
/*
* Upgrade needed?
*/
if (!zfsvfs->z_use_fuids) {
otype = DMU_OT_OLDACL;
} else {
if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
(zfsvfs->z_version >= ZPL_VERSION_FUID))
zfs_acl_xform(zp, aclp, cr);
ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
otype = DMU_OT_ACL;
}
/*
* Arrgh, we have to handle old on disk format
* as well as newer (preferred) SA format.
*/
if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
locate.cb_aclp = aclp;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
NULL, &aclp->z_acl_count, sizeof (uint64_t));
} else { /* Painful legacy way */
zfs_acl_node_t *aclnode;
uint64_t off = 0;
uint64_t aoid;
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
&acl_phys, sizeof (acl_phys))) != 0)
return (error);
aoid = acl_phys.z_acl_extern_obj;
if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
/*
* If ACL was previously external and we are now
* converting to new ACL format then release old
* ACL object and create a new one.
*/
if (aoid &&
aclp->z_version != acl_phys.z_acl_version) {
error = dmu_object_free(zfsvfs->z_os, aoid, tx);
if (error)
return (error);
aoid = 0;
}
if (aoid == 0) {
aoid = dmu_object_alloc(zfsvfs->z_os,
otype, aclp->z_acl_bytes,
otype == DMU_OT_ACL ?
DMU_OT_SYSACL : DMU_OT_NONE,
otype == DMU_OT_ACL ?
DN_OLD_MAX_BONUSLEN : 0, tx);
} else {
(void) dmu_object_set_blocksize(zfsvfs->z_os,
aoid, aclp->z_acl_bytes, 0, tx);
}
acl_phys.z_acl_extern_obj = aoid;
for (aclnode = list_head(&aclp->z_acl); aclnode;
aclnode = list_next(&aclp->z_acl, aclnode)) {
if (aclnode->z_ace_count == 0)
continue;
dmu_write(zfsvfs->z_os, aoid, off,
aclnode->z_size, aclnode->z_acldata, tx);
off += aclnode->z_size;
}
} else {
void *start = acl_phys.z_ace_data;
/*
* Migrating back embedded?
*/
if (acl_phys.z_acl_extern_obj) {
error = dmu_object_free(zfsvfs->z_os,
acl_phys.z_acl_extern_obj, tx);
if (error)
return (error);
acl_phys.z_acl_extern_obj = 0;
}
for (aclnode = list_head(&aclp->z_acl); aclnode;
aclnode = list_next(&aclp->z_acl, aclnode)) {
if (aclnode->z_ace_count == 0)
continue;
bcopy(aclnode->z_acldata, start,
aclnode->z_size);
start = (caddr_t)start + aclnode->z_size;
}
}
/*
* If Old version then swap count/bytes to match old
* layout of znode_acl_phys_t.
*/
if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
acl_phys.z_acl_size = aclp->z_acl_count;
acl_phys.z_acl_count = aclp->z_acl_bytes;
} else {
acl_phys.z_acl_size = aclp->z_acl_bytes;
acl_phys.z_acl_count = aclp->z_acl_count;
}
acl_phys.z_acl_version = aclp->z_version;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
&acl_phys, sizeof (acl_phys));
}
/*
* Replace ACL wide bits, but first clear them.
*/
zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
zp->z_pflags |= aclp->z_hints;
if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
zp->z_pflags |= ZFS_ACL_TRIVIAL;
zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
}
static void
zfs_acl_chmod(boolean_t isdir, uint64_t mode, boolean_t split, boolean_t trim,
zfs_acl_t *aclp)
{
void *acep = NULL;
uint64_t who;
int new_count, new_bytes;
int ace_size;
int entry_type;
uint16_t iflags, type;
uint32_t access_mask;
zfs_acl_node_t *newnode;
size_t abstract_size = aclp->z_ops->ace_abstract_size();
void *zacep;
trivial_acl_t masks;
new_count = new_bytes = 0;
acl_trivial_access_masks((mode_t)mode, isdir, &masks);
newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
zacep = newnode->z_acldata;
if (masks.allow0) {
zfs_set_ace(aclp, zacep, masks.allow0, ALLOW, -1, ACE_OWNER);
zacep = (void *)((uintptr_t)zacep + abstract_size);
new_count++;
new_bytes += abstract_size;
}
if (masks.deny1) {
zfs_set_ace(aclp, zacep, masks.deny1, DENY, -1, ACE_OWNER);
zacep = (void *)((uintptr_t)zacep + abstract_size);
new_count++;
new_bytes += abstract_size;
}
if (masks.deny2) {
zfs_set_ace(aclp, zacep, masks.deny2, DENY, -1, OWNING_GROUP);
zacep = (void *)((uintptr_t)zacep + abstract_size);
new_count++;
new_bytes += abstract_size;
}
while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
&iflags, &type))) {
entry_type = (iflags & ACE_TYPE_FLAGS);
/*
* ACEs used to represent the file mode may be divided
* into an equivalent pair of inherit-only and regular
* ACEs, if they are inheritable.
* Skip regular ACEs, which are replaced by the new mode.
*/
if (split && (entry_type == ACE_OWNER ||
entry_type == OWNING_GROUP ||
entry_type == ACE_EVERYONE)) {
if (!isdir || !(iflags &
(ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
continue;
/*
* We preserve owner@, group@, or @everyone
* permissions, if they are inheritable, by
* copying them to inherit_only ACEs. This
* prevents inheritable permissions from being
* altered along with the file mode.
*/
iflags |= ACE_INHERIT_ONLY_ACE;
}
/*
* If this ACL has any inheritable ACEs, mark that in
* the hints (which are later masked into the pflags)
* so create knows to do inheritance.
*/
if (isdir && (iflags &
(ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
aclp->z_hints |= ZFS_INHERIT_ACE;
if ((type != ALLOW && type != DENY) ||
(iflags & ACE_INHERIT_ONLY_ACE)) {
switch (type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
aclp->z_hints |= ZFS_ACL_OBJ_ACE;
break;
}
} else {
/*
* Limit permissions to be no greater than
* group permissions.
* The "aclinherit" and "aclmode" properties
* affect policy for create and chmod(2),
* respectively.
*/
if ((type == ALLOW) && trim)
access_mask &= masks.group;
}
zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
ace_size = aclp->z_ops->ace_size(acep);
zacep = (void *)((uintptr_t)zacep + ace_size);
new_count++;
new_bytes += ace_size;
}
zfs_set_ace(aclp, zacep, masks.owner, ALLOW, -1, ACE_OWNER);
zacep = (void *)((uintptr_t)zacep + abstract_size);
zfs_set_ace(aclp, zacep, masks.group, ALLOW, -1, OWNING_GROUP);
zacep = (void *)((uintptr_t)zacep + abstract_size);
zfs_set_ace(aclp, zacep, masks.everyone, ALLOW, -1, ACE_EVERYONE);
new_count += 3;
new_bytes += abstract_size * 3;
zfs_acl_release_nodes(aclp);
aclp->z_acl_count = new_count;
aclp->z_acl_bytes = new_bytes;
newnode->z_ace_count = new_count;
newnode->z_size = new_bytes;
list_insert_tail(&aclp->z_acl, newnode);
}
int
zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
{
int error = 0;
mutex_enter(&zp->z_acl_lock);
mutex_enter(&zp->z_lock);
if (ZTOZSB(zp)->z_acl_mode == ZFS_ACL_DISCARD)
*aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
else
error = zfs_acl_node_read(zp, B_TRUE, aclp, B_TRUE);
if (error == 0) {
(*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
zfs_acl_chmod(S_ISDIR(ZTOI(zp)->i_mode), mode, B_TRUE,
(ZTOZSB(zp)->z_acl_mode == ZFS_ACL_GROUPMASK), *aclp);
}
mutex_exit(&zp->z_lock);
mutex_exit(&zp->z_acl_lock);
return (error);
}
/*
* Should ACE be inherited?
*/
static int
zfs_ace_can_use(umode_t obj_mode, uint16_t acep_flags)
{
int iflags = (acep_flags & 0xf);
if (S_ISDIR(obj_mode) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
return (1);
else if (iflags & ACE_FILE_INHERIT_ACE)
return (!(S_ISDIR(obj_mode) &&
(iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
return (0);
}
/*
* inherit inheritable ACEs from parent
*/
static zfs_acl_t *
zfs_acl_inherit(zfsvfs_t *zfsvfs, umode_t va_mode, zfs_acl_t *paclp,
uint64_t mode, boolean_t *need_chmod)
{
void *pacep = NULL;
void *acep;
zfs_acl_node_t *aclnode;
zfs_acl_t *aclp = NULL;
uint64_t who;
uint32_t access_mask;
uint16_t iflags, newflags, type;
size_t ace_size;
void *data1, *data2;
size_t data1sz, data2sz;
uint_t aclinherit;
boolean_t isdir = S_ISDIR(va_mode);
boolean_t isreg = S_ISREG(va_mode);
*need_chmod = B_TRUE;
aclp = zfs_acl_alloc(paclp->z_version);
aclinherit = zfsvfs->z_acl_inherit;
if (aclinherit == ZFS_ACL_DISCARD || S_ISLNK(va_mode))
return (aclp);
while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
&access_mask, &iflags, &type))) {
/*
* don't inherit bogus ACEs
*/
if (!zfs_acl_valid_ace_type(type, iflags))
continue;
/*
* Check if ACE is inheritable by this vnode
*/
if ((aclinherit == ZFS_ACL_NOALLOW && type == ALLOW) ||
!zfs_ace_can_use(va_mode, iflags))
continue;
/*
* If owner@, group@, or everyone@ inheritable
* then zfs_acl_chmod() isn't needed.
*/
if ((aclinherit == ZFS_ACL_PASSTHROUGH ||
aclinherit == ZFS_ACL_PASSTHROUGH_X) &&
((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
((iflags & OWNING_GROUP) == OWNING_GROUP)) &&
(isreg || (isdir && (iflags & ACE_DIRECTORY_INHERIT_ACE))))
*need_chmod = B_FALSE;
/*
* Strip inherited execute permission from file if
* not in mode
*/
if (aclinherit == ZFS_ACL_PASSTHROUGH_X && type == ALLOW &&
!isdir && ((mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0)) {
access_mask &= ~ACE_EXECUTE;
}
/*
* Strip write_acl and write_owner from permissions
* when inheriting an ACE
*/
if (aclinherit == ZFS_ACL_RESTRICTED && type == ALLOW) {
access_mask &= ~RESTRICTED_CLEAR;
}
ace_size = aclp->z_ops->ace_size(pacep);
aclnode = zfs_acl_node_alloc(ace_size);
list_insert_tail(&aclp->z_acl, aclnode);
acep = aclnode->z_acldata;
zfs_set_ace(aclp, acep, access_mask, type,
who, iflags|ACE_INHERITED_ACE);
/*
* Copy special opaque data if any
*/
if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
VERIFY((data2sz = aclp->z_ops->ace_data(acep,
&data2)) == data1sz);
bcopy(data1, data2, data2sz);
}
aclp->z_acl_count++;
aclnode->z_ace_count++;
aclp->z_acl_bytes += aclnode->z_size;
newflags = aclp->z_ops->ace_flags_get(acep);
/*
* If ACE is not to be inherited further, or if the vnode is
* not a directory, remove all inheritance flags
*/
if (!isdir || (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)) {
newflags &= ~ALL_INHERIT;
aclp->z_ops->ace_flags_set(acep,
newflags|ACE_INHERITED_ACE);
continue;
}
/*
* This directory has an inheritable ACE
*/
aclp->z_hints |= ZFS_INHERIT_ACE;
/*
* If only FILE_INHERIT is set then turn on
* inherit_only
*/
if ((iflags & (ACE_FILE_INHERIT_ACE |
ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
newflags |= ACE_INHERIT_ONLY_ACE;
aclp->z_ops->ace_flags_set(acep,
newflags|ACE_INHERITED_ACE);
} else {
newflags &= ~ACE_INHERIT_ONLY_ACE;
aclp->z_ops->ace_flags_set(acep,
newflags|ACE_INHERITED_ACE);
}
}
if (zfsvfs->z_acl_mode == ZFS_ACL_RESTRICTED &&
aclp->z_acl_count != 0) {
*need_chmod = B_FALSE;
}
return (aclp);
}
/*
* Create file system object initial permissions
* including inheritable ACEs.
* Also, create FUIDs for owner and group.
*/
int
zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
{
int error;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
zfs_acl_t *paclp;
gid_t gid = vap->va_gid;
boolean_t need_chmod = B_TRUE;
boolean_t trim = B_FALSE;
boolean_t inherited = B_FALSE;
bzero(acl_ids, sizeof (zfs_acl_ids_t));
acl_ids->z_mode = vap->va_mode;
if (vsecp)
if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_mode, vsecp,
cr, &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
return (error);
acl_ids->z_fuid = vap->va_uid;
acl_ids->z_fgid = vap->va_gid;
#ifdef HAVE_KSID
/*
* Determine uid and gid.
*/
if ((flag & IS_ROOT_NODE) || zfsvfs->z_replay ||
((flag & IS_XATTR) && (S_ISDIR(vap->va_mode)))) {
acl_ids->z_fuid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_uid,
cr, ZFS_OWNER, &acl_ids->z_fuidp);
acl_ids->z_fgid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
cr, ZFS_GROUP, &acl_ids->z_fuidp);
gid = vap->va_gid;
} else {
acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER,
cr, &acl_ids->z_fuidp);
acl_ids->z_fgid = 0;
if (vap->va_mask & AT_GID) {
acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
(uint64_t)vap->va_gid,
cr, ZFS_GROUP, &acl_ids->z_fuidp);
gid = vap->va_gid;
if (acl_ids->z_fgid != KGID_TO_SGID(ZTOI(dzp)->i_gid) &&
!groupmember(vap->va_gid, cr) &&
secpolicy_vnode_create_gid(cr) != 0)
acl_ids->z_fgid = 0;
}
if (acl_ids->z_fgid == 0) {
if (dzp->z_mode & S_ISGID) {
char *domain;
uint32_t rid;
acl_ids->z_fgid = KGID_TO_SGID(
ZTOI(dzp)->i_gid);
gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
cr, ZFS_GROUP);
if (zfsvfs->z_use_fuids &&
IS_EPHEMERAL(acl_ids->z_fgid)) {
domain = zfs_fuid_idx_domain(
&zfsvfs->z_fuid_idx,
FUID_INDEX(acl_ids->z_fgid));
rid = FUID_RID(acl_ids->z_fgid);
zfs_fuid_node_add(&acl_ids->z_fuidp,
domain, rid,
FUID_INDEX(acl_ids->z_fgid),
acl_ids->z_fgid, ZFS_GROUP);
}
} else {
acl_ids->z_fgid = zfs_fuid_create_cred(zfsvfs,
ZFS_GROUP, cr, &acl_ids->z_fuidp);
gid = crgetgid(cr);
}
}
}
#endif /* HAVE_KSID */
/*
* If we're creating a directory, and the parent directory has the
* set-GID bit set, set in on the new directory.
* Otherwise, if the user is neither privileged nor a member of the
* file's new group, clear the file's set-GID bit.
*/
if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
(S_ISDIR(vap->va_mode))) {
acl_ids->z_mode |= S_ISGID;
} else {
if ((acl_ids->z_mode & S_ISGID) &&
secpolicy_vnode_setids_setgids(cr, gid) != 0)
acl_ids->z_mode &= ~S_ISGID;
}
if (acl_ids->z_aclp == NULL) {
mutex_enter(&dzp->z_acl_lock);
mutex_enter(&dzp->z_lock);
if (!(flag & IS_ROOT_NODE) &&
(dzp->z_pflags & ZFS_INHERIT_ACE) &&
!(dzp->z_pflags & ZFS_XATTR)) {
VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
&paclp, B_FALSE));
acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
vap->va_mode, paclp, acl_ids->z_mode, &need_chmod);
inherited = B_TRUE;
} else {
acl_ids->z_aclp =
zfs_acl_alloc(zfs_acl_version_zp(dzp));
acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
}
mutex_exit(&dzp->z_lock);
mutex_exit(&dzp->z_acl_lock);
if (need_chmod) {
if (S_ISDIR(vap->va_mode))
acl_ids->z_aclp->z_hints |=
ZFS_ACL_AUTO_INHERIT;
if (zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK &&
zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH &&
zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH_X)
trim = B_TRUE;
zfs_acl_chmod(vap->va_mode, acl_ids->z_mode, B_FALSE,
trim, acl_ids->z_aclp);
}
}
if (inherited || vsecp) {
acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
acl_ids->z_fuid, acl_ids->z_fgid);
if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
}
return (0);
}
/*
* Free ACL and fuid_infop, but not the acl_ids structure
*/
void
zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
{
if (acl_ids->z_aclp)
zfs_acl_free(acl_ids->z_aclp);
if (acl_ids->z_fuidp)
zfs_fuid_info_free(acl_ids->z_fuidp);
acl_ids->z_aclp = NULL;
acl_ids->z_fuidp = NULL;
}
boolean_t
zfs_acl_ids_overquota(zfsvfs_t *zv, zfs_acl_ids_t *acl_ids, uint64_t projid)
{
return (zfs_id_overquota(zv, DMU_USERUSED_OBJECT, acl_ids->z_fuid) ||
zfs_id_overquota(zv, DMU_GROUPUSED_OBJECT, acl_ids->z_fgid) ||
(projid != ZFS_DEFAULT_PROJID && projid != ZFS_INVALID_PROJID &&
zfs_id_overquota(zv, DMU_PROJECTUSED_OBJECT, projid)));
}
/*
* Retrieve a file's ACL
*/
int
zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
{
zfs_acl_t *aclp;
ulong_t mask;
int error;
int count = 0;
int largeace = 0;
mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
if (mask == 0)
return (SET_ERROR(ENOSYS));
if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
return (error);
mutex_enter(&zp->z_acl_lock);
error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
if (error != 0) {
mutex_exit(&zp->z_acl_lock);
return (error);
}
/*
* Scan ACL to determine number of ACEs
*/
if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
void *zacep = NULL;
uint64_t who;
uint32_t access_mask;
uint16_t type, iflags;
while ((zacep = zfs_acl_next_ace(aclp, zacep,
&who, &access_mask, &iflags, &type))) {
switch (type) {
case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
largeace++;
continue;
default:
count++;
}
}
vsecp->vsa_aclcnt = count;
} else
count = (int)aclp->z_acl_count;
if (mask & VSA_ACECNT) {
vsecp->vsa_aclcnt = count;
}
if (mask & VSA_ACE) {
size_t aclsz;
aclsz = count * sizeof (ace_t) +
sizeof (ace_object_t) * largeace;
vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
vsecp->vsa_aclentsz = aclsz;
if (aclp->z_version == ZFS_ACL_VERSION_FUID)
zfs_copy_fuid_2_ace(ZTOZSB(zp), aclp, cr,
vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
else {
zfs_acl_node_t *aclnode;
void *start = vsecp->vsa_aclentp;
for (aclnode = list_head(&aclp->z_acl); aclnode;
aclnode = list_next(&aclp->z_acl, aclnode)) {
bcopy(aclnode->z_acldata, start,
aclnode->z_size);
start = (caddr_t)start + aclnode->z_size;
}
ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
aclp->z_acl_bytes);
}
}
if (mask & VSA_ACE_ACLFLAGS) {
vsecp->vsa_aclflags = 0;
if (zp->z_pflags & ZFS_ACL_DEFAULTED)
vsecp->vsa_aclflags |= ACL_DEFAULTED;
if (zp->z_pflags & ZFS_ACL_PROTECTED)
vsecp->vsa_aclflags |= ACL_PROTECTED;
if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
}
mutex_exit(&zp->z_acl_lock);
return (0);
}
int
zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, umode_t obj_mode,
vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
{
zfs_acl_t *aclp;
zfs_acl_node_t *aclnode;
int aclcnt = vsecp->vsa_aclcnt;
int error;
if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
return (SET_ERROR(EINVAL));
aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
aclp->z_hints = 0;
aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
if ((error = zfs_copy_ace_2_oldace(obj_mode, aclp,
(ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
aclcnt, &aclnode->z_size)) != 0) {
zfs_acl_free(aclp);
zfs_acl_node_free(aclnode);
return (error);
}
} else {
if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_mode, aclp,
vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
&aclnode->z_size, fuidp, cr)) != 0) {
zfs_acl_free(aclp);
zfs_acl_node_free(aclnode);
return (error);
}
}
aclp->z_acl_bytes = aclnode->z_size;
aclnode->z_ace_count = aclcnt;
aclp->z_acl_count = aclcnt;
list_insert_head(&aclp->z_acl, aclnode);
/*
* If flags are being set then add them to z_hints
*/
if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
if (vsecp->vsa_aclflags & ACL_PROTECTED)
aclp->z_hints |= ZFS_ACL_PROTECTED;
if (vsecp->vsa_aclflags & ACL_DEFAULTED)
aclp->z_hints |= ZFS_ACL_DEFAULTED;
if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
}
*zaclp = aclp;
return (0);
}
/*
* Set a file's ACL
*/
int
zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
zilog_t *zilog = zfsvfs->z_log;
ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
dmu_tx_t *tx;
int error;
zfs_acl_t *aclp;
zfs_fuid_info_t *fuidp = NULL;
boolean_t fuid_dirtied;
uint64_t acl_obj;
if (mask == 0)
return (SET_ERROR(ENOSYS));
if (zp->z_pflags & ZFS_IMMUTABLE)
return (SET_ERROR(EPERM));
if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
return (error);
error = zfs_vsec_2_aclp(zfsvfs, ZTOI(zp)->i_mode, vsecp, cr, &fuidp,
&aclp);
if (error)
return (error);
/*
* If ACL wide flags aren't being set then preserve any
* existing flags.
*/
if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
aclp->z_hints |=
(zp->z_pflags & V4_ACL_WIDE_FLAGS);
}
top:
mutex_enter(&zp->z_acl_lock);
mutex_enter(&zp->z_lock);
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
fuid_dirtied = zfsvfs->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
/*
* If old version and ACL won't fit in bonus and we aren't
* upgrading then take out necessary DMU holds
*/
if ((acl_obj = zfs_external_acl(zp)) != 0) {
if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
dmu_tx_hold_free(tx, acl_obj, 0,
DMU_OBJECT_END);
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
aclp->z_acl_bytes);
} else {
dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
}
} else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
}
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_NOWAIT);
if (error) {
mutex_exit(&zp->z_acl_lock);
mutex_exit(&zp->z_lock);
if (error == ERESTART) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
zfs_acl_free(aclp);
return (error);
}
error = zfs_aclset_common(zp, aclp, cr, tx);
ASSERT(error == 0);
ASSERT(zp->z_acl_cached == NULL);
zp->z_acl_cached = aclp;
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
if (fuidp)
zfs_fuid_info_free(fuidp);
dmu_tx_commit(tx);
mutex_exit(&zp->z_lock);
mutex_exit(&zp->z_acl_lock);
return (error);
}
/*
* Check accesses of interest (AoI) against attributes of the dataset
* such as read-only. Returns zero if no AoI conflict with dataset
* attributes, otherwise an appropriate errno is returned.
*/
static int
zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
{
if ((v4_mode & WRITE_MASK) && (zfs_is_readonly(ZTOZSB(zp))) &&
(!Z_ISDEV(ZTOI(zp)->i_mode) ||
(Z_ISDEV(ZTOI(zp)->i_mode) && (v4_mode & WRITE_MASK_ATTRS)))) {
return (SET_ERROR(EROFS));
}
/*
* Intentionally allow ZFS_READONLY through here.
* See zfs_zaccess_common().
*/
if ((v4_mode & WRITE_MASK_DATA) &&
(zp->z_pflags & ZFS_IMMUTABLE)) {
return (SET_ERROR(EPERM));
}
if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
(zp->z_pflags & ZFS_NOUNLINK)) {
return (SET_ERROR(EPERM));
}
if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
(zp->z_pflags & ZFS_AV_QUARANTINED))) {
return (SET_ERROR(EACCES));
}
return (0);
}
/*
* The primary usage of this function is to loop through all of the
* ACEs in the znode, determining what accesses of interest (AoI) to
* the caller are allowed or denied. The AoI are expressed as bits in
* the working_mode parameter. As each ACE is processed, bits covered
* by that ACE are removed from the working_mode. This removal
* facilitates two things. The first is that when the working mode is
* empty (= 0), we know we've looked at all the AoI. The second is
* that the ACE interpretation rules don't allow a later ACE to undo
* something granted or denied by an earlier ACE. Removing the
* discovered access or denial enforces this rule. At the end of
* processing the ACEs, all AoI that were found to be denied are
* placed into the working_mode, giving the caller a mask of denied
* accesses. Returns:
* 0 if all AoI granted
* EACCES if the denied mask is non-zero
* other error if abnormal failure (e.g., IO error)
*
* A secondary usage of the function is to determine if any of the
* AoI are granted. If an ACE grants any access in
* the working_mode, we immediately short circuit out of the function.
* This mode is chosen by setting anyaccess to B_TRUE. The
* working_mode is not a denied access mask upon exit if the function
* is used in this manner.
*/
static int
zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
boolean_t anyaccess, cred_t *cr)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
zfs_acl_t *aclp;
int error;
uid_t uid = crgetuid(cr);
uint64_t who;
uint16_t type, iflags;
uint16_t entry_type;
uint32_t access_mask;
uint32_t deny_mask = 0;
zfs_ace_hdr_t *acep = NULL;
boolean_t checkit;
uid_t gowner;
uid_t fowner;
zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
mutex_enter(&zp->z_acl_lock);
error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
if (error != 0) {
mutex_exit(&zp->z_acl_lock);
return (error);
}
ASSERT(zp->z_acl_cached);
while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
&iflags, &type))) {
uint32_t mask_matched;
if (!zfs_acl_valid_ace_type(type, iflags))
continue;
if (S_ISDIR(ZTOI(zp)->i_mode) &&
(iflags & ACE_INHERIT_ONLY_ACE))
continue;
/* Skip ACE if it does not affect any AoI */
mask_matched = (access_mask & *working_mode);
if (!mask_matched)
continue;
entry_type = (iflags & ACE_TYPE_FLAGS);
checkit = B_FALSE;
switch (entry_type) {
case ACE_OWNER:
if (uid == fowner)
checkit = B_TRUE;
break;
case OWNING_GROUP:
who = gowner;
fallthrough;
case ACE_IDENTIFIER_GROUP:
checkit = zfs_groupmember(zfsvfs, who, cr);
break;
case ACE_EVERYONE:
checkit = B_TRUE;
break;
/* USER Entry */
default:
if (entry_type == 0) {
uid_t newid;
newid = zfs_fuid_map_id(zfsvfs, who, cr,
ZFS_ACE_USER);
if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
uid == newid)
checkit = B_TRUE;
break;
} else {
mutex_exit(&zp->z_acl_lock);
return (SET_ERROR(EIO));
}
}
if (checkit) {
if (type == DENY) {
DTRACE_PROBE3(zfs__ace__denies,
znode_t *, zp,
zfs_ace_hdr_t *, acep,
uint32_t, mask_matched);
deny_mask |= mask_matched;
} else {
DTRACE_PROBE3(zfs__ace__allows,
znode_t *, zp,
zfs_ace_hdr_t *, acep,
uint32_t, mask_matched);
if (anyaccess) {
mutex_exit(&zp->z_acl_lock);
return (0);
}
}
*working_mode &= ~mask_matched;
}
/* Are we done? */
if (*working_mode == 0)
break;
}
mutex_exit(&zp->z_acl_lock);
/* Put the found 'denies' back on the working mode */
if (deny_mask) {
*working_mode |= deny_mask;
return (SET_ERROR(EACCES));
} else if (*working_mode) {
return (-1);
}
return (0);
}
/*
* Return true if any access whatsoever granted, we don't actually
* care what access is granted.
*/
boolean_t
zfs_has_access(znode_t *zp, cred_t *cr)
{
uint32_t have = ACE_ALL_PERMS;
if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
uid_t owner;
owner = zfs_fuid_map_id(ZTOZSB(zp),
KUID_TO_SUID(ZTOI(zp)->i_uid), cr, ZFS_OWNER);
return (secpolicy_vnode_any_access(cr, ZTOI(zp), owner) == 0);
}
return (B_TRUE);
}
static int
zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
int err;
*working_mode = v4_mode;
*check_privs = B_TRUE;
/*
* Short circuit empty requests
*/
if (v4_mode == 0 || zfsvfs->z_replay) {
*working_mode = 0;
return (0);
}
if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
*check_privs = B_FALSE;
return (err);
}
/*
* The caller requested that the ACL check be skipped. This
* would only happen if the caller checked VOP_ACCESS() with a
* 32 bit ACE mask and already had the appropriate permissions.
*/
if (skipaclchk) {
*working_mode = 0;
return (0);
}
/*
* Note: ZFS_READONLY represents the "DOS R/O" attribute.
* When that flag is set, we should behave as if write access
* were not granted by anything in the ACL. In particular:
* We _must_ allow writes after opening the file r/w, then
* setting the DOS R/O attribute, and writing some more.
* (Similar to how you can write after fchmod(fd, 0444).)
*
* Therefore ZFS_READONLY is ignored in the dataset check
* above, and checked here as if part of the ACL check.
* Also note: DOS R/O is ignored for directories.
*/
if ((v4_mode & WRITE_MASK_DATA) &&
S_ISDIR(ZTOI(zp)->i_mode) &&
(zp->z_pflags & ZFS_READONLY)) {
return (SET_ERROR(EPERM));
}
return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
}
static int
zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
cred_t *cr)
{
if (*working_mode != ACE_WRITE_DATA)
return (SET_ERROR(EACCES));
return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
check_privs, B_FALSE, cr));
}
int
zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
{
boolean_t owner = B_FALSE;
boolean_t groupmbr = B_FALSE;
boolean_t is_attr;
uid_t uid = crgetuid(cr);
int error;
if (zdp->z_pflags & ZFS_AV_QUARANTINED)
return (SET_ERROR(EACCES));
is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
(S_ISDIR(ZTOI(zdp)->i_mode)));
if (is_attr)
goto slow;
mutex_enter(&zdp->z_acl_lock);
if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
mutex_exit(&zdp->z_acl_lock);
return (0);
}
if (KUID_TO_SUID(ZTOI(zdp)->i_uid) != 0 ||
KGID_TO_SGID(ZTOI(zdp)->i_gid) != 0) {
mutex_exit(&zdp->z_acl_lock);
goto slow;
}
if (uid == KUID_TO_SUID(ZTOI(zdp)->i_uid)) {
owner = B_TRUE;
if (zdp->z_mode & S_IXUSR) {
mutex_exit(&zdp->z_acl_lock);
return (0);
} else {
mutex_exit(&zdp->z_acl_lock);
goto slow;
}
}
if (groupmember(KGID_TO_SGID(ZTOI(zdp)->i_gid), cr)) {
groupmbr = B_TRUE;
if (zdp->z_mode & S_IXGRP) {
mutex_exit(&zdp->z_acl_lock);
return (0);
} else {
mutex_exit(&zdp->z_acl_lock);
goto slow;
}
}
if (!owner && !groupmbr) {
if (zdp->z_mode & S_IXOTH) {
mutex_exit(&zdp->z_acl_lock);
return (0);
}
}
mutex_exit(&zdp->z_acl_lock);
slow:
DTRACE_PROBE(zfs__fastpath__execute__access__miss);
ZFS_ENTER(ZTOZSB(zdp));
error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
ZFS_EXIT(ZTOZSB(zdp));
return (error);
}
/*
* Determine whether Access should be granted/denied.
*
* The least priv subsystem is always consulted as a basic privilege
* can define any form of access.
*/
int
zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
{
uint32_t working_mode;
int error;
int is_attr;
boolean_t check_privs;
znode_t *xzp;
znode_t *check_zp = zp;
mode_t needed_bits;
uid_t owner;
is_attr = ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode));
/*
* If attribute then validate against base file
*/
if (is_attr) {
if ((error = zfs_zget(ZTOZSB(zp),
zp->z_xattr_parent, &xzp)) != 0) {
return (error);
}
check_zp = xzp;
/*
* fixup mode to map to xattr perms
*/
if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
mode |= ACE_WRITE_NAMED_ATTRS;
}
if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
mode |= ACE_READ_NAMED_ATTRS;
}
}
owner = zfs_fuid_map_id(ZTOZSB(zp), KUID_TO_SUID(ZTOI(zp)->i_uid),
cr, ZFS_OWNER);
/*
* Map the bits required to the standard inode flags
* S_IRUSR|S_IWUSR|S_IXUSR in the needed_bits. Map the bits
* mapped by working_mode (currently missing) in missing_bits.
* Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
* needed_bits.
*/
needed_bits = 0;
working_mode = mode;
if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
owner == crgetuid(cr))
working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
needed_bits |= S_IRUSR;
if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
needed_bits |= S_IWUSR;
if (working_mode & ACE_EXECUTE)
needed_bits |= S_IXUSR;
if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
&check_privs, skipaclchk, cr)) == 0) {
if (is_attr)
zrele(xzp);
return (secpolicy_vnode_access2(cr, ZTOI(zp), owner,
needed_bits, needed_bits));
}
if (error && !check_privs) {
if (is_attr)
zrele(xzp);
return (error);
}
if (error && (flags & V_APPEND)) {
error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
}
if (error && check_privs) {
mode_t checkmode = 0;
/*
* First check for implicit owner permission on
* read_acl/read_attributes
*/
error = 0;
ASSERT(working_mode != 0);
if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
owner == crgetuid(cr)))
working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
checkmode |= S_IRUSR;
if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
checkmode |= S_IWUSR;
if (working_mode & ACE_EXECUTE)
checkmode |= S_IXUSR;
error = secpolicy_vnode_access2(cr, ZTOI(check_zp), owner,
needed_bits & ~checkmode, needed_bits);
if (error == 0 && (working_mode & ACE_WRITE_OWNER))
error = secpolicy_vnode_chown(cr, owner);
if (error == 0 && (working_mode & ACE_WRITE_ACL))
error = secpolicy_vnode_setdac(cr, owner);
if (error == 0 && (working_mode &
(ACE_DELETE|ACE_DELETE_CHILD)))
error = secpolicy_vnode_remove(cr);
if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
error = secpolicy_vnode_chown(cr, owner);
}
if (error == 0) {
/*
* See if any bits other than those already checked
* for are still present. If so then return EACCES
*/
if (working_mode & ~(ZFS_CHECKED_MASKS)) {
error = SET_ERROR(EACCES);
}
}
} else if (error == 0) {
error = secpolicy_vnode_access2(cr, ZTOI(zp), owner,
needed_bits, needed_bits);
}
if (is_attr)
zrele(xzp);
return (error);
}
/*
* Translate traditional unix S_IRUSR/S_IWUSR/S_IXUSR mode into
* NFSv4-style ZFS ACL format and call zfs_zaccess()
*/
int
zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
{
return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
}
/*
* Access function for secpolicy_vnode_setattr
*/
int
zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
{
int v4_mode = zfs_unix_to_v4(mode >> 6);
return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
}
/* See zfs_zaccess_delete() */
-int zfs_write_implies_delete_child = 1;
+static const boolean_t zfs_write_implies_delete_child = B_TRUE;
/*
* Determine whether delete access should be granted.
*
* The following chart outlines how we handle delete permissions which is
* how recent versions of windows (Windows 2008) handles it. The efficiency
* comes from not having to check the parent ACL where the object itself grants
* delete:
*
* -------------------------------------------------------
* | Parent Dir | Target Object Permissions |
* | permissions | |
* -------------------------------------------------------
* | | ACL Allows | ACL Denies| Delete |
* | | Delete | Delete | unspecified|
* -------------------------------------------------------
* | ACL Allows | Permit | Deny * | Permit |
* | DELETE_CHILD | | | |
* -------------------------------------------------------
* | ACL Denies | Permit | Deny | Deny |
* | DELETE_CHILD | | | |
* -------------------------------------------------------
* | ACL specifies | | | |
* | only allow | Permit | Deny * | Permit |
* | write and | | | |
* | execute | | | |
* -------------------------------------------------------
* | ACL denies | | | |
* | write and | Permit | Deny | Deny |
* | execute | | | |
* -------------------------------------------------------
* ^
* |
* Re. execute permission on the directory: if that's missing,
* the vnode lookup of the target will fail before we get here.
*
* Re [*] in the table above: NFSv4 would normally Permit delete for
* these two cells of the matrix.
* See acl.h for notes on which ACE_... flags should be checked for which
* operations. Specifically, the NFSv4 committee recommendation is in
* conflict with the Windows interpretation of DENY ACEs, where DENY ACEs
* should take precedence ahead of ALLOW ACEs.
*
* This implementation always consults the target object's ACL first.
* If a DENY ACE is present on the target object that specifies ACE_DELETE,
* delete access is denied. If an ALLOW ACE with ACE_DELETE is present on
* the target object, access is allowed. If and only if no entries with
* ACE_DELETE are present in the object's ACL, check the container's ACL
* for entries with ACE_DELETE_CHILD.
*
* A summary of the logic implemented from the table above is as follows:
*
* First check for DENY ACEs that apply.
* If either target or container has a deny, EACCES.
*
* Delete access can then be summarized as follows:
* 1: The object to be deleted grants ACE_DELETE, or
* 2: The containing directory grants ACE_DELETE_CHILD.
* In a Windows system, that would be the end of the story.
* In this system, (2) has some complications...
* 2a: "sticky" bit on a directory adds restrictions, and
* 2b: existing ACEs from previous versions of ZFS may
* not carry ACE_DELETE_CHILD where they should, so we
* also allow delete when ACE_WRITE_DATA is granted.
*
* Note: 2b is technically a work-around for a prior bug,
* which hopefully can go away some day. For those who
* no longer need the work around, and for testing, this
* work-around is made conditional via the tunable:
* zfs_write_implies_delete_child
*/
int
zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
{
uint32_t wanted_dirperms;
uint32_t dzp_working_mode = 0;
uint32_t zp_working_mode = 0;
int dzp_error, zp_error;
boolean_t dzpcheck_privs;
boolean_t zpcheck_privs;
if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
return (SET_ERROR(EPERM));
/*
* Case 1:
* If target object grants ACE_DELETE then we are done. This is
* indicated by a return value of 0. For this case we don't worry
* about the sticky bit because sticky only applies to the parent
* directory and this is the child access result.
*
* If we encounter a DENY ACE here, we're also done (EACCES).
* Note that if we hit a DENY ACE here (on the target) it should
* take precedence over a DENY ACE on the container, so that when
* we have more complete auditing support we will be able to
* report an access failure against the specific target.
* (This is part of why we're checking the target first.)
*/
zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
&zpcheck_privs, B_FALSE, cr);
if (zp_error == EACCES) {
/* We hit a DENY ACE. */
if (!zpcheck_privs)
return (SET_ERROR(zp_error));
return (secpolicy_vnode_remove(cr));
}
if (zp_error == 0)
return (0);
/*
* Case 2:
* If the containing directory grants ACE_DELETE_CHILD,
* or we're in backward compatibility mode and the
* containing directory has ACE_WRITE_DATA, allow.
* Case 2b is handled with wanted_dirperms.
*/
wanted_dirperms = ACE_DELETE_CHILD;
if (zfs_write_implies_delete_child)
wanted_dirperms |= ACE_WRITE_DATA;
dzp_error = zfs_zaccess_common(dzp, wanted_dirperms,
&dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
if (dzp_error == EACCES) {
/* We hit a DENY ACE. */
if (!dzpcheck_privs)
return (SET_ERROR(dzp_error));
return (secpolicy_vnode_remove(cr));
}
/*
* Cases 2a, 2b (continued)
*
* Note: dzp_working_mode now contains any permissions
* that were NOT granted. Therefore, if any of the
* wanted_dirperms WERE granted, we will have:
* dzp_working_mode != wanted_dirperms
* We're really asking if ANY of those permissions
* were granted, and if so, grant delete access.
*/
if (dzp_working_mode != wanted_dirperms)
dzp_error = 0;
/*
* dzp_error is 0 if the container granted us permissions to "modify".
* If we do not have permission via one or more ACEs, our current
* privileges may still permit us to modify the container.
*
* dzpcheck_privs is false when i.e. the FS is read-only.
* Otherwise, do privilege checks for the container.
*/
if (dzp_error != 0 && dzpcheck_privs) {
uid_t owner;
/*
* The secpolicy call needs the requested access and
* the current access mode of the container, but it
* only knows about Unix-style modes (VEXEC, VWRITE),
* so this must condense the fine-grained ACE bits into
* Unix modes.
*
* The VEXEC flag is easy, because we know that has
* always been checked before we get here (during the
* lookup of the target vnode). The container has not
* granted us permissions to "modify", so we do not set
* the VWRITE flag in the current access mode.
*/
owner = zfs_fuid_map_id(ZTOZSB(dzp),
KUID_TO_SUID(ZTOI(dzp)->i_uid), cr, ZFS_OWNER);
dzp_error = secpolicy_vnode_access2(cr, ZTOI(dzp),
owner, S_IXUSR, S_IWUSR|S_IXUSR);
}
if (dzp_error != 0) {
/*
* Note: We may have dzp_error = -1 here (from
* zfs_zacess_common). Don't return that.
*/
return (SET_ERROR(EACCES));
}
/*
* At this point, we know that the directory permissions allow
* us to modify, but we still need to check for the additional
* restrictions that apply when the "sticky bit" is set.
*
* Yes, zfs_sticky_remove_access() also checks this bit, but
* checking it here and skipping the call below is nice when
* you're watching all of this with dtrace.
*/
if ((dzp->z_mode & S_ISVTX) == 0)
return (0);
/*
* zfs_sticky_remove_access will succeed if:
* 1. The sticky bit is absent.
* 2. We pass the sticky bit restrictions.
* 3. We have privileges that always allow file removal.
*/
return (zfs_sticky_remove_access(dzp, zp, cr));
}
int
zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
znode_t *tzp, cred_t *cr)
{
int add_perm;
int error;
if (szp->z_pflags & ZFS_AV_QUARANTINED)
return (SET_ERROR(EACCES));
add_perm = S_ISDIR(ZTOI(szp)->i_mode) ?
ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
/*
* Rename permissions are combination of delete permission +
* add file/subdir permission.
*/
/*
* first make sure we do the delete portion.
*
* If that succeeds then check for add_file/add_subdir permissions
*/
if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
return (error);
/*
* If we have a tzp, see if we can delete it?
*/
if (tzp) {
if ((error = zfs_zaccess_delete(tdzp, tzp, cr)))
return (error);
}
/*
* Now check for add permissions
*/
error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
return (error);
}
diff --git a/module/os/linux/zfs/zfs_ctldir.c b/module/os/linux/zfs/zfs_ctldir.c
index c58d851d77d3..f7e71461a3bd 100644
--- a/module/os/linux/zfs/zfs_ctldir.c
+++ b/module/os/linux/zfs/zfs_ctldir.c
@@ -1,1278 +1,1278 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (C) 2011 Lawrence Livermore National Security, LLC.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* LLNL-CODE-403049.
* Rewritten for Linux by:
* Rohan Puri <rohan.puri15@gmail.com>
* Brian Behlendorf <behlendorf1@llnl.gov>
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
* Copyright (c) 2018 George Melikov. All Rights Reserved.
* Copyright (c) 2019 Datto, Inc. All rights reserved.
* Copyright (c) 2020 The MathWorks, Inc. All rights reserved.
*/
/*
* ZFS control directory (a.k.a. ".zfs")
*
* This directory provides a common location for all ZFS meta-objects.
* Currently, this is only the 'snapshot' and 'shares' directory, but this may
* expand in the future. The elements are built dynamically, as the hierarchy
* does not actually exist on disk.
*
* For 'snapshot', we don't want to have all snapshots always mounted, because
* this would take up a huge amount of space in /etc/mnttab. We have three
* types of objects:
*
* ctldir ------> snapshotdir -------> snapshot
* |
* |
* V
* mounted fs
*
* The 'snapshot' node contains just enough information to lookup '..' and act
* as a mountpoint for the snapshot. Whenever we lookup a specific snapshot, we
* perform an automount of the underlying filesystem and return the
* corresponding inode.
*
* All mounts are handled automatically by an user mode helper which invokes
* the mount procedure. Unmounts are handled by allowing the mount
* point to expire so the kernel may automatically unmount it.
*
* The '.zfs', '.zfs/snapshot', and all directories created under
* '.zfs/snapshot' (ie: '.zfs/snapshot/<snapname>') all share the same
* zfsvfs_t as the head filesystem (what '.zfs' lives under).
*
* File systems mounted on top of the '.zfs/snapshot/<snapname>' paths
* (ie: snapshots) are complete ZFS filesystems and have their own unique
* zfsvfs_t. However, the fsid reported by these mounts will be the same
* as that used by the parent zfsvfs_t to make NFS happy.
*/
#include <sys/types.h>
#include <sys/param.h>
#include <sys/time.h>
#include <sys/sysmacros.h>
#include <sys/pathname.h>
#include <sys/vfs.h>
#include <sys/zfs_ctldir.h>
#include <sys/zfs_ioctl.h>
#include <sys/zfs_vfsops.h>
#include <sys/zfs_vnops.h>
#include <sys/stat.h>
#include <sys/dmu.h>
#include <sys/dmu_objset.h>
#include <sys/dsl_destroy.h>
#include <sys/dsl_deleg.h>
#include <sys/zpl.h>
#include <sys/mntent.h>
#include "zfs_namecheck.h"
/*
* Two AVL trees are maintained which contain all currently automounted
* snapshots. Every automounted snapshots maps to a single zfs_snapentry_t
* entry which MUST:
*
* - be attached to both trees, and
* - be unique, no duplicate entries are allowed.
*
* The zfs_snapshots_by_name tree is indexed by the full dataset name
* while the zfs_snapshots_by_objsetid tree is indexed by the unique
* objsetid. This allows for fast lookups either by name or objsetid.
*/
static avl_tree_t zfs_snapshots_by_name;
static avl_tree_t zfs_snapshots_by_objsetid;
static krwlock_t zfs_snapshot_lock;
/*
* Control Directory Tunables (.zfs)
*/
int zfs_expire_snapshot = ZFSCTL_EXPIRE_SNAPSHOT;
-int zfs_admin_snapshot = 0;
+static int zfs_admin_snapshot = 0;
typedef struct {
char *se_name; /* full snapshot name */
char *se_path; /* full mount path */
spa_t *se_spa; /* pool spa */
uint64_t se_objsetid; /* snapshot objset id */
struct dentry *se_root_dentry; /* snapshot root dentry */
krwlock_t se_taskqid_lock; /* scheduled unmount taskqid lock */
taskqid_t se_taskqid; /* scheduled unmount taskqid */
avl_node_t se_node_name; /* zfs_snapshots_by_name link */
avl_node_t se_node_objsetid; /* zfs_snapshots_by_objsetid link */
zfs_refcount_t se_refcount; /* reference count */
} zfs_snapentry_t;
static void zfsctl_snapshot_unmount_delay_impl(zfs_snapentry_t *se, int delay);
/*
* Allocate a new zfs_snapentry_t being careful to make a copy of the
* the snapshot name and provided mount point. No reference is taken.
*/
static zfs_snapentry_t *
zfsctl_snapshot_alloc(const char *full_name, const char *full_path, spa_t *spa,
uint64_t objsetid, struct dentry *root_dentry)
{
zfs_snapentry_t *se;
se = kmem_zalloc(sizeof (zfs_snapentry_t), KM_SLEEP);
se->se_name = kmem_strdup(full_name);
se->se_path = kmem_strdup(full_path);
se->se_spa = spa;
se->se_objsetid = objsetid;
se->se_root_dentry = root_dentry;
se->se_taskqid = TASKQID_INVALID;
rw_init(&se->se_taskqid_lock, NULL, RW_DEFAULT, NULL);
zfs_refcount_create(&se->se_refcount);
return (se);
}
/*
* Free a zfs_snapentry_t the caller must ensure there are no active
* references.
*/
static void
zfsctl_snapshot_free(zfs_snapentry_t *se)
{
zfs_refcount_destroy(&se->se_refcount);
kmem_strfree(se->se_name);
kmem_strfree(se->se_path);
rw_destroy(se->se_taskqid_lock);
kmem_free(se, sizeof (zfs_snapentry_t));
}
/*
* Hold a reference on the zfs_snapentry_t.
*/
static void
zfsctl_snapshot_hold(zfs_snapentry_t *se)
{
zfs_refcount_add(&se->se_refcount, NULL);
}
/*
* Release a reference on the zfs_snapentry_t. When the number of
* references drops to zero the structure will be freed.
*/
static void
zfsctl_snapshot_rele(zfs_snapentry_t *se)
{
if (zfs_refcount_remove(&se->se_refcount, NULL) == 0)
zfsctl_snapshot_free(se);
}
/*
* Add a zfs_snapentry_t to both the zfs_snapshots_by_name and
* zfs_snapshots_by_objsetid trees. While the zfs_snapentry_t is part
* of the trees a reference is held.
*/
static void
zfsctl_snapshot_add(zfs_snapentry_t *se)
{
ASSERT(RW_WRITE_HELD(&zfs_snapshot_lock));
zfsctl_snapshot_hold(se);
avl_add(&zfs_snapshots_by_name, se);
avl_add(&zfs_snapshots_by_objsetid, se);
}
/*
* Remove a zfs_snapentry_t from both the zfs_snapshots_by_name and
* zfs_snapshots_by_objsetid trees. Upon removal a reference is dropped,
* this can result in the structure being freed if that was the last
* remaining reference.
*/
static void
zfsctl_snapshot_remove(zfs_snapentry_t *se)
{
ASSERT(RW_WRITE_HELD(&zfs_snapshot_lock));
avl_remove(&zfs_snapshots_by_name, se);
avl_remove(&zfs_snapshots_by_objsetid, se);
zfsctl_snapshot_rele(se);
}
/*
* Snapshot name comparison function for the zfs_snapshots_by_name.
*/
static int
snapentry_compare_by_name(const void *a, const void *b)
{
const zfs_snapentry_t *se_a = a;
const zfs_snapentry_t *se_b = b;
int ret;
ret = strcmp(se_a->se_name, se_b->se_name);
if (ret < 0)
return (-1);
else if (ret > 0)
return (1);
else
return (0);
}
/*
* Snapshot name comparison function for the zfs_snapshots_by_objsetid.
*/
static int
snapentry_compare_by_objsetid(const void *a, const void *b)
{
const zfs_snapentry_t *se_a = a;
const zfs_snapentry_t *se_b = b;
if (se_a->se_spa != se_b->se_spa)
return ((ulong_t)se_a->se_spa < (ulong_t)se_b->se_spa ? -1 : 1);
if (se_a->se_objsetid < se_b->se_objsetid)
return (-1);
else if (se_a->se_objsetid > se_b->se_objsetid)
return (1);
else
return (0);
}
/*
* Find a zfs_snapentry_t in zfs_snapshots_by_name. If the snapname
* is found a pointer to the zfs_snapentry_t is returned and a reference
* taken on the structure. The caller is responsible for dropping the
* reference with zfsctl_snapshot_rele(). If the snapname is not found
* NULL will be returned.
*/
static zfs_snapentry_t *
zfsctl_snapshot_find_by_name(const char *snapname)
{
zfs_snapentry_t *se, search;
ASSERT(RW_LOCK_HELD(&zfs_snapshot_lock));
search.se_name = (char *)snapname;
se = avl_find(&zfs_snapshots_by_name, &search, NULL);
if (se)
zfsctl_snapshot_hold(se);
return (se);
}
/*
* Find a zfs_snapentry_t in zfs_snapshots_by_objsetid given the objset id
* rather than the snapname. In all other respects it behaves the same
* as zfsctl_snapshot_find_by_name().
*/
static zfs_snapentry_t *
zfsctl_snapshot_find_by_objsetid(spa_t *spa, uint64_t objsetid)
{
zfs_snapentry_t *se, search;
ASSERT(RW_LOCK_HELD(&zfs_snapshot_lock));
search.se_spa = spa;
search.se_objsetid = objsetid;
se = avl_find(&zfs_snapshots_by_objsetid, &search, NULL);
if (se)
zfsctl_snapshot_hold(se);
return (se);
}
/*
* Rename a zfs_snapentry_t in the zfs_snapshots_by_name. The structure is
* removed, renamed, and added back to the new correct location in the tree.
*/
static int
zfsctl_snapshot_rename(const char *old_snapname, const char *new_snapname)
{
zfs_snapentry_t *se;
ASSERT(RW_WRITE_HELD(&zfs_snapshot_lock));
se = zfsctl_snapshot_find_by_name(old_snapname);
if (se == NULL)
return (SET_ERROR(ENOENT));
zfsctl_snapshot_remove(se);
kmem_strfree(se->se_name);
se->se_name = kmem_strdup(new_snapname);
zfsctl_snapshot_add(se);
zfsctl_snapshot_rele(se);
return (0);
}
/*
* Delayed task responsible for unmounting an expired automounted snapshot.
*/
static void
snapentry_expire(void *data)
{
zfs_snapentry_t *se = (zfs_snapentry_t *)data;
spa_t *spa = se->se_spa;
uint64_t objsetid = se->se_objsetid;
if (zfs_expire_snapshot <= 0) {
zfsctl_snapshot_rele(se);
return;
}
rw_enter(&se->se_taskqid_lock, RW_WRITER);
se->se_taskqid = TASKQID_INVALID;
rw_exit(&se->se_taskqid_lock);
(void) zfsctl_snapshot_unmount(se->se_name, MNT_EXPIRE);
zfsctl_snapshot_rele(se);
/*
* Reschedule the unmount if the zfs_snapentry_t wasn't removed.
* This can occur when the snapshot is busy.
*/
rw_enter(&zfs_snapshot_lock, RW_READER);
if ((se = zfsctl_snapshot_find_by_objsetid(spa, objsetid)) != NULL) {
zfsctl_snapshot_unmount_delay_impl(se, zfs_expire_snapshot);
zfsctl_snapshot_rele(se);
}
rw_exit(&zfs_snapshot_lock);
}
/*
* Cancel an automatic unmount of a snapname. This callback is responsible
* for dropping the reference on the zfs_snapentry_t which was taken when
* during dispatch.
*/
static void
zfsctl_snapshot_unmount_cancel(zfs_snapentry_t *se)
{
int err = 0;
rw_enter(&se->se_taskqid_lock, RW_WRITER);
err = taskq_cancel_id(system_delay_taskq, se->se_taskqid);
/*
* if we get ENOENT, the taskq couldn't be found to be
* canceled, so we can just mark it as invalid because
* it's already gone. If we got EBUSY, then we already
* blocked until it was gone _anyway_, so we don't care.
*/
se->se_taskqid = TASKQID_INVALID;
rw_exit(&se->se_taskqid_lock);
if (err == 0) {
zfsctl_snapshot_rele(se);
}
}
/*
* Dispatch the unmount task for delayed handling with a hold protecting it.
*/
static void
zfsctl_snapshot_unmount_delay_impl(zfs_snapentry_t *se, int delay)
{
if (delay <= 0)
return;
zfsctl_snapshot_hold(se);
rw_enter(&se->se_taskqid_lock, RW_WRITER);
ASSERT3S(se->se_taskqid, ==, TASKQID_INVALID);
se->se_taskqid = taskq_dispatch_delay(system_delay_taskq,
snapentry_expire, se, TQ_SLEEP, ddi_get_lbolt() + delay * HZ);
rw_exit(&se->se_taskqid_lock);
}
/*
* Schedule an automatic unmount of objset id to occur in delay seconds from
* now. Any previous delayed unmount will be cancelled in favor of the
* updated deadline. A reference is taken by zfsctl_snapshot_find_by_name()
* and held until the outstanding task is handled or cancelled.
*/
int
zfsctl_snapshot_unmount_delay(spa_t *spa, uint64_t objsetid, int delay)
{
zfs_snapentry_t *se;
int error = ENOENT;
rw_enter(&zfs_snapshot_lock, RW_READER);
if ((se = zfsctl_snapshot_find_by_objsetid(spa, objsetid)) != NULL) {
zfsctl_snapshot_unmount_cancel(se);
zfsctl_snapshot_unmount_delay_impl(se, delay);
zfsctl_snapshot_rele(se);
error = 0;
}
rw_exit(&zfs_snapshot_lock);
return (error);
}
/*
* Check if snapname is currently mounted. Returned non-zero when mounted
* and zero when unmounted.
*/
static boolean_t
zfsctl_snapshot_ismounted(const char *snapname)
{
zfs_snapentry_t *se;
boolean_t ismounted = B_FALSE;
rw_enter(&zfs_snapshot_lock, RW_READER);
if ((se = zfsctl_snapshot_find_by_name(snapname)) != NULL) {
zfsctl_snapshot_rele(se);
ismounted = B_TRUE;
}
rw_exit(&zfs_snapshot_lock);
return (ismounted);
}
/*
* Check if the given inode is a part of the virtual .zfs directory.
*/
boolean_t
zfsctl_is_node(struct inode *ip)
{
return (ITOZ(ip)->z_is_ctldir);
}
/*
* Check if the given inode is a .zfs/snapshots/snapname directory.
*/
boolean_t
zfsctl_is_snapdir(struct inode *ip)
{
return (zfsctl_is_node(ip) && (ip->i_ino <= ZFSCTL_INO_SNAPDIRS));
}
/*
* Allocate a new inode with the passed id and ops.
*/
static struct inode *
zfsctl_inode_alloc(zfsvfs_t *zfsvfs, uint64_t id,
const struct file_operations *fops, const struct inode_operations *ops)
{
inode_timespec_t now;
struct inode *ip;
znode_t *zp;
ip = new_inode(zfsvfs->z_sb);
if (ip == NULL)
return (NULL);
now = current_time(ip);
zp = ITOZ(ip);
ASSERT3P(zp->z_dirlocks, ==, NULL);
ASSERT3P(zp->z_acl_cached, ==, NULL);
ASSERT3P(zp->z_xattr_cached, ==, NULL);
zp->z_id = id;
zp->z_unlinked = B_FALSE;
zp->z_atime_dirty = B_FALSE;
zp->z_zn_prefetch = B_FALSE;
zp->z_is_sa = B_FALSE;
zp->z_is_mapped = B_FALSE;
zp->z_is_ctldir = B_TRUE;
zp->z_is_stale = B_FALSE;
zp->z_sa_hdl = NULL;
zp->z_blksz = 0;
zp->z_seq = 0;
zp->z_mapcnt = 0;
zp->z_size = 0;
zp->z_pflags = 0;
zp->z_mode = 0;
zp->z_sync_cnt = 0;
ip->i_generation = 0;
ip->i_ino = id;
ip->i_mode = (S_IFDIR | S_IRWXUGO);
ip->i_uid = SUID_TO_KUID(0);
ip->i_gid = SGID_TO_KGID(0);
ip->i_blkbits = SPA_MINBLOCKSHIFT;
ip->i_atime = now;
ip->i_mtime = now;
ip->i_ctime = now;
ip->i_fop = fops;
ip->i_op = ops;
#if defined(IOP_XATTR)
ip->i_opflags &= ~IOP_XATTR;
#endif
if (insert_inode_locked(ip)) {
unlock_new_inode(ip);
iput(ip);
return (NULL);
}
mutex_enter(&zfsvfs->z_znodes_lock);
list_insert_tail(&zfsvfs->z_all_znodes, zp);
zfsvfs->z_nr_znodes++;
membar_producer();
mutex_exit(&zfsvfs->z_znodes_lock);
unlock_new_inode(ip);
return (ip);
}
/*
* Lookup the inode with given id, it will be allocated if needed.
*/
static struct inode *
zfsctl_inode_lookup(zfsvfs_t *zfsvfs, uint64_t id,
const struct file_operations *fops, const struct inode_operations *ops)
{
struct inode *ip = NULL;
while (ip == NULL) {
ip = ilookup(zfsvfs->z_sb, (unsigned long)id);
if (ip)
break;
/* May fail due to concurrent zfsctl_inode_alloc() */
ip = zfsctl_inode_alloc(zfsvfs, id, fops, ops);
}
return (ip);
}
/*
* Create the '.zfs' directory. This directory is cached as part of the VFS
* structure. This results in a hold on the zfsvfs_t. The code in zfs_umount()
* therefore checks against a vfs_count of 2 instead of 1. This reference
* is removed when the ctldir is destroyed in the unmount. All other entities
* under the '.zfs' directory are created dynamically as needed.
*
* Because the dynamically created '.zfs' directory entries assume the use
* of 64-bit inode numbers this support must be disabled on 32-bit systems.
*/
int
zfsctl_create(zfsvfs_t *zfsvfs)
{
ASSERT(zfsvfs->z_ctldir == NULL);
zfsvfs->z_ctldir = zfsctl_inode_alloc(zfsvfs, ZFSCTL_INO_ROOT,
&zpl_fops_root, &zpl_ops_root);
if (zfsvfs->z_ctldir == NULL)
return (SET_ERROR(ENOENT));
return (0);
}
/*
* Destroy the '.zfs' directory or remove a snapshot from zfs_snapshots_by_name.
* Only called when the filesystem is unmounted.
*/
void
zfsctl_destroy(zfsvfs_t *zfsvfs)
{
if (zfsvfs->z_issnap) {
zfs_snapentry_t *se;
spa_t *spa = zfsvfs->z_os->os_spa;
uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
rw_enter(&zfs_snapshot_lock, RW_WRITER);
se = zfsctl_snapshot_find_by_objsetid(spa, objsetid);
if (se != NULL)
zfsctl_snapshot_remove(se);
rw_exit(&zfs_snapshot_lock);
if (se != NULL) {
zfsctl_snapshot_unmount_cancel(se);
zfsctl_snapshot_rele(se);
}
} else if (zfsvfs->z_ctldir) {
iput(zfsvfs->z_ctldir);
zfsvfs->z_ctldir = NULL;
}
}
/*
* Given a root znode, retrieve the associated .zfs directory.
* Add a hold to the vnode and return it.
*/
struct inode *
zfsctl_root(znode_t *zp)
{
ASSERT(zfs_has_ctldir(zp));
/* Must have an existing ref, so igrab() cannot return NULL */
VERIFY3P(igrab(ZTOZSB(zp)->z_ctldir), !=, NULL);
return (ZTOZSB(zp)->z_ctldir);
}
/*
* Generate a long fid to indicate a snapdir. We encode whether snapdir is
* already mounted in gen field. We do this because nfsd lookup will not
* trigger automount. Next time the nfsd does fh_to_dentry, we will notice
* this and do automount and return ESTALE to force nfsd revalidate and follow
* mount.
*/
static int
zfsctl_snapdir_fid(struct inode *ip, fid_t *fidp)
{
zfid_short_t *zfid = (zfid_short_t *)fidp;
zfid_long_t *zlfid = (zfid_long_t *)fidp;
uint32_t gen = 0;
uint64_t object;
uint64_t objsetid;
int i;
struct dentry *dentry;
if (fidp->fid_len < LONG_FID_LEN) {
fidp->fid_len = LONG_FID_LEN;
return (SET_ERROR(ENOSPC));
}
object = ip->i_ino;
objsetid = ZFSCTL_INO_SNAPDIRS - ip->i_ino;
zfid->zf_len = LONG_FID_LEN;
dentry = d_obtain_alias(igrab(ip));
if (!IS_ERR(dentry)) {
gen = !!d_mountpoint(dentry);
dput(dentry);
}
for (i = 0; i < sizeof (zfid->zf_object); i++)
zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
for (i = 0; i < sizeof (zfid->zf_gen); i++)
zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
for (i = 0; i < sizeof (zlfid->zf_setid); i++)
zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
zlfid->zf_setgen[i] = 0;
return (0);
}
/*
* Generate an appropriate fid for an entry in the .zfs directory.
*/
int
zfsctl_fid(struct inode *ip, fid_t *fidp)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
uint64_t object = zp->z_id;
zfid_short_t *zfid;
int i;
ZFS_ENTER(zfsvfs);
if (zfsctl_is_snapdir(ip)) {
ZFS_EXIT(zfsvfs);
return (zfsctl_snapdir_fid(ip, fidp));
}
if (fidp->fid_len < SHORT_FID_LEN) {
fidp->fid_len = SHORT_FID_LEN;
ZFS_EXIT(zfsvfs);
return (SET_ERROR(ENOSPC));
}
zfid = (zfid_short_t *)fidp;
zfid->zf_len = SHORT_FID_LEN;
for (i = 0; i < sizeof (zfid->zf_object); i++)
zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
/* .zfs znodes always have a generation number of 0 */
for (i = 0; i < sizeof (zfid->zf_gen); i++)
zfid->zf_gen[i] = 0;
ZFS_EXIT(zfsvfs);
return (0);
}
/*
* Construct a full dataset name in full_name: "pool/dataset@snap_name"
*/
static int
zfsctl_snapshot_name(zfsvfs_t *zfsvfs, const char *snap_name, int len,
char *full_name)
{
objset_t *os = zfsvfs->z_os;
if (zfs_component_namecheck(snap_name, NULL, NULL) != 0)
return (SET_ERROR(EILSEQ));
dmu_objset_name(os, full_name);
if ((strlen(full_name) + 1 + strlen(snap_name)) >= len)
return (SET_ERROR(ENAMETOOLONG));
(void) strcat(full_name, "@");
(void) strcat(full_name, snap_name);
return (0);
}
/*
* Returns full path in full_path: "/pool/dataset/.zfs/snapshot/snap_name/"
*/
static int
zfsctl_snapshot_path_objset(zfsvfs_t *zfsvfs, uint64_t objsetid,
int path_len, char *full_path)
{
objset_t *os = zfsvfs->z_os;
fstrans_cookie_t cookie;
char *snapname;
boolean_t case_conflict;
uint64_t id, pos = 0;
int error = 0;
if (zfsvfs->z_vfs->vfs_mntpoint == NULL)
return (SET_ERROR(ENOENT));
cookie = spl_fstrans_mark();
snapname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
while (error == 0) {
dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
error = dmu_snapshot_list_next(zfsvfs->z_os,
ZFS_MAX_DATASET_NAME_LEN, snapname, &id, &pos,
&case_conflict);
dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
if (error)
goto out;
if (id == objsetid)
break;
}
snprintf(full_path, path_len, "%s/.zfs/snapshot/%s",
zfsvfs->z_vfs->vfs_mntpoint, snapname);
out:
kmem_free(snapname, ZFS_MAX_DATASET_NAME_LEN);
spl_fstrans_unmark(cookie);
return (error);
}
/*
* Special case the handling of "..".
*/
int
zfsctl_root_lookup(struct inode *dip, const char *name, struct inode **ipp,
int flags, cred_t *cr, int *direntflags, pathname_t *realpnp)
{
zfsvfs_t *zfsvfs = ITOZSB(dip);
int error = 0;
ZFS_ENTER(zfsvfs);
if (strcmp(name, "..") == 0) {
*ipp = dip->i_sb->s_root->d_inode;
} else if (strcmp(name, ZFS_SNAPDIR_NAME) == 0) {
*ipp = zfsctl_inode_lookup(zfsvfs, ZFSCTL_INO_SNAPDIR,
&zpl_fops_snapdir, &zpl_ops_snapdir);
} else if (strcmp(name, ZFS_SHAREDIR_NAME) == 0) {
*ipp = zfsctl_inode_lookup(zfsvfs, ZFSCTL_INO_SHARES,
&zpl_fops_shares, &zpl_ops_shares);
} else {
*ipp = NULL;
}
if (*ipp == NULL)
error = SET_ERROR(ENOENT);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Lookup entry point for the 'snapshot' directory. Try to open the
* snapshot if it exist, creating the pseudo filesystem inode as necessary.
*/
int
zfsctl_snapdir_lookup(struct inode *dip, const char *name, struct inode **ipp,
int flags, cred_t *cr, int *direntflags, pathname_t *realpnp)
{
zfsvfs_t *zfsvfs = ITOZSB(dip);
uint64_t id;
int error;
ZFS_ENTER(zfsvfs);
error = dmu_snapshot_lookup(zfsvfs->z_os, name, &id);
if (error) {
ZFS_EXIT(zfsvfs);
return (error);
}
*ipp = zfsctl_inode_lookup(zfsvfs, ZFSCTL_INO_SNAPDIRS - id,
&simple_dir_operations, &simple_dir_inode_operations);
if (*ipp == NULL)
error = SET_ERROR(ENOENT);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Renaming a directory under '.zfs/snapshot' will automatically trigger
* a rename of the snapshot to the new given name. The rename is confined
* to the '.zfs/snapshot' directory snapshots cannot be moved elsewhere.
*/
int
zfsctl_snapdir_rename(struct inode *sdip, const char *snm,
struct inode *tdip, const char *tnm, cred_t *cr, int flags)
{
zfsvfs_t *zfsvfs = ITOZSB(sdip);
char *to, *from, *real, *fsname;
int error;
if (!zfs_admin_snapshot)
return (SET_ERROR(EACCES));
ZFS_ENTER(zfsvfs);
to = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
from = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
real = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
fsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
error = dmu_snapshot_realname(zfsvfs->z_os, snm, real,
ZFS_MAX_DATASET_NAME_LEN, NULL);
if (error == 0) {
snm = real;
} else if (error != ENOTSUP) {
goto out;
}
}
dmu_objset_name(zfsvfs->z_os, fsname);
error = zfsctl_snapshot_name(ITOZSB(sdip), snm,
ZFS_MAX_DATASET_NAME_LEN, from);
if (error == 0)
error = zfsctl_snapshot_name(ITOZSB(tdip), tnm,
ZFS_MAX_DATASET_NAME_LEN, to);
if (error == 0)
error = zfs_secpolicy_rename_perms(from, to, cr);
if (error != 0)
goto out;
/*
* Cannot move snapshots out of the snapdir.
*/
if (sdip != tdip) {
error = SET_ERROR(EINVAL);
goto out;
}
/*
* No-op when names are identical.
*/
if (strcmp(snm, tnm) == 0) {
error = 0;
goto out;
}
rw_enter(&zfs_snapshot_lock, RW_WRITER);
error = dsl_dataset_rename_snapshot(fsname, snm, tnm, B_FALSE);
if (error == 0)
(void) zfsctl_snapshot_rename(snm, tnm);
rw_exit(&zfs_snapshot_lock);
out:
kmem_free(from, ZFS_MAX_DATASET_NAME_LEN);
kmem_free(to, ZFS_MAX_DATASET_NAME_LEN);
kmem_free(real, ZFS_MAX_DATASET_NAME_LEN);
kmem_free(fsname, ZFS_MAX_DATASET_NAME_LEN);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Removing a directory under '.zfs/snapshot' will automatically trigger
* the removal of the snapshot with the given name.
*/
int
zfsctl_snapdir_remove(struct inode *dip, const char *name, cred_t *cr,
int flags)
{
zfsvfs_t *zfsvfs = ITOZSB(dip);
char *snapname, *real;
int error;
if (!zfs_admin_snapshot)
return (SET_ERROR(EACCES));
ZFS_ENTER(zfsvfs);
snapname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
real = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
error = dmu_snapshot_realname(zfsvfs->z_os, name, real,
ZFS_MAX_DATASET_NAME_LEN, NULL);
if (error == 0) {
name = real;
} else if (error != ENOTSUP) {
goto out;
}
}
error = zfsctl_snapshot_name(ITOZSB(dip), name,
ZFS_MAX_DATASET_NAME_LEN, snapname);
if (error == 0)
error = zfs_secpolicy_destroy_perms(snapname, cr);
if (error != 0)
goto out;
error = zfsctl_snapshot_unmount(snapname, MNT_FORCE);
if ((error == 0) || (error == ENOENT))
error = dsl_destroy_snapshot(snapname, B_FALSE);
out:
kmem_free(snapname, ZFS_MAX_DATASET_NAME_LEN);
kmem_free(real, ZFS_MAX_DATASET_NAME_LEN);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Creating a directory under '.zfs/snapshot' will automatically trigger
* the creation of a new snapshot with the given name.
*/
int
zfsctl_snapdir_mkdir(struct inode *dip, const char *dirname, vattr_t *vap,
struct inode **ipp, cred_t *cr, int flags)
{
zfsvfs_t *zfsvfs = ITOZSB(dip);
char *dsname;
int error;
if (!zfs_admin_snapshot)
return (SET_ERROR(EACCES));
dsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
if (zfs_component_namecheck(dirname, NULL, NULL) != 0) {
error = SET_ERROR(EILSEQ);
goto out;
}
dmu_objset_name(zfsvfs->z_os, dsname);
error = zfs_secpolicy_snapshot_perms(dsname, cr);
if (error != 0)
goto out;
if (error == 0) {
error = dmu_objset_snapshot_one(dsname, dirname);
if (error != 0)
goto out;
error = zfsctl_snapdir_lookup(dip, dirname, ipp,
0, cr, NULL, NULL);
}
out:
kmem_free(dsname, ZFS_MAX_DATASET_NAME_LEN);
return (error);
}
/*
* Flush everything out of the kernel's export table and such.
* This is needed as once the snapshot is used over NFS, its
* entries in svc_export and svc_expkey caches hold reference
* to the snapshot mount point. There is no known way of flushing
* only the entries related to the snapshot.
*/
static void
exportfs_flush(void)
{
char *argv[] = { "/usr/sbin/exportfs", "-f", NULL };
char *envp[] = { NULL };
(void) call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
}
/*
* Attempt to unmount a snapshot by making a call to user space.
* There is no assurance that this can or will succeed, is just a
* best effort. In the case where it does fail, perhaps because
* it's in use, the unmount will fail harmlessly.
*/
int
zfsctl_snapshot_unmount(const char *snapname, int flags)
{
char *argv[] = { "/usr/bin/env", "umount", "-t", "zfs", "-n", NULL,
NULL };
char *envp[] = { NULL };
zfs_snapentry_t *se;
int error;
rw_enter(&zfs_snapshot_lock, RW_READER);
if ((se = zfsctl_snapshot_find_by_name(snapname)) == NULL) {
rw_exit(&zfs_snapshot_lock);
return (SET_ERROR(ENOENT));
}
rw_exit(&zfs_snapshot_lock);
exportfs_flush();
if (flags & MNT_FORCE)
argv[4] = "-fn";
argv[5] = se->se_path;
dprintf("unmount; path=%s\n", se->se_path);
error = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
zfsctl_snapshot_rele(se);
/*
* The umount system utility will return 256 on error. We must
* assume this error is because the file system is busy so it is
* converted to the more sensible EBUSY.
*/
if (error)
error = SET_ERROR(EBUSY);
return (error);
}
int
zfsctl_snapshot_mount(struct path *path, int flags)
{
struct dentry *dentry = path->dentry;
struct inode *ip = dentry->d_inode;
zfsvfs_t *zfsvfs;
zfsvfs_t *snap_zfsvfs;
zfs_snapentry_t *se;
char *full_name, *full_path;
char *argv[] = { "/usr/bin/env", "mount", "-t", "zfs", "-n", NULL, NULL,
NULL };
char *envp[] = { NULL };
int error;
struct path spath;
if (ip == NULL)
return (SET_ERROR(EISDIR));
zfsvfs = ITOZSB(ip);
ZFS_ENTER(zfsvfs);
full_name = kmem_zalloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
full_path = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
error = zfsctl_snapshot_name(zfsvfs, dname(dentry),
ZFS_MAX_DATASET_NAME_LEN, full_name);
if (error)
goto error;
/*
* Construct a mount point path from sb of the ctldir inode and dirent
* name, instead of from d_path(), so that chroot'd process doesn't fail
* on mount.zfs(8).
*/
snprintf(full_path, MAXPATHLEN, "%s/.zfs/snapshot/%s",
zfsvfs->z_vfs->vfs_mntpoint ? zfsvfs->z_vfs->vfs_mntpoint : "",
dname(dentry));
/*
* Multiple concurrent automounts of a snapshot are never allowed.
* The snapshot may be manually mounted as many times as desired.
*/
if (zfsctl_snapshot_ismounted(full_name)) {
error = 0;
goto error;
}
/*
* Attempt to mount the snapshot from user space. Normally this
* would be done using the vfs_kern_mount() function, however that
* function is marked GPL-only and cannot be used. On error we
* careful to log the real error to the console and return EISDIR
* to safely abort the automount. This should be very rare.
*
* If the user mode helper happens to return EBUSY, a concurrent
* mount is already in progress in which case the error is ignored.
* Take note that if the program was executed successfully the return
* value from call_usermodehelper() will be (exitcode << 8 + signal).
*/
dprintf("mount; name=%s path=%s\n", full_name, full_path);
argv[5] = full_name;
argv[6] = full_path;
error = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
if (error) {
if (!(error & MOUNT_BUSY << 8)) {
zfs_dbgmsg("Unable to automount %s error=%d",
full_path, error);
error = SET_ERROR(EISDIR);
} else {
/*
* EBUSY, this could mean a concurrent mount, or the
* snapshot has already been mounted at completely
* different place. We return 0 so VFS will retry. For
* the latter case the VFS will retry several times
* and return ELOOP, which is probably not a very good
* behavior.
*/
error = 0;
}
goto error;
}
/*
* Follow down in to the mounted snapshot and set MNT_SHRINKABLE
* to identify this as an automounted filesystem.
*/
spath = *path;
path_get(&spath);
if (follow_down_one(&spath)) {
snap_zfsvfs = ITOZSB(spath.dentry->d_inode);
snap_zfsvfs->z_parent = zfsvfs;
dentry = spath.dentry;
spath.mnt->mnt_flags |= MNT_SHRINKABLE;
rw_enter(&zfs_snapshot_lock, RW_WRITER);
se = zfsctl_snapshot_alloc(full_name, full_path,
snap_zfsvfs->z_os->os_spa, dmu_objset_id(snap_zfsvfs->z_os),
dentry);
zfsctl_snapshot_add(se);
zfsctl_snapshot_unmount_delay_impl(se, zfs_expire_snapshot);
rw_exit(&zfs_snapshot_lock);
}
path_put(&spath);
error:
kmem_free(full_name, ZFS_MAX_DATASET_NAME_LEN);
kmem_free(full_path, MAXPATHLEN);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Get the snapdir inode from fid
*/
int
zfsctl_snapdir_vget(struct super_block *sb, uint64_t objsetid, int gen,
struct inode **ipp)
{
int error;
struct path path;
char *mnt;
struct dentry *dentry;
mnt = kmem_alloc(MAXPATHLEN, KM_SLEEP);
error = zfsctl_snapshot_path_objset(sb->s_fs_info, objsetid,
MAXPATHLEN, mnt);
if (error)
goto out;
/* Trigger automount */
error = -kern_path(mnt, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &path);
if (error)
goto out;
path_put(&path);
/*
* Get the snapdir inode. Note, we don't want to use the above
* path because it contains the root of the snapshot rather
* than the snapdir.
*/
*ipp = ilookup(sb, ZFSCTL_INO_SNAPDIRS - objsetid);
if (*ipp == NULL) {
error = SET_ERROR(ENOENT);
goto out;
}
/* check gen, see zfsctl_snapdir_fid */
dentry = d_obtain_alias(igrab(*ipp));
if (gen != (!IS_ERR(dentry) && d_mountpoint(dentry))) {
iput(*ipp);
*ipp = NULL;
error = SET_ERROR(ENOENT);
}
if (!IS_ERR(dentry))
dput(dentry);
out:
kmem_free(mnt, MAXPATHLEN);
return (error);
}
int
zfsctl_shares_lookup(struct inode *dip, char *name, struct inode **ipp,
int flags, cred_t *cr, int *direntflags, pathname_t *realpnp)
{
zfsvfs_t *zfsvfs = ITOZSB(dip);
znode_t *zp;
znode_t *dzp;
int error;
ZFS_ENTER(zfsvfs);
if (zfsvfs->z_shares_dir == 0) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(ENOTSUP));
}
if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
error = zfs_lookup(dzp, name, &zp, 0, cr, NULL, NULL);
zrele(dzp);
}
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Initialize the various pieces we'll need to create and manipulate .zfs
* directories. Currently this is unused but available.
*/
void
zfsctl_init(void)
{
avl_create(&zfs_snapshots_by_name, snapentry_compare_by_name,
sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t,
se_node_name));
avl_create(&zfs_snapshots_by_objsetid, snapentry_compare_by_objsetid,
sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t,
se_node_objsetid));
rw_init(&zfs_snapshot_lock, NULL, RW_DEFAULT, NULL);
}
/*
* Cleanup the various pieces we needed for .zfs directories. In particular
* ensure the expiry timer is canceled safely.
*/
void
zfsctl_fini(void)
{
avl_destroy(&zfs_snapshots_by_name);
avl_destroy(&zfs_snapshots_by_objsetid);
rw_destroy(&zfs_snapshot_lock);
}
module_param(zfs_admin_snapshot, int, 0644);
MODULE_PARM_DESC(zfs_admin_snapshot, "Enable mkdir/rmdir/mv in .zfs/snapshot");
module_param(zfs_expire_snapshot, int, 0644);
MODULE_PARM_DESC(zfs_expire_snapshot, "Seconds to expire .zfs/snapshot");
diff --git a/module/os/linux/zfs/zfs_debug.c b/module/os/linux/zfs/zfs_debug.c
index a32a4663ebe7..be65f0a2e245 100644
--- a/module/os/linux/zfs/zfs_debug.c
+++ b/module/os/linux/zfs/zfs_debug.c
@@ -1,257 +1,257 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/trace_zfs.h>
typedef struct zfs_dbgmsg {
procfs_list_node_t zdm_node;
uint64_t zdm_timestamp;
int zdm_size;
char zdm_msg[1]; /* variable length allocation */
} zfs_dbgmsg_t;
-procfs_list_t zfs_dbgmsgs;
-int zfs_dbgmsg_size = 0;
+static procfs_list_t zfs_dbgmsgs;
+static int zfs_dbgmsg_size = 0;
int zfs_dbgmsg_maxsize = 4<<20; /* 4MB */
/*
* Internal ZFS debug messages are enabled by default.
*
* # Print debug messages
* cat /proc/spl/kstat/zfs/dbgmsg
*
* # Disable the kernel debug message log.
* echo 0 > /sys/module/zfs/parameters/zfs_dbgmsg_enable
*
* # Clear the kernel debug message log.
* echo 0 >/proc/spl/kstat/zfs/dbgmsg
*/
-int zfs_dbgmsg_enable = 1;
+int zfs_dbgmsg_enable = B_TRUE;
static int
zfs_dbgmsg_show_header(struct seq_file *f)
{
seq_printf(f, "%-12s %-8s\n", "timestamp", "message");
return (0);
}
static int
zfs_dbgmsg_show(struct seq_file *f, void *p)
{
zfs_dbgmsg_t *zdm = (zfs_dbgmsg_t *)p;
seq_printf(f, "%-12llu %-s\n",
(u_longlong_t)zdm->zdm_timestamp, zdm->zdm_msg);
return (0);
}
static void
zfs_dbgmsg_purge(int max_size)
{
while (zfs_dbgmsg_size > max_size) {
zfs_dbgmsg_t *zdm = list_remove_head(&zfs_dbgmsgs.pl_list);
if (zdm == NULL)
return;
int size = zdm->zdm_size;
kmem_free(zdm, size);
zfs_dbgmsg_size -= size;
}
}
static int
zfs_dbgmsg_clear(procfs_list_t *procfs_list)
{
(void) procfs_list;
mutex_enter(&zfs_dbgmsgs.pl_lock);
zfs_dbgmsg_purge(0);
mutex_exit(&zfs_dbgmsgs.pl_lock);
return (0);
}
void
zfs_dbgmsg_init(void)
{
procfs_list_install("zfs",
NULL,
"dbgmsg",
0600,
&zfs_dbgmsgs,
zfs_dbgmsg_show,
zfs_dbgmsg_show_header,
zfs_dbgmsg_clear,
offsetof(zfs_dbgmsg_t, zdm_node));
}
void
zfs_dbgmsg_fini(void)
{
procfs_list_uninstall(&zfs_dbgmsgs);
zfs_dbgmsg_purge(0);
/*
* TODO - decide how to make this permanent
*/
#ifdef _KERNEL
procfs_list_destroy(&zfs_dbgmsgs);
#endif
}
void
__set_error(const char *file, const char *func, int line, int err)
{
/*
* To enable this:
*
* $ echo 512 >/sys/module/zfs/parameters/zfs_flags
*/
if (zfs_flags & ZFS_DEBUG_SET_ERROR)
__dprintf(B_FALSE, file, func, line, "error %lu",
(ulong_t)err);
}
void
__zfs_dbgmsg(char *buf)
{
int size = sizeof (zfs_dbgmsg_t) + strlen(buf);
zfs_dbgmsg_t *zdm = kmem_zalloc(size, KM_SLEEP);
zdm->zdm_size = size;
zdm->zdm_timestamp = gethrestime_sec();
strcpy(zdm->zdm_msg, buf);
mutex_enter(&zfs_dbgmsgs.pl_lock);
procfs_list_add(&zfs_dbgmsgs, zdm);
zfs_dbgmsg_size += size;
zfs_dbgmsg_purge(MAX(zfs_dbgmsg_maxsize, 0));
mutex_exit(&zfs_dbgmsgs.pl_lock);
}
#ifdef _KERNEL
void
__dprintf(boolean_t dprint, const char *file, const char *func,
int line, const char *fmt, ...)
{
const char *newfile;
va_list adx;
size_t size;
char *buf;
char *nl;
int i;
char *prefix = (dprint) ? "dprintf: " : "";
size = 1024;
buf = kmem_alloc(size, KM_SLEEP);
/*
* Get rid of annoying prefix to filename.
*/
newfile = strrchr(file, '/');
if (newfile != NULL) {
newfile = newfile + 1; /* Get rid of leading / */
} else {
newfile = file;
}
i = snprintf(buf, size, "%s%s:%d:%s(): ", prefix, newfile, line, func);
if (i < size) {
va_start(adx, fmt);
(void) vsnprintf(buf + i, size - i, fmt, adx);
va_end(adx);
}
/*
* Get rid of trailing newline for dprintf logs.
*/
if (dprint && buf[0] != '\0') {
nl = &buf[strlen(buf) - 1];
if (*nl == '\n')
*nl = '\0';
}
/*
* To get this data enable the zfs__dprintf trace point as shown:
*
* # Enable zfs__dprintf tracepoint, clear the tracepoint ring buffer
* $ echo 1 > /sys/kernel/debug/tracing/events/zfs/enable
* $ echo 0 > /sys/kernel/debug/tracing/trace
*
* # Dump the ring buffer.
* $ cat /sys/kernel/debug/tracing/trace
*/
DTRACE_PROBE1(zfs__dprintf, char *, buf);
/*
* To get this data:
*
* $ cat /proc/spl/kstat/zfs/dbgmsg
*
* To clear the buffer:
* $ echo 0 > /proc/spl/kstat/zfs/dbgmsg
*/
__zfs_dbgmsg(buf);
kmem_free(buf, size);
}
#else
void
zfs_dbgmsg_print(const char *tag)
{
ssize_t ret __attribute__((unused));
/*
* We use write() in this function instead of printf()
* so it is safe to call from a signal handler.
*/
ret = write(STDOUT_FILENO, "ZFS_DBGMSG(", 11);
ret = write(STDOUT_FILENO, tag, strlen(tag));
ret = write(STDOUT_FILENO, ") START:\n", 9);
mutex_enter(&zfs_dbgmsgs.pl_lock);
for (zfs_dbgmsg_t *zdm = list_head(&zfs_dbgmsgs.pl_list); zdm != NULL;
zdm = list_next(&zfs_dbgmsgs.pl_list, zdm)) {
ret = write(STDOUT_FILENO, zdm->zdm_msg,
strlen(zdm->zdm_msg));
ret = write(STDOUT_FILENO, "\n", 1);
}
ret = write(STDOUT_FILENO, "ZFS_DBGMSG(", 11);
ret = write(STDOUT_FILENO, tag, strlen(tag));
ret = write(STDOUT_FILENO, ") END\n", 6);
mutex_exit(&zfs_dbgmsgs.pl_lock);
}
#endif /* _KERNEL */
#ifdef _KERNEL
module_param(zfs_dbgmsg_enable, int, 0644);
MODULE_PARM_DESC(zfs_dbgmsg_enable, "Enable ZFS debug message log");
module_param(zfs_dbgmsg_maxsize, int, 0644);
MODULE_PARM_DESC(zfs_dbgmsg_maxsize, "Maximum ZFS debug log size");
#endif
diff --git a/module/os/linux/zfs/zfs_vnops_os.c b/module/os/linux/zfs/zfs_vnops_os.c
index 6c7de9830c15..aff3c4ad4e4c 100644
--- a/module/os/linux/zfs/zfs_vnops_os.c
+++ b/module/os/linux/zfs/zfs_vnops_os.c
@@ -1,4003 +1,4003 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
* Copyright (c) 2015 by Chunwei Chen. All rights reserved.
* Copyright 2017 Nexenta Systems, Inc.
*/
/* Portions Copyright 2007 Jeremy Teo */
/* Portions Copyright 2010 Robert Milkowski */
#include <sys/types.h>
#include <sys/param.h>
#include <sys/time.h>
#include <sys/sysmacros.h>
#include <sys/vfs.h>
#include <sys/file.h>
#include <sys/stat.h>
#include <sys/kmem.h>
#include <sys/taskq.h>
#include <sys/uio.h>
#include <sys/vmsystm.h>
#include <sys/atomic.h>
#include <sys/pathname.h>
#include <sys/cmn_err.h>
#include <sys/errno.h>
#include <sys/zfs_dir.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_ioctl.h>
#include <sys/fs/zfs.h>
#include <sys/dmu.h>
#include <sys/dmu_objset.h>
#include <sys/spa.h>
#include <sys/txg.h>
#include <sys/dbuf.h>
#include <sys/zap.h>
#include <sys/sa.h>
#include <sys/policy.h>
#include <sys/sunddi.h>
#include <sys/sid.h>
#include <sys/zfs_ctldir.h>
#include <sys/zfs_fuid.h>
#include <sys/zfs_quota.h>
#include <sys/zfs_sa.h>
#include <sys/zfs_vnops.h>
#include <sys/zfs_rlock.h>
#include <sys/cred.h>
#include <sys/zpl.h>
#include <sys/zil.h>
#include <sys/sa_impl.h>
/*
* Programming rules.
*
* Each vnode op performs some logical unit of work. To do this, the ZPL must
* properly lock its in-core state, create a DMU transaction, do the work,
* record this work in the intent log (ZIL), commit the DMU transaction,
* and wait for the intent log to commit if it is a synchronous operation.
* Moreover, the vnode ops must work in both normal and log replay context.
* The ordering of events is important to avoid deadlocks and references
* to freed memory. The example below illustrates the following Big Rules:
*
* (1) A check must be made in each zfs thread for a mounted file system.
* This is done avoiding races using ZFS_ENTER(zfsvfs).
* A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes
* must be checked with ZFS_VERIFY_ZP(zp). Both of these macros
* can return EIO from the calling function.
*
* (2) zrele() should always be the last thing except for zil_commit() (if
* necessary) and ZFS_EXIT(). This is for 3 reasons: First, if it's the
* last reference, the vnode/znode can be freed, so the zp may point to
* freed memory. Second, the last reference will call zfs_zinactive(),
* which may induce a lot of work -- pushing cached pages (which acquires
* range locks) and syncing out cached atime changes. Third,
* zfs_zinactive() may require a new tx, which could deadlock the system
* if you were already holding one. This deadlock occurs because the tx
* currently being operated on prevents a txg from syncing, which
* prevents the new tx from progressing, resulting in a deadlock. If you
* must call zrele() within a tx, use zfs_zrele_async(). Note that iput()
* is a synonym for zrele().
*
* (3) All range locks must be grabbed before calling dmu_tx_assign(),
* as they can span dmu_tx_assign() calls.
*
* (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to
* dmu_tx_assign(). This is critical because we don't want to block
* while holding locks.
*
* If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This
* reduces lock contention and CPU usage when we must wait (note that if
* throughput is constrained by the storage, nearly every transaction
* must wait).
*
* Note, in particular, that if a lock is sometimes acquired before
* the tx assigns, and sometimes after (e.g. z_lock), then failing
* to use a non-blocking assign can deadlock the system. The scenario:
*
* Thread A has grabbed a lock before calling dmu_tx_assign().
* Thread B is in an already-assigned tx, and blocks for this lock.
* Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
* forever, because the previous txg can't quiesce until B's tx commits.
*
* If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
* then drop all locks, call dmu_tx_wait(), and try again. On subsequent
* calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT,
* to indicate that this operation has already called dmu_tx_wait().
* This will ensure that we don't retry forever, waiting a short bit
* each time.
*
* (5) If the operation succeeded, generate the intent log entry for it
* before dropping locks. This ensures that the ordering of events
* in the intent log matches the order in which they actually occurred.
* During ZIL replay the zfs_log_* functions will update the sequence
* number to indicate the zil transaction has replayed.
*
* (6) At the end of each vnode op, the DMU tx must always commit,
* regardless of whether there were any errors.
*
* (7) After dropping all locks, invoke zil_commit(zilog, foid)
* to ensure that synchronous semantics are provided when necessary.
*
* In general, this is how things should be ordered in each vnode op:
*
* ZFS_ENTER(zfsvfs); // exit if unmounted
* top:
* zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab())
* rw_enter(...); // grab any other locks you need
* tx = dmu_tx_create(...); // get DMU tx
* dmu_tx_hold_*(); // hold each object you might modify
* error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
* if (error) {
* rw_exit(...); // drop locks
* zfs_dirent_unlock(dl); // unlock directory entry
* zrele(...); // release held znodes
* if (error == ERESTART) {
* waited = B_TRUE;
* dmu_tx_wait(tx);
* dmu_tx_abort(tx);
* goto top;
* }
* dmu_tx_abort(tx); // abort DMU tx
* ZFS_EXIT(zfsvfs); // finished in zfs
* return (error); // really out of space
* }
* error = do_real_work(); // do whatever this VOP does
* if (error == 0)
* zfs_log_*(...); // on success, make ZIL entry
* dmu_tx_commit(tx); // commit DMU tx -- error or not
* rw_exit(...); // drop locks
* zfs_dirent_unlock(dl); // unlock directory entry
* zrele(...); // release held znodes
* zil_commit(zilog, foid); // synchronous when necessary
* ZFS_EXIT(zfsvfs); // finished in zfs
* return (error); // done, report error
*/
/* ARGSUSED */
int
zfs_open(struct inode *ip, int mode, int flag, cred_t *cr)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
/* Honor ZFS_APPENDONLY file attribute */
if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) &&
((flag & O_APPEND) == 0)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EPERM));
}
/* Keep a count of the synchronous opens in the znode */
if (flag & O_SYNC)
atomic_inc_32(&zp->z_sync_cnt);
ZFS_EXIT(zfsvfs);
return (0);
}
/* ARGSUSED */
int
zfs_close(struct inode *ip, int flag, cred_t *cr)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
/* Decrement the synchronous opens in the znode */
if (flag & O_SYNC)
atomic_dec_32(&zp->z_sync_cnt);
ZFS_EXIT(zfsvfs);
return (0);
}
#if defined(_KERNEL)
/*
* When a file is memory mapped, we must keep the IO data synchronized
* between the DMU cache and the memory mapped pages. What this means:
*
* On Write: If we find a memory mapped page, we write to *both*
* the page and the dmu buffer.
*/
void
update_pages(znode_t *zp, int64_t start, int len, objset_t *os)
{
struct inode *ip = ZTOI(zp);
struct address_space *mp = ip->i_mapping;
struct page *pp;
uint64_t nbytes;
int64_t off;
void *pb;
off = start & (PAGE_SIZE-1);
for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) {
nbytes = MIN(PAGE_SIZE - off, len);
pp = find_lock_page(mp, start >> PAGE_SHIFT);
if (pp) {
if (mapping_writably_mapped(mp))
flush_dcache_page(pp);
pb = kmap(pp);
(void) dmu_read(os, zp->z_id, start + off, nbytes,
pb + off, DMU_READ_PREFETCH);
kunmap(pp);
if (mapping_writably_mapped(mp))
flush_dcache_page(pp);
mark_page_accessed(pp);
SetPageUptodate(pp);
ClearPageError(pp);
unlock_page(pp);
put_page(pp);
}
len -= nbytes;
off = 0;
}
}
/*
* When a file is memory mapped, we must keep the IO data synchronized
* between the DMU cache and the memory mapped pages. What this means:
*
* On Read: We "read" preferentially from memory mapped pages,
* else we default from the dmu buffer.
*
* NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
* the file is memory mapped.
*/
int
mappedread(znode_t *zp, int nbytes, zfs_uio_t *uio)
{
struct inode *ip = ZTOI(zp);
struct address_space *mp = ip->i_mapping;
struct page *pp;
int64_t start, off;
uint64_t bytes;
int len = nbytes;
int error = 0;
void *pb;
start = uio->uio_loffset;
off = start & (PAGE_SIZE-1);
for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) {
bytes = MIN(PAGE_SIZE - off, len);
pp = find_lock_page(mp, start >> PAGE_SHIFT);
if (pp) {
ASSERT(PageUptodate(pp));
unlock_page(pp);
pb = kmap(pp);
error = zfs_uiomove(pb + off, bytes, UIO_READ, uio);
kunmap(pp);
if (mapping_writably_mapped(mp))
flush_dcache_page(pp);
mark_page_accessed(pp);
put_page(pp);
} else {
error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
uio, bytes);
}
len -= bytes;
off = 0;
if (error)
break;
}
return (error);
}
#endif /* _KERNEL */
-unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT;
+static unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT;
/*
* Write the bytes to a file.
*
* IN: zp - znode of file to be written to
* data - bytes to write
* len - number of bytes to write
* pos - offset to start writing at
*
* OUT: resid - remaining bytes to write
*
* RETURN: 0 if success
* positive error code if failure. EIO is returned
* for a short write when residp isn't provided.
*
* Timestamps:
* zp - ctime|mtime updated if byte count > 0
*/
int
zfs_write_simple(znode_t *zp, const void *data, size_t len,
loff_t pos, size_t *residp)
{
fstrans_cookie_t cookie;
int error;
struct iovec iov;
iov.iov_base = (void *)data;
iov.iov_len = len;
zfs_uio_t uio;
zfs_uio_iovec_init(&uio, &iov, 1, pos, UIO_SYSSPACE, len, 0);
cookie = spl_fstrans_mark();
error = zfs_write(zp, &uio, 0, kcred);
spl_fstrans_unmark(cookie);
if (error == 0) {
if (residp != NULL)
*residp = zfs_uio_resid(&uio);
else if (zfs_uio_resid(&uio) != 0)
error = SET_ERROR(EIO);
}
return (error);
}
static void
zfs_rele_async_task(void *arg)
{
iput(arg);
}
void
zfs_zrele_async(znode_t *zp)
{
struct inode *ip = ZTOI(zp);
objset_t *os = ITOZSB(ip)->z_os;
ASSERT(atomic_read(&ip->i_count) > 0);
ASSERT(os != NULL);
/*
* If decrementing the count would put us at 0, we can't do it inline
* here, because that would be synchronous. Instead, dispatch an iput
* to run later.
*
* For more information on the dangers of a synchronous iput, see the
* header comment of this file.
*/
if (!atomic_add_unless(&ip->i_count, -1, 1)) {
VERIFY(taskq_dispatch(dsl_pool_zrele_taskq(dmu_objset_pool(os)),
zfs_rele_async_task, ip, TQ_SLEEP) != TASKQID_INVALID);
}
}
/*
* Lookup an entry in a directory, or an extended attribute directory.
* If it exists, return a held inode reference for it.
*
* IN: zdp - znode of directory to search.
* nm - name of entry to lookup.
* flags - LOOKUP_XATTR set if looking for an attribute.
* cr - credentials of caller.
* direntflags - directory lookup flags
* realpnp - returned pathname.
*
* OUT: zpp - znode of located entry, NULL if not found.
*
* RETURN: 0 on success, error code on failure.
*
* Timestamps:
* NA
*/
/* ARGSUSED */
int
zfs_lookup(znode_t *zdp, char *nm, znode_t **zpp, int flags, cred_t *cr,
int *direntflags, pathname_t *realpnp)
{
zfsvfs_t *zfsvfs = ZTOZSB(zdp);
int error = 0;
/*
* Fast path lookup, however we must skip DNLC lookup
* for case folding or normalizing lookups because the
* DNLC code only stores the passed in name. This means
* creating 'a' and removing 'A' on a case insensitive
* file system would work, but DNLC still thinks 'a'
* exists and won't let you create it again on the next
* pass through fast path.
*/
if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {
if (!S_ISDIR(ZTOI(zdp)->i_mode)) {
return (SET_ERROR(ENOTDIR));
} else if (zdp->z_sa_hdl == NULL) {
return (SET_ERROR(EIO));
}
if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
error = zfs_fastaccesschk_execute(zdp, cr);
if (!error) {
*zpp = zdp;
zhold(*zpp);
return (0);
}
return (error);
}
}
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zdp);
*zpp = NULL;
if (flags & LOOKUP_XATTR) {
/*
* We don't allow recursive attributes..
* Maybe someday we will.
*/
if (zdp->z_pflags & ZFS_XATTR) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EINVAL));
}
if ((error = zfs_get_xattrdir(zdp, zpp, cr, flags))) {
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Do we have permission to get into attribute directory?
*/
if ((error = zfs_zaccess(*zpp, ACE_EXECUTE, 0,
B_FALSE, cr))) {
zrele(*zpp);
*zpp = NULL;
}
ZFS_EXIT(zfsvfs);
return (error);
}
if (!S_ISDIR(ZTOI(zdp)->i_mode)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(ENOTDIR));
}
/*
* Check accessibility of directory.
*/
if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) {
ZFS_EXIT(zfsvfs);
return (error);
}
if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EILSEQ));
}
error = zfs_dirlook(zdp, nm, zpp, flags, direntflags, realpnp);
if ((error == 0) && (*zpp))
zfs_znode_update_vfs(*zpp);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Attempt to create a new entry in a directory. If the entry
* already exists, truncate the file if permissible, else return
* an error. Return the ip of the created or trunc'd file.
*
* IN: dzp - znode of directory to put new file entry in.
* name - name of new file entry.
* vap - attributes of new file.
* excl - flag indicating exclusive or non-exclusive mode.
* mode - mode to open file with.
* cr - credentials of caller.
* flag - file flag.
* vsecp - ACL to be set
*
* OUT: zpp - znode of created or trunc'd entry.
*
* RETURN: 0 on success, error code on failure.
*
* Timestamps:
* dzp - ctime|mtime updated if new entry created
* zp - ctime|mtime always, atime if new
*/
/* ARGSUSED */
int
zfs_create(znode_t *dzp, char *name, vattr_t *vap, int excl,
int mode, znode_t **zpp, cred_t *cr, int flag, vsecattr_t *vsecp)
{
znode_t *zp;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
zilog_t *zilog;
objset_t *os;
zfs_dirlock_t *dl;
dmu_tx_t *tx;
int error;
uid_t uid;
gid_t gid;
zfs_acl_ids_t acl_ids;
boolean_t fuid_dirtied;
boolean_t have_acl = B_FALSE;
boolean_t waited = B_FALSE;
/*
* If we have an ephemeral id, ACL, or XVATTR then
* make sure file system is at proper version
*/
gid = crgetgid(cr);
uid = crgetuid(cr);
if (zfsvfs->z_use_fuids == B_FALSE &&
(vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
return (SET_ERROR(EINVAL));
if (name == NULL)
return (SET_ERROR(EINVAL));
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(dzp);
os = zfsvfs->z_os;
zilog = zfsvfs->z_log;
if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EILSEQ));
}
if (vap->va_mask & ATTR_XVATTR) {
if ((error = secpolicy_xvattr((xvattr_t *)vap,
crgetuid(cr), cr, vap->va_mode)) != 0) {
ZFS_EXIT(zfsvfs);
return (error);
}
}
top:
*zpp = NULL;
if (*name == '\0') {
/*
* Null component name refers to the directory itself.
*/
zhold(dzp);
zp = dzp;
dl = NULL;
error = 0;
} else {
/* possible igrab(zp) */
int zflg = 0;
if (flag & FIGNORECASE)
zflg |= ZCILOOK;
error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
NULL, NULL);
if (error) {
if (have_acl)
zfs_acl_ids_free(&acl_ids);
if (strcmp(name, "..") == 0)
error = SET_ERROR(EISDIR);
ZFS_EXIT(zfsvfs);
return (error);
}
}
if (zp == NULL) {
uint64_t txtype;
uint64_t projid = ZFS_DEFAULT_PROJID;
/*
* Create a new file object and update the directory
* to reference it.
*/
if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
if (have_acl)
zfs_acl_ids_free(&acl_ids);
goto out;
}
/*
* We only support the creation of regular files in
* extended attribute directories.
*/
if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) {
if (have_acl)
zfs_acl_ids_free(&acl_ids);
error = SET_ERROR(EINVAL);
goto out;
}
if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap,
cr, vsecp, &acl_ids)) != 0)
goto out;
have_acl = B_TRUE;
if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode))
projid = zfs_inherit_projid(dzp);
if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) {
zfs_acl_ids_free(&acl_ids);
error = SET_ERROR(EDQUOT);
goto out;
}
tx = dmu_tx_create(os);
dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
ZFS_SA_BASE_ATTR_SIZE);
fuid_dirtied = zfsvfs->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
if (!zfsvfs->z_use_sa &&
acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
0, acl_ids.z_aclp->z_acl_bytes);
}
error = dmu_tx_assign(tx,
(waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
if (error) {
zfs_dirent_unlock(dl);
if (error == ERESTART) {
waited = B_TRUE;
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
zfs_acl_ids_free(&acl_ids);
dmu_tx_abort(tx);
ZFS_EXIT(zfsvfs);
return (error);
}
zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
error = zfs_link_create(dl, zp, tx, ZNEW);
if (error != 0) {
/*
* Since, we failed to add the directory entry for it,
* delete the newly created dnode.
*/
zfs_znode_delete(zp, tx);
remove_inode_hash(ZTOI(zp));
zfs_acl_ids_free(&acl_ids);
dmu_tx_commit(tx);
goto out;
}
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
if (flag & FIGNORECASE)
txtype |= TX_CI;
zfs_log_create(zilog, tx, txtype, dzp, zp, name,
vsecp, acl_ids.z_fuidp, vap);
zfs_acl_ids_free(&acl_ids);
dmu_tx_commit(tx);
} else {
int aflags = (flag & O_APPEND) ? V_APPEND : 0;
if (have_acl)
zfs_acl_ids_free(&acl_ids);
have_acl = B_FALSE;
/*
* A directory entry already exists for this name.
*/
/*
* Can't truncate an existing file if in exclusive mode.
*/
if (excl) {
error = SET_ERROR(EEXIST);
goto out;
}
/*
* Can't open a directory for writing.
*/
if (S_ISDIR(ZTOI(zp)->i_mode)) {
error = SET_ERROR(EISDIR);
goto out;
}
/*
* Verify requested access to file.
*/
if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
goto out;
}
mutex_enter(&dzp->z_lock);
dzp->z_seq++;
mutex_exit(&dzp->z_lock);
/*
* Truncate regular files if requested.
*/
if (S_ISREG(ZTOI(zp)->i_mode) &&
(vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) {
/* we can't hold any locks when calling zfs_freesp() */
if (dl) {
zfs_dirent_unlock(dl);
dl = NULL;
}
error = zfs_freesp(zp, 0, 0, mode, TRUE);
}
}
out:
if (dl)
zfs_dirent_unlock(dl);
if (error) {
if (zp)
zrele(zp);
} else {
zfs_znode_update_vfs(dzp);
zfs_znode_update_vfs(zp);
*zpp = zp;
}
if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zilog, 0);
ZFS_EXIT(zfsvfs);
return (error);
}
/* ARGSUSED */
int
zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl,
int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp)
{
znode_t *zp = NULL, *dzp = ITOZ(dip);
zfsvfs_t *zfsvfs = ITOZSB(dip);
objset_t *os;
dmu_tx_t *tx;
int error;
uid_t uid;
gid_t gid;
zfs_acl_ids_t acl_ids;
uint64_t projid = ZFS_DEFAULT_PROJID;
boolean_t fuid_dirtied;
boolean_t have_acl = B_FALSE;
boolean_t waited = B_FALSE;
/*
* If we have an ephemeral id, ACL, or XVATTR then
* make sure file system is at proper version
*/
gid = crgetgid(cr);
uid = crgetuid(cr);
if (zfsvfs->z_use_fuids == B_FALSE &&
(vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
return (SET_ERROR(EINVAL));
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(dzp);
os = zfsvfs->z_os;
if (vap->va_mask & ATTR_XVATTR) {
if ((error = secpolicy_xvattr((xvattr_t *)vap,
crgetuid(cr), cr, vap->va_mode)) != 0) {
ZFS_EXIT(zfsvfs);
return (error);
}
}
top:
*ipp = NULL;
/*
* Create a new file object and update the directory
* to reference it.
*/
if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
if (have_acl)
zfs_acl_ids_free(&acl_ids);
goto out;
}
if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap,
cr, vsecp, &acl_ids)) != 0)
goto out;
have_acl = B_TRUE;
if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode))
projid = zfs_inherit_projid(dzp);
if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) {
zfs_acl_ids_free(&acl_ids);
error = SET_ERROR(EDQUOT);
goto out;
}
tx = dmu_tx_create(os);
dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
ZFS_SA_BASE_ATTR_SIZE);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
fuid_dirtied = zfsvfs->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
if (!zfsvfs->z_use_sa &&
acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
0, acl_ids.z_aclp->z_acl_bytes);
}
error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
if (error) {
if (error == ERESTART) {
waited = B_TRUE;
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
zfs_acl_ids_free(&acl_ids);
dmu_tx_abort(tx);
ZFS_EXIT(zfsvfs);
return (error);
}
zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids);
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
/* Add to unlinked set */
zp->z_unlinked = B_TRUE;
zfs_unlinked_add(zp, tx);
zfs_acl_ids_free(&acl_ids);
dmu_tx_commit(tx);
out:
if (error) {
if (zp)
zrele(zp);
} else {
zfs_znode_update_vfs(dzp);
zfs_znode_update_vfs(zp);
*ipp = ZTOI(zp);
}
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Remove an entry from a directory.
*
* IN: dzp - znode of directory to remove entry from.
* name - name of entry to remove.
* cr - credentials of caller.
* flags - case flags.
*
* RETURN: 0 if success
* error code if failure
*
* Timestamps:
* dzp - ctime|mtime
* ip - ctime (if nlink > 0)
*/
uint64_t null_xattr = 0;
/*ARGSUSED*/
int
zfs_remove(znode_t *dzp, char *name, cred_t *cr, int flags)
{
znode_t *zp;
znode_t *xzp;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
zilog_t *zilog;
uint64_t acl_obj, xattr_obj;
uint64_t xattr_obj_unlinked = 0;
uint64_t obj = 0;
uint64_t links;
zfs_dirlock_t *dl;
dmu_tx_t *tx;
boolean_t may_delete_now, delete_now = FALSE;
boolean_t unlinked, toobig = FALSE;
uint64_t txtype;
pathname_t *realnmp = NULL;
pathname_t realnm;
int error;
int zflg = ZEXISTS;
boolean_t waited = B_FALSE;
if (name == NULL)
return (SET_ERROR(EINVAL));
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(dzp);
zilog = zfsvfs->z_log;
if (flags & FIGNORECASE) {
zflg |= ZCILOOK;
pn_alloc(&realnm);
realnmp = &realnm;
}
top:
xattr_obj = 0;
xzp = NULL;
/*
* Attempt to lock directory; fail if entry doesn't exist.
*/
if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
NULL, realnmp))) {
if (realnmp)
pn_free(realnmp);
ZFS_EXIT(zfsvfs);
return (error);
}
if ((error = zfs_zaccess_delete(dzp, zp, cr))) {
goto out;
}
/*
* Need to use rmdir for removing directories.
*/
if (S_ISDIR(ZTOI(zp)->i_mode)) {
error = SET_ERROR(EPERM);
goto out;
}
mutex_enter(&zp->z_lock);
may_delete_now = atomic_read(&ZTOI(zp)->i_count) == 1 &&
!(zp->z_is_mapped);
mutex_exit(&zp->z_lock);
/*
* We may delete the znode now, or we may put it in the unlinked set;
* it depends on whether we're the last link, and on whether there are
* other holds on the inode. So we dmu_tx_hold() the right things to
* allow for either case.
*/
obj = zp->z_id;
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
zfs_sa_upgrade_txholds(tx, dzp);
if (may_delete_now) {
toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks;
/* if the file is too big, only hold_free a token amount */
dmu_tx_hold_free(tx, zp->z_id, 0,
(toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
}
/* are there any extended attributes? */
error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
&xattr_obj, sizeof (xattr_obj));
if (error == 0 && xattr_obj) {
error = zfs_zget(zfsvfs, xattr_obj, &xzp);
ASSERT0(error);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
}
mutex_enter(&zp->z_lock);
if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now)
dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
mutex_exit(&zp->z_lock);
/* charge as an update -- would be nice not to charge at all */
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
/*
* Mark this transaction as typically resulting in a net free of space
*/
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
if (error) {
zfs_dirent_unlock(dl);
if (error == ERESTART) {
waited = B_TRUE;
dmu_tx_wait(tx);
dmu_tx_abort(tx);
zrele(zp);
if (xzp)
zrele(xzp);
goto top;
}
if (realnmp)
pn_free(realnmp);
dmu_tx_abort(tx);
zrele(zp);
if (xzp)
zrele(xzp);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Remove the directory entry.
*/
error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
if (error) {
dmu_tx_commit(tx);
goto out;
}
if (unlinked) {
/*
* Hold z_lock so that we can make sure that the ACL obj
* hasn't changed. Could have been deleted due to
* zfs_sa_upgrade().
*/
mutex_enter(&zp->z_lock);
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
&xattr_obj_unlinked, sizeof (xattr_obj_unlinked));
delete_now = may_delete_now && !toobig &&
atomic_read(&ZTOI(zp)->i_count) == 1 &&
!(zp->z_is_mapped) && xattr_obj == xattr_obj_unlinked &&
zfs_external_acl(zp) == acl_obj;
}
if (delete_now) {
if (xattr_obj_unlinked) {
ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2);
mutex_enter(&xzp->z_lock);
xzp->z_unlinked = B_TRUE;
clear_nlink(ZTOI(xzp));
links = 0;
error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
&links, sizeof (links), tx);
ASSERT3U(error, ==, 0);
mutex_exit(&xzp->z_lock);
zfs_unlinked_add(xzp, tx);
if (zp->z_is_sa)
error = sa_remove(zp->z_sa_hdl,
SA_ZPL_XATTR(zfsvfs), tx);
else
error = sa_update(zp->z_sa_hdl,
SA_ZPL_XATTR(zfsvfs), &null_xattr,
sizeof (uint64_t), tx);
ASSERT0(error);
}
/*
* Add to the unlinked set because a new reference could be
* taken concurrently resulting in a deferred destruction.
*/
zfs_unlinked_add(zp, tx);
mutex_exit(&zp->z_lock);
} else if (unlinked) {
mutex_exit(&zp->z_lock);
zfs_unlinked_add(zp, tx);
}
txtype = TX_REMOVE;
if (flags & FIGNORECASE)
txtype |= TX_CI;
zfs_log_remove(zilog, tx, txtype, dzp, name, obj, unlinked);
dmu_tx_commit(tx);
out:
if (realnmp)
pn_free(realnmp);
zfs_dirent_unlock(dl);
zfs_znode_update_vfs(dzp);
zfs_znode_update_vfs(zp);
if (delete_now)
zrele(zp);
else
zfs_zrele_async(zp);
if (xzp) {
zfs_znode_update_vfs(xzp);
zfs_zrele_async(xzp);
}
if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zilog, 0);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Create a new directory and insert it into dzp using the name
* provided. Return a pointer to the inserted directory.
*
* IN: dzp - znode of directory to add subdir to.
* dirname - name of new directory.
* vap - attributes of new directory.
* cr - credentials of caller.
* flags - case flags.
* vsecp - ACL to be set
*
* OUT: zpp - znode of created directory.
*
* RETURN: 0 if success
* error code if failure
*
* Timestamps:
* dzp - ctime|mtime updated
* zpp - ctime|mtime|atime updated
*/
/*ARGSUSED*/
int
zfs_mkdir(znode_t *dzp, char *dirname, vattr_t *vap, znode_t **zpp,
cred_t *cr, int flags, vsecattr_t *vsecp)
{
znode_t *zp;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
zilog_t *zilog;
zfs_dirlock_t *dl;
uint64_t txtype;
dmu_tx_t *tx;
int error;
int zf = ZNEW;
uid_t uid;
gid_t gid = crgetgid(cr);
zfs_acl_ids_t acl_ids;
boolean_t fuid_dirtied;
boolean_t waited = B_FALSE;
ASSERT(S_ISDIR(vap->va_mode));
/*
* If we have an ephemeral id, ACL, or XVATTR then
* make sure file system is at proper version
*/
uid = crgetuid(cr);
if (zfsvfs->z_use_fuids == B_FALSE &&
(vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
return (SET_ERROR(EINVAL));
if (dirname == NULL)
return (SET_ERROR(EINVAL));
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(dzp);
zilog = zfsvfs->z_log;
if (dzp->z_pflags & ZFS_XATTR) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EINVAL));
}
if (zfsvfs->z_utf8 && u8_validate(dirname,
strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EILSEQ));
}
if (flags & FIGNORECASE)
zf |= ZCILOOK;
if (vap->va_mask & ATTR_XVATTR) {
if ((error = secpolicy_xvattr((xvattr_t *)vap,
crgetuid(cr), cr, vap->va_mode)) != 0) {
ZFS_EXIT(zfsvfs);
return (error);
}
}
if ((error = zfs_acl_ids_create(dzp, 0, vap, cr,
vsecp, &acl_ids)) != 0) {
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* First make sure the new directory doesn't exist.
*
* Existence is checked first to make sure we don't return
* EACCES instead of EEXIST which can cause some applications
* to fail.
*/
top:
*zpp = NULL;
if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
NULL, NULL))) {
zfs_acl_ids_free(&acl_ids);
ZFS_EXIT(zfsvfs);
return (error);
}
if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) {
zfs_acl_ids_free(&acl_ids);
zfs_dirent_unlock(dl);
ZFS_EXIT(zfsvfs);
return (error);
}
if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) {
zfs_acl_ids_free(&acl_ids);
zfs_dirent_unlock(dl);
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EDQUOT));
}
/*
* Add a new entry to the directory.
*/
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
fuid_dirtied = zfsvfs->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
acl_ids.z_aclp->z_acl_bytes);
}
dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
ZFS_SA_BASE_ATTR_SIZE);
error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
if (error) {
zfs_dirent_unlock(dl);
if (error == ERESTART) {
waited = B_TRUE;
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
zfs_acl_ids_free(&acl_ids);
dmu_tx_abort(tx);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Create new node.
*/
zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
/*
* Now put new name in parent dir.
*/
error = zfs_link_create(dl, zp, tx, ZNEW);
if (error != 0) {
zfs_znode_delete(zp, tx);
remove_inode_hash(ZTOI(zp));
goto out;
}
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
*zpp = zp;
txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
if (flags & FIGNORECASE)
txtype |= TX_CI;
zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
acl_ids.z_fuidp, vap);
out:
zfs_acl_ids_free(&acl_ids);
dmu_tx_commit(tx);
zfs_dirent_unlock(dl);
if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zilog, 0);
if (error != 0) {
zrele(zp);
} else {
zfs_znode_update_vfs(dzp);
zfs_znode_update_vfs(zp);
}
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Remove a directory subdir entry. If the current working
* directory is the same as the subdir to be removed, the
* remove will fail.
*
* IN: dzp - znode of directory to remove from.
* name - name of directory to be removed.
* cwd - inode of current working directory.
* cr - credentials of caller.
* flags - case flags
*
* RETURN: 0 on success, error code on failure.
*
* Timestamps:
* dzp - ctime|mtime updated
*/
/*ARGSUSED*/
int
zfs_rmdir(znode_t *dzp, char *name, znode_t *cwd, cred_t *cr,
int flags)
{
znode_t *zp;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
zilog_t *zilog;
zfs_dirlock_t *dl;
dmu_tx_t *tx;
int error;
int zflg = ZEXISTS;
boolean_t waited = B_FALSE;
if (name == NULL)
return (SET_ERROR(EINVAL));
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(dzp);
zilog = zfsvfs->z_log;
if (flags & FIGNORECASE)
zflg |= ZCILOOK;
top:
zp = NULL;
/*
* Attempt to lock directory; fail if entry doesn't exist.
*/
if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
NULL, NULL))) {
ZFS_EXIT(zfsvfs);
return (error);
}
if ((error = zfs_zaccess_delete(dzp, zp, cr))) {
goto out;
}
if (!S_ISDIR(ZTOI(zp)->i_mode)) {
error = SET_ERROR(ENOTDIR);
goto out;
}
if (zp == cwd) {
error = SET_ERROR(EINVAL);
goto out;
}
/*
* Grab a lock on the directory to make sure that no one is
* trying to add (or lookup) entries while we are removing it.
*/
rw_enter(&zp->z_name_lock, RW_WRITER);
/*
* Grab a lock on the parent pointer to make sure we play well
* with the treewalk and directory rename code.
*/
rw_enter(&zp->z_parent_lock, RW_WRITER);
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
zfs_sa_upgrade_txholds(tx, zp);
zfs_sa_upgrade_txholds(tx, dzp);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
if (error) {
rw_exit(&zp->z_parent_lock);
rw_exit(&zp->z_name_lock);
zfs_dirent_unlock(dl);
if (error == ERESTART) {
waited = B_TRUE;
dmu_tx_wait(tx);
dmu_tx_abort(tx);
zrele(zp);
goto top;
}
dmu_tx_abort(tx);
zrele(zp);
ZFS_EXIT(zfsvfs);
return (error);
}
error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
if (error == 0) {
uint64_t txtype = TX_RMDIR;
if (flags & FIGNORECASE)
txtype |= TX_CI;
zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT,
B_FALSE);
}
dmu_tx_commit(tx);
rw_exit(&zp->z_parent_lock);
rw_exit(&zp->z_name_lock);
out:
zfs_dirent_unlock(dl);
zfs_znode_update_vfs(dzp);
zfs_znode_update_vfs(zp);
zrele(zp);
if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zilog, 0);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Read directory entries from the given directory cursor position and emit
* name and position for each entry.
*
* IN: ip - inode of directory to read.
* ctx - directory entry context.
* cr - credentials of caller.
*
* RETURN: 0 if success
* error code if failure
*
* Timestamps:
* ip - atime updated
*
* Note that the low 4 bits of the cookie returned by zap is always zero.
* This allows us to use the low range for "special" directory entries:
* We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
* we use the offset 2 for the '.zfs' directory.
*/
/* ARGSUSED */
int
zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
objset_t *os;
zap_cursor_t zc;
zap_attribute_t zap;
int error;
uint8_t prefetch;
uint8_t type;
int done = 0;
uint64_t parent;
uint64_t offset; /* must be unsigned; checks for < 1 */
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&parent, sizeof (parent))) != 0)
goto out;
/*
* Quit if directory has been removed (posix)
*/
if (zp->z_unlinked)
goto out;
error = 0;
os = zfsvfs->z_os;
offset = ctx->pos;
prefetch = zp->z_zn_prefetch;
/*
* Initialize the iterator cursor.
*/
if (offset <= 3) {
/*
* Start iteration from the beginning of the directory.
*/
zap_cursor_init(&zc, os, zp->z_id);
} else {
/*
* The offset is a serialized cursor.
*/
zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
}
/*
* Transform to file-system independent format
*/
while (!done) {
uint64_t objnum;
/*
* Special case `.', `..', and `.zfs'.
*/
if (offset == 0) {
(void) strcpy(zap.za_name, ".");
zap.za_normalization_conflict = 0;
objnum = zp->z_id;
type = DT_DIR;
} else if (offset == 1) {
(void) strcpy(zap.za_name, "..");
zap.za_normalization_conflict = 0;
objnum = parent;
type = DT_DIR;
} else if (offset == 2 && zfs_show_ctldir(zp)) {
(void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
zap.za_normalization_conflict = 0;
objnum = ZFSCTL_INO_ROOT;
type = DT_DIR;
} else {
/*
* Grab next entry.
*/
if ((error = zap_cursor_retrieve(&zc, &zap))) {
if (error == ENOENT)
break;
else
goto update;
}
/*
* Allow multiple entries provided the first entry is
* the object id. Non-zpl consumers may safely make
* use of the additional space.
*
* XXX: This should be a feature flag for compatibility
*/
if (zap.za_integer_length != 8 ||
zap.za_num_integers == 0) {
cmn_err(CE_WARN, "zap_readdir: bad directory "
"entry, obj = %lld, offset = %lld, "
"length = %d, num = %lld\n",
(u_longlong_t)zp->z_id,
(u_longlong_t)offset,
zap.za_integer_length,
(u_longlong_t)zap.za_num_integers);
error = SET_ERROR(ENXIO);
goto update;
}
objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
type = ZFS_DIRENT_TYPE(zap.za_first_integer);
}
done = !zpl_dir_emit(ctx, zap.za_name, strlen(zap.za_name),
objnum, type);
if (done)
break;
/* Prefetch znode */
if (prefetch) {
dmu_prefetch(os, objnum, 0, 0, 0,
ZIO_PRIORITY_SYNC_READ);
}
/*
* Move to the next entry, fill in the previous offset.
*/
if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
zap_cursor_advance(&zc);
offset = zap_cursor_serialize(&zc);
} else {
offset += 1;
}
ctx->pos = offset;
}
zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
update:
zap_cursor_fini(&zc);
if (error == ENOENT)
error = 0;
out:
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Get the basic file attributes and place them in the provided kstat
* structure. The inode is assumed to be the authoritative source
* for most of the attributes. However, the znode currently has the
* authoritative atime, blksize, and block count.
*
* IN: ip - inode of file.
*
* OUT: sp - kstat values.
*
* RETURN: 0 (always succeeds)
*/
/* ARGSUSED */
int
zfs_getattr_fast(struct user_namespace *user_ns, struct inode *ip,
struct kstat *sp)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
uint32_t blksize;
u_longlong_t nblocks;
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
mutex_enter(&zp->z_lock);
zpl_generic_fillattr(user_ns, ip, sp);
/*
* +1 link count for root inode with visible '.zfs' directory.
*/
if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp))
if (sp->nlink < ZFS_LINK_MAX)
sp->nlink++;
sa_object_size(zp->z_sa_hdl, &blksize, &nblocks);
sp->blksize = blksize;
sp->blocks = nblocks;
if (unlikely(zp->z_blksz == 0)) {
/*
* Block size hasn't been set; suggest maximal I/O transfers.
*/
sp->blksize = zfsvfs->z_max_blksz;
}
mutex_exit(&zp->z_lock);
/*
* Required to prevent NFS client from detecting different inode
* numbers of snapshot root dentry before and after snapshot mount.
*/
if (zfsvfs->z_issnap) {
if (ip->i_sb->s_root->d_inode == ip)
sp->ino = ZFSCTL_INO_SNAPDIRS -
dmu_objset_id(zfsvfs->z_os);
}
ZFS_EXIT(zfsvfs);
return (0);
}
/*
* For the operation of changing file's user/group/project, we need to
* handle not only the main object that is assigned to the file directly,
* but also the ones that are used by the file via hidden xattr directory.
*
* Because the xattr directory may contains many EA entries, as to it may
* be impossible to change all of them via the transaction of changing the
* main object's user/group/project attributes. Then we have to change them
* via other multiple independent transactions one by one. It may be not good
* solution, but we have no better idea yet.
*/
static int
zfs_setattr_dir(znode_t *dzp)
{
struct inode *dxip = ZTOI(dzp);
struct inode *xip = NULL;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
objset_t *os = zfsvfs->z_os;
zap_cursor_t zc;
zap_attribute_t zap;
zfs_dirlock_t *dl;
znode_t *zp = NULL;
dmu_tx_t *tx = NULL;
uint64_t uid, gid;
sa_bulk_attr_t bulk[4];
int count;
int err;
zap_cursor_init(&zc, os, dzp->z_id);
while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) {
count = 0;
if (zap.za_integer_length != 8 || zap.za_num_integers != 1) {
err = ENXIO;
break;
}
err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp,
ZEXISTS, NULL, NULL);
if (err == ENOENT)
goto next;
if (err)
break;
xip = ZTOI(zp);
if (KUID_TO_SUID(xip->i_uid) == KUID_TO_SUID(dxip->i_uid) &&
KGID_TO_SGID(xip->i_gid) == KGID_TO_SGID(dxip->i_gid) &&
zp->z_projid == dzp->z_projid)
goto next;
tx = dmu_tx_create(os);
if (!(zp->z_pflags & ZFS_PROJID))
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
else
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err)
break;
mutex_enter(&dzp->z_lock);
if (KUID_TO_SUID(xip->i_uid) != KUID_TO_SUID(dxip->i_uid)) {
xip->i_uid = dxip->i_uid;
uid = zfs_uid_read(dxip);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
&uid, sizeof (uid));
}
if (KGID_TO_SGID(xip->i_gid) != KGID_TO_SGID(dxip->i_gid)) {
xip->i_gid = dxip->i_gid;
gid = zfs_gid_read(dxip);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
&gid, sizeof (gid));
}
if (zp->z_projid != dzp->z_projid) {
if (!(zp->z_pflags & ZFS_PROJID)) {
zp->z_pflags |= ZFS_PROJID;
SA_ADD_BULK_ATTR(bulk, count,
SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags,
sizeof (zp->z_pflags));
}
zp->z_projid = dzp->z_projid;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs),
NULL, &zp->z_projid, sizeof (zp->z_projid));
}
mutex_exit(&dzp->z_lock);
if (likely(count > 0)) {
err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
dmu_tx_commit(tx);
} else {
dmu_tx_abort(tx);
}
tx = NULL;
if (err != 0 && err != ENOENT)
break;
next:
if (zp) {
zrele(zp);
zp = NULL;
zfs_dirent_unlock(dl);
}
zap_cursor_advance(&zc);
}
if (tx)
dmu_tx_abort(tx);
if (zp) {
zrele(zp);
zfs_dirent_unlock(dl);
}
zap_cursor_fini(&zc);
return (err == ENOENT ? 0 : err);
}
/*
* Set the file attributes to the values contained in the
* vattr structure.
*
* IN: zp - znode of file to be modified.
* vap - new attribute values.
* If ATTR_XVATTR set, then optional attrs are being set
* flags - ATTR_UTIME set if non-default time values provided.
* - ATTR_NOACLCHECK (CIFS context only).
* cr - credentials of caller.
*
* RETURN: 0 if success
* error code if failure
*
* Timestamps:
* ip - ctime updated, mtime updated if size changed.
*/
/* ARGSUSED */
int
zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr)
{
struct inode *ip;
zfsvfs_t *zfsvfs = ZTOZSB(zp);
objset_t *os = zfsvfs->z_os;
zilog_t *zilog;
dmu_tx_t *tx;
vattr_t oldva;
xvattr_t *tmpxvattr;
uint_t mask = vap->va_mask;
uint_t saved_mask = 0;
int trim_mask = 0;
uint64_t new_mode;
uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid;
uint64_t xattr_obj;
uint64_t mtime[2], ctime[2], atime[2];
uint64_t projid = ZFS_INVALID_PROJID;
znode_t *attrzp;
int need_policy = FALSE;
int err, err2 = 0;
zfs_fuid_info_t *fuidp = NULL;
xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
xoptattr_t *xoap;
zfs_acl_t *aclp;
boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
boolean_t fuid_dirtied = B_FALSE;
boolean_t handle_eadir = B_FALSE;
sa_bulk_attr_t *bulk, *xattr_bulk;
int count = 0, xattr_count = 0, bulks = 8;
if (mask == 0)
return (0);
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
ip = ZTOI(zp);
/*
* If this is a xvattr_t, then get a pointer to the structure of
* optional attributes. If this is NULL, then we have a vattr_t.
*/
xoap = xva_getxoptattr(xvap);
if (xoap != NULL && (mask & ATTR_XVATTR)) {
if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
if (!dmu_objset_projectquota_enabled(os) ||
(!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode))) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(ENOTSUP));
}
projid = xoap->xoa_projid;
if (unlikely(projid == ZFS_INVALID_PROJID)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EINVAL));
}
if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID)
projid = ZFS_INVALID_PROJID;
else
need_policy = TRUE;
}
if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) &&
(xoap->xoa_projinherit !=
((zp->z_pflags & ZFS_PROJINHERIT) != 0)) &&
(!dmu_objset_projectquota_enabled(os) ||
(!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode)))) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(ENOTSUP));
}
}
zilog = zfsvfs->z_log;
/*
* Make sure that if we have ephemeral uid/gid or xvattr specified
* that file system is at proper version level
*/
if (zfsvfs->z_use_fuids == B_FALSE &&
(((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) ||
((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) ||
(mask & ATTR_XVATTR))) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EINVAL));
}
if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EISDIR));
}
if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EINVAL));
}
tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP);
xva_init(tmpxvattr);
bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP);
xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP);
/*
* Immutable files can only alter immutable bit and atime
*/
if ((zp->z_pflags & ZFS_IMMUTABLE) &&
((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) ||
((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
err = SET_ERROR(EPERM);
goto out3;
}
if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) {
err = SET_ERROR(EPERM);
goto out3;
}
/*
* Verify timestamps doesn't overflow 32 bits.
* ZFS can handle large timestamps, but 32bit syscalls can't
* handle times greater than 2039. This check should be removed
* once large timestamps are fully supported.
*/
if (mask & (ATTR_ATIME | ATTR_MTIME)) {
if (((mask & ATTR_ATIME) &&
TIMESPEC_OVERFLOW(&vap->va_atime)) ||
((mask & ATTR_MTIME) &&
TIMESPEC_OVERFLOW(&vap->va_mtime))) {
err = SET_ERROR(EOVERFLOW);
goto out3;
}
}
top:
attrzp = NULL;
aclp = NULL;
/* Can this be moved to before the top label? */
if (zfs_is_readonly(zfsvfs)) {
err = SET_ERROR(EROFS);
goto out3;
}
/*
* First validate permissions
*/
if (mask & ATTR_SIZE) {
err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
if (err)
goto out3;
/*
* XXX - Note, we are not providing any open
* mode flags here (like FNDELAY), so we may
* block if there are locks present... this
* should be addressed in openat().
*/
/* XXX - would it be OK to generate a log record here? */
err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
if (err)
goto out3;
}
if (mask & (ATTR_ATIME|ATTR_MTIME) ||
((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
XVA_ISSET_REQ(xvap, XAT_READONLY) ||
XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
XVA_ISSET_REQ(xvap, XAT_OFFLINE) ||
XVA_ISSET_REQ(xvap, XAT_SPARSE) ||
XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) {
need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
skipaclchk, cr);
}
if (mask & (ATTR_UID|ATTR_GID)) {
int idmask = (mask & (ATTR_UID|ATTR_GID));
int take_owner;
int take_group;
/*
* NOTE: even if a new mode is being set,
* we may clear S_ISUID/S_ISGID bits.
*/
if (!(mask & ATTR_MODE))
vap->va_mode = zp->z_mode;
/*
* Take ownership or chgrp to group we are a member of
*/
take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr));
take_group = (mask & ATTR_GID) &&
zfs_groupmember(zfsvfs, vap->va_gid, cr);
/*
* If both ATTR_UID and ATTR_GID are set then take_owner and
* take_group must both be set in order to allow taking
* ownership.
*
* Otherwise, send the check through secpolicy_vnode_setattr()
*
*/
if (((idmask == (ATTR_UID|ATTR_GID)) &&
take_owner && take_group) ||
((idmask == ATTR_UID) && take_owner) ||
((idmask == ATTR_GID) && take_group)) {
if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
skipaclchk, cr) == 0) {
/*
* Remove setuid/setgid for non-privileged users
*/
(void) secpolicy_setid_clear(vap, cr);
trim_mask = (mask & (ATTR_UID|ATTR_GID));
} else {
need_policy = TRUE;
}
} else {
need_policy = TRUE;
}
}
mutex_enter(&zp->z_lock);
oldva.va_mode = zp->z_mode;
zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
if (mask & ATTR_XVATTR) {
/*
* Update xvattr mask to include only those attributes
* that are actually changing.
*
* the bits will be restored prior to actually setting
* the attributes so the caller thinks they were set.
*/
if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
if (xoap->xoa_appendonly !=
((zp->z_pflags & ZFS_APPENDONLY) != 0)) {
need_policy = TRUE;
} else {
XVA_CLR_REQ(xvap, XAT_APPENDONLY);
XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY);
}
}
if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) {
if (xoap->xoa_projinherit !=
((zp->z_pflags & ZFS_PROJINHERIT) != 0)) {
need_policy = TRUE;
} else {
XVA_CLR_REQ(xvap, XAT_PROJINHERIT);
XVA_SET_REQ(tmpxvattr, XAT_PROJINHERIT);
}
}
if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
if (xoap->xoa_nounlink !=
((zp->z_pflags & ZFS_NOUNLINK) != 0)) {
need_policy = TRUE;
} else {
XVA_CLR_REQ(xvap, XAT_NOUNLINK);
XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK);
}
}
if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
if (xoap->xoa_immutable !=
((zp->z_pflags & ZFS_IMMUTABLE) != 0)) {
need_policy = TRUE;
} else {
XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE);
}
}
if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
if (xoap->xoa_nodump !=
((zp->z_pflags & ZFS_NODUMP) != 0)) {
need_policy = TRUE;
} else {
XVA_CLR_REQ(xvap, XAT_NODUMP);
XVA_SET_REQ(tmpxvattr, XAT_NODUMP);
}
}
if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
if (xoap->xoa_av_modified !=
((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) {
need_policy = TRUE;
} else {
XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED);
}
}
if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
if ((!S_ISREG(ip->i_mode) &&
xoap->xoa_av_quarantined) ||
xoap->xoa_av_quarantined !=
((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) {
need_policy = TRUE;
} else {
XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED);
}
}
if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
mutex_exit(&zp->z_lock);
err = SET_ERROR(EPERM);
goto out3;
}
if (need_policy == FALSE &&
(XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
need_policy = TRUE;
}
}
mutex_exit(&zp->z_lock);
if (mask & ATTR_MODE) {
if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
err = secpolicy_setid_setsticky_clear(ip, vap,
&oldva, cr);
if (err)
goto out3;
trim_mask |= ATTR_MODE;
} else {
need_policy = TRUE;
}
}
if (need_policy) {
/*
* If trim_mask is set then take ownership
* has been granted or write_acl is present and user
* has the ability to modify mode. In that case remove
* UID|GID and or MODE from mask so that
* secpolicy_vnode_setattr() doesn't revoke it.
*/
if (trim_mask) {
saved_mask = vap->va_mask;
vap->va_mask &= ~trim_mask;
}
err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags,
(int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
if (err)
goto out3;
if (trim_mask)
vap->va_mask |= saved_mask;
}
/*
* secpolicy_vnode_setattr, or take ownership may have
* changed va_mask
*/
mask = vap->va_mask;
if ((mask & (ATTR_UID | ATTR_GID)) || projid != ZFS_INVALID_PROJID) {
handle_eadir = B_TRUE;
err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
&xattr_obj, sizeof (xattr_obj));
if (err == 0 && xattr_obj) {
err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp);
if (err)
goto out2;
}
if (mask & ATTR_UID) {
new_kuid = zfs_fuid_create(zfsvfs,
(uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) &&
zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT,
new_kuid)) {
if (attrzp)
zrele(attrzp);
err = SET_ERROR(EDQUOT);
goto out2;
}
}
if (mask & ATTR_GID) {
new_kgid = zfs_fuid_create(zfsvfs,
(uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp);
if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) &&
zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT,
new_kgid)) {
if (attrzp)
zrele(attrzp);
err = SET_ERROR(EDQUOT);
goto out2;
}
}
if (projid != ZFS_INVALID_PROJID &&
zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) {
if (attrzp)
zrele(attrzp);
err = EDQUOT;
goto out2;
}
}
tx = dmu_tx_create(os);
if (mask & ATTR_MODE) {
uint64_t pmode = zp->z_mode;
uint64_t acl_obj;
new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
if (ZTOZSB(zp)->z_acl_mode == ZFS_ACL_RESTRICTED &&
!(zp->z_pflags & ZFS_ACL_TRIVIAL)) {
err = EPERM;
goto out;
}
if ((err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)))
goto out;
mutex_enter(&zp->z_lock);
if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) {
/*
* Are we upgrading ACL from old V0 format
* to V1 format?
*/
if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
zfs_znode_acl_version(zp) ==
ZFS_ACL_VERSION_INITIAL) {
dmu_tx_hold_free(tx, acl_obj, 0,
DMU_OBJECT_END);
dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
0, aclp->z_acl_bytes);
} else {
dmu_tx_hold_write(tx, acl_obj, 0,
aclp->z_acl_bytes);
}
} else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
0, aclp->z_acl_bytes);
}
mutex_exit(&zp->z_lock);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
} else {
if (((mask & ATTR_XVATTR) &&
XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) ||
(projid != ZFS_INVALID_PROJID &&
!(zp->z_pflags & ZFS_PROJID)))
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
else
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
}
if (attrzp) {
dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE);
}
fuid_dirtied = zfsvfs->z_fuid_dirty;
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
zfs_sa_upgrade_txholds(tx, zp);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err)
goto out;
count = 0;
/*
* Set each attribute requested.
* We group settings according to the locks they need to acquire.
*
* Note: you cannot set ctime directly, although it will be
* updated as a side-effect of calling this function.
*/
if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) {
/*
* For the existed object that is upgraded from old system,
* its on-disk layout has no slot for the project ID attribute.
* But quota accounting logic needs to access related slots by
* offset directly. So we need to adjust old objects' layout
* to make the project ID to some unified and fixed offset.
*/
if (attrzp)
err = sa_add_projid(attrzp->z_sa_hdl, tx, projid);
if (err == 0)
err = sa_add_projid(zp->z_sa_hdl, tx, projid);
if (unlikely(err == EEXIST))
err = 0;
else if (err != 0)
goto out;
else
projid = ZFS_INVALID_PROJID;
}
if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE))
mutex_enter(&zp->z_acl_lock);
mutex_enter(&zp->z_lock);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
if (attrzp) {
if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE))
mutex_enter(&attrzp->z_acl_lock);
mutex_enter(&attrzp->z_lock);
SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags,
sizeof (attrzp->z_pflags));
if (projid != ZFS_INVALID_PROJID) {
attrzp->z_projid = projid;
SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid,
sizeof (attrzp->z_projid));
}
}
if (mask & (ATTR_UID|ATTR_GID)) {
if (mask & ATTR_UID) {
ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid);
new_uid = zfs_uid_read(ZTOI(zp));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
&new_uid, sizeof (new_uid));
if (attrzp) {
SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
SA_ZPL_UID(zfsvfs), NULL, &new_uid,
sizeof (new_uid));
ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid);
}
}
if (mask & ATTR_GID) {
ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid);
new_gid = zfs_gid_read(ZTOI(zp));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs),
NULL, &new_gid, sizeof (new_gid));
if (attrzp) {
SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
SA_ZPL_GID(zfsvfs), NULL, &new_gid,
sizeof (new_gid));
ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid);
}
}
if (!(mask & ATTR_MODE)) {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs),
NULL, &new_mode, sizeof (new_mode));
new_mode = zp->z_mode;
}
err = zfs_acl_chown_setattr(zp);
ASSERT(err == 0);
if (attrzp) {
err = zfs_acl_chown_setattr(attrzp);
ASSERT(err == 0);
}
}
if (mask & ATTR_MODE) {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
&new_mode, sizeof (new_mode));
zp->z_mode = ZTOI(zp)->i_mode = new_mode;
ASSERT3P(aclp, !=, NULL);
err = zfs_aclset_common(zp, aclp, cr, tx);
ASSERT0(err);
if (zp->z_acl_cached)
zfs_acl_free(zp->z_acl_cached);
zp->z_acl_cached = aclp;
aclp = NULL;
}
if ((mask & ATTR_ATIME) || zp->z_atime_dirty) {
zp->z_atime_dirty = B_FALSE;
ZFS_TIME_ENCODE(&ip->i_atime, atime);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
&atime, sizeof (atime));
}
if (mask & (ATTR_MTIME | ATTR_SIZE)) {
ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
ZTOI(zp)->i_mtime = zpl_inode_timestamp_truncate(
vap->va_mtime, ZTOI(zp));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
mtime, sizeof (mtime));
}
if (mask & (ATTR_CTIME | ATTR_SIZE)) {
ZFS_TIME_ENCODE(&vap->va_ctime, ctime);
ZTOI(zp)->i_ctime = zpl_inode_timestamp_truncate(vap->va_ctime,
ZTOI(zp));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
ctime, sizeof (ctime));
}
if (projid != ZFS_INVALID_PROJID) {
zp->z_projid = projid;
SA_ADD_BULK_ATTR(bulk, count,
SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid,
sizeof (zp->z_projid));
}
if (attrzp && mask) {
SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
SA_ZPL_CTIME(zfsvfs), NULL, &ctime,
sizeof (ctime));
}
/*
* Do this after setting timestamps to prevent timestamp
* update from toggling bit
*/
if (xoap && (mask & ATTR_XVATTR)) {
/*
* restore trimmed off masks
* so that return masks can be set for caller.
*/
if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) {
XVA_SET_REQ(xvap, XAT_APPENDONLY);
}
if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) {
XVA_SET_REQ(xvap, XAT_NOUNLINK);
}
if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) {
XVA_SET_REQ(xvap, XAT_IMMUTABLE);
}
if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) {
XVA_SET_REQ(xvap, XAT_NODUMP);
}
if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) {
XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
}
if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) {
XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
}
if (XVA_ISSET_REQ(tmpxvattr, XAT_PROJINHERIT)) {
XVA_SET_REQ(xvap, XAT_PROJINHERIT);
}
if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
ASSERT(S_ISREG(ip->i_mode));
zfs_xvattr_set(zp, xvap, tx);
}
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
if (mask != 0)
zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
mutex_exit(&zp->z_lock);
if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE))
mutex_exit(&zp->z_acl_lock);
if (attrzp) {
if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE))
mutex_exit(&attrzp->z_acl_lock);
mutex_exit(&attrzp->z_lock);
}
out:
if (err == 0 && xattr_count > 0) {
err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk,
xattr_count, tx);
ASSERT(err2 == 0);
}
if (aclp)
zfs_acl_free(aclp);
if (fuidp) {
zfs_fuid_info_free(fuidp);
fuidp = NULL;
}
if (err) {
dmu_tx_abort(tx);
if (attrzp)
zrele(attrzp);
if (err == ERESTART)
goto top;
} else {
if (count > 0)
err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
dmu_tx_commit(tx);
if (attrzp) {
if (err2 == 0 && handle_eadir)
err2 = zfs_setattr_dir(attrzp);
zrele(attrzp);
}
zfs_znode_update_vfs(zp);
}
out2:
if (os->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zilog, 0);
out3:
kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks);
kmem_free(bulk, sizeof (sa_bulk_attr_t) * bulks);
kmem_free(tmpxvattr, sizeof (xvattr_t));
ZFS_EXIT(zfsvfs);
return (err);
}
typedef struct zfs_zlock {
krwlock_t *zl_rwlock; /* lock we acquired */
znode_t *zl_znode; /* znode we held */
struct zfs_zlock *zl_next; /* next in list */
} zfs_zlock_t;
/*
* Drop locks and release vnodes that were held by zfs_rename_lock().
*/
static void
zfs_rename_unlock(zfs_zlock_t **zlpp)
{
zfs_zlock_t *zl;
while ((zl = *zlpp) != NULL) {
if (zl->zl_znode != NULL)
zfs_zrele_async(zl->zl_znode);
rw_exit(zl->zl_rwlock);
*zlpp = zl->zl_next;
kmem_free(zl, sizeof (*zl));
}
}
/*
* Search back through the directory tree, using the ".." entries.
* Lock each directory in the chain to prevent concurrent renames.
* Fail any attempt to move a directory into one of its own descendants.
* XXX - z_parent_lock can overlap with map or grow locks
*/
static int
zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
{
zfs_zlock_t *zl;
znode_t *zp = tdzp;
uint64_t rootid = ZTOZSB(zp)->z_root;
uint64_t oidp = zp->z_id;
krwlock_t *rwlp = &szp->z_parent_lock;
krw_t rw = RW_WRITER;
/*
* First pass write-locks szp and compares to zp->z_id.
* Later passes read-lock zp and compare to zp->z_parent.
*/
do {
if (!rw_tryenter(rwlp, rw)) {
/*
* Another thread is renaming in this path.
* Note that if we are a WRITER, we don't have any
* parent_locks held yet.
*/
if (rw == RW_READER && zp->z_id > szp->z_id) {
/*
* Drop our locks and restart
*/
zfs_rename_unlock(&zl);
*zlpp = NULL;
zp = tdzp;
oidp = zp->z_id;
rwlp = &szp->z_parent_lock;
rw = RW_WRITER;
continue;
} else {
/*
* Wait for other thread to drop its locks
*/
rw_enter(rwlp, rw);
}
}
zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
zl->zl_rwlock = rwlp;
zl->zl_znode = NULL;
zl->zl_next = *zlpp;
*zlpp = zl;
if (oidp == szp->z_id) /* We're a descendant of szp */
return (SET_ERROR(EINVAL));
if (oidp == rootid) /* We've hit the top */
return (0);
if (rw == RW_READER) { /* i.e. not the first pass */
int error = zfs_zget(ZTOZSB(zp), oidp, &zp);
if (error)
return (error);
zl->zl_znode = zp;
}
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)),
&oidp, sizeof (oidp));
rwlp = &zp->z_parent_lock;
rw = RW_READER;
} while (zp->z_id != sdzp->z_id);
return (0);
}
/*
* Move an entry from the provided source directory to the target
* directory. Change the entry name as indicated.
*
* IN: sdzp - Source directory containing the "old entry".
* snm - Old entry name.
* tdzp - Target directory to contain the "new entry".
* tnm - New entry name.
* cr - credentials of caller.
* flags - case flags
*
* RETURN: 0 on success, error code on failure.
*
* Timestamps:
* sdzp,tdzp - ctime|mtime updated
*/
/*ARGSUSED*/
int
zfs_rename(znode_t *sdzp, char *snm, znode_t *tdzp, char *tnm,
cred_t *cr, int flags)
{
znode_t *szp, *tzp;
zfsvfs_t *zfsvfs = ZTOZSB(sdzp);
zilog_t *zilog;
zfs_dirlock_t *sdl, *tdl;
dmu_tx_t *tx;
zfs_zlock_t *zl;
int cmp, serr, terr;
int error = 0;
int zflg = 0;
boolean_t waited = B_FALSE;
if (snm == NULL || tnm == NULL)
return (SET_ERROR(EINVAL));
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(sdzp);
zilog = zfsvfs->z_log;
ZFS_VERIFY_ZP(tdzp);
/*
* We check i_sb because snapshots and the ctldir must have different
* super blocks.
*/
if (ZTOI(tdzp)->i_sb != ZTOI(sdzp)->i_sb ||
zfsctl_is_node(ZTOI(tdzp))) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EXDEV));
}
if (zfsvfs->z_utf8 && u8_validate(tnm,
strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EILSEQ));
}
if (flags & FIGNORECASE)
zflg |= ZCILOOK;
top:
szp = NULL;
tzp = NULL;
zl = NULL;
/*
* This is to prevent the creation of links into attribute space
* by renaming a linked file into/outof an attribute directory.
* See the comment in zfs_link() for why this is considered bad.
*/
if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EINVAL));
}
/*
* Lock source and target directory entries. To prevent deadlock,
* a lock ordering must be defined. We lock the directory with
* the smallest object id first, or if it's a tie, the one with
* the lexically first name.
*/
if (sdzp->z_id < tdzp->z_id) {
cmp = -1;
} else if (sdzp->z_id > tdzp->z_id) {
cmp = 1;
} else {
/*
* First compare the two name arguments without
* considering any case folding.
*/
int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
ASSERT(error == 0 || !zfsvfs->z_utf8);
if (cmp == 0) {
/*
* POSIX: "If the old argument and the new argument
* both refer to links to the same existing file,
* the rename() function shall return successfully
* and perform no other action."
*/
ZFS_EXIT(zfsvfs);
return (0);
}
/*
* If the file system is case-folding, then we may
* have some more checking to do. A case-folding file
* system is either supporting mixed case sensitivity
* access or is completely case-insensitive. Note
* that the file system is always case preserving.
*
* In mixed sensitivity mode case sensitive behavior
* is the default. FIGNORECASE must be used to
* explicitly request case insensitive behavior.
*
* If the source and target names provided differ only
* by case (e.g., a request to rename 'tim' to 'Tim'),
* we will treat this as a special case in the
* case-insensitive mode: as long as the source name
* is an exact match, we will allow this to proceed as
* a name-change request.
*/
if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
(zfsvfs->z_case == ZFS_CASE_MIXED &&
flags & FIGNORECASE)) &&
u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
&error) == 0) {
/*
* case preserving rename request, require exact
* name matches
*/
zflg |= ZCIEXACT;
zflg &= ~ZCILOOK;
}
}
/*
* If the source and destination directories are the same, we should
* grab the z_name_lock of that directory only once.
*/
if (sdzp == tdzp) {
zflg |= ZHAVELOCK;
rw_enter(&sdzp->z_name_lock, RW_READER);
}
if (cmp < 0) {
serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
ZEXISTS | zflg, NULL, NULL);
terr = zfs_dirent_lock(&tdl,
tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
} else {
terr = zfs_dirent_lock(&tdl,
tdzp, tnm, &tzp, zflg, NULL, NULL);
serr = zfs_dirent_lock(&sdl,
sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
NULL, NULL);
}
if (serr) {
/*
* Source entry invalid or not there.
*/
if (!terr) {
zfs_dirent_unlock(tdl);
if (tzp)
zrele(tzp);
}
if (sdzp == tdzp)
rw_exit(&sdzp->z_name_lock);
if (strcmp(snm, "..") == 0)
serr = EINVAL;
ZFS_EXIT(zfsvfs);
return (serr);
}
if (terr) {
zfs_dirent_unlock(sdl);
zrele(szp);
if (sdzp == tdzp)
rw_exit(&sdzp->z_name_lock);
if (strcmp(tnm, "..") == 0)
terr = EINVAL;
ZFS_EXIT(zfsvfs);
return (terr);
}
/*
* If we are using project inheritance, means if the directory has
* ZFS_PROJINHERIT set, then its descendant directories will inherit
* not only the project ID, but also the ZFS_PROJINHERIT flag. Under
* such case, we only allow renames into our tree when the project
* IDs are the same.
*/
if (tdzp->z_pflags & ZFS_PROJINHERIT &&
tdzp->z_projid != szp->z_projid) {
error = SET_ERROR(EXDEV);
goto out;
}
/*
* Must have write access at the source to remove the old entry
* and write access at the target to create the new entry.
* Note that if target and source are the same, this can be
* done in a single check.
*/
if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)))
goto out;
if (S_ISDIR(ZTOI(szp)->i_mode)) {
/*
* Check to make sure rename is valid.
* Can't do a move like this: /usr/a/b to /usr/a/b/c/d
*/
if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl)))
goto out;
}
/*
* Does target exist?
*/
if (tzp) {
/*
* Source and target must be the same type.
*/
if (S_ISDIR(ZTOI(szp)->i_mode)) {
if (!S_ISDIR(ZTOI(tzp)->i_mode)) {
error = SET_ERROR(ENOTDIR);
goto out;
}
} else {
if (S_ISDIR(ZTOI(tzp)->i_mode)) {
error = SET_ERROR(EISDIR);
goto out;
}
}
/*
* POSIX dictates that when the source and target
* entries refer to the same file object, rename
* must do nothing and exit without error.
*/
if (szp->z_id == tzp->z_id) {
error = 0;
goto out;
}
}
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE);
dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
if (sdzp != tdzp) {
dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, tdzp);
}
if (tzp) {
dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, tzp);
}
zfs_sa_upgrade_txholds(tx, szp);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
if (error) {
if (zl != NULL)
zfs_rename_unlock(&zl);
zfs_dirent_unlock(sdl);
zfs_dirent_unlock(tdl);
if (sdzp == tdzp)
rw_exit(&sdzp->z_name_lock);
if (error == ERESTART) {
waited = B_TRUE;
dmu_tx_wait(tx);
dmu_tx_abort(tx);
zrele(szp);
if (tzp)
zrele(tzp);
goto top;
}
dmu_tx_abort(tx);
zrele(szp);
if (tzp)
zrele(tzp);
ZFS_EXIT(zfsvfs);
return (error);
}
if (tzp) /* Attempt to remove the existing target */
error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
if (error == 0) {
error = zfs_link_create(tdl, szp, tx, ZRENAMING);
if (error == 0) {
szp->z_pflags |= ZFS_AV_MODIFIED;
if (tdzp->z_pflags & ZFS_PROJINHERIT)
szp->z_pflags |= ZFS_PROJINHERIT;
error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
(void *)&szp->z_pflags, sizeof (uint64_t), tx);
ASSERT0(error);
error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
if (error == 0) {
zfs_log_rename(zilog, tx, TX_RENAME |
(flags & FIGNORECASE ? TX_CI : 0), sdzp,
sdl->dl_name, tdzp, tdl->dl_name, szp);
} else {
/*
* At this point, we have successfully created
* the target name, but have failed to remove
* the source name. Since the create was done
* with the ZRENAMING flag, there are
* complications; for one, the link count is
* wrong. The easiest way to deal with this
* is to remove the newly created target, and
* return the original error. This must
* succeed; fortunately, it is very unlikely to
* fail, since we just created it.
*/
VERIFY3U(zfs_link_destroy(tdl, szp, tx,
ZRENAMING, NULL), ==, 0);
}
} else {
/*
* If we had removed the existing target, subsequent
* call to zfs_link_create() to add back the same entry
* but, the new dnode (szp) should not fail.
*/
ASSERT(tzp == NULL);
}
}
dmu_tx_commit(tx);
out:
if (zl != NULL)
zfs_rename_unlock(&zl);
zfs_dirent_unlock(sdl);
zfs_dirent_unlock(tdl);
zfs_znode_update_vfs(sdzp);
if (sdzp == tdzp)
rw_exit(&sdzp->z_name_lock);
if (sdzp != tdzp)
zfs_znode_update_vfs(tdzp);
zfs_znode_update_vfs(szp);
zrele(szp);
if (tzp) {
zfs_znode_update_vfs(tzp);
zrele(tzp);
}
if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zilog, 0);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Insert the indicated symbolic reference entry into the directory.
*
* IN: dzp - Directory to contain new symbolic link.
* name - Name of directory entry in dip.
* vap - Attributes of new entry.
* link - Name for new symlink entry.
* cr - credentials of caller.
* flags - case flags
*
* OUT: zpp - Znode for new symbolic link.
*
* RETURN: 0 on success, error code on failure.
*
* Timestamps:
* dip - ctime|mtime updated
*/
/*ARGSUSED*/
int
zfs_symlink(znode_t *dzp, char *name, vattr_t *vap, char *link,
znode_t **zpp, cred_t *cr, int flags)
{
znode_t *zp;
zfs_dirlock_t *dl;
dmu_tx_t *tx;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
zilog_t *zilog;
uint64_t len = strlen(link);
int error;
int zflg = ZNEW;
zfs_acl_ids_t acl_ids;
boolean_t fuid_dirtied;
uint64_t txtype = TX_SYMLINK;
boolean_t waited = B_FALSE;
ASSERT(S_ISLNK(vap->va_mode));
if (name == NULL)
return (SET_ERROR(EINVAL));
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(dzp);
zilog = zfsvfs->z_log;
if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EILSEQ));
}
if (flags & FIGNORECASE)
zflg |= ZCILOOK;
if (len > MAXPATHLEN) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(ENAMETOOLONG));
}
if ((error = zfs_acl_ids_create(dzp, 0,
vap, cr, NULL, &acl_ids)) != 0) {
ZFS_EXIT(zfsvfs);
return (error);
}
top:
*zpp = NULL;
/*
* Attempt to lock directory; fail if entry already exists.
*/
error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
if (error) {
zfs_acl_ids_free(&acl_ids);
ZFS_EXIT(zfsvfs);
return (error);
}
if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
zfs_acl_ids_free(&acl_ids);
zfs_dirent_unlock(dl);
ZFS_EXIT(zfsvfs);
return (error);
}
if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, ZFS_DEFAULT_PROJID)) {
zfs_acl_ids_free(&acl_ids);
zfs_dirent_unlock(dl);
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EDQUOT));
}
tx = dmu_tx_create(zfsvfs->z_os);
fuid_dirtied = zfsvfs->z_fuid_dirty;
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
ZFS_SA_BASE_ATTR_SIZE + len);
dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
acl_ids.z_aclp->z_acl_bytes);
}
if (fuid_dirtied)
zfs_fuid_txhold(zfsvfs, tx);
error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
if (error) {
zfs_dirent_unlock(dl);
if (error == ERESTART) {
waited = B_TRUE;
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
zfs_acl_ids_free(&acl_ids);
dmu_tx_abort(tx);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Create a new object for the symlink.
* for version 4 ZPL datasets the symlink will be an SA attribute
*/
zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
if (fuid_dirtied)
zfs_fuid_sync(zfsvfs, tx);
mutex_enter(&zp->z_lock);
if (zp->z_is_sa)
error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs),
link, len, tx);
else
zfs_sa_symlink(zp, link, len, tx);
mutex_exit(&zp->z_lock);
zp->z_size = len;
(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
&zp->z_size, sizeof (zp->z_size), tx);
/*
* Insert the new object into the directory.
*/
error = zfs_link_create(dl, zp, tx, ZNEW);
if (error != 0) {
zfs_znode_delete(zp, tx);
remove_inode_hash(ZTOI(zp));
} else {
if (flags & FIGNORECASE)
txtype |= TX_CI;
zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
zfs_znode_update_vfs(dzp);
zfs_znode_update_vfs(zp);
}
zfs_acl_ids_free(&acl_ids);
dmu_tx_commit(tx);
zfs_dirent_unlock(dl);
if (error == 0) {
*zpp = zp;
if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zilog, 0);
} else {
zrele(zp);
}
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Return, in the buffer contained in the provided uio structure,
* the symbolic path referred to by ip.
*
* IN: ip - inode of symbolic link
* uio - structure to contain the link path.
* cr - credentials of caller.
*
* RETURN: 0 if success
* error code if failure
*
* Timestamps:
* ip - atime updated
*/
/* ARGSUSED */
int
zfs_readlink(struct inode *ip, zfs_uio_t *uio, cred_t *cr)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
int error;
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
mutex_enter(&zp->z_lock);
if (zp->z_is_sa)
error = sa_lookup_uio(zp->z_sa_hdl,
SA_ZPL_SYMLINK(zfsvfs), uio);
else
error = zfs_sa_readlink(zp, uio);
mutex_exit(&zp->z_lock);
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* Insert a new entry into directory tdzp referencing szp.
*
* IN: tdzp - Directory to contain new entry.
* szp - znode of new entry.
* name - name of new entry.
* cr - credentials of caller.
* flags - case flags.
*
* RETURN: 0 if success
* error code if failure
*
* Timestamps:
* tdzp - ctime|mtime updated
* szp - ctime updated
*/
/* ARGSUSED */
int
zfs_link(znode_t *tdzp, znode_t *szp, char *name, cred_t *cr,
int flags)
{
struct inode *sip = ZTOI(szp);
znode_t *tzp;
zfsvfs_t *zfsvfs = ZTOZSB(tdzp);
zilog_t *zilog;
zfs_dirlock_t *dl;
dmu_tx_t *tx;
int error;
int zf = ZNEW;
uint64_t parent;
uid_t owner;
boolean_t waited = B_FALSE;
boolean_t is_tmpfile = 0;
uint64_t txg;
#ifdef HAVE_TMPFILE
is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE));
#endif
ASSERT(S_ISDIR(ZTOI(tdzp)->i_mode));
if (name == NULL)
return (SET_ERROR(EINVAL));
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(tdzp);
zilog = zfsvfs->z_log;
/*
* POSIX dictates that we return EPERM here.
* Better choices include ENOTSUP or EISDIR.
*/
if (S_ISDIR(sip->i_mode)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EPERM));
}
ZFS_VERIFY_ZP(szp);
/*
* If we are using project inheritance, means if the directory has
* ZFS_PROJINHERIT set, then its descendant directories will inherit
* not only the project ID, but also the ZFS_PROJINHERIT flag. Under
* such case, we only allow hard link creation in our tree when the
* project IDs are the same.
*/
if (tdzp->z_pflags & ZFS_PROJINHERIT &&
tdzp->z_projid != szp->z_projid) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EXDEV));
}
/*
* We check i_sb because snapshots and the ctldir must have different
* super blocks.
*/
if (sip->i_sb != ZTOI(tdzp)->i_sb || zfsctl_is_node(sip)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EXDEV));
}
/* Prevent links to .zfs/shares files */
if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
&parent, sizeof (uint64_t))) != 0) {
ZFS_EXIT(zfsvfs);
return (error);
}
if (parent == zfsvfs->z_shares_dir) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EPERM));
}
if (zfsvfs->z_utf8 && u8_validate(name,
strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EILSEQ));
}
if (flags & FIGNORECASE)
zf |= ZCILOOK;
/*
* We do not support links between attributes and non-attributes
* because of the potential security risk of creating links
* into "normal" file space in order to circumvent restrictions
* imposed in attribute space.
*/
if ((szp->z_pflags & ZFS_XATTR) != (tdzp->z_pflags & ZFS_XATTR)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EINVAL));
}
owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid),
cr, ZFS_OWNER);
if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EPERM));
}
if ((error = zfs_zaccess(tdzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
ZFS_EXIT(zfsvfs);
return (error);
}
top:
/*
* Attempt to lock directory; fail if entry already exists.
*/
error = zfs_dirent_lock(&dl, tdzp, name, &tzp, zf, NULL, NULL);
if (error) {
ZFS_EXIT(zfsvfs);
return (error);
}
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, name);
if (is_tmpfile)
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
zfs_sa_upgrade_txholds(tx, szp);
zfs_sa_upgrade_txholds(tx, tdzp);
error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
if (error) {
zfs_dirent_unlock(dl);
if (error == ERESTART) {
waited = B_TRUE;
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
ZFS_EXIT(zfsvfs);
return (error);
}
/* unmark z_unlinked so zfs_link_create will not reject */
if (is_tmpfile)
szp->z_unlinked = B_FALSE;
error = zfs_link_create(dl, szp, tx, 0);
if (error == 0) {
uint64_t txtype = TX_LINK;
/*
* tmpfile is created to be in z_unlinkedobj, so remove it.
* Also, we don't log in ZIL, because all previous file
* operation on the tmpfile are ignored by ZIL. Instead we
* always wait for txg to sync to make sure all previous
* operation are sync safe.
*/
if (is_tmpfile) {
VERIFY(zap_remove_int(zfsvfs->z_os,
zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0);
} else {
if (flags & FIGNORECASE)
txtype |= TX_CI;
zfs_log_link(zilog, tx, txtype, tdzp, szp, name);
}
} else if (is_tmpfile) {
/* restore z_unlinked since when linking failed */
szp->z_unlinked = B_TRUE;
}
txg = dmu_tx_get_txg(tx);
dmu_tx_commit(tx);
zfs_dirent_unlock(dl);
if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zilog, 0);
if (is_tmpfile && zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED)
txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg);
zfs_znode_update_vfs(tdzp);
zfs_znode_update_vfs(szp);
ZFS_EXIT(zfsvfs);
return (error);
}
static void
zfs_putpage_commit_cb(void *arg)
{
struct page *pp = arg;
ClearPageError(pp);
end_page_writeback(pp);
}
/*
* Push a page out to disk, once the page is on stable storage the
* registered commit callback will be run as notification of completion.
*
* IN: ip - page mapped for inode.
* pp - page to push (page is locked)
* wbc - writeback control data
*
* RETURN: 0 if success
* error code if failure
*
* Timestamps:
* ip - ctime|mtime updated
*/
/* ARGSUSED */
int
zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
loff_t offset;
loff_t pgoff;
unsigned int pglen;
dmu_tx_t *tx;
caddr_t va;
int err = 0;
uint64_t mtime[2], ctime[2];
sa_bulk_attr_t bulk[3];
int cnt = 0;
struct address_space *mapping;
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
ASSERT(PageLocked(pp));
pgoff = page_offset(pp); /* Page byte-offset in file */
offset = i_size_read(ip); /* File length in bytes */
pglen = MIN(PAGE_SIZE, /* Page length in bytes */
P2ROUNDUP(offset, PAGE_SIZE)-pgoff);
/* Page is beyond end of file */
if (pgoff >= offset) {
unlock_page(pp);
ZFS_EXIT(zfsvfs);
return (0);
}
/* Truncate page length to end of file */
if (pgoff + pglen > offset)
pglen = offset - pgoff;
#if 0
/*
* FIXME: Allow mmap writes past its quota. The correct fix
* is to register a page_mkwrite() handler to count the page
* against its quota when it is about to be dirtied.
*/
if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT,
KUID_TO_SUID(ip->i_uid)) ||
zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT,
KGID_TO_SGID(ip->i_gid)) ||
(zp->z_projid != ZFS_DEFAULT_PROJID &&
zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT,
zp->z_projid))) {
err = EDQUOT;
}
#endif
/*
* The ordering here is critical and must adhere to the following
* rules in order to avoid deadlocking in either zfs_read() or
* zfs_free_range() due to a lock inversion.
*
* 1) The page must be unlocked prior to acquiring the range lock.
* This is critical because zfs_read() calls find_lock_page()
* which may block on the page lock while holding the range lock.
*
* 2) Before setting or clearing write back on a page the range lock
* must be held in order to prevent a lock inversion with the
* zfs_free_range() function.
*
* This presents a problem because upon entering this function the
* page lock is already held. To safely acquire the range lock the
* page lock must be dropped. This creates a window where another
* process could truncate, invalidate, dirty, or write out the page.
*
* Therefore, after successfully reacquiring the range and page locks
* the current page state is checked. In the common case everything
* will be as is expected and it can be written out. However, if
* the page state has changed it must be handled accordingly.
*/
mapping = pp->mapping;
redirty_page_for_writepage(wbc, pp);
unlock_page(pp);
zfs_locked_range_t *lr = zfs_rangelock_enter(&zp->z_rangelock,
pgoff, pglen, RL_WRITER);
lock_page(pp);
/* Page mapping changed or it was no longer dirty, we're done */
if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) {
unlock_page(pp);
zfs_rangelock_exit(lr);
ZFS_EXIT(zfsvfs);
return (0);
}
/* Another process started write block if required */
if (PageWriteback(pp)) {
unlock_page(pp);
zfs_rangelock_exit(lr);
if (wbc->sync_mode != WB_SYNC_NONE) {
if (PageWriteback(pp))
#ifdef HAVE_PAGEMAP_FOLIO_WAIT_BIT
folio_wait_bit(page_folio(pp), PG_writeback);
#else
wait_on_page_bit(pp, PG_writeback);
#endif
}
ZFS_EXIT(zfsvfs);
return (0);
}
/* Clear the dirty flag the required locks are held */
if (!clear_page_dirty_for_io(pp)) {
unlock_page(pp);
zfs_rangelock_exit(lr);
ZFS_EXIT(zfsvfs);
return (0);
}
/*
* Counterpart for redirty_page_for_writepage() above. This page
* was in fact not skipped and should not be counted as if it were.
*/
wbc->pages_skipped--;
set_page_writeback(pp);
unlock_page(pp);
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
err = dmu_tx_assign(tx, TXG_NOWAIT);
if (err != 0) {
if (err == ERESTART)
dmu_tx_wait(tx);
dmu_tx_abort(tx);
__set_page_dirty_nobuffers(pp);
ClearPageError(pp);
end_page_writeback(pp);
zfs_rangelock_exit(lr);
ZFS_EXIT(zfsvfs);
return (err);
}
va = kmap(pp);
ASSERT3U(pglen, <=, PAGE_SIZE);
dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx);
kunmap(pp);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, 8);
/* Preserve the mtime and ctime provided by the inode */
ZFS_TIME_ENCODE(&ip->i_mtime, mtime);
ZFS_TIME_ENCODE(&ip->i_ctime, ctime);
zp->z_atime_dirty = B_FALSE;
zp->z_seq++;
err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx);
zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0,
zfs_putpage_commit_cb, pp);
dmu_tx_commit(tx);
zfs_rangelock_exit(lr);
if (wbc->sync_mode != WB_SYNC_NONE) {
/*
* Note that this is rarely called under writepages(), because
* writepages() normally handles the entire commit for
* performance reasons.
*/
zil_commit(zfsvfs->z_log, zp->z_id);
}
ZFS_EXIT(zfsvfs);
return (err);
}
/*
* Update the system attributes when the inode has been dirtied. For the
* moment we only update the mode, atime, mtime, and ctime.
*/
int
zfs_dirty_inode(struct inode *ip, int flags)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
dmu_tx_t *tx;
uint64_t mode, atime[2], mtime[2], ctime[2];
sa_bulk_attr_t bulk[4];
int error = 0;
int cnt = 0;
if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os))
return (0);
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
#ifdef I_DIRTY_TIME
/*
* This is the lazytime semantic introduced in Linux 4.0
* This flag will only be called from update_time when lazytime is set.
* (Note, I_DIRTY_SYNC will also set if not lazytime)
* Fortunately mtime and ctime are managed within ZFS itself, so we
* only need to dirty atime.
*/
if (flags == I_DIRTY_TIME) {
zp->z_atime_dirty = B_TRUE;
goto out;
}
#endif
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
goto out;
}
mutex_enter(&zp->z_lock);
zp->z_atime_dirty = B_FALSE;
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
/* Preserve the mode, mtime and ctime provided by the inode */
ZFS_TIME_ENCODE(&ip->i_atime, atime);
ZFS_TIME_ENCODE(&ip->i_mtime, mtime);
ZFS_TIME_ENCODE(&ip->i_ctime, ctime);
mode = ip->i_mode;
zp->z_mode = mode;
error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx);
mutex_exit(&zp->z_lock);
dmu_tx_commit(tx);
out:
ZFS_EXIT(zfsvfs);
return (error);
}
/*ARGSUSED*/
void
zfs_inactive(struct inode *ip)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
uint64_t atime[2];
int error;
int need_unlock = 0;
/* Only read lock if we haven't already write locked, e.g. rollback */
if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) {
need_unlock = 1;
rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
}
if (zp->z_sa_hdl == NULL) {
if (need_unlock)
rw_exit(&zfsvfs->z_teardown_inactive_lock);
return;
}
if (zp->z_atime_dirty && zp->z_unlinked == B_FALSE) {
dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
} else {
ZFS_TIME_ENCODE(&ip->i_atime, atime);
mutex_enter(&zp->z_lock);
(void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs),
(void *)&atime, sizeof (atime), tx);
zp->z_atime_dirty = B_FALSE;
mutex_exit(&zp->z_lock);
dmu_tx_commit(tx);
}
}
zfs_zinactive(zp);
if (need_unlock)
rw_exit(&zfsvfs->z_teardown_inactive_lock);
}
/*
* Fill pages with data from the disk.
*/
static int
zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
objset_t *os;
struct page *cur_pp;
u_offset_t io_off, total;
size_t io_len;
loff_t i_size;
unsigned page_idx;
int err;
os = zfsvfs->z_os;
io_len = nr_pages << PAGE_SHIFT;
i_size = i_size_read(ip);
io_off = page_offset(pl[0]);
if (io_off + io_len > i_size)
io_len = i_size - io_off;
/*
* Iterate over list of pages and read each page individually.
*/
page_idx = 0;
for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) {
caddr_t va;
cur_pp = pl[page_idx++];
va = kmap(cur_pp);
err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va,
DMU_READ_PREFETCH);
kunmap(cur_pp);
if (err) {
/* convert checksum errors into IO errors */
if (err == ECKSUM)
err = SET_ERROR(EIO);
return (err);
}
}
return (0);
}
/*
* Uses zfs_fillpage to read data from the file and fill the pages.
*
* IN: ip - inode of file to get data from.
* pl - list of pages to read
* nr_pages - number of pages to read
*
* RETURN: 0 on success, error code on failure.
*
* Timestamps:
* vp - atime updated
*/
/* ARGSUSED */
int
zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
int err;
if (pl == NULL)
return (0);
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
err = zfs_fillpage(ip, pl, nr_pages);
ZFS_EXIT(zfsvfs);
return (err);
}
/*
* Check ZFS specific permissions to memory map a section of a file.
*
* IN: ip - inode of the file to mmap
* off - file offset
* addrp - start address in memory region
* len - length of memory region
* vm_flags- address flags
*
* RETURN: 0 if success
* error code if failure
*/
/*ARGSUSED*/
int
zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len,
unsigned long vm_flags)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
if ((vm_flags & VM_WRITE) && (zp->z_pflags &
(ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EPERM));
}
if ((vm_flags & (VM_READ | VM_EXEC)) &&
(zp->z_pflags & ZFS_AV_QUARANTINED)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EACCES));
}
if (off < 0 || len > MAXOFFSET_T - off) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(ENXIO));
}
ZFS_EXIT(zfsvfs);
return (0);
}
/*
* Free or allocate space in a file. Currently, this function only
* supports the `F_FREESP' command. However, this command is somewhat
* misnamed, as its functionality includes the ability to allocate as
* well as free space.
*
* IN: zp - znode of file to free data in.
* cmd - action to take (only F_FREESP supported).
* bfp - section of file to free/alloc.
* flag - current file open mode flags.
* offset - current file offset.
* cr - credentials of caller.
*
* RETURN: 0 on success, error code on failure.
*
* Timestamps:
* zp - ctime|mtime updated
*/
/* ARGSUSED */
int
zfs_space(znode_t *zp, int cmd, flock64_t *bfp, int flag,
offset_t offset, cred_t *cr)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
uint64_t off, len;
int error;
ZFS_ENTER(zfsvfs);
ZFS_VERIFY_ZP(zp);
if (cmd != F_FREESP) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EINVAL));
}
/*
* Callers might not be able to detect properly that we are read-only,
* so check it explicitly here.
*/
if (zfs_is_readonly(zfsvfs)) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EROFS));
}
if (bfp->l_len < 0) {
ZFS_EXIT(zfsvfs);
return (SET_ERROR(EINVAL));
}
/*
* Permissions aren't checked on Solaris because on this OS
* zfs_space() can only be called with an opened file handle.
* On Linux we can get here through truncate_range() which
* operates directly on inodes, so we need to check access rights.
*/
if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) {
ZFS_EXIT(zfsvfs);
return (error);
}
off = bfp->l_start;
len = bfp->l_len; /* 0 means from off to end of file */
error = zfs_freesp(zp, off, len, flag, TRUE);
ZFS_EXIT(zfsvfs);
return (error);
}
/*ARGSUSED*/
int
zfs_fid(struct inode *ip, fid_t *fidp)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ITOZSB(ip);
uint32_t gen;
uint64_t gen64;
uint64_t object = zp->z_id;
zfid_short_t *zfid;
int size, i, error;
ZFS_ENTER(zfsvfs);
if (fidp->fid_len < SHORT_FID_LEN) {
fidp->fid_len = SHORT_FID_LEN;
ZFS_EXIT(zfsvfs);
return (SET_ERROR(ENOSPC));
}
ZFS_VERIFY_ZP(zp);
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs),
&gen64, sizeof (uint64_t))) != 0) {
ZFS_EXIT(zfsvfs);
return (error);
}
gen = (uint32_t)gen64;
size = SHORT_FID_LEN;
zfid = (zfid_short_t *)fidp;
zfid->zf_len = size;
for (i = 0; i < sizeof (zfid->zf_object); i++)
zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
/* Must have a non-zero generation number to distinguish from .zfs */
if (gen == 0)
gen = 1;
for (i = 0; i < sizeof (zfid->zf_gen); i++)
zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
ZFS_EXIT(zfsvfs);
return (0);
}
#if defined(_KERNEL)
EXPORT_SYMBOL(zfs_open);
EXPORT_SYMBOL(zfs_close);
EXPORT_SYMBOL(zfs_lookup);
EXPORT_SYMBOL(zfs_create);
EXPORT_SYMBOL(zfs_tmpfile);
EXPORT_SYMBOL(zfs_remove);
EXPORT_SYMBOL(zfs_mkdir);
EXPORT_SYMBOL(zfs_rmdir);
EXPORT_SYMBOL(zfs_readdir);
EXPORT_SYMBOL(zfs_getattr_fast);
EXPORT_SYMBOL(zfs_setattr);
EXPORT_SYMBOL(zfs_rename);
EXPORT_SYMBOL(zfs_symlink);
EXPORT_SYMBOL(zfs_readlink);
EXPORT_SYMBOL(zfs_link);
EXPORT_SYMBOL(zfs_inactive);
EXPORT_SYMBOL(zfs_space);
EXPORT_SYMBOL(zfs_fid);
EXPORT_SYMBOL(zfs_getpage);
EXPORT_SYMBOL(zfs_putpage);
EXPORT_SYMBOL(zfs_dirty_inode);
EXPORT_SYMBOL(zfs_map);
/* BEGIN CSTYLED */
module_param(zfs_delete_blocks, ulong, 0644);
MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async");
/* END CSTYLED */
#endif
diff --git a/module/os/linux/zfs/zfs_znode.c b/module/os/linux/zfs/zfs_znode.c
index 859c51baffd8..5b1573a6df1a 100644
--- a/module/os/linux/zfs/zfs_znode.c
+++ b/module/os/linux/zfs/zfs_znode.c
@@ -1,2255 +1,2255 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
*/
/* Portions Copyright 2007 Jeremy Teo */
#ifdef _KERNEL
#include <sys/types.h>
#include <sys/param.h>
#include <sys/time.h>
#include <sys/sysmacros.h>
#include <sys/mntent.h>
#include <sys/u8_textprep.h>
#include <sys/dsl_dataset.h>
#include <sys/vfs.h>
#include <sys/vnode.h>
#include <sys/file.h>
#include <sys/kmem.h>
#include <sys/errno.h>
#include <sys/atomic.h>
#include <sys/zfs_dir.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_ioctl.h>
#include <sys/zfs_rlock.h>
#include <sys/zfs_fuid.h>
#include <sys/zfs_vnops.h>
#include <sys/zfs_ctldir.h>
#include <sys/dnode.h>
#include <sys/fs/zfs.h>
#include <sys/zpl.h>
#endif /* _KERNEL */
#include <sys/dmu.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_tx.h>
#include <sys/zfs_refcount.h>
#include <sys/stat.h>
#include <sys/zap.h>
#include <sys/zfs_znode.h>
#include <sys/sa.h>
#include <sys/zfs_sa.h>
#include <sys/zfs_stat.h>
#include "zfs_prop.h"
#include "zfs_comutil.h"
/*
* Functions needed for userland (ie: libzpool) are not put under
* #ifdef_KERNEL; the rest of the functions have dependencies
* (such as VFS logic) that will not compile easily in userland.
*/
#ifdef _KERNEL
static kmem_cache_t *znode_cache = NULL;
static kmem_cache_t *znode_hold_cache = NULL;
unsigned int zfs_object_mutex_size = ZFS_OBJ_MTX_SZ;
/*
* This is used by the test suite so that it can delay znodes from being
* freed in order to inspect the unlinked set.
*/
-int zfs_unlink_suspend_progress = 0;
+static int zfs_unlink_suspend_progress = 0;
/*
* This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on
* z_rangelock. It will modify the offset and length of the lock to reflect
* znode-specific information, and convert RL_APPEND to RL_WRITER. This is
* called with the rangelock_t's rl_lock held, which avoids races.
*/
static void
zfs_rangelock_cb(zfs_locked_range_t *new, void *arg)
{
znode_t *zp = arg;
/*
* If in append mode, convert to writer and lock starting at the
* current end of file.
*/
if (new->lr_type == RL_APPEND) {
new->lr_offset = zp->z_size;
new->lr_type = RL_WRITER;
}
/*
* If we need to grow the block size then lock the whole file range.
*/
uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length);
if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) ||
zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) {
new->lr_offset = 0;
new->lr_length = UINT64_MAX;
}
}
/*ARGSUSED*/
static int
zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
{
znode_t *zp = buf;
inode_init_once(ZTOI(zp));
list_link_init(&zp->z_link_node);
mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
rw_init(&zp->z_name_lock, NULL, RW_NOLOCKDEP, NULL);
mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL);
zfs_rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp);
zp->z_dirlocks = NULL;
zp->z_acl_cached = NULL;
zp->z_xattr_cached = NULL;
zp->z_xattr_parent = 0;
return (0);
}
/*ARGSUSED*/
static void
zfs_znode_cache_destructor(void *buf, void *arg)
{
znode_t *zp = buf;
ASSERT(!list_link_active(&zp->z_link_node));
mutex_destroy(&zp->z_lock);
rw_destroy(&zp->z_parent_lock);
rw_destroy(&zp->z_name_lock);
mutex_destroy(&zp->z_acl_lock);
rw_destroy(&zp->z_xattr_lock);
zfs_rangelock_fini(&zp->z_rangelock);
ASSERT3P(zp->z_dirlocks, ==, NULL);
ASSERT3P(zp->z_acl_cached, ==, NULL);
ASSERT3P(zp->z_xattr_cached, ==, NULL);
}
static int
zfs_znode_hold_cache_constructor(void *buf, void *arg, int kmflags)
{
znode_hold_t *zh = buf;
mutex_init(&zh->zh_lock, NULL, MUTEX_DEFAULT, NULL);
zfs_refcount_create(&zh->zh_refcount);
zh->zh_obj = ZFS_NO_OBJECT;
return (0);
}
static void
zfs_znode_hold_cache_destructor(void *buf, void *arg)
{
znode_hold_t *zh = buf;
mutex_destroy(&zh->zh_lock);
zfs_refcount_destroy(&zh->zh_refcount);
}
void
zfs_znode_init(void)
{
/*
* Initialize zcache. The KMC_SLAB hint is used in order that it be
* backed by kmalloc() when on the Linux slab in order that any
* wait_on_bit() operations on the related inode operate properly.
*/
ASSERT(znode_cache == NULL);
znode_cache = kmem_cache_create("zfs_znode_cache",
sizeof (znode_t), 0, zfs_znode_cache_constructor,
zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_SLAB);
ASSERT(znode_hold_cache == NULL);
znode_hold_cache = kmem_cache_create("zfs_znode_hold_cache",
sizeof (znode_hold_t), 0, zfs_znode_hold_cache_constructor,
zfs_znode_hold_cache_destructor, NULL, NULL, NULL, 0);
}
void
zfs_znode_fini(void)
{
/*
* Cleanup zcache
*/
if (znode_cache)
kmem_cache_destroy(znode_cache);
znode_cache = NULL;
if (znode_hold_cache)
kmem_cache_destroy(znode_hold_cache);
znode_hold_cache = NULL;
}
/*
* The zfs_znode_hold_enter() / zfs_znode_hold_exit() functions are used to
* serialize access to a znode and its SA buffer while the object is being
* created or destroyed. This kind of locking would normally reside in the
* znode itself but in this case that's impossible because the znode and SA
* buffer may not yet exist. Therefore the locking is handled externally
* with an array of mutexes and AVLs trees which contain per-object locks.
*
* In zfs_znode_hold_enter() a per-object lock is created as needed, inserted
* in to the correct AVL tree and finally the per-object lock is held. In
* zfs_znode_hold_exit() the process is reversed. The per-object lock is
* released, removed from the AVL tree and destroyed if there are no waiters.
*
* This scheme has two important properties:
*
* 1) No memory allocations are performed while holding one of the z_hold_locks.
* This ensures evict(), which can be called from direct memory reclaim, will
* never block waiting on a z_hold_locks which just happens to have hashed
* to the same index.
*
* 2) All locks used to serialize access to an object are per-object and never
* shared. This minimizes lock contention without creating a large number
* of dedicated locks.
*
* On the downside it does require znode_lock_t structures to be frequently
* allocated and freed. However, because these are backed by a kmem cache
* and very short lived this cost is minimal.
*/
int
zfs_znode_hold_compare(const void *a, const void *b)
{
const znode_hold_t *zh_a = (const znode_hold_t *)a;
const znode_hold_t *zh_b = (const znode_hold_t *)b;
return (TREE_CMP(zh_a->zh_obj, zh_b->zh_obj));
}
static boolean_t __maybe_unused
zfs_znode_held(zfsvfs_t *zfsvfs, uint64_t obj)
{
znode_hold_t *zh, search;
int i = ZFS_OBJ_HASH(zfsvfs, obj);
boolean_t held;
search.zh_obj = obj;
mutex_enter(&zfsvfs->z_hold_locks[i]);
zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL);
held = (zh && MUTEX_HELD(&zh->zh_lock)) ? B_TRUE : B_FALSE;
mutex_exit(&zfsvfs->z_hold_locks[i]);
return (held);
}
static znode_hold_t *
zfs_znode_hold_enter(zfsvfs_t *zfsvfs, uint64_t obj)
{
znode_hold_t *zh, *zh_new, search;
int i = ZFS_OBJ_HASH(zfsvfs, obj);
boolean_t found = B_FALSE;
zh_new = kmem_cache_alloc(znode_hold_cache, KM_SLEEP);
zh_new->zh_obj = obj;
search.zh_obj = obj;
mutex_enter(&zfsvfs->z_hold_locks[i]);
zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL);
if (likely(zh == NULL)) {
zh = zh_new;
avl_add(&zfsvfs->z_hold_trees[i], zh);
} else {
ASSERT3U(zh->zh_obj, ==, obj);
found = B_TRUE;
}
zfs_refcount_add(&zh->zh_refcount, NULL);
mutex_exit(&zfsvfs->z_hold_locks[i]);
if (found == B_TRUE)
kmem_cache_free(znode_hold_cache, zh_new);
ASSERT(MUTEX_NOT_HELD(&zh->zh_lock));
ASSERT3S(zfs_refcount_count(&zh->zh_refcount), >, 0);
mutex_enter(&zh->zh_lock);
return (zh);
}
static void
zfs_znode_hold_exit(zfsvfs_t *zfsvfs, znode_hold_t *zh)
{
int i = ZFS_OBJ_HASH(zfsvfs, zh->zh_obj);
boolean_t remove = B_FALSE;
ASSERT(zfs_znode_held(zfsvfs, zh->zh_obj));
ASSERT3S(zfs_refcount_count(&zh->zh_refcount), >, 0);
mutex_exit(&zh->zh_lock);
mutex_enter(&zfsvfs->z_hold_locks[i]);
if (zfs_refcount_remove(&zh->zh_refcount, NULL) == 0) {
avl_remove(&zfsvfs->z_hold_trees[i], zh);
remove = B_TRUE;
}
mutex_exit(&zfsvfs->z_hold_locks[i]);
if (remove == B_TRUE)
kmem_cache_free(znode_hold_cache, zh);
}
dev_t
zfs_cmpldev(uint64_t dev)
{
return (dev);
}
static void
zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
{
ASSERT(zfs_znode_held(zfsvfs, zp->z_id));
mutex_enter(&zp->z_lock);
ASSERT(zp->z_sa_hdl == NULL);
ASSERT(zp->z_acl_cached == NULL);
if (sa_hdl == NULL) {
VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
SA_HDL_SHARED, &zp->z_sa_hdl));
} else {
zp->z_sa_hdl = sa_hdl;
sa_set_userp(sa_hdl, zp);
}
zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
mutex_exit(&zp->z_lock);
}
void
zfs_znode_dmu_fini(znode_t *zp)
{
ASSERT(zfs_znode_held(ZTOZSB(zp), zp->z_id) || zp->z_unlinked ||
RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock));
sa_handle_destroy(zp->z_sa_hdl);
zp->z_sa_hdl = NULL;
}
/*
* Called by new_inode() to allocate a new inode.
*/
int
zfs_inode_alloc(struct super_block *sb, struct inode **ip)
{
znode_t *zp;
zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
*ip = ZTOI(zp);
return (0);
}
/*
* Called in multiple places when an inode should be destroyed.
*/
void
zfs_inode_destroy(struct inode *ip)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ZTOZSB(zp);
mutex_enter(&zfsvfs->z_znodes_lock);
if (list_link_active(&zp->z_link_node)) {
list_remove(&zfsvfs->z_all_znodes, zp);
zfsvfs->z_nr_znodes--;
}
mutex_exit(&zfsvfs->z_znodes_lock);
if (zp->z_acl_cached) {
zfs_acl_free(zp->z_acl_cached);
zp->z_acl_cached = NULL;
}
if (zp->z_xattr_cached) {
nvlist_free(zp->z_xattr_cached);
zp->z_xattr_cached = NULL;
}
kmem_cache_free(znode_cache, zp);
}
static void
zfs_inode_set_ops(zfsvfs_t *zfsvfs, struct inode *ip)
{
uint64_t rdev = 0;
switch (ip->i_mode & S_IFMT) {
case S_IFREG:
ip->i_op = &zpl_inode_operations;
ip->i_fop = &zpl_file_operations;
ip->i_mapping->a_ops = &zpl_address_space_operations;
break;
case S_IFDIR:
ip->i_op = &zpl_dir_inode_operations;
ip->i_fop = &zpl_dir_file_operations;
ITOZ(ip)->z_zn_prefetch = B_TRUE;
break;
case S_IFLNK:
ip->i_op = &zpl_symlink_inode_operations;
break;
/*
* rdev is only stored in a SA only for device files.
*/
case S_IFCHR:
case S_IFBLK:
(void) sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zfsvfs), &rdev,
sizeof (rdev));
fallthrough;
case S_IFIFO:
case S_IFSOCK:
init_special_inode(ip, ip->i_mode, rdev);
ip->i_op = &zpl_special_inode_operations;
break;
default:
zfs_panic_recover("inode %llu has invalid mode: 0x%x\n",
(u_longlong_t)ip->i_ino, ip->i_mode);
/* Assume the inode is a file and attempt to continue */
ip->i_mode = S_IFREG | 0644;
ip->i_op = &zpl_inode_operations;
ip->i_fop = &zpl_file_operations;
ip->i_mapping->a_ops = &zpl_address_space_operations;
break;
}
}
static void
zfs_set_inode_flags(znode_t *zp, struct inode *ip)
{
/*
* Linux and Solaris have different sets of file attributes, so we
* restrict this conversion to the intersection of the two.
*/
#ifdef HAVE_INODE_SET_FLAGS
unsigned int flags = 0;
if (zp->z_pflags & ZFS_IMMUTABLE)
flags |= S_IMMUTABLE;
if (zp->z_pflags & ZFS_APPENDONLY)
flags |= S_APPEND;
inode_set_flags(ip, flags, S_IMMUTABLE|S_APPEND);
#else
if (zp->z_pflags & ZFS_IMMUTABLE)
ip->i_flags |= S_IMMUTABLE;
else
ip->i_flags &= ~S_IMMUTABLE;
if (zp->z_pflags & ZFS_APPENDONLY)
ip->i_flags |= S_APPEND;
else
ip->i_flags &= ~S_APPEND;
#endif
}
/*
* Update the embedded inode given the znode.
*/
void
zfs_znode_update_vfs(znode_t *zp)
{
zfsvfs_t *zfsvfs;
struct inode *ip;
uint32_t blksize;
u_longlong_t i_blocks;
ASSERT(zp != NULL);
zfsvfs = ZTOZSB(zp);
ip = ZTOI(zp);
/* Skip .zfs control nodes which do not exist on disk. */
if (zfsctl_is_node(ip))
return;
dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, &i_blocks);
spin_lock(&ip->i_lock);
ip->i_mode = zp->z_mode;
ip->i_blocks = i_blocks;
i_size_write(ip, zp->z_size);
spin_unlock(&ip->i_lock);
}
/*
* Construct a znode+inode and initialize.
*
* This does not do a call to dmu_set_user() that is
* up to the caller to do, in case you don't want to
* return the znode
*/
static znode_t *
zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
dmu_object_type_t obj_type, sa_handle_t *hdl)
{
znode_t *zp;
struct inode *ip;
uint64_t mode;
uint64_t parent;
uint64_t tmp_gen;
uint64_t links;
uint64_t z_uid, z_gid;
uint64_t atime[2], mtime[2], ctime[2], btime[2];
uint64_t projid = ZFS_DEFAULT_PROJID;
sa_bulk_attr_t bulk[12];
int count = 0;
ASSERT(zfsvfs != NULL);
ip = new_inode(zfsvfs->z_sb);
if (ip == NULL)
return (NULL);
zp = ITOZ(ip);
ASSERT(zp->z_dirlocks == NULL);
ASSERT3P(zp->z_acl_cached, ==, NULL);
ASSERT3P(zp->z_xattr_cached, ==, NULL);
zp->z_unlinked = B_FALSE;
zp->z_atime_dirty = B_FALSE;
zp->z_is_mapped = B_FALSE;
zp->z_is_ctldir = B_FALSE;
zp->z_is_stale = B_FALSE;
zp->z_suspended = B_FALSE;
zp->z_sa_hdl = NULL;
zp->z_mapcnt = 0;
zp->z_id = db->db_object;
zp->z_blksz = blksz;
zp->z_seq = 0x7A4653;
zp->z_sync_cnt = 0;
zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
&zp->z_size, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
&parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &z_uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &z_gid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &btime, 16);
if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || tmp_gen == 0 ||
(dmu_objset_projectquota_enabled(zfsvfs->z_os) &&
(zp->z_pflags & ZFS_PROJID) &&
sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) {
if (hdl == NULL)
sa_handle_destroy(zp->z_sa_hdl);
zp->z_sa_hdl = NULL;
goto error;
}
zp->z_projid = projid;
zp->z_mode = ip->i_mode = mode;
ip->i_generation = (uint32_t)tmp_gen;
ip->i_blkbits = SPA_MINBLOCKSHIFT;
set_nlink(ip, (uint32_t)links);
zfs_uid_write(ip, z_uid);
zfs_gid_write(ip, z_gid);
zfs_set_inode_flags(zp, ip);
/* Cache the xattr parent id */
if (zp->z_pflags & ZFS_XATTR)
zp->z_xattr_parent = parent;
ZFS_TIME_DECODE(&ip->i_atime, atime);
ZFS_TIME_DECODE(&ip->i_mtime, mtime);
ZFS_TIME_DECODE(&ip->i_ctime, ctime);
ZFS_TIME_DECODE(&zp->z_btime, btime);
ip->i_ino = zp->z_id;
zfs_znode_update_vfs(zp);
zfs_inode_set_ops(zfsvfs, ip);
/*
* The only way insert_inode_locked() can fail is if the ip->i_ino
* number is already hashed for this super block. This can never
* happen because the inode numbers map 1:1 with the object numbers.
*
* Exceptions include rolling back a mounted file system, either
* from the zfs rollback or zfs recv command.
*
* Active inodes are unhashed during the rollback, but since zrele
* can happen asynchronously, we can't guarantee they've been
* unhashed. This can cause hash collisions in unlinked drain
* processing so do not hash unlinked znodes.
*/
if (links > 0)
VERIFY3S(insert_inode_locked(ip), ==, 0);
mutex_enter(&zfsvfs->z_znodes_lock);
list_insert_tail(&zfsvfs->z_all_znodes, zp);
zfsvfs->z_nr_znodes++;
mutex_exit(&zfsvfs->z_znodes_lock);
if (links > 0)
unlock_new_inode(ip);
return (zp);
error:
iput(ip);
return (NULL);
}
/*
* Safely mark an inode dirty. Inodes which are part of a read-only
* file system or snapshot may not be dirtied.
*/
void
zfs_mark_inode_dirty(struct inode *ip)
{
zfsvfs_t *zfsvfs = ITOZSB(ip);
if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os))
return;
mark_inode_dirty(ip);
}
static uint64_t empty_xattr;
static uint64_t pad[4];
static zfs_acl_phys_t acl_phys;
/*
* Create a new DMU object to hold a zfs znode.
*
* IN: dzp - parent directory for new znode
* vap - file attributes for new znode
* tx - dmu transaction id for zap operations
* cr - credentials of caller
* flag - flags:
* IS_ROOT_NODE - new object will be root
* IS_TMPFILE - new object is of O_TMPFILE
* IS_XATTR - new object is an attribute
* acl_ids - ACL related attributes
*
* OUT: zpp - allocated znode (set to dzp if IS_ROOT_NODE)
*
*/
void
zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
{
uint64_t crtime[2], atime[2], mtime[2], ctime[2];
uint64_t mode, size, links, parent, pflags;
uint64_t projid = ZFS_DEFAULT_PROJID;
uint64_t rdev = 0;
zfsvfs_t *zfsvfs = ZTOZSB(dzp);
dmu_buf_t *db;
inode_timespec_t now;
uint64_t gen, obj;
int bonuslen;
int dnodesize;
sa_handle_t *sa_hdl;
dmu_object_type_t obj_type;
sa_bulk_attr_t *sa_attrs;
int cnt = 0;
zfs_acl_locator_cb_t locate = { 0 };
znode_hold_t *zh;
if (zfsvfs->z_replay) {
obj = vap->va_nodeid;
now = vap->va_ctime; /* see zfs_replay_create() */
gen = vap->va_nblocks; /* ditto */
dnodesize = vap->va_fsid; /* ditto */
} else {
obj = 0;
gethrestime(&now);
gen = dmu_tx_get_txg(tx);
dnodesize = dmu_objset_dnodesize(zfsvfs->z_os);
}
if (dnodesize == 0)
dnodesize = DNODE_MIN_SIZE;
obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
bonuslen = (obj_type == DMU_OT_SA) ?
DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE;
/*
* Create a new DMU object.
*/
/*
* There's currently no mechanism for pre-reading the blocks that will
* be needed to allocate a new object, so we accept the small chance
* that there will be an i/o error and we will fail one of the
* assertions below.
*/
if (S_ISDIR(vap->va_mode)) {
if (zfsvfs->z_replay) {
VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj,
zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
obj_type, bonuslen, dnodesize, tx));
} else {
obj = zap_create_norm_dnsize(zfsvfs->z_os,
zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
obj_type, bonuslen, dnodesize, tx);
}
} else {
if (zfsvfs->z_replay) {
VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj,
DMU_OT_PLAIN_FILE_CONTENTS, 0,
obj_type, bonuslen, dnodesize, tx));
} else {
obj = dmu_object_alloc_dnsize(zfsvfs->z_os,
DMU_OT_PLAIN_FILE_CONTENTS, 0,
obj_type, bonuslen, dnodesize, tx);
}
}
zh = zfs_znode_hold_enter(zfsvfs, obj);
VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
/*
* If this is the root, fix up the half-initialized parent pointer
* to reference the just-allocated physical data area.
*/
if (flag & IS_ROOT_NODE) {
dzp->z_id = obj;
}
/*
* If parent is an xattr, so am I.
*/
if (dzp->z_pflags & ZFS_XATTR) {
flag |= IS_XATTR;
}
if (zfsvfs->z_use_fuids)
pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
else
pflags = 0;
if (S_ISDIR(vap->va_mode)) {
size = 2; /* contents ("." and "..") */
links = 2;
} else {
size = 0;
links = (flag & IS_TMPFILE) ? 0 : 1;
}
if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))
rdev = vap->va_rdev;
parent = dzp->z_id;
mode = acl_ids->z_mode;
if (flag & IS_XATTR)
pflags |= ZFS_XATTR;
if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) {
/*
* With ZFS_PROJID flag, we can easily know whether there is
* project ID stored on disk or not. See zfs_space_delta_cb().
*/
if (obj_type != DMU_OT_ZNODE &&
dmu_objset_projectquota_enabled(zfsvfs->z_os))
pflags |= ZFS_PROJID;
/*
* Inherit project ID from parent if required.
*/
projid = zfs_inherit_projid(dzp);
if (dzp->z_pflags & ZFS_PROJINHERIT)
pflags |= ZFS_PROJINHERIT;
}
/*
* No execs denied will be determined when zfs_mode_compute() is called.
*/
pflags |= acl_ids->z_aclp->z_hints &
(ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
ZFS_TIME_ENCODE(&now, crtime);
ZFS_TIME_ENCODE(&now, ctime);
if (vap->va_mask & ATTR_ATIME) {
ZFS_TIME_ENCODE(&vap->va_atime, atime);
} else {
ZFS_TIME_ENCODE(&now, atime);
}
if (vap->va_mask & ATTR_MTIME) {
ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
} else {
ZFS_TIME_ENCODE(&now, mtime);
}
/* Now add in all of the "SA" attributes */
VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
&sa_hdl));
/*
* Setup the array of attributes to be replaced/set on the new file
*
* order for DMU_OT_ZNODE is critical since it needs to be constructed
* in the old znode_phys_t format. Don't change this ordering
*/
sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
if (obj_type == DMU_OT_ZNODE) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
NULL, &atime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
NULL, &mtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
NULL, &ctime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
NULL, &crtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
NULL, &gen, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
NULL, &mode, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
NULL, &size, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
NULL, &parent, 8);
} else {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
NULL, &mode, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
NULL, &size, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
NULL, &gen, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs),
NULL, &acl_ids->z_fuid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs),
NULL, &acl_ids->z_fgid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
NULL, &parent, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
NULL, &pflags, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
NULL, &atime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
NULL, &mtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
NULL, &ctime, 16);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
NULL, &crtime, 16);
}
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
if (obj_type == DMU_OT_ZNODE) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
&empty_xattr, 8);
} else if (dmu_objset_projectquota_enabled(zfsvfs->z_os) &&
pflags & ZFS_PROJID) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PROJID(zfsvfs),
NULL, &projid, 8);
}
if (obj_type == DMU_OT_ZNODE ||
(S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
NULL, &rdev, 8);
}
if (obj_type == DMU_OT_ZNODE) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
NULL, &pflags, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
&acl_ids->z_fuid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
&acl_ids->z_fgid, 8);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
sizeof (uint64_t) * 4);
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
&acl_phys, sizeof (zfs_acl_phys_t));
} else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
&acl_ids->z_aclp->z_acl_count, 8);
locate.cb_aclp = acl_ids->z_aclp;
SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
zfs_acl_data_locator, &locate,
acl_ids->z_aclp->z_acl_bytes);
mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
acl_ids->z_fuid, acl_ids->z_fgid);
}
VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
if (!(flag & IS_ROOT_NODE)) {
/*
* The call to zfs_znode_alloc() may fail if memory is low
* via the call path: alloc_inode() -> inode_init_always() ->
* security_inode_alloc() -> inode_alloc_security(). Since
* the existing code is written such that zfs_mknode() can
* not fail retry until sufficient memory has been reclaimed.
*/
do {
*zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
} while (*zpp == NULL);
VERIFY(*zpp != NULL);
VERIFY(dzp != NULL);
} else {
/*
* If we are creating the root node, the "parent" we
* passed in is the znode for the root.
*/
*zpp = dzp;
(*zpp)->z_sa_hdl = sa_hdl;
}
(*zpp)->z_pflags = pflags;
(*zpp)->z_mode = ZTOI(*zpp)->i_mode = mode;
(*zpp)->z_dnodesize = dnodesize;
(*zpp)->z_projid = projid;
if (obj_type == DMU_OT_ZNODE ||
acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
}
kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
zfs_znode_hold_exit(zfsvfs, zh);
}
/*
* Update in-core attributes. It is assumed the caller will be doing an
* sa_bulk_update to push the changes out.
*/
void
zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
{
xoptattr_t *xoap;
boolean_t update_inode = B_FALSE;
xoap = xva_getxoptattr(xvap);
ASSERT(xoap);
if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
uint64_t times[2];
ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
(void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
&times, sizeof (times), tx);
XVA_SET_RTN(xvap, XAT_CREATETIME);
}
if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_READONLY);
}
if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_HIDDEN);
}
if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_SYSTEM);
}
if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_ARCHIVE);
}
if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_IMMUTABLE);
update_inode = B_TRUE;
}
if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_NOUNLINK);
}
if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_APPENDONLY);
update_inode = B_TRUE;
}
if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_NODUMP);
}
if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_OPAQUE);
}
if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
xoap->xoa_av_quarantined, zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
}
if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
}
if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
zfs_sa_set_scanstamp(zp, xvap, tx);
XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
}
if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_REPARSE);
}
if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_OFFLINE);
}
if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_SPARSE);
}
if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) {
ZFS_ATTR_SET(zp, ZFS_PROJINHERIT, xoap->xoa_projinherit,
zp->z_pflags, tx);
XVA_SET_RTN(xvap, XAT_PROJINHERIT);
}
if (update_inode)
zfs_set_inode_flags(zp, ZTOI(zp));
}
int
zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
{
dmu_object_info_t doi;
dmu_buf_t *db;
znode_t *zp;
znode_hold_t *zh;
int err;
sa_handle_t *hdl;
*zpp = NULL;
again:
zh = zfs_znode_hold_enter(zfsvfs, obj_num);
err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
if (err) {
zfs_znode_hold_exit(zfsvfs, zh);
return (err);
}
dmu_object_info_from_db(db, &doi);
if (doi.doi_bonus_type != DMU_OT_SA &&
(doi.doi_bonus_type != DMU_OT_ZNODE ||
(doi.doi_bonus_type == DMU_OT_ZNODE &&
doi.doi_bonus_size < sizeof (znode_phys_t)))) {
sa_buf_rele(db, NULL);
zfs_znode_hold_exit(zfsvfs, zh);
return (SET_ERROR(EINVAL));
}
hdl = dmu_buf_get_user(db);
if (hdl != NULL) {
zp = sa_get_userdata(hdl);
/*
* Since "SA" does immediate eviction we
* should never find a sa handle that doesn't
* know about the znode.
*/
ASSERT3P(zp, !=, NULL);
mutex_enter(&zp->z_lock);
ASSERT3U(zp->z_id, ==, obj_num);
/*
* If zp->z_unlinked is set, the znode is already marked
* for deletion and should not be discovered. Check this
* after checking igrab() due to fsetxattr() & O_TMPFILE.
*
* If igrab() returns NULL the VFS has independently
* determined the inode should be evicted and has
* called iput_final() to start the eviction process.
* The SA handle is still valid but because the VFS
* requires that the eviction succeed we must drop
* our locks and references to allow the eviction to
* complete. The zfs_zget() may then be retried.
*
* This unlikely case could be optimized by registering
* a sops->drop_inode() callback. The callback would
* need to detect the active SA hold thereby informing
* the VFS that this inode should not be evicted.
*/
if (igrab(ZTOI(zp)) == NULL) {
if (zp->z_unlinked)
err = SET_ERROR(ENOENT);
else
err = SET_ERROR(EAGAIN);
} else {
*zpp = zp;
err = 0;
}
mutex_exit(&zp->z_lock);
sa_buf_rele(db, NULL);
zfs_znode_hold_exit(zfsvfs, zh);
if (err == EAGAIN) {
/* inode might need this to finish evict */
cond_resched();
goto again;
}
return (err);
}
/*
* Not found create new znode/vnode but only if file exists.
*
* There is a small window where zfs_vget() could
* find this object while a file create is still in
* progress. This is checked for in zfs_znode_alloc()
*
* if zfs_znode_alloc() fails it will drop the hold on the
* bonus buffer.
*/
zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
doi.doi_bonus_type, NULL);
if (zp == NULL) {
err = SET_ERROR(ENOENT);
} else {
*zpp = zp;
}
zfs_znode_hold_exit(zfsvfs, zh);
return (err);
}
int
zfs_rezget(znode_t *zp)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
dmu_object_info_t doi;
dmu_buf_t *db;
uint64_t obj_num = zp->z_id;
uint64_t mode;
uint64_t links;
sa_bulk_attr_t bulk[11];
int err;
int count = 0;
uint64_t gen;
uint64_t z_uid, z_gid;
uint64_t atime[2], mtime[2], ctime[2], btime[2];
uint64_t projid = ZFS_DEFAULT_PROJID;
znode_hold_t *zh;
/*
* skip ctldir, otherwise they will always get invalidated. This will
* cause funny behaviour for the mounted snapdirs. Especially for
* Linux >= 3.18, d_invalidate will detach the mountpoint and prevent
* anyone automount it again as long as someone is still using the
* detached mount.
*/
if (zp->z_is_ctldir)
return (0);
zh = zfs_znode_hold_enter(zfsvfs, obj_num);
mutex_enter(&zp->z_acl_lock);
if (zp->z_acl_cached) {
zfs_acl_free(zp->z_acl_cached);
zp->z_acl_cached = NULL;
}
mutex_exit(&zp->z_acl_lock);
rw_enter(&zp->z_xattr_lock, RW_WRITER);
if (zp->z_xattr_cached) {
nvlist_free(zp->z_xattr_cached);
zp->z_xattr_cached = NULL;
}
rw_exit(&zp->z_xattr_lock);
ASSERT(zp->z_sa_hdl == NULL);
err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
if (err) {
zfs_znode_hold_exit(zfsvfs, zh);
return (err);
}
dmu_object_info_from_db(db, &doi);
if (doi.doi_bonus_type != DMU_OT_SA &&
(doi.doi_bonus_type != DMU_OT_ZNODE ||
(doi.doi_bonus_type == DMU_OT_ZNODE &&
doi.doi_bonus_size < sizeof (znode_phys_t)))) {
sa_buf_rele(db, NULL);
zfs_znode_hold_exit(zfsvfs, zh);
return (SET_ERROR(EINVAL));
}
zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
/* reload cached values */
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
&gen, sizeof (gen));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
&zp->z_size, sizeof (zp->z_size));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
&links, sizeof (links));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, sizeof (zp->z_pflags));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
&z_uid, sizeof (z_uid));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
&z_gid, sizeof (z_gid));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
&mode, sizeof (mode));
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
&atime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
&mtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
&ctime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &btime, 16);
if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
zfs_znode_dmu_fini(zp);
zfs_znode_hold_exit(zfsvfs, zh);
return (SET_ERROR(EIO));
}
if (dmu_objset_projectquota_enabled(zfsvfs->z_os)) {
err = sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs),
&projid, 8);
if (err != 0 && err != ENOENT) {
zfs_znode_dmu_fini(zp);
zfs_znode_hold_exit(zfsvfs, zh);
return (SET_ERROR(err));
}
}
zp->z_projid = projid;
zp->z_mode = ZTOI(zp)->i_mode = mode;
zfs_uid_write(ZTOI(zp), z_uid);
zfs_gid_write(ZTOI(zp), z_gid);
ZFS_TIME_DECODE(&ZTOI(zp)->i_atime, atime);
ZFS_TIME_DECODE(&ZTOI(zp)->i_mtime, mtime);
ZFS_TIME_DECODE(&ZTOI(zp)->i_ctime, ctime);
ZFS_TIME_DECODE(&zp->z_btime, btime);
if ((uint32_t)gen != ZTOI(zp)->i_generation) {
zfs_znode_dmu_fini(zp);
zfs_znode_hold_exit(zfsvfs, zh);
return (SET_ERROR(EIO));
}
set_nlink(ZTOI(zp), (uint32_t)links);
zfs_set_inode_flags(zp, ZTOI(zp));
zp->z_blksz = doi.doi_data_block_size;
zp->z_atime_dirty = B_FALSE;
zfs_znode_update_vfs(zp);
/*
* If the file has zero links, then it has been unlinked on the send
* side and it must be in the received unlinked set.
* We call zfs_znode_dmu_fini() now to prevent any accesses to the
* stale data and to prevent automatic removal of the file in
* zfs_zinactive(). The file will be removed either when it is removed
* on the send side and the next incremental stream is received or
* when the unlinked set gets processed.
*/
zp->z_unlinked = (ZTOI(zp)->i_nlink == 0);
if (zp->z_unlinked)
zfs_znode_dmu_fini(zp);
zfs_znode_hold_exit(zfsvfs, zh);
return (0);
}
void
zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
objset_t *os = zfsvfs->z_os;
uint64_t obj = zp->z_id;
uint64_t acl_obj = zfs_external_acl(zp);
znode_hold_t *zh;
zh = zfs_znode_hold_enter(zfsvfs, obj);
if (acl_obj) {
VERIFY(!zp->z_is_sa);
VERIFY(0 == dmu_object_free(os, acl_obj, tx));
}
VERIFY(0 == dmu_object_free(os, obj, tx));
zfs_znode_dmu_fini(zp);
zfs_znode_hold_exit(zfsvfs, zh);
}
void
zfs_zinactive(znode_t *zp)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
uint64_t z_id = zp->z_id;
znode_hold_t *zh;
ASSERT(zp->z_sa_hdl);
/*
* Don't allow a zfs_zget() while were trying to release this znode.
*/
zh = zfs_znode_hold_enter(zfsvfs, z_id);
mutex_enter(&zp->z_lock);
/*
* If this was the last reference to a file with no links, remove
* the file from the file system unless the file system is mounted
* read-only. That can happen, for example, if the file system was
* originally read-write, the file was opened, then unlinked and
* the file system was made read-only before the file was finally
* closed. The file will remain in the unlinked set.
*/
if (zp->z_unlinked) {
ASSERT(!zfsvfs->z_issnap);
if (!zfs_is_readonly(zfsvfs) && !zfs_unlink_suspend_progress) {
mutex_exit(&zp->z_lock);
zfs_znode_hold_exit(zfsvfs, zh);
zfs_rmnode(zp);
return;
}
}
mutex_exit(&zp->z_lock);
zfs_znode_dmu_fini(zp);
zfs_znode_hold_exit(zfsvfs, zh);
}
#if defined(HAVE_INODE_TIMESPEC64_TIMES)
#define zfs_compare_timespec timespec64_compare
#else
#define zfs_compare_timespec timespec_compare
#endif
/*
* Determine whether the znode's atime must be updated. The logic mostly
* duplicates the Linux kernel's relatime_need_update() functionality.
* This function is only called if the underlying filesystem actually has
* atime updates enabled.
*/
boolean_t
zfs_relatime_need_update(const struct inode *ip)
{
inode_timespec_t now;
gethrestime(&now);
/*
* In relatime mode, only update the atime if the previous atime
* is earlier than either the ctime or mtime or if at least a day
* has passed since the last update of atime.
*/
if (zfs_compare_timespec(&ip->i_mtime, &ip->i_atime) >= 0)
return (B_TRUE);
if (zfs_compare_timespec(&ip->i_ctime, &ip->i_atime) >= 0)
return (B_TRUE);
if ((hrtime_t)now.tv_sec - (hrtime_t)ip->i_atime.tv_sec >= 24*60*60)
return (B_TRUE);
return (B_FALSE);
}
/*
* Prepare to update znode time stamps.
*
* IN: zp - znode requiring timestamp update
* flag - ATTR_MTIME, ATTR_CTIME flags
*
* OUT: zp - z_seq
* mtime - new mtime
* ctime - new ctime
*
* Note: We don't update atime here, because we rely on Linux VFS to do
* atime updating.
*/
void
zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
uint64_t ctime[2])
{
inode_timespec_t now;
gethrestime(&now);
zp->z_seq++;
if (flag & ATTR_MTIME) {
ZFS_TIME_ENCODE(&now, mtime);
ZFS_TIME_DECODE(&(ZTOI(zp)->i_mtime), mtime);
if (ZTOZSB(zp)->z_use_fuids) {
zp->z_pflags |= (ZFS_ARCHIVE |
ZFS_AV_MODIFIED);
}
}
if (flag & ATTR_CTIME) {
ZFS_TIME_ENCODE(&now, ctime);
ZFS_TIME_DECODE(&(ZTOI(zp)->i_ctime), ctime);
if (ZTOZSB(zp)->z_use_fuids)
zp->z_pflags |= ZFS_ARCHIVE;
}
}
/*
* Grow the block size for a file.
*
* IN: zp - znode of file to free data in.
* size - requested block size
* tx - open transaction.
*
* NOTE: this function assumes that the znode is write locked.
*/
void
zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
{
int error;
u_longlong_t dummy;
if (size <= zp->z_blksz)
return;
/*
* If the file size is already greater than the current blocksize,
* we will not grow. If there is more than one block in a file,
* the blocksize cannot change.
*/
if (zp->z_blksz && zp->z_size > zp->z_blksz)
return;
error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id,
size, 0, tx);
if (error == ENOTSUP)
return;
ASSERT0(error);
/* What blocksize did we actually get? */
dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
}
/*
* Increase the file length
*
* IN: zp - znode of file to free data in.
* end - new end-of-file
*
* RETURN: 0 on success, error code on failure
*/
static int
zfs_extend(znode_t *zp, uint64_t end)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
dmu_tx_t *tx;
zfs_locked_range_t *lr;
uint64_t newblksz;
int error;
/*
* We will change zp_size, lock the whole file.
*/
lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER);
/*
* Nothing to do if file already at desired length.
*/
if (end <= zp->z_size) {
zfs_rangelock_exit(lr);
return (0);
}
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
if (end > zp->z_blksz &&
(!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
/*
* We are growing the file past the current block size.
*/
if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
/*
* File's blocksize is already larger than the
* "recordsize" property. Only let it grow to
* the next power of 2.
*/
ASSERT(!ISP2(zp->z_blksz));
newblksz = MIN(end, 1 << highbit64(zp->z_blksz));
} else {
newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
}
dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
} else {
newblksz = 0;
}
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_rangelock_exit(lr);
return (error);
}
if (newblksz)
zfs_grow_blocksize(zp, newblksz, tx);
zp->z_size = end;
VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)),
&zp->z_size, sizeof (zp->z_size), tx));
zfs_rangelock_exit(lr);
dmu_tx_commit(tx);
return (0);
}
/*
* zfs_zero_partial_page - Modeled after update_pages() but
* with different arguments and semantics for use by zfs_freesp().
*
* Zeroes a piece of a single page cache entry for zp at offset
* start and length len.
*
* Caller must acquire a range lock on the file for the region
* being zeroed in order that the ARC and page cache stay in sync.
*/
static void
zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len)
{
struct address_space *mp = ZTOI(zp)->i_mapping;
struct page *pp;
int64_t off;
void *pb;
ASSERT((start & PAGE_MASK) == ((start + len - 1) & PAGE_MASK));
off = start & (PAGE_SIZE - 1);
start &= PAGE_MASK;
pp = find_lock_page(mp, start >> PAGE_SHIFT);
if (pp) {
if (mapping_writably_mapped(mp))
flush_dcache_page(pp);
pb = kmap(pp);
bzero(pb + off, len);
kunmap(pp);
if (mapping_writably_mapped(mp))
flush_dcache_page(pp);
mark_page_accessed(pp);
SetPageUptodate(pp);
ClearPageError(pp);
unlock_page(pp);
put_page(pp);
}
}
/*
* Free space in a file.
*
* IN: zp - znode of file to free data in.
* off - start of section to free.
* len - length of section to free.
*
* RETURN: 0 on success, error code on failure
*/
static int
zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
zfs_locked_range_t *lr;
int error;
/*
* Lock the range being freed.
*/
lr = zfs_rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER);
/*
* Nothing to do if file already at desired length.
*/
if (off >= zp->z_size) {
zfs_rangelock_exit(lr);
return (0);
}
if (off + len > zp->z_size)
len = zp->z_size - off;
error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
/*
* Zero partial page cache entries. This must be done under a
* range lock in order to keep the ARC and page cache in sync.
*/
if (zp->z_is_mapped) {
loff_t first_page, last_page, page_len;
loff_t first_page_offset, last_page_offset;
/* first possible full page in hole */
first_page = (off + PAGE_SIZE - 1) >> PAGE_SHIFT;
/* last page of hole */
last_page = (off + len) >> PAGE_SHIFT;
/* offset of first_page */
first_page_offset = first_page << PAGE_SHIFT;
/* offset of last_page */
last_page_offset = last_page << PAGE_SHIFT;
/* truncate whole pages */
if (last_page_offset > first_page_offset) {
truncate_inode_pages_range(ZTOI(zp)->i_mapping,
first_page_offset, last_page_offset - 1);
}
/* truncate sub-page ranges */
if (first_page > last_page) {
/* entire punched area within a single page */
zfs_zero_partial_page(zp, off, len);
} else {
/* beginning of punched area at the end of a page */
page_len = first_page_offset - off;
if (page_len > 0)
zfs_zero_partial_page(zp, off, page_len);
/* end of punched area at the beginning of a page */
page_len = off + len - last_page_offset;
if (page_len > 0)
zfs_zero_partial_page(zp, last_page_offset,
page_len);
}
}
zfs_rangelock_exit(lr);
return (error);
}
/*
* Truncate a file
*
* IN: zp - znode of file to free data in.
* end - new end-of-file.
*
* RETURN: 0 on success, error code on failure
*/
static int
zfs_trunc(znode_t *zp, uint64_t end)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
dmu_tx_t *tx;
zfs_locked_range_t *lr;
int error;
sa_bulk_attr_t bulk[2];
int count = 0;
/*
* We will change zp_size, lock the whole file.
*/
lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER);
/*
* Nothing to do if file already at desired length.
*/
if (end >= zp->z_size) {
zfs_rangelock_exit(lr);
return (0);
}
error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end,
DMU_OBJECT_END);
if (error) {
zfs_rangelock_exit(lr);
return (error);
}
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_rangelock_exit(lr);
return (error);
}
zp->z_size = end;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
NULL, &zp->z_size, sizeof (zp->z_size));
if (end == 0) {
zp->z_pflags &= ~ZFS_SPARSE;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
NULL, &zp->z_pflags, 8);
}
VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
dmu_tx_commit(tx);
zfs_rangelock_exit(lr);
return (0);
}
/*
* Free space in a file
*
* IN: zp - znode of file to free data in.
* off - start of range
* len - end of range (0 => EOF)
* flag - current file open mode flags.
* log - TRUE if this action should be logged
*
* RETURN: 0 on success, error code on failure
*/
int
zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
{
dmu_tx_t *tx;
zfsvfs_t *zfsvfs = ZTOZSB(zp);
zilog_t *zilog = zfsvfs->z_log;
uint64_t mode;
uint64_t mtime[2], ctime[2];
sa_bulk_attr_t bulk[3];
int count = 0;
int error;
if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
sizeof (mode))) != 0)
return (error);
if (off > zp->z_size) {
error = zfs_extend(zp, off+len);
if (error == 0 && log)
goto log;
goto out;
}
if (len == 0) {
error = zfs_trunc(zp, off);
} else {
if ((error = zfs_free_range(zp, off, len)) == 0 &&
off + len > zp->z_size)
error = zfs_extend(zp, off+len);
}
if (error || !log)
goto out;
log:
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
goto out;
}
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
NULL, &zp->z_pflags, 8);
zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
ASSERT(error == 0);
zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
dmu_tx_commit(tx);
zfs_znode_update_vfs(zp);
error = 0;
out:
/*
* Truncate the page cache - for file truncate operations, use
* the purpose-built API for truncations. For punching operations,
* the truncation is handled under a range lock in zfs_free_range.
*/
if (len == 0)
truncate_setsize(ZTOI(zp), off);
return (error);
}
void
zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
{
struct super_block *sb;
zfsvfs_t *zfsvfs;
uint64_t moid, obj, sa_obj, version;
uint64_t sense = ZFS_CASE_SENSITIVE;
uint64_t norm = 0;
nvpair_t *elem;
int size;
int error;
int i;
znode_t *rootzp = NULL;
vattr_t vattr;
znode_t *zp;
zfs_acl_ids_t acl_ids;
/*
* First attempt to create master node.
*/
/*
* In an empty objset, there are no blocks to read and thus
* there can be no i/o errors (which we assert below).
*/
moid = MASTER_NODE_OBJ;
error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
/*
* Set starting attributes.
*/
version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
elem = NULL;
while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
/* For the moment we expect all zpl props to be uint64_ts */
uint64_t val;
char *name;
ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
VERIFY(nvpair_value_uint64(elem, &val) == 0);
name = nvpair_name(elem);
if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
if (val < version)
version = val;
} else {
error = zap_update(os, moid, name, 8, 1, &val, tx);
}
ASSERT(error == 0);
if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
norm = val;
else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
sense = val;
}
ASSERT(version != 0);
error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
/*
* Create zap object used for SA attribute registration
*/
if (version >= ZPL_VERSION_SA) {
sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
DMU_OT_NONE, 0, tx);
error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
ASSERT(error == 0);
} else {
sa_obj = 0;
}
/*
* Create a delete queue.
*/
obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
ASSERT(error == 0);
/*
* Create root znode. Create minimal znode/inode/zfsvfs/sb
* to allow zfs_mknode to work.
*/
vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID;
vattr.va_mode = S_IFDIR|0755;
vattr.va_uid = crgetuid(cr);
vattr.va_gid = crgetgid(cr);
rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
rootzp->z_unlinked = B_FALSE;
rootzp->z_atime_dirty = B_FALSE;
rootzp->z_is_sa = USE_SA(version, os);
rootzp->z_pflags = 0;
zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
zfsvfs->z_os = os;
zfsvfs->z_parent = zfsvfs;
zfsvfs->z_version = version;
zfsvfs->z_use_fuids = USE_FUIDS(version, os);
zfsvfs->z_use_sa = USE_SA(version, os);
zfsvfs->z_norm = norm;
sb = kmem_zalloc(sizeof (struct super_block), KM_SLEEP);
sb->s_fs_info = zfsvfs;
ZTOI(rootzp)->i_sb = sb;
error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
&zfsvfs->z_attr_table);
ASSERT(error == 0);
/*
* Fold case on file systems that are always or sometimes case
* insensitive.
*/
if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
offsetof(znode_t, z_link_node));
size = MIN(1 << (highbit64(zfs_object_mutex_size)-1), ZFS_OBJ_MTX_MAX);
zfsvfs->z_hold_size = size;
zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size,
KM_SLEEP);
zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP);
for (i = 0; i != size; i++) {
avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare,
sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node));
mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL);
}
VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
cr, NULL, &acl_ids));
zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
ASSERT3P(zp, ==, rootzp);
error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
ASSERT(error == 0);
zfs_acl_ids_free(&acl_ids);
atomic_set(&ZTOI(rootzp)->i_count, 0);
sa_handle_destroy(rootzp->z_sa_hdl);
kmem_cache_free(znode_cache, rootzp);
for (i = 0; i != size; i++) {
avl_destroy(&zfsvfs->z_hold_trees[i]);
mutex_destroy(&zfsvfs->z_hold_locks[i]);
}
mutex_destroy(&zfsvfs->z_znodes_lock);
vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size);
vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size);
kmem_free(sb, sizeof (struct super_block));
kmem_free(zfsvfs, sizeof (zfsvfs_t));
}
#endif /* _KERNEL */
static int
zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
{
uint64_t sa_obj = 0;
int error;
error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
if (error != 0 && error != ENOENT)
return (error);
error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
return (error);
}
static int
zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
dmu_buf_t **db, void *tag)
{
dmu_object_info_t doi;
int error;
if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
return (error);
dmu_object_info_from_db(*db, &doi);
if ((doi.doi_bonus_type != DMU_OT_SA &&
doi.doi_bonus_type != DMU_OT_ZNODE) ||
(doi.doi_bonus_type == DMU_OT_ZNODE &&
doi.doi_bonus_size < sizeof (znode_phys_t))) {
sa_buf_rele(*db, tag);
return (SET_ERROR(ENOTSUP));
}
error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
if (error != 0) {
sa_buf_rele(*db, tag);
return (error);
}
return (0);
}
static void
zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
{
sa_handle_destroy(hdl);
sa_buf_rele(db, tag);
}
/*
* Given an object number, return its parent object number and whether
* or not the object is an extended attribute directory.
*/
static int
zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table,
uint64_t *pobjp, int *is_xattrdir)
{
uint64_t parent;
uint64_t pflags;
uint64_t mode;
uint64_t parent_mode;
sa_bulk_attr_t bulk[3];
sa_handle_t *sa_hdl;
dmu_buf_t *sa_db;
int count = 0;
int error;
SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
&parent, sizeof (parent));
SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
&pflags, sizeof (pflags));
SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
&mode, sizeof (mode));
if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
return (error);
/*
* When a link is removed its parent pointer is not changed and will
* be invalid. There are two cases where a link is removed but the
* file stays around, when it goes to the delete queue and when there
* are additional links.
*/
error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG);
if (error != 0)
return (error);
error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode));
zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
if (error != 0)
return (error);
*is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
/*
* Extended attributes can be applied to files, directories, etc.
* Otherwise the parent must be a directory.
*/
if (!*is_xattrdir && !S_ISDIR(parent_mode))
return (SET_ERROR(EINVAL));
*pobjp = parent;
return (0);
}
/*
* Given an object number, return some zpl level statistics
*/
static int
zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
zfs_stat_t *sb)
{
sa_bulk_attr_t bulk[4];
int count = 0;
SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
&sb->zs_mode, sizeof (sb->zs_mode));
SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
&sb->zs_gen, sizeof (sb->zs_gen));
SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
&sb->zs_links, sizeof (sb->zs_links));
SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
&sb->zs_ctime, sizeof (sb->zs_ctime));
return (sa_bulk_lookup(hdl, bulk, count));
}
static int
zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
sa_attr_type_t *sa_table, char *buf, int len)
{
sa_handle_t *sa_hdl;
sa_handle_t *prevhdl = NULL;
dmu_buf_t *prevdb = NULL;
dmu_buf_t *sa_db = NULL;
char *path = buf + len - 1;
int error;
*path = '\0';
sa_hdl = hdl;
uint64_t deleteq_obj;
VERIFY0(zap_lookup(osp, MASTER_NODE_OBJ,
ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj));
error = zap_lookup_int(osp, deleteq_obj, obj);
if (error == 0) {
return (ESTALE);
} else if (error != ENOENT) {
return (error);
}
error = 0;
for (;;) {
uint64_t pobj = 0;
char component[MAXNAMELEN + 2];
size_t complen;
int is_xattrdir = 0;
if (prevdb) {
ASSERT(prevhdl != NULL);
zfs_release_sa_handle(prevhdl, prevdb, FTAG);
}
if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj,
&is_xattrdir)) != 0)
break;
if (pobj == obj) {
if (path[0] != '/')
*--path = '/';
break;
}
component[0] = '/';
if (is_xattrdir) {
(void) sprintf(component + 1, "<xattrdir>");
} else {
error = zap_value_search(osp, pobj, obj,
ZFS_DIRENT_OBJ(-1ULL), component + 1);
if (error != 0)
break;
}
complen = strlen(component);
path -= complen;
ASSERT(path >= buf);
bcopy(component, path, complen);
obj = pobj;
if (sa_hdl != hdl) {
prevhdl = sa_hdl;
prevdb = sa_db;
}
error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
if (error != 0) {
sa_hdl = prevhdl;
sa_db = prevdb;
break;
}
}
if (sa_hdl != NULL && sa_hdl != hdl) {
ASSERT(sa_db != NULL);
zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
}
if (error == 0)
(void) memmove(buf, path, buf + len - path);
return (error);
}
int
zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
{
sa_attr_type_t *sa_table;
sa_handle_t *hdl;
dmu_buf_t *db;
int error;
error = zfs_sa_setup(osp, &sa_table);
if (error != 0)
return (error);
error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
if (error != 0)
return (error);
error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
zfs_release_sa_handle(hdl, db, FTAG);
return (error);
}
int
zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
char *buf, int len)
{
char *path = buf + len - 1;
sa_attr_type_t *sa_table;
sa_handle_t *hdl;
dmu_buf_t *db;
int error;
*path = '\0';
error = zfs_sa_setup(osp, &sa_table);
if (error != 0)
return (error);
error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
if (error != 0)
return (error);
error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
if (error != 0) {
zfs_release_sa_handle(hdl, db, FTAG);
return (error);
}
error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
zfs_release_sa_handle(hdl, db, FTAG);
return (error);
}
#if defined(_KERNEL)
EXPORT_SYMBOL(zfs_create_fs);
EXPORT_SYMBOL(zfs_obj_to_path);
/* CSTYLED */
module_param(zfs_object_mutex_size, uint, 0644);
MODULE_PARM_DESC(zfs_object_mutex_size, "Size of znode hold array");
module_param(zfs_unlink_suspend_progress, int, 0644);
MODULE_PARM_DESC(zfs_unlink_suspend_progress, "Set to prevent async unlinks "
"(debug - leaks space into the unlinked set)");
#endif
diff --git a/module/os/linux/zfs/zio_crypt.c b/module/os/linux/zfs/zio_crypt.c
index 2c82be2d7d82..9f8b9f53e965 100644
--- a/module/os/linux/zfs/zio_crypt.c
+++ b/module/os/linux/zfs/zio_crypt.c
@@ -1,2044 +1,2044 @@
/*
* CDDL HEADER START
*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2017, Datto, Inc. All rights reserved.
*/
#include <sys/zio_crypt.h>
#include <sys/dmu.h>
#include <sys/dmu_objset.h>
#include <sys/dnode.h>
#include <sys/fs/zfs.h>
#include <sys/zio.h>
#include <sys/zil.h>
#include <sys/sha2.h>
#include <sys/hkdf.h>
#include <sys/qat.h>
/*
* This file is responsible for handling all of the details of generating
* encryption parameters and performing encryption and authentication.
*
* BLOCK ENCRYPTION PARAMETERS:
* Encryption /Authentication Algorithm Suite (crypt):
* The encryption algorithm, mode, and key length we are going to use. We
* currently support AES in either GCM or CCM modes with 128, 192, and 256 bit
* keys. All authentication is currently done with SHA512-HMAC.
*
* Plaintext:
* The unencrypted data that we want to encrypt.
*
* Initialization Vector (IV):
* An initialization vector for the encryption algorithms. This is used to
* "tweak" the encryption algorithms so that two blocks of the same data are
* encrypted into different ciphertext outputs, thus obfuscating block patterns.
* The supported encryption modes (AES-GCM and AES-CCM) require that an IV is
* never reused with the same encryption key. This value is stored unencrypted
* and must simply be provided to the decryption function. We use a 96 bit IV
* (as recommended by NIST) for all block encryption. For non-dedup blocks we
* derive the IV randomly. The first 64 bits of the IV are stored in the second
* word of DVA[2] and the remaining 32 bits are stored in the upper 32 bits of
* blk_fill. This is safe because encrypted blocks can't use the upper 32 bits
* of blk_fill. We only encrypt level 0 blocks, which normally have a fill count
* of 1. The only exception is for DMU_OT_DNODE objects, where the fill count of
* level 0 blocks is the number of allocated dnodes in that block. The on-disk
* format supports at most 2^15 slots per L0 dnode block, because the maximum
* block size is 16MB (2^24). In either case, for level 0 blocks this number
* will still be smaller than UINT32_MAX so it is safe to store the IV in the
* top 32 bits of blk_fill, while leaving the bottom 32 bits of the fill count
* for the dnode code.
*
* Master key:
* This is the most important secret data of an encrypted dataset. It is used
* along with the salt to generate that actual encryption keys via HKDF. We
* do not use the master key to directly encrypt any data because there are
* theoretical limits on how much data can actually be safely encrypted with
* any encryption mode. The master key is stored encrypted on disk with the
* user's wrapping key. Its length is determined by the encryption algorithm.
* For details on how this is stored see the block comment in dsl_crypt.c
*
* Salt:
* Used as an input to the HKDF function, along with the master key. We use a
* 64 bit salt, stored unencrypted in the first word of DVA[2]. Any given salt
* can be used for encrypting many blocks, so we cache the current salt and the
* associated derived key in zio_crypt_t so we do not need to derive it again
* needlessly.
*
* Encryption Key:
* A secret binary key, generated from an HKDF function used to encrypt and
* decrypt data.
*
* Message Authentication Code (MAC)
* The MAC is an output of authenticated encryption modes such as AES-GCM and
* AES-CCM. Its purpose is to ensure that an attacker cannot modify encrypted
* data on disk and return garbage to the application. Effectively, it is a
* checksum that can not be reproduced by an attacker. We store the MAC in the
* second 128 bits of blk_cksum, leaving the first 128 bits for a truncated
* regular checksum of the ciphertext which can be used for scrubbing.
*
* OBJECT AUTHENTICATION:
* Some object types, such as DMU_OT_MASTER_NODE cannot be encrypted because
* they contain some info that always needs to be readable. To prevent this
* data from being altered, we authenticate this data using SHA512-HMAC. This
* will produce a MAC (similar to the one produced via encryption) which can
* be used to verify the object was not modified. HMACs do not require key
* rotation or IVs, so we can keep up to the full 3 copies of authenticated
* data.
*
* ZIL ENCRYPTION:
* ZIL blocks have their bp written to disk ahead of the associated data, so we
* cannot store the MAC there as we normally do. For these blocks the MAC is
* stored in the embedded checksum within the zil_chain_t header. The salt and
* IV are generated for the block on bp allocation instead of at encryption
* time. In addition, ZIL blocks have some pieces that must be left in plaintext
* for claiming even though all of the sensitive user data still needs to be
* encrypted. The function zio_crypt_init_uios_zil() handles parsing which
* pieces of the block need to be encrypted. All data that is not encrypted is
* authenticated using the AAD mechanisms that the supported encryption modes
* provide for. In order to preserve the semantics of the ZIL for encrypted
* datasets, the ZIL is not protected at the objset level as described below.
*
* DNODE ENCRYPTION:
* Similarly to ZIL blocks, the core part of each dnode_phys_t needs to be left
* in plaintext for scrubbing and claiming, but the bonus buffers might contain
* sensitive user data. The function zio_crypt_init_uios_dnode() handles parsing
* which pieces of the block need to be encrypted. For more details about
* dnode authentication and encryption, see zio_crypt_init_uios_dnode().
*
* OBJECT SET AUTHENTICATION:
* Up to this point, everything we have encrypted and authenticated has been
* at level 0 (or -2 for the ZIL). If we did not do any further work the
* on-disk format would be susceptible to attacks that deleted or rearranged
* the order of level 0 blocks. Ideally, the cleanest solution would be to
* maintain a tree of authentication MACs going up the bp tree. However, this
* presents a problem for raw sends. Send files do not send information about
* indirect blocks so there would be no convenient way to transfer the MACs and
* they cannot be recalculated on the receive side without the master key which
* would defeat one of the purposes of raw sends in the first place. Instead,
* for the indirect levels of the bp tree, we use a regular SHA512 of the MACs
* from the level below. We also include some portable fields from blk_prop such
* as the lsize and compression algorithm to prevent the data from being
* misinterpreted.
*
* At the objset level, we maintain 2 separate 256 bit MACs in the
* objset_phys_t. The first one is "portable" and is the logical root of the
* MAC tree maintained in the metadnode's bps. The second, is "local" and is
* used as the root MAC for the user accounting objects, which are also not
* transferred via "zfs send". The portable MAC is sent in the DRR_BEGIN payload
* of the send file. The useraccounting code ensures that the useraccounting
* info is not present upon a receive, so the local MAC can simply be cleared
* out at that time. For more info about objset_phys_t authentication, see
* zio_crypt_do_objset_hmacs().
*
* CONSIDERATIONS FOR DEDUP:
* In order for dedup to work, blocks that we want to dedup with one another
* need to use the same IV and encryption key, so that they will have the same
* ciphertext. Normally, one should never reuse an IV with the same encryption
* key or else AES-GCM and AES-CCM can both actually leak the plaintext of both
* blocks. In this case, however, since we are using the same plaintext as
* well all that we end up with is a duplicate of the original ciphertext we
* already had. As a result, an attacker with read access to the raw disk will
* be able to tell which blocks are the same but this information is given away
* by dedup anyway. In order to get the same IVs and encryption keys for
* equivalent blocks of data we use an HMAC of the plaintext. We use an HMAC
* here so that a reproducible checksum of the plaintext is never available to
* the attacker. The HMAC key is kept alongside the master key, encrypted on
* disk. The first 64 bits of the HMAC are used in place of the random salt, and
* the next 96 bits are used as the IV. As a result of this mechanism, dedup
* will only work within a clone family since encrypted dedup requires use of
* the same master and HMAC keys.
*/
/*
* After encrypting many blocks with the same key we may start to run up
* against the theoretical limits of how much data can securely be encrypted
* with a single key using the supported encryption modes. The most obvious
* limitation is that our risk of generating 2 equivalent 96 bit IVs increases
* the more IVs we generate (which both GCM and CCM modes strictly forbid).
* This risk actually grows surprisingly quickly over time according to the
* Birthday Problem. With a total IV space of 2^(96 bits), and assuming we have
* generated n IVs with a cryptographically secure RNG, the approximate
* probability p(n) of a collision is given as:
*
* p(n) ~= e^(-n*(n-1)/(2*(2^96)))
*
* [http://www.math.cornell.edu/~mec/2008-2009/TianyiZheng/Birthday.html]
*
* Assuming that we want to ensure that p(n) never goes over 1 / 1 trillion
* we must not write more than 398,065,730 blocks with the same encryption key.
* Therefore, we rotate our keys after 400,000,000 blocks have been written by
* generating a new random 64 bit salt for our HKDF encryption key generation
* function.
*/
#define ZFS_KEY_MAX_SALT_USES_DEFAULT 400000000
#define ZFS_CURRENT_MAX_SALT_USES \
(MIN(zfs_key_max_salt_uses, ZFS_KEY_MAX_SALT_USES_DEFAULT))
-unsigned long zfs_key_max_salt_uses = ZFS_KEY_MAX_SALT_USES_DEFAULT;
+static unsigned long zfs_key_max_salt_uses = ZFS_KEY_MAX_SALT_USES_DEFAULT;
typedef struct blkptr_auth_buf {
uint64_t bab_prop; /* blk_prop - portable mask */
uint8_t bab_mac[ZIO_DATA_MAC_LEN]; /* MAC from blk_cksum */
uint64_t bab_pad; /* reserved for future use */
} blkptr_auth_buf_t;
-zio_crypt_info_t zio_crypt_table[ZIO_CRYPT_FUNCTIONS] = {
+const zio_crypt_info_t zio_crypt_table[ZIO_CRYPT_FUNCTIONS] = {
{"", ZC_TYPE_NONE, 0, "inherit"},
{"", ZC_TYPE_NONE, 0, "on"},
{"", ZC_TYPE_NONE, 0, "off"},
{SUN_CKM_AES_CCM, ZC_TYPE_CCM, 16, "aes-128-ccm"},
{SUN_CKM_AES_CCM, ZC_TYPE_CCM, 24, "aes-192-ccm"},
{SUN_CKM_AES_CCM, ZC_TYPE_CCM, 32, "aes-256-ccm"},
{SUN_CKM_AES_GCM, ZC_TYPE_GCM, 16, "aes-128-gcm"},
{SUN_CKM_AES_GCM, ZC_TYPE_GCM, 24, "aes-192-gcm"},
{SUN_CKM_AES_GCM, ZC_TYPE_GCM, 32, "aes-256-gcm"}
};
void
zio_crypt_key_destroy(zio_crypt_key_t *key)
{
rw_destroy(&key->zk_salt_lock);
/* free crypto templates */
crypto_destroy_ctx_template(key->zk_current_tmpl);
crypto_destroy_ctx_template(key->zk_hmac_tmpl);
/* zero out sensitive data */
bzero(key, sizeof (zio_crypt_key_t));
}
int
zio_crypt_key_init(uint64_t crypt, zio_crypt_key_t *key)
{
int ret;
crypto_mechanism_t mech;
uint_t keydata_len;
ASSERT(key != NULL);
ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);
keydata_len = zio_crypt_table[crypt].ci_keylen;
bzero(key, sizeof (zio_crypt_key_t));
/* fill keydata buffers and salt with random data */
ret = random_get_bytes((uint8_t *)&key->zk_guid, sizeof (uint64_t));
if (ret != 0)
goto error;
ret = random_get_bytes(key->zk_master_keydata, keydata_len);
if (ret != 0)
goto error;
ret = random_get_bytes(key->zk_hmac_keydata, SHA512_HMAC_KEYLEN);
if (ret != 0)
goto error;
ret = random_get_bytes(key->zk_salt, ZIO_DATA_SALT_LEN);
if (ret != 0)
goto error;
/* derive the current key from the master key */
ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
key->zk_salt, ZIO_DATA_SALT_LEN, key->zk_current_keydata,
keydata_len);
if (ret != 0)
goto error;
/* initialize keys for the ICP */
key->zk_current_key.ck_format = CRYPTO_KEY_RAW;
key->zk_current_key.ck_data = key->zk_current_keydata;
key->zk_current_key.ck_length = CRYPTO_BYTES2BITS(keydata_len);
key->zk_hmac_key.ck_format = CRYPTO_KEY_RAW;
key->zk_hmac_key.ck_data = &key->zk_hmac_key;
key->zk_hmac_key.ck_length = CRYPTO_BYTES2BITS(SHA512_HMAC_KEYLEN);
/*
* Initialize the crypto templates. It's ok if this fails because
* this is just an optimization.
*/
mech.cm_type = crypto_mech2id(zio_crypt_table[crypt].ci_mechname);
ret = crypto_create_ctx_template(&mech, &key->zk_current_key,
&key->zk_current_tmpl, KM_SLEEP);
if (ret != CRYPTO_SUCCESS)
key->zk_current_tmpl = NULL;
mech.cm_type = crypto_mech2id(SUN_CKM_SHA512_HMAC);
ret = crypto_create_ctx_template(&mech, &key->zk_hmac_key,
&key->zk_hmac_tmpl, KM_SLEEP);
if (ret != CRYPTO_SUCCESS)
key->zk_hmac_tmpl = NULL;
key->zk_crypt = crypt;
key->zk_version = ZIO_CRYPT_KEY_CURRENT_VERSION;
key->zk_salt_count = 0;
rw_init(&key->zk_salt_lock, NULL, RW_DEFAULT, NULL);
return (0);
error:
zio_crypt_key_destroy(key);
return (ret);
}
static int
zio_crypt_key_change_salt(zio_crypt_key_t *key)
{
int ret = 0;
uint8_t salt[ZIO_DATA_SALT_LEN];
crypto_mechanism_t mech;
uint_t keydata_len = zio_crypt_table[key->zk_crypt].ci_keylen;
/* generate a new salt */
ret = random_get_bytes(salt, ZIO_DATA_SALT_LEN);
if (ret != 0)
goto error;
rw_enter(&key->zk_salt_lock, RW_WRITER);
/* someone beat us to the salt rotation, just unlock and return */
if (key->zk_salt_count < ZFS_CURRENT_MAX_SALT_USES)
goto out_unlock;
/* derive the current key from the master key and the new salt */
ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
salt, ZIO_DATA_SALT_LEN, key->zk_current_keydata, keydata_len);
if (ret != 0)
goto out_unlock;
/* assign the salt and reset the usage count */
bcopy(salt, key->zk_salt, ZIO_DATA_SALT_LEN);
key->zk_salt_count = 0;
/* destroy the old context template and create the new one */
crypto_destroy_ctx_template(key->zk_current_tmpl);
ret = crypto_create_ctx_template(&mech, &key->zk_current_key,
&key->zk_current_tmpl, KM_SLEEP);
if (ret != CRYPTO_SUCCESS)
key->zk_current_tmpl = NULL;
rw_exit(&key->zk_salt_lock);
return (0);
out_unlock:
rw_exit(&key->zk_salt_lock);
error:
return (ret);
}
/* See comment above zfs_key_max_salt_uses definition for details */
int
zio_crypt_key_get_salt(zio_crypt_key_t *key, uint8_t *salt)
{
int ret;
boolean_t salt_change;
rw_enter(&key->zk_salt_lock, RW_READER);
bcopy(key->zk_salt, salt, ZIO_DATA_SALT_LEN);
salt_change = (atomic_inc_64_nv(&key->zk_salt_count) >=
ZFS_CURRENT_MAX_SALT_USES);
rw_exit(&key->zk_salt_lock);
if (salt_change) {
ret = zio_crypt_key_change_salt(key);
if (ret != 0)
goto error;
}
return (0);
error:
return (ret);
}
/*
* This function handles all encryption and decryption in zfs. When
* encrypting it expects puio to reference the plaintext and cuio to
* reference the ciphertext. cuio must have enough space for the
* ciphertext + room for a MAC. datalen should be the length of the
* plaintext / ciphertext alone.
*/
static int
zio_do_crypt_uio(boolean_t encrypt, uint64_t crypt, crypto_key_t *key,
crypto_ctx_template_t tmpl, uint8_t *ivbuf, uint_t datalen,
zfs_uio_t *puio, zfs_uio_t *cuio, uint8_t *authbuf, uint_t auth_len)
{
int ret;
crypto_data_t plaindata, cipherdata;
CK_AES_CCM_PARAMS ccmp;
CK_AES_GCM_PARAMS gcmp;
crypto_mechanism_t mech;
zio_crypt_info_t crypt_info;
uint_t plain_full_len, maclen;
ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);
ASSERT3U(key->ck_format, ==, CRYPTO_KEY_RAW);
/* lookup the encryption info */
crypt_info = zio_crypt_table[crypt];
/* the mac will always be the last iovec_t in the cipher uio */
maclen = cuio->uio_iov[cuio->uio_iovcnt - 1].iov_len;
ASSERT(maclen <= ZIO_DATA_MAC_LEN);
/* setup encryption mechanism (same as crypt) */
mech.cm_type = crypto_mech2id(crypt_info.ci_mechname);
/*
* Strangely, the ICP requires that plain_full_len must include
* the MAC length when decrypting, even though the UIO does not
* need to have the extra space allocated.
*/
if (encrypt) {
plain_full_len = datalen;
} else {
plain_full_len = datalen + maclen;
}
/*
* setup encryption params (currently only AES CCM and AES GCM
* are supported)
*/
if (crypt_info.ci_crypt_type == ZC_TYPE_CCM) {
ccmp.ulNonceSize = ZIO_DATA_IV_LEN;
ccmp.ulAuthDataSize = auth_len;
ccmp.authData = authbuf;
ccmp.ulMACSize = maclen;
ccmp.nonce = ivbuf;
ccmp.ulDataSize = plain_full_len;
mech.cm_param = (char *)(&ccmp);
mech.cm_param_len = sizeof (CK_AES_CCM_PARAMS);
} else {
gcmp.ulIvLen = ZIO_DATA_IV_LEN;
gcmp.ulIvBits = CRYPTO_BYTES2BITS(ZIO_DATA_IV_LEN);
gcmp.ulAADLen = auth_len;
gcmp.pAAD = authbuf;
gcmp.ulTagBits = CRYPTO_BYTES2BITS(maclen);
gcmp.pIv = ivbuf;
mech.cm_param = (char *)(&gcmp);
mech.cm_param_len = sizeof (CK_AES_GCM_PARAMS);
}
/* populate the cipher and plain data structs. */
plaindata.cd_format = CRYPTO_DATA_UIO;
plaindata.cd_offset = 0;
plaindata.cd_uio = puio;
plaindata.cd_miscdata = NULL;
plaindata.cd_length = plain_full_len;
cipherdata.cd_format = CRYPTO_DATA_UIO;
cipherdata.cd_offset = 0;
cipherdata.cd_uio = cuio;
cipherdata.cd_miscdata = NULL;
cipherdata.cd_length = datalen + maclen;
/* perform the actual encryption */
if (encrypt) {
ret = crypto_encrypt(&mech, &plaindata, key, tmpl, &cipherdata,
NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
} else {
ret = crypto_decrypt(&mech, &cipherdata, key, tmpl, &plaindata,
NULL);
if (ret != CRYPTO_SUCCESS) {
ASSERT3U(ret, ==, CRYPTO_INVALID_MAC);
ret = SET_ERROR(ECKSUM);
goto error;
}
}
return (0);
error:
return (ret);
}
int
zio_crypt_key_wrap(crypto_key_t *cwkey, zio_crypt_key_t *key, uint8_t *iv,
uint8_t *mac, uint8_t *keydata_out, uint8_t *hmac_keydata_out)
{
int ret;
zfs_uio_t puio, cuio;
uint64_t aad[3];
iovec_t plain_iovecs[2], cipher_iovecs[3];
uint64_t crypt = key->zk_crypt;
uint_t enc_len, keydata_len, aad_len;
ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);
ASSERT3U(cwkey->ck_format, ==, CRYPTO_KEY_RAW);
keydata_len = zio_crypt_table[crypt].ci_keylen;
/* generate iv for wrapping the master and hmac key */
ret = random_get_pseudo_bytes(iv, WRAPPING_IV_LEN);
if (ret != 0)
goto error;
/* initialize zfs_uio_ts */
plain_iovecs[0].iov_base = key->zk_master_keydata;
plain_iovecs[0].iov_len = keydata_len;
plain_iovecs[1].iov_base = key->zk_hmac_keydata;
plain_iovecs[1].iov_len = SHA512_HMAC_KEYLEN;
cipher_iovecs[0].iov_base = keydata_out;
cipher_iovecs[0].iov_len = keydata_len;
cipher_iovecs[1].iov_base = hmac_keydata_out;
cipher_iovecs[1].iov_len = SHA512_HMAC_KEYLEN;
cipher_iovecs[2].iov_base = mac;
cipher_iovecs[2].iov_len = WRAPPING_MAC_LEN;
/*
* Although we don't support writing to the old format, we do
* support rewrapping the key so that the user can move and
* quarantine datasets on the old format.
*/
if (key->zk_version == 0) {
aad_len = sizeof (uint64_t);
aad[0] = LE_64(key->zk_guid);
} else {
ASSERT3U(key->zk_version, ==, ZIO_CRYPT_KEY_CURRENT_VERSION);
aad_len = sizeof (uint64_t) * 3;
aad[0] = LE_64(key->zk_guid);
aad[1] = LE_64(crypt);
aad[2] = LE_64(key->zk_version);
}
enc_len = zio_crypt_table[crypt].ci_keylen + SHA512_HMAC_KEYLEN;
puio.uio_iov = plain_iovecs;
puio.uio_iovcnt = 2;
puio.uio_segflg = UIO_SYSSPACE;
cuio.uio_iov = cipher_iovecs;
cuio.uio_iovcnt = 3;
cuio.uio_segflg = UIO_SYSSPACE;
/* encrypt the keys and store the resulting ciphertext and mac */
ret = zio_do_crypt_uio(B_TRUE, crypt, cwkey, NULL, iv, enc_len,
&puio, &cuio, (uint8_t *)aad, aad_len);
if (ret != 0)
goto error;
return (0);
error:
return (ret);
}
int
zio_crypt_key_unwrap(crypto_key_t *cwkey, uint64_t crypt, uint64_t version,
uint64_t guid, uint8_t *keydata, uint8_t *hmac_keydata, uint8_t *iv,
uint8_t *mac, zio_crypt_key_t *key)
{
crypto_mechanism_t mech;
zfs_uio_t puio, cuio;
uint64_t aad[3];
iovec_t plain_iovecs[2], cipher_iovecs[3];
uint_t enc_len, keydata_len, aad_len;
int ret;
ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);
ASSERT3U(cwkey->ck_format, ==, CRYPTO_KEY_RAW);
rw_init(&key->zk_salt_lock, NULL, RW_DEFAULT, NULL);
keydata_len = zio_crypt_table[crypt].ci_keylen;
/* initialize zfs_uio_ts */
plain_iovecs[0].iov_base = key->zk_master_keydata;
plain_iovecs[0].iov_len = keydata_len;
plain_iovecs[1].iov_base = key->zk_hmac_keydata;
plain_iovecs[1].iov_len = SHA512_HMAC_KEYLEN;
cipher_iovecs[0].iov_base = keydata;
cipher_iovecs[0].iov_len = keydata_len;
cipher_iovecs[1].iov_base = hmac_keydata;
cipher_iovecs[1].iov_len = SHA512_HMAC_KEYLEN;
cipher_iovecs[2].iov_base = mac;
cipher_iovecs[2].iov_len = WRAPPING_MAC_LEN;
if (version == 0) {
aad_len = sizeof (uint64_t);
aad[0] = LE_64(guid);
} else {
ASSERT3U(version, ==, ZIO_CRYPT_KEY_CURRENT_VERSION);
aad_len = sizeof (uint64_t) * 3;
aad[0] = LE_64(guid);
aad[1] = LE_64(crypt);
aad[2] = LE_64(version);
}
enc_len = keydata_len + SHA512_HMAC_KEYLEN;
puio.uio_iov = plain_iovecs;
puio.uio_segflg = UIO_SYSSPACE;
puio.uio_iovcnt = 2;
cuio.uio_iov = cipher_iovecs;
cuio.uio_iovcnt = 3;
cuio.uio_segflg = UIO_SYSSPACE;
/* decrypt the keys and store the result in the output buffers */
ret = zio_do_crypt_uio(B_FALSE, crypt, cwkey, NULL, iv, enc_len,
&puio, &cuio, (uint8_t *)aad, aad_len);
if (ret != 0)
goto error;
/* generate a fresh salt */
ret = random_get_bytes(key->zk_salt, ZIO_DATA_SALT_LEN);
if (ret != 0)
goto error;
/* derive the current key from the master key */
ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
key->zk_salt, ZIO_DATA_SALT_LEN, key->zk_current_keydata,
keydata_len);
if (ret != 0)
goto error;
/* initialize keys for ICP */
key->zk_current_key.ck_format = CRYPTO_KEY_RAW;
key->zk_current_key.ck_data = key->zk_current_keydata;
key->zk_current_key.ck_length = CRYPTO_BYTES2BITS(keydata_len);
key->zk_hmac_key.ck_format = CRYPTO_KEY_RAW;
key->zk_hmac_key.ck_data = key->zk_hmac_keydata;
key->zk_hmac_key.ck_length = CRYPTO_BYTES2BITS(SHA512_HMAC_KEYLEN);
/*
* Initialize the crypto templates. It's ok if this fails because
* this is just an optimization.
*/
mech.cm_type = crypto_mech2id(zio_crypt_table[crypt].ci_mechname);
ret = crypto_create_ctx_template(&mech, &key->zk_current_key,
&key->zk_current_tmpl, KM_SLEEP);
if (ret != CRYPTO_SUCCESS)
key->zk_current_tmpl = NULL;
mech.cm_type = crypto_mech2id(SUN_CKM_SHA512_HMAC);
ret = crypto_create_ctx_template(&mech, &key->zk_hmac_key,
&key->zk_hmac_tmpl, KM_SLEEP);
if (ret != CRYPTO_SUCCESS)
key->zk_hmac_tmpl = NULL;
key->zk_crypt = crypt;
key->zk_version = version;
key->zk_guid = guid;
key->zk_salt_count = 0;
return (0);
error:
zio_crypt_key_destroy(key);
return (ret);
}
int
zio_crypt_generate_iv(uint8_t *ivbuf)
{
int ret;
/* randomly generate the IV */
ret = random_get_pseudo_bytes(ivbuf, ZIO_DATA_IV_LEN);
if (ret != 0)
goto error;
return (0);
error:
bzero(ivbuf, ZIO_DATA_IV_LEN);
return (ret);
}
int
zio_crypt_do_hmac(zio_crypt_key_t *key, uint8_t *data, uint_t datalen,
uint8_t *digestbuf, uint_t digestlen)
{
int ret;
crypto_mechanism_t mech;
crypto_data_t in_data, digest_data;
uint8_t raw_digestbuf[SHA512_DIGEST_LENGTH];
ASSERT3U(digestlen, <=, SHA512_DIGEST_LENGTH);
/* initialize sha512-hmac mechanism and crypto data */
mech.cm_type = crypto_mech2id(SUN_CKM_SHA512_HMAC);
mech.cm_param = NULL;
mech.cm_param_len = 0;
/* initialize the crypto data */
in_data.cd_format = CRYPTO_DATA_RAW;
in_data.cd_offset = 0;
in_data.cd_length = datalen;
in_data.cd_raw.iov_base = (char *)data;
in_data.cd_raw.iov_len = in_data.cd_length;
digest_data.cd_format = CRYPTO_DATA_RAW;
digest_data.cd_offset = 0;
digest_data.cd_length = SHA512_DIGEST_LENGTH;
digest_data.cd_raw.iov_base = (char *)raw_digestbuf;
digest_data.cd_raw.iov_len = digest_data.cd_length;
/* generate the hmac */
ret = crypto_mac(&mech, &in_data, &key->zk_hmac_key, key->zk_hmac_tmpl,
&digest_data, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
bcopy(raw_digestbuf, digestbuf, digestlen);
return (0);
error:
bzero(digestbuf, digestlen);
return (ret);
}
int
zio_crypt_generate_iv_salt_dedup(zio_crypt_key_t *key, uint8_t *data,
uint_t datalen, uint8_t *ivbuf, uint8_t *salt)
{
int ret;
uint8_t digestbuf[SHA512_DIGEST_LENGTH];
ret = zio_crypt_do_hmac(key, data, datalen,
digestbuf, SHA512_DIGEST_LENGTH);
if (ret != 0)
return (ret);
bcopy(digestbuf, salt, ZIO_DATA_SALT_LEN);
bcopy(digestbuf + ZIO_DATA_SALT_LEN, ivbuf, ZIO_DATA_IV_LEN);
return (0);
}
/*
* The following functions are used to encode and decode encryption parameters
* into blkptr_t and zil_header_t. The ICP wants to use these parameters as
* byte strings, which normally means that these strings would not need to deal
* with byteswapping at all. However, both blkptr_t and zil_header_t may be
* byteswapped by lower layers and so we must "undo" that byteswap here upon
* decoding and encoding in a non-native byteorder. These functions require
* that the byteorder bit is correct before being called.
*/
void
zio_crypt_encode_params_bp(blkptr_t *bp, uint8_t *salt, uint8_t *iv)
{
uint64_t val64;
uint32_t val32;
ASSERT(BP_IS_ENCRYPTED(bp));
if (!BP_SHOULD_BYTESWAP(bp)) {
bcopy(salt, &bp->blk_dva[2].dva_word[0], sizeof (uint64_t));
bcopy(iv, &bp->blk_dva[2].dva_word[1], sizeof (uint64_t));
bcopy(iv + sizeof (uint64_t), &val32, sizeof (uint32_t));
BP_SET_IV2(bp, val32);
} else {
bcopy(salt, &val64, sizeof (uint64_t));
bp->blk_dva[2].dva_word[0] = BSWAP_64(val64);
bcopy(iv, &val64, sizeof (uint64_t));
bp->blk_dva[2].dva_word[1] = BSWAP_64(val64);
bcopy(iv + sizeof (uint64_t), &val32, sizeof (uint32_t));
BP_SET_IV2(bp, BSWAP_32(val32));
}
}
void
zio_crypt_decode_params_bp(const blkptr_t *bp, uint8_t *salt, uint8_t *iv)
{
uint64_t val64;
uint32_t val32;
ASSERT(BP_IS_PROTECTED(bp));
/* for convenience, so callers don't need to check */
if (BP_IS_AUTHENTICATED(bp)) {
bzero(salt, ZIO_DATA_SALT_LEN);
bzero(iv, ZIO_DATA_IV_LEN);
return;
}
if (!BP_SHOULD_BYTESWAP(bp)) {
bcopy(&bp->blk_dva[2].dva_word[0], salt, sizeof (uint64_t));
bcopy(&bp->blk_dva[2].dva_word[1], iv, sizeof (uint64_t));
val32 = (uint32_t)BP_GET_IV2(bp);
bcopy(&val32, iv + sizeof (uint64_t), sizeof (uint32_t));
} else {
val64 = BSWAP_64(bp->blk_dva[2].dva_word[0]);
bcopy(&val64, salt, sizeof (uint64_t));
val64 = BSWAP_64(bp->blk_dva[2].dva_word[1]);
bcopy(&val64, iv, sizeof (uint64_t));
val32 = BSWAP_32((uint32_t)BP_GET_IV2(bp));
bcopy(&val32, iv + sizeof (uint64_t), sizeof (uint32_t));
}
}
void
zio_crypt_encode_mac_bp(blkptr_t *bp, uint8_t *mac)
{
uint64_t val64;
ASSERT(BP_USES_CRYPT(bp));
ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_OBJSET);
if (!BP_SHOULD_BYTESWAP(bp)) {
bcopy(mac, &bp->blk_cksum.zc_word[2], sizeof (uint64_t));
bcopy(mac + sizeof (uint64_t), &bp->blk_cksum.zc_word[3],
sizeof (uint64_t));
} else {
bcopy(mac, &val64, sizeof (uint64_t));
bp->blk_cksum.zc_word[2] = BSWAP_64(val64);
bcopy(mac + sizeof (uint64_t), &val64, sizeof (uint64_t));
bp->blk_cksum.zc_word[3] = BSWAP_64(val64);
}
}
void
zio_crypt_decode_mac_bp(const blkptr_t *bp, uint8_t *mac)
{
uint64_t val64;
ASSERT(BP_USES_CRYPT(bp) || BP_IS_HOLE(bp));
/* for convenience, so callers don't need to check */
if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
bzero(mac, ZIO_DATA_MAC_LEN);
return;
}
if (!BP_SHOULD_BYTESWAP(bp)) {
bcopy(&bp->blk_cksum.zc_word[2], mac, sizeof (uint64_t));
bcopy(&bp->blk_cksum.zc_word[3], mac + sizeof (uint64_t),
sizeof (uint64_t));
} else {
val64 = BSWAP_64(bp->blk_cksum.zc_word[2]);
bcopy(&val64, mac, sizeof (uint64_t));
val64 = BSWAP_64(bp->blk_cksum.zc_word[3]);
bcopy(&val64, mac + sizeof (uint64_t), sizeof (uint64_t));
}
}
void
zio_crypt_encode_mac_zil(void *data, uint8_t *mac)
{
zil_chain_t *zilc = data;
bcopy(mac, &zilc->zc_eck.zec_cksum.zc_word[2], sizeof (uint64_t));
bcopy(mac + sizeof (uint64_t), &zilc->zc_eck.zec_cksum.zc_word[3],
sizeof (uint64_t));
}
void
zio_crypt_decode_mac_zil(const void *data, uint8_t *mac)
{
/*
* The ZIL MAC is embedded in the block it protects, which will
* not have been byteswapped by the time this function has been called.
* As a result, we don't need to worry about byteswapping the MAC.
*/
const zil_chain_t *zilc = data;
bcopy(&zilc->zc_eck.zec_cksum.zc_word[2], mac, sizeof (uint64_t));
bcopy(&zilc->zc_eck.zec_cksum.zc_word[3], mac + sizeof (uint64_t),
sizeof (uint64_t));
}
/*
* This routine takes a block of dnodes (src_abd) and copies only the bonus
* buffers to the same offsets in the dst buffer. datalen should be the size
* of both the src_abd and the dst buffer (not just the length of the bonus
* buffers).
*/
void
zio_crypt_copy_dnode_bonus(abd_t *src_abd, uint8_t *dst, uint_t datalen)
{
uint_t i, max_dnp = datalen >> DNODE_SHIFT;
uint8_t *src;
dnode_phys_t *dnp, *sdnp, *ddnp;
src = abd_borrow_buf_copy(src_abd, datalen);
sdnp = (dnode_phys_t *)src;
ddnp = (dnode_phys_t *)dst;
for (i = 0; i < max_dnp; i += sdnp[i].dn_extra_slots + 1) {
dnp = &sdnp[i];
if (dnp->dn_type != DMU_OT_NONE &&
DMU_OT_IS_ENCRYPTED(dnp->dn_bonustype) &&
dnp->dn_bonuslen != 0) {
bcopy(DN_BONUS(dnp), DN_BONUS(&ddnp[i]),
DN_MAX_BONUS_LEN(dnp));
}
}
abd_return_buf(src_abd, src, datalen);
}
/*
* This function decides what fields from blk_prop are included in
* the on-disk various MAC algorithms.
*/
static void
zio_crypt_bp_zero_nonportable_blkprop(blkptr_t *bp, uint64_t version)
{
/*
* Version 0 did not properly zero out all non-portable fields
* as it should have done. We maintain this code so that we can
* do read-only imports of pools on this version.
*/
if (version == 0) {
BP_SET_DEDUP(bp, 0);
BP_SET_CHECKSUM(bp, 0);
BP_SET_PSIZE(bp, SPA_MINBLOCKSIZE);
return;
}
ASSERT3U(version, ==, ZIO_CRYPT_KEY_CURRENT_VERSION);
/*
* The hole_birth feature might set these fields even if this bp
* is a hole. We zero them out here to guarantee that raw sends
* will function with or without the feature.
*/
if (BP_IS_HOLE(bp)) {
bp->blk_prop = 0ULL;
return;
}
/*
* At L0 we want to verify these fields to ensure that data blocks
* can not be reinterpreted. For instance, we do not want an attacker
* to trick us into returning raw lz4 compressed data to the user
* by modifying the compression bits. At higher levels, we cannot
* enforce this policy since raw sends do not convey any information
* about indirect blocks, so these values might be different on the
* receive side. Fortunately, this does not open any new attack
* vectors, since any alterations that can be made to a higher level
* bp must still verify the correct order of the layer below it.
*/
if (BP_GET_LEVEL(bp) != 0) {
BP_SET_BYTEORDER(bp, 0);
BP_SET_COMPRESS(bp, 0);
/*
* psize cannot be set to zero or it will trigger
* asserts, but the value doesn't really matter as
* long as it is constant.
*/
BP_SET_PSIZE(bp, SPA_MINBLOCKSIZE);
}
BP_SET_DEDUP(bp, 0);
BP_SET_CHECKSUM(bp, 0);
}
static void
zio_crypt_bp_auth_init(uint64_t version, boolean_t should_bswap, blkptr_t *bp,
blkptr_auth_buf_t *bab, uint_t *bab_len)
{
blkptr_t tmpbp = *bp;
if (should_bswap)
byteswap_uint64_array(&tmpbp, sizeof (blkptr_t));
ASSERT(BP_USES_CRYPT(&tmpbp) || BP_IS_HOLE(&tmpbp));
ASSERT0(BP_IS_EMBEDDED(&tmpbp));
zio_crypt_decode_mac_bp(&tmpbp, bab->bab_mac);
/*
* We always MAC blk_prop in LE to ensure portability. This
* must be done after decoding the mac, since the endianness
* will get zero'd out here.
*/
zio_crypt_bp_zero_nonportable_blkprop(&tmpbp, version);
bab->bab_prop = LE_64(tmpbp.blk_prop);
bab->bab_pad = 0ULL;
/* version 0 did not include the padding */
*bab_len = sizeof (blkptr_auth_buf_t);
if (version == 0)
*bab_len -= sizeof (uint64_t);
}
static int
zio_crypt_bp_do_hmac_updates(crypto_context_t ctx, uint64_t version,
boolean_t should_bswap, blkptr_t *bp)
{
int ret;
uint_t bab_len;
blkptr_auth_buf_t bab;
crypto_data_t cd;
zio_crypt_bp_auth_init(version, should_bswap, bp, &bab, &bab_len);
cd.cd_format = CRYPTO_DATA_RAW;
cd.cd_offset = 0;
cd.cd_length = bab_len;
cd.cd_raw.iov_base = (char *)&bab;
cd.cd_raw.iov_len = cd.cd_length;
ret = crypto_mac_update(ctx, &cd, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
return (0);
error:
return (ret);
}
static void
zio_crypt_bp_do_indrect_checksum_updates(SHA2_CTX *ctx, uint64_t version,
boolean_t should_bswap, blkptr_t *bp)
{
uint_t bab_len;
blkptr_auth_buf_t bab;
zio_crypt_bp_auth_init(version, should_bswap, bp, &bab, &bab_len);
SHA2Update(ctx, &bab, bab_len);
}
static void
zio_crypt_bp_do_aad_updates(uint8_t **aadp, uint_t *aad_len, uint64_t version,
boolean_t should_bswap, blkptr_t *bp)
{
uint_t bab_len;
blkptr_auth_buf_t bab;
zio_crypt_bp_auth_init(version, should_bswap, bp, &bab, &bab_len);
bcopy(&bab, *aadp, bab_len);
*aadp += bab_len;
*aad_len += bab_len;
}
static int
zio_crypt_do_dnode_hmac_updates(crypto_context_t ctx, uint64_t version,
boolean_t should_bswap, dnode_phys_t *dnp)
{
int ret, i;
dnode_phys_t *adnp, tmp_dncore;
size_t dn_core_size = offsetof(dnode_phys_t, dn_blkptr);
boolean_t le_bswap = (should_bswap == ZFS_HOST_BYTEORDER);
crypto_data_t cd;
cd.cd_format = CRYPTO_DATA_RAW;
cd.cd_offset = 0;
/*
* Authenticate the core dnode (masking out non-portable bits).
* We only copy the first 64 bytes we operate on to avoid the overhead
* of copying 512-64 unneeded bytes. The compiler seems to be fine
* with that.
*/
bcopy(dnp, &tmp_dncore, dn_core_size);
adnp = &tmp_dncore;
if (le_bswap) {
adnp->dn_datablkszsec = BSWAP_16(adnp->dn_datablkszsec);
adnp->dn_bonuslen = BSWAP_16(adnp->dn_bonuslen);
adnp->dn_maxblkid = BSWAP_64(adnp->dn_maxblkid);
adnp->dn_used = BSWAP_64(adnp->dn_used);
}
adnp->dn_flags &= DNODE_CRYPT_PORTABLE_FLAGS_MASK;
adnp->dn_used = 0;
cd.cd_length = dn_core_size;
cd.cd_raw.iov_base = (char *)adnp;
cd.cd_raw.iov_len = cd.cd_length;
ret = crypto_mac_update(ctx, &cd, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
for (i = 0; i < dnp->dn_nblkptr; i++) {
ret = zio_crypt_bp_do_hmac_updates(ctx, version,
should_bswap, &dnp->dn_blkptr[i]);
if (ret != 0)
goto error;
}
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
ret = zio_crypt_bp_do_hmac_updates(ctx, version,
should_bswap, DN_SPILL_BLKPTR(dnp));
if (ret != 0)
goto error;
}
return (0);
error:
return (ret);
}
/*
* objset_phys_t blocks introduce a number of exceptions to the normal
* authentication process. objset_phys_t's contain 2 separate HMACS for
* protecting the integrity of their data. The portable_mac protects the
* metadnode. This MAC can be sent with a raw send and protects against
* reordering of data within the metadnode. The local_mac protects the user
* accounting objects which are not sent from one system to another.
*
* In addition, objset blocks are the only blocks that can be modified and
* written to disk without the key loaded under certain circumstances. During
* zil_claim() we need to be able to update the zil_header_t to complete
* claiming log blocks and during raw receives we need to write out the
* portable_mac from the send file. Both of these actions are possible
* because these fields are not protected by either MAC so neither one will
* need to modify the MACs without the key. However, when the modified blocks
* are written out they will be byteswapped into the host machine's native
* endianness which will modify fields protected by the MAC. As a result, MAC
* calculation for objset blocks works slightly differently from other block
* types. Where other block types MAC the data in whatever endianness is
* written to disk, objset blocks always MAC little endian version of their
* values. In the code, should_bswap is the value from BP_SHOULD_BYTESWAP()
* and le_bswap indicates whether a byteswap is needed to get this block
* into little endian format.
*/
int
zio_crypt_do_objset_hmacs(zio_crypt_key_t *key, void *data, uint_t datalen,
boolean_t should_bswap, uint8_t *portable_mac, uint8_t *local_mac)
{
int ret;
crypto_mechanism_t mech;
crypto_context_t ctx;
crypto_data_t cd;
objset_phys_t *osp = data;
uint64_t intval;
boolean_t le_bswap = (should_bswap == ZFS_HOST_BYTEORDER);
uint8_t raw_portable_mac[SHA512_DIGEST_LENGTH];
uint8_t raw_local_mac[SHA512_DIGEST_LENGTH];
/* initialize HMAC mechanism */
mech.cm_type = crypto_mech2id(SUN_CKM_SHA512_HMAC);
mech.cm_param = NULL;
mech.cm_param_len = 0;
cd.cd_format = CRYPTO_DATA_RAW;
cd.cd_offset = 0;
/* calculate the portable MAC from the portable fields and metadnode */
ret = crypto_mac_init(&mech, &key->zk_hmac_key, NULL, &ctx, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
/* add in the os_type */
intval = (le_bswap) ? osp->os_type : BSWAP_64(osp->os_type);
cd.cd_length = sizeof (uint64_t);
cd.cd_raw.iov_base = (char *)&intval;
cd.cd_raw.iov_len = cd.cd_length;
ret = crypto_mac_update(ctx, &cd, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
/* add in the portable os_flags */
intval = osp->os_flags;
if (should_bswap)
intval = BSWAP_64(intval);
intval &= OBJSET_CRYPT_PORTABLE_FLAGS_MASK;
if (!ZFS_HOST_BYTEORDER)
intval = BSWAP_64(intval);
cd.cd_length = sizeof (uint64_t);
cd.cd_raw.iov_base = (char *)&intval;
cd.cd_raw.iov_len = cd.cd_length;
ret = crypto_mac_update(ctx, &cd, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
/* add in fields from the metadnode */
ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
should_bswap, &osp->os_meta_dnode);
if (ret)
goto error;
/* store the final digest in a temporary buffer and copy what we need */
cd.cd_length = SHA512_DIGEST_LENGTH;
cd.cd_raw.iov_base = (char *)raw_portable_mac;
cd.cd_raw.iov_len = cd.cd_length;
ret = crypto_mac_final(ctx, &cd, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
bcopy(raw_portable_mac, portable_mac, ZIO_OBJSET_MAC_LEN);
/*
* The local MAC protects the user, group and project accounting.
* If these objects are not present, the local MAC is zeroed out.
*/
if ((datalen >= OBJSET_PHYS_SIZE_V3 &&
osp->os_userused_dnode.dn_type == DMU_OT_NONE &&
osp->os_groupused_dnode.dn_type == DMU_OT_NONE &&
osp->os_projectused_dnode.dn_type == DMU_OT_NONE) ||
(datalen >= OBJSET_PHYS_SIZE_V2 &&
osp->os_userused_dnode.dn_type == DMU_OT_NONE &&
osp->os_groupused_dnode.dn_type == DMU_OT_NONE) ||
(datalen <= OBJSET_PHYS_SIZE_V1)) {
bzero(local_mac, ZIO_OBJSET_MAC_LEN);
return (0);
}
/* calculate the local MAC from the userused and groupused dnodes */
ret = crypto_mac_init(&mech, &key->zk_hmac_key, NULL, &ctx, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
/* add in the non-portable os_flags */
intval = osp->os_flags;
if (should_bswap)
intval = BSWAP_64(intval);
intval &= ~OBJSET_CRYPT_PORTABLE_FLAGS_MASK;
if (!ZFS_HOST_BYTEORDER)
intval = BSWAP_64(intval);
cd.cd_length = sizeof (uint64_t);
cd.cd_raw.iov_base = (char *)&intval;
cd.cd_raw.iov_len = cd.cd_length;
ret = crypto_mac_update(ctx, &cd, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
/* add in fields from the user accounting dnodes */
if (osp->os_userused_dnode.dn_type != DMU_OT_NONE) {
ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
should_bswap, &osp->os_userused_dnode);
if (ret)
goto error;
}
if (osp->os_groupused_dnode.dn_type != DMU_OT_NONE) {
ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
should_bswap, &osp->os_groupused_dnode);
if (ret)
goto error;
}
if (osp->os_projectused_dnode.dn_type != DMU_OT_NONE &&
datalen >= OBJSET_PHYS_SIZE_V3) {
ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
should_bswap, &osp->os_projectused_dnode);
if (ret)
goto error;
}
/* store the final digest in a temporary buffer and copy what we need */
cd.cd_length = SHA512_DIGEST_LENGTH;
cd.cd_raw.iov_base = (char *)raw_local_mac;
cd.cd_raw.iov_len = cd.cd_length;
ret = crypto_mac_final(ctx, &cd, NULL);
if (ret != CRYPTO_SUCCESS) {
ret = SET_ERROR(EIO);
goto error;
}
bcopy(raw_local_mac, local_mac, ZIO_OBJSET_MAC_LEN);
return (0);
error:
bzero(portable_mac, ZIO_OBJSET_MAC_LEN);
bzero(local_mac, ZIO_OBJSET_MAC_LEN);
return (ret);
}
static void
zio_crypt_destroy_uio(zfs_uio_t *uio)
{
if (uio->uio_iov)
kmem_free(uio->uio_iov, uio->uio_iovcnt * sizeof (iovec_t));
}
/*
* This function parses an uncompressed indirect block and returns a checksum
* of all the portable fields from all of the contained bps. The portable
* fields are the MAC and all of the fields from blk_prop except for the dedup,
* checksum, and psize bits. For an explanation of the purpose of this, see
* the comment block on object set authentication.
*/
static int
zio_crypt_do_indirect_mac_checksum_impl(boolean_t generate, void *buf,
uint_t datalen, uint64_t version, boolean_t byteswap, uint8_t *cksum)
{
blkptr_t *bp;
int i, epb = datalen >> SPA_BLKPTRSHIFT;
SHA2_CTX ctx;
uint8_t digestbuf[SHA512_DIGEST_LENGTH];
/* checksum all of the MACs from the layer below */
SHA2Init(SHA512, &ctx);
for (i = 0, bp = buf; i < epb; i++, bp++) {
zio_crypt_bp_do_indrect_checksum_updates(&ctx, version,
byteswap, bp);
}
SHA2Final(digestbuf, &ctx);
if (generate) {
bcopy(digestbuf, cksum, ZIO_DATA_MAC_LEN);
return (0);
}
if (bcmp(digestbuf, cksum, ZIO_DATA_MAC_LEN) != 0)
return (SET_ERROR(ECKSUM));
return (0);
}
int
zio_crypt_do_indirect_mac_checksum(boolean_t generate, void *buf,
uint_t datalen, boolean_t byteswap, uint8_t *cksum)
{
int ret;
/*
* Unfortunately, callers of this function will not always have
* easy access to the on-disk format version. This info is
* normally found in the DSL Crypto Key, but the checksum-of-MACs
* is expected to be verifiable even when the key isn't loaded.
* Here, instead of doing a ZAP lookup for the version for each
* zio, we simply try both existing formats.
*/
ret = zio_crypt_do_indirect_mac_checksum_impl(generate, buf,
datalen, ZIO_CRYPT_KEY_CURRENT_VERSION, byteswap, cksum);
if (ret == ECKSUM) {
ASSERT(!generate);
ret = zio_crypt_do_indirect_mac_checksum_impl(generate,
buf, datalen, 0, byteswap, cksum);
}
return (ret);
}
int
zio_crypt_do_indirect_mac_checksum_abd(boolean_t generate, abd_t *abd,
uint_t datalen, boolean_t byteswap, uint8_t *cksum)
{
int ret;
void *buf;
buf = abd_borrow_buf_copy(abd, datalen);
ret = zio_crypt_do_indirect_mac_checksum(generate, buf, datalen,
byteswap, cksum);
abd_return_buf(abd, buf, datalen);
return (ret);
}
/*
* Special case handling routine for encrypting / decrypting ZIL blocks.
* We do not check for the older ZIL chain because the encryption feature
* was not available before the newer ZIL chain was introduced. The goal
* here is to encrypt everything except the blkptr_t of a lr_write_t and
* the zil_chain_t header. Everything that is not encrypted is authenticated.
*/
static int
zio_crypt_init_uios_zil(boolean_t encrypt, uint8_t *plainbuf,
uint8_t *cipherbuf, uint_t datalen, boolean_t byteswap, zfs_uio_t *puio,
zfs_uio_t *cuio, uint_t *enc_len, uint8_t **authbuf, uint_t *auth_len,
boolean_t *no_crypt)
{
int ret;
uint64_t txtype, lr_len;
uint_t nr_src, nr_dst, crypt_len;
uint_t aad_len = 0, nr_iovecs = 0, total_len = 0;
iovec_t *src_iovecs = NULL, *dst_iovecs = NULL;
uint8_t *src, *dst, *slrp, *dlrp, *blkend, *aadp;
zil_chain_t *zilc;
lr_t *lr;
uint8_t *aadbuf = zio_buf_alloc(datalen);
/* cipherbuf always needs an extra iovec for the MAC */
if (encrypt) {
src = plainbuf;
dst = cipherbuf;
nr_src = 0;
nr_dst = 1;
} else {
src = cipherbuf;
dst = plainbuf;
nr_src = 1;
nr_dst = 0;
}
bzero(dst, datalen);
/* find the start and end record of the log block */
zilc = (zil_chain_t *)src;
slrp = src + sizeof (zil_chain_t);
aadp = aadbuf;
blkend = src + ((byteswap) ? BSWAP_64(zilc->zc_nused) : zilc->zc_nused);
/* calculate the number of encrypted iovecs we will need */
for (; slrp < blkend; slrp += lr_len) {
lr = (lr_t *)slrp;
if (!byteswap) {
txtype = lr->lrc_txtype;
lr_len = lr->lrc_reclen;
} else {
txtype = BSWAP_64(lr->lrc_txtype);
lr_len = BSWAP_64(lr->lrc_reclen);
}
nr_iovecs++;
if (txtype == TX_WRITE && lr_len != sizeof (lr_write_t))
nr_iovecs++;
}
nr_src += nr_iovecs;
nr_dst += nr_iovecs;
/* allocate the iovec arrays */
if (nr_src != 0) {
src_iovecs = kmem_alloc(nr_src * sizeof (iovec_t), KM_SLEEP);
if (src_iovecs == NULL) {
ret = SET_ERROR(ENOMEM);
goto error;
}
}
if (nr_dst != 0) {
dst_iovecs = kmem_alloc(nr_dst * sizeof (iovec_t), KM_SLEEP);
if (dst_iovecs == NULL) {
ret = SET_ERROR(ENOMEM);
goto error;
}
}
/*
* Copy the plain zil header over and authenticate everything except
* the checksum that will store our MAC. If we are writing the data
* the embedded checksum will not have been calculated yet, so we don't
* authenticate that.
*/
bcopy(src, dst, sizeof (zil_chain_t));
bcopy(src, aadp, sizeof (zil_chain_t) - sizeof (zio_eck_t));
aadp += sizeof (zil_chain_t) - sizeof (zio_eck_t);
aad_len += sizeof (zil_chain_t) - sizeof (zio_eck_t);
/* loop over records again, filling in iovecs */
nr_iovecs = 0;
slrp = src + sizeof (zil_chain_t);
dlrp = dst + sizeof (zil_chain_t);
for (; slrp < blkend; slrp += lr_len, dlrp += lr_len) {
lr = (lr_t *)slrp;
if (!byteswap) {
txtype = lr->lrc_txtype;
lr_len = lr->lrc_reclen;
} else {
txtype = BSWAP_64(lr->lrc_txtype);
lr_len = BSWAP_64(lr->lrc_reclen);
}
/* copy the common lr_t */
bcopy(slrp, dlrp, sizeof (lr_t));
bcopy(slrp, aadp, sizeof (lr_t));
aadp += sizeof (lr_t);
aad_len += sizeof (lr_t);
ASSERT3P(src_iovecs, !=, NULL);
ASSERT3P(dst_iovecs, !=, NULL);
/*
* If this is a TX_WRITE record we want to encrypt everything
* except the bp if exists. If the bp does exist we want to
* authenticate it.
*/
if (txtype == TX_WRITE) {
crypt_len = sizeof (lr_write_t) -
sizeof (lr_t) - sizeof (blkptr_t);
src_iovecs[nr_iovecs].iov_base = slrp + sizeof (lr_t);
src_iovecs[nr_iovecs].iov_len = crypt_len;
dst_iovecs[nr_iovecs].iov_base = dlrp + sizeof (lr_t);
dst_iovecs[nr_iovecs].iov_len = crypt_len;
/* copy the bp now since it will not be encrypted */
bcopy(slrp + sizeof (lr_write_t) - sizeof (blkptr_t),
dlrp + sizeof (lr_write_t) - sizeof (blkptr_t),
sizeof (blkptr_t));
bcopy(slrp + sizeof (lr_write_t) - sizeof (blkptr_t),
aadp, sizeof (blkptr_t));
aadp += sizeof (blkptr_t);
aad_len += sizeof (blkptr_t);
nr_iovecs++;
total_len += crypt_len;
if (lr_len != sizeof (lr_write_t)) {
crypt_len = lr_len - sizeof (lr_write_t);
src_iovecs[nr_iovecs].iov_base =
slrp + sizeof (lr_write_t);
src_iovecs[nr_iovecs].iov_len = crypt_len;
dst_iovecs[nr_iovecs].iov_base =
dlrp + sizeof (lr_write_t);
dst_iovecs[nr_iovecs].iov_len = crypt_len;
nr_iovecs++;
total_len += crypt_len;
}
} else {
crypt_len = lr_len - sizeof (lr_t);
src_iovecs[nr_iovecs].iov_base = slrp + sizeof (lr_t);
src_iovecs[nr_iovecs].iov_len = crypt_len;
dst_iovecs[nr_iovecs].iov_base = dlrp + sizeof (lr_t);
dst_iovecs[nr_iovecs].iov_len = crypt_len;
nr_iovecs++;
total_len += crypt_len;
}
}
*no_crypt = (nr_iovecs == 0);
*enc_len = total_len;
*authbuf = aadbuf;
*auth_len = aad_len;
if (encrypt) {
puio->uio_iov = src_iovecs;
puio->uio_iovcnt = nr_src;
cuio->uio_iov = dst_iovecs;
cuio->uio_iovcnt = nr_dst;
} else {
puio->uio_iov = dst_iovecs;
puio->uio_iovcnt = nr_dst;
cuio->uio_iov = src_iovecs;
cuio->uio_iovcnt = nr_src;
}
return (0);
error:
zio_buf_free(aadbuf, datalen);
if (src_iovecs != NULL)
kmem_free(src_iovecs, nr_src * sizeof (iovec_t));
if (dst_iovecs != NULL)
kmem_free(dst_iovecs, nr_dst * sizeof (iovec_t));
*enc_len = 0;
*authbuf = NULL;
*auth_len = 0;
*no_crypt = B_FALSE;
puio->uio_iov = NULL;
puio->uio_iovcnt = 0;
cuio->uio_iov = NULL;
cuio->uio_iovcnt = 0;
return (ret);
}
/*
* Special case handling routine for encrypting / decrypting dnode blocks.
*/
static int
zio_crypt_init_uios_dnode(boolean_t encrypt, uint64_t version,
uint8_t *plainbuf, uint8_t *cipherbuf, uint_t datalen, boolean_t byteswap,
zfs_uio_t *puio, zfs_uio_t *cuio, uint_t *enc_len, uint8_t **authbuf,
uint_t *auth_len, boolean_t *no_crypt)
{
int ret;
uint_t nr_src, nr_dst, crypt_len;
uint_t aad_len = 0, nr_iovecs = 0, total_len = 0;
uint_t i, j, max_dnp = datalen >> DNODE_SHIFT;
iovec_t *src_iovecs = NULL, *dst_iovecs = NULL;
uint8_t *src, *dst, *aadp;
dnode_phys_t *dnp, *adnp, *sdnp, *ddnp;
uint8_t *aadbuf = zio_buf_alloc(datalen);
if (encrypt) {
src = plainbuf;
dst = cipherbuf;
nr_src = 0;
nr_dst = 1;
} else {
src = cipherbuf;
dst = plainbuf;
nr_src = 1;
nr_dst = 0;
}
sdnp = (dnode_phys_t *)src;
ddnp = (dnode_phys_t *)dst;
aadp = aadbuf;
/*
* Count the number of iovecs we will need to do the encryption by
* counting the number of bonus buffers that need to be encrypted.
*/
for (i = 0; i < max_dnp; i += sdnp[i].dn_extra_slots + 1) {
/*
* This block may still be byteswapped. However, all of the
* values we use are either uint8_t's (for which byteswapping
* is a noop) or a * != 0 check, which will work regardless
* of whether or not we byteswap.
*/
if (sdnp[i].dn_type != DMU_OT_NONE &&
DMU_OT_IS_ENCRYPTED(sdnp[i].dn_bonustype) &&
sdnp[i].dn_bonuslen != 0) {
nr_iovecs++;
}
}
nr_src += nr_iovecs;
nr_dst += nr_iovecs;
if (nr_src != 0) {
src_iovecs = kmem_alloc(nr_src * sizeof (iovec_t), KM_SLEEP);
if (src_iovecs == NULL) {
ret = SET_ERROR(ENOMEM);
goto error;
}
}
if (nr_dst != 0) {
dst_iovecs = kmem_alloc(nr_dst * sizeof (iovec_t), KM_SLEEP);
if (dst_iovecs == NULL) {
ret = SET_ERROR(ENOMEM);
goto error;
}
}
nr_iovecs = 0;
/*
* Iterate through the dnodes again, this time filling in the uios
* we allocated earlier. We also concatenate any data we want to
* authenticate onto aadbuf.
*/
for (i = 0; i < max_dnp; i += sdnp[i].dn_extra_slots + 1) {
dnp = &sdnp[i];
/* copy over the core fields and blkptrs (kept as plaintext) */
bcopy(dnp, &ddnp[i], (uint8_t *)DN_BONUS(dnp) - (uint8_t *)dnp);
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
bcopy(DN_SPILL_BLKPTR(dnp), DN_SPILL_BLKPTR(&ddnp[i]),
sizeof (blkptr_t));
}
/*
* Handle authenticated data. We authenticate everything in
* the dnode that can be brought over when we do a raw send.
* This includes all of the core fields as well as the MACs
* stored in the bp checksums and all of the portable bits
* from blk_prop. We include the dnode padding here in case it
* ever gets used in the future. Some dn_flags and dn_used are
* not portable so we mask those out values out of the
* authenticated data.
*/
crypt_len = offsetof(dnode_phys_t, dn_blkptr);
bcopy(dnp, aadp, crypt_len);
adnp = (dnode_phys_t *)aadp;
adnp->dn_flags &= DNODE_CRYPT_PORTABLE_FLAGS_MASK;
adnp->dn_used = 0;
aadp += crypt_len;
aad_len += crypt_len;
for (j = 0; j < dnp->dn_nblkptr; j++) {
zio_crypt_bp_do_aad_updates(&aadp, &aad_len,
version, byteswap, &dnp->dn_blkptr[j]);
}
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
zio_crypt_bp_do_aad_updates(&aadp, &aad_len,
version, byteswap, DN_SPILL_BLKPTR(dnp));
}
/*
* If this bonus buffer needs to be encrypted, we prepare an
* iovec_t. The encryption / decryption functions will fill
* this in for us with the encrypted or decrypted data.
* Otherwise we add the bonus buffer to the authenticated
* data buffer and copy it over to the destination. The
* encrypted iovec extends to DN_MAX_BONUS_LEN(dnp) so that
* we can guarantee alignment with the AES block size
* (128 bits).
*/
crypt_len = DN_MAX_BONUS_LEN(dnp);
if (dnp->dn_type != DMU_OT_NONE &&
DMU_OT_IS_ENCRYPTED(dnp->dn_bonustype) &&
dnp->dn_bonuslen != 0) {
ASSERT3U(nr_iovecs, <, nr_src);
ASSERT3U(nr_iovecs, <, nr_dst);
ASSERT3P(src_iovecs, !=, NULL);
ASSERT3P(dst_iovecs, !=, NULL);
src_iovecs[nr_iovecs].iov_base = DN_BONUS(dnp);
src_iovecs[nr_iovecs].iov_len = crypt_len;
dst_iovecs[nr_iovecs].iov_base = DN_BONUS(&ddnp[i]);
dst_iovecs[nr_iovecs].iov_len = crypt_len;
nr_iovecs++;
total_len += crypt_len;
} else {
bcopy(DN_BONUS(dnp), DN_BONUS(&ddnp[i]), crypt_len);
bcopy(DN_BONUS(dnp), aadp, crypt_len);
aadp += crypt_len;
aad_len += crypt_len;
}
}
*no_crypt = (nr_iovecs == 0);
*enc_len = total_len;
*authbuf = aadbuf;
*auth_len = aad_len;
if (encrypt) {
puio->uio_iov = src_iovecs;
puio->uio_iovcnt = nr_src;
cuio->uio_iov = dst_iovecs;
cuio->uio_iovcnt = nr_dst;
} else {
puio->uio_iov = dst_iovecs;
puio->uio_iovcnt = nr_dst;
cuio->uio_iov = src_iovecs;
cuio->uio_iovcnt = nr_src;
}
return (0);
error:
zio_buf_free(aadbuf, datalen);
if (src_iovecs != NULL)
kmem_free(src_iovecs, nr_src * sizeof (iovec_t));
if (dst_iovecs != NULL)
kmem_free(dst_iovecs, nr_dst * sizeof (iovec_t));
*enc_len = 0;
*authbuf = NULL;
*auth_len = 0;
*no_crypt = B_FALSE;
puio->uio_iov = NULL;
puio->uio_iovcnt = 0;
cuio->uio_iov = NULL;
cuio->uio_iovcnt = 0;
return (ret);
}
static int
zio_crypt_init_uios_normal(boolean_t encrypt, uint8_t *plainbuf,
uint8_t *cipherbuf, uint_t datalen, zfs_uio_t *puio, zfs_uio_t *cuio,
uint_t *enc_len)
{
(void) encrypt;
int ret;
uint_t nr_plain = 1, nr_cipher = 2;
iovec_t *plain_iovecs = NULL, *cipher_iovecs = NULL;
/* allocate the iovecs for the plain and cipher data */
plain_iovecs = kmem_alloc(nr_plain * sizeof (iovec_t),
KM_SLEEP);
if (!plain_iovecs) {
ret = SET_ERROR(ENOMEM);
goto error;
}
cipher_iovecs = kmem_alloc(nr_cipher * sizeof (iovec_t),
KM_SLEEP);
if (!cipher_iovecs) {
ret = SET_ERROR(ENOMEM);
goto error;
}
plain_iovecs[0].iov_base = plainbuf;
plain_iovecs[0].iov_len = datalen;
cipher_iovecs[0].iov_base = cipherbuf;
cipher_iovecs[0].iov_len = datalen;
*enc_len = datalen;
puio->uio_iov = plain_iovecs;
puio->uio_iovcnt = nr_plain;
cuio->uio_iov = cipher_iovecs;
cuio->uio_iovcnt = nr_cipher;
return (0);
error:
if (plain_iovecs != NULL)
kmem_free(plain_iovecs, nr_plain * sizeof (iovec_t));
if (cipher_iovecs != NULL)
kmem_free(cipher_iovecs, nr_cipher * sizeof (iovec_t));
*enc_len = 0;
puio->uio_iov = NULL;
puio->uio_iovcnt = 0;
cuio->uio_iov = NULL;
cuio->uio_iovcnt = 0;
return (ret);
}
/*
* This function builds up the plaintext (puio) and ciphertext (cuio) uios so
* that they can be used for encryption and decryption by zio_do_crypt_uio().
* Most blocks will use zio_crypt_init_uios_normal(), with ZIL and dnode blocks
* requiring special handling to parse out pieces that are to be encrypted. The
* authbuf is used by these special cases to store additional authenticated
* data (AAD) for the encryption modes.
*/
static int
zio_crypt_init_uios(boolean_t encrypt, uint64_t version, dmu_object_type_t ot,
uint8_t *plainbuf, uint8_t *cipherbuf, uint_t datalen, boolean_t byteswap,
uint8_t *mac, zfs_uio_t *puio, zfs_uio_t *cuio, uint_t *enc_len,
uint8_t **authbuf, uint_t *auth_len, boolean_t *no_crypt)
{
int ret;
iovec_t *mac_iov;
ASSERT(DMU_OT_IS_ENCRYPTED(ot) || ot == DMU_OT_NONE);
/* route to handler */
switch (ot) {
case DMU_OT_INTENT_LOG:
ret = zio_crypt_init_uios_zil(encrypt, plainbuf, cipherbuf,
datalen, byteswap, puio, cuio, enc_len, authbuf, auth_len,
no_crypt);
break;
case DMU_OT_DNODE:
ret = zio_crypt_init_uios_dnode(encrypt, version, plainbuf,
cipherbuf, datalen, byteswap, puio, cuio, enc_len, authbuf,
auth_len, no_crypt);
break;
default:
ret = zio_crypt_init_uios_normal(encrypt, plainbuf, cipherbuf,
datalen, puio, cuio, enc_len);
*authbuf = NULL;
*auth_len = 0;
*no_crypt = B_FALSE;
break;
}
if (ret != 0)
goto error;
/* populate the uios */
puio->uio_segflg = UIO_SYSSPACE;
cuio->uio_segflg = UIO_SYSSPACE;
mac_iov = ((iovec_t *)&cuio->uio_iov[cuio->uio_iovcnt - 1]);
mac_iov->iov_base = mac;
mac_iov->iov_len = ZIO_DATA_MAC_LEN;
return (0);
error:
return (ret);
}
/*
* Primary encryption / decryption entrypoint for zio data.
*/
int
zio_do_crypt_data(boolean_t encrypt, zio_crypt_key_t *key,
dmu_object_type_t ot, boolean_t byteswap, uint8_t *salt, uint8_t *iv,
uint8_t *mac, uint_t datalen, uint8_t *plainbuf, uint8_t *cipherbuf,
boolean_t *no_crypt)
{
int ret;
boolean_t locked = B_FALSE;
uint64_t crypt = key->zk_crypt;
uint_t keydata_len = zio_crypt_table[crypt].ci_keylen;
uint_t enc_len, auth_len;
zfs_uio_t puio, cuio;
uint8_t enc_keydata[MASTER_KEY_MAX_LEN];
crypto_key_t tmp_ckey, *ckey = NULL;
crypto_ctx_template_t tmpl;
uint8_t *authbuf = NULL;
/*
* If the needed key is the current one, just use it. Otherwise we
* need to generate a temporary one from the given salt + master key.
* If we are encrypting, we must return a copy of the current salt
* so that it can be stored in the blkptr_t.
*/
rw_enter(&key->zk_salt_lock, RW_READER);
locked = B_TRUE;
if (bcmp(salt, key->zk_salt, ZIO_DATA_SALT_LEN) == 0) {
ckey = &key->zk_current_key;
tmpl = key->zk_current_tmpl;
} else {
rw_exit(&key->zk_salt_lock);
locked = B_FALSE;
ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
salt, ZIO_DATA_SALT_LEN, enc_keydata, keydata_len);
if (ret != 0)
goto error;
tmp_ckey.ck_format = CRYPTO_KEY_RAW;
tmp_ckey.ck_data = enc_keydata;
tmp_ckey.ck_length = CRYPTO_BYTES2BITS(keydata_len);
ckey = &tmp_ckey;
tmpl = NULL;
}
/*
* Attempt to use QAT acceleration if we can. We currently don't
* do this for metadnode and ZIL blocks, since they have a much
* more involved buffer layout and the qat_crypt() function only
* works in-place.
*/
if (qat_crypt_use_accel(datalen) &&
ot != DMU_OT_INTENT_LOG && ot != DMU_OT_DNODE) {
uint8_t *srcbuf, *dstbuf;
if (encrypt) {
srcbuf = plainbuf;
dstbuf = cipherbuf;
} else {
srcbuf = cipherbuf;
dstbuf = plainbuf;
}
ret = qat_crypt((encrypt) ? QAT_ENCRYPT : QAT_DECRYPT, srcbuf,
dstbuf, NULL, 0, iv, mac, ckey, key->zk_crypt, datalen);
if (ret == CPA_STATUS_SUCCESS) {
if (locked) {
rw_exit(&key->zk_salt_lock);
locked = B_FALSE;
}
return (0);
}
/* If the hardware implementation fails fall back to software */
}
bzero(&puio, sizeof (zfs_uio_t));
bzero(&cuio, sizeof (zfs_uio_t));
/* create uios for encryption */
ret = zio_crypt_init_uios(encrypt, key->zk_version, ot, plainbuf,
cipherbuf, datalen, byteswap, mac, &puio, &cuio, &enc_len,
&authbuf, &auth_len, no_crypt);
if (ret != 0)
goto error;
/* perform the encryption / decryption in software */
ret = zio_do_crypt_uio(encrypt, key->zk_crypt, ckey, tmpl, iv, enc_len,
&puio, &cuio, authbuf, auth_len);
if (ret != 0)
goto error;
if (locked) {
rw_exit(&key->zk_salt_lock);
locked = B_FALSE;
}
if (authbuf != NULL)
zio_buf_free(authbuf, datalen);
if (ckey == &tmp_ckey)
bzero(enc_keydata, keydata_len);
zio_crypt_destroy_uio(&puio);
zio_crypt_destroy_uio(&cuio);
return (0);
error:
if (locked)
rw_exit(&key->zk_salt_lock);
if (authbuf != NULL)
zio_buf_free(authbuf, datalen);
if (ckey == &tmp_ckey)
bzero(enc_keydata, keydata_len);
zio_crypt_destroy_uio(&puio);
zio_crypt_destroy_uio(&cuio);
return (ret);
}
/*
* Simple wrapper around zio_do_crypt_data() to work with abd's instead of
* linear buffers.
*/
int
zio_do_crypt_abd(boolean_t encrypt, zio_crypt_key_t *key, dmu_object_type_t ot,
boolean_t byteswap, uint8_t *salt, uint8_t *iv, uint8_t *mac,
uint_t datalen, abd_t *pabd, abd_t *cabd, boolean_t *no_crypt)
{
int ret;
void *ptmp, *ctmp;
if (encrypt) {
ptmp = abd_borrow_buf_copy(pabd, datalen);
ctmp = abd_borrow_buf(cabd, datalen);
} else {
ptmp = abd_borrow_buf(pabd, datalen);
ctmp = abd_borrow_buf_copy(cabd, datalen);
}
ret = zio_do_crypt_data(encrypt, key, ot, byteswap, salt, iv, mac,
datalen, ptmp, ctmp, no_crypt);
if (ret != 0)
goto error;
if (encrypt) {
abd_return_buf(pabd, ptmp, datalen);
abd_return_buf_copy(cabd, ctmp, datalen);
} else {
abd_return_buf_copy(pabd, ptmp, datalen);
abd_return_buf(cabd, ctmp, datalen);
}
return (0);
error:
if (encrypt) {
abd_return_buf(pabd, ptmp, datalen);
abd_return_buf_copy(cabd, ctmp, datalen);
} else {
abd_return_buf_copy(pabd, ptmp, datalen);
abd_return_buf(cabd, ctmp, datalen);
}
return (ret);
}
#if defined(_KERNEL)
/* BEGIN CSTYLED */
module_param(zfs_key_max_salt_uses, ulong, 0644);
MODULE_PARM_DESC(zfs_key_max_salt_uses, "Max number of times a salt value "
"can be used for generating encryption keys before it is rotated");
/* END CSTYLED */
#endif
diff --git a/module/os/linux/zfs/zpl_ctldir.c b/module/os/linux/zfs/zpl_ctldir.c
index 9b526afd0002..a640930a02e1 100644
--- a/module/os/linux/zfs/zpl_ctldir.c
+++ b/module/os/linux/zfs/zpl_ctldir.c
@@ -1,597 +1,597 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (C) 2011 Lawrence Livermore National Security, LLC.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* LLNL-CODE-403049.
* Rewritten for Linux by:
* Rohan Puri <rohan.puri15@gmail.com>
* Brian Behlendorf <behlendorf1@llnl.gov>
*/
#include <sys/zfs_znode.h>
#include <sys/zfs_vfsops.h>
#include <sys/zfs_vnops.h>
#include <sys/zfs_ctldir.h>
#include <sys/zpl.h>
/*
* Common open routine. Disallow any write access.
*/
/* ARGSUSED */
static int
zpl_common_open(struct inode *ip, struct file *filp)
{
if (filp->f_mode & FMODE_WRITE)
return (-EACCES);
return (generic_file_open(ip, filp));
}
/*
* Get root directory contents.
*/
static int
zpl_root_iterate(struct file *filp, zpl_dir_context_t *ctx)
{
zfsvfs_t *zfsvfs = ITOZSB(file_inode(filp));
int error = 0;
ZPL_ENTER(zfsvfs);
if (!zpl_dir_emit_dots(filp, ctx))
goto out;
if (ctx->pos == 2) {
if (!zpl_dir_emit(ctx, ZFS_SNAPDIR_NAME,
strlen(ZFS_SNAPDIR_NAME), ZFSCTL_INO_SNAPDIR, DT_DIR))
goto out;
ctx->pos++;
}
if (ctx->pos == 3) {
if (!zpl_dir_emit(ctx, ZFS_SHAREDIR_NAME,
strlen(ZFS_SHAREDIR_NAME), ZFSCTL_INO_SHARES, DT_DIR))
goto out;
ctx->pos++;
}
out:
ZPL_EXIT(zfsvfs);
return (error);
}
#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
static int
zpl_root_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
zpl_dir_context_t ctx =
ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos);
int error;
error = zpl_root_iterate(filp, &ctx);
filp->f_pos = ctx.pos;
return (error);
}
#endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */
/*
* Get root directory attributes.
*/
/* ARGSUSED */
static int
#ifdef HAVE_USERNS_IOPS_GETATTR
zpl_root_getattr_impl(struct user_namespace *user_ns,
const struct path *path, struct kstat *stat, u32 request_mask,
unsigned int query_flags)
#else
zpl_root_getattr_impl(const struct path *path, struct kstat *stat,
u32 request_mask, unsigned int query_flags)
#endif
{
struct inode *ip = path->dentry->d_inode;
#if defined(HAVE_GENERIC_FILLATTR_USERNS) && defined(HAVE_USERNS_IOPS_GETATTR)
generic_fillattr(user_ns, ip, stat);
#else
generic_fillattr(ip, stat);
#endif
stat->atime = current_time(ip);
return (0);
}
ZPL_GETATTR_WRAPPER(zpl_root_getattr);
static struct dentry *
zpl_root_lookup(struct inode *dip, struct dentry *dentry, unsigned int flags)
{
cred_t *cr = CRED();
struct inode *ip;
int error;
crhold(cr);
error = -zfsctl_root_lookup(dip, dname(dentry), &ip, 0, cr, NULL, NULL);
ASSERT3S(error, <=, 0);
crfree(cr);
if (error) {
if (error == -ENOENT)
return (d_splice_alias(NULL, dentry));
else
return (ERR_PTR(error));
}
return (d_splice_alias(ip, dentry));
}
/*
* The '.zfs' control directory file and inode operations.
*/
const struct file_operations zpl_fops_root = {
.open = zpl_common_open,
.llseek = generic_file_llseek,
.read = generic_read_dir,
#ifdef HAVE_VFS_ITERATE_SHARED
.iterate_shared = zpl_root_iterate,
#elif defined(HAVE_VFS_ITERATE)
.iterate = zpl_root_iterate,
#else
.readdir = zpl_root_readdir,
#endif
};
const struct inode_operations zpl_ops_root = {
.lookup = zpl_root_lookup,
.getattr = zpl_root_getattr,
};
static struct vfsmount *
zpl_snapdir_automount(struct path *path)
{
int error;
error = -zfsctl_snapshot_mount(path, 0);
if (error)
return (ERR_PTR(error));
/*
* Rather than returning the new vfsmount for the snapshot we must
* return NULL to indicate a mount collision. This is done because
* the user space mount calls do_add_mount() which adds the vfsmount
* to the name space. If we returned the new mount here it would be
* added again to the vfsmount list resulting in list corruption.
*/
return (NULL);
}
/*
* Negative dentries must always be revalidated so newly created snapshots
* can be detected and automounted. Normal dentries should be kept because
* as of the 3.18 kernel revaliding the mountpoint dentry will result in
* the snapshot being immediately unmounted.
*/
static int
#ifdef HAVE_D_REVALIDATE_NAMEIDATA
zpl_snapdir_revalidate(struct dentry *dentry, struct nameidata *i)
#else
zpl_snapdir_revalidate(struct dentry *dentry, unsigned int flags)
#endif
{
return (!!dentry->d_inode);
}
-dentry_operations_t zpl_dops_snapdirs = {
+static const dentry_operations_t zpl_dops_snapdirs = {
/*
* Auto mounting of snapshots is only supported for 2.6.37 and
* newer kernels. Prior to this kernel the ops->follow_link()
* callback was used as a hack to trigger the mount. The
* resulting vfsmount was then explicitly grafted in to the
* name space. While it might be possible to add compatibility
* code to accomplish this it would require considerable care.
*/
.d_automount = zpl_snapdir_automount,
.d_revalidate = zpl_snapdir_revalidate,
};
static struct dentry *
zpl_snapdir_lookup(struct inode *dip, struct dentry *dentry,
unsigned int flags)
{
fstrans_cookie_t cookie;
cred_t *cr = CRED();
struct inode *ip = NULL;
int error;
crhold(cr);
cookie = spl_fstrans_mark();
error = -zfsctl_snapdir_lookup(dip, dname(dentry), &ip,
0, cr, NULL, NULL);
ASSERT3S(error, <=, 0);
spl_fstrans_unmark(cookie);
crfree(cr);
if (error && error != -ENOENT)
return (ERR_PTR(error));
ASSERT(error == 0 || ip == NULL);
d_clear_d_op(dentry);
d_set_d_op(dentry, &zpl_dops_snapdirs);
dentry->d_flags |= DCACHE_NEED_AUTOMOUNT;
return (d_splice_alias(ip, dentry));
}
static int
zpl_snapdir_iterate(struct file *filp, zpl_dir_context_t *ctx)
{
zfsvfs_t *zfsvfs = ITOZSB(file_inode(filp));
fstrans_cookie_t cookie;
char snapname[MAXNAMELEN];
boolean_t case_conflict;
uint64_t id, pos;
int error = 0;
ZPL_ENTER(zfsvfs);
cookie = spl_fstrans_mark();
if (!zpl_dir_emit_dots(filp, ctx))
goto out;
/* Start the position at 0 if it already emitted . and .. */
pos = (ctx->pos == 2 ? 0 : ctx->pos);
while (error == 0) {
dsl_pool_config_enter(dmu_objset_pool(zfsvfs->z_os), FTAG);
error = -dmu_snapshot_list_next(zfsvfs->z_os, MAXNAMELEN,
snapname, &id, &pos, &case_conflict);
dsl_pool_config_exit(dmu_objset_pool(zfsvfs->z_os), FTAG);
if (error)
goto out;
if (!zpl_dir_emit(ctx, snapname, strlen(snapname),
ZFSCTL_INO_SHARES - id, DT_DIR))
goto out;
ctx->pos = pos;
}
out:
spl_fstrans_unmark(cookie);
ZPL_EXIT(zfsvfs);
if (error == -ENOENT)
return (0);
return (error);
}
#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
static int
zpl_snapdir_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
zpl_dir_context_t ctx =
ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos);
int error;
error = zpl_snapdir_iterate(filp, &ctx);
filp->f_pos = ctx.pos;
return (error);
}
#endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */
static int
#ifdef HAVE_IOPS_RENAME_USERNS
zpl_snapdir_rename2(struct user_namespace *user_ns, struct inode *sdip,
struct dentry *sdentry, struct inode *tdip, struct dentry *tdentry,
unsigned int flags)
#else
zpl_snapdir_rename2(struct inode *sdip, struct dentry *sdentry,
struct inode *tdip, struct dentry *tdentry, unsigned int flags)
#endif
{
cred_t *cr = CRED();
int error;
/* We probably don't want to support renameat2(2) in ctldir */
if (flags)
return (-EINVAL);
crhold(cr);
error = -zfsctl_snapdir_rename(sdip, dname(sdentry),
tdip, dname(tdentry), cr, 0);
ASSERT3S(error, <=, 0);
crfree(cr);
return (error);
}
#if !defined(HAVE_RENAME_WANTS_FLAGS) && !defined(HAVE_IOPS_RENAME_USERNS)
static int
zpl_snapdir_rename(struct inode *sdip, struct dentry *sdentry,
struct inode *tdip, struct dentry *tdentry)
{
return (zpl_snapdir_rename2(sdip, sdentry, tdip, tdentry, 0));
}
#endif
static int
zpl_snapdir_rmdir(struct inode *dip, struct dentry *dentry)
{
cred_t *cr = CRED();
int error;
crhold(cr);
error = -zfsctl_snapdir_remove(dip, dname(dentry), cr, 0);
ASSERT3S(error, <=, 0);
crfree(cr);
return (error);
}
static int
#ifdef HAVE_IOPS_MKDIR_USERNS
zpl_snapdir_mkdir(struct user_namespace *user_ns, struct inode *dip,
struct dentry *dentry, umode_t mode)
#else
zpl_snapdir_mkdir(struct inode *dip, struct dentry *dentry, umode_t mode)
#endif
{
cred_t *cr = CRED();
vattr_t *vap;
struct inode *ip;
int error;
crhold(cr);
vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP);
zpl_vap_init(vap, dip, mode | S_IFDIR, cr);
error = -zfsctl_snapdir_mkdir(dip, dname(dentry), vap, &ip, cr, 0);
if (error == 0) {
d_clear_d_op(dentry);
d_set_d_op(dentry, &zpl_dops_snapdirs);
d_instantiate(dentry, ip);
}
kmem_free(vap, sizeof (vattr_t));
ASSERT3S(error, <=, 0);
crfree(cr);
return (error);
}
/*
* Get snapshot directory attributes.
*/
/* ARGSUSED */
static int
#ifdef HAVE_USERNS_IOPS_GETATTR
zpl_snapdir_getattr_impl(struct user_namespace *user_ns,
const struct path *path, struct kstat *stat, u32 request_mask,
unsigned int query_flags)
#else
zpl_snapdir_getattr_impl(const struct path *path, struct kstat *stat,
u32 request_mask, unsigned int query_flags)
#endif
{
struct inode *ip = path->dentry->d_inode;
zfsvfs_t *zfsvfs = ITOZSB(ip);
ZPL_ENTER(zfsvfs);
#if defined(HAVE_GENERIC_FILLATTR_USERNS) && defined(HAVE_USERNS_IOPS_GETATTR)
generic_fillattr(user_ns, ip, stat);
#else
generic_fillattr(ip, stat);
#endif
stat->nlink = stat->size = 2;
stat->ctime = stat->mtime = dmu_objset_snap_cmtime(zfsvfs->z_os);
stat->atime = current_time(ip);
ZPL_EXIT(zfsvfs);
return (0);
}
ZPL_GETATTR_WRAPPER(zpl_snapdir_getattr);
/*
* The '.zfs/snapshot' directory file operations. These mainly control
* generating the list of available snapshots when doing an 'ls' in the
* directory. See zpl_snapdir_readdir().
*/
const struct file_operations zpl_fops_snapdir = {
.open = zpl_common_open,
.llseek = generic_file_llseek,
.read = generic_read_dir,
#ifdef HAVE_VFS_ITERATE_SHARED
.iterate_shared = zpl_snapdir_iterate,
#elif defined(HAVE_VFS_ITERATE)
.iterate = zpl_snapdir_iterate,
#else
.readdir = zpl_snapdir_readdir,
#endif
};
/*
* The '.zfs/snapshot' directory inode operations. These mainly control
* creating an inode for a snapshot directory and initializing the needed
* infrastructure to automount the snapshot. See zpl_snapdir_lookup().
*/
const struct inode_operations zpl_ops_snapdir = {
.lookup = zpl_snapdir_lookup,
.getattr = zpl_snapdir_getattr,
#if defined(HAVE_RENAME_WANTS_FLAGS) || defined(HAVE_IOPS_RENAME_USERNS)
.rename = zpl_snapdir_rename2,
#else
.rename = zpl_snapdir_rename,
#endif
.rmdir = zpl_snapdir_rmdir,
.mkdir = zpl_snapdir_mkdir,
};
static struct dentry *
zpl_shares_lookup(struct inode *dip, struct dentry *dentry,
unsigned int flags)
{
fstrans_cookie_t cookie;
cred_t *cr = CRED();
struct inode *ip = NULL;
int error;
crhold(cr);
cookie = spl_fstrans_mark();
error = -zfsctl_shares_lookup(dip, dname(dentry), &ip,
0, cr, NULL, NULL);
ASSERT3S(error, <=, 0);
spl_fstrans_unmark(cookie);
crfree(cr);
if (error) {
if (error == -ENOENT)
return (d_splice_alias(NULL, dentry));
else
return (ERR_PTR(error));
}
return (d_splice_alias(ip, dentry));
}
static int
zpl_shares_iterate(struct file *filp, zpl_dir_context_t *ctx)
{
fstrans_cookie_t cookie;
cred_t *cr = CRED();
zfsvfs_t *zfsvfs = ITOZSB(file_inode(filp));
znode_t *dzp;
int error = 0;
ZPL_ENTER(zfsvfs);
cookie = spl_fstrans_mark();
if (zfsvfs->z_shares_dir == 0) {
zpl_dir_emit_dots(filp, ctx);
goto out;
}
error = -zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp);
if (error)
goto out;
crhold(cr);
error = -zfs_readdir(ZTOI(dzp), ctx, cr);
crfree(cr);
iput(ZTOI(dzp));
out:
spl_fstrans_unmark(cookie);
ZPL_EXIT(zfsvfs);
ASSERT3S(error, <=, 0);
return (error);
}
#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
static int
zpl_shares_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
zpl_dir_context_t ctx =
ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos);
int error;
error = zpl_shares_iterate(filp, &ctx);
filp->f_pos = ctx.pos;
return (error);
}
#endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */
/* ARGSUSED */
static int
#ifdef HAVE_USERNS_IOPS_GETATTR
zpl_shares_getattr_impl(struct user_namespace *user_ns,
const struct path *path, struct kstat *stat, u32 request_mask,
unsigned int query_flags)
#else
zpl_shares_getattr_impl(const struct path *path, struct kstat *stat,
u32 request_mask, unsigned int query_flags)
#endif
{
struct inode *ip = path->dentry->d_inode;
zfsvfs_t *zfsvfs = ITOZSB(ip);
znode_t *dzp;
int error;
ZPL_ENTER(zfsvfs);
if (zfsvfs->z_shares_dir == 0) {
#if defined(HAVE_GENERIC_FILLATTR_USERNS) && defined(HAVE_USERNS_IOPS_GETATTR)
generic_fillattr(user_ns, path->dentry->d_inode, stat);
#else
generic_fillattr(path->dentry->d_inode, stat);
#endif
stat->nlink = stat->size = 2;
stat->atime = current_time(ip);
ZPL_EXIT(zfsvfs);
return (0);
}
error = -zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp);
if (error == 0) {
#if defined(HAVE_GENERIC_FILLATTR_USERNS) && defined(HAVE_USERNS_IOPS_GETATTR)
error = -zfs_getattr_fast(user_ns, ZTOI(dzp), stat);
#else
error = -zfs_getattr_fast(kcred->user_ns, ZTOI(dzp), stat);
#endif
iput(ZTOI(dzp));
}
ZPL_EXIT(zfsvfs);
ASSERT3S(error, <=, 0);
return (error);
}
ZPL_GETATTR_WRAPPER(zpl_shares_getattr);
/*
* The '.zfs/shares' directory file operations.
*/
const struct file_operations zpl_fops_shares = {
.open = zpl_common_open,
.llseek = generic_file_llseek,
.read = generic_read_dir,
#ifdef HAVE_VFS_ITERATE_SHARED
.iterate_shared = zpl_shares_iterate,
#elif defined(HAVE_VFS_ITERATE)
.iterate = zpl_shares_iterate,
#else
.readdir = zpl_shares_readdir,
#endif
};
/*
* The '.zfs/shares' directory inode operations.
*/
const struct inode_operations zpl_ops_shares = {
.lookup = zpl_shares_lookup,
.getattr = zpl_shares_getattr,
};
diff --git a/module/os/linux/zfs/zpl_file.c b/module/os/linux/zfs/zpl_file.c
index ff324222d15d..21926f170c5c 100644
--- a/module/os/linux/zfs/zpl_file.c
+++ b/module/os/linux/zfs/zpl_file.c
@@ -1,1102 +1,1102 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2011, Lawrence Livermore National Security, LLC.
* Copyright (c) 2015 by Chunwei Chen. All rights reserved.
*/
#ifdef CONFIG_COMPAT
#include <linux/compat.h>
#endif
#include <sys/file.h>
#include <sys/dmu_objset.h>
#include <sys/zfs_znode.h>
#include <sys/zfs_vfsops.h>
#include <sys/zfs_vnops.h>
#include <sys/zfs_project.h>
#ifdef HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS
#include <linux/pagemap.h>
#endif
/*
* When using fallocate(2) to preallocate space, inflate the requested
* capacity check by 10% to account for the required metadata blocks.
*/
-unsigned int zfs_fallocate_reserve_percent = 110;
+static unsigned int zfs_fallocate_reserve_percent = 110;
static int
zpl_open(struct inode *ip, struct file *filp)
{
cred_t *cr = CRED();
int error;
fstrans_cookie_t cookie;
error = generic_file_open(ip, filp);
if (error)
return (error);
crhold(cr);
cookie = spl_fstrans_mark();
error = -zfs_open(ip, filp->f_mode, filp->f_flags, cr);
spl_fstrans_unmark(cookie);
crfree(cr);
ASSERT3S(error, <=, 0);
return (error);
}
static int
zpl_release(struct inode *ip, struct file *filp)
{
cred_t *cr = CRED();
int error;
fstrans_cookie_t cookie;
cookie = spl_fstrans_mark();
if (ITOZ(ip)->z_atime_dirty)
zfs_mark_inode_dirty(ip);
crhold(cr);
error = -zfs_close(ip, filp->f_flags, cr);
spl_fstrans_unmark(cookie);
crfree(cr);
ASSERT3S(error, <=, 0);
return (error);
}
static int
zpl_iterate(struct file *filp, zpl_dir_context_t *ctx)
{
cred_t *cr = CRED();
int error;
fstrans_cookie_t cookie;
crhold(cr);
cookie = spl_fstrans_mark();
error = -zfs_readdir(file_inode(filp), ctx, cr);
spl_fstrans_unmark(cookie);
crfree(cr);
ASSERT3S(error, <=, 0);
return (error);
}
#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
static int
zpl_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
zpl_dir_context_t ctx =
ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos);
int error;
error = zpl_iterate(filp, &ctx);
filp->f_pos = ctx.pos;
return (error);
}
#endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */
#if defined(HAVE_FSYNC_WITHOUT_DENTRY)
/*
* Linux 2.6.35 - 3.0 API,
* As of 2.6.35 the dentry argument to the fops->fsync() hook was deemed
* redundant. The dentry is still accessible via filp->f_path.dentry,
* and we are guaranteed that filp will never be NULL.
*/
static int
zpl_fsync(struct file *filp, int datasync)
{
struct inode *inode = filp->f_mapping->host;
cred_t *cr = CRED();
int error;
fstrans_cookie_t cookie;
crhold(cr);
cookie = spl_fstrans_mark();
error = -zfs_fsync(ITOZ(inode), datasync, cr);
spl_fstrans_unmark(cookie);
crfree(cr);
ASSERT3S(error, <=, 0);
return (error);
}
#ifdef HAVE_FILE_AIO_FSYNC
static int
zpl_aio_fsync(struct kiocb *kiocb, int datasync)
{
return (zpl_fsync(kiocb->ki_filp, datasync));
}
#endif
#elif defined(HAVE_FSYNC_RANGE)
/*
* Linux 3.1 API,
* As of 3.1 the responsibility to call filemap_write_and_wait_range() has
* been pushed down in to the .fsync() vfs hook. Additionally, the i_mutex
* lock is no longer held by the caller, for zfs we don't require the lock
* to be held so we don't acquire it.
*/
static int
zpl_fsync(struct file *filp, loff_t start, loff_t end, int datasync)
{
struct inode *inode = filp->f_mapping->host;
cred_t *cr = CRED();
int error;
fstrans_cookie_t cookie;
error = filemap_write_and_wait_range(inode->i_mapping, start, end);
if (error)
return (error);
crhold(cr);
cookie = spl_fstrans_mark();
error = -zfs_fsync(ITOZ(inode), datasync, cr);
spl_fstrans_unmark(cookie);
crfree(cr);
ASSERT3S(error, <=, 0);
return (error);
}
#ifdef HAVE_FILE_AIO_FSYNC
static int
zpl_aio_fsync(struct kiocb *kiocb, int datasync)
{
return (zpl_fsync(kiocb->ki_filp, kiocb->ki_pos, -1, datasync));
}
#endif
#else
#error "Unsupported fops->fsync() implementation"
#endif
static inline int
zfs_io_flags(struct kiocb *kiocb)
{
int flags = 0;
#if defined(IOCB_DSYNC)
if (kiocb->ki_flags & IOCB_DSYNC)
flags |= O_DSYNC;
#endif
#if defined(IOCB_SYNC)
if (kiocb->ki_flags & IOCB_SYNC)
flags |= O_SYNC;
#endif
#if defined(IOCB_APPEND)
if (kiocb->ki_flags & IOCB_APPEND)
flags |= O_APPEND;
#endif
#if defined(IOCB_DIRECT)
if (kiocb->ki_flags & IOCB_DIRECT)
flags |= O_DIRECT;
#endif
return (flags);
}
/*
* If relatime is enabled, call file_accessed() if zfs_relatime_need_update()
* is true. This is needed since datasets with inherited "relatime" property
* aren't necessarily mounted with the MNT_RELATIME flag (e.g. after
* `zfs set relatime=...`), which is what relatime test in VFS by
* relatime_need_update() is based on.
*/
static inline void
zpl_file_accessed(struct file *filp)
{
struct inode *ip = filp->f_mapping->host;
if (!IS_NOATIME(ip) && ITOZSB(ip)->z_relatime) {
if (zfs_relatime_need_update(ip))
file_accessed(filp);
} else {
file_accessed(filp);
}
}
#if defined(HAVE_VFS_RW_ITERATE)
/*
* When HAVE_VFS_IOV_ITER is defined the iov_iter structure supports
* iovecs, kvevs, bvecs and pipes, plus all the required interfaces to
* manipulate the iov_iter are available. In which case the full iov_iter
* can be attached to the uio and correctly handled in the lower layers.
* Otherwise, for older kernels extract the iovec and pass it instead.
*/
static void
zpl_uio_init(zfs_uio_t *uio, struct kiocb *kiocb, struct iov_iter *to,
loff_t pos, ssize_t count, size_t skip)
{
#if defined(HAVE_VFS_IOV_ITER)
zfs_uio_iov_iter_init(uio, to, pos, count, skip);
#else
#ifdef HAVE_IOV_ITER_TYPE
zfs_uio_iovec_init(uio, to->iov, to->nr_segs, pos,
iov_iter_type(to) & ITER_KVEC ? UIO_SYSSPACE : UIO_USERSPACE,
count, skip);
#else
zfs_uio_iovec_init(uio, to->iov, to->nr_segs, pos,
to->type & ITER_KVEC ? UIO_SYSSPACE : UIO_USERSPACE,
count, skip);
#endif
#endif
}
static ssize_t
zpl_iter_read(struct kiocb *kiocb, struct iov_iter *to)
{
cred_t *cr = CRED();
fstrans_cookie_t cookie;
struct file *filp = kiocb->ki_filp;
ssize_t count = iov_iter_count(to);
zfs_uio_t uio;
zpl_uio_init(&uio, kiocb, to, kiocb->ki_pos, count, 0);
crhold(cr);
cookie = spl_fstrans_mark();
int error = -zfs_read(ITOZ(filp->f_mapping->host), &uio,
filp->f_flags | zfs_io_flags(kiocb), cr);
spl_fstrans_unmark(cookie);
crfree(cr);
if (error < 0)
return (error);
ssize_t read = count - uio.uio_resid;
kiocb->ki_pos += read;
zpl_file_accessed(filp);
return (read);
}
static inline ssize_t
zpl_generic_write_checks(struct kiocb *kiocb, struct iov_iter *from,
size_t *countp)
{
#ifdef HAVE_GENERIC_WRITE_CHECKS_KIOCB
ssize_t ret = generic_write_checks(kiocb, from);
if (ret <= 0)
return (ret);
*countp = ret;
#else
struct file *file = kiocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *ip = mapping->host;
int isblk = S_ISBLK(ip->i_mode);
*countp = iov_iter_count(from);
ssize_t ret = generic_write_checks(file, &kiocb->ki_pos, countp, isblk);
if (ret)
return (ret);
#endif
return (0);
}
static ssize_t
zpl_iter_write(struct kiocb *kiocb, struct iov_iter *from)
{
cred_t *cr = CRED();
fstrans_cookie_t cookie;
struct file *filp = kiocb->ki_filp;
struct inode *ip = filp->f_mapping->host;
zfs_uio_t uio;
size_t count = 0;
ssize_t ret;
ret = zpl_generic_write_checks(kiocb, from, &count);
if (ret)
return (ret);
zpl_uio_init(&uio, kiocb, from, kiocb->ki_pos, count, from->iov_offset);
crhold(cr);
cookie = spl_fstrans_mark();
int error = -zfs_write(ITOZ(ip), &uio,
filp->f_flags | zfs_io_flags(kiocb), cr);
spl_fstrans_unmark(cookie);
crfree(cr);
if (error < 0)
return (error);
ssize_t wrote = count - uio.uio_resid;
kiocb->ki_pos += wrote;
return (wrote);
}
#else /* !HAVE_VFS_RW_ITERATE */
static ssize_t
zpl_aio_read(struct kiocb *kiocb, const struct iovec *iov,
unsigned long nr_segs, loff_t pos)
{
cred_t *cr = CRED();
fstrans_cookie_t cookie;
struct file *filp = kiocb->ki_filp;
size_t count;
ssize_t ret;
ret = generic_segment_checks(iov, &nr_segs, &count, VERIFY_WRITE);
if (ret)
return (ret);
zfs_uio_t uio;
zfs_uio_iovec_init(&uio, iov, nr_segs, kiocb->ki_pos, UIO_USERSPACE,
count, 0);
crhold(cr);
cookie = spl_fstrans_mark();
int error = -zfs_read(ITOZ(filp->f_mapping->host), &uio,
filp->f_flags | zfs_io_flags(kiocb), cr);
spl_fstrans_unmark(cookie);
crfree(cr);
if (error < 0)
return (error);
ssize_t read = count - uio.uio_resid;
kiocb->ki_pos += read;
zpl_file_accessed(filp);
return (read);
}
static ssize_t
zpl_aio_write(struct kiocb *kiocb, const struct iovec *iov,
unsigned long nr_segs, loff_t pos)
{
cred_t *cr = CRED();
fstrans_cookie_t cookie;
struct file *filp = kiocb->ki_filp;
struct inode *ip = filp->f_mapping->host;
size_t count;
ssize_t ret;
ret = generic_segment_checks(iov, &nr_segs, &count, VERIFY_READ);
if (ret)
return (ret);
ret = generic_write_checks(filp, &pos, &count, S_ISBLK(ip->i_mode));
if (ret)
return (ret);
zfs_uio_t uio;
zfs_uio_iovec_init(&uio, iov, nr_segs, kiocb->ki_pos, UIO_USERSPACE,
count, 0);
crhold(cr);
cookie = spl_fstrans_mark();
int error = -zfs_write(ITOZ(ip), &uio,
filp->f_flags | zfs_io_flags(kiocb), cr);
spl_fstrans_unmark(cookie);
crfree(cr);
if (error < 0)
return (error);
ssize_t wrote = count - uio.uio_resid;
kiocb->ki_pos += wrote;
return (wrote);
}
#endif /* HAVE_VFS_RW_ITERATE */
#if defined(HAVE_VFS_RW_ITERATE)
static ssize_t
zpl_direct_IO_impl(int rw, struct kiocb *kiocb, struct iov_iter *iter)
{
if (rw == WRITE)
return (zpl_iter_write(kiocb, iter));
else
return (zpl_iter_read(kiocb, iter));
}
#if defined(HAVE_VFS_DIRECT_IO_ITER)
static ssize_t
zpl_direct_IO(struct kiocb *kiocb, struct iov_iter *iter)
{
return (zpl_direct_IO_impl(iov_iter_rw(iter), kiocb, iter));
}
#elif defined(HAVE_VFS_DIRECT_IO_ITER_OFFSET)
static ssize_t
zpl_direct_IO(struct kiocb *kiocb, struct iov_iter *iter, loff_t pos)
{
ASSERT3S(pos, ==, kiocb->ki_pos);
return (zpl_direct_IO_impl(iov_iter_rw(iter), kiocb, iter));
}
#elif defined(HAVE_VFS_DIRECT_IO_ITER_RW_OFFSET)
static ssize_t
zpl_direct_IO(int rw, struct kiocb *kiocb, struct iov_iter *iter, loff_t pos)
{
ASSERT3S(pos, ==, kiocb->ki_pos);
return (zpl_direct_IO_impl(rw, kiocb, iter));
}
#else
#error "Unknown direct IO interface"
#endif
#else /* HAVE_VFS_RW_ITERATE */
#if defined(HAVE_VFS_DIRECT_IO_IOVEC)
static ssize_t
zpl_direct_IO(int rw, struct kiocb *kiocb, const struct iovec *iov,
loff_t pos, unsigned long nr_segs)
{
if (rw == WRITE)
return (zpl_aio_write(kiocb, iov, nr_segs, pos));
else
return (zpl_aio_read(kiocb, iov, nr_segs, pos));
}
#elif defined(HAVE_VFS_DIRECT_IO_ITER_RW_OFFSET)
static ssize_t
zpl_direct_IO(int rw, struct kiocb *kiocb, struct iov_iter *iter, loff_t pos)
{
const struct iovec *iovp = iov_iter_iovec(iter);
unsigned long nr_segs = iter->nr_segs;
ASSERT3S(pos, ==, kiocb->ki_pos);
if (rw == WRITE)
return (zpl_aio_write(kiocb, iovp, nr_segs, pos));
else
return (zpl_aio_read(kiocb, iovp, nr_segs, pos));
}
#else
#error "Unknown direct IO interface"
#endif
#endif /* HAVE_VFS_RW_ITERATE */
static loff_t
zpl_llseek(struct file *filp, loff_t offset, int whence)
{
#if defined(SEEK_HOLE) && defined(SEEK_DATA)
fstrans_cookie_t cookie;
if (whence == SEEK_DATA || whence == SEEK_HOLE) {
struct inode *ip = filp->f_mapping->host;
loff_t maxbytes = ip->i_sb->s_maxbytes;
loff_t error;
spl_inode_lock_shared(ip);
cookie = spl_fstrans_mark();
error = -zfs_holey(ITOZ(ip), whence, &offset);
spl_fstrans_unmark(cookie);
if (error == 0)
error = lseek_execute(filp, ip, offset, maxbytes);
spl_inode_unlock_shared(ip);
return (error);
}
#endif /* SEEK_HOLE && SEEK_DATA */
return (generic_file_llseek(filp, offset, whence));
}
/*
* It's worth taking a moment to describe how mmap is implemented
* for zfs because it differs considerably from other Linux filesystems.
* However, this issue is handled the same way under OpenSolaris.
*
* The issue is that by design zfs bypasses the Linux page cache and
* leaves all caching up to the ARC. This has been shown to work
* well for the common read(2)/write(2) case. However, mmap(2)
* is problem because it relies on being tightly integrated with the
* page cache. To handle this we cache mmap'ed files twice, once in
* the ARC and a second time in the page cache. The code is careful
* to keep both copies synchronized.
*
* When a file with an mmap'ed region is written to using write(2)
* both the data in the ARC and existing pages in the page cache
* are updated. For a read(2) data will be read first from the page
* cache then the ARC if needed. Neither a write(2) or read(2) will
* will ever result in new pages being added to the page cache.
*
* New pages are added to the page cache only via .readpage() which
* is called when the vfs needs to read a page off disk to back the
* virtual memory region. These pages may be modified without
* notifying the ARC and will be written out periodically via
* .writepage(). This will occur due to either a sync or the usual
* page aging behavior. Note because a read(2) of a mmap'ed file
* will always check the page cache first even when the ARC is out
* of date correct data will still be returned.
*
* While this implementation ensures correct behavior it does have
* have some drawbacks. The most obvious of which is that it
* increases the required memory footprint when access mmap'ed
* files. It also adds additional complexity to the code keeping
* both caches synchronized.
*
* Longer term it may be possible to cleanly resolve this wart by
* mapping page cache pages directly on to the ARC buffers. The
* Linux address space operations are flexible enough to allow
* selection of which pages back a particular index. The trick
* would be working out the details of which subsystem is in
* charge, the ARC, the page cache, or both. It may also prove
* helpful to move the ARC buffers to a scatter-gather lists
* rather than a vmalloc'ed region.
*/
static int
zpl_mmap(struct file *filp, struct vm_area_struct *vma)
{
struct inode *ip = filp->f_mapping->host;
znode_t *zp = ITOZ(ip);
int error;
fstrans_cookie_t cookie;
cookie = spl_fstrans_mark();
error = -zfs_map(ip, vma->vm_pgoff, (caddr_t *)vma->vm_start,
(size_t)(vma->vm_end - vma->vm_start), vma->vm_flags);
spl_fstrans_unmark(cookie);
if (error)
return (error);
error = generic_file_mmap(filp, vma);
if (error)
return (error);
mutex_enter(&zp->z_lock);
zp->z_is_mapped = B_TRUE;
mutex_exit(&zp->z_lock);
return (error);
}
/*
* Populate a page with data for the Linux page cache. This function is
* only used to support mmap(2). There will be an identical copy of the
* data in the ARC which is kept up to date via .write() and .writepage().
*/
static inline int
zpl_readpage_common(struct page *pp)
{
struct inode *ip;
struct page *pl[1];
int error = 0;
fstrans_cookie_t cookie;
ASSERT(PageLocked(pp));
ip = pp->mapping->host;
pl[0] = pp;
cookie = spl_fstrans_mark();
error = -zfs_getpage(ip, pl, 1);
spl_fstrans_unmark(cookie);
if (error) {
SetPageError(pp);
ClearPageUptodate(pp);
} else {
ClearPageError(pp);
SetPageUptodate(pp);
flush_dcache_page(pp);
}
unlock_page(pp);
return (error);
}
static int
zpl_readpage(struct file *filp, struct page *pp)
{
return (zpl_readpage_common(pp));
}
static int
zpl_readpage_filler(void *data, struct page *pp)
{
return (zpl_readpage_common(pp));
}
/*
* Populate a set of pages with data for the Linux page cache. This
* function will only be called for read ahead and never for demand
* paging. For simplicity, the code relies on read_cache_pages() to
* correctly lock each page for IO and call zpl_readpage().
*/
static int
zpl_readpages(struct file *filp, struct address_space *mapping,
struct list_head *pages, unsigned nr_pages)
{
return (read_cache_pages(mapping, pages, zpl_readpage_filler, NULL));
}
static int
zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data)
{
struct address_space *mapping = data;
fstrans_cookie_t cookie;
ASSERT(PageLocked(pp));
ASSERT(!PageWriteback(pp));
cookie = spl_fstrans_mark();
(void) zfs_putpage(mapping->host, pp, wbc);
spl_fstrans_unmark(cookie);
return (0);
}
static int
zpl_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
znode_t *zp = ITOZ(mapping->host);
zfsvfs_t *zfsvfs = ITOZSB(mapping->host);
enum writeback_sync_modes sync_mode;
int result;
ZPL_ENTER(zfsvfs);
if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
wbc->sync_mode = WB_SYNC_ALL;
ZPL_EXIT(zfsvfs);
sync_mode = wbc->sync_mode;
/*
* We don't want to run write_cache_pages() in SYNC mode here, because
* that would make putpage() wait for a single page to be committed to
* disk every single time, resulting in atrocious performance. Instead
* we run it once in non-SYNC mode so that the ZIL gets all the data,
* and then we commit it all in one go.
*/
wbc->sync_mode = WB_SYNC_NONE;
result = write_cache_pages(mapping, wbc, zpl_putpage, mapping);
if (sync_mode != wbc->sync_mode) {
ZPL_ENTER(zfsvfs);
ZPL_VERIFY_ZP(zp);
if (zfsvfs->z_log != NULL)
zil_commit(zfsvfs->z_log, zp->z_id);
ZPL_EXIT(zfsvfs);
/*
* We need to call write_cache_pages() again (we can't just
* return after the commit) because the previous call in
* non-SYNC mode does not guarantee that we got all the dirty
* pages (see the implementation of write_cache_pages() for
* details). That being said, this is a no-op in most cases.
*/
wbc->sync_mode = sync_mode;
result = write_cache_pages(mapping, wbc, zpl_putpage, mapping);
}
return (result);
}
/*
* Write out dirty pages to the ARC, this function is only required to
* support mmap(2). Mapped pages may be dirtied by memory operations
* which never call .write(). These dirty pages are kept in sync with
* the ARC buffers via this hook.
*/
static int
zpl_writepage(struct page *pp, struct writeback_control *wbc)
{
if (ITOZSB(pp->mapping->host)->z_os->os_sync == ZFS_SYNC_ALWAYS)
wbc->sync_mode = WB_SYNC_ALL;
return (zpl_putpage(pp, wbc, pp->mapping));
}
/*
* The flag combination which matches the behavior of zfs_space() is
* FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE. The FALLOC_FL_PUNCH_HOLE
* flag was introduced in the 2.6.38 kernel.
*
* The original mode=0 (allocate space) behavior can be reasonably emulated
* by checking if enough space exists and creating a sparse file, as real
* persistent space reservation is not possible due to COW, snapshots, etc.
*/
static long
zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len)
{
cred_t *cr = CRED();
loff_t olen;
fstrans_cookie_t cookie;
int error = 0;
if ((mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)) != 0)
return (-EOPNOTSUPP);
if (offset < 0 || len <= 0)
return (-EINVAL);
spl_inode_lock(ip);
olen = i_size_read(ip);
crhold(cr);
cookie = spl_fstrans_mark();
if (mode & FALLOC_FL_PUNCH_HOLE) {
flock64_t bf;
if (offset > olen)
goto out_unmark;
if (offset + len > olen)
len = olen - offset;
bf.l_type = F_WRLCK;
bf.l_whence = SEEK_SET;
bf.l_start = offset;
bf.l_len = len;
bf.l_pid = 0;
error = -zfs_space(ITOZ(ip), F_FREESP, &bf, O_RDWR, offset, cr);
} else if ((mode & ~FALLOC_FL_KEEP_SIZE) == 0) {
unsigned int percent = zfs_fallocate_reserve_percent;
struct kstatfs statfs;
/* Legacy mode, disable fallocate compatibility. */
if (percent == 0) {
error = -EOPNOTSUPP;
goto out_unmark;
}
/*
* Use zfs_statvfs() instead of dmu_objset_space() since it
* also checks project quota limits, which are relevant here.
*/
error = zfs_statvfs(ip, &statfs);
if (error)
goto out_unmark;
/*
* Shrink available space a bit to account for overhead/races.
* We know the product previously fit into availbytes from
* dmu_objset_space(), so the smaller product will also fit.
*/
if (len > statfs.f_bavail * (statfs.f_bsize * 100 / percent)) {
error = -ENOSPC;
goto out_unmark;
}
if (!(mode & FALLOC_FL_KEEP_SIZE) && offset + len > olen)
error = zfs_freesp(ITOZ(ip), offset + len, 0, 0, FALSE);
}
out_unmark:
spl_fstrans_unmark(cookie);
spl_inode_unlock(ip);
crfree(cr);
return (error);
}
static long
zpl_fallocate(struct file *filp, int mode, loff_t offset, loff_t len)
{
return zpl_fallocate_common(file_inode(filp),
mode, offset, len);
}
static int
zpl_ioctl_getversion(struct file *filp, void __user *arg)
{
uint32_t generation = file_inode(filp)->i_generation;
return (copy_to_user(arg, &generation, sizeof (generation)));
}
#define ZFS_FL_USER_VISIBLE (FS_FL_USER_VISIBLE | ZFS_PROJINHERIT_FL)
#define ZFS_FL_USER_MODIFIABLE (FS_FL_USER_MODIFIABLE | ZFS_PROJINHERIT_FL)
static uint32_t
__zpl_ioctl_getflags(struct inode *ip)
{
uint64_t zfs_flags = ITOZ(ip)->z_pflags;
uint32_t ioctl_flags = 0;
if (zfs_flags & ZFS_IMMUTABLE)
ioctl_flags |= FS_IMMUTABLE_FL;
if (zfs_flags & ZFS_APPENDONLY)
ioctl_flags |= FS_APPEND_FL;
if (zfs_flags & ZFS_NODUMP)
ioctl_flags |= FS_NODUMP_FL;
if (zfs_flags & ZFS_PROJINHERIT)
ioctl_flags |= ZFS_PROJINHERIT_FL;
return (ioctl_flags & ZFS_FL_USER_VISIBLE);
}
/*
* Map zfs file z_pflags (xvattr_t) to linux file attributes. Only file
* attributes common to both Linux and Solaris are mapped.
*/
static int
zpl_ioctl_getflags(struct file *filp, void __user *arg)
{
uint32_t flags;
int err;
flags = __zpl_ioctl_getflags(file_inode(filp));
err = copy_to_user(arg, &flags, sizeof (flags));
return (err);
}
/*
* fchange() is a helper macro to detect if we have been asked to change a
* flag. This is ugly, but the requirement that we do this is a consequence of
* how the Linux file attribute interface was designed. Another consequence is
* that concurrent modification of files suffers from a TOCTOU race. Neither
* are things we can fix without modifying the kernel-userland interface, which
* is outside of our jurisdiction.
*/
#define fchange(f0, f1, b0, b1) (!((f0) & (b0)) != !((f1) & (b1)))
static int
__zpl_ioctl_setflags(struct inode *ip, uint32_t ioctl_flags, xvattr_t *xva)
{
uint64_t zfs_flags = ITOZ(ip)->z_pflags;
xoptattr_t *xoap;
if (ioctl_flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | FS_NODUMP_FL |
ZFS_PROJINHERIT_FL))
return (-EOPNOTSUPP);
if (ioctl_flags & ~ZFS_FL_USER_MODIFIABLE)
return (-EACCES);
if ((fchange(ioctl_flags, zfs_flags, FS_IMMUTABLE_FL, ZFS_IMMUTABLE) ||
fchange(ioctl_flags, zfs_flags, FS_APPEND_FL, ZFS_APPENDONLY)) &&
!capable(CAP_LINUX_IMMUTABLE))
return (-EPERM);
if (!zpl_inode_owner_or_capable(kcred->user_ns, ip))
return (-EACCES);
xva_init(xva);
xoap = xva_getxoptattr(xva);
XVA_SET_REQ(xva, XAT_IMMUTABLE);
if (ioctl_flags & FS_IMMUTABLE_FL)
xoap->xoa_immutable = B_TRUE;
XVA_SET_REQ(xva, XAT_APPENDONLY);
if (ioctl_flags & FS_APPEND_FL)
xoap->xoa_appendonly = B_TRUE;
XVA_SET_REQ(xva, XAT_NODUMP);
if (ioctl_flags & FS_NODUMP_FL)
xoap->xoa_nodump = B_TRUE;
XVA_SET_REQ(xva, XAT_PROJINHERIT);
if (ioctl_flags & ZFS_PROJINHERIT_FL)
xoap->xoa_projinherit = B_TRUE;
return (0);
}
static int
zpl_ioctl_setflags(struct file *filp, void __user *arg)
{
struct inode *ip = file_inode(filp);
uint32_t flags;
cred_t *cr = CRED();
xvattr_t xva;
int err;
fstrans_cookie_t cookie;
if (copy_from_user(&flags, arg, sizeof (flags)))
return (-EFAULT);
err = __zpl_ioctl_setflags(ip, flags, &xva);
if (err)
return (err);
crhold(cr);
cookie = spl_fstrans_mark();
err = -zfs_setattr(ITOZ(ip), (vattr_t *)&xva, 0, cr);
spl_fstrans_unmark(cookie);
crfree(cr);
return (err);
}
static int
zpl_ioctl_getxattr(struct file *filp, void __user *arg)
{
zfsxattr_t fsx = { 0 };
struct inode *ip = file_inode(filp);
int err;
fsx.fsx_xflags = __zpl_ioctl_getflags(ip);
fsx.fsx_projid = ITOZ(ip)->z_projid;
err = copy_to_user(arg, &fsx, sizeof (fsx));
return (err);
}
static int
zpl_ioctl_setxattr(struct file *filp, void __user *arg)
{
struct inode *ip = file_inode(filp);
zfsxattr_t fsx;
cred_t *cr = CRED();
xvattr_t xva;
xoptattr_t *xoap;
int err;
fstrans_cookie_t cookie;
if (copy_from_user(&fsx, arg, sizeof (fsx)))
return (-EFAULT);
if (!zpl_is_valid_projid(fsx.fsx_projid))
return (-EINVAL);
err = __zpl_ioctl_setflags(ip, fsx.fsx_xflags, &xva);
if (err)
return (err);
xoap = xva_getxoptattr(&xva);
XVA_SET_REQ(&xva, XAT_PROJID);
xoap->xoa_projid = fsx.fsx_projid;
crhold(cr);
cookie = spl_fstrans_mark();
err = -zfs_setattr(ITOZ(ip), (vattr_t *)&xva, 0, cr);
spl_fstrans_unmark(cookie);
crfree(cr);
return (err);
}
static long
zpl_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
{
switch (cmd) {
case FS_IOC_GETVERSION:
return (zpl_ioctl_getversion(filp, (void *)arg));
case FS_IOC_GETFLAGS:
return (zpl_ioctl_getflags(filp, (void *)arg));
case FS_IOC_SETFLAGS:
return (zpl_ioctl_setflags(filp, (void *)arg));
case ZFS_IOC_FSGETXATTR:
return (zpl_ioctl_getxattr(filp, (void *)arg));
case ZFS_IOC_FSSETXATTR:
return (zpl_ioctl_setxattr(filp, (void *)arg));
default:
return (-ENOTTY);
}
}
#ifdef CONFIG_COMPAT
static long
zpl_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
{
switch (cmd) {
case FS_IOC32_GETVERSION:
cmd = FS_IOC_GETVERSION;
break;
case FS_IOC32_GETFLAGS:
cmd = FS_IOC_GETFLAGS;
break;
case FS_IOC32_SETFLAGS:
cmd = FS_IOC_SETFLAGS;
break;
default:
return (-ENOTTY);
}
return (zpl_ioctl(filp, cmd, (unsigned long)compat_ptr(arg)));
}
#endif /* CONFIG_COMPAT */
const struct address_space_operations zpl_address_space_operations = {
.readpages = zpl_readpages,
.readpage = zpl_readpage,
.writepage = zpl_writepage,
.writepages = zpl_writepages,
.direct_IO = zpl_direct_IO,
#ifdef HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS
.set_page_dirty = __set_page_dirty_nobuffers,
#endif
};
const struct file_operations zpl_file_operations = {
.open = zpl_open,
.release = zpl_release,
.llseek = zpl_llseek,
#ifdef HAVE_VFS_RW_ITERATE
#ifdef HAVE_NEW_SYNC_READ
.read = new_sync_read,
.write = new_sync_write,
#endif
.read_iter = zpl_iter_read,
.write_iter = zpl_iter_write,
#ifdef HAVE_VFS_IOV_ITER
.splice_read = generic_file_splice_read,
.splice_write = iter_file_splice_write,
#endif
#else
.read = do_sync_read,
.write = do_sync_write,
.aio_read = zpl_aio_read,
.aio_write = zpl_aio_write,
#endif
.mmap = zpl_mmap,
.fsync = zpl_fsync,
#ifdef HAVE_FILE_AIO_FSYNC
.aio_fsync = zpl_aio_fsync,
#endif
.fallocate = zpl_fallocate,
.unlocked_ioctl = zpl_ioctl,
#ifdef CONFIG_COMPAT
.compat_ioctl = zpl_compat_ioctl,
#endif
};
const struct file_operations zpl_dir_file_operations = {
.llseek = generic_file_llseek,
.read = generic_read_dir,
#if defined(HAVE_VFS_ITERATE_SHARED)
.iterate_shared = zpl_iterate,
#elif defined(HAVE_VFS_ITERATE)
.iterate = zpl_iterate,
#else
.readdir = zpl_readdir,
#endif
.fsync = zpl_fsync,
.unlocked_ioctl = zpl_ioctl,
#ifdef CONFIG_COMPAT
.compat_ioctl = zpl_compat_ioctl,
#endif
};
/* BEGIN CSTYLED */
module_param(zfs_fallocate_reserve_percent, uint, 0644);
MODULE_PARM_DESC(zfs_fallocate_reserve_percent,
"Percentage of length to use for the available capacity check");
/* END CSTYLED */
diff --git a/module/os/linux/zfs/zpl_xattr.c b/module/os/linux/zfs/zpl_xattr.c
index e7726e8458af..a1921ed08863 100644
--- a/module/os/linux/zfs/zpl_xattr.c
+++ b/module/os/linux/zfs/zpl_xattr.c
@@ -1,1515 +1,1512 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2011, Lawrence Livermore National Security, LLC.
*
* Extended attributes (xattr) on Solaris are implemented as files
* which exist in a hidden xattr directory. These extended attributes
* can be accessed using the attropen() system call which opens
* the extended attribute. It can then be manipulated just like
* a standard file descriptor. This has a couple advantages such
* as practically no size limit on the file, and the extended
* attributes permissions may differ from those of the parent file.
* This interface is really quite clever, but it's also completely
* different than what is supported on Linux. It also comes with a
* steep performance penalty when accessing small xattrs because they
* are not stored with the parent file.
*
* Under Linux extended attributes are manipulated by the system
* calls getxattr(2), setxattr(2), and listxattr(2). They consider
* extended attributes to be name/value pairs where the name is a
* NULL terminated string. The name must also include one of the
* following namespace prefixes:
*
* user - No restrictions and is available to user applications.
* trusted - Restricted to kernel and root (CAP_SYS_ADMIN) use.
* system - Used for access control lists (system.nfs4_acl, etc).
* security - Used by SELinux to store a files security context.
*
* The value under Linux to limited to 65536 bytes of binary data.
* In practice, individual xattrs tend to be much smaller than this
* and are typically less than 100 bytes. A good example of this
* are the security.selinux xattrs which are less than 100 bytes and
* exist for every file when xattr labeling is enabled.
*
* The Linux xattr implementation has been written to take advantage of
* this typical usage. When the dataset property 'xattr=sa' is set,
* then xattrs will be preferentially stored as System Attributes (SA).
* This allows tiny xattrs (~100 bytes) to be stored with the dnode and
* up to 64k of xattrs to be stored in the spill block. If additional
* xattr space is required, which is unlikely under Linux, they will
* be stored using the traditional directory approach.
*
* This optimization results in roughly a 3x performance improvement
* when accessing xattrs because it avoids the need to perform a seek
* for every xattr value. When multiple xattrs are stored per-file
* the performance improvements are even greater because all of the
* xattrs stored in the spill block will be cached.
*
* However, by default SA based xattrs are disabled in the Linux port
* to maximize compatibility with other implementations. If you do
* enable SA based xattrs then they will not be visible on platforms
* which do not support this feature.
*
* NOTE: One additional consequence of the xattr directory implementation
* is that when an extended attribute is manipulated an inode is created.
* This inode will exist in the Linux inode cache but there will be no
* associated entry in the dentry cache which references it. This is
* safe but it may result in some confusion. Enabling SA based xattrs
* largely avoids the issue except in the overflow case.
*/
#include <sys/zfs_znode.h>
#include <sys/zfs_vfsops.h>
#include <sys/zfs_vnops.h>
#include <sys/zap.h>
#include <sys/vfs.h>
#include <sys/zpl.h>
typedef struct xattr_filldir {
size_t size;
size_t offset;
char *buf;
struct dentry *dentry;
} xattr_filldir_t;
static const struct xattr_handler *zpl_xattr_handler(const char *);
static int
zpl_xattr_permission(xattr_filldir_t *xf, const char *name, int name_len)
{
static const struct xattr_handler *handler;
struct dentry *d = xf->dentry;
handler = zpl_xattr_handler(name);
if (!handler)
return (0);
if (handler->list) {
#if defined(HAVE_XATTR_LIST_SIMPLE)
if (!handler->list(d))
return (0);
#elif defined(HAVE_XATTR_LIST_DENTRY)
if (!handler->list(d, NULL, 0, name, name_len, 0))
return (0);
#elif defined(HAVE_XATTR_LIST_HANDLER)
if (!handler->list(handler, d, NULL, 0, name, name_len))
return (0);
#endif
}
return (1);
}
/*
* Determine is a given xattr name should be visible and if so copy it
* in to the provided buffer (xf->buf).
*/
static int
zpl_xattr_filldir(xattr_filldir_t *xf, const char *name, int name_len)
{
/* Check permissions using the per-namespace list xattr handler. */
if (!zpl_xattr_permission(xf, name, name_len))
return (0);
/* When xf->buf is NULL only calculate the required size. */
if (xf->buf) {
if (xf->offset + name_len + 1 > xf->size)
return (-ERANGE);
memcpy(xf->buf + xf->offset, name, name_len);
xf->buf[xf->offset + name_len] = '\0';
}
xf->offset += (name_len + 1);
return (0);
}
/*
* Read as many directory entry names as will fit in to the provided buffer,
* or when no buffer is provided calculate the required buffer size.
*/
static int
zpl_xattr_readdir(struct inode *dxip, xattr_filldir_t *xf)
{
zap_cursor_t zc;
zap_attribute_t zap;
int error;
zap_cursor_init(&zc, ITOZSB(dxip)->z_os, ITOZ(dxip)->z_id);
while ((error = -zap_cursor_retrieve(&zc, &zap)) == 0) {
if (zap.za_integer_length != 8 || zap.za_num_integers != 1) {
error = -ENXIO;
break;
}
error = zpl_xattr_filldir(xf, zap.za_name, strlen(zap.za_name));
if (error)
break;
zap_cursor_advance(&zc);
}
zap_cursor_fini(&zc);
if (error == -ENOENT)
error = 0;
return (error);
}
static ssize_t
zpl_xattr_list_dir(xattr_filldir_t *xf, cred_t *cr)
{
struct inode *ip = xf->dentry->d_inode;
struct inode *dxip = NULL;
znode_t *dxzp;
int error;
/* Lookup the xattr directory */
error = -zfs_lookup(ITOZ(ip), NULL, &dxzp, LOOKUP_XATTR,
cr, NULL, NULL);
if (error) {
if (error == -ENOENT)
error = 0;
return (error);
}
dxip = ZTOI(dxzp);
error = zpl_xattr_readdir(dxip, xf);
iput(dxip);
return (error);
}
static ssize_t
zpl_xattr_list_sa(xattr_filldir_t *xf)
{
znode_t *zp = ITOZ(xf->dentry->d_inode);
nvpair_t *nvp = NULL;
int error = 0;
mutex_enter(&zp->z_lock);
if (zp->z_xattr_cached == NULL)
error = -zfs_sa_get_xattr(zp);
mutex_exit(&zp->z_lock);
if (error)
return (error);
ASSERT(zp->z_xattr_cached);
while ((nvp = nvlist_next_nvpair(zp->z_xattr_cached, nvp)) != NULL) {
ASSERT3U(nvpair_type(nvp), ==, DATA_TYPE_BYTE_ARRAY);
error = zpl_xattr_filldir(xf, nvpair_name(nvp),
strlen(nvpair_name(nvp)));
if (error)
return (error);
}
return (0);
}
ssize_t
zpl_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
{
znode_t *zp = ITOZ(dentry->d_inode);
zfsvfs_t *zfsvfs = ZTOZSB(zp);
xattr_filldir_t xf = { buffer_size, 0, buffer, dentry };
cred_t *cr = CRED();
fstrans_cookie_t cookie;
int error = 0;
crhold(cr);
cookie = spl_fstrans_mark();
ZPL_ENTER(zfsvfs);
ZPL_VERIFY_ZP(zp);
rw_enter(&zp->z_xattr_lock, RW_READER);
if (zfsvfs->z_use_sa && zp->z_is_sa) {
error = zpl_xattr_list_sa(&xf);
if (error)
goto out;
}
error = zpl_xattr_list_dir(&xf, cr);
if (error)
goto out;
error = xf.offset;
out:
rw_exit(&zp->z_xattr_lock);
ZPL_EXIT(zfsvfs);
spl_fstrans_unmark(cookie);
crfree(cr);
return (error);
}
static int
zpl_xattr_get_dir(struct inode *ip, const char *name, void *value,
size_t size, cred_t *cr)
{
fstrans_cookie_t cookie;
struct inode *xip = NULL;
znode_t *dxzp = NULL;
znode_t *xzp = NULL;
int error;
/* Lookup the xattr directory */
error = -zfs_lookup(ITOZ(ip), NULL, &dxzp, LOOKUP_XATTR,
cr, NULL, NULL);
if (error)
goto out;
/* Lookup a specific xattr name in the directory */
error = -zfs_lookup(dxzp, (char *)name, &xzp, 0, cr, NULL, NULL);
if (error)
goto out;
xip = ZTOI(xzp);
if (!size) {
error = i_size_read(xip);
goto out;
}
if (size < i_size_read(xip)) {
error = -ERANGE;
goto out;
}
struct iovec iov;
iov.iov_base = (void *)value;
iov.iov_len = size;
zfs_uio_t uio;
zfs_uio_iovec_init(&uio, &iov, 1, 0, UIO_SYSSPACE, size, 0);
cookie = spl_fstrans_mark();
error = -zfs_read(ITOZ(xip), &uio, 0, cr);
spl_fstrans_unmark(cookie);
if (error == 0)
error = size - zfs_uio_resid(&uio);
out:
if (xzp)
zrele(xzp);
if (dxzp)
zrele(dxzp);
return (error);
}
static int
zpl_xattr_get_sa(struct inode *ip, const char *name, void *value, size_t size)
{
znode_t *zp = ITOZ(ip);
uchar_t *nv_value;
uint_t nv_size;
int error = 0;
ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
mutex_enter(&zp->z_lock);
if (zp->z_xattr_cached == NULL)
error = -zfs_sa_get_xattr(zp);
mutex_exit(&zp->z_lock);
if (error)
return (error);
ASSERT(zp->z_xattr_cached);
error = -nvlist_lookup_byte_array(zp->z_xattr_cached, name,
&nv_value, &nv_size);
if (error)
return (error);
if (size == 0 || value == NULL)
return (nv_size);
if (size < nv_size)
return (-ERANGE);
memcpy(value, nv_value, nv_size);
return (nv_size);
}
static int
__zpl_xattr_get(struct inode *ip, const char *name, void *value, size_t size,
cred_t *cr)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ZTOZSB(zp);
int error;
ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
if (zfsvfs->z_use_sa && zp->z_is_sa) {
error = zpl_xattr_get_sa(ip, name, value, size);
if (error != -ENOENT)
goto out;
}
error = zpl_xattr_get_dir(ip, name, value, size, cr);
out:
if (error == -ENOENT)
error = -ENODATA;
return (error);
}
#define XATTR_NOENT 0x0
#define XATTR_IN_SA 0x1
#define XATTR_IN_DIR 0x2
/* check where the xattr resides */
static int
__zpl_xattr_where(struct inode *ip, const char *name, int *where, cred_t *cr)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ZTOZSB(zp);
int error;
ASSERT(where);
ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
*where = XATTR_NOENT;
if (zfsvfs->z_use_sa && zp->z_is_sa) {
error = zpl_xattr_get_sa(ip, name, NULL, 0);
if (error >= 0)
*where |= XATTR_IN_SA;
else if (error != -ENOENT)
return (error);
}
error = zpl_xattr_get_dir(ip, name, NULL, 0, cr);
if (error >= 0)
*where |= XATTR_IN_DIR;
else if (error != -ENOENT)
return (error);
if (*where == (XATTR_IN_SA|XATTR_IN_DIR))
cmn_err(CE_WARN, "ZFS: inode %p has xattr \"%s\""
" in both SA and dir", ip, name);
if (*where == XATTR_NOENT)
error = -ENODATA;
else
error = 0;
return (error);
}
static int
zpl_xattr_get(struct inode *ip, const char *name, void *value, size_t size)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ZTOZSB(zp);
cred_t *cr = CRED();
fstrans_cookie_t cookie;
int error;
crhold(cr);
cookie = spl_fstrans_mark();
ZPL_ENTER(zfsvfs);
ZPL_VERIFY_ZP(zp);
rw_enter(&zp->z_xattr_lock, RW_READER);
error = __zpl_xattr_get(ip, name, value, size, cr);
rw_exit(&zp->z_xattr_lock);
ZPL_EXIT(zfsvfs);
spl_fstrans_unmark(cookie);
crfree(cr);
return (error);
}
static int
zpl_xattr_set_dir(struct inode *ip, const char *name, const void *value,
size_t size, int flags, cred_t *cr)
{
znode_t *dxzp = NULL;
znode_t *xzp = NULL;
vattr_t *vap = NULL;
int lookup_flags, error;
const int xattr_mode = S_IFREG | 0644;
loff_t pos = 0;
/*
* Lookup the xattr directory. When we're adding an entry pass
* CREATE_XATTR_DIR to ensure the xattr directory is created.
* When removing an entry this flag is not passed to avoid
* unnecessarily creating a new xattr directory.
*/
lookup_flags = LOOKUP_XATTR;
if (value != NULL)
lookup_flags |= CREATE_XATTR_DIR;
error = -zfs_lookup(ITOZ(ip), NULL, &dxzp, lookup_flags,
cr, NULL, NULL);
if (error)
goto out;
/* Lookup a specific xattr name in the directory */
error = -zfs_lookup(dxzp, (char *)name, &xzp, 0, cr, NULL, NULL);
if (error && (error != -ENOENT))
goto out;
error = 0;
/* Remove a specific name xattr when value is set to NULL. */
if (value == NULL) {
if (xzp)
error = -zfs_remove(dxzp, (char *)name, cr, 0);
goto out;
}
/* Lookup failed create a new xattr. */
if (xzp == NULL) {
vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP);
vap->va_mode = xattr_mode;
vap->va_mask = ATTR_MODE;
vap->va_uid = crgetfsuid(cr);
vap->va_gid = crgetfsgid(cr);
error = -zfs_create(dxzp, (char *)name, vap, 0, 0644, &xzp,
cr, 0, NULL);
if (error)
goto out;
}
ASSERT(xzp != NULL);
error = -zfs_freesp(xzp, 0, 0, xattr_mode, TRUE);
if (error)
goto out;
error = -zfs_write_simple(xzp, value, size, pos, NULL);
out:
if (error == 0) {
ip->i_ctime = current_time(ip);
zfs_mark_inode_dirty(ip);
}
if (vap)
kmem_free(vap, sizeof (vattr_t));
if (xzp)
zrele(xzp);
if (dxzp)
zrele(dxzp);
if (error == -ENOENT)
error = -ENODATA;
ASSERT3S(error, <=, 0);
return (error);
}
static int
zpl_xattr_set_sa(struct inode *ip, const char *name, const void *value,
size_t size, int flags, cred_t *cr)
{
znode_t *zp = ITOZ(ip);
nvlist_t *nvl;
size_t sa_size;
int error = 0;
mutex_enter(&zp->z_lock);
if (zp->z_xattr_cached == NULL)
error = -zfs_sa_get_xattr(zp);
mutex_exit(&zp->z_lock);
if (error)
return (error);
ASSERT(zp->z_xattr_cached);
nvl = zp->z_xattr_cached;
if (value == NULL) {
error = -nvlist_remove(nvl, name, DATA_TYPE_BYTE_ARRAY);
if (error == -ENOENT)
error = zpl_xattr_set_dir(ip, name, NULL, 0, flags, cr);
} else {
/* Limited to 32k to keep nvpair memory allocations small */
if (size > DXATTR_MAX_ENTRY_SIZE)
return (-EFBIG);
/* Prevent the DXATTR SA from consuming the entire SA region */
error = -nvlist_size(nvl, &sa_size, NV_ENCODE_XDR);
if (error)
return (error);
if (sa_size > DXATTR_MAX_SA_SIZE)
return (-EFBIG);
error = -nvlist_add_byte_array(nvl, name,
(uchar_t *)value, size);
}
/*
* Update the SA for additions, modifications, and removals. On
* error drop the inconsistent cached version of the nvlist, it
* will be reconstructed from the ARC when next accessed.
*/
if (error == 0)
error = -zfs_sa_set_xattr(zp);
if (error) {
nvlist_free(nvl);
zp->z_xattr_cached = NULL;
}
ASSERT3S(error, <=, 0);
return (error);
}
static int
zpl_xattr_set(struct inode *ip, const char *name, const void *value,
size_t size, int flags)
{
znode_t *zp = ITOZ(ip);
zfsvfs_t *zfsvfs = ZTOZSB(zp);
cred_t *cr = CRED();
fstrans_cookie_t cookie;
int where;
int error;
crhold(cr);
cookie = spl_fstrans_mark();
ZPL_ENTER(zfsvfs);
ZPL_VERIFY_ZP(zp);
rw_enter(&zp->z_xattr_lock, RW_WRITER);
/*
* Before setting the xattr check to see if it already exists.
* This is done to ensure the following optional flags are honored.
*
* XATTR_CREATE: fail if xattr already exists
* XATTR_REPLACE: fail if xattr does not exist
*
* We also want to know if it resides in sa or dir, so we can make
* sure we don't end up with duplicate in both places.
*/
error = __zpl_xattr_where(ip, name, &where, cr);
if (error < 0) {
if (error != -ENODATA)
goto out;
if (flags & XATTR_REPLACE)
goto out;
/* The xattr to be removed already doesn't exist */
error = 0;
if (value == NULL)
goto out;
} else {
error = -EEXIST;
if (flags & XATTR_CREATE)
goto out;
}
/* Preferentially store the xattr as a SA for better performance */
if (zfsvfs->z_use_sa && zp->z_is_sa &&
(zfsvfs->z_xattr_sa || (value == NULL && where & XATTR_IN_SA))) {
error = zpl_xattr_set_sa(ip, name, value, size, flags, cr);
if (error == 0) {
/*
* Successfully put into SA, we need to clear the one
* in dir.
*/
if (where & XATTR_IN_DIR)
zpl_xattr_set_dir(ip, name, NULL, 0, 0, cr);
goto out;
}
}
error = zpl_xattr_set_dir(ip, name, value, size, flags, cr);
/*
* Successfully put into dir, we need to clear the one in SA.
*/
if (error == 0 && (where & XATTR_IN_SA))
zpl_xattr_set_sa(ip, name, NULL, 0, 0, cr);
out:
rw_exit(&zp->z_xattr_lock);
ZPL_EXIT(zfsvfs);
spl_fstrans_unmark(cookie);
crfree(cr);
ASSERT3S(error, <=, 0);
return (error);
}
/*
* Extended user attributes
*
* "Extended user attributes may be assigned to files and directories for
* storing arbitrary additional information such as the mime type,
* character set or encoding of a file. The access permissions for user
* attributes are defined by the file permission bits: read permission
* is required to retrieve the attribute value, and writer permission is
* required to change it.
*
* The file permission bits of regular files and directories are
* interpreted differently from the file permission bits of special
* files and symbolic links. For regular files and directories the file
* permission bits define access to the file's contents, while for
* device special files they define access to the device described by
* the special file. The file permissions of symbolic links are not
* used in access checks. These differences would allow users to
* consume filesystem resources in a way not controllable by disk quotas
* for group or world writable special files and directories.
*
* For this reason, extended user attributes are allowed only for
* regular files and directories, and access to extended user attributes
* is restricted to the owner and to users with appropriate capabilities
* for directories with the sticky bit set (see the chmod(1) manual page
* for an explanation of the sticky bit)." - xattr(7)
*
* ZFS allows extended user attributes to be disabled administratively
* by setting the 'xattr=off' property on the dataset.
*/
static int
__zpl_xattr_user_list(struct inode *ip, char *list, size_t list_size,
const char *name, size_t name_len)
{
return (ITOZSB(ip)->z_flags & ZSB_XATTR);
}
ZPL_XATTR_LIST_WRAPPER(zpl_xattr_user_list);
static int
__zpl_xattr_user_get(struct inode *ip, const char *name,
void *value, size_t size)
{
char *xattr_name;
int error;
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") == 0)
return (-EINVAL);
#endif
if (!(ITOZSB(ip)->z_flags & ZSB_XATTR))
return (-EOPNOTSUPP);
xattr_name = kmem_asprintf("%s%s", XATTR_USER_PREFIX, name);
error = zpl_xattr_get(ip, xattr_name, value, size);
kmem_strfree(xattr_name);
return (error);
}
ZPL_XATTR_GET_WRAPPER(zpl_xattr_user_get);
static int
__zpl_xattr_user_set(struct inode *ip, const char *name,
const void *value, size_t size, int flags)
{
char *xattr_name;
int error;
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") == 0)
return (-EINVAL);
#endif
if (!(ITOZSB(ip)->z_flags & ZSB_XATTR))
return (-EOPNOTSUPP);
xattr_name = kmem_asprintf("%s%s", XATTR_USER_PREFIX, name);
error = zpl_xattr_set(ip, xattr_name, value, size, flags);
kmem_strfree(xattr_name);
return (error);
}
ZPL_XATTR_SET_WRAPPER(zpl_xattr_user_set);
-xattr_handler_t zpl_xattr_user_handler =
+static xattr_handler_t zpl_xattr_user_handler =
{
.prefix = XATTR_USER_PREFIX,
.list = zpl_xattr_user_list,
.get = zpl_xattr_user_get,
.set = zpl_xattr_user_set,
};
/*
* Trusted extended attributes
*
* "Trusted extended attributes are visible and accessible only to
* processes that have the CAP_SYS_ADMIN capability. Attributes in this
* class are used to implement mechanisms in user space (i.e., outside
* the kernel) which keep information in extended attributes to which
* ordinary processes should not have access." - xattr(7)
*/
static int
__zpl_xattr_trusted_list(struct inode *ip, char *list, size_t list_size,
const char *name, size_t name_len)
{
return (capable(CAP_SYS_ADMIN));
}
ZPL_XATTR_LIST_WRAPPER(zpl_xattr_trusted_list);
static int
__zpl_xattr_trusted_get(struct inode *ip, const char *name,
void *value, size_t size)
{
char *xattr_name;
int error;
if (!capable(CAP_SYS_ADMIN))
return (-EACCES);
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") == 0)
return (-EINVAL);
#endif
xattr_name = kmem_asprintf("%s%s", XATTR_TRUSTED_PREFIX, name);
error = zpl_xattr_get(ip, xattr_name, value, size);
kmem_strfree(xattr_name);
return (error);
}
ZPL_XATTR_GET_WRAPPER(zpl_xattr_trusted_get);
static int
__zpl_xattr_trusted_set(struct inode *ip, const char *name,
const void *value, size_t size, int flags)
{
char *xattr_name;
int error;
if (!capable(CAP_SYS_ADMIN))
return (-EACCES);
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") == 0)
return (-EINVAL);
#endif
xattr_name = kmem_asprintf("%s%s", XATTR_TRUSTED_PREFIX, name);
error = zpl_xattr_set(ip, xattr_name, value, size, flags);
kmem_strfree(xattr_name);
return (error);
}
ZPL_XATTR_SET_WRAPPER(zpl_xattr_trusted_set);
-xattr_handler_t zpl_xattr_trusted_handler =
-{
+static xattr_handler_t zpl_xattr_trusted_handler = {
.prefix = XATTR_TRUSTED_PREFIX,
.list = zpl_xattr_trusted_list,
.get = zpl_xattr_trusted_get,
.set = zpl_xattr_trusted_set,
};
/*
* Extended security attributes
*
* "The security attribute namespace is used by kernel security modules,
* such as Security Enhanced Linux, and also to implement file
* capabilities (see capabilities(7)). Read and write access
* permissions to security attributes depend on the policy implemented
* for each security attribute by the security module. When no security
* module is loaded, all processes have read access to extended security
* attributes, and write access is limited to processes that have the
* CAP_SYS_ADMIN capability." - xattr(7)
*/
static int
__zpl_xattr_security_list(struct inode *ip, char *list, size_t list_size,
const char *name, size_t name_len)
{
return (1);
}
ZPL_XATTR_LIST_WRAPPER(zpl_xattr_security_list);
static int
__zpl_xattr_security_get(struct inode *ip, const char *name,
void *value, size_t size)
{
char *xattr_name;
int error;
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") == 0)
return (-EINVAL);
#endif
xattr_name = kmem_asprintf("%s%s", XATTR_SECURITY_PREFIX, name);
error = zpl_xattr_get(ip, xattr_name, value, size);
kmem_strfree(xattr_name);
return (error);
}
ZPL_XATTR_GET_WRAPPER(zpl_xattr_security_get);
static int
__zpl_xattr_security_set(struct inode *ip, const char *name,
const void *value, size_t size, int flags)
{
char *xattr_name;
int error;
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") == 0)
return (-EINVAL);
#endif
xattr_name = kmem_asprintf("%s%s", XATTR_SECURITY_PREFIX, name);
error = zpl_xattr_set(ip, xattr_name, value, size, flags);
kmem_strfree(xattr_name);
return (error);
}
ZPL_XATTR_SET_WRAPPER(zpl_xattr_security_set);
static int
zpl_xattr_security_init_impl(struct inode *ip, const struct xattr *xattrs,
void *fs_info)
{
const struct xattr *xattr;
int error = 0;
for (xattr = xattrs; xattr->name != NULL; xattr++) {
error = __zpl_xattr_security_set(ip,
xattr->name, xattr->value, xattr->value_len, 0);
if (error < 0)
break;
}
return (error);
}
int
zpl_xattr_security_init(struct inode *ip, struct inode *dip,
const struct qstr *qstr)
{
return security_inode_init_security(ip, dip, qstr,
&zpl_xattr_security_init_impl, NULL);
}
/*
* Security xattr namespace handlers.
*/
-xattr_handler_t zpl_xattr_security_handler = {
+static xattr_handler_t zpl_xattr_security_handler = {
.prefix = XATTR_SECURITY_PREFIX,
.list = zpl_xattr_security_list,
.get = zpl_xattr_security_get,
.set = zpl_xattr_security_set,
};
/*
* Extended system attributes
*
* "Extended system attributes are used by the kernel to store system
* objects such as Access Control Lists. Read and write access permissions
* to system attributes depend on the policy implemented for each system
* attribute implemented by filesystems in the kernel." - xattr(7)
*/
#ifdef CONFIG_FS_POSIX_ACL
static int
zpl_set_acl_impl(struct inode *ip, struct posix_acl *acl, int type)
{
char *name, *value = NULL;
int error = 0;
size_t size = 0;
if (S_ISLNK(ip->i_mode))
return (-EOPNOTSUPP);
switch (type) {
case ACL_TYPE_ACCESS:
name = XATTR_NAME_POSIX_ACL_ACCESS;
if (acl) {
umode_t mode = ip->i_mode;
error = posix_acl_equiv_mode(acl, &mode);
if (error < 0) {
return (error);
} else {
/*
* The mode bits will have been set by
* ->zfs_setattr()->zfs_acl_chmod_setattr()
* using the ZFS ACL conversion. If they
* differ from the Posix ACL conversion dirty
* the inode to write the Posix mode bits.
*/
if (ip->i_mode != mode) {
ip->i_mode = mode;
ip->i_ctime = current_time(ip);
zfs_mark_inode_dirty(ip);
}
if (error == 0)
acl = NULL;
}
}
break;
case ACL_TYPE_DEFAULT:
name = XATTR_NAME_POSIX_ACL_DEFAULT;
if (!S_ISDIR(ip->i_mode))
return (acl ? -EACCES : 0);
break;
default:
return (-EINVAL);
}
if (acl) {
size = posix_acl_xattr_size(acl->a_count);
value = kmem_alloc(size, KM_SLEEP);
error = zpl_acl_to_xattr(acl, value, size);
if (error < 0) {
kmem_free(value, size);
return (error);
}
}
error = zpl_xattr_set(ip, name, value, size, 0);
if (value)
kmem_free(value, size);
if (!error) {
if (acl)
zpl_set_cached_acl(ip, type, acl);
else
zpl_forget_cached_acl(ip, type);
}
return (error);
}
#ifdef HAVE_SET_ACL
int
#ifdef HAVE_SET_ACL_USERNS
zpl_set_acl(struct user_namespace *userns, struct inode *ip,
struct posix_acl *acl, int type)
#else
zpl_set_acl(struct inode *ip, struct posix_acl *acl, int type)
#endif /* HAVE_SET_ACL_USERNS */
{
return (zpl_set_acl_impl(ip, acl, type));
}
#endif /* HAVE_SET_ACL */
static struct posix_acl *
zpl_get_acl_impl(struct inode *ip, int type)
{
struct posix_acl *acl;
void *value = NULL;
char *name;
/*
* As of Linux 3.14, the kernel get_acl will check this for us.
* Also as of Linux 4.7, comparing against ACL_NOT_CACHED is wrong
* as the kernel get_acl will set it to temporary sentinel value.
*/
#ifndef HAVE_KERNEL_GET_ACL_HANDLE_CACHE
acl = get_cached_acl(ip, type);
if (acl != ACL_NOT_CACHED)
return (acl);
#endif
switch (type) {
case ACL_TYPE_ACCESS:
name = XATTR_NAME_POSIX_ACL_ACCESS;
break;
case ACL_TYPE_DEFAULT:
name = XATTR_NAME_POSIX_ACL_DEFAULT;
break;
default:
return (ERR_PTR(-EINVAL));
}
int size = zpl_xattr_get(ip, name, NULL, 0);
if (size > 0) {
value = kmem_alloc(size, KM_SLEEP);
size = zpl_xattr_get(ip, name, value, size);
}
if (size > 0) {
acl = zpl_acl_from_xattr(value, size);
} else if (size == -ENODATA || size == -ENOSYS) {
acl = NULL;
} else {
acl = ERR_PTR(-EIO);
}
if (size > 0)
kmem_free(value, size);
/* As of Linux 4.7, the kernel get_acl will set this for us */
#ifndef HAVE_KERNEL_GET_ACL_HANDLE_CACHE
if (!IS_ERR(acl))
zpl_set_cached_acl(ip, type, acl);
#endif
return (acl);
}
#if defined(HAVE_GET_ACL_RCU)
struct posix_acl *
zpl_get_acl(struct inode *ip, int type, bool rcu)
{
if (rcu)
return (ERR_PTR(-ECHILD));
return (zpl_get_acl_impl(ip, type));
}
#elif defined(HAVE_GET_ACL)
struct posix_acl *
zpl_get_acl(struct inode *ip, int type)
{
return (zpl_get_acl_impl(ip, type));
}
#else
#error "Unsupported iops->get_acl() implementation"
#endif /* HAVE_GET_ACL_RCU */
int
zpl_init_acl(struct inode *ip, struct inode *dir)
{
struct posix_acl *acl = NULL;
int error = 0;
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
return (0);
if (!S_ISLNK(ip->i_mode)) {
acl = zpl_get_acl_impl(dir, ACL_TYPE_DEFAULT);
if (IS_ERR(acl))
return (PTR_ERR(acl));
if (!acl) {
ip->i_mode &= ~current_umask();
ip->i_ctime = current_time(ip);
zfs_mark_inode_dirty(ip);
return (0);
}
}
if (acl) {
umode_t mode;
if (S_ISDIR(ip->i_mode)) {
error = zpl_set_acl_impl(ip, acl, ACL_TYPE_DEFAULT);
if (error)
goto out;
}
mode = ip->i_mode;
error = __posix_acl_create(&acl, GFP_KERNEL, &mode);
if (error >= 0) {
ip->i_mode = mode;
zfs_mark_inode_dirty(ip);
if (error > 0) {
error = zpl_set_acl_impl(ip, acl,
ACL_TYPE_ACCESS);
}
}
}
out:
zpl_posix_acl_release(acl);
return (error);
}
int
zpl_chmod_acl(struct inode *ip)
{
struct posix_acl *acl;
int error;
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
return (0);
if (S_ISLNK(ip->i_mode))
return (-EOPNOTSUPP);
acl = zpl_get_acl_impl(ip, ACL_TYPE_ACCESS);
if (IS_ERR(acl) || !acl)
return (PTR_ERR(acl));
error = __posix_acl_chmod(&acl, GFP_KERNEL, ip->i_mode);
if (!error)
error = zpl_set_acl_impl(ip, acl, ACL_TYPE_ACCESS);
zpl_posix_acl_release(acl);
return (error);
}
static int
__zpl_xattr_acl_list_access(struct inode *ip, char *list, size_t list_size,
const char *name, size_t name_len)
{
char *xattr_name = XATTR_NAME_POSIX_ACL_ACCESS;
size_t xattr_size = sizeof (XATTR_NAME_POSIX_ACL_ACCESS);
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
return (0);
if (list && xattr_size <= list_size)
memcpy(list, xattr_name, xattr_size);
return (xattr_size);
}
ZPL_XATTR_LIST_WRAPPER(zpl_xattr_acl_list_access);
static int
__zpl_xattr_acl_list_default(struct inode *ip, char *list, size_t list_size,
const char *name, size_t name_len)
{
char *xattr_name = XATTR_NAME_POSIX_ACL_DEFAULT;
size_t xattr_size = sizeof (XATTR_NAME_POSIX_ACL_DEFAULT);
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
return (0);
if (list && xattr_size <= list_size)
memcpy(list, xattr_name, xattr_size);
return (xattr_size);
}
ZPL_XATTR_LIST_WRAPPER(zpl_xattr_acl_list_default);
static int
__zpl_xattr_acl_get_access(struct inode *ip, const char *name,
void *buffer, size_t size)
{
struct posix_acl *acl;
int type = ACL_TYPE_ACCESS;
int error;
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") != 0)
return (-EINVAL);
#endif
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
return (-EOPNOTSUPP);
acl = zpl_get_acl_impl(ip, type);
if (IS_ERR(acl))
return (PTR_ERR(acl));
if (acl == NULL)
return (-ENODATA);
error = zpl_acl_to_xattr(acl, buffer, size);
zpl_posix_acl_release(acl);
return (error);
}
ZPL_XATTR_GET_WRAPPER(zpl_xattr_acl_get_access);
static int
__zpl_xattr_acl_get_default(struct inode *ip, const char *name,
void *buffer, size_t size)
{
struct posix_acl *acl;
int type = ACL_TYPE_DEFAULT;
int error;
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") != 0)
return (-EINVAL);
#endif
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
return (-EOPNOTSUPP);
acl = zpl_get_acl_impl(ip, type);
if (IS_ERR(acl))
return (PTR_ERR(acl));
if (acl == NULL)
return (-ENODATA);
error = zpl_acl_to_xattr(acl, buffer, size);
zpl_posix_acl_release(acl);
return (error);
}
ZPL_XATTR_GET_WRAPPER(zpl_xattr_acl_get_default);
static int
__zpl_xattr_acl_set_access(struct inode *ip, const char *name,
const void *value, size_t size, int flags)
{
struct posix_acl *acl;
int type = ACL_TYPE_ACCESS;
int error = 0;
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") != 0)
return (-EINVAL);
#endif
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
return (-EOPNOTSUPP);
if (!zpl_inode_owner_or_capable(kcred->user_ns, ip))
return (-EPERM);
if (value) {
acl = zpl_acl_from_xattr(value, size);
if (IS_ERR(acl))
return (PTR_ERR(acl));
else if (acl) {
error = zpl_posix_acl_valid(ip, acl);
if (error) {
zpl_posix_acl_release(acl);
return (error);
}
}
} else {
acl = NULL;
}
error = zpl_set_acl_impl(ip, acl, type);
zpl_posix_acl_release(acl);
return (error);
}
ZPL_XATTR_SET_WRAPPER(zpl_xattr_acl_set_access);
static int
__zpl_xattr_acl_set_default(struct inode *ip, const char *name,
const void *value, size_t size, int flags)
{
struct posix_acl *acl;
int type = ACL_TYPE_DEFAULT;
int error = 0;
/* xattr_resolve_name will do this for us if this is defined */
#ifndef HAVE_XATTR_HANDLER_NAME
if (strcmp(name, "") != 0)
return (-EINVAL);
#endif
if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIX)
return (-EOPNOTSUPP);
if (!zpl_inode_owner_or_capable(kcred->user_ns, ip))
return (-EPERM);
if (value) {
acl = zpl_acl_from_xattr(value, size);
if (IS_ERR(acl))
return (PTR_ERR(acl));
else if (acl) {
error = zpl_posix_acl_valid(ip, acl);
if (error) {
zpl_posix_acl_release(acl);
return (error);
}
}
} else {
acl = NULL;
}
error = zpl_set_acl_impl(ip, acl, type);
zpl_posix_acl_release(acl);
return (error);
}
ZPL_XATTR_SET_WRAPPER(zpl_xattr_acl_set_default);
/*
* ACL access xattr namespace handlers.
*
* Use .name instead of .prefix when available. xattr_resolve_name will match
* whole name and reject anything that has .name only as prefix.
*/
-xattr_handler_t zpl_xattr_acl_access_handler =
-{
+static xattr_handler_t zpl_xattr_acl_access_handler = {
#ifdef HAVE_XATTR_HANDLER_NAME
.name = XATTR_NAME_POSIX_ACL_ACCESS,
#else
.prefix = XATTR_NAME_POSIX_ACL_ACCESS,
#endif
.list = zpl_xattr_acl_list_access,
.get = zpl_xattr_acl_get_access,
.set = zpl_xattr_acl_set_access,
#if defined(HAVE_XATTR_LIST_SIMPLE) || \
defined(HAVE_XATTR_LIST_DENTRY) || \
defined(HAVE_XATTR_LIST_HANDLER)
.flags = ACL_TYPE_ACCESS,
#endif
};
/*
* ACL default xattr namespace handlers.
*
* Use .name instead of .prefix when available. xattr_resolve_name will match
* whole name and reject anything that has .name only as prefix.
*/
-xattr_handler_t zpl_xattr_acl_default_handler =
-{
+static xattr_handler_t zpl_xattr_acl_default_handler = {
#ifdef HAVE_XATTR_HANDLER_NAME
.name = XATTR_NAME_POSIX_ACL_DEFAULT,
#else
.prefix = XATTR_NAME_POSIX_ACL_DEFAULT,
#endif
.list = zpl_xattr_acl_list_default,
.get = zpl_xattr_acl_get_default,
.set = zpl_xattr_acl_set_default,
#if defined(HAVE_XATTR_LIST_SIMPLE) || \
defined(HAVE_XATTR_LIST_DENTRY) || \
defined(HAVE_XATTR_LIST_HANDLER)
.flags = ACL_TYPE_DEFAULT,
#endif
};
#endif /* CONFIG_FS_POSIX_ACL */
xattr_handler_t *zpl_xattr_handlers[] = {
&zpl_xattr_security_handler,
&zpl_xattr_trusted_handler,
&zpl_xattr_user_handler,
#ifdef CONFIG_FS_POSIX_ACL
&zpl_xattr_acl_access_handler,
&zpl_xattr_acl_default_handler,
#endif /* CONFIG_FS_POSIX_ACL */
NULL
};
static const struct xattr_handler *
zpl_xattr_handler(const char *name)
{
if (strncmp(name, XATTR_USER_PREFIX,
XATTR_USER_PREFIX_LEN) == 0)
return (&zpl_xattr_user_handler);
if (strncmp(name, XATTR_TRUSTED_PREFIX,
XATTR_TRUSTED_PREFIX_LEN) == 0)
return (&zpl_xattr_trusted_handler);
if (strncmp(name, XATTR_SECURITY_PREFIX,
XATTR_SECURITY_PREFIX_LEN) == 0)
return (&zpl_xattr_security_handler);
#ifdef CONFIG_FS_POSIX_ACL
if (strncmp(name, XATTR_NAME_POSIX_ACL_ACCESS,
sizeof (XATTR_NAME_POSIX_ACL_ACCESS)) == 0)
return (&zpl_xattr_acl_access_handler);
if (strncmp(name, XATTR_NAME_POSIX_ACL_DEFAULT,
sizeof (XATTR_NAME_POSIX_ACL_DEFAULT)) == 0)
return (&zpl_xattr_acl_default_handler);
#endif /* CONFIG_FS_POSIX_ACL */
return (NULL);
}
#if !defined(HAVE_POSIX_ACL_RELEASE) || defined(HAVE_POSIX_ACL_RELEASE_GPL_ONLY)
struct acl_rel_struct {
struct acl_rel_struct *next;
struct posix_acl *acl;
clock_t time;
};
#define ACL_REL_GRACE (60*HZ)
#define ACL_REL_WINDOW (1*HZ)
#define ACL_REL_SCHED (ACL_REL_GRACE+ACL_REL_WINDOW)
/*
* Lockless multi-producer single-consumer fifo list.
* Nodes are added to tail and removed from head. Tail pointer is our
* synchronization point. It always points to the next pointer of the last
* node, or head if list is empty.
*/
static struct acl_rel_struct *acl_rel_head = NULL;
static struct acl_rel_struct **acl_rel_tail = &acl_rel_head;
static void
zpl_posix_acl_free(void *arg)
{
struct acl_rel_struct *freelist = NULL;
struct acl_rel_struct *a;
clock_t new_time;
boolean_t refire = B_FALSE;
ASSERT3P(acl_rel_head, !=, NULL);
while (acl_rel_head) {
a = acl_rel_head;
if (ddi_get_lbolt() - a->time >= ACL_REL_GRACE) {
/*
* If a is the last node we need to reset tail, but we
* need to use cmpxchg to make sure it is still the
* last node.
*/
if (acl_rel_tail == &a->next) {
acl_rel_head = NULL;
if (cmpxchg(&acl_rel_tail, &a->next,
&acl_rel_head) == &a->next) {
ASSERT3P(a->next, ==, NULL);
a->next = freelist;
freelist = a;
break;
}
}
/*
* a is not last node, make sure next pointer is set
* by the adder and advance the head.
*/
while (READ_ONCE(a->next) == NULL)
cpu_relax();
acl_rel_head = a->next;
a->next = freelist;
freelist = a;
} else {
/*
* a is still in grace period. We are responsible to
* reschedule the free task, since adder will only do
* so if list is empty.
*/
new_time = a->time + ACL_REL_SCHED;
refire = B_TRUE;
break;
}
}
if (refire)
taskq_dispatch_delay(system_delay_taskq, zpl_posix_acl_free,
NULL, TQ_SLEEP, new_time);
while (freelist) {
a = freelist;
freelist = a->next;
kfree(a->acl);
kmem_free(a, sizeof (struct acl_rel_struct));
}
}
void
zpl_posix_acl_release_impl(struct posix_acl *acl)
{
struct acl_rel_struct *a, **prev;
a = kmem_alloc(sizeof (struct acl_rel_struct), KM_SLEEP);
a->next = NULL;
a->acl = acl;
a->time = ddi_get_lbolt();
/* atomically points tail to us and get the previous tail */
prev = xchg(&acl_rel_tail, &a->next);
ASSERT3P(*prev, ==, NULL);
*prev = a;
/* if it was empty before, schedule the free task */
if (prev == &acl_rel_head)
taskq_dispatch_delay(system_delay_taskq, zpl_posix_acl_free,
NULL, TQ_SLEEP, ddi_get_lbolt() + ACL_REL_SCHED);
}
#endif
diff --git a/module/os/linux/zfs/zvol_os.c b/module/os/linux/zfs/zvol_os.c
index 69479b3f7988..cef52e22483a 100644
--- a/module/os/linux/zfs/zvol_os.c
+++ b/module/os/linux/zfs/zvol_os.c
@@ -1,1226 +1,1226 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
*/
#include <sys/dataset_kstats.h>
#include <sys/dbuf.h>
#include <sys/dmu_traverse.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_dir.h>
#include <sys/zap.h>
#include <sys/zfeature.h>
#include <sys/zil_impl.h>
#include <sys/dmu_tx.h>
#include <sys/zio.h>
#include <sys/zfs_rlock.h>
#include <sys/spa_impl.h>
#include <sys/zvol.h>
#include <sys/zvol_impl.h>
#include <linux/blkdev_compat.h>
#include <linux/task_io_accounting_ops.h>
-unsigned int zvol_major = ZVOL_MAJOR;
-unsigned int zvol_request_sync = 0;
-unsigned int zvol_prefetch_bytes = (128 * 1024);
-unsigned long zvol_max_discard_blocks = 16384;
-unsigned int zvol_threads = 32;
-unsigned int zvol_open_timeout_ms = 1000;
+static unsigned int zvol_major = ZVOL_MAJOR;
+static unsigned int zvol_request_sync = 0;
+static unsigned int zvol_prefetch_bytes = (128 * 1024);
+static unsigned long zvol_max_discard_blocks = 16384;
+static unsigned int zvol_threads = 32;
+static const unsigned int zvol_open_timeout_ms = 1000;
struct zvol_state_os {
struct gendisk *zvo_disk; /* generic disk */
struct request_queue *zvo_queue; /* request queue */
dev_t zvo_dev; /* device id */
};
taskq_t *zvol_taskq;
static struct ida zvol_ida;
typedef struct zv_request_stack {
zvol_state_t *zv;
struct bio *bio;
} zv_request_t;
typedef struct zv_request_task {
zv_request_t zvr;
taskq_ent_t ent;
} zv_request_task_t;
static zv_request_task_t *
zv_request_task_create(zv_request_t zvr)
{
zv_request_task_t *task;
task = kmem_alloc(sizeof (zv_request_task_t), KM_SLEEP);
taskq_init_ent(&task->ent);
task->zvr = zvr;
return (task);
}
static void
zv_request_task_free(zv_request_task_t *task)
{
kmem_free(task, sizeof (*task));
}
/*
* Given a path, return TRUE if path is a ZVOL.
*/
static boolean_t
zvol_is_zvol_impl(const char *path)
{
dev_t dev = 0;
if (vdev_lookup_bdev(path, &dev) != 0)
return (B_FALSE);
if (MAJOR(dev) == zvol_major)
return (B_TRUE);
return (B_FALSE);
}
static void
zvol_write(zv_request_t *zvr)
{
struct bio *bio = zvr->bio;
int error = 0;
zfs_uio_t uio;
zfs_uio_bvec_init(&uio, bio);
zvol_state_t *zv = zvr->zv;
ASSERT3P(zv, !=, NULL);
ASSERT3U(zv->zv_open_count, >, 0);
ASSERT3P(zv->zv_zilog, !=, NULL);
/* bio marked as FLUSH need to flush before write */
if (bio_is_flush(bio))
zil_commit(zv->zv_zilog, ZVOL_OBJ);
/* Some requests are just for flush and nothing else. */
if (uio.uio_resid == 0) {
rw_exit(&zv->zv_suspend_lock);
BIO_END_IO(bio, 0);
return;
}
struct request_queue *q = zv->zv_zso->zvo_queue;
struct gendisk *disk = zv->zv_zso->zvo_disk;
ssize_t start_resid = uio.uio_resid;
unsigned long start_time;
boolean_t acct = blk_queue_io_stat(q);
if (acct)
start_time = blk_generic_start_io_acct(q, disk, WRITE, bio);
boolean_t sync =
bio_is_fua(bio) || zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS;
zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock,
uio.uio_loffset, uio.uio_resid, RL_WRITER);
uint64_t volsize = zv->zv_volsize;
while (uio.uio_resid > 0 && uio.uio_loffset < volsize) {
uint64_t bytes = MIN(uio.uio_resid, DMU_MAX_ACCESS >> 1);
uint64_t off = uio.uio_loffset;
dmu_tx_t *tx = dmu_tx_create(zv->zv_objset);
if (bytes > volsize - off) /* don't write past the end */
bytes = volsize - off;
dmu_tx_hold_write_by_dnode(tx, zv->zv_dn, off, bytes);
/* This will only fail for ENOSPC */
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
break;
}
error = dmu_write_uio_dnode(zv->zv_dn, &uio, bytes, tx);
if (error == 0) {
zvol_log_write(zv, tx, off, bytes, sync);
}
dmu_tx_commit(tx);
if (error)
break;
}
zfs_rangelock_exit(lr);
int64_t nwritten = start_resid - uio.uio_resid;
dataset_kstats_update_write_kstats(&zv->zv_kstat, nwritten);
task_io_account_write(nwritten);
if (sync)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
rw_exit(&zv->zv_suspend_lock);
if (acct)
blk_generic_end_io_acct(q, disk, WRITE, bio, start_time);
BIO_END_IO(bio, -error);
}
static void
zvol_write_task(void *arg)
{
zv_request_task_t *task = arg;
zvol_write(&task->zvr);
zv_request_task_free(task);
}
static void
zvol_discard(zv_request_t *zvr)
{
struct bio *bio = zvr->bio;
zvol_state_t *zv = zvr->zv;
uint64_t start = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
uint64_t end = start + size;
boolean_t sync;
int error = 0;
dmu_tx_t *tx;
ASSERT3P(zv, !=, NULL);
ASSERT3U(zv->zv_open_count, >, 0);
ASSERT3P(zv->zv_zilog, !=, NULL);
struct request_queue *q = zv->zv_zso->zvo_queue;
struct gendisk *disk = zv->zv_zso->zvo_disk;
unsigned long start_time;
boolean_t acct = blk_queue_io_stat(q);
if (acct)
start_time = blk_generic_start_io_acct(q, disk, WRITE, bio);
sync = bio_is_fua(bio) || zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS;
if (end > zv->zv_volsize) {
error = SET_ERROR(EIO);
goto unlock;
}
/*
* Align the request to volume block boundaries when a secure erase is
* not required. This will prevent dnode_free_range() from zeroing out
* the unaligned parts which is slow (read-modify-write) and useless
* since we are not freeing any space by doing so.
*/
if (!bio_is_secure_erase(bio)) {
start = P2ROUNDUP(start, zv->zv_volblocksize);
end = P2ALIGN(end, zv->zv_volblocksize);
size = end - start;
}
if (start >= end)
goto unlock;
zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock,
start, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error != 0) {
dmu_tx_abort(tx);
} else {
zvol_log_truncate(zv, tx, start, size, B_TRUE);
dmu_tx_commit(tx);
error = dmu_free_long_range(zv->zv_objset,
ZVOL_OBJ, start, size);
}
zfs_rangelock_exit(lr);
if (error == 0 && sync)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
unlock:
rw_exit(&zv->zv_suspend_lock);
if (acct)
blk_generic_end_io_acct(q, disk, WRITE, bio, start_time);
BIO_END_IO(bio, -error);
}
static void
zvol_discard_task(void *arg)
{
zv_request_task_t *task = arg;
zvol_discard(&task->zvr);
zv_request_task_free(task);
}
static void
zvol_read(zv_request_t *zvr)
{
struct bio *bio = zvr->bio;
int error = 0;
zfs_uio_t uio;
zfs_uio_bvec_init(&uio, bio);
zvol_state_t *zv = zvr->zv;
ASSERT3P(zv, !=, NULL);
ASSERT3U(zv->zv_open_count, >, 0);
struct request_queue *q = zv->zv_zso->zvo_queue;
struct gendisk *disk = zv->zv_zso->zvo_disk;
ssize_t start_resid = uio.uio_resid;
unsigned long start_time;
boolean_t acct = blk_queue_io_stat(q);
if (acct)
start_time = blk_generic_start_io_acct(q, disk, READ, bio);
zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock,
uio.uio_loffset, uio.uio_resid, RL_READER);
uint64_t volsize = zv->zv_volsize;
while (uio.uio_resid > 0 && uio.uio_loffset < volsize) {
uint64_t bytes = MIN(uio.uio_resid, DMU_MAX_ACCESS >> 1);
/* don't read past the end */
if (bytes > volsize - uio.uio_loffset)
bytes = volsize - uio.uio_loffset;
error = dmu_read_uio_dnode(zv->zv_dn, &uio, bytes);
if (error) {
/* convert checksum errors into IO errors */
if (error == ECKSUM)
error = SET_ERROR(EIO);
break;
}
}
zfs_rangelock_exit(lr);
int64_t nread = start_resid - uio.uio_resid;
dataset_kstats_update_read_kstats(&zv->zv_kstat, nread);
task_io_account_read(nread);
rw_exit(&zv->zv_suspend_lock);
if (acct)
blk_generic_end_io_acct(q, disk, READ, bio, start_time);
BIO_END_IO(bio, -error);
}
static void
zvol_read_task(void *arg)
{
zv_request_task_t *task = arg;
zvol_read(&task->zvr);
zv_request_task_free(task);
}
#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
#ifdef HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID
static void
zvol_submit_bio(struct bio *bio)
#else
static blk_qc_t
zvol_submit_bio(struct bio *bio)
#endif
#else
static MAKE_REQUEST_FN_RET
zvol_request(struct request_queue *q, struct bio *bio)
#endif
{
#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
#if defined(HAVE_BIO_BDEV_DISK)
struct request_queue *q = bio->bi_bdev->bd_disk->queue;
#else
struct request_queue *q = bio->bi_disk->queue;
#endif
#endif
zvol_state_t *zv = q->queuedata;
fstrans_cookie_t cookie = spl_fstrans_mark();
uint64_t offset = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
int rw = bio_data_dir(bio);
if (bio_has_data(bio) && offset + size > zv->zv_volsize) {
printk(KERN_INFO
"%s: bad access: offset=%llu, size=%lu\n",
zv->zv_zso->zvo_disk->disk_name,
(long long unsigned)offset,
(long unsigned)size);
BIO_END_IO(bio, -SET_ERROR(EIO));
goto out;
}
zv_request_t zvr = {
.zv = zv,
.bio = bio,
};
zv_request_task_t *task;
if (rw == WRITE) {
if (unlikely(zv->zv_flags & ZVOL_RDONLY)) {
BIO_END_IO(bio, -SET_ERROR(EROFS));
goto out;
}
/*
* Prevents the zvol from being suspended, or the ZIL being
* concurrently opened. Will be released after the i/o
* completes.
*/
rw_enter(&zv->zv_suspend_lock, RW_READER);
/*
* Open a ZIL if this is the first time we have written to this
* zvol. We protect zv->zv_zilog with zv_suspend_lock rather
* than zv_state_lock so that we don't need to acquire an
* additional lock in this path.
*/
if (zv->zv_zilog == NULL) {
rw_exit(&zv->zv_suspend_lock);
rw_enter(&zv->zv_suspend_lock, RW_WRITER);
if (zv->zv_zilog == NULL) {
zv->zv_zilog = zil_open(zv->zv_objset,
zvol_get_data);
zv->zv_flags |= ZVOL_WRITTEN_TO;
/* replay / destroy done in zvol_create_minor */
VERIFY0((zv->zv_zilog->zl_header->zh_flags &
ZIL_REPLAY_NEEDED));
}
rw_downgrade(&zv->zv_suspend_lock);
}
/*
* We don't want this thread to be blocked waiting for i/o to
* complete, so we instead wait from a taskq callback. The
* i/o may be a ZIL write (via zil_commit()), or a read of an
* indirect block, or a read of a data block (if this is a
* partial-block write). We will indicate that the i/o is
* complete by calling BIO_END_IO() from the taskq callback.
*
* This design allows the calling thread to continue and
* initiate more concurrent operations by calling
* zvol_request() again. There are typically only a small
* number of threads available to call zvol_request() (e.g.
* one per iSCSI target), so keeping the latency of
* zvol_request() low is important for performance.
*
* The zvol_request_sync module parameter allows this
* behavior to be altered, for performance evaluation
* purposes. If the callback blocks, setting
* zvol_request_sync=1 will result in much worse performance.
*
* We can have up to zvol_threads concurrent i/o's being
* processed for all zvols on the system. This is typically
* a vast improvement over the zvol_request_sync=1 behavior
* of one i/o at a time per zvol. However, an even better
* design would be for zvol_request() to initiate the zio
* directly, and then be notified by the zio_done callback,
* which would call BIO_END_IO(). Unfortunately, the DMU/ZIL
* interfaces lack this functionality (they block waiting for
* the i/o to complete).
*/
if (bio_is_discard(bio) || bio_is_secure_erase(bio)) {
if (zvol_request_sync) {
zvol_discard(&zvr);
} else {
task = zv_request_task_create(zvr);
taskq_dispatch_ent(zvol_taskq,
zvol_discard_task, task, 0, &task->ent);
}
} else {
if (zvol_request_sync) {
zvol_write(&zvr);
} else {
task = zv_request_task_create(zvr);
taskq_dispatch_ent(zvol_taskq,
zvol_write_task, task, 0, &task->ent);
}
}
} else {
/*
* The SCST driver, and possibly others, may issue READ I/Os
* with a length of zero bytes. These empty I/Os contain no
* data and require no additional handling.
*/
if (size == 0) {
BIO_END_IO(bio, 0);
goto out;
}
rw_enter(&zv->zv_suspend_lock, RW_READER);
/* See comment in WRITE case above. */
if (zvol_request_sync) {
zvol_read(&zvr);
} else {
task = zv_request_task_create(zvr);
taskq_dispatch_ent(zvol_taskq,
zvol_read_task, task, 0, &task->ent);
}
}
out:
spl_fstrans_unmark(cookie);
#if (defined(HAVE_MAKE_REQUEST_FN_RET_QC) || \
defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS)) && \
!defined(HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID)
return (BLK_QC_T_NONE);
#endif
}
static int
zvol_open(struct block_device *bdev, fmode_t flag)
{
zvol_state_t *zv;
int error = 0;
boolean_t drop_suspend = B_FALSE;
#ifndef HAVE_BLKDEV_GET_ERESTARTSYS
hrtime_t timeout = MSEC2NSEC(zvol_open_timeout_ms);
hrtime_t start = gethrtime();
retry:
#endif
rw_enter(&zvol_state_lock, RW_READER);
/*
* Obtain a copy of private_data under the zvol_state_lock to make
* sure that either the result of zvol free code path setting
* bdev->bd_disk->private_data to NULL is observed, or zvol_free()
* is not called on this zv because of the positive zv_open_count.
*/
zv = bdev->bd_disk->private_data;
if (zv == NULL) {
rw_exit(&zvol_state_lock);
return (SET_ERROR(-ENXIO));
}
mutex_enter(&zv->zv_state_lock);
/*
* Make sure zvol is not suspended during first open
* (hold zv_suspend_lock) and respect proper lock acquisition
* ordering - zv_suspend_lock before zv_state_lock
*/
if (zv->zv_open_count == 0) {
if (!rw_tryenter(&zv->zv_suspend_lock, RW_READER)) {
mutex_exit(&zv->zv_state_lock);
rw_enter(&zv->zv_suspend_lock, RW_READER);
mutex_enter(&zv->zv_state_lock);
/* check to see if zv_suspend_lock is needed */
if (zv->zv_open_count != 0) {
rw_exit(&zv->zv_suspend_lock);
} else {
drop_suspend = B_TRUE;
}
} else {
drop_suspend = B_TRUE;
}
}
rw_exit(&zvol_state_lock);
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
if (zv->zv_open_count == 0) {
boolean_t drop_namespace = B_FALSE;
ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));
/*
* In all other call paths the spa_namespace_lock is taken
* before the bdev->bd_mutex lock. However, on open(2)
* the __blkdev_get() function calls fops->open() with the
* bdev->bd_mutex lock held. This can result in a deadlock
* when zvols from one pool are used as vdevs in another.
*
* To prevent a lock inversion deadlock we preemptively
* take the spa_namespace_lock. Normally the lock will not
* be contended and this is safe because spa_open_common()
* handles the case where the caller already holds the
* spa_namespace_lock.
*
* When the lock cannot be aquired after multiple retries
* this must be the vdev on zvol deadlock case and we have
* no choice but to return an error. For 5.12 and older
* kernels returning -ERESTARTSYS will result in the
* bdev->bd_mutex being dropped, then reacquired, and
* fops->open() being called again. This process can be
* repeated safely until both locks are acquired. For 5.13
* and newer the -ERESTARTSYS retry logic was removed from
* the kernel so the only option is to return the error for
* the caller to handle it.
*/
if (!mutex_owned(&spa_namespace_lock)) {
if (!mutex_tryenter(&spa_namespace_lock)) {
mutex_exit(&zv->zv_state_lock);
rw_exit(&zv->zv_suspend_lock);
#ifdef HAVE_BLKDEV_GET_ERESTARTSYS
schedule();
return (SET_ERROR(-ERESTARTSYS));
#else
if ((gethrtime() - start) > timeout)
return (SET_ERROR(-ERESTARTSYS));
schedule_timeout(MSEC_TO_TICK(10));
goto retry;
#endif
} else {
drop_namespace = B_TRUE;
}
}
error = -zvol_first_open(zv, !(flag & FMODE_WRITE));
if (drop_namespace)
mutex_exit(&spa_namespace_lock);
}
if (error == 0) {
if ((flag & FMODE_WRITE) && (zv->zv_flags & ZVOL_RDONLY)) {
if (zv->zv_open_count == 0)
zvol_last_close(zv);
error = SET_ERROR(-EROFS);
} else {
zv->zv_open_count++;
}
}
mutex_exit(&zv->zv_state_lock);
if (drop_suspend)
rw_exit(&zv->zv_suspend_lock);
if (error == 0)
zfs_check_media_change(bdev);
return (error);
}
static void
zvol_release(struct gendisk *disk, fmode_t mode)
{
zvol_state_t *zv;
boolean_t drop_suspend = B_TRUE;
rw_enter(&zvol_state_lock, RW_READER);
zv = disk->private_data;
mutex_enter(&zv->zv_state_lock);
ASSERT3U(zv->zv_open_count, >, 0);
/*
* make sure zvol is not suspended during last close
* (hold zv_suspend_lock) and respect proper lock acquisition
* ordering - zv_suspend_lock before zv_state_lock
*/
if (zv->zv_open_count == 1) {
if (!rw_tryenter(&zv->zv_suspend_lock, RW_READER)) {
mutex_exit(&zv->zv_state_lock);
rw_enter(&zv->zv_suspend_lock, RW_READER);
mutex_enter(&zv->zv_state_lock);
/* check to see if zv_suspend_lock is needed */
if (zv->zv_open_count != 1) {
rw_exit(&zv->zv_suspend_lock);
drop_suspend = B_FALSE;
}
}
} else {
drop_suspend = B_FALSE;
}
rw_exit(&zvol_state_lock);
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
zv->zv_open_count--;
if (zv->zv_open_count == 0) {
ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));
zvol_last_close(zv);
}
mutex_exit(&zv->zv_state_lock);
if (drop_suspend)
rw_exit(&zv->zv_suspend_lock);
}
static int
zvol_ioctl(struct block_device *bdev, fmode_t mode,
unsigned int cmd, unsigned long arg)
{
zvol_state_t *zv = bdev->bd_disk->private_data;
int error = 0;
ASSERT3U(zv->zv_open_count, >, 0);
switch (cmd) {
case BLKFLSBUF:
fsync_bdev(bdev);
invalidate_bdev(bdev);
rw_enter(&zv->zv_suspend_lock, RW_READER);
if (!(zv->zv_flags & ZVOL_RDONLY))
txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);
rw_exit(&zv->zv_suspend_lock);
break;
case BLKZNAME:
mutex_enter(&zv->zv_state_lock);
error = copy_to_user((void *)arg, zv->zv_name, MAXNAMELEN);
mutex_exit(&zv->zv_state_lock);
break;
default:
error = -ENOTTY;
break;
}
return (SET_ERROR(error));
}
#ifdef CONFIG_COMPAT
static int
zvol_compat_ioctl(struct block_device *bdev, fmode_t mode,
unsigned cmd, unsigned long arg)
{
return (zvol_ioctl(bdev, mode, cmd, arg));
}
#else
#define zvol_compat_ioctl NULL
#endif
static unsigned int
zvol_check_events(struct gendisk *disk, unsigned int clearing)
{
unsigned int mask = 0;
rw_enter(&zvol_state_lock, RW_READER);
zvol_state_t *zv = disk->private_data;
if (zv != NULL) {
mutex_enter(&zv->zv_state_lock);
mask = zv->zv_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
zv->zv_changed = 0;
mutex_exit(&zv->zv_state_lock);
}
rw_exit(&zvol_state_lock);
return (mask);
}
static int
zvol_revalidate_disk(struct gendisk *disk)
{
rw_enter(&zvol_state_lock, RW_READER);
zvol_state_t *zv = disk->private_data;
if (zv != NULL) {
mutex_enter(&zv->zv_state_lock);
set_capacity(zv->zv_zso->zvo_disk,
zv->zv_volsize >> SECTOR_BITS);
mutex_exit(&zv->zv_state_lock);
}
rw_exit(&zvol_state_lock);
return (0);
}
static int
zvol_update_volsize(zvol_state_t *zv, uint64_t volsize)
{
struct gendisk *disk = zv->zv_zso->zvo_disk;
#if defined(HAVE_REVALIDATE_DISK_SIZE)
revalidate_disk_size(disk, zvol_revalidate_disk(disk) == 0);
#elif defined(HAVE_REVALIDATE_DISK)
revalidate_disk(disk);
#else
zvol_revalidate_disk(disk);
#endif
return (0);
}
static void
zvol_clear_private(zvol_state_t *zv)
{
/*
* Cleared while holding zvol_state_lock as a writer
* which will prevent zvol_open() from opening it.
*/
zv->zv_zso->zvo_disk->private_data = NULL;
}
/*
* Provide a simple virtual geometry for legacy compatibility. For devices
* smaller than 1 MiB a small head and sector count is used to allow very
* tiny devices. For devices over 1 Mib a standard head and sector count
* is used to keep the cylinders count reasonable.
*/
static int
zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo)
{
zvol_state_t *zv = bdev->bd_disk->private_data;
sector_t sectors;
ASSERT3U(zv->zv_open_count, >, 0);
sectors = get_capacity(zv->zv_zso->zvo_disk);
if (sectors > 2048) {
geo->heads = 16;
geo->sectors = 63;
} else {
geo->heads = 2;
geo->sectors = 4;
}
geo->start = 0;
geo->cylinders = sectors / (geo->heads * geo->sectors);
return (0);
}
-static struct block_device_operations zvol_ops = {
+static const struct block_device_operations zvol_ops = {
.open = zvol_open,
.release = zvol_release,
.ioctl = zvol_ioctl,
.compat_ioctl = zvol_compat_ioctl,
.check_events = zvol_check_events,
#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK
.revalidate_disk = zvol_revalidate_disk,
#endif
.getgeo = zvol_getgeo,
.owner = THIS_MODULE,
#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
.submit_bio = zvol_submit_bio,
#endif
};
/*
* Allocate memory for a new zvol_state_t and setup the required
* request queue and generic disk structures for the block device.
*/
static zvol_state_t *
zvol_alloc(dev_t dev, const char *name)
{
zvol_state_t *zv;
struct zvol_state_os *zso;
uint64_t volmode;
if (dsl_prop_get_integer(name, "volmode", &volmode, NULL) != 0)
return (NULL);
if (volmode == ZFS_VOLMODE_DEFAULT)
volmode = zvol_volmode;
if (volmode == ZFS_VOLMODE_NONE)
return (NULL);
zv = kmem_zalloc(sizeof (zvol_state_t), KM_SLEEP);
zso = kmem_zalloc(sizeof (struct zvol_state_os), KM_SLEEP);
zv->zv_zso = zso;
zv->zv_volmode = volmode;
list_link_init(&zv->zv_next);
mutex_init(&zv->zv_state_lock, NULL, MUTEX_DEFAULT, NULL);
#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
#ifdef HAVE_BLK_ALLOC_DISK
zso->zvo_disk = blk_alloc_disk(NUMA_NO_NODE);
if (zso->zvo_disk == NULL)
goto out_kmem;
zso->zvo_disk->minors = ZVOL_MINORS;
zso->zvo_queue = zso->zvo_disk->queue;
#else
zso->zvo_queue = blk_alloc_queue(NUMA_NO_NODE);
if (zso->zvo_queue == NULL)
goto out_kmem;
zso->zvo_disk = alloc_disk(ZVOL_MINORS);
if (zso->zvo_disk == NULL) {
blk_cleanup_queue(zso->zvo_queue);
goto out_kmem;
}
zso->zvo_disk->queue = zso->zvo_queue;
#endif /* HAVE_BLK_ALLOC_DISK */
#else
zso->zvo_queue = blk_generic_alloc_queue(zvol_request, NUMA_NO_NODE);
if (zso->zvo_queue == NULL)
goto out_kmem;
zso->zvo_disk = alloc_disk(ZVOL_MINORS);
if (zso->zvo_disk == NULL) {
blk_cleanup_queue(zso->zvo_queue);
goto out_kmem;
}
zso->zvo_disk->queue = zso->zvo_queue;
#endif /* HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */
blk_queue_set_write_cache(zso->zvo_queue, B_TRUE, B_TRUE);
/* Limit read-ahead to a single page to prevent over-prefetching. */
blk_queue_set_read_ahead(zso->zvo_queue, 1);
/* Disable write merging in favor of the ZIO pipeline. */
blk_queue_flag_set(QUEUE_FLAG_NOMERGES, zso->zvo_queue);
/* Enable /proc/diskstats */
blk_queue_flag_set(QUEUE_FLAG_IO_STAT, zso->zvo_queue);
zso->zvo_queue->queuedata = zv;
zso->zvo_dev = dev;
zv->zv_open_count = 0;
strlcpy(zv->zv_name, name, MAXNAMELEN);
zfs_rangelock_init(&zv->zv_rangelock, NULL, NULL);
rw_init(&zv->zv_suspend_lock, NULL, RW_DEFAULT, NULL);
zso->zvo_disk->major = zvol_major;
zso->zvo_disk->events = DISK_EVENT_MEDIA_CHANGE;
if (volmode == ZFS_VOLMODE_DEV) {
/*
* ZFS_VOLMODE_DEV disable partitioning on ZVOL devices: set
* gendisk->minors = 1 as noted in include/linux/genhd.h.
* Also disable extended partition numbers (GENHD_FL_EXT_DEVT)
* and suppresses partition scanning (GENHD_FL_NO_PART_SCAN)
* setting gendisk->flags accordingly.
*/
zso->zvo_disk->minors = 1;
#if defined(GENHD_FL_EXT_DEVT)
zso->zvo_disk->flags &= ~GENHD_FL_EXT_DEVT;
#endif
#if defined(GENHD_FL_NO_PART_SCAN)
zso->zvo_disk->flags |= GENHD_FL_NO_PART_SCAN;
#endif
}
zso->zvo_disk->first_minor = (dev & MINORMASK);
zso->zvo_disk->fops = &zvol_ops;
zso->zvo_disk->private_data = zv;
snprintf(zso->zvo_disk->disk_name, DISK_NAME_LEN, "%s%d",
ZVOL_DEV_NAME, (dev & MINORMASK));
return (zv);
out_kmem:
kmem_free(zso, sizeof (struct zvol_state_os));
kmem_free(zv, sizeof (zvol_state_t));
return (NULL);
}
/*
* Cleanup then free a zvol_state_t which was created by zvol_alloc().
* At this time, the structure is not opened by anyone, is taken off
* the zvol_state_list, and has its private data set to NULL.
* The zvol_state_lock is dropped.
*
* This function may take many milliseconds to complete (e.g. we've seen
* it take over 256ms), due to the calls to "blk_cleanup_queue" and
* "del_gendisk". Thus, consumers need to be careful to account for this
* latency when calling this function.
*/
static void
zvol_free(zvol_state_t *zv)
{
ASSERT(!RW_LOCK_HELD(&zv->zv_suspend_lock));
ASSERT(!MUTEX_HELD(&zv->zv_state_lock));
ASSERT0(zv->zv_open_count);
ASSERT3P(zv->zv_zso->zvo_disk->private_data, ==, NULL);
rw_destroy(&zv->zv_suspend_lock);
zfs_rangelock_fini(&zv->zv_rangelock);
del_gendisk(zv->zv_zso->zvo_disk);
#if defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS) && \
defined(HAVE_BLK_ALLOC_DISK)
blk_cleanup_disk(zv->zv_zso->zvo_disk);
#else
blk_cleanup_queue(zv->zv_zso->zvo_queue);
put_disk(zv->zv_zso->zvo_disk);
#endif
ida_simple_remove(&zvol_ida,
MINOR(zv->zv_zso->zvo_dev) >> ZVOL_MINOR_BITS);
mutex_destroy(&zv->zv_state_lock);
dataset_kstats_destroy(&zv->zv_kstat);
kmem_free(zv->zv_zso, sizeof (struct zvol_state_os));
kmem_free(zv, sizeof (zvol_state_t));
}
void
zvol_wait_close(zvol_state_t *zv)
{
}
/*
* Create a block device minor node and setup the linkage between it
* and the specified volume. Once this function returns the block
* device is live and ready for use.
*/
static int
zvol_os_create_minor(const char *name)
{
zvol_state_t *zv;
objset_t *os;
dmu_object_info_t *doi;
uint64_t volsize;
uint64_t len;
unsigned minor = 0;
int error = 0;
int idx;
uint64_t hash = zvol_name_hash(name);
if (zvol_inhibit_dev)
return (0);
idx = ida_simple_get(&zvol_ida, 0, 0, kmem_flags_convert(KM_SLEEP));
if (idx < 0)
return (SET_ERROR(-idx));
minor = idx << ZVOL_MINOR_BITS;
zv = zvol_find_by_name_hash(name, hash, RW_NONE);
if (zv) {
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
mutex_exit(&zv->zv_state_lock);
ida_simple_remove(&zvol_ida, idx);
return (SET_ERROR(EEXIST));
}
doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);
error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, B_TRUE, FTAG, &os);
if (error)
goto out_doi;
error = dmu_object_info(os, ZVOL_OBJ, doi);
if (error)
goto out_dmu_objset_disown;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_dmu_objset_disown;
zv = zvol_alloc(MKDEV(zvol_major, minor), name);
if (zv == NULL) {
error = SET_ERROR(EAGAIN);
goto out_dmu_objset_disown;
}
zv->zv_hash = hash;
if (dmu_objset_is_snapshot(os))
zv->zv_flags |= ZVOL_RDONLY;
zv->zv_volblocksize = doi->doi_data_block_size;
zv->zv_volsize = volsize;
zv->zv_objset = os;
set_capacity(zv->zv_zso->zvo_disk, zv->zv_volsize >> 9);
blk_queue_max_hw_sectors(zv->zv_zso->zvo_queue,
(DMU_MAX_ACCESS / 4) >> 9);
blk_queue_max_segments(zv->zv_zso->zvo_queue, UINT16_MAX);
blk_queue_max_segment_size(zv->zv_zso->zvo_queue, UINT_MAX);
blk_queue_physical_block_size(zv->zv_zso->zvo_queue,
zv->zv_volblocksize);
blk_queue_io_opt(zv->zv_zso->zvo_queue, zv->zv_volblocksize);
blk_queue_max_discard_sectors(zv->zv_zso->zvo_queue,
(zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
blk_queue_discard_granularity(zv->zv_zso->zvo_queue,
zv->zv_volblocksize);
blk_queue_flag_set(QUEUE_FLAG_DISCARD, zv->zv_zso->zvo_queue);
#ifdef QUEUE_FLAG_NONROT
blk_queue_flag_set(QUEUE_FLAG_NONROT, zv->zv_zso->zvo_queue);
#endif
#ifdef QUEUE_FLAG_ADD_RANDOM
blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, zv->zv_zso->zvo_queue);
#endif
/* This flag was introduced in kernel version 4.12. */
#ifdef QUEUE_FLAG_SCSI_PASSTHROUGH
blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, zv->zv_zso->zvo_queue);
#endif
ASSERT3P(zv->zv_zilog, ==, NULL);
zv->zv_zilog = zil_open(os, zvol_get_data);
if (spa_writeable(dmu_objset_spa(os))) {
if (zil_replay_disable)
zil_destroy(zv->zv_zilog, B_FALSE);
else
zil_replay(os, zv, zvol_replay_vector);
}
zil_close(zv->zv_zilog);
zv->zv_zilog = NULL;
ASSERT3P(zv->zv_kstat.dk_kstats, ==, NULL);
dataset_kstats_create(&zv->zv_kstat, zv->zv_objset);
/*
* When udev detects the addition of the device it will immediately
* invoke blkid(8) to determine the type of content on the device.
* Prefetching the blocks commonly scanned by blkid(8) will speed
* up this process.
*/
len = MIN(MAX(zvol_prefetch_bytes, 0), SPA_MAXBLOCKSIZE);
if (len > 0) {
dmu_prefetch(os, ZVOL_OBJ, 0, 0, len, ZIO_PRIORITY_SYNC_READ);
dmu_prefetch(os, ZVOL_OBJ, 0, volsize - len, len,
ZIO_PRIORITY_SYNC_READ);
}
zv->zv_objset = NULL;
out_dmu_objset_disown:
dmu_objset_disown(os, B_TRUE, FTAG);
out_doi:
kmem_free(doi, sizeof (dmu_object_info_t));
/*
* Keep in mind that once add_disk() is called, the zvol is
* announced to the world, and zvol_open()/zvol_release() can
* be called at any time. Incidentally, add_disk() itself calls
* zvol_open()->zvol_first_open() and zvol_release()->zvol_last_close()
* directly as well.
*/
if (error == 0) {
rw_enter(&zvol_state_lock, RW_WRITER);
zvol_insert(zv);
rw_exit(&zvol_state_lock);
add_disk(zv->zv_zso->zvo_disk);
} else {
ida_simple_remove(&zvol_ida, idx);
}
return (error);
}
static void
zvol_rename_minor(zvol_state_t *zv, const char *newname)
{
int readonly = get_disk_ro(zv->zv_zso->zvo_disk);
ASSERT(RW_LOCK_HELD(&zvol_state_lock));
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
strlcpy(zv->zv_name, newname, sizeof (zv->zv_name));
/* move to new hashtable entry */
zv->zv_hash = zvol_name_hash(zv->zv_name);
hlist_del(&zv->zv_hlink);
hlist_add_head(&zv->zv_hlink, ZVOL_HT_HEAD(zv->zv_hash));
/*
* The block device's read-only state is briefly changed causing
* a KOBJ_CHANGE uevent to be issued. This ensures udev detects
* the name change and fixes the symlinks. This does not change
* ZVOL_RDONLY in zv->zv_flags so the actual read-only state never
* changes. This would normally be done using kobject_uevent() but
* that is a GPL-only symbol which is why we need this workaround.
*/
set_disk_ro(zv->zv_zso->zvo_disk, !readonly);
set_disk_ro(zv->zv_zso->zvo_disk, readonly);
}
static void
zvol_set_disk_ro_impl(zvol_state_t *zv, int flags)
{
set_disk_ro(zv->zv_zso->zvo_disk, flags);
}
static void
zvol_set_capacity_impl(zvol_state_t *zv, uint64_t capacity)
{
set_capacity(zv->zv_zso->zvo_disk, capacity);
}
const static zvol_platform_ops_t zvol_linux_ops = {
.zv_free = zvol_free,
.zv_rename_minor = zvol_rename_minor,
.zv_create_minor = zvol_os_create_minor,
.zv_update_volsize = zvol_update_volsize,
.zv_clear_private = zvol_clear_private,
.zv_is_zvol = zvol_is_zvol_impl,
.zv_set_disk_ro = zvol_set_disk_ro_impl,
.zv_set_capacity = zvol_set_capacity_impl,
};
int
zvol_init(void)
{
int error;
int threads = MIN(MAX(zvol_threads, 1), 1024);
error = register_blkdev(zvol_major, ZVOL_DRIVER);
if (error) {
printk(KERN_INFO "ZFS: register_blkdev() failed %d\n", error);
return (error);
}
zvol_taskq = taskq_create(ZVOL_DRIVER, threads, maxclsyspri,
threads * 2, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC);
if (zvol_taskq == NULL) {
unregister_blkdev(zvol_major, ZVOL_DRIVER);
return (-ENOMEM);
}
zvol_init_impl();
ida_init(&zvol_ida);
zvol_register_ops(&zvol_linux_ops);
return (0);
}
void
zvol_fini(void)
{
zvol_fini_impl();
unregister_blkdev(zvol_major, ZVOL_DRIVER);
taskq_destroy(zvol_taskq);
ida_destroy(&zvol_ida);
}
/* BEGIN CSTYLED */
module_param(zvol_inhibit_dev, uint, 0644);
MODULE_PARM_DESC(zvol_inhibit_dev, "Do not create zvol device nodes");
module_param(zvol_major, uint, 0444);
MODULE_PARM_DESC(zvol_major, "Major number for zvol device");
module_param(zvol_threads, uint, 0444);
MODULE_PARM_DESC(zvol_threads, "Max number of threads to handle I/O requests");
module_param(zvol_request_sync, uint, 0644);
MODULE_PARM_DESC(zvol_request_sync, "Synchronously handle bio requests");
module_param(zvol_max_discard_blocks, ulong, 0444);
MODULE_PARM_DESC(zvol_max_discard_blocks, "Max number of blocks to discard");
module_param(zvol_prefetch_bytes, uint, 0644);
MODULE_PARM_DESC(zvol_prefetch_bytes, "Prefetch N bytes at zvol start+end");
module_param(zvol_volmode, uint, 0644);
MODULE_PARM_DESC(zvol_volmode, "Default volmode property value");
/* END CSTYLED */
diff --git a/module/unicode/u8_textprep.c b/module/unicode/u8_textprep.c
index bce5f19625cb..863f69f7ba82 100644
--- a/module/unicode/u8_textprep.c
+++ b/module/unicode/u8_textprep.c
@@ -1,2155 +1,2155 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* UTF-8 text preparation functions (PSARC/2007/149, PSARC/2007/458).
*
* Man pages: u8_textprep_open(9F), u8_textprep_buf(9F), u8_textprep_close(9F),
* u8_textprep_str(9F), u8_strcmp(9F), and u8_validate(9F). See also
* the section 3C man pages.
* Interface stability: Committed.
*/
#include <sys/types.h>
#include <sys/strings.h>
#include <sys/param.h>
#include <sys/sysmacros.h>
#include <sys/debug.h>
#include <sys/kmem.h>
#include <sys/sunddi.h>
#include <sys/u8_textprep.h>
#include <sys/byteorder.h>
#include <sys/errno.h>
#include <sys/u8_textprep_data.h>
#include <sys/mod.h>
/* The maximum possible number of bytes in a UTF-8 character. */
#define U8_MB_CUR_MAX (4)
/*
* The maximum number of bytes needed for a UTF-8 character to cover
* U+0000 - U+FFFF, i.e., the coding space of now deprecated UCS-2.
*/
#define U8_MAX_BYTES_UCS2 (3)
/* The maximum possible number of bytes in a Stream-Safe Text. */
#define U8_STREAM_SAFE_TEXT_MAX (128)
/*
* The maximum number of characters in a combining/conjoining sequence and
* the actual upperbound limit of a combining/conjoining sequence.
*/
#define U8_MAX_CHARS_A_SEQ (32)
#define U8_UPPER_LIMIT_IN_A_SEQ (31)
/* The combining class value for Starter. */
#define U8_COMBINING_CLASS_STARTER (0)
/*
* Some Hangul related macros at below.
*
* The first and the last of Hangul syllables, Hangul Jamo Leading consonants,
* Vowels, and optional Trailing consonants in Unicode scalar values.
*
* Please be noted that the U8_HANGUL_JAMO_T_FIRST is 0x11A7 at below not
* the actual U+11A8. This is due to that the trailing consonant is optional
* and thus we are doing a pre-calculation of subtracting one.
*
* Each of 19 modern leading consonants has total 588 possible syllables since
* Hangul has 21 modern vowels and 27 modern trailing consonants plus 1 for
* no trailing consonant case, i.e., 21 x 28 = 588.
*
* We also have bunch of Hangul related macros at below. Please bear in mind
* that the U8_HANGUL_JAMO_1ST_BYTE can be used to check whether it is
* a Hangul Jamo or not but the value does not guarantee that it is a Hangul
* Jamo; it just guarantee that it will be most likely.
*/
#define U8_HANGUL_SYL_FIRST (0xAC00U)
#define U8_HANGUL_SYL_LAST (0xD7A3U)
#define U8_HANGUL_JAMO_L_FIRST (0x1100U)
#define U8_HANGUL_JAMO_L_LAST (0x1112U)
#define U8_HANGUL_JAMO_V_FIRST (0x1161U)
#define U8_HANGUL_JAMO_V_LAST (0x1175U)
#define U8_HANGUL_JAMO_T_FIRST (0x11A7U)
#define U8_HANGUL_JAMO_T_LAST (0x11C2U)
#define U8_HANGUL_V_COUNT (21)
#define U8_HANGUL_VT_COUNT (588)
#define U8_HANGUL_T_COUNT (28)
#define U8_HANGUL_JAMO_1ST_BYTE (0xE1U)
#define U8_SAVE_HANGUL_AS_UTF8(s, i, j, k, b) \
(s)[(i)] = (uchar_t)(0xE0U | ((uint32_t)(b) & 0xF000U) >> 12); \
(s)[(j)] = (uchar_t)(0x80U | ((uint32_t)(b) & 0x0FC0U) >> 6); \
(s)[(k)] = (uchar_t)(0x80U | ((uint32_t)(b) & 0x003FU));
#define U8_HANGUL_JAMO_L(u) \
((u) >= U8_HANGUL_JAMO_L_FIRST && (u) <= U8_HANGUL_JAMO_L_LAST)
#define U8_HANGUL_JAMO_V(u) \
((u) >= U8_HANGUL_JAMO_V_FIRST && (u) <= U8_HANGUL_JAMO_V_LAST)
#define U8_HANGUL_JAMO_T(u) \
((u) > U8_HANGUL_JAMO_T_FIRST && (u) <= U8_HANGUL_JAMO_T_LAST)
#define U8_HANGUL_JAMO(u) \
((u) >= U8_HANGUL_JAMO_L_FIRST && (u) <= U8_HANGUL_JAMO_T_LAST)
#define U8_HANGUL_SYLLABLE(u) \
((u) >= U8_HANGUL_SYL_FIRST && (u) <= U8_HANGUL_SYL_LAST)
#define U8_HANGUL_COMPOSABLE_L_V(s, u) \
((s) == U8_STATE_HANGUL_L && U8_HANGUL_JAMO_V((u)))
#define U8_HANGUL_COMPOSABLE_LV_T(s, u) \
((s) == U8_STATE_HANGUL_LV && U8_HANGUL_JAMO_T((u)))
/* The types of decomposition mappings. */
#define U8_DECOMP_BOTH (0xF5U)
#define U8_DECOMP_CANONICAL (0xF6U)
/* The indicator for 16-bit table. */
#define U8_16BIT_TABLE_INDICATOR (0x8000U)
/* The following are some convenience macros. */
#define U8_PUT_3BYTES_INTO_UTF32(u, b1, b2, b3) \
(u) = ((((uint32_t)(b1) & 0x0F) << 12) | \
(((uint32_t)(b2) & 0x3F) << 6) | \
((uint32_t)(b3) & 0x3F));
#define U8_SIMPLE_SWAP(a, b, t) \
(t) = (a); \
(a) = (b); \
(b) = (t);
#define U8_ASCII_TOUPPER(c) \
(((c) >= 'a' && (c) <= 'z') ? (c) - 'a' + 'A' : (c))
#define U8_ASCII_TOLOWER(c) \
(((c) >= 'A' && (c) <= 'Z') ? (c) - 'A' + 'a' : (c))
#define U8_ISASCII(c) (((uchar_t)(c)) < 0x80U)
/*
* The following macro assumes that the two characters that are to be
* swapped are adjacent to each other and 'a' comes before 'b'.
*
* If the assumptions are not met, then, the macro will fail.
*/
#define U8_SWAP_COMB_MARKS(a, b) \
for (k = 0; k < disp[(a)]; k++) \
u8t[k] = u8s[start[(a)] + k]; \
for (k = 0; k < disp[(b)]; k++) \
u8s[start[(a)] + k] = u8s[start[(b)] + k]; \
start[(b)] = start[(a)] + disp[(b)]; \
for (k = 0; k < disp[(a)]; k++) \
u8s[start[(b)] + k] = u8t[k]; \
U8_SIMPLE_SWAP(comb_class[(a)], comb_class[(b)], tc); \
U8_SIMPLE_SWAP(disp[(a)], disp[(b)], tc);
/* The possible states during normalization. */
typedef enum {
U8_STATE_START = 0,
U8_STATE_HANGUL_L = 1,
U8_STATE_HANGUL_LV = 2,
U8_STATE_HANGUL_LVT = 3,
U8_STATE_HANGUL_V = 4,
U8_STATE_HANGUL_T = 5,
U8_STATE_COMBINING_MARK = 6
} u8_normalization_states_t;
/*
* The three vectors at below are used to check bytes of a given UTF-8
* character are valid and not containing any malformed byte values.
*
* We used to have a quite relaxed UTF-8 binary representation but then there
* was some security related issues and so the Unicode Consortium defined
* and announced the UTF-8 Corrigendum at Unicode 3.1 and then refined it
* one more time at the Unicode 3.2. The following three tables are based on
* that.
*/
#define U8_ILLEGAL_NEXT_BYTE_COMMON(c) ((c) < 0x80 || (c) > 0xBF)
#define I_ U8_ILLEGAL_CHAR
#define O_ U8_OUT_OF_RANGE_CHAR
-const int8_t u8_number_of_bytes[0x100] = {
+static const int8_t u8_number_of_bytes[0x100] = {
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
/* 80 81 82 83 84 85 86 87 88 89 8A 8B 8C 8D 8E 8F */
I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_,
/* 90 91 92 93 94 95 96 97 98 99 9A 9B 9C 9D 9E 9F */
I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_,
/* A0 A1 A2 A3 A4 A5 A6 A7 A8 A9 AA AB AC AD AE AF */
I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_,
/* B0 B1 B2 B3 B4 B5 B6 B7 B8 B9 BA BB BC BD BE BF */
I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_, I_,
/* C0 C1 C2 C3 C4 C5 C6 C7 C8 C9 CA CB CC CD CE CF */
I_, I_, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
/* D0 D1 D2 D3 D4 D5 D6 D7 D8 D9 DA DB DC DD DE DF */
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
/* E0 E1 E2 E3 E4 E5 E6 E7 E8 E9 EA EB EC ED EE EF */
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
/* F0 F1 F2 F3 F4 F5 F6 F7 F8 F9 FA FB FC FD FE FF */
4, 4, 4, 4, 4, O_, O_, O_, O_, O_, O_, O_, O_, O_, O_, O_,
};
#undef I_
#undef O_
-const uint8_t u8_valid_min_2nd_byte[0x100] = {
+static const uint8_t u8_valid_min_2nd_byte[0x100] = {
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
/* C0 C1 C2 C3 C4 C5 C6 C7 */
0, 0, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* C8 C9 CA CB CC CD CE CF */
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* D0 D1 D2 D3 D4 D5 D6 D7 */
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* D8 D9 DA DB DC DD DE DF */
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* E0 E1 E2 E3 E4 E5 E6 E7 */
0xa0, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* E8 E9 EA EB EC ED EE EF */
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* F0 F1 F2 F3 F4 F5 F6 F7 */
0x90, 0x80, 0x80, 0x80, 0x80, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
};
-const uint8_t u8_valid_max_2nd_byte[0x100] = {
+static const uint8_t u8_valid_max_2nd_byte[0x100] = {
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
/* C0 C1 C2 C3 C4 C5 C6 C7 */
0, 0, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* C8 C9 CA CB CC CD CE CF */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* D0 D1 D2 D3 D4 D5 D6 D7 */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* D8 D9 DA DB DC DD DE DF */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* E0 E1 E2 E3 E4 E5 E6 E7 */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* E8 E9 EA EB EC ED EE EF */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0x9f, 0xbf, 0xbf,
/* F0 F1 F2 F3 F4 F5 F6 F7 */
0xbf, 0xbf, 0xbf, 0xbf, 0x8f, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
};
/*
* The u8_validate() validates on the given UTF-8 character string and
* calculate the byte length. It is quite similar to mblen(3C) except that
* this will validate against the list of characters if required and
* specific to UTF-8 and Unicode.
*/
int
u8_validate(const char *u8str, size_t n, char **list, int flag, int *errnum)
{
uchar_t *ib;
uchar_t *ibtail;
uchar_t **p;
uchar_t *s1;
uchar_t *s2;
uchar_t f;
int sz;
size_t i;
int ret_val;
boolean_t second;
boolean_t no_need_to_validate_entire;
boolean_t check_additional;
boolean_t validate_ucs2_range_only;
if (! u8str)
return (0);
ib = (uchar_t *)u8str;
ibtail = ib + n;
ret_val = 0;
no_need_to_validate_entire = ! (flag & U8_VALIDATE_ENTIRE);
check_additional = flag & U8_VALIDATE_CHECK_ADDITIONAL;
validate_ucs2_range_only = flag & U8_VALIDATE_UCS2_RANGE;
while (ib < ibtail) {
/*
* The first byte of a UTF-8 character tells how many
* bytes will follow for the character. If the first byte
* is an illegal byte value or out of range value, we just
* return -1 with an appropriate error number.
*/
sz = u8_number_of_bytes[*ib];
if (sz == U8_ILLEGAL_CHAR) {
*errnum = EILSEQ;
return (-1);
}
if (sz == U8_OUT_OF_RANGE_CHAR ||
(validate_ucs2_range_only && sz > U8_MAX_BYTES_UCS2)) {
*errnum = ERANGE;
return (-1);
}
/*
* If we don't have enough bytes to check on, that's also
* an error. As you can see, we give illegal byte sequence
* checking higher priority then EINVAL cases.
*/
if ((ibtail - ib) < sz) {
*errnum = EINVAL;
return (-1);
}
if (sz == 1) {
ib++;
ret_val++;
} else {
/*
* Check on the multi-byte UTF-8 character. For more
* details on this, see comment added for the used
* data structures at the beginning of the file.
*/
f = *ib++;
ret_val++;
second = B_TRUE;
for (i = 1; i < sz; i++) {
if (second) {
if (*ib < u8_valid_min_2nd_byte[f] ||
*ib > u8_valid_max_2nd_byte[f]) {
*errnum = EILSEQ;
return (-1);
}
second = B_FALSE;
} else if (U8_ILLEGAL_NEXT_BYTE_COMMON(*ib)) {
*errnum = EILSEQ;
return (-1);
}
ib++;
ret_val++;
}
}
if (check_additional) {
for (p = (uchar_t **)list, i = 0; p[i]; i++) {
s1 = ib - sz;
s2 = p[i];
while (s1 < ib) {
if (*s1 != *s2 || *s2 == '\0')
break;
s1++;
s2++;
}
if (s1 >= ib && *s2 == '\0') {
*errnum = EBADF;
return (-1);
}
}
}
if (no_need_to_validate_entire)
break;
}
return (ret_val);
}
/*
* The do_case_conv() looks at the mapping tables and returns found
* bytes if any. If not found, the input bytes are returned. The function
* always terminate the return bytes with a null character assuming that
* there are plenty of room to do so.
*
* The case conversions are simple case conversions mapping a character to
* another character as specified in the Unicode data. The byte size of
* the mapped character could be different from that of the input character.
*
* The return value is the byte length of the returned character excluding
* the terminating null byte.
*/
static size_t
do_case_conv(int uv, uchar_t *u8s, uchar_t *s, int sz, boolean_t is_it_toupper)
{
size_t i;
uint16_t b1 = 0;
uint16_t b2 = 0;
uint16_t b3 = 0;
uint16_t b3_tbl;
uint16_t b3_base;
uint16_t b4 = 0;
size_t start_id;
size_t end_id;
/*
* At this point, the only possible values for sz are 2, 3, and 4.
* The u8s should point to a vector that is well beyond the size of
* 5 bytes.
*/
if (sz == 2) {
b3 = u8s[0] = s[0];
b4 = u8s[1] = s[1];
} else if (sz == 3) {
b2 = u8s[0] = s[0];
b3 = u8s[1] = s[1];
b4 = u8s[2] = s[2];
} else if (sz == 4) {
b1 = u8s[0] = s[0];
b2 = u8s[1] = s[1];
b3 = u8s[2] = s[2];
b4 = u8s[3] = s[3];
} else {
/* This is not possible but just in case as a fallback. */
if (is_it_toupper)
*u8s = U8_ASCII_TOUPPER(*s);
else
*u8s = U8_ASCII_TOLOWER(*s);
u8s[1] = '\0';
return (1);
}
u8s[sz] = '\0';
/*
* Let's find out if we have a corresponding character.
*/
b1 = u8_common_b1_tbl[uv][b1];
if (b1 == U8_TBL_ELEMENT_NOT_DEF)
return ((size_t)sz);
b2 = u8_case_common_b2_tbl[uv][b1][b2];
if (b2 == U8_TBL_ELEMENT_NOT_DEF)
return ((size_t)sz);
if (is_it_toupper) {
b3_tbl = u8_toupper_b3_tbl[uv][b2][b3].tbl_id;
if (b3_tbl == U8_TBL_ELEMENT_NOT_DEF)
return ((size_t)sz);
start_id = u8_toupper_b4_tbl[uv][b3_tbl][b4];
end_id = u8_toupper_b4_tbl[uv][b3_tbl][b4 + 1];
/* Either there is no match or an error at the table. */
if (start_id >= end_id || (end_id - start_id) > U8_MB_CUR_MAX)
return ((size_t)sz);
b3_base = u8_toupper_b3_tbl[uv][b2][b3].base;
for (i = 0; start_id < end_id; start_id++)
u8s[i++] = u8_toupper_final_tbl[uv][b3_base + start_id];
} else {
b3_tbl = u8_tolower_b3_tbl[uv][b2][b3].tbl_id;
if (b3_tbl == U8_TBL_ELEMENT_NOT_DEF)
return ((size_t)sz);
start_id = u8_tolower_b4_tbl[uv][b3_tbl][b4];
end_id = u8_tolower_b4_tbl[uv][b3_tbl][b4 + 1];
if (start_id >= end_id || (end_id - start_id) > U8_MB_CUR_MAX)
return ((size_t)sz);
b3_base = u8_tolower_b3_tbl[uv][b2][b3].base;
for (i = 0; start_id < end_id; start_id++)
u8s[i++] = u8_tolower_final_tbl[uv][b3_base + start_id];
}
/*
* If i is still zero, that means there is no corresponding character.
*/
if (i == 0)
return ((size_t)sz);
u8s[i] = '\0';
return (i);
}
/*
* The do_case_compare() function compares the two input strings, s1 and s2,
* one character at a time doing case conversions if applicable and return
* the comparison result as like strcmp().
*
* Since, in empirical sense, most of text data are 7-bit ASCII characters,
* we treat the 7-bit ASCII characters as a special case trying to yield
* faster processing time.
*/
static int
do_case_compare(size_t uv, uchar_t *s1, uchar_t *s2, size_t n1,
size_t n2, boolean_t is_it_toupper, int *errnum)
{
int f;
int sz1;
int sz2;
size_t j;
size_t i1;
size_t i2;
uchar_t u8s1[U8_MB_CUR_MAX + 1];
uchar_t u8s2[U8_MB_CUR_MAX + 1];
i1 = i2 = 0;
while (i1 < n1 && i2 < n2) {
/*
* Find out what would be the byte length for this UTF-8
* character at string s1 and also find out if this is
* an illegal start byte or not and if so, issue a proper
* error number and yet treat this byte as a character.
*/
sz1 = u8_number_of_bytes[*s1];
if (sz1 < 0) {
*errnum = EILSEQ;
sz1 = 1;
}
/*
* For 7-bit ASCII characters mainly, we do a quick case
* conversion right at here.
*
* If we don't have enough bytes for this character, issue
* an EINVAL error and use what are available.
*
* If we have enough bytes, find out if there is
* a corresponding uppercase character and if so, copy over
* the bytes for a comparison later. If there is no
* corresponding uppercase character, then, use what we have
* for the comparison.
*/
if (sz1 == 1) {
if (is_it_toupper)
u8s1[0] = U8_ASCII_TOUPPER(*s1);
else
u8s1[0] = U8_ASCII_TOLOWER(*s1);
s1++;
u8s1[1] = '\0';
} else if ((i1 + sz1) > n1) {
*errnum = EINVAL;
for (j = 0; (i1 + j) < n1; )
u8s1[j++] = *s1++;
u8s1[j] = '\0';
} else {
(void) do_case_conv(uv, u8s1, s1, sz1, is_it_toupper);
s1 += sz1;
}
/* Do the same for the string s2. */
sz2 = u8_number_of_bytes[*s2];
if (sz2 < 0) {
*errnum = EILSEQ;
sz2 = 1;
}
if (sz2 == 1) {
if (is_it_toupper)
u8s2[0] = U8_ASCII_TOUPPER(*s2);
else
u8s2[0] = U8_ASCII_TOLOWER(*s2);
s2++;
u8s2[1] = '\0';
} else if ((i2 + sz2) > n2) {
*errnum = EINVAL;
for (j = 0; (i2 + j) < n2; )
u8s2[j++] = *s2++;
u8s2[j] = '\0';
} else {
(void) do_case_conv(uv, u8s2, s2, sz2, is_it_toupper);
s2 += sz2;
}
/* Now compare the two characters. */
if (sz1 == 1 && sz2 == 1) {
if (*u8s1 > *u8s2)
return (1);
if (*u8s1 < *u8s2)
return (-1);
} else {
f = strcmp((const char *)u8s1, (const char *)u8s2);
if (f != 0)
return (f);
}
/*
* They were the same. Let's move on to the next
* characters then.
*/
i1 += sz1;
i2 += sz2;
}
/*
* We compared until the end of either or both strings.
*
* If we reached to or went over the ends for the both, that means
* they are the same.
*
* If we reached only one of the two ends, that means the other string
* has something which then the fact can be used to determine
* the return value.
*/
if (i1 >= n1) {
if (i2 >= n2)
return (0);
return (-1);
}
return (1);
}
/*
* The combining_class() function checks on the given bytes and find out
* the corresponding Unicode combining class value. The return value 0 means
* it is a Starter. Any illegal UTF-8 character will also be treated as
* a Starter.
*/
static uchar_t
combining_class(size_t uv, uchar_t *s, size_t sz)
{
uint16_t b1 = 0;
uint16_t b2 = 0;
uint16_t b3 = 0;
uint16_t b4 = 0;
if (sz == 1 || sz > 4)
return (0);
if (sz == 2) {
b3 = s[0];
b4 = s[1];
} else if (sz == 3) {
b2 = s[0];
b3 = s[1];
b4 = s[2];
} else if (sz == 4) {
b1 = s[0];
b2 = s[1];
b3 = s[2];
b4 = s[3];
}
b1 = u8_common_b1_tbl[uv][b1];
if (b1 == U8_TBL_ELEMENT_NOT_DEF)
return (0);
b2 = u8_combining_class_b2_tbl[uv][b1][b2];
if (b2 == U8_TBL_ELEMENT_NOT_DEF)
return (0);
b3 = u8_combining_class_b3_tbl[uv][b2][b3];
if (b3 == U8_TBL_ELEMENT_NOT_DEF)
return (0);
return (u8_combining_class_b4_tbl[uv][b3][b4]);
}
/*
* The do_decomp() function finds out a matching decomposition if any
* and return. If there is no match, the input bytes are copied and returned.
* The function also checks if there is a Hangul, decomposes it if necessary
* and returns.
*
* To save time, a single byte 7-bit ASCII character should be handled by
* the caller.
*
* The function returns the number of bytes returned sans always terminating
* the null byte. It will also return a state that will tell if there was
* a Hangul character decomposed which then will be used by the caller.
*/
static size_t
do_decomp(size_t uv, uchar_t *u8s, uchar_t *s, int sz,
boolean_t canonical_decomposition, u8_normalization_states_t *state)
{
uint16_t b1 = 0;
uint16_t b2 = 0;
uint16_t b3 = 0;
uint16_t b3_tbl;
uint16_t b3_base;
uint16_t b4 = 0;
size_t start_id;
size_t end_id;
size_t i;
uint32_t u1;
if (sz == 2) {
b3 = u8s[0] = s[0];
b4 = u8s[1] = s[1];
u8s[2] = '\0';
} else if (sz == 3) {
/* Convert it to a Unicode scalar value. */
U8_PUT_3BYTES_INTO_UTF32(u1, s[0], s[1], s[2]);
/*
* If this is a Hangul syllable, we decompose it into
* a leading consonant, a vowel, and an optional trailing
* consonant and then return.
*/
if (U8_HANGUL_SYLLABLE(u1)) {
u1 -= U8_HANGUL_SYL_FIRST;
b1 = U8_HANGUL_JAMO_L_FIRST + u1 / U8_HANGUL_VT_COUNT;
b2 = U8_HANGUL_JAMO_V_FIRST + (u1 % U8_HANGUL_VT_COUNT)
/ U8_HANGUL_T_COUNT;
b3 = u1 % U8_HANGUL_T_COUNT;
U8_SAVE_HANGUL_AS_UTF8(u8s, 0, 1, 2, b1);
U8_SAVE_HANGUL_AS_UTF8(u8s, 3, 4, 5, b2);
if (b3) {
b3 += U8_HANGUL_JAMO_T_FIRST;
U8_SAVE_HANGUL_AS_UTF8(u8s, 6, 7, 8, b3);
u8s[9] = '\0';
*state = U8_STATE_HANGUL_LVT;
return (9);
}
u8s[6] = '\0';
*state = U8_STATE_HANGUL_LV;
return (6);
}
b2 = u8s[0] = s[0];
b3 = u8s[1] = s[1];
b4 = u8s[2] = s[2];
u8s[3] = '\0';
/*
* If this is a Hangul Jamo, we know there is nothing
* further that we can decompose.
*/
if (U8_HANGUL_JAMO_L(u1)) {
*state = U8_STATE_HANGUL_L;
return (3);
}
if (U8_HANGUL_JAMO_V(u1)) {
if (*state == U8_STATE_HANGUL_L)
*state = U8_STATE_HANGUL_LV;
else
*state = U8_STATE_HANGUL_V;
return (3);
}
if (U8_HANGUL_JAMO_T(u1)) {
if (*state == U8_STATE_HANGUL_LV)
*state = U8_STATE_HANGUL_LVT;
else
*state = U8_STATE_HANGUL_T;
return (3);
}
} else if (sz == 4) {
b1 = u8s[0] = s[0];
b2 = u8s[1] = s[1];
b3 = u8s[2] = s[2];
b4 = u8s[3] = s[3];
u8s[4] = '\0';
} else {
/*
* This is a fallback and should not happen if the function
* was called properly.
*/
u8s[0] = s[0];
u8s[1] = '\0';
*state = U8_STATE_START;
return (1);
}
/*
* At this point, this routine does not know what it would get.
* The caller should sort it out if the state isn't a Hangul one.
*/
*state = U8_STATE_START;
/* Try to find matching decomposition mapping byte sequence. */
b1 = u8_common_b1_tbl[uv][b1];
if (b1 == U8_TBL_ELEMENT_NOT_DEF)
return ((size_t)sz);
b2 = u8_decomp_b2_tbl[uv][b1][b2];
if (b2 == U8_TBL_ELEMENT_NOT_DEF)
return ((size_t)sz);
b3_tbl = u8_decomp_b3_tbl[uv][b2][b3].tbl_id;
if (b3_tbl == U8_TBL_ELEMENT_NOT_DEF)
return ((size_t)sz);
/*
* If b3_tbl is bigger than or equal to U8_16BIT_TABLE_INDICATOR
* which is 0x8000, this means we couldn't fit the mappings into
* the cardinality of a unsigned byte.
*/
if (b3_tbl >= U8_16BIT_TABLE_INDICATOR) {
b3_tbl -= U8_16BIT_TABLE_INDICATOR;
start_id = u8_decomp_b4_16bit_tbl[uv][b3_tbl][b4];
end_id = u8_decomp_b4_16bit_tbl[uv][b3_tbl][b4 + 1];
} else {
// cppcheck-suppress arrayIndexOutOfBoundsCond
start_id = u8_decomp_b4_tbl[uv][b3_tbl][b4];
// cppcheck-suppress arrayIndexOutOfBoundsCond
end_id = u8_decomp_b4_tbl[uv][b3_tbl][b4 + 1];
}
/* This also means there wasn't any matching decomposition. */
if (start_id >= end_id)
return ((size_t)sz);
/*
* The final table for decomposition mappings has three types of
* byte sequences depending on whether a mapping is for compatibility
* decomposition, canonical decomposition, or both like the following:
*
* (1) Compatibility decomposition mappings:
*
* +---+---+-...-+---+
* | B0| B1| ... | Bm|
* +---+---+-...-+---+
*
* The first byte, B0, is always less than 0xF5 (U8_DECOMP_BOTH).
*
* (2) Canonical decomposition mappings:
*
* +---+---+---+-...-+---+
* | T | b0| b1| ... | bn|
* +---+---+---+-...-+---+
*
* where the first byte, T, is 0xF6 (U8_DECOMP_CANONICAL).
*
* (3) Both mappings:
*
* +---+---+---+---+-...-+---+---+---+-...-+---+
* | T | D | b0| b1| ... | bn| B0| B1| ... | Bm|
* +---+---+---+---+-...-+---+---+---+-...-+---+
*
* where T is 0xF5 (U8_DECOMP_BOTH) and D is a displacement
* byte, b0 to bn are canonical mapping bytes and B0 to Bm are
* compatibility mapping bytes.
*
* Note that compatibility decomposition means doing recursive
* decompositions using both compatibility decomposition mappings and
* canonical decomposition mappings. On the other hand, canonical
* decomposition means doing recursive decompositions using only
* canonical decomposition mappings. Since the table we have has gone
* through the recursions already, we do not need to do so during
* runtime, i.e., the table has been completely flattened out
* already.
*/
b3_base = u8_decomp_b3_tbl[uv][b2][b3].base;
/* Get the type, T, of the byte sequence. */
b1 = u8_decomp_final_tbl[uv][b3_base + start_id];
/*
* If necessary, adjust start_id, end_id, or both. Note that if
* this is compatibility decomposition mapping, there is no
* adjustment.
*/
if (canonical_decomposition) {
/* Is the mapping only for compatibility decomposition? */
if (b1 < U8_DECOMP_BOTH)
return ((size_t)sz);
start_id++;
if (b1 == U8_DECOMP_BOTH) {
end_id = start_id +
u8_decomp_final_tbl[uv][b3_base + start_id];
start_id++;
}
} else {
/*
* Unless this is a compatibility decomposition mapping,
* we adjust the start_id.
*/
if (b1 == U8_DECOMP_BOTH) {
start_id++;
start_id += u8_decomp_final_tbl[uv][b3_base + start_id];
} else if (b1 == U8_DECOMP_CANONICAL) {
start_id++;
}
}
for (i = 0; start_id < end_id; start_id++)
u8s[i++] = u8_decomp_final_tbl[uv][b3_base + start_id];
u8s[i] = '\0';
return (i);
}
/*
* The find_composition_start() function uses the character bytes given and
* find out the matching composition mappings if any and return the address
* to the composition mappings as explained in the do_composition().
*/
static uchar_t *
find_composition_start(size_t uv, uchar_t *s, size_t sz)
{
uint16_t b1 = 0;
uint16_t b2 = 0;
uint16_t b3 = 0;
uint16_t b3_tbl;
uint16_t b3_base;
uint16_t b4 = 0;
size_t start_id;
size_t end_id;
if (sz == 1) {
b4 = s[0];
} else if (sz == 2) {
b3 = s[0];
b4 = s[1];
} else if (sz == 3) {
b2 = s[0];
b3 = s[1];
b4 = s[2];
} else if (sz == 4) {
b1 = s[0];
b2 = s[1];
b3 = s[2];
b4 = s[3];
} else {
/*
* This is a fallback and should not happen if the function
* was called properly.
*/
return (NULL);
}
b1 = u8_composition_b1_tbl[uv][b1];
if (b1 == U8_TBL_ELEMENT_NOT_DEF)
return (NULL);
b2 = u8_composition_b2_tbl[uv][b1][b2];
if (b2 == U8_TBL_ELEMENT_NOT_DEF)
return (NULL);
b3_tbl = u8_composition_b3_tbl[uv][b2][b3].tbl_id;
if (b3_tbl == U8_TBL_ELEMENT_NOT_DEF)
return (NULL);
if (b3_tbl >= U8_16BIT_TABLE_INDICATOR) {
b3_tbl -= U8_16BIT_TABLE_INDICATOR;
start_id = u8_composition_b4_16bit_tbl[uv][b3_tbl][b4];
end_id = u8_composition_b4_16bit_tbl[uv][b3_tbl][b4 + 1];
} else {
// cppcheck-suppress arrayIndexOutOfBoundsCond
start_id = u8_composition_b4_tbl[uv][b3_tbl][b4];
// cppcheck-suppress arrayIndexOutOfBoundsCond
end_id = u8_composition_b4_tbl[uv][b3_tbl][b4 + 1];
}
if (start_id >= end_id)
return (NULL);
b3_base = u8_composition_b3_tbl[uv][b2][b3].base;
return ((uchar_t *)&(u8_composition_final_tbl[uv][b3_base + start_id]));
}
/*
* The blocked() function checks on the combining class values of previous
* characters in this sequence and return whether it is blocked or not.
*/
static boolean_t
blocked(uchar_t *comb_class, size_t last)
{
uchar_t my_comb_class;
size_t i;
my_comb_class = comb_class[last];
for (i = 1; i < last; i++)
if (comb_class[i] >= my_comb_class ||
comb_class[i] == U8_COMBINING_CLASS_STARTER)
return (B_TRUE);
return (B_FALSE);
}
/*
* The do_composition() reads the character string pointed by 's' and
* do necessary canonical composition and then copy over the result back to
* the 's'.
*
* The input argument 's' cannot contain more than 32 characters.
*/
static size_t
do_composition(size_t uv, uchar_t *s, uchar_t *comb_class, uchar_t *start,
uchar_t *disp, size_t last, uchar_t **os, uchar_t *oslast)
{
uchar_t t[U8_STREAM_SAFE_TEXT_MAX + 1];
uchar_t tc[U8_MB_CUR_MAX] = { '\0' };
uint8_t saved_marks[U8_MAX_CHARS_A_SEQ];
size_t saved_marks_count;
uchar_t *p;
uchar_t *saved_p;
uchar_t *q;
size_t i;
size_t saved_i;
size_t j;
size_t k;
size_t l;
size_t C;
size_t saved_l;
size_t size;
uint32_t u1;
uint32_t u2;
boolean_t match_not_found = B_TRUE;
/*
* This should never happen unless the callers are doing some strange
* and unexpected things.
*
* The "last" is the index pointing to the last character not last + 1.
*/
if (last >= U8_MAX_CHARS_A_SEQ)
last = U8_UPPER_LIMIT_IN_A_SEQ;
for (i = l = 0; i <= last; i++) {
/*
* The last or any non-Starters at the beginning, we don't
* have any chance to do composition and so we just copy them
* to the temporary buffer.
*/
if (i >= last || comb_class[i] != U8_COMBINING_CLASS_STARTER) {
SAVE_THE_CHAR:
p = s + start[i];
size = disp[i];
for (k = 0; k < size; k++)
t[l++] = *p++;
continue;
}
/*
* If this could be a start of Hangul Jamos, then, we try to
* conjoin them.
*/
if (s[start[i]] == U8_HANGUL_JAMO_1ST_BYTE) {
U8_PUT_3BYTES_INTO_UTF32(u1, s[start[i]],
s[start[i] + 1], s[start[i] + 2]);
U8_PUT_3BYTES_INTO_UTF32(u2, s[start[i] + 3],
s[start[i] + 4], s[start[i] + 5]);
if (U8_HANGUL_JAMO_L(u1) && U8_HANGUL_JAMO_V(u2)) {
u1 -= U8_HANGUL_JAMO_L_FIRST;
u2 -= U8_HANGUL_JAMO_V_FIRST;
u1 = U8_HANGUL_SYL_FIRST +
(u1 * U8_HANGUL_V_COUNT + u2) *
U8_HANGUL_T_COUNT;
i += 2;
if (i <= last) {
U8_PUT_3BYTES_INTO_UTF32(u2,
s[start[i]], s[start[i] + 1],
s[start[i] + 2]);
if (U8_HANGUL_JAMO_T(u2)) {
u1 += u2 -
U8_HANGUL_JAMO_T_FIRST;
i++;
}
}
U8_SAVE_HANGUL_AS_UTF8(t + l, 0, 1, 2, u1);
i--;
l += 3;
continue;
}
}
/*
* Let's then find out if this Starter has composition
* mapping.
*/
p = find_composition_start(uv, s + start[i], disp[i]);
if (p == NULL)
goto SAVE_THE_CHAR;
/*
* We have a Starter with composition mapping and the next
* character is a non-Starter. Let's try to find out if
* we can do composition.
*/
saved_p = p;
saved_i = i;
saved_l = l;
saved_marks_count = 0;
TRY_THE_NEXT_MARK:
q = s + start[++i];
size = disp[i];
/*
* The next for() loop compares the non-Starter pointed by
* 'q' with the possible (joinable) characters pointed by 'p'.
*
* The composition final table entry pointed by the 'p'
* looks like the following:
*
* +---+---+---+-...-+---+---+---+---+-...-+---+---+
* | C | b0| b2| ... | bn| F | B0| B1| ... | Bm| F |
* +---+---+---+-...-+---+---+---+---+-...-+---+---+
*
* where C is the count byte indicating the number of
* mapping pairs where each pair would be look like
* (b0-bn F, B0-Bm F). The b0-bn are the bytes of the second
* character of a canonical decomposition and the B0-Bm are
* the bytes of a matching composite character. The F is
* a filler byte after each character as the separator.
*/
match_not_found = B_TRUE;
for (C = *p++; C > 0; C--) {
for (k = 0; k < size; p++, k++)
if (*p != q[k])
break;
/* Have we found it? */
if (k >= size && *p == U8_TBL_ELEMENT_FILLER) {
match_not_found = B_FALSE;
l = saved_l;
while (*++p != U8_TBL_ELEMENT_FILLER)
t[l++] = *p;
break;
}
/* We didn't find; skip to the next pair. */
if (*p != U8_TBL_ELEMENT_FILLER)
while (*++p != U8_TBL_ELEMENT_FILLER)
;
while (*++p != U8_TBL_ELEMENT_FILLER)
;
p++;
}
/*
* If there was no match, we will need to save the combining
* mark for later appending. After that, if the next one
* is a non-Starter and not blocked, then, we try once
* again to do composition with the next non-Starter.
*
* If there was no match and this was a Starter, then,
* this is a new start.
*
* If there was a match and a composition done and we have
* more to check on, then, we retrieve a new composition final
* table entry for the composite and then try to do the
* composition again.
*/
if (match_not_found) {
if (comb_class[i] == U8_COMBINING_CLASS_STARTER) {
i--;
goto SAVE_THE_CHAR;
}
saved_marks[saved_marks_count++] = i;
}
if (saved_l == l) {
while (i < last) {
if (blocked(comb_class, i + 1))
saved_marks[saved_marks_count++] = ++i;
else
break;
}
if (i < last) {
p = saved_p;
goto TRY_THE_NEXT_MARK;
}
} else if (i < last) {
p = find_composition_start(uv, t + saved_l,
l - saved_l);
if (p != NULL) {
saved_p = p;
goto TRY_THE_NEXT_MARK;
}
}
/*
* There is no more composition possible.
*
* If there was no composition what so ever then we copy
* over the original Starter and then append any non-Starters
* remaining at the target string sequentially after that.
*/
if (saved_l == l) {
p = s + start[saved_i];
size = disp[saved_i];
for (j = 0; j < size; j++)
t[l++] = *p++;
}
for (k = 0; k < saved_marks_count; k++) {
p = s + start[saved_marks[k]];
size = disp[saved_marks[k]];
for (j = 0; j < size; j++)
t[l++] = *p++;
}
}
/*
* If the last character is a Starter and if we have a character
* (possibly another Starter) that can be turned into a composite,
* we do so and we do so until there is no more of composition
* possible.
*/
if (comb_class[last] == U8_COMBINING_CLASS_STARTER) {
p = *os;
saved_l = l - disp[last];
while (p < oslast) {
size = u8_number_of_bytes[*p];
if (size <= 1 || (p + size) > oslast)
break;
saved_p = p;
for (i = 0; i < size; i++)
tc[i] = *p++;
q = find_composition_start(uv, t + saved_l,
l - saved_l);
if (q == NULL) {
p = saved_p;
break;
}
match_not_found = B_TRUE;
for (C = *q++; C > 0; C--) {
for (k = 0; k < size; q++, k++)
if (*q != tc[k])
break;
if (k >= size && *q == U8_TBL_ELEMENT_FILLER) {
match_not_found = B_FALSE;
l = saved_l;
while (*++q != U8_TBL_ELEMENT_FILLER) {
/*
* This is practically
* impossible but we don't
* want to take any chances.
*/
if (l >=
U8_STREAM_SAFE_TEXT_MAX) {
p = saved_p;
goto SAFE_RETURN;
}
t[l++] = *q;
}
break;
}
if (*q != U8_TBL_ELEMENT_FILLER)
while (*++q != U8_TBL_ELEMENT_FILLER)
;
while (*++q != U8_TBL_ELEMENT_FILLER)
;
q++;
}
if (match_not_found) {
p = saved_p;
break;
}
}
SAFE_RETURN:
*os = p;
}
/*
* Now we copy over the temporary string to the target string.
* Since composition always reduces the number of characters or
* the number of characters stay, we don't need to worry about
* the buffer overflow here.
*/
for (i = 0; i < l; i++)
s[i] = t[i];
s[l] = '\0';
return (l);
}
/*
* The collect_a_seq() function checks on the given string s, collect
* a sequence of characters at u8s, and return the sequence. While it collects
* a sequence, it also applies case conversion, canonical or compatibility
* decomposition, canonical decomposition, or some or all of them and
* in that order.
*
* The collected sequence cannot be bigger than 32 characters since if
* it is having more than 31 characters, the sequence will be terminated
* with a U+034F COMBINING GRAPHEME JOINER (CGJ) character and turned into
* a Stream-Safe Text. The collected sequence is always terminated with
* a null byte and the return value is the byte length of the sequence
* including 0. The return value does not include the terminating
* null byte.
*/
static size_t
collect_a_seq(size_t uv, uchar_t *u8s, uchar_t **source, uchar_t *slast,
boolean_t is_it_toupper, boolean_t is_it_tolower,
boolean_t canonical_decomposition, boolean_t compatibility_decomposition,
boolean_t canonical_composition,
int *errnum, u8_normalization_states_t *state)
{
uchar_t *s;
int sz;
int saved_sz;
size_t i;
size_t j;
size_t k;
size_t l;
uchar_t comb_class[U8_MAX_CHARS_A_SEQ];
uchar_t disp[U8_MAX_CHARS_A_SEQ];
uchar_t start[U8_MAX_CHARS_A_SEQ];
uchar_t u8t[U8_MB_CUR_MAX] = { '\0' };
uchar_t uts[U8_STREAM_SAFE_TEXT_MAX + 1];
uchar_t tc;
size_t last;
size_t saved_last;
uint32_t u1;
/*
* Save the source string pointer which we will return a changed
* pointer if we do processing.
*/
s = *source;
/*
* The following is a fallback for just in case callers are not
* checking the string boundaries before the calling.
*/
if (s >= slast) {
u8s[0] = '\0';
return (0);
}
/*
* As the first thing, let's collect a character and do case
* conversion if necessary.
*/
sz = u8_number_of_bytes[*s];
if (sz < 0) {
*errnum = EILSEQ;
u8s[0] = *s++;
u8s[1] = '\0';
*source = s;
return (1);
}
if (sz == 1) {
if (is_it_toupper)
u8s[0] = U8_ASCII_TOUPPER(*s);
else if (is_it_tolower)
u8s[0] = U8_ASCII_TOLOWER(*s);
else
u8s[0] = *s;
s++;
u8s[1] = '\0';
} else if ((s + sz) > slast) {
*errnum = EINVAL;
for (i = 0; s < slast; )
u8s[i++] = *s++;
u8s[i] = '\0';
*source = s;
return (i);
} else {
if (is_it_toupper || is_it_tolower) {
i = do_case_conv(uv, u8s, s, sz, is_it_toupper);
s += sz;
sz = i;
} else {
for (i = 0; i < sz; )
u8s[i++] = *s++;
u8s[i] = '\0';
}
}
/*
* And then canonical/compatibility decomposition followed by
* an optional canonical composition. Please be noted that
* canonical composition is done only when a decomposition is
* done.
*/
if (canonical_decomposition || compatibility_decomposition) {
if (sz == 1) {
*state = U8_STATE_START;
saved_sz = 1;
comb_class[0] = 0;
start[0] = 0;
disp[0] = 1;
last = 1;
} else {
saved_sz = do_decomp(uv, u8s, u8s, sz,
canonical_decomposition, state);
last = 0;
for (i = 0; i < saved_sz; ) {
sz = u8_number_of_bytes[u8s[i]];
comb_class[last] = combining_class(uv,
u8s + i, sz);
start[last] = i;
disp[last] = sz;
last++;
i += sz;
}
/*
* Decomposition yields various Hangul related
* states but not on combining marks. We need to
* find out at here by checking on the last
* character.
*/
if (*state == U8_STATE_START) {
if (comb_class[last - 1])
*state = U8_STATE_COMBINING_MARK;
}
}
saved_last = last;
while (s < slast) {
sz = u8_number_of_bytes[*s];
/*
* If this is an illegal character, an incomplete
* character, or an 7-bit ASCII Starter character,
* then we have collected a sequence; break and let
* the next call deal with the two cases.
*
* Note that this is okay only if you are using this
* function with a fixed length string, not on
* a buffer with multiple calls of one chunk at a time.
*/
if (sz <= 1) {
break;
} else if ((s + sz) > slast) {
break;
} else {
/*
* If the previous character was a Hangul Jamo
* and this character is a Hangul Jamo that
* can be conjoined, we collect the Jamo.
*/
if (*s == U8_HANGUL_JAMO_1ST_BYTE) {
U8_PUT_3BYTES_INTO_UTF32(u1,
*s, *(s + 1), *(s + 2));
if (U8_HANGUL_COMPOSABLE_L_V(*state,
u1)) {
i = 0;
*state = U8_STATE_HANGUL_LV;
goto COLLECT_A_HANGUL;
}
if (U8_HANGUL_COMPOSABLE_LV_T(*state,
u1)) {
i = 0;
*state = U8_STATE_HANGUL_LVT;
goto COLLECT_A_HANGUL;
}
}
/*
* Regardless of whatever it was, if this is
* a Starter, we don't collect the character
* since that's a new start and we will deal
* with it at the next time.
*/
i = combining_class(uv, s, sz);
if (i == U8_COMBINING_CLASS_STARTER)
break;
/*
* We know the current character is a combining
* mark. If the previous character wasn't
* a Starter (not Hangul) or a combining mark,
* then, we don't collect this combining mark.
*/
if (*state != U8_STATE_START &&
*state != U8_STATE_COMBINING_MARK)
break;
*state = U8_STATE_COMBINING_MARK;
COLLECT_A_HANGUL:
/*
* If we collected a Starter and combining
* marks up to 30, i.e., total 31 characters,
* then, we terminate this degenerately long
* combining sequence with a U+034F COMBINING
* GRAPHEME JOINER (CGJ) which is 0xCD 0x8F in
* UTF-8 and turn this into a Stream-Safe
* Text. This will be extremely rare but
* possible.
*
* The following will also guarantee that
* we are not writing more than 32 characters
* plus a NULL at u8s[].
*/
if (last >= U8_UPPER_LIMIT_IN_A_SEQ) {
TURN_STREAM_SAFE:
*state = U8_STATE_START;
comb_class[last] = 0;
start[last] = saved_sz;
disp[last] = 2;
last++;
u8s[saved_sz++] = 0xCD;
u8s[saved_sz++] = 0x8F;
break;
}
/*
* Some combining marks also do decompose into
* another combining mark or marks.
*/
if (*state == U8_STATE_COMBINING_MARK) {
k = last;
l = sz;
i = do_decomp(uv, uts, s, sz,
canonical_decomposition, state);
for (j = 0; j < i; ) {
sz = u8_number_of_bytes[uts[j]];
comb_class[last] =
combining_class(uv,
uts + j, sz);
start[last] = saved_sz + j;
disp[last] = sz;
last++;
if (last >=
U8_UPPER_LIMIT_IN_A_SEQ) {
last = k;
goto TURN_STREAM_SAFE;
}
j += sz;
}
*state = U8_STATE_COMBINING_MARK;
sz = i;
s += l;
for (i = 0; i < sz; i++)
u8s[saved_sz++] = uts[i];
} else {
comb_class[last] = i;
start[last] = saved_sz;
disp[last] = sz;
last++;
for (i = 0; i < sz; i++)
u8s[saved_sz++] = *s++;
}
/*
* If this is U+0345 COMBINING GREEK
* YPOGEGRAMMENI (0xCD 0x85 in UTF-8), a.k.a.,
* iota subscript, and need to be converted to
* uppercase letter, convert it to U+0399 GREEK
* CAPITAL LETTER IOTA (0xCE 0x99 in UTF-8),
* i.e., convert to capital adscript form as
* specified in the Unicode standard.
*
* This is the only special case of (ambiguous)
* case conversion at combining marks and
* probably the standard will never have
* anything similar like this in future.
*/
if (is_it_toupper && sz >= 2 &&
u8s[saved_sz - 2] == 0xCD &&
u8s[saved_sz - 1] == 0x85) {
u8s[saved_sz - 2] = 0xCE;
u8s[saved_sz - 1] = 0x99;
}
}
}
/*
* Let's try to ensure a canonical ordering for the collected
* combining marks. We do this only if we have collected
* at least one more non-Starter. (The decomposition mapping
* data tables have fully (and recursively) expanded and
* canonically ordered decompositions.)
*
* The U8_SWAP_COMB_MARKS() convenience macro has some
* assumptions and we are meeting the assumptions.
*/
last--;
if (last >= saved_last) {
for (i = 0; i < last; i++)
for (j = last; j > i; j--)
if (comb_class[j] &&
comb_class[j - 1] > comb_class[j]) {
U8_SWAP_COMB_MARKS(j - 1, j);
}
}
*source = s;
if (! canonical_composition) {
u8s[saved_sz] = '\0';
return (saved_sz);
}
/*
* Now do the canonical composition. Note that we do this
* only after a canonical or compatibility decomposition to
* finish up NFC or NFKC.
*/
sz = do_composition(uv, u8s, comb_class, start, disp, last,
&s, slast);
}
*source = s;
return ((size_t)sz);
}
/*
* The do_norm_compare() function does string comparison based on Unicode
* simple case mappings and Unicode Normalization definitions.
*
* It does so by collecting a sequence of character at a time and comparing
* the collected sequences from the strings.
*
* The meanings on the return values are the same as the usual strcmp().
*/
static int
do_norm_compare(size_t uv, uchar_t *s1, uchar_t *s2, size_t n1, size_t n2,
int flag, int *errnum)
{
int result;
size_t sz1;
size_t sz2;
uchar_t u8s1[U8_STREAM_SAFE_TEXT_MAX + 1];
uchar_t u8s2[U8_STREAM_SAFE_TEXT_MAX + 1];
uchar_t *s1last;
uchar_t *s2last;
boolean_t is_it_toupper;
boolean_t is_it_tolower;
boolean_t canonical_decomposition;
boolean_t compatibility_decomposition;
boolean_t canonical_composition;
u8_normalization_states_t state;
s1last = s1 + n1;
s2last = s2 + n2;
is_it_toupper = flag & U8_TEXTPREP_TOUPPER;
is_it_tolower = flag & U8_TEXTPREP_TOLOWER;
canonical_decomposition = flag & U8_CANON_DECOMP;
compatibility_decomposition = flag & U8_COMPAT_DECOMP;
canonical_composition = flag & U8_CANON_COMP;
while (s1 < s1last && s2 < s2last) {
/*
* If the current character is a 7-bit ASCII and the last
* character, or, if the current character and the next
* character are both some 7-bit ASCII characters then
* we treat the current character as a sequence.
*
* In any other cases, we need to call collect_a_seq().
*/
if (U8_ISASCII(*s1) && ((s1 + 1) >= s1last ||
((s1 + 1) < s1last && U8_ISASCII(*(s1 + 1))))) {
if (is_it_toupper)
u8s1[0] = U8_ASCII_TOUPPER(*s1);
else if (is_it_tolower)
u8s1[0] = U8_ASCII_TOLOWER(*s1);
else
u8s1[0] = *s1;
u8s1[1] = '\0';
sz1 = 1;
s1++;
} else {
state = U8_STATE_START;
sz1 = collect_a_seq(uv, u8s1, &s1, s1last,
is_it_toupper, is_it_tolower,
canonical_decomposition,
compatibility_decomposition,
canonical_composition, errnum, &state);
}
if (U8_ISASCII(*s2) && ((s2 + 1) >= s2last ||
((s2 + 1) < s2last && U8_ISASCII(*(s2 + 1))))) {
if (is_it_toupper)
u8s2[0] = U8_ASCII_TOUPPER(*s2);
else if (is_it_tolower)
u8s2[0] = U8_ASCII_TOLOWER(*s2);
else
u8s2[0] = *s2;
u8s2[1] = '\0';
sz2 = 1;
s2++;
} else {
state = U8_STATE_START;
sz2 = collect_a_seq(uv, u8s2, &s2, s2last,
is_it_toupper, is_it_tolower,
canonical_decomposition,
compatibility_decomposition,
canonical_composition, errnum, &state);
}
/*
* Now compare the two characters. If they are the same,
* we move on to the next character sequences.
*/
if (sz1 == 1 && sz2 == 1) {
if (*u8s1 > *u8s2)
return (1);
if (*u8s1 < *u8s2)
return (-1);
} else {
result = strcmp((const char *)u8s1, (const char *)u8s2);
if (result != 0)
return (result);
}
}
/*
* We compared until the end of either or both strings.
*
* If we reached to or went over the ends for the both, that means
* they are the same.
*
* If we reached only one end, that means the other string has
* something which then can be used to determine the return value.
*/
if (s1 >= s1last) {
if (s2 >= s2last)
return (0);
return (-1);
}
return (1);
}
/*
* The u8_strcmp() function compares two UTF-8 strings quite similar to
* the strcmp(). For the comparison, however, Unicode Normalization specific
* equivalency and Unicode simple case conversion mappings based equivalency
* can be requested and checked against.
*/
int
u8_strcmp(const char *s1, const char *s2, size_t n, int flag, size_t uv,
int *errnum)
{
int f;
size_t n1;
size_t n2;
*errnum = 0;
/*
* Check on the requested Unicode version, case conversion, and
* normalization flag values.
*/
if (uv > U8_UNICODE_LATEST) {
*errnum = ERANGE;
uv = U8_UNICODE_LATEST;
}
if (flag == 0) {
flag = U8_STRCMP_CS;
} else {
f = flag & (U8_STRCMP_CS | U8_STRCMP_CI_UPPER |
U8_STRCMP_CI_LOWER);
if (f == 0) {
flag |= U8_STRCMP_CS;
} else if (f != U8_STRCMP_CS && f != U8_STRCMP_CI_UPPER &&
f != U8_STRCMP_CI_LOWER) {
*errnum = EBADF;
flag = U8_STRCMP_CS;
}
f = flag & (U8_CANON_DECOMP | U8_COMPAT_DECOMP | U8_CANON_COMP);
if (f && f != U8_STRCMP_NFD && f != U8_STRCMP_NFC &&
f != U8_STRCMP_NFKD && f != U8_STRCMP_NFKC) {
*errnum = EBADF;
flag = U8_STRCMP_CS;
}
}
if (flag == U8_STRCMP_CS) {
return (n == 0 ? strcmp(s1, s2) : strncmp(s1, s2, n));
}
n1 = strlen(s1);
n2 = strlen(s2);
if (n != 0) {
if (n < n1)
n1 = n;
if (n < n2)
n2 = n;
}
/*
* Simple case conversion can be done much faster and so we do
* them separately here.
*/
if (flag == U8_STRCMP_CI_UPPER) {
return (do_case_compare(uv, (uchar_t *)s1, (uchar_t *)s2,
n1, n2, B_TRUE, errnum));
} else if (flag == U8_STRCMP_CI_LOWER) {
return (do_case_compare(uv, (uchar_t *)s1, (uchar_t *)s2,
n1, n2, B_FALSE, errnum));
}
return (do_norm_compare(uv, (uchar_t *)s1, (uchar_t *)s2, n1, n2,
flag, errnum));
}
size_t
u8_textprep_str(char *inarray, size_t *inlen, char *outarray, size_t *outlen,
int flag, size_t unicode_version, int *errnum)
{
int f;
int sz;
uchar_t *ib;
uchar_t *ibtail;
uchar_t *ob;
uchar_t *obtail;
boolean_t do_not_ignore_null;
boolean_t do_not_ignore_invalid;
boolean_t is_it_toupper;
boolean_t is_it_tolower;
boolean_t canonical_decomposition;
boolean_t compatibility_decomposition;
boolean_t canonical_composition;
size_t ret_val;
size_t i;
size_t j;
uchar_t u8s[U8_STREAM_SAFE_TEXT_MAX + 1];
u8_normalization_states_t state;
if (unicode_version > U8_UNICODE_LATEST) {
*errnum = ERANGE;
return ((size_t)-1);
}
f = flag & (U8_TEXTPREP_TOUPPER | U8_TEXTPREP_TOLOWER);
if (f == (U8_TEXTPREP_TOUPPER | U8_TEXTPREP_TOLOWER)) {
*errnum = EBADF;
return ((size_t)-1);
}
f = flag & (U8_CANON_DECOMP | U8_COMPAT_DECOMP | U8_CANON_COMP);
if (f && f != U8_TEXTPREP_NFD && f != U8_TEXTPREP_NFC &&
f != U8_TEXTPREP_NFKD && f != U8_TEXTPREP_NFKC) {
*errnum = EBADF;
return ((size_t)-1);
}
if (inarray == NULL || *inlen == 0)
return (0);
if (outarray == NULL) {
*errnum = E2BIG;
return ((size_t)-1);
}
ib = (uchar_t *)inarray;
ob = (uchar_t *)outarray;
ibtail = ib + *inlen;
obtail = ob + *outlen;
do_not_ignore_null = !(flag & U8_TEXTPREP_IGNORE_NULL);
do_not_ignore_invalid = !(flag & U8_TEXTPREP_IGNORE_INVALID);
is_it_toupper = flag & U8_TEXTPREP_TOUPPER;
is_it_tolower = flag & U8_TEXTPREP_TOLOWER;
ret_val = 0;
/*
* If we don't have a normalization flag set, we do the simple case
* conversion based text preparation separately below. Text
* preparation involving Normalization will be done in the false task
* block, again, separately since it will take much more time and
* resource than doing simple case conversions.
*/
if (f == 0) {
while (ib < ibtail) {
if (*ib == '\0' && do_not_ignore_null)
break;
sz = u8_number_of_bytes[*ib];
if (sz < 0) {
if (do_not_ignore_invalid) {
*errnum = EILSEQ;
ret_val = (size_t)-1;
break;
}
sz = 1;
ret_val++;
}
if (sz == 1) {
if (ob >= obtail) {
*errnum = E2BIG;
ret_val = (size_t)-1;
break;
}
if (is_it_toupper)
*ob = U8_ASCII_TOUPPER(*ib);
else if (is_it_tolower)
*ob = U8_ASCII_TOLOWER(*ib);
else
*ob = *ib;
ib++;
ob++;
} else if ((ib + sz) > ibtail) {
if (do_not_ignore_invalid) {
*errnum = EINVAL;
ret_val = (size_t)-1;
break;
}
if ((obtail - ob) < (ibtail - ib)) {
*errnum = E2BIG;
ret_val = (size_t)-1;
break;
}
/*
* We treat the remaining incomplete character
* bytes as a character.
*/
ret_val++;
while (ib < ibtail)
*ob++ = *ib++;
} else {
if (is_it_toupper || is_it_tolower) {
i = do_case_conv(unicode_version, u8s,
ib, sz, is_it_toupper);
if ((obtail - ob) < i) {
*errnum = E2BIG;
ret_val = (size_t)-1;
break;
}
ib += sz;
for (sz = 0; sz < i; sz++)
*ob++ = u8s[sz];
} else {
if ((obtail - ob) < sz) {
*errnum = E2BIG;
ret_val = (size_t)-1;
break;
}
for (i = 0; i < sz; i++)
*ob++ = *ib++;
}
}
}
} else {
canonical_decomposition = flag & U8_CANON_DECOMP;
compatibility_decomposition = flag & U8_COMPAT_DECOMP;
canonical_composition = flag & U8_CANON_COMP;
while (ib < ibtail) {
if (*ib == '\0' && do_not_ignore_null)
break;
/*
* If the current character is a 7-bit ASCII
* character and it is the last character, or,
* if the current character is a 7-bit ASCII
* character and the next character is also a 7-bit
* ASCII character, then, we copy over this
* character without going through collect_a_seq().
*
* In any other cases, we need to look further with
* the collect_a_seq() function.
*/
if (U8_ISASCII(*ib) && ((ib + 1) >= ibtail ||
((ib + 1) < ibtail && U8_ISASCII(*(ib + 1))))) {
if (ob >= obtail) {
*errnum = E2BIG;
ret_val = (size_t)-1;
break;
}
if (is_it_toupper)
*ob = U8_ASCII_TOUPPER(*ib);
else if (is_it_tolower)
*ob = U8_ASCII_TOLOWER(*ib);
else
*ob = *ib;
ib++;
ob++;
} else {
*errnum = 0;
state = U8_STATE_START;
j = collect_a_seq(unicode_version, u8s,
&ib, ibtail,
is_it_toupper,
is_it_tolower,
canonical_decomposition,
compatibility_decomposition,
canonical_composition,
errnum, &state);
if (*errnum && do_not_ignore_invalid) {
ret_val = (size_t)-1;
break;
}
if ((obtail - ob) < j) {
*errnum = E2BIG;
ret_val = (size_t)-1;
break;
}
for (i = 0; i < j; i++)
*ob++ = u8s[i];
}
}
}
*inlen = ibtail - ib;
*outlen = obtail - ob;
return (ret_val);
}
#if defined(_KERNEL)
static int __init
unicode_init(void)
{
return (0);
}
static void __exit
unicode_fini(void)
{
}
module_init(unicode_init);
module_exit(unicode_fini);
#endif
ZFS_MODULE_DESCRIPTION("Unicode implementation");
ZFS_MODULE_AUTHOR(ZFS_META_AUTHOR);
ZFS_MODULE_LICENSE(ZFS_META_LICENSE);
ZFS_MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
EXPORT_SYMBOL(u8_validate);
EXPORT_SYMBOL(u8_strcmp);
EXPORT_SYMBOL(u8_textprep_str);
diff --git a/module/unicode/uconv.c b/module/unicode/uconv.c
index fe84979d08b2..6854aeab277f 100644
--- a/module/unicode/uconv.c
+++ b/module/unicode/uconv.c
@@ -1,863 +1,859 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Unicode encoding conversion functions among UTF-8, UTF-16, and UTF-32.
* (PSARC/2005/446, PSARC/2007/038, PSARC/2007/517)
* Man pages: uconv_u16tou32(9F), uconv_u16tou8(9F), uconv_u32tou16(9F),
* uconv_u32tou8(9F), uconv_u8tou16(9F), and uconv_u8tou32(9F). See also
* the section 3C man pages.
* Interface stability: Committed
*/
#include <sys/types.h>
#ifdef _KERNEL
#include <sys/param.h>
#include <sys/sysmacros.h>
#include <sys/debug.h>
#include <sys/kmem.h>
#include <sys/sunddi.h>
#else
#include <sys/u8_textprep.h>
#endif /* _KERNEL */
#include <sys/byteorder.h>
#include <sys/errno.h>
/*
* The max and min values of high and low surrogate pairs of UTF-16,
* UTF-16 bit shift value, bit mask, and starting value outside of BMP.
*/
#define UCONV_U16_HI_MIN (0xd800U)
#define UCONV_U16_HI_MAX (0xdbffU)
#define UCONV_U16_LO_MIN (0xdc00U)
#define UCONV_U16_LO_MAX (0xdfffU)
#define UCONV_U16_BIT_SHIFT (0x0400U)
#define UCONV_U16_BIT_MASK (0x0fffffU)
#define UCONV_U16_START (0x010000U)
/* The maximum value of Unicode coding space and ASCII coding space. */
#define UCONV_UNICODE_MAX (0x10ffffU)
#define UCONV_ASCII_MAX (0x7fU)
/* The mask values for input and output endians. */
#define UCONV_IN_ENDIAN_MASKS (UCONV_IN_BIG_ENDIAN | UCONV_IN_LITTLE_ENDIAN)
#define UCONV_OUT_ENDIAN_MASKS (UCONV_OUT_BIG_ENDIAN | UCONV_OUT_LITTLE_ENDIAN)
/* Native and reversed endian macros. */
#ifdef _ZFS_BIG_ENDIAN
#define UCONV_IN_NAT_ENDIAN UCONV_IN_BIG_ENDIAN
#define UCONV_IN_REV_ENDIAN UCONV_IN_LITTLE_ENDIAN
#define UCONV_OUT_NAT_ENDIAN UCONV_OUT_BIG_ENDIAN
#define UCONV_OUT_REV_ENDIAN UCONV_OUT_LITTLE_ENDIAN
#else
#define UCONV_IN_NAT_ENDIAN UCONV_IN_LITTLE_ENDIAN
#define UCONV_IN_REV_ENDIAN UCONV_IN_BIG_ENDIAN
#define UCONV_OUT_NAT_ENDIAN UCONV_OUT_LITTLE_ENDIAN
#define UCONV_OUT_REV_ENDIAN UCONV_OUT_BIG_ENDIAN
#endif /* _BIG_ENDIAN */
/* The Byte Order Mark (BOM) character in normal and reversed byte orderings. */
#define UCONV_BOM_NORMAL (0xfeffU)
#define UCONV_BOM_SWAPPED (0xfffeU)
#define UCONV_BOM_SWAPPED_32 (0xfffe0000U)
/* UTF-32 boundaries based on UTF-8 character byte lengths. */
#define UCONV_U8_ONE_BYTE (0x7fU)
#define UCONV_U8_TWO_BYTES (0x7ffU)
#define UCONV_U8_THREE_BYTES (0xffffU)
#define UCONV_U8_FOUR_BYTES (0x10ffffU)
/* The common minimum and maximum values at the UTF-8 character bytes. */
#define UCONV_U8_BYTE_MIN (0x80U)
#define UCONV_U8_BYTE_MAX (0xbfU)
/*
* The following "6" and "0x3f" came from "10xx xxxx" bit representation of
* UTF-8 character bytes.
*/
#define UCONV_U8_BIT_SHIFT 6
#define UCONV_U8_BIT_MASK 0x3f
/*
* The following vector shows remaining bytes in a UTF-8 character.
* Index will be the first byte of the character.
*/
static const uchar_t remaining_bytes_tbl[0x100] = {
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
/* C0 C1 C2 C3 C4 C5 C6 C7 C8 C9 CA CB CC CD CE CF */
0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
/* D0 D1 D2 D3 D4 D5 D6 D7 D8 D9 DA DB DC DD DE DF */
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
/* E0 E1 E2 E3 E4 E5 E6 E7 E8 E9 EA EB EC ED EE EF */
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
/* F0 F1 F2 F3 F4 F5 F6 F7 F8 F9 FA FB FC FD FE FF */
3, 3, 3, 3, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
};
/*
* The following is a vector of bit-masks to get used bits in
* the first byte of a UTF-8 character. Index is remaining bytes at above of
* the character.
*/
-#ifdef _KERNEL
-const uchar_t u8_masks_tbl[6] = { 0x00, 0x1f, 0x0f, 0x07, 0x03, 0x01 };
-#else
static const uchar_t u8_masks_tbl[6] = { 0x00, 0x1f, 0x0f, 0x07, 0x03, 0x01 };
-#endif /* _KERNEL */
/*
* The following two vectors are to provide valid minimum and
* maximum values for the 2'nd byte of a multibyte UTF-8 character for
* better illegal sequence checking. The index value must be the value of
* the first byte of the UTF-8 character.
*/
static const uchar_t valid_min_2nd_byte[0x100] = {
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
/* C0 C1 C2 C3 C4 C5 C6 C7 */
0, 0, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* C8 C9 CA CB CC CD CE CF */
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* D0 D1 D2 D3 D4 D5 D6 D7 */
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* D8 D9 DA DB DC DD DE DF */
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* E0 E1 E2 E3 E4 E5 E6 E7 */
0xa0, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* E8 E9 EA EB EC ED EE EF */
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
/* F0 F1 F2 F3 F4 F5 F6 F7 */
0x90, 0x80, 0x80, 0x80, 0x80, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0
};
static const uchar_t valid_max_2nd_byte[0x100] = {
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
/* C0 C1 C2 C3 C4 C5 C6 C7 */
0, 0, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* C8 C9 CA CB CC CD CE CF */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* D0 D1 D2 D3 D4 D5 D6 D7 */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* D8 D9 DA DB DC DD DE DF */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* E0 E1 E2 E3 E4 E5 E6 E7 */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
/* E8 E9 EA EB EC ED EE EF */
0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0x9f, 0xbf, 0xbf,
/* F0 F1 F2 F3 F4 F5 F6 F7 */
0xbf, 0xbf, 0xbf, 0xbf, 0x8f, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0
};
static int
check_endian(int flag, int *in, int *out)
{
*in = flag & UCONV_IN_ENDIAN_MASKS;
/* You cannot have both. */
if (*in == UCONV_IN_ENDIAN_MASKS)
return (EBADF);
if (*in == 0)
*in = UCONV_IN_NAT_ENDIAN;
*out = flag & UCONV_OUT_ENDIAN_MASKS;
/* You cannot have both. */
if (*out == UCONV_OUT_ENDIAN_MASKS)
return (EBADF);
if (*out == 0)
*out = UCONV_OUT_NAT_ENDIAN;
return (0);
}
static boolean_t
check_bom16(const uint16_t *u16s, size_t u16l, int *in)
{
if (u16l > 0) {
if (*u16s == UCONV_BOM_NORMAL) {
*in = UCONV_IN_NAT_ENDIAN;
return (B_TRUE);
}
if (*u16s == UCONV_BOM_SWAPPED) {
*in = UCONV_IN_REV_ENDIAN;
return (B_TRUE);
}
}
return (B_FALSE);
}
static boolean_t
check_bom32(const uint32_t *u32s, size_t u32l, int *in)
{
if (u32l > 0) {
if (*u32s == UCONV_BOM_NORMAL) {
*in = UCONV_IN_NAT_ENDIAN;
return (B_TRUE);
}
if (*u32s == UCONV_BOM_SWAPPED_32) {
*in = UCONV_IN_REV_ENDIAN;
return (B_TRUE);
}
}
return (B_FALSE);
}
int
uconv_u16tou32(const uint16_t *u16s, size_t *utf16len,
uint32_t *u32s, size_t *utf32len, int flag)
{
int inendian;
int outendian;
size_t u16l;
size_t u32l;
uint32_t hi;
uint32_t lo;
boolean_t do_not_ignore_null;
/*
* Do preliminary validity checks on parameters and collect info on
* endians.
*/
if (u16s == NULL || utf16len == NULL)
return (EILSEQ);
if (u32s == NULL || utf32len == NULL)
return (E2BIG);
if (check_endian(flag, &inendian, &outendian) != 0)
return (EBADF);
/*
* Initialize input and output parameter buffer indices and
* temporary variables.
*/
u16l = u32l = 0;
hi = 0;
do_not_ignore_null = ((flag & UCONV_IGNORE_NULL) == 0);
/*
* Check on the BOM at the beginning of the input buffer if required
* and if there is indeed one, process it.
*/
if ((flag & UCONV_IN_ACCEPT_BOM) &&
check_bom16(u16s, *utf16len, &inendian))
u16l++;
/*
* Reset inendian and outendian so that after this point, those can be
* used as condition values.
*/
inendian &= UCONV_IN_NAT_ENDIAN;
outendian &= UCONV_OUT_NAT_ENDIAN;
/*
* If there is something in the input buffer and if necessary and
* requested, save the BOM at the output buffer.
*/
if (*utf16len > 0 && *utf32len > 0 && (flag & UCONV_OUT_EMIT_BOM))
u32s[u32l++] = (outendian) ? UCONV_BOM_NORMAL :
UCONV_BOM_SWAPPED_32;
/*
* Do conversion; if encounter a surrogate pair, assemble high and
* low pair values to form a UTF-32 character. If a half of a pair
* exists alone, then, either it is an illegal (EILSEQ) or
* invalid (EINVAL) value.
*/
for (; u16l < *utf16len; u16l++) {
if (u16s[u16l] == 0 && do_not_ignore_null)
break;
lo = (uint32_t)((inendian) ? u16s[u16l] : BSWAP_16(u16s[u16l]));
if (lo >= UCONV_U16_HI_MIN && lo <= UCONV_U16_HI_MAX) {
if (hi)
return (EILSEQ);
hi = lo;
continue;
} else if (lo >= UCONV_U16_LO_MIN && lo <= UCONV_U16_LO_MAX) {
if (! hi)
return (EILSEQ);
lo = (((hi - UCONV_U16_HI_MIN) * UCONV_U16_BIT_SHIFT +
lo - UCONV_U16_LO_MIN) & UCONV_U16_BIT_MASK)
+ UCONV_U16_START;
hi = 0;
} else if (hi) {
return (EILSEQ);
}
if (u32l >= *utf32len)
return (E2BIG);
u32s[u32l++] = (outendian) ? lo : BSWAP_32(lo);
}
/*
* If high half didn't see low half, then, it's most likely the input
* parameter is incomplete.
*/
if (hi)
return (EINVAL);
/*
* Save the number of consumed and saved characters. They do not
* include terminating NULL character (U+0000) at the end of
* the input buffer (even when UCONV_IGNORE_NULL isn't specified and
* the input buffer length is big enough to include the terminating
* NULL character).
*/
*utf16len = u16l;
*utf32len = u32l;
return (0);
}
int
uconv_u16tou8(const uint16_t *u16s, size_t *utf16len,
uchar_t *u8s, size_t *utf8len, int flag)
{
int inendian;
int outendian;
size_t u16l;
size_t u8l;
uint32_t hi;
uint32_t lo;
boolean_t do_not_ignore_null;
if (u16s == NULL || utf16len == NULL)
return (EILSEQ);
if (u8s == NULL || utf8len == NULL)
return (E2BIG);
if (check_endian(flag, &inendian, &outendian) != 0)
return (EBADF);
u16l = u8l = 0;
hi = 0;
do_not_ignore_null = ((flag & UCONV_IGNORE_NULL) == 0);
if ((flag & UCONV_IN_ACCEPT_BOM) &&
check_bom16(u16s, *utf16len, &inendian))
u16l++;
inendian &= UCONV_IN_NAT_ENDIAN;
for (; u16l < *utf16len; u16l++) {
if (u16s[u16l] == 0 && do_not_ignore_null)
break;
lo = (uint32_t)((inendian) ? u16s[u16l] : BSWAP_16(u16s[u16l]));
if (lo >= UCONV_U16_HI_MIN && lo <= UCONV_U16_HI_MAX) {
if (hi)
return (EILSEQ);
hi = lo;
continue;
} else if (lo >= UCONV_U16_LO_MIN && lo <= UCONV_U16_LO_MAX) {
if (! hi)
return (EILSEQ);
lo = (((hi - UCONV_U16_HI_MIN) * UCONV_U16_BIT_SHIFT +
lo - UCONV_U16_LO_MIN) & UCONV_U16_BIT_MASK)
+ UCONV_U16_START;
hi = 0;
} else if (hi) {
return (EILSEQ);
}
/*
* Now we convert a UTF-32 character into a UTF-8 character.
* Unicode coding space is between U+0000 and U+10FFFF;
* anything bigger is an illegal character.
*/
if (lo <= UCONV_U8_ONE_BYTE) {
if (u8l >= *utf8len)
return (E2BIG);
u8s[u8l++] = (uchar_t)lo;
} else if (lo <= UCONV_U8_TWO_BYTES) {
if ((u8l + 1) >= *utf8len)
return (E2BIG);
u8s[u8l++] = (uchar_t)(0xc0 | ((lo & 0x07c0) >> 6));
u8s[u8l++] = (uchar_t)(0x80 | (lo & 0x003f));
} else if (lo <= UCONV_U8_THREE_BYTES) {
if ((u8l + 2) >= *utf8len)
return (E2BIG);
u8s[u8l++] = (uchar_t)(0xe0 | ((lo & 0x0f000) >> 12));
u8s[u8l++] = (uchar_t)(0x80 | ((lo & 0x00fc0) >> 6));
u8s[u8l++] = (uchar_t)(0x80 | (lo & 0x0003f));
} else if (lo <= UCONV_U8_FOUR_BYTES) {
if ((u8l + 3) >= *utf8len)
return (E2BIG);
u8s[u8l++] = (uchar_t)(0xf0 | ((lo & 0x01c0000) >> 18));
u8s[u8l++] = (uchar_t)(0x80 | ((lo & 0x003f000) >> 12));
u8s[u8l++] = (uchar_t)(0x80 | ((lo & 0x0000fc0) >> 6));
u8s[u8l++] = (uchar_t)(0x80 | (lo & 0x000003f));
} else {
return (EILSEQ);
}
}
if (hi)
return (EINVAL);
*utf16len = u16l;
*utf8len = u8l;
return (0);
}
int
uconv_u32tou16(const uint32_t *u32s, size_t *utf32len,
uint16_t *u16s, size_t *utf16len, int flag)
{
int inendian;
int outendian;
size_t u16l;
size_t u32l;
uint32_t hi;
uint32_t lo;
boolean_t do_not_ignore_null;
if (u32s == NULL || utf32len == NULL)
return (EILSEQ);
if (u16s == NULL || utf16len == NULL)
return (E2BIG);
if (check_endian(flag, &inendian, &outendian) != 0)
return (EBADF);
u16l = u32l = 0;
do_not_ignore_null = ((flag & UCONV_IGNORE_NULL) == 0);
if ((flag & UCONV_IN_ACCEPT_BOM) &&
check_bom32(u32s, *utf32len, &inendian))
u32l++;
inendian &= UCONV_IN_NAT_ENDIAN;
outendian &= UCONV_OUT_NAT_ENDIAN;
if (*utf32len > 0 && *utf16len > 0 && (flag & UCONV_OUT_EMIT_BOM))
u16s[u16l++] = (outendian) ? UCONV_BOM_NORMAL :
UCONV_BOM_SWAPPED;
for (; u32l < *utf32len; u32l++) {
if (u32s[u32l] == 0 && do_not_ignore_null)
break;
hi = (inendian) ? u32s[u32l] : BSWAP_32(u32s[u32l]);
/*
* Anything bigger than the Unicode coding space, i.e.,
* Unicode scalar value bigger than U+10FFFF, is an illegal
* character.
*/
if (hi > UCONV_UNICODE_MAX)
return (EILSEQ);
/*
* Anything bigger than U+FFFF must be converted into
* a surrogate pair in UTF-16.
*/
if (hi >= UCONV_U16_START) {
lo = ((hi - UCONV_U16_START) % UCONV_U16_BIT_SHIFT) +
UCONV_U16_LO_MIN;
hi = ((hi - UCONV_U16_START) / UCONV_U16_BIT_SHIFT) +
UCONV_U16_HI_MIN;
if ((u16l + 1) >= *utf16len)
return (E2BIG);
if (outendian) {
u16s[u16l++] = (uint16_t)hi;
u16s[u16l++] = (uint16_t)lo;
} else {
u16s[u16l++] = BSWAP_16(((uint16_t)hi));
u16s[u16l++] = BSWAP_16(((uint16_t)lo));
}
} else {
if (u16l >= *utf16len)
return (E2BIG);
u16s[u16l++] = (outendian) ? (uint16_t)hi :
BSWAP_16(((uint16_t)hi));
}
}
*utf16len = u16l;
*utf32len = u32l;
return (0);
}
int
uconv_u32tou8(const uint32_t *u32s, size_t *utf32len,
uchar_t *u8s, size_t *utf8len, int flag)
{
int inendian;
int outendian;
size_t u32l;
size_t u8l;
uint32_t lo;
boolean_t do_not_ignore_null;
if (u32s == NULL || utf32len == NULL)
return (EILSEQ);
if (u8s == NULL || utf8len == NULL)
return (E2BIG);
if (check_endian(flag, &inendian, &outendian) != 0)
return (EBADF);
u32l = u8l = 0;
do_not_ignore_null = ((flag & UCONV_IGNORE_NULL) == 0);
if ((flag & UCONV_IN_ACCEPT_BOM) &&
check_bom32(u32s, *utf32len, &inendian))
u32l++;
inendian &= UCONV_IN_NAT_ENDIAN;
for (; u32l < *utf32len; u32l++) {
if (u32s[u32l] == 0 && do_not_ignore_null)
break;
lo = (inendian) ? u32s[u32l] : BSWAP_32(u32s[u32l]);
if (lo <= UCONV_U8_ONE_BYTE) {
if (u8l >= *utf8len)
return (E2BIG);
u8s[u8l++] = (uchar_t)lo;
} else if (lo <= UCONV_U8_TWO_BYTES) {
if ((u8l + 1) >= *utf8len)
return (E2BIG);
u8s[u8l++] = (uchar_t)(0xc0 | ((lo & 0x07c0) >> 6));
u8s[u8l++] = (uchar_t)(0x80 | (lo & 0x003f));
} else if (lo <= UCONV_U8_THREE_BYTES) {
if ((u8l + 2) >= *utf8len)
return (E2BIG);
u8s[u8l++] = (uchar_t)(0xe0 | ((lo & 0x0f000) >> 12));
u8s[u8l++] = (uchar_t)(0x80 | ((lo & 0x00fc0) >> 6));
u8s[u8l++] = (uchar_t)(0x80 | (lo & 0x0003f));
} else if (lo <= UCONV_U8_FOUR_BYTES) {
if ((u8l + 3) >= *utf8len)
return (E2BIG);
u8s[u8l++] = (uchar_t)(0xf0 | ((lo & 0x01c0000) >> 18));
u8s[u8l++] = (uchar_t)(0x80 | ((lo & 0x003f000) >> 12));
u8s[u8l++] = (uchar_t)(0x80 | ((lo & 0x0000fc0) >> 6));
u8s[u8l++] = (uchar_t)(0x80 | (lo & 0x000003f));
} else {
return (EILSEQ);
}
}
*utf32len = u32l;
*utf8len = u8l;
return (0);
}
int
uconv_u8tou16(const uchar_t *u8s, size_t *utf8len,
uint16_t *u16s, size_t *utf16len, int flag)
{
int inendian;
int outendian;
size_t u16l;
size_t u8l;
uint32_t hi;
uint32_t lo;
int remaining_bytes;
int first_b;
boolean_t do_not_ignore_null;
if (u8s == NULL || utf8len == NULL)
return (EILSEQ);
if (u16s == NULL || utf16len == NULL)
return (E2BIG);
if (check_endian(flag, &inendian, &outendian) != 0)
return (EBADF);
u16l = u8l = 0;
do_not_ignore_null = ((flag & UCONV_IGNORE_NULL) == 0);
outendian &= UCONV_OUT_NAT_ENDIAN;
if (*utf8len > 0 && *utf16len > 0 && (flag & UCONV_OUT_EMIT_BOM))
u16s[u16l++] = (outendian) ? UCONV_BOM_NORMAL :
UCONV_BOM_SWAPPED;
for (; u8l < *utf8len; ) {
if (u8s[u8l] == 0 && do_not_ignore_null)
break;
/*
* Collect a UTF-8 character and convert it to a UTF-32
* character. In doing so, we screen out illegally formed
* UTF-8 characters and treat such as illegal characters.
* The algorithm at below also screens out anything bigger
* than the U+10FFFF.
*
* See Unicode 3.1 UTF-8 Corrigendum and Unicode 3.2 for
* more details on the illegal values of UTF-8 character
* bytes.
*/
hi = (uint32_t)u8s[u8l++];
if (hi > UCONV_ASCII_MAX) {
if ((remaining_bytes = remaining_bytes_tbl[hi]) == 0)
return (EILSEQ);
first_b = hi;
hi = hi & u8_masks_tbl[remaining_bytes];
for (; remaining_bytes > 0; remaining_bytes--) {
/*
* If we have no more bytes, the current
* UTF-8 character is incomplete.
*/
if (u8l >= *utf8len)
return (EINVAL);
lo = (uint32_t)u8s[u8l++];
if (first_b) {
if (lo < valid_min_2nd_byte[first_b] ||
lo > valid_max_2nd_byte[first_b])
return (EILSEQ);
first_b = 0;
} else if (lo < UCONV_U8_BYTE_MIN ||
lo > UCONV_U8_BYTE_MAX) {
return (EILSEQ);
}
hi = (hi << UCONV_U8_BIT_SHIFT) |
(lo & UCONV_U8_BIT_MASK);
}
}
if (hi >= UCONV_U16_START) {
lo = ((hi - UCONV_U16_START) % UCONV_U16_BIT_SHIFT) +
UCONV_U16_LO_MIN;
hi = ((hi - UCONV_U16_START) / UCONV_U16_BIT_SHIFT) +
UCONV_U16_HI_MIN;
if ((u16l + 1) >= *utf16len)
return (E2BIG);
if (outendian) {
u16s[u16l++] = (uint16_t)hi;
u16s[u16l++] = (uint16_t)lo;
} else {
u16s[u16l++] = BSWAP_16(((uint16_t)hi));
u16s[u16l++] = BSWAP_16(((uint16_t)lo));
}
} else {
if (u16l >= *utf16len)
return (E2BIG);
u16s[u16l++] = (outendian) ? (uint16_t)hi :
BSWAP_16(((uint16_t)hi));
}
}
*utf16len = u16l;
*utf8len = u8l;
return (0);
}
int
uconv_u8tou32(const uchar_t *u8s, size_t *utf8len,
uint32_t *u32s, size_t *utf32len, int flag)
{
int inendian;
int outendian;
size_t u32l;
size_t u8l;
uint32_t hi;
uint32_t c;
int remaining_bytes;
int first_b;
boolean_t do_not_ignore_null;
if (u8s == NULL || utf8len == NULL)
return (EILSEQ);
if (u32s == NULL || utf32len == NULL)
return (E2BIG);
if (check_endian(flag, &inendian, &outendian) != 0)
return (EBADF);
u32l = u8l = 0;
do_not_ignore_null = ((flag & UCONV_IGNORE_NULL) == 0);
outendian &= UCONV_OUT_NAT_ENDIAN;
if (*utf8len > 0 && *utf32len > 0 && (flag & UCONV_OUT_EMIT_BOM))
u32s[u32l++] = (outendian) ? UCONV_BOM_NORMAL :
UCONV_BOM_SWAPPED_32;
for (; u8l < *utf8len; ) {
if (u8s[u8l] == 0 && do_not_ignore_null)
break;
hi = (uint32_t)u8s[u8l++];
if (hi > UCONV_ASCII_MAX) {
if ((remaining_bytes = remaining_bytes_tbl[hi]) == 0)
return (EILSEQ);
first_b = hi;
hi = hi & u8_masks_tbl[remaining_bytes];
for (; remaining_bytes > 0; remaining_bytes--) {
if (u8l >= *utf8len)
return (EINVAL);
c = (uint32_t)u8s[u8l++];
if (first_b) {
if (c < valid_min_2nd_byte[first_b] ||
c > valid_max_2nd_byte[first_b])
return (EILSEQ);
first_b = 0;
} else if (c < UCONV_U8_BYTE_MIN ||
c > UCONV_U8_BYTE_MAX) {
return (EILSEQ);
}
hi = (hi << UCONV_U8_BIT_SHIFT) |
(c & UCONV_U8_BIT_MASK);
}
}
if (u32l >= *utf32len)
return (E2BIG);
u32s[u32l++] = (outendian) ? hi : BSWAP_32(hi);
}
*utf32len = u32l;
*utf8len = u8l;
return (0);
}
#if defined(_KERNEL)
EXPORT_SYMBOL(uconv_u16tou32);
EXPORT_SYMBOL(uconv_u16tou8);
EXPORT_SYMBOL(uconv_u32tou16);
EXPORT_SYMBOL(uconv_u32tou8);
EXPORT_SYMBOL(uconv_u8tou16);
EXPORT_SYMBOL(uconv_u8tou32);
#endif
diff --git a/module/zcommon/zfeature_common.c b/module/zcommon/zfeature_common.c
index c7278fa0076a..529c52316f3e 100644
--- a/module/zcommon/zfeature_common.c
+++ b/module/zcommon/zfeature_common.c
@@ -1,709 +1,709 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
* Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2017, Intel Corporation.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
*/
#ifndef _KERNEL
#include <errno.h>
#include <string.h>
#include <dirent.h>
#include <search.h>
#include <sys/stat.h>
#endif
#include <sys/debug.h>
#include <sys/fs/zfs.h>
#include <sys/inttypes.h>
#include <sys/types.h>
#include <sys/param.h>
#include <sys/zfs_sysfs.h>
#include "zfeature_common.h"
/*
* Set to disable all feature checks while opening pools, allowing pools with
* unsupported features to be opened. Set for testing only.
*/
boolean_t zfeature_checks_disable = B_FALSE;
zfeature_info_t spa_feature_table[SPA_FEATURES];
/*
* Valid characters for feature guids. This list is mainly for aesthetic
* purposes and could be expanded in the future. There are different allowed
* characters in the guids reverse dns portion (before the colon) and its
* short name (after the colon).
*/
static int
valid_char(char c, boolean_t after_colon)
{
return ((c >= 'a' && c <= 'z') ||
(c >= '0' && c <= '9') ||
(after_colon && c == '_') ||
(!after_colon && (c == '.' || c == '-')));
}
/*
* Every feature guid must contain exactly one colon which separates a reverse
* dns organization name from the feature's "short" name (e.g.
* "com.company:feature_name").
*/
boolean_t
zfeature_is_valid_guid(const char *name)
{
int i;
boolean_t has_colon = B_FALSE;
i = 0;
while (name[i] != '\0') {
char c = name[i++];
if (c == ':') {
if (has_colon)
return (B_FALSE);
has_colon = B_TRUE;
continue;
}
if (!valid_char(c, has_colon))
return (B_FALSE);
}
return (has_colon);
}
boolean_t
zfeature_is_supported(const char *guid)
{
if (zfeature_checks_disable)
return (B_TRUE);
for (spa_feature_t i = 0; i < SPA_FEATURES; i++) {
zfeature_info_t *feature = &spa_feature_table[i];
if (!feature->fi_zfs_mod_supported)
continue;
if (strcmp(guid, feature->fi_guid) == 0)
return (B_TRUE);
}
return (B_FALSE);
}
int
zfeature_lookup_guid(const char *guid, spa_feature_t *res)
{
for (spa_feature_t i = 0; i < SPA_FEATURES; i++) {
zfeature_info_t *feature = &spa_feature_table[i];
if (!feature->fi_zfs_mod_supported)
continue;
if (strcmp(guid, feature->fi_guid) == 0) {
if (res != NULL)
*res = i;
return (0);
}
}
return (ENOENT);
}
int
zfeature_lookup_name(const char *name, spa_feature_t *res)
{
for (spa_feature_t i = 0; i < SPA_FEATURES; i++) {
zfeature_info_t *feature = &spa_feature_table[i];
if (!feature->fi_zfs_mod_supported)
continue;
if (strcmp(name, feature->fi_uname) == 0) {
if (res != NULL)
*res = i;
return (0);
}
}
return (ENOENT);
}
boolean_t
zfeature_depends_on(spa_feature_t fid, spa_feature_t check)
{
zfeature_info_t *feature = &spa_feature_table[fid];
for (int i = 0; feature->fi_depends[i] != SPA_FEATURE_NONE; i++) {
if (feature->fi_depends[i] == check)
return (B_TRUE);
}
return (B_FALSE);
}
static boolean_t
deps_contains_feature(const spa_feature_t *deps, const spa_feature_t feature)
{
for (int i = 0; deps[i] != SPA_FEATURE_NONE; i++)
if (deps[i] == feature)
return (B_TRUE);
return (B_FALSE);
}
#define STRCMP ((int(*)(const void *, const void *))&strcmp)
struct zfs_mod_supported_features {
void *tree;
boolean_t all_features;
};
struct zfs_mod_supported_features *
zfs_mod_list_supported(const char *scope)
{
#if defined(__FreeBSD__) || defined(_KERNEL) || defined(LIB_ZPOOL_BUILD)
(void) scope;
return (NULL);
#else
struct zfs_mod_supported_features *ret = calloc(1, sizeof (*ret));
if (ret == NULL)
return (NULL);
DIR *sysfs_dir = NULL;
char path[128];
if (snprintf(path, sizeof (path), "%s/%s",
ZFS_SYSFS_DIR, scope) < sizeof (path))
sysfs_dir = opendir(path);
if (sysfs_dir == NULL && errno == ENOENT) {
if (snprintf(path, sizeof (path), "%s/%s",
ZFS_SYSFS_ALT_DIR, scope) < sizeof (path))
sysfs_dir = opendir(path);
}
if (sysfs_dir == NULL) {
ret->all_features = errno == ENOENT &&
(access(ZFS_SYSFS_DIR, F_OK) == 0 ||
access(ZFS_SYSFS_ALT_DIR, F_OK) == 0);
return (ret);
}
struct dirent *node;
while ((node = readdir(sysfs_dir)) != NULL) {
if (strcmp(node->d_name, ".") == 0 ||
strcmp(node->d_name, "..") == 0)
continue;
char *name = strdup(node->d_name);
if (name == NULL) {
goto nomem;
}
if (tsearch(name, &ret->tree, STRCMP) == NULL) {
/*
* Don't bother checking for duplicate entries:
* we're iterating a single directory.
*/
free(name);
goto nomem;
}
}
end:
closedir(sysfs_dir);
return (ret);
nomem:
zfs_mod_list_supported_free(ret);
ret = NULL;
goto end;
#endif
}
void
zfs_mod_list_supported_free(struct zfs_mod_supported_features *list)
{
#if !defined(__FreeBSD__) && !defined(_KERNEL) && !defined(LIB_ZPOOL_BUILD)
if (list) {
tdestroy(list->tree, free);
free(list);
}
#else
(void) list;
#endif
}
#if !defined(_KERNEL) && !defined(LIB_ZPOOL_BUILD)
static boolean_t
zfs_mod_supported_impl(const char *scope, const char *name, const char *sysfs)
{
char path[128];
if (snprintf(path, sizeof (path), "%s%s%s%s%s", sysfs,
scope == NULL ? "" : "/", scope ?: "",
name == NULL ? "" : "/", name ?: "") < sizeof (path))
return (access(path, F_OK) == 0);
else
return (B_FALSE);
}
boolean_t
zfs_mod_supported(const char *scope, const char *name,
const struct zfs_mod_supported_features *sfeatures)
{
boolean_t supported;
if (sfeatures != NULL)
return (sfeatures->all_features ||
tfind(name, &sfeatures->tree, STRCMP));
/*
* Check both the primary and alternate sysfs locations to determine
* if the required functionality is supported.
*/
supported = (zfs_mod_supported_impl(scope, name, ZFS_SYSFS_DIR) ||
zfs_mod_supported_impl(scope, name, ZFS_SYSFS_ALT_DIR));
/*
* For backwards compatibility with kernel modules that predate
* supported feature/property checking. Report the feature/property
* as supported if the kernel module is loaded but the requested
* scope directory does not exist.
*/
if (supported == B_FALSE) {
if ((access(ZFS_SYSFS_DIR, F_OK) == 0 &&
!zfs_mod_supported_impl(scope, NULL, ZFS_SYSFS_DIR)) ||
(access(ZFS_SYSFS_ALT_DIR, F_OK) == 0 &&
!zfs_mod_supported_impl(scope, NULL, ZFS_SYSFS_ALT_DIR))) {
supported = B_TRUE;
}
}
return (supported);
}
#endif
static boolean_t
zfs_mod_supported_feature(const char *name,
const struct zfs_mod_supported_features *sfeatures)
{
/*
* The zfs module spa_feature_table[], whether in-kernel or in
* libzpool, always supports all the features. libzfs needs to
* query the running module, via sysfs, to determine which
* features are supported.
*
* The equivalent _can_ be done on FreeBSD by way of the sysctl
* tree, but this has not been done yet. Therefore, we return
* that all features are supported.
*/
#if defined(_KERNEL) || defined(LIB_ZPOOL_BUILD) || defined(__FreeBSD__)
(void) name, (void) sfeatures;
return (B_TRUE);
#else
return (zfs_mod_supported(ZFS_SYSFS_POOL_FEATURES, name, sfeatures));
#endif
}
static void
zfeature_register(spa_feature_t fid, const char *guid, const char *name,
const char *desc, zfeature_flags_t flags, zfeature_type_t type,
const spa_feature_t *deps,
const struct zfs_mod_supported_features *sfeatures)
{
zfeature_info_t *feature = &spa_feature_table[fid];
- static spa_feature_t nodeps[] = { SPA_FEATURE_NONE };
+ static const spa_feature_t nodeps[] = { SPA_FEATURE_NONE };
ASSERT(name != NULL);
ASSERT(desc != NULL);
ASSERT((flags & ZFEATURE_FLAG_READONLY_COMPAT) == 0 ||
(flags & ZFEATURE_FLAG_MOS) == 0);
ASSERT3U(fid, <, SPA_FEATURES);
ASSERT(zfeature_is_valid_guid(guid));
if (deps == NULL)
deps = nodeps;
VERIFY(((flags & ZFEATURE_FLAG_PER_DATASET) == 0) ||
(deps_contains_feature(deps, SPA_FEATURE_EXTENSIBLE_DATASET)));
feature->fi_feature = fid;
feature->fi_guid = guid;
feature->fi_uname = name;
feature->fi_desc = desc;
feature->fi_flags = flags;
feature->fi_type = type;
feature->fi_depends = deps;
feature->fi_zfs_mod_supported =
zfs_mod_supported_feature(guid, sfeatures);
}
/*
* Every feature has a GUID of the form com.example:feature_name. The
* reversed DNS name ensures that the feature's GUID is unique across all ZFS
* implementations. This allows companies to independently develop and
* release features. Examples include org.delphix and org.datto. Previously,
* features developed on one implementation have used that implementation's
* domain name (e.g. org.illumos and org.zfsonlinux). Use of the org.openzfs
* domain name is recommended for new features which are developed by the
* OpenZFS community and its platforms. This domain may optionally be used by
* companies developing features for initial release through an OpenZFS
* implementation. Use of the org.openzfs domain requires reserving the
* feature name in advance with the OpenZFS project.
*/
void
zpool_feature_init(void)
{
struct zfs_mod_supported_features *sfeatures =
zfs_mod_list_supported(ZFS_SYSFS_POOL_FEATURES);
zfeature_register(SPA_FEATURE_ASYNC_DESTROY,
"com.delphix:async_destroy", "async_destroy",
"Destroy filesystems asynchronously.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL,
sfeatures);
zfeature_register(SPA_FEATURE_EMPTY_BPOBJ,
"com.delphix:empty_bpobj", "empty_bpobj",
"Snapshots use less space.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL,
sfeatures);
zfeature_register(SPA_FEATURE_LZ4_COMPRESS,
"org.illumos:lz4_compress", "lz4_compress",
"LZ4 compression algorithm support.",
ZFEATURE_FLAG_ACTIVATE_ON_ENABLE, ZFEATURE_TYPE_BOOLEAN, NULL,
sfeatures);
zfeature_register(SPA_FEATURE_MULTI_VDEV_CRASH_DUMP,
"com.joyent:multi_vdev_crash_dump", "multi_vdev_crash_dump",
"Crash dumps to multiple vdev pools.",
0, ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures);
zfeature_register(SPA_FEATURE_SPACEMAP_HISTOGRAM,
"com.delphix:spacemap_histogram", "spacemap_histogram",
"Spacemaps maintain space histograms.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL,
sfeatures);
zfeature_register(SPA_FEATURE_ENABLED_TXG,
"com.delphix:enabled_txg", "enabled_txg",
"Record txg at which a feature is enabled",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL,
sfeatures);
{
static const spa_feature_t hole_birth_deps[] = {
SPA_FEATURE_ENABLED_TXG,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_HOLE_BIRTH,
"com.delphix:hole_birth", "hole_birth",
"Retain hole birth txg for more precise zfs send",
ZFEATURE_FLAG_MOS | ZFEATURE_FLAG_ACTIVATE_ON_ENABLE,
ZFEATURE_TYPE_BOOLEAN, hole_birth_deps, sfeatures);
}
zfeature_register(SPA_FEATURE_POOL_CHECKPOINT,
"com.delphix:zpool_checkpoint", "zpool_checkpoint",
"Pool state can be checkpointed, allowing rewind later.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL,
sfeatures);
zfeature_register(SPA_FEATURE_SPACEMAP_V2,
"com.delphix:spacemap_v2", "spacemap_v2",
"Space maps representing large segments are more efficient.",
ZFEATURE_FLAG_READONLY_COMPAT | ZFEATURE_FLAG_ACTIVATE_ON_ENABLE,
ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures);
zfeature_register(SPA_FEATURE_EXTENSIBLE_DATASET,
"com.delphix:extensible_dataset", "extensible_dataset",
"Enhanced dataset functionality, used by other features.",
0, ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures);
{
static const spa_feature_t bookmarks_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_BOOKMARKS,
"com.delphix:bookmarks", "bookmarks",
"\"zfs bookmark\" command",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN,
bookmarks_deps, sfeatures);
}
{
static const spa_feature_t filesystem_limits_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_FS_SS_LIMIT,
"com.joyent:filesystem_limits", "filesystem_limits",
"Filesystem and snapshot limits.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN,
filesystem_limits_deps, sfeatures);
}
zfeature_register(SPA_FEATURE_EMBEDDED_DATA,
"com.delphix:embedded_data", "embedded_data",
"Blocks which compress very well use even less space.",
ZFEATURE_FLAG_MOS | ZFEATURE_FLAG_ACTIVATE_ON_ENABLE,
ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures);
{
static const spa_feature_t livelist_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_LIVELIST,
"com.delphix:livelist", "livelist",
"Improved clone deletion performance.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN,
livelist_deps, sfeatures);
}
{
static const spa_feature_t log_spacemap_deps[] = {
SPA_FEATURE_SPACEMAP_V2,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_LOG_SPACEMAP,
"com.delphix:log_spacemap", "log_spacemap",
"Log metaslab changes on a single spacemap and "
"flush them periodically.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN,
log_spacemap_deps, sfeatures);
}
{
static const spa_feature_t large_blocks_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_LARGE_BLOCKS,
"org.open-zfs:large_blocks", "large_blocks",
"Support for blocks larger than 128KB.",
ZFEATURE_FLAG_PER_DATASET, ZFEATURE_TYPE_BOOLEAN,
large_blocks_deps, sfeatures);
}
{
static const spa_feature_t large_dnode_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_LARGE_DNODE,
"org.zfsonlinux:large_dnode", "large_dnode",
"Variable on-disk size of dnodes.",
ZFEATURE_FLAG_PER_DATASET, ZFEATURE_TYPE_BOOLEAN,
large_dnode_deps, sfeatures);
}
{
static const spa_feature_t sha512_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_SHA512,
"org.illumos:sha512", "sha512",
"SHA-512/256 hash algorithm.",
ZFEATURE_FLAG_PER_DATASET, ZFEATURE_TYPE_BOOLEAN,
sha512_deps, sfeatures);
}
{
static const spa_feature_t skein_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_SKEIN,
"org.illumos:skein", "skein",
"Skein hash algorithm.",
ZFEATURE_FLAG_PER_DATASET, ZFEATURE_TYPE_BOOLEAN,
skein_deps, sfeatures);
}
{
static const spa_feature_t edonr_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_EDONR,
"org.illumos:edonr", "edonr",
"Edon-R hash algorithm.",
ZFEATURE_FLAG_PER_DATASET, ZFEATURE_TYPE_BOOLEAN,
edonr_deps, sfeatures);
}
{
static const spa_feature_t redact_books_deps[] = {
SPA_FEATURE_BOOKMARK_V2,
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_BOOKMARKS,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_REDACTION_BOOKMARKS,
"com.delphix:redaction_bookmarks", "redaction_bookmarks",
"Support for bookmarks which store redaction lists for zfs "
"redacted send/recv.", 0, ZFEATURE_TYPE_BOOLEAN,
redact_books_deps, sfeatures);
}
{
static const spa_feature_t redact_datasets_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_REDACTED_DATASETS,
"com.delphix:redacted_datasets", "redacted_datasets",
"Support for redacted datasets, produced by receiving "
"a redacted zfs send stream.",
ZFEATURE_FLAG_PER_DATASET, ZFEATURE_TYPE_UINT64_ARRAY,
redact_datasets_deps, sfeatures);
}
{
static const spa_feature_t bookmark_written_deps[] = {
SPA_FEATURE_BOOKMARK_V2,
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_BOOKMARKS,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_BOOKMARK_WRITTEN,
"com.delphix:bookmark_written", "bookmark_written",
"Additional accounting, enabling the written#<bookmark> "
"property (space written since a bookmark), "
"and estimates of send stream sizes for incrementals from "
"bookmarks.",
0, ZFEATURE_TYPE_BOOLEAN, bookmark_written_deps, sfeatures);
}
zfeature_register(SPA_FEATURE_DEVICE_REMOVAL,
"com.delphix:device_removal", "device_removal",
"Top-level vdevs can be removed, reducing logical pool size.",
ZFEATURE_FLAG_MOS, ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures);
{
static const spa_feature_t obsolete_counts_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_DEVICE_REMOVAL,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_OBSOLETE_COUNTS,
"com.delphix:obsolete_counts", "obsolete_counts",
"Reduce memory used by removed devices when their blocks "
"are freed or remapped.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN,
obsolete_counts_deps, sfeatures);
}
{
static const spa_feature_t userobj_accounting_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_USEROBJ_ACCOUNTING,
"org.zfsonlinux:userobj_accounting", "userobj_accounting",
"User/Group object accounting.",
ZFEATURE_FLAG_READONLY_COMPAT | ZFEATURE_FLAG_PER_DATASET,
ZFEATURE_TYPE_BOOLEAN, userobj_accounting_deps, sfeatures);
}
{
static const spa_feature_t bookmark_v2_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_BOOKMARKS,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_BOOKMARK_V2,
"com.datto:bookmark_v2", "bookmark_v2",
"Support for larger bookmarks",
0, ZFEATURE_TYPE_BOOLEAN, bookmark_v2_deps, sfeatures);
}
{
static const spa_feature_t encryption_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_BOOKMARK_V2,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_ENCRYPTION,
"com.datto:encryption", "encryption",
"Support for dataset level encryption",
ZFEATURE_FLAG_PER_DATASET, ZFEATURE_TYPE_BOOLEAN,
encryption_deps, sfeatures);
}
{
static const spa_feature_t project_quota_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_PROJECT_QUOTA,
"org.zfsonlinux:project_quota", "project_quota",
"space/object accounting based on project ID.",
ZFEATURE_FLAG_READONLY_COMPAT | ZFEATURE_FLAG_PER_DATASET,
ZFEATURE_TYPE_BOOLEAN, project_quota_deps, sfeatures);
}
zfeature_register(SPA_FEATURE_ALLOCATION_CLASSES,
"org.zfsonlinux:allocation_classes", "allocation_classes",
"Support for separate allocation classes.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL,
sfeatures);
zfeature_register(SPA_FEATURE_RESILVER_DEFER,
"com.datto:resilver_defer", "resilver_defer",
"Support for deferring new resilvers when one is already running.",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL,
sfeatures);
zfeature_register(SPA_FEATURE_DEVICE_REBUILD,
"org.openzfs:device_rebuild", "device_rebuild",
"Support for sequential mirror/dRAID device rebuilds",
ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL,
sfeatures);
{
static const spa_feature_t zstd_deps[] = {
SPA_FEATURE_EXTENSIBLE_DATASET,
SPA_FEATURE_NONE
};
zfeature_register(SPA_FEATURE_ZSTD_COMPRESS,
"org.freebsd:zstd_compress", "zstd_compress",
"zstd compression algorithm support.",
ZFEATURE_FLAG_PER_DATASET, ZFEATURE_TYPE_BOOLEAN, zstd_deps,
sfeatures);
}
zfeature_register(SPA_FEATURE_DRAID,
"org.openzfs:draid", "draid", "Support for distributed spare RAID",
ZFEATURE_FLAG_MOS, ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures);
zfs_mod_list_supported_free(sfeatures);
}
#if defined(_KERNEL)
EXPORT_SYMBOL(zfeature_lookup_guid);
EXPORT_SYMBOL(zfeature_lookup_name);
EXPORT_SYMBOL(zfeature_is_supported);
EXPORT_SYMBOL(zfeature_is_valid_guid);
EXPORT_SYMBOL(zfeature_depends_on);
EXPORT_SYMBOL(zpool_feature_init);
EXPORT_SYMBOL(spa_feature_table);
#endif
diff --git a/module/zcommon/zfs_comutil.c b/module/zcommon/zfs_comutil.c
index 886167759be8..020e7e86c11f 100644
--- a/module/zcommon/zfs_comutil.c
+++ b/module/zcommon/zfs_comutil.c
@@ -1,263 +1,255 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2017 by Delphix. All rights reserved.
*/
/*
* This file is intended for functions that ought to be common between user
* land (libzfs) and the kernel. When many common routines need to be shared
* then a separate file should be created.
*/
#if !defined(_KERNEL)
#include <string.h>
#endif
#include <sys/types.h>
#include <sys/fs/zfs.h>
#include <sys/nvpair.h>
#include "zfs_comutil.h"
#include <sys/zfs_ratelimit.h>
/*
* Are there allocatable vdevs?
*/
boolean_t
zfs_allocatable_devs(nvlist_t *nv)
{
uint64_t is_log;
uint_t c;
nvlist_t **child;
uint_t children;
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0) {
return (B_FALSE);
}
for (c = 0; c < children; c++) {
is_log = 0;
(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
&is_log);
if (!is_log)
return (B_TRUE);
}
return (B_FALSE);
}
/*
* Are there special vdevs?
*/
boolean_t
zfs_special_devs(nvlist_t *nv, char *type)
{
char *bias;
uint_t c;
nvlist_t **child;
uint_t children;
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) != 0) {
return (B_FALSE);
}
for (c = 0; c < children; c++) {
if (nvlist_lookup_string(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS,
&bias) == 0) {
if (strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0 ||
strcmp(bias, VDEV_ALLOC_BIAS_DEDUP) == 0) {
if (type != NULL && strcmp(bias, type) == 0) {
return (B_TRUE);
} else if (type == NULL) {
return (B_TRUE);
}
}
}
}
return (B_FALSE);
}
void
zpool_get_load_policy(nvlist_t *nvl, zpool_load_policy_t *zlpp)
{
nvlist_t *policy;
nvpair_t *elem;
char *nm;
/* Defaults */
zlpp->zlp_rewind = ZPOOL_NO_REWIND;
zlpp->zlp_maxmeta = 0;
zlpp->zlp_maxdata = UINT64_MAX;
zlpp->zlp_txg = UINT64_MAX;
if (nvl == NULL)
return;
elem = NULL;
while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
nm = nvpair_name(elem);
if (strcmp(nm, ZPOOL_LOAD_POLICY) == 0) {
if (nvpair_value_nvlist(elem, &policy) == 0)
zpool_get_load_policy(policy, zlpp);
return;
} else if (strcmp(nm, ZPOOL_LOAD_REWIND_POLICY) == 0) {
if (nvpair_value_uint32(elem, &zlpp->zlp_rewind) == 0)
if (zlpp->zlp_rewind & ~ZPOOL_REWIND_POLICIES)
zlpp->zlp_rewind = ZPOOL_NO_REWIND;
} else if (strcmp(nm, ZPOOL_LOAD_REQUEST_TXG) == 0) {
(void) nvpair_value_uint64(elem, &zlpp->zlp_txg);
} else if (strcmp(nm, ZPOOL_LOAD_META_THRESH) == 0) {
(void) nvpair_value_uint64(elem, &zlpp->zlp_maxmeta);
} else if (strcmp(nm, ZPOOL_LOAD_DATA_THRESH) == 0) {
(void) nvpair_value_uint64(elem, &zlpp->zlp_maxdata);
}
}
if (zlpp->zlp_rewind == 0)
zlpp->zlp_rewind = ZPOOL_NO_REWIND;
}
typedef struct zfs_version_spa_map {
int version_zpl;
int version_spa;
} zfs_version_spa_map_t;
/*
* Keep this table in monotonically increasing version number order.
*/
static zfs_version_spa_map_t zfs_version_table[] = {
{ZPL_VERSION_INITIAL, SPA_VERSION_INITIAL},
{ZPL_VERSION_DIRENT_TYPE, SPA_VERSION_INITIAL},
{ZPL_VERSION_FUID, SPA_VERSION_FUID},
{ZPL_VERSION_USERSPACE, SPA_VERSION_USERSPACE},
{ZPL_VERSION_SA, SPA_VERSION_SA},
{0, 0}
};
/*
* Return the max zpl version for a corresponding spa version
* -1 is returned if no mapping exists.
*/
int
zfs_zpl_version_map(int spa_version)
{
- int i;
int version = -1;
- for (i = 0; zfs_version_table[i].version_spa; i++) {
+ for (int i = 0; zfs_version_table[i].version_spa; i++)
if (spa_version >= zfs_version_table[i].version_spa)
version = zfs_version_table[i].version_zpl;
- }
return (version);
}
/*
* Return the min spa version for a corresponding spa version
* -1 is returned if no mapping exists.
*/
int
zfs_spa_version_map(int zpl_version)
{
- int i;
- int version = -1;
-
- for (i = 0; zfs_version_table[i].version_zpl; i++) {
+ for (int i = 0; zfs_version_table[i].version_zpl; i++)
if (zfs_version_table[i].version_zpl >= zpl_version)
return (zfs_version_table[i].version_spa);
- }
- return (version);
+ return (-1);
}
/*
* This is the table of legacy internal event names; it should not be modified.
* The internal events are now stored in the history log as strings.
*/
-const char *zfs_history_event_names[ZFS_NUM_LEGACY_HISTORY_EVENTS] = {
+const char *const zfs_history_event_names[ZFS_NUM_LEGACY_HISTORY_EVENTS] = {
"invalid event",
"pool create",
"vdev add",
"pool remove",
"pool destroy",
"pool export",
"pool import",
"vdev attach",
"vdev replace",
"vdev detach",
"vdev online",
"vdev offline",
"vdev upgrade",
"pool clear",
"pool scrub",
"pool property set",
"create",
"clone",
"destroy",
"destroy_begin_sync",
"inherit",
"property set",
"quota set",
"permission update",
"permission remove",
"permission who remove",
"promote",
"receive",
"rename",
"reservation set",
"replay_inc_sync",
"replay_full_sync",
"rollback",
"snapshot",
"filesystem version upgrade",
"refquota set",
"refreservation set",
"pool scrub done",
"user hold",
"user release",
"pool split",
};
boolean_t
zfs_dataset_name_hidden(const char *name)
{
/*
* Skip over datasets that are not visible in this zone,
* internal datasets (which have a $ in their name), and
* temporary datasets (which have a % in their name).
*/
- if (strchr(name, '$') != NULL)
- return (B_TRUE);
- if (strchr(name, '%') != NULL)
+ if (strpbrk(name, "$%") != NULL)
return (B_TRUE);
if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
return (B_TRUE);
return (B_FALSE);
}
#if defined(_KERNEL)
EXPORT_SYMBOL(zfs_allocatable_devs);
EXPORT_SYMBOL(zfs_special_devs);
EXPORT_SYMBOL(zpool_get_load_policy);
EXPORT_SYMBOL(zfs_zpl_version_map);
EXPORT_SYMBOL(zfs_spa_version_map);
EXPORT_SYMBOL(zfs_history_event_names);
EXPORT_SYMBOL(zfs_dataset_name_hidden);
#endif
diff --git a/module/zcommon/zfs_deleg.c b/module/zcommon/zfs_deleg.c
index e1f5a353b7a4..8a4a6ca86e13 100644
--- a/module/zcommon/zfs_deleg.c
+++ b/module/zcommon/zfs_deleg.c
@@ -1,249 +1,246 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2010 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2013, 2015 by Delphix. All rights reserved.
* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
*/
#include <sys/zfs_context.h>
#if defined(_KERNEL)
#include <sys/sunddi.h>
#include <sys/ctype.h>
#else
#include <stdio.h>
#include <unistd.h>
#include <libnvpair.h>
#include <ctype.h>
#endif
#include <sys/strings.h>
#include <sys/dsl_deleg.h>
#include "zfs_prop.h"
#include "zfs_deleg.h"
#include "zfs_namecheck.h"
-zfs_deleg_perm_tab_t zfs_deleg_perm_tab[] = {
+const zfs_deleg_perm_tab_t zfs_deleg_perm_tab[] = {
{ZFS_DELEG_PERM_ALLOW},
{ZFS_DELEG_PERM_BOOKMARK},
{ZFS_DELEG_PERM_CLONE},
{ZFS_DELEG_PERM_CREATE},
{ZFS_DELEG_PERM_DESTROY},
{ZFS_DELEG_PERM_DIFF},
{ZFS_DELEG_PERM_MOUNT},
{ZFS_DELEG_PERM_PROMOTE},
{ZFS_DELEG_PERM_RECEIVE},
{ZFS_DELEG_PERM_RENAME},
{ZFS_DELEG_PERM_ROLLBACK},
{ZFS_DELEG_PERM_SNAPSHOT},
{ZFS_DELEG_PERM_SHARE},
{ZFS_DELEG_PERM_SEND},
{ZFS_DELEG_PERM_USERPROP},
{ZFS_DELEG_PERM_USERQUOTA},
{ZFS_DELEG_PERM_GROUPQUOTA},
{ZFS_DELEG_PERM_USERUSED},
{ZFS_DELEG_PERM_GROUPUSED},
{ZFS_DELEG_PERM_USEROBJQUOTA},
{ZFS_DELEG_PERM_GROUPOBJQUOTA},
{ZFS_DELEG_PERM_USEROBJUSED},
{ZFS_DELEG_PERM_GROUPOBJUSED},
{ZFS_DELEG_PERM_HOLD},
{ZFS_DELEG_PERM_RELEASE},
{ZFS_DELEG_PERM_LOAD_KEY},
{ZFS_DELEG_PERM_CHANGE_KEY},
{ZFS_DELEG_PERM_PROJECTUSED},
{ZFS_DELEG_PERM_PROJECTQUOTA},
{ZFS_DELEG_PERM_PROJECTOBJUSED},
{ZFS_DELEG_PERM_PROJECTOBJQUOTA},
{NULL}
};
static int
zfs_valid_permission_name(const char *perm)
{
if (zfs_deleg_canonicalize_perm(perm))
return (0);
return (permset_namecheck(perm, NULL, NULL));
}
const char *
zfs_deleg_canonicalize_perm(const char *perm)
{
- int i;
- zfs_prop_t prop;
-
- for (i = 0; zfs_deleg_perm_tab[i].z_perm != NULL; i++) {
+ for (int i = 0; zfs_deleg_perm_tab[i].z_perm != NULL; i++) {
if (strcmp(perm, zfs_deleg_perm_tab[i].z_perm) == 0)
return (perm);
}
- prop = zfs_name_to_prop(perm);
+ zfs_prop_t prop = zfs_name_to_prop(perm);
if (prop != ZPROP_INVAL && zfs_prop_delegatable(prop))
return (zfs_prop_to_name(prop));
return (NULL);
}
static int
zfs_validate_who(char *who)
{
char *p;
if (who[2] != ZFS_DELEG_FIELD_SEP_CHR)
return (-1);
switch (who[0]) {
case ZFS_DELEG_USER:
case ZFS_DELEG_GROUP:
case ZFS_DELEG_USER_SETS:
case ZFS_DELEG_GROUP_SETS:
if (who[1] != ZFS_DELEG_LOCAL && who[1] != ZFS_DELEG_DESCENDENT)
return (-1);
for (p = &who[3]; *p; p++)
if (!isdigit(*p))
return (-1);
break;
case ZFS_DELEG_NAMED_SET:
case ZFS_DELEG_NAMED_SET_SETS:
if (who[1] != ZFS_DELEG_NA)
return (-1);
return (permset_namecheck(&who[3], NULL, NULL));
case ZFS_DELEG_CREATE:
case ZFS_DELEG_CREATE_SETS:
if (who[1] != ZFS_DELEG_NA)
return (-1);
if (who[3] != '\0')
return (-1);
break;
case ZFS_DELEG_EVERYONE:
case ZFS_DELEG_EVERYONE_SETS:
if (who[1] != ZFS_DELEG_LOCAL && who[1] != ZFS_DELEG_DESCENDENT)
return (-1);
if (who[3] != '\0')
return (-1);
break;
default:
return (-1);
}
return (0);
}
int
zfs_deleg_verify_nvlist(nvlist_t *nvp)
{
nvpair_t *who, *perm_name;
nvlist_t *perms;
int error;
if (nvp == NULL)
return (-1);
who = nvlist_next_nvpair(nvp, NULL);
if (who == NULL)
return (-1);
do {
if (zfs_validate_who(nvpair_name(who)))
return (-1);
error = nvlist_lookup_nvlist(nvp, nvpair_name(who), &perms);
if (error && error != ENOENT)
return (-1);
if (error == ENOENT)
continue;
perm_name = nvlist_next_nvpair(perms, NULL);
if (perm_name == NULL) {
return (-1);
}
do {
error = zfs_valid_permission_name(
nvpair_name(perm_name));
if (error)
return (-1);
} while ((perm_name = nvlist_next_nvpair(perms, perm_name))
!= NULL);
} while ((who = nvlist_next_nvpair(nvp, who)) != NULL);
return (0);
}
/*
* Construct the base attribute name. The base attribute names
* are the "key" to locate the jump objects which contain the actual
* permissions. The base attribute names are encoded based on
* type of entry and whether it is a local or descendent permission.
*
* Arguments:
* attr - attribute name return string, attribute is assumed to be
* ZFS_MAX_DELEG_NAME long.
* type - type of entry to construct
* inheritchr - inheritance type (local,descendent, or NA for create and
* permission set definitions
* data - is either a permission set name or a 64 bit uid/gid.
*/
void
zfs_deleg_whokey(char *attr, zfs_deleg_who_type_t type,
char inheritchr, void *data)
{
int len = ZFS_MAX_DELEG_NAME;
uint64_t *id = data;
switch (type) {
case ZFS_DELEG_USER:
case ZFS_DELEG_GROUP:
case ZFS_DELEG_USER_SETS:
case ZFS_DELEG_GROUP_SETS:
(void) snprintf(attr, len, "%c%c%c%lld", type, inheritchr,
ZFS_DELEG_FIELD_SEP_CHR, (longlong_t)*id);
break;
case ZFS_DELEG_NAMED_SET_SETS:
case ZFS_DELEG_NAMED_SET:
(void) snprintf(attr, len, "%c-%c%s", type,
ZFS_DELEG_FIELD_SEP_CHR, (char *)data);
break;
case ZFS_DELEG_CREATE:
case ZFS_DELEG_CREATE_SETS:
(void) snprintf(attr, len, "%c-%c", type,
ZFS_DELEG_FIELD_SEP_CHR);
break;
case ZFS_DELEG_EVERYONE:
case ZFS_DELEG_EVERYONE_SETS:
(void) snprintf(attr, len, "%c%c%c", type, inheritchr,
ZFS_DELEG_FIELD_SEP_CHR);
break;
default:
ASSERT(!"bad zfs_deleg_who_type_t");
}
}
#if defined(_KERNEL)
EXPORT_SYMBOL(zfs_deleg_verify_nvlist);
EXPORT_SYMBOL(zfs_deleg_whokey);
EXPORT_SYMBOL(zfs_deleg_canonicalize_perm);
#endif
diff --git a/module/zcommon/zfs_prop.c b/module/zcommon/zfs_prop.c
index 800885ee6486..36f30859df04 100644
--- a/module/zcommon/zfs_prop.c
+++ b/module/zcommon/zfs_prop.c
@@ -1,1053 +1,1053 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright 2016, Joyent, Inc.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
*/
/* Portions Copyright 2010 Robert Milkowski */
#include <sys/zio.h>
#include <sys/spa.h>
#include <sys/u8_textprep.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_ioctl.h>
#include <sys/zfs_znode.h>
#include <sys/dsl_crypt.h>
#include "zfs_prop.h"
#include "zfs_deleg.h"
#include "zfs_fletcher.h"
#if !defined(_KERNEL)
#include <stdlib.h>
#include <string.h>
#include <ctype.h>
#endif
static zprop_desc_t zfs_prop_table[ZFS_NUM_PROPS];
/* Note this is indexed by zfs_userquota_prop_t, keep the order the same */
-const char *zfs_userquota_prop_prefixes[] = {
+const char *const zfs_userquota_prop_prefixes[] = {
"userused@",
"userquota@",
"groupused@",
"groupquota@",
"userobjused@",
"userobjquota@",
"groupobjused@",
"groupobjquota@",
"projectused@",
"projectquota@",
"projectobjused@",
"projectobjquota@"
};
zprop_desc_t *
zfs_prop_get_table(void)
{
return (zfs_prop_table);
}
void
zfs_prop_init(void)
{
static const zprop_index_t checksum_table[] = {
{ "on", ZIO_CHECKSUM_ON },
{ "off", ZIO_CHECKSUM_OFF },
{ "fletcher2", ZIO_CHECKSUM_FLETCHER_2 },
{ "fletcher4", ZIO_CHECKSUM_FLETCHER_4 },
{ "sha256", ZIO_CHECKSUM_SHA256 },
{ "noparity", ZIO_CHECKSUM_NOPARITY },
{ "sha512", ZIO_CHECKSUM_SHA512 },
{ "skein", ZIO_CHECKSUM_SKEIN },
{ "edonr", ZIO_CHECKSUM_EDONR },
{ NULL }
};
static const zprop_index_t dedup_table[] = {
{ "on", ZIO_CHECKSUM_ON },
{ "off", ZIO_CHECKSUM_OFF },
{ "verify", ZIO_CHECKSUM_ON | ZIO_CHECKSUM_VERIFY },
{ "sha256", ZIO_CHECKSUM_SHA256 },
{ "sha256,verify",
ZIO_CHECKSUM_SHA256 | ZIO_CHECKSUM_VERIFY },
{ "sha512", ZIO_CHECKSUM_SHA512 },
{ "sha512,verify",
ZIO_CHECKSUM_SHA512 | ZIO_CHECKSUM_VERIFY },
{ "skein", ZIO_CHECKSUM_SKEIN },
{ "skein,verify",
ZIO_CHECKSUM_SKEIN | ZIO_CHECKSUM_VERIFY },
{ "edonr,verify",
ZIO_CHECKSUM_EDONR | ZIO_CHECKSUM_VERIFY },
{ NULL }
};
static const zprop_index_t compress_table[] = {
{ "on", ZIO_COMPRESS_ON },
{ "off", ZIO_COMPRESS_OFF },
{ "lzjb", ZIO_COMPRESS_LZJB },
{ "gzip", ZIO_COMPRESS_GZIP_6 }, /* gzip default */
{ "gzip-1", ZIO_COMPRESS_GZIP_1 },
{ "gzip-2", ZIO_COMPRESS_GZIP_2 },
{ "gzip-3", ZIO_COMPRESS_GZIP_3 },
{ "gzip-4", ZIO_COMPRESS_GZIP_4 },
{ "gzip-5", ZIO_COMPRESS_GZIP_5 },
{ "gzip-6", ZIO_COMPRESS_GZIP_6 },
{ "gzip-7", ZIO_COMPRESS_GZIP_7 },
{ "gzip-8", ZIO_COMPRESS_GZIP_8 },
{ "gzip-9", ZIO_COMPRESS_GZIP_9 },
{ "zle", ZIO_COMPRESS_ZLE },
{ "lz4", ZIO_COMPRESS_LZ4 },
{ "zstd", ZIO_COMPRESS_ZSTD },
{ "zstd-fast",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_DEFAULT) },
/*
* ZSTD 1-19 are synthetic. We store the compression level in a
* separate hidden property to avoid wasting a large amount of
* space in the ZIO_COMPRESS enum.
*
* The compression level is also stored within the header of the
* compressed block since we may need it for later recompression
* to avoid checksum errors (L2ARC).
*
* Note that the level here is defined as bit shifted mask on
* top of the method.
*/
{ "zstd-1", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_1) },
{ "zstd-2", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_2) },
{ "zstd-3", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_3) },
{ "zstd-4", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_4) },
{ "zstd-5", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_5) },
{ "zstd-6", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_6) },
{ "zstd-7", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_7) },
{ "zstd-8", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_8) },
{ "zstd-9", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_9) },
{ "zstd-10", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_10) },
{ "zstd-11", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_11) },
{ "zstd-12", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_12) },
{ "zstd-13", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_13) },
{ "zstd-14", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_14) },
{ "zstd-15", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_15) },
{ "zstd-16", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_16) },
{ "zstd-17", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_17) },
{ "zstd-18", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_18) },
{ "zstd-19", ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_19) },
/*
* The ZSTD-Fast levels are also synthetic.
*/
{ "zstd-fast-1",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_1) },
{ "zstd-fast-2",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_2) },
{ "zstd-fast-3",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_3) },
{ "zstd-fast-4",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_4) },
{ "zstd-fast-5",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_5) },
{ "zstd-fast-6",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_6) },
{ "zstd-fast-7",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_7) },
{ "zstd-fast-8",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_8) },
{ "zstd-fast-9",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_9) },
{ "zstd-fast-10",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_10) },
{ "zstd-fast-20",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_20) },
{ "zstd-fast-30",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_30) },
{ "zstd-fast-40",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_40) },
{ "zstd-fast-50",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_50) },
{ "zstd-fast-60",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_60) },
{ "zstd-fast-70",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_70) },
{ "zstd-fast-80",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_80) },
{ "zstd-fast-90",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_90) },
{ "zstd-fast-100",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_100) },
{ "zstd-fast-500",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_500) },
{ "zstd-fast-1000",
ZIO_COMPLEVEL_ZSTD(ZIO_ZSTD_LEVEL_FAST_1000) },
{ NULL }
};
static const zprop_index_t crypto_table[] = {
{ "on", ZIO_CRYPT_ON },
{ "off", ZIO_CRYPT_OFF },
{ "aes-128-ccm", ZIO_CRYPT_AES_128_CCM },
{ "aes-192-ccm", ZIO_CRYPT_AES_192_CCM },
{ "aes-256-ccm", ZIO_CRYPT_AES_256_CCM },
{ "aes-128-gcm", ZIO_CRYPT_AES_128_GCM },
{ "aes-192-gcm", ZIO_CRYPT_AES_192_GCM },
{ "aes-256-gcm", ZIO_CRYPT_AES_256_GCM },
{ NULL }
};
static const zprop_index_t keyformat_table[] = {
{ "none", ZFS_KEYFORMAT_NONE },
{ "raw", ZFS_KEYFORMAT_RAW },
{ "hex", ZFS_KEYFORMAT_HEX },
{ "passphrase", ZFS_KEYFORMAT_PASSPHRASE },
{ NULL }
};
static const zprop_index_t snapdir_table[] = {
{ "hidden", ZFS_SNAPDIR_HIDDEN },
{ "visible", ZFS_SNAPDIR_VISIBLE },
{ NULL }
};
static const zprop_index_t snapdev_table[] = {
{ "hidden", ZFS_SNAPDEV_HIDDEN },
{ "visible", ZFS_SNAPDEV_VISIBLE },
{ NULL }
};
static const zprop_index_t acl_mode_table[] = {
{ "discard", ZFS_ACL_DISCARD },
{ "groupmask", ZFS_ACL_GROUPMASK },
{ "passthrough", ZFS_ACL_PASSTHROUGH },
{ "restricted", ZFS_ACL_RESTRICTED },
{ NULL }
};
static const zprop_index_t acltype_table[] = {
{ "off", ZFS_ACLTYPE_OFF },
{ "posix", ZFS_ACLTYPE_POSIX },
{ "nfsv4", ZFS_ACLTYPE_NFSV4 },
{ "disabled", ZFS_ACLTYPE_OFF }, /* bkwrd compatibility */
{ "noacl", ZFS_ACLTYPE_OFF }, /* bkwrd compatibility */
{ "posixacl", ZFS_ACLTYPE_POSIX }, /* bkwrd compatibility */
{ NULL }
};
static const zprop_index_t acl_inherit_table[] = {
{ "discard", ZFS_ACL_DISCARD },
{ "noallow", ZFS_ACL_NOALLOW },
{ "restricted", ZFS_ACL_RESTRICTED },
{ "passthrough", ZFS_ACL_PASSTHROUGH },
{ "secure", ZFS_ACL_RESTRICTED }, /* bkwrd compatibility */
{ "passthrough-x", ZFS_ACL_PASSTHROUGH_X },
{ NULL }
};
static const zprop_index_t case_table[] = {
{ "sensitive", ZFS_CASE_SENSITIVE },
{ "insensitive", ZFS_CASE_INSENSITIVE },
{ "mixed", ZFS_CASE_MIXED },
{ NULL }
};
static const zprop_index_t copies_table[] = {
{ "1", 1 },
{ "2", 2 },
{ "3", 3 },
{ NULL }
};
/*
* Use the unique flags we have to send to u8_strcmp() and/or
* u8_textprep() to represent the various normalization property
* values.
*/
static const zprop_index_t normalize_table[] = {
{ "none", 0 },
{ "formD", U8_TEXTPREP_NFD },
{ "formKC", U8_TEXTPREP_NFKC },
{ "formC", U8_TEXTPREP_NFC },
{ "formKD", U8_TEXTPREP_NFKD },
{ NULL }
};
static const zprop_index_t version_table[] = {
{ "1", 1 },
{ "2", 2 },
{ "3", 3 },
{ "4", 4 },
{ "5", 5 },
{ "current", ZPL_VERSION },
{ NULL }
};
static const zprop_index_t boolean_table[] = {
{ "off", 0 },
{ "on", 1 },
{ NULL }
};
static const zprop_index_t keystatus_table[] = {
{ "none", ZFS_KEYSTATUS_NONE},
{ "unavailable", ZFS_KEYSTATUS_UNAVAILABLE},
{ "available", ZFS_KEYSTATUS_AVAILABLE},
{ NULL }
};
static const zprop_index_t logbias_table[] = {
{ "latency", ZFS_LOGBIAS_LATENCY },
{ "throughput", ZFS_LOGBIAS_THROUGHPUT },
{ NULL }
};
static const zprop_index_t canmount_table[] = {
{ "off", ZFS_CANMOUNT_OFF },
{ "on", ZFS_CANMOUNT_ON },
{ "noauto", ZFS_CANMOUNT_NOAUTO },
{ NULL }
};
static const zprop_index_t cache_table[] = {
{ "none", ZFS_CACHE_NONE },
{ "metadata", ZFS_CACHE_METADATA },
{ "all", ZFS_CACHE_ALL },
{ NULL }
};
static const zprop_index_t sync_table[] = {
{ "standard", ZFS_SYNC_STANDARD },
{ "always", ZFS_SYNC_ALWAYS },
{ "disabled", ZFS_SYNC_DISABLED },
{ NULL }
};
static const zprop_index_t xattr_table[] = {
{ "off", ZFS_XATTR_OFF },
{ "on", ZFS_XATTR_DIR },
{ "sa", ZFS_XATTR_SA },
{ "dir", ZFS_XATTR_DIR },
{ NULL }
};
static const zprop_index_t dnsize_table[] = {
{ "legacy", ZFS_DNSIZE_LEGACY },
{ "auto", ZFS_DNSIZE_AUTO },
{ "1k", ZFS_DNSIZE_1K },
{ "2k", ZFS_DNSIZE_2K },
{ "4k", ZFS_DNSIZE_4K },
{ "8k", ZFS_DNSIZE_8K },
{ "16k", ZFS_DNSIZE_16K },
{ NULL }
};
static const zprop_index_t redundant_metadata_table[] = {
{ "all", ZFS_REDUNDANT_METADATA_ALL },
{ "most", ZFS_REDUNDANT_METADATA_MOST },
{ NULL }
};
static const zprop_index_t volmode_table[] = {
{ "default", ZFS_VOLMODE_DEFAULT },
{ "full", ZFS_VOLMODE_GEOM },
{ "geom", ZFS_VOLMODE_GEOM },
{ "dev", ZFS_VOLMODE_DEV },
{ "none", ZFS_VOLMODE_NONE },
{ NULL }
};
struct zfs_mod_supported_features *sfeatures =
zfs_mod_list_supported(ZFS_SYSFS_DATASET_PROPERTIES);
/* inherit index properties */
zprop_register_index(ZFS_PROP_REDUNDANT_METADATA, "redundant_metadata",
ZFS_REDUNDANT_METADATA_ALL,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"all | most", "REDUND_MD",
redundant_metadata_table, sfeatures);
zprop_register_index(ZFS_PROP_SYNC, "sync", ZFS_SYNC_STANDARD,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"standard | always | disabled", "SYNC",
sync_table, sfeatures);
zprop_register_index(ZFS_PROP_CHECKSUM, "checksum",
ZIO_CHECKSUM_DEFAULT, PROP_INHERIT, ZFS_TYPE_FILESYSTEM |
ZFS_TYPE_VOLUME,
"on | off | fletcher2 | fletcher4 | sha256 | sha512 | skein"
" | edonr",
"CHECKSUM", checksum_table, sfeatures);
zprop_register_index(ZFS_PROP_DEDUP, "dedup", ZIO_CHECKSUM_OFF,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"on | off | verify | sha256[,verify] | sha512[,verify] | "
"skein[,verify] | edonr,verify",
"DEDUP", dedup_table, sfeatures);
zprop_register_index(ZFS_PROP_COMPRESSION, "compression",
ZIO_COMPRESS_DEFAULT, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"on | off | lzjb | gzip | gzip-[1-9] | zle | lz4 | "
"zstd | zstd-[1-19] | "
"zstd-fast | zstd-fast-[1-10,20,30,40,50,60,70,80,90,100,500,1000]",
"COMPRESS", compress_table, sfeatures);
zprop_register_index(ZFS_PROP_SNAPDIR, "snapdir", ZFS_SNAPDIR_HIDDEN,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM,
"hidden | visible", "SNAPDIR", snapdir_table, sfeatures);
zprop_register_index(ZFS_PROP_SNAPDEV, "snapdev", ZFS_SNAPDEV_HIDDEN,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"hidden | visible", "SNAPDEV", snapdev_table, sfeatures);
zprop_register_index(ZFS_PROP_ACLMODE, "aclmode", ZFS_ACL_DISCARD,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM,
"discard | groupmask | passthrough | restricted", "ACLMODE",
acl_mode_table, sfeatures);
zprop_register_index(ZFS_PROP_ACLTYPE, "acltype",
#ifdef __linux__
/* Linux doesn't natively support ZFS's NFSv4-style ACLs. */
ZFS_ACLTYPE_OFF,
#else
ZFS_ACLTYPE_NFSV4,
#endif
PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT,
"off | nfsv4 | posix", "ACLTYPE", acltype_table, sfeatures);
zprop_register_index(ZFS_PROP_ACLINHERIT, "aclinherit",
ZFS_ACL_RESTRICTED, PROP_INHERIT, ZFS_TYPE_FILESYSTEM,
"discard | noallow | restricted | passthrough | passthrough-x",
"ACLINHERIT", acl_inherit_table, sfeatures);
zprop_register_index(ZFS_PROP_COPIES, "copies", 1, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"1 | 2 | 3", "COPIES", copies_table, sfeatures);
zprop_register_index(ZFS_PROP_PRIMARYCACHE, "primarycache",
ZFS_CACHE_ALL, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT | ZFS_TYPE_VOLUME,
"all | none | metadata", "PRIMARYCACHE", cache_table, sfeatures);
zprop_register_index(ZFS_PROP_SECONDARYCACHE, "secondarycache",
ZFS_CACHE_ALL, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT | ZFS_TYPE_VOLUME,
"all | none | metadata", "SECONDARYCACHE", cache_table, sfeatures);
zprop_register_index(ZFS_PROP_LOGBIAS, "logbias", ZFS_LOGBIAS_LATENCY,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"latency | throughput", "LOGBIAS", logbias_table, sfeatures);
zprop_register_index(ZFS_PROP_XATTR, "xattr", ZFS_XATTR_DIR,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT,
"on | off | dir | sa", "XATTR", xattr_table, sfeatures);
zprop_register_index(ZFS_PROP_DNODESIZE, "dnodesize",
ZFS_DNSIZE_LEGACY, PROP_INHERIT, ZFS_TYPE_FILESYSTEM,
"legacy | auto | 1k | 2k | 4k | 8k | 16k", "DNSIZE", dnsize_table,
sfeatures);
zprop_register_index(ZFS_PROP_VOLMODE, "volmode",
ZFS_VOLMODE_DEFAULT, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"default | full | geom | dev | none", "VOLMODE", volmode_table,
sfeatures);
/* inherit index (boolean) properties */
zprop_register_index(ZFS_PROP_ATIME, "atime", 1, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM, "on | off", "ATIME", boolean_table, sfeatures);
zprop_register_index(ZFS_PROP_RELATIME, "relatime", 0, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM, "on | off", "RELATIME", boolean_table,
sfeatures);
zprop_register_index(ZFS_PROP_DEVICES, "devices", 1, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "DEVICES",
boolean_table, sfeatures);
zprop_register_index(ZFS_PROP_EXEC, "exec", 1, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "EXEC",
boolean_table, sfeatures);
zprop_register_index(ZFS_PROP_SETUID, "setuid", 1, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "SETUID",
boolean_table, sfeatures);
zprop_register_index(ZFS_PROP_READONLY, "readonly", 0, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "on | off", "RDONLY",
boolean_table, sfeatures);
#ifdef __FreeBSD__
zprop_register_index(ZFS_PROP_ZONED, "jailed", 0, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM, "on | off", "JAILED", boolean_table,
sfeatures);
#else
zprop_register_index(ZFS_PROP_ZONED, "zoned", 0, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM, "on | off", "ZONED", boolean_table, sfeatures);
#endif
zprop_register_index(ZFS_PROP_VSCAN, "vscan", 0, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM, "on | off", "VSCAN", boolean_table, sfeatures);
zprop_register_index(ZFS_PROP_NBMAND, "nbmand", 0, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT, "on | off", "NBMAND",
boolean_table, sfeatures);
zprop_register_index(ZFS_PROP_OVERLAY, "overlay", 1, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM, "on | off", "OVERLAY", boolean_table,
sfeatures);
/* default index properties */
zprop_register_index(ZFS_PROP_VERSION, "version", 0, PROP_DEFAULT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT,
"1 | 2 | 3 | 4 | 5 | current", "VERSION", version_table, sfeatures);
zprop_register_index(ZFS_PROP_CANMOUNT, "canmount", ZFS_CANMOUNT_ON,
PROP_DEFAULT, ZFS_TYPE_FILESYSTEM, "on | off | noauto",
"CANMOUNT", canmount_table, sfeatures);
/* readonly index properties */
zprop_register_index(ZFS_PROP_MOUNTED, "mounted", 0, PROP_READONLY,
ZFS_TYPE_FILESYSTEM, "yes | no", "MOUNTED", boolean_table,
sfeatures);
zprop_register_index(ZFS_PROP_DEFER_DESTROY, "defer_destroy", 0,
PROP_READONLY, ZFS_TYPE_SNAPSHOT, "yes | no", "DEFER_DESTROY",
boolean_table, sfeatures);
zprop_register_index(ZFS_PROP_KEYSTATUS, "keystatus",
ZFS_KEYSTATUS_NONE, PROP_READONLY, ZFS_TYPE_DATASET,
"none | unavailable | available",
"KEYSTATUS", keystatus_table, sfeatures);
/* set once index properties */
zprop_register_index(ZFS_PROP_NORMALIZE, "normalization", 0,
PROP_ONETIME, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT,
"none | formC | formD | formKC | formKD", "NORMALIZATION",
normalize_table, sfeatures);
zprop_register_index(ZFS_PROP_CASE, "casesensitivity",
ZFS_CASE_SENSITIVE, PROP_ONETIME, ZFS_TYPE_FILESYSTEM |
ZFS_TYPE_SNAPSHOT,
"sensitive | insensitive | mixed", "CASE", case_table, sfeatures);
zprop_register_index(ZFS_PROP_KEYFORMAT, "keyformat",
ZFS_KEYFORMAT_NONE, PROP_ONETIME_DEFAULT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"none | raw | hex | passphrase", "KEYFORMAT", keyformat_table,
sfeatures);
zprop_register_index(ZFS_PROP_ENCRYPTION, "encryption",
ZIO_CRYPT_DEFAULT, PROP_ONETIME, ZFS_TYPE_DATASET,
"on | off | aes-128-ccm | aes-192-ccm | aes-256-ccm | "
"aes-128-gcm | aes-192-gcm | aes-256-gcm", "ENCRYPTION",
crypto_table, sfeatures);
/* set once index (boolean) properties */
zprop_register_index(ZFS_PROP_UTF8ONLY, "utf8only", 0, PROP_ONETIME,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_SNAPSHOT,
"on | off", "UTF8ONLY", boolean_table, sfeatures);
/* string properties */
zprop_register_string(ZFS_PROP_ORIGIN, "origin", NULL, PROP_READONLY,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "<snapshot>", "ORIGIN",
sfeatures);
zprop_register_string(ZFS_PROP_CLONES, "clones", NULL, PROP_READONLY,
ZFS_TYPE_SNAPSHOT, "<dataset>[,...]", "CLONES", sfeatures);
zprop_register_string(ZFS_PROP_MOUNTPOINT, "mountpoint", "/",
PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "<path> | legacy | none",
"MOUNTPOINT", sfeatures);
zprop_register_string(ZFS_PROP_SHARENFS, "sharenfs", "off",
PROP_INHERIT, ZFS_TYPE_FILESYSTEM, "on | off | NFS share options",
"SHARENFS", sfeatures);
zprop_register_string(ZFS_PROP_TYPE, "type", NULL, PROP_READONLY,
ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK,
"filesystem | volume | snapshot | bookmark", "TYPE", sfeatures);
zprop_register_string(ZFS_PROP_SHARESMB, "sharesmb", "off",
PROP_INHERIT, ZFS_TYPE_FILESYSTEM,
"on | off | SMB share options", "SHARESMB", sfeatures);
zprop_register_string(ZFS_PROP_MLSLABEL, "mlslabel",
ZFS_MLSLABEL_DEFAULT, PROP_INHERIT, ZFS_TYPE_DATASET,
"<sensitivity label>", "MLSLABEL", sfeatures);
zprop_register_string(ZFS_PROP_SELINUX_CONTEXT, "context",
"none", PROP_DEFAULT, ZFS_TYPE_DATASET, "<selinux context>",
"CONTEXT", sfeatures);
zprop_register_string(ZFS_PROP_SELINUX_FSCONTEXT, "fscontext",
"none", PROP_DEFAULT, ZFS_TYPE_DATASET, "<selinux fscontext>",
"FSCONTEXT", sfeatures);
zprop_register_string(ZFS_PROP_SELINUX_DEFCONTEXT, "defcontext",
"none", PROP_DEFAULT, ZFS_TYPE_DATASET, "<selinux defcontext>",
"DEFCONTEXT", sfeatures);
zprop_register_string(ZFS_PROP_SELINUX_ROOTCONTEXT, "rootcontext",
"none", PROP_DEFAULT, ZFS_TYPE_DATASET, "<selinux rootcontext>",
"ROOTCONTEXT", sfeatures);
zprop_register_string(ZFS_PROP_RECEIVE_RESUME_TOKEN,
"receive_resume_token",
NULL, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"<string token>", "RESUMETOK", sfeatures);
zprop_register_string(ZFS_PROP_ENCRYPTION_ROOT, "encryptionroot", NULL,
PROP_READONLY, ZFS_TYPE_DATASET, "<filesystem | volume>",
"ENCROOT", sfeatures);
zprop_register_string(ZFS_PROP_KEYLOCATION, "keylocation",
"none", PROP_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"prompt | <file URI> | <https URL> | <http URL>", "KEYLOCATION",
sfeatures);
zprop_register_string(ZFS_PROP_REDACT_SNAPS,
"redact_snaps", NULL, PROP_READONLY,
ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "<snapshot>[,...]",
"RSNAPS", sfeatures);
/* readonly number properties */
zprop_register_number(ZFS_PROP_USED, "used", 0, PROP_READONLY,
ZFS_TYPE_DATASET, "<size>", "USED", sfeatures);
zprop_register_number(ZFS_PROP_AVAILABLE, "available", 0, PROP_READONLY,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "<size>", "AVAIL",
sfeatures);
zprop_register_number(ZFS_PROP_REFERENCED, "referenced", 0,
PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "<size>",
"REFER", sfeatures);
zprop_register_number(ZFS_PROP_COMPRESSRATIO, "compressratio", 0,
PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK,
"<1.00x or higher if compressed>", "RATIO", sfeatures);
zprop_register_number(ZFS_PROP_REFRATIO, "refcompressratio", 0,
PROP_READONLY, ZFS_TYPE_DATASET,
"<1.00x or higher if compressed>", "REFRATIO", sfeatures);
zprop_register_number(ZFS_PROP_VOLBLOCKSIZE, "volblocksize",
ZVOL_DEFAULT_BLOCKSIZE, PROP_ONETIME,
ZFS_TYPE_VOLUME, "512 to 128k, power of 2", "VOLBLOCK", sfeatures);
zprop_register_number(ZFS_PROP_USEDSNAP, "usedbysnapshots", 0,
PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "<size>",
"USEDSNAP", sfeatures);
zprop_register_number(ZFS_PROP_USEDDS, "usedbydataset", 0,
PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "<size>",
"USEDDS", sfeatures);
zprop_register_number(ZFS_PROP_USEDCHILD, "usedbychildren", 0,
PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "<size>",
"USEDCHILD", sfeatures);
zprop_register_number(ZFS_PROP_USEDREFRESERV, "usedbyrefreservation", 0,
PROP_READONLY,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "<size>", "USEDREFRESERV",
sfeatures);
zprop_register_number(ZFS_PROP_USERREFS, "userrefs", 0, PROP_READONLY,
ZFS_TYPE_SNAPSHOT, "<count>", "USERREFS", sfeatures);
zprop_register_number(ZFS_PROP_WRITTEN, "written", 0, PROP_READONLY,
ZFS_TYPE_DATASET, "<size>", "WRITTEN", sfeatures);
zprop_register_number(ZFS_PROP_LOGICALUSED, "logicalused", 0,
PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "<size>",
"LUSED", sfeatures);
zprop_register_number(ZFS_PROP_LOGICALREFERENCED, "logicalreferenced",
0, PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "<size>",
"LREFER", sfeatures);
zprop_register_number(ZFS_PROP_FILESYSTEM_COUNT, "filesystem_count",
UINT64_MAX, PROP_READONLY, ZFS_TYPE_FILESYSTEM,
"<count>", "FSCOUNT", sfeatures);
zprop_register_number(ZFS_PROP_SNAPSHOT_COUNT, "snapshot_count",
UINT64_MAX, PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"<count>", "SSCOUNT", sfeatures);
zprop_register_number(ZFS_PROP_GUID, "guid", 0, PROP_READONLY,
ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "<uint64>", "GUID",
sfeatures);
zprop_register_number(ZFS_PROP_CREATETXG, "createtxg", 0, PROP_READONLY,
ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "<uint64>", "CREATETXG",
sfeatures);
zprop_register_number(ZFS_PROP_PBKDF2_ITERS, "pbkdf2iters",
0, PROP_ONETIME_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"<iters>", "PBKDF2ITERS", sfeatures);
zprop_register_number(ZFS_PROP_OBJSETID, "objsetid", 0,
PROP_READONLY, ZFS_TYPE_DATASET, "<uint64>", "OBJSETID", sfeatures);
/* default number properties */
zprop_register_number(ZFS_PROP_QUOTA, "quota", 0, PROP_DEFAULT,
ZFS_TYPE_FILESYSTEM, "<size> | none", "QUOTA", sfeatures);
zprop_register_number(ZFS_PROP_RESERVATION, "reservation", 0,
PROP_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"<size> | none", "RESERV", sfeatures);
zprop_register_number(ZFS_PROP_VOLSIZE, "volsize", 0, PROP_DEFAULT,
ZFS_TYPE_SNAPSHOT | ZFS_TYPE_VOLUME, "<size>", "VOLSIZE",
sfeatures);
zprop_register_number(ZFS_PROP_REFQUOTA, "refquota", 0, PROP_DEFAULT,
ZFS_TYPE_FILESYSTEM, "<size> | none", "REFQUOTA", sfeatures);
zprop_register_number(ZFS_PROP_REFRESERVATION, "refreservation", 0,
PROP_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"<size> | none", "REFRESERV", sfeatures);
zprop_register_number(ZFS_PROP_FILESYSTEM_LIMIT, "filesystem_limit",
UINT64_MAX, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM,
"<count> | none", "FSLIMIT", sfeatures);
zprop_register_number(ZFS_PROP_SNAPSHOT_LIMIT, "snapshot_limit",
UINT64_MAX, PROP_DEFAULT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"<count> | none", "SSLIMIT", sfeatures);
/* inherit number properties */
zprop_register_number(ZFS_PROP_RECORDSIZE, "recordsize",
SPA_OLD_MAXBLOCKSIZE, PROP_INHERIT,
ZFS_TYPE_FILESYSTEM, "512 to 1M, power of 2", "RECSIZE", sfeatures);
zprop_register_number(ZFS_PROP_SPECIAL_SMALL_BLOCKS,
"special_small_blocks", 0, PROP_INHERIT, ZFS_TYPE_FILESYSTEM,
"zero or 512 to 1M, power of 2", "SPECIAL_SMALL_BLOCKS", sfeatures);
/* hidden properties */
zprop_register_hidden(ZFS_PROP_NUMCLONES, "numclones", PROP_TYPE_NUMBER,
PROP_READONLY, ZFS_TYPE_SNAPSHOT, "NUMCLONES", sfeatures);
zprop_register_hidden(ZFS_PROP_NAME, "name", PROP_TYPE_STRING,
PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "NAME",
sfeatures);
zprop_register_hidden(ZFS_PROP_ISCSIOPTIONS, "iscsioptions",
PROP_TYPE_STRING, PROP_INHERIT, ZFS_TYPE_VOLUME, "ISCSIOPTIONS",
sfeatures);
zprop_register_hidden(ZFS_PROP_STMF_SHAREINFO, "stmf_sbd_lu",
PROP_TYPE_STRING, PROP_INHERIT, ZFS_TYPE_VOLUME,
"STMF_SBD_LU", sfeatures);
zprop_register_hidden(ZFS_PROP_USERACCOUNTING, "useraccounting",
PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET,
"USERACCOUNTING", sfeatures);
zprop_register_hidden(ZFS_PROP_UNIQUE, "unique", PROP_TYPE_NUMBER,
PROP_READONLY, ZFS_TYPE_DATASET, "UNIQUE", sfeatures);
zprop_register_hidden(ZFS_PROP_INCONSISTENT, "inconsistent",
PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_DATASET, "INCONSISTENT",
sfeatures);
zprop_register_hidden(ZFS_PROP_IVSET_GUID, "ivsetguid",
PROP_TYPE_NUMBER, PROP_READONLY,
ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK, "IVSETGUID", sfeatures);
zprop_register_hidden(ZFS_PROP_PREV_SNAP, "prevsnap", PROP_TYPE_STRING,
PROP_READONLY, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "PREVSNAP",
sfeatures);
zprop_register_hidden(ZFS_PROP_PBKDF2_SALT, "pbkdf2salt",
PROP_TYPE_NUMBER, PROP_ONETIME_DEFAULT,
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "PBKDF2SALT", sfeatures);
zprop_register_hidden(ZFS_PROP_KEY_GUID, "keyguid", PROP_TYPE_NUMBER,
PROP_READONLY, ZFS_TYPE_DATASET, "KEYGUID", sfeatures);
zprop_register_hidden(ZFS_PROP_REDACTED, "redacted", PROP_TYPE_NUMBER,
PROP_READONLY, ZFS_TYPE_DATASET, "REDACTED", sfeatures);
/*
* Properties that are obsolete and not used. These are retained so
* that we don't have to change the values of the zfs_prop_t enum, or
* have NULL pointers in the zfs_prop_table[].
*/
zprop_register_hidden(ZFS_PROP_REMAPTXG, "remaptxg", PROP_TYPE_NUMBER,
PROP_READONLY, ZFS_TYPE_DATASET, "REMAPTXG", sfeatures);
/* oddball properties */
zprop_register_impl(ZFS_PROP_CREATION, "creation", PROP_TYPE_NUMBER, 0,
NULL, PROP_READONLY, ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK,
"<date>", "CREATION", B_FALSE, B_TRUE, NULL, sfeatures);
zfs_mod_list_supported_free(sfeatures);
}
boolean_t
zfs_prop_delegatable(zfs_prop_t prop)
{
zprop_desc_t *pd = &zfs_prop_table[prop];
/* The mlslabel property is never delegatable. */
if (prop == ZFS_PROP_MLSLABEL)
return (B_FALSE);
return (pd->pd_attr != PROP_READONLY);
}
/*
* Given a zfs dataset property name, returns the corresponding property ID.
*/
zfs_prop_t
zfs_name_to_prop(const char *propname)
{
return (zprop_name_to_prop(propname, ZFS_TYPE_DATASET));
}
/*
* Returns true if this is a valid user-defined property (one with a ':').
*/
boolean_t
zfs_prop_user(const char *name)
{
int i;
char c;
boolean_t foundsep = B_FALSE;
for (i = 0; i < strlen(name); i++) {
c = name[i];
if (!zprop_valid_char(c))
return (B_FALSE);
if (c == ':')
foundsep = B_TRUE;
}
if (!foundsep)
return (B_FALSE);
return (B_TRUE);
}
/*
* Returns true if this is a valid userspace-type property (one with a '@').
* Note that after the @, any character is valid (eg, another @, for SID
* user@domain).
*/
boolean_t
zfs_prop_userquota(const char *name)
{
zfs_userquota_prop_t prop;
for (prop = 0; prop < ZFS_NUM_USERQUOTA_PROPS; prop++) {
if (strncmp(name, zfs_userquota_prop_prefixes[prop],
strlen(zfs_userquota_prop_prefixes[prop])) == 0) {
return (B_TRUE);
}
}
return (B_FALSE);
}
/*
* Returns true if this is a valid written@ property.
* Note that after the @, any character is valid (eg, another @, for
* written@pool/fs@origin).
*/
boolean_t
zfs_prop_written(const char *name)
{
static const char *prop_prefix = "written@";
static const char *book_prefix = "written#";
return (strncmp(name, prop_prefix, strlen(prop_prefix)) == 0 ||
strncmp(name, book_prefix, strlen(book_prefix)) == 0);
}
/*
* Tables of index types, plus functions to convert between the user view
* (strings) and internal representation (uint64_t).
*/
int
zfs_prop_string_to_index(zfs_prop_t prop, const char *string, uint64_t *index)
{
return (zprop_string_to_index(prop, string, index, ZFS_TYPE_DATASET));
}
int
zfs_prop_index_to_string(zfs_prop_t prop, uint64_t index, const char **string)
{
return (zprop_index_to_string(prop, index, string, ZFS_TYPE_DATASET));
}
uint64_t
zfs_prop_random_value(zfs_prop_t prop, uint64_t seed)
{
return (zprop_random_value(prop, seed, ZFS_TYPE_DATASET));
}
/*
* Returns TRUE if the property applies to any of the given dataset types.
*/
boolean_t
zfs_prop_valid_for_type(int prop, zfs_type_t types, boolean_t headcheck)
{
return (zprop_valid_for_type(prop, types, headcheck));
}
zprop_type_t
zfs_prop_get_type(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_proptype);
}
/*
* Returns TRUE if the property is readonly.
*/
boolean_t
zfs_prop_readonly(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_attr == PROP_READONLY ||
zfs_prop_table[prop].pd_attr == PROP_ONETIME ||
zfs_prop_table[prop].pd_attr == PROP_ONETIME_DEFAULT);
}
/*
* Returns TRUE if the property is visible (not hidden).
*/
boolean_t
zfs_prop_visible(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_visible &&
zfs_prop_table[prop].pd_zfs_mod_supported);
}
/*
* Returns TRUE if the property is only allowed to be set once.
*/
boolean_t
zfs_prop_setonce(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_attr == PROP_ONETIME ||
zfs_prop_table[prop].pd_attr == PROP_ONETIME_DEFAULT);
}
const char *
zfs_prop_default_string(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_strdefault);
}
uint64_t
zfs_prop_default_numeric(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_numdefault);
}
/*
* Given a dataset property ID, returns the corresponding name.
* Assuming the zfs dataset property ID is valid.
*/
const char *
zfs_prop_to_name(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_name);
}
/*
* Returns TRUE if the property is inheritable.
*/
boolean_t
zfs_prop_inheritable(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_attr == PROP_INHERIT ||
zfs_prop_table[prop].pd_attr == PROP_ONETIME);
}
/*
* Returns TRUE if property is one of the encryption properties that requires
* a loaded encryption key to modify.
*/
boolean_t
zfs_prop_encryption_key_param(zfs_prop_t prop)
{
/*
* keylocation does not count as an encryption property. It can be
* changed at will without needing the master keys.
*/
return (prop == ZFS_PROP_PBKDF2_SALT || prop == ZFS_PROP_PBKDF2_ITERS ||
prop == ZFS_PROP_KEYFORMAT);
}
/*
* Helper function used by both kernelspace and userspace to check the
* keylocation property. If encrypted is set, the keylocation must be valid
* for an encrypted dataset.
*/
boolean_t
zfs_prop_valid_keylocation(const char *str, boolean_t encrypted)
{
if (strcmp("none", str) == 0)
return (!encrypted);
else if (strcmp("prompt", str) == 0)
return (B_TRUE);
else if (strlen(str) > 8 && strncmp("file:///", str, 8) == 0)
return (B_TRUE);
else if (strlen(str) > 8 && strncmp("https://", str, 8) == 0)
return (B_TRUE);
else if (strlen(str) > 7 && strncmp("http://", str, 7) == 0)
return (B_TRUE);
return (B_FALSE);
}
#ifndef _KERNEL
#include <libzfs.h>
/*
* Returns a string describing the set of acceptable values for the given
* zfs property, or NULL if it cannot be set.
*/
const char *
zfs_prop_values(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_values);
}
/*
* Returns TRUE if this property is a string type. Note that index types
* (compression, checksum) are treated as strings in userland, even though they
* are stored numerically on disk.
*/
int
zfs_prop_is_string(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_proptype == PROP_TYPE_STRING ||
zfs_prop_table[prop].pd_proptype == PROP_TYPE_INDEX);
}
/*
* Returns the column header for the given property. Used only in
* 'zfs list -o', but centralized here with the other property information.
*/
const char *
zfs_prop_column_name(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_colname);
}
/*
* Returns whether the given property should be displayed right-justified for
* 'zfs list'.
*/
boolean_t
zfs_prop_align_right(zfs_prop_t prop)
{
return (zfs_prop_table[prop].pd_rightalign);
}
#endif
#if defined(_KERNEL)
#include <sys/simd.h>
#if defined(HAVE_KERNEL_FPU_INTERNAL)
union fpregs_state **zfs_kfpu_fpregs;
EXPORT_SYMBOL(zfs_kfpu_fpregs);
#endif /* HAVE_KERNEL_FPU_INTERNAL */
static int __init
zcommon_init(void)
{
int error = kfpu_init();
if (error)
return (error);
fletcher_4_init();
return (0);
}
static void __exit
zcommon_fini(void)
{
fletcher_4_fini();
kfpu_fini();
}
module_init_early(zcommon_init);
module_exit(zcommon_fini);
#endif
ZFS_MODULE_DESCRIPTION("Generic ZFS support");
ZFS_MODULE_AUTHOR(ZFS_META_AUTHOR);
ZFS_MODULE_LICENSE(ZFS_META_LICENSE);
ZFS_MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
/* zfs dataset property functions */
EXPORT_SYMBOL(zfs_userquota_prop_prefixes);
EXPORT_SYMBOL(zfs_prop_init);
EXPORT_SYMBOL(zfs_prop_get_type);
EXPORT_SYMBOL(zfs_prop_get_table);
EXPORT_SYMBOL(zfs_prop_delegatable);
EXPORT_SYMBOL(zfs_prop_visible);
/* Dataset property functions shared between libzfs and kernel. */
EXPORT_SYMBOL(zfs_prop_default_string);
EXPORT_SYMBOL(zfs_prop_default_numeric);
EXPORT_SYMBOL(zfs_prop_readonly);
EXPORT_SYMBOL(zfs_prop_inheritable);
EXPORT_SYMBOL(zfs_prop_encryption_key_param);
EXPORT_SYMBOL(zfs_prop_valid_keylocation);
EXPORT_SYMBOL(zfs_prop_setonce);
EXPORT_SYMBOL(zfs_prop_to_name);
EXPORT_SYMBOL(zfs_name_to_prop);
EXPORT_SYMBOL(zfs_prop_user);
EXPORT_SYMBOL(zfs_prop_userquota);
EXPORT_SYMBOL(zfs_prop_index_to_string);
EXPORT_SYMBOL(zfs_prop_string_to_index);
EXPORT_SYMBOL(zfs_prop_valid_for_type);
EXPORT_SYMBOL(zfs_prop_written);
diff --git a/module/zfs/arc.c b/module/zfs/arc.c
index 46d7788d6ee8..bf019df16540 100644
--- a/module/zfs/arc.c
+++ b/module/zfs/arc.c
@@ -1,11137 +1,11138 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2018, Joyent, Inc.
* Copyright (c) 2011, 2020, Delphix. All rights reserved.
* Copyright (c) 2014, Saso Kiselkov. All rights reserved.
* Copyright (c) 2017, Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
* Copyright (c) 2020, George Amanakis. All rights reserved.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
* Copyright (c) 2020, The FreeBSD Foundation [1]
*
* [1] Portions of this software were developed by Allan Jude
* under sponsorship from the FreeBSD Foundation.
*/
/*
* DVA-based Adjustable Replacement Cache
*
* While much of the theory of operation used here is
* based on the self-tuning, low overhead replacement cache
* presented by Megiddo and Modha at FAST 2003, there are some
* significant differences:
*
* 1. The Megiddo and Modha model assumes any page is evictable.
* Pages in its cache cannot be "locked" into memory. This makes
* the eviction algorithm simple: evict the last page in the list.
* This also make the performance characteristics easy to reason
* about. Our cache is not so simple. At any given moment, some
* subset of the blocks in the cache are un-evictable because we
* have handed out a reference to them. Blocks are only evictable
* when there are no external references active. This makes
* eviction far more problematic: we choose to evict the evictable
* blocks that are the "lowest" in the list.
*
* There are times when it is not possible to evict the requested
* space. In these circumstances we are unable to adjust the cache
* size. To prevent the cache growing unbounded at these times we
* implement a "cache throttle" that slows the flow of new data
* into the cache until we can make space available.
*
* 2. The Megiddo and Modha model assumes a fixed cache size.
* Pages are evicted when the cache is full and there is a cache
* miss. Our model has a variable sized cache. It grows with
* high use, but also tries to react to memory pressure from the
* operating system: decreasing its size when system memory is
* tight.
*
* 3. The Megiddo and Modha model assumes a fixed page size. All
* elements of the cache are therefore exactly the same size. So
* when adjusting the cache size following a cache miss, its simply
* a matter of choosing a single page to evict. In our model, we
* have variable sized cache blocks (ranging from 512 bytes to
* 128K bytes). We therefore choose a set of blocks to evict to make
* space for a cache miss that approximates as closely as possible
* the space used by the new block.
*
* See also: "ARC: A Self-Tuning, Low Overhead Replacement Cache"
* by N. Megiddo & D. Modha, FAST 2003
*/
/*
* The locking model:
*
* A new reference to a cache buffer can be obtained in two
* ways: 1) via a hash table lookup using the DVA as a key,
* or 2) via one of the ARC lists. The arc_read() interface
* uses method 1, while the internal ARC algorithms for
* adjusting the cache use method 2. We therefore provide two
* types of locks: 1) the hash table lock array, and 2) the
* ARC list locks.
*
* Buffers do not have their own mutexes, rather they rely on the
* hash table mutexes for the bulk of their protection (i.e. most
* fields in the arc_buf_hdr_t are protected by these mutexes).
*
* buf_hash_find() returns the appropriate mutex (held) when it
* locates the requested buffer in the hash table. It returns
* NULL for the mutex if the buffer was not in the table.
*
* buf_hash_remove() expects the appropriate hash mutex to be
* already held before it is invoked.
*
* Each ARC state also has a mutex which is used to protect the
* buffer list associated with the state. When attempting to
* obtain a hash table lock while holding an ARC list lock you
* must use: mutex_tryenter() to avoid deadlock. Also note that
* the active state mutex must be held before the ghost state mutex.
*
* It as also possible to register a callback which is run when the
* arc_meta_limit is reached and no buffers can be safely evicted. In
* this case the arc user should drop a reference on some arc buffers so
* they can be reclaimed and the arc_meta_limit honored. For example,
* when using the ZPL each dentry holds a references on a znode. These
* dentries must be pruned before the arc buffer holding the znode can
* be safely evicted.
*
* Note that the majority of the performance stats are manipulated
* with atomic operations.
*
* The L2ARC uses the l2ad_mtx on each vdev for the following:
*
* - L2ARC buflist creation
* - L2ARC buflist eviction
* - L2ARC write completion, which walks L2ARC buflists
* - ARC header destruction, as it removes from L2ARC buflists
* - ARC header release, as it removes from L2ARC buflists
*/
/*
* ARC operation:
*
* Every block that is in the ARC is tracked by an arc_buf_hdr_t structure.
* This structure can point either to a block that is still in the cache or to
* one that is only accessible in an L2 ARC device, or it can provide
* information about a block that was recently evicted. If a block is
* only accessible in the L2ARC, then the arc_buf_hdr_t only has enough
* information to retrieve it from the L2ARC device. This information is
* stored in the l2arc_buf_hdr_t sub-structure of the arc_buf_hdr_t. A block
* that is in this state cannot access the data directly.
*
* Blocks that are actively being referenced or have not been evicted
* are cached in the L1ARC. The L1ARC (l1arc_buf_hdr_t) is a structure within
* the arc_buf_hdr_t that will point to the data block in memory. A block can
* only be read by a consumer if it has an l1arc_buf_hdr_t. The L1ARC
* caches data in two ways -- in a list of ARC buffers (arc_buf_t) and
* also in the arc_buf_hdr_t's private physical data block pointer (b_pabd).
*
* The L1ARC's data pointer may or may not be uncompressed. The ARC has the
* ability to store the physical data (b_pabd) associated with the DVA of the
* arc_buf_hdr_t. Since the b_pabd is a copy of the on-disk physical block,
* it will match its on-disk compression characteristics. This behavior can be
* disabled by setting 'zfs_compressed_arc_enabled' to B_FALSE. When the
* compressed ARC functionality is disabled, the b_pabd will point to an
* uncompressed version of the on-disk data.
*
* Data in the L1ARC is not accessed by consumers of the ARC directly. Each
* arc_buf_hdr_t can have multiple ARC buffers (arc_buf_t) which reference it.
* Each ARC buffer (arc_buf_t) is being actively accessed by a specific ARC
* consumer. The ARC will provide references to this data and will keep it
* cached until it is no longer in use. The ARC caches only the L1ARC's physical
* data block and will evict any arc_buf_t that is no longer referenced. The
* amount of memory consumed by the arc_buf_ts' data buffers can be seen via the
* "overhead_size" kstat.
*
* Depending on the consumer, an arc_buf_t can be requested in uncompressed or
* compressed form. The typical case is that consumers will want uncompressed
* data, and when that happens a new data buffer is allocated where the data is
* decompressed for them to use. Currently the only consumer who wants
* compressed arc_buf_t's is "zfs send", when it streams data exactly as it
* exists on disk. When this happens, the arc_buf_t's data buffer is shared
* with the arc_buf_hdr_t.
*
* Here is a diagram showing an arc_buf_hdr_t referenced by two arc_buf_t's. The
* first one is owned by a compressed send consumer (and therefore references
* the same compressed data buffer as the arc_buf_hdr_t) and the second could be
* used by any other consumer (and has its own uncompressed copy of the data
* buffer).
*
* arc_buf_hdr_t
* +-----------+
* | fields |
* | common to |
* | L1- and |
* | L2ARC |
* +-----------+
* | l2arc_buf_hdr_t
* | |
* +-----------+
* | l1arc_buf_hdr_t
* | | arc_buf_t
* | b_buf +------------>+-----------+ arc_buf_t
* | b_pabd +-+ |b_next +---->+-----------+
* +-----------+ | |-----------| |b_next +-->NULL
* | |b_comp = T | +-----------+
* | |b_data +-+ |b_comp = F |
* | +-----------+ | |b_data +-+
* +->+------+ | +-----------+ |
* compressed | | | |
* data | |<--------------+ | uncompressed
* +------+ compressed, | data
* shared +-->+------+
* data | |
* | |
* +------+
*
* When a consumer reads a block, the ARC must first look to see if the
* arc_buf_hdr_t is cached. If the hdr is cached then the ARC allocates a new
* arc_buf_t and either copies uncompressed data into a new data buffer from an
* existing uncompressed arc_buf_t, decompresses the hdr's b_pabd buffer into a
* new data buffer, or shares the hdr's b_pabd buffer, depending on whether the
* hdr is compressed and the desired compression characteristics of the
* arc_buf_t consumer. If the arc_buf_t ends up sharing data with the
* arc_buf_hdr_t and both of them are uncompressed then the arc_buf_t must be
* the last buffer in the hdr's b_buf list, however a shared compressed buf can
* be anywhere in the hdr's list.
*
* The diagram below shows an example of an uncompressed ARC hdr that is
* sharing its data with an arc_buf_t (note that the shared uncompressed buf is
* the last element in the buf list):
*
* arc_buf_hdr_t
* +-----------+
* | |
* | |
* | |
* +-----------+
* l2arc_buf_hdr_t| |
* | |
* +-----------+
* l1arc_buf_hdr_t| |
* | | arc_buf_t (shared)
* | b_buf +------------>+---------+ arc_buf_t
* | | |b_next +---->+---------+
* | b_pabd +-+ |---------| |b_next +-->NULL
* +-----------+ | | | +---------+
* | |b_data +-+ | |
* | +---------+ | |b_data +-+
* +->+------+ | +---------+ |
* | | | |
* uncompressed | | | |
* data +------+ | |
* ^ +->+------+ |
* | uncompressed | | |
* | data | | |
* | +------+ |
* +---------------------------------+
*
* Writing to the ARC requires that the ARC first discard the hdr's b_pabd
* since the physical block is about to be rewritten. The new data contents
* will be contained in the arc_buf_t. As the I/O pipeline performs the write,
* it may compress the data before writing it to disk. The ARC will be called
* with the transformed data and will bcopy the transformed on-disk block into
* a newly allocated b_pabd. Writes are always done into buffers which have
* either been loaned (and hence are new and don't have other readers) or
* buffers which have been released (and hence have their own hdr, if there
* were originally other readers of the buf's original hdr). This ensures that
* the ARC only needs to update a single buf and its hdr after a write occurs.
*
* When the L2ARC is in use, it will also take advantage of the b_pabd. The
* L2ARC will always write the contents of b_pabd to the L2ARC. This means
* that when compressed ARC is enabled that the L2ARC blocks are identical
* to the on-disk block in the main data pool. This provides a significant
* advantage since the ARC can leverage the bp's checksum when reading from the
* L2ARC to determine if the contents are valid. However, if the compressed
* ARC is disabled, then the L2ARC's block must be transformed to look
* like the physical block in the main data pool before comparing the
* checksum and determining its validity.
*
* The L1ARC has a slightly different system for storing encrypted data.
* Raw (encrypted + possibly compressed) data has a few subtle differences from
* data that is just compressed. The biggest difference is that it is not
* possible to decrypt encrypted data (or vice-versa) if the keys aren't loaded.
* The other difference is that encryption cannot be treated as a suggestion.
* If a caller would prefer compressed data, but they actually wind up with
* uncompressed data the worst thing that could happen is there might be a
* performance hit. If the caller requests encrypted data, however, we must be
* sure they actually get it or else secret information could be leaked. Raw
* data is stored in hdr->b_crypt_hdr.b_rabd. An encrypted header, therefore,
* may have both an encrypted version and a decrypted version of its data at
* once. When a caller needs a raw arc_buf_t, it is allocated and the data is
* copied out of this header. To avoid complications with b_pabd, raw buffers
* cannot be shared.
*/
#include <sys/spa.h>
#include <sys/zio.h>
#include <sys/spa_impl.h>
#include <sys/zio_compress.h>
#include <sys/zio_checksum.h>
#include <sys/zfs_context.h>
#include <sys/arc.h>
#include <sys/zfs_refcount.h>
#include <sys/vdev.h>
#include <sys/vdev_impl.h>
#include <sys/dsl_pool.h>
#include <sys/multilist.h>
#include <sys/abd.h>
#include <sys/zil.h>
#include <sys/fm/fs/zfs.h>
#include <sys/callb.h>
#include <sys/kstat.h>
#include <sys/zthr.h>
#include <zfs_fletcher.h>
#include <sys/arc_impl.h>
#include <sys/trace_zfs.h>
#include <sys/aggsum.h>
#include <sys/wmsum.h>
#include <cityhash.h>
#include <sys/vdev_trim.h>
#include <sys/zfs_racct.h>
#include <sys/zstd/zstd.h>
#ifndef _KERNEL
/* set with ZFS_DEBUG=watch, to enable watchpoints on frozen buffers */
boolean_t arc_watch = B_FALSE;
#endif
/*
* This thread's job is to keep enough free memory in the system, by
* calling arc_kmem_reap_soon() plus arc_reduce_target_size(), which improves
* arc_available_memory().
*/
static zthr_t *arc_reap_zthr;
/*
* This thread's job is to keep arc_size under arc_c, by calling
* arc_evict(), which improves arc_is_overflowing().
*/
static zthr_t *arc_evict_zthr;
static kmutex_t arc_evict_lock;
static boolean_t arc_evict_needed = B_FALSE;
/*
* Count of bytes evicted since boot.
*/
static uint64_t arc_evict_count;
/*
* List of arc_evict_waiter_t's, representing threads waiting for the
* arc_evict_count to reach specific values.
*/
static list_t arc_evict_waiters;
/*
* When arc_is_overflowing(), arc_get_data_impl() waits for this percent of
* the requested amount of data to be evicted. For example, by default for
* every 2KB that's evicted, 1KB of it may be "reused" by a new allocation.
* Since this is above 100%, it ensures that progress is made towards getting
* arc_size under arc_c. Since this is finite, it ensures that allocations
* can still happen, even during the potentially long time that arc_size is
* more than arc_c.
*/
-int zfs_arc_eviction_pct = 200;
+static int zfs_arc_eviction_pct = 200;
/*
* The number of headers to evict in arc_evict_state_impl() before
* dropping the sublist lock and evicting from another sublist. A lower
* value means we're more likely to evict the "correct" header (i.e. the
* oldest header in the arc state), but comes with higher overhead
* (i.e. more invocations of arc_evict_state_impl()).
*/
-int zfs_arc_evict_batch_limit = 10;
+static int zfs_arc_evict_batch_limit = 10;
/* number of seconds before growing cache again */
int arc_grow_retry = 5;
/*
* Minimum time between calls to arc_kmem_reap_soon().
*/
-int arc_kmem_cache_reap_retry_ms = 1000;
+static const int arc_kmem_cache_reap_retry_ms = 1000;
/* shift of arc_c for calculating overflow limit in arc_get_data_impl */
-int zfs_arc_overflow_shift = 8;
+static int zfs_arc_overflow_shift = 8;
/* shift of arc_c for calculating both min and max arc_p */
-int arc_p_min_shift = 4;
+static int arc_p_min_shift = 4;
/* log2(fraction of arc to reclaim) */
int arc_shrink_shift = 7;
/* percent of pagecache to reclaim arc to */
#ifdef _KERNEL
uint_t zfs_arc_pc_percent = 0;
#endif
/*
* log2(fraction of ARC which must be free to allow growing).
* I.e. If there is less than arc_c >> arc_no_grow_shift free memory,
* when reading a new block into the ARC, we will evict an equal-sized block
* from the ARC.
*
* This must be less than arc_shrink_shift, so that when we shrink the ARC,
* we will still not allow it to grow.
*/
int arc_no_grow_shift = 5;
/*
* minimum lifespan of a prefetch block in clock ticks
* (initialized in arc_init())
*/
static int arc_min_prefetch_ms;
static int arc_min_prescient_prefetch_ms;
/*
* If this percent of memory is free, don't throttle.
*/
int arc_lotsfree_percent = 10;
/*
* The arc has filled available memory and has now warmed up.
*/
boolean_t arc_warm;
/*
* These tunables are for performance analysis.
*/
unsigned long zfs_arc_max = 0;
unsigned long zfs_arc_min = 0;
unsigned long zfs_arc_meta_limit = 0;
unsigned long zfs_arc_meta_min = 0;
-unsigned long zfs_arc_dnode_limit = 0;
-unsigned long zfs_arc_dnode_reduce_percent = 10;
-int zfs_arc_grow_retry = 0;
-int zfs_arc_shrink_shift = 0;
-int zfs_arc_p_min_shift = 0;
+static unsigned long zfs_arc_dnode_limit = 0;
+static unsigned long zfs_arc_dnode_reduce_percent = 10;
+static int zfs_arc_grow_retry = 0;
+static int zfs_arc_shrink_shift = 0;
+static int zfs_arc_p_min_shift = 0;
int zfs_arc_average_blocksize = 8 * 1024; /* 8KB */
/*
- * ARC dirty data constraints for arc_tempreserve_space() throttle.
+ * ARC dirty data constraints for arc_tempreserve_space() throttle:
+ * * total dirty data limit
+ * * anon block dirty limit
+ * * each pool's anon allowance
*/
-unsigned long zfs_arc_dirty_limit_percent = 50; /* total dirty data limit */
-unsigned long zfs_arc_anon_limit_percent = 25; /* anon block dirty limit */
-unsigned long zfs_arc_pool_dirty_percent = 20; /* each pool's anon allowance */
+static const unsigned long zfs_arc_dirty_limit_percent = 50;
+static const unsigned long zfs_arc_anon_limit_percent = 25;
+static const unsigned long zfs_arc_pool_dirty_percent = 20;
/*
* Enable or disable compressed arc buffers.
*/
int zfs_compressed_arc_enabled = B_TRUE;
/*
* ARC will evict meta buffers that exceed arc_meta_limit. This
* tunable make arc_meta_limit adjustable for different workloads.
*/
-unsigned long zfs_arc_meta_limit_percent = 75;
+static unsigned long zfs_arc_meta_limit_percent = 75;
/*
* Percentage that can be consumed by dnodes of ARC meta buffers.
*/
-unsigned long zfs_arc_dnode_limit_percent = 10;
+static unsigned long zfs_arc_dnode_limit_percent = 10;
/*
- * These tunables are Linux specific
+ * These tunables are Linux-specific
*/
-unsigned long zfs_arc_sys_free = 0;
-int zfs_arc_min_prefetch_ms = 0;
-int zfs_arc_min_prescient_prefetch_ms = 0;
-int zfs_arc_p_dampener_disable = 1;
-int zfs_arc_meta_prune = 10000;
-int zfs_arc_meta_strategy = ARC_STRATEGY_META_BALANCED;
-int zfs_arc_meta_adjust_restarts = 4096;
-int zfs_arc_lotsfree_percent = 10;
+static unsigned long zfs_arc_sys_free = 0;
+static int zfs_arc_min_prefetch_ms = 0;
+static int zfs_arc_min_prescient_prefetch_ms = 0;
+static int zfs_arc_p_dampener_disable = 1;
+static int zfs_arc_meta_prune = 10000;
+static int zfs_arc_meta_strategy = ARC_STRATEGY_META_BALANCED;
+static int zfs_arc_meta_adjust_restarts = 4096;
+static int zfs_arc_lotsfree_percent = 10;
/*
* Number of arc_prune threads
*/
static int zfs_arc_prune_task_threads = 1;
/* The 6 states: */
arc_state_t ARC_anon;
arc_state_t ARC_mru;
arc_state_t ARC_mru_ghost;
arc_state_t ARC_mfu;
arc_state_t ARC_mfu_ghost;
arc_state_t ARC_l2c_only;
arc_stats_t arc_stats = {
{ "hits", KSTAT_DATA_UINT64 },
{ "misses", KSTAT_DATA_UINT64 },
{ "demand_data_hits", KSTAT_DATA_UINT64 },
{ "demand_data_misses", KSTAT_DATA_UINT64 },
{ "demand_metadata_hits", KSTAT_DATA_UINT64 },
{ "demand_metadata_misses", KSTAT_DATA_UINT64 },
{ "prefetch_data_hits", KSTAT_DATA_UINT64 },
{ "prefetch_data_misses", KSTAT_DATA_UINT64 },
{ "prefetch_metadata_hits", KSTAT_DATA_UINT64 },
{ "prefetch_metadata_misses", KSTAT_DATA_UINT64 },
{ "mru_hits", KSTAT_DATA_UINT64 },
{ "mru_ghost_hits", KSTAT_DATA_UINT64 },
{ "mfu_hits", KSTAT_DATA_UINT64 },
{ "mfu_ghost_hits", KSTAT_DATA_UINT64 },
{ "deleted", KSTAT_DATA_UINT64 },
{ "mutex_miss", KSTAT_DATA_UINT64 },
{ "access_skip", KSTAT_DATA_UINT64 },
{ "evict_skip", KSTAT_DATA_UINT64 },
{ "evict_not_enough", KSTAT_DATA_UINT64 },
{ "evict_l2_cached", KSTAT_DATA_UINT64 },
{ "evict_l2_eligible", KSTAT_DATA_UINT64 },
{ "evict_l2_eligible_mfu", KSTAT_DATA_UINT64 },
{ "evict_l2_eligible_mru", KSTAT_DATA_UINT64 },
{ "evict_l2_ineligible", KSTAT_DATA_UINT64 },
{ "evict_l2_skip", KSTAT_DATA_UINT64 },
{ "hash_elements", KSTAT_DATA_UINT64 },
{ "hash_elements_max", KSTAT_DATA_UINT64 },
{ "hash_collisions", KSTAT_DATA_UINT64 },
{ "hash_chains", KSTAT_DATA_UINT64 },
{ "hash_chain_max", KSTAT_DATA_UINT64 },
{ "p", KSTAT_DATA_UINT64 },
{ "c", KSTAT_DATA_UINT64 },
{ "c_min", KSTAT_DATA_UINT64 },
{ "c_max", KSTAT_DATA_UINT64 },
{ "size", KSTAT_DATA_UINT64 },
{ "compressed_size", KSTAT_DATA_UINT64 },
{ "uncompressed_size", KSTAT_DATA_UINT64 },
{ "overhead_size", KSTAT_DATA_UINT64 },
{ "hdr_size", KSTAT_DATA_UINT64 },
{ "data_size", KSTAT_DATA_UINT64 },
{ "metadata_size", KSTAT_DATA_UINT64 },
{ "dbuf_size", KSTAT_DATA_UINT64 },
{ "dnode_size", KSTAT_DATA_UINT64 },
{ "bonus_size", KSTAT_DATA_UINT64 },
#if defined(COMPAT_FREEBSD11)
{ "other_size", KSTAT_DATA_UINT64 },
#endif
{ "anon_size", KSTAT_DATA_UINT64 },
{ "anon_evictable_data", KSTAT_DATA_UINT64 },
{ "anon_evictable_metadata", KSTAT_DATA_UINT64 },
{ "mru_size", KSTAT_DATA_UINT64 },
{ "mru_evictable_data", KSTAT_DATA_UINT64 },
{ "mru_evictable_metadata", KSTAT_DATA_UINT64 },
{ "mru_ghost_size", KSTAT_DATA_UINT64 },
{ "mru_ghost_evictable_data", KSTAT_DATA_UINT64 },
{ "mru_ghost_evictable_metadata", KSTAT_DATA_UINT64 },
{ "mfu_size", KSTAT_DATA_UINT64 },
{ "mfu_evictable_data", KSTAT_DATA_UINT64 },
{ "mfu_evictable_metadata", KSTAT_DATA_UINT64 },
{ "mfu_ghost_size", KSTAT_DATA_UINT64 },
{ "mfu_ghost_evictable_data", KSTAT_DATA_UINT64 },
{ "mfu_ghost_evictable_metadata", KSTAT_DATA_UINT64 },
{ "l2_hits", KSTAT_DATA_UINT64 },
{ "l2_misses", KSTAT_DATA_UINT64 },
{ "l2_prefetch_asize", KSTAT_DATA_UINT64 },
{ "l2_mru_asize", KSTAT_DATA_UINT64 },
{ "l2_mfu_asize", KSTAT_DATA_UINT64 },
{ "l2_bufc_data_asize", KSTAT_DATA_UINT64 },
{ "l2_bufc_metadata_asize", KSTAT_DATA_UINT64 },
{ "l2_feeds", KSTAT_DATA_UINT64 },
{ "l2_rw_clash", KSTAT_DATA_UINT64 },
{ "l2_read_bytes", KSTAT_DATA_UINT64 },
{ "l2_write_bytes", KSTAT_DATA_UINT64 },
{ "l2_writes_sent", KSTAT_DATA_UINT64 },
{ "l2_writes_done", KSTAT_DATA_UINT64 },
{ "l2_writes_error", KSTAT_DATA_UINT64 },
{ "l2_writes_lock_retry", KSTAT_DATA_UINT64 },
{ "l2_evict_lock_retry", KSTAT_DATA_UINT64 },
{ "l2_evict_reading", KSTAT_DATA_UINT64 },
{ "l2_evict_l1cached", KSTAT_DATA_UINT64 },
{ "l2_free_on_write", KSTAT_DATA_UINT64 },
{ "l2_abort_lowmem", KSTAT_DATA_UINT64 },
{ "l2_cksum_bad", KSTAT_DATA_UINT64 },
{ "l2_io_error", KSTAT_DATA_UINT64 },
{ "l2_size", KSTAT_DATA_UINT64 },
{ "l2_asize", KSTAT_DATA_UINT64 },
{ "l2_hdr_size", KSTAT_DATA_UINT64 },
{ "l2_log_blk_writes", KSTAT_DATA_UINT64 },
{ "l2_log_blk_avg_asize", KSTAT_DATA_UINT64 },
{ "l2_log_blk_asize", KSTAT_DATA_UINT64 },
{ "l2_log_blk_count", KSTAT_DATA_UINT64 },
{ "l2_data_to_meta_ratio", KSTAT_DATA_UINT64 },
{ "l2_rebuild_success", KSTAT_DATA_UINT64 },
{ "l2_rebuild_unsupported", KSTAT_DATA_UINT64 },
{ "l2_rebuild_io_errors", KSTAT_DATA_UINT64 },
{ "l2_rebuild_dh_errors", KSTAT_DATA_UINT64 },
{ "l2_rebuild_cksum_lb_errors", KSTAT_DATA_UINT64 },
{ "l2_rebuild_lowmem", KSTAT_DATA_UINT64 },
{ "l2_rebuild_size", KSTAT_DATA_UINT64 },
{ "l2_rebuild_asize", KSTAT_DATA_UINT64 },
{ "l2_rebuild_bufs", KSTAT_DATA_UINT64 },
{ "l2_rebuild_bufs_precached", KSTAT_DATA_UINT64 },
{ "l2_rebuild_log_blks", KSTAT_DATA_UINT64 },
{ "memory_throttle_count", KSTAT_DATA_UINT64 },
{ "memory_direct_count", KSTAT_DATA_UINT64 },
{ "memory_indirect_count", KSTAT_DATA_UINT64 },
{ "memory_all_bytes", KSTAT_DATA_UINT64 },
{ "memory_free_bytes", KSTAT_DATA_UINT64 },
{ "memory_available_bytes", KSTAT_DATA_INT64 },
{ "arc_no_grow", KSTAT_DATA_UINT64 },
{ "arc_tempreserve", KSTAT_DATA_UINT64 },
{ "arc_loaned_bytes", KSTAT_DATA_UINT64 },
{ "arc_prune", KSTAT_DATA_UINT64 },
{ "arc_meta_used", KSTAT_DATA_UINT64 },
{ "arc_meta_limit", KSTAT_DATA_UINT64 },
{ "arc_dnode_limit", KSTAT_DATA_UINT64 },
{ "arc_meta_max", KSTAT_DATA_UINT64 },
{ "arc_meta_min", KSTAT_DATA_UINT64 },
{ "async_upgrade_sync", KSTAT_DATA_UINT64 },
{ "demand_hit_predictive_prefetch", KSTAT_DATA_UINT64 },
{ "demand_hit_prescient_prefetch", KSTAT_DATA_UINT64 },
{ "arc_need_free", KSTAT_DATA_UINT64 },
{ "arc_sys_free", KSTAT_DATA_UINT64 },
{ "arc_raw_size", KSTAT_DATA_UINT64 },
{ "cached_only_in_progress", KSTAT_DATA_UINT64 },
{ "abd_chunk_waste_size", KSTAT_DATA_UINT64 },
};
arc_sums_t arc_sums;
#define ARCSTAT_MAX(stat, val) { \
uint64_t m; \
while ((val) > (m = arc_stats.stat.value.ui64) && \
(m != atomic_cas_64(&arc_stats.stat.value.ui64, m, (val)))) \
continue; \
}
/*
* We define a macro to allow ARC hits/misses to be easily broken down by
* two separate conditions, giving a total of four different subtypes for
* each of hits and misses (so eight statistics total).
*/
#define ARCSTAT_CONDSTAT(cond1, stat1, notstat1, cond2, stat2, notstat2, stat) \
if (cond1) { \
if (cond2) { \
ARCSTAT_BUMP(arcstat_##stat1##_##stat2##_##stat); \
} else { \
ARCSTAT_BUMP(arcstat_##stat1##_##notstat2##_##stat); \
} \
} else { \
if (cond2) { \
ARCSTAT_BUMP(arcstat_##notstat1##_##stat2##_##stat); \
} else { \
ARCSTAT_BUMP(arcstat_##notstat1##_##notstat2##_##stat);\
} \
}
/*
* This macro allows us to use kstats as floating averages. Each time we
* update this kstat, we first factor it and the update value by
* ARCSTAT_AVG_FACTOR to shrink the new value's contribution to the overall
* average. This macro assumes that integer loads and stores are atomic, but
* is not safe for multiple writers updating the kstat in parallel (only the
* last writer's update will remain).
*/
#define ARCSTAT_F_AVG_FACTOR 3
#define ARCSTAT_F_AVG(stat, value) \
do { \
uint64_t x = ARCSTAT(stat); \
x = x - x / ARCSTAT_F_AVG_FACTOR + \
(value) / ARCSTAT_F_AVG_FACTOR; \
ARCSTAT(stat) = x; \
} while (0)
-kstat_t *arc_ksp;
+static kstat_t *arc_ksp;
/*
* There are several ARC variables that are critical to export as kstats --
* but we don't want to have to grovel around in the kstat whenever we wish to
* manipulate them. For these variables, we therefore define them to be in
* terms of the statistic variable. This assures that we are not introducing
* the possibility of inconsistency by having shadow copies of the variables,
* while still allowing the code to be readable.
*/
#define arc_tempreserve ARCSTAT(arcstat_tempreserve)
#define arc_loaned_bytes ARCSTAT(arcstat_loaned_bytes)
#define arc_meta_limit ARCSTAT(arcstat_meta_limit) /* max size for metadata */
/* max size for dnodes */
#define arc_dnode_size_limit ARCSTAT(arcstat_dnode_limit)
#define arc_meta_min ARCSTAT(arcstat_meta_min) /* min size for metadata */
#define arc_need_free ARCSTAT(arcstat_need_free) /* waiting to be evicted */
hrtime_t arc_growtime;
list_t arc_prune_list;
kmutex_t arc_prune_mtx;
taskq_t *arc_prune_taskq;
#define GHOST_STATE(state) \
((state) == arc_mru_ghost || (state) == arc_mfu_ghost || \
(state) == arc_l2c_only)
#define HDR_IN_HASH_TABLE(hdr) ((hdr)->b_flags & ARC_FLAG_IN_HASH_TABLE)
#define HDR_IO_IN_PROGRESS(hdr) ((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS)
#define HDR_IO_ERROR(hdr) ((hdr)->b_flags & ARC_FLAG_IO_ERROR)
#define HDR_PREFETCH(hdr) ((hdr)->b_flags & ARC_FLAG_PREFETCH)
#define HDR_PRESCIENT_PREFETCH(hdr) \
((hdr)->b_flags & ARC_FLAG_PRESCIENT_PREFETCH)
#define HDR_COMPRESSION_ENABLED(hdr) \
((hdr)->b_flags & ARC_FLAG_COMPRESSED_ARC)
#define HDR_L2CACHE(hdr) ((hdr)->b_flags & ARC_FLAG_L2CACHE)
#define HDR_L2_READING(hdr) \
(((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) && \
((hdr)->b_flags & ARC_FLAG_HAS_L2HDR))
#define HDR_L2_WRITING(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITING)
#define HDR_L2_EVICTED(hdr) ((hdr)->b_flags & ARC_FLAG_L2_EVICTED)
#define HDR_L2_WRITE_HEAD(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITE_HEAD)
#define HDR_PROTECTED(hdr) ((hdr)->b_flags & ARC_FLAG_PROTECTED)
#define HDR_NOAUTH(hdr) ((hdr)->b_flags & ARC_FLAG_NOAUTH)
#define HDR_SHARED_DATA(hdr) ((hdr)->b_flags & ARC_FLAG_SHARED_DATA)
#define HDR_ISTYPE_METADATA(hdr) \
((hdr)->b_flags & ARC_FLAG_BUFC_METADATA)
#define HDR_ISTYPE_DATA(hdr) (!HDR_ISTYPE_METADATA(hdr))
#define HDR_HAS_L1HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L1HDR)
#define HDR_HAS_L2HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR)
#define HDR_HAS_RABD(hdr) \
(HDR_HAS_L1HDR(hdr) && HDR_PROTECTED(hdr) && \
(hdr)->b_crypt_hdr.b_rabd != NULL)
#define HDR_ENCRYPTED(hdr) \
(HDR_PROTECTED(hdr) && DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot))
#define HDR_AUTHENTICATED(hdr) \
(HDR_PROTECTED(hdr) && !DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot))
/* For storing compression mode in b_flags */
#define HDR_COMPRESS_OFFSET (highbit64(ARC_FLAG_COMPRESS_0) - 1)
#define HDR_GET_COMPRESS(hdr) ((enum zio_compress)BF32_GET((hdr)->b_flags, \
HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS))
#define HDR_SET_COMPRESS(hdr, cmp) BF32_SET((hdr)->b_flags, \
HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS, (cmp));
#define ARC_BUF_LAST(buf) ((buf)->b_next == NULL)
#define ARC_BUF_SHARED(buf) ((buf)->b_flags & ARC_BUF_FLAG_SHARED)
#define ARC_BUF_COMPRESSED(buf) ((buf)->b_flags & ARC_BUF_FLAG_COMPRESSED)
#define ARC_BUF_ENCRYPTED(buf) ((buf)->b_flags & ARC_BUF_FLAG_ENCRYPTED)
/*
* Other sizes
*/
#define HDR_FULL_CRYPT_SIZE ((int64_t)sizeof (arc_buf_hdr_t))
#define HDR_FULL_SIZE ((int64_t)offsetof(arc_buf_hdr_t, b_crypt_hdr))
#define HDR_L2ONLY_SIZE ((int64_t)offsetof(arc_buf_hdr_t, b_l1hdr))
/*
* Hash table routines
*/
#define BUF_LOCKS 2048
typedef struct buf_hash_table {
uint64_t ht_mask;
arc_buf_hdr_t **ht_table;
kmutex_t ht_locks[BUF_LOCKS] ____cacheline_aligned;
} buf_hash_table_t;
static buf_hash_table_t buf_hash_table;
#define BUF_HASH_INDEX(spa, dva, birth) \
(buf_hash(spa, dva, birth) & buf_hash_table.ht_mask)
#define BUF_HASH_LOCK(idx) (&buf_hash_table.ht_locks[idx & (BUF_LOCKS-1)])
#define HDR_LOCK(hdr) \
(BUF_HASH_LOCK(BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth)))
uint64_t zfs_crc64_table[256];
/*
* Level 2 ARC
*/
#define L2ARC_WRITE_SIZE (8 * 1024 * 1024) /* initial write max */
#define L2ARC_HEADROOM 2 /* num of writes */
/*
* If we discover during ARC scan any buffers to be compressed, we boost
* our headroom for the next scanning cycle by this percentage multiple.
*/
#define L2ARC_HEADROOM_BOOST 200
#define L2ARC_FEED_SECS 1 /* caching interval secs */
#define L2ARC_FEED_MIN_MS 200 /* min caching interval ms */
/*
* We can feed L2ARC from two states of ARC buffers, mru and mfu,
* and each of the state has two types: data and metadata.
*/
#define L2ARC_FEED_TYPES 4
/* L2ARC Performance Tunables */
unsigned long l2arc_write_max = L2ARC_WRITE_SIZE; /* def max write size */
unsigned long l2arc_write_boost = L2ARC_WRITE_SIZE; /* extra warmup write */
unsigned long l2arc_headroom = L2ARC_HEADROOM; /* # of dev writes */
unsigned long l2arc_headroom_boost = L2ARC_HEADROOM_BOOST;
unsigned long l2arc_feed_secs = L2ARC_FEED_SECS; /* interval seconds */
unsigned long l2arc_feed_min_ms = L2ARC_FEED_MIN_MS; /* min interval msecs */
int l2arc_noprefetch = B_TRUE; /* don't cache prefetch bufs */
int l2arc_feed_again = B_TRUE; /* turbo warmup */
int l2arc_norw = B_FALSE; /* no reads during writes */
-int l2arc_meta_percent = 33; /* limit on headers size */
+static int l2arc_meta_percent = 33; /* limit on headers size */
/*
* L2ARC Internals
*/
static list_t L2ARC_dev_list; /* device list */
static list_t *l2arc_dev_list; /* device list pointer */
static kmutex_t l2arc_dev_mtx; /* device list mutex */
static l2arc_dev_t *l2arc_dev_last; /* last device used */
static list_t L2ARC_free_on_write; /* free after write buf list */
static list_t *l2arc_free_on_write; /* free after write list ptr */
static kmutex_t l2arc_free_on_write_mtx; /* mutex for list */
static uint64_t l2arc_ndev; /* number of devices */
typedef struct l2arc_read_callback {
arc_buf_hdr_t *l2rcb_hdr; /* read header */
blkptr_t l2rcb_bp; /* original blkptr */
zbookmark_phys_t l2rcb_zb; /* original bookmark */
int l2rcb_flags; /* original flags */
abd_t *l2rcb_abd; /* temporary buffer */
} l2arc_read_callback_t;
typedef struct l2arc_data_free {
/* protected by l2arc_free_on_write_mtx */
abd_t *l2df_abd;
size_t l2df_size;
arc_buf_contents_t l2df_type;
list_node_t l2df_list_node;
} l2arc_data_free_t;
typedef enum arc_fill_flags {
ARC_FILL_LOCKED = 1 << 0, /* hdr lock is held */
ARC_FILL_COMPRESSED = 1 << 1, /* fill with compressed data */
ARC_FILL_ENCRYPTED = 1 << 2, /* fill with encrypted data */
ARC_FILL_NOAUTH = 1 << 3, /* don't attempt to authenticate */
ARC_FILL_IN_PLACE = 1 << 4 /* fill in place (special case) */
} arc_fill_flags_t;
typedef enum arc_ovf_level {
ARC_OVF_NONE, /* ARC within target size. */
ARC_OVF_SOME, /* ARC is slightly overflowed. */
ARC_OVF_SEVERE /* ARC is severely overflowed. */
} arc_ovf_level_t;
static kmutex_t l2arc_feed_thr_lock;
static kcondvar_t l2arc_feed_thr_cv;
static uint8_t l2arc_thread_exit;
static kmutex_t l2arc_rebuild_thr_lock;
static kcondvar_t l2arc_rebuild_thr_cv;
enum arc_hdr_alloc_flags {
ARC_HDR_ALLOC_RDATA = 0x1,
ARC_HDR_DO_ADAPT = 0x2,
ARC_HDR_USE_RESERVE = 0x4,
};
static abd_t *arc_get_data_abd(arc_buf_hdr_t *, uint64_t, void *, int);
static void *arc_get_data_buf(arc_buf_hdr_t *, uint64_t, void *);
static void arc_get_data_impl(arc_buf_hdr_t *, uint64_t, void *, int);
static void arc_free_data_abd(arc_buf_hdr_t *, abd_t *, uint64_t, void *);
static void arc_free_data_buf(arc_buf_hdr_t *, void *, uint64_t, void *);
static void arc_free_data_impl(arc_buf_hdr_t *hdr, uint64_t size, void *tag);
static void arc_hdr_free_abd(arc_buf_hdr_t *, boolean_t);
static void arc_hdr_alloc_abd(arc_buf_hdr_t *, int);
static void arc_access(arc_buf_hdr_t *, kmutex_t *);
static void arc_buf_watch(arc_buf_t *);
static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *);
static uint32_t arc_bufc_to_flags(arc_buf_contents_t);
static inline void arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags);
static inline void arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags);
static boolean_t l2arc_write_eligible(uint64_t, arc_buf_hdr_t *);
static void l2arc_read_done(zio_t *);
static void l2arc_do_free_on_write(void);
static void l2arc_hdr_arcstats_update(arc_buf_hdr_t *hdr, boolean_t incr,
boolean_t state_only);
#define l2arc_hdr_arcstats_increment(hdr) \
l2arc_hdr_arcstats_update((hdr), B_TRUE, B_FALSE)
#define l2arc_hdr_arcstats_decrement(hdr) \
l2arc_hdr_arcstats_update((hdr), B_FALSE, B_FALSE)
#define l2arc_hdr_arcstats_increment_state(hdr) \
l2arc_hdr_arcstats_update((hdr), B_TRUE, B_TRUE)
#define l2arc_hdr_arcstats_decrement_state(hdr) \
l2arc_hdr_arcstats_update((hdr), B_FALSE, B_TRUE)
/*
* l2arc_exclude_special : A zfs module parameter that controls whether buffers
* present on special vdevs are eligibile for caching in L2ARC. If
* set to 1, exclude dbufs on special vdevs from being cached to
* L2ARC.
*/
int l2arc_exclude_special = 0;
/*
* l2arc_mfuonly : A ZFS module parameter that controls whether only MFU
* metadata and data are cached from ARC into L2ARC.
*/
-int l2arc_mfuonly = 0;
+static int l2arc_mfuonly = 0;
/*
* L2ARC TRIM
* l2arc_trim_ahead : A ZFS module parameter that controls how much ahead of
* the current write size (l2arc_write_max) we should TRIM if we
* have filled the device. It is defined as a percentage of the
* write size. If set to 100 we trim twice the space required to
* accommodate upcoming writes. A minimum of 64MB will be trimmed.
* It also enables TRIM of the whole L2ARC device upon creation or
* addition to an existing pool or if the header of the device is
* invalid upon importing a pool or onlining a cache device. The
* default is 0, which disables TRIM on L2ARC altogether as it can
* put significant stress on the underlying storage devices. This
* will vary depending of how well the specific device handles
* these commands.
*/
-unsigned long l2arc_trim_ahead = 0;
+static unsigned long l2arc_trim_ahead = 0;
/*
* Performance tuning of L2ARC persistence:
*
* l2arc_rebuild_enabled : A ZFS module parameter that controls whether adding
* an L2ARC device (either at pool import or later) will attempt
* to rebuild L2ARC buffer contents.
* l2arc_rebuild_blocks_min_l2size : A ZFS module parameter that controls
* whether log blocks are written to the L2ARC device. If the L2ARC
* device is less than 1GB, the amount of data l2arc_evict()
* evicts is significant compared to the amount of restored L2ARC
* data. In this case do not write log blocks in L2ARC in order
* not to waste space.
*/
-int l2arc_rebuild_enabled = B_TRUE;
-unsigned long l2arc_rebuild_blocks_min_l2size = 1024 * 1024 * 1024;
+static int l2arc_rebuild_enabled = B_TRUE;
+static unsigned long l2arc_rebuild_blocks_min_l2size = 1024 * 1024 * 1024;
/* L2ARC persistence rebuild control routines. */
void l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen);
static void l2arc_dev_rebuild_thread(void *arg);
static int l2arc_rebuild(l2arc_dev_t *dev);
/* L2ARC persistence read I/O routines. */
static int l2arc_dev_hdr_read(l2arc_dev_t *dev);
static int l2arc_log_blk_read(l2arc_dev_t *dev,
const l2arc_log_blkptr_t *this_lp, const l2arc_log_blkptr_t *next_lp,
l2arc_log_blk_phys_t *this_lb, l2arc_log_blk_phys_t *next_lb,
zio_t *this_io, zio_t **next_io);
static zio_t *l2arc_log_blk_fetch(vdev_t *vd,
const l2arc_log_blkptr_t *lp, l2arc_log_blk_phys_t *lb);
static void l2arc_log_blk_fetch_abort(zio_t *zio);
/* L2ARC persistence block restoration routines. */
static void l2arc_log_blk_restore(l2arc_dev_t *dev,
const l2arc_log_blk_phys_t *lb, uint64_t lb_asize);
static void l2arc_hdr_restore(const l2arc_log_ent_phys_t *le,
l2arc_dev_t *dev);
/* L2ARC persistence write I/O routines. */
static void l2arc_log_blk_commit(l2arc_dev_t *dev, zio_t *pio,
l2arc_write_callback_t *cb);
/* L2ARC persistence auxiliary routines. */
boolean_t l2arc_log_blkptr_valid(l2arc_dev_t *dev,
const l2arc_log_blkptr_t *lbp);
static boolean_t l2arc_log_blk_insert(l2arc_dev_t *dev,
const arc_buf_hdr_t *ab);
boolean_t l2arc_range_check_overlap(uint64_t bottom,
uint64_t top, uint64_t check);
static void l2arc_blk_fetch_done(zio_t *zio);
static inline uint64_t
l2arc_log_blk_overhead(uint64_t write_sz, l2arc_dev_t *dev);
/*
* We use Cityhash for this. It's fast, and has good hash properties without
* requiring any large static buffers.
*/
static uint64_t
buf_hash(uint64_t spa, const dva_t *dva, uint64_t birth)
{
return (cityhash4(spa, dva->dva_word[0], dva->dva_word[1], birth));
}
#define HDR_EMPTY(hdr) \
((hdr)->b_dva.dva_word[0] == 0 && \
(hdr)->b_dva.dva_word[1] == 0)
#define HDR_EMPTY_OR_LOCKED(hdr) \
(HDR_EMPTY(hdr) || MUTEX_HELD(HDR_LOCK(hdr)))
#define HDR_EQUAL(spa, dva, birth, hdr) \
((hdr)->b_dva.dva_word[0] == (dva)->dva_word[0]) && \
((hdr)->b_dva.dva_word[1] == (dva)->dva_word[1]) && \
((hdr)->b_birth == birth) && ((hdr)->b_spa == spa)
static void
buf_discard_identity(arc_buf_hdr_t *hdr)
{
hdr->b_dva.dva_word[0] = 0;
hdr->b_dva.dva_word[1] = 0;
hdr->b_birth = 0;
}
static arc_buf_hdr_t *
buf_hash_find(uint64_t spa, const blkptr_t *bp, kmutex_t **lockp)
{
const dva_t *dva = BP_IDENTITY(bp);
uint64_t birth = BP_PHYSICAL_BIRTH(bp);
uint64_t idx = BUF_HASH_INDEX(spa, dva, birth);
kmutex_t *hash_lock = BUF_HASH_LOCK(idx);
arc_buf_hdr_t *hdr;
mutex_enter(hash_lock);
for (hdr = buf_hash_table.ht_table[idx]; hdr != NULL;
hdr = hdr->b_hash_next) {
if (HDR_EQUAL(spa, dva, birth, hdr)) {
*lockp = hash_lock;
return (hdr);
}
}
mutex_exit(hash_lock);
*lockp = NULL;
return (NULL);
}
/*
* Insert an entry into the hash table. If there is already an element
* equal to elem in the hash table, then the already existing element
* will be returned and the new element will not be inserted.
* Otherwise returns NULL.
* If lockp == NULL, the caller is assumed to already hold the hash lock.
*/
static arc_buf_hdr_t *
buf_hash_insert(arc_buf_hdr_t *hdr, kmutex_t **lockp)
{
uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth);
kmutex_t *hash_lock = BUF_HASH_LOCK(idx);
arc_buf_hdr_t *fhdr;
uint32_t i;
ASSERT(!DVA_IS_EMPTY(&hdr->b_dva));
ASSERT(hdr->b_birth != 0);
ASSERT(!HDR_IN_HASH_TABLE(hdr));
if (lockp != NULL) {
*lockp = hash_lock;
mutex_enter(hash_lock);
} else {
ASSERT(MUTEX_HELD(hash_lock));
}
for (fhdr = buf_hash_table.ht_table[idx], i = 0; fhdr != NULL;
fhdr = fhdr->b_hash_next, i++) {
if (HDR_EQUAL(hdr->b_spa, &hdr->b_dva, hdr->b_birth, fhdr))
return (fhdr);
}
hdr->b_hash_next = buf_hash_table.ht_table[idx];
buf_hash_table.ht_table[idx] = hdr;
arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE);
/* collect some hash table performance data */
if (i > 0) {
ARCSTAT_BUMP(arcstat_hash_collisions);
if (i == 1)
ARCSTAT_BUMP(arcstat_hash_chains);
ARCSTAT_MAX(arcstat_hash_chain_max, i);
}
uint64_t he = atomic_inc_64_nv(
&arc_stats.arcstat_hash_elements.value.ui64);
ARCSTAT_MAX(arcstat_hash_elements_max, he);
return (NULL);
}
static void
buf_hash_remove(arc_buf_hdr_t *hdr)
{
arc_buf_hdr_t *fhdr, **hdrp;
uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth);
ASSERT(MUTEX_HELD(BUF_HASH_LOCK(idx)));
ASSERT(HDR_IN_HASH_TABLE(hdr));
hdrp = &buf_hash_table.ht_table[idx];
while ((fhdr = *hdrp) != hdr) {
ASSERT3P(fhdr, !=, NULL);
hdrp = &fhdr->b_hash_next;
}
*hdrp = hdr->b_hash_next;
hdr->b_hash_next = NULL;
arc_hdr_clear_flags(hdr, ARC_FLAG_IN_HASH_TABLE);
/* collect some hash table performance data */
atomic_dec_64(&arc_stats.arcstat_hash_elements.value.ui64);
if (buf_hash_table.ht_table[idx] &&
buf_hash_table.ht_table[idx]->b_hash_next == NULL)
ARCSTAT_BUMPDOWN(arcstat_hash_chains);
}
/*
* Global data structures and functions for the buf kmem cache.
*/
static kmem_cache_t *hdr_full_cache;
static kmem_cache_t *hdr_full_crypt_cache;
static kmem_cache_t *hdr_l2only_cache;
static kmem_cache_t *buf_cache;
static void
buf_fini(void)
{
#if defined(_KERNEL)
/*
* Large allocations which do not require contiguous pages
* should be using vmem_free() in the linux kernel\
*/
vmem_free(buf_hash_table.ht_table,
(buf_hash_table.ht_mask + 1) * sizeof (void *));
#else
kmem_free(buf_hash_table.ht_table,
(buf_hash_table.ht_mask + 1) * sizeof (void *));
#endif
for (int i = 0; i < BUF_LOCKS; i++)
mutex_destroy(BUF_HASH_LOCK(i));
kmem_cache_destroy(hdr_full_cache);
kmem_cache_destroy(hdr_full_crypt_cache);
kmem_cache_destroy(hdr_l2only_cache);
kmem_cache_destroy(buf_cache);
}
/*
* Constructor callback - called when the cache is empty
* and a new buf is requested.
*/
static int
hdr_full_cons(void *vbuf, void *unused, int kmflag)
{
(void) unused, (void) kmflag;
arc_buf_hdr_t *hdr = vbuf;
bzero(hdr, HDR_FULL_SIZE);
hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
cv_init(&hdr->b_l1hdr.b_cv, NULL, CV_DEFAULT, NULL);
zfs_refcount_create(&hdr->b_l1hdr.b_refcnt);
mutex_init(&hdr->b_l1hdr.b_freeze_lock, NULL, MUTEX_DEFAULT, NULL);
list_link_init(&hdr->b_l1hdr.b_arc_node);
list_link_init(&hdr->b_l2hdr.b_l2node);
multilist_link_init(&hdr->b_l1hdr.b_arc_node);
arc_space_consume(HDR_FULL_SIZE, ARC_SPACE_HDRS);
return (0);
}
static int
hdr_full_crypt_cons(void *vbuf, void *unused, int kmflag)
{
(void) unused;
arc_buf_hdr_t *hdr = vbuf;
hdr_full_cons(vbuf, unused, kmflag);
bzero(&hdr->b_crypt_hdr, sizeof (hdr->b_crypt_hdr));
arc_space_consume(sizeof (hdr->b_crypt_hdr), ARC_SPACE_HDRS);
return (0);
}
static int
hdr_l2only_cons(void *vbuf, void *unused, int kmflag)
{
(void) unused, (void) kmflag;
arc_buf_hdr_t *hdr = vbuf;
bzero(hdr, HDR_L2ONLY_SIZE);
arc_space_consume(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS);
return (0);
}
static int
buf_cons(void *vbuf, void *unused, int kmflag)
{
(void) unused, (void) kmflag;
arc_buf_t *buf = vbuf;
bzero(buf, sizeof (arc_buf_t));
mutex_init(&buf->b_evict_lock, NULL, MUTEX_DEFAULT, NULL);
arc_space_consume(sizeof (arc_buf_t), ARC_SPACE_HDRS);
return (0);
}
/*
* Destructor callback - called when a cached buf is
* no longer required.
*/
static void
hdr_full_dest(void *vbuf, void *unused)
{
(void) unused;
arc_buf_hdr_t *hdr = vbuf;
ASSERT(HDR_EMPTY(hdr));
cv_destroy(&hdr->b_l1hdr.b_cv);
zfs_refcount_destroy(&hdr->b_l1hdr.b_refcnt);
mutex_destroy(&hdr->b_l1hdr.b_freeze_lock);
ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
arc_space_return(HDR_FULL_SIZE, ARC_SPACE_HDRS);
}
static void
hdr_full_crypt_dest(void *vbuf, void *unused)
{
(void) unused;
arc_buf_hdr_t *hdr = vbuf;
hdr_full_dest(vbuf, unused);
arc_space_return(sizeof (hdr->b_crypt_hdr), ARC_SPACE_HDRS);
}
static void
hdr_l2only_dest(void *vbuf, void *unused)
{
(void) unused;
arc_buf_hdr_t *hdr = vbuf;
ASSERT(HDR_EMPTY(hdr));
arc_space_return(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS);
}
static void
buf_dest(void *vbuf, void *unused)
{
(void) unused;
arc_buf_t *buf = vbuf;
mutex_destroy(&buf->b_evict_lock);
arc_space_return(sizeof (arc_buf_t), ARC_SPACE_HDRS);
}
static void
buf_init(void)
{
uint64_t *ct = NULL;
uint64_t hsize = 1ULL << 12;
int i, j;
/*
* The hash table is big enough to fill all of physical memory
* with an average block size of zfs_arc_average_blocksize (default 8K).
* By default, the table will take up
* totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers).
*/
while (hsize * zfs_arc_average_blocksize < arc_all_memory())
hsize <<= 1;
retry:
buf_hash_table.ht_mask = hsize - 1;
#if defined(_KERNEL)
/*
* Large allocations which do not require contiguous pages
* should be using vmem_alloc() in the linux kernel
*/
buf_hash_table.ht_table =
vmem_zalloc(hsize * sizeof (void*), KM_SLEEP);
#else
buf_hash_table.ht_table =
kmem_zalloc(hsize * sizeof (void*), KM_NOSLEEP);
#endif
if (buf_hash_table.ht_table == NULL) {
ASSERT(hsize > (1ULL << 8));
hsize >>= 1;
goto retry;
}
hdr_full_cache = kmem_cache_create("arc_buf_hdr_t_full", HDR_FULL_SIZE,
0, hdr_full_cons, hdr_full_dest, NULL, NULL, NULL, 0);
hdr_full_crypt_cache = kmem_cache_create("arc_buf_hdr_t_full_crypt",
HDR_FULL_CRYPT_SIZE, 0, hdr_full_crypt_cons, hdr_full_crypt_dest,
NULL, NULL, NULL, 0);
hdr_l2only_cache = kmem_cache_create("arc_buf_hdr_t_l2only",
HDR_L2ONLY_SIZE, 0, hdr_l2only_cons, hdr_l2only_dest, NULL,
NULL, NULL, 0);
buf_cache = kmem_cache_create("arc_buf_t", sizeof (arc_buf_t),
0, buf_cons, buf_dest, NULL, NULL, NULL, 0);
for (i = 0; i < 256; i++)
for (ct = zfs_crc64_table + i, *ct = i, j = 8; j > 0; j--)
*ct = (*ct >> 1) ^ (-(*ct & 1) & ZFS_CRC64_POLY);
for (i = 0; i < BUF_LOCKS; i++)
mutex_init(BUF_HASH_LOCK(i), NULL, MUTEX_DEFAULT, NULL);
}
#define ARC_MINTIME (hz>>4) /* 62 ms */
/*
* This is the size that the buf occupies in memory. If the buf is compressed,
* it will correspond to the compressed size. You should use this method of
* getting the buf size unless you explicitly need the logical size.
*/
uint64_t
arc_buf_size(arc_buf_t *buf)
{
return (ARC_BUF_COMPRESSED(buf) ?
HDR_GET_PSIZE(buf->b_hdr) : HDR_GET_LSIZE(buf->b_hdr));
}
uint64_t
arc_buf_lsize(arc_buf_t *buf)
{
return (HDR_GET_LSIZE(buf->b_hdr));
}
/*
* This function will return B_TRUE if the buffer is encrypted in memory.
* This buffer can be decrypted by calling arc_untransform().
*/
boolean_t
arc_is_encrypted(arc_buf_t *buf)
{
return (ARC_BUF_ENCRYPTED(buf) != 0);
}
/*
* Returns B_TRUE if the buffer represents data that has not had its MAC
* verified yet.
*/
boolean_t
arc_is_unauthenticated(arc_buf_t *buf)
{
return (HDR_NOAUTH(buf->b_hdr) != 0);
}
void
arc_get_raw_params(arc_buf_t *buf, boolean_t *byteorder, uint8_t *salt,
uint8_t *iv, uint8_t *mac)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
ASSERT(HDR_PROTECTED(hdr));
bcopy(hdr->b_crypt_hdr.b_salt, salt, ZIO_DATA_SALT_LEN);
bcopy(hdr->b_crypt_hdr.b_iv, iv, ZIO_DATA_IV_LEN);
bcopy(hdr->b_crypt_hdr.b_mac, mac, ZIO_DATA_MAC_LEN);
*byteorder = (hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ?
ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER;
}
/*
* Indicates how this buffer is compressed in memory. If it is not compressed
* the value will be ZIO_COMPRESS_OFF. It can be made normally readable with
* arc_untransform() as long as it is also unencrypted.
*/
enum zio_compress
arc_get_compression(arc_buf_t *buf)
{
return (ARC_BUF_COMPRESSED(buf) ?
HDR_GET_COMPRESS(buf->b_hdr) : ZIO_COMPRESS_OFF);
}
/*
* Return the compression algorithm used to store this data in the ARC. If ARC
* compression is enabled or this is an encrypted block, this will be the same
* as what's used to store it on-disk. Otherwise, this will be ZIO_COMPRESS_OFF.
*/
static inline enum zio_compress
arc_hdr_get_compress(arc_buf_hdr_t *hdr)
{
return (HDR_COMPRESSION_ENABLED(hdr) ?
HDR_GET_COMPRESS(hdr) : ZIO_COMPRESS_OFF);
}
uint8_t
arc_get_complevel(arc_buf_t *buf)
{
return (buf->b_hdr->b_complevel);
}
static inline boolean_t
arc_buf_is_shared(arc_buf_t *buf)
{
boolean_t shared = (buf->b_data != NULL &&
buf->b_hdr->b_l1hdr.b_pabd != NULL &&
abd_is_linear(buf->b_hdr->b_l1hdr.b_pabd) &&
buf->b_data == abd_to_buf(buf->b_hdr->b_l1hdr.b_pabd));
IMPLY(shared, HDR_SHARED_DATA(buf->b_hdr));
IMPLY(shared, ARC_BUF_SHARED(buf));
IMPLY(shared, ARC_BUF_COMPRESSED(buf) || ARC_BUF_LAST(buf));
/*
* It would be nice to assert arc_can_share() too, but the "hdr isn't
* already being shared" requirement prevents us from doing that.
*/
return (shared);
}
/*
* Free the checksum associated with this header. If there is no checksum, this
* is a no-op.
*/
static inline void
arc_cksum_free(arc_buf_hdr_t *hdr)
{
ASSERT(HDR_HAS_L1HDR(hdr));
mutex_enter(&hdr->b_l1hdr.b_freeze_lock);
if (hdr->b_l1hdr.b_freeze_cksum != NULL) {
kmem_free(hdr->b_l1hdr.b_freeze_cksum, sizeof (zio_cksum_t));
hdr->b_l1hdr.b_freeze_cksum = NULL;
}
mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
}
/*
* Return true iff at least one of the bufs on hdr is not compressed.
* Encrypted buffers count as compressed.
*/
static boolean_t
arc_hdr_has_uncompressed_buf(arc_buf_hdr_t *hdr)
{
ASSERT(hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY_OR_LOCKED(hdr));
for (arc_buf_t *b = hdr->b_l1hdr.b_buf; b != NULL; b = b->b_next) {
if (!ARC_BUF_COMPRESSED(b)) {
return (B_TRUE);
}
}
return (B_FALSE);
}
/*
* If we've turned on the ZFS_DEBUG_MODIFY flag, verify that the buf's data
* matches the checksum that is stored in the hdr. If there is no checksum,
* or if the buf is compressed, this is a no-op.
*/
static void
arc_cksum_verify(arc_buf_t *buf)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
zio_cksum_t zc;
if (!(zfs_flags & ZFS_DEBUG_MODIFY))
return;
if (ARC_BUF_COMPRESSED(buf))
return;
ASSERT(HDR_HAS_L1HDR(hdr));
mutex_enter(&hdr->b_l1hdr.b_freeze_lock);
if (hdr->b_l1hdr.b_freeze_cksum == NULL || HDR_IO_ERROR(hdr)) {
mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
return;
}
fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL, &zc);
if (!ZIO_CHECKSUM_EQUAL(*hdr->b_l1hdr.b_freeze_cksum, zc))
panic("buffer modified while frozen!");
mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
}
/*
* This function makes the assumption that data stored in the L2ARC
* will be transformed exactly as it is in the main pool. Because of
* this we can verify the checksum against the reading process's bp.
*/
static boolean_t
arc_cksum_is_equal(arc_buf_hdr_t *hdr, zio_t *zio)
{
ASSERT(!BP_IS_EMBEDDED(zio->io_bp));
VERIFY3U(BP_GET_PSIZE(zio->io_bp), ==, HDR_GET_PSIZE(hdr));
/*
* Block pointers always store the checksum for the logical data.
* If the block pointer has the gang bit set, then the checksum
* it represents is for the reconstituted data and not for an
* individual gang member. The zio pipeline, however, must be able to
* determine the checksum of each of the gang constituents so it
* treats the checksum comparison differently than what we need
* for l2arc blocks. This prevents us from using the
* zio_checksum_error() interface directly. Instead we must call the
* zio_checksum_error_impl() so that we can ensure the checksum is
* generated using the correct checksum algorithm and accounts for the
* logical I/O size and not just a gang fragment.
*/
return (zio_checksum_error_impl(zio->io_spa, zio->io_bp,
BP_GET_CHECKSUM(zio->io_bp), zio->io_abd, zio->io_size,
zio->io_offset, NULL) == 0);
}
/*
* Given a buf full of data, if ZFS_DEBUG_MODIFY is enabled this computes a
* checksum and attaches it to the buf's hdr so that we can ensure that the buf
* isn't modified later on. If buf is compressed or there is already a checksum
* on the hdr, this is a no-op (we only checksum uncompressed bufs).
*/
static void
arc_cksum_compute(arc_buf_t *buf)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
if (!(zfs_flags & ZFS_DEBUG_MODIFY))
return;
ASSERT(HDR_HAS_L1HDR(hdr));
mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock);
if (hdr->b_l1hdr.b_freeze_cksum != NULL || ARC_BUF_COMPRESSED(buf)) {
mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
return;
}
ASSERT(!ARC_BUF_ENCRYPTED(buf));
ASSERT(!ARC_BUF_COMPRESSED(buf));
hdr->b_l1hdr.b_freeze_cksum = kmem_alloc(sizeof (zio_cksum_t),
KM_SLEEP);
fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL,
hdr->b_l1hdr.b_freeze_cksum);
mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
arc_buf_watch(buf);
}
#ifndef _KERNEL
void
arc_buf_sigsegv(int sig, siginfo_t *si, void *unused)
{
(void) sig, (void) unused;
panic("Got SIGSEGV at address: 0x%lx\n", (long)si->si_addr);
}
#endif
static void
arc_buf_unwatch(arc_buf_t *buf)
{
#ifndef _KERNEL
if (arc_watch) {
ASSERT0(mprotect(buf->b_data, arc_buf_size(buf),
PROT_READ | PROT_WRITE));
}
#else
(void) buf;
#endif
}
static void
arc_buf_watch(arc_buf_t *buf)
{
#ifndef _KERNEL
if (arc_watch)
ASSERT0(mprotect(buf->b_data, arc_buf_size(buf),
PROT_READ));
#else
(void) buf;
#endif
}
static arc_buf_contents_t
arc_buf_type(arc_buf_hdr_t *hdr)
{
arc_buf_contents_t type;
if (HDR_ISTYPE_METADATA(hdr)) {
type = ARC_BUFC_METADATA;
} else {
type = ARC_BUFC_DATA;
}
VERIFY3U(hdr->b_type, ==, type);
return (type);
}
boolean_t
arc_is_metadata(arc_buf_t *buf)
{
return (HDR_ISTYPE_METADATA(buf->b_hdr) != 0);
}
static uint32_t
arc_bufc_to_flags(arc_buf_contents_t type)
{
switch (type) {
case ARC_BUFC_DATA:
/* metadata field is 0 if buffer contains normal data */
return (0);
case ARC_BUFC_METADATA:
return (ARC_FLAG_BUFC_METADATA);
default:
break;
}
panic("undefined ARC buffer type!");
return ((uint32_t)-1);
}
void
arc_buf_thaw(arc_buf_t *buf)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
arc_cksum_verify(buf);
/*
* Compressed buffers do not manipulate the b_freeze_cksum.
*/
if (ARC_BUF_COMPRESSED(buf))
return;
ASSERT(HDR_HAS_L1HDR(hdr));
arc_cksum_free(hdr);
arc_buf_unwatch(buf);
}
void
arc_buf_freeze(arc_buf_t *buf)
{
if (!(zfs_flags & ZFS_DEBUG_MODIFY))
return;
if (ARC_BUF_COMPRESSED(buf))
return;
ASSERT(HDR_HAS_L1HDR(buf->b_hdr));
arc_cksum_compute(buf);
}
/*
* The arc_buf_hdr_t's b_flags should never be modified directly. Instead,
* the following functions should be used to ensure that the flags are
* updated in a thread-safe way. When manipulating the flags either
* the hash_lock must be held or the hdr must be undiscoverable. This
* ensures that we're not racing with any other threads when updating
* the flags.
*/
static inline void
arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags)
{
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
hdr->b_flags |= flags;
}
static inline void
arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags)
{
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
hdr->b_flags &= ~flags;
}
/*
* Setting the compression bits in the arc_buf_hdr_t's b_flags is
* done in a special way since we have to clear and set bits
* at the same time. Consumers that wish to set the compression bits
* must use this function to ensure that the flags are updated in
* thread-safe manner.
*/
static void
arc_hdr_set_compress(arc_buf_hdr_t *hdr, enum zio_compress cmp)
{
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
/*
* Holes and embedded blocks will always have a psize = 0 so
* we ignore the compression of the blkptr and set the
* want to uncompress them. Mark them as uncompressed.
*/
if (!zfs_compressed_arc_enabled || HDR_GET_PSIZE(hdr) == 0) {
arc_hdr_clear_flags(hdr, ARC_FLAG_COMPRESSED_ARC);
ASSERT(!HDR_COMPRESSION_ENABLED(hdr));
} else {
arc_hdr_set_flags(hdr, ARC_FLAG_COMPRESSED_ARC);
ASSERT(HDR_COMPRESSION_ENABLED(hdr));
}
HDR_SET_COMPRESS(hdr, cmp);
ASSERT3U(HDR_GET_COMPRESS(hdr), ==, cmp);
}
/*
* Looks for another buf on the same hdr which has the data decompressed, copies
* from it, and returns true. If no such buf exists, returns false.
*/
static boolean_t
arc_buf_try_copy_decompressed_data(arc_buf_t *buf)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
boolean_t copied = B_FALSE;
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT3P(buf->b_data, !=, NULL);
ASSERT(!ARC_BUF_COMPRESSED(buf));
for (arc_buf_t *from = hdr->b_l1hdr.b_buf; from != NULL;
from = from->b_next) {
/* can't use our own data buffer */
if (from == buf) {
continue;
}
if (!ARC_BUF_COMPRESSED(from)) {
bcopy(from->b_data, buf->b_data, arc_buf_size(buf));
copied = B_TRUE;
break;
}
}
/*
* There were no decompressed bufs, so there should not be a
* checksum on the hdr either.
*/
if (zfs_flags & ZFS_DEBUG_MODIFY)
EQUIV(!copied, hdr->b_l1hdr.b_freeze_cksum == NULL);
return (copied);
}
/*
* Allocates an ARC buf header that's in an evicted & L2-cached state.
* This is used during l2arc reconstruction to make empty ARC buffers
* which circumvent the regular disk->arc->l2arc path and instead come
* into being in the reverse order, i.e. l2arc->arc.
*/
static arc_buf_hdr_t *
arc_buf_alloc_l2only(size_t size, arc_buf_contents_t type, l2arc_dev_t *dev,
dva_t dva, uint64_t daddr, int32_t psize, uint64_t birth,
enum zio_compress compress, uint8_t complevel, boolean_t protected,
boolean_t prefetch, arc_state_type_t arcs_state)
{
arc_buf_hdr_t *hdr;
ASSERT(size != 0);
hdr = kmem_cache_alloc(hdr_l2only_cache, KM_SLEEP);
hdr->b_birth = birth;
hdr->b_type = type;
hdr->b_flags = 0;
arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) | ARC_FLAG_HAS_L2HDR);
HDR_SET_LSIZE(hdr, size);
HDR_SET_PSIZE(hdr, psize);
arc_hdr_set_compress(hdr, compress);
hdr->b_complevel = complevel;
if (protected)
arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);
if (prefetch)
arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
hdr->b_spa = spa_load_guid(dev->l2ad_vdev->vdev_spa);
hdr->b_dva = dva;
hdr->b_l2hdr.b_dev = dev;
hdr->b_l2hdr.b_daddr = daddr;
hdr->b_l2hdr.b_arcs_state = arcs_state;
return (hdr);
}
/*
* Return the size of the block, b_pabd, that is stored in the arc_buf_hdr_t.
*/
static uint64_t
arc_hdr_size(arc_buf_hdr_t *hdr)
{
uint64_t size;
if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF &&
HDR_GET_PSIZE(hdr) > 0) {
size = HDR_GET_PSIZE(hdr);
} else {
ASSERT3U(HDR_GET_LSIZE(hdr), !=, 0);
size = HDR_GET_LSIZE(hdr);
}
return (size);
}
static int
arc_hdr_authenticate(arc_buf_hdr_t *hdr, spa_t *spa, uint64_t dsobj)
{
int ret;
uint64_t csize;
uint64_t lsize = HDR_GET_LSIZE(hdr);
uint64_t psize = HDR_GET_PSIZE(hdr);
void *tmpbuf = NULL;
abd_t *abd = hdr->b_l1hdr.b_pabd;
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
ASSERT(HDR_AUTHENTICATED(hdr));
ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
/*
* The MAC is calculated on the compressed data that is stored on disk.
* However, if compressed arc is disabled we will only have the
* decompressed data available to us now. Compress it into a temporary
* abd so we can verify the MAC. The performance overhead of this will
* be relatively low, since most objects in an encrypted objset will
* be encrypted (instead of authenticated) anyway.
*/
if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
!HDR_COMPRESSION_ENABLED(hdr)) {
tmpbuf = zio_buf_alloc(lsize);
abd = abd_get_from_buf(tmpbuf, lsize);
abd_take_ownership_of_buf(abd, B_TRUE);
csize = zio_compress_data(HDR_GET_COMPRESS(hdr),
hdr->b_l1hdr.b_pabd, tmpbuf, lsize, hdr->b_complevel);
ASSERT3U(csize, <=, psize);
abd_zero_off(abd, csize, psize - csize);
}
/*
* Authentication is best effort. We authenticate whenever the key is
* available. If we succeed we clear ARC_FLAG_NOAUTH.
*/
if (hdr->b_crypt_hdr.b_ot == DMU_OT_OBJSET) {
ASSERT3U(HDR_GET_COMPRESS(hdr), ==, ZIO_COMPRESS_OFF);
ASSERT3U(lsize, ==, psize);
ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa, dsobj, abd,
psize, hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS);
} else {
ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj, abd, psize,
hdr->b_crypt_hdr.b_mac);
}
if (ret == 0)
arc_hdr_clear_flags(hdr, ARC_FLAG_NOAUTH);
else if (ret != ENOENT)
goto error;
if (tmpbuf != NULL)
abd_free(abd);
return (0);
error:
if (tmpbuf != NULL)
abd_free(abd);
return (ret);
}
/*
* This function will take a header that only has raw encrypted data in
* b_crypt_hdr.b_rabd and decrypt it into a new buffer which is stored in
* b_l1hdr.b_pabd. If designated in the header flags, this function will
* also decompress the data.
*/
static int
arc_hdr_decrypt(arc_buf_hdr_t *hdr, spa_t *spa, const zbookmark_phys_t *zb)
{
int ret;
abd_t *cabd = NULL;
void *tmp = NULL;
boolean_t no_crypt = B_FALSE;
boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS);
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
ASSERT(HDR_ENCRYPTED(hdr));
arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT);
ret = spa_do_crypt_abd(B_FALSE, spa, zb, hdr->b_crypt_hdr.b_ot,
B_FALSE, bswap, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv,
hdr->b_crypt_hdr.b_mac, HDR_GET_PSIZE(hdr), hdr->b_l1hdr.b_pabd,
hdr->b_crypt_hdr.b_rabd, &no_crypt);
if (ret != 0)
goto error;
if (no_crypt) {
abd_copy(hdr->b_l1hdr.b_pabd, hdr->b_crypt_hdr.b_rabd,
HDR_GET_PSIZE(hdr));
}
/*
* If this header has disabled arc compression but the b_pabd is
* compressed after decrypting it, we need to decompress the newly
* decrypted data.
*/
if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
!HDR_COMPRESSION_ENABLED(hdr)) {
/*
* We want to make sure that we are correctly honoring the
* zfs_abd_scatter_enabled setting, so we allocate an abd here
* and then loan a buffer from it, rather than allocating a
* linear buffer and wrapping it in an abd later.
*/
cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr,
ARC_HDR_DO_ADAPT);
tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr));
ret = zio_decompress_data(HDR_GET_COMPRESS(hdr),
hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr),
HDR_GET_LSIZE(hdr), &hdr->b_complevel);
if (ret != 0) {
abd_return_buf(cabd, tmp, arc_hdr_size(hdr));
goto error;
}
abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd,
arc_hdr_size(hdr), hdr);
hdr->b_l1hdr.b_pabd = cabd;
}
return (0);
error:
arc_hdr_free_abd(hdr, B_FALSE);
if (cabd != NULL)
arc_free_data_buf(hdr, cabd, arc_hdr_size(hdr), hdr);
return (ret);
}
/*
* This function is called during arc_buf_fill() to prepare the header's
* abd plaintext pointer for use. This involves authenticated protected
* data and decrypting encrypted data into the plaintext abd.
*/
static int
arc_fill_hdr_crypt(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, spa_t *spa,
const zbookmark_phys_t *zb, boolean_t noauth)
{
int ret;
ASSERT(HDR_PROTECTED(hdr));
if (hash_lock != NULL)
mutex_enter(hash_lock);
if (HDR_NOAUTH(hdr) && !noauth) {
/*
* The caller requested authenticated data but our data has
* not been authenticated yet. Verify the MAC now if we can.
*/
ret = arc_hdr_authenticate(hdr, spa, zb->zb_objset);
if (ret != 0)
goto error;
} else if (HDR_HAS_RABD(hdr) && hdr->b_l1hdr.b_pabd == NULL) {
/*
* If we only have the encrypted version of the data, but the
* unencrypted version was requested we take this opportunity
* to store the decrypted version in the header for future use.
*/
ret = arc_hdr_decrypt(hdr, spa, zb);
if (ret != 0)
goto error;
}
ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
if (hash_lock != NULL)
mutex_exit(hash_lock);
return (0);
error:
if (hash_lock != NULL)
mutex_exit(hash_lock);
return (ret);
}
/*
* This function is used by the dbuf code to decrypt bonus buffers in place.
* The dbuf code itself doesn't have any locking for decrypting a shared dnode
* block, so we use the hash lock here to protect against concurrent calls to
* arc_buf_fill().
*/
static void
arc_buf_untransform_in_place(arc_buf_t *buf)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
ASSERT(HDR_ENCRYPTED(hdr));
ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE);
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
zio_crypt_copy_dnode_bonus(hdr->b_l1hdr.b_pabd, buf->b_data,
arc_buf_size(buf));
buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED;
buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
hdr->b_crypt_hdr.b_ebufcnt -= 1;
}
/*
* Given a buf that has a data buffer attached to it, this function will
* efficiently fill the buf with data of the specified compression setting from
* the hdr and update the hdr's b_freeze_cksum if necessary. If the buf and hdr
* are already sharing a data buf, no copy is performed.
*
* If the buf is marked as compressed but uncompressed data was requested, this
* will allocate a new data buffer for the buf, remove that flag, and fill the
* buf with uncompressed data. You can't request a compressed buf on a hdr with
* uncompressed data, and (since we haven't added support for it yet) if you
* want compressed data your buf must already be marked as compressed and have
* the correct-sized data buffer.
*/
static int
arc_buf_fill(arc_buf_t *buf, spa_t *spa, const zbookmark_phys_t *zb,
arc_fill_flags_t flags)
{
int error = 0;
arc_buf_hdr_t *hdr = buf->b_hdr;
boolean_t hdr_compressed =
(arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF);
boolean_t compressed = (flags & ARC_FILL_COMPRESSED) != 0;
boolean_t encrypted = (flags & ARC_FILL_ENCRYPTED) != 0;
dmu_object_byteswap_t bswap = hdr->b_l1hdr.b_byteswap;
kmutex_t *hash_lock = (flags & ARC_FILL_LOCKED) ? NULL : HDR_LOCK(hdr);
ASSERT3P(buf->b_data, !=, NULL);
IMPLY(compressed, hdr_compressed || ARC_BUF_ENCRYPTED(buf));
IMPLY(compressed, ARC_BUF_COMPRESSED(buf));
IMPLY(encrypted, HDR_ENCRYPTED(hdr));
IMPLY(encrypted, ARC_BUF_ENCRYPTED(buf));
IMPLY(encrypted, ARC_BUF_COMPRESSED(buf));
IMPLY(encrypted, !ARC_BUF_SHARED(buf));
/*
* If the caller wanted encrypted data we just need to copy it from
* b_rabd and potentially byteswap it. We won't be able to do any
* further transforms on it.
*/
if (encrypted) {
ASSERT(HDR_HAS_RABD(hdr));
abd_copy_to_buf(buf->b_data, hdr->b_crypt_hdr.b_rabd,
HDR_GET_PSIZE(hdr));
goto byteswap;
}
/*
* Adjust encrypted and authenticated headers to accommodate
* the request if needed. Dnode blocks (ARC_FILL_IN_PLACE) are
* allowed to fail decryption due to keys not being loaded
* without being marked as an IO error.
*/
if (HDR_PROTECTED(hdr)) {
error = arc_fill_hdr_crypt(hdr, hash_lock, spa,
zb, !!(flags & ARC_FILL_NOAUTH));
if (error == EACCES && (flags & ARC_FILL_IN_PLACE) != 0) {
return (error);
} else if (error != 0) {
if (hash_lock != NULL)
mutex_enter(hash_lock);
arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
if (hash_lock != NULL)
mutex_exit(hash_lock);
return (error);
}
}
/*
* There is a special case here for dnode blocks which are
* decrypting their bonus buffers. These blocks may request to
* be decrypted in-place. This is necessary because there may
* be many dnodes pointing into this buffer and there is
* currently no method to synchronize replacing the backing
* b_data buffer and updating all of the pointers. Here we use
* the hash lock to ensure there are no races. If the need
* arises for other types to be decrypted in-place, they must
* add handling here as well.
*/
if ((flags & ARC_FILL_IN_PLACE) != 0) {
ASSERT(!hdr_compressed);
ASSERT(!compressed);
ASSERT(!encrypted);
if (HDR_ENCRYPTED(hdr) && ARC_BUF_ENCRYPTED(buf)) {
ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE);
if (hash_lock != NULL)
mutex_enter(hash_lock);
arc_buf_untransform_in_place(buf);
if (hash_lock != NULL)
mutex_exit(hash_lock);
/* Compute the hdr's checksum if necessary */
arc_cksum_compute(buf);
}
return (0);
}
if (hdr_compressed == compressed) {
if (!arc_buf_is_shared(buf)) {
abd_copy_to_buf(buf->b_data, hdr->b_l1hdr.b_pabd,
arc_buf_size(buf));
}
} else {
ASSERT(hdr_compressed);
ASSERT(!compressed);
/*
* If the buf is sharing its data with the hdr, unlink it and
* allocate a new data buffer for the buf.
*/
if (arc_buf_is_shared(buf)) {
ASSERT(ARC_BUF_COMPRESSED(buf));
/* We need to give the buf its own b_data */
buf->b_flags &= ~ARC_BUF_FLAG_SHARED;
buf->b_data =
arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf);
arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
/* Previously overhead was 0; just add new overhead */
ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr));
} else if (ARC_BUF_COMPRESSED(buf)) {
/* We need to reallocate the buf's b_data */
arc_free_data_buf(hdr, buf->b_data, HDR_GET_PSIZE(hdr),
buf);
buf->b_data =
arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf);
/* We increased the size of b_data; update overhead */
ARCSTAT_INCR(arcstat_overhead_size,
HDR_GET_LSIZE(hdr) - HDR_GET_PSIZE(hdr));
}
/*
* Regardless of the buf's previous compression settings, it
* should not be compressed at the end of this function.
*/
buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
/*
* Try copying the data from another buf which already has a
* decompressed version. If that's not possible, it's time to
* bite the bullet and decompress the data from the hdr.
*/
if (arc_buf_try_copy_decompressed_data(buf)) {
/* Skip byteswapping and checksumming (already done) */
return (0);
} else {
error = zio_decompress_data(HDR_GET_COMPRESS(hdr),
hdr->b_l1hdr.b_pabd, buf->b_data,
HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr),
&hdr->b_complevel);
/*
* Absent hardware errors or software bugs, this should
* be impossible, but log it anyway so we can debug it.
*/
if (error != 0) {
zfs_dbgmsg(
"hdr %px, compress %d, psize %d, lsize %d",
hdr, arc_hdr_get_compress(hdr),
HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr));
if (hash_lock != NULL)
mutex_enter(hash_lock);
arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
if (hash_lock != NULL)
mutex_exit(hash_lock);
return (SET_ERROR(EIO));
}
}
}
byteswap:
/* Byteswap the buf's data if necessary */
if (bswap != DMU_BSWAP_NUMFUNCS) {
ASSERT(!HDR_SHARED_DATA(hdr));
ASSERT3U(bswap, <, DMU_BSWAP_NUMFUNCS);
dmu_ot_byteswap[bswap].ob_func(buf->b_data, HDR_GET_LSIZE(hdr));
}
/* Compute the hdr's checksum if necessary */
arc_cksum_compute(buf);
return (0);
}
/*
* If this function is being called to decrypt an encrypted buffer or verify an
* authenticated one, the key must be loaded and a mapping must be made
* available in the keystore via spa_keystore_create_mapping() or one of its
* callers.
*/
int
arc_untransform(arc_buf_t *buf, spa_t *spa, const zbookmark_phys_t *zb,
boolean_t in_place)
{
int ret;
arc_fill_flags_t flags = 0;
if (in_place)
flags |= ARC_FILL_IN_PLACE;
ret = arc_buf_fill(buf, spa, zb, flags);
if (ret == ECKSUM) {
/*
* Convert authentication and decryption errors to EIO
* (and generate an ereport) before leaving the ARC.
*/
ret = SET_ERROR(EIO);
spa_log_error(spa, zb);
(void) zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION,
spa, NULL, zb, NULL, 0);
}
return (ret);
}
/*
* Increment the amount of evictable space in the arc_state_t's refcount.
* We account for the space used by the hdr and the arc buf individually
* so that we can add and remove them from the refcount individually.
*/
static void
arc_evictable_space_increment(arc_buf_hdr_t *hdr, arc_state_t *state)
{
arc_buf_contents_t type = arc_buf_type(hdr);
ASSERT(HDR_HAS_L1HDR(hdr));
if (GHOST_STATE(state)) {
ASSERT0(hdr->b_l1hdr.b_bufcnt);
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
ASSERT(!HDR_HAS_RABD(hdr));
(void) zfs_refcount_add_many(&state->arcs_esize[type],
HDR_GET_LSIZE(hdr), hdr);
return;
}
if (hdr->b_l1hdr.b_pabd != NULL) {
(void) zfs_refcount_add_many(&state->arcs_esize[type],
arc_hdr_size(hdr), hdr);
}
if (HDR_HAS_RABD(hdr)) {
(void) zfs_refcount_add_many(&state->arcs_esize[type],
HDR_GET_PSIZE(hdr), hdr);
}
for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
buf = buf->b_next) {
if (arc_buf_is_shared(buf))
continue;
(void) zfs_refcount_add_many(&state->arcs_esize[type],
arc_buf_size(buf), buf);
}
}
/*
* Decrement the amount of evictable space in the arc_state_t's refcount.
* We account for the space used by the hdr and the arc buf individually
* so that we can add and remove them from the refcount individually.
*/
static void
arc_evictable_space_decrement(arc_buf_hdr_t *hdr, arc_state_t *state)
{
arc_buf_contents_t type = arc_buf_type(hdr);
ASSERT(HDR_HAS_L1HDR(hdr));
if (GHOST_STATE(state)) {
ASSERT0(hdr->b_l1hdr.b_bufcnt);
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
ASSERT(!HDR_HAS_RABD(hdr));
(void) zfs_refcount_remove_many(&state->arcs_esize[type],
HDR_GET_LSIZE(hdr), hdr);
return;
}
if (hdr->b_l1hdr.b_pabd != NULL) {
(void) zfs_refcount_remove_many(&state->arcs_esize[type],
arc_hdr_size(hdr), hdr);
}
if (HDR_HAS_RABD(hdr)) {
(void) zfs_refcount_remove_many(&state->arcs_esize[type],
HDR_GET_PSIZE(hdr), hdr);
}
for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
buf = buf->b_next) {
if (arc_buf_is_shared(buf))
continue;
(void) zfs_refcount_remove_many(&state->arcs_esize[type],
arc_buf_size(buf), buf);
}
}
/*
* Add a reference to this hdr indicating that someone is actively
* referencing that memory. When the refcount transitions from 0 to 1,
* we remove it from the respective arc_state_t list to indicate that
* it is not evictable.
*/
static void
add_reference(arc_buf_hdr_t *hdr, void *tag)
{
arc_state_t *state;
ASSERT(HDR_HAS_L1HDR(hdr));
if (!HDR_EMPTY(hdr) && !MUTEX_HELD(HDR_LOCK(hdr))) {
ASSERT(hdr->b_l1hdr.b_state == arc_anon);
ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
}
state = hdr->b_l1hdr.b_state;
if ((zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, tag) == 1) &&
(state != arc_anon)) {
/* We don't use the L2-only state list. */
if (state != arc_l2c_only) {
multilist_remove(&state->arcs_list[arc_buf_type(hdr)],
hdr);
arc_evictable_space_decrement(hdr, state);
}
/* remove the prefetch flag if we get a reference */
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_decrement_state(hdr);
arc_hdr_clear_flags(hdr, ARC_FLAG_PREFETCH);
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_increment_state(hdr);
}
}
/*
* Remove a reference from this hdr. When the reference transitions from
* 1 to 0 and we're not anonymous, then we add this hdr to the arc_state_t's
* list making it eligible for eviction.
*/
static int
remove_reference(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, void *tag)
{
int cnt;
arc_state_t *state = hdr->b_l1hdr.b_state;
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT(state == arc_anon || MUTEX_HELD(hash_lock));
ASSERT(!GHOST_STATE(state));
/*
* arc_l2c_only counts as a ghost state so we don't need to explicitly
* check to prevent usage of the arc_l2c_only list.
*/
if (((cnt = zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, tag)) == 0) &&
(state != arc_anon)) {
multilist_insert(&state->arcs_list[arc_buf_type(hdr)], hdr);
ASSERT3U(hdr->b_l1hdr.b_bufcnt, >, 0);
arc_evictable_space_increment(hdr, state);
}
return (cnt);
}
/*
* Returns detailed information about a specific arc buffer. When the
* state_index argument is set the function will calculate the arc header
* list position for its arc state. Since this requires a linear traversal
* callers are strongly encourage not to do this. However, it can be helpful
* for targeted analysis so the functionality is provided.
*/
void
arc_buf_info(arc_buf_t *ab, arc_buf_info_t *abi, int state_index)
{
(void) state_index;
arc_buf_hdr_t *hdr = ab->b_hdr;
l1arc_buf_hdr_t *l1hdr = NULL;
l2arc_buf_hdr_t *l2hdr = NULL;
arc_state_t *state = NULL;
memset(abi, 0, sizeof (arc_buf_info_t));
if (hdr == NULL)
return;
abi->abi_flags = hdr->b_flags;
if (HDR_HAS_L1HDR(hdr)) {
l1hdr = &hdr->b_l1hdr;
state = l1hdr->b_state;
}
if (HDR_HAS_L2HDR(hdr))
l2hdr = &hdr->b_l2hdr;
if (l1hdr) {
abi->abi_bufcnt = l1hdr->b_bufcnt;
abi->abi_access = l1hdr->b_arc_access;
abi->abi_mru_hits = l1hdr->b_mru_hits;
abi->abi_mru_ghost_hits = l1hdr->b_mru_ghost_hits;
abi->abi_mfu_hits = l1hdr->b_mfu_hits;
abi->abi_mfu_ghost_hits = l1hdr->b_mfu_ghost_hits;
abi->abi_holds = zfs_refcount_count(&l1hdr->b_refcnt);
}
if (l2hdr) {
abi->abi_l2arc_dattr = l2hdr->b_daddr;
abi->abi_l2arc_hits = l2hdr->b_hits;
}
abi->abi_state_type = state ? state->arcs_state : ARC_STATE_ANON;
abi->abi_state_contents = arc_buf_type(hdr);
abi->abi_size = arc_hdr_size(hdr);
}
/*
* Move the supplied buffer to the indicated state. The hash lock
* for the buffer must be held by the caller.
*/
static void
arc_change_state(arc_state_t *new_state, arc_buf_hdr_t *hdr,
kmutex_t *hash_lock)
{
arc_state_t *old_state;
int64_t refcnt;
uint32_t bufcnt;
boolean_t update_old, update_new;
arc_buf_contents_t buftype = arc_buf_type(hdr);
/*
* We almost always have an L1 hdr here, since we call arc_hdr_realloc()
* in arc_read() when bringing a buffer out of the L2ARC. However, the
* L1 hdr doesn't always exist when we change state to arc_anon before
* destroying a header, in which case reallocating to add the L1 hdr is
* pointless.
*/
if (HDR_HAS_L1HDR(hdr)) {
old_state = hdr->b_l1hdr.b_state;
refcnt = zfs_refcount_count(&hdr->b_l1hdr.b_refcnt);
bufcnt = hdr->b_l1hdr.b_bufcnt;
update_old = (bufcnt > 0 || hdr->b_l1hdr.b_pabd != NULL ||
HDR_HAS_RABD(hdr));
} else {
old_state = arc_l2c_only;
refcnt = 0;
bufcnt = 0;
update_old = B_FALSE;
}
update_new = update_old;
ASSERT(MUTEX_HELD(hash_lock));
ASSERT3P(new_state, !=, old_state);
ASSERT(!GHOST_STATE(new_state) || bufcnt == 0);
ASSERT(old_state != arc_anon || bufcnt <= 1);
/*
* If this buffer is evictable, transfer it from the
* old state list to the new state list.
*/
if (refcnt == 0) {
if (old_state != arc_anon && old_state != arc_l2c_only) {
ASSERT(HDR_HAS_L1HDR(hdr));
multilist_remove(&old_state->arcs_list[buftype], hdr);
if (GHOST_STATE(old_state)) {
ASSERT0(bufcnt);
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
update_old = B_TRUE;
}
arc_evictable_space_decrement(hdr, old_state);
}
if (new_state != arc_anon && new_state != arc_l2c_only) {
/*
* An L1 header always exists here, since if we're
* moving to some L1-cached state (i.e. not l2c_only or
* anonymous), we realloc the header to add an L1hdr
* beforehand.
*/
ASSERT(HDR_HAS_L1HDR(hdr));
multilist_insert(&new_state->arcs_list[buftype], hdr);
if (GHOST_STATE(new_state)) {
ASSERT0(bufcnt);
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
update_new = B_TRUE;
}
arc_evictable_space_increment(hdr, new_state);
}
}
ASSERT(!HDR_EMPTY(hdr));
if (new_state == arc_anon && HDR_IN_HASH_TABLE(hdr))
buf_hash_remove(hdr);
/* adjust state sizes (ignore arc_l2c_only) */
if (update_new && new_state != arc_l2c_only) {
ASSERT(HDR_HAS_L1HDR(hdr));
if (GHOST_STATE(new_state)) {
ASSERT0(bufcnt);
/*
* When moving a header to a ghost state, we first
* remove all arc buffers. Thus, we'll have a
* bufcnt of zero, and no arc buffer to use for
* the reference. As a result, we use the arc
* header pointer for the reference.
*/
(void) zfs_refcount_add_many(&new_state->arcs_size,
HDR_GET_LSIZE(hdr), hdr);
ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
ASSERT(!HDR_HAS_RABD(hdr));
} else {
uint32_t buffers = 0;
/*
* Each individual buffer holds a unique reference,
* thus we must remove each of these references one
* at a time.
*/
for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
buf = buf->b_next) {
ASSERT3U(bufcnt, !=, 0);
buffers++;
/*
* When the arc_buf_t is sharing the data
* block with the hdr, the owner of the
* reference belongs to the hdr. Only
* add to the refcount if the arc_buf_t is
* not shared.
*/
if (arc_buf_is_shared(buf))
continue;
(void) zfs_refcount_add_many(
&new_state->arcs_size,
arc_buf_size(buf), buf);
}
ASSERT3U(bufcnt, ==, buffers);
if (hdr->b_l1hdr.b_pabd != NULL) {
(void) zfs_refcount_add_many(
&new_state->arcs_size,
arc_hdr_size(hdr), hdr);
}
if (HDR_HAS_RABD(hdr)) {
(void) zfs_refcount_add_many(
&new_state->arcs_size,
HDR_GET_PSIZE(hdr), hdr);
}
}
}
if (update_old && old_state != arc_l2c_only) {
ASSERT(HDR_HAS_L1HDR(hdr));
if (GHOST_STATE(old_state)) {
ASSERT0(bufcnt);
ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
ASSERT(!HDR_HAS_RABD(hdr));
/*
* When moving a header off of a ghost state,
* the header will not contain any arc buffers.
* We use the arc header pointer for the reference
* which is exactly what we did when we put the
* header on the ghost state.
*/
(void) zfs_refcount_remove_many(&old_state->arcs_size,
HDR_GET_LSIZE(hdr), hdr);
} else {
uint32_t buffers = 0;
/*
* Each individual buffer holds a unique reference,
* thus we must remove each of these references one
* at a time.
*/
for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
buf = buf->b_next) {
ASSERT3U(bufcnt, !=, 0);
buffers++;
/*
* When the arc_buf_t is sharing the data
* block with the hdr, the owner of the
* reference belongs to the hdr. Only
* add to the refcount if the arc_buf_t is
* not shared.
*/
if (arc_buf_is_shared(buf))
continue;
(void) zfs_refcount_remove_many(
&old_state->arcs_size, arc_buf_size(buf),
buf);
}
ASSERT3U(bufcnt, ==, buffers);
ASSERT(hdr->b_l1hdr.b_pabd != NULL ||
HDR_HAS_RABD(hdr));
if (hdr->b_l1hdr.b_pabd != NULL) {
(void) zfs_refcount_remove_many(
&old_state->arcs_size, arc_hdr_size(hdr),
hdr);
}
if (HDR_HAS_RABD(hdr)) {
(void) zfs_refcount_remove_many(
&old_state->arcs_size, HDR_GET_PSIZE(hdr),
hdr);
}
}
}
if (HDR_HAS_L1HDR(hdr)) {
hdr->b_l1hdr.b_state = new_state;
if (HDR_HAS_L2HDR(hdr) && new_state != arc_l2c_only) {
l2arc_hdr_arcstats_decrement_state(hdr);
hdr->b_l2hdr.b_arcs_state = new_state->arcs_state;
l2arc_hdr_arcstats_increment_state(hdr);
}
}
}
void
arc_space_consume(uint64_t space, arc_space_type_t type)
{
ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES);
switch (type) {
default:
break;
case ARC_SPACE_DATA:
ARCSTAT_INCR(arcstat_data_size, space);
break;
case ARC_SPACE_META:
ARCSTAT_INCR(arcstat_metadata_size, space);
break;
case ARC_SPACE_BONUS:
ARCSTAT_INCR(arcstat_bonus_size, space);
break;
case ARC_SPACE_DNODE:
aggsum_add(&arc_sums.arcstat_dnode_size, space);
break;
case ARC_SPACE_DBUF:
ARCSTAT_INCR(arcstat_dbuf_size, space);
break;
case ARC_SPACE_HDRS:
ARCSTAT_INCR(arcstat_hdr_size, space);
break;
case ARC_SPACE_L2HDRS:
aggsum_add(&arc_sums.arcstat_l2_hdr_size, space);
break;
case ARC_SPACE_ABD_CHUNK_WASTE:
/*
* Note: this includes space wasted by all scatter ABD's, not
* just those allocated by the ARC. But the vast majority of
* scatter ABD's come from the ARC, because other users are
* very short-lived.
*/
ARCSTAT_INCR(arcstat_abd_chunk_waste_size, space);
break;
}
if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE)
aggsum_add(&arc_sums.arcstat_meta_used, space);
aggsum_add(&arc_sums.arcstat_size, space);
}
void
arc_space_return(uint64_t space, arc_space_type_t type)
{
ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES);
switch (type) {
default:
break;
case ARC_SPACE_DATA:
ARCSTAT_INCR(arcstat_data_size, -space);
break;
case ARC_SPACE_META:
ARCSTAT_INCR(arcstat_metadata_size, -space);
break;
case ARC_SPACE_BONUS:
ARCSTAT_INCR(arcstat_bonus_size, -space);
break;
case ARC_SPACE_DNODE:
aggsum_add(&arc_sums.arcstat_dnode_size, -space);
break;
case ARC_SPACE_DBUF:
ARCSTAT_INCR(arcstat_dbuf_size, -space);
break;
case ARC_SPACE_HDRS:
ARCSTAT_INCR(arcstat_hdr_size, -space);
break;
case ARC_SPACE_L2HDRS:
aggsum_add(&arc_sums.arcstat_l2_hdr_size, -space);
break;
case ARC_SPACE_ABD_CHUNK_WASTE:
ARCSTAT_INCR(arcstat_abd_chunk_waste_size, -space);
break;
}
if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE) {
ASSERT(aggsum_compare(&arc_sums.arcstat_meta_used,
space) >= 0);
ARCSTAT_MAX(arcstat_meta_max,
aggsum_upper_bound(&arc_sums.arcstat_meta_used));
aggsum_add(&arc_sums.arcstat_meta_used, -space);
}
ASSERT(aggsum_compare(&arc_sums.arcstat_size, space) >= 0);
aggsum_add(&arc_sums.arcstat_size, -space);
}
/*
* Given a hdr and a buf, returns whether that buf can share its b_data buffer
* with the hdr's b_pabd.
*/
static boolean_t
arc_can_share(arc_buf_hdr_t *hdr, arc_buf_t *buf)
{
/*
* The criteria for sharing a hdr's data are:
* 1. the buffer is not encrypted
* 2. the hdr's compression matches the buf's compression
* 3. the hdr doesn't need to be byteswapped
* 4. the hdr isn't already being shared
* 5. the buf is either compressed or it is the last buf in the hdr list
*
* Criterion #5 maintains the invariant that shared uncompressed
* bufs must be the final buf in the hdr's b_buf list. Reading this, you
* might ask, "if a compressed buf is allocated first, won't that be the
* last thing in the list?", but in that case it's impossible to create
* a shared uncompressed buf anyway (because the hdr must be compressed
* to have the compressed buf). You might also think that #3 is
* sufficient to make this guarantee, however it's possible
* (specifically in the rare L2ARC write race mentioned in
* arc_buf_alloc_impl()) there will be an existing uncompressed buf that
* is shareable, but wasn't at the time of its allocation. Rather than
* allow a new shared uncompressed buf to be created and then shuffle
* the list around to make it the last element, this simply disallows
* sharing if the new buf isn't the first to be added.
*/
ASSERT3P(buf->b_hdr, ==, hdr);
boolean_t hdr_compressed =
arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF;
boolean_t buf_compressed = ARC_BUF_COMPRESSED(buf) != 0;
return (!ARC_BUF_ENCRYPTED(buf) &&
buf_compressed == hdr_compressed &&
hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS &&
!HDR_SHARED_DATA(hdr) &&
(ARC_BUF_LAST(buf) || ARC_BUF_COMPRESSED(buf)));
}
/*
* Allocate a buf for this hdr. If you care about the data that's in the hdr,
* or if you want a compressed buffer, pass those flags in. Returns 0 if the
* copy was made successfully, or an error code otherwise.
*/
static int
arc_buf_alloc_impl(arc_buf_hdr_t *hdr, spa_t *spa, const zbookmark_phys_t *zb,
void *tag, boolean_t encrypted, boolean_t compressed, boolean_t noauth,
boolean_t fill, arc_buf_t **ret)
{
arc_buf_t *buf;
arc_fill_flags_t flags = ARC_FILL_LOCKED;
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT3U(HDR_GET_LSIZE(hdr), >, 0);
VERIFY(hdr->b_type == ARC_BUFC_DATA ||
hdr->b_type == ARC_BUFC_METADATA);
ASSERT3P(ret, !=, NULL);
ASSERT3P(*ret, ==, NULL);
IMPLY(encrypted, compressed);
buf = *ret = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
buf->b_hdr = hdr;
buf->b_data = NULL;
buf->b_next = hdr->b_l1hdr.b_buf;
buf->b_flags = 0;
add_reference(hdr, tag);
/*
* We're about to change the hdr's b_flags. We must either
* hold the hash_lock or be undiscoverable.
*/
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
/*
* Only honor requests for compressed bufs if the hdr is actually
* compressed. This must be overridden if the buffer is encrypted since
* encrypted buffers cannot be decompressed.
*/
if (encrypted) {
buf->b_flags |= ARC_BUF_FLAG_COMPRESSED;
buf->b_flags |= ARC_BUF_FLAG_ENCRYPTED;
flags |= ARC_FILL_COMPRESSED | ARC_FILL_ENCRYPTED;
} else if (compressed &&
arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) {
buf->b_flags |= ARC_BUF_FLAG_COMPRESSED;
flags |= ARC_FILL_COMPRESSED;
}
if (noauth) {
ASSERT0(encrypted);
flags |= ARC_FILL_NOAUTH;
}
/*
* If the hdr's data can be shared then we share the data buffer and
* set the appropriate bit in the hdr's b_flags to indicate the hdr is
* sharing it's b_pabd with the arc_buf_t. Otherwise, we allocate a new
* buffer to store the buf's data.
*
* There are two additional restrictions here because we're sharing
* hdr -> buf instead of the usual buf -> hdr. First, the hdr can't be
* actively involved in an L2ARC write, because if this buf is used by
* an arc_write() then the hdr's data buffer will be released when the
* write completes, even though the L2ARC write might still be using it.
* Second, the hdr's ABD must be linear so that the buf's user doesn't
* need to be ABD-aware. It must be allocated via
* zio_[data_]buf_alloc(), not as a page, because we need to be able
* to abd_release_ownership_of_buf(), which isn't allowed on "linear
* page" buffers because the ABD code needs to handle freeing them
* specially.
*/
boolean_t can_share = arc_can_share(hdr, buf) &&
!HDR_L2_WRITING(hdr) &&
hdr->b_l1hdr.b_pabd != NULL &&
abd_is_linear(hdr->b_l1hdr.b_pabd) &&
!abd_is_linear_page(hdr->b_l1hdr.b_pabd);
/* Set up b_data and sharing */
if (can_share) {
buf->b_data = abd_to_buf(hdr->b_l1hdr.b_pabd);
buf->b_flags |= ARC_BUF_FLAG_SHARED;
arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA);
} else {
buf->b_data =
arc_get_data_buf(hdr, arc_buf_size(buf), buf);
ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf));
}
VERIFY3P(buf->b_data, !=, NULL);
hdr->b_l1hdr.b_buf = buf;
hdr->b_l1hdr.b_bufcnt += 1;
if (encrypted)
hdr->b_crypt_hdr.b_ebufcnt += 1;
/*
* If the user wants the data from the hdr, we need to either copy or
* decompress the data.
*/
if (fill) {
ASSERT3P(zb, !=, NULL);
return (arc_buf_fill(buf, spa, zb, flags));
}
return (0);
}
static char *arc_onloan_tag = "onloan";
static inline void
arc_loaned_bytes_update(int64_t delta)
{
atomic_add_64(&arc_loaned_bytes, delta);
/* assert that it did not wrap around */
ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0);
}
/*
* Loan out an anonymous arc buffer. Loaned buffers are not counted as in
* flight data by arc_tempreserve_space() until they are "returned". Loaned
* buffers must be returned to the arc before they can be used by the DMU or
* freed.
*/
arc_buf_t *
arc_loan_buf(spa_t *spa, boolean_t is_metadata, int size)
{
arc_buf_t *buf = arc_alloc_buf(spa, arc_onloan_tag,
is_metadata ? ARC_BUFC_METADATA : ARC_BUFC_DATA, size);
arc_loaned_bytes_update(arc_buf_size(buf));
return (buf);
}
arc_buf_t *
arc_loan_compressed_buf(spa_t *spa, uint64_t psize, uint64_t lsize,
enum zio_compress compression_type, uint8_t complevel)
{
arc_buf_t *buf = arc_alloc_compressed_buf(spa, arc_onloan_tag,
psize, lsize, compression_type, complevel);
arc_loaned_bytes_update(arc_buf_size(buf));
return (buf);
}
arc_buf_t *
arc_loan_raw_buf(spa_t *spa, uint64_t dsobj, boolean_t byteorder,
const uint8_t *salt, const uint8_t *iv, const uint8_t *mac,
dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
enum zio_compress compression_type, uint8_t complevel)
{
arc_buf_t *buf = arc_alloc_raw_buf(spa, arc_onloan_tag, dsobj,
byteorder, salt, iv, mac, ot, psize, lsize, compression_type,
complevel);
atomic_add_64(&arc_loaned_bytes, psize);
return (buf);
}
/*
* Return a loaned arc buffer to the arc.
*/
void
arc_return_buf(arc_buf_t *buf, void *tag)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
ASSERT3P(buf->b_data, !=, NULL);
ASSERT(HDR_HAS_L1HDR(hdr));
(void) zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, tag);
(void) zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag);
arc_loaned_bytes_update(-arc_buf_size(buf));
}
/* Detach an arc_buf from a dbuf (tag) */
void
arc_loan_inuse_buf(arc_buf_t *buf, void *tag)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
ASSERT3P(buf->b_data, !=, NULL);
ASSERT(HDR_HAS_L1HDR(hdr));
(void) zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag);
(void) zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, tag);
arc_loaned_bytes_update(arc_buf_size(buf));
}
static void
l2arc_free_abd_on_write(abd_t *abd, size_t size, arc_buf_contents_t type)
{
l2arc_data_free_t *df = kmem_alloc(sizeof (*df), KM_SLEEP);
df->l2df_abd = abd;
df->l2df_size = size;
df->l2df_type = type;
mutex_enter(&l2arc_free_on_write_mtx);
list_insert_head(l2arc_free_on_write, df);
mutex_exit(&l2arc_free_on_write_mtx);
}
static void
arc_hdr_free_on_write(arc_buf_hdr_t *hdr, boolean_t free_rdata)
{
arc_state_t *state = hdr->b_l1hdr.b_state;
arc_buf_contents_t type = arc_buf_type(hdr);
uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr);
/* protected by hash lock, if in the hash table */
if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
ASSERT(state != arc_anon && state != arc_l2c_only);
(void) zfs_refcount_remove_many(&state->arcs_esize[type],
size, hdr);
}
(void) zfs_refcount_remove_many(&state->arcs_size, size, hdr);
if (type == ARC_BUFC_METADATA) {
arc_space_return(size, ARC_SPACE_META);
} else {
ASSERT(type == ARC_BUFC_DATA);
arc_space_return(size, ARC_SPACE_DATA);
}
if (free_rdata) {
l2arc_free_abd_on_write(hdr->b_crypt_hdr.b_rabd, size, type);
} else {
l2arc_free_abd_on_write(hdr->b_l1hdr.b_pabd, size, type);
}
}
/*
* Share the arc_buf_t's data with the hdr. Whenever we are sharing the
* data buffer, we transfer the refcount ownership to the hdr and update
* the appropriate kstats.
*/
static void
arc_share_buf(arc_buf_hdr_t *hdr, arc_buf_t *buf)
{
ASSERT(arc_can_share(hdr, buf));
ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
ASSERT(!ARC_BUF_ENCRYPTED(buf));
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
/*
* Start sharing the data buffer. We transfer the
* refcount ownership to the hdr since it always owns
* the refcount whenever an arc_buf_t is shared.
*/
zfs_refcount_transfer_ownership_many(&hdr->b_l1hdr.b_state->arcs_size,
arc_hdr_size(hdr), buf, hdr);
hdr->b_l1hdr.b_pabd = abd_get_from_buf(buf->b_data, arc_buf_size(buf));
abd_take_ownership_of_buf(hdr->b_l1hdr.b_pabd,
HDR_ISTYPE_METADATA(hdr));
arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA);
buf->b_flags |= ARC_BUF_FLAG_SHARED;
/*
* Since we've transferred ownership to the hdr we need
* to increment its compressed and uncompressed kstats and
* decrement the overhead size.
*/
ARCSTAT_INCR(arcstat_compressed_size, arc_hdr_size(hdr));
ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr));
ARCSTAT_INCR(arcstat_overhead_size, -arc_buf_size(buf));
}
static void
arc_unshare_buf(arc_buf_hdr_t *hdr, arc_buf_t *buf)
{
ASSERT(arc_buf_is_shared(buf));
ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
/*
* We are no longer sharing this buffer so we need
* to transfer its ownership to the rightful owner.
*/
zfs_refcount_transfer_ownership_many(&hdr->b_l1hdr.b_state->arcs_size,
arc_hdr_size(hdr), hdr, buf);
arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
abd_release_ownership_of_buf(hdr->b_l1hdr.b_pabd);
abd_free(hdr->b_l1hdr.b_pabd);
hdr->b_l1hdr.b_pabd = NULL;
buf->b_flags &= ~ARC_BUF_FLAG_SHARED;
/*
* Since the buffer is no longer shared between
* the arc buf and the hdr, count it as overhead.
*/
ARCSTAT_INCR(arcstat_compressed_size, -arc_hdr_size(hdr));
ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr));
ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf));
}
/*
* Remove an arc_buf_t from the hdr's buf list and return the last
* arc_buf_t on the list. If no buffers remain on the list then return
* NULL.
*/
static arc_buf_t *
arc_buf_remove(arc_buf_hdr_t *hdr, arc_buf_t *buf)
{
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
arc_buf_t **bufp = &hdr->b_l1hdr.b_buf;
arc_buf_t *lastbuf = NULL;
/*
* Remove the buf from the hdr list and locate the last
* remaining buffer on the list.
*/
while (*bufp != NULL) {
if (*bufp == buf)
*bufp = buf->b_next;
/*
* If we've removed a buffer in the middle of
* the list then update the lastbuf and update
* bufp.
*/
if (*bufp != NULL) {
lastbuf = *bufp;
bufp = &(*bufp)->b_next;
}
}
buf->b_next = NULL;
ASSERT3P(lastbuf, !=, buf);
IMPLY(hdr->b_l1hdr.b_bufcnt > 0, lastbuf != NULL);
IMPLY(hdr->b_l1hdr.b_bufcnt > 0, hdr->b_l1hdr.b_buf != NULL);
IMPLY(lastbuf != NULL, ARC_BUF_LAST(lastbuf));
return (lastbuf);
}
/*
* Free up buf->b_data and pull the arc_buf_t off of the arc_buf_hdr_t's
* list and free it.
*/
static void
arc_buf_destroy_impl(arc_buf_t *buf)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
/*
* Free up the data associated with the buf but only if we're not
* sharing this with the hdr. If we are sharing it with the hdr, the
* hdr is responsible for doing the free.
*/
if (buf->b_data != NULL) {
/*
* We're about to change the hdr's b_flags. We must either
* hold the hash_lock or be undiscoverable.
*/
ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
arc_cksum_verify(buf);
arc_buf_unwatch(buf);
if (arc_buf_is_shared(buf)) {
arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
} else {
uint64_t size = arc_buf_size(buf);
arc_free_data_buf(hdr, buf->b_data, size, buf);
ARCSTAT_INCR(arcstat_overhead_size, -size);
}
buf->b_data = NULL;
ASSERT(hdr->b_l1hdr.b_bufcnt > 0);
hdr->b_l1hdr.b_bufcnt -= 1;
if (ARC_BUF_ENCRYPTED(buf)) {
hdr->b_crypt_hdr.b_ebufcnt -= 1;
/*
* If we have no more encrypted buffers and we've
* already gotten a copy of the decrypted data we can
* free b_rabd to save some space.
*/
if (hdr->b_crypt_hdr.b_ebufcnt == 0 &&
HDR_HAS_RABD(hdr) && hdr->b_l1hdr.b_pabd != NULL &&
!HDR_IO_IN_PROGRESS(hdr)) {
arc_hdr_free_abd(hdr, B_TRUE);
}
}
}
arc_buf_t *lastbuf = arc_buf_remove(hdr, buf);
if (ARC_BUF_SHARED(buf) && !ARC_BUF_COMPRESSED(buf)) {
/*
* If the current arc_buf_t is sharing its data buffer with the
* hdr, then reassign the hdr's b_pabd to share it with the new
* buffer at the end of the list. The shared buffer is always
* the last one on the hdr's buffer list.
*
* There is an equivalent case for compressed bufs, but since
* they aren't guaranteed to be the last buf in the list and
* that is an exceedingly rare case, we just allow that space be
* wasted temporarily. We must also be careful not to share
* encrypted buffers, since they cannot be shared.
*/
if (lastbuf != NULL && !ARC_BUF_ENCRYPTED(lastbuf)) {
/* Only one buf can be shared at once */
VERIFY(!arc_buf_is_shared(lastbuf));
/* hdr is uncompressed so can't have compressed buf */
VERIFY(!ARC_BUF_COMPRESSED(lastbuf));
ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
arc_hdr_free_abd(hdr, B_FALSE);
/*
* We must setup a new shared block between the
* last buffer and the hdr. The data would have
* been allocated by the arc buf so we need to transfer
* ownership to the hdr since it's now being shared.
*/
arc_share_buf(hdr, lastbuf);
}
} else if (HDR_SHARED_DATA(hdr)) {
/*
* Uncompressed shared buffers are always at the end
* of the list. Compressed buffers don't have the
* same requirements. This makes it hard to
* simply assert that the lastbuf is shared so
* we rely on the hdr's compression flags to determine
* if we have a compressed, shared buffer.
*/
ASSERT3P(lastbuf, !=, NULL);
ASSERT(arc_buf_is_shared(lastbuf) ||
arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF);
}
/*
* Free the checksum if we're removing the last uncompressed buf from
* this hdr.
*/
if (!arc_hdr_has_uncompressed_buf(hdr)) {
arc_cksum_free(hdr);
}
/* clean up the buf */
buf->b_hdr = NULL;
kmem_cache_free(buf_cache, buf);
}
static void
arc_hdr_alloc_abd(arc_buf_hdr_t *hdr, int alloc_flags)
{
uint64_t size;
boolean_t alloc_rdata = ((alloc_flags & ARC_HDR_ALLOC_RDATA) != 0);
ASSERT3U(HDR_GET_LSIZE(hdr), >, 0);
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT(!HDR_SHARED_DATA(hdr) || alloc_rdata);
IMPLY(alloc_rdata, HDR_PROTECTED(hdr));
if (alloc_rdata) {
size = HDR_GET_PSIZE(hdr);
ASSERT3P(hdr->b_crypt_hdr.b_rabd, ==, NULL);
hdr->b_crypt_hdr.b_rabd = arc_get_data_abd(hdr, size, hdr,
alloc_flags);
ASSERT3P(hdr->b_crypt_hdr.b_rabd, !=, NULL);
ARCSTAT_INCR(arcstat_raw_size, size);
} else {
size = arc_hdr_size(hdr);
ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
hdr->b_l1hdr.b_pabd = arc_get_data_abd(hdr, size, hdr,
alloc_flags);
ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
}
ARCSTAT_INCR(arcstat_compressed_size, size);
ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr));
}
static void
arc_hdr_free_abd(arc_buf_hdr_t *hdr, boolean_t free_rdata)
{
uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr);
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr));
IMPLY(free_rdata, HDR_HAS_RABD(hdr));
/*
* If the hdr is currently being written to the l2arc then
* we defer freeing the data by adding it to the l2arc_free_on_write
* list. The l2arc will free the data once it's finished
* writing it to the l2arc device.
*/
if (HDR_L2_WRITING(hdr)) {
arc_hdr_free_on_write(hdr, free_rdata);
ARCSTAT_BUMP(arcstat_l2_free_on_write);
} else if (free_rdata) {
arc_free_data_abd(hdr, hdr->b_crypt_hdr.b_rabd, size, hdr);
} else {
arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, size, hdr);
}
if (free_rdata) {
hdr->b_crypt_hdr.b_rabd = NULL;
ARCSTAT_INCR(arcstat_raw_size, -size);
} else {
hdr->b_l1hdr.b_pabd = NULL;
}
if (hdr->b_l1hdr.b_pabd == NULL && !HDR_HAS_RABD(hdr))
hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
ARCSTAT_INCR(arcstat_compressed_size, -size);
ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr));
}
/*
* Allocate empty anonymous ARC header. The header will get its identity
* assigned and buffers attached later as part of read or write operations.
*
* In case of read arc_read() assigns header its identify (b_dva + b_birth),
* inserts it into ARC hash to become globally visible and allocates physical
* (b_pabd) or raw (b_rabd) ABD buffer to read into from disk. On disk read
* completion arc_read_done() allocates ARC buffer(s) as needed, potentially
* sharing one of them with the physical ABD buffer.
*
* In case of write arc_alloc_buf() allocates ARC buffer to be filled with
* data. Then after compression and/or encryption arc_write_ready() allocates
* and fills (or potentially shares) physical (b_pabd) or raw (b_rabd) ABD
* buffer. On disk write completion arc_write_done() assigns the header its
* new identity (b_dva + b_birth) and inserts into ARC hash.
*
* In case of partial overwrite the old data is read first as described. Then
* arc_release() either allocates new anonymous ARC header and moves the ARC
* buffer to it, or reuses the old ARC header by discarding its identity and
* removing it from ARC hash. After buffer modification normal write process
* follows as described.
*/
static arc_buf_hdr_t *
arc_hdr_alloc(uint64_t spa, int32_t psize, int32_t lsize,
boolean_t protected, enum zio_compress compression_type, uint8_t complevel,
arc_buf_contents_t type)
{
arc_buf_hdr_t *hdr;
VERIFY(type == ARC_BUFC_DATA || type == ARC_BUFC_METADATA);
if (protected) {
hdr = kmem_cache_alloc(hdr_full_crypt_cache, KM_PUSHPAGE);
} else {
hdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE);
}
ASSERT(HDR_EMPTY(hdr));
ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
HDR_SET_PSIZE(hdr, psize);
HDR_SET_LSIZE(hdr, lsize);
hdr->b_spa = spa;
hdr->b_type = type;
hdr->b_flags = 0;
arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) | ARC_FLAG_HAS_L1HDR);
arc_hdr_set_compress(hdr, compression_type);
hdr->b_complevel = complevel;
if (protected)
arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);
hdr->b_l1hdr.b_state = arc_anon;
hdr->b_l1hdr.b_arc_access = 0;
hdr->b_l1hdr.b_mru_hits = 0;
hdr->b_l1hdr.b_mru_ghost_hits = 0;
hdr->b_l1hdr.b_mfu_hits = 0;
hdr->b_l1hdr.b_mfu_ghost_hits = 0;
hdr->b_l1hdr.b_bufcnt = 0;
hdr->b_l1hdr.b_buf = NULL;
ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
return (hdr);
}
/*
* Transition between the two allocation states for the arc_buf_hdr struct.
* The arc_buf_hdr struct can be allocated with (hdr_full_cache) or without
* (hdr_l2only_cache) the fields necessary for the L1 cache - the smaller
* version is used when a cache buffer is only in the L2ARC in order to reduce
* memory usage.
*/
static arc_buf_hdr_t *
arc_hdr_realloc(arc_buf_hdr_t *hdr, kmem_cache_t *old, kmem_cache_t *new)
{
ASSERT(HDR_HAS_L2HDR(hdr));
arc_buf_hdr_t *nhdr;
l2arc_dev_t *dev = hdr->b_l2hdr.b_dev;
ASSERT((old == hdr_full_cache && new == hdr_l2only_cache) ||
(old == hdr_l2only_cache && new == hdr_full_cache));
/*
* if the caller wanted a new full header and the header is to be
* encrypted we will actually allocate the header from the full crypt
* cache instead. The same applies to freeing from the old cache.
*/
if (HDR_PROTECTED(hdr) && new == hdr_full_cache)
new = hdr_full_crypt_cache;
if (HDR_PROTECTED(hdr) && old == hdr_full_cache)
old = hdr_full_crypt_cache;
nhdr = kmem_cache_alloc(new, KM_PUSHPAGE);
ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
buf_hash_remove(hdr);
bcopy(hdr, nhdr, HDR_L2ONLY_SIZE);
if (new == hdr_full_cache || new == hdr_full_crypt_cache) {
arc_hdr_set_flags(nhdr, ARC_FLAG_HAS_L1HDR);
/*
* arc_access and arc_change_state need to be aware that a
* header has just come out of L2ARC, so we set its state to
* l2c_only even though it's about to change.
*/
nhdr->b_l1hdr.b_state = arc_l2c_only;
/* Verify previous threads set to NULL before freeing */
ASSERT3P(nhdr->b_l1hdr.b_pabd, ==, NULL);
ASSERT(!HDR_HAS_RABD(hdr));
} else {
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
ASSERT0(hdr->b_l1hdr.b_bufcnt);
ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
/*
* If we've reached here, We must have been called from
* arc_evict_hdr(), as such we should have already been
* removed from any ghost list we were previously on
* (which protects us from racing with arc_evict_state),
* thus no locking is needed during this check.
*/
ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
/*
* A buffer must not be moved into the arc_l2c_only
* state if it's not finished being written out to the
* l2arc device. Otherwise, the b_l1hdr.b_pabd field
* might try to be accessed, even though it was removed.
*/
VERIFY(!HDR_L2_WRITING(hdr));
VERIFY3P(hdr->b_l1hdr.b_pabd, ==, NULL);
ASSERT(!HDR_HAS_RABD(hdr));
arc_hdr_clear_flags(nhdr, ARC_FLAG_HAS_L1HDR);
}
/*
* The header has been reallocated so we need to re-insert it into any
* lists it was on.
*/
(void) buf_hash_insert(nhdr, NULL);
ASSERT(list_link_active(&hdr->b_l2hdr.b_l2node));
mutex_enter(&dev->l2ad_mtx);
/*
* We must place the realloc'ed header back into the list at
* the same spot. Otherwise, if it's placed earlier in the list,
* l2arc_write_buffers() could find it during the function's
* write phase, and try to write it out to the l2arc.
*/
list_insert_after(&dev->l2ad_buflist, hdr, nhdr);
list_remove(&dev->l2ad_buflist, hdr);
mutex_exit(&dev->l2ad_mtx);
/*
* Since we're using the pointer address as the tag when
* incrementing and decrementing the l2ad_alloc refcount, we
* must remove the old pointer (that we're about to destroy) and
* add the new pointer to the refcount. Otherwise we'd remove
* the wrong pointer address when calling arc_hdr_destroy() later.
*/
(void) zfs_refcount_remove_many(&dev->l2ad_alloc,
arc_hdr_size(hdr), hdr);
(void) zfs_refcount_add_many(&dev->l2ad_alloc,
arc_hdr_size(nhdr), nhdr);
buf_discard_identity(hdr);
kmem_cache_free(old, hdr);
return (nhdr);
}
/*
* This function allows an L1 header to be reallocated as a crypt
* header and vice versa. If we are going to a crypt header, the
* new fields will be zeroed out.
*/
static arc_buf_hdr_t *
arc_hdr_realloc_crypt(arc_buf_hdr_t *hdr, boolean_t need_crypt)
{
arc_buf_hdr_t *nhdr;
arc_buf_t *buf;
kmem_cache_t *ncache, *ocache;
/*
* This function requires that hdr is in the arc_anon state.
* Therefore it won't have any L2ARC data for us to worry
* about copying.
*/
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT(!HDR_HAS_L2HDR(hdr));
ASSERT3U(!!HDR_PROTECTED(hdr), !=, need_crypt);
ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
ASSERT(!list_link_active(&hdr->b_l2hdr.b_l2node));
ASSERT3P(hdr->b_hash_next, ==, NULL);
if (need_crypt) {
ncache = hdr_full_crypt_cache;
ocache = hdr_full_cache;
} else {
ncache = hdr_full_cache;
ocache = hdr_full_crypt_cache;
}
nhdr = kmem_cache_alloc(ncache, KM_PUSHPAGE);
/*
* Copy all members that aren't locks or condvars to the new header.
* No lists are pointing to us (as we asserted above), so we don't
* need to worry about the list nodes.
*/
nhdr->b_dva = hdr->b_dva;
nhdr->b_birth = hdr->b_birth;
nhdr->b_type = hdr->b_type;
nhdr->b_flags = hdr->b_flags;
nhdr->b_psize = hdr->b_psize;
nhdr->b_lsize = hdr->b_lsize;
nhdr->b_spa = hdr->b_spa;
nhdr->b_l1hdr.b_freeze_cksum = hdr->b_l1hdr.b_freeze_cksum;
nhdr->b_l1hdr.b_bufcnt = hdr->b_l1hdr.b_bufcnt;
nhdr->b_l1hdr.b_byteswap = hdr->b_l1hdr.b_byteswap;
nhdr->b_l1hdr.b_state = hdr->b_l1hdr.b_state;
nhdr->b_l1hdr.b_arc_access = hdr->b_l1hdr.b_arc_access;
nhdr->b_l1hdr.b_mru_hits = hdr->b_l1hdr.b_mru_hits;
nhdr->b_l1hdr.b_mru_ghost_hits = hdr->b_l1hdr.b_mru_ghost_hits;
nhdr->b_l1hdr.b_mfu_hits = hdr->b_l1hdr.b_mfu_hits;
nhdr->b_l1hdr.b_mfu_ghost_hits = hdr->b_l1hdr.b_mfu_ghost_hits;
nhdr->b_l1hdr.b_acb = hdr->b_l1hdr.b_acb;
nhdr->b_l1hdr.b_pabd = hdr->b_l1hdr.b_pabd;
/*
* This zfs_refcount_add() exists only to ensure that the individual
* arc buffers always point to a header that is referenced, avoiding
* a small race condition that could trigger ASSERTs.
*/
(void) zfs_refcount_add(&nhdr->b_l1hdr.b_refcnt, FTAG);
nhdr->b_l1hdr.b_buf = hdr->b_l1hdr.b_buf;
for (buf = nhdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) {
mutex_enter(&buf->b_evict_lock);
buf->b_hdr = nhdr;
mutex_exit(&buf->b_evict_lock);
}
zfs_refcount_transfer(&nhdr->b_l1hdr.b_refcnt, &hdr->b_l1hdr.b_refcnt);
(void) zfs_refcount_remove(&nhdr->b_l1hdr.b_refcnt, FTAG);
ASSERT0(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt));
if (need_crypt) {
arc_hdr_set_flags(nhdr, ARC_FLAG_PROTECTED);
} else {
arc_hdr_clear_flags(nhdr, ARC_FLAG_PROTECTED);
}
/* unset all members of the original hdr */
bzero(&hdr->b_dva, sizeof (dva_t));
hdr->b_birth = 0;
hdr->b_type = ARC_BUFC_INVALID;
hdr->b_flags = 0;
hdr->b_psize = 0;
hdr->b_lsize = 0;
hdr->b_spa = 0;
hdr->b_l1hdr.b_freeze_cksum = NULL;
hdr->b_l1hdr.b_buf = NULL;
hdr->b_l1hdr.b_bufcnt = 0;
hdr->b_l1hdr.b_byteswap = 0;
hdr->b_l1hdr.b_state = NULL;
hdr->b_l1hdr.b_arc_access = 0;
hdr->b_l1hdr.b_mru_hits = 0;
hdr->b_l1hdr.b_mru_ghost_hits = 0;
hdr->b_l1hdr.b_mfu_hits = 0;
hdr->b_l1hdr.b_mfu_ghost_hits = 0;
hdr->b_l1hdr.b_acb = NULL;
hdr->b_l1hdr.b_pabd = NULL;
if (ocache == hdr_full_crypt_cache) {
ASSERT(!HDR_HAS_RABD(hdr));
hdr->b_crypt_hdr.b_ot = DMU_OT_NONE;
hdr->b_crypt_hdr.b_ebufcnt = 0;
hdr->b_crypt_hdr.b_dsobj = 0;
bzero(hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN);
bzero(hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN);
bzero(hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN);
}
buf_discard_identity(hdr);
kmem_cache_free(ocache, hdr);
return (nhdr);
}
/*
* This function is used by the send / receive code to convert a newly
* allocated arc_buf_t to one that is suitable for a raw encrypted write. It
* is also used to allow the root objset block to be updated without altering
* its embedded MACs. Both block types will always be uncompressed so we do not
* have to worry about compression type or psize.
*/
void
arc_convert_to_raw(arc_buf_t *buf, uint64_t dsobj, boolean_t byteorder,
dmu_object_type_t ot, const uint8_t *salt, const uint8_t *iv,
const uint8_t *mac)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
ASSERT(ot == DMU_OT_DNODE || ot == DMU_OT_OBJSET);
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
buf->b_flags |= (ARC_BUF_FLAG_COMPRESSED | ARC_BUF_FLAG_ENCRYPTED);
if (!HDR_PROTECTED(hdr))
hdr = arc_hdr_realloc_crypt(hdr, B_TRUE);
hdr->b_crypt_hdr.b_dsobj = dsobj;
hdr->b_crypt_hdr.b_ot = ot;
hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ?
DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot);
if (!arc_hdr_has_uncompressed_buf(hdr))
arc_cksum_free(hdr);
if (salt != NULL)
bcopy(salt, hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN);
if (iv != NULL)
bcopy(iv, hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN);
if (mac != NULL)
bcopy(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN);
}
/*
* Allocate a new arc_buf_hdr_t and arc_buf_t and return the buf to the caller.
* The buf is returned thawed since we expect the consumer to modify it.
*/
arc_buf_t *
arc_alloc_buf(spa_t *spa, void *tag, arc_buf_contents_t type, int32_t size)
{
arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), size, size,
B_FALSE, ZIO_COMPRESS_OFF, 0, type);
arc_buf_t *buf = NULL;
VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE, B_FALSE,
B_FALSE, B_FALSE, &buf));
arc_buf_thaw(buf);
return (buf);
}
/*
* Allocate a compressed buf in the same manner as arc_alloc_buf. Don't use this
* for bufs containing metadata.
*/
arc_buf_t *
arc_alloc_compressed_buf(spa_t *spa, void *tag, uint64_t psize, uint64_t lsize,
enum zio_compress compression_type, uint8_t complevel)
{
ASSERT3U(lsize, >, 0);
ASSERT3U(lsize, >=, psize);
ASSERT3U(compression_type, >, ZIO_COMPRESS_OFF);
ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS);
arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize,
B_FALSE, compression_type, complevel, ARC_BUFC_DATA);
arc_buf_t *buf = NULL;
VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE,
B_TRUE, B_FALSE, B_FALSE, &buf));
arc_buf_thaw(buf);
ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
/*
* To ensure that the hdr has the correct data in it if we call
* arc_untransform() on this buf before it's been written to disk,
* it's easiest if we just set up sharing between the buf and the hdr.
*/
arc_share_buf(hdr, buf);
return (buf);
}
arc_buf_t *
arc_alloc_raw_buf(spa_t *spa, void *tag, uint64_t dsobj, boolean_t byteorder,
const uint8_t *salt, const uint8_t *iv, const uint8_t *mac,
dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
enum zio_compress compression_type, uint8_t complevel)
{
arc_buf_hdr_t *hdr;
arc_buf_t *buf;
arc_buf_contents_t type = DMU_OT_IS_METADATA(ot) ?
ARC_BUFC_METADATA : ARC_BUFC_DATA;
ASSERT3U(lsize, >, 0);
ASSERT3U(lsize, >=, psize);
ASSERT3U(compression_type, >=, ZIO_COMPRESS_OFF);
ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS);
hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize, B_TRUE,
compression_type, complevel, type);
hdr->b_crypt_hdr.b_dsobj = dsobj;
hdr->b_crypt_hdr.b_ot = ot;
hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ?
DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot);
bcopy(salt, hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN);
bcopy(iv, hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN);
bcopy(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN);
/*
* This buffer will be considered encrypted even if the ot is not an
* encrypted type. It will become authenticated instead in
* arc_write_ready().
*/
buf = NULL;
VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_TRUE, B_TRUE,
B_FALSE, B_FALSE, &buf));
arc_buf_thaw(buf);
ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
return (buf);
}
static void
l2arc_hdr_arcstats_update(arc_buf_hdr_t *hdr, boolean_t incr,
boolean_t state_only)
{
l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr;
l2arc_dev_t *dev = l2hdr->b_dev;
uint64_t lsize = HDR_GET_LSIZE(hdr);
uint64_t psize = HDR_GET_PSIZE(hdr);
uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize);
arc_buf_contents_t type = hdr->b_type;
int64_t lsize_s;
int64_t psize_s;
int64_t asize_s;
if (incr) {
lsize_s = lsize;
psize_s = psize;
asize_s = asize;
} else {
lsize_s = -lsize;
psize_s = -psize;
asize_s = -asize;
}
/* If the buffer is a prefetch, count it as such. */
if (HDR_PREFETCH(hdr)) {
ARCSTAT_INCR(arcstat_l2_prefetch_asize, asize_s);
} else {
/*
* We use the value stored in the L2 header upon initial
* caching in L2ARC. This value will be updated in case
* an MRU/MRU_ghost buffer transitions to MFU but the L2ARC
* metadata (log entry) cannot currently be updated. Having
* the ARC state in the L2 header solves the problem of a
* possibly absent L1 header (apparent in buffers restored
* from persistent L2ARC).
*/
switch (hdr->b_l2hdr.b_arcs_state) {
case ARC_STATE_MRU_GHOST:
case ARC_STATE_MRU:
ARCSTAT_INCR(arcstat_l2_mru_asize, asize_s);
break;
case ARC_STATE_MFU_GHOST:
case ARC_STATE_MFU:
ARCSTAT_INCR(arcstat_l2_mfu_asize, asize_s);
break;
default:
break;
}
}
if (state_only)
return;
ARCSTAT_INCR(arcstat_l2_psize, psize_s);
ARCSTAT_INCR(arcstat_l2_lsize, lsize_s);
switch (type) {
case ARC_BUFC_DATA:
ARCSTAT_INCR(arcstat_l2_bufc_data_asize, asize_s);
break;
case ARC_BUFC_METADATA:
ARCSTAT_INCR(arcstat_l2_bufc_metadata_asize, asize_s);
break;
default:
break;
}
}
static void
arc_hdr_l2hdr_destroy(arc_buf_hdr_t *hdr)
{
l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr;
l2arc_dev_t *dev = l2hdr->b_dev;
uint64_t psize = HDR_GET_PSIZE(hdr);
uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize);
ASSERT(MUTEX_HELD(&dev->l2ad_mtx));
ASSERT(HDR_HAS_L2HDR(hdr));
list_remove(&dev->l2ad_buflist, hdr);
l2arc_hdr_arcstats_decrement(hdr);
vdev_space_update(dev->l2ad_vdev, -asize, 0, 0);
(void) zfs_refcount_remove_many(&dev->l2ad_alloc, arc_hdr_size(hdr),
hdr);
arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR);
}
static void
arc_hdr_destroy(arc_buf_hdr_t *hdr)
{
if (HDR_HAS_L1HDR(hdr)) {
ASSERT(hdr->b_l1hdr.b_buf == NULL ||
hdr->b_l1hdr.b_bufcnt > 0);
ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
}
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT(!HDR_IN_HASH_TABLE(hdr));
if (HDR_HAS_L2HDR(hdr)) {
l2arc_dev_t *dev = hdr->b_l2hdr.b_dev;
boolean_t buflist_held = MUTEX_HELD(&dev->l2ad_mtx);
if (!buflist_held)
mutex_enter(&dev->l2ad_mtx);
/*
* Even though we checked this conditional above, we
* need to check this again now that we have the
* l2ad_mtx. This is because we could be racing with
* another thread calling l2arc_evict() which might have
* destroyed this header's L2 portion as we were waiting
* to acquire the l2ad_mtx. If that happens, we don't
* want to re-destroy the header's L2 portion.
*/
if (HDR_HAS_L2HDR(hdr)) {
if (!HDR_EMPTY(hdr))
buf_discard_identity(hdr);
arc_hdr_l2hdr_destroy(hdr);
}
if (!buflist_held)
mutex_exit(&dev->l2ad_mtx);
}
/*
* The header's identify can only be safely discarded once it is no
* longer discoverable. This requires removing it from the hash table
* and the l2arc header list. After this point the hash lock can not
* be used to protect the header.
*/
if (!HDR_EMPTY(hdr))
buf_discard_identity(hdr);
if (HDR_HAS_L1HDR(hdr)) {
arc_cksum_free(hdr);
while (hdr->b_l1hdr.b_buf != NULL)
arc_buf_destroy_impl(hdr->b_l1hdr.b_buf);
if (hdr->b_l1hdr.b_pabd != NULL)
arc_hdr_free_abd(hdr, B_FALSE);
if (HDR_HAS_RABD(hdr))
arc_hdr_free_abd(hdr, B_TRUE);
}
ASSERT3P(hdr->b_hash_next, ==, NULL);
if (HDR_HAS_L1HDR(hdr)) {
ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
if (!HDR_PROTECTED(hdr)) {
kmem_cache_free(hdr_full_cache, hdr);
} else {
kmem_cache_free(hdr_full_crypt_cache, hdr);
}
} else {
kmem_cache_free(hdr_l2only_cache, hdr);
}
}
void
arc_buf_destroy(arc_buf_t *buf, void* tag)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
if (hdr->b_l1hdr.b_state == arc_anon) {
ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1);
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
VERIFY0(remove_reference(hdr, NULL, tag));
arc_hdr_destroy(hdr);
return;
}
kmutex_t *hash_lock = HDR_LOCK(hdr);
mutex_enter(hash_lock);
ASSERT3P(hdr, ==, buf->b_hdr);
ASSERT(hdr->b_l1hdr.b_bufcnt > 0);
ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
ASSERT3P(hdr->b_l1hdr.b_state, !=, arc_anon);
ASSERT3P(buf->b_data, !=, NULL);
(void) remove_reference(hdr, hash_lock, tag);
arc_buf_destroy_impl(buf);
mutex_exit(hash_lock);
}
/*
* Evict the arc_buf_hdr that is provided as a parameter. The resultant
* state of the header is dependent on its state prior to entering this
* function. The following transitions are possible:
*
* - arc_mru -> arc_mru_ghost
* - arc_mfu -> arc_mfu_ghost
* - arc_mru_ghost -> arc_l2c_only
* - arc_mru_ghost -> deleted
* - arc_mfu_ghost -> arc_l2c_only
* - arc_mfu_ghost -> deleted
*
* Return total size of evicted data buffers for eviction progress tracking.
* When evicting from ghost states return logical buffer size to make eviction
* progress at the same (or at least comparable) rate as from non-ghost states.
*
* Return *real_evicted for actual ARC size reduction to wake up threads
* waiting for it. For non-ghost states it includes size of evicted data
* buffers (the headers are not freed there). For ghost states it includes
* only the evicted headers size.
*/
static int64_t
arc_evict_hdr(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, uint64_t *real_evicted)
{
arc_state_t *evicted_state, *state;
int64_t bytes_evicted = 0;
int min_lifetime = HDR_PRESCIENT_PREFETCH(hdr) ?
arc_min_prescient_prefetch_ms : arc_min_prefetch_ms;
ASSERT(MUTEX_HELD(hash_lock));
ASSERT(HDR_HAS_L1HDR(hdr));
*real_evicted = 0;
state = hdr->b_l1hdr.b_state;
if (GHOST_STATE(state)) {
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
/*
* l2arc_write_buffers() relies on a header's L1 portion
* (i.e. its b_pabd field) during it's write phase.
* Thus, we cannot push a header onto the arc_l2c_only
* state (removing its L1 piece) until the header is
* done being written to the l2arc.
*/
if (HDR_HAS_L2HDR(hdr) && HDR_L2_WRITING(hdr)) {
ARCSTAT_BUMP(arcstat_evict_l2_skip);
return (bytes_evicted);
}
ARCSTAT_BUMP(arcstat_deleted);
bytes_evicted += HDR_GET_LSIZE(hdr);
DTRACE_PROBE1(arc__delete, arc_buf_hdr_t *, hdr);
if (HDR_HAS_L2HDR(hdr)) {
ASSERT(hdr->b_l1hdr.b_pabd == NULL);
ASSERT(!HDR_HAS_RABD(hdr));
/*
* This buffer is cached on the 2nd Level ARC;
* don't destroy the header.
*/
arc_change_state(arc_l2c_only, hdr, hash_lock);
/*
* dropping from L1+L2 cached to L2-only,
* realloc to remove the L1 header.
*/
hdr = arc_hdr_realloc(hdr, hdr_full_cache,
hdr_l2only_cache);
*real_evicted += HDR_FULL_SIZE - HDR_L2ONLY_SIZE;
} else {
arc_change_state(arc_anon, hdr, hash_lock);
arc_hdr_destroy(hdr);
*real_evicted += HDR_FULL_SIZE;
}
return (bytes_evicted);
}
ASSERT(state == arc_mru || state == arc_mfu);
evicted_state = (state == arc_mru) ? arc_mru_ghost : arc_mfu_ghost;
/* prefetch buffers have a minimum lifespan */
if (HDR_IO_IN_PROGRESS(hdr) ||
((hdr->b_flags & (ARC_FLAG_PREFETCH | ARC_FLAG_INDIRECT)) &&
ddi_get_lbolt() - hdr->b_l1hdr.b_arc_access <
MSEC_TO_TICK(min_lifetime))) {
ARCSTAT_BUMP(arcstat_evict_skip);
return (bytes_evicted);
}
ASSERT0(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt));
while (hdr->b_l1hdr.b_buf) {
arc_buf_t *buf = hdr->b_l1hdr.b_buf;
if (!mutex_tryenter(&buf->b_evict_lock)) {
ARCSTAT_BUMP(arcstat_mutex_miss);
break;
}
if (buf->b_data != NULL) {
bytes_evicted += HDR_GET_LSIZE(hdr);
*real_evicted += HDR_GET_LSIZE(hdr);
}
mutex_exit(&buf->b_evict_lock);
arc_buf_destroy_impl(buf);
}
if (HDR_HAS_L2HDR(hdr)) {
ARCSTAT_INCR(arcstat_evict_l2_cached, HDR_GET_LSIZE(hdr));
} else {
if (l2arc_write_eligible(hdr->b_spa, hdr)) {
ARCSTAT_INCR(arcstat_evict_l2_eligible,
HDR_GET_LSIZE(hdr));
switch (state->arcs_state) {
case ARC_STATE_MRU:
ARCSTAT_INCR(
arcstat_evict_l2_eligible_mru,
HDR_GET_LSIZE(hdr));
break;
case ARC_STATE_MFU:
ARCSTAT_INCR(
arcstat_evict_l2_eligible_mfu,
HDR_GET_LSIZE(hdr));
break;
default:
break;
}
} else {
ARCSTAT_INCR(arcstat_evict_l2_ineligible,
HDR_GET_LSIZE(hdr));
}
}
if (hdr->b_l1hdr.b_bufcnt == 0) {
arc_cksum_free(hdr);
bytes_evicted += arc_hdr_size(hdr);
*real_evicted += arc_hdr_size(hdr);
/*
* If this hdr is being evicted and has a compressed
* buffer then we discard it here before we change states.
* This ensures that the accounting is updated correctly
* in arc_free_data_impl().
*/
if (hdr->b_l1hdr.b_pabd != NULL)
arc_hdr_free_abd(hdr, B_FALSE);
if (HDR_HAS_RABD(hdr))
arc_hdr_free_abd(hdr, B_TRUE);
arc_change_state(evicted_state, hdr, hash_lock);
ASSERT(HDR_IN_HASH_TABLE(hdr));
arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE);
DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, hdr);
}
return (bytes_evicted);
}
static void
arc_set_need_free(void)
{
ASSERT(MUTEX_HELD(&arc_evict_lock));
int64_t remaining = arc_free_memory() - arc_sys_free / 2;
arc_evict_waiter_t *aw = list_tail(&arc_evict_waiters);
if (aw == NULL) {
arc_need_free = MAX(-remaining, 0);
} else {
arc_need_free =
MAX(-remaining, (int64_t)(aw->aew_count - arc_evict_count));
}
}
static uint64_t
arc_evict_state_impl(multilist_t *ml, int idx, arc_buf_hdr_t *marker,
uint64_t spa, uint64_t bytes)
{
multilist_sublist_t *mls;
uint64_t bytes_evicted = 0, real_evicted = 0;
arc_buf_hdr_t *hdr;
kmutex_t *hash_lock;
int evict_count = zfs_arc_evict_batch_limit;
ASSERT3P(marker, !=, NULL);
mls = multilist_sublist_lock(ml, idx);
for (hdr = multilist_sublist_prev(mls, marker); likely(hdr != NULL);
hdr = multilist_sublist_prev(mls, marker)) {
if ((evict_count <= 0) || (bytes_evicted >= bytes))
break;
/*
* To keep our iteration location, move the marker
* forward. Since we're not holding hdr's hash lock, we
* must be very careful and not remove 'hdr' from the
* sublist. Otherwise, other consumers might mistake the
* 'hdr' as not being on a sublist when they call the
* multilist_link_active() function (they all rely on
* the hash lock protecting concurrent insertions and
* removals). multilist_sublist_move_forward() was
* specifically implemented to ensure this is the case
* (only 'marker' will be removed and re-inserted).
*/
multilist_sublist_move_forward(mls, marker);
/*
* The only case where the b_spa field should ever be
* zero, is the marker headers inserted by
* arc_evict_state(). It's possible for multiple threads
* to be calling arc_evict_state() concurrently (e.g.
* dsl_pool_close() and zio_inject_fault()), so we must
* skip any markers we see from these other threads.
*/
if (hdr->b_spa == 0)
continue;
/* we're only interested in evicting buffers of a certain spa */
if (spa != 0 && hdr->b_spa != spa) {
ARCSTAT_BUMP(arcstat_evict_skip);
continue;
}
hash_lock = HDR_LOCK(hdr);
/*
* We aren't calling this function from any code path
* that would already be holding a hash lock, so we're
* asserting on this assumption to be defensive in case
* this ever changes. Without this check, it would be
* possible to incorrectly increment arcstat_mutex_miss
* below (e.g. if the code changed such that we called
* this function with a hash lock held).
*/
ASSERT(!MUTEX_HELD(hash_lock));
if (mutex_tryenter(hash_lock)) {
uint64_t revicted;
uint64_t evicted = arc_evict_hdr(hdr, hash_lock,
&revicted);
mutex_exit(hash_lock);
bytes_evicted += evicted;
real_evicted += revicted;
/*
* If evicted is zero, arc_evict_hdr() must have
* decided to skip this header, don't increment
* evict_count in this case.
*/
if (evicted != 0)
evict_count--;
} else {
ARCSTAT_BUMP(arcstat_mutex_miss);
}
}
multilist_sublist_unlock(mls);
/*
* Increment the count of evicted bytes, and wake up any threads that
* are waiting for the count to reach this value. Since the list is
* ordered by ascending aew_count, we pop off the beginning of the
* list until we reach the end, or a waiter that's past the current
* "count". Doing this outside the loop reduces the number of times
* we need to acquire the global arc_evict_lock.
*
* Only wake when there's sufficient free memory in the system
* (specifically, arc_sys_free/2, which by default is a bit more than
* 1/64th of RAM). See the comments in arc_wait_for_eviction().
*/
mutex_enter(&arc_evict_lock);
arc_evict_count += real_evicted;
if (arc_free_memory() > arc_sys_free / 2) {
arc_evict_waiter_t *aw;
while ((aw = list_head(&arc_evict_waiters)) != NULL &&
aw->aew_count <= arc_evict_count) {
list_remove(&arc_evict_waiters, aw);
cv_broadcast(&aw->aew_cv);
}
}
arc_set_need_free();
mutex_exit(&arc_evict_lock);
/*
* If the ARC size is reduced from arc_c_max to arc_c_min (especially
* if the average cached block is small), eviction can be on-CPU for
* many seconds. To ensure that other threads that may be bound to
* this CPU are able to make progress, make a voluntary preemption
* call here.
*/
cond_resched();
return (bytes_evicted);
}
/*
* Evict buffers from the given arc state, until we've removed the
* specified number of bytes. Move the removed buffers to the
* appropriate evict state.
*
* This function makes a "best effort". It skips over any buffers
* it can't get a hash_lock on, and so, may not catch all candidates.
* It may also return without evicting as much space as requested.
*
* If bytes is specified using the special value ARC_EVICT_ALL, this
* will evict all available (i.e. unlocked and evictable) buffers from
* the given arc state; which is used by arc_flush().
*/
static uint64_t
arc_evict_state(arc_state_t *state, uint64_t spa, uint64_t bytes,
arc_buf_contents_t type)
{
uint64_t total_evicted = 0;
multilist_t *ml = &state->arcs_list[type];
int num_sublists;
arc_buf_hdr_t **markers;
num_sublists = multilist_get_num_sublists(ml);
/*
* If we've tried to evict from each sublist, made some
* progress, but still have not hit the target number of bytes
* to evict, we want to keep trying. The markers allow us to
* pick up where we left off for each individual sublist, rather
* than starting from the tail each time.
*/
markers = kmem_zalloc(sizeof (*markers) * num_sublists, KM_SLEEP);
for (int i = 0; i < num_sublists; i++) {
multilist_sublist_t *mls;
markers[i] = kmem_cache_alloc(hdr_full_cache, KM_SLEEP);
/*
* A b_spa of 0 is used to indicate that this header is
* a marker. This fact is used in arc_evict_type() and
* arc_evict_state_impl().
*/
markers[i]->b_spa = 0;
mls = multilist_sublist_lock(ml, i);
multilist_sublist_insert_tail(mls, markers[i]);
multilist_sublist_unlock(mls);
}
/*
* While we haven't hit our target number of bytes to evict, or
* we're evicting all available buffers.
*/
while (total_evicted < bytes) {
int sublist_idx = multilist_get_random_index(ml);
uint64_t scan_evicted = 0;
/*
* Try to reduce pinned dnodes with a floor of arc_dnode_limit.
* Request that 10% of the LRUs be scanned by the superblock
* shrinker.
*/
if (type == ARC_BUFC_DATA && aggsum_compare(
&arc_sums.arcstat_dnode_size, arc_dnode_size_limit) > 0) {
arc_prune_async((aggsum_upper_bound(
&arc_sums.arcstat_dnode_size) -
arc_dnode_size_limit) / sizeof (dnode_t) /
zfs_arc_dnode_reduce_percent);
}
/*
* Start eviction using a randomly selected sublist,
* this is to try and evenly balance eviction across all
* sublists. Always starting at the same sublist
* (e.g. index 0) would cause evictions to favor certain
* sublists over others.
*/
for (int i = 0; i < num_sublists; i++) {
uint64_t bytes_remaining;
uint64_t bytes_evicted;
if (total_evicted < bytes)
bytes_remaining = bytes - total_evicted;
else
break;
bytes_evicted = arc_evict_state_impl(ml, sublist_idx,
markers[sublist_idx], spa, bytes_remaining);
scan_evicted += bytes_evicted;
total_evicted += bytes_evicted;
/* we've reached the end, wrap to the beginning */
if (++sublist_idx >= num_sublists)
sublist_idx = 0;
}
/*
* If we didn't evict anything during this scan, we have
* no reason to believe we'll evict more during another
* scan, so break the loop.
*/
if (scan_evicted == 0) {
/* This isn't possible, let's make that obvious */
ASSERT3S(bytes, !=, 0);
/*
* When bytes is ARC_EVICT_ALL, the only way to
* break the loop is when scan_evicted is zero.
* In that case, we actually have evicted enough,
* so we don't want to increment the kstat.
*/
if (bytes != ARC_EVICT_ALL) {
ASSERT3S(total_evicted, <, bytes);
ARCSTAT_BUMP(arcstat_evict_not_enough);
}
break;
}
}
for (int i = 0; i < num_sublists; i++) {
multilist_sublist_t *mls = multilist_sublist_lock(ml, i);
multilist_sublist_remove(mls, markers[i]);
multilist_sublist_unlock(mls);
kmem_cache_free(hdr_full_cache, markers[i]);
}
kmem_free(markers, sizeof (*markers) * num_sublists);
return (total_evicted);
}
/*
* Flush all "evictable" data of the given type from the arc state
* specified. This will not evict any "active" buffers (i.e. referenced).
*
* When 'retry' is set to B_FALSE, the function will make a single pass
* over the state and evict any buffers that it can. Since it doesn't
* continually retry the eviction, it might end up leaving some buffers
* in the ARC due to lock misses.
*
* When 'retry' is set to B_TRUE, the function will continually retry the
* eviction until *all* evictable buffers have been removed from the
* state. As a result, if concurrent insertions into the state are
* allowed (e.g. if the ARC isn't shutting down), this function might
* wind up in an infinite loop, continually trying to evict buffers.
*/
static uint64_t
arc_flush_state(arc_state_t *state, uint64_t spa, arc_buf_contents_t type,
boolean_t retry)
{
uint64_t evicted = 0;
while (zfs_refcount_count(&state->arcs_esize[type]) != 0) {
evicted += arc_evict_state(state, spa, ARC_EVICT_ALL, type);
if (!retry)
break;
}
return (evicted);
}
/*
* Evict the specified number of bytes from the state specified,
* restricting eviction to the spa and type given. This function
* prevents us from trying to evict more from a state's list than
* is "evictable", and to skip evicting altogether when passed a
* negative value for "bytes". In contrast, arc_evict_state() will
* evict everything it can, when passed a negative value for "bytes".
*/
static uint64_t
arc_evict_impl(arc_state_t *state, uint64_t spa, int64_t bytes,
arc_buf_contents_t type)
{
uint64_t delta;
if (bytes > 0 && zfs_refcount_count(&state->arcs_esize[type]) > 0) {
delta = MIN(zfs_refcount_count(&state->arcs_esize[type]),
bytes);
return (arc_evict_state(state, spa, delta, type));
}
return (0);
}
/*
* The goal of this function is to evict enough meta data buffers from the
* ARC in order to enforce the arc_meta_limit. Achieving this is slightly
* more complicated than it appears because it is common for data buffers
* to have holds on meta data buffers. In addition, dnode meta data buffers
* will be held by the dnodes in the block preventing them from being freed.
* This means we can't simply traverse the ARC and expect to always find
* enough unheld meta data buffer to release.
*
* Therefore, this function has been updated to make alternating passes
* over the ARC releasing data buffers and then newly unheld meta data
* buffers. This ensures forward progress is maintained and meta_used
* will decrease. Normally this is sufficient, but if required the ARC
* will call the registered prune callbacks causing dentry and inodes to
* be dropped from the VFS cache. This will make dnode meta data buffers
* available for reclaim.
*/
static uint64_t
arc_evict_meta_balanced(uint64_t meta_used)
{
int64_t delta, prune = 0, adjustmnt;
uint64_t total_evicted = 0;
arc_buf_contents_t type = ARC_BUFC_DATA;
int restarts = MAX(zfs_arc_meta_adjust_restarts, 0);
restart:
/*
* This slightly differs than the way we evict from the mru in
* arc_evict because we don't have a "target" value (i.e. no
* "meta" arc_p). As a result, I think we can completely
* cannibalize the metadata in the MRU before we evict the
* metadata from the MFU. I think we probably need to implement a
* "metadata arc_p" value to do this properly.
*/
adjustmnt = meta_used - arc_meta_limit;
if (adjustmnt > 0 &&
zfs_refcount_count(&arc_mru->arcs_esize[type]) > 0) {
delta = MIN(zfs_refcount_count(&arc_mru->arcs_esize[type]),
adjustmnt);
total_evicted += arc_evict_impl(arc_mru, 0, delta, type);
adjustmnt -= delta;
}
/*
* We can't afford to recalculate adjustmnt here. If we do,
* new metadata buffers can sneak into the MRU or ANON lists,
* thus penalize the MFU metadata. Although the fudge factor is
* small, it has been empirically shown to be significant for
* certain workloads (e.g. creating many empty directories). As
* such, we use the original calculation for adjustmnt, and
* simply decrement the amount of data evicted from the MRU.
*/
if (adjustmnt > 0 &&
zfs_refcount_count(&arc_mfu->arcs_esize[type]) > 0) {
delta = MIN(zfs_refcount_count(&arc_mfu->arcs_esize[type]),
adjustmnt);
total_evicted += arc_evict_impl(arc_mfu, 0, delta, type);
}
adjustmnt = meta_used - arc_meta_limit;
if (adjustmnt > 0 &&
zfs_refcount_count(&arc_mru_ghost->arcs_esize[type]) > 0) {
delta = MIN(adjustmnt,
zfs_refcount_count(&arc_mru_ghost->arcs_esize[type]));
total_evicted += arc_evict_impl(arc_mru_ghost, 0, delta, type);
adjustmnt -= delta;
}
if (adjustmnt > 0 &&
zfs_refcount_count(&arc_mfu_ghost->arcs_esize[type]) > 0) {
delta = MIN(adjustmnt,
zfs_refcount_count(&arc_mfu_ghost->arcs_esize[type]));
total_evicted += arc_evict_impl(arc_mfu_ghost, 0, delta, type);
}
/*
* If after attempting to make the requested adjustment to the ARC
* the meta limit is still being exceeded then request that the
* higher layers drop some cached objects which have holds on ARC
* meta buffers. Requests to the upper layers will be made with
* increasingly large scan sizes until the ARC is below the limit.
*/
if (meta_used > arc_meta_limit) {
if (type == ARC_BUFC_DATA) {
type = ARC_BUFC_METADATA;
} else {
type = ARC_BUFC_DATA;
if (zfs_arc_meta_prune) {
prune += zfs_arc_meta_prune;
arc_prune_async(prune);
}
}
if (restarts > 0) {
restarts--;
goto restart;
}
}
return (total_evicted);
}
/*
* Evict metadata buffers from the cache, such that arcstat_meta_used is
* capped by the arc_meta_limit tunable.
*/
static uint64_t
arc_evict_meta_only(uint64_t meta_used)
{
uint64_t total_evicted = 0;
int64_t target;
/*
* If we're over the meta limit, we want to evict enough
* metadata to get back under the meta limit. We don't want to
* evict so much that we drop the MRU below arc_p, though. If
* we're over the meta limit more than we're over arc_p, we
* evict some from the MRU here, and some from the MFU below.
*/
target = MIN((int64_t)(meta_used - arc_meta_limit),
(int64_t)(zfs_refcount_count(&arc_anon->arcs_size) +
zfs_refcount_count(&arc_mru->arcs_size) - arc_p));
total_evicted += arc_evict_impl(arc_mru, 0, target, ARC_BUFC_METADATA);
/*
* Similar to the above, we want to evict enough bytes to get us
* below the meta limit, but not so much as to drop us below the
* space allotted to the MFU (which is defined as arc_c - arc_p).
*/
target = MIN((int64_t)(meta_used - arc_meta_limit),
(int64_t)(zfs_refcount_count(&arc_mfu->arcs_size) -
(arc_c - arc_p)));
total_evicted += arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_METADATA);
return (total_evicted);
}
static uint64_t
arc_evict_meta(uint64_t meta_used)
{
if (zfs_arc_meta_strategy == ARC_STRATEGY_META_ONLY)
return (arc_evict_meta_only(meta_used));
else
return (arc_evict_meta_balanced(meta_used));
}
/*
* Return the type of the oldest buffer in the given arc state
*
* This function will select a random sublist of type ARC_BUFC_DATA and
* a random sublist of type ARC_BUFC_METADATA. The tail of each sublist
* is compared, and the type which contains the "older" buffer will be
* returned.
*/
static arc_buf_contents_t
arc_evict_type(arc_state_t *state)
{
multilist_t *data_ml = &state->arcs_list[ARC_BUFC_DATA];
multilist_t *meta_ml = &state->arcs_list[ARC_BUFC_METADATA];
int data_idx = multilist_get_random_index(data_ml);
int meta_idx = multilist_get_random_index(meta_ml);
multilist_sublist_t *data_mls;
multilist_sublist_t *meta_mls;
arc_buf_contents_t type;
arc_buf_hdr_t *data_hdr;
arc_buf_hdr_t *meta_hdr;
/*
* We keep the sublist lock until we're finished, to prevent
* the headers from being destroyed via arc_evict_state().
*/
data_mls = multilist_sublist_lock(data_ml, data_idx);
meta_mls = multilist_sublist_lock(meta_ml, meta_idx);
/*
* These two loops are to ensure we skip any markers that
* might be at the tail of the lists due to arc_evict_state().
*/
for (data_hdr = multilist_sublist_tail(data_mls); data_hdr != NULL;
data_hdr = multilist_sublist_prev(data_mls, data_hdr)) {
if (data_hdr->b_spa != 0)
break;
}
for (meta_hdr = multilist_sublist_tail(meta_mls); meta_hdr != NULL;
meta_hdr = multilist_sublist_prev(meta_mls, meta_hdr)) {
if (meta_hdr->b_spa != 0)
break;
}
if (data_hdr == NULL && meta_hdr == NULL) {
type = ARC_BUFC_DATA;
} else if (data_hdr == NULL) {
ASSERT3P(meta_hdr, !=, NULL);
type = ARC_BUFC_METADATA;
} else if (meta_hdr == NULL) {
ASSERT3P(data_hdr, !=, NULL);
type = ARC_BUFC_DATA;
} else {
ASSERT3P(data_hdr, !=, NULL);
ASSERT3P(meta_hdr, !=, NULL);
/* The headers can't be on the sublist without an L1 header */
ASSERT(HDR_HAS_L1HDR(data_hdr));
ASSERT(HDR_HAS_L1HDR(meta_hdr));
if (data_hdr->b_l1hdr.b_arc_access <
meta_hdr->b_l1hdr.b_arc_access) {
type = ARC_BUFC_DATA;
} else {
type = ARC_BUFC_METADATA;
}
}
multilist_sublist_unlock(meta_mls);
multilist_sublist_unlock(data_mls);
return (type);
}
/*
* Evict buffers from the cache, such that arcstat_size is capped by arc_c.
*/
static uint64_t
arc_evict(void)
{
uint64_t total_evicted = 0;
uint64_t bytes;
int64_t target;
uint64_t asize = aggsum_value(&arc_sums.arcstat_size);
uint64_t ameta = aggsum_value(&arc_sums.arcstat_meta_used);
/*
* If we're over arc_meta_limit, we want to correct that before
* potentially evicting data buffers below.
*/
total_evicted += arc_evict_meta(ameta);
/*
* Adjust MRU size
*
* If we're over the target cache size, we want to evict enough
* from the list to get back to our target size. We don't want
* to evict too much from the MRU, such that it drops below
* arc_p. So, if we're over our target cache size more than
* the MRU is over arc_p, we'll evict enough to get back to
* arc_p here, and then evict more from the MFU below.
*/
target = MIN((int64_t)(asize - arc_c),
(int64_t)(zfs_refcount_count(&arc_anon->arcs_size) +
zfs_refcount_count(&arc_mru->arcs_size) + ameta - arc_p));
/*
* If we're below arc_meta_min, always prefer to evict data.
* Otherwise, try to satisfy the requested number of bytes to
* evict from the type which contains older buffers; in an
* effort to keep newer buffers in the cache regardless of their
* type. If we cannot satisfy the number of bytes from this
* type, spill over into the next type.
*/
if (arc_evict_type(arc_mru) == ARC_BUFC_METADATA &&
ameta > arc_meta_min) {
bytes = arc_evict_impl(arc_mru, 0, target, ARC_BUFC_METADATA);
total_evicted += bytes;
/*
* If we couldn't evict our target number of bytes from
* metadata, we try to get the rest from data.
*/
target -= bytes;
total_evicted +=
arc_evict_impl(arc_mru, 0, target, ARC_BUFC_DATA);
} else {
bytes = arc_evict_impl(arc_mru, 0, target, ARC_BUFC_DATA);
total_evicted += bytes;
/*
* If we couldn't evict our target number of bytes from
* data, we try to get the rest from metadata.
*/
target -= bytes;
total_evicted +=
arc_evict_impl(arc_mru, 0, target, ARC_BUFC_METADATA);
}
/*
* Re-sum ARC stats after the first round of evictions.
*/
asize = aggsum_value(&arc_sums.arcstat_size);
ameta = aggsum_value(&arc_sums.arcstat_meta_used);
/*
* Adjust MFU size
*
* Now that we've tried to evict enough from the MRU to get its
* size back to arc_p, if we're still above the target cache
* size, we evict the rest from the MFU.
*/
target = asize - arc_c;
if (arc_evict_type(arc_mfu) == ARC_BUFC_METADATA &&
ameta > arc_meta_min) {
bytes = arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_METADATA);
total_evicted += bytes;
/*
* If we couldn't evict our target number of bytes from
* metadata, we try to get the rest from data.
*/
target -= bytes;
total_evicted +=
arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_DATA);
} else {
bytes = arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_DATA);
total_evicted += bytes;
/*
* If we couldn't evict our target number of bytes from
* data, we try to get the rest from data.
*/
target -= bytes;
total_evicted +=
arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_METADATA);
}
/*
* Adjust ghost lists
*
* In addition to the above, the ARC also defines target values
* for the ghost lists. The sum of the mru list and mru ghost
* list should never exceed the target size of the cache, and
* the sum of the mru list, mfu list, mru ghost list, and mfu
* ghost list should never exceed twice the target size of the
* cache. The following logic enforces these limits on the ghost
* caches, and evicts from them as needed.
*/
target = zfs_refcount_count(&arc_mru->arcs_size) +
zfs_refcount_count(&arc_mru_ghost->arcs_size) - arc_c;
bytes = arc_evict_impl(arc_mru_ghost, 0, target, ARC_BUFC_DATA);
total_evicted += bytes;
target -= bytes;
total_evicted +=
arc_evict_impl(arc_mru_ghost, 0, target, ARC_BUFC_METADATA);
/*
* We assume the sum of the mru list and mfu list is less than
* or equal to arc_c (we enforced this above), which means we
* can use the simpler of the two equations below:
*
* mru + mfu + mru ghost + mfu ghost <= 2 * arc_c
* mru ghost + mfu ghost <= arc_c
*/
target = zfs_refcount_count(&arc_mru_ghost->arcs_size) +
zfs_refcount_count(&arc_mfu_ghost->arcs_size) - arc_c;
bytes = arc_evict_impl(arc_mfu_ghost, 0, target, ARC_BUFC_DATA);
total_evicted += bytes;
target -= bytes;
total_evicted +=
arc_evict_impl(arc_mfu_ghost, 0, target, ARC_BUFC_METADATA);
return (total_evicted);
}
void
arc_flush(spa_t *spa, boolean_t retry)
{
uint64_t guid = 0;
/*
* If retry is B_TRUE, a spa must not be specified since we have
* no good way to determine if all of a spa's buffers have been
* evicted from an arc state.
*/
ASSERT(!retry || spa == 0);
if (spa != NULL)
guid = spa_load_guid(spa);
(void) arc_flush_state(arc_mru, guid, ARC_BUFC_DATA, retry);
(void) arc_flush_state(arc_mru, guid, ARC_BUFC_METADATA, retry);
(void) arc_flush_state(arc_mfu, guid, ARC_BUFC_DATA, retry);
(void) arc_flush_state(arc_mfu, guid, ARC_BUFC_METADATA, retry);
(void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_DATA, retry);
(void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_METADATA, retry);
(void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_DATA, retry);
(void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_METADATA, retry);
}
void
arc_reduce_target_size(int64_t to_free)
{
uint64_t asize = aggsum_value(&arc_sums.arcstat_size);
/*
* All callers want the ARC to actually evict (at least) this much
* memory. Therefore we reduce from the lower of the current size and
* the target size. This way, even if arc_c is much higher than
* arc_size (as can be the case after many calls to arc_freed(), we will
* immediately have arc_c < arc_size and therefore the arc_evict_zthr
* will evict.
*/
uint64_t c = MIN(arc_c, asize);
if (c > to_free && c - to_free > arc_c_min) {
arc_c = c - to_free;
atomic_add_64(&arc_p, -(arc_p >> arc_shrink_shift));
if (arc_p > arc_c)
arc_p = (arc_c >> 1);
ASSERT(arc_c >= arc_c_min);
ASSERT((int64_t)arc_p >= 0);
} else {
arc_c = arc_c_min;
}
if (asize > arc_c) {
/* See comment in arc_evict_cb_check() on why lock+flag */
mutex_enter(&arc_evict_lock);
arc_evict_needed = B_TRUE;
mutex_exit(&arc_evict_lock);
zthr_wakeup(arc_evict_zthr);
}
}
/*
* Determine if the system is under memory pressure and is asking
* to reclaim memory. A return value of B_TRUE indicates that the system
* is under memory pressure and that the arc should adjust accordingly.
*/
boolean_t
arc_reclaim_needed(void)
{
return (arc_available_memory() < 0);
}
void
arc_kmem_reap_soon(void)
{
size_t i;
kmem_cache_t *prev_cache = NULL;
kmem_cache_t *prev_data_cache = NULL;
- extern kmem_cache_t *zio_buf_cache[];
- extern kmem_cache_t *zio_data_buf_cache[];
#ifdef _KERNEL
if ((aggsum_compare(&arc_sums.arcstat_meta_used,
arc_meta_limit) >= 0) && zfs_arc_meta_prune) {
/*
* We are exceeding our meta-data cache limit.
* Prune some entries to release holds on meta-data.
*/
arc_prune_async(zfs_arc_meta_prune);
}
#if defined(_ILP32)
/*
* Reclaim unused memory from all kmem caches.
*/
kmem_reap();
#endif
#endif
for (i = 0; i < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; i++) {
#if defined(_ILP32)
/* reach upper limit of cache size on 32-bit */
if (zio_buf_cache[i] == NULL)
break;
#endif
if (zio_buf_cache[i] != prev_cache) {
prev_cache = zio_buf_cache[i];
kmem_cache_reap_now(zio_buf_cache[i]);
}
if (zio_data_buf_cache[i] != prev_data_cache) {
prev_data_cache = zio_data_buf_cache[i];
kmem_cache_reap_now(zio_data_buf_cache[i]);
}
}
kmem_cache_reap_now(buf_cache);
kmem_cache_reap_now(hdr_full_cache);
kmem_cache_reap_now(hdr_l2only_cache);
kmem_cache_reap_now(zfs_btree_leaf_cache);
abd_cache_reap_now();
}
static boolean_t
arc_evict_cb_check(void *arg, zthr_t *zthr)
{
(void) arg, (void) zthr;
#ifdef ZFS_DEBUG
/*
* This is necessary in order to keep the kstat information
* up to date for tools that display kstat data such as the
* mdb ::arc dcmd and the Linux crash utility. These tools
* typically do not call kstat's update function, but simply
* dump out stats from the most recent update. Without
* this call, these commands may show stale stats for the
* anon, mru, mru_ghost, mfu, and mfu_ghost lists. Even
* with this call, the data might be out of date if the
* evict thread hasn't been woken recently; but that should
* suffice. The arc_state_t structures can be queried
* directly if more accurate information is needed.
*/
if (arc_ksp != NULL)
arc_ksp->ks_update(arc_ksp, KSTAT_READ);
#endif
/*
* We have to rely on arc_wait_for_eviction() to tell us when to
* evict, rather than checking if we are overflowing here, so that we
* are sure to not leave arc_wait_for_eviction() waiting on aew_cv.
* If we have become "not overflowing" since arc_wait_for_eviction()
* checked, we need to wake it up. We could broadcast the CV here,
* but arc_wait_for_eviction() may have not yet gone to sleep. We
* would need to use a mutex to ensure that this function doesn't
* broadcast until arc_wait_for_eviction() has gone to sleep (e.g.
* the arc_evict_lock). However, the lock ordering of such a lock
* would necessarily be incorrect with respect to the zthr_lock,
* which is held before this function is called, and is held by
* arc_wait_for_eviction() when it calls zthr_wakeup().
*/
return (arc_evict_needed);
}
/*
* Keep arc_size under arc_c by running arc_evict which evicts data
* from the ARC.
*/
static void
arc_evict_cb(void *arg, zthr_t *zthr)
{
(void) arg, (void) zthr;
uint64_t evicted = 0;
fstrans_cookie_t cookie = spl_fstrans_mark();
/* Evict from cache */
evicted = arc_evict();
/*
* If evicted is zero, we couldn't evict anything
* via arc_evict(). This could be due to hash lock
* collisions, but more likely due to the majority of
* arc buffers being unevictable. Therefore, even if
* arc_size is above arc_c, another pass is unlikely to
* be helpful and could potentially cause us to enter an
* infinite loop. Additionally, zthr_iscancelled() is
* checked here so that if the arc is shutting down, the
* broadcast will wake any remaining arc evict waiters.
*/
mutex_enter(&arc_evict_lock);
arc_evict_needed = !zthr_iscancelled(arc_evict_zthr) &&
evicted > 0 && aggsum_compare(&arc_sums.arcstat_size, arc_c) > 0;
if (!arc_evict_needed) {
/*
* We're either no longer overflowing, or we
* can't evict anything more, so we should wake
* arc_get_data_impl() sooner.
*/
arc_evict_waiter_t *aw;
while ((aw = list_remove_head(&arc_evict_waiters)) != NULL) {
cv_broadcast(&aw->aew_cv);
}
arc_set_need_free();
}
mutex_exit(&arc_evict_lock);
spl_fstrans_unmark(cookie);
}
static boolean_t
arc_reap_cb_check(void *arg, zthr_t *zthr)
{
(void) arg, (void) zthr;
int64_t free_memory = arc_available_memory();
static int reap_cb_check_counter = 0;
/*
* If a kmem reap is already active, don't schedule more. We must
* check for this because kmem_cache_reap_soon() won't actually
* block on the cache being reaped (this is to prevent callers from
* becoming implicitly blocked by a system-wide kmem reap -- which,
* on a system with many, many full magazines, can take minutes).
*/
if (!kmem_cache_reap_active() && free_memory < 0) {
arc_no_grow = B_TRUE;
arc_warm = B_TRUE;
/*
* Wait at least zfs_grow_retry (default 5) seconds
* before considering growing.
*/
arc_growtime = gethrtime() + SEC2NSEC(arc_grow_retry);
return (B_TRUE);
} else if (free_memory < arc_c >> arc_no_grow_shift) {
arc_no_grow = B_TRUE;
} else if (gethrtime() >= arc_growtime) {
arc_no_grow = B_FALSE;
}
/*
* Called unconditionally every 60 seconds to reclaim unused
* zstd compression and decompression context. This is done
* here to avoid the need for an independent thread.
*/
if (!((reap_cb_check_counter++) % 60))
zfs_zstd_cache_reap_now();
return (B_FALSE);
}
/*
* Keep enough free memory in the system by reaping the ARC's kmem
* caches. To cause more slabs to be reapable, we may reduce the
* target size of the cache (arc_c), causing the arc_evict_cb()
* to free more buffers.
*/
static void
arc_reap_cb(void *arg, zthr_t *zthr)
{
(void) arg, (void) zthr;
int64_t free_memory;
fstrans_cookie_t cookie = spl_fstrans_mark();
/*
* Kick off asynchronous kmem_reap()'s of all our caches.
*/
arc_kmem_reap_soon();
/*
* Wait at least arc_kmem_cache_reap_retry_ms between
* arc_kmem_reap_soon() calls. Without this check it is possible to
* end up in a situation where we spend lots of time reaping
* caches, while we're near arc_c_min. Waiting here also gives the
* subsequent free memory check a chance of finding that the
* asynchronous reap has already freed enough memory, and we don't
* need to call arc_reduce_target_size().
*/
delay((hz * arc_kmem_cache_reap_retry_ms + 999) / 1000);
/*
* Reduce the target size as needed to maintain the amount of free
* memory in the system at a fraction of the arc_size (1/128th by
* default). If oversubscribed (free_memory < 0) then reduce the
* target arc_size by the deficit amount plus the fractional
* amount. If free memory is positive but less than the fractional
* amount, reduce by what is needed to hit the fractional amount.
*/
free_memory = arc_available_memory();
int64_t to_free =
(arc_c >> arc_shrink_shift) - free_memory;
if (to_free > 0) {
arc_reduce_target_size(to_free);
}
spl_fstrans_unmark(cookie);
}
#ifdef _KERNEL
/*
* Determine the amount of memory eligible for eviction contained in the
* ARC. All clean data reported by the ghost lists can always be safely
* evicted. Due to arc_c_min, the same does not hold for all clean data
* contained by the regular mru and mfu lists.
*
* In the case of the regular mru and mfu lists, we need to report as
* much clean data as possible, such that evicting that same reported
* data will not bring arc_size below arc_c_min. Thus, in certain
* circumstances, the total amount of clean data in the mru and mfu
* lists might not actually be evictable.
*
* The following two distinct cases are accounted for:
*
* 1. The sum of the amount of dirty data contained by both the mru and
* mfu lists, plus the ARC's other accounting (e.g. the anon list),
* is greater than or equal to arc_c_min.
* (i.e. amount of dirty data >= arc_c_min)
*
* This is the easy case; all clean data contained by the mru and mfu
* lists is evictable. Evicting all clean data can only drop arc_size
* to the amount of dirty data, which is greater than arc_c_min.
*
* 2. The sum of the amount of dirty data contained by both the mru and
* mfu lists, plus the ARC's other accounting (e.g. the anon list),
* is less than arc_c_min.
* (i.e. arc_c_min > amount of dirty data)
*
* 2.1. arc_size is greater than or equal arc_c_min.
* (i.e. arc_size >= arc_c_min > amount of dirty data)
*
* In this case, not all clean data from the regular mru and mfu
* lists is actually evictable; we must leave enough clean data
* to keep arc_size above arc_c_min. Thus, the maximum amount of
* evictable data from the two lists combined, is exactly the
* difference between arc_size and arc_c_min.
*
* 2.2. arc_size is less than arc_c_min
* (i.e. arc_c_min > arc_size > amount of dirty data)
*
* In this case, none of the data contained in the mru and mfu
* lists is evictable, even if it's clean. Since arc_size is
* already below arc_c_min, evicting any more would only
* increase this negative difference.
*/
#endif /* _KERNEL */
/*
* Adapt arc info given the number of bytes we are trying to add and
* the state that we are coming from. This function is only called
* when we are adding new content to the cache.
*/
static void
arc_adapt(int bytes, arc_state_t *state)
{
int mult;
uint64_t arc_p_min = (arc_c >> arc_p_min_shift);
int64_t mrug_size = zfs_refcount_count(&arc_mru_ghost->arcs_size);
int64_t mfug_size = zfs_refcount_count(&arc_mfu_ghost->arcs_size);
ASSERT(bytes > 0);
/*
* Adapt the target size of the MRU list:
* - if we just hit in the MRU ghost list, then increase
* the target size of the MRU list.
* - if we just hit in the MFU ghost list, then increase
* the target size of the MFU list by decreasing the
* target size of the MRU list.
*/
if (state == arc_mru_ghost) {
mult = (mrug_size >= mfug_size) ? 1 : (mfug_size / mrug_size);
if (!zfs_arc_p_dampener_disable)
mult = MIN(mult, 10); /* avoid wild arc_p adjustment */
arc_p = MIN(arc_c - arc_p_min, arc_p + bytes * mult);
} else if (state == arc_mfu_ghost) {
uint64_t delta;
mult = (mfug_size >= mrug_size) ? 1 : (mrug_size / mfug_size);
if (!zfs_arc_p_dampener_disable)
mult = MIN(mult, 10);
delta = MIN(bytes * mult, arc_p);
arc_p = MAX(arc_p_min, arc_p - delta);
}
ASSERT((int64_t)arc_p >= 0);
/*
* Wake reap thread if we do not have any available memory
*/
if (arc_reclaim_needed()) {
zthr_wakeup(arc_reap_zthr);
return;
}
if (arc_no_grow)
return;
if (arc_c >= arc_c_max)
return;
/*
* If we're within (2 * maxblocksize) bytes of the target
* cache size, increment the target cache size
*/
ASSERT3U(arc_c, >=, 2ULL << SPA_MAXBLOCKSHIFT);
if (aggsum_upper_bound(&arc_sums.arcstat_size) >=
arc_c - (2ULL << SPA_MAXBLOCKSHIFT)) {
atomic_add_64(&arc_c, (int64_t)bytes);
if (arc_c > arc_c_max)
arc_c = arc_c_max;
else if (state == arc_anon)
atomic_add_64(&arc_p, (int64_t)bytes);
if (arc_p > arc_c)
arc_p = arc_c;
}
ASSERT((int64_t)arc_p >= 0);
}
/*
* Check if arc_size has grown past our upper threshold, determined by
* zfs_arc_overflow_shift.
*/
static arc_ovf_level_t
arc_is_overflowing(boolean_t use_reserve)
{
/* Always allow at least one block of overflow */
int64_t overflow = MAX(SPA_MAXBLOCKSIZE,
arc_c >> zfs_arc_overflow_shift);
/*
* We just compare the lower bound here for performance reasons. Our
* primary goals are to make sure that the arc never grows without
* bound, and that it can reach its maximum size. This check
* accomplishes both goals. The maximum amount we could run over by is
* 2 * aggsum_borrow_multiplier * NUM_CPUS * the average size of a block
* in the ARC. In practice, that's in the tens of MB, which is low
* enough to be safe.
*/
int64_t over = aggsum_lower_bound(&arc_sums.arcstat_size) -
arc_c - overflow / 2;
if (!use_reserve)
overflow /= 2;
return (over < 0 ? ARC_OVF_NONE :
over < overflow ? ARC_OVF_SOME : ARC_OVF_SEVERE);
}
static abd_t *
arc_get_data_abd(arc_buf_hdr_t *hdr, uint64_t size, void *tag,
int alloc_flags)
{
arc_buf_contents_t type = arc_buf_type(hdr);
arc_get_data_impl(hdr, size, tag, alloc_flags);
if (type == ARC_BUFC_METADATA) {
return (abd_alloc(size, B_TRUE));
} else {
ASSERT(type == ARC_BUFC_DATA);
return (abd_alloc(size, B_FALSE));
}
}
static void *
arc_get_data_buf(arc_buf_hdr_t *hdr, uint64_t size, void *tag)
{
arc_buf_contents_t type = arc_buf_type(hdr);
arc_get_data_impl(hdr, size, tag, ARC_HDR_DO_ADAPT);
if (type == ARC_BUFC_METADATA) {
return (zio_buf_alloc(size));
} else {
ASSERT(type == ARC_BUFC_DATA);
return (zio_data_buf_alloc(size));
}
}
/*
* Wait for the specified amount of data (in bytes) to be evicted from the
* ARC, and for there to be sufficient free memory in the system. Waiting for
* eviction ensures that the memory used by the ARC decreases. Waiting for
* free memory ensures that the system won't run out of free pages, regardless
* of ARC behavior and settings. See arc_lowmem_init().
*/
void
arc_wait_for_eviction(uint64_t amount, boolean_t use_reserve)
{
switch (arc_is_overflowing(use_reserve)) {
case ARC_OVF_NONE:
return;
case ARC_OVF_SOME:
/*
* This is a bit racy without taking arc_evict_lock, but the
* worst that can happen is we either call zthr_wakeup() extra
* time due to race with other thread here, or the set flag
* get cleared by arc_evict_cb(), which is unlikely due to
* big hysteresis, but also not important since at this level
* of overflow the eviction is purely advisory. Same time
* taking the global lock here every time without waiting for
* the actual eviction creates a significant lock contention.
*/
if (!arc_evict_needed) {
arc_evict_needed = B_TRUE;
zthr_wakeup(arc_evict_zthr);
}
return;
case ARC_OVF_SEVERE:
default:
{
arc_evict_waiter_t aw;
list_link_init(&aw.aew_node);
cv_init(&aw.aew_cv, NULL, CV_DEFAULT, NULL);
uint64_t last_count = 0;
mutex_enter(&arc_evict_lock);
if (!list_is_empty(&arc_evict_waiters)) {
arc_evict_waiter_t *last =
list_tail(&arc_evict_waiters);
last_count = last->aew_count;
} else if (!arc_evict_needed) {
arc_evict_needed = B_TRUE;
zthr_wakeup(arc_evict_zthr);
}
/*
* Note, the last waiter's count may be less than
* arc_evict_count if we are low on memory in which
* case arc_evict_state_impl() may have deferred
* wakeups (but still incremented arc_evict_count).
*/
aw.aew_count = MAX(last_count, arc_evict_count) + amount;
list_insert_tail(&arc_evict_waiters, &aw);
arc_set_need_free();
DTRACE_PROBE3(arc__wait__for__eviction,
uint64_t, amount,
uint64_t, arc_evict_count,
uint64_t, aw.aew_count);
/*
* We will be woken up either when arc_evict_count reaches
* aew_count, or when the ARC is no longer overflowing and
* eviction completes.
* In case of "false" wakeup, we will still be on the list.
*/
do {
cv_wait(&aw.aew_cv, &arc_evict_lock);
} while (list_link_active(&aw.aew_node));
mutex_exit(&arc_evict_lock);
cv_destroy(&aw.aew_cv);
}
}
}
/*
* Allocate a block and return it to the caller. If we are hitting the
* hard limit for the cache size, we must sleep, waiting for the eviction
* thread to catch up. If we're past the target size but below the hard
* limit, we'll only signal the reclaim thread and continue on.
*/
static void
arc_get_data_impl(arc_buf_hdr_t *hdr, uint64_t size, void *tag,
int alloc_flags)
{
arc_state_t *state = hdr->b_l1hdr.b_state;
arc_buf_contents_t type = arc_buf_type(hdr);
if (alloc_flags & ARC_HDR_DO_ADAPT)
arc_adapt(size, state);
/*
* If arc_size is currently overflowing, we must be adding data
* faster than we are evicting. To ensure we don't compound the
* problem by adding more data and forcing arc_size to grow even
* further past it's target size, we wait for the eviction thread to
* make some progress. We also wait for there to be sufficient free
* memory in the system, as measured by arc_free_memory().
*
* Specifically, we wait for zfs_arc_eviction_pct percent of the
* requested size to be evicted. This should be more than 100%, to
* ensure that that progress is also made towards getting arc_size
* under arc_c. See the comment above zfs_arc_eviction_pct.
*/
arc_wait_for_eviction(size * zfs_arc_eviction_pct / 100,
alloc_flags & ARC_HDR_USE_RESERVE);
VERIFY3U(hdr->b_type, ==, type);
if (type == ARC_BUFC_METADATA) {
arc_space_consume(size, ARC_SPACE_META);
} else {
arc_space_consume(size, ARC_SPACE_DATA);
}
/*
* Update the state size. Note that ghost states have a
* "ghost size" and so don't need to be updated.
*/
if (!GHOST_STATE(state)) {
(void) zfs_refcount_add_many(&state->arcs_size, size, tag);
/*
* If this is reached via arc_read, the link is
* protected by the hash lock. If reached via
* arc_buf_alloc, the header should not be accessed by
* any other thread. And, if reached via arc_read_done,
* the hash lock will protect it if it's found in the
* hash table; otherwise no other thread should be
* trying to [add|remove]_reference it.
*/
if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
(void) zfs_refcount_add_many(&state->arcs_esize[type],
size, tag);
}
/*
* If we are growing the cache, and we are adding anonymous
* data, and we have outgrown arc_p, update arc_p
*/
if (aggsum_upper_bound(&arc_sums.arcstat_size) < arc_c &&
hdr->b_l1hdr.b_state == arc_anon &&
(zfs_refcount_count(&arc_anon->arcs_size) +
zfs_refcount_count(&arc_mru->arcs_size) > arc_p))
arc_p = MIN(arc_c, arc_p + size);
}
}
static void
arc_free_data_abd(arc_buf_hdr_t *hdr, abd_t *abd, uint64_t size, void *tag)
{
arc_free_data_impl(hdr, size, tag);
abd_free(abd);
}
static void
arc_free_data_buf(arc_buf_hdr_t *hdr, void *buf, uint64_t size, void *tag)
{
arc_buf_contents_t type = arc_buf_type(hdr);
arc_free_data_impl(hdr, size, tag);
if (type == ARC_BUFC_METADATA) {
zio_buf_free(buf, size);
} else {
ASSERT(type == ARC_BUFC_DATA);
zio_data_buf_free(buf, size);
}
}
/*
* Free the arc data buffer.
*/
static void
arc_free_data_impl(arc_buf_hdr_t *hdr, uint64_t size, void *tag)
{
arc_state_t *state = hdr->b_l1hdr.b_state;
arc_buf_contents_t type = arc_buf_type(hdr);
/* protected by hash lock, if in the hash table */
if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
ASSERT(state != arc_anon && state != arc_l2c_only);
(void) zfs_refcount_remove_many(&state->arcs_esize[type],
size, tag);
}
(void) zfs_refcount_remove_many(&state->arcs_size, size, tag);
VERIFY3U(hdr->b_type, ==, type);
if (type == ARC_BUFC_METADATA) {
arc_space_return(size, ARC_SPACE_META);
} else {
ASSERT(type == ARC_BUFC_DATA);
arc_space_return(size, ARC_SPACE_DATA);
}
}
/*
* This routine is called whenever a buffer is accessed.
* NOTE: the hash lock is dropped in this function.
*/
static void
arc_access(arc_buf_hdr_t *hdr, kmutex_t *hash_lock)
{
clock_t now;
ASSERT(MUTEX_HELD(hash_lock));
ASSERT(HDR_HAS_L1HDR(hdr));
if (hdr->b_l1hdr.b_state == arc_anon) {
/*
* This buffer is not in the cache, and does not
* appear in our "ghost" list. Add the new buffer
* to the MRU state.
*/
ASSERT0(hdr->b_l1hdr.b_arc_access);
hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr);
arc_change_state(arc_mru, hdr, hash_lock);
} else if (hdr->b_l1hdr.b_state == arc_mru) {
now = ddi_get_lbolt();
/*
* If this buffer is here because of a prefetch, then either:
* - clear the flag if this is a "referencing" read
* (any subsequent access will bump this into the MFU state).
* or
* - move the buffer to the head of the list if this is
* another prefetch (to make it less likely to be evicted).
*/
if (HDR_PREFETCH(hdr) || HDR_PRESCIENT_PREFETCH(hdr)) {
if (zfs_refcount_count(&hdr->b_l1hdr.b_refcnt) == 0) {
/* link protected by hash lock */
ASSERT(multilist_link_active(
&hdr->b_l1hdr.b_arc_node));
} else {
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_decrement_state(hdr);
arc_hdr_clear_flags(hdr,
ARC_FLAG_PREFETCH |
ARC_FLAG_PRESCIENT_PREFETCH);
hdr->b_l1hdr.b_mru_hits++;
ARCSTAT_BUMP(arcstat_mru_hits);
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_increment_state(hdr);
}
hdr->b_l1hdr.b_arc_access = now;
return;
}
/*
* This buffer has been "accessed" only once so far,
* but it is still in the cache. Move it to the MFU
* state.
*/
if (ddi_time_after(now, hdr->b_l1hdr.b_arc_access +
ARC_MINTIME)) {
/*
* More than 125ms have passed since we
* instantiated this buffer. Move it to the
* most frequently used state.
*/
hdr->b_l1hdr.b_arc_access = now;
DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
arc_change_state(arc_mfu, hdr, hash_lock);
}
hdr->b_l1hdr.b_mru_hits++;
ARCSTAT_BUMP(arcstat_mru_hits);
} else if (hdr->b_l1hdr.b_state == arc_mru_ghost) {
arc_state_t *new_state;
/*
* This buffer has been "accessed" recently, but
* was evicted from the cache. Move it to the
* MFU state.
*/
if (HDR_PREFETCH(hdr) || HDR_PRESCIENT_PREFETCH(hdr)) {
new_state = arc_mru;
if (zfs_refcount_count(&hdr->b_l1hdr.b_refcnt) > 0) {
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_decrement_state(hdr);
arc_hdr_clear_flags(hdr,
ARC_FLAG_PREFETCH |
ARC_FLAG_PRESCIENT_PREFETCH);
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_increment_state(hdr);
}
DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr);
} else {
new_state = arc_mfu;
DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
}
hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
arc_change_state(new_state, hdr, hash_lock);
hdr->b_l1hdr.b_mru_ghost_hits++;
ARCSTAT_BUMP(arcstat_mru_ghost_hits);
} else if (hdr->b_l1hdr.b_state == arc_mfu) {
/*
* This buffer has been accessed more than once and is
* still in the cache. Keep it in the MFU state.
*
* NOTE: an add_reference() that occurred when we did
* the arc_read() will have kicked this off the list.
* If it was a prefetch, we will explicitly move it to
* the head of the list now.
*/
hdr->b_l1hdr.b_mfu_hits++;
ARCSTAT_BUMP(arcstat_mfu_hits);
hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
} else if (hdr->b_l1hdr.b_state == arc_mfu_ghost) {
arc_state_t *new_state = arc_mfu;
/*
* This buffer has been accessed more than once but has
* been evicted from the cache. Move it back to the
* MFU state.
*/
if (HDR_PREFETCH(hdr) || HDR_PRESCIENT_PREFETCH(hdr)) {
/*
* This is a prefetch access...
* move this block back to the MRU state.
*/
new_state = arc_mru;
}
hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
arc_change_state(new_state, hdr, hash_lock);
hdr->b_l1hdr.b_mfu_ghost_hits++;
ARCSTAT_BUMP(arcstat_mfu_ghost_hits);
} else if (hdr->b_l1hdr.b_state == arc_l2c_only) {
/*
* This buffer is on the 2nd Level ARC.
*/
hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
arc_change_state(arc_mfu, hdr, hash_lock);
} else {
cmn_err(CE_PANIC, "invalid arc state 0x%p",
hdr->b_l1hdr.b_state);
}
}
/*
* This routine is called by dbuf_hold() to update the arc_access() state
* which otherwise would be skipped for entries in the dbuf cache.
*/
void
arc_buf_access(arc_buf_t *buf)
{
mutex_enter(&buf->b_evict_lock);
arc_buf_hdr_t *hdr = buf->b_hdr;
/*
* Avoid taking the hash_lock when possible as an optimization.
* The header must be checked again under the hash_lock in order
* to handle the case where it is concurrently being released.
*/
if (hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY(hdr)) {
mutex_exit(&buf->b_evict_lock);
return;
}
kmutex_t *hash_lock = HDR_LOCK(hdr);
mutex_enter(hash_lock);
if (hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY(hdr)) {
mutex_exit(hash_lock);
mutex_exit(&buf->b_evict_lock);
ARCSTAT_BUMP(arcstat_access_skip);
return;
}
mutex_exit(&buf->b_evict_lock);
ASSERT(hdr->b_l1hdr.b_state == arc_mru ||
hdr->b_l1hdr.b_state == arc_mfu);
DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr);
arc_access(hdr, hash_lock);
mutex_exit(hash_lock);
ARCSTAT_BUMP(arcstat_hits);
ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr) && !HDR_PRESCIENT_PREFETCH(hdr),
demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, hits);
}
/* a generic arc_read_done_func_t which you can use */
void
arc_bcopy_func(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp,
arc_buf_t *buf, void *arg)
{
(void) zio, (void) zb, (void) bp;
if (buf == NULL)
return;
bcopy(buf->b_data, arg, arc_buf_size(buf));
arc_buf_destroy(buf, arg);
}
/* a generic arc_read_done_func_t */
void
arc_getbuf_func(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp,
arc_buf_t *buf, void *arg)
{
(void) zb, (void) bp;
arc_buf_t **bufp = arg;
if (buf == NULL) {
ASSERT(zio == NULL || zio->io_error != 0);
*bufp = NULL;
} else {
ASSERT(zio == NULL || zio->io_error == 0);
*bufp = buf;
ASSERT(buf->b_data != NULL);
}
}
static void
arc_hdr_verify(arc_buf_hdr_t *hdr, blkptr_t *bp)
{
if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) {
ASSERT3U(HDR_GET_PSIZE(hdr), ==, 0);
ASSERT3U(arc_hdr_get_compress(hdr), ==, ZIO_COMPRESS_OFF);
} else {
if (HDR_COMPRESSION_ENABLED(hdr)) {
ASSERT3U(arc_hdr_get_compress(hdr), ==,
BP_GET_COMPRESS(bp));
}
ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp));
ASSERT3U(HDR_GET_PSIZE(hdr), ==, BP_GET_PSIZE(bp));
ASSERT3U(!!HDR_PROTECTED(hdr), ==, BP_IS_PROTECTED(bp));
}
}
static void
arc_read_done(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
arc_buf_hdr_t *hdr = zio->io_private;
kmutex_t *hash_lock = NULL;
arc_callback_t *callback_list;
arc_callback_t *acb;
boolean_t freeable = B_FALSE;
/*
* The hdr was inserted into hash-table and removed from lists
* prior to starting I/O. We should find this header, since
* it's in the hash table, and it should be legit since it's
* not possible to evict it during the I/O. The only possible
* reason for it not to be found is if we were freed during the
* read.
*/
if (HDR_IN_HASH_TABLE(hdr)) {
arc_buf_hdr_t *found;
ASSERT3U(hdr->b_birth, ==, BP_PHYSICAL_BIRTH(zio->io_bp));
ASSERT3U(hdr->b_dva.dva_word[0], ==,
BP_IDENTITY(zio->io_bp)->dva_word[0]);
ASSERT3U(hdr->b_dva.dva_word[1], ==,
BP_IDENTITY(zio->io_bp)->dva_word[1]);
found = buf_hash_find(hdr->b_spa, zio->io_bp, &hash_lock);
ASSERT((found == hdr &&
DVA_EQUAL(&hdr->b_dva, BP_IDENTITY(zio->io_bp))) ||
(found == hdr && HDR_L2_READING(hdr)));
ASSERT3P(hash_lock, !=, NULL);
}
if (BP_IS_PROTECTED(bp)) {
hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp);
hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset;
zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt,
hdr->b_crypt_hdr.b_iv);
if (zio->io_error == 0) {
if (BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG) {
void *tmpbuf;
tmpbuf = abd_borrow_buf_copy(zio->io_abd,
sizeof (zil_chain_t));
zio_crypt_decode_mac_zil(tmpbuf,
hdr->b_crypt_hdr.b_mac);
abd_return_buf(zio->io_abd, tmpbuf,
sizeof (zil_chain_t));
} else {
zio_crypt_decode_mac_bp(bp,
hdr->b_crypt_hdr.b_mac);
}
}
}
if (zio->io_error == 0) {
/* byteswap if necessary */
if (BP_SHOULD_BYTESWAP(zio->io_bp)) {
if (BP_GET_LEVEL(zio->io_bp) > 0) {
hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64;
} else {
hdr->b_l1hdr.b_byteswap =
DMU_OT_BYTESWAP(BP_GET_TYPE(zio->io_bp));
}
} else {
hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
}
if (!HDR_L2_READING(hdr)) {
hdr->b_complevel = zio->io_prop.zp_complevel;
}
}
arc_hdr_clear_flags(hdr, ARC_FLAG_L2_EVICTED);
if (l2arc_noprefetch && HDR_PREFETCH(hdr))
arc_hdr_clear_flags(hdr, ARC_FLAG_L2CACHE);
callback_list = hdr->b_l1hdr.b_acb;
ASSERT3P(callback_list, !=, NULL);
if (hash_lock && zio->io_error == 0 &&
hdr->b_l1hdr.b_state == arc_anon) {
/*
* Only call arc_access on anonymous buffers. This is because
* if we've issued an I/O for an evicted buffer, we've already
* called arc_access (to prevent any simultaneous readers from
* getting confused).
*/
arc_access(hdr, hash_lock);
}
/*
* If a read request has a callback (i.e. acb_done is not NULL), then we
* make a buf containing the data according to the parameters which were
* passed in. The implementation of arc_buf_alloc_impl() ensures that we
* aren't needlessly decompressing the data multiple times.
*/
int callback_cnt = 0;
for (acb = callback_list; acb != NULL; acb = acb->acb_next) {
if (!acb->acb_done || acb->acb_nobuf)
continue;
callback_cnt++;
if (zio->io_error != 0)
continue;
int error = arc_buf_alloc_impl(hdr, zio->io_spa,
&acb->acb_zb, acb->acb_private, acb->acb_encrypted,
acb->acb_compressed, acb->acb_noauth, B_TRUE,
&acb->acb_buf);
/*
* Assert non-speculative zios didn't fail because an
* encryption key wasn't loaded
*/
ASSERT((zio->io_flags & ZIO_FLAG_SPECULATIVE) ||
error != EACCES);
/*
* If we failed to decrypt, report an error now (as the zio
* layer would have done if it had done the transforms).
*/
if (error == ECKSUM) {
ASSERT(BP_IS_PROTECTED(bp));
error = SET_ERROR(EIO);
if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) {
spa_log_error(zio->io_spa, &acb->acb_zb);
(void) zfs_ereport_post(
FM_EREPORT_ZFS_AUTHENTICATION,
zio->io_spa, NULL, &acb->acb_zb, zio, 0);
}
}
if (error != 0) {
/*
* Decompression or decryption failed. Set
* io_error so that when we call acb_done
* (below), we will indicate that the read
* failed. Note that in the unusual case
* where one callback is compressed and another
* uncompressed, we will mark all of them
* as failed, even though the uncompressed
* one can't actually fail. In this case,
* the hdr will not be anonymous, because
* if there are multiple callbacks, it's
* because multiple threads found the same
* arc buf in the hash table.
*/
zio->io_error = error;
}
}
/*
* If there are multiple callbacks, we must have the hash lock,
* because the only way for multiple threads to find this hdr is
* in the hash table. This ensures that if there are multiple
* callbacks, the hdr is not anonymous. If it were anonymous,
* we couldn't use arc_buf_destroy() in the error case below.
*/
ASSERT(callback_cnt < 2 || hash_lock != NULL);
hdr->b_l1hdr.b_acb = NULL;
arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
if (callback_cnt == 0)
ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr));
ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt) ||
callback_list != NULL);
if (zio->io_error == 0) {
arc_hdr_verify(hdr, zio->io_bp);
} else {
arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
if (hdr->b_l1hdr.b_state != arc_anon)
arc_change_state(arc_anon, hdr, hash_lock);
if (HDR_IN_HASH_TABLE(hdr))
buf_hash_remove(hdr);
freeable = zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt);
}
/*
* Broadcast before we drop the hash_lock to avoid the possibility
* that the hdr (and hence the cv) might be freed before we get to
* the cv_broadcast().
*/
cv_broadcast(&hdr->b_l1hdr.b_cv);
if (hash_lock != NULL) {
mutex_exit(hash_lock);
} else {
/*
* This block was freed while we waited for the read to
* complete. It has been removed from the hash table and
* moved to the anonymous state (so that it won't show up
* in the cache).
*/
ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
freeable = zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt);
}
/* execute each callback and free its structure */
while ((acb = callback_list) != NULL) {
if (acb->acb_done != NULL) {
if (zio->io_error != 0 && acb->acb_buf != NULL) {
/*
* If arc_buf_alloc_impl() fails during
* decompression, the buf will still be
* allocated, and needs to be freed here.
*/
arc_buf_destroy(acb->acb_buf,
acb->acb_private);
acb->acb_buf = NULL;
}
acb->acb_done(zio, &zio->io_bookmark, zio->io_bp,
acb->acb_buf, acb->acb_private);
}
if (acb->acb_zio_dummy != NULL) {
acb->acb_zio_dummy->io_error = zio->io_error;
zio_nowait(acb->acb_zio_dummy);
}
callback_list = acb->acb_next;
kmem_free(acb, sizeof (arc_callback_t));
}
if (freeable)
arc_hdr_destroy(hdr);
}
/*
* "Read" the block at the specified DVA (in bp) via the
* cache. If the block is found in the cache, invoke the provided
* callback immediately and return. Note that the `zio' parameter
* in the callback will be NULL in this case, since no IO was
* required. If the block is not in the cache pass the read request
* on to the spa with a substitute callback function, so that the
* requested block will be added to the cache.
*
* If a read request arrives for a block that has a read in-progress,
* either wait for the in-progress read to complete (and return the
* results); or, if this is a read with a "done" func, add a record
* to the read to invoke the "done" func when the read completes,
* and return; or just return.
*
* arc_read_done() will invoke all the requested "done" functions
* for readers of this block.
*/
int
arc_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
arc_read_done_func_t *done, void *private, zio_priority_t priority,
int zio_flags, arc_flags_t *arc_flags, const zbookmark_phys_t *zb)
{
arc_buf_hdr_t *hdr = NULL;
kmutex_t *hash_lock = NULL;
zio_t *rzio;
uint64_t guid = spa_load_guid(spa);
boolean_t compressed_read = (zio_flags & ZIO_FLAG_RAW_COMPRESS) != 0;
boolean_t encrypted_read = BP_IS_ENCRYPTED(bp) &&
(zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0;
boolean_t noauth_read = BP_IS_AUTHENTICATED(bp) &&
(zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0;
boolean_t embedded_bp = !!BP_IS_EMBEDDED(bp);
boolean_t no_buf = *arc_flags & ARC_FLAG_NO_BUF;
int rc = 0;
ASSERT(!embedded_bp ||
BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA);
ASSERT(!BP_IS_HOLE(bp));
ASSERT(!BP_IS_REDACTED(bp));
/*
* Normally SPL_FSTRANS will already be set since kernel threads which
* expect to call the DMU interfaces will set it when created. System
* calls are similarly handled by setting/cleaning the bit in the
* registered callback (module/os/.../zfs/zpl_*).
*
* External consumers such as Lustre which call the exported DMU
* interfaces may not have set SPL_FSTRANS. To avoid a deadlock
* on the hash_lock always set and clear the bit.
*/
fstrans_cookie_t cookie = spl_fstrans_mark();
top:
/*
* Verify the block pointer contents are reasonable. This should
* always be the case since the blkptr is protected by a checksum.
* However, if there is damage it's desirable to detect this early
* and treat it as a checksum error. This allows an alternate blkptr
* to be tried when one is available (e.g. ditto blocks).
*/
if (!zfs_blkptr_verify(spa, bp, zio_flags & ZIO_FLAG_CONFIG_WRITER,
BLK_VERIFY_LOG)) {
rc = SET_ERROR(ECKSUM);
goto out;
}
if (!embedded_bp) {
/*
* Embedded BP's have no DVA and require no I/O to "read".
* Create an anonymous arc buf to back it.
*/
hdr = buf_hash_find(guid, bp, &hash_lock);
}
/*
* Determine if we have an L1 cache hit or a cache miss. For simplicity
* we maintain encrypted data separately from compressed / uncompressed
* data. If the user is requesting raw encrypted data and we don't have
* that in the header we will read from disk to guarantee that we can
* get it even if the encryption keys aren't loaded.
*/
if (hdr != NULL && HDR_HAS_L1HDR(hdr) && (HDR_HAS_RABD(hdr) ||
(hdr->b_l1hdr.b_pabd != NULL && !encrypted_read))) {
arc_buf_t *buf = NULL;
*arc_flags |= ARC_FLAG_CACHED;
if (HDR_IO_IN_PROGRESS(hdr)) {
zio_t *head_zio = hdr->b_l1hdr.b_acb->acb_zio_head;
if (*arc_flags & ARC_FLAG_CACHED_ONLY) {
mutex_exit(hash_lock);
ARCSTAT_BUMP(arcstat_cached_only_in_progress);
rc = SET_ERROR(ENOENT);
goto out;
}
ASSERT3P(head_zio, !=, NULL);
if ((hdr->b_flags & ARC_FLAG_PRIO_ASYNC_READ) &&
priority == ZIO_PRIORITY_SYNC_READ) {
/*
* This is a sync read that needs to wait for
* an in-flight async read. Request that the
* zio have its priority upgraded.
*/
zio_change_priority(head_zio, priority);
DTRACE_PROBE1(arc__async__upgrade__sync,
arc_buf_hdr_t *, hdr);
ARCSTAT_BUMP(arcstat_async_upgrade_sync);
}
if (hdr->b_flags & ARC_FLAG_PREDICTIVE_PREFETCH) {
arc_hdr_clear_flags(hdr,
ARC_FLAG_PREDICTIVE_PREFETCH);
}
if (*arc_flags & ARC_FLAG_WAIT) {
cv_wait(&hdr->b_l1hdr.b_cv, hash_lock);
mutex_exit(hash_lock);
goto top;
}
ASSERT(*arc_flags & ARC_FLAG_NOWAIT);
if (done) {
arc_callback_t *acb = NULL;
acb = kmem_zalloc(sizeof (arc_callback_t),
KM_SLEEP);
acb->acb_done = done;
acb->acb_private = private;
acb->acb_compressed = compressed_read;
acb->acb_encrypted = encrypted_read;
acb->acb_noauth = noauth_read;
acb->acb_nobuf = no_buf;
acb->acb_zb = *zb;
if (pio != NULL)
acb->acb_zio_dummy = zio_null(pio,
spa, NULL, NULL, NULL, zio_flags);
ASSERT3P(acb->acb_done, !=, NULL);
acb->acb_zio_head = head_zio;
acb->acb_next = hdr->b_l1hdr.b_acb;
hdr->b_l1hdr.b_acb = acb;
}
mutex_exit(hash_lock);
goto out;
}
ASSERT(hdr->b_l1hdr.b_state == arc_mru ||
hdr->b_l1hdr.b_state == arc_mfu);
if (done && !no_buf) {
if (hdr->b_flags & ARC_FLAG_PREDICTIVE_PREFETCH) {
/*
* This is a demand read which does not have to
* wait for i/o because we did a predictive
* prefetch i/o for it, which has completed.
*/
DTRACE_PROBE1(
arc__demand__hit__predictive__prefetch,
arc_buf_hdr_t *, hdr);
ARCSTAT_BUMP(
arcstat_demand_hit_predictive_prefetch);
arc_hdr_clear_flags(hdr,
ARC_FLAG_PREDICTIVE_PREFETCH);
}
if (hdr->b_flags & ARC_FLAG_PRESCIENT_PREFETCH) {
ARCSTAT_BUMP(
arcstat_demand_hit_prescient_prefetch);
arc_hdr_clear_flags(hdr,
ARC_FLAG_PRESCIENT_PREFETCH);
}
ASSERT(!embedded_bp || !BP_IS_HOLE(bp));
/* Get a buf with the desired data in it. */
rc = arc_buf_alloc_impl(hdr, spa, zb, private,
encrypted_read, compressed_read, noauth_read,
B_TRUE, &buf);
if (rc == ECKSUM) {
/*
* Convert authentication and decryption errors
* to EIO (and generate an ereport if needed)
* before leaving the ARC.
*/
rc = SET_ERROR(EIO);
if ((zio_flags & ZIO_FLAG_SPECULATIVE) == 0) {
spa_log_error(spa, zb);
(void) zfs_ereport_post(
FM_EREPORT_ZFS_AUTHENTICATION,
spa, NULL, zb, NULL, 0);
}
}
if (rc != 0) {
(void) remove_reference(hdr, hash_lock,
private);
arc_buf_destroy_impl(buf);
buf = NULL;
}
/* assert any errors weren't due to unloaded keys */
ASSERT((zio_flags & ZIO_FLAG_SPECULATIVE) ||
rc != EACCES);
} else if (*arc_flags & ARC_FLAG_PREFETCH &&
zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) {
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_decrement_state(hdr);
arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_increment_state(hdr);
}
DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr);
arc_access(hdr, hash_lock);
if (*arc_flags & ARC_FLAG_PRESCIENT_PREFETCH)
arc_hdr_set_flags(hdr, ARC_FLAG_PRESCIENT_PREFETCH);
if (*arc_flags & ARC_FLAG_L2CACHE)
arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);
mutex_exit(hash_lock);
ARCSTAT_BUMP(arcstat_hits);
ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr),
demand, prefetch, !HDR_ISTYPE_METADATA(hdr),
data, metadata, hits);
if (done)
done(NULL, zb, bp, buf, private);
} else {
uint64_t lsize = BP_GET_LSIZE(bp);
uint64_t psize = BP_GET_PSIZE(bp);
arc_callback_t *acb;
vdev_t *vd = NULL;
uint64_t addr = 0;
boolean_t devw = B_FALSE;
uint64_t size;
abd_t *hdr_abd;
int alloc_flags = encrypted_read ? ARC_HDR_ALLOC_RDATA : 0;
if (*arc_flags & ARC_FLAG_CACHED_ONLY) {
rc = SET_ERROR(ENOENT);
if (hash_lock != NULL)
mutex_exit(hash_lock);
goto out;
}
if (hdr == NULL) {
/*
* This block is not in the cache or it has
* embedded data.
*/
arc_buf_hdr_t *exists = NULL;
arc_buf_contents_t type = BP_GET_BUFC_TYPE(bp);
hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize,
BP_IS_PROTECTED(bp), BP_GET_COMPRESS(bp), 0, type);
if (!embedded_bp) {
hdr->b_dva = *BP_IDENTITY(bp);
hdr->b_birth = BP_PHYSICAL_BIRTH(bp);
exists = buf_hash_insert(hdr, &hash_lock);
}
if (exists != NULL) {
/* somebody beat us to the hash insert */
mutex_exit(hash_lock);
buf_discard_identity(hdr);
arc_hdr_destroy(hdr);
goto top; /* restart the IO request */
}
alloc_flags |= ARC_HDR_DO_ADAPT;
} else {
/*
* This block is in the ghost cache or encrypted data
* was requested and we didn't have it. If it was
* L2-only (and thus didn't have an L1 hdr),
* we realloc the header to add an L1 hdr.
*/
if (!HDR_HAS_L1HDR(hdr)) {
hdr = arc_hdr_realloc(hdr, hdr_l2only_cache,
hdr_full_cache);
}
if (GHOST_STATE(hdr->b_l1hdr.b_state)) {
ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
ASSERT(!HDR_HAS_RABD(hdr));
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT0(zfs_refcount_count(
&hdr->b_l1hdr.b_refcnt));
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
} else if (HDR_IO_IN_PROGRESS(hdr)) {
/*
* If this header already had an IO in progress
* and we are performing another IO to fetch
* encrypted data we must wait until the first
* IO completes so as not to confuse
* arc_read_done(). This should be very rare
* and so the performance impact shouldn't
* matter.
*/
cv_wait(&hdr->b_l1hdr.b_cv, hash_lock);
mutex_exit(hash_lock);
goto top;
}
/*
* This is a delicate dance that we play here.
* This hdr might be in the ghost list so we access
* it to move it out of the ghost list before we
* initiate the read. If it's a prefetch then
* it won't have a callback so we'll remove the
* reference that arc_buf_alloc_impl() created. We
* do this after we've called arc_access() to
* avoid hitting an assert in remove_reference().
*/
arc_adapt(arc_hdr_size(hdr), hdr->b_l1hdr.b_state);
arc_access(hdr, hash_lock);
}
arc_hdr_alloc_abd(hdr, alloc_flags);
if (encrypted_read) {
ASSERT(HDR_HAS_RABD(hdr));
size = HDR_GET_PSIZE(hdr);
hdr_abd = hdr->b_crypt_hdr.b_rabd;
zio_flags |= ZIO_FLAG_RAW;
} else {
ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
size = arc_hdr_size(hdr);
hdr_abd = hdr->b_l1hdr.b_pabd;
if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) {
zio_flags |= ZIO_FLAG_RAW_COMPRESS;
}
/*
* For authenticated bp's, we do not ask the ZIO layer
* to authenticate them since this will cause the entire
* IO to fail if the key isn't loaded. Instead, we
* defer authentication until arc_buf_fill(), which will
* verify the data when the key is available.
*/
if (BP_IS_AUTHENTICATED(bp))
zio_flags |= ZIO_FLAG_RAW_ENCRYPT;
}
if (*arc_flags & ARC_FLAG_PREFETCH &&
zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) {
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_decrement_state(hdr);
arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
if (HDR_HAS_L2HDR(hdr))
l2arc_hdr_arcstats_increment_state(hdr);
}
if (*arc_flags & ARC_FLAG_PRESCIENT_PREFETCH)
arc_hdr_set_flags(hdr, ARC_FLAG_PRESCIENT_PREFETCH);
if (*arc_flags & ARC_FLAG_L2CACHE)
arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);
if (BP_IS_AUTHENTICATED(bp))
arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH);
if (BP_GET_LEVEL(bp) > 0)
arc_hdr_set_flags(hdr, ARC_FLAG_INDIRECT);
if (*arc_flags & ARC_FLAG_PREDICTIVE_PREFETCH)
arc_hdr_set_flags(hdr, ARC_FLAG_PREDICTIVE_PREFETCH);
ASSERT(!GHOST_STATE(hdr->b_l1hdr.b_state));
acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP);
acb->acb_done = done;
acb->acb_private = private;
acb->acb_compressed = compressed_read;
acb->acb_encrypted = encrypted_read;
acb->acb_noauth = noauth_read;
acb->acb_zb = *zb;
ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
hdr->b_l1hdr.b_acb = acb;
arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
if (HDR_HAS_L2HDR(hdr) &&
(vd = hdr->b_l2hdr.b_dev->l2ad_vdev) != NULL) {
devw = hdr->b_l2hdr.b_dev->l2ad_writing;
addr = hdr->b_l2hdr.b_daddr;
/*
* Lock out L2ARC device removal.
*/
if (vdev_is_dead(vd) ||
!spa_config_tryenter(spa, SCL_L2ARC, vd, RW_READER))
vd = NULL;
}
/*
* We count both async reads and scrub IOs as asynchronous so
* that both can be upgraded in the event of a cache hit while
* the read IO is still in-flight.
*/
if (priority == ZIO_PRIORITY_ASYNC_READ ||
priority == ZIO_PRIORITY_SCRUB)
arc_hdr_set_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ);
else
arc_hdr_clear_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ);
/*
* At this point, we have a level 1 cache miss or a blkptr
* with embedded data. Try again in L2ARC if possible.
*/
ASSERT3U(HDR_GET_LSIZE(hdr), ==, lsize);
/*
* Skip ARC stat bump for block pointers with embedded
* data. The data are read from the blkptr itself via
* decode_embedded_bp_compressed().
*/
if (!embedded_bp) {
DTRACE_PROBE4(arc__miss, arc_buf_hdr_t *, hdr,
blkptr_t *, bp, uint64_t, lsize,
zbookmark_phys_t *, zb);
ARCSTAT_BUMP(arcstat_misses);
ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr),
demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data,
metadata, misses);
zfs_racct_read(size, 1);
}
/* Check if the spa even has l2 configured */
const boolean_t spa_has_l2 = l2arc_ndev != 0 &&
spa->spa_l2cache.sav_count > 0;
if (vd != NULL && spa_has_l2 && !(l2arc_norw && devw)) {
/*
* Read from the L2ARC if the following are true:
* 1. The L2ARC vdev was previously cached.
* 2. This buffer still has L2ARC metadata.
* 3. This buffer isn't currently writing to the L2ARC.
* 4. The L2ARC entry wasn't evicted, which may
* also have invalidated the vdev.
* 5. This isn't prefetch or l2arc_noprefetch is 0.
*/
if (HDR_HAS_L2HDR(hdr) &&
!HDR_L2_WRITING(hdr) && !HDR_L2_EVICTED(hdr) &&
!(l2arc_noprefetch && HDR_PREFETCH(hdr))) {
l2arc_read_callback_t *cb;
abd_t *abd;
uint64_t asize;
DTRACE_PROBE1(l2arc__hit, arc_buf_hdr_t *, hdr);
ARCSTAT_BUMP(arcstat_l2_hits);
hdr->b_l2hdr.b_hits++;
cb = kmem_zalloc(sizeof (l2arc_read_callback_t),
KM_SLEEP);
cb->l2rcb_hdr = hdr;
cb->l2rcb_bp = *bp;
cb->l2rcb_zb = *zb;
cb->l2rcb_flags = zio_flags;
/*
* When Compressed ARC is disabled, but the
* L2ARC block is compressed, arc_hdr_size()
* will have returned LSIZE rather than PSIZE.
*/
if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
!HDR_COMPRESSION_ENABLED(hdr) &&
HDR_GET_PSIZE(hdr) != 0) {
size = HDR_GET_PSIZE(hdr);
}
asize = vdev_psize_to_asize(vd, size);
if (asize != size) {
abd = abd_alloc_for_io(asize,
HDR_ISTYPE_METADATA(hdr));
cb->l2rcb_abd = abd;
} else {
abd = hdr_abd;
}
ASSERT(addr >= VDEV_LABEL_START_SIZE &&
addr + asize <= vd->vdev_psize -
VDEV_LABEL_END_SIZE);
/*
* l2arc read. The SCL_L2ARC lock will be
* released by l2arc_read_done().
* Issue a null zio if the underlying buffer
* was squashed to zero size by compression.
*/
ASSERT3U(arc_hdr_get_compress(hdr), !=,
ZIO_COMPRESS_EMPTY);
rzio = zio_read_phys(pio, vd, addr,
asize, abd,
ZIO_CHECKSUM_OFF,
l2arc_read_done, cb, priority,
zio_flags | ZIO_FLAG_DONT_CACHE |
ZIO_FLAG_CANFAIL |
ZIO_FLAG_DONT_PROPAGATE |
ZIO_FLAG_DONT_RETRY, B_FALSE);
acb->acb_zio_head = rzio;
if (hash_lock != NULL)
mutex_exit(hash_lock);
DTRACE_PROBE2(l2arc__read, vdev_t *, vd,
zio_t *, rzio);
ARCSTAT_INCR(arcstat_l2_read_bytes,
HDR_GET_PSIZE(hdr));
if (*arc_flags & ARC_FLAG_NOWAIT) {
zio_nowait(rzio);
goto out;
}
ASSERT(*arc_flags & ARC_FLAG_WAIT);
if (zio_wait(rzio) == 0)
goto out;
/* l2arc read error; goto zio_read() */
if (hash_lock != NULL)
mutex_enter(hash_lock);
} else {
DTRACE_PROBE1(l2arc__miss,
arc_buf_hdr_t *, hdr);
ARCSTAT_BUMP(arcstat_l2_misses);
if (HDR_L2_WRITING(hdr))
ARCSTAT_BUMP(arcstat_l2_rw_clash);
spa_config_exit(spa, SCL_L2ARC, vd);
}
} else {
if (vd != NULL)
spa_config_exit(spa, SCL_L2ARC, vd);
/*
* Only a spa with l2 should contribute to l2
* miss stats. (Including the case of having a
* faulted cache device - that's also a miss.)
*/
if (spa_has_l2) {
/*
* Skip ARC stat bump for block pointers with
* embedded data. The data are read from the
* blkptr itself via
* decode_embedded_bp_compressed().
*/
if (!embedded_bp) {
DTRACE_PROBE1(l2arc__miss,
arc_buf_hdr_t *, hdr);
ARCSTAT_BUMP(arcstat_l2_misses);
}
}
}
rzio = zio_read(pio, spa, bp, hdr_abd, size,
arc_read_done, hdr, priority, zio_flags, zb);
acb->acb_zio_head = rzio;
if (hash_lock != NULL)
mutex_exit(hash_lock);
if (*arc_flags & ARC_FLAG_WAIT) {
rc = zio_wait(rzio);
goto out;
}
ASSERT(*arc_flags & ARC_FLAG_NOWAIT);
zio_nowait(rzio);
}
out:
/* embedded bps don't actually go to disk */
if (!embedded_bp)
spa_read_history_add(spa, zb, *arc_flags);
spl_fstrans_unmark(cookie);
return (rc);
}
arc_prune_t *
arc_add_prune_callback(arc_prune_func_t *func, void *private)
{
arc_prune_t *p;
p = kmem_alloc(sizeof (*p), KM_SLEEP);
p->p_pfunc = func;
p->p_private = private;
list_link_init(&p->p_node);
zfs_refcount_create(&p->p_refcnt);
mutex_enter(&arc_prune_mtx);
zfs_refcount_add(&p->p_refcnt, &arc_prune_list);
list_insert_head(&arc_prune_list, p);
mutex_exit(&arc_prune_mtx);
return (p);
}
void
arc_remove_prune_callback(arc_prune_t *p)
{
boolean_t wait = B_FALSE;
mutex_enter(&arc_prune_mtx);
list_remove(&arc_prune_list, p);
if (zfs_refcount_remove(&p->p_refcnt, &arc_prune_list) > 0)
wait = B_TRUE;
mutex_exit(&arc_prune_mtx);
/* wait for arc_prune_task to finish */
if (wait)
taskq_wait_outstanding(arc_prune_taskq, 0);
ASSERT0(zfs_refcount_count(&p->p_refcnt));
zfs_refcount_destroy(&p->p_refcnt);
kmem_free(p, sizeof (*p));
}
/*
* Notify the arc that a block was freed, and thus will never be used again.
*/
void
arc_freed(spa_t *spa, const blkptr_t *bp)
{
arc_buf_hdr_t *hdr;
kmutex_t *hash_lock;
uint64_t guid = spa_load_guid(spa);
ASSERT(!BP_IS_EMBEDDED(bp));
hdr = buf_hash_find(guid, bp, &hash_lock);
if (hdr == NULL)
return;
/*
* We might be trying to free a block that is still doing I/O
* (i.e. prefetch) or has a reference (i.e. a dedup-ed,
* dmu_sync-ed block). If this block is being prefetched, then it
* would still have the ARC_FLAG_IO_IN_PROGRESS flag set on the hdr
* until the I/O completes. A block may also have a reference if it is
* part of a dedup-ed, dmu_synced write. The dmu_sync() function would
* have written the new block to its final resting place on disk but
* without the dedup flag set. This would have left the hdr in the MRU
* state and discoverable. When the txg finally syncs it detects that
* the block was overridden in open context and issues an override I/O.
* Since this is a dedup block, the override I/O will determine if the
* block is already in the DDT. If so, then it will replace the io_bp
* with the bp from the DDT and allow the I/O to finish. When the I/O
* reaches the done callback, dbuf_write_override_done, it will
* check to see if the io_bp and io_bp_override are identical.
* If they are not, then it indicates that the bp was replaced with
* the bp in the DDT and the override bp is freed. This allows
* us to arrive here with a reference on a block that is being
* freed. So if we have an I/O in progress, or a reference to
* this hdr, then we don't destroy the hdr.
*/
if (!HDR_HAS_L1HDR(hdr) || (!HDR_IO_IN_PROGRESS(hdr) &&
zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt))) {
arc_change_state(arc_anon, hdr, hash_lock);
arc_hdr_destroy(hdr);
mutex_exit(hash_lock);
} else {
mutex_exit(hash_lock);
}
}
/*
* Release this buffer from the cache, making it an anonymous buffer. This
* must be done after a read and prior to modifying the buffer contents.
* If the buffer has more than one reference, we must make
* a new hdr for the buffer.
*/
void
arc_release(arc_buf_t *buf, void *tag)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
/*
* It would be nice to assert that if its DMU metadata (level >
* 0 || it's the dnode file), then it must be syncing context.
* But we don't know that information at this level.
*/
mutex_enter(&buf->b_evict_lock);
ASSERT(HDR_HAS_L1HDR(hdr));
/*
* We don't grab the hash lock prior to this check, because if
* the buffer's header is in the arc_anon state, it won't be
* linked into the hash table.
*/
if (hdr->b_l1hdr.b_state == arc_anon) {
mutex_exit(&buf->b_evict_lock);
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT(!HDR_IN_HASH_TABLE(hdr));
ASSERT(!HDR_HAS_L2HDR(hdr));
ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1);
ASSERT3S(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt), ==, 1);
ASSERT(!list_link_active(&hdr->b_l1hdr.b_arc_node));
hdr->b_l1hdr.b_arc_access = 0;
/*
* If the buf is being overridden then it may already
* have a hdr that is not empty.
*/
buf_discard_identity(hdr);
arc_buf_thaw(buf);
return;
}
kmutex_t *hash_lock = HDR_LOCK(hdr);
mutex_enter(hash_lock);
/*
* This assignment is only valid as long as the hash_lock is
* held, we must be careful not to reference state or the
* b_state field after dropping the lock.
*/
arc_state_t *state = hdr->b_l1hdr.b_state;
ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
ASSERT3P(state, !=, arc_anon);
/* this buffer is not on any list */
ASSERT3S(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt), >, 0);
if (HDR_HAS_L2HDR(hdr)) {
mutex_enter(&hdr->b_l2hdr.b_dev->l2ad_mtx);
/*
* We have to recheck this conditional again now that
* we're holding the l2ad_mtx to prevent a race with
* another thread which might be concurrently calling
* l2arc_evict(). In that case, l2arc_evict() might have
* destroyed the header's L2 portion as we were waiting
* to acquire the l2ad_mtx.
*/
if (HDR_HAS_L2HDR(hdr))
arc_hdr_l2hdr_destroy(hdr);
mutex_exit(&hdr->b_l2hdr.b_dev->l2ad_mtx);
}
/*
* Do we have more than one buf?
*/
if (hdr->b_l1hdr.b_bufcnt > 1) {
arc_buf_hdr_t *nhdr;
uint64_t spa = hdr->b_spa;
uint64_t psize = HDR_GET_PSIZE(hdr);
uint64_t lsize = HDR_GET_LSIZE(hdr);
boolean_t protected = HDR_PROTECTED(hdr);
enum zio_compress compress = arc_hdr_get_compress(hdr);
arc_buf_contents_t type = arc_buf_type(hdr);
VERIFY3U(hdr->b_type, ==, type);
ASSERT(hdr->b_l1hdr.b_buf != buf || buf->b_next != NULL);
(void) remove_reference(hdr, hash_lock, tag);
if (arc_buf_is_shared(buf) && !ARC_BUF_COMPRESSED(buf)) {
ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf);
ASSERT(ARC_BUF_LAST(buf));
}
/*
* Pull the data off of this hdr and attach it to
* a new anonymous hdr. Also find the last buffer
* in the hdr's buffer list.
*/
arc_buf_t *lastbuf = arc_buf_remove(hdr, buf);
ASSERT3P(lastbuf, !=, NULL);
/*
* If the current arc_buf_t and the hdr are sharing their data
* buffer, then we must stop sharing that block.
*/
if (arc_buf_is_shared(buf)) {
ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf);
VERIFY(!arc_buf_is_shared(lastbuf));
/*
* First, sever the block sharing relationship between
* buf and the arc_buf_hdr_t.
*/
arc_unshare_buf(hdr, buf);
/*
* Now we need to recreate the hdr's b_pabd. Since we
* have lastbuf handy, we try to share with it, but if
* we can't then we allocate a new b_pabd and copy the
* data from buf into it.
*/
if (arc_can_share(hdr, lastbuf)) {
arc_share_buf(hdr, lastbuf);
} else {
arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT);
abd_copy_from_buf(hdr->b_l1hdr.b_pabd,
buf->b_data, psize);
}
VERIFY3P(lastbuf->b_data, !=, NULL);
} else if (HDR_SHARED_DATA(hdr)) {
/*
* Uncompressed shared buffers are always at the end
* of the list. Compressed buffers don't have the
* same requirements. This makes it hard to
* simply assert that the lastbuf is shared so
* we rely on the hdr's compression flags to determine
* if we have a compressed, shared buffer.
*/
ASSERT(arc_buf_is_shared(lastbuf) ||
arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF);
ASSERT(!ARC_BUF_SHARED(buf));
}
ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr));
ASSERT3P(state, !=, arc_l2c_only);
(void) zfs_refcount_remove_many(&state->arcs_size,
arc_buf_size(buf), buf);
if (zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) {
ASSERT3P(state, !=, arc_l2c_only);
(void) zfs_refcount_remove_many(
&state->arcs_esize[type],
arc_buf_size(buf), buf);
}
hdr->b_l1hdr.b_bufcnt -= 1;
if (ARC_BUF_ENCRYPTED(buf))
hdr->b_crypt_hdr.b_ebufcnt -= 1;
arc_cksum_verify(buf);
arc_buf_unwatch(buf);
/* if this is the last uncompressed buf free the checksum */
if (!arc_hdr_has_uncompressed_buf(hdr))
arc_cksum_free(hdr);
mutex_exit(hash_lock);
/*
* Allocate a new hdr. The new hdr will contain a b_pabd
* buffer which will be freed in arc_write().
*/
nhdr = arc_hdr_alloc(spa, psize, lsize, protected,
compress, hdr->b_complevel, type);
ASSERT3P(nhdr->b_l1hdr.b_buf, ==, NULL);
ASSERT0(nhdr->b_l1hdr.b_bufcnt);
ASSERT0(zfs_refcount_count(&nhdr->b_l1hdr.b_refcnt));
VERIFY3U(nhdr->b_type, ==, type);
ASSERT(!HDR_SHARED_DATA(nhdr));
nhdr->b_l1hdr.b_buf = buf;
nhdr->b_l1hdr.b_bufcnt = 1;
if (ARC_BUF_ENCRYPTED(buf))
nhdr->b_crypt_hdr.b_ebufcnt = 1;
(void) zfs_refcount_add(&nhdr->b_l1hdr.b_refcnt, tag);
buf->b_hdr = nhdr;
mutex_exit(&buf->b_evict_lock);
(void) zfs_refcount_add_many(&arc_anon->arcs_size,
arc_buf_size(buf), buf);
} else {
mutex_exit(&buf->b_evict_lock);
ASSERT(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt) == 1);
/* protected by hash lock, or hdr is on arc_anon */
ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
hdr->b_l1hdr.b_mru_hits = 0;
hdr->b_l1hdr.b_mru_ghost_hits = 0;
hdr->b_l1hdr.b_mfu_hits = 0;
hdr->b_l1hdr.b_mfu_ghost_hits = 0;
arc_change_state(arc_anon, hdr, hash_lock);
hdr->b_l1hdr.b_arc_access = 0;
mutex_exit(hash_lock);
buf_discard_identity(hdr);
arc_buf_thaw(buf);
}
}
int
arc_released(arc_buf_t *buf)
{
int released;
mutex_enter(&buf->b_evict_lock);
released = (buf->b_data != NULL &&
buf->b_hdr->b_l1hdr.b_state == arc_anon);
mutex_exit(&buf->b_evict_lock);
return (released);
}
#ifdef ZFS_DEBUG
int
arc_referenced(arc_buf_t *buf)
{
int referenced;
mutex_enter(&buf->b_evict_lock);
referenced = (zfs_refcount_count(&buf->b_hdr->b_l1hdr.b_refcnt));
mutex_exit(&buf->b_evict_lock);
return (referenced);
}
#endif
static void
arc_write_ready(zio_t *zio)
{
arc_write_callback_t *callback = zio->io_private;
arc_buf_t *buf = callback->awcb_buf;
arc_buf_hdr_t *hdr = buf->b_hdr;
blkptr_t *bp = zio->io_bp;
uint64_t psize = BP_IS_HOLE(bp) ? 0 : BP_GET_PSIZE(bp);
fstrans_cookie_t cookie = spl_fstrans_mark();
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT(!zfs_refcount_is_zero(&buf->b_hdr->b_l1hdr.b_refcnt));
ASSERT(hdr->b_l1hdr.b_bufcnt > 0);
/*
* If we're reexecuting this zio because the pool suspended, then
* cleanup any state that was previously set the first time the
* callback was invoked.
*/
if (zio->io_flags & ZIO_FLAG_REEXECUTED) {
arc_cksum_free(hdr);
arc_buf_unwatch(buf);
if (hdr->b_l1hdr.b_pabd != NULL) {
if (arc_buf_is_shared(buf)) {
arc_unshare_buf(hdr, buf);
} else {
arc_hdr_free_abd(hdr, B_FALSE);
}
}
if (HDR_HAS_RABD(hdr))
arc_hdr_free_abd(hdr, B_TRUE);
}
ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
ASSERT(!HDR_HAS_RABD(hdr));
ASSERT(!HDR_SHARED_DATA(hdr));
ASSERT(!arc_buf_is_shared(buf));
callback->awcb_ready(zio, buf, callback->awcb_private);
if (HDR_IO_IN_PROGRESS(hdr))
ASSERT(zio->io_flags & ZIO_FLAG_REEXECUTED);
arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
if (BP_IS_PROTECTED(bp) != !!HDR_PROTECTED(hdr))
hdr = arc_hdr_realloc_crypt(hdr, BP_IS_PROTECTED(bp));
if (BP_IS_PROTECTED(bp)) {
/* ZIL blocks are written through zio_rewrite */
ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG);
ASSERT(HDR_PROTECTED(hdr));
if (BP_SHOULD_BYTESWAP(bp)) {
if (BP_GET_LEVEL(bp) > 0) {
hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64;
} else {
hdr->b_l1hdr.b_byteswap =
DMU_OT_BYTESWAP(BP_GET_TYPE(bp));
}
} else {
hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
}
hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp);
hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset;
zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt,
hdr->b_crypt_hdr.b_iv);
zio_crypt_decode_mac_bp(bp, hdr->b_crypt_hdr.b_mac);
}
/*
* If this block was written for raw encryption but the zio layer
* ended up only authenticating it, adjust the buffer flags now.
*/
if (BP_IS_AUTHENTICATED(bp) && ARC_BUF_ENCRYPTED(buf)) {
arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH);
buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED;
if (BP_GET_COMPRESS(bp) == ZIO_COMPRESS_OFF)
buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
} else if (BP_IS_HOLE(bp) && ARC_BUF_ENCRYPTED(buf)) {
buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED;
buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
}
/* this must be done after the buffer flags are adjusted */
arc_cksum_compute(buf);
enum zio_compress compress;
if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) {
compress = ZIO_COMPRESS_OFF;
} else {
ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp));
compress = BP_GET_COMPRESS(bp);
}
HDR_SET_PSIZE(hdr, psize);
arc_hdr_set_compress(hdr, compress);
hdr->b_complevel = zio->io_prop.zp_complevel;
if (zio->io_error != 0 || psize == 0)
goto out;
/*
* Fill the hdr with data. If the buffer is encrypted we have no choice
* but to copy the data into b_radb. If the hdr is compressed, the data
* we want is available from the zio, otherwise we can take it from
* the buf.
*
* We might be able to share the buf's data with the hdr here. However,
* doing so would cause the ARC to be full of linear ABDs if we write a
* lot of shareable data. As a compromise, we check whether scattered
* ABDs are allowed, and assume that if they are then the user wants
* the ARC to be primarily filled with them regardless of the data being
* written. Therefore, if they're allowed then we allocate one and copy
* the data into it; otherwise, we share the data directly if we can.
*/
if (ARC_BUF_ENCRYPTED(buf)) {
ASSERT3U(psize, >, 0);
ASSERT(ARC_BUF_COMPRESSED(buf));
arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT | ARC_HDR_ALLOC_RDATA |
ARC_HDR_USE_RESERVE);
abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize);
} else if (!abd_size_alloc_linear(arc_buf_size(buf)) ||
!arc_can_share(hdr, buf)) {
/*
* Ideally, we would always copy the io_abd into b_pabd, but the
* user may have disabled compressed ARC, thus we must check the
* hdr's compression setting rather than the io_bp's.
*/
if (BP_IS_ENCRYPTED(bp)) {
ASSERT3U(psize, >, 0);
arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT |
ARC_HDR_ALLOC_RDATA | ARC_HDR_USE_RESERVE);
abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize);
} else if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF &&
!ARC_BUF_COMPRESSED(buf)) {
ASSERT3U(psize, >, 0);
arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT |
ARC_HDR_USE_RESERVE);
abd_copy(hdr->b_l1hdr.b_pabd, zio->io_abd, psize);
} else {
ASSERT3U(zio->io_orig_size, ==, arc_hdr_size(hdr));
arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT |
ARC_HDR_USE_RESERVE);
abd_copy_from_buf(hdr->b_l1hdr.b_pabd, buf->b_data,
arc_buf_size(buf));
}
} else {
ASSERT3P(buf->b_data, ==, abd_to_buf(zio->io_orig_abd));
ASSERT3U(zio->io_orig_size, ==, arc_buf_size(buf));
ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1);
arc_share_buf(hdr, buf);
}
out:
arc_hdr_verify(hdr, bp);
spl_fstrans_unmark(cookie);
}
static void
arc_write_children_ready(zio_t *zio)
{
arc_write_callback_t *callback = zio->io_private;
arc_buf_t *buf = callback->awcb_buf;
callback->awcb_children_ready(zio, buf, callback->awcb_private);
}
/*
* The SPA calls this callback for each physical write that happens on behalf
* of a logical write. See the comment in dbuf_write_physdone() for details.
*/
static void
arc_write_physdone(zio_t *zio)
{
arc_write_callback_t *cb = zio->io_private;
if (cb->awcb_physdone != NULL)
cb->awcb_physdone(zio, cb->awcb_buf, cb->awcb_private);
}
static void
arc_write_done(zio_t *zio)
{
arc_write_callback_t *callback = zio->io_private;
arc_buf_t *buf = callback->awcb_buf;
arc_buf_hdr_t *hdr = buf->b_hdr;
ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
if (zio->io_error == 0) {
arc_hdr_verify(hdr, zio->io_bp);
if (BP_IS_HOLE(zio->io_bp) || BP_IS_EMBEDDED(zio->io_bp)) {
buf_discard_identity(hdr);
} else {
hdr->b_dva = *BP_IDENTITY(zio->io_bp);
hdr->b_birth = BP_PHYSICAL_BIRTH(zio->io_bp);
}
} else {
ASSERT(HDR_EMPTY(hdr));
}
/*
* If the block to be written was all-zero or compressed enough to be
* embedded in the BP, no write was performed so there will be no
* dva/birth/checksum. The buffer must therefore remain anonymous
* (and uncached).
*/
if (!HDR_EMPTY(hdr)) {
arc_buf_hdr_t *exists;
kmutex_t *hash_lock;
ASSERT3U(zio->io_error, ==, 0);
arc_cksum_verify(buf);
exists = buf_hash_insert(hdr, &hash_lock);
if (exists != NULL) {
/*
* This can only happen if we overwrite for
* sync-to-convergence, because we remove
* buffers from the hash table when we arc_free().
*/
if (zio->io_flags & ZIO_FLAG_IO_REWRITE) {
if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp))
panic("bad overwrite, hdr=%p exists=%p",
(void *)hdr, (void *)exists);
ASSERT(zfs_refcount_is_zero(
&exists->b_l1hdr.b_refcnt));
arc_change_state(arc_anon, exists, hash_lock);
arc_hdr_destroy(exists);
mutex_exit(hash_lock);
exists = buf_hash_insert(hdr, &hash_lock);
ASSERT3P(exists, ==, NULL);
} else if (zio->io_flags & ZIO_FLAG_NOPWRITE) {
/* nopwrite */
ASSERT(zio->io_prop.zp_nopwrite);
if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp))
panic("bad nopwrite, hdr=%p exists=%p",
(void *)hdr, (void *)exists);
} else {
/* Dedup */
ASSERT(hdr->b_l1hdr.b_bufcnt == 1);
ASSERT(hdr->b_l1hdr.b_state == arc_anon);
ASSERT(BP_GET_DEDUP(zio->io_bp));
ASSERT(BP_GET_LEVEL(zio->io_bp) == 0);
}
}
arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
/* if it's not anon, we are doing a scrub */
if (exists == NULL && hdr->b_l1hdr.b_state == arc_anon)
arc_access(hdr, hash_lock);
mutex_exit(hash_lock);
} else {
arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
}
ASSERT(!zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
callback->awcb_done(zio, buf, callback->awcb_private);
abd_free(zio->io_abd);
kmem_free(callback, sizeof (arc_write_callback_t));
}
zio_t *
arc_write(zio_t *pio, spa_t *spa, uint64_t txg,
blkptr_t *bp, arc_buf_t *buf, boolean_t l2arc,
const zio_prop_t *zp, arc_write_done_func_t *ready,
arc_write_done_func_t *children_ready, arc_write_done_func_t *physdone,
arc_write_done_func_t *done, void *private, zio_priority_t priority,
int zio_flags, const zbookmark_phys_t *zb)
{
arc_buf_hdr_t *hdr = buf->b_hdr;
arc_write_callback_t *callback;
zio_t *zio;
zio_prop_t localprop = *zp;
ASSERT3P(ready, !=, NULL);
ASSERT3P(done, !=, NULL);
ASSERT(!HDR_IO_ERROR(hdr));
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
ASSERT3U(hdr->b_l1hdr.b_bufcnt, >, 0);
if (l2arc)
arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);
if (ARC_BUF_ENCRYPTED(buf)) {
ASSERT(ARC_BUF_COMPRESSED(buf));
localprop.zp_encrypt = B_TRUE;
localprop.zp_compress = HDR_GET_COMPRESS(hdr);
localprop.zp_complevel = hdr->b_complevel;
localprop.zp_byteorder =
(hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ?
ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER;
bcopy(hdr->b_crypt_hdr.b_salt, localprop.zp_salt,
ZIO_DATA_SALT_LEN);
bcopy(hdr->b_crypt_hdr.b_iv, localprop.zp_iv,
ZIO_DATA_IV_LEN);
bcopy(hdr->b_crypt_hdr.b_mac, localprop.zp_mac,
ZIO_DATA_MAC_LEN);
if (DMU_OT_IS_ENCRYPTED(localprop.zp_type)) {
localprop.zp_nopwrite = B_FALSE;
localprop.zp_copies =
MIN(localprop.zp_copies, SPA_DVAS_PER_BP - 1);
}
zio_flags |= ZIO_FLAG_RAW;
} else if (ARC_BUF_COMPRESSED(buf)) {
ASSERT3U(HDR_GET_LSIZE(hdr), !=, arc_buf_size(buf));
localprop.zp_compress = HDR_GET_COMPRESS(hdr);
localprop.zp_complevel = hdr->b_complevel;
zio_flags |= ZIO_FLAG_RAW_COMPRESS;
}
callback = kmem_zalloc(sizeof (arc_write_callback_t), KM_SLEEP);
callback->awcb_ready = ready;
callback->awcb_children_ready = children_ready;
callback->awcb_physdone = physdone;
callback->awcb_done = done;
callback->awcb_private = private;
callback->awcb_buf = buf;
/*
* The hdr's b_pabd is now stale, free it now. A new data block
* will be allocated when the zio pipeline calls arc_write_ready().
*/
if (hdr->b_l1hdr.b_pabd != NULL) {
/*
* If the buf is currently sharing the data block with
* the hdr then we need to break that relationship here.
* The hdr will remain with a NULL data pointer and the
* buf will take sole ownership of the block.
*/
if (arc_buf_is_shared(buf)) {
arc_unshare_buf(hdr, buf);
} else {
arc_hdr_free_abd(hdr, B_FALSE);
}
VERIFY3P(buf->b_data, !=, NULL);
}
if (HDR_HAS_RABD(hdr))
arc_hdr_free_abd(hdr, B_TRUE);
if (!(zio_flags & ZIO_FLAG_RAW))
arc_hdr_set_compress(hdr, ZIO_COMPRESS_OFF);
ASSERT(!arc_buf_is_shared(buf));
ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
zio = zio_write(pio, spa, txg, bp,
abd_get_from_buf(buf->b_data, HDR_GET_LSIZE(hdr)),
HDR_GET_LSIZE(hdr), arc_buf_size(buf), &localprop, arc_write_ready,
(children_ready != NULL) ? arc_write_children_ready : NULL,
arc_write_physdone, arc_write_done, callback,
priority, zio_flags, zb);
return (zio);
}
void
arc_tempreserve_clear(uint64_t reserve)
{
atomic_add_64(&arc_tempreserve, -reserve);
ASSERT((int64_t)arc_tempreserve >= 0);
}
int
arc_tempreserve_space(spa_t *spa, uint64_t reserve, uint64_t txg)
{
int error;
uint64_t anon_size;
if (!arc_no_grow &&
reserve > arc_c/4 &&
reserve * 4 > (2ULL << SPA_MAXBLOCKSHIFT))
arc_c = MIN(arc_c_max, reserve * 4);
/*
* Throttle when the calculated memory footprint for the TXG
* exceeds the target ARC size.
*/
if (reserve > arc_c) {
DMU_TX_STAT_BUMP(dmu_tx_memory_reserve);
return (SET_ERROR(ERESTART));
}
/*
* Don't count loaned bufs as in flight dirty data to prevent long
* network delays from blocking transactions that are ready to be
* assigned to a txg.
*/
/* assert that it has not wrapped around */
ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0);
anon_size = MAX((int64_t)(zfs_refcount_count(&arc_anon->arcs_size) -
arc_loaned_bytes), 0);
/*
* Writes will, almost always, require additional memory allocations
* in order to compress/encrypt/etc the data. We therefore need to
* make sure that there is sufficient available memory for this.
*/
error = arc_memory_throttle(spa, reserve, txg);
if (error != 0)
return (error);
/*
* Throttle writes when the amount of dirty data in the cache
* gets too large. We try to keep the cache less than half full
* of dirty blocks so that our sync times don't grow too large.
*
* In the case of one pool being built on another pool, we want
* to make sure we don't end up throttling the lower (backing)
* pool when the upper pool is the majority contributor to dirty
* data. To insure we make forward progress during throttling, we
* also check the current pool's net dirty data and only throttle
* if it exceeds zfs_arc_pool_dirty_percent of the anonymous dirty
* data in the cache.
*
* Note: if two requests come in concurrently, we might let them
* both succeed, when one of them should fail. Not a huge deal.
*/
uint64_t total_dirty = reserve + arc_tempreserve + anon_size;
uint64_t spa_dirty_anon = spa_dirty_data(spa);
uint64_t rarc_c = arc_warm ? arc_c : arc_c_max;
if (total_dirty > rarc_c * zfs_arc_dirty_limit_percent / 100 &&
anon_size > rarc_c * zfs_arc_anon_limit_percent / 100 &&
spa_dirty_anon > anon_size * zfs_arc_pool_dirty_percent / 100) {
#ifdef ZFS_DEBUG
uint64_t meta_esize = zfs_refcount_count(
&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
uint64_t data_esize =
zfs_refcount_count(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
dprintf("failing, arc_tempreserve=%lluK anon_meta=%lluK "
"anon_data=%lluK tempreserve=%lluK rarc_c=%lluK\n",
(u_longlong_t)arc_tempreserve >> 10,
(u_longlong_t)meta_esize >> 10,
(u_longlong_t)data_esize >> 10,
(u_longlong_t)reserve >> 10,
(u_longlong_t)rarc_c >> 10);
#endif
DMU_TX_STAT_BUMP(dmu_tx_dirty_throttle);
return (SET_ERROR(ERESTART));
}
atomic_add_64(&arc_tempreserve, reserve);
return (0);
}
static void
arc_kstat_update_state(arc_state_t *state, kstat_named_t *size,
kstat_named_t *evict_data, kstat_named_t *evict_metadata)
{
size->value.ui64 = zfs_refcount_count(&state->arcs_size);
evict_data->value.ui64 =
zfs_refcount_count(&state->arcs_esize[ARC_BUFC_DATA]);
evict_metadata->value.ui64 =
zfs_refcount_count(&state->arcs_esize[ARC_BUFC_METADATA]);
}
static int
arc_kstat_update(kstat_t *ksp, int rw)
{
arc_stats_t *as = ksp->ks_data;
if (rw == KSTAT_WRITE)
return (SET_ERROR(EACCES));
as->arcstat_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_hits);
as->arcstat_misses.value.ui64 =
wmsum_value(&arc_sums.arcstat_misses);
as->arcstat_demand_data_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_demand_data_hits);
as->arcstat_demand_data_misses.value.ui64 =
wmsum_value(&arc_sums.arcstat_demand_data_misses);
as->arcstat_demand_metadata_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_demand_metadata_hits);
as->arcstat_demand_metadata_misses.value.ui64 =
wmsum_value(&arc_sums.arcstat_demand_metadata_misses);
as->arcstat_prefetch_data_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_prefetch_data_hits);
as->arcstat_prefetch_data_misses.value.ui64 =
wmsum_value(&arc_sums.arcstat_prefetch_data_misses);
as->arcstat_prefetch_metadata_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_prefetch_metadata_hits);
as->arcstat_prefetch_metadata_misses.value.ui64 =
wmsum_value(&arc_sums.arcstat_prefetch_metadata_misses);
as->arcstat_mru_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_mru_hits);
as->arcstat_mru_ghost_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_mru_ghost_hits);
as->arcstat_mfu_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_mfu_hits);
as->arcstat_mfu_ghost_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_mfu_ghost_hits);
as->arcstat_deleted.value.ui64 =
wmsum_value(&arc_sums.arcstat_deleted);
as->arcstat_mutex_miss.value.ui64 =
wmsum_value(&arc_sums.arcstat_mutex_miss);
as->arcstat_access_skip.value.ui64 =
wmsum_value(&arc_sums.arcstat_access_skip);
as->arcstat_evict_skip.value.ui64 =
wmsum_value(&arc_sums.arcstat_evict_skip);
as->arcstat_evict_not_enough.value.ui64 =
wmsum_value(&arc_sums.arcstat_evict_not_enough);
as->arcstat_evict_l2_cached.value.ui64 =
wmsum_value(&arc_sums.arcstat_evict_l2_cached);
as->arcstat_evict_l2_eligible.value.ui64 =
wmsum_value(&arc_sums.arcstat_evict_l2_eligible);
as->arcstat_evict_l2_eligible_mfu.value.ui64 =
wmsum_value(&arc_sums.arcstat_evict_l2_eligible_mfu);
as->arcstat_evict_l2_eligible_mru.value.ui64 =
wmsum_value(&arc_sums.arcstat_evict_l2_eligible_mru);
as->arcstat_evict_l2_ineligible.value.ui64 =
wmsum_value(&arc_sums.arcstat_evict_l2_ineligible);
as->arcstat_evict_l2_skip.value.ui64 =
wmsum_value(&arc_sums.arcstat_evict_l2_skip);
as->arcstat_hash_collisions.value.ui64 =
wmsum_value(&arc_sums.arcstat_hash_collisions);
as->arcstat_hash_chains.value.ui64 =
wmsum_value(&arc_sums.arcstat_hash_chains);
as->arcstat_size.value.ui64 =
aggsum_value(&arc_sums.arcstat_size);
as->arcstat_compressed_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_compressed_size);
as->arcstat_uncompressed_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_uncompressed_size);
as->arcstat_overhead_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_overhead_size);
as->arcstat_hdr_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_hdr_size);
as->arcstat_data_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_data_size);
as->arcstat_metadata_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_metadata_size);
as->arcstat_dbuf_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_dbuf_size);
#if defined(COMPAT_FREEBSD11)
as->arcstat_other_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_bonus_size) +
aggsum_value(&arc_sums.arcstat_dnode_size) +
wmsum_value(&arc_sums.arcstat_dbuf_size);
#endif
arc_kstat_update_state(arc_anon,
&as->arcstat_anon_size,
&as->arcstat_anon_evictable_data,
&as->arcstat_anon_evictable_metadata);
arc_kstat_update_state(arc_mru,
&as->arcstat_mru_size,
&as->arcstat_mru_evictable_data,
&as->arcstat_mru_evictable_metadata);
arc_kstat_update_state(arc_mru_ghost,
&as->arcstat_mru_ghost_size,
&as->arcstat_mru_ghost_evictable_data,
&as->arcstat_mru_ghost_evictable_metadata);
arc_kstat_update_state(arc_mfu,
&as->arcstat_mfu_size,
&as->arcstat_mfu_evictable_data,
&as->arcstat_mfu_evictable_metadata);
arc_kstat_update_state(arc_mfu_ghost,
&as->arcstat_mfu_ghost_size,
&as->arcstat_mfu_ghost_evictable_data,
&as->arcstat_mfu_ghost_evictable_metadata);
as->arcstat_dnode_size.value.ui64 =
aggsum_value(&arc_sums.arcstat_dnode_size);
as->arcstat_bonus_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_bonus_size);
as->arcstat_l2_hits.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_hits);
as->arcstat_l2_misses.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_misses);
as->arcstat_l2_prefetch_asize.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_prefetch_asize);
as->arcstat_l2_mru_asize.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_mru_asize);
as->arcstat_l2_mfu_asize.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_mfu_asize);
as->arcstat_l2_bufc_data_asize.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_bufc_data_asize);
as->arcstat_l2_bufc_metadata_asize.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_bufc_metadata_asize);
as->arcstat_l2_feeds.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_feeds);
as->arcstat_l2_rw_clash.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rw_clash);
as->arcstat_l2_read_bytes.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_read_bytes);
as->arcstat_l2_write_bytes.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_write_bytes);
as->arcstat_l2_writes_sent.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_writes_sent);
as->arcstat_l2_writes_done.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_writes_done);
as->arcstat_l2_writes_error.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_writes_error);
as->arcstat_l2_writes_lock_retry.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_writes_lock_retry);
as->arcstat_l2_evict_lock_retry.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_evict_lock_retry);
as->arcstat_l2_evict_reading.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_evict_reading);
as->arcstat_l2_evict_l1cached.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_evict_l1cached);
as->arcstat_l2_free_on_write.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_free_on_write);
as->arcstat_l2_abort_lowmem.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_abort_lowmem);
as->arcstat_l2_cksum_bad.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_cksum_bad);
as->arcstat_l2_io_error.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_io_error);
as->arcstat_l2_lsize.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_lsize);
as->arcstat_l2_psize.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_psize);
as->arcstat_l2_hdr_size.value.ui64 =
aggsum_value(&arc_sums.arcstat_l2_hdr_size);
as->arcstat_l2_log_blk_writes.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_log_blk_writes);
as->arcstat_l2_log_blk_asize.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_log_blk_asize);
as->arcstat_l2_log_blk_count.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_log_blk_count);
as->arcstat_l2_rebuild_success.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_success);
as->arcstat_l2_rebuild_abort_unsupported.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_unsupported);
as->arcstat_l2_rebuild_abort_io_errors.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_io_errors);
as->arcstat_l2_rebuild_abort_dh_errors.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_dh_errors);
as->arcstat_l2_rebuild_abort_cksum_lb_errors.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors);
as->arcstat_l2_rebuild_abort_lowmem.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_lowmem);
as->arcstat_l2_rebuild_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_size);
as->arcstat_l2_rebuild_asize.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_asize);
as->arcstat_l2_rebuild_bufs.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_bufs);
as->arcstat_l2_rebuild_bufs_precached.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_bufs_precached);
as->arcstat_l2_rebuild_log_blks.value.ui64 =
wmsum_value(&arc_sums.arcstat_l2_rebuild_log_blks);
as->arcstat_memory_throttle_count.value.ui64 =
wmsum_value(&arc_sums.arcstat_memory_throttle_count);
as->arcstat_memory_direct_count.value.ui64 =
wmsum_value(&arc_sums.arcstat_memory_direct_count);
as->arcstat_memory_indirect_count.value.ui64 =
wmsum_value(&arc_sums.arcstat_memory_indirect_count);
as->arcstat_memory_all_bytes.value.ui64 =
arc_all_memory();
as->arcstat_memory_free_bytes.value.ui64 =
arc_free_memory();
as->arcstat_memory_available_bytes.value.i64 =
arc_available_memory();
as->arcstat_prune.value.ui64 =
wmsum_value(&arc_sums.arcstat_prune);
as->arcstat_meta_used.value.ui64 =
aggsum_value(&arc_sums.arcstat_meta_used);
as->arcstat_async_upgrade_sync.value.ui64 =
wmsum_value(&arc_sums.arcstat_async_upgrade_sync);
as->arcstat_demand_hit_predictive_prefetch.value.ui64 =
wmsum_value(&arc_sums.arcstat_demand_hit_predictive_prefetch);
as->arcstat_demand_hit_prescient_prefetch.value.ui64 =
wmsum_value(&arc_sums.arcstat_demand_hit_prescient_prefetch);
as->arcstat_raw_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_raw_size);
as->arcstat_cached_only_in_progress.value.ui64 =
wmsum_value(&arc_sums.arcstat_cached_only_in_progress);
as->arcstat_abd_chunk_waste_size.value.ui64 =
wmsum_value(&arc_sums.arcstat_abd_chunk_waste_size);
return (0);
}
/*
* This function *must* return indices evenly distributed between all
* sublists of the multilist. This is needed due to how the ARC eviction
* code is laid out; arc_evict_state() assumes ARC buffers are evenly
* distributed between all sublists and uses this assumption when
* deciding which sublist to evict from and how much to evict from it.
*/
static unsigned int
arc_state_multilist_index_func(multilist_t *ml, void *obj)
{
arc_buf_hdr_t *hdr = obj;
/*
* We rely on b_dva to generate evenly distributed index
* numbers using buf_hash below. So, as an added precaution,
* let's make sure we never add empty buffers to the arc lists.
*/
ASSERT(!HDR_EMPTY(hdr));
/*
* The assumption here, is the hash value for a given
* arc_buf_hdr_t will remain constant throughout its lifetime
* (i.e. its b_spa, b_dva, and b_birth fields don't change).
* Thus, we don't need to store the header's sublist index
* on insertion, as this index can be recalculated on removal.
*
* Also, the low order bits of the hash value are thought to be
* distributed evenly. Otherwise, in the case that the multilist
* has a power of two number of sublists, each sublists' usage
* would not be evenly distributed. In this context full 64bit
* division would be a waste of time, so limit it to 32 bits.
*/
return ((unsigned int)buf_hash(hdr->b_spa, &hdr->b_dva, hdr->b_birth) %
multilist_get_num_sublists(ml));
}
static unsigned int
arc_state_l2c_multilist_index_func(multilist_t *ml, void *obj)
{
panic("Header %p insert into arc_l2c_only %p", obj, ml);
}
#define WARN_IF_TUNING_IGNORED(tuning, value, do_warn) do { \
if ((do_warn) && (tuning) && ((tuning) != (value))) { \
cmn_err(CE_WARN, \
"ignoring tunable %s (using %llu instead)", \
(#tuning), (u_longlong_t)(value)); \
} \
} while (0)
/*
* Called during module initialization and periodically thereafter to
* apply reasonable changes to the exposed performance tunings. Can also be
* called explicitly by param_set_arc_*() functions when ARC tunables are
* updated manually. Non-zero zfs_* values which differ from the currently set
* values will be applied.
*/
void
arc_tuning_update(boolean_t verbose)
{
uint64_t allmem = arc_all_memory();
unsigned long limit;
/* Valid range: 32M - <arc_c_max> */
if ((zfs_arc_min) && (zfs_arc_min != arc_c_min) &&
(zfs_arc_min >= 2ULL << SPA_MAXBLOCKSHIFT) &&
(zfs_arc_min <= arc_c_max)) {
arc_c_min = zfs_arc_min;
arc_c = MAX(arc_c, arc_c_min);
}
WARN_IF_TUNING_IGNORED(zfs_arc_min, arc_c_min, verbose);
/* Valid range: 64M - <all physical memory> */
if ((zfs_arc_max) && (zfs_arc_max != arc_c_max) &&
(zfs_arc_max >= MIN_ARC_MAX) && (zfs_arc_max < allmem) &&
(zfs_arc_max > arc_c_min)) {
arc_c_max = zfs_arc_max;
arc_c = MIN(arc_c, arc_c_max);
arc_p = (arc_c >> 1);
if (arc_meta_limit > arc_c_max)
arc_meta_limit = arc_c_max;
if (arc_dnode_size_limit > arc_meta_limit)
arc_dnode_size_limit = arc_meta_limit;
}
WARN_IF_TUNING_IGNORED(zfs_arc_max, arc_c_max, verbose);
/* Valid range: 16M - <arc_c_max> */
if ((zfs_arc_meta_min) && (zfs_arc_meta_min != arc_meta_min) &&
(zfs_arc_meta_min >= 1ULL << SPA_MAXBLOCKSHIFT) &&
(zfs_arc_meta_min <= arc_c_max)) {
arc_meta_min = zfs_arc_meta_min;
if (arc_meta_limit < arc_meta_min)
arc_meta_limit = arc_meta_min;
if (arc_dnode_size_limit < arc_meta_min)
arc_dnode_size_limit = arc_meta_min;
}
WARN_IF_TUNING_IGNORED(zfs_arc_meta_min, arc_meta_min, verbose);
/* Valid range: <arc_meta_min> - <arc_c_max> */
limit = zfs_arc_meta_limit ? zfs_arc_meta_limit :
MIN(zfs_arc_meta_limit_percent, 100) * arc_c_max / 100;
if ((limit != arc_meta_limit) &&
(limit >= arc_meta_min) &&
(limit <= arc_c_max))
arc_meta_limit = limit;
WARN_IF_TUNING_IGNORED(zfs_arc_meta_limit, arc_meta_limit, verbose);
/* Valid range: <arc_meta_min> - <arc_meta_limit> */
limit = zfs_arc_dnode_limit ? zfs_arc_dnode_limit :
MIN(zfs_arc_dnode_limit_percent, 100) * arc_meta_limit / 100;
if ((limit != arc_dnode_size_limit) &&
(limit >= arc_meta_min) &&
(limit <= arc_meta_limit))
arc_dnode_size_limit = limit;
WARN_IF_TUNING_IGNORED(zfs_arc_dnode_limit, arc_dnode_size_limit,
verbose);
/* Valid range: 1 - N */
if (zfs_arc_grow_retry)
arc_grow_retry = zfs_arc_grow_retry;
/* Valid range: 1 - N */
if (zfs_arc_shrink_shift) {
arc_shrink_shift = zfs_arc_shrink_shift;
arc_no_grow_shift = MIN(arc_no_grow_shift, arc_shrink_shift -1);
}
/* Valid range: 1 - N */
if (zfs_arc_p_min_shift)
arc_p_min_shift = zfs_arc_p_min_shift;
/* Valid range: 1 - N ms */
if (zfs_arc_min_prefetch_ms)
arc_min_prefetch_ms = zfs_arc_min_prefetch_ms;
/* Valid range: 1 - N ms */
if (zfs_arc_min_prescient_prefetch_ms) {
arc_min_prescient_prefetch_ms =
zfs_arc_min_prescient_prefetch_ms;
}
/* Valid range: 0 - 100 */
if ((zfs_arc_lotsfree_percent >= 0) &&
(zfs_arc_lotsfree_percent <= 100))
arc_lotsfree_percent = zfs_arc_lotsfree_percent;
WARN_IF_TUNING_IGNORED(zfs_arc_lotsfree_percent, arc_lotsfree_percent,
verbose);
/* Valid range: 0 - <all physical memory> */
if ((zfs_arc_sys_free) && (zfs_arc_sys_free != arc_sys_free))
arc_sys_free = MIN(MAX(zfs_arc_sys_free, 0), allmem);
WARN_IF_TUNING_IGNORED(zfs_arc_sys_free, arc_sys_free, verbose);
}
static void
arc_state_init(void)
{
multilist_create(&arc_mru->arcs_list[ARC_BUFC_METADATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_multilist_index_func);
multilist_create(&arc_mru->arcs_list[ARC_BUFC_DATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_multilist_index_func);
multilist_create(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_multilist_index_func);
multilist_create(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_multilist_index_func);
multilist_create(&arc_mfu->arcs_list[ARC_BUFC_METADATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_multilist_index_func);
multilist_create(&arc_mfu->arcs_list[ARC_BUFC_DATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_multilist_index_func);
multilist_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_multilist_index_func);
multilist_create(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_multilist_index_func);
/*
* L2 headers should never be on the L2 state list since they don't
* have L1 headers allocated. Special index function asserts that.
*/
multilist_create(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_l2c_multilist_index_func);
multilist_create(&arc_l2c_only->arcs_list[ARC_BUFC_DATA],
sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node),
arc_state_l2c_multilist_index_func);
zfs_refcount_create(&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_create(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_create(&arc_mru->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_create(&arc_mru->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_create(&arc_anon->arcs_size);
zfs_refcount_create(&arc_mru->arcs_size);
zfs_refcount_create(&arc_mru_ghost->arcs_size);
zfs_refcount_create(&arc_mfu->arcs_size);
zfs_refcount_create(&arc_mfu_ghost->arcs_size);
zfs_refcount_create(&arc_l2c_only->arcs_size);
wmsum_init(&arc_sums.arcstat_hits, 0);
wmsum_init(&arc_sums.arcstat_misses, 0);
wmsum_init(&arc_sums.arcstat_demand_data_hits, 0);
wmsum_init(&arc_sums.arcstat_demand_data_misses, 0);
wmsum_init(&arc_sums.arcstat_demand_metadata_hits, 0);
wmsum_init(&arc_sums.arcstat_demand_metadata_misses, 0);
wmsum_init(&arc_sums.arcstat_prefetch_data_hits, 0);
wmsum_init(&arc_sums.arcstat_prefetch_data_misses, 0);
wmsum_init(&arc_sums.arcstat_prefetch_metadata_hits, 0);
wmsum_init(&arc_sums.arcstat_prefetch_metadata_misses, 0);
wmsum_init(&arc_sums.arcstat_mru_hits, 0);
wmsum_init(&arc_sums.arcstat_mru_ghost_hits, 0);
wmsum_init(&arc_sums.arcstat_mfu_hits, 0);
wmsum_init(&arc_sums.arcstat_mfu_ghost_hits, 0);
wmsum_init(&arc_sums.arcstat_deleted, 0);
wmsum_init(&arc_sums.arcstat_mutex_miss, 0);
wmsum_init(&arc_sums.arcstat_access_skip, 0);
wmsum_init(&arc_sums.arcstat_evict_skip, 0);
wmsum_init(&arc_sums.arcstat_evict_not_enough, 0);
wmsum_init(&arc_sums.arcstat_evict_l2_cached, 0);
wmsum_init(&arc_sums.arcstat_evict_l2_eligible, 0);
wmsum_init(&arc_sums.arcstat_evict_l2_eligible_mfu, 0);
wmsum_init(&arc_sums.arcstat_evict_l2_eligible_mru, 0);
wmsum_init(&arc_sums.arcstat_evict_l2_ineligible, 0);
wmsum_init(&arc_sums.arcstat_evict_l2_skip, 0);
wmsum_init(&arc_sums.arcstat_hash_collisions, 0);
wmsum_init(&arc_sums.arcstat_hash_chains, 0);
aggsum_init(&arc_sums.arcstat_size, 0);
wmsum_init(&arc_sums.arcstat_compressed_size, 0);
wmsum_init(&arc_sums.arcstat_uncompressed_size, 0);
wmsum_init(&arc_sums.arcstat_overhead_size, 0);
wmsum_init(&arc_sums.arcstat_hdr_size, 0);
wmsum_init(&arc_sums.arcstat_data_size, 0);
wmsum_init(&arc_sums.arcstat_metadata_size, 0);
wmsum_init(&arc_sums.arcstat_dbuf_size, 0);
aggsum_init(&arc_sums.arcstat_dnode_size, 0);
wmsum_init(&arc_sums.arcstat_bonus_size, 0);
wmsum_init(&arc_sums.arcstat_l2_hits, 0);
wmsum_init(&arc_sums.arcstat_l2_misses, 0);
wmsum_init(&arc_sums.arcstat_l2_prefetch_asize, 0);
wmsum_init(&arc_sums.arcstat_l2_mru_asize, 0);
wmsum_init(&arc_sums.arcstat_l2_mfu_asize, 0);
wmsum_init(&arc_sums.arcstat_l2_bufc_data_asize, 0);
wmsum_init(&arc_sums.arcstat_l2_bufc_metadata_asize, 0);
wmsum_init(&arc_sums.arcstat_l2_feeds, 0);
wmsum_init(&arc_sums.arcstat_l2_rw_clash, 0);
wmsum_init(&arc_sums.arcstat_l2_read_bytes, 0);
wmsum_init(&arc_sums.arcstat_l2_write_bytes, 0);
wmsum_init(&arc_sums.arcstat_l2_writes_sent, 0);
wmsum_init(&arc_sums.arcstat_l2_writes_done, 0);
wmsum_init(&arc_sums.arcstat_l2_writes_error, 0);
wmsum_init(&arc_sums.arcstat_l2_writes_lock_retry, 0);
wmsum_init(&arc_sums.arcstat_l2_evict_lock_retry, 0);
wmsum_init(&arc_sums.arcstat_l2_evict_reading, 0);
wmsum_init(&arc_sums.arcstat_l2_evict_l1cached, 0);
wmsum_init(&arc_sums.arcstat_l2_free_on_write, 0);
wmsum_init(&arc_sums.arcstat_l2_abort_lowmem, 0);
wmsum_init(&arc_sums.arcstat_l2_cksum_bad, 0);
wmsum_init(&arc_sums.arcstat_l2_io_error, 0);
wmsum_init(&arc_sums.arcstat_l2_lsize, 0);
wmsum_init(&arc_sums.arcstat_l2_psize, 0);
aggsum_init(&arc_sums.arcstat_l2_hdr_size, 0);
wmsum_init(&arc_sums.arcstat_l2_log_blk_writes, 0);
wmsum_init(&arc_sums.arcstat_l2_log_blk_asize, 0);
wmsum_init(&arc_sums.arcstat_l2_log_blk_count, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_success, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_unsupported, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_io_errors, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_dh_errors, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_lowmem, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_size, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_asize, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_bufs, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_bufs_precached, 0);
wmsum_init(&arc_sums.arcstat_l2_rebuild_log_blks, 0);
wmsum_init(&arc_sums.arcstat_memory_throttle_count, 0);
wmsum_init(&arc_sums.arcstat_memory_direct_count, 0);
wmsum_init(&arc_sums.arcstat_memory_indirect_count, 0);
wmsum_init(&arc_sums.arcstat_prune, 0);
aggsum_init(&arc_sums.arcstat_meta_used, 0);
wmsum_init(&arc_sums.arcstat_async_upgrade_sync, 0);
wmsum_init(&arc_sums.arcstat_demand_hit_predictive_prefetch, 0);
wmsum_init(&arc_sums.arcstat_demand_hit_prescient_prefetch, 0);
wmsum_init(&arc_sums.arcstat_raw_size, 0);
wmsum_init(&arc_sums.arcstat_cached_only_in_progress, 0);
wmsum_init(&arc_sums.arcstat_abd_chunk_waste_size, 0);
arc_anon->arcs_state = ARC_STATE_ANON;
arc_mru->arcs_state = ARC_STATE_MRU;
arc_mru_ghost->arcs_state = ARC_STATE_MRU_GHOST;
arc_mfu->arcs_state = ARC_STATE_MFU;
arc_mfu_ghost->arcs_state = ARC_STATE_MFU_GHOST;
arc_l2c_only->arcs_state = ARC_STATE_L2C_ONLY;
}
static void
arc_state_fini(void)
{
zfs_refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]);
zfs_refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]);
zfs_refcount_destroy(&arc_anon->arcs_size);
zfs_refcount_destroy(&arc_mru->arcs_size);
zfs_refcount_destroy(&arc_mru_ghost->arcs_size);
zfs_refcount_destroy(&arc_mfu->arcs_size);
zfs_refcount_destroy(&arc_mfu_ghost->arcs_size);
zfs_refcount_destroy(&arc_l2c_only->arcs_size);
multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_METADATA]);
multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA]);
multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_METADATA]);
multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA]);
multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_DATA]);
multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA]);
multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_DATA]);
multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA]);
multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA]);
multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_DATA]);
wmsum_fini(&arc_sums.arcstat_hits);
wmsum_fini(&arc_sums.arcstat_misses);
wmsum_fini(&arc_sums.arcstat_demand_data_hits);
wmsum_fini(&arc_sums.arcstat_demand_data_misses);
wmsum_fini(&arc_sums.arcstat_demand_metadata_hits);
wmsum_fini(&arc_sums.arcstat_demand_metadata_misses);
wmsum_fini(&arc_sums.arcstat_prefetch_data_hits);
wmsum_fini(&arc_sums.arcstat_prefetch_data_misses);
wmsum_fini(&arc_sums.arcstat_prefetch_metadata_hits);
wmsum_fini(&arc_sums.arcstat_prefetch_metadata_misses);
wmsum_fini(&arc_sums.arcstat_mru_hits);
wmsum_fini(&arc_sums.arcstat_mru_ghost_hits);
wmsum_fini(&arc_sums.arcstat_mfu_hits);
wmsum_fini(&arc_sums.arcstat_mfu_ghost_hits);
wmsum_fini(&arc_sums.arcstat_deleted);
wmsum_fini(&arc_sums.arcstat_mutex_miss);
wmsum_fini(&arc_sums.arcstat_access_skip);
wmsum_fini(&arc_sums.arcstat_evict_skip);
wmsum_fini(&arc_sums.arcstat_evict_not_enough);
wmsum_fini(&arc_sums.arcstat_evict_l2_cached);
wmsum_fini(&arc_sums.arcstat_evict_l2_eligible);
wmsum_fini(&arc_sums.arcstat_evict_l2_eligible_mfu);
wmsum_fini(&arc_sums.arcstat_evict_l2_eligible_mru);
wmsum_fini(&arc_sums.arcstat_evict_l2_ineligible);
wmsum_fini(&arc_sums.arcstat_evict_l2_skip);
wmsum_fini(&arc_sums.arcstat_hash_collisions);
wmsum_fini(&arc_sums.arcstat_hash_chains);
aggsum_fini(&arc_sums.arcstat_size);
wmsum_fini(&arc_sums.arcstat_compressed_size);
wmsum_fini(&arc_sums.arcstat_uncompressed_size);
wmsum_fini(&arc_sums.arcstat_overhead_size);
wmsum_fini(&arc_sums.arcstat_hdr_size);
wmsum_fini(&arc_sums.arcstat_data_size);
wmsum_fini(&arc_sums.arcstat_metadata_size);
wmsum_fini(&arc_sums.arcstat_dbuf_size);
aggsum_fini(&arc_sums.arcstat_dnode_size);
wmsum_fini(&arc_sums.arcstat_bonus_size);
wmsum_fini(&arc_sums.arcstat_l2_hits);
wmsum_fini(&arc_sums.arcstat_l2_misses);
wmsum_fini(&arc_sums.arcstat_l2_prefetch_asize);
wmsum_fini(&arc_sums.arcstat_l2_mru_asize);
wmsum_fini(&arc_sums.arcstat_l2_mfu_asize);
wmsum_fini(&arc_sums.arcstat_l2_bufc_data_asize);
wmsum_fini(&arc_sums.arcstat_l2_bufc_metadata_asize);
wmsum_fini(&arc_sums.arcstat_l2_feeds);
wmsum_fini(&arc_sums.arcstat_l2_rw_clash);
wmsum_fini(&arc_sums.arcstat_l2_read_bytes);
wmsum_fini(&arc_sums.arcstat_l2_write_bytes);
wmsum_fini(&arc_sums.arcstat_l2_writes_sent);
wmsum_fini(&arc_sums.arcstat_l2_writes_done);
wmsum_fini(&arc_sums.arcstat_l2_writes_error);
wmsum_fini(&arc_sums.arcstat_l2_writes_lock_retry);
wmsum_fini(&arc_sums.arcstat_l2_evict_lock_retry);
wmsum_fini(&arc_sums.arcstat_l2_evict_reading);
wmsum_fini(&arc_sums.arcstat_l2_evict_l1cached);
wmsum_fini(&arc_sums.arcstat_l2_free_on_write);
wmsum_fini(&arc_sums.arcstat_l2_abort_lowmem);
wmsum_fini(&arc_sums.arcstat_l2_cksum_bad);
wmsum_fini(&arc_sums.arcstat_l2_io_error);
wmsum_fini(&arc_sums.arcstat_l2_lsize);
wmsum_fini(&arc_sums.arcstat_l2_psize);
aggsum_fini(&arc_sums.arcstat_l2_hdr_size);
wmsum_fini(&arc_sums.arcstat_l2_log_blk_writes);
wmsum_fini(&arc_sums.arcstat_l2_log_blk_asize);
wmsum_fini(&arc_sums.arcstat_l2_log_blk_count);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_success);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_unsupported);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_io_errors);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_dh_errors);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_lowmem);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_size);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_asize);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_bufs);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_bufs_precached);
wmsum_fini(&arc_sums.arcstat_l2_rebuild_log_blks);
wmsum_fini(&arc_sums.arcstat_memory_throttle_count);
wmsum_fini(&arc_sums.arcstat_memory_direct_count);
wmsum_fini(&arc_sums.arcstat_memory_indirect_count);
wmsum_fini(&arc_sums.arcstat_prune);
aggsum_fini(&arc_sums.arcstat_meta_used);
wmsum_fini(&arc_sums.arcstat_async_upgrade_sync);
wmsum_fini(&arc_sums.arcstat_demand_hit_predictive_prefetch);
wmsum_fini(&arc_sums.arcstat_demand_hit_prescient_prefetch);
wmsum_fini(&arc_sums.arcstat_raw_size);
wmsum_fini(&arc_sums.arcstat_cached_only_in_progress);
wmsum_fini(&arc_sums.arcstat_abd_chunk_waste_size);
}
uint64_t
arc_target_bytes(void)
{
return (arc_c);
}
void
arc_set_limits(uint64_t allmem)
{
/* Set min cache to 1/32 of all memory, or 32MB, whichever is more. */
arc_c_min = MAX(allmem / 32, 2ULL << SPA_MAXBLOCKSHIFT);
/* How to set default max varies by platform. */
arc_c_max = arc_default_max(arc_c_min, allmem);
}
void
arc_init(void)
{
uint64_t percent, allmem = arc_all_memory();
mutex_init(&arc_evict_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&arc_evict_waiters, sizeof (arc_evict_waiter_t),
offsetof(arc_evict_waiter_t, aew_node));
arc_min_prefetch_ms = 1000;
arc_min_prescient_prefetch_ms = 6000;
#if defined(_KERNEL)
arc_lowmem_init();
#endif
arc_set_limits(allmem);
#ifdef _KERNEL
/*
* If zfs_arc_max is non-zero at init, meaning it was set in the kernel
* environment before the module was loaded, don't block setting the
* maximum because it is less than arc_c_min, instead, reset arc_c_min
* to a lower value.
* zfs_arc_min will be handled by arc_tuning_update().
*/
if (zfs_arc_max != 0 && zfs_arc_max >= MIN_ARC_MAX &&
zfs_arc_max < allmem) {
arc_c_max = zfs_arc_max;
if (arc_c_min >= arc_c_max) {
arc_c_min = MAX(zfs_arc_max / 2,
2ULL << SPA_MAXBLOCKSHIFT);
}
}
#else
/*
* In userland, there's only the memory pressure that we artificially
* create (see arc_available_memory()). Don't let arc_c get too
* small, because it can cause transactions to be larger than
* arc_c, causing arc_tempreserve_space() to fail.
*/
arc_c_min = MAX(arc_c_max / 2, 2ULL << SPA_MAXBLOCKSHIFT);
#endif
arc_c = arc_c_min;
arc_p = (arc_c >> 1);
/* Set min to 1/2 of arc_c_min */
arc_meta_min = 1ULL << SPA_MAXBLOCKSHIFT;
/*
* Set arc_meta_limit to a percent of arc_c_max with a floor of
* arc_meta_min, and a ceiling of arc_c_max.
*/
percent = MIN(zfs_arc_meta_limit_percent, 100);
arc_meta_limit = MAX(arc_meta_min, (percent * arc_c_max) / 100);
percent = MIN(zfs_arc_dnode_limit_percent, 100);
arc_dnode_size_limit = (percent * arc_meta_limit) / 100;
/* Apply user specified tunings */
arc_tuning_update(B_TRUE);
/* if kmem_flags are set, lets try to use less memory */
if (kmem_debugging())
arc_c = arc_c / 2;
if (arc_c < arc_c_min)
arc_c = arc_c_min;
arc_register_hotplug();
arc_state_init();
buf_init();
list_create(&arc_prune_list, sizeof (arc_prune_t),
offsetof(arc_prune_t, p_node));
mutex_init(&arc_prune_mtx, NULL, MUTEX_DEFAULT, NULL);
arc_prune_taskq = taskq_create("arc_prune", zfs_arc_prune_task_threads,
defclsyspri, 100, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC);
arc_ksp = kstat_create("zfs", 0, "arcstats", "misc", KSTAT_TYPE_NAMED,
sizeof (arc_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
if (arc_ksp != NULL) {
arc_ksp->ks_data = &arc_stats;
arc_ksp->ks_update = arc_kstat_update;
kstat_install(arc_ksp);
}
arc_evict_zthr = zthr_create("arc_evict",
arc_evict_cb_check, arc_evict_cb, NULL, defclsyspri);
arc_reap_zthr = zthr_create_timer("arc_reap",
arc_reap_cb_check, arc_reap_cb, NULL, SEC2NSEC(1), minclsyspri);
arc_warm = B_FALSE;
/*
* Calculate maximum amount of dirty data per pool.
*
* If it has been set by a module parameter, take that.
* Otherwise, use a percentage of physical memory defined by
* zfs_dirty_data_max_percent (default 10%) with a cap at
* zfs_dirty_data_max_max (default 4G or 25% of physical memory).
*/
#ifdef __LP64__
if (zfs_dirty_data_max_max == 0)
zfs_dirty_data_max_max = MIN(4ULL * 1024 * 1024 * 1024,
allmem * zfs_dirty_data_max_max_percent / 100);
#else
if (zfs_dirty_data_max_max == 0)
zfs_dirty_data_max_max = MIN(1ULL * 1024 * 1024 * 1024,
allmem * zfs_dirty_data_max_max_percent / 100);
#endif
if (zfs_dirty_data_max == 0) {
zfs_dirty_data_max = allmem *
zfs_dirty_data_max_percent / 100;
zfs_dirty_data_max = MIN(zfs_dirty_data_max,
zfs_dirty_data_max_max);
}
if (zfs_wrlog_data_max == 0) {
/*
* dp_wrlog_total is reduced for each txg at the end of
* spa_sync(). However, dp_dirty_total is reduced every time
* a block is written out. Thus under normal operation,
* dp_wrlog_total could grow 2 times as big as
* zfs_dirty_data_max.
*/
zfs_wrlog_data_max = zfs_dirty_data_max * 2;
}
}
void
arc_fini(void)
{
arc_prune_t *p;
#ifdef _KERNEL
arc_lowmem_fini();
#endif /* _KERNEL */
/* Use B_TRUE to ensure *all* buffers are evicted */
arc_flush(NULL, B_TRUE);
if (arc_ksp != NULL) {
kstat_delete(arc_ksp);
arc_ksp = NULL;
}
taskq_wait(arc_prune_taskq);
taskq_destroy(arc_prune_taskq);
mutex_enter(&arc_prune_mtx);
while ((p = list_head(&arc_prune_list)) != NULL) {
list_remove(&arc_prune_list, p);
zfs_refcount_remove(&p->p_refcnt, &arc_prune_list);
zfs_refcount_destroy(&p->p_refcnt);
kmem_free(p, sizeof (*p));
}
mutex_exit(&arc_prune_mtx);
list_destroy(&arc_prune_list);
mutex_destroy(&arc_prune_mtx);
(void) zthr_cancel(arc_evict_zthr);
(void) zthr_cancel(arc_reap_zthr);
mutex_destroy(&arc_evict_lock);
list_destroy(&arc_evict_waiters);
/*
* Free any buffers that were tagged for destruction. This needs
* to occur before arc_state_fini() runs and destroys the aggsum
* values which are updated when freeing scatter ABDs.
*/
l2arc_do_free_on_write();
/*
* buf_fini() must proceed arc_state_fini() because buf_fin() may
* trigger the release of kmem magazines, which can callback to
* arc_space_return() which accesses aggsums freed in act_state_fini().
*/
buf_fini();
arc_state_fini();
arc_unregister_hotplug();
/*
* We destroy the zthrs after all the ARC state has been
* torn down to avoid the case of them receiving any
* wakeup() signals after they are destroyed.
*/
zthr_destroy(arc_evict_zthr);
zthr_destroy(arc_reap_zthr);
ASSERT0(arc_loaned_bytes);
}
/*
* Level 2 ARC
*
* The level 2 ARC (L2ARC) is a cache layer in-between main memory and disk.
* It uses dedicated storage devices to hold cached data, which are populated
* using large infrequent writes. The main role of this cache is to boost
* the performance of random read workloads. The intended L2ARC devices
* include short-stroked disks, solid state disks, and other media with
* substantially faster read latency than disk.
*
* +-----------------------+
* | ARC |
* +-----------------------+
* | ^ ^
* | | |
* l2arc_feed_thread() arc_read()
* | | |
* | l2arc read |
* V | |
* +---------------+ |
* | L2ARC | |
* +---------------+ |
* | ^ |
* l2arc_write() | |
* | | |
* V | |
* +-------+ +-------+
* | vdev | | vdev |
* | cache | | cache |
* +-------+ +-------+
* +=========+ .-----.
* : L2ARC : |-_____-|
* : devices : | Disks |
* +=========+ `-_____-'
*
* Read requests are satisfied from the following sources, in order:
*
* 1) ARC
* 2) vdev cache of L2ARC devices
* 3) L2ARC devices
* 4) vdev cache of disks
* 5) disks
*
* Some L2ARC device types exhibit extremely slow write performance.
* To accommodate for this there are some significant differences between
* the L2ARC and traditional cache design:
*
* 1. There is no eviction path from the ARC to the L2ARC. Evictions from
* the ARC behave as usual, freeing buffers and placing headers on ghost
* lists. The ARC does not send buffers to the L2ARC during eviction as
* this would add inflated write latencies for all ARC memory pressure.
*
* 2. The L2ARC attempts to cache data from the ARC before it is evicted.
* It does this by periodically scanning buffers from the eviction-end of
* the MFU and MRU ARC lists, copying them to the L2ARC devices if they are
* not already there. It scans until a headroom of buffers is satisfied,
* which itself is a buffer for ARC eviction. If a compressible buffer is
* found during scanning and selected for writing to an L2ARC device, we
* temporarily boost scanning headroom during the next scan cycle to make
* sure we adapt to compression effects (which might significantly reduce
* the data volume we write to L2ARC). The thread that does this is
* l2arc_feed_thread(), illustrated below; example sizes are included to
* provide a better sense of ratio than this diagram:
*
* head --> tail
* +---------------------+----------+
* ARC_mfu |:::::#:::::::::::::::|o#o###o###|-->. # already on L2ARC
* +---------------------+----------+ | o L2ARC eligible
* ARC_mru |:#:::::::::::::::::::|#o#ooo####|-->| : ARC buffer
* +---------------------+----------+ |
* 15.9 Gbytes ^ 32 Mbytes |
* headroom |
* l2arc_feed_thread()
* |
* l2arc write hand <--[oooo]--'
* | 8 Mbyte
* | write max
* V
* +==============================+
* L2ARC dev |####|#|###|###| |####| ... |
* +==============================+
* 32 Gbytes
*
* 3. If an ARC buffer is copied to the L2ARC but then hit instead of
* evicted, then the L2ARC has cached a buffer much sooner than it probably
* needed to, potentially wasting L2ARC device bandwidth and storage. It is
* safe to say that this is an uncommon case, since buffers at the end of
* the ARC lists have moved there due to inactivity.
*
* 4. If the ARC evicts faster than the L2ARC can maintain a headroom,
* then the L2ARC simply misses copying some buffers. This serves as a
* pressure valve to prevent heavy read workloads from both stalling the ARC
* with waits and clogging the L2ARC with writes. This also helps prevent
* the potential for the L2ARC to churn if it attempts to cache content too
* quickly, such as during backups of the entire pool.
*
* 5. After system boot and before the ARC has filled main memory, there are
* no evictions from the ARC and so the tails of the ARC_mfu and ARC_mru
* lists can remain mostly static. Instead of searching from tail of these
* lists as pictured, the l2arc_feed_thread() will search from the list heads
* for eligible buffers, greatly increasing its chance of finding them.
*
* The L2ARC device write speed is also boosted during this time so that
* the L2ARC warms up faster. Since there have been no ARC evictions yet,
* there are no L2ARC reads, and no fear of degrading read performance
* through increased writes.
*
* 6. Writes to the L2ARC devices are grouped and sent in-sequence, so that
* the vdev queue can aggregate them into larger and fewer writes. Each
* device is written to in a rotor fashion, sweeping writes through
* available space then repeating.
*
* 7. The L2ARC does not store dirty content. It never needs to flush
* write buffers back to disk based storage.
*
* 8. If an ARC buffer is written (and dirtied) which also exists in the
* L2ARC, the now stale L2ARC buffer is immediately dropped.
*
* The performance of the L2ARC can be tweaked by a number of tunables, which
* may be necessary for different workloads:
*
* l2arc_write_max max write bytes per interval
* l2arc_write_boost extra write bytes during device warmup
* l2arc_noprefetch skip caching prefetched buffers
* l2arc_headroom number of max device writes to precache
* l2arc_headroom_boost when we find compressed buffers during ARC
* scanning, we multiply headroom by this
* percentage factor for the next scan cycle,
* since more compressed buffers are likely to
* be present
* l2arc_feed_secs seconds between L2ARC writing
*
* Tunables may be removed or added as future performance improvements are
* integrated, and also may become zpool properties.
*
* There are three key functions that control how the L2ARC warms up:
*
* l2arc_write_eligible() check if a buffer is eligible to cache
* l2arc_write_size() calculate how much to write
* l2arc_write_interval() calculate sleep delay between writes
*
* These three functions determine what to write, how much, and how quickly
* to send writes.
*
* L2ARC persistence:
*
* When writing buffers to L2ARC, we periodically add some metadata to
* make sure we can pick them up after reboot, thus dramatically reducing
* the impact that any downtime has on the performance of storage systems
* with large caches.
*
* The implementation works fairly simply by integrating the following two
* modifications:
*
* *) When writing to the L2ARC, we occasionally write a "l2arc log block",
* which is an additional piece of metadata which describes what's been
* written. This allows us to rebuild the arc_buf_hdr_t structures of the
* main ARC buffers. There are 2 linked-lists of log blocks headed by
* dh_start_lbps[2]. We alternate which chain we append to, so they are
* time-wise and offset-wise interleaved, but that is an optimization rather
* than for correctness. The log block also includes a pointer to the
* previous block in its chain.
*
* *) We reserve SPA_MINBLOCKSIZE of space at the start of each L2ARC device
* for our header bookkeeping purposes. This contains a device header,
* which contains our top-level reference structures. We update it each
* time we write a new log block, so that we're able to locate it in the
* L2ARC device. If this write results in an inconsistent device header
* (e.g. due to power failure), we detect this by verifying the header's
* checksum and simply fail to reconstruct the L2ARC after reboot.
*
* Implementation diagram:
*
* +=== L2ARC device (not to scale) ======================================+
* | ___two newest log block pointers__.__________ |
* | / \dh_start_lbps[1] |
* | / \ \dh_start_lbps[0]|
* |.___/__. V V |
* ||L2 dev|....|lb |bufs |lb |bufs |lb |bufs |lb |bufs |lb |---(empty)---|
* || hdr| ^ /^ /^ / / |
* |+------+ ...--\-------/ \-----/--\------/ / |
* | \--------------/ \--------------/ |
* +======================================================================+
*
* As can be seen on the diagram, rather than using a simple linked list,
* we use a pair of linked lists with alternating elements. This is a
* performance enhancement due to the fact that we only find out the
* address of the next log block access once the current block has been
* completely read in. Obviously, this hurts performance, because we'd be
* keeping the device's I/O queue at only a 1 operation deep, thus
* incurring a large amount of I/O round-trip latency. Having two lists
* allows us to fetch two log blocks ahead of where we are currently
* rebuilding L2ARC buffers.
*
* On-device data structures:
*
* L2ARC device header: l2arc_dev_hdr_phys_t
* L2ARC log block: l2arc_log_blk_phys_t
*
* L2ARC reconstruction:
*
* When writing data, we simply write in the standard rotary fashion,
* evicting buffers as we go and simply writing new data over them (writing
* a new log block every now and then). This obviously means that once we
* loop around the end of the device, we will start cutting into an already
* committed log block (and its referenced data buffers), like so:
*
* current write head__ __old tail
* \ /
* V V
* <--|bufs |lb |bufs |lb | |bufs |lb |bufs |lb |-->
* ^ ^^^^^^^^^___________________________________
* | \
* <<nextwrite>> may overwrite this blk and/or its bufs --'
*
* When importing the pool, we detect this situation and use it to stop
* our scanning process (see l2arc_rebuild).
*
* There is one significant caveat to consider when rebuilding ARC contents
* from an L2ARC device: what about invalidated buffers? Given the above
* construction, we cannot update blocks which we've already written to amend
* them to remove buffers which were invalidated. Thus, during reconstruction,
* we might be populating the cache with buffers for data that's not on the
* main pool anymore, or may have been overwritten!
*
* As it turns out, this isn't a problem. Every arc_read request includes
* both the DVA and, crucially, the birth TXG of the BP the caller is
* looking for. So even if the cache were populated by completely rotten
* blocks for data that had been long deleted and/or overwritten, we'll
* never actually return bad data from the cache, since the DVA with the
* birth TXG uniquely identify a block in space and time - once created,
* a block is immutable on disk. The worst thing we have done is wasted
* some time and memory at l2arc rebuild to reconstruct outdated ARC
* entries that will get dropped from the l2arc as it is being updated
* with new blocks.
*
* L2ARC buffers that have been evicted by l2arc_evict() ahead of the write
* hand are not restored. This is done by saving the offset (in bytes)
* l2arc_evict() has evicted to in the L2ARC device header and taking it
* into account when restoring buffers.
*/
static boolean_t
l2arc_write_eligible(uint64_t spa_guid, arc_buf_hdr_t *hdr)
{
/*
* A buffer is *not* eligible for the L2ARC if it:
* 1. belongs to a different spa.
* 2. is already cached on the L2ARC.
* 3. has an I/O in progress (it may be an incomplete read).
* 4. is flagged not eligible (zfs property).
*/
if (hdr->b_spa != spa_guid || HDR_HAS_L2HDR(hdr) ||
HDR_IO_IN_PROGRESS(hdr) || !HDR_L2CACHE(hdr))
return (B_FALSE);
return (B_TRUE);
}
static uint64_t
l2arc_write_size(l2arc_dev_t *dev)
{
uint64_t size, dev_size, tsize;
/*
* Make sure our globals have meaningful values in case the user
* altered them.
*/
size = l2arc_write_max;
if (size == 0) {
cmn_err(CE_NOTE, "Bad value for l2arc_write_max, value must "
"be greater than zero, resetting it to the default (%d)",
L2ARC_WRITE_SIZE);
size = l2arc_write_max = L2ARC_WRITE_SIZE;
}
if (arc_warm == B_FALSE)
size += l2arc_write_boost;
/*
* Make sure the write size does not exceed the size of the cache
* device. This is important in l2arc_evict(), otherwise infinite
* iteration can occur.
*/
dev_size = dev->l2ad_end - dev->l2ad_start;
tsize = size + l2arc_log_blk_overhead(size, dev);
if (dev->l2ad_vdev->vdev_has_trim && l2arc_trim_ahead > 0)
tsize += MAX(64 * 1024 * 1024,
(tsize * l2arc_trim_ahead) / 100);
if (tsize >= dev_size) {
cmn_err(CE_NOTE, "l2arc_write_max or l2arc_write_boost "
"plus the overhead of log blocks (persistent L2ARC, "
"%llu bytes) exceeds the size of the cache device "
"(guid %llu), resetting them to the default (%d)",
(u_longlong_t)l2arc_log_blk_overhead(size, dev),
(u_longlong_t)dev->l2ad_vdev->vdev_guid, L2ARC_WRITE_SIZE);
size = l2arc_write_max = l2arc_write_boost = L2ARC_WRITE_SIZE;
if (arc_warm == B_FALSE)
size += l2arc_write_boost;
}
return (size);
}
static clock_t
l2arc_write_interval(clock_t began, uint64_t wanted, uint64_t wrote)
{
clock_t interval, next, now;
/*
* If the ARC lists are busy, increase our write rate; if the
* lists are stale, idle back. This is achieved by checking
* how much we previously wrote - if it was more than half of
* what we wanted, schedule the next write much sooner.
*/
if (l2arc_feed_again && wrote > (wanted / 2))
interval = (hz * l2arc_feed_min_ms) / 1000;
else
interval = hz * l2arc_feed_secs;
now = ddi_get_lbolt();
next = MAX(now, MIN(now + interval, began + interval));
return (next);
}
/*
* Cycle through L2ARC devices. This is how L2ARC load balances.
* If a device is returned, this also returns holding the spa config lock.
*/
static l2arc_dev_t *
l2arc_dev_get_next(void)
{
l2arc_dev_t *first, *next = NULL;
/*
* Lock out the removal of spas (spa_namespace_lock), then removal
* of cache devices (l2arc_dev_mtx). Once a device has been selected,
* both locks will be dropped and a spa config lock held instead.
*/
mutex_enter(&spa_namespace_lock);
mutex_enter(&l2arc_dev_mtx);
/* if there are no vdevs, there is nothing to do */
if (l2arc_ndev == 0)
goto out;
first = NULL;
next = l2arc_dev_last;
do {
/* loop around the list looking for a non-faulted vdev */
if (next == NULL) {
next = list_head(l2arc_dev_list);
} else {
next = list_next(l2arc_dev_list, next);
if (next == NULL)
next = list_head(l2arc_dev_list);
}
/* if we have come back to the start, bail out */
if (first == NULL)
first = next;
else if (next == first)
break;
} while (vdev_is_dead(next->l2ad_vdev) || next->l2ad_rebuild ||
next->l2ad_trim_all);
/* if we were unable to find any usable vdevs, return NULL */
if (vdev_is_dead(next->l2ad_vdev) || next->l2ad_rebuild ||
next->l2ad_trim_all)
next = NULL;
l2arc_dev_last = next;
out:
mutex_exit(&l2arc_dev_mtx);
/*
* Grab the config lock to prevent the 'next' device from being
* removed while we are writing to it.
*/
if (next != NULL)
spa_config_enter(next->l2ad_spa, SCL_L2ARC, next, RW_READER);
mutex_exit(&spa_namespace_lock);
return (next);
}
/*
* Free buffers that were tagged for destruction.
*/
static void
l2arc_do_free_on_write(void)
{
list_t *buflist;
l2arc_data_free_t *df, *df_prev;
mutex_enter(&l2arc_free_on_write_mtx);
buflist = l2arc_free_on_write;
for (df = list_tail(buflist); df; df = df_prev) {
df_prev = list_prev(buflist, df);
ASSERT3P(df->l2df_abd, !=, NULL);
abd_free(df->l2df_abd);
list_remove(buflist, df);
kmem_free(df, sizeof (l2arc_data_free_t));
}
mutex_exit(&l2arc_free_on_write_mtx);
}
/*
* A write to a cache device has completed. Update all headers to allow
* reads from these buffers to begin.
*/
static void
l2arc_write_done(zio_t *zio)
{
l2arc_write_callback_t *cb;
l2arc_lb_abd_buf_t *abd_buf;
l2arc_lb_ptr_buf_t *lb_ptr_buf;
l2arc_dev_t *dev;
l2arc_dev_hdr_phys_t *l2dhdr;
list_t *buflist;
arc_buf_hdr_t *head, *hdr, *hdr_prev;
kmutex_t *hash_lock;
int64_t bytes_dropped = 0;
cb = zio->io_private;
ASSERT3P(cb, !=, NULL);
dev = cb->l2wcb_dev;
l2dhdr = dev->l2ad_dev_hdr;
ASSERT3P(dev, !=, NULL);
head = cb->l2wcb_head;
ASSERT3P(head, !=, NULL);
buflist = &dev->l2ad_buflist;
ASSERT3P(buflist, !=, NULL);
DTRACE_PROBE2(l2arc__iodone, zio_t *, zio,
l2arc_write_callback_t *, cb);
/*
* All writes completed, or an error was hit.
*/
top:
mutex_enter(&dev->l2ad_mtx);
for (hdr = list_prev(buflist, head); hdr; hdr = hdr_prev) {
hdr_prev = list_prev(buflist, hdr);
hash_lock = HDR_LOCK(hdr);
/*
* We cannot use mutex_enter or else we can deadlock
* with l2arc_write_buffers (due to swapping the order
* the hash lock and l2ad_mtx are taken).
*/
if (!mutex_tryenter(hash_lock)) {
/*
* Missed the hash lock. We must retry so we
* don't leave the ARC_FLAG_L2_WRITING bit set.
*/
ARCSTAT_BUMP(arcstat_l2_writes_lock_retry);
/*
* We don't want to rescan the headers we've
* already marked as having been written out, so
* we reinsert the head node so we can pick up
* where we left off.
*/
list_remove(buflist, head);
list_insert_after(buflist, hdr, head);
mutex_exit(&dev->l2ad_mtx);
/*
* We wait for the hash lock to become available
* to try and prevent busy waiting, and increase
* the chance we'll be able to acquire the lock
* the next time around.
*/
mutex_enter(hash_lock);
mutex_exit(hash_lock);
goto top;
}
/*
* We could not have been moved into the arc_l2c_only
* state while in-flight due to our ARC_FLAG_L2_WRITING
* bit being set. Let's just ensure that's being enforced.
*/
ASSERT(HDR_HAS_L1HDR(hdr));
/*
* Skipped - drop L2ARC entry and mark the header as no
* longer L2 eligibile.
*/
if (zio->io_error != 0) {
/*
* Error - drop L2ARC entry.
*/
list_remove(buflist, hdr);
arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR);
uint64_t psize = HDR_GET_PSIZE(hdr);
l2arc_hdr_arcstats_decrement(hdr);
bytes_dropped +=
vdev_psize_to_asize(dev->l2ad_vdev, psize);
(void) zfs_refcount_remove_many(&dev->l2ad_alloc,
arc_hdr_size(hdr), hdr);
}
/*
* Allow ARC to begin reads and ghost list evictions to
* this L2ARC entry.
*/
arc_hdr_clear_flags(hdr, ARC_FLAG_L2_WRITING);
mutex_exit(hash_lock);
}
/*
* Free the allocated abd buffers for writing the log blocks.
* If the zio failed reclaim the allocated space and remove the
* pointers to these log blocks from the log block pointer list
* of the L2ARC device.
*/
while ((abd_buf = list_remove_tail(&cb->l2wcb_abd_list)) != NULL) {
abd_free(abd_buf->abd);
zio_buf_free(abd_buf, sizeof (*abd_buf));
if (zio->io_error != 0) {
lb_ptr_buf = list_remove_head(&dev->l2ad_lbptr_list);
/*
* L2BLK_GET_PSIZE returns aligned size for log
* blocks.
*/
uint64_t asize =
L2BLK_GET_PSIZE((lb_ptr_buf->lb_ptr)->lbp_prop);
bytes_dropped += asize;
ARCSTAT_INCR(arcstat_l2_log_blk_asize, -asize);
ARCSTAT_BUMPDOWN(arcstat_l2_log_blk_count);
zfs_refcount_remove_many(&dev->l2ad_lb_asize, asize,
lb_ptr_buf);
zfs_refcount_remove(&dev->l2ad_lb_count, lb_ptr_buf);
kmem_free(lb_ptr_buf->lb_ptr,
sizeof (l2arc_log_blkptr_t));
kmem_free(lb_ptr_buf, sizeof (l2arc_lb_ptr_buf_t));
}
}
list_destroy(&cb->l2wcb_abd_list);
if (zio->io_error != 0) {
ARCSTAT_BUMP(arcstat_l2_writes_error);
/*
* Restore the lbps array in the header to its previous state.
* If the list of log block pointers is empty, zero out the
* log block pointers in the device header.
*/
lb_ptr_buf = list_head(&dev->l2ad_lbptr_list);
for (int i = 0; i < 2; i++) {
if (lb_ptr_buf == NULL) {
/*
* If the list is empty zero out the device
* header. Otherwise zero out the second log
* block pointer in the header.
*/
if (i == 0) {
bzero(l2dhdr, dev->l2ad_dev_hdr_asize);
} else {
bzero(&l2dhdr->dh_start_lbps[i],
sizeof (l2arc_log_blkptr_t));
}
break;
}
bcopy(lb_ptr_buf->lb_ptr, &l2dhdr->dh_start_lbps[i],
sizeof (l2arc_log_blkptr_t));
lb_ptr_buf = list_next(&dev->l2ad_lbptr_list,
lb_ptr_buf);
}
}
ARCSTAT_BUMP(arcstat_l2_writes_done);
list_remove(buflist, head);
ASSERT(!HDR_HAS_L1HDR(head));
kmem_cache_free(hdr_l2only_cache, head);
mutex_exit(&dev->l2ad_mtx);
ASSERT(dev->l2ad_vdev != NULL);
vdev_space_update(dev->l2ad_vdev, -bytes_dropped, 0, 0);
l2arc_do_free_on_write();
kmem_free(cb, sizeof (l2arc_write_callback_t));
}
static int
l2arc_untransform(zio_t *zio, l2arc_read_callback_t *cb)
{
int ret;
spa_t *spa = zio->io_spa;
arc_buf_hdr_t *hdr = cb->l2rcb_hdr;
blkptr_t *bp = zio->io_bp;
uint8_t salt[ZIO_DATA_SALT_LEN];
uint8_t iv[ZIO_DATA_IV_LEN];
uint8_t mac[ZIO_DATA_MAC_LEN];
boolean_t no_crypt = B_FALSE;
/*
* ZIL data is never be written to the L2ARC, so we don't need
* special handling for its unique MAC storage.
*/
ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG);
ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
/*
* If the data was encrypted, decrypt it now. Note that
* we must check the bp here and not the hdr, since the
* hdr does not have its encryption parameters updated
* until arc_read_done().
*/
if (BP_IS_ENCRYPTED(bp)) {
abd_t *eabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr,
ARC_HDR_DO_ADAPT | ARC_HDR_USE_RESERVE);
zio_crypt_decode_params_bp(bp, salt, iv);
zio_crypt_decode_mac_bp(bp, mac);
ret = spa_do_crypt_abd(B_FALSE, spa, &cb->l2rcb_zb,
BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp),
salt, iv, mac, HDR_GET_PSIZE(hdr), eabd,
hdr->b_l1hdr.b_pabd, &no_crypt);
if (ret != 0) {
arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr);
goto error;
}
/*
* If we actually performed decryption, replace b_pabd
* with the decrypted data. Otherwise we can just throw
* our decryption buffer away.
*/
if (!no_crypt) {
arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd,
arc_hdr_size(hdr), hdr);
hdr->b_l1hdr.b_pabd = eabd;
zio->io_abd = eabd;
} else {
arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr);
}
}
/*
* If the L2ARC block was compressed, but ARC compression
* is disabled we decompress the data into a new buffer and
* replace the existing data.
*/
if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
!HDR_COMPRESSION_ENABLED(hdr)) {
abd_t *cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr,
ARC_HDR_DO_ADAPT | ARC_HDR_USE_RESERVE);
void *tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr));
ret = zio_decompress_data(HDR_GET_COMPRESS(hdr),
hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr),
HDR_GET_LSIZE(hdr), &hdr->b_complevel);
if (ret != 0) {
abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
arc_free_data_abd(hdr, cabd, arc_hdr_size(hdr), hdr);
goto error;
}
abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd,
arc_hdr_size(hdr), hdr);
hdr->b_l1hdr.b_pabd = cabd;
zio->io_abd = cabd;
zio->io_size = HDR_GET_LSIZE(hdr);
}
return (0);
error:
return (ret);
}
/*
* A read to a cache device completed. Validate buffer contents before
* handing over to the regular ARC routines.
*/
static void
l2arc_read_done(zio_t *zio)
{
int tfm_error = 0;
l2arc_read_callback_t *cb = zio->io_private;
arc_buf_hdr_t *hdr;
kmutex_t *hash_lock;
boolean_t valid_cksum;
boolean_t using_rdata = (BP_IS_ENCRYPTED(&cb->l2rcb_bp) &&
(cb->l2rcb_flags & ZIO_FLAG_RAW_ENCRYPT));
ASSERT3P(zio->io_vd, !=, NULL);
ASSERT(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE);
spa_config_exit(zio->io_spa, SCL_L2ARC, zio->io_vd);
ASSERT3P(cb, !=, NULL);
hdr = cb->l2rcb_hdr;
ASSERT3P(hdr, !=, NULL);
hash_lock = HDR_LOCK(hdr);
mutex_enter(hash_lock);
ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
/*
* If the data was read into a temporary buffer,
* move it and free the buffer.
*/
if (cb->l2rcb_abd != NULL) {
ASSERT3U(arc_hdr_size(hdr), <, zio->io_size);
if (zio->io_error == 0) {
if (using_rdata) {
abd_copy(hdr->b_crypt_hdr.b_rabd,
cb->l2rcb_abd, arc_hdr_size(hdr));
} else {
abd_copy(hdr->b_l1hdr.b_pabd,
cb->l2rcb_abd, arc_hdr_size(hdr));
}
}
/*
* The following must be done regardless of whether
* there was an error:
* - free the temporary buffer
* - point zio to the real ARC buffer
* - set zio size accordingly
* These are required because zio is either re-used for
* an I/O of the block in the case of the error
* or the zio is passed to arc_read_done() and it
* needs real data.
*/
abd_free(cb->l2rcb_abd);
zio->io_size = zio->io_orig_size = arc_hdr_size(hdr);
if (using_rdata) {
ASSERT(HDR_HAS_RABD(hdr));
zio->io_abd = zio->io_orig_abd =
hdr->b_crypt_hdr.b_rabd;
} else {
ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
zio->io_abd = zio->io_orig_abd = hdr->b_l1hdr.b_pabd;
}
}
ASSERT3P(zio->io_abd, !=, NULL);
/*
* Check this survived the L2ARC journey.
*/
ASSERT(zio->io_abd == hdr->b_l1hdr.b_pabd ||
(HDR_HAS_RABD(hdr) && zio->io_abd == hdr->b_crypt_hdr.b_rabd));
zio->io_bp_copy = cb->l2rcb_bp; /* XXX fix in L2ARC 2.0 */
zio->io_bp = &zio->io_bp_copy; /* XXX fix in L2ARC 2.0 */
zio->io_prop.zp_complevel = hdr->b_complevel;
valid_cksum = arc_cksum_is_equal(hdr, zio);
/*
* b_rabd will always match the data as it exists on disk if it is
* being used. Therefore if we are reading into b_rabd we do not
* attempt to untransform the data.
*/
if (valid_cksum && !using_rdata)
tfm_error = l2arc_untransform(zio, cb);
if (valid_cksum && tfm_error == 0 && zio->io_error == 0 &&
!HDR_L2_EVICTED(hdr)) {
mutex_exit(hash_lock);
zio->io_private = hdr;
arc_read_done(zio);
} else {
/*
* Buffer didn't survive caching. Increment stats and
* reissue to the original storage device.
*/
if (zio->io_error != 0) {
ARCSTAT_BUMP(arcstat_l2_io_error);
} else {
zio->io_error = SET_ERROR(EIO);
}
if (!valid_cksum || tfm_error != 0)
ARCSTAT_BUMP(arcstat_l2_cksum_bad);
/*
* If there's no waiter, issue an async i/o to the primary
* storage now. If there *is* a waiter, the caller must
* issue the i/o in a context where it's OK to block.
*/
if (zio->io_waiter == NULL) {
zio_t *pio = zio_unique_parent(zio);
void *abd = (using_rdata) ?
hdr->b_crypt_hdr.b_rabd : hdr->b_l1hdr.b_pabd;
ASSERT(!pio || pio->io_child_type == ZIO_CHILD_LOGICAL);
zio = zio_read(pio, zio->io_spa, zio->io_bp,
abd, zio->io_size, arc_read_done,
hdr, zio->io_priority, cb->l2rcb_flags,
&cb->l2rcb_zb);
/*
* Original ZIO will be freed, so we need to update
* ARC header with the new ZIO pointer to be used
* by zio_change_priority() in arc_read().
*/
for (struct arc_callback *acb = hdr->b_l1hdr.b_acb;
acb != NULL; acb = acb->acb_next)
acb->acb_zio_head = zio;
mutex_exit(hash_lock);
zio_nowait(zio);
} else {
mutex_exit(hash_lock);
}
}
kmem_free(cb, sizeof (l2arc_read_callback_t));
}
/*
* This is the list priority from which the L2ARC will search for pages to
* cache. This is used within loops (0..3) to cycle through lists in the
* desired order. This order can have a significant effect on cache
* performance.
*
* Currently the metadata lists are hit first, MFU then MRU, followed by
* the data lists. This function returns a locked list, and also returns
* the lock pointer.
*/
static multilist_sublist_t *
l2arc_sublist_lock(int list_num)
{
multilist_t *ml = NULL;
unsigned int idx;
ASSERT(list_num >= 0 && list_num < L2ARC_FEED_TYPES);
switch (list_num) {
case 0:
ml = &arc_mfu->arcs_list[ARC_BUFC_METADATA];
break;
case 1:
ml = &arc_mru->arcs_list[ARC_BUFC_METADATA];
break;
case 2:
ml = &arc_mfu->arcs_list[ARC_BUFC_DATA];
break;
case 3:
ml = &arc_mru->arcs_list[ARC_BUFC_DATA];
break;
default:
return (NULL);
}
/*
* Return a randomly-selected sublist. This is acceptable
* because the caller feeds only a little bit of data for each
* call (8MB). Subsequent calls will result in different
* sublists being selected.
*/
idx = multilist_get_random_index(ml);
return (multilist_sublist_lock(ml, idx));
}
/*
* Calculates the maximum overhead of L2ARC metadata log blocks for a given
* L2ARC write size. l2arc_evict and l2arc_write_size need to include this
* overhead in processing to make sure there is enough headroom available
* when writing buffers.
*/
static inline uint64_t
l2arc_log_blk_overhead(uint64_t write_sz, l2arc_dev_t *dev)
{
if (dev->l2ad_log_entries == 0) {
return (0);
} else {
uint64_t log_entries = write_sz >> SPA_MINBLOCKSHIFT;
uint64_t log_blocks = (log_entries +
dev->l2ad_log_entries - 1) /
dev->l2ad_log_entries;
return (vdev_psize_to_asize(dev->l2ad_vdev,
sizeof (l2arc_log_blk_phys_t)) * log_blocks);
}
}
/*
* Evict buffers from the device write hand to the distance specified in
* bytes. This distance may span populated buffers, it may span nothing.
* This is clearing a region on the L2ARC device ready for writing.
* If the 'all' boolean is set, every buffer is evicted.
*/
static void
l2arc_evict(l2arc_dev_t *dev, uint64_t distance, boolean_t all)
{
list_t *buflist;
arc_buf_hdr_t *hdr, *hdr_prev;
kmutex_t *hash_lock;
uint64_t taddr;
l2arc_lb_ptr_buf_t *lb_ptr_buf, *lb_ptr_buf_prev;
vdev_t *vd = dev->l2ad_vdev;
boolean_t rerun;
buflist = &dev->l2ad_buflist;
/*
* We need to add in the worst case scenario of log block overhead.
*/
distance += l2arc_log_blk_overhead(distance, dev);
if (vd->vdev_has_trim && l2arc_trim_ahead > 0) {
/*
* Trim ahead of the write size 64MB or (l2arc_trim_ahead/100)
* times the write size, whichever is greater.
*/
distance += MAX(64 * 1024 * 1024,
(distance * l2arc_trim_ahead) / 100);
}
top:
rerun = B_FALSE;
if (dev->l2ad_hand >= (dev->l2ad_end - distance)) {
/*
* When there is no space to accommodate upcoming writes,
* evict to the end. Then bump the write and evict hands
* to the start and iterate. This iteration does not
* happen indefinitely as we make sure in
* l2arc_write_size() that when the write hand is reset,
* the write size does not exceed the end of the device.
*/
rerun = B_TRUE;
taddr = dev->l2ad_end;
} else {
taddr = dev->l2ad_hand + distance;
}
DTRACE_PROBE4(l2arc__evict, l2arc_dev_t *, dev, list_t *, buflist,
uint64_t, taddr, boolean_t, all);
if (!all) {
/*
* This check has to be placed after deciding whether to
* iterate (rerun).
*/
if (dev->l2ad_first) {
/*
* This is the first sweep through the device. There is
* nothing to evict. We have already trimmmed the
* whole device.
*/
goto out;
} else {
/*
* Trim the space to be evicted.
*/
if (vd->vdev_has_trim && dev->l2ad_evict < taddr &&
l2arc_trim_ahead > 0) {
/*
* We have to drop the spa_config lock because
* vdev_trim_range() will acquire it.
* l2ad_evict already accounts for the label
* size. To prevent vdev_trim_ranges() from
* adding it again, we subtract it from
* l2ad_evict.
*/
spa_config_exit(dev->l2ad_spa, SCL_L2ARC, dev);
vdev_trim_simple(vd,
dev->l2ad_evict - VDEV_LABEL_START_SIZE,
taddr - dev->l2ad_evict);
spa_config_enter(dev->l2ad_spa, SCL_L2ARC, dev,
RW_READER);
}
/*
* When rebuilding L2ARC we retrieve the evict hand
* from the header of the device. Of note, l2arc_evict()
* does not actually delete buffers from the cache
* device, but trimming may do so depending on the
* hardware implementation. Thus keeping track of the
* evict hand is useful.
*/
dev->l2ad_evict = MAX(dev->l2ad_evict, taddr);
}
}
retry:
mutex_enter(&dev->l2ad_mtx);
/*
* We have to account for evicted log blocks. Run vdev_space_update()
* on log blocks whose offset (in bytes) is before the evicted offset
* (in bytes) by searching in the list of pointers to log blocks
* present in the L2ARC device.
*/
for (lb_ptr_buf = list_tail(&dev->l2ad_lbptr_list); lb_ptr_buf;
lb_ptr_buf = lb_ptr_buf_prev) {
lb_ptr_buf_prev = list_prev(&dev->l2ad_lbptr_list, lb_ptr_buf);
/* L2BLK_GET_PSIZE returns aligned size for log blocks */
uint64_t asize = L2BLK_GET_PSIZE(
(lb_ptr_buf->lb_ptr)->lbp_prop);
/*
* We don't worry about log blocks left behind (ie
* lbp_payload_start < l2ad_hand) because l2arc_write_buffers()
* will never write more than l2arc_evict() evicts.
*/
if (!all && l2arc_log_blkptr_valid(dev, lb_ptr_buf->lb_ptr)) {
break;
} else {
vdev_space_update(vd, -asize, 0, 0);
ARCSTAT_INCR(arcstat_l2_log_blk_asize, -asize);
ARCSTAT_BUMPDOWN(arcstat_l2_log_blk_count);
zfs_refcount_remove_many(&dev->l2ad_lb_asize, asize,
lb_ptr_buf);
zfs_refcount_remove(&dev->l2ad_lb_count, lb_ptr_buf);
list_remove(&dev->l2ad_lbptr_list, lb_ptr_buf);
kmem_free(lb_ptr_buf->lb_ptr,
sizeof (l2arc_log_blkptr_t));
kmem_free(lb_ptr_buf, sizeof (l2arc_lb_ptr_buf_t));
}
}
for (hdr = list_tail(buflist); hdr; hdr = hdr_prev) {
hdr_prev = list_prev(buflist, hdr);
ASSERT(!HDR_EMPTY(hdr));
hash_lock = HDR_LOCK(hdr);
/*
* We cannot use mutex_enter or else we can deadlock
* with l2arc_write_buffers (due to swapping the order
* the hash lock and l2ad_mtx are taken).
*/
if (!mutex_tryenter(hash_lock)) {
/*
* Missed the hash lock. Retry.
*/
ARCSTAT_BUMP(arcstat_l2_evict_lock_retry);
mutex_exit(&dev->l2ad_mtx);
mutex_enter(hash_lock);
mutex_exit(hash_lock);
goto retry;
}
/*
* A header can't be on this list if it doesn't have L2 header.
*/
ASSERT(HDR_HAS_L2HDR(hdr));
/* Ensure this header has finished being written. */
ASSERT(!HDR_L2_WRITING(hdr));
ASSERT(!HDR_L2_WRITE_HEAD(hdr));
if (!all && (hdr->b_l2hdr.b_daddr >= dev->l2ad_evict ||
hdr->b_l2hdr.b_daddr < dev->l2ad_hand)) {
/*
* We've evicted to the target address,
* or the end of the device.
*/
mutex_exit(hash_lock);
break;
}
if (!HDR_HAS_L1HDR(hdr)) {
ASSERT(!HDR_L2_READING(hdr));
/*
* This doesn't exist in the ARC. Destroy.
* arc_hdr_destroy() will call list_remove()
* and decrement arcstat_l2_lsize.
*/
arc_change_state(arc_anon, hdr, hash_lock);
arc_hdr_destroy(hdr);
} else {
ASSERT(hdr->b_l1hdr.b_state != arc_l2c_only);
ARCSTAT_BUMP(arcstat_l2_evict_l1cached);
/*
* Invalidate issued or about to be issued
* reads, since we may be about to write
* over this location.
*/
if (HDR_L2_READING(hdr)) {
ARCSTAT_BUMP(arcstat_l2_evict_reading);
arc_hdr_set_flags(hdr, ARC_FLAG_L2_EVICTED);
}
arc_hdr_l2hdr_destroy(hdr);
}
mutex_exit(hash_lock);
}
mutex_exit(&dev->l2ad_mtx);
out:
/*
* We need to check if we evict all buffers, otherwise we may iterate
* unnecessarily.
*/
if (!all && rerun) {
/*
* Bump device hand to the device start if it is approaching the
* end. l2arc_evict() has already evicted ahead for this case.
*/
dev->l2ad_hand = dev->l2ad_start;
dev->l2ad_evict = dev->l2ad_start;
dev->l2ad_first = B_FALSE;
goto top;
}
if (!all) {
/*
* In case of cache device removal (all) the following
* assertions may be violated without functional consequences
* as the device is about to be removed.
*/
ASSERT3U(dev->l2ad_hand + distance, <, dev->l2ad_end);
if (!dev->l2ad_first)
ASSERT3U(dev->l2ad_hand, <, dev->l2ad_evict);
}
}
/*
* Handle any abd transforms that might be required for writing to the L2ARC.
* If successful, this function will always return an abd with the data
* transformed as it is on disk in a new abd of asize bytes.
*/
static int
l2arc_apply_transforms(spa_t *spa, arc_buf_hdr_t *hdr, uint64_t asize,
abd_t **abd_out)
{
int ret;
void *tmp = NULL;
abd_t *cabd = NULL, *eabd = NULL, *to_write = hdr->b_l1hdr.b_pabd;
enum zio_compress compress = HDR_GET_COMPRESS(hdr);
uint64_t psize = HDR_GET_PSIZE(hdr);
uint64_t size = arc_hdr_size(hdr);
boolean_t ismd = HDR_ISTYPE_METADATA(hdr);
boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS);
dsl_crypto_key_t *dck = NULL;
uint8_t mac[ZIO_DATA_MAC_LEN] = { 0 };
boolean_t no_crypt = B_FALSE;
ASSERT((HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
!HDR_COMPRESSION_ENABLED(hdr)) ||
HDR_ENCRYPTED(hdr) || HDR_SHARED_DATA(hdr) || psize != asize);
ASSERT3U(psize, <=, asize);
/*
* If this data simply needs its own buffer, we simply allocate it
* and copy the data. This may be done to eliminate a dependency on a
* shared buffer or to reallocate the buffer to match asize.
*/
if (HDR_HAS_RABD(hdr) && asize != psize) {
ASSERT3U(asize, >=, psize);
to_write = abd_alloc_for_io(asize, ismd);
abd_copy(to_write, hdr->b_crypt_hdr.b_rabd, psize);
if (psize != asize)
abd_zero_off(to_write, psize, asize - psize);
goto out;
}
if ((compress == ZIO_COMPRESS_OFF || HDR_COMPRESSION_ENABLED(hdr)) &&
!HDR_ENCRYPTED(hdr)) {
ASSERT3U(size, ==, psize);
to_write = abd_alloc_for_io(asize, ismd);
abd_copy(to_write, hdr->b_l1hdr.b_pabd, size);
if (size != asize)
abd_zero_off(to_write, size, asize - size);
goto out;
}
if (compress != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) {
cabd = abd_alloc_for_io(asize, ismd);
tmp = abd_borrow_buf(cabd, asize);
psize = zio_compress_data(compress, to_write, tmp, size,
hdr->b_complevel);
if (psize >= size) {
abd_return_buf(cabd, tmp, asize);
HDR_SET_COMPRESS(hdr, ZIO_COMPRESS_OFF);
to_write = cabd;
abd_copy(to_write, hdr->b_l1hdr.b_pabd, size);
if (size != asize)
abd_zero_off(to_write, size, asize - size);
goto encrypt;
}
ASSERT3U(psize, <=, HDR_GET_PSIZE(hdr));
if (psize < asize)
bzero((char *)tmp + psize, asize - psize);
psize = HDR_GET_PSIZE(hdr);
abd_return_buf_copy(cabd, tmp, asize);
to_write = cabd;
}
encrypt:
if (HDR_ENCRYPTED(hdr)) {
eabd = abd_alloc_for_io(asize, ismd);
/*
* If the dataset was disowned before the buffer
* made it to this point, the key to re-encrypt
* it won't be available. In this case we simply
* won't write the buffer to the L2ARC.
*/
ret = spa_keystore_lookup_key(spa, hdr->b_crypt_hdr.b_dsobj,
FTAG, &dck);
if (ret != 0)
goto error;
ret = zio_do_crypt_abd(B_TRUE, &dck->dck_key,
hdr->b_crypt_hdr.b_ot, bswap, hdr->b_crypt_hdr.b_salt,
hdr->b_crypt_hdr.b_iv, mac, psize, to_write, eabd,
&no_crypt);
if (ret != 0)
goto error;
if (no_crypt)
abd_copy(eabd, to_write, psize);
if (psize != asize)
abd_zero_off(eabd, psize, asize - psize);
/* assert that the MAC we got here matches the one we saved */
ASSERT0(bcmp(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN));
spa_keystore_dsl_key_rele(spa, dck, FTAG);
if (to_write == cabd)
abd_free(cabd);
to_write = eabd;
}
out:
ASSERT3P(to_write, !=, hdr->b_l1hdr.b_pabd);
*abd_out = to_write;
return (0);
error:
if (dck != NULL)
spa_keystore_dsl_key_rele(spa, dck, FTAG);
if (cabd != NULL)
abd_free(cabd);
if (eabd != NULL)
abd_free(eabd);
*abd_out = NULL;
return (ret);
}
static void
l2arc_blk_fetch_done(zio_t *zio)
{
l2arc_read_callback_t *cb;
cb = zio->io_private;
if (cb->l2rcb_abd != NULL)
abd_free(cb->l2rcb_abd);
kmem_free(cb, sizeof (l2arc_read_callback_t));
}
/*
* Find and write ARC buffers to the L2ARC device.
*
* An ARC_FLAG_L2_WRITING flag is set so that the L2ARC buffers are not valid
* for reading until they have completed writing.
* The headroom_boost is an in-out parameter used to maintain headroom boost
* state between calls to this function.
*
* Returns the number of bytes actually written (which may be smaller than
* the delta by which the device hand has changed due to alignment and the
* writing of log blocks).
*/
static uint64_t
l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz)
{
arc_buf_hdr_t *hdr, *hdr_prev, *head;
uint64_t write_asize, write_psize, write_lsize, headroom;
boolean_t full;
l2arc_write_callback_t *cb = NULL;
zio_t *pio, *wzio;
uint64_t guid = spa_load_guid(spa);
l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
ASSERT3P(dev->l2ad_vdev, !=, NULL);
pio = NULL;
write_lsize = write_asize = write_psize = 0;
full = B_FALSE;
head = kmem_cache_alloc(hdr_l2only_cache, KM_PUSHPAGE);
arc_hdr_set_flags(head, ARC_FLAG_L2_WRITE_HEAD | ARC_FLAG_HAS_L2HDR);
/*
* Copy buffers for L2ARC writing.
*/
for (int pass = 0; pass < L2ARC_FEED_TYPES; pass++) {
/*
* If pass == 1 or 3, we cache MRU metadata and data
* respectively.
*/
if (l2arc_mfuonly) {
if (pass == 1 || pass == 3)
continue;
}
multilist_sublist_t *mls = l2arc_sublist_lock(pass);
uint64_t passed_sz = 0;
VERIFY3P(mls, !=, NULL);
/*
* L2ARC fast warmup.
*
* Until the ARC is warm and starts to evict, read from the
* head of the ARC lists rather than the tail.
*/
if (arc_warm == B_FALSE)
hdr = multilist_sublist_head(mls);
else
hdr = multilist_sublist_tail(mls);
headroom = target_sz * l2arc_headroom;
if (zfs_compressed_arc_enabled)
headroom = (headroom * l2arc_headroom_boost) / 100;
for (; hdr; hdr = hdr_prev) {
kmutex_t *hash_lock;
abd_t *to_write = NULL;
if (arc_warm == B_FALSE)
hdr_prev = multilist_sublist_next(mls, hdr);
else
hdr_prev = multilist_sublist_prev(mls, hdr);
hash_lock = HDR_LOCK(hdr);
if (!mutex_tryenter(hash_lock)) {
/*
* Skip this buffer rather than waiting.
*/
continue;
}
passed_sz += HDR_GET_LSIZE(hdr);
if (l2arc_headroom != 0 && passed_sz > headroom) {
/*
* Searched too far.
*/
mutex_exit(hash_lock);
break;
}
if (!l2arc_write_eligible(guid, hdr)) {
mutex_exit(hash_lock);
continue;
}
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT3U(HDR_GET_PSIZE(hdr), >, 0);
ASSERT3U(arc_hdr_size(hdr), >, 0);
ASSERT(hdr->b_l1hdr.b_pabd != NULL ||
HDR_HAS_RABD(hdr));
uint64_t psize = HDR_GET_PSIZE(hdr);
uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev,
psize);
if ((write_asize + asize) > target_sz) {
full = B_TRUE;
mutex_exit(hash_lock);
break;
}
/*
* We rely on the L1 portion of the header below, so
* it's invalid for this header to have been evicted out
* of the ghost cache, prior to being written out. The
* ARC_FLAG_L2_WRITING bit ensures this won't happen.
*/
arc_hdr_set_flags(hdr, ARC_FLAG_L2_WRITING);
/*
* If this header has b_rabd, we can use this since it
* must always match the data exactly as it exists on
* disk. Otherwise, the L2ARC can normally use the
* hdr's data, but if we're sharing data between the
* hdr and one of its bufs, L2ARC needs its own copy of
* the data so that the ZIO below can't race with the
* buf consumer. To ensure that this copy will be
* available for the lifetime of the ZIO and be cleaned
* up afterwards, we add it to the l2arc_free_on_write
* queue. If we need to apply any transforms to the
* data (compression, encryption) we will also need the
* extra buffer.
*/
if (HDR_HAS_RABD(hdr) && psize == asize) {
to_write = hdr->b_crypt_hdr.b_rabd;
} else if ((HDR_COMPRESSION_ENABLED(hdr) ||
HDR_GET_COMPRESS(hdr) == ZIO_COMPRESS_OFF) &&
!HDR_ENCRYPTED(hdr) && !HDR_SHARED_DATA(hdr) &&
psize == asize) {
to_write = hdr->b_l1hdr.b_pabd;
} else {
int ret;
arc_buf_contents_t type = arc_buf_type(hdr);
ret = l2arc_apply_transforms(spa, hdr, asize,
&to_write);
if (ret != 0) {
arc_hdr_clear_flags(hdr,
ARC_FLAG_L2_WRITING);
mutex_exit(hash_lock);
continue;
}
l2arc_free_abd_on_write(to_write, asize, type);
}
if (pio == NULL) {
/*
* Insert a dummy header on the buflist so
* l2arc_write_done() can find where the
* write buffers begin without searching.
*/
mutex_enter(&dev->l2ad_mtx);
list_insert_head(&dev->l2ad_buflist, head);
mutex_exit(&dev->l2ad_mtx);
cb = kmem_alloc(
sizeof (l2arc_write_callback_t), KM_SLEEP);
cb->l2wcb_dev = dev;
cb->l2wcb_head = head;
/*
* Create a list to save allocated abd buffers
* for l2arc_log_blk_commit().
*/
list_create(&cb->l2wcb_abd_list,
sizeof (l2arc_lb_abd_buf_t),
offsetof(l2arc_lb_abd_buf_t, node));
pio = zio_root(spa, l2arc_write_done, cb,
ZIO_FLAG_CANFAIL);
}
hdr->b_l2hdr.b_dev = dev;
hdr->b_l2hdr.b_hits = 0;
hdr->b_l2hdr.b_daddr = dev->l2ad_hand;
hdr->b_l2hdr.b_arcs_state =
hdr->b_l1hdr.b_state->arcs_state;
arc_hdr_set_flags(hdr, ARC_FLAG_HAS_L2HDR);
mutex_enter(&dev->l2ad_mtx);
list_insert_head(&dev->l2ad_buflist, hdr);
mutex_exit(&dev->l2ad_mtx);
(void) zfs_refcount_add_many(&dev->l2ad_alloc,
arc_hdr_size(hdr), hdr);
wzio = zio_write_phys(pio, dev->l2ad_vdev,
hdr->b_l2hdr.b_daddr, asize, to_write,
ZIO_CHECKSUM_OFF, NULL, hdr,
ZIO_PRIORITY_ASYNC_WRITE,
ZIO_FLAG_CANFAIL, B_FALSE);
write_lsize += HDR_GET_LSIZE(hdr);
DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev,
zio_t *, wzio);
write_psize += psize;
write_asize += asize;
dev->l2ad_hand += asize;
l2arc_hdr_arcstats_increment(hdr);
vdev_space_update(dev->l2ad_vdev, asize, 0, 0);
mutex_exit(hash_lock);
/*
* Append buf info to current log and commit if full.
* arcstat_l2_{size,asize} kstats are updated
* internally.
*/
if (l2arc_log_blk_insert(dev, hdr))
l2arc_log_blk_commit(dev, pio, cb);
zio_nowait(wzio);
}
multilist_sublist_unlock(mls);
if (full == B_TRUE)
break;
}
/* No buffers selected for writing? */
if (pio == NULL) {
ASSERT0(write_lsize);
ASSERT(!HDR_HAS_L1HDR(head));
kmem_cache_free(hdr_l2only_cache, head);
/*
* Although we did not write any buffers l2ad_evict may
* have advanced.
*/
if (dev->l2ad_evict != l2dhdr->dh_evict)
l2arc_dev_hdr_update(dev);
return (0);
}
if (!dev->l2ad_first)
ASSERT3U(dev->l2ad_hand, <=, dev->l2ad_evict);
ASSERT3U(write_asize, <=, target_sz);
ARCSTAT_BUMP(arcstat_l2_writes_sent);
ARCSTAT_INCR(arcstat_l2_write_bytes, write_psize);
dev->l2ad_writing = B_TRUE;
(void) zio_wait(pio);
dev->l2ad_writing = B_FALSE;
/*
* Update the device header after the zio completes as
* l2arc_write_done() may have updated the memory holding the log block
* pointers in the device header.
*/
l2arc_dev_hdr_update(dev);
return (write_asize);
}
static boolean_t
l2arc_hdr_limit_reached(void)
{
int64_t s = aggsum_upper_bound(&arc_sums.arcstat_l2_hdr_size);
return (arc_reclaim_needed() || (s > arc_meta_limit * 3 / 4) ||
(s > (arc_warm ? arc_c : arc_c_max) * l2arc_meta_percent / 100));
}
/*
* This thread feeds the L2ARC at regular intervals. This is the beating
* heart of the L2ARC.
*/
static void
l2arc_feed_thread(void *unused)
{
(void) unused;
callb_cpr_t cpr;
l2arc_dev_t *dev;
spa_t *spa;
uint64_t size, wrote;
clock_t begin, next = ddi_get_lbolt();
fstrans_cookie_t cookie;
CALLB_CPR_INIT(&cpr, &l2arc_feed_thr_lock, callb_generic_cpr, FTAG);
mutex_enter(&l2arc_feed_thr_lock);
cookie = spl_fstrans_mark();
while (l2arc_thread_exit == 0) {
CALLB_CPR_SAFE_BEGIN(&cpr);
(void) cv_timedwait_idle(&l2arc_feed_thr_cv,
&l2arc_feed_thr_lock, next);
CALLB_CPR_SAFE_END(&cpr, &l2arc_feed_thr_lock);
next = ddi_get_lbolt() + hz;
/*
* Quick check for L2ARC devices.
*/
mutex_enter(&l2arc_dev_mtx);
if (l2arc_ndev == 0) {
mutex_exit(&l2arc_dev_mtx);
continue;
}
mutex_exit(&l2arc_dev_mtx);
begin = ddi_get_lbolt();
/*
* This selects the next l2arc device to write to, and in
* doing so the next spa to feed from: dev->l2ad_spa. This
* will return NULL if there are now no l2arc devices or if
* they are all faulted.
*
* If a device is returned, its spa's config lock is also
* held to prevent device removal. l2arc_dev_get_next()
* will grab and release l2arc_dev_mtx.
*/
if ((dev = l2arc_dev_get_next()) == NULL)
continue;
spa = dev->l2ad_spa;
ASSERT3P(spa, !=, NULL);
/*
* If the pool is read-only then force the feed thread to
* sleep a little longer.
*/
if (!spa_writeable(spa)) {
next = ddi_get_lbolt() + 5 * l2arc_feed_secs * hz;
spa_config_exit(spa, SCL_L2ARC, dev);
continue;
}
/*
* Avoid contributing to memory pressure.
*/
if (l2arc_hdr_limit_reached()) {
ARCSTAT_BUMP(arcstat_l2_abort_lowmem);
spa_config_exit(spa, SCL_L2ARC, dev);
continue;
}
ARCSTAT_BUMP(arcstat_l2_feeds);
size = l2arc_write_size(dev);
/*
* Evict L2ARC buffers that will be overwritten.
*/
l2arc_evict(dev, size, B_FALSE);
/*
* Write ARC buffers.
*/
wrote = l2arc_write_buffers(spa, dev, size);
/*
* Calculate interval between writes.
*/
next = l2arc_write_interval(begin, size, wrote);
spa_config_exit(spa, SCL_L2ARC, dev);
}
spl_fstrans_unmark(cookie);
l2arc_thread_exit = 0;
cv_broadcast(&l2arc_feed_thr_cv);
CALLB_CPR_EXIT(&cpr); /* drops l2arc_feed_thr_lock */
thread_exit();
}
boolean_t
l2arc_vdev_present(vdev_t *vd)
{
return (l2arc_vdev_get(vd) != NULL);
}
/*
* Returns the l2arc_dev_t associated with a particular vdev_t or NULL if
* the vdev_t isn't an L2ARC device.
*/
l2arc_dev_t *
l2arc_vdev_get(vdev_t *vd)
{
l2arc_dev_t *dev;
mutex_enter(&l2arc_dev_mtx);
for (dev = list_head(l2arc_dev_list); dev != NULL;
dev = list_next(l2arc_dev_list, dev)) {
if (dev->l2ad_vdev == vd)
break;
}
mutex_exit(&l2arc_dev_mtx);
return (dev);
}
static void
l2arc_rebuild_dev(l2arc_dev_t *dev, boolean_t reopen)
{
l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize;
spa_t *spa = dev->l2ad_spa;
/*
* The L2ARC has to hold at least the payload of one log block for
* them to be restored (persistent L2ARC). The payload of a log block
* depends on the amount of its log entries. We always write log blocks
* with 1022 entries. How many of them are committed or restored depends
* on the size of the L2ARC device. Thus the maximum payload of
* one log block is 1022 * SPA_MAXBLOCKSIZE = 16GB. If the L2ARC device
* is less than that, we reduce the amount of committed and restored
* log entries per block so as to enable persistence.
*/
if (dev->l2ad_end < l2arc_rebuild_blocks_min_l2size) {
dev->l2ad_log_entries = 0;
} else {
dev->l2ad_log_entries = MIN((dev->l2ad_end -
dev->l2ad_start) >> SPA_MAXBLOCKSHIFT,
L2ARC_LOG_BLK_MAX_ENTRIES);
}
/*
* Read the device header, if an error is returned do not rebuild L2ARC.
*/
if (l2arc_dev_hdr_read(dev) == 0 && dev->l2ad_log_entries > 0) {
/*
* If we are onlining a cache device (vdev_reopen) that was
* still present (l2arc_vdev_present()) and rebuild is enabled,
* we should evict all ARC buffers and pointers to log blocks
* and reclaim their space before restoring its contents to
* L2ARC.
*/
if (reopen) {
if (!l2arc_rebuild_enabled) {
return;
} else {
l2arc_evict(dev, 0, B_TRUE);
/* start a new log block */
dev->l2ad_log_ent_idx = 0;
dev->l2ad_log_blk_payload_asize = 0;
dev->l2ad_log_blk_payload_start = 0;
}
}
/*
* Just mark the device as pending for a rebuild. We won't
* be starting a rebuild in line here as it would block pool
* import. Instead spa_load_impl will hand that off to an
* async task which will call l2arc_spa_rebuild_start.
*/
dev->l2ad_rebuild = B_TRUE;
} else if (spa_writeable(spa)) {
/*
* In this case TRIM the whole device if l2arc_trim_ahead > 0,
* otherwise create a new header. We zero out the memory holding
* the header to reset dh_start_lbps. If we TRIM the whole
* device the new header will be written by
* vdev_trim_l2arc_thread() at the end of the TRIM to update the
* trim_state in the header too. When reading the header, if
* trim_state is not VDEV_TRIM_COMPLETE and l2arc_trim_ahead > 0
* we opt to TRIM the whole device again.
*/
if (l2arc_trim_ahead > 0) {
dev->l2ad_trim_all = B_TRUE;
} else {
bzero(l2dhdr, l2dhdr_asize);
l2arc_dev_hdr_update(dev);
}
}
}
/*
* Add a vdev for use by the L2ARC. By this point the spa has already
* validated the vdev and opened it.
*/
void
l2arc_add_vdev(spa_t *spa, vdev_t *vd)
{
l2arc_dev_t *adddev;
uint64_t l2dhdr_asize;
ASSERT(!l2arc_vdev_present(vd));
/*
* Create a new l2arc device entry.
*/
adddev = vmem_zalloc(sizeof (l2arc_dev_t), KM_SLEEP);
adddev->l2ad_spa = spa;
adddev->l2ad_vdev = vd;
/* leave extra size for an l2arc device header */
l2dhdr_asize = adddev->l2ad_dev_hdr_asize =
MAX(sizeof (*adddev->l2ad_dev_hdr), 1 << vd->vdev_ashift);
adddev->l2ad_start = VDEV_LABEL_START_SIZE + l2dhdr_asize;
adddev->l2ad_end = VDEV_LABEL_START_SIZE + vdev_get_min_asize(vd);
ASSERT3U(adddev->l2ad_start, <, adddev->l2ad_end);
adddev->l2ad_hand = adddev->l2ad_start;
adddev->l2ad_evict = adddev->l2ad_start;
adddev->l2ad_first = B_TRUE;
adddev->l2ad_writing = B_FALSE;
adddev->l2ad_trim_all = B_FALSE;
list_link_init(&adddev->l2ad_node);
adddev->l2ad_dev_hdr = kmem_zalloc(l2dhdr_asize, KM_SLEEP);
mutex_init(&adddev->l2ad_mtx, NULL, MUTEX_DEFAULT, NULL);
/*
* This is a list of all ARC buffers that are still valid on the
* device.
*/
list_create(&adddev->l2ad_buflist, sizeof (arc_buf_hdr_t),
offsetof(arc_buf_hdr_t, b_l2hdr.b_l2node));
/*
* This is a list of pointers to log blocks that are still present
* on the device.
*/
list_create(&adddev->l2ad_lbptr_list, sizeof (l2arc_lb_ptr_buf_t),
offsetof(l2arc_lb_ptr_buf_t, node));
vdev_space_update(vd, 0, 0, adddev->l2ad_end - adddev->l2ad_hand);
zfs_refcount_create(&adddev->l2ad_alloc);
zfs_refcount_create(&adddev->l2ad_lb_asize);
zfs_refcount_create(&adddev->l2ad_lb_count);
/*
* Decide if dev is eligible for L2ARC rebuild or whole device
* trimming. This has to happen before the device is added in the
* cache device list and l2arc_dev_mtx is released. Otherwise
* l2arc_feed_thread() might already start writing on the
* device.
*/
l2arc_rebuild_dev(adddev, B_FALSE);
/*
* Add device to global list
*/
mutex_enter(&l2arc_dev_mtx);
list_insert_head(l2arc_dev_list, adddev);
atomic_inc_64(&l2arc_ndev);
mutex_exit(&l2arc_dev_mtx);
}
/*
* Decide if a vdev is eligible for L2ARC rebuild, called from vdev_reopen()
* in case of onlining a cache device.
*/
void
l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen)
{
l2arc_dev_t *dev = NULL;
dev = l2arc_vdev_get(vd);
ASSERT3P(dev, !=, NULL);
/*
* In contrast to l2arc_add_vdev() we do not have to worry about
* l2arc_feed_thread() invalidating previous content when onlining a
* cache device. The device parameters (l2ad*) are not cleared when
* offlining the device and writing new buffers will not invalidate
* all previous content. In worst case only buffers that have not had
* their log block written to the device will be lost.
* When onlining the cache device (ie offline->online without exporting
* the pool in between) this happens:
* vdev_reopen() -> vdev_open() -> l2arc_rebuild_vdev()
* | |
* vdev_is_dead() = B_FALSE l2ad_rebuild = B_TRUE
* During the time where vdev_is_dead = B_FALSE and until l2ad_rebuild
* is set to B_TRUE we might write additional buffers to the device.
*/
l2arc_rebuild_dev(dev, reopen);
}
/*
* Remove a vdev from the L2ARC.
*/
void
l2arc_remove_vdev(vdev_t *vd)
{
l2arc_dev_t *remdev = NULL;
/*
* Find the device by vdev
*/
remdev = l2arc_vdev_get(vd);
ASSERT3P(remdev, !=, NULL);
/*
* Cancel any ongoing or scheduled rebuild.
*/
mutex_enter(&l2arc_rebuild_thr_lock);
if (remdev->l2ad_rebuild_began == B_TRUE) {
remdev->l2ad_rebuild_cancel = B_TRUE;
while (remdev->l2ad_rebuild == B_TRUE)
cv_wait(&l2arc_rebuild_thr_cv, &l2arc_rebuild_thr_lock);
}
mutex_exit(&l2arc_rebuild_thr_lock);
/*
* Remove device from global list
*/
mutex_enter(&l2arc_dev_mtx);
list_remove(l2arc_dev_list, remdev);
l2arc_dev_last = NULL; /* may have been invalidated */
atomic_dec_64(&l2arc_ndev);
mutex_exit(&l2arc_dev_mtx);
/*
* Clear all buflists and ARC references. L2ARC device flush.
*/
l2arc_evict(remdev, 0, B_TRUE);
list_destroy(&remdev->l2ad_buflist);
ASSERT(list_is_empty(&remdev->l2ad_lbptr_list));
list_destroy(&remdev->l2ad_lbptr_list);
mutex_destroy(&remdev->l2ad_mtx);
zfs_refcount_destroy(&remdev->l2ad_alloc);
zfs_refcount_destroy(&remdev->l2ad_lb_asize);
zfs_refcount_destroy(&remdev->l2ad_lb_count);
kmem_free(remdev->l2ad_dev_hdr, remdev->l2ad_dev_hdr_asize);
vmem_free(remdev, sizeof (l2arc_dev_t));
}
void
l2arc_init(void)
{
l2arc_thread_exit = 0;
l2arc_ndev = 0;
mutex_init(&l2arc_feed_thr_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&l2arc_feed_thr_cv, NULL, CV_DEFAULT, NULL);
mutex_init(&l2arc_rebuild_thr_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&l2arc_rebuild_thr_cv, NULL, CV_DEFAULT, NULL);
mutex_init(&l2arc_dev_mtx, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&l2arc_free_on_write_mtx, NULL, MUTEX_DEFAULT, NULL);
l2arc_dev_list = &L2ARC_dev_list;
l2arc_free_on_write = &L2ARC_free_on_write;
list_create(l2arc_dev_list, sizeof (l2arc_dev_t),
offsetof(l2arc_dev_t, l2ad_node));
list_create(l2arc_free_on_write, sizeof (l2arc_data_free_t),
offsetof(l2arc_data_free_t, l2df_list_node));
}
void
l2arc_fini(void)
{
mutex_destroy(&l2arc_feed_thr_lock);
cv_destroy(&l2arc_feed_thr_cv);
mutex_destroy(&l2arc_rebuild_thr_lock);
cv_destroy(&l2arc_rebuild_thr_cv);
mutex_destroy(&l2arc_dev_mtx);
mutex_destroy(&l2arc_free_on_write_mtx);
list_destroy(l2arc_dev_list);
list_destroy(l2arc_free_on_write);
}
void
l2arc_start(void)
{
if (!(spa_mode_global & SPA_MODE_WRITE))
return;
(void) thread_create(NULL, 0, l2arc_feed_thread, NULL, 0, &p0,
TS_RUN, defclsyspri);
}
void
l2arc_stop(void)
{
if (!(spa_mode_global & SPA_MODE_WRITE))
return;
mutex_enter(&l2arc_feed_thr_lock);
cv_signal(&l2arc_feed_thr_cv); /* kick thread out of startup */
l2arc_thread_exit = 1;
while (l2arc_thread_exit != 0)
cv_wait(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock);
mutex_exit(&l2arc_feed_thr_lock);
}
/*
* Punches out rebuild threads for the L2ARC devices in a spa. This should
* be called after pool import from the spa async thread, since starting
* these threads directly from spa_import() will make them part of the
* "zpool import" context and delay process exit (and thus pool import).
*/
void
l2arc_spa_rebuild_start(spa_t *spa)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
/*
* Locate the spa's l2arc devices and kick off rebuild threads.
*/
for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
l2arc_dev_t *dev =
l2arc_vdev_get(spa->spa_l2cache.sav_vdevs[i]);
if (dev == NULL) {
/* Don't attempt a rebuild if the vdev is UNAVAIL */
continue;
}
mutex_enter(&l2arc_rebuild_thr_lock);
if (dev->l2ad_rebuild && !dev->l2ad_rebuild_cancel) {
dev->l2ad_rebuild_began = B_TRUE;
(void) thread_create(NULL, 0, l2arc_dev_rebuild_thread,
dev, 0, &p0, TS_RUN, minclsyspri);
}
mutex_exit(&l2arc_rebuild_thr_lock);
}
}
/*
* Main entry point for L2ARC rebuilding.
*/
static void
l2arc_dev_rebuild_thread(void *arg)
{
l2arc_dev_t *dev = arg;
VERIFY(!dev->l2ad_rebuild_cancel);
VERIFY(dev->l2ad_rebuild);
(void) l2arc_rebuild(dev);
mutex_enter(&l2arc_rebuild_thr_lock);
dev->l2ad_rebuild_began = B_FALSE;
dev->l2ad_rebuild = B_FALSE;
mutex_exit(&l2arc_rebuild_thr_lock);
thread_exit();
}
/*
* This function implements the actual L2ARC metadata rebuild. It:
* starts reading the log block chain and restores each block's contents
* to memory (reconstructing arc_buf_hdr_t's).
*
* Operation stops under any of the following conditions:
*
* 1) We reach the end of the log block chain.
* 2) We encounter *any* error condition (cksum errors, io errors)
*/
static int
l2arc_rebuild(l2arc_dev_t *dev)
{
vdev_t *vd = dev->l2ad_vdev;
spa_t *spa = vd->vdev_spa;
int err = 0;
l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
l2arc_log_blk_phys_t *this_lb, *next_lb;
zio_t *this_io = NULL, *next_io = NULL;
l2arc_log_blkptr_t lbps[2];
l2arc_lb_ptr_buf_t *lb_ptr_buf;
boolean_t lock_held;
this_lb = vmem_zalloc(sizeof (*this_lb), KM_SLEEP);
next_lb = vmem_zalloc(sizeof (*next_lb), KM_SLEEP);
/*
* We prevent device removal while issuing reads to the device,
* then during the rebuilding phases we drop this lock again so
* that a spa_unload or device remove can be initiated - this is
* safe, because the spa will signal us to stop before removing
* our device and wait for us to stop.
*/
spa_config_enter(spa, SCL_L2ARC, vd, RW_READER);
lock_held = B_TRUE;
/*
* Retrieve the persistent L2ARC device state.
* L2BLK_GET_PSIZE returns aligned size for log blocks.
*/
dev->l2ad_evict = MAX(l2dhdr->dh_evict, dev->l2ad_start);
dev->l2ad_hand = MAX(l2dhdr->dh_start_lbps[0].lbp_daddr +
L2BLK_GET_PSIZE((&l2dhdr->dh_start_lbps[0])->lbp_prop),
dev->l2ad_start);
dev->l2ad_first = !!(l2dhdr->dh_flags & L2ARC_DEV_HDR_EVICT_FIRST);
vd->vdev_trim_action_time = l2dhdr->dh_trim_action_time;
vd->vdev_trim_state = l2dhdr->dh_trim_state;
/*
* In case the zfs module parameter l2arc_rebuild_enabled is false
* we do not start the rebuild process.
*/
if (!l2arc_rebuild_enabled)
goto out;
/* Prepare the rebuild process */
bcopy(l2dhdr->dh_start_lbps, lbps, sizeof (lbps));
/* Start the rebuild process */
for (;;) {
if (!l2arc_log_blkptr_valid(dev, &lbps[0]))
break;
if ((err = l2arc_log_blk_read(dev, &lbps[0], &lbps[1],
this_lb, next_lb, this_io, &next_io)) != 0)
goto out;
/*
* Our memory pressure valve. If the system is running low
* on memory, rather than swamping memory with new ARC buf
* hdrs, we opt not to rebuild the L2ARC. At this point,
* however, we have already set up our L2ARC dev to chain in
* new metadata log blocks, so the user may choose to offline/
* online the L2ARC dev at a later time (or re-import the pool)
* to reconstruct it (when there's less memory pressure).
*/
if (l2arc_hdr_limit_reached()) {
ARCSTAT_BUMP(arcstat_l2_rebuild_abort_lowmem);
cmn_err(CE_NOTE, "System running low on memory, "
"aborting L2ARC rebuild.");
err = SET_ERROR(ENOMEM);
goto out;
}
spa_config_exit(spa, SCL_L2ARC, vd);
lock_held = B_FALSE;
/*
* Now that we know that the next_lb checks out alright, we
* can start reconstruction from this log block.
* L2BLK_GET_PSIZE returns aligned size for log blocks.
*/
uint64_t asize = L2BLK_GET_PSIZE((&lbps[0])->lbp_prop);
l2arc_log_blk_restore(dev, this_lb, asize);
/*
* log block restored, include its pointer in the list of
* pointers to log blocks present in the L2ARC device.
*/
lb_ptr_buf = kmem_zalloc(sizeof (l2arc_lb_ptr_buf_t), KM_SLEEP);
lb_ptr_buf->lb_ptr = kmem_zalloc(sizeof (l2arc_log_blkptr_t),
KM_SLEEP);
bcopy(&lbps[0], lb_ptr_buf->lb_ptr,
sizeof (l2arc_log_blkptr_t));
mutex_enter(&dev->l2ad_mtx);
list_insert_tail(&dev->l2ad_lbptr_list, lb_ptr_buf);
ARCSTAT_INCR(arcstat_l2_log_blk_asize, asize);
ARCSTAT_BUMP(arcstat_l2_log_blk_count);
zfs_refcount_add_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf);
zfs_refcount_add(&dev->l2ad_lb_count, lb_ptr_buf);
mutex_exit(&dev->l2ad_mtx);
vdev_space_update(vd, asize, 0, 0);
/*
* Protection against loops of log blocks:
*
* l2ad_hand l2ad_evict
* V V
* l2ad_start |=======================================| l2ad_end
* -----|||----|||---|||----|||
* (3) (2) (1) (0)
* ---|||---|||----|||---|||
* (7) (6) (5) (4)
*
* In this situation the pointer of log block (4) passes
* l2arc_log_blkptr_valid() but the log block should not be
* restored as it is overwritten by the payload of log block
* (0). Only log blocks (0)-(3) should be restored. We check
* whether l2ad_evict lies in between the payload starting
* offset of the next log block (lbps[1].lbp_payload_start)
* and the payload starting offset of the present log block
* (lbps[0].lbp_payload_start). If true and this isn't the
* first pass, we are looping from the beginning and we should
* stop.
*/
if (l2arc_range_check_overlap(lbps[1].lbp_payload_start,
lbps[0].lbp_payload_start, dev->l2ad_evict) &&
!dev->l2ad_first)
goto out;
cond_resched();
for (;;) {
mutex_enter(&l2arc_rebuild_thr_lock);
if (dev->l2ad_rebuild_cancel) {
dev->l2ad_rebuild = B_FALSE;
cv_signal(&l2arc_rebuild_thr_cv);
mutex_exit(&l2arc_rebuild_thr_lock);
err = SET_ERROR(ECANCELED);
goto out;
}
mutex_exit(&l2arc_rebuild_thr_lock);
if (spa_config_tryenter(spa, SCL_L2ARC, vd,
RW_READER)) {
lock_held = B_TRUE;
break;
}
/*
* L2ARC config lock held by somebody in writer,
* possibly due to them trying to remove us. They'll
* likely to want us to shut down, so after a little
* delay, we check l2ad_rebuild_cancel and retry
* the lock again.
*/
delay(1);
}
/*
* Continue with the next log block.
*/
lbps[0] = lbps[1];
lbps[1] = this_lb->lb_prev_lbp;
PTR_SWAP(this_lb, next_lb);
this_io = next_io;
next_io = NULL;
}
if (this_io != NULL)
l2arc_log_blk_fetch_abort(this_io);
out:
if (next_io != NULL)
l2arc_log_blk_fetch_abort(next_io);
vmem_free(this_lb, sizeof (*this_lb));
vmem_free(next_lb, sizeof (*next_lb));
if (!l2arc_rebuild_enabled) {
spa_history_log_internal(spa, "L2ARC rebuild", NULL,
"disabled");
} else if (err == 0 && zfs_refcount_count(&dev->l2ad_lb_count) > 0) {
ARCSTAT_BUMP(arcstat_l2_rebuild_success);
spa_history_log_internal(spa, "L2ARC rebuild", NULL,
"successful, restored %llu blocks",
(u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count));
} else if (err == 0 && zfs_refcount_count(&dev->l2ad_lb_count) == 0) {
/*
* No error but also nothing restored, meaning the lbps array
* in the device header points to invalid/non-present log
* blocks. Reset the header.
*/
spa_history_log_internal(spa, "L2ARC rebuild", NULL,
"no valid log blocks");
bzero(l2dhdr, dev->l2ad_dev_hdr_asize);
l2arc_dev_hdr_update(dev);
} else if (err == ECANCELED) {
/*
* In case the rebuild was canceled do not log to spa history
* log as the pool may be in the process of being removed.
*/
zfs_dbgmsg("L2ARC rebuild aborted, restored %llu blocks",
(u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count));
} else if (err != 0) {
spa_history_log_internal(spa, "L2ARC rebuild", NULL,
"aborted, restored %llu blocks",
(u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count));
}
if (lock_held)
spa_config_exit(spa, SCL_L2ARC, vd);
return (err);
}
/*
* Attempts to read the device header on the provided L2ARC device and writes
* it to `hdr'. On success, this function returns 0, otherwise the appropriate
* error code is returned.
*/
static int
l2arc_dev_hdr_read(l2arc_dev_t *dev)
{
int err;
uint64_t guid;
l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
const uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize;
abd_t *abd;
guid = spa_guid(dev->l2ad_vdev->vdev_spa);
abd = abd_get_from_buf(l2dhdr, l2dhdr_asize);
err = zio_wait(zio_read_phys(NULL, dev->l2ad_vdev,
VDEV_LABEL_START_SIZE, l2dhdr_asize, abd,
ZIO_CHECKSUM_LABEL, NULL, NULL, ZIO_PRIORITY_SYNC_READ,
ZIO_FLAG_DONT_CACHE | ZIO_FLAG_CANFAIL |
ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY |
ZIO_FLAG_SPECULATIVE, B_FALSE));
abd_free(abd);
if (err != 0) {
ARCSTAT_BUMP(arcstat_l2_rebuild_abort_dh_errors);
zfs_dbgmsg("L2ARC IO error (%d) while reading device header, "
"vdev guid: %llu", err,
(u_longlong_t)dev->l2ad_vdev->vdev_guid);
return (err);
}
if (l2dhdr->dh_magic == BSWAP_64(L2ARC_DEV_HDR_MAGIC))
byteswap_uint64_array(l2dhdr, sizeof (*l2dhdr));
if (l2dhdr->dh_magic != L2ARC_DEV_HDR_MAGIC ||
l2dhdr->dh_spa_guid != guid ||
l2dhdr->dh_vdev_guid != dev->l2ad_vdev->vdev_guid ||
l2dhdr->dh_version != L2ARC_PERSISTENT_VERSION ||
l2dhdr->dh_log_entries != dev->l2ad_log_entries ||
l2dhdr->dh_end != dev->l2ad_end ||
!l2arc_range_check_overlap(dev->l2ad_start, dev->l2ad_end,
l2dhdr->dh_evict) ||
(l2dhdr->dh_trim_state != VDEV_TRIM_COMPLETE &&
l2arc_trim_ahead > 0)) {
/*
* Attempt to rebuild a device containing no actual dev hdr
* or containing a header from some other pool or from another
* version of persistent L2ARC.
*/
ARCSTAT_BUMP(arcstat_l2_rebuild_abort_unsupported);
return (SET_ERROR(ENOTSUP));
}
return (0);
}
/*
* Reads L2ARC log blocks from storage and validates their contents.
*
* This function implements a simple fetcher to make sure that while
* we're processing one buffer the L2ARC is already fetching the next
* one in the chain.
*
* The arguments this_lp and next_lp point to the current and next log block
* address in the block chain. Similarly, this_lb and next_lb hold the
* l2arc_log_blk_phys_t's of the current and next L2ARC blk.
*
* The `this_io' and `next_io' arguments are used for block fetching.
* When issuing the first blk IO during rebuild, you should pass NULL for
* `this_io'. This function will then issue a sync IO to read the block and
* also issue an async IO to fetch the next block in the block chain. The
* fetched IO is returned in `next_io'. On subsequent calls to this
* function, pass the value returned in `next_io' from the previous call
* as `this_io' and a fresh `next_io' pointer to hold the next fetch IO.
* Prior to the call, you should initialize your `next_io' pointer to be
* NULL. If no fetch IO was issued, the pointer is left set at NULL.
*
* On success, this function returns 0, otherwise it returns an appropriate
* error code. On error the fetching IO is aborted and cleared before
* returning from this function. Therefore, if we return `success', the
* caller can assume that we have taken care of cleanup of fetch IOs.
*/
static int
l2arc_log_blk_read(l2arc_dev_t *dev,
const l2arc_log_blkptr_t *this_lbp, const l2arc_log_blkptr_t *next_lbp,
l2arc_log_blk_phys_t *this_lb, l2arc_log_blk_phys_t *next_lb,
zio_t *this_io, zio_t **next_io)
{
int err = 0;
zio_cksum_t cksum;
abd_t *abd = NULL;
uint64_t asize;
ASSERT(this_lbp != NULL && next_lbp != NULL);
ASSERT(this_lb != NULL && next_lb != NULL);
ASSERT(next_io != NULL && *next_io == NULL);
ASSERT(l2arc_log_blkptr_valid(dev, this_lbp));
/*
* Check to see if we have issued the IO for this log block in a
* previous run. If not, this is the first call, so issue it now.
*/
if (this_io == NULL) {
this_io = l2arc_log_blk_fetch(dev->l2ad_vdev, this_lbp,
this_lb);
}
/*
* Peek to see if we can start issuing the next IO immediately.
*/
if (l2arc_log_blkptr_valid(dev, next_lbp)) {
/*
* Start issuing IO for the next log block early - this
* should help keep the L2ARC device busy while we
* decompress and restore this log block.
*/
*next_io = l2arc_log_blk_fetch(dev->l2ad_vdev, next_lbp,
next_lb);
}
/* Wait for the IO to read this log block to complete */
if ((err = zio_wait(this_io)) != 0) {
ARCSTAT_BUMP(arcstat_l2_rebuild_abort_io_errors);
zfs_dbgmsg("L2ARC IO error (%d) while reading log block, "
"offset: %llu, vdev guid: %llu", err,
(u_longlong_t)this_lbp->lbp_daddr,
(u_longlong_t)dev->l2ad_vdev->vdev_guid);
goto cleanup;
}
/*
* Make sure the buffer checks out.
* L2BLK_GET_PSIZE returns aligned size for log blocks.
*/
asize = L2BLK_GET_PSIZE((this_lbp)->lbp_prop);
fletcher_4_native(this_lb, asize, NULL, &cksum);
if (!ZIO_CHECKSUM_EQUAL(cksum, this_lbp->lbp_cksum)) {
ARCSTAT_BUMP(arcstat_l2_rebuild_abort_cksum_lb_errors);
zfs_dbgmsg("L2ARC log block cksum failed, offset: %llu, "
"vdev guid: %llu, l2ad_hand: %llu, l2ad_evict: %llu",
(u_longlong_t)this_lbp->lbp_daddr,
(u_longlong_t)dev->l2ad_vdev->vdev_guid,
(u_longlong_t)dev->l2ad_hand,
(u_longlong_t)dev->l2ad_evict);
err = SET_ERROR(ECKSUM);
goto cleanup;
}
/* Now we can take our time decoding this buffer */
switch (L2BLK_GET_COMPRESS((this_lbp)->lbp_prop)) {
case ZIO_COMPRESS_OFF:
break;
case ZIO_COMPRESS_LZ4:
abd = abd_alloc_for_io(asize, B_TRUE);
abd_copy_from_buf_off(abd, this_lb, 0, asize);
if ((err = zio_decompress_data(
L2BLK_GET_COMPRESS((this_lbp)->lbp_prop),
abd, this_lb, asize, sizeof (*this_lb), NULL)) != 0) {
err = SET_ERROR(EINVAL);
goto cleanup;
}
break;
default:
err = SET_ERROR(EINVAL);
goto cleanup;
}
if (this_lb->lb_magic == BSWAP_64(L2ARC_LOG_BLK_MAGIC))
byteswap_uint64_array(this_lb, sizeof (*this_lb));
if (this_lb->lb_magic != L2ARC_LOG_BLK_MAGIC) {
err = SET_ERROR(EINVAL);
goto cleanup;
}
cleanup:
/* Abort an in-flight fetch I/O in case of error */
if (err != 0 && *next_io != NULL) {
l2arc_log_blk_fetch_abort(*next_io);
*next_io = NULL;
}
if (abd != NULL)
abd_free(abd);
return (err);
}
/*
* Restores the payload of a log block to ARC. This creates empty ARC hdr
* entries which only contain an l2arc hdr, essentially restoring the
* buffers to their L2ARC evicted state. This function also updates space
* usage on the L2ARC vdev to make sure it tracks restored buffers.
*/
static void
l2arc_log_blk_restore(l2arc_dev_t *dev, const l2arc_log_blk_phys_t *lb,
uint64_t lb_asize)
{
uint64_t size = 0, asize = 0;
uint64_t log_entries = dev->l2ad_log_entries;
/*
* Usually arc_adapt() is called only for data, not headers, but
* since we may allocate significant amount of memory here, let ARC
* grow its arc_c.
*/
arc_adapt(log_entries * HDR_L2ONLY_SIZE, arc_l2c_only);
for (int i = log_entries - 1; i >= 0; i--) {
/*
* Restore goes in the reverse temporal direction to preserve
* correct temporal ordering of buffers in the l2ad_buflist.
* l2arc_hdr_restore also does a list_insert_tail instead of
* list_insert_head on the l2ad_buflist:
*
* LIST l2ad_buflist LIST
* HEAD <------ (time) ------ TAIL
* direction +-----+-----+-----+-----+-----+ direction
* of l2arc <== | buf | buf | buf | buf | buf | ===> of rebuild
* fill +-----+-----+-----+-----+-----+
* ^ ^
* | |
* | |
* l2arc_feed_thread l2arc_rebuild
* will place new bufs here restores bufs here
*
* During l2arc_rebuild() the device is not used by
* l2arc_feed_thread() as dev->l2ad_rebuild is set to true.
*/
size += L2BLK_GET_LSIZE((&lb->lb_entries[i])->le_prop);
asize += vdev_psize_to_asize(dev->l2ad_vdev,
L2BLK_GET_PSIZE((&lb->lb_entries[i])->le_prop));
l2arc_hdr_restore(&lb->lb_entries[i], dev);
}
/*
* Record rebuild stats:
* size Logical size of restored buffers in the L2ARC
* asize Aligned size of restored buffers in the L2ARC
*/
ARCSTAT_INCR(arcstat_l2_rebuild_size, size);
ARCSTAT_INCR(arcstat_l2_rebuild_asize, asize);
ARCSTAT_INCR(arcstat_l2_rebuild_bufs, log_entries);
ARCSTAT_F_AVG(arcstat_l2_log_blk_avg_asize, lb_asize);
ARCSTAT_F_AVG(arcstat_l2_data_to_meta_ratio, asize / lb_asize);
ARCSTAT_BUMP(arcstat_l2_rebuild_log_blks);
}
/*
* Restores a single ARC buf hdr from a log entry. The ARC buffer is put
* into a state indicating that it has been evicted to L2ARC.
*/
static void
l2arc_hdr_restore(const l2arc_log_ent_phys_t *le, l2arc_dev_t *dev)
{
arc_buf_hdr_t *hdr, *exists;
kmutex_t *hash_lock;
arc_buf_contents_t type = L2BLK_GET_TYPE((le)->le_prop);
uint64_t asize;
/*
* Do all the allocation before grabbing any locks, this lets us
* sleep if memory is full and we don't have to deal with failed
* allocations.
*/
hdr = arc_buf_alloc_l2only(L2BLK_GET_LSIZE((le)->le_prop), type,
dev, le->le_dva, le->le_daddr,
L2BLK_GET_PSIZE((le)->le_prop), le->le_birth,
L2BLK_GET_COMPRESS((le)->le_prop), le->le_complevel,
L2BLK_GET_PROTECTED((le)->le_prop),
L2BLK_GET_PREFETCH((le)->le_prop),
L2BLK_GET_STATE((le)->le_prop));
asize = vdev_psize_to_asize(dev->l2ad_vdev,
L2BLK_GET_PSIZE((le)->le_prop));
/*
* vdev_space_update() has to be called before arc_hdr_destroy() to
* avoid underflow since the latter also calls vdev_space_update().
*/
l2arc_hdr_arcstats_increment(hdr);
vdev_space_update(dev->l2ad_vdev, asize, 0, 0);
mutex_enter(&dev->l2ad_mtx);
list_insert_tail(&dev->l2ad_buflist, hdr);
(void) zfs_refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr);
mutex_exit(&dev->l2ad_mtx);
exists = buf_hash_insert(hdr, &hash_lock);
if (exists) {
/* Buffer was already cached, no need to restore it. */
arc_hdr_destroy(hdr);
/*
* If the buffer is already cached, check whether it has
* L2ARC metadata. If not, enter them and update the flag.
* This is important is case of onlining a cache device, since
* we previously evicted all L2ARC metadata from ARC.
*/
if (!HDR_HAS_L2HDR(exists)) {
arc_hdr_set_flags(exists, ARC_FLAG_HAS_L2HDR);
exists->b_l2hdr.b_dev = dev;
exists->b_l2hdr.b_daddr = le->le_daddr;
exists->b_l2hdr.b_arcs_state =
L2BLK_GET_STATE((le)->le_prop);
mutex_enter(&dev->l2ad_mtx);
list_insert_tail(&dev->l2ad_buflist, exists);
(void) zfs_refcount_add_many(&dev->l2ad_alloc,
arc_hdr_size(exists), exists);
mutex_exit(&dev->l2ad_mtx);
l2arc_hdr_arcstats_increment(exists);
vdev_space_update(dev->l2ad_vdev, asize, 0, 0);
}
ARCSTAT_BUMP(arcstat_l2_rebuild_bufs_precached);
}
mutex_exit(hash_lock);
}
/*
* Starts an asynchronous read IO to read a log block. This is used in log
* block reconstruction to start reading the next block before we are done
* decoding and reconstructing the current block, to keep the l2arc device
* nice and hot with read IO to process.
* The returned zio will contain a newly allocated memory buffers for the IO
* data which should then be freed by the caller once the zio is no longer
* needed (i.e. due to it having completed). If you wish to abort this
* zio, you should do so using l2arc_log_blk_fetch_abort, which takes
* care of disposing of the allocated buffers correctly.
*/
static zio_t *
l2arc_log_blk_fetch(vdev_t *vd, const l2arc_log_blkptr_t *lbp,
l2arc_log_blk_phys_t *lb)
{
uint32_t asize;
zio_t *pio;
l2arc_read_callback_t *cb;
/* L2BLK_GET_PSIZE returns aligned size for log blocks */
asize = L2BLK_GET_PSIZE((lbp)->lbp_prop);
ASSERT(asize <= sizeof (l2arc_log_blk_phys_t));
cb = kmem_zalloc(sizeof (l2arc_read_callback_t), KM_SLEEP);
cb->l2rcb_abd = abd_get_from_buf(lb, asize);
pio = zio_root(vd->vdev_spa, l2arc_blk_fetch_done, cb,
ZIO_FLAG_DONT_CACHE | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE |
ZIO_FLAG_DONT_RETRY);
(void) zio_nowait(zio_read_phys(pio, vd, lbp->lbp_daddr, asize,
cb->l2rcb_abd, ZIO_CHECKSUM_OFF, NULL, NULL,
ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_DONT_CACHE | ZIO_FLAG_CANFAIL |
ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY, B_FALSE));
return (pio);
}
/*
* Aborts a zio returned from l2arc_log_blk_fetch and frees the data
* buffers allocated for it.
*/
static void
l2arc_log_blk_fetch_abort(zio_t *zio)
{
(void) zio_wait(zio);
}
/*
* Creates a zio to update the device header on an l2arc device.
*/
void
l2arc_dev_hdr_update(l2arc_dev_t *dev)
{
l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
const uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize;
abd_t *abd;
int err;
VERIFY(spa_config_held(dev->l2ad_spa, SCL_STATE_ALL, RW_READER));
l2dhdr->dh_magic = L2ARC_DEV_HDR_MAGIC;
l2dhdr->dh_version = L2ARC_PERSISTENT_VERSION;
l2dhdr->dh_spa_guid = spa_guid(dev->l2ad_vdev->vdev_spa);
l2dhdr->dh_vdev_guid = dev->l2ad_vdev->vdev_guid;
l2dhdr->dh_log_entries = dev->l2ad_log_entries;
l2dhdr->dh_evict = dev->l2ad_evict;
l2dhdr->dh_start = dev->l2ad_start;
l2dhdr->dh_end = dev->l2ad_end;
l2dhdr->dh_lb_asize = zfs_refcount_count(&dev->l2ad_lb_asize);
l2dhdr->dh_lb_count = zfs_refcount_count(&dev->l2ad_lb_count);
l2dhdr->dh_flags = 0;
l2dhdr->dh_trim_action_time = dev->l2ad_vdev->vdev_trim_action_time;
l2dhdr->dh_trim_state = dev->l2ad_vdev->vdev_trim_state;
if (dev->l2ad_first)
l2dhdr->dh_flags |= L2ARC_DEV_HDR_EVICT_FIRST;
abd = abd_get_from_buf(l2dhdr, l2dhdr_asize);
err = zio_wait(zio_write_phys(NULL, dev->l2ad_vdev,
VDEV_LABEL_START_SIZE, l2dhdr_asize, abd, ZIO_CHECKSUM_LABEL, NULL,
NULL, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE));
abd_free(abd);
if (err != 0) {
zfs_dbgmsg("L2ARC IO error (%d) while writing device header, "
"vdev guid: %llu", err,
(u_longlong_t)dev->l2ad_vdev->vdev_guid);
}
}
/*
* Commits a log block to the L2ARC device. This routine is invoked from
* l2arc_write_buffers when the log block fills up.
* This function allocates some memory to temporarily hold the serialized
* buffer to be written. This is then released in l2arc_write_done.
*/
static void
l2arc_log_blk_commit(l2arc_dev_t *dev, zio_t *pio, l2arc_write_callback_t *cb)
{
l2arc_log_blk_phys_t *lb = &dev->l2ad_log_blk;
l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
uint64_t psize, asize;
zio_t *wzio;
l2arc_lb_abd_buf_t *abd_buf;
uint8_t *tmpbuf;
l2arc_lb_ptr_buf_t *lb_ptr_buf;
VERIFY3S(dev->l2ad_log_ent_idx, ==, dev->l2ad_log_entries);
tmpbuf = zio_buf_alloc(sizeof (*lb));
abd_buf = zio_buf_alloc(sizeof (*abd_buf));
abd_buf->abd = abd_get_from_buf(lb, sizeof (*lb));
lb_ptr_buf = kmem_zalloc(sizeof (l2arc_lb_ptr_buf_t), KM_SLEEP);
lb_ptr_buf->lb_ptr = kmem_zalloc(sizeof (l2arc_log_blkptr_t), KM_SLEEP);
/* link the buffer into the block chain */
lb->lb_prev_lbp = l2dhdr->dh_start_lbps[1];
lb->lb_magic = L2ARC_LOG_BLK_MAGIC;
/*
* l2arc_log_blk_commit() may be called multiple times during a single
* l2arc_write_buffers() call. Save the allocated abd buffers in a list
* so we can free them in l2arc_write_done() later on.
*/
list_insert_tail(&cb->l2wcb_abd_list, abd_buf);
/* try to compress the buffer */
psize = zio_compress_data(ZIO_COMPRESS_LZ4,
abd_buf->abd, tmpbuf, sizeof (*lb), 0);
/* a log block is never entirely zero */
ASSERT(psize != 0);
asize = vdev_psize_to_asize(dev->l2ad_vdev, psize);
ASSERT(asize <= sizeof (*lb));
/*
* Update the start log block pointer in the device header to point
* to the log block we're about to write.
*/
l2dhdr->dh_start_lbps[1] = l2dhdr->dh_start_lbps[0];
l2dhdr->dh_start_lbps[0].lbp_daddr = dev->l2ad_hand;
l2dhdr->dh_start_lbps[0].lbp_payload_asize =
dev->l2ad_log_blk_payload_asize;
l2dhdr->dh_start_lbps[0].lbp_payload_start =
dev->l2ad_log_blk_payload_start;
L2BLK_SET_LSIZE(
(&l2dhdr->dh_start_lbps[0])->lbp_prop, sizeof (*lb));
L2BLK_SET_PSIZE(
(&l2dhdr->dh_start_lbps[0])->lbp_prop, asize);
L2BLK_SET_CHECKSUM(
(&l2dhdr->dh_start_lbps[0])->lbp_prop,
ZIO_CHECKSUM_FLETCHER_4);
if (asize < sizeof (*lb)) {
/* compression succeeded */
bzero(tmpbuf + psize, asize - psize);
L2BLK_SET_COMPRESS(
(&l2dhdr->dh_start_lbps[0])->lbp_prop,
ZIO_COMPRESS_LZ4);
} else {
/* compression failed */
bcopy(lb, tmpbuf, sizeof (*lb));
L2BLK_SET_COMPRESS(
(&l2dhdr->dh_start_lbps[0])->lbp_prop,
ZIO_COMPRESS_OFF);
}
/* checksum what we're about to write */
fletcher_4_native(tmpbuf, asize, NULL,
&l2dhdr->dh_start_lbps[0].lbp_cksum);
abd_free(abd_buf->abd);
/* perform the write itself */
abd_buf->abd = abd_get_from_buf(tmpbuf, sizeof (*lb));
abd_take_ownership_of_buf(abd_buf->abd, B_TRUE);
wzio = zio_write_phys(pio, dev->l2ad_vdev, dev->l2ad_hand,
asize, abd_buf->abd, ZIO_CHECKSUM_OFF, NULL, NULL,
ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE);
DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev, zio_t *, wzio);
(void) zio_nowait(wzio);
dev->l2ad_hand += asize;
/*
* Include the committed log block's pointer in the list of pointers
* to log blocks present in the L2ARC device.
*/
bcopy(&l2dhdr->dh_start_lbps[0], lb_ptr_buf->lb_ptr,
sizeof (l2arc_log_blkptr_t));
mutex_enter(&dev->l2ad_mtx);
list_insert_head(&dev->l2ad_lbptr_list, lb_ptr_buf);
ARCSTAT_INCR(arcstat_l2_log_blk_asize, asize);
ARCSTAT_BUMP(arcstat_l2_log_blk_count);
zfs_refcount_add_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf);
zfs_refcount_add(&dev->l2ad_lb_count, lb_ptr_buf);
mutex_exit(&dev->l2ad_mtx);
vdev_space_update(dev->l2ad_vdev, asize, 0, 0);
/* bump the kstats */
ARCSTAT_INCR(arcstat_l2_write_bytes, asize);
ARCSTAT_BUMP(arcstat_l2_log_blk_writes);
ARCSTAT_F_AVG(arcstat_l2_log_blk_avg_asize, asize);
ARCSTAT_F_AVG(arcstat_l2_data_to_meta_ratio,
dev->l2ad_log_blk_payload_asize / asize);
/* start a new log block */
dev->l2ad_log_ent_idx = 0;
dev->l2ad_log_blk_payload_asize = 0;
dev->l2ad_log_blk_payload_start = 0;
}
/*
* Validates an L2ARC log block address to make sure that it can be read
* from the provided L2ARC device.
*/
boolean_t
l2arc_log_blkptr_valid(l2arc_dev_t *dev, const l2arc_log_blkptr_t *lbp)
{
/* L2BLK_GET_PSIZE returns aligned size for log blocks */
uint64_t asize = L2BLK_GET_PSIZE((lbp)->lbp_prop);
uint64_t end = lbp->lbp_daddr + asize - 1;
uint64_t start = lbp->lbp_payload_start;
boolean_t evicted = B_FALSE;
/*
* A log block is valid if all of the following conditions are true:
* - it fits entirely (including its payload) between l2ad_start and
* l2ad_end
* - it has a valid size
* - neither the log block itself nor part of its payload was evicted
* by l2arc_evict():
*
* l2ad_hand l2ad_evict
* | | lbp_daddr
* | start | | end
* | | | | |
* V V V V V
* l2ad_start ============================================ l2ad_end
* --------------------------||||
* ^ ^
* | log block
* payload
*/
evicted =
l2arc_range_check_overlap(start, end, dev->l2ad_hand) ||
l2arc_range_check_overlap(start, end, dev->l2ad_evict) ||
l2arc_range_check_overlap(dev->l2ad_hand, dev->l2ad_evict, start) ||
l2arc_range_check_overlap(dev->l2ad_hand, dev->l2ad_evict, end);
return (start >= dev->l2ad_start && end <= dev->l2ad_end &&
asize > 0 && asize <= sizeof (l2arc_log_blk_phys_t) &&
(!evicted || dev->l2ad_first));
}
/*
* Inserts ARC buffer header `hdr' into the current L2ARC log block on
* the device. The buffer being inserted must be present in L2ARC.
* Returns B_TRUE if the L2ARC log block is full and needs to be committed
* to L2ARC, or B_FALSE if it still has room for more ARC buffers.
*/
static boolean_t
l2arc_log_blk_insert(l2arc_dev_t *dev, const arc_buf_hdr_t *hdr)
{
l2arc_log_blk_phys_t *lb = &dev->l2ad_log_blk;
l2arc_log_ent_phys_t *le;
if (dev->l2ad_log_entries == 0)
return (B_FALSE);
int index = dev->l2ad_log_ent_idx++;
ASSERT3S(index, <, dev->l2ad_log_entries);
ASSERT(HDR_HAS_L2HDR(hdr));
le = &lb->lb_entries[index];
bzero(le, sizeof (*le));
le->le_dva = hdr->b_dva;
le->le_birth = hdr->b_birth;
le->le_daddr = hdr->b_l2hdr.b_daddr;
if (index == 0)
dev->l2ad_log_blk_payload_start = le->le_daddr;
L2BLK_SET_LSIZE((le)->le_prop, HDR_GET_LSIZE(hdr));
L2BLK_SET_PSIZE((le)->le_prop, HDR_GET_PSIZE(hdr));
L2BLK_SET_COMPRESS((le)->le_prop, HDR_GET_COMPRESS(hdr));
le->le_complevel = hdr->b_complevel;
L2BLK_SET_TYPE((le)->le_prop, hdr->b_type);
L2BLK_SET_PROTECTED((le)->le_prop, !!(HDR_PROTECTED(hdr)));
L2BLK_SET_PREFETCH((le)->le_prop, !!(HDR_PREFETCH(hdr)));
L2BLK_SET_STATE((le)->le_prop, hdr->b_l1hdr.b_state->arcs_state);
dev->l2ad_log_blk_payload_asize += vdev_psize_to_asize(dev->l2ad_vdev,
HDR_GET_PSIZE(hdr));
return (dev->l2ad_log_ent_idx == dev->l2ad_log_entries);
}
/*
* Checks whether a given L2ARC device address sits in a time-sequential
* range. The trick here is that the L2ARC is a rotary buffer, so we can't
* just do a range comparison, we need to handle the situation in which the
* range wraps around the end of the L2ARC device. Arguments:
* bottom -- Lower end of the range to check (written to earlier).
* top -- Upper end of the range to check (written to later).
* check -- The address for which we want to determine if it sits in
* between the top and bottom.
*
* The 3-way conditional below represents the following cases:
*
* bottom < top : Sequentially ordered case:
* <check>--------+-------------------+
* | (overlap here?) |
* L2ARC dev V V
* |---------------<bottom>============<top>--------------|
*
* bottom > top: Looped-around case:
* <check>--------+------------------+
* | (overlap here?) |
* L2ARC dev V V
* |===============<top>---------------<bottom>===========|
* ^ ^
* | (or here?) |
* +---------------+---------<check>
*
* top == bottom : Just a single address comparison.
*/
boolean_t
l2arc_range_check_overlap(uint64_t bottom, uint64_t top, uint64_t check)
{
if (bottom < top)
return (bottom <= check && check <= top);
else if (bottom > top)
return (check <= top || bottom <= check);
else
return (check == top);
}
EXPORT_SYMBOL(arc_buf_size);
EXPORT_SYMBOL(arc_write);
EXPORT_SYMBOL(arc_read);
EXPORT_SYMBOL(arc_buf_info);
EXPORT_SYMBOL(arc_getbuf_func);
EXPORT_SYMBOL(arc_add_prune_callback);
EXPORT_SYMBOL(arc_remove_prune_callback);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min, param_set_arc_min,
param_get_long, ZMOD_RW, "Min arc size");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, max, param_set_arc_max,
param_get_long, ZMOD_RW, "Max arc size");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, meta_limit, param_set_arc_long,
param_get_long, ZMOD_RW, "Metadata limit for arc size");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, meta_limit_percent,
param_set_arc_long, param_get_long, ZMOD_RW,
"Percent of arc size for arc meta limit");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, meta_min, param_set_arc_long,
param_get_long, ZMOD_RW, "Min arc metadata");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_prune, INT, ZMOD_RW,
"Meta objects to scan for prune");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_adjust_restarts, INT, ZMOD_RW,
"Limit number of restarts in arc_evict_meta");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_strategy, INT, ZMOD_RW,
"Meta reclaim strategy");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, grow_retry, param_set_arc_int,
param_get_int, ZMOD_RW, "Seconds before growing arc size");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, p_dampener_disable, INT, ZMOD_RW,
"Disable arc_p adapt dampener");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, shrink_shift, param_set_arc_int,
param_get_int, ZMOD_RW, "log2(fraction of arc to reclaim)");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, pc_percent, UINT, ZMOD_RW,
"Percent of pagecache to reclaim arc to");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, p_min_shift, param_set_arc_int,
param_get_int, ZMOD_RW, "arc_c shift to calc min/max arc_p");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, average_blocksize, INT, ZMOD_RD,
"Target average block size");
ZFS_MODULE_PARAM(zfs, zfs_, compressed_arc_enabled, INT, ZMOD_RW,
"Disable compressed arc buffers");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min_prefetch_ms, param_set_arc_int,
param_get_int, ZMOD_RW, "Min life of prefetch block in ms");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min_prescient_prefetch_ms,
param_set_arc_int, param_get_int, ZMOD_RW,
"Min life of prescient prefetched block in ms");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, write_max, ULONG, ZMOD_RW,
"Max write bytes per interval");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, write_boost, ULONG, ZMOD_RW,
"Extra write bytes during device warmup");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, headroom, ULONG, ZMOD_RW,
"Number of max device writes to precache");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, headroom_boost, ULONG, ZMOD_RW,
"Compressed l2arc_headroom multiplier");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, trim_ahead, ULONG, ZMOD_RW,
"TRIM ahead L2ARC write size multiplier");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_secs, ULONG, ZMOD_RW,
"Seconds between L2ARC writing");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_min_ms, ULONG, ZMOD_RW,
"Min feed interval in milliseconds");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, noprefetch, INT, ZMOD_RW,
"Skip caching prefetched buffers");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_again, INT, ZMOD_RW,
"Turbo L2ARC warmup");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, norw, INT, ZMOD_RW,
"No reads during writes");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, meta_percent, INT, ZMOD_RW,
"Percent of ARC size allowed for L2ARC-only headers");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, rebuild_enabled, INT, ZMOD_RW,
"Rebuild the L2ARC when importing a pool");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, rebuild_blocks_min_l2size, ULONG, ZMOD_RW,
"Min size in bytes to write rebuild log blocks in L2ARC");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, mfuonly, INT, ZMOD_RW,
"Cache only MFU data from ARC into L2ARC");
ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, exclude_special, INT, ZMOD_RW,
"If set to 1 exclude dbufs on special vdevs from being cached to "
"L2ARC.");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, lotsfree_percent, param_set_arc_int,
param_get_int, ZMOD_RW, "System free memory I/O throttle in bytes");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, sys_free, param_set_arc_long,
param_get_long, ZMOD_RW, "System free memory target size in bytes");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, dnode_limit, param_set_arc_long,
param_get_long, ZMOD_RW, "Minimum bytes of dnodes in arc");
ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, dnode_limit_percent,
param_set_arc_long, param_get_long, ZMOD_RW,
"Percent of ARC meta buffers for dnodes");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, dnode_reduce_percent, ULONG, ZMOD_RW,
"Percentage of excess dnodes to try to unpin");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, eviction_pct, INT, ZMOD_RW,
"When full, ARC allocation waits for eviction of this % of alloc size");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, evict_batch_limit, INT, ZMOD_RW,
"The number of headers to evict per sublist before moving to the next");
ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, prune_task_threads, INT, ZMOD_RW,
"Number of arc_prune threads");
/* END CSTYLED */
diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c
index 8443704c740d..96dcb2564754 100644
--- a/module/zfs/dbuf.c
+++ b/module/zfs/dbuf.c
@@ -1,5120 +1,5120 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
*/
#include <sys/zfs_context.h>
#include <sys/arc.h>
#include <sys/dmu.h>
#include <sys/dmu_send.h>
#include <sys/dmu_impl.h>
#include <sys/dbuf.h>
#include <sys/dmu_objset.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dmu_tx.h>
#include <sys/spa.h>
#include <sys/zio.h>
#include <sys/dmu_zfetch.h>
#include <sys/sa.h>
#include <sys/sa_impl.h>
#include <sys/zfeature.h>
#include <sys/blkptr.h>
#include <sys/range_tree.h>
#include <sys/trace_zfs.h>
#include <sys/callb.h>
#include <sys/abd.h>
#include <sys/vdev.h>
#include <cityhash.h>
#include <sys/spa_impl.h>
#include <sys/wmsum.h>
#include <sys/vdev_impl.h>
-kstat_t *dbuf_ksp;
+static kstat_t *dbuf_ksp;
typedef struct dbuf_stats {
/*
* Various statistics about the size of the dbuf cache.
*/
kstat_named_t cache_count;
kstat_named_t cache_size_bytes;
kstat_named_t cache_size_bytes_max;
/*
* Statistics regarding the bounds on the dbuf cache size.
*/
kstat_named_t cache_target_bytes;
kstat_named_t cache_lowater_bytes;
kstat_named_t cache_hiwater_bytes;
/*
* Total number of dbuf cache evictions that have occurred.
*/
kstat_named_t cache_total_evicts;
/*
* The distribution of dbuf levels in the dbuf cache and
* the total size of all dbufs at each level.
*/
kstat_named_t cache_levels[DN_MAX_LEVELS];
kstat_named_t cache_levels_bytes[DN_MAX_LEVELS];
/*
* Statistics about the dbuf hash table.
*/
kstat_named_t hash_hits;
kstat_named_t hash_misses;
kstat_named_t hash_collisions;
kstat_named_t hash_elements;
kstat_named_t hash_elements_max;
/*
* Number of sublists containing more than one dbuf in the dbuf
* hash table. Keep track of the longest hash chain.
*/
kstat_named_t hash_chains;
kstat_named_t hash_chain_max;
/*
* Number of times a dbuf_create() discovers that a dbuf was
* already created and in the dbuf hash table.
*/
kstat_named_t hash_insert_race;
/*
* Statistics about the size of the metadata dbuf cache.
*/
kstat_named_t metadata_cache_count;
kstat_named_t metadata_cache_size_bytes;
kstat_named_t metadata_cache_size_bytes_max;
/*
* For diagnostic purposes, this is incremented whenever we can't add
* something to the metadata cache because it's full, and instead put
* the data in the regular dbuf cache.
*/
kstat_named_t metadata_cache_overflow;
} dbuf_stats_t;
dbuf_stats_t dbuf_stats = {
{ "cache_count", KSTAT_DATA_UINT64 },
{ "cache_size_bytes", KSTAT_DATA_UINT64 },
{ "cache_size_bytes_max", KSTAT_DATA_UINT64 },
{ "cache_target_bytes", KSTAT_DATA_UINT64 },
{ "cache_lowater_bytes", KSTAT_DATA_UINT64 },
{ "cache_hiwater_bytes", KSTAT_DATA_UINT64 },
{ "cache_total_evicts", KSTAT_DATA_UINT64 },
{ { "cache_levels_N", KSTAT_DATA_UINT64 } },
{ { "cache_levels_bytes_N", KSTAT_DATA_UINT64 } },
{ "hash_hits", KSTAT_DATA_UINT64 },
{ "hash_misses", KSTAT_DATA_UINT64 },
{ "hash_collisions", KSTAT_DATA_UINT64 },
{ "hash_elements", KSTAT_DATA_UINT64 },
{ "hash_elements_max", KSTAT_DATA_UINT64 },
{ "hash_chains", KSTAT_DATA_UINT64 },
{ "hash_chain_max", KSTAT_DATA_UINT64 },
{ "hash_insert_race", KSTAT_DATA_UINT64 },
{ "metadata_cache_count", KSTAT_DATA_UINT64 },
{ "metadata_cache_size_bytes", KSTAT_DATA_UINT64 },
{ "metadata_cache_size_bytes_max", KSTAT_DATA_UINT64 },
{ "metadata_cache_overflow", KSTAT_DATA_UINT64 }
};
struct {
wmsum_t cache_count;
wmsum_t cache_total_evicts;
wmsum_t cache_levels[DN_MAX_LEVELS];
wmsum_t cache_levels_bytes[DN_MAX_LEVELS];
wmsum_t hash_hits;
wmsum_t hash_misses;
wmsum_t hash_collisions;
wmsum_t hash_chains;
wmsum_t hash_insert_race;
wmsum_t metadata_cache_count;
wmsum_t metadata_cache_overflow;
} dbuf_sums;
#define DBUF_STAT_INCR(stat, val) \
wmsum_add(&dbuf_sums.stat, val);
#define DBUF_STAT_DECR(stat, val) \
DBUF_STAT_INCR(stat, -(val));
#define DBUF_STAT_BUMP(stat) \
DBUF_STAT_INCR(stat, 1);
#define DBUF_STAT_BUMPDOWN(stat) \
DBUF_STAT_INCR(stat, -1);
#define DBUF_STAT_MAX(stat, v) { \
uint64_t _m; \
while ((v) > (_m = dbuf_stats.stat.value.ui64) && \
(_m != atomic_cas_64(&dbuf_stats.stat.value.ui64, _m, (v))))\
continue; \
}
static boolean_t dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx);
static void dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx);
static void dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr);
static int dbuf_read_verify_dnode_crypt(dmu_buf_impl_t *db, uint32_t flags);
extern inline void dmu_buf_init_user(dmu_buf_user_t *dbu,
dmu_buf_evict_func_t *evict_func_sync,
dmu_buf_evict_func_t *evict_func_async,
dmu_buf_t **clear_on_evict_dbufp);
/*
* Global data structures and functions for the dbuf cache.
*/
static kmem_cache_t *dbuf_kmem_cache;
static taskq_t *dbu_evict_taskq;
static kthread_t *dbuf_cache_evict_thread;
static kmutex_t dbuf_evict_lock;
static kcondvar_t dbuf_evict_cv;
static boolean_t dbuf_evict_thread_exit;
/*
* There are two dbuf caches; each dbuf can only be in one of them at a time.
*
* 1. Cache of metadata dbufs, to help make read-heavy administrative commands
* from /sbin/zfs run faster. The "metadata cache" specifically stores dbufs
* that represent the metadata that describes filesystems/snapshots/
* bookmarks/properties/etc. We only evict from this cache when we export a
* pool, to short-circuit as much I/O as possible for all administrative
* commands that need the metadata. There is no eviction policy for this
* cache, because we try to only include types in it which would occupy a
* very small amount of space per object but create a large impact on the
* performance of these commands. Instead, after it reaches a maximum size
* (which should only happen on very small memory systems with a very large
* number of filesystem objects), we stop taking new dbufs into the
* metadata cache, instead putting them in the normal dbuf cache.
*
* 2. LRU cache of dbufs. The dbuf cache maintains a list of dbufs that
* are not currently held but have been recently released. These dbufs
* are not eligible for arc eviction until they are aged out of the cache.
* Dbufs that are aged out of the cache will be immediately destroyed and
* become eligible for arc eviction.
*
* Dbufs are added to these caches once the last hold is released. If a dbuf is
* later accessed and still exists in the dbuf cache, then it will be removed
* from the cache and later re-added to the head of the cache.
*
* If a given dbuf meets the requirements for the metadata cache, it will go
* there, otherwise it will be considered for the generic LRU dbuf cache. The
* caches and the refcounts tracking their sizes are stored in an array indexed
* by those caches' matching enum values (from dbuf_cached_state_t).
*/
typedef struct dbuf_cache {
multilist_t cache;
zfs_refcount_t size ____cacheline_aligned;
} dbuf_cache_t;
dbuf_cache_t dbuf_caches[DB_CACHE_MAX];
/* Size limits for the caches */
-unsigned long dbuf_cache_max_bytes = ULONG_MAX;
-unsigned long dbuf_metadata_cache_max_bytes = ULONG_MAX;
+static unsigned long dbuf_cache_max_bytes = ULONG_MAX;
+static unsigned long dbuf_metadata_cache_max_bytes = ULONG_MAX;
/* Set the default sizes of the caches to log2 fraction of arc size */
-int dbuf_cache_shift = 5;
-int dbuf_metadata_cache_shift = 6;
+static int dbuf_cache_shift = 5;
+static int dbuf_metadata_cache_shift = 6;
static unsigned long dbuf_cache_target_bytes(void);
static unsigned long dbuf_metadata_cache_target_bytes(void);
/*
* The LRU dbuf cache uses a three-stage eviction policy:
* - A low water marker designates when the dbuf eviction thread
* should stop evicting from the dbuf cache.
* - When we reach the maximum size (aka mid water mark), we
* signal the eviction thread to run.
* - The high water mark indicates when the eviction thread
* is unable to keep up with the incoming load and eviction must
* happen in the context of the calling thread.
*
* The dbuf cache:
* (max size)
* low water mid water hi water
* +----------------------------------------+----------+----------+
* | | | |
* | | | |
* | | | |
* | | | |
* +----------------------------------------+----------+----------+
* stop signal evict
* evicting eviction directly
* thread
*
* The high and low water marks indicate the operating range for the eviction
* thread. The low water mark is, by default, 90% of the total size of the
* cache and the high water mark is at 110% (both of these percentages can be
* changed by setting dbuf_cache_lowater_pct and dbuf_cache_hiwater_pct,
* respectively). The eviction thread will try to ensure that the cache remains
* within this range by waking up every second and checking if the cache is
* above the low water mark. The thread can also be woken up by callers adding
* elements into the cache if the cache is larger than the mid water (i.e max
* cache size). Once the eviction thread is woken up and eviction is required,
* it will continue evicting buffers until it's able to reduce the cache size
* to the low water mark. If the cache size continues to grow and hits the high
* water mark, then callers adding elements to the cache will begin to evict
* directly from the cache until the cache is no longer above the high water
* mark.
*/
/*
* The percentage above and below the maximum cache size.
*/
-uint_t dbuf_cache_hiwater_pct = 10;
-uint_t dbuf_cache_lowater_pct = 10;
+static uint_t dbuf_cache_hiwater_pct = 10;
+static uint_t dbuf_cache_lowater_pct = 10;
static int
dbuf_cons(void *vdb, void *unused, int kmflag)
{
(void) unused, (void) kmflag;
dmu_buf_impl_t *db = vdb;
bzero(db, sizeof (dmu_buf_impl_t));
mutex_init(&db->db_mtx, NULL, MUTEX_DEFAULT, NULL);
rw_init(&db->db_rwlock, NULL, RW_DEFAULT, NULL);
cv_init(&db->db_changed, NULL, CV_DEFAULT, NULL);
multilist_link_init(&db->db_cache_link);
zfs_refcount_create(&db->db_holds);
return (0);
}
static void
dbuf_dest(void *vdb, void *unused)
{
(void) unused;
dmu_buf_impl_t *db = vdb;
mutex_destroy(&db->db_mtx);
rw_destroy(&db->db_rwlock);
cv_destroy(&db->db_changed);
ASSERT(!multilist_link_active(&db->db_cache_link));
zfs_refcount_destroy(&db->db_holds);
}
/*
* dbuf hash table routines
*/
static dbuf_hash_table_t dbuf_hash_table;
/*
* We use Cityhash for this. It's fast, and has good hash properties without
* requiring any large static buffers.
*/
static uint64_t
dbuf_hash(void *os, uint64_t obj, uint8_t lvl, uint64_t blkid)
{
return (cityhash4((uintptr_t)os, obj, (uint64_t)lvl, blkid));
}
#define DTRACE_SET_STATE(db, why) \
DTRACE_PROBE2(dbuf__state_change, dmu_buf_impl_t *, db, \
const char *, why)
#define DBUF_EQUAL(dbuf, os, obj, level, blkid) \
((dbuf)->db.db_object == (obj) && \
(dbuf)->db_objset == (os) && \
(dbuf)->db_level == (level) && \
(dbuf)->db_blkid == (blkid))
dmu_buf_impl_t *
dbuf_find(objset_t *os, uint64_t obj, uint8_t level, uint64_t blkid)
{
dbuf_hash_table_t *h = &dbuf_hash_table;
uint64_t hv;
uint64_t idx;
dmu_buf_impl_t *db;
hv = dbuf_hash(os, obj, level, blkid);
idx = hv & h->hash_table_mask;
mutex_enter(DBUF_HASH_MUTEX(h, idx));
for (db = h->hash_table[idx]; db != NULL; db = db->db_hash_next) {
if (DBUF_EQUAL(db, os, obj, level, blkid)) {
mutex_enter(&db->db_mtx);
if (db->db_state != DB_EVICTING) {
mutex_exit(DBUF_HASH_MUTEX(h, idx));
return (db);
}
mutex_exit(&db->db_mtx);
}
}
mutex_exit(DBUF_HASH_MUTEX(h, idx));
return (NULL);
}
static dmu_buf_impl_t *
dbuf_find_bonus(objset_t *os, uint64_t object)
{
dnode_t *dn;
dmu_buf_impl_t *db = NULL;
if (dnode_hold(os, object, FTAG, &dn) == 0) {
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (dn->dn_bonus != NULL) {
db = dn->dn_bonus;
mutex_enter(&db->db_mtx);
}
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
}
return (db);
}
/*
* Insert an entry into the hash table. If there is already an element
* equal to elem in the hash table, then the already existing element
* will be returned and the new element will not be inserted.
* Otherwise returns NULL.
*/
static dmu_buf_impl_t *
dbuf_hash_insert(dmu_buf_impl_t *db)
{
dbuf_hash_table_t *h = &dbuf_hash_table;
objset_t *os = db->db_objset;
uint64_t obj = db->db.db_object;
int level = db->db_level;
uint64_t blkid, hv, idx;
dmu_buf_impl_t *dbf;
uint32_t i;
blkid = db->db_blkid;
hv = dbuf_hash(os, obj, level, blkid);
idx = hv & h->hash_table_mask;
mutex_enter(DBUF_HASH_MUTEX(h, idx));
for (dbf = h->hash_table[idx], i = 0; dbf != NULL;
dbf = dbf->db_hash_next, i++) {
if (DBUF_EQUAL(dbf, os, obj, level, blkid)) {
mutex_enter(&dbf->db_mtx);
if (dbf->db_state != DB_EVICTING) {
mutex_exit(DBUF_HASH_MUTEX(h, idx));
return (dbf);
}
mutex_exit(&dbf->db_mtx);
}
}
if (i > 0) {
DBUF_STAT_BUMP(hash_collisions);
if (i == 1)
DBUF_STAT_BUMP(hash_chains);
DBUF_STAT_MAX(hash_chain_max, i);
}
mutex_enter(&db->db_mtx);
db->db_hash_next = h->hash_table[idx];
h->hash_table[idx] = db;
mutex_exit(DBUF_HASH_MUTEX(h, idx));
uint64_t he = atomic_inc_64_nv(&dbuf_stats.hash_elements.value.ui64);
DBUF_STAT_MAX(hash_elements_max, he);
return (NULL);
}
/*
* This returns whether this dbuf should be stored in the metadata cache, which
* is based on whether it's from one of the dnode types that store data related
* to traversing dataset hierarchies.
*/
static boolean_t
dbuf_include_in_metadata_cache(dmu_buf_impl_t *db)
{
DB_DNODE_ENTER(db);
dmu_object_type_t type = DB_DNODE(db)->dn_type;
DB_DNODE_EXIT(db);
/* Check if this dbuf is one of the types we care about */
if (DMU_OT_IS_METADATA_CACHED(type)) {
/* If we hit this, then we set something up wrong in dmu_ot */
ASSERT(DMU_OT_IS_METADATA(type));
/*
* Sanity check for small-memory systems: don't allocate too
* much memory for this purpose.
*/
if (zfs_refcount_count(
&dbuf_caches[DB_DBUF_METADATA_CACHE].size) >
dbuf_metadata_cache_target_bytes()) {
DBUF_STAT_BUMP(metadata_cache_overflow);
return (B_FALSE);
}
return (B_TRUE);
}
return (B_FALSE);
}
/*
* Remove an entry from the hash table. It must be in the EVICTING state.
*/
static void
dbuf_hash_remove(dmu_buf_impl_t *db)
{
dbuf_hash_table_t *h = &dbuf_hash_table;
uint64_t hv, idx;
dmu_buf_impl_t *dbf, **dbp;
hv = dbuf_hash(db->db_objset, db->db.db_object,
db->db_level, db->db_blkid);
idx = hv & h->hash_table_mask;
/*
* We mustn't hold db_mtx to maintain lock ordering:
* DBUF_HASH_MUTEX > db_mtx.
*/
ASSERT(zfs_refcount_is_zero(&db->db_holds));
ASSERT(db->db_state == DB_EVICTING);
ASSERT(!MUTEX_HELD(&db->db_mtx));
mutex_enter(DBUF_HASH_MUTEX(h, idx));
dbp = &h->hash_table[idx];
while ((dbf = *dbp) != db) {
dbp = &dbf->db_hash_next;
ASSERT(dbf != NULL);
}
*dbp = db->db_hash_next;
db->db_hash_next = NULL;
if (h->hash_table[idx] &&
h->hash_table[idx]->db_hash_next == NULL)
DBUF_STAT_BUMPDOWN(hash_chains);
mutex_exit(DBUF_HASH_MUTEX(h, idx));
atomic_dec_64(&dbuf_stats.hash_elements.value.ui64);
}
typedef enum {
DBVU_EVICTING,
DBVU_NOT_EVICTING
} dbvu_verify_type_t;
static void
dbuf_verify_user(dmu_buf_impl_t *db, dbvu_verify_type_t verify_type)
{
#ifdef ZFS_DEBUG
int64_t holds;
if (db->db_user == NULL)
return;
/* Only data blocks support the attachment of user data. */
ASSERT(db->db_level == 0);
/* Clients must resolve a dbuf before attaching user data. */
ASSERT(db->db.db_data != NULL);
ASSERT3U(db->db_state, ==, DB_CACHED);
holds = zfs_refcount_count(&db->db_holds);
if (verify_type == DBVU_EVICTING) {
/*
* Immediate eviction occurs when holds == dirtycnt.
* For normal eviction buffers, holds is zero on
* eviction, except when dbuf_fix_old_data() calls
* dbuf_clear_data(). However, the hold count can grow
* during eviction even though db_mtx is held (see
* dmu_bonus_hold() for an example), so we can only
* test the generic invariant that holds >= dirtycnt.
*/
ASSERT3U(holds, >=, db->db_dirtycnt);
} else {
if (db->db_user_immediate_evict == TRUE)
ASSERT3U(holds, >=, db->db_dirtycnt);
else
ASSERT3U(holds, >, 0);
}
#endif
}
static void
dbuf_evict_user(dmu_buf_impl_t *db)
{
dmu_buf_user_t *dbu = db->db_user;
ASSERT(MUTEX_HELD(&db->db_mtx));
if (dbu == NULL)
return;
dbuf_verify_user(db, DBVU_EVICTING);
db->db_user = NULL;
#ifdef ZFS_DEBUG
if (dbu->dbu_clear_on_evict_dbufp != NULL)
*dbu->dbu_clear_on_evict_dbufp = NULL;
#endif
/*
* There are two eviction callbacks - one that we call synchronously
* and one that we invoke via a taskq. The async one is useful for
* avoiding lock order reversals and limiting stack depth.
*
* Note that if we have a sync callback but no async callback,
* it's likely that the sync callback will free the structure
* containing the dbu. In that case we need to take care to not
* dereference dbu after calling the sync evict func.
*/
boolean_t has_async = (dbu->dbu_evict_func_async != NULL);
if (dbu->dbu_evict_func_sync != NULL)
dbu->dbu_evict_func_sync(dbu);
if (has_async) {
taskq_dispatch_ent(dbu_evict_taskq, dbu->dbu_evict_func_async,
dbu, 0, &dbu->dbu_tqent);
}
}
boolean_t
dbuf_is_metadata(dmu_buf_impl_t *db)
{
/*
* Consider indirect blocks and spill blocks to be meta data.
*/
if (db->db_level > 0 || db->db_blkid == DMU_SPILL_BLKID) {
return (B_TRUE);
} else {
boolean_t is_metadata;
DB_DNODE_ENTER(db);
is_metadata = DMU_OT_IS_METADATA(DB_DNODE(db)->dn_type);
DB_DNODE_EXIT(db);
return (is_metadata);
}
}
/*
* We want to exclude buffers that are on a special allocation class from
* L2ARC.
*/
boolean_t
dbuf_is_l2cacheable(dmu_buf_impl_t *db)
{
vdev_t *vd = NULL;
zfs_cache_type_t cache = db->db_objset->os_secondary_cache;
blkptr_t *bp = db->db_blkptr;
if (bp != NULL && !BP_IS_HOLE(bp)) {
uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
vdev_t *rvd = db->db_objset->os_spa->spa_root_vdev;
if (vdev < rvd->vdev_children)
vd = rvd->vdev_child[vdev];
if (cache == ZFS_CACHE_ALL ||
(dbuf_is_metadata(db) && cache == ZFS_CACHE_METADATA)) {
if (vd == NULL)
return (B_TRUE);
if ((vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) ||
l2arc_exclude_special == 0)
return (B_TRUE);
}
}
return (B_FALSE);
}
static inline boolean_t
dnode_level_is_l2cacheable(blkptr_t *bp, dnode_t *dn, int64_t level)
{
vdev_t *vd = NULL;
zfs_cache_type_t cache = dn->dn_objset->os_secondary_cache;
if (bp != NULL && !BP_IS_HOLE(bp)) {
uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
vdev_t *rvd = dn->dn_objset->os_spa->spa_root_vdev;
if (vdev < rvd->vdev_children)
vd = rvd->vdev_child[vdev];
if (cache == ZFS_CACHE_ALL || ((level > 0 ||
DMU_OT_IS_METADATA(dn->dn_handle->dnh_dnode->dn_type)) &&
cache == ZFS_CACHE_METADATA)) {
if (vd == NULL)
return (B_TRUE);
if ((vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) ||
l2arc_exclude_special == 0)
return (B_TRUE);
}
}
return (B_FALSE);
}
/*
* This function *must* return indices evenly distributed between all
* sublists of the multilist. This is needed due to how the dbuf eviction
* code is laid out; dbuf_evict_thread() assumes dbufs are evenly
* distributed between all sublists and uses this assumption when
* deciding which sublist to evict from and how much to evict from it.
*/
static unsigned int
dbuf_cache_multilist_index_func(multilist_t *ml, void *obj)
{
dmu_buf_impl_t *db = obj;
/*
* The assumption here, is the hash value for a given
* dmu_buf_impl_t will remain constant throughout it's lifetime
* (i.e. it's objset, object, level and blkid fields don't change).
* Thus, we don't need to store the dbuf's sublist index
* on insertion, as this index can be recalculated on removal.
*
* Also, the low order bits of the hash value are thought to be
* distributed evenly. Otherwise, in the case that the multilist
* has a power of two number of sublists, each sublists' usage
* would not be evenly distributed. In this context full 64bit
* division would be a waste of time, so limit it to 32 bits.
*/
return ((unsigned int)dbuf_hash(db->db_objset, db->db.db_object,
db->db_level, db->db_blkid) %
multilist_get_num_sublists(ml));
}
/*
* The target size of the dbuf cache can grow with the ARC target,
* unless limited by the tunable dbuf_cache_max_bytes.
*/
static inline unsigned long
dbuf_cache_target_bytes(void)
{
return (MIN(dbuf_cache_max_bytes,
arc_target_bytes() >> dbuf_cache_shift));
}
/*
* The target size of the dbuf metadata cache can grow with the ARC target,
* unless limited by the tunable dbuf_metadata_cache_max_bytes.
*/
static inline unsigned long
dbuf_metadata_cache_target_bytes(void)
{
return (MIN(dbuf_metadata_cache_max_bytes,
arc_target_bytes() >> dbuf_metadata_cache_shift));
}
static inline uint64_t
dbuf_cache_hiwater_bytes(void)
{
uint64_t dbuf_cache_target = dbuf_cache_target_bytes();
return (dbuf_cache_target +
(dbuf_cache_target * dbuf_cache_hiwater_pct) / 100);
}
static inline uint64_t
dbuf_cache_lowater_bytes(void)
{
uint64_t dbuf_cache_target = dbuf_cache_target_bytes();
return (dbuf_cache_target -
(dbuf_cache_target * dbuf_cache_lowater_pct) / 100);
}
static inline boolean_t
dbuf_cache_above_lowater(void)
{
return (zfs_refcount_count(&dbuf_caches[DB_DBUF_CACHE].size) >
dbuf_cache_lowater_bytes());
}
/*
* Evict the oldest eligible dbuf from the dbuf cache.
*/
static void
dbuf_evict_one(void)
{
int idx = multilist_get_random_index(&dbuf_caches[DB_DBUF_CACHE].cache);
multilist_sublist_t *mls = multilist_sublist_lock(
&dbuf_caches[DB_DBUF_CACHE].cache, idx);
ASSERT(!MUTEX_HELD(&dbuf_evict_lock));
dmu_buf_impl_t *db = multilist_sublist_tail(mls);
while (db != NULL && mutex_tryenter(&db->db_mtx) == 0) {
db = multilist_sublist_prev(mls, db);
}
DTRACE_PROBE2(dbuf__evict__one, dmu_buf_impl_t *, db,
multilist_sublist_t *, mls);
if (db != NULL) {
multilist_sublist_remove(mls, db);
multilist_sublist_unlock(mls);
(void) zfs_refcount_remove_many(
&dbuf_caches[DB_DBUF_CACHE].size, db->db.db_size, db);
DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]);
DBUF_STAT_BUMPDOWN(cache_count);
DBUF_STAT_DECR(cache_levels_bytes[db->db_level],
db->db.db_size);
ASSERT3U(db->db_caching_status, ==, DB_DBUF_CACHE);
db->db_caching_status = DB_NO_CACHE;
dbuf_destroy(db);
DBUF_STAT_BUMP(cache_total_evicts);
} else {
multilist_sublist_unlock(mls);
}
}
/*
* The dbuf evict thread is responsible for aging out dbufs from the
* cache. Once the cache has reached it's maximum size, dbufs are removed
* and destroyed. The eviction thread will continue running until the size
* of the dbuf cache is at or below the maximum size. Once the dbuf is aged
* out of the cache it is destroyed and becomes eligible for arc eviction.
*/
static void
dbuf_evict_thread(void *unused)
{
(void) unused;
callb_cpr_t cpr;
CALLB_CPR_INIT(&cpr, &dbuf_evict_lock, callb_generic_cpr, FTAG);
mutex_enter(&dbuf_evict_lock);
while (!dbuf_evict_thread_exit) {
while (!dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) {
CALLB_CPR_SAFE_BEGIN(&cpr);
(void) cv_timedwait_idle_hires(&dbuf_evict_cv,
&dbuf_evict_lock, SEC2NSEC(1), MSEC2NSEC(1), 0);
CALLB_CPR_SAFE_END(&cpr, &dbuf_evict_lock);
}
mutex_exit(&dbuf_evict_lock);
/*
* Keep evicting as long as we're above the low water mark
* for the cache. We do this without holding the locks to
* minimize lock contention.
*/
while (dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) {
dbuf_evict_one();
}
mutex_enter(&dbuf_evict_lock);
}
dbuf_evict_thread_exit = B_FALSE;
cv_broadcast(&dbuf_evict_cv);
CALLB_CPR_EXIT(&cpr); /* drops dbuf_evict_lock */
thread_exit();
}
/*
* Wake up the dbuf eviction thread if the dbuf cache is at its max size.
* If the dbuf cache is at its high water mark, then evict a dbuf from the
* dbuf cache using the callers context.
*/
static void
dbuf_evict_notify(uint64_t size)
{
/*
* We check if we should evict without holding the dbuf_evict_lock,
* because it's OK to occasionally make the wrong decision here,
* and grabbing the lock results in massive lock contention.
*/
if (size > dbuf_cache_target_bytes()) {
if (size > dbuf_cache_hiwater_bytes())
dbuf_evict_one();
cv_signal(&dbuf_evict_cv);
}
}
static int
dbuf_kstat_update(kstat_t *ksp, int rw)
{
dbuf_stats_t *ds = ksp->ks_data;
if (rw == KSTAT_WRITE)
return (SET_ERROR(EACCES));
ds->cache_count.value.ui64 =
wmsum_value(&dbuf_sums.cache_count);
ds->cache_size_bytes.value.ui64 =
zfs_refcount_count(&dbuf_caches[DB_DBUF_CACHE].size);
ds->cache_target_bytes.value.ui64 = dbuf_cache_target_bytes();
ds->cache_hiwater_bytes.value.ui64 = dbuf_cache_hiwater_bytes();
ds->cache_lowater_bytes.value.ui64 = dbuf_cache_lowater_bytes();
ds->cache_total_evicts.value.ui64 =
wmsum_value(&dbuf_sums.cache_total_evicts);
for (int i = 0; i < DN_MAX_LEVELS; i++) {
ds->cache_levels[i].value.ui64 =
wmsum_value(&dbuf_sums.cache_levels[i]);
ds->cache_levels_bytes[i].value.ui64 =
wmsum_value(&dbuf_sums.cache_levels_bytes[i]);
}
ds->hash_hits.value.ui64 =
wmsum_value(&dbuf_sums.hash_hits);
ds->hash_misses.value.ui64 =
wmsum_value(&dbuf_sums.hash_misses);
ds->hash_collisions.value.ui64 =
wmsum_value(&dbuf_sums.hash_collisions);
ds->hash_chains.value.ui64 =
wmsum_value(&dbuf_sums.hash_chains);
ds->hash_insert_race.value.ui64 =
wmsum_value(&dbuf_sums.hash_insert_race);
ds->metadata_cache_count.value.ui64 =
wmsum_value(&dbuf_sums.metadata_cache_count);
ds->metadata_cache_size_bytes.value.ui64 = zfs_refcount_count(
&dbuf_caches[DB_DBUF_METADATA_CACHE].size);
ds->metadata_cache_overflow.value.ui64 =
wmsum_value(&dbuf_sums.metadata_cache_overflow);
return (0);
}
void
dbuf_init(void)
{
uint64_t hsize = 1ULL << 16;
dbuf_hash_table_t *h = &dbuf_hash_table;
int i;
/*
* The hash table is big enough to fill one eighth of physical memory
* with an average block size of zfs_arc_average_blocksize (default 8K).
* By default, the table will take up
* totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers).
*/
while (hsize * zfs_arc_average_blocksize < arc_all_memory() / 8)
hsize <<= 1;
retry:
h->hash_table_mask = hsize - 1;
#if defined(_KERNEL)
/*
* Large allocations which do not require contiguous pages
* should be using vmem_alloc() in the linux kernel
*/
h->hash_table = vmem_zalloc(hsize * sizeof (void *), KM_SLEEP);
#else
h->hash_table = kmem_zalloc(hsize * sizeof (void *), KM_NOSLEEP);
#endif
if (h->hash_table == NULL) {
/* XXX - we should really return an error instead of assert */
ASSERT(hsize > (1ULL << 10));
hsize >>= 1;
goto retry;
}
dbuf_kmem_cache = kmem_cache_create("dmu_buf_impl_t",
sizeof (dmu_buf_impl_t),
0, dbuf_cons, dbuf_dest, NULL, NULL, NULL, 0);
for (i = 0; i < DBUF_MUTEXES; i++)
mutex_init(&h->hash_mutexes[i], NULL, MUTEX_DEFAULT, NULL);
dbuf_stats_init(h);
/*
* All entries are queued via taskq_dispatch_ent(), so min/maxalloc
* configuration is not required.
*/
dbu_evict_taskq = taskq_create("dbu_evict", 1, defclsyspri, 0, 0, 0);
for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) {
multilist_create(&dbuf_caches[dcs].cache,
sizeof (dmu_buf_impl_t),
offsetof(dmu_buf_impl_t, db_cache_link),
dbuf_cache_multilist_index_func);
zfs_refcount_create(&dbuf_caches[dcs].size);
}
dbuf_evict_thread_exit = B_FALSE;
mutex_init(&dbuf_evict_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&dbuf_evict_cv, NULL, CV_DEFAULT, NULL);
dbuf_cache_evict_thread = thread_create(NULL, 0, dbuf_evict_thread,
NULL, 0, &p0, TS_RUN, minclsyspri);
wmsum_init(&dbuf_sums.cache_count, 0);
wmsum_init(&dbuf_sums.cache_total_evicts, 0);
for (i = 0; i < DN_MAX_LEVELS; i++) {
wmsum_init(&dbuf_sums.cache_levels[i], 0);
wmsum_init(&dbuf_sums.cache_levels_bytes[i], 0);
}
wmsum_init(&dbuf_sums.hash_hits, 0);
wmsum_init(&dbuf_sums.hash_misses, 0);
wmsum_init(&dbuf_sums.hash_collisions, 0);
wmsum_init(&dbuf_sums.hash_chains, 0);
wmsum_init(&dbuf_sums.hash_insert_race, 0);
wmsum_init(&dbuf_sums.metadata_cache_count, 0);
wmsum_init(&dbuf_sums.metadata_cache_overflow, 0);
dbuf_ksp = kstat_create("zfs", 0, "dbufstats", "misc",
KSTAT_TYPE_NAMED, sizeof (dbuf_stats) / sizeof (kstat_named_t),
KSTAT_FLAG_VIRTUAL);
if (dbuf_ksp != NULL) {
for (i = 0; i < DN_MAX_LEVELS; i++) {
snprintf(dbuf_stats.cache_levels[i].name,
KSTAT_STRLEN, "cache_level_%d", i);
dbuf_stats.cache_levels[i].data_type =
KSTAT_DATA_UINT64;
snprintf(dbuf_stats.cache_levels_bytes[i].name,
KSTAT_STRLEN, "cache_level_%d_bytes", i);
dbuf_stats.cache_levels_bytes[i].data_type =
KSTAT_DATA_UINT64;
}
dbuf_ksp->ks_data = &dbuf_stats;
dbuf_ksp->ks_update = dbuf_kstat_update;
kstat_install(dbuf_ksp);
}
}
void
dbuf_fini(void)
{
dbuf_hash_table_t *h = &dbuf_hash_table;
int i;
dbuf_stats_destroy();
for (i = 0; i < DBUF_MUTEXES; i++)
mutex_destroy(&h->hash_mutexes[i]);
#if defined(_KERNEL)
/*
* Large allocations which do not require contiguous pages
* should be using vmem_free() in the linux kernel
*/
vmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *));
#else
kmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *));
#endif
kmem_cache_destroy(dbuf_kmem_cache);
taskq_destroy(dbu_evict_taskq);
mutex_enter(&dbuf_evict_lock);
dbuf_evict_thread_exit = B_TRUE;
while (dbuf_evict_thread_exit) {
cv_signal(&dbuf_evict_cv);
cv_wait(&dbuf_evict_cv, &dbuf_evict_lock);
}
mutex_exit(&dbuf_evict_lock);
mutex_destroy(&dbuf_evict_lock);
cv_destroy(&dbuf_evict_cv);
for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) {
zfs_refcount_destroy(&dbuf_caches[dcs].size);
multilist_destroy(&dbuf_caches[dcs].cache);
}
if (dbuf_ksp != NULL) {
kstat_delete(dbuf_ksp);
dbuf_ksp = NULL;
}
wmsum_fini(&dbuf_sums.cache_count);
wmsum_fini(&dbuf_sums.cache_total_evicts);
for (i = 0; i < DN_MAX_LEVELS; i++) {
wmsum_fini(&dbuf_sums.cache_levels[i]);
wmsum_fini(&dbuf_sums.cache_levels_bytes[i]);
}
wmsum_fini(&dbuf_sums.hash_hits);
wmsum_fini(&dbuf_sums.hash_misses);
wmsum_fini(&dbuf_sums.hash_collisions);
wmsum_fini(&dbuf_sums.hash_chains);
wmsum_fini(&dbuf_sums.hash_insert_race);
wmsum_fini(&dbuf_sums.metadata_cache_count);
wmsum_fini(&dbuf_sums.metadata_cache_overflow);
}
/*
* Other stuff.
*/
#ifdef ZFS_DEBUG
static void
dbuf_verify(dmu_buf_impl_t *db)
{
dnode_t *dn;
dbuf_dirty_record_t *dr;
uint32_t txg_prev;
ASSERT(MUTEX_HELD(&db->db_mtx));
if (!(zfs_flags & ZFS_DEBUG_DBUF_VERIFY))
return;
ASSERT(db->db_objset != NULL);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
if (dn == NULL) {
ASSERT(db->db_parent == NULL);
ASSERT(db->db_blkptr == NULL);
} else {
ASSERT3U(db->db.db_object, ==, dn->dn_object);
ASSERT3P(db->db_objset, ==, dn->dn_objset);
ASSERT3U(db->db_level, <, dn->dn_nlevels);
ASSERT(db->db_blkid == DMU_BONUS_BLKID ||
db->db_blkid == DMU_SPILL_BLKID ||
!avl_is_empty(&dn->dn_dbufs));
}
if (db->db_blkid == DMU_BONUS_BLKID) {
ASSERT(dn != NULL);
ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen);
ASSERT3U(db->db.db_offset, ==, DMU_BONUS_BLKID);
} else if (db->db_blkid == DMU_SPILL_BLKID) {
ASSERT(dn != NULL);
ASSERT0(db->db.db_offset);
} else {
ASSERT3U(db->db.db_offset, ==, db->db_blkid * db->db.db_size);
}
if ((dr = list_head(&db->db_dirty_records)) != NULL) {
ASSERT(dr->dr_dbuf == db);
txg_prev = dr->dr_txg;
for (dr = list_next(&db->db_dirty_records, dr); dr != NULL;
dr = list_next(&db->db_dirty_records, dr)) {
ASSERT(dr->dr_dbuf == db);
ASSERT(txg_prev > dr->dr_txg);
txg_prev = dr->dr_txg;
}
}
/*
* We can't assert that db_size matches dn_datablksz because it
* can be momentarily different when another thread is doing
* dnode_set_blksz().
*/
if (db->db_level == 0 && db->db.db_object == DMU_META_DNODE_OBJECT) {
dr = db->db_data_pending;
/*
* It should only be modified in syncing context, so
* make sure we only have one copy of the data.
*/
ASSERT(dr == NULL || dr->dt.dl.dr_data == db->db_buf);
}
/* verify db->db_blkptr */
if (db->db_blkptr) {
if (db->db_parent == dn->dn_dbuf) {
/* db is pointed to by the dnode */
/* ASSERT3U(db->db_blkid, <, dn->dn_nblkptr); */
if (DMU_OBJECT_IS_SPECIAL(db->db.db_object))
ASSERT(db->db_parent == NULL);
else
ASSERT(db->db_parent != NULL);
if (db->db_blkid != DMU_SPILL_BLKID)
ASSERT3P(db->db_blkptr, ==,
&dn->dn_phys->dn_blkptr[db->db_blkid]);
} else {
/* db is pointed to by an indirect block */
int epb __maybe_unused = db->db_parent->db.db_size >>
SPA_BLKPTRSHIFT;
ASSERT3U(db->db_parent->db_level, ==, db->db_level+1);
ASSERT3U(db->db_parent->db.db_object, ==,
db->db.db_object);
/*
* dnode_grow_indblksz() can make this fail if we don't
* have the parent's rwlock. XXX indblksz no longer
* grows. safe to do this now?
*/
if (RW_LOCK_HELD(&db->db_parent->db_rwlock)) {
ASSERT3P(db->db_blkptr, ==,
((blkptr_t *)db->db_parent->db.db_data +
db->db_blkid % epb));
}
}
}
if ((db->db_blkptr == NULL || BP_IS_HOLE(db->db_blkptr)) &&
(db->db_buf == NULL || db->db_buf->b_data) &&
db->db.db_data && db->db_blkid != DMU_BONUS_BLKID &&
db->db_state != DB_FILL && !dn->dn_free_txg) {
/*
* If the blkptr isn't set but they have nonzero data,
* it had better be dirty, otherwise we'll lose that
* data when we evict this buffer.
*
* There is an exception to this rule for indirect blocks; in
* this case, if the indirect block is a hole, we fill in a few
* fields on each of the child blocks (importantly, birth time)
* to prevent hole birth times from being lost when you
* partially fill in a hole.
*/
if (db->db_dirtycnt == 0) {
if (db->db_level == 0) {
uint64_t *buf = db->db.db_data;
int i;
for (i = 0; i < db->db.db_size >> 3; i++) {
ASSERT(buf[i] == 0);
}
} else {
blkptr_t *bps = db->db.db_data;
ASSERT3U(1 << DB_DNODE(db)->dn_indblkshift, ==,
db->db.db_size);
/*
* We want to verify that all the blkptrs in the
* indirect block are holes, but we may have
* automatically set up a few fields for them.
* We iterate through each blkptr and verify
* they only have those fields set.
*/
for (int i = 0;
i < db->db.db_size / sizeof (blkptr_t);
i++) {
blkptr_t *bp = &bps[i];
ASSERT(ZIO_CHECKSUM_IS_ZERO(
&bp->blk_cksum));
ASSERT(
DVA_IS_EMPTY(&bp->blk_dva[0]) &&
DVA_IS_EMPTY(&bp->blk_dva[1]) &&
DVA_IS_EMPTY(&bp->blk_dva[2]));
ASSERT0(bp->blk_fill);
ASSERT0(bp->blk_pad[0]);
ASSERT0(bp->blk_pad[1]);
ASSERT(!BP_IS_EMBEDDED(bp));
ASSERT(BP_IS_HOLE(bp));
ASSERT0(bp->blk_phys_birth);
}
}
}
}
DB_DNODE_EXIT(db);
}
#endif
static void
dbuf_clear_data(dmu_buf_impl_t *db)
{
ASSERT(MUTEX_HELD(&db->db_mtx));
dbuf_evict_user(db);
ASSERT3P(db->db_buf, ==, NULL);
db->db.db_data = NULL;
if (db->db_state != DB_NOFILL) {
db->db_state = DB_UNCACHED;
DTRACE_SET_STATE(db, "clear data");
}
}
static void
dbuf_set_data(dmu_buf_impl_t *db, arc_buf_t *buf)
{
ASSERT(MUTEX_HELD(&db->db_mtx));
ASSERT(buf != NULL);
db->db_buf = buf;
ASSERT(buf->b_data != NULL);
db->db.db_data = buf->b_data;
}
static arc_buf_t *
dbuf_alloc_arcbuf(dmu_buf_impl_t *db)
{
spa_t *spa = db->db_objset->os_spa;
return (arc_alloc_buf(spa, db, DBUF_GET_BUFC_TYPE(db), db->db.db_size));
}
/*
* Loan out an arc_buf for read. Return the loaned arc_buf.
*/
arc_buf_t *
dbuf_loan_arcbuf(dmu_buf_impl_t *db)
{
arc_buf_t *abuf;
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
mutex_enter(&db->db_mtx);
if (arc_released(db->db_buf) || zfs_refcount_count(&db->db_holds) > 1) {
int blksz = db->db.db_size;
spa_t *spa = db->db_objset->os_spa;
mutex_exit(&db->db_mtx);
abuf = arc_loan_buf(spa, B_FALSE, blksz);
bcopy(db->db.db_data, abuf->b_data, blksz);
} else {
abuf = db->db_buf;
arc_loan_inuse_buf(abuf, db);
db->db_buf = NULL;
dbuf_clear_data(db);
mutex_exit(&db->db_mtx);
}
return (abuf);
}
/*
* Calculate which level n block references the data at the level 0 offset
* provided.
*/
uint64_t
dbuf_whichblock(const dnode_t *dn, const int64_t level, const uint64_t offset)
{
if (dn->dn_datablkshift != 0 && dn->dn_indblkshift != 0) {
/*
* The level n blkid is equal to the level 0 blkid divided by
* the number of level 0s in a level n block.
*
* The level 0 blkid is offset >> datablkshift =
* offset / 2^datablkshift.
*
* The number of level 0s in a level n is the number of block
* pointers in an indirect block, raised to the power of level.
* This is 2^(indblkshift - SPA_BLKPTRSHIFT)^level =
* 2^(level*(indblkshift - SPA_BLKPTRSHIFT)).
*
* Thus, the level n blkid is: offset /
* ((2^datablkshift)*(2^(level*(indblkshift-SPA_BLKPTRSHIFT))))
* = offset / 2^(datablkshift + level *
* (indblkshift - SPA_BLKPTRSHIFT))
* = offset >> (datablkshift + level *
* (indblkshift - SPA_BLKPTRSHIFT))
*/
const unsigned exp = dn->dn_datablkshift +
level * (dn->dn_indblkshift - SPA_BLKPTRSHIFT);
if (exp >= 8 * sizeof (offset)) {
/* This only happens on the highest indirection level */
ASSERT3U(level, ==, dn->dn_nlevels - 1);
return (0);
}
ASSERT3U(exp, <, 8 * sizeof (offset));
return (offset >> exp);
} else {
ASSERT3U(offset, <, dn->dn_datablksz);
return (0);
}
}
/*
* This function is used to lock the parent of the provided dbuf. This should be
* used when modifying or reading db_blkptr.
*/
db_lock_type_t
dmu_buf_lock_parent(dmu_buf_impl_t *db, krw_t rw, void *tag)
{
enum db_lock_type ret = DLT_NONE;
if (db->db_parent != NULL) {
rw_enter(&db->db_parent->db_rwlock, rw);
ret = DLT_PARENT;
} else if (dmu_objset_ds(db->db_objset) != NULL) {
rrw_enter(&dmu_objset_ds(db->db_objset)->ds_bp_rwlock, rw,
tag);
ret = DLT_OBJSET;
}
/*
* We only return a DLT_NONE lock when it's the top-most indirect block
* of the meta-dnode of the MOS.
*/
return (ret);
}
/*
* We need to pass the lock type in because it's possible that the block will
* move from being the topmost indirect block in a dnode (and thus, have no
* parent) to not the top-most via an indirection increase. This would cause a
* panic if we didn't pass the lock type in.
*/
void
dmu_buf_unlock_parent(dmu_buf_impl_t *db, db_lock_type_t type, void *tag)
{
if (type == DLT_PARENT)
rw_exit(&db->db_parent->db_rwlock);
else if (type == DLT_OBJSET)
rrw_exit(&dmu_objset_ds(db->db_objset)->ds_bp_rwlock, tag);
}
static void
dbuf_read_done(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp,
arc_buf_t *buf, void *vdb)
{
(void) zb, (void) bp;
dmu_buf_impl_t *db = vdb;
mutex_enter(&db->db_mtx);
ASSERT3U(db->db_state, ==, DB_READ);
/*
* All reads are synchronous, so we must have a hold on the dbuf
*/
ASSERT(zfs_refcount_count(&db->db_holds) > 0);
ASSERT(db->db_buf == NULL);
ASSERT(db->db.db_data == NULL);
if (buf == NULL) {
/* i/o error */
ASSERT(zio == NULL || zio->io_error != 0);
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
ASSERT3P(db->db_buf, ==, NULL);
db->db_state = DB_UNCACHED;
DTRACE_SET_STATE(db, "i/o error");
} else if (db->db_level == 0 && db->db_freed_in_flight) {
/* freed in flight */
ASSERT(zio == NULL || zio->io_error == 0);
arc_release(buf, db);
bzero(buf->b_data, db->db.db_size);
arc_buf_freeze(buf);
db->db_freed_in_flight = FALSE;
dbuf_set_data(db, buf);
db->db_state = DB_CACHED;
DTRACE_SET_STATE(db, "freed in flight");
} else {
/* success */
ASSERT(zio == NULL || zio->io_error == 0);
dbuf_set_data(db, buf);
db->db_state = DB_CACHED;
DTRACE_SET_STATE(db, "successful read");
}
cv_broadcast(&db->db_changed);
dbuf_rele_and_unlock(db, NULL, B_FALSE);
}
/*
* Shortcut for performing reads on bonus dbufs. Returns
* an error if we fail to verify the dnode associated with
* a decrypted block. Otherwise success.
*/
static int
dbuf_read_bonus(dmu_buf_impl_t *db, dnode_t *dn, uint32_t flags)
{
int bonuslen, max_bonuslen, err;
err = dbuf_read_verify_dnode_crypt(db, flags);
if (err)
return (err);
bonuslen = MIN(dn->dn_bonuslen, dn->dn_phys->dn_bonuslen);
max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
ASSERT(MUTEX_HELD(&db->db_mtx));
ASSERT(DB_DNODE_HELD(db));
ASSERT3U(bonuslen, <=, db->db.db_size);
db->db.db_data = kmem_alloc(max_bonuslen, KM_SLEEP);
arc_space_consume(max_bonuslen, ARC_SPACE_BONUS);
if (bonuslen < max_bonuslen)
bzero(db->db.db_data, max_bonuslen);
if (bonuslen)
bcopy(DN_BONUS(dn->dn_phys), db->db.db_data, bonuslen);
db->db_state = DB_CACHED;
DTRACE_SET_STATE(db, "bonus buffer filled");
return (0);
}
static void
dbuf_handle_indirect_hole(dmu_buf_impl_t *db, dnode_t *dn)
{
blkptr_t *bps = db->db.db_data;
uint32_t indbs = 1ULL << dn->dn_indblkshift;
int n_bps = indbs >> SPA_BLKPTRSHIFT;
for (int i = 0; i < n_bps; i++) {
blkptr_t *bp = &bps[i];
ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==, indbs);
BP_SET_LSIZE(bp, BP_GET_LEVEL(db->db_blkptr) == 1 ?
dn->dn_datablksz : BP_GET_LSIZE(db->db_blkptr));
BP_SET_TYPE(bp, BP_GET_TYPE(db->db_blkptr));
BP_SET_LEVEL(bp, BP_GET_LEVEL(db->db_blkptr) - 1);
BP_SET_BIRTH(bp, db->db_blkptr->blk_birth, 0);
}
}
/*
* Handle reads on dbufs that are holes, if necessary. This function
* requires that the dbuf's mutex is held. Returns success (0) if action
* was taken, ENOENT if no action was taken.
*/
static int
dbuf_read_hole(dmu_buf_impl_t *db, dnode_t *dn)
{
ASSERT(MUTEX_HELD(&db->db_mtx));
int is_hole = db->db_blkptr == NULL || BP_IS_HOLE(db->db_blkptr);
/*
* For level 0 blocks only, if the above check fails:
* Recheck BP_IS_HOLE() after dnode_block_freed() in case dnode_sync()
* processes the delete record and clears the bp while we are waiting
* for the dn_mtx (resulting in a "no" from block_freed).
*/
if (!is_hole && db->db_level == 0) {
is_hole = dnode_block_freed(dn, db->db_blkid) ||
BP_IS_HOLE(db->db_blkptr);
}
if (is_hole) {
dbuf_set_data(db, dbuf_alloc_arcbuf(db));
bzero(db->db.db_data, db->db.db_size);
if (db->db_blkptr != NULL && db->db_level > 0 &&
BP_IS_HOLE(db->db_blkptr) &&
db->db_blkptr->blk_birth != 0) {
dbuf_handle_indirect_hole(db, dn);
}
db->db_state = DB_CACHED;
DTRACE_SET_STATE(db, "hole read satisfied");
return (0);
}
return (ENOENT);
}
/*
* This function ensures that, when doing a decrypting read of a block,
* we make sure we have decrypted the dnode associated with it. We must do
* this so that we ensure we are fully authenticating the checksum-of-MACs
* tree from the root of the objset down to this block. Indirect blocks are
* always verified against their secure checksum-of-MACs assuming that the
* dnode containing them is correct. Now that we are doing a decrypting read,
* we can be sure that the key is loaded and verify that assumption. This is
* especially important considering that we always read encrypted dnode
* blocks as raw data (without verifying their MACs) to start, and
* decrypt / authenticate them when we need to read an encrypted bonus buffer.
*/
static int
dbuf_read_verify_dnode_crypt(dmu_buf_impl_t *db, uint32_t flags)
{
int err = 0;
objset_t *os = db->db_objset;
arc_buf_t *dnode_abuf;
dnode_t *dn;
zbookmark_phys_t zb;
ASSERT(MUTEX_HELD(&db->db_mtx));
if (!os->os_encrypted || os->os_raw_receive ||
(flags & DB_RF_NO_DECRYPT) != 0)
return (0);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
dnode_abuf = (dn->dn_dbuf != NULL) ? dn->dn_dbuf->db_buf : NULL;
if (dnode_abuf == NULL || !arc_is_encrypted(dnode_abuf)) {
DB_DNODE_EXIT(db);
return (0);
}
SET_BOOKMARK(&zb, dmu_objset_id(os),
DMU_META_DNODE_OBJECT, 0, dn->dn_dbuf->db_blkid);
err = arc_untransform(dnode_abuf, os->os_spa, &zb, B_TRUE);
/*
* An error code of EACCES tells us that the key is still not
* available. This is ok if we are only reading authenticated
* (and therefore non-encrypted) blocks.
*/
if (err == EACCES && ((db->db_blkid != DMU_BONUS_BLKID &&
!DMU_OT_IS_ENCRYPTED(dn->dn_type)) ||
(db->db_blkid == DMU_BONUS_BLKID &&
!DMU_OT_IS_ENCRYPTED(dn->dn_bonustype))))
err = 0;
DB_DNODE_EXIT(db);
return (err);
}
/*
* Drops db_mtx and the parent lock specified by dblt and tag before
* returning.
*/
static int
dbuf_read_impl(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags,
db_lock_type_t dblt, void *tag)
{
dnode_t *dn;
zbookmark_phys_t zb;
uint32_t aflags = ARC_FLAG_NOWAIT;
int err, zio_flags;
err = zio_flags = 0;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
ASSERT(!zfs_refcount_is_zero(&db->db_holds));
ASSERT(MUTEX_HELD(&db->db_mtx));
ASSERT(db->db_state == DB_UNCACHED);
ASSERT(db->db_buf == NULL);
ASSERT(db->db_parent == NULL ||
RW_LOCK_HELD(&db->db_parent->db_rwlock));
if (db->db_blkid == DMU_BONUS_BLKID) {
err = dbuf_read_bonus(db, dn, flags);
goto early_unlock;
}
err = dbuf_read_hole(db, dn);
if (err == 0)
goto early_unlock;
/*
* Any attempt to read a redacted block should result in an error. This
* will never happen under normal conditions, but can be useful for
* debugging purposes.
*/
if (BP_IS_REDACTED(db->db_blkptr)) {
ASSERT(dsl_dataset_feature_is_active(
db->db_objset->os_dsl_dataset,
SPA_FEATURE_REDACTED_DATASETS));
err = SET_ERROR(EIO);
goto early_unlock;
}
SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset),
db->db.db_object, db->db_level, db->db_blkid);
/*
* All bps of an encrypted os should have the encryption bit set.
* If this is not true it indicates tampering and we report an error.
*/
if (db->db_objset->os_encrypted && !BP_USES_CRYPT(db->db_blkptr)) {
spa_log_error(db->db_objset->os_spa, &zb);
zfs_panic_recover("unencrypted block in encrypted "
"object set %llu", dmu_objset_id(db->db_objset));
err = SET_ERROR(EIO);
goto early_unlock;
}
err = dbuf_read_verify_dnode_crypt(db, flags);
if (err != 0)
goto early_unlock;
DB_DNODE_EXIT(db);
db->db_state = DB_READ;
DTRACE_SET_STATE(db, "read issued");
mutex_exit(&db->db_mtx);
if (dbuf_is_l2cacheable(db))
aflags |= ARC_FLAG_L2CACHE;
dbuf_add_ref(db, NULL);
zio_flags = (flags & DB_RF_CANFAIL) ?
ZIO_FLAG_CANFAIL : ZIO_FLAG_MUSTSUCCEED;
if ((flags & DB_RF_NO_DECRYPT) && BP_IS_PROTECTED(db->db_blkptr))
zio_flags |= ZIO_FLAG_RAW;
/*
* The zio layer will copy the provided blkptr later, but we need to
* do this now so that we can release the parent's rwlock. We have to
* do that now so that if dbuf_read_done is called synchronously (on
* an l1 cache hit) we don't acquire the db_mtx while holding the
* parent's rwlock, which would be a lock ordering violation.
*/
blkptr_t bp = *db->db_blkptr;
dmu_buf_unlock_parent(db, dblt, tag);
(void) arc_read(zio, db->db_objset->os_spa, &bp,
dbuf_read_done, db, ZIO_PRIORITY_SYNC_READ, zio_flags,
&aflags, &zb);
return (err);
early_unlock:
DB_DNODE_EXIT(db);
mutex_exit(&db->db_mtx);
dmu_buf_unlock_parent(db, dblt, tag);
return (err);
}
/*
* This is our just-in-time copy function. It makes a copy of buffers that
* have been modified in a previous transaction group before we access them in
* the current active group.
*
* This function is used in three places: when we are dirtying a buffer for the
* first time in a txg, when we are freeing a range in a dnode that includes
* this buffer, and when we are accessing a buffer which was received compressed
* and later referenced in a WRITE_BYREF record.
*
* Note that when we are called from dbuf_free_range() we do not put a hold on
* the buffer, we just traverse the active dbuf list for the dnode.
*/
static void
dbuf_fix_old_data(dmu_buf_impl_t *db, uint64_t txg)
{
dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records);
ASSERT(MUTEX_HELD(&db->db_mtx));
ASSERT(db->db.db_data != NULL);
ASSERT(db->db_level == 0);
ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT);
if (dr == NULL ||
(dr->dt.dl.dr_data !=
((db->db_blkid == DMU_BONUS_BLKID) ? db->db.db_data : db->db_buf)))
return;
/*
* If the last dirty record for this dbuf has not yet synced
* and its referencing the dbuf data, either:
* reset the reference to point to a new copy,
* or (if there a no active holders)
* just null out the current db_data pointer.
*/
ASSERT3U(dr->dr_txg, >=, txg - 2);
if (db->db_blkid == DMU_BONUS_BLKID) {
dnode_t *dn = DB_DNODE(db);
int bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
dr->dt.dl.dr_data = kmem_alloc(bonuslen, KM_SLEEP);
arc_space_consume(bonuslen, ARC_SPACE_BONUS);
bcopy(db->db.db_data, dr->dt.dl.dr_data, bonuslen);
} else if (zfs_refcount_count(&db->db_holds) > db->db_dirtycnt) {
dnode_t *dn = DB_DNODE(db);
int size = arc_buf_size(db->db_buf);
arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
spa_t *spa = db->db_objset->os_spa;
enum zio_compress compress_type =
arc_get_compression(db->db_buf);
uint8_t complevel = arc_get_complevel(db->db_buf);
if (arc_is_encrypted(db->db_buf)) {
boolean_t byteorder;
uint8_t salt[ZIO_DATA_SALT_LEN];
uint8_t iv[ZIO_DATA_IV_LEN];
uint8_t mac[ZIO_DATA_MAC_LEN];
arc_get_raw_params(db->db_buf, &byteorder, salt,
iv, mac);
dr->dt.dl.dr_data = arc_alloc_raw_buf(spa, db,
dmu_objset_id(dn->dn_objset), byteorder, salt, iv,
mac, dn->dn_type, size, arc_buf_lsize(db->db_buf),
compress_type, complevel);
} else if (compress_type != ZIO_COMPRESS_OFF) {
ASSERT3U(type, ==, ARC_BUFC_DATA);
dr->dt.dl.dr_data = arc_alloc_compressed_buf(spa, db,
size, arc_buf_lsize(db->db_buf), compress_type,
complevel);
} else {
dr->dt.dl.dr_data = arc_alloc_buf(spa, db, type, size);
}
bcopy(db->db.db_data, dr->dt.dl.dr_data->b_data, size);
} else {
db->db_buf = NULL;
dbuf_clear_data(db);
}
}
int
dbuf_read(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags)
{
int err = 0;
boolean_t prefetch;
dnode_t *dn;
/*
* We don't have to hold the mutex to check db_state because it
* can't be freed while we have a hold on the buffer.
*/
ASSERT(!zfs_refcount_is_zero(&db->db_holds));
if (db->db_state == DB_NOFILL)
return (SET_ERROR(EIO));
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
prefetch = db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
(flags & DB_RF_NOPREFETCH) == 0 && dn != NULL &&
DBUF_IS_CACHEABLE(db);
mutex_enter(&db->db_mtx);
if (db->db_state == DB_CACHED) {
spa_t *spa = dn->dn_objset->os_spa;
/*
* Ensure that this block's dnode has been decrypted if
* the caller has requested decrypted data.
*/
err = dbuf_read_verify_dnode_crypt(db, flags);
/*
* If the arc buf is compressed or encrypted and the caller
* requested uncompressed data, we need to untransform it
* before returning. We also call arc_untransform() on any
* unauthenticated blocks, which will verify their MAC if
* the key is now available.
*/
if (err == 0 && db->db_buf != NULL &&
(flags & DB_RF_NO_DECRYPT) == 0 &&
(arc_is_encrypted(db->db_buf) ||
arc_is_unauthenticated(db->db_buf) ||
arc_get_compression(db->db_buf) != ZIO_COMPRESS_OFF)) {
zbookmark_phys_t zb;
SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset),
db->db.db_object, db->db_level, db->db_blkid);
dbuf_fix_old_data(db, spa_syncing_txg(spa));
err = arc_untransform(db->db_buf, spa, &zb, B_FALSE);
dbuf_set_data(db, db->db_buf);
}
mutex_exit(&db->db_mtx);
if (err == 0 && prefetch) {
dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE,
B_FALSE, flags & DB_RF_HAVESTRUCT);
}
DB_DNODE_EXIT(db);
DBUF_STAT_BUMP(hash_hits);
} else if (db->db_state == DB_UNCACHED) {
spa_t *spa = dn->dn_objset->os_spa;
boolean_t need_wait = B_FALSE;
db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_READER, FTAG);
if (zio == NULL &&
db->db_blkptr != NULL && !BP_IS_HOLE(db->db_blkptr)) {
zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
need_wait = B_TRUE;
}
err = dbuf_read_impl(db, zio, flags, dblt, FTAG);
/*
* dbuf_read_impl has dropped db_mtx and our parent's rwlock
* for us
*/
if (!err && prefetch) {
dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE,
db->db_state != DB_CACHED,
flags & DB_RF_HAVESTRUCT);
}
DB_DNODE_EXIT(db);
DBUF_STAT_BUMP(hash_misses);
/*
* If we created a zio_root we must execute it to avoid
* leaking it, even if it isn't attached to any work due
* to an error in dbuf_read_impl().
*/
if (need_wait) {
if (err == 0)
err = zio_wait(zio);
else
VERIFY0(zio_wait(zio));
}
} else {
/*
* Another reader came in while the dbuf was in flight
* between UNCACHED and CACHED. Either a writer will finish
* writing the buffer (sending the dbuf to CACHED) or the
* first reader's request will reach the read_done callback
* and send the dbuf to CACHED. Otherwise, a failure
* occurred and the dbuf went to UNCACHED.
*/
mutex_exit(&db->db_mtx);
if (prefetch) {
dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE,
B_TRUE, flags & DB_RF_HAVESTRUCT);
}
DB_DNODE_EXIT(db);
DBUF_STAT_BUMP(hash_misses);
/* Skip the wait per the caller's request. */
if ((flags & DB_RF_NEVERWAIT) == 0) {
mutex_enter(&db->db_mtx);
while (db->db_state == DB_READ ||
db->db_state == DB_FILL) {
ASSERT(db->db_state == DB_READ ||
(flags & DB_RF_HAVESTRUCT) == 0);
DTRACE_PROBE2(blocked__read, dmu_buf_impl_t *,
db, zio_t *, zio);
cv_wait(&db->db_changed, &db->db_mtx);
}
if (db->db_state == DB_UNCACHED)
err = SET_ERROR(EIO);
mutex_exit(&db->db_mtx);
}
}
return (err);
}
static void
dbuf_noread(dmu_buf_impl_t *db)
{
ASSERT(!zfs_refcount_is_zero(&db->db_holds));
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
mutex_enter(&db->db_mtx);
while (db->db_state == DB_READ || db->db_state == DB_FILL)
cv_wait(&db->db_changed, &db->db_mtx);
if (db->db_state == DB_UNCACHED) {
ASSERT(db->db_buf == NULL);
ASSERT(db->db.db_data == NULL);
dbuf_set_data(db, dbuf_alloc_arcbuf(db));
db->db_state = DB_FILL;
DTRACE_SET_STATE(db, "assigning filled buffer");
} else if (db->db_state == DB_NOFILL) {
dbuf_clear_data(db);
} else {
ASSERT3U(db->db_state, ==, DB_CACHED);
}
mutex_exit(&db->db_mtx);
}
void
dbuf_unoverride(dbuf_dirty_record_t *dr)
{
dmu_buf_impl_t *db = dr->dr_dbuf;
blkptr_t *bp = &dr->dt.dl.dr_overridden_by;
uint64_t txg = dr->dr_txg;
ASSERT(MUTEX_HELD(&db->db_mtx));
/*
* This assert is valid because dmu_sync() expects to be called by
* a zilog's get_data while holding a range lock. This call only
* comes from dbuf_dirty() callers who must also hold a range lock.
*/
ASSERT(dr->dt.dl.dr_override_state != DR_IN_DMU_SYNC);
ASSERT(db->db_level == 0);
if (db->db_blkid == DMU_BONUS_BLKID ||
dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN)
return;
ASSERT(db->db_data_pending != dr);
/* free this block */
if (!BP_IS_HOLE(bp) && !dr->dt.dl.dr_nopwrite)
zio_free(db->db_objset->os_spa, txg, bp);
dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
dr->dt.dl.dr_nopwrite = B_FALSE;
dr->dt.dl.dr_has_raw_params = B_FALSE;
/*
* Release the already-written buffer, so we leave it in
* a consistent dirty state. Note that all callers are
* modifying the buffer, so they will immediately do
* another (redundant) arc_release(). Therefore, leave
* the buf thawed to save the effort of freezing &
* immediately re-thawing it.
*/
arc_release(dr->dt.dl.dr_data, db);
}
/*
* Evict (if its unreferenced) or clear (if its referenced) any level-0
* data blocks in the free range, so that any future readers will find
* empty blocks.
*/
void
dbuf_free_range(dnode_t *dn, uint64_t start_blkid, uint64_t end_blkid,
dmu_tx_t *tx)
{
dmu_buf_impl_t *db_search;
dmu_buf_impl_t *db, *db_next;
uint64_t txg = tx->tx_txg;
avl_index_t where;
dbuf_dirty_record_t *dr;
if (end_blkid > dn->dn_maxblkid &&
!(start_blkid == DMU_SPILL_BLKID || end_blkid == DMU_SPILL_BLKID))
end_blkid = dn->dn_maxblkid;
dprintf_dnode(dn, "start=%llu end=%llu\n", (u_longlong_t)start_blkid,
(u_longlong_t)end_blkid);
db_search = kmem_alloc(sizeof (dmu_buf_impl_t), KM_SLEEP);
db_search->db_level = 0;
db_search->db_blkid = start_blkid;
db_search->db_state = DB_SEARCH;
mutex_enter(&dn->dn_dbufs_mtx);
db = avl_find(&dn->dn_dbufs, db_search, &where);
ASSERT3P(db, ==, NULL);
db = avl_nearest(&dn->dn_dbufs, where, AVL_AFTER);
for (; db != NULL; db = db_next) {
db_next = AVL_NEXT(&dn->dn_dbufs, db);
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
if (db->db_level != 0 || db->db_blkid > end_blkid) {
break;
}
ASSERT3U(db->db_blkid, >=, start_blkid);
/* found a level 0 buffer in the range */
mutex_enter(&db->db_mtx);
if (dbuf_undirty(db, tx)) {
/* mutex has been dropped and dbuf destroyed */
continue;
}
if (db->db_state == DB_UNCACHED ||
db->db_state == DB_NOFILL ||
db->db_state == DB_EVICTING) {
ASSERT(db->db.db_data == NULL);
mutex_exit(&db->db_mtx);
continue;
}
if (db->db_state == DB_READ || db->db_state == DB_FILL) {
/* will be handled in dbuf_read_done or dbuf_rele */
db->db_freed_in_flight = TRUE;
mutex_exit(&db->db_mtx);
continue;
}
if (zfs_refcount_count(&db->db_holds) == 0) {
ASSERT(db->db_buf);
dbuf_destroy(db);
continue;
}
/* The dbuf is referenced */
dr = list_head(&db->db_dirty_records);
if (dr != NULL) {
if (dr->dr_txg == txg) {
/*
* This buffer is "in-use", re-adjust the file
* size to reflect that this buffer may
* contain new data when we sync.
*/
if (db->db_blkid != DMU_SPILL_BLKID &&
db->db_blkid > dn->dn_maxblkid)
dn->dn_maxblkid = db->db_blkid;
dbuf_unoverride(dr);
} else {
/*
* This dbuf is not dirty in the open context.
* Either uncache it (if its not referenced in
* the open context) or reset its contents to
* empty.
*/
dbuf_fix_old_data(db, txg);
}
}
/* clear the contents if its cached */
if (db->db_state == DB_CACHED) {
ASSERT(db->db.db_data != NULL);
arc_release(db->db_buf, db);
rw_enter(&db->db_rwlock, RW_WRITER);
bzero(db->db.db_data, db->db.db_size);
rw_exit(&db->db_rwlock);
arc_buf_freeze(db->db_buf);
}
mutex_exit(&db->db_mtx);
}
mutex_exit(&dn->dn_dbufs_mtx);
kmem_free(db_search, sizeof (dmu_buf_impl_t));
}
void
dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx)
{
arc_buf_t *buf, *old_buf;
dbuf_dirty_record_t *dr;
int osize = db->db.db_size;
arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
dnode_t *dn;
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
/*
* XXX we should be doing a dbuf_read, checking the return
* value and returning that up to our callers
*/
dmu_buf_will_dirty(&db->db, tx);
/* create the data buffer for the new block */
buf = arc_alloc_buf(dn->dn_objset->os_spa, db, type, size);
/* copy old block data to the new block */
old_buf = db->db_buf;
bcopy(old_buf->b_data, buf->b_data, MIN(osize, size));
/* zero the remainder */
if (size > osize)
bzero((uint8_t *)buf->b_data + osize, size - osize);
mutex_enter(&db->db_mtx);
dbuf_set_data(db, buf);
arc_buf_destroy(old_buf, db);
db->db.db_size = size;
dr = list_head(&db->db_dirty_records);
/* dirty record added by dmu_buf_will_dirty() */
VERIFY(dr != NULL);
if (db->db_level == 0)
dr->dt.dl.dr_data = buf;
ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
ASSERT3U(dr->dr_accounted, ==, osize);
dr->dr_accounted = size;
mutex_exit(&db->db_mtx);
dmu_objset_willuse_space(dn->dn_objset, size - osize, tx);
DB_DNODE_EXIT(db);
}
void
dbuf_release_bp(dmu_buf_impl_t *db)
{
objset_t *os __maybe_unused = db->db_objset;
ASSERT(dsl_pool_sync_context(dmu_objset_pool(os)));
ASSERT(arc_released(os->os_phys_buf) ||
list_link_active(&os->os_dsl_dataset->ds_synced_link));
ASSERT(db->db_parent == NULL || arc_released(db->db_parent->db_buf));
(void) arc_release(db->db_buf, db);
}
/*
* We already have a dirty record for this TXG, and we are being
* dirtied again.
*/
static void
dbuf_redirty(dbuf_dirty_record_t *dr)
{
dmu_buf_impl_t *db = dr->dr_dbuf;
ASSERT(MUTEX_HELD(&db->db_mtx));
if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID) {
/*
* If this buffer has already been written out,
* we now need to reset its state.
*/
dbuf_unoverride(dr);
if (db->db.db_object != DMU_META_DNODE_OBJECT &&
db->db_state != DB_NOFILL) {
/* Already released on initial dirty, so just thaw. */
ASSERT(arc_released(db->db_buf));
arc_buf_thaw(db->db_buf);
}
}
}
dbuf_dirty_record_t *
dbuf_dirty_lightweight(dnode_t *dn, uint64_t blkid, dmu_tx_t *tx)
{
rw_enter(&dn->dn_struct_rwlock, RW_READER);
IMPLY(dn->dn_objset->os_raw_receive, dn->dn_maxblkid >= blkid);
dnode_new_blkid(dn, blkid, tx, B_TRUE, B_FALSE);
ASSERT(dn->dn_maxblkid >= blkid);
dbuf_dirty_record_t *dr = kmem_zalloc(sizeof (*dr), KM_SLEEP);
list_link_init(&dr->dr_dirty_node);
list_link_init(&dr->dr_dbuf_node);
dr->dr_dnode = dn;
dr->dr_txg = tx->tx_txg;
dr->dt.dll.dr_blkid = blkid;
dr->dr_accounted = dn->dn_datablksz;
/*
* There should not be any dbuf for the block that we're dirtying.
* Otherwise the buffer contents could be inconsistent between the
* dbuf and the lightweight dirty record.
*/
ASSERT3P(NULL, ==, dbuf_find(dn->dn_objset, dn->dn_object, 0, blkid));
mutex_enter(&dn->dn_mtx);
int txgoff = tx->tx_txg & TXG_MASK;
if (dn->dn_free_ranges[txgoff] != NULL) {
range_tree_clear(dn->dn_free_ranges[txgoff], blkid, 1);
}
if (dn->dn_nlevels == 1) {
ASSERT3U(blkid, <, dn->dn_nblkptr);
list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
mutex_exit(&dn->dn_mtx);
rw_exit(&dn->dn_struct_rwlock);
dnode_setdirty(dn, tx);
} else {
mutex_exit(&dn->dn_mtx);
int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
dmu_buf_impl_t *parent_db = dbuf_hold_level(dn,
1, blkid >> epbs, FTAG);
rw_exit(&dn->dn_struct_rwlock);
if (parent_db == NULL) {
kmem_free(dr, sizeof (*dr));
return (NULL);
}
int err = dbuf_read(parent_db, NULL,
(DB_RF_NOPREFETCH | DB_RF_CANFAIL));
if (err != 0) {
dbuf_rele(parent_db, FTAG);
kmem_free(dr, sizeof (*dr));
return (NULL);
}
dbuf_dirty_record_t *parent_dr = dbuf_dirty(parent_db, tx);
dbuf_rele(parent_db, FTAG);
mutex_enter(&parent_dr->dt.di.dr_mtx);
ASSERT3U(parent_dr->dr_txg, ==, tx->tx_txg);
list_insert_tail(&parent_dr->dt.di.dr_children, dr);
mutex_exit(&parent_dr->dt.di.dr_mtx);
dr->dr_parent = parent_dr;
}
dmu_objset_willuse_space(dn->dn_objset, dr->dr_accounted, tx);
return (dr);
}
dbuf_dirty_record_t *
dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
{
dnode_t *dn;
objset_t *os;
dbuf_dirty_record_t *dr, *dr_next, *dr_head;
int txgoff = tx->tx_txg & TXG_MASK;
boolean_t drop_struct_rwlock = B_FALSE;
ASSERT(tx->tx_txg != 0);
ASSERT(!zfs_refcount_is_zero(&db->db_holds));
DMU_TX_DIRTY_BUF(tx, db);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
/*
* Shouldn't dirty a regular buffer in syncing context. Private
* objects may be dirtied in syncing context, but only if they
* were already pre-dirtied in open context.
*/
#ifdef ZFS_DEBUG
if (dn->dn_objset->os_dsl_dataset != NULL) {
rrw_enter(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock,
RW_READER, FTAG);
}
ASSERT(!dmu_tx_is_syncing(tx) ||
BP_IS_HOLE(dn->dn_objset->os_rootbp) ||
DMU_OBJECT_IS_SPECIAL(dn->dn_object) ||
dn->dn_objset->os_dsl_dataset == NULL);
if (dn->dn_objset->os_dsl_dataset != NULL)
rrw_exit(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, FTAG);
#endif
/*
* We make this assert for private objects as well, but after we
* check if we're already dirty. They are allowed to re-dirty
* in syncing context.
*/
ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT ||
dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx ==
(dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN));
mutex_enter(&db->db_mtx);
/*
* XXX make this true for indirects too? The problem is that
* transactions created with dmu_tx_create_assigned() from
* syncing context don't bother holding ahead.
*/
ASSERT(db->db_level != 0 ||
db->db_state == DB_CACHED || db->db_state == DB_FILL ||
db->db_state == DB_NOFILL);
mutex_enter(&dn->dn_mtx);
dnode_set_dirtyctx(dn, tx, db);
if (tx->tx_txg > dn->dn_dirty_txg)
dn->dn_dirty_txg = tx->tx_txg;
mutex_exit(&dn->dn_mtx);
if (db->db_blkid == DMU_SPILL_BLKID)
dn->dn_have_spill = B_TRUE;
/*
* If this buffer is already dirty, we're done.
*/
dr_head = list_head(&db->db_dirty_records);
ASSERT(dr_head == NULL || dr_head->dr_txg <= tx->tx_txg ||
db->db.db_object == DMU_META_DNODE_OBJECT);
dr_next = dbuf_find_dirty_lte(db, tx->tx_txg);
if (dr_next && dr_next->dr_txg == tx->tx_txg) {
DB_DNODE_EXIT(db);
dbuf_redirty(dr_next);
mutex_exit(&db->db_mtx);
return (dr_next);
}
/*
* Only valid if not already dirty.
*/
ASSERT(dn->dn_object == 0 ||
dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx ==
(dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN));
ASSERT3U(dn->dn_nlevels, >, db->db_level);
/*
* We should only be dirtying in syncing context if it's the
* mos or we're initializing the os or it's a special object.
* However, we are allowed to dirty in syncing context provided
* we already dirtied it in open context. Hence we must make
* this assertion only if we're not already dirty.
*/
os = dn->dn_objset;
VERIFY3U(tx->tx_txg, <=, spa_final_dirty_txg(os->os_spa));
#ifdef ZFS_DEBUG
if (dn->dn_objset->os_dsl_dataset != NULL)
rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG);
ASSERT(!dmu_tx_is_syncing(tx) || DMU_OBJECT_IS_SPECIAL(dn->dn_object) ||
os->os_dsl_dataset == NULL || BP_IS_HOLE(os->os_rootbp));
if (dn->dn_objset->os_dsl_dataset != NULL)
rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG);
#endif
ASSERT(db->db.db_size != 0);
dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size);
if (db->db_blkid != DMU_BONUS_BLKID) {
dmu_objset_willuse_space(os, db->db.db_size, tx);
}
/*
* If this buffer is dirty in an old transaction group we need
* to make a copy of it so that the changes we make in this
* transaction group won't leak out when we sync the older txg.
*/
dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_SLEEP);
list_link_init(&dr->dr_dirty_node);
list_link_init(&dr->dr_dbuf_node);
dr->dr_dnode = dn;
if (db->db_level == 0) {
void *data_old = db->db_buf;
if (db->db_state != DB_NOFILL) {
if (db->db_blkid == DMU_BONUS_BLKID) {
dbuf_fix_old_data(db, tx->tx_txg);
data_old = db->db.db_data;
} else if (db->db.db_object != DMU_META_DNODE_OBJECT) {
/*
* Release the data buffer from the cache so
* that we can modify it without impacting
* possible other users of this cached data
* block. Note that indirect blocks and
* private objects are not released until the
* syncing state (since they are only modified
* then).
*/
arc_release(db->db_buf, db);
dbuf_fix_old_data(db, tx->tx_txg);
data_old = db->db_buf;
}
ASSERT(data_old != NULL);
}
dr->dt.dl.dr_data = data_old;
} else {
mutex_init(&dr->dt.di.dr_mtx, NULL, MUTEX_NOLOCKDEP, NULL);
list_create(&dr->dt.di.dr_children,
sizeof (dbuf_dirty_record_t),
offsetof(dbuf_dirty_record_t, dr_dirty_node));
}
if (db->db_blkid != DMU_BONUS_BLKID)
dr->dr_accounted = db->db.db_size;
dr->dr_dbuf = db;
dr->dr_txg = tx->tx_txg;
list_insert_before(&db->db_dirty_records, dr_next, dr);
/*
* We could have been freed_in_flight between the dbuf_noread
* and dbuf_dirty. We win, as though the dbuf_noread() had
* happened after the free.
*/
if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
db->db_blkid != DMU_SPILL_BLKID) {
mutex_enter(&dn->dn_mtx);
if (dn->dn_free_ranges[txgoff] != NULL) {
range_tree_clear(dn->dn_free_ranges[txgoff],
db->db_blkid, 1);
}
mutex_exit(&dn->dn_mtx);
db->db_freed_in_flight = FALSE;
}
/*
* This buffer is now part of this txg
*/
dbuf_add_ref(db, (void *)(uintptr_t)tx->tx_txg);
db->db_dirtycnt += 1;
ASSERT3U(db->db_dirtycnt, <=, 3);
mutex_exit(&db->db_mtx);
if (db->db_blkid == DMU_BONUS_BLKID ||
db->db_blkid == DMU_SPILL_BLKID) {
mutex_enter(&dn->dn_mtx);
ASSERT(!list_link_active(&dr->dr_dirty_node));
list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
mutex_exit(&dn->dn_mtx);
dnode_setdirty(dn, tx);
DB_DNODE_EXIT(db);
return (dr);
}
if (!RW_WRITE_HELD(&dn->dn_struct_rwlock)) {
rw_enter(&dn->dn_struct_rwlock, RW_READER);
drop_struct_rwlock = B_TRUE;
}
/*
* If we are overwriting a dedup BP, then unless it is snapshotted,
* when we get to syncing context we will need to decrement its
* refcount in the DDT. Prefetch the relevant DDT block so that
* syncing context won't have to wait for the i/o.
*/
if (db->db_blkptr != NULL) {
db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_READER, FTAG);
ddt_prefetch(os->os_spa, db->db_blkptr);
dmu_buf_unlock_parent(db, dblt, FTAG);
}
/*
* We need to hold the dn_struct_rwlock to make this assertion,
* because it protects dn_phys / dn_next_nlevels from changing.
*/
ASSERT((dn->dn_phys->dn_nlevels == 0 && db->db_level == 0) ||
dn->dn_phys->dn_nlevels > db->db_level ||
dn->dn_next_nlevels[txgoff] > db->db_level ||
dn->dn_next_nlevels[(tx->tx_txg-1) & TXG_MASK] > db->db_level ||
dn->dn_next_nlevels[(tx->tx_txg-2) & TXG_MASK] > db->db_level);
if (db->db_level == 0) {
ASSERT(!db->db_objset->os_raw_receive ||
dn->dn_maxblkid >= db->db_blkid);
dnode_new_blkid(dn, db->db_blkid, tx,
drop_struct_rwlock, B_FALSE);
ASSERT(dn->dn_maxblkid >= db->db_blkid);
}
if (db->db_level+1 < dn->dn_nlevels) {
dmu_buf_impl_t *parent = db->db_parent;
dbuf_dirty_record_t *di;
int parent_held = FALSE;
if (db->db_parent == NULL || db->db_parent == dn->dn_dbuf) {
int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
parent = dbuf_hold_level(dn, db->db_level + 1,
db->db_blkid >> epbs, FTAG);
ASSERT(parent != NULL);
parent_held = TRUE;
}
if (drop_struct_rwlock)
rw_exit(&dn->dn_struct_rwlock);
ASSERT3U(db->db_level + 1, ==, parent->db_level);
di = dbuf_dirty(parent, tx);
if (parent_held)
dbuf_rele(parent, FTAG);
mutex_enter(&db->db_mtx);
/*
* Since we've dropped the mutex, it's possible that
* dbuf_undirty() might have changed this out from under us.
*/
if (list_head(&db->db_dirty_records) == dr ||
dn->dn_object == DMU_META_DNODE_OBJECT) {
mutex_enter(&di->dt.di.dr_mtx);
ASSERT3U(di->dr_txg, ==, tx->tx_txg);
ASSERT(!list_link_active(&dr->dr_dirty_node));
list_insert_tail(&di->dt.di.dr_children, dr);
mutex_exit(&di->dt.di.dr_mtx);
dr->dr_parent = di;
}
mutex_exit(&db->db_mtx);
} else {
ASSERT(db->db_level + 1 == dn->dn_nlevels);
ASSERT(db->db_blkid < dn->dn_nblkptr);
ASSERT(db->db_parent == NULL || db->db_parent == dn->dn_dbuf);
mutex_enter(&dn->dn_mtx);
ASSERT(!list_link_active(&dr->dr_dirty_node));
list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
mutex_exit(&dn->dn_mtx);
if (drop_struct_rwlock)
rw_exit(&dn->dn_struct_rwlock);
}
dnode_setdirty(dn, tx);
DB_DNODE_EXIT(db);
return (dr);
}
static void
dbuf_undirty_bonus(dbuf_dirty_record_t *dr)
{
dmu_buf_impl_t *db = dr->dr_dbuf;
if (dr->dt.dl.dr_data != db->db.db_data) {
struct dnode *dn = dr->dr_dnode;
int max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
kmem_free(dr->dt.dl.dr_data, max_bonuslen);
arc_space_return(max_bonuslen, ARC_SPACE_BONUS);
}
db->db_data_pending = NULL;
ASSERT(list_next(&db->db_dirty_records, dr) == NULL);
list_remove(&db->db_dirty_records, dr);
if (dr->dr_dbuf->db_level != 0) {
mutex_destroy(&dr->dt.di.dr_mtx);
list_destroy(&dr->dt.di.dr_children);
}
kmem_free(dr, sizeof (dbuf_dirty_record_t));
ASSERT3U(db->db_dirtycnt, >, 0);
db->db_dirtycnt -= 1;
}
/*
* Undirty a buffer in the transaction group referenced by the given
* transaction. Return whether this evicted the dbuf.
*/
static boolean_t
dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
{
uint64_t txg = tx->tx_txg;
ASSERT(txg != 0);
/*
* Due to our use of dn_nlevels below, this can only be called
* in open context, unless we are operating on the MOS.
* From syncing context, dn_nlevels may be different from the
* dn_nlevels used when dbuf was dirtied.
*/
ASSERT(db->db_objset ==
dmu_objset_pool(db->db_objset)->dp_meta_objset ||
txg != spa_syncing_txg(dmu_objset_spa(db->db_objset)));
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
ASSERT0(db->db_level);
ASSERT(MUTEX_HELD(&db->db_mtx));
/*
* If this buffer is not dirty, we're done.
*/
dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, txg);
if (dr == NULL)
return (B_FALSE);
ASSERT(dr->dr_dbuf == db);
dnode_t *dn = dr->dr_dnode;
dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size);
ASSERT(db->db.db_size != 0);
dsl_pool_undirty_space(dmu_objset_pool(dn->dn_objset),
dr->dr_accounted, txg);
list_remove(&db->db_dirty_records, dr);
/*
* Note that there are three places in dbuf_dirty()
* where this dirty record may be put on a list.
* Make sure to do a list_remove corresponding to
* every one of those list_insert calls.
*/
if (dr->dr_parent) {
mutex_enter(&dr->dr_parent->dt.di.dr_mtx);
list_remove(&dr->dr_parent->dt.di.dr_children, dr);
mutex_exit(&dr->dr_parent->dt.di.dr_mtx);
} else if (db->db_blkid == DMU_SPILL_BLKID ||
db->db_level + 1 == dn->dn_nlevels) {
ASSERT(db->db_blkptr == NULL || db->db_parent == dn->dn_dbuf);
mutex_enter(&dn->dn_mtx);
list_remove(&dn->dn_dirty_records[txg & TXG_MASK], dr);
mutex_exit(&dn->dn_mtx);
}
if (db->db_state != DB_NOFILL) {
dbuf_unoverride(dr);
ASSERT(db->db_buf != NULL);
ASSERT(dr->dt.dl.dr_data != NULL);
if (dr->dt.dl.dr_data != db->db_buf)
arc_buf_destroy(dr->dt.dl.dr_data, db);
}
kmem_free(dr, sizeof (dbuf_dirty_record_t));
ASSERT(db->db_dirtycnt > 0);
db->db_dirtycnt -= 1;
if (zfs_refcount_remove(&db->db_holds, (void *)(uintptr_t)txg) == 0) {
ASSERT(db->db_state == DB_NOFILL || arc_released(db->db_buf));
dbuf_destroy(db);
return (B_TRUE);
}
return (B_FALSE);
}
static void
dmu_buf_will_dirty_impl(dmu_buf_t *db_fake, int flags, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
ASSERT(tx->tx_txg != 0);
ASSERT(!zfs_refcount_is_zero(&db->db_holds));
/*
* Quick check for dirtiness. For already dirty blocks, this
* reduces runtime of this function by >90%, and overall performance
* by 50% for some workloads (e.g. file deletion with indirect blocks
* cached).
*/
mutex_enter(&db->db_mtx);
if (db->db_state == DB_CACHED) {
dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, tx->tx_txg);
/*
* It's possible that it is already dirty but not cached,
* because there are some calls to dbuf_dirty() that don't
* go through dmu_buf_will_dirty().
*/
if (dr != NULL) {
/* This dbuf is already dirty and cached. */
dbuf_redirty(dr);
mutex_exit(&db->db_mtx);
return;
}
}
mutex_exit(&db->db_mtx);
DB_DNODE_ENTER(db);
if (RW_WRITE_HELD(&DB_DNODE(db)->dn_struct_rwlock))
flags |= DB_RF_HAVESTRUCT;
DB_DNODE_EXIT(db);
(void) dbuf_read(db, NULL, flags);
(void) dbuf_dirty(db, tx);
}
void
dmu_buf_will_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx)
{
dmu_buf_will_dirty_impl(db_fake,
DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH, tx);
}
boolean_t
dmu_buf_is_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dbuf_dirty_record_t *dr;
mutex_enter(&db->db_mtx);
dr = dbuf_find_dirty_eq(db, tx->tx_txg);
mutex_exit(&db->db_mtx);
return (dr != NULL);
}
void
dmu_buf_will_not_fill(dmu_buf_t *db_fake, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
db->db_state = DB_NOFILL;
DTRACE_SET_STATE(db, "allocating NOFILL buffer");
dmu_buf_will_fill(db_fake, tx);
}
void
dmu_buf_will_fill(dmu_buf_t *db_fake, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
ASSERT(tx->tx_txg != 0);
ASSERT(db->db_level == 0);
ASSERT(!zfs_refcount_is_zero(&db->db_holds));
ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT ||
dmu_tx_private_ok(tx));
dbuf_noread(db);
(void) dbuf_dirty(db, tx);
}
/*
* This function is effectively the same as dmu_buf_will_dirty(), but
* indicates the caller expects raw encrypted data in the db, and provides
* the crypt params (byteorder, salt, iv, mac) which should be stored in the
* blkptr_t when this dbuf is written. This is only used for blocks of
* dnodes, during raw receive.
*/
void
dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder,
const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dbuf_dirty_record_t *dr;
/*
* dr_has_raw_params is only processed for blocks of dnodes
* (see dbuf_sync_dnode_leaf_crypt()).
*/
ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT);
ASSERT3U(db->db_level, ==, 0);
ASSERT(db->db_objset->os_raw_receive);
dmu_buf_will_dirty_impl(db_fake,
DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH | DB_RF_NO_DECRYPT, tx);
dr = dbuf_find_dirty_eq(db, tx->tx_txg);
ASSERT3P(dr, !=, NULL);
dr->dt.dl.dr_has_raw_params = B_TRUE;
dr->dt.dl.dr_byteorder = byteorder;
bcopy(salt, dr->dt.dl.dr_salt, ZIO_DATA_SALT_LEN);
bcopy(iv, dr->dt.dl.dr_iv, ZIO_DATA_IV_LEN);
bcopy(mac, dr->dt.dl.dr_mac, ZIO_DATA_MAC_LEN);
}
static void
dbuf_override_impl(dmu_buf_impl_t *db, const blkptr_t *bp, dmu_tx_t *tx)
{
struct dirty_leaf *dl;
dbuf_dirty_record_t *dr;
dr = list_head(&db->db_dirty_records);
ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
dl = &dr->dt.dl;
dl->dr_overridden_by = *bp;
dl->dr_override_state = DR_OVERRIDDEN;
dl->dr_overridden_by.blk_birth = dr->dr_txg;
}
void
dmu_buf_fill_done(dmu_buf_t *dbuf, dmu_tx_t *tx)
{
(void) tx;
dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbuf;
dbuf_states_t old_state;
mutex_enter(&db->db_mtx);
DBUF_VERIFY(db);
old_state = db->db_state;
db->db_state = DB_CACHED;
if (old_state == DB_FILL) {
if (db->db_level == 0 && db->db_freed_in_flight) {
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
/* we were freed while filling */
/* XXX dbuf_undirty? */
bzero(db->db.db_data, db->db.db_size);
db->db_freed_in_flight = FALSE;
DTRACE_SET_STATE(db,
"fill done handling freed in flight");
} else {
DTRACE_SET_STATE(db, "fill done");
}
cv_broadcast(&db->db_changed);
}
mutex_exit(&db->db_mtx);
}
void
dmu_buf_write_embedded(dmu_buf_t *dbuf, void *data,
bp_embedded_type_t etype, enum zio_compress comp,
int uncompressed_size, int compressed_size, int byteorder,
dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbuf;
struct dirty_leaf *dl;
dmu_object_type_t type;
dbuf_dirty_record_t *dr;
if (etype == BP_EMBEDDED_TYPE_DATA) {
ASSERT(spa_feature_is_active(dmu_objset_spa(db->db_objset),
SPA_FEATURE_EMBEDDED_DATA));
}
DB_DNODE_ENTER(db);
type = DB_DNODE(db)->dn_type;
DB_DNODE_EXIT(db);
ASSERT0(db->db_level);
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
dmu_buf_will_not_fill(dbuf, tx);
dr = list_head(&db->db_dirty_records);
ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
dl = &dr->dt.dl;
encode_embedded_bp_compressed(&dl->dr_overridden_by,
data, comp, uncompressed_size, compressed_size);
BPE_SET_ETYPE(&dl->dr_overridden_by, etype);
BP_SET_TYPE(&dl->dr_overridden_by, type);
BP_SET_LEVEL(&dl->dr_overridden_by, 0);
BP_SET_BYTEORDER(&dl->dr_overridden_by, byteorder);
dl->dr_override_state = DR_OVERRIDDEN;
dl->dr_overridden_by.blk_birth = dr->dr_txg;
}
void
dmu_buf_redact(dmu_buf_t *dbuf, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbuf;
dmu_object_type_t type;
ASSERT(dsl_dataset_feature_is_active(db->db_objset->os_dsl_dataset,
SPA_FEATURE_REDACTED_DATASETS));
DB_DNODE_ENTER(db);
type = DB_DNODE(db)->dn_type;
DB_DNODE_EXIT(db);
ASSERT0(db->db_level);
dmu_buf_will_not_fill(dbuf, tx);
blkptr_t bp = { { { {0} } } };
BP_SET_TYPE(&bp, type);
BP_SET_LEVEL(&bp, 0);
BP_SET_BIRTH(&bp, tx->tx_txg, 0);
BP_SET_REDACTED(&bp);
BPE_SET_LSIZE(&bp, dbuf->db_size);
dbuf_override_impl(db, &bp, tx);
}
/*
* Directly assign a provided arc buf to a given dbuf if it's not referenced
* by anybody except our caller. Otherwise copy arcbuf's contents to dbuf.
*/
void
dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx)
{
ASSERT(!zfs_refcount_is_zero(&db->db_holds));
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
ASSERT(db->db_level == 0);
ASSERT3U(dbuf_is_metadata(db), ==, arc_is_metadata(buf));
ASSERT(buf != NULL);
ASSERT3U(arc_buf_lsize(buf), ==, db->db.db_size);
ASSERT(tx->tx_txg != 0);
arc_return_buf(buf, db);
ASSERT(arc_released(buf));
mutex_enter(&db->db_mtx);
while (db->db_state == DB_READ || db->db_state == DB_FILL)
cv_wait(&db->db_changed, &db->db_mtx);
ASSERT(db->db_state == DB_CACHED || db->db_state == DB_UNCACHED);
if (db->db_state == DB_CACHED &&
zfs_refcount_count(&db->db_holds) - 1 > db->db_dirtycnt) {
/*
* In practice, we will never have a case where we have an
* encrypted arc buffer while additional holds exist on the
* dbuf. We don't handle this here so we simply assert that
* fact instead.
*/
ASSERT(!arc_is_encrypted(buf));
mutex_exit(&db->db_mtx);
(void) dbuf_dirty(db, tx);
bcopy(buf->b_data, db->db.db_data, db->db.db_size);
arc_buf_destroy(buf, db);
return;
}
if (db->db_state == DB_CACHED) {
dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records);
ASSERT(db->db_buf != NULL);
if (dr != NULL && dr->dr_txg == tx->tx_txg) {
ASSERT(dr->dt.dl.dr_data == db->db_buf);
if (!arc_released(db->db_buf)) {
ASSERT(dr->dt.dl.dr_override_state ==
DR_OVERRIDDEN);
arc_release(db->db_buf, db);
}
dr->dt.dl.dr_data = buf;
arc_buf_destroy(db->db_buf, db);
} else if (dr == NULL || dr->dt.dl.dr_data != db->db_buf) {
arc_release(db->db_buf, db);
arc_buf_destroy(db->db_buf, db);
}
db->db_buf = NULL;
}
ASSERT(db->db_buf == NULL);
dbuf_set_data(db, buf);
db->db_state = DB_FILL;
DTRACE_SET_STATE(db, "filling assigned arcbuf");
mutex_exit(&db->db_mtx);
(void) dbuf_dirty(db, tx);
dmu_buf_fill_done(&db->db, tx);
}
void
dbuf_destroy(dmu_buf_impl_t *db)
{
dnode_t *dn;
dmu_buf_impl_t *parent = db->db_parent;
dmu_buf_impl_t *dndb;
ASSERT(MUTEX_HELD(&db->db_mtx));
ASSERT(zfs_refcount_is_zero(&db->db_holds));
if (db->db_buf != NULL) {
arc_buf_destroy(db->db_buf, db);
db->db_buf = NULL;
}
if (db->db_blkid == DMU_BONUS_BLKID) {
int slots = DB_DNODE(db)->dn_num_slots;
int bonuslen = DN_SLOTS_TO_BONUSLEN(slots);
if (db->db.db_data != NULL) {
kmem_free(db->db.db_data, bonuslen);
arc_space_return(bonuslen, ARC_SPACE_BONUS);
db->db_state = DB_UNCACHED;
DTRACE_SET_STATE(db, "buffer cleared");
}
}
dbuf_clear_data(db);
if (multilist_link_active(&db->db_cache_link)) {
ASSERT(db->db_caching_status == DB_DBUF_CACHE ||
db->db_caching_status == DB_DBUF_METADATA_CACHE);
multilist_remove(&dbuf_caches[db->db_caching_status].cache, db);
(void) zfs_refcount_remove_many(
&dbuf_caches[db->db_caching_status].size,
db->db.db_size, db);
if (db->db_caching_status == DB_DBUF_METADATA_CACHE) {
DBUF_STAT_BUMPDOWN(metadata_cache_count);
} else {
DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]);
DBUF_STAT_BUMPDOWN(cache_count);
DBUF_STAT_DECR(cache_levels_bytes[db->db_level],
db->db.db_size);
}
db->db_caching_status = DB_NO_CACHE;
}
ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL);
ASSERT(db->db_data_pending == NULL);
ASSERT(list_is_empty(&db->db_dirty_records));
db->db_state = DB_EVICTING;
DTRACE_SET_STATE(db, "buffer eviction started");
db->db_blkptr = NULL;
/*
* Now that db_state is DB_EVICTING, nobody else can find this via
* the hash table. We can now drop db_mtx, which allows us to
* acquire the dn_dbufs_mtx.
*/
mutex_exit(&db->db_mtx);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
dndb = dn->dn_dbuf;
if (db->db_blkid != DMU_BONUS_BLKID) {
boolean_t needlock = !MUTEX_HELD(&dn->dn_dbufs_mtx);
if (needlock)
mutex_enter_nested(&dn->dn_dbufs_mtx,
NESTED_SINGLE);
avl_remove(&dn->dn_dbufs, db);
membar_producer();
DB_DNODE_EXIT(db);
if (needlock)
mutex_exit(&dn->dn_dbufs_mtx);
/*
* Decrementing the dbuf count means that the hold corresponding
* to the removed dbuf is no longer discounted in dnode_move(),
* so the dnode cannot be moved until after we release the hold.
* The membar_producer() ensures visibility of the decremented
* value in dnode_move(), since DB_DNODE_EXIT doesn't actually
* release any lock.
*/
mutex_enter(&dn->dn_mtx);
dnode_rele_and_unlock(dn, db, B_TRUE);
db->db_dnode_handle = NULL;
dbuf_hash_remove(db);
} else {
DB_DNODE_EXIT(db);
}
ASSERT(zfs_refcount_is_zero(&db->db_holds));
db->db_parent = NULL;
ASSERT(db->db_buf == NULL);
ASSERT(db->db.db_data == NULL);
ASSERT(db->db_hash_next == NULL);
ASSERT(db->db_blkptr == NULL);
ASSERT(db->db_data_pending == NULL);
ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE);
ASSERT(!multilist_link_active(&db->db_cache_link));
kmem_cache_free(dbuf_kmem_cache, db);
arc_space_return(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF);
/*
* If this dbuf is referenced from an indirect dbuf,
* decrement the ref count on the indirect dbuf.
*/
if (parent && parent != dndb) {
mutex_enter(&parent->db_mtx);
dbuf_rele_and_unlock(parent, db, B_TRUE);
}
}
/*
* Note: While bpp will always be updated if the function returns success,
* parentp will not be updated if the dnode does not have dn_dbuf filled in;
* this happens when the dnode is the meta-dnode, or {user|group|project}used
* object.
*/
__attribute__((always_inline))
static inline int
dbuf_findbp(dnode_t *dn, int level, uint64_t blkid, int fail_sparse,
dmu_buf_impl_t **parentp, blkptr_t **bpp)
{
*parentp = NULL;
*bpp = NULL;
ASSERT(blkid != DMU_BONUS_BLKID);
if (blkid == DMU_SPILL_BLKID) {
mutex_enter(&dn->dn_mtx);
if (dn->dn_have_spill &&
(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR))
*bpp = DN_SPILL_BLKPTR(dn->dn_phys);
else
*bpp = NULL;
dbuf_add_ref(dn->dn_dbuf, NULL);
*parentp = dn->dn_dbuf;
mutex_exit(&dn->dn_mtx);
return (0);
}
int nlevels =
(dn->dn_phys->dn_nlevels == 0) ? 1 : dn->dn_phys->dn_nlevels;
int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
ASSERT3U(level * epbs, <, 64);
ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
/*
* This assertion shouldn't trip as long as the max indirect block size
* is less than 1M. The reason for this is that up to that point,
* the number of levels required to address an entire object with blocks
* of size SPA_MINBLOCKSIZE satisfies nlevels * epbs + 1 <= 64. In
* other words, if N * epbs + 1 > 64, then if (N-1) * epbs + 1 > 55
* (i.e. we can address the entire object), objects will all use at most
* N-1 levels and the assertion won't overflow. However, once epbs is
* 13, 4 * 13 + 1 = 53, but 5 * 13 + 1 = 66. Then, 4 levels will not be
* enough to address an entire object, so objects will have 5 levels,
* but then this assertion will overflow.
*
* All this is to say that if we ever increase DN_MAX_INDBLKSHIFT, we
* need to redo this logic to handle overflows.
*/
ASSERT(level >= nlevels ||
((nlevels - level - 1) * epbs) +
highbit64(dn->dn_phys->dn_nblkptr) <= 64);
if (level >= nlevels ||
blkid >= ((uint64_t)dn->dn_phys->dn_nblkptr <<
((nlevels - level - 1) * epbs)) ||
(fail_sparse &&
blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))) {
/* the buffer has no parent yet */
return (SET_ERROR(ENOENT));
} else if (level < nlevels-1) {
/* this block is referenced from an indirect block */
int err;
err = dbuf_hold_impl(dn, level + 1,
blkid >> epbs, fail_sparse, FALSE, NULL, parentp);
if (err)
return (err);
err = dbuf_read(*parentp, NULL,
(DB_RF_HAVESTRUCT | DB_RF_NOPREFETCH | DB_RF_CANFAIL));
if (err) {
dbuf_rele(*parentp, NULL);
*parentp = NULL;
return (err);
}
rw_enter(&(*parentp)->db_rwlock, RW_READER);
*bpp = ((blkptr_t *)(*parentp)->db.db_data) +
(blkid & ((1ULL << epbs) - 1));
if (blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))
ASSERT(BP_IS_HOLE(*bpp));
rw_exit(&(*parentp)->db_rwlock);
return (0);
} else {
/* the block is referenced from the dnode */
ASSERT3U(level, ==, nlevels-1);
ASSERT(dn->dn_phys->dn_nblkptr == 0 ||
blkid < dn->dn_phys->dn_nblkptr);
if (dn->dn_dbuf) {
dbuf_add_ref(dn->dn_dbuf, NULL);
*parentp = dn->dn_dbuf;
}
*bpp = &dn->dn_phys->dn_blkptr[blkid];
return (0);
}
}
static dmu_buf_impl_t *
dbuf_create(dnode_t *dn, uint8_t level, uint64_t blkid,
dmu_buf_impl_t *parent, blkptr_t *blkptr)
{
objset_t *os = dn->dn_objset;
dmu_buf_impl_t *db, *odb;
ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
ASSERT(dn->dn_type != DMU_OT_NONE);
db = kmem_cache_alloc(dbuf_kmem_cache, KM_SLEEP);
list_create(&db->db_dirty_records, sizeof (dbuf_dirty_record_t),
offsetof(dbuf_dirty_record_t, dr_dbuf_node));
db->db_objset = os;
db->db.db_object = dn->dn_object;
db->db_level = level;
db->db_blkid = blkid;
db->db_dirtycnt = 0;
db->db_dnode_handle = dn->dn_handle;
db->db_parent = parent;
db->db_blkptr = blkptr;
db->db_user = NULL;
db->db_user_immediate_evict = FALSE;
db->db_freed_in_flight = FALSE;
db->db_pending_evict = FALSE;
if (blkid == DMU_BONUS_BLKID) {
ASSERT3P(parent, ==, dn->dn_dbuf);
db->db.db_size = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots) -
(dn->dn_nblkptr-1) * sizeof (blkptr_t);
ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen);
db->db.db_offset = DMU_BONUS_BLKID;
db->db_state = DB_UNCACHED;
DTRACE_SET_STATE(db, "bonus buffer created");
db->db_caching_status = DB_NO_CACHE;
/* the bonus dbuf is not placed in the hash table */
arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF);
return (db);
} else if (blkid == DMU_SPILL_BLKID) {
db->db.db_size = (blkptr != NULL) ?
BP_GET_LSIZE(blkptr) : SPA_MINBLOCKSIZE;
db->db.db_offset = 0;
} else {
int blocksize =
db->db_level ? 1 << dn->dn_indblkshift : dn->dn_datablksz;
db->db.db_size = blocksize;
db->db.db_offset = db->db_blkid * blocksize;
}
/*
* Hold the dn_dbufs_mtx while we get the new dbuf
* in the hash table *and* added to the dbufs list.
* This prevents a possible deadlock with someone
* trying to look up this dbuf before it's added to the
* dn_dbufs list.
*/
mutex_enter(&dn->dn_dbufs_mtx);
db->db_state = DB_EVICTING; /* not worth logging this state change */
if ((odb = dbuf_hash_insert(db)) != NULL) {
/* someone else inserted it first */
mutex_exit(&dn->dn_dbufs_mtx);
kmem_cache_free(dbuf_kmem_cache, db);
DBUF_STAT_BUMP(hash_insert_race);
return (odb);
}
avl_add(&dn->dn_dbufs, db);
db->db_state = DB_UNCACHED;
DTRACE_SET_STATE(db, "regular buffer created");
db->db_caching_status = DB_NO_CACHE;
mutex_exit(&dn->dn_dbufs_mtx);
arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF);
if (parent && parent != dn->dn_dbuf)
dbuf_add_ref(parent, db);
ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT ||
zfs_refcount_count(&dn->dn_holds) > 0);
(void) zfs_refcount_add(&dn->dn_holds, db);
dprintf_dbuf(db, "db=%p\n", db);
return (db);
}
/*
* This function returns a block pointer and information about the object,
* given a dnode and a block. This is a publicly accessible version of
* dbuf_findbp that only returns some information, rather than the
* dbuf. Note that the dnode passed in must be held, and the dn_struct_rwlock
* should be locked as (at least) a reader.
*/
int
dbuf_dnode_findbp(dnode_t *dn, uint64_t level, uint64_t blkid,
blkptr_t *bp, uint16_t *datablkszsec, uint8_t *indblkshift)
{
dmu_buf_impl_t *dbp = NULL;
blkptr_t *bp2;
int err = 0;
ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
err = dbuf_findbp(dn, level, blkid, B_FALSE, &dbp, &bp2);
if (err == 0) {
*bp = *bp2;
if (dbp != NULL)
dbuf_rele(dbp, NULL);
if (datablkszsec != NULL)
*datablkszsec = dn->dn_phys->dn_datablkszsec;
if (indblkshift != NULL)
*indblkshift = dn->dn_phys->dn_indblkshift;
}
return (err);
}
typedef struct dbuf_prefetch_arg {
spa_t *dpa_spa; /* The spa to issue the prefetch in. */
zbookmark_phys_t dpa_zb; /* The target block to prefetch. */
int dpa_epbs; /* Entries (blkptr_t's) Per Block Shift. */
int dpa_curlevel; /* The current level that we're reading */
dnode_t *dpa_dnode; /* The dnode associated with the prefetch */
zio_priority_t dpa_prio; /* The priority I/Os should be issued at. */
zio_t *dpa_zio; /* The parent zio_t for all prefetches. */
arc_flags_t dpa_aflags; /* Flags to pass to the final prefetch. */
dbuf_prefetch_fn dpa_cb; /* prefetch completion callback */
void *dpa_arg; /* prefetch completion arg */
} dbuf_prefetch_arg_t;
static void
dbuf_prefetch_fini(dbuf_prefetch_arg_t *dpa, boolean_t io_done)
{
if (dpa->dpa_cb != NULL)
dpa->dpa_cb(dpa->dpa_arg, io_done);
kmem_free(dpa, sizeof (*dpa));
}
static void
dbuf_issue_final_prefetch_done(zio_t *zio, const zbookmark_phys_t *zb,
const blkptr_t *iobp, arc_buf_t *abuf, void *private)
{
(void) zio, (void) zb, (void) iobp;
dbuf_prefetch_arg_t *dpa = private;
dbuf_prefetch_fini(dpa, B_TRUE);
if (abuf != NULL)
arc_buf_destroy(abuf, private);
}
/*
* Actually issue the prefetch read for the block given.
*/
static void
dbuf_issue_final_prefetch(dbuf_prefetch_arg_t *dpa, blkptr_t *bp)
{
ASSERT(!BP_IS_REDACTED(bp) ||
dsl_dataset_feature_is_active(
dpa->dpa_dnode->dn_objset->os_dsl_dataset,
SPA_FEATURE_REDACTED_DATASETS));
if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp) || BP_IS_REDACTED(bp))
return (dbuf_prefetch_fini(dpa, B_FALSE));
int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE;
arc_flags_t aflags =
dpa->dpa_aflags | ARC_FLAG_NOWAIT | ARC_FLAG_PREFETCH |
ARC_FLAG_NO_BUF;
/* dnodes are always read as raw and then converted later */
if (BP_GET_TYPE(bp) == DMU_OT_DNODE && BP_IS_PROTECTED(bp) &&
dpa->dpa_curlevel == 0)
zio_flags |= ZIO_FLAG_RAW;
ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp));
ASSERT3U(dpa->dpa_curlevel, ==, dpa->dpa_zb.zb_level);
ASSERT(dpa->dpa_zio != NULL);
(void) arc_read(dpa->dpa_zio, dpa->dpa_spa, bp,
dbuf_issue_final_prefetch_done, dpa,
dpa->dpa_prio, zio_flags, &aflags, &dpa->dpa_zb);
}
/*
* Called when an indirect block above our prefetch target is read in. This
* will either read in the next indirect block down the tree or issue the actual
* prefetch if the next block down is our target.
*/
static void
dbuf_prefetch_indirect_done(zio_t *zio, const zbookmark_phys_t *zb,
const blkptr_t *iobp, arc_buf_t *abuf, void *private)
{
(void) zb, (void) iobp;
dbuf_prefetch_arg_t *dpa = private;
ASSERT3S(dpa->dpa_zb.zb_level, <, dpa->dpa_curlevel);
ASSERT3S(dpa->dpa_curlevel, >, 0);
if (abuf == NULL) {
ASSERT(zio == NULL || zio->io_error != 0);
return (dbuf_prefetch_fini(dpa, B_TRUE));
}
ASSERT(zio == NULL || zio->io_error == 0);
/*
* The dpa_dnode is only valid if we are called with a NULL
* zio. This indicates that the arc_read() returned without
* first calling zio_read() to issue a physical read. Once
* a physical read is made the dpa_dnode must be invalidated
* as the locks guarding it may have been dropped. If the
* dpa_dnode is still valid, then we want to add it to the dbuf
* cache. To do so, we must hold the dbuf associated with the block
* we just prefetched, read its contents so that we associate it
* with an arc_buf_t, and then release it.
*/
if (zio != NULL) {
ASSERT3S(BP_GET_LEVEL(zio->io_bp), ==, dpa->dpa_curlevel);
if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS) {
ASSERT3U(BP_GET_PSIZE(zio->io_bp), ==, zio->io_size);
} else {
ASSERT3U(BP_GET_LSIZE(zio->io_bp), ==, zio->io_size);
}
ASSERT3P(zio->io_spa, ==, dpa->dpa_spa);
dpa->dpa_dnode = NULL;
} else if (dpa->dpa_dnode != NULL) {
uint64_t curblkid = dpa->dpa_zb.zb_blkid >>
(dpa->dpa_epbs * (dpa->dpa_curlevel -
dpa->dpa_zb.zb_level));
dmu_buf_impl_t *db = dbuf_hold_level(dpa->dpa_dnode,
dpa->dpa_curlevel, curblkid, FTAG);
if (db == NULL) {
arc_buf_destroy(abuf, private);
return (dbuf_prefetch_fini(dpa, B_TRUE));
}
(void) dbuf_read(db, NULL,
DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH | DB_RF_HAVESTRUCT);
dbuf_rele(db, FTAG);
}
dpa->dpa_curlevel--;
uint64_t nextblkid = dpa->dpa_zb.zb_blkid >>
(dpa->dpa_epbs * (dpa->dpa_curlevel - dpa->dpa_zb.zb_level));
blkptr_t *bp = ((blkptr_t *)abuf->b_data) +
P2PHASE(nextblkid, 1ULL << dpa->dpa_epbs);
ASSERT(!BP_IS_REDACTED(bp) ||
dsl_dataset_feature_is_active(
dpa->dpa_dnode->dn_objset->os_dsl_dataset,
SPA_FEATURE_REDACTED_DATASETS));
if (BP_IS_HOLE(bp) || BP_IS_REDACTED(bp)) {
dbuf_prefetch_fini(dpa, B_TRUE);
} else if (dpa->dpa_curlevel == dpa->dpa_zb.zb_level) {
ASSERT3U(nextblkid, ==, dpa->dpa_zb.zb_blkid);
dbuf_issue_final_prefetch(dpa, bp);
} else {
arc_flags_t iter_aflags = ARC_FLAG_NOWAIT;
zbookmark_phys_t zb;
/* flag if L2ARC eligible, l2arc_noprefetch then decides */
if (dpa->dpa_aflags & ARC_FLAG_L2CACHE)
iter_aflags |= ARC_FLAG_L2CACHE;
ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp));
SET_BOOKMARK(&zb, dpa->dpa_zb.zb_objset,
dpa->dpa_zb.zb_object, dpa->dpa_curlevel, nextblkid);
(void) arc_read(dpa->dpa_zio, dpa->dpa_spa,
bp, dbuf_prefetch_indirect_done, dpa, dpa->dpa_prio,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
&iter_aflags, &zb);
}
arc_buf_destroy(abuf, private);
}
/*
* Issue prefetch reads for the given block on the given level. If the indirect
* blocks above that block are not in memory, we will read them in
* asynchronously. As a result, this call never blocks waiting for a read to
* complete. Note that the prefetch might fail if the dataset is encrypted and
* the encryption key is unmapped before the IO completes.
*/
int
dbuf_prefetch_impl(dnode_t *dn, int64_t level, uint64_t blkid,
zio_priority_t prio, arc_flags_t aflags, dbuf_prefetch_fn cb,
void *arg)
{
blkptr_t bp;
int epbs, nlevels, curlevel;
uint64_t curblkid;
ASSERT(blkid != DMU_BONUS_BLKID);
ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
if (blkid > dn->dn_maxblkid)
goto no_issue;
if (level == 0 && dnode_block_freed(dn, blkid))
goto no_issue;
/*
* This dnode hasn't been written to disk yet, so there's nothing to
* prefetch.
*/
nlevels = dn->dn_phys->dn_nlevels;
if (level >= nlevels || dn->dn_phys->dn_nblkptr == 0)
goto no_issue;
epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
if (dn->dn_phys->dn_maxblkid < blkid << (epbs * level))
goto no_issue;
dmu_buf_impl_t *db = dbuf_find(dn->dn_objset, dn->dn_object,
level, blkid);
if (db != NULL) {
mutex_exit(&db->db_mtx);
/*
* This dbuf already exists. It is either CACHED, or
* (we assume) about to be read or filled.
*/
goto no_issue;
}
/*
* Find the closest ancestor (indirect block) of the target block
* that is present in the cache. In this indirect block, we will
* find the bp that is at curlevel, curblkid.
*/
curlevel = level;
curblkid = blkid;
while (curlevel < nlevels - 1) {
int parent_level = curlevel + 1;
uint64_t parent_blkid = curblkid >> epbs;
dmu_buf_impl_t *db;
if (dbuf_hold_impl(dn, parent_level, parent_blkid,
FALSE, TRUE, FTAG, &db) == 0) {
blkptr_t *bpp = db->db_buf->b_data;
bp = bpp[P2PHASE(curblkid, 1 << epbs)];
dbuf_rele(db, FTAG);
break;
}
curlevel = parent_level;
curblkid = parent_blkid;
}
if (curlevel == nlevels - 1) {
/* No cached indirect blocks found. */
ASSERT3U(curblkid, <, dn->dn_phys->dn_nblkptr);
bp = dn->dn_phys->dn_blkptr[curblkid];
}
ASSERT(!BP_IS_REDACTED(&bp) ||
dsl_dataset_feature_is_active(dn->dn_objset->os_dsl_dataset,
SPA_FEATURE_REDACTED_DATASETS));
if (BP_IS_HOLE(&bp) || BP_IS_REDACTED(&bp))
goto no_issue;
ASSERT3U(curlevel, ==, BP_GET_LEVEL(&bp));
zio_t *pio = zio_root(dmu_objset_spa(dn->dn_objset), NULL, NULL,
ZIO_FLAG_CANFAIL);
dbuf_prefetch_arg_t *dpa = kmem_zalloc(sizeof (*dpa), KM_SLEEP);
dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset;
SET_BOOKMARK(&dpa->dpa_zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET,
dn->dn_object, level, blkid);
dpa->dpa_curlevel = curlevel;
dpa->dpa_prio = prio;
dpa->dpa_aflags = aflags;
dpa->dpa_spa = dn->dn_objset->os_spa;
dpa->dpa_dnode = dn;
dpa->dpa_epbs = epbs;
dpa->dpa_zio = pio;
dpa->dpa_cb = cb;
dpa->dpa_arg = arg;
/* flag if L2ARC eligible, l2arc_noprefetch then decides */
if (dnode_level_is_l2cacheable(&bp, dn, level))
dpa->dpa_aflags |= ARC_FLAG_L2CACHE;
/*
* If we have the indirect just above us, no need to do the asynchronous
* prefetch chain; we'll just run the last step ourselves. If we're at
* a higher level, though, we want to issue the prefetches for all the
* indirect blocks asynchronously, so we can go on with whatever we were
* doing.
*/
if (curlevel == level) {
ASSERT3U(curblkid, ==, blkid);
dbuf_issue_final_prefetch(dpa, &bp);
} else {
arc_flags_t iter_aflags = ARC_FLAG_NOWAIT;
zbookmark_phys_t zb;
/* flag if L2ARC eligible, l2arc_noprefetch then decides */
if (dnode_level_is_l2cacheable(&bp, dn, level))
iter_aflags |= ARC_FLAG_L2CACHE;
SET_BOOKMARK(&zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET,
dn->dn_object, curlevel, curblkid);
(void) arc_read(dpa->dpa_zio, dpa->dpa_spa,
&bp, dbuf_prefetch_indirect_done, dpa, prio,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
&iter_aflags, &zb);
}
/*
* We use pio here instead of dpa_zio since it's possible that
* dpa may have already been freed.
*/
zio_nowait(pio);
return (1);
no_issue:
if (cb != NULL)
cb(arg, B_FALSE);
return (0);
}
int
dbuf_prefetch(dnode_t *dn, int64_t level, uint64_t blkid, zio_priority_t prio,
arc_flags_t aflags)
{
return (dbuf_prefetch_impl(dn, level, blkid, prio, aflags, NULL, NULL));
}
/*
* Helper function for dbuf_hold_impl() to copy a buffer. Handles
* the case of encrypted, compressed and uncompressed buffers by
* allocating the new buffer, respectively, with arc_alloc_raw_buf(),
* arc_alloc_compressed_buf() or arc_alloc_buf().*
*
* NOTE: Declared noinline to avoid stack bloat in dbuf_hold_impl().
*/
noinline static void
dbuf_hold_copy(dnode_t *dn, dmu_buf_impl_t *db)
{
dbuf_dirty_record_t *dr = db->db_data_pending;
arc_buf_t *data = dr->dt.dl.dr_data;
enum zio_compress compress_type = arc_get_compression(data);
uint8_t complevel = arc_get_complevel(data);
if (arc_is_encrypted(data)) {
boolean_t byteorder;
uint8_t salt[ZIO_DATA_SALT_LEN];
uint8_t iv[ZIO_DATA_IV_LEN];
uint8_t mac[ZIO_DATA_MAC_LEN];
arc_get_raw_params(data, &byteorder, salt, iv, mac);
dbuf_set_data(db, arc_alloc_raw_buf(dn->dn_objset->os_spa, db,
dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac,
dn->dn_type, arc_buf_size(data), arc_buf_lsize(data),
compress_type, complevel));
} else if (compress_type != ZIO_COMPRESS_OFF) {
dbuf_set_data(db, arc_alloc_compressed_buf(
dn->dn_objset->os_spa, db, arc_buf_size(data),
arc_buf_lsize(data), compress_type, complevel));
} else {
dbuf_set_data(db, arc_alloc_buf(dn->dn_objset->os_spa, db,
DBUF_GET_BUFC_TYPE(db), db->db.db_size));
}
rw_enter(&db->db_rwlock, RW_WRITER);
bcopy(data->b_data, db->db.db_data, arc_buf_size(data));
rw_exit(&db->db_rwlock);
}
/*
* Returns with db_holds incremented, and db_mtx not held.
* Note: dn_struct_rwlock must be held.
*/
int
dbuf_hold_impl(dnode_t *dn, uint8_t level, uint64_t blkid,
boolean_t fail_sparse, boolean_t fail_uncached,
void *tag, dmu_buf_impl_t **dbp)
{
dmu_buf_impl_t *db, *parent = NULL;
/* If the pool has been created, verify the tx_sync_lock is not held */
spa_t *spa = dn->dn_objset->os_spa;
dsl_pool_t *dp = spa->spa_dsl_pool;
if (dp != NULL) {
ASSERT(!MUTEX_HELD(&dp->dp_tx.tx_sync_lock));
}
ASSERT(blkid != DMU_BONUS_BLKID);
ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
ASSERT3U(dn->dn_nlevels, >, level);
*dbp = NULL;
/* dbuf_find() returns with db_mtx held */
db = dbuf_find(dn->dn_objset, dn->dn_object, level, blkid);
if (db == NULL) {
blkptr_t *bp = NULL;
int err;
if (fail_uncached)
return (SET_ERROR(ENOENT));
ASSERT3P(parent, ==, NULL);
err = dbuf_findbp(dn, level, blkid, fail_sparse, &parent, &bp);
if (fail_sparse) {
if (err == 0 && bp && BP_IS_HOLE(bp))
err = SET_ERROR(ENOENT);
if (err) {
if (parent)
dbuf_rele(parent, NULL);
return (err);
}
}
if (err && err != ENOENT)
return (err);
db = dbuf_create(dn, level, blkid, parent, bp);
}
if (fail_uncached && db->db_state != DB_CACHED) {
mutex_exit(&db->db_mtx);
return (SET_ERROR(ENOENT));
}
if (db->db_buf != NULL) {
arc_buf_access(db->db_buf);
ASSERT3P(db->db.db_data, ==, db->db_buf->b_data);
}
ASSERT(db->db_buf == NULL || arc_referenced(db->db_buf));
/*
* If this buffer is currently syncing out, and we are
* still referencing it from db_data, we need to make a copy
* of it in case we decide we want to dirty it again in this txg.
*/
if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
dn->dn_object != DMU_META_DNODE_OBJECT &&
db->db_state == DB_CACHED && db->db_data_pending) {
dbuf_dirty_record_t *dr = db->db_data_pending;
if (dr->dt.dl.dr_data == db->db_buf)
dbuf_hold_copy(dn, db);
}
if (multilist_link_active(&db->db_cache_link)) {
ASSERT(zfs_refcount_is_zero(&db->db_holds));
ASSERT(db->db_caching_status == DB_DBUF_CACHE ||
db->db_caching_status == DB_DBUF_METADATA_CACHE);
multilist_remove(&dbuf_caches[db->db_caching_status].cache, db);
(void) zfs_refcount_remove_many(
&dbuf_caches[db->db_caching_status].size,
db->db.db_size, db);
if (db->db_caching_status == DB_DBUF_METADATA_CACHE) {
DBUF_STAT_BUMPDOWN(metadata_cache_count);
} else {
DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]);
DBUF_STAT_BUMPDOWN(cache_count);
DBUF_STAT_DECR(cache_levels_bytes[db->db_level],
db->db.db_size);
}
db->db_caching_status = DB_NO_CACHE;
}
(void) zfs_refcount_add(&db->db_holds, tag);
DBUF_VERIFY(db);
mutex_exit(&db->db_mtx);
/* NOTE: we can't rele the parent until after we drop the db_mtx */
if (parent)
dbuf_rele(parent, NULL);
ASSERT3P(DB_DNODE(db), ==, dn);
ASSERT3U(db->db_blkid, ==, blkid);
ASSERT3U(db->db_level, ==, level);
*dbp = db;
return (0);
}
dmu_buf_impl_t *
dbuf_hold(dnode_t *dn, uint64_t blkid, void *tag)
{
return (dbuf_hold_level(dn, 0, blkid, tag));
}
dmu_buf_impl_t *
dbuf_hold_level(dnode_t *dn, int level, uint64_t blkid, void *tag)
{
dmu_buf_impl_t *db;
int err = dbuf_hold_impl(dn, level, blkid, FALSE, FALSE, tag, &db);
return (err ? NULL : db);
}
void
dbuf_create_bonus(dnode_t *dn)
{
ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock));
ASSERT(dn->dn_bonus == NULL);
dn->dn_bonus = dbuf_create(dn, 0, DMU_BONUS_BLKID, dn->dn_dbuf, NULL);
}
int
dbuf_spill_set_blksz(dmu_buf_t *db_fake, uint64_t blksz, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
if (db->db_blkid != DMU_SPILL_BLKID)
return (SET_ERROR(ENOTSUP));
if (blksz == 0)
blksz = SPA_MINBLOCKSIZE;
ASSERT3U(blksz, <=, spa_maxblocksize(dmu_objset_spa(db->db_objset)));
blksz = P2ROUNDUP(blksz, SPA_MINBLOCKSIZE);
dbuf_new_size(db, blksz, tx);
return (0);
}
void
dbuf_rm_spill(dnode_t *dn, dmu_tx_t *tx)
{
dbuf_free_range(dn, DMU_SPILL_BLKID, DMU_SPILL_BLKID, tx);
}
#pragma weak dmu_buf_add_ref = dbuf_add_ref
void
dbuf_add_ref(dmu_buf_impl_t *db, void *tag)
{
int64_t holds = zfs_refcount_add(&db->db_holds, tag);
VERIFY3S(holds, >, 1);
}
#pragma weak dmu_buf_try_add_ref = dbuf_try_add_ref
boolean_t
dbuf_try_add_ref(dmu_buf_t *db_fake, objset_t *os, uint64_t obj, uint64_t blkid,
void *tag)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dmu_buf_impl_t *found_db;
boolean_t result = B_FALSE;
if (blkid == DMU_BONUS_BLKID)
found_db = dbuf_find_bonus(os, obj);
else
found_db = dbuf_find(os, obj, 0, blkid);
if (found_db != NULL) {
if (db == found_db && dbuf_refcount(db) > db->db_dirtycnt) {
(void) zfs_refcount_add(&db->db_holds, tag);
result = B_TRUE;
}
mutex_exit(&found_db->db_mtx);
}
return (result);
}
/*
* If you call dbuf_rele() you had better not be referencing the dnode handle
* unless you have some other direct or indirect hold on the dnode. (An indirect
* hold is a hold on one of the dnode's dbufs, including the bonus buffer.)
* Without that, the dbuf_rele() could lead to a dnode_rele() followed by the
* dnode's parent dbuf evicting its dnode handles.
*/
void
dbuf_rele(dmu_buf_impl_t *db, void *tag)
{
mutex_enter(&db->db_mtx);
dbuf_rele_and_unlock(db, tag, B_FALSE);
}
void
dmu_buf_rele(dmu_buf_t *db, void *tag)
{
dbuf_rele((dmu_buf_impl_t *)db, tag);
}
/*
* dbuf_rele() for an already-locked dbuf. This is necessary to allow
* db_dirtycnt and db_holds to be updated atomically. The 'evicting'
* argument should be set if we are already in the dbuf-evicting code
* path, in which case we don't want to recursively evict. This allows us to
* avoid deeply nested stacks that would have a call flow similar to this:
*
* dbuf_rele()-->dbuf_rele_and_unlock()-->dbuf_evict_notify()
* ^ |
* | |
* +-----dbuf_destroy()<--dbuf_evict_one()<--------+
*
*/
void
dbuf_rele_and_unlock(dmu_buf_impl_t *db, void *tag, boolean_t evicting)
{
int64_t holds;
uint64_t size;
ASSERT(MUTEX_HELD(&db->db_mtx));
DBUF_VERIFY(db);
/*
* Remove the reference to the dbuf before removing its hold on the
* dnode so we can guarantee in dnode_move() that a referenced bonus
* buffer has a corresponding dnode hold.
*/
holds = zfs_refcount_remove(&db->db_holds, tag);
ASSERT(holds >= 0);
/*
* We can't freeze indirects if there is a possibility that they
* may be modified in the current syncing context.
*/
if (db->db_buf != NULL &&
holds == (db->db_level == 0 ? db->db_dirtycnt : 0)) {
arc_buf_freeze(db->db_buf);
}
if (holds == db->db_dirtycnt &&
db->db_level == 0 && db->db_user_immediate_evict)
dbuf_evict_user(db);
if (holds == 0) {
if (db->db_blkid == DMU_BONUS_BLKID) {
dnode_t *dn;
boolean_t evict_dbuf = db->db_pending_evict;
/*
* If the dnode moves here, we cannot cross this
* barrier until the move completes.
*/
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
atomic_dec_32(&dn->dn_dbufs_count);
/*
* Decrementing the dbuf count means that the bonus
* buffer's dnode hold is no longer discounted in
* dnode_move(). The dnode cannot move until after
* the dnode_rele() below.
*/
DB_DNODE_EXIT(db);
/*
* Do not reference db after its lock is dropped.
* Another thread may evict it.
*/
mutex_exit(&db->db_mtx);
if (evict_dbuf)
dnode_evict_bonus(dn);
dnode_rele(dn, db);
} else if (db->db_buf == NULL) {
/*
* This is a special case: we never associated this
* dbuf with any data allocated from the ARC.
*/
ASSERT(db->db_state == DB_UNCACHED ||
db->db_state == DB_NOFILL);
dbuf_destroy(db);
} else if (arc_released(db->db_buf)) {
/*
* This dbuf has anonymous data associated with it.
*/
dbuf_destroy(db);
} else {
boolean_t do_arc_evict = B_FALSE;
blkptr_t bp;
spa_t *spa = dmu_objset_spa(db->db_objset);
if (!DBUF_IS_CACHEABLE(db) &&
db->db_blkptr != NULL &&
!BP_IS_HOLE(db->db_blkptr) &&
!BP_IS_EMBEDDED(db->db_blkptr)) {
do_arc_evict = B_TRUE;
bp = *db->db_blkptr;
}
if (!DBUF_IS_CACHEABLE(db) ||
db->db_pending_evict) {
dbuf_destroy(db);
} else if (!multilist_link_active(&db->db_cache_link)) {
ASSERT3U(db->db_caching_status, ==,
DB_NO_CACHE);
dbuf_cached_state_t dcs =
dbuf_include_in_metadata_cache(db) ?
DB_DBUF_METADATA_CACHE : DB_DBUF_CACHE;
db->db_caching_status = dcs;
multilist_insert(&dbuf_caches[dcs].cache, db);
uint64_t db_size = db->db.db_size;
size = zfs_refcount_add_many(
&dbuf_caches[dcs].size, db_size, db);
uint8_t db_level = db->db_level;
mutex_exit(&db->db_mtx);
if (dcs == DB_DBUF_METADATA_CACHE) {
DBUF_STAT_BUMP(metadata_cache_count);
DBUF_STAT_MAX(
metadata_cache_size_bytes_max,
size);
} else {
DBUF_STAT_BUMP(cache_count);
DBUF_STAT_MAX(cache_size_bytes_max,
size);
DBUF_STAT_BUMP(cache_levels[db_level]);
DBUF_STAT_INCR(
cache_levels_bytes[db_level],
db_size);
}
if (dcs == DB_DBUF_CACHE && !evicting)
dbuf_evict_notify(size);
}
if (do_arc_evict)
arc_freed(spa, &bp);
}
} else {
mutex_exit(&db->db_mtx);
}
}
#pragma weak dmu_buf_refcount = dbuf_refcount
uint64_t
dbuf_refcount(dmu_buf_impl_t *db)
{
return (zfs_refcount_count(&db->db_holds));
}
uint64_t
dmu_buf_user_refcount(dmu_buf_t *db_fake)
{
uint64_t holds;
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
mutex_enter(&db->db_mtx);
ASSERT3U(zfs_refcount_count(&db->db_holds), >=, db->db_dirtycnt);
holds = zfs_refcount_count(&db->db_holds) - db->db_dirtycnt;
mutex_exit(&db->db_mtx);
return (holds);
}
void *
dmu_buf_replace_user(dmu_buf_t *db_fake, dmu_buf_user_t *old_user,
dmu_buf_user_t *new_user)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
mutex_enter(&db->db_mtx);
dbuf_verify_user(db, DBVU_NOT_EVICTING);
if (db->db_user == old_user)
db->db_user = new_user;
else
old_user = db->db_user;
dbuf_verify_user(db, DBVU_NOT_EVICTING);
mutex_exit(&db->db_mtx);
return (old_user);
}
void *
dmu_buf_set_user(dmu_buf_t *db_fake, dmu_buf_user_t *user)
{
return (dmu_buf_replace_user(db_fake, NULL, user));
}
void *
dmu_buf_set_user_ie(dmu_buf_t *db_fake, dmu_buf_user_t *user)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
db->db_user_immediate_evict = TRUE;
return (dmu_buf_set_user(db_fake, user));
}
void *
dmu_buf_remove_user(dmu_buf_t *db_fake, dmu_buf_user_t *user)
{
return (dmu_buf_replace_user(db_fake, user, NULL));
}
void *
dmu_buf_get_user(dmu_buf_t *db_fake)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dbuf_verify_user(db, DBVU_NOT_EVICTING);
return (db->db_user);
}
void
dmu_buf_user_evict_wait()
{
taskq_wait(dbu_evict_taskq);
}
blkptr_t *
dmu_buf_get_blkptr(dmu_buf_t *db)
{
dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db;
return (dbi->db_blkptr);
}
objset_t *
dmu_buf_get_objset(dmu_buf_t *db)
{
dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db;
return (dbi->db_objset);
}
dnode_t *
dmu_buf_dnode_enter(dmu_buf_t *db)
{
dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db;
DB_DNODE_ENTER(dbi);
return (DB_DNODE(dbi));
}
void
dmu_buf_dnode_exit(dmu_buf_t *db)
{
dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db;
DB_DNODE_EXIT(dbi);
}
static void
dbuf_check_blkptr(dnode_t *dn, dmu_buf_impl_t *db)
{
/* ASSERT(dmu_tx_is_syncing(tx) */
ASSERT(MUTEX_HELD(&db->db_mtx));
if (db->db_blkptr != NULL)
return;
if (db->db_blkid == DMU_SPILL_BLKID) {
db->db_blkptr = DN_SPILL_BLKPTR(dn->dn_phys);
BP_ZERO(db->db_blkptr);
return;
}
if (db->db_level == dn->dn_phys->dn_nlevels-1) {
/*
* This buffer was allocated at a time when there was
* no available blkptrs from the dnode, or it was
* inappropriate to hook it in (i.e., nlevels mismatch).
*/
ASSERT(db->db_blkid < dn->dn_phys->dn_nblkptr);
ASSERT(db->db_parent == NULL);
db->db_parent = dn->dn_dbuf;
db->db_blkptr = &dn->dn_phys->dn_blkptr[db->db_blkid];
DBUF_VERIFY(db);
} else {
dmu_buf_impl_t *parent = db->db_parent;
int epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
ASSERT(dn->dn_phys->dn_nlevels > 1);
if (parent == NULL) {
mutex_exit(&db->db_mtx);
rw_enter(&dn->dn_struct_rwlock, RW_READER);
parent = dbuf_hold_level(dn, db->db_level + 1,
db->db_blkid >> epbs, db);
rw_exit(&dn->dn_struct_rwlock);
mutex_enter(&db->db_mtx);
db->db_parent = parent;
}
db->db_blkptr = (blkptr_t *)parent->db.db_data +
(db->db_blkid & ((1ULL << epbs) - 1));
DBUF_VERIFY(db);
}
}
static void
dbuf_sync_bonus(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = dr->dr_dbuf;
void *data = dr->dt.dl.dr_data;
ASSERT0(db->db_level);
ASSERT(MUTEX_HELD(&db->db_mtx));
ASSERT(db->db_blkid == DMU_BONUS_BLKID);
ASSERT(data != NULL);
dnode_t *dn = dr->dr_dnode;
ASSERT3U(DN_MAX_BONUS_LEN(dn->dn_phys), <=,
DN_SLOTS_TO_BONUSLEN(dn->dn_phys->dn_extra_slots + 1));
bcopy(data, DN_BONUS(dn->dn_phys), DN_MAX_BONUS_LEN(dn->dn_phys));
dbuf_sync_leaf_verify_bonus_dnode(dr);
dbuf_undirty_bonus(dr);
dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg, B_FALSE);
}
/*
* When syncing out a blocks of dnodes, adjust the block to deal with
* encryption. Normally, we make sure the block is decrypted before writing
* it. If we have crypt params, then we are writing a raw (encrypted) block,
* from a raw receive. In this case, set the ARC buf's crypt params so
* that the BP will be filled with the correct byteorder, salt, iv, and mac.
*/
static void
dbuf_prepare_encrypted_dnode_leaf(dbuf_dirty_record_t *dr)
{
int err;
dmu_buf_impl_t *db = dr->dr_dbuf;
ASSERT(MUTEX_HELD(&db->db_mtx));
ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT);
ASSERT3U(db->db_level, ==, 0);
if (!db->db_objset->os_raw_receive && arc_is_encrypted(db->db_buf)) {
zbookmark_phys_t zb;
/*
* Unfortunately, there is currently no mechanism for
* syncing context to handle decryption errors. An error
* here is only possible if an attacker maliciously
* changed a dnode block and updated the associated
* checksums going up the block tree.
*/
SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset),
db->db.db_object, db->db_level, db->db_blkid);
err = arc_untransform(db->db_buf, db->db_objset->os_spa,
&zb, B_TRUE);
if (err)
panic("Invalid dnode block MAC");
} else if (dr->dt.dl.dr_has_raw_params) {
(void) arc_release(dr->dt.dl.dr_data, db);
arc_convert_to_raw(dr->dt.dl.dr_data,
dmu_objset_id(db->db_objset),
dr->dt.dl.dr_byteorder, DMU_OT_DNODE,
dr->dt.dl.dr_salt, dr->dt.dl.dr_iv, dr->dt.dl.dr_mac);
}
}
/*
* dbuf_sync_indirect() is called recursively from dbuf_sync_list() so it
* is critical the we not allow the compiler to inline this function in to
* dbuf_sync_list() thereby drastically bloating the stack usage.
*/
noinline static void
dbuf_sync_indirect(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = dr->dr_dbuf;
dnode_t *dn = dr->dr_dnode;
ASSERT(dmu_tx_is_syncing(tx));
dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr);
mutex_enter(&db->db_mtx);
ASSERT(db->db_level > 0);
DBUF_VERIFY(db);
/* Read the block if it hasn't been read yet. */
if (db->db_buf == NULL) {
mutex_exit(&db->db_mtx);
(void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED);
mutex_enter(&db->db_mtx);
}
ASSERT3U(db->db_state, ==, DB_CACHED);
ASSERT(db->db_buf != NULL);
/* Indirect block size must match what the dnode thinks it is. */
ASSERT3U(db->db.db_size, ==, 1<<dn->dn_phys->dn_indblkshift);
dbuf_check_blkptr(dn, db);
/* Provide the pending dirty record to child dbufs */
db->db_data_pending = dr;
mutex_exit(&db->db_mtx);
dbuf_write(dr, db->db_buf, tx);
zio_t *zio = dr->dr_zio;
mutex_enter(&dr->dt.di.dr_mtx);
dbuf_sync_list(&dr->dt.di.dr_children, db->db_level - 1, tx);
ASSERT(list_head(&dr->dt.di.dr_children) == NULL);
mutex_exit(&dr->dt.di.dr_mtx);
zio_nowait(zio);
}
/*
* Verify that the size of the data in our bonus buffer does not exceed
* its recorded size.
*
* The purpose of this verification is to catch any cases in development
* where the size of a phys structure (i.e space_map_phys_t) grows and,
* due to incorrect feature management, older pools expect to read more
* data even though they didn't actually write it to begin with.
*
* For a example, this would catch an error in the feature logic where we
* open an older pool and we expect to write the space map histogram of
* a space map with size SPACE_MAP_SIZE_V0.
*/
static void
dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr)
{
#ifdef ZFS_DEBUG
dnode_t *dn = dr->dr_dnode;
/*
* Encrypted bonus buffers can have data past their bonuslen.
* Skip the verification of these blocks.
*/
if (DMU_OT_IS_ENCRYPTED(dn->dn_bonustype))
return;
uint16_t bonuslen = dn->dn_phys->dn_bonuslen;
uint16_t maxbonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
ASSERT3U(bonuslen, <=, maxbonuslen);
arc_buf_t *datap = dr->dt.dl.dr_data;
char *datap_end = ((char *)datap) + bonuslen;
char *datap_max = ((char *)datap) + maxbonuslen;
/* ensure that everything is zero after our data */
for (; datap_end < datap_max; datap_end++)
ASSERT(*datap_end == 0);
#endif
}
static blkptr_t *
dbuf_lightweight_bp(dbuf_dirty_record_t *dr)
{
/* This must be a lightweight dirty record. */
ASSERT3P(dr->dr_dbuf, ==, NULL);
dnode_t *dn = dr->dr_dnode;
if (dn->dn_phys->dn_nlevels == 1) {
VERIFY3U(dr->dt.dll.dr_blkid, <, dn->dn_phys->dn_nblkptr);
return (&dn->dn_phys->dn_blkptr[dr->dt.dll.dr_blkid]);
} else {
dmu_buf_impl_t *parent_db = dr->dr_parent->dr_dbuf;
int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
VERIFY3U(parent_db->db_level, ==, 1);
VERIFY3P(parent_db->db_dnode_handle->dnh_dnode, ==, dn);
VERIFY3U(dr->dt.dll.dr_blkid >> epbs, ==, parent_db->db_blkid);
blkptr_t *bp = parent_db->db.db_data;
return (&bp[dr->dt.dll.dr_blkid & ((1 << epbs) - 1)]);
}
}
static void
dbuf_lightweight_ready(zio_t *zio)
{
dbuf_dirty_record_t *dr = zio->io_private;
blkptr_t *bp = zio->io_bp;
if (zio->io_error != 0)
return;
dnode_t *dn = dr->dr_dnode;
blkptr_t *bp_orig = dbuf_lightweight_bp(dr);
spa_t *spa = dmu_objset_spa(dn->dn_objset);
int64_t delta = bp_get_dsize_sync(spa, bp) -
bp_get_dsize_sync(spa, bp_orig);
dnode_diduse_space(dn, delta);
uint64_t blkid = dr->dt.dll.dr_blkid;
mutex_enter(&dn->dn_mtx);
if (blkid > dn->dn_phys->dn_maxblkid) {
ASSERT0(dn->dn_objset->os_raw_receive);
dn->dn_phys->dn_maxblkid = blkid;
}
mutex_exit(&dn->dn_mtx);
if (!BP_IS_EMBEDDED(bp)) {
uint64_t fill = BP_IS_HOLE(bp) ? 0 : 1;
BP_SET_FILL(bp, fill);
}
dmu_buf_impl_t *parent_db;
EQUIV(dr->dr_parent == NULL, dn->dn_phys->dn_nlevels == 1);
if (dr->dr_parent == NULL) {
parent_db = dn->dn_dbuf;
} else {
parent_db = dr->dr_parent->dr_dbuf;
}
rw_enter(&parent_db->db_rwlock, RW_WRITER);
*bp_orig = *bp;
rw_exit(&parent_db->db_rwlock);
}
static void
dbuf_lightweight_physdone(zio_t *zio)
{
dbuf_dirty_record_t *dr = zio->io_private;
dsl_pool_t *dp = spa_get_dsl(zio->io_spa);
ASSERT3U(dr->dr_txg, ==, zio->io_txg);
/*
* The callback will be called io_phys_children times. Retire one
* portion of our dirty space each time we are called. Any rounding
* error will be cleaned up by dbuf_lightweight_done().
*/
int delta = dr->dr_accounted / zio->io_phys_children;
dsl_pool_undirty_space(dp, delta, zio->io_txg);
}
static void
dbuf_lightweight_done(zio_t *zio)
{
dbuf_dirty_record_t *dr = zio->io_private;
VERIFY0(zio->io_error);
objset_t *os = dr->dr_dnode->dn_objset;
dmu_tx_t *tx = os->os_synctx;
if (zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE)) {
ASSERT(BP_EQUAL(zio->io_bp, &zio->io_bp_orig));
} else {
dsl_dataset_t *ds = os->os_dsl_dataset;
(void) dsl_dataset_block_kill(ds, &zio->io_bp_orig, tx, B_TRUE);
dsl_dataset_block_born(ds, zio->io_bp, tx);
}
/*
* See comment in dbuf_write_done().
*/
if (zio->io_phys_children == 0) {
dsl_pool_undirty_space(dmu_objset_pool(os),
dr->dr_accounted, zio->io_txg);
} else {
dsl_pool_undirty_space(dmu_objset_pool(os),
dr->dr_accounted % zio->io_phys_children, zio->io_txg);
}
abd_free(dr->dt.dll.dr_abd);
kmem_free(dr, sizeof (*dr));
}
noinline static void
dbuf_sync_lightweight(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
{
dnode_t *dn = dr->dr_dnode;
zio_t *pio;
if (dn->dn_phys->dn_nlevels == 1) {
pio = dn->dn_zio;
} else {
pio = dr->dr_parent->dr_zio;
}
zbookmark_phys_t zb = {
.zb_objset = dmu_objset_id(dn->dn_objset),
.zb_object = dn->dn_object,
.zb_level = 0,
.zb_blkid = dr->dt.dll.dr_blkid,
};
/*
* See comment in dbuf_write(). This is so that zio->io_bp_orig
* will have the old BP in dbuf_lightweight_done().
*/
dr->dr_bp_copy = *dbuf_lightweight_bp(dr);
dr->dr_zio = zio_write(pio, dmu_objset_spa(dn->dn_objset),
dmu_tx_get_txg(tx), &dr->dr_bp_copy, dr->dt.dll.dr_abd,
dn->dn_datablksz, abd_get_size(dr->dt.dll.dr_abd),
&dr->dt.dll.dr_props, dbuf_lightweight_ready, NULL,
dbuf_lightweight_physdone, dbuf_lightweight_done, dr,
ZIO_PRIORITY_ASYNC_WRITE,
ZIO_FLAG_MUSTSUCCEED | dr->dt.dll.dr_flags, &zb);
zio_nowait(dr->dr_zio);
}
/*
* dbuf_sync_leaf() is called recursively from dbuf_sync_list() so it is
* critical the we not allow the compiler to inline this function in to
* dbuf_sync_list() thereby drastically bloating the stack usage.
*/
noinline static void
dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
{
arc_buf_t **datap = &dr->dt.dl.dr_data;
dmu_buf_impl_t *db = dr->dr_dbuf;
dnode_t *dn = dr->dr_dnode;
objset_t *os;
uint64_t txg = tx->tx_txg;
ASSERT(dmu_tx_is_syncing(tx));
dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr);
mutex_enter(&db->db_mtx);
/*
* To be synced, we must be dirtied. But we
* might have been freed after the dirty.
*/
if (db->db_state == DB_UNCACHED) {
/* This buffer has been freed since it was dirtied */
ASSERT(db->db.db_data == NULL);
} else if (db->db_state == DB_FILL) {
/* This buffer was freed and is now being re-filled */
ASSERT(db->db.db_data != dr->dt.dl.dr_data);
} else {
ASSERT(db->db_state == DB_CACHED || db->db_state == DB_NOFILL);
}
DBUF_VERIFY(db);
if (db->db_blkid == DMU_SPILL_BLKID) {
mutex_enter(&dn->dn_mtx);
if (!(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) {
/*
* In the previous transaction group, the bonus buffer
* was entirely used to store the attributes for the
* dnode which overrode the dn_spill field. However,
* when adding more attributes to the file a spill
* block was required to hold the extra attributes.
*
* Make sure to clear the garbage left in the dn_spill
* field from the previous attributes in the bonus
* buffer. Otherwise, after writing out the spill
* block to the new allocated dva, it will free
* the old block pointed to by the invalid dn_spill.
*/
db->db_blkptr = NULL;
}
dn->dn_phys->dn_flags |= DNODE_FLAG_SPILL_BLKPTR;
mutex_exit(&dn->dn_mtx);
}
/*
* If this is a bonus buffer, simply copy the bonus data into the
* dnode. It will be written out when the dnode is synced (and it
* will be synced, since it must have been dirty for dbuf_sync to
* be called).
*/
if (db->db_blkid == DMU_BONUS_BLKID) {
ASSERT(dr->dr_dbuf == db);
dbuf_sync_bonus(dr, tx);
return;
}
os = dn->dn_objset;
/*
* This function may have dropped the db_mtx lock allowing a dmu_sync
* operation to sneak in. As a result, we need to ensure that we
* don't check the dr_override_state until we have returned from
* dbuf_check_blkptr.
*/
dbuf_check_blkptr(dn, db);
/*
* If this buffer is in the middle of an immediate write,
* wait for the synchronous IO to complete.
*/
while (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) {
ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT);
cv_wait(&db->db_changed, &db->db_mtx);
ASSERT(dr->dt.dl.dr_override_state != DR_NOT_OVERRIDDEN);
}
/*
* If this is a dnode block, ensure it is appropriately encrypted
* or decrypted, depending on what we are writing to it this txg.
*/
if (os->os_encrypted && dn->dn_object == DMU_META_DNODE_OBJECT)
dbuf_prepare_encrypted_dnode_leaf(dr);
if (db->db_state != DB_NOFILL &&
dn->dn_object != DMU_META_DNODE_OBJECT &&
zfs_refcount_count(&db->db_holds) > 1 &&
dr->dt.dl.dr_override_state != DR_OVERRIDDEN &&
*datap == db->db_buf) {
/*
* If this buffer is currently "in use" (i.e., there
* are active holds and db_data still references it),
* then make a copy before we start the write so that
* any modifications from the open txg will not leak
* into this write.
*
* NOTE: this copy does not need to be made for
* objects only modified in the syncing context (e.g.
* DNONE_DNODE blocks).
*/
int psize = arc_buf_size(*datap);
int lsize = arc_buf_lsize(*datap);
arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
enum zio_compress compress_type = arc_get_compression(*datap);
uint8_t complevel = arc_get_complevel(*datap);
if (arc_is_encrypted(*datap)) {
boolean_t byteorder;
uint8_t salt[ZIO_DATA_SALT_LEN];
uint8_t iv[ZIO_DATA_IV_LEN];
uint8_t mac[ZIO_DATA_MAC_LEN];
arc_get_raw_params(*datap, &byteorder, salt, iv, mac);
*datap = arc_alloc_raw_buf(os->os_spa, db,
dmu_objset_id(os), byteorder, salt, iv, mac,
dn->dn_type, psize, lsize, compress_type,
complevel);
} else if (compress_type != ZIO_COMPRESS_OFF) {
ASSERT3U(type, ==, ARC_BUFC_DATA);
*datap = arc_alloc_compressed_buf(os->os_spa, db,
psize, lsize, compress_type, complevel);
} else {
*datap = arc_alloc_buf(os->os_spa, db, type, psize);
}
bcopy(db->db.db_data, (*datap)->b_data, psize);
}
db->db_data_pending = dr;
mutex_exit(&db->db_mtx);
dbuf_write(dr, *datap, tx);
ASSERT(!list_link_active(&dr->dr_dirty_node));
if (dn->dn_object == DMU_META_DNODE_OBJECT) {
list_insert_tail(&dn->dn_dirty_records[txg & TXG_MASK], dr);
} else {
zio_nowait(dr->dr_zio);
}
}
void
dbuf_sync_list(list_t *list, int level, dmu_tx_t *tx)
{
dbuf_dirty_record_t *dr;
while ((dr = list_head(list))) {
if (dr->dr_zio != NULL) {
/*
* If we find an already initialized zio then we
* are processing the meta-dnode, and we have finished.
* The dbufs for all dnodes are put back on the list
* during processing, so that we can zio_wait()
* these IOs after initiating all child IOs.
*/
ASSERT3U(dr->dr_dbuf->db.db_object, ==,
DMU_META_DNODE_OBJECT);
break;
}
list_remove(list, dr);
if (dr->dr_dbuf == NULL) {
dbuf_sync_lightweight(dr, tx);
} else {
if (dr->dr_dbuf->db_blkid != DMU_BONUS_BLKID &&
dr->dr_dbuf->db_blkid != DMU_SPILL_BLKID) {
VERIFY3U(dr->dr_dbuf->db_level, ==, level);
}
if (dr->dr_dbuf->db_level > 0)
dbuf_sync_indirect(dr, tx);
else
dbuf_sync_leaf(dr, tx);
}
}
}
static void
dbuf_write_ready(zio_t *zio, arc_buf_t *buf, void *vdb)
{
(void) buf;
dmu_buf_impl_t *db = vdb;
dnode_t *dn;
blkptr_t *bp = zio->io_bp;
blkptr_t *bp_orig = &zio->io_bp_orig;
spa_t *spa = zio->io_spa;
int64_t delta;
uint64_t fill = 0;
int i;
ASSERT3P(db->db_blkptr, !=, NULL);
ASSERT3P(&db->db_data_pending->dr_bp_copy, ==, bp);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
delta = bp_get_dsize_sync(spa, bp) - bp_get_dsize_sync(spa, bp_orig);
dnode_diduse_space(dn, delta - zio->io_prev_space_delta);
zio->io_prev_space_delta = delta;
if (bp->blk_birth != 0) {
ASSERT((db->db_blkid != DMU_SPILL_BLKID &&
BP_GET_TYPE(bp) == dn->dn_type) ||
(db->db_blkid == DMU_SPILL_BLKID &&
BP_GET_TYPE(bp) == dn->dn_bonustype) ||
BP_IS_EMBEDDED(bp));
ASSERT(BP_GET_LEVEL(bp) == db->db_level);
}
mutex_enter(&db->db_mtx);
#ifdef ZFS_DEBUG
if (db->db_blkid == DMU_SPILL_BLKID) {
ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR);
ASSERT(!(BP_IS_HOLE(bp)) &&
db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys));
}
#endif
if (db->db_level == 0) {
mutex_enter(&dn->dn_mtx);
if (db->db_blkid > dn->dn_phys->dn_maxblkid &&
db->db_blkid != DMU_SPILL_BLKID) {
ASSERT0(db->db_objset->os_raw_receive);
dn->dn_phys->dn_maxblkid = db->db_blkid;
}
mutex_exit(&dn->dn_mtx);
if (dn->dn_type == DMU_OT_DNODE) {
i = 0;
while (i < db->db.db_size) {
dnode_phys_t *dnp =
(void *)(((char *)db->db.db_data) + i);
i += DNODE_MIN_SIZE;
if (dnp->dn_type != DMU_OT_NONE) {
fill++;
i += dnp->dn_extra_slots *
DNODE_MIN_SIZE;
}
}
} else {
if (BP_IS_HOLE(bp)) {
fill = 0;
} else {
fill = 1;
}
}
} else {
blkptr_t *ibp = db->db.db_data;
ASSERT3U(db->db.db_size, ==, 1<<dn->dn_phys->dn_indblkshift);
for (i = db->db.db_size >> SPA_BLKPTRSHIFT; i > 0; i--, ibp++) {
if (BP_IS_HOLE(ibp))
continue;
fill += BP_GET_FILL(ibp);
}
}
DB_DNODE_EXIT(db);
if (!BP_IS_EMBEDDED(bp))
BP_SET_FILL(bp, fill);
mutex_exit(&db->db_mtx);
db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_WRITER, FTAG);
*db->db_blkptr = *bp;
dmu_buf_unlock_parent(db, dblt, FTAG);
}
/*
* This function gets called just prior to running through the compression
* stage of the zio pipeline. If we're an indirect block comprised of only
* holes, then we want this indirect to be compressed away to a hole. In
* order to do that we must zero out any information about the holes that
* this indirect points to prior to before we try to compress it.
*/
static void
dbuf_write_children_ready(zio_t *zio, arc_buf_t *buf, void *vdb)
{
(void) zio, (void) buf;
dmu_buf_impl_t *db = vdb;
dnode_t *dn;
blkptr_t *bp;
unsigned int epbs, i;
ASSERT3U(db->db_level, >, 0);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
ASSERT3U(epbs, <, 31);
/* Determine if all our children are holes */
for (i = 0, bp = db->db.db_data; i < 1ULL << epbs; i++, bp++) {
if (!BP_IS_HOLE(bp))
break;
}
/*
* If all the children are holes, then zero them all out so that
* we may get compressed away.
*/
if (i == 1ULL << epbs) {
/*
* We only found holes. Grab the rwlock to prevent
* anybody from reading the blocks we're about to
* zero out.
*/
rw_enter(&db->db_rwlock, RW_WRITER);
bzero(db->db.db_data, db->db.db_size);
rw_exit(&db->db_rwlock);
}
DB_DNODE_EXIT(db);
}
/*
* The SPA will call this callback several times for each zio - once
* for every physical child i/o (zio->io_phys_children times). This
* allows the DMU to monitor the progress of each logical i/o. For example,
* there may be 2 copies of an indirect block, or many fragments of a RAID-Z
* block. There may be a long delay before all copies/fragments are completed,
* so this callback allows us to retire dirty space gradually, as the physical
* i/os complete.
*/
static void
dbuf_write_physdone(zio_t *zio, arc_buf_t *buf, void *arg)
{
(void) buf;
dmu_buf_impl_t *db = arg;
objset_t *os = db->db_objset;
dsl_pool_t *dp = dmu_objset_pool(os);
dbuf_dirty_record_t *dr;
int delta = 0;
dr = db->db_data_pending;
ASSERT3U(dr->dr_txg, ==, zio->io_txg);
/*
* The callback will be called io_phys_children times. Retire one
* portion of our dirty space each time we are called. Any rounding
* error will be cleaned up by dbuf_write_done().
*/
delta = dr->dr_accounted / zio->io_phys_children;
dsl_pool_undirty_space(dp, delta, zio->io_txg);
}
static void
dbuf_write_done(zio_t *zio, arc_buf_t *buf, void *vdb)
{
(void) buf;
dmu_buf_impl_t *db = vdb;
blkptr_t *bp_orig = &zio->io_bp_orig;
blkptr_t *bp = db->db_blkptr;
objset_t *os = db->db_objset;
dmu_tx_t *tx = os->os_synctx;
ASSERT0(zio->io_error);
ASSERT(db->db_blkptr == bp);
/*
* For nopwrites and rewrites we ensure that the bp matches our
* original and bypass all the accounting.
*/
if (zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE)) {
ASSERT(BP_EQUAL(bp, bp_orig));
} else {
dsl_dataset_t *ds = os->os_dsl_dataset;
(void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE);
dsl_dataset_block_born(ds, bp, tx);
}
mutex_enter(&db->db_mtx);
DBUF_VERIFY(db);
dbuf_dirty_record_t *dr = db->db_data_pending;
dnode_t *dn = dr->dr_dnode;
ASSERT(!list_link_active(&dr->dr_dirty_node));
ASSERT(dr->dr_dbuf == db);
ASSERT(list_next(&db->db_dirty_records, dr) == NULL);
list_remove(&db->db_dirty_records, dr);
#ifdef ZFS_DEBUG
if (db->db_blkid == DMU_SPILL_BLKID) {
ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR);
ASSERT(!(BP_IS_HOLE(db->db_blkptr)) &&
db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys));
}
#endif
if (db->db_level == 0) {
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN);
if (db->db_state != DB_NOFILL) {
if (dr->dt.dl.dr_data != db->db_buf)
arc_buf_destroy(dr->dt.dl.dr_data, db);
}
} else {
ASSERT(list_head(&dr->dt.di.dr_children) == NULL);
ASSERT3U(db->db.db_size, ==, 1 << dn->dn_phys->dn_indblkshift);
if (!BP_IS_HOLE(db->db_blkptr)) {
int epbs __maybe_unused = dn->dn_phys->dn_indblkshift -
SPA_BLKPTRSHIFT;
ASSERT3U(db->db_blkid, <=,
dn->dn_phys->dn_maxblkid >> (db->db_level * epbs));
ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==,
db->db.db_size);
}
mutex_destroy(&dr->dt.di.dr_mtx);
list_destroy(&dr->dt.di.dr_children);
}
cv_broadcast(&db->db_changed);
ASSERT(db->db_dirtycnt > 0);
db->db_dirtycnt -= 1;
db->db_data_pending = NULL;
dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg, B_FALSE);
/*
* If we didn't do a physical write in this ZIO and we
* still ended up here, it means that the space of the
* dbuf that we just released (and undirtied) above hasn't
* been marked as undirtied in the pool's accounting.
*
* Thus, we undirty that space in the pool's view of the
* world here. For physical writes this type of update
* happens in dbuf_write_physdone().
*
* If we did a physical write, cleanup any rounding errors
* that came up due to writing multiple copies of a block
* on disk [see dbuf_write_physdone()].
*/
if (zio->io_phys_children == 0) {
dsl_pool_undirty_space(dmu_objset_pool(os),
dr->dr_accounted, zio->io_txg);
} else {
dsl_pool_undirty_space(dmu_objset_pool(os),
dr->dr_accounted % zio->io_phys_children, zio->io_txg);
}
kmem_free(dr, sizeof (dbuf_dirty_record_t));
}
static void
dbuf_write_nofill_ready(zio_t *zio)
{
dbuf_write_ready(zio, NULL, zio->io_private);
}
static void
dbuf_write_nofill_done(zio_t *zio)
{
dbuf_write_done(zio, NULL, zio->io_private);
}
static void
dbuf_write_override_ready(zio_t *zio)
{
dbuf_dirty_record_t *dr = zio->io_private;
dmu_buf_impl_t *db = dr->dr_dbuf;
dbuf_write_ready(zio, NULL, db);
}
static void
dbuf_write_override_done(zio_t *zio)
{
dbuf_dirty_record_t *dr = zio->io_private;
dmu_buf_impl_t *db = dr->dr_dbuf;
blkptr_t *obp = &dr->dt.dl.dr_overridden_by;
mutex_enter(&db->db_mtx);
if (!BP_EQUAL(zio->io_bp, obp)) {
if (!BP_IS_HOLE(obp))
dsl_free(spa_get_dsl(zio->io_spa), zio->io_txg, obp);
arc_release(dr->dt.dl.dr_data, db);
}
mutex_exit(&db->db_mtx);
dbuf_write_done(zio, NULL, db);
if (zio->io_abd != NULL)
abd_free(zio->io_abd);
}
typedef struct dbuf_remap_impl_callback_arg {
objset_t *drica_os;
uint64_t drica_blk_birth;
dmu_tx_t *drica_tx;
} dbuf_remap_impl_callback_arg_t;
static void
dbuf_remap_impl_callback(uint64_t vdev, uint64_t offset, uint64_t size,
void *arg)
{
dbuf_remap_impl_callback_arg_t *drica = arg;
objset_t *os = drica->drica_os;
spa_t *spa = dmu_objset_spa(os);
dmu_tx_t *tx = drica->drica_tx;
ASSERT(dsl_pool_sync_context(spa_get_dsl(spa)));
if (os == spa_meta_objset(spa)) {
spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx);
} else {
dsl_dataset_block_remapped(dmu_objset_ds(os), vdev, offset,
size, drica->drica_blk_birth, tx);
}
}
static void
dbuf_remap_impl(dnode_t *dn, blkptr_t *bp, krwlock_t *rw, dmu_tx_t *tx)
{
blkptr_t bp_copy = *bp;
spa_t *spa = dmu_objset_spa(dn->dn_objset);
dbuf_remap_impl_callback_arg_t drica;
ASSERT(dsl_pool_sync_context(spa_get_dsl(spa)));
drica.drica_os = dn->dn_objset;
drica.drica_blk_birth = bp->blk_birth;
drica.drica_tx = tx;
if (spa_remap_blkptr(spa, &bp_copy, dbuf_remap_impl_callback,
&drica)) {
/*
* If the blkptr being remapped is tracked by a livelist,
* then we need to make sure the livelist reflects the update.
* First, cancel out the old blkptr by appending a 'FREE'
* entry. Next, add an 'ALLOC' to track the new version. This
* way we avoid trying to free an inaccurate blkptr at delete.
* Note that embedded blkptrs are not tracked in livelists.
*/
if (dn->dn_objset != spa_meta_objset(spa)) {
dsl_dataset_t *ds = dmu_objset_ds(dn->dn_objset);
if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) &&
bp->blk_birth > ds->ds_dir->dd_origin_txg) {
ASSERT(!BP_IS_EMBEDDED(bp));
ASSERT(dsl_dir_is_clone(ds->ds_dir));
ASSERT(spa_feature_is_enabled(spa,
SPA_FEATURE_LIVELIST));
bplist_append(&ds->ds_dir->dd_pending_frees,
bp);
bplist_append(&ds->ds_dir->dd_pending_allocs,
&bp_copy);
}
}
/*
* The db_rwlock prevents dbuf_read_impl() from
* dereferencing the BP while we are changing it. To
* avoid lock contention, only grab it when we are actually
* changing the BP.
*/
if (rw != NULL)
rw_enter(rw, RW_WRITER);
*bp = bp_copy;
if (rw != NULL)
rw_exit(rw);
}
}
/*
* Remap any existing BP's to concrete vdevs, if possible.
*/
static void
dbuf_remap(dnode_t *dn, dmu_buf_impl_t *db, dmu_tx_t *tx)
{
spa_t *spa = dmu_objset_spa(db->db_objset);
ASSERT(dsl_pool_sync_context(spa_get_dsl(spa)));
if (!spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL))
return;
if (db->db_level > 0) {
blkptr_t *bp = db->db.db_data;
for (int i = 0; i < db->db.db_size >> SPA_BLKPTRSHIFT; i++) {
dbuf_remap_impl(dn, &bp[i], &db->db_rwlock, tx);
}
} else if (db->db.db_object == DMU_META_DNODE_OBJECT) {
dnode_phys_t *dnp = db->db.db_data;
ASSERT3U(db->db_dnode_handle->dnh_dnode->dn_type, ==,
DMU_OT_DNODE);
for (int i = 0; i < db->db.db_size >> DNODE_SHIFT;
i += dnp[i].dn_extra_slots + 1) {
for (int j = 0; j < dnp[i].dn_nblkptr; j++) {
krwlock_t *lock = (dn->dn_dbuf == NULL ? NULL :
&dn->dn_dbuf->db_rwlock);
dbuf_remap_impl(dn, &dnp[i].dn_blkptr[j], lock,
tx);
}
}
}
}
/* Issue I/O to commit a dirty buffer to disk. */
static void
dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = dr->dr_dbuf;
dnode_t *dn = dr->dr_dnode;
objset_t *os;
dmu_buf_impl_t *parent = db->db_parent;
uint64_t txg = tx->tx_txg;
zbookmark_phys_t zb;
zio_prop_t zp;
zio_t *pio; /* parent I/O */
int wp_flag = 0;
ASSERT(dmu_tx_is_syncing(tx));
os = dn->dn_objset;
if (db->db_state != DB_NOFILL) {
if (db->db_level > 0 || dn->dn_type == DMU_OT_DNODE) {
/*
* Private object buffers are released here rather
* than in dbuf_dirty() since they are only modified
* in the syncing context and we don't want the
* overhead of making multiple copies of the data.
*/
if (BP_IS_HOLE(db->db_blkptr)) {
arc_buf_thaw(data);
} else {
dbuf_release_bp(db);
}
dbuf_remap(dn, db, tx);
}
}
if (parent != dn->dn_dbuf) {
/* Our parent is an indirect block. */
/* We have a dirty parent that has been scheduled for write. */
ASSERT(parent && parent->db_data_pending);
/* Our parent's buffer is one level closer to the dnode. */
ASSERT(db->db_level == parent->db_level-1);
/*
* We're about to modify our parent's db_data by modifying
* our block pointer, so the parent must be released.
*/
ASSERT(arc_released(parent->db_buf));
pio = parent->db_data_pending->dr_zio;
} else {
/* Our parent is the dnode itself. */
ASSERT((db->db_level == dn->dn_phys->dn_nlevels-1 &&
db->db_blkid != DMU_SPILL_BLKID) ||
(db->db_blkid == DMU_SPILL_BLKID && db->db_level == 0));
if (db->db_blkid != DMU_SPILL_BLKID)
ASSERT3P(db->db_blkptr, ==,
&dn->dn_phys->dn_blkptr[db->db_blkid]);
pio = dn->dn_zio;
}
ASSERT(db->db_level == 0 || data == db->db_buf);
ASSERT3U(db->db_blkptr->blk_birth, <=, txg);
ASSERT(pio);
SET_BOOKMARK(&zb, os->os_dsl_dataset ?
os->os_dsl_dataset->ds_object : DMU_META_OBJSET,
db->db.db_object, db->db_level, db->db_blkid);
if (db->db_blkid == DMU_SPILL_BLKID)
wp_flag = WP_SPILL;
wp_flag |= (db->db_state == DB_NOFILL) ? WP_NOFILL : 0;
dmu_write_policy(os, dn, db->db_level, wp_flag, &zp);
/*
* We copy the blkptr now (rather than when we instantiate the dirty
* record), because its value can change between open context and
* syncing context. We do not need to hold dn_struct_rwlock to read
* db_blkptr because we are in syncing context.
*/
dr->dr_bp_copy = *db->db_blkptr;
if (db->db_level == 0 &&
dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
/*
* The BP for this block has been provided by open context
* (by dmu_sync() or dmu_buf_write_embedded()).
*/
abd_t *contents = (data != NULL) ?
abd_get_from_buf(data->b_data, arc_buf_size(data)) : NULL;
dr->dr_zio = zio_write(pio, os->os_spa, txg, &dr->dr_bp_copy,
contents, db->db.db_size, db->db.db_size, &zp,
dbuf_write_override_ready, NULL, NULL,
dbuf_write_override_done,
dr, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
mutex_enter(&db->db_mtx);
dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
zio_write_override(dr->dr_zio, &dr->dt.dl.dr_overridden_by,
dr->dt.dl.dr_copies, dr->dt.dl.dr_nopwrite);
mutex_exit(&db->db_mtx);
} else if (db->db_state == DB_NOFILL) {
ASSERT(zp.zp_checksum == ZIO_CHECKSUM_OFF ||
zp.zp_checksum == ZIO_CHECKSUM_NOPARITY);
dr->dr_zio = zio_write(pio, os->os_spa, txg,
&dr->dr_bp_copy, NULL, db->db.db_size, db->db.db_size, &zp,
dbuf_write_nofill_ready, NULL, NULL,
dbuf_write_nofill_done, db,
ZIO_PRIORITY_ASYNC_WRITE,
ZIO_FLAG_MUSTSUCCEED | ZIO_FLAG_NODATA, &zb);
} else {
ASSERT(arc_released(data));
/*
* For indirect blocks, we want to setup the children
* ready callback so that we can properly handle an indirect
* block that only contains holes.
*/
arc_write_done_func_t *children_ready_cb = NULL;
if (db->db_level != 0)
children_ready_cb = dbuf_write_children_ready;
dr->dr_zio = arc_write(pio, os->os_spa, txg,
&dr->dr_bp_copy, data, dbuf_is_l2cacheable(db),
&zp, dbuf_write_ready,
children_ready_cb, dbuf_write_physdone,
dbuf_write_done, db, ZIO_PRIORITY_ASYNC_WRITE,
ZIO_FLAG_MUSTSUCCEED, &zb);
}
}
EXPORT_SYMBOL(dbuf_find);
EXPORT_SYMBOL(dbuf_is_metadata);
EXPORT_SYMBOL(dbuf_destroy);
EXPORT_SYMBOL(dbuf_loan_arcbuf);
EXPORT_SYMBOL(dbuf_whichblock);
EXPORT_SYMBOL(dbuf_read);
EXPORT_SYMBOL(dbuf_unoverride);
EXPORT_SYMBOL(dbuf_free_range);
EXPORT_SYMBOL(dbuf_new_size);
EXPORT_SYMBOL(dbuf_release_bp);
EXPORT_SYMBOL(dbuf_dirty);
EXPORT_SYMBOL(dmu_buf_set_crypt_params);
EXPORT_SYMBOL(dmu_buf_will_dirty);
EXPORT_SYMBOL(dmu_buf_is_dirty);
EXPORT_SYMBOL(dmu_buf_will_not_fill);
EXPORT_SYMBOL(dmu_buf_will_fill);
EXPORT_SYMBOL(dmu_buf_fill_done);
EXPORT_SYMBOL(dmu_buf_rele);
EXPORT_SYMBOL(dbuf_assign_arcbuf);
EXPORT_SYMBOL(dbuf_prefetch);
EXPORT_SYMBOL(dbuf_hold_impl);
EXPORT_SYMBOL(dbuf_hold);
EXPORT_SYMBOL(dbuf_hold_level);
EXPORT_SYMBOL(dbuf_create_bonus);
EXPORT_SYMBOL(dbuf_spill_set_blksz);
EXPORT_SYMBOL(dbuf_rm_spill);
EXPORT_SYMBOL(dbuf_add_ref);
EXPORT_SYMBOL(dbuf_rele);
EXPORT_SYMBOL(dbuf_rele_and_unlock);
EXPORT_SYMBOL(dbuf_refcount);
EXPORT_SYMBOL(dbuf_sync_list);
EXPORT_SYMBOL(dmu_buf_set_user);
EXPORT_SYMBOL(dmu_buf_set_user_ie);
EXPORT_SYMBOL(dmu_buf_get_user);
EXPORT_SYMBOL(dmu_buf_get_blkptr);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, max_bytes, ULONG, ZMOD_RW,
"Maximum size in bytes of the dbuf cache.");
ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, hiwater_pct, UINT, ZMOD_RW,
"Percentage over dbuf_cache_max_bytes when dbufs must be evicted "
"directly.");
ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, lowater_pct, UINT, ZMOD_RW,
"Percentage below dbuf_cache_max_bytes when the evict thread stops "
"evicting dbufs.");
ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, metadata_cache_max_bytes, ULONG, ZMOD_RW,
"Maximum size in bytes of the dbuf metadata cache.");
ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, cache_shift, INT, ZMOD_RW,
"Set the size of the dbuf cache to a log2 fraction of arc size.");
ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, metadata_cache_shift, INT, ZMOD_RW,
"Set the size of the dbuf metadata cache to a log2 fraction of arc "
"size.");
/* END CSTYLED */
diff --git a/module/zfs/ddt_zap.c b/module/zfs/ddt_zap.c
index c5c9eda0b2d0..d0127f22e5ff 100644
--- a/module/zfs/ddt_zap.c
+++ b/module/zfs/ddt_zap.c
@@ -1,168 +1,168 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2018 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/zio.h>
#include <sys/ddt.h>
#include <sys/zap.h>
#include <sys/dmu_tx.h>
-int ddt_zap_leaf_blockshift = 12;
-int ddt_zap_indirect_blockshift = 12;
+static const int ddt_zap_leaf_blockshift = 12;
+static const int ddt_zap_indirect_blockshift = 12;
static int
ddt_zap_create(objset_t *os, uint64_t *objectp, dmu_tx_t *tx, boolean_t prehash)
{
zap_flags_t flags = ZAP_FLAG_HASH64 | ZAP_FLAG_UINT64_KEY;
if (prehash)
flags |= ZAP_FLAG_PRE_HASHED_KEY;
*objectp = zap_create_flags(os, 0, flags, DMU_OT_DDT_ZAP,
ddt_zap_leaf_blockshift, ddt_zap_indirect_blockshift,
DMU_OT_NONE, 0, tx);
return (*objectp == 0 ? SET_ERROR(ENOTSUP) : 0);
}
static int
ddt_zap_destroy(objset_t *os, uint64_t object, dmu_tx_t *tx)
{
return (zap_destroy(os, object, tx));
}
static int
ddt_zap_lookup(objset_t *os, uint64_t object, ddt_entry_t *dde)
{
uchar_t *cbuf;
uint64_t one, csize;
int error;
cbuf = kmem_alloc(sizeof (dde->dde_phys) + 1, KM_SLEEP);
error = zap_length_uint64(os, object, (uint64_t *)&dde->dde_key,
DDT_KEY_WORDS, &one, &csize);
if (error)
goto out;
ASSERT(one == 1);
ASSERT(csize <= (sizeof (dde->dde_phys) + 1));
error = zap_lookup_uint64(os, object, (uint64_t *)&dde->dde_key,
DDT_KEY_WORDS, 1, csize, cbuf);
if (error)
goto out;
ddt_decompress(cbuf, dde->dde_phys, csize, sizeof (dde->dde_phys));
out:
kmem_free(cbuf, sizeof (dde->dde_phys) + 1);
return (error);
}
static void
ddt_zap_prefetch(objset_t *os, uint64_t object, ddt_entry_t *dde)
{
(void) zap_prefetch_uint64(os, object, (uint64_t *)&dde->dde_key,
DDT_KEY_WORDS);
}
static int
ddt_zap_update(objset_t *os, uint64_t object, ddt_entry_t *dde, dmu_tx_t *tx)
{
uchar_t cbuf[sizeof (dde->dde_phys) + 1];
uint64_t csize;
csize = ddt_compress(dde->dde_phys, cbuf,
sizeof (dde->dde_phys), sizeof (cbuf));
return (zap_update_uint64(os, object, (uint64_t *)&dde->dde_key,
DDT_KEY_WORDS, 1, csize, cbuf, tx));
}
static int
ddt_zap_remove(objset_t *os, uint64_t object, ddt_entry_t *dde, dmu_tx_t *tx)
{
return (zap_remove_uint64(os, object, (uint64_t *)&dde->dde_key,
DDT_KEY_WORDS, tx));
}
static int
ddt_zap_walk(objset_t *os, uint64_t object, ddt_entry_t *dde, uint64_t *walk)
{
zap_cursor_t zc;
zap_attribute_t za;
int error;
if (*walk == 0) {
/*
* We don't want to prefetch the entire ZAP object, because
* it can be enormous. Also the primary use of DDT iteration
* is for scrubbing, in which case we will be issuing many
* scrub I/Os for each ZAP block that we read in, so
* reading the ZAP is unlikely to be the bottleneck.
*/
zap_cursor_init_noprefetch(&zc, os, object);
} else {
zap_cursor_init_serialized(&zc, os, object, *walk);
}
if ((error = zap_cursor_retrieve(&zc, &za)) == 0) {
uchar_t cbuf[sizeof (dde->dde_phys) + 1];
uint64_t csize = za.za_num_integers;
ASSERT(za.za_integer_length == 1);
error = zap_lookup_uint64(os, object, (uint64_t *)za.za_name,
DDT_KEY_WORDS, 1, csize, cbuf);
ASSERT(error == 0);
if (error == 0) {
ddt_decompress(cbuf, dde->dde_phys, csize,
sizeof (dde->dde_phys));
dde->dde_key = *(ddt_key_t *)za.za_name;
}
zap_cursor_advance(&zc);
*walk = zap_cursor_serialize(&zc);
}
zap_cursor_fini(&zc);
return (error);
}
static int
ddt_zap_count(objset_t *os, uint64_t object, uint64_t *count)
{
return (zap_count(os, object, count));
}
const ddt_ops_t ddt_zap_ops = {
"zap",
ddt_zap_create,
ddt_zap_destroy,
ddt_zap_lookup,
ddt_zap_prefetch,
ddt_zap_update,
ddt_zap_remove,
ddt_zap_walk,
ddt_zap_count,
};
diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c
index e6f391066de0..d802ce557657 100644
--- a/module/zfs/dmu.c
+++ b/module/zfs/dmu.c
@@ -1,2361 +1,2361 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
* Copyright (c) 2016, Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2015 by Chunwei Chen. All rights reserved.
* Copyright (c) 2019 Datto Inc.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
*/
#include <sys/dmu.h>
#include <sys/dmu_impl.h>
#include <sys/dmu_tx.h>
#include <sys/dbuf.h>
#include <sys/dnode.h>
#include <sys/zfs_context.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_traverse.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_synctask.h>
#include <sys/dsl_prop.h>
#include <sys/dmu_zfetch.h>
#include <sys/zfs_ioctl.h>
#include <sys/zap.h>
#include <sys/zio_checksum.h>
#include <sys/zio_compress.h>
#include <sys/sa.h>
#include <sys/zfeature.h>
#include <sys/abd.h>
#include <sys/trace_zfs.h>
#include <sys/zfs_racct.h>
#include <sys/zfs_rlock.h>
#ifdef _KERNEL
#include <sys/vmsystm.h>
#include <sys/zfs_znode.h>
#endif
/*
* Enable/disable nopwrite feature.
*/
-int zfs_nopwrite_enabled = 1;
+static int zfs_nopwrite_enabled = 1;
/*
* Tunable to control percentage of dirtied L1 blocks from frees allowed into
* one TXG. After this threshold is crossed, additional dirty blocks from frees
* will wait until the next TXG.
* A value of zero will disable this throttle.
*/
-unsigned long zfs_per_txg_dirty_frees_percent = 5;
+static unsigned long zfs_per_txg_dirty_frees_percent = 5;
/*
* Enable/disable forcing txg sync when dirty checking for holes with lseek().
* By default this is enabled to ensure accurate hole reporting, it can result
* in a significant performance penalty for lseek(SEEK_HOLE) heavy workloads.
* Disabling this option will result in holes never being reported in dirty
* files which is always safe.
*/
-int zfs_dmu_offset_next_sync = 1;
+static int zfs_dmu_offset_next_sync = 1;
/*
* Limit the amount we can prefetch with one call to this amount. This
* helps to limit the amount of memory that can be used by prefetching.
* Larger objects should be prefetched a bit at a time.
*/
-int dmu_prefetch_max = 8 * SPA_MAXBLOCKSIZE;
+static int dmu_prefetch_max = 8 * SPA_MAXBLOCKSIZE;
const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
{DMU_BSWAP_UINT8, TRUE, FALSE, FALSE, "unallocated" },
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "object directory" },
{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "object array" },
{DMU_BSWAP_UINT8, TRUE, FALSE, FALSE, "packed nvlist" },
{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "packed nvlist size" },
{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "bpobj" },
{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "bpobj header" },
{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "SPA space map header" },
{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "SPA space map" },
{DMU_BSWAP_UINT64, TRUE, FALSE, TRUE, "ZIL intent log" },
{DMU_BSWAP_DNODE, TRUE, FALSE, TRUE, "DMU dnode" },
{DMU_BSWAP_OBJSET, TRUE, TRUE, FALSE, "DMU objset" },
{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "DSL directory" },
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL directory child map"},
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL dataset snap map" },
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL props" },
{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "DSL dataset" },
{DMU_BSWAP_ZNODE, TRUE, FALSE, FALSE, "ZFS znode" },
{DMU_BSWAP_OLDACL, TRUE, FALSE, TRUE, "ZFS V0 ACL" },
{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "ZFS plain file" },
{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS directory" },
{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "ZFS master node" },
{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS delete queue" },
{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "zvol object" },
{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "zvol prop" },
{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "other uint8[]" },
{DMU_BSWAP_UINT64, FALSE, FALSE, TRUE, "other uint64[]" },
{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "other ZAP" },
{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "persistent error log" },
{DMU_BSWAP_UINT8, TRUE, FALSE, FALSE, "SPA history" },
{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "SPA history offsets" },
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "Pool properties" },
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL permissions" },
{DMU_BSWAP_ACL, TRUE, FALSE, TRUE, "ZFS ACL" },
{DMU_BSWAP_UINT8, TRUE, FALSE, TRUE, "ZFS SYSACL" },
{DMU_BSWAP_UINT8, TRUE, FALSE, TRUE, "FUID table" },
{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "FUID table size" },
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL dataset next clones"},
{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "scan work queue" },
{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS user/group/project used" },
{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS user/group/project quota"},
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "snapshot refcount tags"},
{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "DDT ZAP algorithm" },
{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "DDT statistics" },
{DMU_BSWAP_UINT8, TRUE, FALSE, TRUE, "System attributes" },
{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "SA master node" },
{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "SA attr registration" },
{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "SA attr layouts" },
{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "scan translations" },
{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "deduplicated block" },
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL deadlist map" },
{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "DSL deadlist map hdr" },
{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL dir clones" },
{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "bpobj subobj" }
};
const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS] = {
{ byteswap_uint8_array, "uint8" },
{ byteswap_uint16_array, "uint16" },
{ byteswap_uint32_array, "uint32" },
{ byteswap_uint64_array, "uint64" },
{ zap_byteswap, "zap" },
{ dnode_buf_byteswap, "dnode" },
{ dmu_objset_byteswap, "objset" },
{ zfs_znode_byteswap, "znode" },
{ zfs_oldacl_byteswap, "oldacl" },
{ zfs_acl_byteswap, "acl" }
};
static int
dmu_buf_hold_noread_by_dnode(dnode_t *dn, uint64_t offset,
void *tag, dmu_buf_t **dbp)
{
uint64_t blkid;
dmu_buf_impl_t *db;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
blkid = dbuf_whichblock(dn, 0, offset);
db = dbuf_hold(dn, blkid, tag);
rw_exit(&dn->dn_struct_rwlock);
if (db == NULL) {
*dbp = NULL;
return (SET_ERROR(EIO));
}
*dbp = &db->db;
return (0);
}
int
dmu_buf_hold_noread(objset_t *os, uint64_t object, uint64_t offset,
void *tag, dmu_buf_t **dbp)
{
dnode_t *dn;
uint64_t blkid;
dmu_buf_impl_t *db;
int err;
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
rw_enter(&dn->dn_struct_rwlock, RW_READER);
blkid = dbuf_whichblock(dn, 0, offset);
db = dbuf_hold(dn, blkid, tag);
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
if (db == NULL) {
*dbp = NULL;
return (SET_ERROR(EIO));
}
*dbp = &db->db;
return (err);
}
int
dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
void *tag, dmu_buf_t **dbp, int flags)
{
int err;
int db_flags = DB_RF_CANFAIL;
if (flags & DMU_READ_NO_PREFETCH)
db_flags |= DB_RF_NOPREFETCH;
if (flags & DMU_READ_NO_DECRYPT)
db_flags |= DB_RF_NO_DECRYPT;
err = dmu_buf_hold_noread_by_dnode(dn, offset, tag, dbp);
if (err == 0) {
dmu_buf_impl_t *db = (dmu_buf_impl_t *)(*dbp);
err = dbuf_read(db, NULL, db_flags);
if (err != 0) {
dbuf_rele(db, tag);
*dbp = NULL;
}
}
return (err);
}
int
dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
void *tag, dmu_buf_t **dbp, int flags)
{
int err;
int db_flags = DB_RF_CANFAIL;
if (flags & DMU_READ_NO_PREFETCH)
db_flags |= DB_RF_NOPREFETCH;
if (flags & DMU_READ_NO_DECRYPT)
db_flags |= DB_RF_NO_DECRYPT;
err = dmu_buf_hold_noread(os, object, offset, tag, dbp);
if (err == 0) {
dmu_buf_impl_t *db = (dmu_buf_impl_t *)(*dbp);
err = dbuf_read(db, NULL, db_flags);
if (err != 0) {
dbuf_rele(db, tag);
*dbp = NULL;
}
}
return (err);
}
int
dmu_bonus_max(void)
{
return (DN_OLD_MAX_BONUSLEN);
}
int
dmu_set_bonus(dmu_buf_t *db_fake, int newsize, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dnode_t *dn;
int error;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
if (dn->dn_bonus != db) {
error = SET_ERROR(EINVAL);
} else if (newsize < 0 || newsize > db_fake->db_size) {
error = SET_ERROR(EINVAL);
} else {
dnode_setbonuslen(dn, newsize, tx);
error = 0;
}
DB_DNODE_EXIT(db);
return (error);
}
int
dmu_set_bonustype(dmu_buf_t *db_fake, dmu_object_type_t type, dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dnode_t *dn;
int error;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
if (!DMU_OT_IS_VALID(type)) {
error = SET_ERROR(EINVAL);
} else if (dn->dn_bonus != db) {
error = SET_ERROR(EINVAL);
} else {
dnode_setbonus_type(dn, type, tx);
error = 0;
}
DB_DNODE_EXIT(db);
return (error);
}
dmu_object_type_t
dmu_get_bonustype(dmu_buf_t *db_fake)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dnode_t *dn;
dmu_object_type_t type;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
type = dn->dn_bonustype;
DB_DNODE_EXIT(db);
return (type);
}
int
dmu_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx)
{
dnode_t *dn;
int error;
error = dnode_hold(os, object, FTAG, &dn);
dbuf_rm_spill(dn, tx);
rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
dnode_rm_spill(dn, tx);
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
return (error);
}
/*
* Lookup and hold the bonus buffer for the provided dnode. If the dnode
* has not yet been allocated a new bonus dbuf a will be allocated.
* Returns ENOENT, EIO, or 0.
*/
int dmu_bonus_hold_by_dnode(dnode_t *dn, void *tag, dmu_buf_t **dbp,
uint32_t flags)
{
dmu_buf_impl_t *db;
int error;
uint32_t db_flags = DB_RF_MUST_SUCCEED;
if (flags & DMU_READ_NO_PREFETCH)
db_flags |= DB_RF_NOPREFETCH;
if (flags & DMU_READ_NO_DECRYPT)
db_flags |= DB_RF_NO_DECRYPT;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (dn->dn_bonus == NULL) {
rw_exit(&dn->dn_struct_rwlock);
rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
if (dn->dn_bonus == NULL)
dbuf_create_bonus(dn);
}
db = dn->dn_bonus;
/* as long as the bonus buf is held, the dnode will be held */
if (zfs_refcount_add(&db->db_holds, tag) == 1) {
VERIFY(dnode_add_ref(dn, db));
atomic_inc_32(&dn->dn_dbufs_count);
}
/*
* Wait to drop dn_struct_rwlock until after adding the bonus dbuf's
* hold and incrementing the dbuf count to ensure that dnode_move() sees
* a dnode hold for every dbuf.
*/
rw_exit(&dn->dn_struct_rwlock);
error = dbuf_read(db, NULL, db_flags);
if (error) {
dnode_evict_bonus(dn);
dbuf_rele(db, tag);
*dbp = NULL;
return (error);
}
*dbp = &db->db;
return (0);
}
int
dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp)
{
dnode_t *dn;
int error;
error = dnode_hold(os, object, FTAG, &dn);
if (error)
return (error);
error = dmu_bonus_hold_by_dnode(dn, tag, dbp, DMU_READ_NO_PREFETCH);
dnode_rele(dn, FTAG);
return (error);
}
/*
* returns ENOENT, EIO, or 0.
*
* This interface will allocate a blank spill dbuf when a spill blk
* doesn't already exist on the dnode.
*
* if you only want to find an already existing spill db, then
* dmu_spill_hold_existing() should be used.
*/
int
dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, void *tag, dmu_buf_t **dbp)
{
dmu_buf_impl_t *db = NULL;
int err;
if ((flags & DB_RF_HAVESTRUCT) == 0)
rw_enter(&dn->dn_struct_rwlock, RW_READER);
db = dbuf_hold(dn, DMU_SPILL_BLKID, tag);
if ((flags & DB_RF_HAVESTRUCT) == 0)
rw_exit(&dn->dn_struct_rwlock);
if (db == NULL) {
*dbp = NULL;
return (SET_ERROR(EIO));
}
err = dbuf_read(db, NULL, flags);
if (err == 0)
*dbp = &db->db;
else {
dbuf_rele(db, tag);
*dbp = NULL;
}
return (err);
}
int
dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)bonus;
dnode_t *dn;
int err;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_SA) {
err = SET_ERROR(EINVAL);
} else {
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (!dn->dn_have_spill) {
err = SET_ERROR(ENOENT);
} else {
err = dmu_spill_hold_by_dnode(dn,
DB_RF_HAVESTRUCT | DB_RF_CANFAIL, tag, dbp);
}
rw_exit(&dn->dn_struct_rwlock);
}
DB_DNODE_EXIT(db);
return (err);
}
int
dmu_spill_hold_by_bonus(dmu_buf_t *bonus, uint32_t flags, void *tag,
dmu_buf_t **dbp)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)bonus;
dnode_t *dn;
int err;
uint32_t db_flags = DB_RF_CANFAIL;
if (flags & DMU_READ_NO_DECRYPT)
db_flags |= DB_RF_NO_DECRYPT;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
err = dmu_spill_hold_by_dnode(dn, db_flags, tag, dbp);
DB_DNODE_EXIT(db);
return (err);
}
/*
* Note: longer-term, we should modify all of the dmu_buf_*() interfaces
* to take a held dnode rather than <os, object> -- the lookup is wasteful,
* and can induce severe lock contention when writing to several files
* whose dnodes are in the same block.
*/
int
dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length,
boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp, uint32_t flags)
{
dmu_buf_t **dbp;
zstream_t *zs = NULL;
uint64_t blkid, nblks, i;
uint32_t dbuf_flags;
int err;
zio_t *zio = NULL;
boolean_t missed = B_FALSE;
ASSERT(length <= DMU_MAX_ACCESS);
/*
* Note: We directly notify the prefetch code of this read, so that
* we can tell it about the multi-block read. dbuf_read() only knows
* about the one block it is accessing.
*/
dbuf_flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT | DB_RF_HAVESTRUCT |
DB_RF_NOPREFETCH;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (dn->dn_datablkshift) {
int blkshift = dn->dn_datablkshift;
nblks = (P2ROUNDUP(offset + length, 1ULL << blkshift) -
P2ALIGN(offset, 1ULL << blkshift)) >> blkshift;
} else {
if (offset + length > dn->dn_datablksz) {
zfs_panic_recover("zfs: accessing past end of object "
"%llx/%llx (size=%u access=%llu+%llu)",
(longlong_t)dn->dn_objset->
os_dsl_dataset->ds_object,
(longlong_t)dn->dn_object, dn->dn_datablksz,
(longlong_t)offset, (longlong_t)length);
rw_exit(&dn->dn_struct_rwlock);
return (SET_ERROR(EIO));
}
nblks = 1;
}
dbp = kmem_zalloc(sizeof (dmu_buf_t *) * nblks, KM_SLEEP);
if (read)
zio = zio_root(dn->dn_objset->os_spa, NULL, NULL,
ZIO_FLAG_CANFAIL);
blkid = dbuf_whichblock(dn, 0, offset);
if ((flags & DMU_READ_NO_PREFETCH) == 0 &&
DNODE_META_IS_CACHEABLE(dn) && length <= zfetch_array_rd_sz) {
/*
* Prepare the zfetch before initiating the demand reads, so
* that if multiple threads block on same indirect block, we
* base predictions on the original less racy request order.
*/
zs = dmu_zfetch_prepare(&dn->dn_zfetch, blkid, nblks,
read && DNODE_IS_CACHEABLE(dn), B_TRUE);
}
for (i = 0; i < nblks; i++) {
dmu_buf_impl_t *db = dbuf_hold(dn, blkid + i, tag);
if (db == NULL) {
if (zs)
dmu_zfetch_run(zs, missed, B_TRUE);
rw_exit(&dn->dn_struct_rwlock);
dmu_buf_rele_array(dbp, nblks, tag);
if (read)
zio_nowait(zio);
return (SET_ERROR(EIO));
}
/*
* Initiate async demand data read.
* We check the db_state after calling dbuf_read() because
* (1) dbuf_read() may change the state to CACHED due to a
* hit in the ARC, and (2) on a cache miss, a child will
* have been added to "zio" but not yet completed, so the
* state will not yet be CACHED.
*/
if (read) {
(void) dbuf_read(db, zio, dbuf_flags);
if (db->db_state != DB_CACHED)
missed = B_TRUE;
}
dbp[i] = &db->db;
}
if (!read)
zfs_racct_write(length, nblks);
if (zs)
dmu_zfetch_run(zs, missed, B_TRUE);
rw_exit(&dn->dn_struct_rwlock);
if (read) {
/* wait for async read i/o */
err = zio_wait(zio);
if (err) {
dmu_buf_rele_array(dbp, nblks, tag);
return (err);
}
/* wait for other io to complete */
for (i = 0; i < nblks; i++) {
dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbp[i];
mutex_enter(&db->db_mtx);
while (db->db_state == DB_READ ||
db->db_state == DB_FILL)
cv_wait(&db->db_changed, &db->db_mtx);
if (db->db_state == DB_UNCACHED)
err = SET_ERROR(EIO);
mutex_exit(&db->db_mtx);
if (err) {
dmu_buf_rele_array(dbp, nblks, tag);
return (err);
}
}
}
*numbufsp = nblks;
*dbpp = dbp;
return (0);
}
int
dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
{
dnode_t *dn;
int err;
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
numbufsp, dbpp, DMU_READ_PREFETCH);
dnode_rele(dn, FTAG);
return (err);
}
int
dmu_buf_hold_array_by_bonus(dmu_buf_t *db_fake, uint64_t offset,
uint64_t length, boolean_t read, void *tag, int *numbufsp,
dmu_buf_t ***dbpp)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dnode_t *dn;
int err;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
numbufsp, dbpp, DMU_READ_PREFETCH);
DB_DNODE_EXIT(db);
return (err);
}
void
dmu_buf_rele_array(dmu_buf_t **dbp_fake, int numbufs, void *tag)
{
int i;
dmu_buf_impl_t **dbp = (dmu_buf_impl_t **)dbp_fake;
if (numbufs == 0)
return;
for (i = 0; i < numbufs; i++) {
if (dbp[i])
dbuf_rele(dbp[i], tag);
}
kmem_free(dbp, sizeof (dmu_buf_t *) * numbufs);
}
/*
* Issue prefetch i/os for the given blocks. If level is greater than 0, the
* indirect blocks prefetched will be those that point to the blocks containing
* the data starting at offset, and continuing to offset + len.
*
* Note that if the indirect blocks above the blocks being prefetched are not
* in cache, they will be asynchronously read in.
*/
void
dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
uint64_t len, zio_priority_t pri)
{
dnode_t *dn;
uint64_t blkid;
int nblks, err;
if (len == 0) { /* they're interested in the bonus buffer */
dn = DMU_META_DNODE(os);
if (object == 0 || object >= DN_MAX_OBJECT)
return;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
blkid = dbuf_whichblock(dn, level,
object * sizeof (dnode_phys_t));
dbuf_prefetch(dn, level, blkid, pri, 0);
rw_exit(&dn->dn_struct_rwlock);
return;
}
/*
* See comment before the definition of dmu_prefetch_max.
*/
len = MIN(len, dmu_prefetch_max);
/*
* XXX - Note, if the dnode for the requested object is not
* already cached, we will do a *synchronous* read in the
* dnode_hold() call. The same is true for any indirects.
*/
err = dnode_hold(os, object, FTAG, &dn);
if (err != 0)
return;
/*
* offset + len - 1 is the last byte we want to prefetch for, and offset
* is the first. Then dbuf_whichblk(dn, level, off + len - 1) is the
* last block we want to prefetch, and dbuf_whichblock(dn, level,
* offset) is the first. Then the number we need to prefetch is the
* last - first + 1.
*/
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (level > 0 || dn->dn_datablkshift != 0) {
nblks = dbuf_whichblock(dn, level, offset + len - 1) -
dbuf_whichblock(dn, level, offset) + 1;
} else {
nblks = (offset < dn->dn_datablksz);
}
if (nblks != 0) {
blkid = dbuf_whichblock(dn, level, offset);
for (int i = 0; i < nblks; i++)
dbuf_prefetch(dn, level, blkid + i, pri, 0);
}
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
}
/*
* Get the next "chunk" of file data to free. We traverse the file from
* the end so that the file gets shorter over time (if we crashes in the
* middle, this will leave us in a better state). We find allocated file
* data by simply searching the allocated level 1 indirects.
*
* On input, *start should be the first offset that does not need to be
* freed (e.g. "offset + length"). On return, *start will be the first
* offset that should be freed and l1blks is set to the number of level 1
* indirect blocks found within the chunk.
*/
static int
get_next_chunk(dnode_t *dn, uint64_t *start, uint64_t minimum, uint64_t *l1blks)
{
uint64_t blks;
uint64_t maxblks = DMU_MAX_ACCESS >> (dn->dn_indblkshift + 1);
/* bytes of data covered by a level-1 indirect block */
uint64_t iblkrange = (uint64_t)dn->dn_datablksz *
EPB(dn->dn_indblkshift, SPA_BLKPTRSHIFT);
ASSERT3U(minimum, <=, *start);
/*
* Check if we can free the entire range assuming that all of the
* L1 blocks in this range have data. If we can, we use this
* worst case value as an estimate so we can avoid having to look
* at the object's actual data.
*/
uint64_t total_l1blks =
(roundup(*start, iblkrange) - (minimum / iblkrange * iblkrange)) /
iblkrange;
if (total_l1blks <= maxblks) {
*l1blks = total_l1blks;
*start = minimum;
return (0);
}
ASSERT(ISP2(iblkrange));
for (blks = 0; *start > minimum && blks < maxblks; blks++) {
int err;
/*
* dnode_next_offset(BACKWARDS) will find an allocated L1
* indirect block at or before the input offset. We must
* decrement *start so that it is at the end of the region
* to search.
*/
(*start)--;
err = dnode_next_offset(dn,
DNODE_FIND_BACKWARDS, start, 2, 1, 0);
/* if there are no indirect blocks before start, we are done */
if (err == ESRCH) {
*start = minimum;
break;
} else if (err != 0) {
*l1blks = blks;
return (err);
}
/* set start to the beginning of this L1 indirect */
*start = P2ALIGN(*start, iblkrange);
}
if (*start < minimum)
*start = minimum;
*l1blks = blks;
return (0);
}
/*
* If this objset is of type OST_ZFS return true if vfs's unmounted flag is set,
* otherwise return false.
* Used below in dmu_free_long_range_impl() to enable abort when unmounting
*/
static boolean_t
dmu_objset_zfs_unmounting(objset_t *os)
{
#ifdef _KERNEL
if (dmu_objset_type(os) == DMU_OST_ZFS)
return (zfs_get_vfs_flag_unmounted(os));
#else
(void) os;
#endif
return (B_FALSE);
}
static int
dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
uint64_t length)
{
uint64_t object_size;
int err;
uint64_t dirty_frees_threshold;
dsl_pool_t *dp = dmu_objset_pool(os);
if (dn == NULL)
return (SET_ERROR(EINVAL));
object_size = (dn->dn_maxblkid + 1) * dn->dn_datablksz;
if (offset >= object_size)
return (0);
if (zfs_per_txg_dirty_frees_percent <= 100)
dirty_frees_threshold =
zfs_per_txg_dirty_frees_percent * zfs_dirty_data_max / 100;
else
dirty_frees_threshold = zfs_dirty_data_max / 20;
if (length == DMU_OBJECT_END || offset + length > object_size)
length = object_size - offset;
while (length != 0) {
uint64_t chunk_end, chunk_begin, chunk_len;
uint64_t l1blks;
dmu_tx_t *tx;
if (dmu_objset_zfs_unmounting(dn->dn_objset))
return (SET_ERROR(EINTR));
chunk_end = chunk_begin = offset + length;
/* move chunk_begin backwards to the beginning of this chunk */
err = get_next_chunk(dn, &chunk_begin, offset, &l1blks);
if (err)
return (err);
ASSERT3U(chunk_begin, >=, offset);
ASSERT3U(chunk_begin, <=, chunk_end);
chunk_len = chunk_end - chunk_begin;
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, dn->dn_object, chunk_begin, chunk_len);
/*
* Mark this transaction as typically resulting in a net
* reduction in space used.
*/
dmu_tx_mark_netfree(tx);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err) {
dmu_tx_abort(tx);
return (err);
}
uint64_t txg = dmu_tx_get_txg(tx);
mutex_enter(&dp->dp_lock);
uint64_t long_free_dirty =
dp->dp_long_free_dirty_pertxg[txg & TXG_MASK];
mutex_exit(&dp->dp_lock);
/*
* To avoid filling up a TXG with just frees, wait for
* the next TXG to open before freeing more chunks if
* we have reached the threshold of frees.
*/
if (dirty_frees_threshold != 0 &&
long_free_dirty >= dirty_frees_threshold) {
DMU_TX_STAT_BUMP(dmu_tx_dirty_frees_delay);
dmu_tx_commit(tx);
txg_wait_open(dp, 0, B_TRUE);
continue;
}
/*
* In order to prevent unnecessary write throttling, for each
* TXG, we track the cumulative size of L1 blocks being dirtied
* in dnode_free_range() below. We compare this number to a
* tunable threshold, past which we prevent new L1 dirty freeing
* blocks from being added into the open TXG. See
* dmu_free_long_range_impl() for details. The threshold
* prevents write throttle activation due to dirty freeing L1
* blocks taking up a large percentage of zfs_dirty_data_max.
*/
mutex_enter(&dp->dp_lock);
dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] +=
l1blks << dn->dn_indblkshift;
mutex_exit(&dp->dp_lock);
DTRACE_PROBE3(free__long__range,
uint64_t, long_free_dirty, uint64_t, chunk_len,
uint64_t, txg);
dnode_free_range(dn, chunk_begin, chunk_len, tx);
dmu_tx_commit(tx);
length -= chunk_len;
}
return (0);
}
int
dmu_free_long_range(objset_t *os, uint64_t object,
uint64_t offset, uint64_t length)
{
dnode_t *dn;
int err;
err = dnode_hold(os, object, FTAG, &dn);
if (err != 0)
return (err);
err = dmu_free_long_range_impl(os, dn, offset, length);
/*
* It is important to zero out the maxblkid when freeing the entire
* file, so that (a) subsequent calls to dmu_free_long_range_impl()
* will take the fast path, and (b) dnode_reallocate() can verify
* that the entire file has been freed.
*/
if (err == 0 && offset == 0 && length == DMU_OBJECT_END)
dn->dn_maxblkid = 0;
dnode_rele(dn, FTAG);
return (err);
}
int
dmu_free_long_object(objset_t *os, uint64_t object)
{
dmu_tx_t *tx;
int err;
err = dmu_free_long_range(os, object, 0, DMU_OBJECT_END);
if (err != 0)
return (err);
tx = dmu_tx_create(os);
dmu_tx_hold_bonus(tx, object);
dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END);
dmu_tx_mark_netfree(tx);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err == 0) {
err = dmu_object_free(os, object, tx);
dmu_tx_commit(tx);
} else {
dmu_tx_abort(tx);
}
return (err);
}
int
dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
uint64_t size, dmu_tx_t *tx)
{
dnode_t *dn;
int err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
ASSERT(offset < UINT64_MAX);
ASSERT(size == DMU_OBJECT_END || size <= UINT64_MAX - offset);
dnode_free_range(dn, offset, size, tx);
dnode_rele(dn, FTAG);
return (0);
}
static int
dmu_read_impl(dnode_t *dn, uint64_t offset, uint64_t size,
void *buf, uint32_t flags)
{
dmu_buf_t **dbp;
int numbufs, err = 0;
/*
* Deal with odd block sizes, where there can't be data past the first
* block. If we ever do the tail block optimization, we will need to
* handle that here as well.
*/
if (dn->dn_maxblkid == 0) {
uint64_t newsz = offset > dn->dn_datablksz ? 0 :
MIN(size, dn->dn_datablksz - offset);
bzero((char *)buf + newsz, size - newsz);
size = newsz;
}
while (size > 0) {
uint64_t mylen = MIN(size, DMU_MAX_ACCESS / 2);
int i;
/*
* NB: we could do this block-at-a-time, but it's nice
* to be reading in parallel.
*/
err = dmu_buf_hold_array_by_dnode(dn, offset, mylen,
TRUE, FTAG, &numbufs, &dbp, flags);
if (err)
break;
for (i = 0; i < numbufs; i++) {
uint64_t tocpy;
int64_t bufoff;
dmu_buf_t *db = dbp[i];
ASSERT(size > 0);
bufoff = offset - db->db_offset;
tocpy = MIN(db->db_size - bufoff, size);
(void) memcpy(buf, (char *)db->db_data + bufoff, tocpy);
offset += tocpy;
size -= tocpy;
buf = (char *)buf + tocpy;
}
dmu_buf_rele_array(dbp, numbufs, FTAG);
}
return (err);
}
int
dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
void *buf, uint32_t flags)
{
dnode_t *dn;
int err;
err = dnode_hold(os, object, FTAG, &dn);
if (err != 0)
return (err);
err = dmu_read_impl(dn, offset, size, buf, flags);
dnode_rele(dn, FTAG);
return (err);
}
int
dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf,
uint32_t flags)
{
return (dmu_read_impl(dn, offset, size, buf, flags));
}
static void
dmu_write_impl(dmu_buf_t **dbp, int numbufs, uint64_t offset, uint64_t size,
const void *buf, dmu_tx_t *tx)
{
int i;
for (i = 0; i < numbufs; i++) {
uint64_t tocpy;
int64_t bufoff;
dmu_buf_t *db = dbp[i];
ASSERT(size > 0);
bufoff = offset - db->db_offset;
tocpy = MIN(db->db_size - bufoff, size);
ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
if (tocpy == db->db_size)
dmu_buf_will_fill(db, tx);
else
dmu_buf_will_dirty(db, tx);
(void) memcpy((char *)db->db_data + bufoff, buf, tocpy);
if (tocpy == db->db_size)
dmu_buf_fill_done(db, tx);
offset += tocpy;
size -= tocpy;
buf = (char *)buf + tocpy;
}
}
void
dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
const void *buf, dmu_tx_t *tx)
{
dmu_buf_t **dbp;
int numbufs;
if (size == 0)
return;
VERIFY0(dmu_buf_hold_array(os, object, offset, size,
FALSE, FTAG, &numbufs, &dbp));
dmu_write_impl(dbp, numbufs, offset, size, buf, tx);
dmu_buf_rele_array(dbp, numbufs, FTAG);
}
/*
* Note: Lustre is an external consumer of this interface.
*/
void
dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size,
const void *buf, dmu_tx_t *tx)
{
dmu_buf_t **dbp;
int numbufs;
if (size == 0)
return;
VERIFY0(dmu_buf_hold_array_by_dnode(dn, offset, size,
FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH));
dmu_write_impl(dbp, numbufs, offset, size, buf, tx);
dmu_buf_rele_array(dbp, numbufs, FTAG);
}
void
dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
dmu_tx_t *tx)
{
dmu_buf_t **dbp;
int numbufs, i;
if (size == 0)
return;
VERIFY(0 == dmu_buf_hold_array(os, object, offset, size,
FALSE, FTAG, &numbufs, &dbp));
for (i = 0; i < numbufs; i++) {
dmu_buf_t *db = dbp[i];
dmu_buf_will_not_fill(db, tx);
}
dmu_buf_rele_array(dbp, numbufs, FTAG);
}
void
dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
int compressed_size, int byteorder, dmu_tx_t *tx)
{
dmu_buf_t *db;
ASSERT3U(etype, <, NUM_BP_EMBEDDED_TYPES);
ASSERT3U(comp, <, ZIO_COMPRESS_FUNCTIONS);
VERIFY0(dmu_buf_hold_noread(os, object, offset,
FTAG, &db));
dmu_buf_write_embedded(db,
data, (bp_embedded_type_t)etype, (enum zio_compress)comp,
uncompressed_size, compressed_size, byteorder, tx);
dmu_buf_rele(db, FTAG);
}
void
dmu_redact(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
dmu_tx_t *tx)
{
int numbufs, i;
dmu_buf_t **dbp;
VERIFY0(dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG,
&numbufs, &dbp));
for (i = 0; i < numbufs; i++)
dmu_buf_redact(dbp[i], tx);
dmu_buf_rele_array(dbp, numbufs, FTAG);
}
#ifdef _KERNEL
int
dmu_read_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size)
{
dmu_buf_t **dbp;
int numbufs, i, err;
/*
* NB: we could do this block-at-a-time, but it's nice
* to be reading in parallel.
*/
err = dmu_buf_hold_array_by_dnode(dn, zfs_uio_offset(uio), size,
TRUE, FTAG, &numbufs, &dbp, 0);
if (err)
return (err);
for (i = 0; i < numbufs; i++) {
uint64_t tocpy;
int64_t bufoff;
dmu_buf_t *db = dbp[i];
ASSERT(size > 0);
bufoff = zfs_uio_offset(uio) - db->db_offset;
tocpy = MIN(db->db_size - bufoff, size);
err = zfs_uio_fault_move((char *)db->db_data + bufoff, tocpy,
UIO_READ, uio);
if (err)
break;
size -= tocpy;
}
dmu_buf_rele_array(dbp, numbufs, FTAG);
return (err);
}
/*
* Read 'size' bytes into the uio buffer.
* From object zdb->db_object.
* Starting at zfs_uio_offset(uio).
*
* If the caller already has a dbuf in the target object
* (e.g. its bonus buffer), this routine is faster than dmu_read_uio(),
* because we don't have to find the dnode_t for the object.
*/
int
dmu_read_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)zdb;
dnode_t *dn;
int err;
if (size == 0)
return (0);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
err = dmu_read_uio_dnode(dn, uio, size);
DB_DNODE_EXIT(db);
return (err);
}
/*
* Read 'size' bytes into the uio buffer.
* From the specified object
* Starting at offset zfs_uio_offset(uio).
*/
int
dmu_read_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size)
{
dnode_t *dn;
int err;
if (size == 0)
return (0);
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
err = dmu_read_uio_dnode(dn, uio, size);
dnode_rele(dn, FTAG);
return (err);
}
int
dmu_write_uio_dnode(dnode_t *dn, zfs_uio_t *uio, uint64_t size, dmu_tx_t *tx)
{
dmu_buf_t **dbp;
int numbufs;
int err = 0;
int i;
err = dmu_buf_hold_array_by_dnode(dn, zfs_uio_offset(uio), size,
FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH);
if (err)
return (err);
for (i = 0; i < numbufs; i++) {
uint64_t tocpy;
int64_t bufoff;
dmu_buf_t *db = dbp[i];
ASSERT(size > 0);
bufoff = zfs_uio_offset(uio) - db->db_offset;
tocpy = MIN(db->db_size - bufoff, size);
ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
if (tocpy == db->db_size)
dmu_buf_will_fill(db, tx);
else
dmu_buf_will_dirty(db, tx);
/*
* XXX zfs_uiomove could block forever (eg.nfs-backed
* pages). There needs to be a uiolockdown() function
* to lock the pages in memory, so that zfs_uiomove won't
* block.
*/
err = zfs_uio_fault_move((char *)db->db_data + bufoff,
tocpy, UIO_WRITE, uio);
if (tocpy == db->db_size)
dmu_buf_fill_done(db, tx);
if (err)
break;
size -= tocpy;
}
dmu_buf_rele_array(dbp, numbufs, FTAG);
return (err);
}
/*
* Write 'size' bytes from the uio buffer.
* To object zdb->db_object.
* Starting at offset zfs_uio_offset(uio).
*
* If the caller already has a dbuf in the target object
* (e.g. its bonus buffer), this routine is faster than dmu_write_uio(),
* because we don't have to find the dnode_t for the object.
*/
int
dmu_write_uio_dbuf(dmu_buf_t *zdb, zfs_uio_t *uio, uint64_t size,
dmu_tx_t *tx)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)zdb;
dnode_t *dn;
int err;
if (size == 0)
return (0);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
err = dmu_write_uio_dnode(dn, uio, size, tx);
DB_DNODE_EXIT(db);
return (err);
}
/*
* Write 'size' bytes from the uio buffer.
* To the specified object.
* Starting at offset zfs_uio_offset(uio).
*/
int
dmu_write_uio(objset_t *os, uint64_t object, zfs_uio_t *uio, uint64_t size,
dmu_tx_t *tx)
{
dnode_t *dn;
int err;
if (size == 0)
return (0);
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
err = dmu_write_uio_dnode(dn, uio, size, tx);
dnode_rele(dn, FTAG);
return (err);
}
#endif /* _KERNEL */
/*
* Allocate a loaned anonymous arc buffer.
*/
arc_buf_t *
dmu_request_arcbuf(dmu_buf_t *handle, int size)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)handle;
return (arc_loan_buf(db->db_objset->os_spa, B_FALSE, size));
}
/*
* Free a loaned arc buffer.
*/
void
dmu_return_arcbuf(arc_buf_t *buf)
{
arc_return_buf(buf, FTAG);
arc_buf_destroy(buf, FTAG);
}
/*
* A "lightweight" write is faster than a regular write (e.g.
* dmu_write_by_dnode() or dmu_assign_arcbuf_by_dnode()), because it avoids the
* CPU cost of creating a dmu_buf_impl_t and arc_buf_[hdr_]_t. However, the
* data can not be read or overwritten until the transaction's txg has been
* synced. This makes it appropriate for workloads that are known to be
* (temporarily) write-only, like "zfs receive".
*
* A single block is written, starting at the specified offset in bytes. If
* the call is successful, it returns 0 and the provided abd has been
* consumed (the caller should not free it).
*/
int
dmu_lightweight_write_by_dnode(dnode_t *dn, uint64_t offset, abd_t *abd,
const zio_prop_t *zp, enum zio_flag flags, dmu_tx_t *tx)
{
dbuf_dirty_record_t *dr =
dbuf_dirty_lightweight(dn, dbuf_whichblock(dn, 0, offset), tx);
if (dr == NULL)
return (SET_ERROR(EIO));
dr->dt.dll.dr_abd = abd;
dr->dt.dll.dr_props = *zp;
dr->dt.dll.dr_flags = flags;
return (0);
}
/*
* When possible directly assign passed loaned arc buffer to a dbuf.
* If this is not possible copy the contents of passed arc buf via
* dmu_write().
*/
int
dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset, arc_buf_t *buf,
dmu_tx_t *tx)
{
dmu_buf_impl_t *db;
objset_t *os = dn->dn_objset;
uint64_t object = dn->dn_object;
uint32_t blksz = (uint32_t)arc_buf_lsize(buf);
uint64_t blkid;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
blkid = dbuf_whichblock(dn, 0, offset);
db = dbuf_hold(dn, blkid, FTAG);
if (db == NULL)
return (SET_ERROR(EIO));
rw_exit(&dn->dn_struct_rwlock);
/*
* We can only assign if the offset is aligned and the arc buf is the
* same size as the dbuf.
*/
if (offset == db->db.db_offset && blksz == db->db.db_size) {
zfs_racct_write(blksz, 1);
dbuf_assign_arcbuf(db, buf, tx);
dbuf_rele(db, FTAG);
} else {
/* compressed bufs must always be assignable to their dbuf */
ASSERT3U(arc_get_compression(buf), ==, ZIO_COMPRESS_OFF);
ASSERT(!(buf->b_flags & ARC_BUF_FLAG_COMPRESSED));
dbuf_rele(db, FTAG);
dmu_write(os, object, offset, blksz, buf->b_data, tx);
dmu_return_arcbuf(buf);
}
return (0);
}
int
dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset, arc_buf_t *buf,
dmu_tx_t *tx)
{
int err;
dmu_buf_impl_t *dbuf = (dmu_buf_impl_t *)handle;
DB_DNODE_ENTER(dbuf);
err = dmu_assign_arcbuf_by_dnode(DB_DNODE(dbuf), offset, buf, tx);
DB_DNODE_EXIT(dbuf);
return (err);
}
typedef struct {
dbuf_dirty_record_t *dsa_dr;
dmu_sync_cb_t *dsa_done;
zgd_t *dsa_zgd;
dmu_tx_t *dsa_tx;
} dmu_sync_arg_t;
static void
dmu_sync_ready(zio_t *zio, arc_buf_t *buf, void *varg)
{
(void) buf;
dmu_sync_arg_t *dsa = varg;
dmu_buf_t *db = dsa->dsa_zgd->zgd_db;
blkptr_t *bp = zio->io_bp;
if (zio->io_error == 0) {
if (BP_IS_HOLE(bp)) {
/*
* A block of zeros may compress to a hole, but the
* block size still needs to be known for replay.
*/
BP_SET_LSIZE(bp, db->db_size);
} else if (!BP_IS_EMBEDDED(bp)) {
ASSERT(BP_GET_LEVEL(bp) == 0);
BP_SET_FILL(bp, 1);
}
}
}
static void
dmu_sync_late_arrival_ready(zio_t *zio)
{
dmu_sync_ready(zio, NULL, zio->io_private);
}
static void
dmu_sync_done(zio_t *zio, arc_buf_t *buf, void *varg)
{
(void) buf;
dmu_sync_arg_t *dsa = varg;
dbuf_dirty_record_t *dr = dsa->dsa_dr;
dmu_buf_impl_t *db = dr->dr_dbuf;
zgd_t *zgd = dsa->dsa_zgd;
/*
* Record the vdev(s) backing this blkptr so they can be flushed after
* the writes for the lwb have completed.
*/
if (zio->io_error == 0) {
zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp);
}
mutex_enter(&db->db_mtx);
ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC);
if (zio->io_error == 0) {
dr->dt.dl.dr_nopwrite = !!(zio->io_flags & ZIO_FLAG_NOPWRITE);
if (dr->dt.dl.dr_nopwrite) {
blkptr_t *bp = zio->io_bp;
blkptr_t *bp_orig = &zio->io_bp_orig;
uint8_t chksum = BP_GET_CHECKSUM(bp_orig);
ASSERT(BP_EQUAL(bp, bp_orig));
VERIFY(BP_EQUAL(bp, db->db_blkptr));
ASSERT(zio->io_prop.zp_compress != ZIO_COMPRESS_OFF);
VERIFY(zio_checksum_table[chksum].ci_flags &
ZCHECKSUM_FLAG_NOPWRITE);
}
dr->dt.dl.dr_overridden_by = *zio->io_bp;
dr->dt.dl.dr_override_state = DR_OVERRIDDEN;
dr->dt.dl.dr_copies = zio->io_prop.zp_copies;
/*
* Old style holes are filled with all zeros, whereas
* new-style holes maintain their lsize, type, level,
* and birth time (see zio_write_compress). While we
* need to reset the BP_SET_LSIZE() call that happened
* in dmu_sync_ready for old style holes, we do *not*
* want to wipe out the information contained in new
* style holes. Thus, only zero out the block pointer if
* it's an old style hole.
*/
if (BP_IS_HOLE(&dr->dt.dl.dr_overridden_by) &&
dr->dt.dl.dr_overridden_by.blk_birth == 0)
BP_ZERO(&dr->dt.dl.dr_overridden_by);
} else {
dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
}
cv_broadcast(&db->db_changed);
mutex_exit(&db->db_mtx);
dsa->dsa_done(dsa->dsa_zgd, zio->io_error);
kmem_free(dsa, sizeof (*dsa));
}
static void
dmu_sync_late_arrival_done(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
dmu_sync_arg_t *dsa = zio->io_private;
zgd_t *zgd = dsa->dsa_zgd;
if (zio->io_error == 0) {
/*
* Record the vdev(s) backing this blkptr so they can be
* flushed after the writes for the lwb have completed.
*/
zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp);
if (!BP_IS_HOLE(bp)) {
blkptr_t *bp_orig __maybe_unused = &zio->io_bp_orig;
ASSERT(!(zio->io_flags & ZIO_FLAG_NOPWRITE));
ASSERT(BP_IS_HOLE(bp_orig) || !BP_EQUAL(bp, bp_orig));
ASSERT(zio->io_bp->blk_birth == zio->io_txg);
ASSERT(zio->io_txg > spa_syncing_txg(zio->io_spa));
zio_free(zio->io_spa, zio->io_txg, zio->io_bp);
}
}
dmu_tx_commit(dsa->dsa_tx);
dsa->dsa_done(dsa->dsa_zgd, zio->io_error);
abd_free(zio->io_abd);
kmem_free(dsa, sizeof (*dsa));
}
static int
dmu_sync_late_arrival(zio_t *pio, objset_t *os, dmu_sync_cb_t *done, zgd_t *zgd,
zio_prop_t *zp, zbookmark_phys_t *zb)
{
dmu_sync_arg_t *dsa;
dmu_tx_t *tx;
tx = dmu_tx_create(os);
dmu_tx_hold_space(tx, zgd->zgd_db->db_size);
if (dmu_tx_assign(tx, TXG_WAIT) != 0) {
dmu_tx_abort(tx);
/* Make zl_get_data do txg_waited_synced() */
return (SET_ERROR(EIO));
}
/*
* In order to prevent the zgd's lwb from being free'd prior to
* dmu_sync_late_arrival_done() being called, we have to ensure
* the lwb's "max txg" takes this tx's txg into account.
*/
zil_lwb_add_txg(zgd->zgd_lwb, dmu_tx_get_txg(tx));
dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
dsa->dsa_dr = NULL;
dsa->dsa_done = done;
dsa->dsa_zgd = zgd;
dsa->dsa_tx = tx;
/*
* Since we are currently syncing this txg, it's nontrivial to
* determine what BP to nopwrite against, so we disable nopwrite.
*
* When syncing, the db_blkptr is initially the BP of the previous
* txg. We can not nopwrite against it because it will be changed
* (this is similar to the non-late-arrival case where the dbuf is
* dirty in a future txg).
*
* Then dbuf_write_ready() sets bp_blkptr to the location we will write.
* We can not nopwrite against it because although the BP will not
* (typically) be changed, the data has not yet been persisted to this
* location.
*
* Finally, when dbuf_write_done() is called, it is theoretically
* possible to always nopwrite, because the data that was written in
* this txg is the same data that we are trying to write. However we
* would need to check that this dbuf is not dirty in any future
* txg's (as we do in the normal dmu_sync() path). For simplicity, we
* don't nopwrite in this case.
*/
zp->zp_nopwrite = B_FALSE;
zio_nowait(zio_write(pio, os->os_spa, dmu_tx_get_txg(tx), zgd->zgd_bp,
abd_get_from_buf(zgd->zgd_db->db_data, zgd->zgd_db->db_size),
zgd->zgd_db->db_size, zgd->zgd_db->db_size, zp,
dmu_sync_late_arrival_ready, NULL, NULL, dmu_sync_late_arrival_done,
dsa, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, zb));
return (0);
}
/*
* Intent log support: sync the block associated with db to disk.
* N.B. and XXX: the caller is responsible for making sure that the
* data isn't changing while dmu_sync() is writing it.
*
* Return values:
*
* EEXIST: this txg has already been synced, so there's nothing to do.
* The caller should not log the write.
*
* ENOENT: the block was dbuf_free_range()'d, so there's nothing to do.
* The caller should not log the write.
*
* EALREADY: this block is already in the process of being synced.
* The caller should track its progress (somehow).
*
* EIO: could not do the I/O.
* The caller should do a txg_wait_synced().
*
* 0: the I/O has been initiated.
* The caller should log this blkptr in the done callback.
* It is possible that the I/O will fail, in which case
* the error will be reported to the done callback and
* propagated to pio from zio_done().
*/
int
dmu_sync(zio_t *pio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)zgd->zgd_db;
objset_t *os = db->db_objset;
dsl_dataset_t *ds = os->os_dsl_dataset;
dbuf_dirty_record_t *dr, *dr_next;
dmu_sync_arg_t *dsa;
zbookmark_phys_t zb;
zio_prop_t zp;
dnode_t *dn;
ASSERT(pio != NULL);
ASSERT(txg != 0);
SET_BOOKMARK(&zb, ds->ds_object,
db->db.db_object, db->db_level, db->db_blkid);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
dmu_write_policy(os, dn, db->db_level, WP_DMU_SYNC, &zp);
DB_DNODE_EXIT(db);
/*
* If we're frozen (running ziltest), we always need to generate a bp.
*/
if (txg > spa_freeze_txg(os->os_spa))
return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb));
/*
* Grabbing db_mtx now provides a barrier between dbuf_sync_leaf()
* and us. If we determine that this txg is not yet syncing,
* but it begins to sync a moment later, that's OK because the
* sync thread will block in dbuf_sync_leaf() until we drop db_mtx.
*/
mutex_enter(&db->db_mtx);
if (txg <= spa_last_synced_txg(os->os_spa)) {
/*
* This txg has already synced. There's nothing to do.
*/
mutex_exit(&db->db_mtx);
return (SET_ERROR(EEXIST));
}
if (txg <= spa_syncing_txg(os->os_spa)) {
/*
* This txg is currently syncing, so we can't mess with
* the dirty record anymore; just write a new log block.
*/
mutex_exit(&db->db_mtx);
return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb));
}
dr = dbuf_find_dirty_eq(db, txg);
if (dr == NULL) {
/*
* There's no dr for this dbuf, so it must have been freed.
* There's no need to log writes to freed blocks, so we're done.
*/
mutex_exit(&db->db_mtx);
return (SET_ERROR(ENOENT));
}
dr_next = list_next(&db->db_dirty_records, dr);
ASSERT(dr_next == NULL || dr_next->dr_txg < txg);
if (db->db_blkptr != NULL) {
/*
* We need to fill in zgd_bp with the current blkptr so that
* the nopwrite code can check if we're writing the same
* data that's already on disk. We can only nopwrite if we
* are sure that after making the copy, db_blkptr will not
* change until our i/o completes. We ensure this by
* holding the db_mtx, and only allowing nopwrite if the
* block is not already dirty (see below). This is verified
* by dmu_sync_done(), which VERIFYs that the db_blkptr has
* not changed.
*/
*zgd->zgd_bp = *db->db_blkptr;
}
/*
* Assume the on-disk data is X, the current syncing data (in
* txg - 1) is Y, and the current in-memory data is Z (currently
* in dmu_sync).
*
* We usually want to perform a nopwrite if X and Z are the
* same. However, if Y is different (i.e. the BP is going to
* change before this write takes effect), then a nopwrite will
* be incorrect - we would override with X, which could have
* been freed when Y was written.
*
* (Note that this is not a concern when we are nop-writing from
* syncing context, because X and Y must be identical, because
* all previous txgs have been synced.)
*
* Therefore, we disable nopwrite if the current BP could change
* before this TXG. There are two ways it could change: by
* being dirty (dr_next is non-NULL), or by being freed
* (dnode_block_freed()). This behavior is verified by
* zio_done(), which VERIFYs that the override BP is identical
* to the on-disk BP.
*/
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
if (dr_next != NULL || dnode_block_freed(dn, db->db_blkid))
zp.zp_nopwrite = B_FALSE;
DB_DNODE_EXIT(db);
ASSERT(dr->dr_txg == txg);
if (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC ||
dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
/*
* We have already issued a sync write for this buffer,
* or this buffer has already been synced. It could not
* have been dirtied since, or we would have cleared the state.
*/
mutex_exit(&db->db_mtx);
return (SET_ERROR(EALREADY));
}
ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN);
dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC;
mutex_exit(&db->db_mtx);
dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
dsa->dsa_dr = dr;
dsa->dsa_done = done;
dsa->dsa_zgd = zgd;
dsa->dsa_tx = NULL;
zio_nowait(arc_write(pio, os->os_spa, txg,
zgd->zgd_bp, dr->dt.dl.dr_data, dbuf_is_l2cacheable(db),
&zp, dmu_sync_ready, NULL, NULL, dmu_sync_done, dsa,
ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, &zb));
return (0);
}
int
dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels, dmu_tx_t *tx)
{
dnode_t *dn;
int err;
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
err = dnode_set_nlevels(dn, nlevels, tx);
dnode_rele(dn, FTAG);
return (err);
}
int
dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs,
dmu_tx_t *tx)
{
dnode_t *dn;
int err;
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
err = dnode_set_blksz(dn, size, ibs, tx);
dnode_rele(dn, FTAG);
return (err);
}
int
dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid,
dmu_tx_t *tx)
{
dnode_t *dn;
int err;
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
dnode_new_blkid(dn, maxblkid, tx, B_FALSE, B_TRUE);
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
return (0);
}
void
dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
dmu_tx_t *tx)
{
dnode_t *dn;
/*
* Send streams include each object's checksum function. This
* check ensures that the receiving system can understand the
* checksum function transmitted.
*/
ASSERT3U(checksum, <, ZIO_CHECKSUM_LEGACY_FUNCTIONS);
VERIFY0(dnode_hold(os, object, FTAG, &dn));
ASSERT3U(checksum, <, ZIO_CHECKSUM_FUNCTIONS);
dn->dn_checksum = checksum;
dnode_setdirty(dn, tx);
dnode_rele(dn, FTAG);
}
void
dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
dmu_tx_t *tx)
{
dnode_t *dn;
/*
* Send streams include each object's compression function. This
* check ensures that the receiving system can understand the
* compression function transmitted.
*/
ASSERT3U(compress, <, ZIO_COMPRESS_LEGACY_FUNCTIONS);
VERIFY0(dnode_hold(os, object, FTAG, &dn));
dn->dn_compress = compress;
dnode_setdirty(dn, tx);
dnode_rele(dn, FTAG);
}
/*
* When the "redundant_metadata" property is set to "most", only indirect
* blocks of this level and higher will have an additional ditto block.
*/
-int zfs_redundant_metadata_most_ditto_level = 2;
+static const int zfs_redundant_metadata_most_ditto_level = 2;
void
dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp)
{
dmu_object_type_t type = dn ? dn->dn_type : DMU_OT_OBJSET;
boolean_t ismd = (level > 0 || DMU_OT_IS_METADATA(type) ||
(wp & WP_SPILL));
enum zio_checksum checksum = os->os_checksum;
enum zio_compress compress = os->os_compress;
uint8_t complevel = os->os_complevel;
enum zio_checksum dedup_checksum = os->os_dedup_checksum;
boolean_t dedup = B_FALSE;
boolean_t nopwrite = B_FALSE;
boolean_t dedup_verify = os->os_dedup_verify;
boolean_t encrypt = B_FALSE;
int copies = os->os_copies;
/*
* We maintain different write policies for each of the following
* types of data:
* 1. metadata
* 2. preallocated blocks (i.e. level-0 blocks of a dump device)
* 3. all other level 0 blocks
*/
if (ismd) {
/*
* XXX -- we should design a compression algorithm
* that specializes in arrays of bps.
*/
compress = zio_compress_select(os->os_spa,
ZIO_COMPRESS_ON, ZIO_COMPRESS_ON);
/*
* Metadata always gets checksummed. If the data
* checksum is multi-bit correctable, and it's not a
* ZBT-style checksum, then it's suitable for metadata
* as well. Otherwise, the metadata checksum defaults
* to fletcher4.
*/
if (!(zio_checksum_table[checksum].ci_flags &
ZCHECKSUM_FLAG_METADATA) ||
(zio_checksum_table[checksum].ci_flags &
ZCHECKSUM_FLAG_EMBEDDED))
checksum = ZIO_CHECKSUM_FLETCHER_4;
if (os->os_redundant_metadata == ZFS_REDUNDANT_METADATA_ALL ||
(os->os_redundant_metadata ==
ZFS_REDUNDANT_METADATA_MOST &&
(level >= zfs_redundant_metadata_most_ditto_level ||
DMU_OT_IS_METADATA(type) || (wp & WP_SPILL))))
copies++;
} else if (wp & WP_NOFILL) {
ASSERT(level == 0);
/*
* If we're writing preallocated blocks, we aren't actually
* writing them so don't set any policy properties. These
* blocks are currently only used by an external subsystem
* outside of zfs (i.e. dump) and not written by the zio
* pipeline.
*/
compress = ZIO_COMPRESS_OFF;
checksum = ZIO_CHECKSUM_OFF;
} else {
compress = zio_compress_select(os->os_spa, dn->dn_compress,
compress);
complevel = zio_complevel_select(os->os_spa, compress,
complevel, complevel);
checksum = (dedup_checksum == ZIO_CHECKSUM_OFF) ?
zio_checksum_select(dn->dn_checksum, checksum) :
dedup_checksum;
/*
* Determine dedup setting. If we are in dmu_sync(),
* we won't actually dedup now because that's all
* done in syncing context; but we do want to use the
* dedup checksum. If the checksum is not strong
* enough to ensure unique signatures, force
* dedup_verify.
*/
if (dedup_checksum != ZIO_CHECKSUM_OFF) {
dedup = (wp & WP_DMU_SYNC) ? B_FALSE : B_TRUE;
if (!(zio_checksum_table[checksum].ci_flags &
ZCHECKSUM_FLAG_DEDUP))
dedup_verify = B_TRUE;
}
/*
* Enable nopwrite if we have secure enough checksum
* algorithm (see comment in zio_nop_write) and
* compression is enabled. We don't enable nopwrite if
* dedup is enabled as the two features are mutually
* exclusive.
*/
nopwrite = (!dedup && (zio_checksum_table[checksum].ci_flags &
ZCHECKSUM_FLAG_NOPWRITE) &&
compress != ZIO_COMPRESS_OFF && zfs_nopwrite_enabled);
}
/*
* All objects in an encrypted objset are protected from modification
* via a MAC. Encrypted objects store their IV and salt in the last DVA
* in the bp, so we cannot use all copies. Encrypted objects are also
* not subject to nopwrite since writing the same data will still
* result in a new ciphertext. Only encrypted blocks can be dedup'd
* to avoid ambiguity in the dedup code since the DDT does not store
* object types.
*/
if (os->os_encrypted && (wp & WP_NOFILL) == 0) {
encrypt = B_TRUE;
if (DMU_OT_IS_ENCRYPTED(type)) {
copies = MIN(copies, SPA_DVAS_PER_BP - 1);
nopwrite = B_FALSE;
} else {
dedup = B_FALSE;
}
if (level <= 0 &&
(type == DMU_OT_DNODE || type == DMU_OT_OBJSET)) {
compress = ZIO_COMPRESS_EMPTY;
}
}
zp->zp_compress = compress;
zp->zp_complevel = complevel;
zp->zp_checksum = checksum;
zp->zp_type = (wp & WP_SPILL) ? dn->dn_bonustype : type;
zp->zp_level = level;
zp->zp_copies = MIN(copies, spa_max_replication(os->os_spa));
zp->zp_dedup = dedup;
zp->zp_dedup_verify = dedup && dedup_verify;
zp->zp_nopwrite = nopwrite;
zp->zp_encrypt = encrypt;
zp->zp_byteorder = ZFS_HOST_BYTEORDER;
bzero(zp->zp_salt, ZIO_DATA_SALT_LEN);
bzero(zp->zp_iv, ZIO_DATA_IV_LEN);
bzero(zp->zp_mac, ZIO_DATA_MAC_LEN);
zp->zp_zpl_smallblk = DMU_OT_IS_FILE(zp->zp_type) ?
os->os_zpl_special_smallblock : 0;
ASSERT3U(zp->zp_compress, !=, ZIO_COMPRESS_INHERIT);
}
/*
* This function is only called from zfs_holey_common() for zpl_llseek()
* in order to determine the location of holes. In order to accurately
* report holes all dirty data must be synced to disk. This causes extremely
* poor performance when seeking for holes in a dirty file. As a compromise,
* only provide hole data when the dnode is clean. When a dnode is dirty
* report the dnode as having no holes which is always a safe thing to do.
*/
int
dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off)
{
dnode_t *dn;
int err;
restart:
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (dnode_is_dirty(dn)) {
/*
* If the zfs_dmu_offset_next_sync module option is enabled
* then strict hole reporting has been requested. Dirty
* dnodes must be synced to disk to accurately report all
* holes. When disabled dirty dnodes are reported to not
* have any holes which is always safe.
*
* When called by zfs_holey_common() the zp->z_rangelock
* is held to prevent zfs_write() and mmap writeback from
* re-dirtying the dnode after txg_wait_synced().
*/
if (zfs_dmu_offset_next_sync) {
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
txg_wait_synced(dmu_objset_pool(os), 0);
goto restart;
}
err = SET_ERROR(EBUSY);
} else {
err = dnode_next_offset(dn, DNODE_FIND_HAVELOCK |
(hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0);
}
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
return (err);
}
void
__dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi)
{
dnode_phys_t *dnp = dn->dn_phys;
doi->doi_data_block_size = dn->dn_datablksz;
doi->doi_metadata_block_size = dn->dn_indblkshift ?
1ULL << dn->dn_indblkshift : 0;
doi->doi_type = dn->dn_type;
doi->doi_bonus_type = dn->dn_bonustype;
doi->doi_bonus_size = dn->dn_bonuslen;
doi->doi_dnodesize = dn->dn_num_slots << DNODE_SHIFT;
doi->doi_indirection = dn->dn_nlevels;
doi->doi_checksum = dn->dn_checksum;
doi->doi_compress = dn->dn_compress;
doi->doi_nblkptr = dn->dn_nblkptr;
doi->doi_physical_blocks_512 = (DN_USED_BYTES(dnp) + 256) >> 9;
doi->doi_max_offset = (dn->dn_maxblkid + 1) * dn->dn_datablksz;
doi->doi_fill_count = 0;
for (int i = 0; i < dnp->dn_nblkptr; i++)
doi->doi_fill_count += BP_GET_FILL(&dnp->dn_blkptr[i]);
}
void
dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi)
{
rw_enter(&dn->dn_struct_rwlock, RW_READER);
mutex_enter(&dn->dn_mtx);
__dmu_object_info_from_dnode(dn, doi);
mutex_exit(&dn->dn_mtx);
rw_exit(&dn->dn_struct_rwlock);
}
/*
* Get information on a DMU object.
* If doi is NULL, just indicates whether the object exists.
*/
int
dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi)
{
dnode_t *dn;
int err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
if (doi != NULL)
dmu_object_info_from_dnode(dn, doi);
dnode_rele(dn, FTAG);
return (0);
}
/*
* As above, but faster; can be used when you have a held dbuf in hand.
*/
void
dmu_object_info_from_db(dmu_buf_t *db_fake, dmu_object_info_t *doi)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
DB_DNODE_ENTER(db);
dmu_object_info_from_dnode(DB_DNODE(db), doi);
DB_DNODE_EXIT(db);
}
/*
* Faster still when you only care about the size.
*/
void
dmu_object_size_from_db(dmu_buf_t *db_fake, uint32_t *blksize,
u_longlong_t *nblk512)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dnode_t *dn;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
*blksize = dn->dn_datablksz;
/* add in number of slots used for the dnode itself */
*nblk512 = ((DN_USED_BYTES(dn->dn_phys) + SPA_MINBLOCKSIZE/2) >>
SPA_MINBLOCKSHIFT) + dn->dn_num_slots;
DB_DNODE_EXIT(db);
}
void
dmu_object_dnsize_from_db(dmu_buf_t *db_fake, int *dnsize)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dnode_t *dn;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
*dnsize = dn->dn_num_slots << DNODE_SHIFT;
DB_DNODE_EXIT(db);
}
void
byteswap_uint64_array(void *vbuf, size_t size)
{
uint64_t *buf = vbuf;
size_t count = size >> 3;
int i;
ASSERT((size & 7) == 0);
for (i = 0; i < count; i++)
buf[i] = BSWAP_64(buf[i]);
}
void
byteswap_uint32_array(void *vbuf, size_t size)
{
uint32_t *buf = vbuf;
size_t count = size >> 2;
int i;
ASSERT((size & 3) == 0);
for (i = 0; i < count; i++)
buf[i] = BSWAP_32(buf[i]);
}
void
byteswap_uint16_array(void *vbuf, size_t size)
{
uint16_t *buf = vbuf;
size_t count = size >> 1;
int i;
ASSERT((size & 1) == 0);
for (i = 0; i < count; i++)
buf[i] = BSWAP_16(buf[i]);
}
void
byteswap_uint8_array(void *vbuf, size_t size)
{
(void) vbuf, (void) size;
}
void
dmu_init(void)
{
abd_init();
zfs_dbgmsg_init();
sa_cache_init();
dmu_objset_init();
dnode_init();
zfetch_init();
dmu_tx_init();
l2arc_init();
arc_init();
dbuf_init();
}
void
dmu_fini(void)
{
arc_fini(); /* arc depends on l2arc, so arc must go first */
l2arc_fini();
dmu_tx_fini();
zfetch_fini();
dbuf_fini();
dnode_fini();
dmu_objset_fini();
sa_cache_fini();
zfs_dbgmsg_fini();
abd_fini();
}
EXPORT_SYMBOL(dmu_bonus_hold);
EXPORT_SYMBOL(dmu_bonus_hold_by_dnode);
EXPORT_SYMBOL(dmu_buf_hold_array_by_bonus);
EXPORT_SYMBOL(dmu_buf_rele_array);
EXPORT_SYMBOL(dmu_prefetch);
EXPORT_SYMBOL(dmu_free_range);
EXPORT_SYMBOL(dmu_free_long_range);
EXPORT_SYMBOL(dmu_free_long_object);
EXPORT_SYMBOL(dmu_read);
EXPORT_SYMBOL(dmu_read_by_dnode);
EXPORT_SYMBOL(dmu_write);
EXPORT_SYMBOL(dmu_write_by_dnode);
EXPORT_SYMBOL(dmu_prealloc);
EXPORT_SYMBOL(dmu_object_info);
EXPORT_SYMBOL(dmu_object_info_from_dnode);
EXPORT_SYMBOL(dmu_object_info_from_db);
EXPORT_SYMBOL(dmu_object_size_from_db);
EXPORT_SYMBOL(dmu_object_dnsize_from_db);
EXPORT_SYMBOL(dmu_object_set_nlevels);
EXPORT_SYMBOL(dmu_object_set_blocksize);
EXPORT_SYMBOL(dmu_object_set_maxblkid);
EXPORT_SYMBOL(dmu_object_set_checksum);
EXPORT_SYMBOL(dmu_object_set_compress);
EXPORT_SYMBOL(dmu_offset_next);
EXPORT_SYMBOL(dmu_write_policy);
EXPORT_SYMBOL(dmu_sync);
EXPORT_SYMBOL(dmu_request_arcbuf);
EXPORT_SYMBOL(dmu_return_arcbuf);
EXPORT_SYMBOL(dmu_assign_arcbuf_by_dnode);
EXPORT_SYMBOL(dmu_assign_arcbuf_by_dbuf);
EXPORT_SYMBOL(dmu_buf_hold);
EXPORT_SYMBOL(dmu_ot);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, nopwrite_enabled, INT, ZMOD_RW,
"Enable NOP writes");
ZFS_MODULE_PARAM(zfs, zfs_, per_txg_dirty_frees_percent, ULONG, ZMOD_RW,
"Percentage of dirtied blocks from frees in one TXG");
ZFS_MODULE_PARAM(zfs, zfs_, dmu_offset_next_sync, INT, ZMOD_RW,
"Enable forcing txg sync to find holes");
ZFS_MODULE_PARAM(zfs, , dmu_prefetch_max, INT, ZMOD_RW,
"Limit one prefetch call to this size");
/* END CSTYLED */
diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c
index a8975797e8af..9a74fa9ce360 100644
--- a/module/zfs/dmu_objset.c
+++ b/module/zfs/dmu_objset.c
@@ -1,3078 +1,3078 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright (c) 2015, STRATO AG, Inc. All rights reserved.
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
* Copyright 2017 Nexenta Systems, Inc.
* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
*/
/* Portions Copyright 2010 Robert Milkowski */
#include <sys/cred.h>
#include <sys/zfs_context.h>
#include <sys/dmu_objset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_synctask.h>
#include <sys/dsl_deleg.h>
#include <sys/dnode.h>
#include <sys/dbuf.h>
#include <sys/zvol.h>
#include <sys/dmu_tx.h>
#include <sys/zap.h>
#include <sys/zil.h>
#include <sys/dmu_impl.h>
#include <sys/zfs_ioctl.h>
#include <sys/sa.h>
#include <sys/zfs_onexit.h>
#include <sys/dsl_destroy.h>
#include <sys/vdev.h>
#include <sys/zfeature.h>
#include <sys/policy.h>
#include <sys/spa_impl.h>
#include <sys/dmu_recv.h>
#include <sys/zfs_project.h>
#include "zfs_namecheck.h"
#include <sys/vdev_impl.h>
#include <sys/arc.h>
/*
* Needed to close a window in dnode_move() that allows the objset to be freed
* before it can be safely accessed.
*/
krwlock_t os_lock;
/*
* Tunable to overwrite the maximum number of threads for the parallelization
* of dmu_objset_find_dp, needed to speed up the import of pools with many
* datasets.
* Default is 4 times the number of leaf vdevs.
*/
-int dmu_find_threads = 0;
+static const int dmu_find_threads = 0;
/*
* Backfill lower metadnode objects after this many have been freed.
* Backfilling negatively impacts object creation rates, so only do it
* if there are enough holes to fill.
*/
-int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT;
+static const int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT;
-static char *upgrade_tag = "upgrade_tag";
+static const char *upgrade_tag = "upgrade_tag";
static void dmu_objset_find_dp_cb(void *arg);
static void dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb);
static void dmu_objset_upgrade_stop(objset_t *os);
void
dmu_objset_init(void)
{
rw_init(&os_lock, NULL, RW_DEFAULT, NULL);
}
void
dmu_objset_fini(void)
{
rw_destroy(&os_lock);
}
spa_t *
dmu_objset_spa(objset_t *os)
{
return (os->os_spa);
}
zilog_t *
dmu_objset_zil(objset_t *os)
{
return (os->os_zil);
}
dsl_pool_t *
dmu_objset_pool(objset_t *os)
{
dsl_dataset_t *ds;
if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir)
return (ds->ds_dir->dd_pool);
else
return (spa_get_dsl(os->os_spa));
}
dsl_dataset_t *
dmu_objset_ds(objset_t *os)
{
return (os->os_dsl_dataset);
}
dmu_objset_type_t
dmu_objset_type(objset_t *os)
{
return (os->os_phys->os_type);
}
void
dmu_objset_name(objset_t *os, char *buf)
{
dsl_dataset_name(os->os_dsl_dataset, buf);
}
uint64_t
dmu_objset_id(objset_t *os)
{
dsl_dataset_t *ds = os->os_dsl_dataset;
return (ds ? ds->ds_object : 0);
}
uint64_t
dmu_objset_dnodesize(objset_t *os)
{
return (os->os_dnodesize);
}
zfs_sync_type_t
dmu_objset_syncprop(objset_t *os)
{
return (os->os_sync);
}
zfs_logbias_op_t
dmu_objset_logbias(objset_t *os)
{
return (os->os_logbias);
}
static void
checksum_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
/*
* Inheritance should have been done by now.
*/
ASSERT(newval != ZIO_CHECKSUM_INHERIT);
os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE);
}
static void
compression_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
/*
* Inheritance and range checking should have been done by now.
*/
ASSERT(newval != ZIO_COMPRESS_INHERIT);
os->os_compress = zio_compress_select(os->os_spa,
ZIO_COMPRESS_ALGO(newval), ZIO_COMPRESS_ON);
os->os_complevel = zio_complevel_select(os->os_spa, os->os_compress,
ZIO_COMPRESS_LEVEL(newval), ZIO_COMPLEVEL_DEFAULT);
}
static void
copies_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
/*
* Inheritance and range checking should have been done by now.
*/
ASSERT(newval > 0);
ASSERT(newval <= spa_max_replication(os->os_spa));
os->os_copies = newval;
}
static void
dedup_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
spa_t *spa = os->os_spa;
enum zio_checksum checksum;
/*
* Inheritance should have been done by now.
*/
ASSERT(newval != ZIO_CHECKSUM_INHERIT);
checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF);
os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK;
os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY);
}
static void
primary_cache_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
/*
* Inheritance and range checking should have been done by now.
*/
ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE ||
newval == ZFS_CACHE_METADATA);
os->os_primary_cache = newval;
}
static void
secondary_cache_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
/*
* Inheritance and range checking should have been done by now.
*/
ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE ||
newval == ZFS_CACHE_METADATA);
os->os_secondary_cache = newval;
}
static void
sync_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
/*
* Inheritance and range checking should have been done by now.
*/
ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS ||
newval == ZFS_SYNC_DISABLED);
os->os_sync = newval;
if (os->os_zil)
zil_set_sync(os->os_zil, newval);
}
static void
redundant_metadata_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
/*
* Inheritance and range checking should have been done by now.
*/
ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL ||
newval == ZFS_REDUNDANT_METADATA_MOST);
os->os_redundant_metadata = newval;
}
static void
dnodesize_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
switch (newval) {
case ZFS_DNSIZE_LEGACY:
os->os_dnodesize = DNODE_MIN_SIZE;
break;
case ZFS_DNSIZE_AUTO:
/*
* Choose a dnode size that will work well for most
* workloads if the user specified "auto". Future code
* improvements could dynamically select a dnode size
* based on observed workload patterns.
*/
os->os_dnodesize = DNODE_MIN_SIZE * 2;
break;
case ZFS_DNSIZE_1K:
case ZFS_DNSIZE_2K:
case ZFS_DNSIZE_4K:
case ZFS_DNSIZE_8K:
case ZFS_DNSIZE_16K:
os->os_dnodesize = newval;
break;
}
}
static void
smallblk_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
/*
* Inheritance and range checking should have been done by now.
*/
ASSERT(newval <= SPA_MAXBLOCKSIZE);
ASSERT(ISP2(newval));
os->os_zpl_special_smallblock = newval;
}
static void
logbias_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
ASSERT(newval == ZFS_LOGBIAS_LATENCY ||
newval == ZFS_LOGBIAS_THROUGHPUT);
os->os_logbias = newval;
if (os->os_zil)
zil_set_logbias(os->os_zil, newval);
}
static void
recordsize_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
os->os_recordsize = newval;
}
void
dmu_objset_byteswap(void *buf, size_t size)
{
objset_phys_t *osp = buf;
ASSERT(size == OBJSET_PHYS_SIZE_V1 || size == OBJSET_PHYS_SIZE_V2 ||
size == sizeof (objset_phys_t));
dnode_byteswap(&osp->os_meta_dnode);
byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t));
osp->os_type = BSWAP_64(osp->os_type);
osp->os_flags = BSWAP_64(osp->os_flags);
if (size >= OBJSET_PHYS_SIZE_V2) {
dnode_byteswap(&osp->os_userused_dnode);
dnode_byteswap(&osp->os_groupused_dnode);
if (size >= sizeof (objset_phys_t))
dnode_byteswap(&osp->os_projectused_dnode);
}
}
/*
* The hash is a CRC-based hash of the objset_t pointer and the object number.
*/
static uint64_t
dnode_hash(const objset_t *os, uint64_t obj)
{
uintptr_t osv = (uintptr_t)os;
uint64_t crc = -1ULL;
ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY);
/*
* The low 6 bits of the pointer don't have much entropy, because
* the objset_t is larger than 2^6 bytes long.
*/
crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (osv >> 6)) & 0xFF];
crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 0)) & 0xFF];
crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 8)) & 0xFF];
crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 16)) & 0xFF];
crc ^= (osv>>14) ^ (obj>>24);
return (crc);
}
static unsigned int
dnode_multilist_index_func(multilist_t *ml, void *obj)
{
dnode_t *dn = obj;
/*
* The low order bits of the hash value are thought to be
* distributed evenly. Otherwise, in the case that the multilist
* has a power of two number of sublists, each sublists' usage
* would not be evenly distributed. In this context full 64bit
* division would be a waste of time, so limit it to 32 bits.
*/
return ((unsigned int)dnode_hash(dn->dn_objset, dn->dn_object) %
multilist_get_num_sublists(ml));
}
static inline boolean_t
dmu_os_is_l2cacheable(objset_t *os)
{
vdev_t *vd = NULL;
zfs_cache_type_t cache = os->os_secondary_cache;
blkptr_t *bp = os->os_rootbp;
if (bp != NULL && !BP_IS_HOLE(bp)) {
uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
vdev_t *rvd = os->os_spa->spa_root_vdev;
if (vdev < rvd->vdev_children)
vd = rvd->vdev_child[vdev];
if (cache == ZFS_CACHE_ALL || cache == ZFS_CACHE_METADATA) {
if (vd == NULL)
return (B_TRUE);
if ((vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) ||
l2arc_exclude_special == 0)
return (B_TRUE);
}
}
return (B_FALSE);
}
/*
* Instantiates the objset_t in-memory structure corresponding to the
* objset_phys_t that's pointed to by the specified blkptr_t.
*/
int
dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
objset_t **osp)
{
objset_t *os;
int i, err;
ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock));
ASSERT(!BP_IS_REDACTED(bp));
/*
* We need the pool config lock to get properties.
*/
ASSERT(ds == NULL || dsl_pool_config_held(ds->ds_dir->dd_pool));
/*
* The $ORIGIN dataset (if it exists) doesn't have an associated
* objset, so there's no reason to open it. The $ORIGIN dataset
* will not exist on pools older than SPA_VERSION_ORIGIN.
*/
if (ds != NULL && spa_get_dsl(spa) != NULL &&
spa_get_dsl(spa)->dp_origin_snap != NULL) {
ASSERT3P(ds->ds_dir, !=,
spa_get_dsl(spa)->dp_origin_snap->ds_dir);
}
os = kmem_zalloc(sizeof (objset_t), KM_SLEEP);
os->os_dsl_dataset = ds;
os->os_spa = spa;
os->os_rootbp = bp;
if (!BP_IS_HOLE(os->os_rootbp)) {
arc_flags_t aflags = ARC_FLAG_WAIT;
zbookmark_phys_t zb;
int size;
enum zio_flag zio_flags = ZIO_FLAG_CANFAIL;
SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET,
ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
if (dmu_os_is_l2cacheable(os))
aflags |= ARC_FLAG_L2CACHE;
if (ds != NULL && ds->ds_dir->dd_crypto_obj != 0) {
ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
ASSERT(BP_IS_AUTHENTICATED(bp));
zio_flags |= ZIO_FLAG_RAW;
}
dprintf_bp(os->os_rootbp, "reading %s", "");
err = arc_read(NULL, spa, os->os_rootbp,
arc_getbuf_func, &os->os_phys_buf,
ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb);
if (err != 0) {
kmem_free(os, sizeof (objset_t));
/* convert checksum errors into IO errors */
if (err == ECKSUM)
err = SET_ERROR(EIO);
return (err);
}
if (spa_version(spa) < SPA_VERSION_USERSPACE)
size = OBJSET_PHYS_SIZE_V1;
else if (!spa_feature_is_enabled(spa,
SPA_FEATURE_PROJECT_QUOTA))
size = OBJSET_PHYS_SIZE_V2;
else
size = sizeof (objset_phys_t);
/* Increase the blocksize if we are permitted. */
if (arc_buf_size(os->os_phys_buf) < size) {
arc_buf_t *buf = arc_alloc_buf(spa, &os->os_phys_buf,
ARC_BUFC_METADATA, size);
bzero(buf->b_data, size);
bcopy(os->os_phys_buf->b_data, buf->b_data,
arc_buf_size(os->os_phys_buf));
arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf);
os->os_phys_buf = buf;
}
os->os_phys = os->os_phys_buf->b_data;
os->os_flags = os->os_phys->os_flags;
} else {
int size = spa_version(spa) >= SPA_VERSION_USERSPACE ?
sizeof (objset_phys_t) : OBJSET_PHYS_SIZE_V1;
os->os_phys_buf = arc_alloc_buf(spa, &os->os_phys_buf,
ARC_BUFC_METADATA, size);
os->os_phys = os->os_phys_buf->b_data;
bzero(os->os_phys, size);
}
/*
* These properties will be filled in by the logic in zfs_get_zplprop()
* when they are queried for the first time.
*/
os->os_version = OBJSET_PROP_UNINITIALIZED;
os->os_normalization = OBJSET_PROP_UNINITIALIZED;
os->os_utf8only = OBJSET_PROP_UNINITIALIZED;
os->os_casesensitivity = OBJSET_PROP_UNINITIALIZED;
/*
* Note: the changed_cb will be called once before the register
* func returns, thus changing the checksum/compression from the
* default (fletcher2/off). Snapshots don't need to know about
* checksum/compression/copies.
*/
if (ds != NULL) {
os->os_encrypted = (ds->ds_dir->dd_crypto_obj != 0);
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE),
primary_cache_changed_cb, os);
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE),
secondary_cache_changed_cb, os);
}
if (!ds->ds_is_snapshot) {
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_CHECKSUM),
checksum_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_COMPRESSION),
compression_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_COPIES),
copies_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_DEDUP),
dedup_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_LOGBIAS),
logbias_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_SYNC),
sync_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(
ZFS_PROP_REDUNDANT_METADATA),
redundant_metadata_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
recordsize_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_DNODESIZE),
dnodesize_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(
ZFS_PROP_SPECIAL_SMALL_BLOCKS),
smallblk_changed_cb, os);
}
}
if (err != 0) {
arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf);
kmem_free(os, sizeof (objset_t));
return (err);
}
} else {
/* It's the meta-objset. */
os->os_checksum = ZIO_CHECKSUM_FLETCHER_4;
os->os_compress = ZIO_COMPRESS_ON;
os->os_complevel = ZIO_COMPLEVEL_DEFAULT;
os->os_encrypted = B_FALSE;
os->os_copies = spa_max_replication(spa);
os->os_dedup_checksum = ZIO_CHECKSUM_OFF;
os->os_dedup_verify = B_FALSE;
os->os_logbias = ZFS_LOGBIAS_LATENCY;
os->os_sync = ZFS_SYNC_STANDARD;
os->os_primary_cache = ZFS_CACHE_ALL;
os->os_secondary_cache = ZFS_CACHE_ALL;
os->os_dnodesize = DNODE_MIN_SIZE;
}
if (ds == NULL || !ds->ds_is_snapshot)
os->os_zil_header = os->os_phys->os_zil_header;
os->os_zil = zil_alloc(os, &os->os_zil_header);
for (i = 0; i < TXG_SIZE; i++) {
multilist_create(&os->os_dirty_dnodes[i], sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[i]),
dnode_multilist_index_func);
}
list_create(&os->os_dnodes, sizeof (dnode_t),
offsetof(dnode_t, dn_link));
list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t),
offsetof(dmu_buf_impl_t, db_link));
list_link_init(&os->os_evicting_node);
mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&os->os_userused_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL);
os->os_obj_next_percpu_len = boot_ncpus;
os->os_obj_next_percpu = kmem_zalloc(os->os_obj_next_percpu_len *
sizeof (os->os_obj_next_percpu[0]), KM_SLEEP);
dnode_special_open(os, &os->os_phys->os_meta_dnode,
DMU_META_DNODE_OBJECT, &os->os_meta_dnode);
if (OBJSET_BUF_HAS_USERUSED(os->os_phys_buf)) {
dnode_special_open(os, &os->os_phys->os_userused_dnode,
DMU_USERUSED_OBJECT, &os->os_userused_dnode);
dnode_special_open(os, &os->os_phys->os_groupused_dnode,
DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode);
if (OBJSET_BUF_HAS_PROJECTUSED(os->os_phys_buf))
dnode_special_open(os,
&os->os_phys->os_projectused_dnode,
DMU_PROJECTUSED_OBJECT, &os->os_projectused_dnode);
}
mutex_init(&os->os_upgrade_lock, NULL, MUTEX_DEFAULT, NULL);
*osp = os;
return (0);
}
int
dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp)
{
int err = 0;
/*
* We need the pool_config lock to manipulate the dsl_dataset_t.
* Even if the dataset is long-held, we need the pool_config lock
* to open the objset, as it needs to get properties.
*/
ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
mutex_enter(&ds->ds_opening_lock);
if (ds->ds_objset == NULL) {
objset_t *os;
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
err = dmu_objset_open_impl(dsl_dataset_get_spa(ds),
ds, dsl_dataset_get_blkptr(ds), &os);
rrw_exit(&ds->ds_bp_rwlock, FTAG);
if (err == 0) {
mutex_enter(&ds->ds_lock);
ASSERT(ds->ds_objset == NULL);
ds->ds_objset = os;
mutex_exit(&ds->ds_lock);
}
}
*osp = ds->ds_objset;
mutex_exit(&ds->ds_opening_lock);
return (err);
}
/*
* Holds the pool while the objset is held. Therefore only one objset
* can be held at a time.
*/
int
dmu_objset_hold_flags(const char *name, boolean_t decrypt, void *tag,
objset_t **osp)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
int err;
ds_hold_flags_t flags;
flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
err = dsl_pool_hold(name, tag, &dp);
if (err != 0)
return (err);
err = dsl_dataset_hold_flags(dp, name, flags, tag, &ds);
if (err != 0) {
dsl_pool_rele(dp, tag);
return (err);
}
err = dmu_objset_from_ds(ds, osp);
if (err != 0) {
dsl_dataset_rele(ds, tag);
dsl_pool_rele(dp, tag);
}
return (err);
}
int
dmu_objset_hold(const char *name, void *tag, objset_t **osp)
{
return (dmu_objset_hold_flags(name, B_FALSE, tag, osp));
}
static int
dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type,
boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp)
{
(void) tag;
int err = dmu_objset_from_ds(ds, osp);
if (err != 0) {
return (err);
} else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) {
return (SET_ERROR(EINVAL));
} else if (!readonly && dsl_dataset_is_snapshot(ds)) {
return (SET_ERROR(EROFS));
} else if (!readonly && decrypt &&
dsl_dir_incompatible_encryption_version(ds->ds_dir)) {
return (SET_ERROR(EROFS));
}
/* if we are decrypting, we can now check MACs in os->os_phys_buf */
if (decrypt && arc_is_unauthenticated((*osp)->os_phys_buf)) {
zbookmark_phys_t zb;
SET_BOOKMARK(&zb, ds->ds_object, ZB_ROOT_OBJECT,
ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
err = arc_untransform((*osp)->os_phys_buf, (*osp)->os_spa,
&zb, B_FALSE);
if (err != 0)
return (err);
ASSERT0(arc_is_unauthenticated((*osp)->os_phys_buf));
}
return (0);
}
/*
* dsl_pool must not be held when this is called.
* Upon successful return, there will be a longhold on the dataset,
* and the dsl_pool will not be held.
*/
int
dmu_objset_own(const char *name, dmu_objset_type_t type,
boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
int err;
ds_hold_flags_t flags;
flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
err = dsl_pool_hold(name, FTAG, &dp);
if (err != 0)
return (err);
err = dsl_dataset_own(dp, name, flags, tag, &ds);
if (err != 0) {
dsl_pool_rele(dp, FTAG);
return (err);
}
err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp);
if (err != 0) {
dsl_dataset_disown(ds, flags, tag);
dsl_pool_rele(dp, FTAG);
return (err);
}
/*
* User accounting requires the dataset to be decrypted and rw.
* We also don't begin user accounting during claiming to help
* speed up pool import times and to keep this txg reserved
* completely for recovery work.
*/
if (!readonly && !dp->dp_spa->spa_claiming &&
(ds->ds_dir->dd_crypto_obj == 0 || decrypt)) {
if (dmu_objset_userobjspace_upgradable(*osp) ||
dmu_objset_projectquota_upgradable(*osp)) {
dmu_objset_id_quota_upgrade(*osp);
} else if (dmu_objset_userused_enabled(*osp)) {
dmu_objset_userspace_upgrade(*osp);
}
}
dsl_pool_rele(dp, FTAG);
return (0);
}
int
dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type,
boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp)
{
dsl_dataset_t *ds;
int err;
ds_hold_flags_t flags;
flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
err = dsl_dataset_own_obj(dp, obj, flags, tag, &ds);
if (err != 0)
return (err);
err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp);
if (err != 0) {
dsl_dataset_disown(ds, flags, tag);
return (err);
}
return (0);
}
void
dmu_objset_rele_flags(objset_t *os, boolean_t decrypt, void *tag)
{
ds_hold_flags_t flags;
dsl_pool_t *dp = dmu_objset_pool(os);
flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
dsl_dataset_rele_flags(os->os_dsl_dataset, flags, tag);
dsl_pool_rele(dp, tag);
}
void
dmu_objset_rele(objset_t *os, void *tag)
{
dmu_objset_rele_flags(os, B_FALSE, tag);
}
/*
* When we are called, os MUST refer to an objset associated with a dataset
* that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner
* == tag. We will then release and reacquire ownership of the dataset while
* holding the pool config_rwlock to avoid intervening namespace or ownership
* changes may occur.
*
* This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to
* release the hold on its dataset and acquire a new one on the dataset of the
* same name so that it can be partially torn down and reconstructed.
*/
void
dmu_objset_refresh_ownership(dsl_dataset_t *ds, dsl_dataset_t **newds,
boolean_t decrypt, void *tag)
{
dsl_pool_t *dp;
char name[ZFS_MAX_DATASET_NAME_LEN];
ds_hold_flags_t flags;
flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
VERIFY3P(ds, !=, NULL);
VERIFY3P(ds->ds_owner, ==, tag);
VERIFY(dsl_dataset_long_held(ds));
dsl_dataset_name(ds, name);
dp = ds->ds_dir->dd_pool;
dsl_pool_config_enter(dp, FTAG);
dsl_dataset_disown(ds, flags, tag);
VERIFY0(dsl_dataset_own(dp, name, flags, tag, newds));
dsl_pool_config_exit(dp, FTAG);
}
void
dmu_objset_disown(objset_t *os, boolean_t decrypt, void *tag)
{
ds_hold_flags_t flags;
flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
/*
* Stop upgrading thread
*/
dmu_objset_upgrade_stop(os);
dsl_dataset_disown(os->os_dsl_dataset, flags, tag);
}
void
dmu_objset_evict_dbufs(objset_t *os)
{
dnode_t *dn_marker;
dnode_t *dn;
dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP);
mutex_enter(&os->os_lock);
dn = list_head(&os->os_dnodes);
while (dn != NULL) {
/*
* Skip dnodes without holds. We have to do this dance
* because dnode_add_ref() only works if there is already a
* hold. If the dnode has no holds, then it has no dbufs.
*/
if (dnode_add_ref(dn, FTAG)) {
list_insert_after(&os->os_dnodes, dn, dn_marker);
mutex_exit(&os->os_lock);
dnode_evict_dbufs(dn);
dnode_rele(dn, FTAG);
mutex_enter(&os->os_lock);
dn = list_next(&os->os_dnodes, dn_marker);
list_remove(&os->os_dnodes, dn_marker);
} else {
dn = list_next(&os->os_dnodes, dn);
}
}
mutex_exit(&os->os_lock);
kmem_free(dn_marker, sizeof (dnode_t));
if (DMU_USERUSED_DNODE(os) != NULL) {
if (DMU_PROJECTUSED_DNODE(os) != NULL)
dnode_evict_dbufs(DMU_PROJECTUSED_DNODE(os));
dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os));
dnode_evict_dbufs(DMU_USERUSED_DNODE(os));
}
dnode_evict_dbufs(DMU_META_DNODE(os));
}
/*
* Objset eviction processing is split into into two pieces.
* The first marks the objset as evicting, evicts any dbufs that
* have a refcount of zero, and then queues up the objset for the
* second phase of eviction. Once os->os_dnodes has been cleared by
* dnode_buf_pageout()->dnode_destroy(), the second phase is executed.
* The second phase closes the special dnodes, dequeues the objset from
* the list of those undergoing eviction, and finally frees the objset.
*
* NOTE: Due to asynchronous eviction processing (invocation of
* dnode_buf_pageout()), it is possible for the meta dnode for the
* objset to have no holds even though os->os_dnodes is not empty.
*/
void
dmu_objset_evict(objset_t *os)
{
dsl_dataset_t *ds = os->os_dsl_dataset;
for (int t = 0; t < TXG_SIZE; t++)
ASSERT(!dmu_objset_is_dirty(os, t));
if (ds)
dsl_prop_unregister_all(ds, os);
if (os->os_sa)
sa_tear_down(os);
dmu_objset_evict_dbufs(os);
mutex_enter(&os->os_lock);
spa_evicting_os_register(os->os_spa, os);
if (list_is_empty(&os->os_dnodes)) {
mutex_exit(&os->os_lock);
dmu_objset_evict_done(os);
} else {
mutex_exit(&os->os_lock);
}
}
void
dmu_objset_evict_done(objset_t *os)
{
ASSERT3P(list_head(&os->os_dnodes), ==, NULL);
dnode_special_close(&os->os_meta_dnode);
if (DMU_USERUSED_DNODE(os)) {
if (DMU_PROJECTUSED_DNODE(os))
dnode_special_close(&os->os_projectused_dnode);
dnode_special_close(&os->os_userused_dnode);
dnode_special_close(&os->os_groupused_dnode);
}
zil_free(os->os_zil);
arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf);
/*
* This is a barrier to prevent the objset from going away in
* dnode_move() until we can safely ensure that the objset is still in
* use. We consider the objset valid before the barrier and invalid
* after the barrier.
*/
rw_enter(&os_lock, RW_READER);
rw_exit(&os_lock);
kmem_free(os->os_obj_next_percpu,
os->os_obj_next_percpu_len * sizeof (os->os_obj_next_percpu[0]));
mutex_destroy(&os->os_lock);
mutex_destroy(&os->os_userused_lock);
mutex_destroy(&os->os_obj_lock);
mutex_destroy(&os->os_user_ptr_lock);
mutex_destroy(&os->os_upgrade_lock);
for (int i = 0; i < TXG_SIZE; i++)
multilist_destroy(&os->os_dirty_dnodes[i]);
spa_evicting_os_deregister(os->os_spa, os);
kmem_free(os, sizeof (objset_t));
}
inode_timespec_t
dmu_objset_snap_cmtime(objset_t *os)
{
return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir));
}
objset_t *
dmu_objset_create_impl_dnstats(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
dmu_objset_type_t type, int levels, int blksz, int ibs, dmu_tx_t *tx)
{
objset_t *os;
dnode_t *mdn;
ASSERT(dmu_tx_is_syncing(tx));
if (blksz == 0)
blksz = DNODE_BLOCK_SIZE;
if (ibs == 0)
ibs = DN_MAX_INDBLKSHIFT;
if (ds != NULL)
VERIFY0(dmu_objset_from_ds(ds, &os));
else
VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os));
mdn = DMU_META_DNODE(os);
dnode_allocate(mdn, DMU_OT_DNODE, blksz, ibs, DMU_OT_NONE, 0,
DNODE_MIN_SLOTS, tx);
/*
* We don't want to have to increase the meta-dnode's nlevels
* later, because then we could do it in quiescing context while
* we are also accessing it in open context.
*
* This precaution is not necessary for the MOS (ds == NULL),
* because the MOS is only updated in syncing context.
* This is most fortunate: the MOS is the only objset that
* needs to be synced multiple times as spa_sync() iterates
* to convergence, so minimizing its dn_nlevels matters.
*/
if (ds != NULL) {
if (levels == 0) {
levels = 1;
/*
* Determine the number of levels necessary for the
* meta-dnode to contain DN_MAX_OBJECT dnodes. Note
* that in order to ensure that we do not overflow
* 64 bits, there has to be a nlevels that gives us a
* number of blocks > DN_MAX_OBJECT but < 2^64.
* Therefore, (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)
* (10) must be less than (64 - log2(DN_MAX_OBJECT))
* (16).
*/
while ((uint64_t)mdn->dn_nblkptr <<
(mdn->dn_datablkshift - DNODE_SHIFT + (levels - 1) *
(mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) <
DN_MAX_OBJECT)
levels++;
}
mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] =
mdn->dn_nlevels = levels;
}
ASSERT(type != DMU_OST_NONE);
ASSERT(type != DMU_OST_ANY);
ASSERT(type < DMU_OST_NUMTYPES);
os->os_phys->os_type = type;
/*
* Enable user accounting if it is enabled and this is not an
* encrypted receive.
*/
if (dmu_objset_userused_enabled(os) &&
(!os->os_encrypted || !dmu_objset_is_receiving(os))) {
os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
if (dmu_objset_userobjused_enabled(os)) {
ds->ds_feature_activation[
SPA_FEATURE_USEROBJ_ACCOUNTING] = (void *)B_TRUE;
os->os_phys->os_flags |=
OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE;
}
if (dmu_objset_projectquota_enabled(os)) {
ds->ds_feature_activation[
SPA_FEATURE_PROJECT_QUOTA] = (void *)B_TRUE;
os->os_phys->os_flags |=
OBJSET_FLAG_PROJECTQUOTA_COMPLETE;
}
os->os_flags = os->os_phys->os_flags;
}
dsl_dataset_dirty(ds, tx);
return (os);
}
/* called from dsl for meta-objset */
objset_t *
dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
dmu_objset_type_t type, dmu_tx_t *tx)
{
return (dmu_objset_create_impl_dnstats(spa, ds, bp, type, 0, 0, 0, tx));
}
typedef struct dmu_objset_create_arg {
const char *doca_name;
cred_t *doca_cred;
proc_t *doca_proc;
void (*doca_userfunc)(objset_t *os, void *arg,
cred_t *cr, dmu_tx_t *tx);
void *doca_userarg;
dmu_objset_type_t doca_type;
uint64_t doca_flags;
dsl_crypto_params_t *doca_dcp;
} dmu_objset_create_arg_t;
static int
dmu_objset_create_check(void *arg, dmu_tx_t *tx)
{
dmu_objset_create_arg_t *doca = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *pdd;
dsl_dataset_t *parentds;
objset_t *parentos;
const char *tail;
int error;
if (strchr(doca->doca_name, '@') != NULL)
return (SET_ERROR(EINVAL));
if (strlen(doca->doca_name) >= ZFS_MAX_DATASET_NAME_LEN)
return (SET_ERROR(ENAMETOOLONG));
if (dataset_nestcheck(doca->doca_name) != 0)
return (SET_ERROR(ENAMETOOLONG));
error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail);
if (error != 0)
return (error);
if (tail == NULL) {
dsl_dir_rele(pdd, FTAG);
return (SET_ERROR(EEXIST));
}
error = dmu_objset_create_crypt_check(pdd, doca->doca_dcp, NULL);
if (error != 0) {
dsl_dir_rele(pdd, FTAG);
return (error);
}
error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL,
doca->doca_cred, doca->doca_proc);
if (error != 0) {
dsl_dir_rele(pdd, FTAG);
return (error);
}
/* can't create below anything but filesystems (eg. no ZVOLs) */
error = dsl_dataset_hold_obj(pdd->dd_pool,
dsl_dir_phys(pdd)->dd_head_dataset_obj, FTAG, &parentds);
if (error != 0) {
dsl_dir_rele(pdd, FTAG);
return (error);
}
error = dmu_objset_from_ds(parentds, &parentos);
if (error != 0) {
dsl_dataset_rele(parentds, FTAG);
dsl_dir_rele(pdd, FTAG);
return (error);
}
if (dmu_objset_type(parentos) != DMU_OST_ZFS) {
dsl_dataset_rele(parentds, FTAG);
dsl_dir_rele(pdd, FTAG);
return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
}
dsl_dataset_rele(parentds, FTAG);
dsl_dir_rele(pdd, FTAG);
return (error);
}
static void
dmu_objset_create_sync(void *arg, dmu_tx_t *tx)
{
dmu_objset_create_arg_t *doca = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
spa_t *spa = dp->dp_spa;
dsl_dir_t *pdd;
const char *tail;
dsl_dataset_t *ds;
uint64_t obj;
blkptr_t *bp;
objset_t *os;
zio_t *rzio;
VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail));
obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags,
doca->doca_cred, doca->doca_dcp, tx);
VERIFY0(dsl_dataset_hold_obj_flags(pdd->dd_pool, obj,
DS_HOLD_FLAG_DECRYPT, FTAG, &ds));
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
bp = dsl_dataset_get_blkptr(ds);
os = dmu_objset_create_impl(spa, ds, bp, doca->doca_type, tx);
rrw_exit(&ds->ds_bp_rwlock, FTAG);
if (doca->doca_userfunc != NULL) {
doca->doca_userfunc(os, doca->doca_userarg,
doca->doca_cred, tx);
}
/*
* The doca_userfunc() may write out some data that needs to be
* encrypted if the dataset is encrypted (specifically the root
* directory). This data must be written out before the encryption
* key mapping is removed by dsl_dataset_rele_flags(). Force the
* I/O to occur immediately by invoking the relevant sections of
* dsl_pool_sync().
*/
if (os->os_encrypted) {
dsl_dataset_t *tmpds = NULL;
boolean_t need_sync_done = B_FALSE;
mutex_enter(&ds->ds_lock);
ds->ds_owner = FTAG;
mutex_exit(&ds->ds_lock);
rzio = zio_root(spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds,
tx->tx_txg);
if (tmpds != NULL) {
dsl_dataset_sync(ds, rzio, tx);
need_sync_done = B_TRUE;
}
VERIFY0(zio_wait(rzio));
dmu_objset_sync_done(os, tx);
taskq_wait(dp->dp_sync_taskq);
if (txg_list_member(&dp->dp_dirty_datasets, ds, tx->tx_txg)) {
ASSERT3P(ds->ds_key_mapping, !=, NULL);
key_mapping_rele(spa, ds->ds_key_mapping, ds);
}
rzio = zio_root(spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds,
tx->tx_txg);
if (tmpds != NULL) {
dmu_buf_rele(ds->ds_dbuf, ds);
dsl_dataset_sync(ds, rzio, tx);
}
VERIFY0(zio_wait(rzio));
if (need_sync_done) {
ASSERT3P(ds->ds_key_mapping, !=, NULL);
key_mapping_rele(spa, ds->ds_key_mapping, ds);
dsl_dataset_sync_done(ds, tx);
}
mutex_enter(&ds->ds_lock);
ds->ds_owner = NULL;
mutex_exit(&ds->ds_lock);
}
spa_history_log_internal_ds(ds, "create", tx, " ");
dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG);
dsl_dir_rele(pdd, FTAG);
}
int
dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
dsl_crypto_params_t *dcp, dmu_objset_create_sync_func_t func, void *arg)
{
dmu_objset_create_arg_t doca;
dsl_crypto_params_t tmp_dcp = { 0 };
doca.doca_name = name;
doca.doca_cred = CRED();
doca.doca_proc = curproc;
doca.doca_flags = flags;
doca.doca_userfunc = func;
doca.doca_userarg = arg;
doca.doca_type = type;
/*
* Some callers (mostly for testing) do not provide a dcp on their
* own but various code inside the sync task will require it to be
* allocated. Rather than adding NULL checks throughout this code
* or adding dummy dcp's to all of the callers we simply create a
* dummy one here and use that. This zero dcp will have the same
* effect as asking for inheritance of all encryption params.
*/
doca.doca_dcp = (dcp != NULL) ? dcp : &tmp_dcp;
int rv = dsl_sync_task(name,
dmu_objset_create_check, dmu_objset_create_sync, &doca,
6, ZFS_SPACE_CHECK_NORMAL);
if (rv == 0)
zvol_create_minor(name);
return (rv);
}
typedef struct dmu_objset_clone_arg {
const char *doca_clone;
const char *doca_origin;
cred_t *doca_cred;
proc_t *doca_proc;
} dmu_objset_clone_arg_t;
static int
dmu_objset_clone_check(void *arg, dmu_tx_t *tx)
{
dmu_objset_clone_arg_t *doca = arg;
dsl_dir_t *pdd;
const char *tail;
int error;
dsl_dataset_t *origin;
dsl_pool_t *dp = dmu_tx_pool(tx);
if (strchr(doca->doca_clone, '@') != NULL)
return (SET_ERROR(EINVAL));
if (strlen(doca->doca_clone) >= ZFS_MAX_DATASET_NAME_LEN)
return (SET_ERROR(ENAMETOOLONG));
error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail);
if (error != 0)
return (error);
if (tail == NULL) {
dsl_dir_rele(pdd, FTAG);
return (SET_ERROR(EEXIST));
}
error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL,
doca->doca_cred, doca->doca_proc);
if (error != 0) {
dsl_dir_rele(pdd, FTAG);
return (SET_ERROR(EDQUOT));
}
error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin);
if (error != 0) {
dsl_dir_rele(pdd, FTAG);
return (error);
}
/* You can only clone snapshots, not the head datasets. */
if (!origin->ds_is_snapshot) {
dsl_dataset_rele(origin, FTAG);
dsl_dir_rele(pdd, FTAG);
return (SET_ERROR(EINVAL));
}
dsl_dataset_rele(origin, FTAG);
dsl_dir_rele(pdd, FTAG);
return (0);
}
static void
dmu_objset_clone_sync(void *arg, dmu_tx_t *tx)
{
dmu_objset_clone_arg_t *doca = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *pdd;
const char *tail;
dsl_dataset_t *origin, *ds;
uint64_t obj;
char namebuf[ZFS_MAX_DATASET_NAME_LEN];
VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail));
VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin));
obj = dsl_dataset_create_sync(pdd, tail, origin, 0,
doca->doca_cred, NULL, tx);
VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds));
dsl_dataset_name(origin, namebuf);
spa_history_log_internal_ds(ds, "clone", tx,
"origin=%s (%llu)", namebuf, (u_longlong_t)origin->ds_object);
dsl_dataset_rele(ds, FTAG);
dsl_dataset_rele(origin, FTAG);
dsl_dir_rele(pdd, FTAG);
}
int
dmu_objset_clone(const char *clone, const char *origin)
{
dmu_objset_clone_arg_t doca;
doca.doca_clone = clone;
doca.doca_origin = origin;
doca.doca_cred = CRED();
doca.doca_proc = curproc;
int rv = dsl_sync_task(clone,
dmu_objset_clone_check, dmu_objset_clone_sync, &doca,
6, ZFS_SPACE_CHECK_NORMAL);
if (rv == 0)
zvol_create_minor(clone);
return (rv);
}
int
dmu_objset_snapshot_one(const char *fsname, const char *snapname)
{
int err;
char *longsnap = kmem_asprintf("%s@%s", fsname, snapname);
nvlist_t *snaps = fnvlist_alloc();
fnvlist_add_boolean(snaps, longsnap);
kmem_strfree(longsnap);
err = dsl_dataset_snapshot(snaps, NULL, NULL);
fnvlist_free(snaps);
return (err);
}
static void
dmu_objset_upgrade_task_cb(void *data)
{
objset_t *os = data;
mutex_enter(&os->os_upgrade_lock);
os->os_upgrade_status = EINTR;
if (!os->os_upgrade_exit) {
int status;
mutex_exit(&os->os_upgrade_lock);
status = os->os_upgrade_cb(os);
mutex_enter(&os->os_upgrade_lock);
os->os_upgrade_status = status;
}
os->os_upgrade_exit = B_TRUE;
os->os_upgrade_id = 0;
mutex_exit(&os->os_upgrade_lock);
dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
}
static void
dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb)
{
if (os->os_upgrade_id != 0)
return;
ASSERT(dsl_pool_config_held(dmu_objset_pool(os)));
dsl_dataset_long_hold(dmu_objset_ds(os), upgrade_tag);
mutex_enter(&os->os_upgrade_lock);
if (os->os_upgrade_id == 0 && os->os_upgrade_status == 0) {
os->os_upgrade_exit = B_FALSE;
os->os_upgrade_cb = cb;
os->os_upgrade_id = taskq_dispatch(
os->os_spa->spa_upgrade_taskq,
dmu_objset_upgrade_task_cb, os, TQ_SLEEP);
if (os->os_upgrade_id == TASKQID_INVALID) {
dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
os->os_upgrade_status = ENOMEM;
}
} else {
dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
}
mutex_exit(&os->os_upgrade_lock);
}
static void
dmu_objset_upgrade_stop(objset_t *os)
{
mutex_enter(&os->os_upgrade_lock);
os->os_upgrade_exit = B_TRUE;
if (os->os_upgrade_id != 0) {
taskqid_t id = os->os_upgrade_id;
os->os_upgrade_id = 0;
mutex_exit(&os->os_upgrade_lock);
if ((taskq_cancel_id(os->os_spa->spa_upgrade_taskq, id)) == 0) {
dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
}
txg_wait_synced(os->os_spa->spa_dsl_pool, 0);
} else {
mutex_exit(&os->os_upgrade_lock);
}
}
static void
dmu_objset_sync_dnodes(multilist_sublist_t *list, dmu_tx_t *tx)
{
dnode_t *dn;
while ((dn = multilist_sublist_head(list)) != NULL) {
ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT);
ASSERT(dn->dn_dbuf->db_data_pending);
/*
* Initialize dn_zio outside dnode_sync() because the
* meta-dnode needs to set it outside dnode_sync().
*/
dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio;
ASSERT(dn->dn_zio);
ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS);
multilist_sublist_remove(list, dn);
/*
* See the comment above dnode_rele_task() for an explanation
* of why this dnode hold is always needed (even when not
* doing user accounting).
*/
multilist_t *newlist = &dn->dn_objset->os_synced_dnodes;
(void) dnode_add_ref(dn, newlist);
multilist_insert(newlist, dn);
dnode_sync(dn, tx);
}
}
static void
dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg)
{
(void) abuf;
blkptr_t *bp = zio->io_bp;
objset_t *os = arg;
dnode_phys_t *dnp = &os->os_phys->os_meta_dnode;
uint64_t fill = 0;
ASSERT(!BP_IS_EMBEDDED(bp));
ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET);
ASSERT0(BP_GET_LEVEL(bp));
/*
* Update rootbp fill count: it should be the number of objects
* allocated in the object set (not counting the "special"
* objects that are stored in the objset_phys_t -- the meta
* dnode and user/group/project accounting objects).
*/
for (int i = 0; i < dnp->dn_nblkptr; i++)
fill += BP_GET_FILL(&dnp->dn_blkptr[i]);
BP_SET_FILL(bp, fill);
if (os->os_dsl_dataset != NULL)
rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_WRITER, FTAG);
*os->os_rootbp = *bp;
if (os->os_dsl_dataset != NULL)
rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG);
}
static void
dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg)
{
(void) abuf;
blkptr_t *bp = zio->io_bp;
blkptr_t *bp_orig = &zio->io_bp_orig;
objset_t *os = arg;
if (zio->io_flags & ZIO_FLAG_IO_REWRITE) {
ASSERT(BP_EQUAL(bp, bp_orig));
} else {
dsl_dataset_t *ds = os->os_dsl_dataset;
dmu_tx_t *tx = os->os_synctx;
(void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE);
dsl_dataset_block_born(ds, bp, tx);
}
kmem_free(bp, sizeof (*bp));
}
typedef struct sync_dnodes_arg {
multilist_t *sda_list;
int sda_sublist_idx;
multilist_t *sda_newlist;
dmu_tx_t *sda_tx;
} sync_dnodes_arg_t;
static void
sync_dnodes_task(void *arg)
{
sync_dnodes_arg_t *sda = arg;
multilist_sublist_t *ms =
multilist_sublist_lock(sda->sda_list, sda->sda_sublist_idx);
dmu_objset_sync_dnodes(ms, sda->sda_tx);
multilist_sublist_unlock(ms);
kmem_free(sda, sizeof (*sda));
}
/* called from dsl */
void
dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx)
{
int txgoff;
zbookmark_phys_t zb;
zio_prop_t zp;
zio_t *zio;
list_t *list;
dbuf_dirty_record_t *dr;
int num_sublists;
multilist_t *ml;
blkptr_t *blkptr_copy = kmem_alloc(sizeof (*os->os_rootbp), KM_SLEEP);
*blkptr_copy = *os->os_rootbp;
dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", (u_longlong_t)tx->tx_txg);
ASSERT(dmu_tx_is_syncing(tx));
/* XXX the write_done callback should really give us the tx... */
os->os_synctx = tx;
if (os->os_dsl_dataset == NULL) {
/*
* This is the MOS. If we have upgraded,
* spa_max_replication() could change, so reset
* os_copies here.
*/
os->os_copies = spa_max_replication(os->os_spa);
}
/*
* Create the root block IO
*/
SET_BOOKMARK(&zb, os->os_dsl_dataset ?
os->os_dsl_dataset->ds_object : DMU_META_OBJSET,
ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
arc_release(os->os_phys_buf, &os->os_phys_buf);
dmu_write_policy(os, NULL, 0, 0, &zp);
/*
* If we are either claiming the ZIL or doing a raw receive, write
* out the os_phys_buf raw. Neither of these actions will effect the
* MAC at this point.
*/
if (os->os_raw_receive ||
os->os_next_write_raw[tx->tx_txg & TXG_MASK]) {
ASSERT(os->os_encrypted);
arc_convert_to_raw(os->os_phys_buf,
os->os_dsl_dataset->ds_object, ZFS_HOST_BYTEORDER,
DMU_OT_OBJSET, NULL, NULL, NULL);
}
zio = arc_write(pio, os->os_spa, tx->tx_txg,
blkptr_copy, os->os_phys_buf, dmu_os_is_l2cacheable(os),
&zp, dmu_objset_write_ready, NULL, NULL, dmu_objset_write_done,
os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
/*
* Sync special dnodes - the parent IO for the sync is the root block
*/
DMU_META_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_META_DNODE(os), tx);
os->os_phys->os_flags = os->os_flags;
if (DMU_USERUSED_DNODE(os) &&
DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) {
DMU_USERUSED_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_USERUSED_DNODE(os), tx);
DMU_GROUPUSED_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_GROUPUSED_DNODE(os), tx);
}
if (DMU_PROJECTUSED_DNODE(os) &&
DMU_PROJECTUSED_DNODE(os)->dn_type != DMU_OT_NONE) {
DMU_PROJECTUSED_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_PROJECTUSED_DNODE(os), tx);
}
txgoff = tx->tx_txg & TXG_MASK;
/*
* We must create the list here because it uses the
* dn_dirty_link[] of this txg. But it may already
* exist because we call dsl_dataset_sync() twice per txg.
*/
if (os->os_synced_dnodes.ml_sublists == NULL) {
multilist_create(&os->os_synced_dnodes, sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[txgoff]),
dnode_multilist_index_func);
} else {
ASSERT3U(os->os_synced_dnodes.ml_offset, ==,
offsetof(dnode_t, dn_dirty_link[txgoff]));
}
ml = &os->os_dirty_dnodes[txgoff];
num_sublists = multilist_get_num_sublists(ml);
for (int i = 0; i < num_sublists; i++) {
if (multilist_sublist_is_empty_idx(ml, i))
continue;
sync_dnodes_arg_t *sda = kmem_alloc(sizeof (*sda), KM_SLEEP);
sda->sda_list = ml;
sda->sda_sublist_idx = i;
sda->sda_tx = tx;
(void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq,
sync_dnodes_task, sda, 0);
/* callback frees sda */
}
taskq_wait(dmu_objset_pool(os)->dp_sync_taskq);
list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff];
while ((dr = list_head(list)) != NULL) {
ASSERT0(dr->dr_dbuf->db_level);
list_remove(list, dr);
zio_nowait(dr->dr_zio);
}
/* Enable dnode backfill if enough objects have been freed. */
if (os->os_freed_dnodes >= dmu_rescan_dnode_threshold) {
os->os_rescan_dnodes = B_TRUE;
os->os_freed_dnodes = 0;
}
/*
* Free intent log blocks up to this tx.
*/
zil_sync(os->os_zil, tx);
os->os_phys->os_zil_header = os->os_zil_header;
zio_nowait(zio);
}
boolean_t
dmu_objset_is_dirty(objset_t *os, uint64_t txg)
{
return (!multilist_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK]));
}
static file_info_cb_t *file_cbs[DMU_OST_NUMTYPES];
void
dmu_objset_register_type(dmu_objset_type_t ost, file_info_cb_t *cb)
{
file_cbs[ost] = cb;
}
int
dmu_get_file_info(objset_t *os, dmu_object_type_t bonustype, const void *data,
zfs_file_info_t *zfi)
{
file_info_cb_t *cb = file_cbs[os->os_phys->os_type];
if (cb == NULL)
return (EINVAL);
return (cb(bonustype, data, zfi));
}
boolean_t
dmu_objset_userused_enabled(objset_t *os)
{
return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE &&
file_cbs[os->os_phys->os_type] != NULL &&
DMU_USERUSED_DNODE(os) != NULL);
}
boolean_t
dmu_objset_userobjused_enabled(objset_t *os)
{
return (dmu_objset_userused_enabled(os) &&
spa_feature_is_enabled(os->os_spa, SPA_FEATURE_USEROBJ_ACCOUNTING));
}
boolean_t
dmu_objset_projectquota_enabled(objset_t *os)
{
return (file_cbs[os->os_phys->os_type] != NULL &&
DMU_PROJECTUSED_DNODE(os) != NULL &&
spa_feature_is_enabled(os->os_spa, SPA_FEATURE_PROJECT_QUOTA));
}
typedef struct userquota_node {
/* must be in the first filed, see userquota_update_cache() */
char uqn_id[20 + DMU_OBJACCT_PREFIX_LEN];
int64_t uqn_delta;
avl_node_t uqn_node;
} userquota_node_t;
typedef struct userquota_cache {
avl_tree_t uqc_user_deltas;
avl_tree_t uqc_group_deltas;
avl_tree_t uqc_project_deltas;
} userquota_cache_t;
static int
userquota_compare(const void *l, const void *r)
{
const userquota_node_t *luqn = l;
const userquota_node_t *ruqn = r;
int rv;
/*
* NB: can only access uqn_id because userquota_update_cache() doesn't
* pass in an entire userquota_node_t.
*/
rv = strcmp(luqn->uqn_id, ruqn->uqn_id);
return (TREE_ISIGN(rv));
}
static void
do_userquota_cacheflush(objset_t *os, userquota_cache_t *cache, dmu_tx_t *tx)
{
void *cookie;
userquota_node_t *uqn;
ASSERT(dmu_tx_is_syncing(tx));
cookie = NULL;
while ((uqn = avl_destroy_nodes(&cache->uqc_user_deltas,
&cookie)) != NULL) {
/*
* os_userused_lock protects against concurrent calls to
* zap_increment_int(). It's needed because zap_increment_int()
* is not thread-safe (i.e. not atomic).
*/
mutex_enter(&os->os_userused_lock);
VERIFY0(zap_increment(os, DMU_USERUSED_OBJECT,
uqn->uqn_id, uqn->uqn_delta, tx));
mutex_exit(&os->os_userused_lock);
kmem_free(uqn, sizeof (*uqn));
}
avl_destroy(&cache->uqc_user_deltas);
cookie = NULL;
while ((uqn = avl_destroy_nodes(&cache->uqc_group_deltas,
&cookie)) != NULL) {
mutex_enter(&os->os_userused_lock);
VERIFY0(zap_increment(os, DMU_GROUPUSED_OBJECT,
uqn->uqn_id, uqn->uqn_delta, tx));
mutex_exit(&os->os_userused_lock);
kmem_free(uqn, sizeof (*uqn));
}
avl_destroy(&cache->uqc_group_deltas);
if (dmu_objset_projectquota_enabled(os)) {
cookie = NULL;
while ((uqn = avl_destroy_nodes(&cache->uqc_project_deltas,
&cookie)) != NULL) {
mutex_enter(&os->os_userused_lock);
VERIFY0(zap_increment(os, DMU_PROJECTUSED_OBJECT,
uqn->uqn_id, uqn->uqn_delta, tx));
mutex_exit(&os->os_userused_lock);
kmem_free(uqn, sizeof (*uqn));
}
avl_destroy(&cache->uqc_project_deltas);
}
}
static void
userquota_update_cache(avl_tree_t *avl, const char *id, int64_t delta)
{
userquota_node_t *uqn;
avl_index_t idx;
ASSERT(strlen(id) < sizeof (uqn->uqn_id));
/*
* Use id directly for searching because uqn_id is the first field of
* userquota_node_t and fields after uqn_id won't be accessed in
* avl_find().
*/
uqn = avl_find(avl, (const void *)id, &idx);
if (uqn == NULL) {
uqn = kmem_zalloc(sizeof (*uqn), KM_SLEEP);
strlcpy(uqn->uqn_id, id, sizeof (uqn->uqn_id));
avl_insert(avl, uqn, idx);
}
uqn->uqn_delta += delta;
}
static void
do_userquota_update(objset_t *os, userquota_cache_t *cache, uint64_t used,
uint64_t flags, uint64_t user, uint64_t group, uint64_t project,
boolean_t subtract)
{
if (flags & DNODE_FLAG_USERUSED_ACCOUNTED) {
int64_t delta = DNODE_MIN_SIZE + used;
char name[20];
if (subtract)
delta = -delta;
(void) snprintf(name, sizeof (name), "%llx", (longlong_t)user);
userquota_update_cache(&cache->uqc_user_deltas, name, delta);
(void) snprintf(name, sizeof (name), "%llx", (longlong_t)group);
userquota_update_cache(&cache->uqc_group_deltas, name, delta);
if (dmu_objset_projectquota_enabled(os)) {
(void) snprintf(name, sizeof (name), "%llx",
(longlong_t)project);
userquota_update_cache(&cache->uqc_project_deltas,
name, delta);
}
}
}
static void
do_userobjquota_update(objset_t *os, userquota_cache_t *cache, uint64_t flags,
uint64_t user, uint64_t group, uint64_t project, boolean_t subtract)
{
if (flags & DNODE_FLAG_USEROBJUSED_ACCOUNTED) {
char name[20 + DMU_OBJACCT_PREFIX_LEN];
int delta = subtract ? -1 : 1;
(void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx",
(longlong_t)user);
userquota_update_cache(&cache->uqc_user_deltas, name, delta);
(void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx",
(longlong_t)group);
userquota_update_cache(&cache->uqc_group_deltas, name, delta);
if (dmu_objset_projectquota_enabled(os)) {
(void) snprintf(name, sizeof (name),
DMU_OBJACCT_PREFIX "%llx", (longlong_t)project);
userquota_update_cache(&cache->uqc_project_deltas,
name, delta);
}
}
}
typedef struct userquota_updates_arg {
objset_t *uua_os;
int uua_sublist_idx;
dmu_tx_t *uua_tx;
} userquota_updates_arg_t;
static void
userquota_updates_task(void *arg)
{
userquota_updates_arg_t *uua = arg;
objset_t *os = uua->uua_os;
dmu_tx_t *tx = uua->uua_tx;
dnode_t *dn;
userquota_cache_t cache = { { 0 } };
multilist_sublist_t *list =
multilist_sublist_lock(&os->os_synced_dnodes, uua->uua_sublist_idx);
ASSERT(multilist_sublist_head(list) == NULL ||
dmu_objset_userused_enabled(os));
avl_create(&cache.uqc_user_deltas, userquota_compare,
sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node));
avl_create(&cache.uqc_group_deltas, userquota_compare,
sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node));
if (dmu_objset_projectquota_enabled(os))
avl_create(&cache.uqc_project_deltas, userquota_compare,
sizeof (userquota_node_t), offsetof(userquota_node_t,
uqn_node));
while ((dn = multilist_sublist_head(list)) != NULL) {
int flags;
ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object));
ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE ||
dn->dn_phys->dn_flags &
DNODE_FLAG_USERUSED_ACCOUNTED);
flags = dn->dn_id_flags;
ASSERT(flags);
if (flags & DN_ID_OLD_EXIST) {
do_userquota_update(os, &cache, dn->dn_oldused,
dn->dn_oldflags, dn->dn_olduid, dn->dn_oldgid,
dn->dn_oldprojid, B_TRUE);
do_userobjquota_update(os, &cache, dn->dn_oldflags,
dn->dn_olduid, dn->dn_oldgid,
dn->dn_oldprojid, B_TRUE);
}
if (flags & DN_ID_NEW_EXIST) {
do_userquota_update(os, &cache,
DN_USED_BYTES(dn->dn_phys), dn->dn_phys->dn_flags,
dn->dn_newuid, dn->dn_newgid,
dn->dn_newprojid, B_FALSE);
do_userobjquota_update(os, &cache,
dn->dn_phys->dn_flags, dn->dn_newuid, dn->dn_newgid,
dn->dn_newprojid, B_FALSE);
}
mutex_enter(&dn->dn_mtx);
dn->dn_oldused = 0;
dn->dn_oldflags = 0;
if (dn->dn_id_flags & DN_ID_NEW_EXIST) {
dn->dn_olduid = dn->dn_newuid;
dn->dn_oldgid = dn->dn_newgid;
dn->dn_oldprojid = dn->dn_newprojid;
dn->dn_id_flags |= DN_ID_OLD_EXIST;
if (dn->dn_bonuslen == 0)
dn->dn_id_flags |= DN_ID_CHKED_SPILL;
else
dn->dn_id_flags |= DN_ID_CHKED_BONUS;
}
dn->dn_id_flags &= ~(DN_ID_NEW_EXIST);
mutex_exit(&dn->dn_mtx);
multilist_sublist_remove(list, dn);
dnode_rele(dn, &os->os_synced_dnodes);
}
do_userquota_cacheflush(os, &cache, tx);
multilist_sublist_unlock(list);
kmem_free(uua, sizeof (*uua));
}
/*
* Release dnode holds from dmu_objset_sync_dnodes(). When the dnode is being
* synced (i.e. we have issued the zio's for blocks in the dnode), it can't be
* evicted because the block containing the dnode can't be evicted until it is
* written out. However, this hold is necessary to prevent the dnode_t from
* being moved (via dnode_move()) while it's still referenced by
* dbuf_dirty_record_t:dr_dnode. And dr_dnode is needed for
* dirty_lightweight_leaf-type dirty records.
*
* If we are doing user-object accounting, the dnode_rele() happens from
* userquota_updates_task() instead.
*/
static void
dnode_rele_task(void *arg)
{
userquota_updates_arg_t *uua = arg;
objset_t *os = uua->uua_os;
multilist_sublist_t *list =
multilist_sublist_lock(&os->os_synced_dnodes, uua->uua_sublist_idx);
dnode_t *dn;
while ((dn = multilist_sublist_head(list)) != NULL) {
multilist_sublist_remove(list, dn);
dnode_rele(dn, &os->os_synced_dnodes);
}
multilist_sublist_unlock(list);
kmem_free(uua, sizeof (*uua));
}
/*
* Return TRUE if userquota updates are needed.
*/
static boolean_t
dmu_objset_do_userquota_updates_prep(objset_t *os, dmu_tx_t *tx)
{
if (!dmu_objset_userused_enabled(os))
return (B_FALSE);
/*
* If this is a raw receive just return and handle accounting
* later when we have the keys loaded. We also don't do user
* accounting during claiming since the datasets are not owned
* for the duration of claiming and this txg should only be
* used for recovery.
*/
if (os->os_encrypted && dmu_objset_is_receiving(os))
return (B_FALSE);
if (tx->tx_txg <= os->os_spa->spa_claim_max_txg)
return (B_FALSE);
/* Allocate the user/group/project used objects if necessary. */
if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) {
VERIFY0(zap_create_claim(os,
DMU_USERUSED_OBJECT,
DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
VERIFY0(zap_create_claim(os,
DMU_GROUPUSED_OBJECT,
DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
}
if (dmu_objset_projectquota_enabled(os) &&
DMU_PROJECTUSED_DNODE(os)->dn_type == DMU_OT_NONE) {
VERIFY0(zap_create_claim(os, DMU_PROJECTUSED_OBJECT,
DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
}
return (B_TRUE);
}
/*
* Dispatch taskq tasks to dp_sync_taskq to update the user accounting, and
* also release the holds on the dnodes from dmu_objset_sync_dnodes().
* The caller must taskq_wait(dp_sync_taskq).
*/
void
dmu_objset_sync_done(objset_t *os, dmu_tx_t *tx)
{
boolean_t need_userquota = dmu_objset_do_userquota_updates_prep(os, tx);
int num_sublists = multilist_get_num_sublists(&os->os_synced_dnodes);
for (int i = 0; i < num_sublists; i++) {
userquota_updates_arg_t *uua =
kmem_alloc(sizeof (*uua), KM_SLEEP);
uua->uua_os = os;
uua->uua_sublist_idx = i;
uua->uua_tx = tx;
/*
* If we don't need to update userquotas, use
* dnode_rele_task() to call dnode_rele()
*/
(void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq,
need_userquota ? userquota_updates_task : dnode_rele_task,
uua, 0);
/* callback frees uua */
}
}
/*
* Returns a pointer to data to find uid/gid from
*
* If a dirty record for transaction group that is syncing can't
* be found then NULL is returned. In the NULL case it is assumed
* the uid/gid aren't changing.
*/
static void *
dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx)
{
dbuf_dirty_record_t *dr;
void *data;
if (db->db_dirtycnt == 0)
return (db->db.db_data); /* Nothing is changing */
dr = dbuf_find_dirty_eq(db, tx->tx_txg);
if (dr == NULL) {
data = NULL;
} else {
if (dr->dr_dnode->dn_bonuslen == 0 &&
dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID)
data = dr->dt.dl.dr_data->b_data;
else
data = dr->dt.dl.dr_data;
}
return (data);
}
void
dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx)
{
objset_t *os = dn->dn_objset;
void *data = NULL;
dmu_buf_impl_t *db = NULL;
int flags = dn->dn_id_flags;
int error;
boolean_t have_spill = B_FALSE;
if (!dmu_objset_userused_enabled(dn->dn_objset))
return;
/*
* Raw receives introduce a problem with user accounting. Raw
* receives cannot update the user accounting info because the
* user ids and the sizes are encrypted. To guarantee that we
* never end up with bad user accounting, we simply disable it
* during raw receives. We also disable this for normal receives
* so that an incremental raw receive may be done on top of an
* existing non-raw receive.
*/
if (os->os_encrypted && dmu_objset_is_receiving(os))
return;
if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST|
DN_ID_CHKED_SPILL)))
return;
if (before && dn->dn_bonuslen != 0)
data = DN_BONUS(dn->dn_phys);
else if (!before && dn->dn_bonuslen != 0) {
if (dn->dn_bonus) {
db = dn->dn_bonus;
mutex_enter(&db->db_mtx);
data = dmu_objset_userquota_find_data(db, tx);
} else {
data = DN_BONUS(dn->dn_phys);
}
} else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) {
int rf = 0;
if (RW_WRITE_HELD(&dn->dn_struct_rwlock))
rf |= DB_RF_HAVESTRUCT;
error = dmu_spill_hold_by_dnode(dn,
rf | DB_RF_MUST_SUCCEED,
FTAG, (dmu_buf_t **)&db);
ASSERT(error == 0);
mutex_enter(&db->db_mtx);
data = (before) ? db->db.db_data :
dmu_objset_userquota_find_data(db, tx);
have_spill = B_TRUE;
} else {
mutex_enter(&dn->dn_mtx);
dn->dn_id_flags |= DN_ID_CHKED_BONUS;
mutex_exit(&dn->dn_mtx);
return;
}
/*
* Must always call the callback in case the object
* type has changed and that type isn't an object type to track
*/
zfs_file_info_t zfi;
error = file_cbs[os->os_phys->os_type](dn->dn_bonustype, data, &zfi);
if (before) {
ASSERT(data);
dn->dn_olduid = zfi.zfi_user;
dn->dn_oldgid = zfi.zfi_group;
dn->dn_oldprojid = zfi.zfi_project;
} else if (data) {
dn->dn_newuid = zfi.zfi_user;
dn->dn_newgid = zfi.zfi_group;
dn->dn_newprojid = zfi.zfi_project;
}
/*
* Preserve existing uid/gid when the callback can't determine
* what the new uid/gid are and the callback returned EEXIST.
* The EEXIST error tells us to just use the existing uid/gid.
* If we don't know what the old values are then just assign
* them to 0, since that is a new file being created.
*/
if (!before && data == NULL && error == EEXIST) {
if (flags & DN_ID_OLD_EXIST) {
dn->dn_newuid = dn->dn_olduid;
dn->dn_newgid = dn->dn_oldgid;
dn->dn_newprojid = dn->dn_oldprojid;
} else {
dn->dn_newuid = 0;
dn->dn_newgid = 0;
dn->dn_newprojid = ZFS_DEFAULT_PROJID;
}
error = 0;
}
if (db)
mutex_exit(&db->db_mtx);
mutex_enter(&dn->dn_mtx);
if (error == 0 && before)
dn->dn_id_flags |= DN_ID_OLD_EXIST;
if (error == 0 && !before)
dn->dn_id_flags |= DN_ID_NEW_EXIST;
if (have_spill) {
dn->dn_id_flags |= DN_ID_CHKED_SPILL;
} else {
dn->dn_id_flags |= DN_ID_CHKED_BONUS;
}
mutex_exit(&dn->dn_mtx);
if (have_spill)
dmu_buf_rele((dmu_buf_t *)db, FTAG);
}
boolean_t
dmu_objset_userspace_present(objset_t *os)
{
return (os->os_phys->os_flags &
OBJSET_FLAG_USERACCOUNTING_COMPLETE);
}
boolean_t
dmu_objset_userobjspace_present(objset_t *os)
{
return (os->os_phys->os_flags &
OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE);
}
boolean_t
dmu_objset_projectquota_present(objset_t *os)
{
return (os->os_phys->os_flags &
OBJSET_FLAG_PROJECTQUOTA_COMPLETE);
}
static int
dmu_objset_space_upgrade(objset_t *os)
{
uint64_t obj;
int err = 0;
/*
* We simply need to mark every object dirty, so that it will be
* synced out and now accounted. If this is called
* concurrently, or if we already did some work before crashing,
* that's fine, since we track each object's accounted state
* independently.
*/
for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) {
dmu_tx_t *tx;
dmu_buf_t *db;
int objerr;
mutex_enter(&os->os_upgrade_lock);
if (os->os_upgrade_exit)
err = SET_ERROR(EINTR);
mutex_exit(&os->os_upgrade_lock);
if (err != 0)
return (err);
if (issig(JUSTLOOKING) && issig(FORREAL))
return (SET_ERROR(EINTR));
objerr = dmu_bonus_hold(os, obj, FTAG, &db);
if (objerr != 0)
continue;
tx = dmu_tx_create(os);
dmu_tx_hold_bonus(tx, obj);
objerr = dmu_tx_assign(tx, TXG_WAIT);
if (objerr != 0) {
dmu_buf_rele(db, FTAG);
dmu_tx_abort(tx);
continue;
}
dmu_buf_will_dirty(db, tx);
dmu_buf_rele(db, FTAG);
dmu_tx_commit(tx);
}
return (0);
}
static int
dmu_objset_userspace_upgrade_cb(objset_t *os)
{
int err = 0;
if (dmu_objset_userspace_present(os))
return (0);
if (dmu_objset_is_snapshot(os))
return (SET_ERROR(EINVAL));
if (!dmu_objset_userused_enabled(os))
return (SET_ERROR(ENOTSUP));
err = dmu_objset_space_upgrade(os);
if (err)
return (err);
os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
txg_wait_synced(dmu_objset_pool(os), 0);
return (0);
}
void
dmu_objset_userspace_upgrade(objset_t *os)
{
dmu_objset_upgrade(os, dmu_objset_userspace_upgrade_cb);
}
static int
dmu_objset_id_quota_upgrade_cb(objset_t *os)
{
int err = 0;
if (dmu_objset_userobjspace_present(os) &&
dmu_objset_projectquota_present(os))
return (0);
if (dmu_objset_is_snapshot(os))
return (SET_ERROR(EINVAL));
if (!dmu_objset_userused_enabled(os))
return (SET_ERROR(ENOTSUP));
if (!dmu_objset_projectquota_enabled(os) &&
dmu_objset_userobjspace_present(os))
return (SET_ERROR(ENOTSUP));
if (dmu_objset_userobjused_enabled(os))
dmu_objset_ds(os)->ds_feature_activation[
SPA_FEATURE_USEROBJ_ACCOUNTING] = (void *)B_TRUE;
if (dmu_objset_projectquota_enabled(os))
dmu_objset_ds(os)->ds_feature_activation[
SPA_FEATURE_PROJECT_QUOTA] = (void *)B_TRUE;
err = dmu_objset_space_upgrade(os);
if (err)
return (err);
os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
if (dmu_objset_userobjused_enabled(os))
os->os_flags |= OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE;
if (dmu_objset_projectquota_enabled(os))
os->os_flags |= OBJSET_FLAG_PROJECTQUOTA_COMPLETE;
txg_wait_synced(dmu_objset_pool(os), 0);
return (0);
}
void
dmu_objset_id_quota_upgrade(objset_t *os)
{
dmu_objset_upgrade(os, dmu_objset_id_quota_upgrade_cb);
}
boolean_t
dmu_objset_userobjspace_upgradable(objset_t *os)
{
return (dmu_objset_type(os) == DMU_OST_ZFS &&
!dmu_objset_is_snapshot(os) &&
dmu_objset_userobjused_enabled(os) &&
!dmu_objset_userobjspace_present(os) &&
spa_writeable(dmu_objset_spa(os)));
}
boolean_t
dmu_objset_projectquota_upgradable(objset_t *os)
{
return (dmu_objset_type(os) == DMU_OST_ZFS &&
!dmu_objset_is_snapshot(os) &&
dmu_objset_projectquota_enabled(os) &&
!dmu_objset_projectquota_present(os) &&
spa_writeable(dmu_objset_spa(os)));
}
void
dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp,
uint64_t *usedobjsp, uint64_t *availobjsp)
{
dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp,
usedobjsp, availobjsp);
}
uint64_t
dmu_objset_fsid_guid(objset_t *os)
{
return (dsl_dataset_fsid_guid(os->os_dsl_dataset));
}
void
dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat)
{
stat->dds_type = os->os_phys->os_type;
if (os->os_dsl_dataset)
dsl_dataset_fast_stat(os->os_dsl_dataset, stat);
}
void
dmu_objset_stats(objset_t *os, nvlist_t *nv)
{
ASSERT(os->os_dsl_dataset ||
os->os_phys->os_type == DMU_OST_META);
if (os->os_dsl_dataset != NULL)
dsl_dataset_stats(os->os_dsl_dataset, nv);
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE,
os->os_phys->os_type);
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING,
dmu_objset_userspace_present(os));
}
int
dmu_objset_is_snapshot(objset_t *os)
{
if (os->os_dsl_dataset != NULL)
return (os->os_dsl_dataset->ds_is_snapshot);
else
return (B_FALSE);
}
int
dmu_snapshot_realname(objset_t *os, const char *name, char *real, int maxlen,
boolean_t *conflict)
{
dsl_dataset_t *ds = os->os_dsl_dataset;
uint64_t ignored;
if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0)
return (SET_ERROR(ENOENT));
return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset,
dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored,
MT_NORMALIZE, real, maxlen, conflict));
}
int
dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
uint64_t *idp, uint64_t *offp, boolean_t *case_conflict)
{
dsl_dataset_t *ds = os->os_dsl_dataset;
zap_cursor_t cursor;
zap_attribute_t attr;
ASSERT(dsl_pool_config_held(dmu_objset_pool(os)));
if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0)
return (SET_ERROR(ENOENT));
zap_cursor_init_serialized(&cursor,
ds->ds_dir->dd_pool->dp_meta_objset,
dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp);
if (zap_cursor_retrieve(&cursor, &attr) != 0) {
zap_cursor_fini(&cursor);
return (SET_ERROR(ENOENT));
}
if (strlen(attr.za_name) + 1 > namelen) {
zap_cursor_fini(&cursor);
return (SET_ERROR(ENAMETOOLONG));
}
(void) strlcpy(name, attr.za_name, namelen);
if (idp)
*idp = attr.za_first_integer;
if (case_conflict)
*case_conflict = attr.za_normalization_conflict;
zap_cursor_advance(&cursor);
*offp = zap_cursor_serialize(&cursor);
zap_cursor_fini(&cursor);
return (0);
}
int
dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value)
{
return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value));
}
int
dmu_dir_list_next(objset_t *os, int namelen, char *name,
uint64_t *idp, uint64_t *offp)
{
dsl_dir_t *dd = os->os_dsl_dataset->ds_dir;
zap_cursor_t cursor;
zap_attribute_t attr;
/* there is no next dir on a snapshot! */
if (os->os_dsl_dataset->ds_object !=
dsl_dir_phys(dd)->dd_head_dataset_obj)
return (SET_ERROR(ENOENT));
zap_cursor_init_serialized(&cursor,
dd->dd_pool->dp_meta_objset,
dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp);
if (zap_cursor_retrieve(&cursor, &attr) != 0) {
zap_cursor_fini(&cursor);
return (SET_ERROR(ENOENT));
}
if (strlen(attr.za_name) + 1 > namelen) {
zap_cursor_fini(&cursor);
return (SET_ERROR(ENAMETOOLONG));
}
(void) strlcpy(name, attr.za_name, namelen);
if (idp)
*idp = attr.za_first_integer;
zap_cursor_advance(&cursor);
*offp = zap_cursor_serialize(&cursor);
zap_cursor_fini(&cursor);
return (0);
}
typedef struct dmu_objset_find_ctx {
taskq_t *dc_tq;
dsl_pool_t *dc_dp;
uint64_t dc_ddobj;
char *dc_ddname; /* last component of ddobj's name */
int (*dc_func)(dsl_pool_t *, dsl_dataset_t *, void *);
void *dc_arg;
int dc_flags;
kmutex_t *dc_error_lock;
int *dc_error;
} dmu_objset_find_ctx_t;
static void
dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp)
{
dsl_pool_t *dp = dcp->dc_dp;
dsl_dir_t *dd;
dsl_dataset_t *ds;
zap_cursor_t zc;
zap_attribute_t *attr;
uint64_t thisobj;
int err = 0;
/* don't process if there already was an error */
if (*dcp->dc_error != 0)
goto out;
/*
* Note: passing the name (dc_ddname) here is optional, but it
* improves performance because we don't need to call
* zap_value_search() to determine the name.
*/
err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, dcp->dc_ddname, FTAG, &dd);
if (err != 0)
goto out;
/* Don't visit hidden ($MOS & $ORIGIN) objsets. */
if (dd->dd_myname[0] == '$') {
dsl_dir_rele(dd, FTAG);
goto out;
}
thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj;
attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
/*
* Iterate over all children.
*/
if (dcp->dc_flags & DS_FIND_CHILDREN) {
for (zap_cursor_init(&zc, dp->dp_meta_objset,
dsl_dir_phys(dd)->dd_child_dir_zapobj);
zap_cursor_retrieve(&zc, attr) == 0;
(void) zap_cursor_advance(&zc)) {
ASSERT3U(attr->za_integer_length, ==,
sizeof (uint64_t));
ASSERT3U(attr->za_num_integers, ==, 1);
dmu_objset_find_ctx_t *child_dcp =
kmem_alloc(sizeof (*child_dcp), KM_SLEEP);
*child_dcp = *dcp;
child_dcp->dc_ddobj = attr->za_first_integer;
child_dcp->dc_ddname = spa_strdup(attr->za_name);
if (dcp->dc_tq != NULL)
(void) taskq_dispatch(dcp->dc_tq,
dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP);
else
dmu_objset_find_dp_impl(child_dcp);
}
zap_cursor_fini(&zc);
}
/*
* Iterate over all snapshots.
*/
if (dcp->dc_flags & DS_FIND_SNAPSHOTS) {
dsl_dataset_t *ds;
err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds);
if (err == 0) {
uint64_t snapobj;
snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj;
dsl_dataset_rele(ds, FTAG);
for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj);
zap_cursor_retrieve(&zc, attr) == 0;
(void) zap_cursor_advance(&zc)) {
ASSERT3U(attr->za_integer_length, ==,
sizeof (uint64_t));
ASSERT3U(attr->za_num_integers, ==, 1);
err = dsl_dataset_hold_obj(dp,
attr->za_first_integer, FTAG, &ds);
if (err != 0)
break;
err = dcp->dc_func(dp, ds, dcp->dc_arg);
dsl_dataset_rele(ds, FTAG);
if (err != 0)
break;
}
zap_cursor_fini(&zc);
}
}
kmem_free(attr, sizeof (zap_attribute_t));
if (err != 0) {
dsl_dir_rele(dd, FTAG);
goto out;
}
/*
* Apply to self.
*/
err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds);
/*
* Note: we hold the dir while calling dsl_dataset_hold_obj() so
* that the dir will remain cached, and we won't have to re-instantiate
* it (which could be expensive due to finding its name via
* zap_value_search()).
*/
dsl_dir_rele(dd, FTAG);
if (err != 0)
goto out;
err = dcp->dc_func(dp, ds, dcp->dc_arg);
dsl_dataset_rele(ds, FTAG);
out:
if (err != 0) {
mutex_enter(dcp->dc_error_lock);
/* only keep first error */
if (*dcp->dc_error == 0)
*dcp->dc_error = err;
mutex_exit(dcp->dc_error_lock);
}
if (dcp->dc_ddname != NULL)
spa_strfree(dcp->dc_ddname);
kmem_free(dcp, sizeof (*dcp));
}
static void
dmu_objset_find_dp_cb(void *arg)
{
dmu_objset_find_ctx_t *dcp = arg;
dsl_pool_t *dp = dcp->dc_dp;
/*
* We need to get a pool_config_lock here, as there are several
* assert(pool_config_held) down the stack. Getting a lock via
* dsl_pool_config_enter is risky, as it might be stalled by a
* pending writer. This would deadlock, as the write lock can
* only be granted when our parent thread gives up the lock.
* The _prio interface gives us priority over a pending writer.
*/
dsl_pool_config_enter_prio(dp, FTAG);
dmu_objset_find_dp_impl(dcp);
dsl_pool_config_exit(dp, FTAG);
}
/*
* Find objsets under and including ddobj, call func(ds) on each.
* The order for the enumeration is completely undefined.
* func is called with dsl_pool_config held.
*/
int
dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj,
int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags)
{
int error = 0;
taskq_t *tq = NULL;
int ntasks;
dmu_objset_find_ctx_t *dcp;
kmutex_t err_lock;
mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL);
dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP);
dcp->dc_tq = NULL;
dcp->dc_dp = dp;
dcp->dc_ddobj = ddobj;
dcp->dc_ddname = NULL;
dcp->dc_func = func;
dcp->dc_arg = arg;
dcp->dc_flags = flags;
dcp->dc_error_lock = &err_lock;
dcp->dc_error = &error;
if ((flags & DS_FIND_SERIALIZE) || dsl_pool_config_held_writer(dp)) {
/*
* In case a write lock is held we can't make use of
* parallelism, as down the stack of the worker threads
* the lock is asserted via dsl_pool_config_held.
* In case of a read lock this is solved by getting a read
* lock in each worker thread, which isn't possible in case
* of a writer lock. So we fall back to the synchronous path
* here.
* In the future it might be possible to get some magic into
* dsl_pool_config_held in a way that it returns true for
* the worker threads so that a single lock held from this
* thread suffices. For now, stay single threaded.
*/
dmu_objset_find_dp_impl(dcp);
mutex_destroy(&err_lock);
return (error);
}
ntasks = dmu_find_threads;
if (ntasks == 0)
ntasks = vdev_count_leaves(dp->dp_spa) * 4;
tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks,
INT_MAX, 0);
if (tq == NULL) {
kmem_free(dcp, sizeof (*dcp));
mutex_destroy(&err_lock);
return (SET_ERROR(ENOMEM));
}
dcp->dc_tq = tq;
/* dcp will be freed by task */
(void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP);
/*
* PORTING: this code relies on the property of taskq_wait to wait
* until no more tasks are queued and no more tasks are active. As
* we always queue new tasks from within other tasks, task_wait
* reliably waits for the full recursion to finish, even though we
* enqueue new tasks after taskq_wait has been called.
* On platforms other than illumos, taskq_wait may not have this
* property.
*/
taskq_wait(tq);
taskq_destroy(tq);
mutex_destroy(&err_lock);
return (error);
}
/*
* Find all objsets under name, and for each, call 'func(child_name, arg)'.
* The dp_config_rwlock must not be held when this is called, and it
* will not be held when the callback is called.
* Therefore this function should only be used when the pool is not changing
* (e.g. in syncing context), or the callback can deal with the possible races.
*/
static int
dmu_objset_find_impl(spa_t *spa, const char *name,
int func(const char *, void *), void *arg, int flags)
{
dsl_dir_t *dd;
dsl_pool_t *dp = spa_get_dsl(spa);
dsl_dataset_t *ds;
zap_cursor_t zc;
zap_attribute_t *attr;
char *child;
uint64_t thisobj;
int err;
dsl_pool_config_enter(dp, FTAG);
err = dsl_dir_hold(dp, name, FTAG, &dd, NULL);
if (err != 0) {
dsl_pool_config_exit(dp, FTAG);
return (err);
}
/* Don't visit hidden ($MOS & $ORIGIN) objsets. */
if (dd->dd_myname[0] == '$') {
dsl_dir_rele(dd, FTAG);
dsl_pool_config_exit(dp, FTAG);
return (0);
}
thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj;
attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
/*
* Iterate over all children.
*/
if (flags & DS_FIND_CHILDREN) {
for (zap_cursor_init(&zc, dp->dp_meta_objset,
dsl_dir_phys(dd)->dd_child_dir_zapobj);
zap_cursor_retrieve(&zc, attr) == 0;
(void) zap_cursor_advance(&zc)) {
ASSERT3U(attr->za_integer_length, ==,
sizeof (uint64_t));
ASSERT3U(attr->za_num_integers, ==, 1);
child = kmem_asprintf("%s/%s", name, attr->za_name);
dsl_pool_config_exit(dp, FTAG);
err = dmu_objset_find_impl(spa, child,
func, arg, flags);
dsl_pool_config_enter(dp, FTAG);
kmem_strfree(child);
if (err != 0)
break;
}
zap_cursor_fini(&zc);
if (err != 0) {
dsl_dir_rele(dd, FTAG);
dsl_pool_config_exit(dp, FTAG);
kmem_free(attr, sizeof (zap_attribute_t));
return (err);
}
}
/*
* Iterate over all snapshots.
*/
if (flags & DS_FIND_SNAPSHOTS) {
err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds);
if (err == 0) {
uint64_t snapobj;
snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj;
dsl_dataset_rele(ds, FTAG);
for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj);
zap_cursor_retrieve(&zc, attr) == 0;
(void) zap_cursor_advance(&zc)) {
ASSERT3U(attr->za_integer_length, ==,
sizeof (uint64_t));
ASSERT3U(attr->za_num_integers, ==, 1);
child = kmem_asprintf("%s@%s",
name, attr->za_name);
dsl_pool_config_exit(dp, FTAG);
err = func(child, arg);
dsl_pool_config_enter(dp, FTAG);
kmem_strfree(child);
if (err != 0)
break;
}
zap_cursor_fini(&zc);
}
}
dsl_dir_rele(dd, FTAG);
kmem_free(attr, sizeof (zap_attribute_t));
dsl_pool_config_exit(dp, FTAG);
if (err != 0)
return (err);
/* Apply to self. */
return (func(name, arg));
}
/*
* See comment above dmu_objset_find_impl().
*/
int
dmu_objset_find(const char *name, int func(const char *, void *), void *arg,
int flags)
{
spa_t *spa;
int error;
error = spa_open(name, &spa, FTAG);
if (error != 0)
return (error);
error = dmu_objset_find_impl(spa, name, func, arg, flags);
spa_close(spa, FTAG);
return (error);
}
boolean_t
dmu_objset_incompatible_encryption_version(objset_t *os)
{
return (dsl_dir_incompatible_encryption_version(
os->os_dsl_dataset->ds_dir));
}
void
dmu_objset_set_user(objset_t *os, void *user_ptr)
{
ASSERT(MUTEX_HELD(&os->os_user_ptr_lock));
os->os_user_ptr = user_ptr;
}
void *
dmu_objset_get_user(objset_t *os)
{
ASSERT(MUTEX_HELD(&os->os_user_ptr_lock));
return (os->os_user_ptr);
}
/*
* Determine name of filesystem, given name of snapshot.
* buf must be at least ZFS_MAX_DATASET_NAME_LEN bytes
*/
int
dmu_fsname(const char *snapname, char *buf)
{
char *atp = strchr(snapname, '@');
if (atp == NULL)
return (SET_ERROR(EINVAL));
if (atp - snapname >= ZFS_MAX_DATASET_NAME_LEN)
return (SET_ERROR(ENAMETOOLONG));
(void) strlcpy(buf, snapname, atp - snapname + 1);
return (0);
}
/*
* Call when we think we're going to write/free space in open context
* to track the amount of dirty data in the open txg, which is also the
* amount of memory that can not be evicted until this txg syncs.
*
* Note that there are two conditions where this can be called from
* syncing context:
*
* [1] When we just created the dataset, in which case we go on with
* updating any accounting of dirty data as usual.
* [2] When we are dirtying MOS data, in which case we only update the
* pool's accounting of dirty data.
*/
void
dmu_objset_willuse_space(objset_t *os, int64_t space, dmu_tx_t *tx)
{
dsl_dataset_t *ds = os->os_dsl_dataset;
int64_t aspace = spa_get_worst_case_asize(os->os_spa, space);
if (ds != NULL) {
dsl_dir_willuse_space(ds->ds_dir, aspace, tx);
}
dsl_pool_dirty_space(dmu_tx_pool(tx), space, tx);
}
#if defined(_KERNEL)
EXPORT_SYMBOL(dmu_objset_zil);
EXPORT_SYMBOL(dmu_objset_pool);
EXPORT_SYMBOL(dmu_objset_ds);
EXPORT_SYMBOL(dmu_objset_type);
EXPORT_SYMBOL(dmu_objset_name);
EXPORT_SYMBOL(dmu_objset_hold);
EXPORT_SYMBOL(dmu_objset_hold_flags);
EXPORT_SYMBOL(dmu_objset_own);
EXPORT_SYMBOL(dmu_objset_rele);
EXPORT_SYMBOL(dmu_objset_rele_flags);
EXPORT_SYMBOL(dmu_objset_disown);
EXPORT_SYMBOL(dmu_objset_from_ds);
EXPORT_SYMBOL(dmu_objset_create);
EXPORT_SYMBOL(dmu_objset_clone);
EXPORT_SYMBOL(dmu_objset_stats);
EXPORT_SYMBOL(dmu_objset_fast_stat);
EXPORT_SYMBOL(dmu_objset_spa);
EXPORT_SYMBOL(dmu_objset_space);
EXPORT_SYMBOL(dmu_objset_fsid_guid);
EXPORT_SYMBOL(dmu_objset_find);
EXPORT_SYMBOL(dmu_objset_byteswap);
EXPORT_SYMBOL(dmu_objset_evict_dbufs);
EXPORT_SYMBOL(dmu_objset_snap_cmtime);
EXPORT_SYMBOL(dmu_objset_dnodesize);
EXPORT_SYMBOL(dmu_objset_sync);
EXPORT_SYMBOL(dmu_objset_is_dirty);
EXPORT_SYMBOL(dmu_objset_create_impl_dnstats);
EXPORT_SYMBOL(dmu_objset_create_impl);
EXPORT_SYMBOL(dmu_objset_open_impl);
EXPORT_SYMBOL(dmu_objset_evict);
EXPORT_SYMBOL(dmu_objset_register_type);
EXPORT_SYMBOL(dmu_objset_sync_done);
EXPORT_SYMBOL(dmu_objset_userquota_get_ids);
EXPORT_SYMBOL(dmu_objset_userused_enabled);
EXPORT_SYMBOL(dmu_objset_userspace_upgrade);
EXPORT_SYMBOL(dmu_objset_userspace_present);
EXPORT_SYMBOL(dmu_objset_userobjused_enabled);
EXPORT_SYMBOL(dmu_objset_userobjspace_upgradable);
EXPORT_SYMBOL(dmu_objset_userobjspace_present);
EXPORT_SYMBOL(dmu_objset_projectquota_enabled);
EXPORT_SYMBOL(dmu_objset_projectquota_present);
EXPORT_SYMBOL(dmu_objset_projectquota_upgradable);
EXPORT_SYMBOL(dmu_objset_id_quota_upgrade);
#endif
diff --git a/module/zfs/dmu_recv.c b/module/zfs/dmu_recv.c
index 0ec46bdb4f47..f132219c9e1d 100644
--- a/module/zfs/dmu_recv.c
+++ b/module/zfs/dmu_recv.c
@@ -1,3401 +1,3401 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright (c) 2014, Joyent, Inc. All rights reserved.
* Copyright 2014 HybridCluster. All rights reserved.
* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
*/
#include <sys/dmu.h>
#include <sys/dmu_impl.h>
#include <sys/dmu_send.h>
#include <sys/dmu_recv.h>
#include <sys/dmu_tx.h>
#include <sys/dbuf.h>
#include <sys/dnode.h>
#include <sys/zfs_context.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_traverse.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_synctask.h>
#include <sys/zfs_ioctl.h>
#include <sys/zap.h>
#include <sys/zvol.h>
#include <sys/zio_checksum.h>
#include <sys/zfs_znode.h>
#include <zfs_fletcher.h>
#include <sys/avl.h>
#include <sys/ddt.h>
#include <sys/zfs_onexit.h>
#include <sys/dsl_destroy.h>
#include <sys/blkptr.h>
#include <sys/dsl_bookmark.h>
#include <sys/zfeature.h>
#include <sys/bqueue.h>
#include <sys/objlist.h>
#ifdef _KERNEL
#include <sys/zfs_vfsops.h>
#endif
#include <sys/zfs_file.h>
-int zfs_recv_queue_length = SPA_MAXBLOCKSIZE;
-int zfs_recv_queue_ff = 20;
-int zfs_recv_write_batch_size = 1024 * 1024;
+static int zfs_recv_queue_length = SPA_MAXBLOCKSIZE;
+static int zfs_recv_queue_ff = 20;
+static int zfs_recv_write_batch_size = 1024 * 1024;
-static char *dmu_recv_tag = "dmu_recv_tag";
-const char *recv_clone_name = "%recv";
+static void *const dmu_recv_tag = "dmu_recv_tag";
+const char *const recv_clone_name = "%recv";
static int receive_read_payload_and_next_header(dmu_recv_cookie_t *ra, int len,
void *buf);
struct receive_record_arg {
dmu_replay_record_t header;
void *payload; /* Pointer to a buffer containing the payload */
/*
* If the record is a WRITE or SPILL, pointer to the abd containing the
* payload.
*/
abd_t *abd;
int payload_size;
uint64_t bytes_read; /* bytes read from stream when record created */
boolean_t eos_marker; /* Marks the end of the stream */
bqueue_node_t node;
};
struct receive_writer_arg {
objset_t *os;
boolean_t byteswap;
bqueue_t q;
/*
* These three members are used to signal to the main thread when
* we're done.
*/
kmutex_t mutex;
kcondvar_t cv;
boolean_t done;
int err;
boolean_t resumable;
boolean_t raw; /* DMU_BACKUP_FEATURE_RAW set */
boolean_t spill; /* DRR_FLAG_SPILL_BLOCK set */
boolean_t full; /* this is a full send stream */
uint64_t last_object;
uint64_t last_offset;
uint64_t max_object; /* highest object ID referenced in stream */
uint64_t bytes_read; /* bytes read when current record created */
list_t write_batch;
/* Encryption parameters for the last received DRR_OBJECT_RANGE */
boolean_t or_crypt_params_present;
uint64_t or_firstobj;
uint64_t or_numslots;
uint8_t or_salt[ZIO_DATA_SALT_LEN];
uint8_t or_iv[ZIO_DATA_IV_LEN];
uint8_t or_mac[ZIO_DATA_MAC_LEN];
boolean_t or_byteorder;
};
typedef struct dmu_recv_begin_arg {
const char *drba_origin;
dmu_recv_cookie_t *drba_cookie;
cred_t *drba_cred;
proc_t *drba_proc;
dsl_crypto_params_t *drba_dcp;
} dmu_recv_begin_arg_t;
static void
byteswap_record(dmu_replay_record_t *drr)
{
#define DO64(X) (drr->drr_u.X = BSWAP_64(drr->drr_u.X))
#define DO32(X) (drr->drr_u.X = BSWAP_32(drr->drr_u.X))
drr->drr_type = BSWAP_32(drr->drr_type);
drr->drr_payloadlen = BSWAP_32(drr->drr_payloadlen);
switch (drr->drr_type) {
case DRR_BEGIN:
DO64(drr_begin.drr_magic);
DO64(drr_begin.drr_versioninfo);
DO64(drr_begin.drr_creation_time);
DO32(drr_begin.drr_type);
DO32(drr_begin.drr_flags);
DO64(drr_begin.drr_toguid);
DO64(drr_begin.drr_fromguid);
break;
case DRR_OBJECT:
DO64(drr_object.drr_object);
DO32(drr_object.drr_type);
DO32(drr_object.drr_bonustype);
DO32(drr_object.drr_blksz);
DO32(drr_object.drr_bonuslen);
DO32(drr_object.drr_raw_bonuslen);
DO64(drr_object.drr_toguid);
DO64(drr_object.drr_maxblkid);
break;
case DRR_FREEOBJECTS:
DO64(drr_freeobjects.drr_firstobj);
DO64(drr_freeobjects.drr_numobjs);
DO64(drr_freeobjects.drr_toguid);
break;
case DRR_WRITE:
DO64(drr_write.drr_object);
DO32(drr_write.drr_type);
DO64(drr_write.drr_offset);
DO64(drr_write.drr_logical_size);
DO64(drr_write.drr_toguid);
ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_write.drr_key.ddk_cksum);
DO64(drr_write.drr_key.ddk_prop);
DO64(drr_write.drr_compressed_size);
break;
case DRR_WRITE_EMBEDDED:
DO64(drr_write_embedded.drr_object);
DO64(drr_write_embedded.drr_offset);
DO64(drr_write_embedded.drr_length);
DO64(drr_write_embedded.drr_toguid);
DO32(drr_write_embedded.drr_lsize);
DO32(drr_write_embedded.drr_psize);
break;
case DRR_FREE:
DO64(drr_free.drr_object);
DO64(drr_free.drr_offset);
DO64(drr_free.drr_length);
DO64(drr_free.drr_toguid);
break;
case DRR_SPILL:
DO64(drr_spill.drr_object);
DO64(drr_spill.drr_length);
DO64(drr_spill.drr_toguid);
DO64(drr_spill.drr_compressed_size);
DO32(drr_spill.drr_type);
break;
case DRR_OBJECT_RANGE:
DO64(drr_object_range.drr_firstobj);
DO64(drr_object_range.drr_numslots);
DO64(drr_object_range.drr_toguid);
break;
case DRR_REDACT:
DO64(drr_redact.drr_object);
DO64(drr_redact.drr_offset);
DO64(drr_redact.drr_length);
DO64(drr_redact.drr_toguid);
break;
case DRR_END:
DO64(drr_end.drr_toguid);
ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_end.drr_checksum);
break;
default:
break;
}
if (drr->drr_type != DRR_BEGIN) {
ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_checksum.drr_checksum);
}
#undef DO64
#undef DO32
}
static boolean_t
redact_snaps_contains(uint64_t *snaps, uint64_t num_snaps, uint64_t guid)
{
for (int i = 0; i < num_snaps; i++) {
if (snaps[i] == guid)
return (B_TRUE);
}
return (B_FALSE);
}
/*
* Check that the new stream we're trying to receive is redacted with respect to
* a subset of the snapshots that the origin was redacted with respect to. For
* the reasons behind this, see the man page on redacted zfs sends and receives.
*/
static boolean_t
compatible_redact_snaps(uint64_t *origin_snaps, uint64_t origin_num_snaps,
uint64_t *redact_snaps, uint64_t num_redact_snaps)
{
/*
* Short circuit the comparison; if we are redacted with respect to
* more snapshots than the origin, we can't be redacted with respect
* to a subset.
*/
if (num_redact_snaps > origin_num_snaps) {
return (B_FALSE);
}
for (int i = 0; i < num_redact_snaps; i++) {
if (!redact_snaps_contains(origin_snaps, origin_num_snaps,
redact_snaps[i])) {
return (B_FALSE);
}
}
return (B_TRUE);
}
static boolean_t
redact_check(dmu_recv_begin_arg_t *drba, dsl_dataset_t *origin)
{
uint64_t *origin_snaps;
uint64_t origin_num_snaps;
dmu_recv_cookie_t *drc = drba->drba_cookie;
struct drr_begin *drrb = drc->drc_drrb;
int featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
int err = 0;
boolean_t ret = B_TRUE;
uint64_t *redact_snaps;
uint_t numredactsnaps;
/*
* If this is a full send stream, we're safe no matter what.
*/
if (drrb->drr_fromguid == 0)
return (ret);
VERIFY(dsl_dataset_get_uint64_array_feature(origin,
SPA_FEATURE_REDACTED_DATASETS, &origin_num_snaps, &origin_snaps));
if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_FROM_SNAPS, &redact_snaps, &numredactsnaps) ==
0) {
/*
* If the send stream was sent from the redaction bookmark or
* the redacted version of the dataset, then we're safe. Verify
* that this is from the a compatible redaction bookmark or
* redacted dataset.
*/
if (!compatible_redact_snaps(origin_snaps, origin_num_snaps,
redact_snaps, numredactsnaps)) {
err = EINVAL;
}
} else if (featureflags & DMU_BACKUP_FEATURE_REDACTED) {
/*
* If the stream is redacted, it must be redacted with respect
* to a subset of what the origin is redacted with respect to.
* See case number 2 in the zfs man page section on redacted zfs
* send.
*/
err = nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_SNAPS, &redact_snaps, &numredactsnaps);
if (err != 0 || !compatible_redact_snaps(origin_snaps,
origin_num_snaps, redact_snaps, numredactsnaps)) {
err = EINVAL;
}
} else if (!redact_snaps_contains(origin_snaps, origin_num_snaps,
drrb->drr_toguid)) {
/*
* If the stream isn't redacted but the origin is, this must be
* one of the snapshots the origin is redacted with respect to.
* See case number 1 in the zfs man page section on redacted zfs
* send.
*/
err = EINVAL;
}
if (err != 0)
ret = B_FALSE;
return (ret);
}
/*
* If we previously received a stream with --large-block, we don't support
* receiving an incremental on top of it without --large-block. This avoids
* forcing a read-modify-write or trying to re-aggregate a string of WRITE
* records.
*/
static int
recv_check_large_blocks(dsl_dataset_t *ds, uint64_t featureflags)
{
if (dsl_dataset_feature_is_active(ds, SPA_FEATURE_LARGE_BLOCKS) &&
!(featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS))
return (SET_ERROR(ZFS_ERR_STREAM_LARGE_BLOCK_MISMATCH));
return (0);
}
static int
recv_begin_check_existing_impl(dmu_recv_begin_arg_t *drba, dsl_dataset_t *ds,
uint64_t fromguid, uint64_t featureflags)
{
uint64_t val;
uint64_t children;
int error;
dsl_pool_t *dp = ds->ds_dir->dd_pool;
boolean_t encrypted = ds->ds_dir->dd_crypto_obj != 0;
boolean_t raw = (featureflags & DMU_BACKUP_FEATURE_RAW) != 0;
boolean_t embed = (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) != 0;
/* Temporary clone name must not exist. */
error = zap_lookup(dp->dp_meta_objset,
dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, recv_clone_name,
8, 1, &val);
if (error != ENOENT)
return (error == 0 ? SET_ERROR(EBUSY) : error);
/* Resume state must not be set. */
if (dsl_dataset_has_resume_receive_state(ds))
return (SET_ERROR(EBUSY));
/* New snapshot name must not exist. */
error = zap_lookup(dp->dp_meta_objset,
dsl_dataset_phys(ds)->ds_snapnames_zapobj,
drba->drba_cookie->drc_tosnap, 8, 1, &val);
if (error != ENOENT)
return (error == 0 ? SET_ERROR(EEXIST) : error);
/* Must not have children if receiving a ZVOL. */
error = zap_count(dp->dp_meta_objset,
dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, &children);
if (error != 0)
return (error);
if (drba->drba_cookie->drc_drrb->drr_type != DMU_OST_ZFS &&
children > 0)
return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
/*
* Check snapshot limit before receiving. We'll recheck again at the
* end, but might as well abort before receiving if we're already over
* the limit.
*
* Note that we do not check the file system limit with
* dsl_dir_fscount_check because the temporary %clones don't count
* against that limit.
*/
error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_SNAPSHOT_LIMIT,
NULL, drba->drba_cred, drba->drba_proc);
if (error != 0)
return (error);
if (fromguid != 0) {
dsl_dataset_t *snap;
uint64_t obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
/* Can't perform a raw receive on top of a non-raw receive */
if (!encrypted && raw)
return (SET_ERROR(EINVAL));
/* Encryption is incompatible with embedded data */
if (encrypted && embed)
return (SET_ERROR(EINVAL));
/* Find snapshot in this dir that matches fromguid. */
while (obj != 0) {
error = dsl_dataset_hold_obj(dp, obj, FTAG,
&snap);
if (error != 0)
return (SET_ERROR(ENODEV));
if (snap->ds_dir != ds->ds_dir) {
dsl_dataset_rele(snap, FTAG);
return (SET_ERROR(ENODEV));
}
if (dsl_dataset_phys(snap)->ds_guid == fromguid)
break;
obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
dsl_dataset_rele(snap, FTAG);
}
if (obj == 0)
return (SET_ERROR(ENODEV));
if (drba->drba_cookie->drc_force) {
drba->drba_cookie->drc_fromsnapobj = obj;
} else {
/*
* If we are not forcing, there must be no
* changes since fromsnap. Raw sends have an
* additional constraint that requires that
* no "noop" snapshots exist between fromsnap
* and tosnap for the IVset checking code to
* work properly.
*/
if (dsl_dataset_modified_since_snap(ds, snap) ||
(raw &&
dsl_dataset_phys(ds)->ds_prev_snap_obj !=
snap->ds_object)) {
dsl_dataset_rele(snap, FTAG);
return (SET_ERROR(ETXTBSY));
}
drba->drba_cookie->drc_fromsnapobj =
ds->ds_prev->ds_object;
}
if (dsl_dataset_feature_is_active(snap,
SPA_FEATURE_REDACTED_DATASETS) && !redact_check(drba,
snap)) {
dsl_dataset_rele(snap, FTAG);
return (SET_ERROR(EINVAL));
}
error = recv_check_large_blocks(snap, featureflags);
if (error != 0) {
dsl_dataset_rele(snap, FTAG);
return (error);
}
dsl_dataset_rele(snap, FTAG);
} else {
/* if full, then must be forced */
if (!drba->drba_cookie->drc_force)
return (SET_ERROR(EEXIST));
/*
* We don't support using zfs recv -F to blow away
* encrypted filesystems. This would require the
* dsl dir to point to the old encryption key and
* the new one at the same time during the receive.
*/
if ((!encrypted && raw) || encrypted)
return (SET_ERROR(EINVAL));
/*
* Perform the same encryption checks we would if
* we were creating a new dataset from scratch.
*/
if (!raw) {
boolean_t will_encrypt;
error = dmu_objset_create_crypt_check(
ds->ds_dir->dd_parent, drba->drba_dcp,
&will_encrypt);
if (error != 0)
return (error);
if (will_encrypt && embed)
return (SET_ERROR(EINVAL));
}
}
return (0);
}
/*
* Check that any feature flags used in the data stream we're receiving are
* supported by the pool we are receiving into.
*
* Note that some of the features we explicitly check here have additional
* (implicit) features they depend on, but those dependencies are enforced
* through the zfeature_register() calls declaring the features that we
* explicitly check.
*/
static int
recv_begin_check_feature_flags_impl(uint64_t featureflags, spa_t *spa)
{
/*
* Check if there are any unsupported feature flags.
*/
if (!DMU_STREAM_SUPPORTED(featureflags)) {
return (SET_ERROR(ZFS_ERR_UNKNOWN_SEND_STREAM_FEATURE));
}
/* Verify pool version supports SA if SA_SPILL feature set */
if ((featureflags & DMU_BACKUP_FEATURE_SA_SPILL) &&
spa_version(spa) < SPA_VERSION_SA)
return (SET_ERROR(ENOTSUP));
/*
* LZ4 compressed, ZSTD compressed, embedded, mooched, large blocks,
* and large_dnodes in the stream can only be used if those pool
* features are enabled because we don't attempt to decompress /
* un-embed / un-mooch / split up the blocks / dnodes during the
* receive process.
*/
if ((featureflags & DMU_BACKUP_FEATURE_LZ4) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_LZ4_COMPRESS))
return (SET_ERROR(ENOTSUP));
if ((featureflags & DMU_BACKUP_FEATURE_ZSTD) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_ZSTD_COMPRESS))
return (SET_ERROR(ENOTSUP));
if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA))
return (SET_ERROR(ENOTSUP));
if ((featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS))
return (SET_ERROR(ENOTSUP));
if ((featureflags & DMU_BACKUP_FEATURE_LARGE_DNODE) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE))
return (SET_ERROR(ENOTSUP));
/*
* Receiving redacted streams requires that redacted datasets are
* enabled.
*/
if ((featureflags & DMU_BACKUP_FEATURE_REDACTED) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_REDACTED_DATASETS))
return (SET_ERROR(ENOTSUP));
return (0);
}
static int
dmu_recv_begin_check(void *arg, dmu_tx_t *tx)
{
dmu_recv_begin_arg_t *drba = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
struct drr_begin *drrb = drba->drba_cookie->drc_drrb;
uint64_t fromguid = drrb->drr_fromguid;
int flags = drrb->drr_flags;
ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
int error;
uint64_t featureflags = drba->drba_cookie->drc_featureflags;
dsl_dataset_t *ds;
const char *tofs = drba->drba_cookie->drc_tofs;
/* already checked */
ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC);
ASSERT(!(featureflags & DMU_BACKUP_FEATURE_RESUMING));
if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
DMU_COMPOUNDSTREAM ||
drrb->drr_type >= DMU_OST_NUMTYPES ||
((flags & DRR_FLAG_CLONE) && drba->drba_origin == NULL))
return (SET_ERROR(EINVAL));
error = recv_begin_check_feature_flags_impl(featureflags, dp->dp_spa);
if (error != 0)
return (error);
/* Resumable receives require extensible datasets */
if (drba->drba_cookie->drc_resumable &&
!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EXTENSIBLE_DATASET))
return (SET_ERROR(ENOTSUP));
if (featureflags & DMU_BACKUP_FEATURE_RAW) {
/* raw receives require the encryption feature */
if (!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_ENCRYPTION))
return (SET_ERROR(ENOTSUP));
/* embedded data is incompatible with encryption and raw recv */
if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)
return (SET_ERROR(EINVAL));
/* raw receives require spill block allocation flag */
if (!(flags & DRR_FLAG_SPILL_BLOCK))
return (SET_ERROR(ZFS_ERR_SPILL_BLOCK_FLAG_MISSING));
} else {
dsflags |= DS_HOLD_FLAG_DECRYPT;
}
error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
if (error == 0) {
/* target fs already exists; recv into temp clone */
/* Can't recv a clone into an existing fs */
if (flags & DRR_FLAG_CLONE || drba->drba_origin) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
error = recv_begin_check_existing_impl(drba, ds, fromguid,
featureflags);
dsl_dataset_rele_flags(ds, dsflags, FTAG);
} else if (error == ENOENT) {
/* target fs does not exist; must be a full backup or clone */
char buf[ZFS_MAX_DATASET_NAME_LEN];
objset_t *os;
/*
* If it's a non-clone incremental, we are missing the
* target fs, so fail the recv.
*/
if (fromguid != 0 && !((flags & DRR_FLAG_CLONE) ||
drba->drba_origin))
return (SET_ERROR(ENOENT));
/*
* If we're receiving a full send as a clone, and it doesn't
* contain all the necessary free records and freeobject
* records, reject it.
*/
if (fromguid == 0 && drba->drba_origin != NULL &&
!(flags & DRR_FLAG_FREERECORDS))
return (SET_ERROR(EINVAL));
/* Open the parent of tofs */
ASSERT3U(strlen(tofs), <, sizeof (buf));
(void) strlcpy(buf, tofs, strrchr(tofs, '/') - tofs + 1);
error = dsl_dataset_hold(dp, buf, FTAG, &ds);
if (error != 0)
return (error);
if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0 &&
drba->drba_origin == NULL) {
boolean_t will_encrypt;
/*
* Check that we aren't breaking any encryption rules
* and that we have all the parameters we need to
* create an encrypted dataset if necessary. If we are
* making an encrypted dataset the stream can't have
* embedded data.
*/
error = dmu_objset_create_crypt_check(ds->ds_dir,
drba->drba_dcp, &will_encrypt);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
if (will_encrypt &&
(featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EINVAL));
}
}
/*
* Check filesystem and snapshot limits before receiving. We'll
* recheck snapshot limits again at the end (we create the
* filesystems and increment those counts during begin_sync).
*/
error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
ZFS_PROP_FILESYSTEM_LIMIT, NULL,
drba->drba_cred, drba->drba_proc);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
ZFS_PROP_SNAPSHOT_LIMIT, NULL,
drba->drba_cred, drba->drba_proc);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
/* can't recv below anything but filesystems (eg. no ZVOLs) */
error = dmu_objset_from_ds(ds, &os);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
if (dmu_objset_type(os) != DMU_OST_ZFS) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
}
if (drba->drba_origin != NULL) {
dsl_dataset_t *origin;
error = dsl_dataset_hold_flags(dp, drba->drba_origin,
dsflags, FTAG, &origin);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
if (!origin->ds_is_snapshot) {
dsl_dataset_rele_flags(origin, dsflags, FTAG);
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EINVAL));
}
if (dsl_dataset_phys(origin)->ds_guid != fromguid &&
fromguid != 0) {
dsl_dataset_rele_flags(origin, dsflags, FTAG);
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(ENODEV));
}
if (origin->ds_dir->dd_crypto_obj != 0 &&
(featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)) {
dsl_dataset_rele_flags(origin, dsflags, FTAG);
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EINVAL));
}
/*
* If the origin is redacted we need to verify that this
* send stream can safely be received on top of the
* origin.
*/
if (dsl_dataset_feature_is_active(origin,
SPA_FEATURE_REDACTED_DATASETS)) {
if (!redact_check(drba, origin)) {
dsl_dataset_rele_flags(origin, dsflags,
FTAG);
dsl_dataset_rele_flags(ds, dsflags,
FTAG);
return (SET_ERROR(EINVAL));
}
}
error = recv_check_large_blocks(ds, featureflags);
if (error != 0) {
dsl_dataset_rele_flags(origin, dsflags, FTAG);
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (error);
}
dsl_dataset_rele_flags(origin, dsflags, FTAG);
}
dsl_dataset_rele(ds, FTAG);
error = 0;
}
return (error);
}
static void
dmu_recv_begin_sync(void *arg, dmu_tx_t *tx)
{
dmu_recv_begin_arg_t *drba = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
objset_t *mos = dp->dp_meta_objset;
dmu_recv_cookie_t *drc = drba->drba_cookie;
struct drr_begin *drrb = drc->drc_drrb;
const char *tofs = drc->drc_tofs;
uint64_t featureflags = drc->drc_featureflags;
dsl_dataset_t *ds, *newds;
objset_t *os;
uint64_t dsobj;
ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
int error;
uint64_t crflags = 0;
dsl_crypto_params_t dummy_dcp = { 0 };
dsl_crypto_params_t *dcp = drba->drba_dcp;
if (drrb->drr_flags & DRR_FLAG_CI_DATA)
crflags |= DS_FLAG_CI_DATASET;
if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0)
dsflags |= DS_HOLD_FLAG_DECRYPT;
/*
* Raw, non-incremental recvs always use a dummy dcp with
* the raw cmd set. Raw incremental recvs do not use a dcp
* since the encryption parameters are already set in stone.
*/
if (dcp == NULL && drrb->drr_fromguid == 0 &&
drba->drba_origin == NULL) {
ASSERT3P(dcp, ==, NULL);
dcp = &dummy_dcp;
if (featureflags & DMU_BACKUP_FEATURE_RAW)
dcp->cp_cmd = DCP_CMD_RAW_RECV;
}
error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
if (error == 0) {
/* create temporary clone */
dsl_dataset_t *snap = NULL;
if (drba->drba_cookie->drc_fromsnapobj != 0) {
VERIFY0(dsl_dataset_hold_obj(dp,
drba->drba_cookie->drc_fromsnapobj, FTAG, &snap));
ASSERT3P(dcp, ==, NULL);
}
dsobj = dsl_dataset_create_sync(ds->ds_dir, recv_clone_name,
snap, crflags, drba->drba_cred, dcp, tx);
if (drba->drba_cookie->drc_fromsnapobj != 0)
dsl_dataset_rele(snap, FTAG);
dsl_dataset_rele_flags(ds, dsflags, FTAG);
} else {
dsl_dir_t *dd;
const char *tail;
dsl_dataset_t *origin = NULL;
VERIFY0(dsl_dir_hold(dp, tofs, FTAG, &dd, &tail));
if (drba->drba_origin != NULL) {
VERIFY0(dsl_dataset_hold(dp, drba->drba_origin,
FTAG, &origin));
ASSERT3P(dcp, ==, NULL);
}
/* Create new dataset. */
dsobj = dsl_dataset_create_sync(dd, strrchr(tofs, '/') + 1,
origin, crflags, drba->drba_cred, dcp, tx);
if (origin != NULL)
dsl_dataset_rele(origin, FTAG);
dsl_dir_rele(dd, FTAG);
drc->drc_newfs = B_TRUE;
}
VERIFY0(dsl_dataset_own_obj_force(dp, dsobj, dsflags, dmu_recv_tag,
&newds));
if (dsl_dataset_feature_is_active(newds,
SPA_FEATURE_REDACTED_DATASETS)) {
/*
* If the origin dataset is redacted, the child will be redacted
* when we create it. We clear the new dataset's
* redaction info; if it should be redacted, we'll fill
* in its information later.
*/
dsl_dataset_deactivate_feature(newds,
SPA_FEATURE_REDACTED_DATASETS, tx);
}
VERIFY0(dmu_objset_from_ds(newds, &os));
if (drc->drc_resumable) {
dsl_dataset_zapify(newds, tx);
if (drrb->drr_fromguid != 0) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_FROMGUID,
8, 1, &drrb->drr_fromguid, tx));
}
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TOGUID,
8, 1, &drrb->drr_toguid, tx));
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TONAME,
1, strlen(drrb->drr_toname) + 1, drrb->drr_toname, tx));
uint64_t one = 1;
uint64_t zero = 0;
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OBJECT,
8, 1, &one, tx));
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OFFSET,
8, 1, &zero, tx));
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_BYTES,
8, 1, &zero, tx));
if (featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_LARGEBLOCK,
8, 1, &one, tx));
}
if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_EMBEDOK,
8, 1, &one, tx));
}
if (featureflags & DMU_BACKUP_FEATURE_COMPRESSED) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_COMPRESSOK,
8, 1, &one, tx));
}
if (featureflags & DMU_BACKUP_FEATURE_RAW) {
VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_RAWOK,
8, 1, &one, tx));
}
uint64_t *redact_snaps;
uint_t numredactsnaps;
if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_FROM_SNAPS, &redact_snaps,
&numredactsnaps) == 0) {
VERIFY0(zap_add(mos, dsobj,
DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS,
sizeof (*redact_snaps), numredactsnaps,
redact_snaps, tx));
}
}
/*
* Usually the os->os_encrypted value is tied to the presence of a
* DSL Crypto Key object in the dd. However, that will not be received
* until dmu_recv_stream(), so we set the value manually for now.
*/
if (featureflags & DMU_BACKUP_FEATURE_RAW) {
os->os_encrypted = B_TRUE;
drba->drba_cookie->drc_raw = B_TRUE;
}
if (featureflags & DMU_BACKUP_FEATURE_REDACTED) {
uint64_t *redact_snaps;
uint_t numredactsnaps;
VERIFY0(nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_SNAPS, &redact_snaps, &numredactsnaps));
dsl_dataset_activate_redaction(newds, redact_snaps,
numredactsnaps, tx);
}
dmu_buf_will_dirty(newds->ds_dbuf, tx);
dsl_dataset_phys(newds)->ds_flags |= DS_FLAG_INCONSISTENT;
/*
* If we actually created a non-clone, we need to create the objset
* in our new dataset. If this is a raw send we postpone this until
* dmu_recv_stream() so that we can allocate the metadnode with the
* properties from the DRR_BEGIN payload.
*/
rrw_enter(&newds->ds_bp_rwlock, RW_READER, FTAG);
if (BP_IS_HOLE(dsl_dataset_get_blkptr(newds)) &&
(featureflags & DMU_BACKUP_FEATURE_RAW) == 0) {
(void) dmu_objset_create_impl(dp->dp_spa,
newds, dsl_dataset_get_blkptr(newds), drrb->drr_type, tx);
}
rrw_exit(&newds->ds_bp_rwlock, FTAG);
drba->drba_cookie->drc_ds = newds;
drba->drba_cookie->drc_os = os;
spa_history_log_internal_ds(newds, "receive", tx, " ");
}
static int
dmu_recv_resume_begin_check(void *arg, dmu_tx_t *tx)
{
dmu_recv_begin_arg_t *drba = arg;
dmu_recv_cookie_t *drc = drba->drba_cookie;
dsl_pool_t *dp = dmu_tx_pool(tx);
struct drr_begin *drrb = drc->drc_drrb;
int error;
ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
dsl_dataset_t *ds;
const char *tofs = drc->drc_tofs;
/* already checked */
ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC);
ASSERT(drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING);
if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
DMU_COMPOUNDSTREAM ||
drrb->drr_type >= DMU_OST_NUMTYPES)
return (SET_ERROR(EINVAL));
/*
* This is mostly a sanity check since we should have already done these
* checks during a previous attempt to receive the data.
*/
error = recv_begin_check_feature_flags_impl(drc->drc_featureflags,
dp->dp_spa);
if (error != 0)
return (error);
/* 6 extra bytes for /%recv */
char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];
(void) snprintf(recvname, sizeof (recvname), "%s/%s",
tofs, recv_clone_name);
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) {
/* raw receives require spill block allocation flag */
if (!(drrb->drr_flags & DRR_FLAG_SPILL_BLOCK))
return (SET_ERROR(ZFS_ERR_SPILL_BLOCK_FLAG_MISSING));
} else {
dsflags |= DS_HOLD_FLAG_DECRYPT;
}
if (dsl_dataset_hold_flags(dp, recvname, dsflags, FTAG, &ds) != 0) {
/* %recv does not exist; continue in tofs */
error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
if (error != 0)
return (error);
}
/* check that ds is marked inconsistent */
if (!DS_IS_INCONSISTENT(ds)) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
/* check that there is resuming data, and that the toguid matches */
if (!dsl_dataset_is_zapified(ds)) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
uint64_t val;
error = zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val);
if (error != 0 || drrb->drr_toguid != val) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
/*
* Check if the receive is still running. If so, it will be owned.
* Note that nothing else can own the dataset (e.g. after the receive
* fails) because it will be marked inconsistent.
*/
if (dsl_dataset_has_owner(ds)) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EBUSY));
}
/* There should not be any snapshots of this fs yet. */
if (ds->ds_prev != NULL && ds->ds_prev->ds_dir == ds->ds_dir) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
/*
* Note: resume point will be checked when we process the first WRITE
* record.
*/
/* check that the origin matches */
val = 0;
(void) zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val);
if (drrb->drr_fromguid != val) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
if (ds->ds_prev != NULL && drrb->drr_fromguid != 0)
drc->drc_fromsnapobj = ds->ds_prev->ds_object;
/*
* If we're resuming, and the send is redacted, then the original send
* must have been redacted, and must have been redacted with respect to
* the same snapshots.
*/
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_REDACTED) {
uint64_t num_ds_redact_snaps;
uint64_t *ds_redact_snaps;
uint_t num_stream_redact_snaps;
uint64_t *stream_redact_snaps;
if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
BEGINNV_REDACT_SNAPS, &stream_redact_snaps,
&num_stream_redact_snaps) != 0) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
if (!dsl_dataset_get_uint64_array_feature(ds,
SPA_FEATURE_REDACTED_DATASETS, &num_ds_redact_snaps,
&ds_redact_snaps)) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
for (int i = 0; i < num_ds_redact_snaps; i++) {
if (!redact_snaps_contains(ds_redact_snaps,
num_ds_redact_snaps, stream_redact_snaps[i])) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (SET_ERROR(EINVAL));
}
}
}
error = recv_check_large_blocks(ds, drc->drc_featureflags);
if (error != 0) {
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (error);
}
dsl_dataset_rele_flags(ds, dsflags, FTAG);
return (0);
}
static void
dmu_recv_resume_begin_sync(void *arg, dmu_tx_t *tx)
{
dmu_recv_begin_arg_t *drba = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
const char *tofs = drba->drba_cookie->drc_tofs;
uint64_t featureflags = drba->drba_cookie->drc_featureflags;
dsl_dataset_t *ds;
ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
/* 6 extra bytes for /%recv */
char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];
(void) snprintf(recvname, sizeof (recvname), "%s/%s", tofs,
recv_clone_name);
if (featureflags & DMU_BACKUP_FEATURE_RAW) {
drba->drba_cookie->drc_raw = B_TRUE;
} else {
dsflags |= DS_HOLD_FLAG_DECRYPT;
}
if (dsl_dataset_own_force(dp, recvname, dsflags, dmu_recv_tag, &ds)
!= 0) {
/* %recv does not exist; continue in tofs */
VERIFY0(dsl_dataset_own_force(dp, tofs, dsflags, dmu_recv_tag,
&ds));
drba->drba_cookie->drc_newfs = B_TRUE;
}
ASSERT(DS_IS_INCONSISTENT(ds));
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
ASSERT(!BP_IS_HOLE(dsl_dataset_get_blkptr(ds)) ||
drba->drba_cookie->drc_raw);
rrw_exit(&ds->ds_bp_rwlock, FTAG);
drba->drba_cookie->drc_ds = ds;
VERIFY0(dmu_objset_from_ds(ds, &drba->drba_cookie->drc_os));
drba->drba_cookie->drc_should_save = B_TRUE;
spa_history_log_internal_ds(ds, "resume receive", tx, " ");
}
/*
* NB: callers *MUST* call dmu_recv_stream() if dmu_recv_begin()
* succeeds; otherwise we will leak the holds on the datasets.
*/
int
dmu_recv_begin(char *tofs, char *tosnap, dmu_replay_record_t *drr_begin,
boolean_t force, boolean_t resumable, nvlist_t *localprops,
nvlist_t *hidden_args, char *origin, dmu_recv_cookie_t *drc,
zfs_file_t *fp, offset_t *voffp)
{
dmu_recv_begin_arg_t drba = { 0 };
int err;
bzero(drc, sizeof (dmu_recv_cookie_t));
drc->drc_drr_begin = drr_begin;
drc->drc_drrb = &drr_begin->drr_u.drr_begin;
drc->drc_tosnap = tosnap;
drc->drc_tofs = tofs;
drc->drc_force = force;
drc->drc_resumable = resumable;
drc->drc_cred = CRED();
drc->drc_proc = curproc;
drc->drc_clone = (origin != NULL);
if (drc->drc_drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) {
drc->drc_byteswap = B_TRUE;
(void) fletcher_4_incremental_byteswap(drr_begin,
sizeof (dmu_replay_record_t), &drc->drc_cksum);
byteswap_record(drr_begin);
} else if (drc->drc_drrb->drr_magic == DMU_BACKUP_MAGIC) {
(void) fletcher_4_incremental_native(drr_begin,
sizeof (dmu_replay_record_t), &drc->drc_cksum);
} else {
return (SET_ERROR(EINVAL));
}
drc->drc_fp = fp;
drc->drc_voff = *voffp;
drc->drc_featureflags =
DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo);
uint32_t payloadlen = drc->drc_drr_begin->drr_payloadlen;
void *payload = NULL;
if (payloadlen != 0)
payload = kmem_alloc(payloadlen, KM_SLEEP);
err = receive_read_payload_and_next_header(drc, payloadlen,
payload);
if (err != 0) {
kmem_free(payload, payloadlen);
return (err);
}
if (payloadlen != 0) {
err = nvlist_unpack(payload, payloadlen, &drc->drc_begin_nvl,
KM_SLEEP);
kmem_free(payload, payloadlen);
if (err != 0) {
kmem_free(drc->drc_next_rrd,
sizeof (*drc->drc_next_rrd));
return (err);
}
}
if (drc->drc_drrb->drr_flags & DRR_FLAG_SPILL_BLOCK)
drc->drc_spill = B_TRUE;
drba.drba_origin = origin;
drba.drba_cookie = drc;
drba.drba_cred = CRED();
drba.drba_proc = curproc;
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING) {
err = dsl_sync_task(tofs,
dmu_recv_resume_begin_check, dmu_recv_resume_begin_sync,
&drba, 5, ZFS_SPACE_CHECK_NORMAL);
} else {
/*
* For non-raw, non-incremental, non-resuming receives the
* user can specify encryption parameters on the command line
* with "zfs recv -o". For these receives we create a dcp and
* pass it to the sync task. Creating the dcp will implicitly
* remove the encryption params from the localprops nvlist,
* which avoids errors when trying to set these normally
* read-only properties. Any other kind of receive that
* attempts to set these properties will fail as a result.
*/
if ((DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo) &
DMU_BACKUP_FEATURE_RAW) == 0 &&
origin == NULL && drc->drc_drrb->drr_fromguid == 0) {
err = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
localprops, hidden_args, &drba.drba_dcp);
}
if (err == 0) {
err = dsl_sync_task(tofs,
dmu_recv_begin_check, dmu_recv_begin_sync,
&drba, 5, ZFS_SPACE_CHECK_NORMAL);
dsl_crypto_params_free(drba.drba_dcp, !!err);
}
}
if (err != 0) {
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
nvlist_free(drc->drc_begin_nvl);
}
return (err);
}
static int
receive_read(dmu_recv_cookie_t *drc, int len, void *buf)
{
int done = 0;
/*
* The code doesn't rely on this (lengths being multiples of 8). See
* comment in dump_bytes.
*/
ASSERT(len % 8 == 0 ||
(drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) != 0);
while (done < len) {
ssize_t resid;
zfs_file_t *fp = drc->drc_fp;
int err = zfs_file_read(fp, (char *)buf + done,
len - done, &resid);
if (resid == len - done) {
/*
* Note: ECKSUM or ZFS_ERR_STREAM_TRUNCATED indicates
* that the receive was interrupted and can
* potentially be resumed.
*/
err = SET_ERROR(ZFS_ERR_STREAM_TRUNCATED);
}
drc->drc_voff += len - done - resid;
done = len - resid;
if (err != 0)
return (err);
}
drc->drc_bytes_read += len;
ASSERT3U(done, ==, len);
return (0);
}
static inline uint8_t
deduce_nblkptr(dmu_object_type_t bonus_type, uint64_t bonus_size)
{
if (bonus_type == DMU_OT_SA) {
return (1);
} else {
return (1 +
((DN_OLD_MAX_BONUSLEN -
MIN(DN_OLD_MAX_BONUSLEN, bonus_size)) >> SPA_BLKPTRSHIFT));
}
}
static void
save_resume_state(struct receive_writer_arg *rwa,
uint64_t object, uint64_t offset, dmu_tx_t *tx)
{
int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;
if (!rwa->resumable)
return;
/*
* We use ds_resume_bytes[] != 0 to indicate that we need to
* update this on disk, so it must not be 0.
*/
ASSERT(rwa->bytes_read != 0);
/*
* We only resume from write records, which have a valid
* (non-meta-dnode) object number.
*/
ASSERT(object != 0);
/*
* For resuming to work correctly, we must receive records in order,
* sorted by object,offset. This is checked by the callers, but
* assert it here for good measure.
*/
ASSERT3U(object, >=, rwa->os->os_dsl_dataset->ds_resume_object[txgoff]);
ASSERT(object != rwa->os->os_dsl_dataset->ds_resume_object[txgoff] ||
offset >= rwa->os->os_dsl_dataset->ds_resume_offset[txgoff]);
ASSERT3U(rwa->bytes_read, >=,
rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff]);
rwa->os->os_dsl_dataset->ds_resume_object[txgoff] = object;
rwa->os->os_dsl_dataset->ds_resume_offset[txgoff] = offset;
rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff] = rwa->bytes_read;
}
static int
receive_object_is_same_generation(objset_t *os, uint64_t object,
dmu_object_type_t old_bonus_type, dmu_object_type_t new_bonus_type,
const void *new_bonus, boolean_t *samegenp)
{
zfs_file_info_t zoi;
int err;
dmu_buf_t *old_bonus_dbuf;
err = dmu_bonus_hold(os, object, FTAG, &old_bonus_dbuf);
if (err != 0)
return (err);
err = dmu_get_file_info(os, old_bonus_type, old_bonus_dbuf->db_data,
&zoi);
dmu_buf_rele(old_bonus_dbuf, FTAG);
if (err != 0)
return (err);
uint64_t old_gen = zoi.zfi_generation;
err = dmu_get_file_info(os, new_bonus_type, new_bonus, &zoi);
if (err != 0)
return (err);
uint64_t new_gen = zoi.zfi_generation;
*samegenp = (old_gen == new_gen);
return (0);
}
static int
receive_handle_existing_object(const struct receive_writer_arg *rwa,
const struct drr_object *drro, const dmu_object_info_t *doi,
const void *bonus_data,
uint64_t *object_to_hold, uint32_t *new_blksz)
{
uint32_t indblksz = drro->drr_indblkshift ?
1ULL << drro->drr_indblkshift : 0;
int nblkptr = deduce_nblkptr(drro->drr_bonustype,
drro->drr_bonuslen);
uint8_t dn_slots = drro->drr_dn_slots != 0 ?
drro->drr_dn_slots : DNODE_MIN_SLOTS;
boolean_t do_free_range = B_FALSE;
int err;
*object_to_hold = drro->drr_object;
/* nblkptr should be bounded by the bonus size and type */
if (rwa->raw && nblkptr != drro->drr_nblkptr)
return (SET_ERROR(EINVAL));
/*
* After the previous send stream, the sending system may
* have freed this object, and then happened to re-allocate
* this object number in a later txg. In this case, we are
* receiving a different logical file, and the block size may
* appear to be different. i.e. we may have a different
* block size for this object than what the send stream says.
* In this case we need to remove the object's contents,
* so that its structure can be changed and then its contents
* entirely replaced by subsequent WRITE records.
*
* If this is a -L (--large-block) incremental stream, and
* the previous stream was not -L, the block size may appear
* to increase. i.e. we may have a smaller block size for
* this object than what the send stream says. In this case
* we need to keep the object's contents and block size
* intact, so that we don't lose parts of the object's
* contents that are not changed by this incremental send
* stream.
*
* We can distinguish between the two above cases by using
* the ZPL's generation number (see
* receive_object_is_same_generation()). However, we only
* want to rely on the generation number when absolutely
* necessary, because with raw receives, the generation is
* encrypted. We also want to minimize dependence on the
* ZPL, so that other types of datasets can also be received
* (e.g. ZVOLs, although note that ZVOLS currently do not
* reallocate their objects or change their structure).
* Therefore, we check a number of different cases where we
* know it is safe to discard the object's contents, before
* using the ZPL's generation number to make the above
* distinction.
*/
if (drro->drr_blksz != doi->doi_data_block_size) {
if (rwa->raw) {
/*
* RAW streams always have large blocks, so
* we are sure that the data is not needed
* due to changing --large-block to be on.
* Which is fortunate since the bonus buffer
* (which contains the ZPL generation) is
* encrypted, and the key might not be
* loaded.
*/
do_free_range = B_TRUE;
} else if (rwa->full) {
/*
* This is a full send stream, so it always
* replaces what we have. Even if the
* generation numbers happen to match, this
* can not actually be the same logical file.
* This is relevant when receiving a full
* send as a clone.
*/
do_free_range = B_TRUE;
} else if (drro->drr_type !=
DMU_OT_PLAIN_FILE_CONTENTS ||
doi->doi_type != DMU_OT_PLAIN_FILE_CONTENTS) {
/*
* PLAIN_FILE_CONTENTS are the only type of
* objects that have ever been stored with
* large blocks, so we don't need the special
* logic below. ZAP blocks can shrink (when
* there's only one block), so we don't want
* to hit the error below about block size
* only increasing.
*/
do_free_range = B_TRUE;
} else if (doi->doi_max_offset <=
doi->doi_data_block_size) {
/*
* There is only one block. We can free it,
* because its contents will be replaced by a
* WRITE record. This can not be the no-L ->
* -L case, because the no-L case would have
* resulted in multiple blocks. If we
* supported -L -> no-L, it would not be safe
* to free the file's contents. Fortunately,
* that is not allowed (see
* recv_check_large_blocks()).
*/
do_free_range = B_TRUE;
} else {
boolean_t is_same_gen;
err = receive_object_is_same_generation(rwa->os,
drro->drr_object, doi->doi_bonus_type,
drro->drr_bonustype, bonus_data, &is_same_gen);
if (err != 0)
return (SET_ERROR(EINVAL));
if (is_same_gen) {
/*
* This is the same logical file, and
* the block size must be increasing.
* It could only decrease if
* --large-block was changed to be
* off, which is checked in
* recv_check_large_blocks().
*/
if (drro->drr_blksz <=
doi->doi_data_block_size)
return (SET_ERROR(EINVAL));
/*
* We keep the existing blocksize and
* contents.
*/
*new_blksz =
doi->doi_data_block_size;
} else {
do_free_range = B_TRUE;
}
}
}
/* nblkptr can only decrease if the object was reallocated */
if (nblkptr < doi->doi_nblkptr)
do_free_range = B_TRUE;
/* number of slots can only change on reallocation */
if (dn_slots != doi->doi_dnodesize >> DNODE_SHIFT)
do_free_range = B_TRUE;
/*
* For raw sends we also check a few other fields to
* ensure we are preserving the objset structure exactly
* as it was on the receive side:
* - A changed indirect block size
* - A smaller nlevels
*/
if (rwa->raw) {
if (indblksz != doi->doi_metadata_block_size)
do_free_range = B_TRUE;
if (drro->drr_nlevels < doi->doi_indirection)
do_free_range = B_TRUE;
}
if (do_free_range) {
err = dmu_free_long_range(rwa->os, drro->drr_object,
0, DMU_OBJECT_END);
if (err != 0)
return (SET_ERROR(EINVAL));
}
/*
* The dmu does not currently support decreasing nlevels
* or changing the number of dnode slots on an object. For
* non-raw sends, this does not matter and the new object
* can just use the previous one's nlevels. For raw sends,
* however, the structure of the received dnode (including
* nlevels and dnode slots) must match that of the send
* side. Therefore, instead of using dmu_object_reclaim(),
* we must free the object completely and call
* dmu_object_claim_dnsize() instead.
*/
if ((rwa->raw && drro->drr_nlevels < doi->doi_indirection) ||
dn_slots != doi->doi_dnodesize >> DNODE_SHIFT) {
err = dmu_free_long_object(rwa->os, drro->drr_object);
if (err != 0)
return (SET_ERROR(EINVAL));
txg_wait_synced(dmu_objset_pool(rwa->os), 0);
*object_to_hold = DMU_NEW_OBJECT;
}
/*
* For raw receives, free everything beyond the new incoming
* maxblkid. Normally this would be done with a DRR_FREE
* record that would come after this DRR_OBJECT record is
* processed. However, for raw receives we manually set the
* maxblkid from the drr_maxblkid and so we must first free
* everything above that blkid to ensure the DMU is always
* consistent with itself. We will never free the first block
* of the object here because a maxblkid of 0 could indicate
* an object with a single block or one with no blocks. This
* free may be skipped when dmu_free_long_range() was called
* above since it covers the entire object's contents.
*/
if (rwa->raw && *object_to_hold != DMU_NEW_OBJECT && !do_free_range) {
err = dmu_free_long_range(rwa->os, drro->drr_object,
(drro->drr_maxblkid + 1) * doi->doi_data_block_size,
DMU_OBJECT_END);
if (err != 0)
return (SET_ERROR(EINVAL));
}
return (0);
}
noinline static int
receive_object(struct receive_writer_arg *rwa, struct drr_object *drro,
void *data)
{
dmu_object_info_t doi;
dmu_tx_t *tx;
int err;
uint32_t new_blksz = drro->drr_blksz;
uint8_t dn_slots = drro->drr_dn_slots != 0 ?
drro->drr_dn_slots : DNODE_MIN_SLOTS;
if (drro->drr_type == DMU_OT_NONE ||
!DMU_OT_IS_VALID(drro->drr_type) ||
!DMU_OT_IS_VALID(drro->drr_bonustype) ||
drro->drr_checksumtype >= ZIO_CHECKSUM_FUNCTIONS ||
drro->drr_compress >= ZIO_COMPRESS_FUNCTIONS ||
P2PHASE(drro->drr_blksz, SPA_MINBLOCKSIZE) ||
drro->drr_blksz < SPA_MINBLOCKSIZE ||
drro->drr_blksz > spa_maxblocksize(dmu_objset_spa(rwa->os)) ||
drro->drr_bonuslen >
DN_BONUS_SIZE(spa_maxdnodesize(dmu_objset_spa(rwa->os))) ||
dn_slots >
(spa_maxdnodesize(dmu_objset_spa(rwa->os)) >> DNODE_SHIFT)) {
return (SET_ERROR(EINVAL));
}
if (rwa->raw) {
/*
* We should have received a DRR_OBJECT_RANGE record
* containing this block and stored it in rwa.
*/
if (drro->drr_object < rwa->or_firstobj ||
drro->drr_object >= rwa->or_firstobj + rwa->or_numslots ||
drro->drr_raw_bonuslen < drro->drr_bonuslen ||
drro->drr_indblkshift > SPA_MAXBLOCKSHIFT ||
drro->drr_nlevels > DN_MAX_LEVELS ||
drro->drr_nblkptr > DN_MAX_NBLKPTR ||
DN_SLOTS_TO_BONUSLEN(dn_slots) <
drro->drr_raw_bonuslen)
return (SET_ERROR(EINVAL));
} else {
/*
* The DRR_OBJECT_SPILL flag is valid when the DRR_BEGIN
* record indicates this by setting DRR_FLAG_SPILL_BLOCK.
*/
if (((drro->drr_flags & ~(DRR_OBJECT_SPILL))) ||
(!rwa->spill && DRR_OBJECT_HAS_SPILL(drro->drr_flags))) {
return (SET_ERROR(EINVAL));
}
if (drro->drr_raw_bonuslen != 0 || drro->drr_nblkptr != 0 ||
drro->drr_indblkshift != 0 || drro->drr_nlevels != 0) {
return (SET_ERROR(EINVAL));
}
}
err = dmu_object_info(rwa->os, drro->drr_object, &doi);
if (err != 0 && err != ENOENT && err != EEXIST)
return (SET_ERROR(EINVAL));
if (drro->drr_object > rwa->max_object)
rwa->max_object = drro->drr_object;
/*
* If we are losing blkptrs or changing the block size this must
* be a new file instance. We must clear out the previous file
* contents before we can change this type of metadata in the dnode.
* Raw receives will also check that the indirect structure of the
* dnode hasn't changed.
*/
uint64_t object_to_hold;
if (err == 0) {
err = receive_handle_existing_object(rwa, drro, &doi, data,
&object_to_hold, &new_blksz);
} else if (err == EEXIST) {
/*
* The object requested is currently an interior slot of a
* multi-slot dnode. This will be resolved when the next txg
* is synced out, since the send stream will have told us
* to free this slot when we freed the associated dnode
* earlier in the stream.
*/
txg_wait_synced(dmu_objset_pool(rwa->os), 0);
if (dmu_object_info(rwa->os, drro->drr_object, NULL) != ENOENT)
return (SET_ERROR(EINVAL));
/* object was freed and we are about to allocate a new one */
object_to_hold = DMU_NEW_OBJECT;
} else {
/* object is free and we are about to allocate a new one */
object_to_hold = DMU_NEW_OBJECT;
}
/*
* If this is a multi-slot dnode there is a chance that this
* object will expand into a slot that is already used by
* another object from the previous snapshot. We must free
* these objects before we attempt to allocate the new dnode.
*/
if (dn_slots > 1) {
boolean_t need_sync = B_FALSE;
for (uint64_t slot = drro->drr_object + 1;
slot < drro->drr_object + dn_slots;
slot++) {
dmu_object_info_t slot_doi;
err = dmu_object_info(rwa->os, slot, &slot_doi);
if (err == ENOENT || err == EEXIST)
continue;
else if (err != 0)
return (err);
err = dmu_free_long_object(rwa->os, slot);
if (err != 0)
return (err);
need_sync = B_TRUE;
}
if (need_sync)
txg_wait_synced(dmu_objset_pool(rwa->os), 0);
}
tx = dmu_tx_create(rwa->os);
dmu_tx_hold_bonus(tx, object_to_hold);
dmu_tx_hold_write(tx, object_to_hold, 0, 0);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err != 0) {
dmu_tx_abort(tx);
return (err);
}
if (object_to_hold == DMU_NEW_OBJECT) {
/* Currently free, wants to be allocated */
err = dmu_object_claim_dnsize(rwa->os, drro->drr_object,
drro->drr_type, new_blksz,
drro->drr_bonustype, drro->drr_bonuslen,
dn_slots << DNODE_SHIFT, tx);
} else if (drro->drr_type != doi.doi_type ||
new_blksz != doi.doi_data_block_size ||
drro->drr_bonustype != doi.doi_bonus_type ||
drro->drr_bonuslen != doi.doi_bonus_size) {
/* Currently allocated, but with different properties */
err = dmu_object_reclaim_dnsize(rwa->os, drro->drr_object,
drro->drr_type, new_blksz,
drro->drr_bonustype, drro->drr_bonuslen,
dn_slots << DNODE_SHIFT, rwa->spill ?
DRR_OBJECT_HAS_SPILL(drro->drr_flags) : B_FALSE, tx);
} else if (rwa->spill && !DRR_OBJECT_HAS_SPILL(drro->drr_flags)) {
/*
* Currently allocated, the existing version of this object
* may reference a spill block that is no longer allocated
* at the source and needs to be freed.
*/
err = dmu_object_rm_spill(rwa->os, drro->drr_object, tx);
}
if (err != 0) {
dmu_tx_commit(tx);
return (SET_ERROR(EINVAL));
}
if (rwa->or_crypt_params_present) {
/*
* Set the crypt params for the buffer associated with this
* range of dnodes. This causes the blkptr_t to have the
* same crypt params (byteorder, salt, iv, mac) as on the
* sending side.
*
* Since we are committing this tx now, it is possible for
* the dnode block to end up on-disk with the incorrect MAC,
* if subsequent objects in this block are received in a
* different txg. However, since the dataset is marked as
* inconsistent, no code paths will do a non-raw read (or
* decrypt the block / verify the MAC). The receive code and
* scrub code can safely do raw reads and verify the
* checksum. They don't need to verify the MAC.
*/
dmu_buf_t *db = NULL;
uint64_t offset = rwa->or_firstobj * DNODE_MIN_SIZE;
err = dmu_buf_hold_by_dnode(DMU_META_DNODE(rwa->os),
offset, FTAG, &db, DMU_READ_PREFETCH | DMU_READ_NO_DECRYPT);
if (err != 0) {
dmu_tx_commit(tx);
return (SET_ERROR(EINVAL));
}
dmu_buf_set_crypt_params(db, rwa->or_byteorder,
rwa->or_salt, rwa->or_iv, rwa->or_mac, tx);
dmu_buf_rele(db, FTAG);
rwa->or_crypt_params_present = B_FALSE;
}
dmu_object_set_checksum(rwa->os, drro->drr_object,
drro->drr_checksumtype, tx);
dmu_object_set_compress(rwa->os, drro->drr_object,
drro->drr_compress, tx);
/* handle more restrictive dnode structuring for raw recvs */
if (rwa->raw) {
/*
* Set the indirect block size, block shift, nlevels.
* This will not fail because we ensured all of the
* blocks were freed earlier if this is a new object.
* For non-new objects block size and indirect block
* shift cannot change and nlevels can only increase.
*/
ASSERT3U(new_blksz, ==, drro->drr_blksz);
VERIFY0(dmu_object_set_blocksize(rwa->os, drro->drr_object,
drro->drr_blksz, drro->drr_indblkshift, tx));
VERIFY0(dmu_object_set_nlevels(rwa->os, drro->drr_object,
drro->drr_nlevels, tx));
/*
* Set the maxblkid. This will always succeed because
* we freed all blocks beyond the new maxblkid above.
*/
VERIFY0(dmu_object_set_maxblkid(rwa->os, drro->drr_object,
drro->drr_maxblkid, tx));
}
if (data != NULL) {
dmu_buf_t *db;
dnode_t *dn;
uint32_t flags = DMU_READ_NO_PREFETCH;
if (rwa->raw)
flags |= DMU_READ_NO_DECRYPT;
VERIFY0(dnode_hold(rwa->os, drro->drr_object, FTAG, &dn));
VERIFY0(dmu_bonus_hold_by_dnode(dn, FTAG, &db, flags));
dmu_buf_will_dirty(db, tx);
ASSERT3U(db->db_size, >=, drro->drr_bonuslen);
bcopy(data, db->db_data, DRR_OBJECT_PAYLOAD_SIZE(drro));
/*
* Raw bonus buffers have their byteorder determined by the
* DRR_OBJECT_RANGE record.
*/
if (rwa->byteswap && !rwa->raw) {
dmu_object_byteswap_t byteswap =
DMU_OT_BYTESWAP(drro->drr_bonustype);
dmu_ot_byteswap[byteswap].ob_func(db->db_data,
DRR_OBJECT_PAYLOAD_SIZE(drro));
}
dmu_buf_rele(db, FTAG);
dnode_rele(dn, FTAG);
}
dmu_tx_commit(tx);
return (0);
}
/* ARGSUSED */
noinline static int
receive_freeobjects(struct receive_writer_arg *rwa,
struct drr_freeobjects *drrfo)
{
uint64_t obj;
int next_err = 0;
if (drrfo->drr_firstobj + drrfo->drr_numobjs < drrfo->drr_firstobj)
return (SET_ERROR(EINVAL));
for (obj = drrfo->drr_firstobj == 0 ? 1 : drrfo->drr_firstobj;
obj < drrfo->drr_firstobj + drrfo->drr_numobjs &&
obj < DN_MAX_OBJECT && next_err == 0;
next_err = dmu_object_next(rwa->os, &obj, FALSE, 0)) {
dmu_object_info_t doi;
int err;
err = dmu_object_info(rwa->os, obj, &doi);
if (err == ENOENT)
continue;
else if (err != 0)
return (err);
err = dmu_free_long_object(rwa->os, obj);
if (err != 0)
return (err);
}
if (next_err != ESRCH)
return (next_err);
return (0);
}
/*
* Note: if this fails, the caller will clean up any records left on the
* rwa->write_batch list.
*/
static int
flush_write_batch_impl(struct receive_writer_arg *rwa)
{
dnode_t *dn;
int err;
if (dnode_hold(rwa->os, rwa->last_object, FTAG, &dn) != 0)
return (SET_ERROR(EINVAL));
struct receive_record_arg *last_rrd = list_tail(&rwa->write_batch);
struct drr_write *last_drrw = &last_rrd->header.drr_u.drr_write;
struct receive_record_arg *first_rrd = list_head(&rwa->write_batch);
struct drr_write *first_drrw = &first_rrd->header.drr_u.drr_write;
ASSERT3U(rwa->last_object, ==, last_drrw->drr_object);
ASSERT3U(rwa->last_offset, ==, last_drrw->drr_offset);
dmu_tx_t *tx = dmu_tx_create(rwa->os);
dmu_tx_hold_write_by_dnode(tx, dn, first_drrw->drr_offset,
last_drrw->drr_offset - first_drrw->drr_offset +
last_drrw->drr_logical_size);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err != 0) {
dmu_tx_abort(tx);
dnode_rele(dn, FTAG);
return (err);
}
struct receive_record_arg *rrd;
while ((rrd = list_head(&rwa->write_batch)) != NULL) {
struct drr_write *drrw = &rrd->header.drr_u.drr_write;
abd_t *abd = rrd->abd;
ASSERT3U(drrw->drr_object, ==, rwa->last_object);
if (drrw->drr_logical_size != dn->dn_datablksz) {
/*
* The WRITE record is larger than the object's block
* size. We must be receiving an incremental
* large-block stream into a dataset that previously did
* a non-large-block receive. Lightweight writes must
* be exactly one block, so we need to decompress the
* data (if compressed) and do a normal dmu_write().
*/
ASSERT3U(drrw->drr_logical_size, >, dn->dn_datablksz);
if (DRR_WRITE_COMPRESSED(drrw)) {
abd_t *decomp_abd =
abd_alloc_linear(drrw->drr_logical_size,
B_FALSE);
err = zio_decompress_data(
drrw->drr_compressiontype,
abd, abd_to_buf(decomp_abd),
abd_get_size(abd),
abd_get_size(decomp_abd), NULL);
if (err == 0) {
dmu_write_by_dnode(dn,
drrw->drr_offset,
drrw->drr_logical_size,
abd_to_buf(decomp_abd), tx);
}
abd_free(decomp_abd);
} else {
dmu_write_by_dnode(dn,
drrw->drr_offset,
drrw->drr_logical_size,
abd_to_buf(abd), tx);
}
if (err == 0)
abd_free(abd);
} else {
zio_prop_t zp;
dmu_write_policy(rwa->os, dn, 0, 0, &zp);
enum zio_flag zio_flags = 0;
if (rwa->raw) {
zp.zp_encrypt = B_TRUE;
zp.zp_compress = drrw->drr_compressiontype;
zp.zp_byteorder = ZFS_HOST_BYTEORDER ^
!!DRR_IS_RAW_BYTESWAPPED(drrw->drr_flags) ^
rwa->byteswap;
bcopy(drrw->drr_salt, zp.zp_salt,
ZIO_DATA_SALT_LEN);
bcopy(drrw->drr_iv, zp.zp_iv,
ZIO_DATA_IV_LEN);
bcopy(drrw->drr_mac, zp.zp_mac,
ZIO_DATA_MAC_LEN);
if (DMU_OT_IS_ENCRYPTED(zp.zp_type)) {
zp.zp_nopwrite = B_FALSE;
zp.zp_copies = MIN(zp.zp_copies,
SPA_DVAS_PER_BP - 1);
}
zio_flags |= ZIO_FLAG_RAW;
} else if (DRR_WRITE_COMPRESSED(drrw)) {
ASSERT3U(drrw->drr_compressed_size, >, 0);
ASSERT3U(drrw->drr_logical_size, >=,
drrw->drr_compressed_size);
zp.zp_compress = drrw->drr_compressiontype;
zio_flags |= ZIO_FLAG_RAW_COMPRESS;
} else if (rwa->byteswap) {
/*
* Note: compressed blocks never need to be
* byteswapped, because WRITE records for
* metadata blocks are never compressed. The
* exception is raw streams, which are written
* in the original byteorder, and the byteorder
* bit is preserved in the BP by setting
* zp_byteorder above.
*/
dmu_object_byteswap_t byteswap =
DMU_OT_BYTESWAP(drrw->drr_type);
dmu_ot_byteswap[byteswap].ob_func(
abd_to_buf(abd),
DRR_WRITE_PAYLOAD_SIZE(drrw));
}
/*
* Since this data can't be read until the receive
* completes, we can do a "lightweight" write for
* improved performance.
*/
err = dmu_lightweight_write_by_dnode(dn,
drrw->drr_offset, abd, &zp, zio_flags, tx);
}
if (err != 0) {
/*
* This rrd is left on the list, so the caller will
* free it (and the abd).
*/
break;
}
/*
* Note: If the receive fails, we want the resume stream to
* start with the same record that we last successfully
* received (as opposed to the next record), so that we can
* verify that we are resuming from the correct location.
*/
save_resume_state(rwa, drrw->drr_object, drrw->drr_offset, tx);
list_remove(&rwa->write_batch, rrd);
kmem_free(rrd, sizeof (*rrd));
}
dmu_tx_commit(tx);
dnode_rele(dn, FTAG);
return (err);
}
noinline static int
flush_write_batch(struct receive_writer_arg *rwa)
{
if (list_is_empty(&rwa->write_batch))
return (0);
int err = rwa->err;
if (err == 0)
err = flush_write_batch_impl(rwa);
if (err != 0) {
struct receive_record_arg *rrd;
while ((rrd = list_remove_head(&rwa->write_batch)) != NULL) {
abd_free(rrd->abd);
kmem_free(rrd, sizeof (*rrd));
}
}
ASSERT(list_is_empty(&rwa->write_batch));
return (err);
}
noinline static int
receive_process_write_record(struct receive_writer_arg *rwa,
struct receive_record_arg *rrd)
{
int err = 0;
ASSERT3U(rrd->header.drr_type, ==, DRR_WRITE);
struct drr_write *drrw = &rrd->header.drr_u.drr_write;
if (drrw->drr_offset + drrw->drr_logical_size < drrw->drr_offset ||
!DMU_OT_IS_VALID(drrw->drr_type))
return (SET_ERROR(EINVAL));
/*
* For resuming to work, records must be in increasing order
* by (object, offset).
*/
if (drrw->drr_object < rwa->last_object ||
(drrw->drr_object == rwa->last_object &&
drrw->drr_offset < rwa->last_offset)) {
return (SET_ERROR(EINVAL));
}
struct receive_record_arg *first_rrd = list_head(&rwa->write_batch);
struct drr_write *first_drrw = &first_rrd->header.drr_u.drr_write;
uint64_t batch_size =
MIN(zfs_recv_write_batch_size, DMU_MAX_ACCESS / 2);
if (first_rrd != NULL &&
(drrw->drr_object != first_drrw->drr_object ||
drrw->drr_offset >= first_drrw->drr_offset + batch_size)) {
err = flush_write_batch(rwa);
if (err != 0)
return (err);
}
rwa->last_object = drrw->drr_object;
rwa->last_offset = drrw->drr_offset;
if (rwa->last_object > rwa->max_object)
rwa->max_object = rwa->last_object;
list_insert_tail(&rwa->write_batch, rrd);
/*
* Return EAGAIN to indicate that we will use this rrd again,
* so the caller should not free it
*/
return (EAGAIN);
}
static int
receive_write_embedded(struct receive_writer_arg *rwa,
struct drr_write_embedded *drrwe, void *data)
{
dmu_tx_t *tx;
int err;
if (drrwe->drr_offset + drrwe->drr_length < drrwe->drr_offset)
return (SET_ERROR(EINVAL));
if (drrwe->drr_psize > BPE_PAYLOAD_SIZE)
return (SET_ERROR(EINVAL));
if (drrwe->drr_etype >= NUM_BP_EMBEDDED_TYPES)
return (SET_ERROR(EINVAL));
if (drrwe->drr_compression >= ZIO_COMPRESS_FUNCTIONS)
return (SET_ERROR(EINVAL));
if (rwa->raw)
return (SET_ERROR(EINVAL));
if (drrwe->drr_object > rwa->max_object)
rwa->max_object = drrwe->drr_object;
tx = dmu_tx_create(rwa->os);
dmu_tx_hold_write(tx, drrwe->drr_object,
drrwe->drr_offset, drrwe->drr_length);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err != 0) {
dmu_tx_abort(tx);
return (err);
}
dmu_write_embedded(rwa->os, drrwe->drr_object,
drrwe->drr_offset, data, drrwe->drr_etype,
drrwe->drr_compression, drrwe->drr_lsize, drrwe->drr_psize,
rwa->byteswap ^ ZFS_HOST_BYTEORDER, tx);
/* See comment in restore_write. */
save_resume_state(rwa, drrwe->drr_object, drrwe->drr_offset, tx);
dmu_tx_commit(tx);
return (0);
}
static int
receive_spill(struct receive_writer_arg *rwa, struct drr_spill *drrs,
abd_t *abd)
{
dmu_buf_t *db, *db_spill;
int err;
if (drrs->drr_length < SPA_MINBLOCKSIZE ||
drrs->drr_length > spa_maxblocksize(dmu_objset_spa(rwa->os)))
return (SET_ERROR(EINVAL));
/*
* This is an unmodified spill block which was added to the stream
* to resolve an issue with incorrectly removing spill blocks. It
* should be ignored by current versions of the code which support
* the DRR_FLAG_SPILL_BLOCK flag.
*/
if (rwa->spill && DRR_SPILL_IS_UNMODIFIED(drrs->drr_flags)) {
abd_free(abd);
return (0);
}
if (rwa->raw) {
if (!DMU_OT_IS_VALID(drrs->drr_type) ||
drrs->drr_compressiontype >= ZIO_COMPRESS_FUNCTIONS ||
drrs->drr_compressed_size == 0)
return (SET_ERROR(EINVAL));
}
if (dmu_object_info(rwa->os, drrs->drr_object, NULL) != 0)
return (SET_ERROR(EINVAL));
if (drrs->drr_object > rwa->max_object)
rwa->max_object = drrs->drr_object;
VERIFY0(dmu_bonus_hold(rwa->os, drrs->drr_object, FTAG, &db));
if ((err = dmu_spill_hold_by_bonus(db, DMU_READ_NO_DECRYPT, FTAG,
&db_spill)) != 0) {
dmu_buf_rele(db, FTAG);
return (err);
}
dmu_tx_t *tx = dmu_tx_create(rwa->os);
dmu_tx_hold_spill(tx, db->db_object);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err != 0) {
dmu_buf_rele(db, FTAG);
dmu_buf_rele(db_spill, FTAG);
dmu_tx_abort(tx);
return (err);
}
/*
* Spill blocks may both grow and shrink. When a change in size
* occurs any existing dbuf must be updated to match the logical
* size of the provided arc_buf_t.
*/
if (db_spill->db_size != drrs->drr_length) {
dmu_buf_will_fill(db_spill, tx);
VERIFY0(dbuf_spill_set_blksz(db_spill,
drrs->drr_length, tx));
}
arc_buf_t *abuf;
if (rwa->raw) {
boolean_t byteorder = ZFS_HOST_BYTEORDER ^
!!DRR_IS_RAW_BYTESWAPPED(drrs->drr_flags) ^
rwa->byteswap;
abuf = arc_loan_raw_buf(dmu_objset_spa(rwa->os),
drrs->drr_object, byteorder, drrs->drr_salt,
drrs->drr_iv, drrs->drr_mac, drrs->drr_type,
drrs->drr_compressed_size, drrs->drr_length,
drrs->drr_compressiontype, 0);
} else {
abuf = arc_loan_buf(dmu_objset_spa(rwa->os),
DMU_OT_IS_METADATA(drrs->drr_type),
drrs->drr_length);
if (rwa->byteswap) {
dmu_object_byteswap_t byteswap =
DMU_OT_BYTESWAP(drrs->drr_type);
dmu_ot_byteswap[byteswap].ob_func(abd_to_buf(abd),
DRR_SPILL_PAYLOAD_SIZE(drrs));
}
}
bcopy(abd_to_buf(abd), abuf->b_data, DRR_SPILL_PAYLOAD_SIZE(drrs));
abd_free(abd);
dbuf_assign_arcbuf((dmu_buf_impl_t *)db_spill, abuf, tx);
dmu_buf_rele(db, FTAG);
dmu_buf_rele(db_spill, FTAG);
dmu_tx_commit(tx);
return (0);
}
/* ARGSUSED */
noinline static int
receive_free(struct receive_writer_arg *rwa, struct drr_free *drrf)
{
int err;
if (drrf->drr_length != -1ULL &&
drrf->drr_offset + drrf->drr_length < drrf->drr_offset)
return (SET_ERROR(EINVAL));
if (dmu_object_info(rwa->os, drrf->drr_object, NULL) != 0)
return (SET_ERROR(EINVAL));
if (drrf->drr_object > rwa->max_object)
rwa->max_object = drrf->drr_object;
err = dmu_free_long_range(rwa->os, drrf->drr_object,
drrf->drr_offset, drrf->drr_length);
return (err);
}
static int
receive_object_range(struct receive_writer_arg *rwa,
struct drr_object_range *drror)
{
/*
* By default, we assume this block is in our native format
* (ZFS_HOST_BYTEORDER). We then take into account whether
* the send stream is byteswapped (rwa->byteswap). Finally,
* we need to byteswap again if this particular block was
* in non-native format on the send side.
*/
boolean_t byteorder = ZFS_HOST_BYTEORDER ^ rwa->byteswap ^
!!DRR_IS_RAW_BYTESWAPPED(drror->drr_flags);
/*
* Since dnode block sizes are constant, we should not need to worry
* about making sure that the dnode block size is the same on the
* sending and receiving sides for the time being. For non-raw sends,
* this does not matter (and in fact we do not send a DRR_OBJECT_RANGE
* record at all). Raw sends require this record type because the
* encryption parameters are used to protect an entire block of bonus
* buffers. If the size of dnode blocks ever becomes variable,
* handling will need to be added to ensure that dnode block sizes
* match on the sending and receiving side.
*/
if (drror->drr_numslots != DNODES_PER_BLOCK ||
P2PHASE(drror->drr_firstobj, DNODES_PER_BLOCK) != 0 ||
!rwa->raw)
return (SET_ERROR(EINVAL));
if (drror->drr_firstobj > rwa->max_object)
rwa->max_object = drror->drr_firstobj;
/*
* The DRR_OBJECT_RANGE handling must be deferred to receive_object()
* so that the block of dnodes is not written out when it's empty,
* and converted to a HOLE BP.
*/
rwa->or_crypt_params_present = B_TRUE;
rwa->or_firstobj = drror->drr_firstobj;
rwa->or_numslots = drror->drr_numslots;
bcopy(drror->drr_salt, rwa->or_salt, ZIO_DATA_SALT_LEN);
bcopy(drror->drr_iv, rwa->or_iv, ZIO_DATA_IV_LEN);
bcopy(drror->drr_mac, rwa->or_mac, ZIO_DATA_MAC_LEN);
rwa->or_byteorder = byteorder;
return (0);
}
/*
* Until we have the ability to redact large ranges of data efficiently, we
* process these records as frees.
*/
/* ARGSUSED */
noinline static int
receive_redact(struct receive_writer_arg *rwa, struct drr_redact *drrr)
{
struct drr_free drrf = {0};
drrf.drr_length = drrr->drr_length;
drrf.drr_object = drrr->drr_object;
drrf.drr_offset = drrr->drr_offset;
drrf.drr_toguid = drrr->drr_toguid;
return (receive_free(rwa, &drrf));
}
/* used to destroy the drc_ds on error */
static void
dmu_recv_cleanup_ds(dmu_recv_cookie_t *drc)
{
dsl_dataset_t *ds = drc->drc_ds;
ds_hold_flags_t dsflags;
dsflags = (drc->drc_raw) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
/*
* Wait for the txg sync before cleaning up the receive. For
* resumable receives, this ensures that our resume state has
* been written out to disk. For raw receives, this ensures
* that the user accounting code will not attempt to do anything
* after we stopped receiving the dataset.
*/
txg_wait_synced(ds->ds_dir->dd_pool, 0);
ds->ds_objset->os_raw_receive = B_FALSE;
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
if (drc->drc_resumable && drc->drc_should_save &&
!BP_IS_HOLE(dsl_dataset_get_blkptr(ds))) {
rrw_exit(&ds->ds_bp_rwlock, FTAG);
dsl_dataset_disown(ds, dsflags, dmu_recv_tag);
} else {
char name[ZFS_MAX_DATASET_NAME_LEN];
rrw_exit(&ds->ds_bp_rwlock, FTAG);
dsl_dataset_name(ds, name);
dsl_dataset_disown(ds, dsflags, dmu_recv_tag);
(void) dsl_destroy_head(name);
}
}
static void
receive_cksum(dmu_recv_cookie_t *drc, int len, void *buf)
{
if (drc->drc_byteswap) {
(void) fletcher_4_incremental_byteswap(buf, len,
&drc->drc_cksum);
} else {
(void) fletcher_4_incremental_native(buf, len, &drc->drc_cksum);
}
}
/*
* Read the payload into a buffer of size len, and update the current record's
* payload field.
* Allocate drc->drc_next_rrd and read the next record's header into
* drc->drc_next_rrd->header.
* Verify checksum of payload and next record.
*/
static int
receive_read_payload_and_next_header(dmu_recv_cookie_t *drc, int len, void *buf)
{
int err;
if (len != 0) {
ASSERT3U(len, <=, SPA_MAXBLOCKSIZE);
err = receive_read(drc, len, buf);
if (err != 0)
return (err);
receive_cksum(drc, len, buf);
/* note: rrd is NULL when reading the begin record's payload */
if (drc->drc_rrd != NULL) {
drc->drc_rrd->payload = buf;
drc->drc_rrd->payload_size = len;
drc->drc_rrd->bytes_read = drc->drc_bytes_read;
}
} else {
ASSERT3P(buf, ==, NULL);
}
drc->drc_prev_cksum = drc->drc_cksum;
drc->drc_next_rrd = kmem_zalloc(sizeof (*drc->drc_next_rrd), KM_SLEEP);
err = receive_read(drc, sizeof (drc->drc_next_rrd->header),
&drc->drc_next_rrd->header);
drc->drc_next_rrd->bytes_read = drc->drc_bytes_read;
if (err != 0) {
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
drc->drc_next_rrd = NULL;
return (err);
}
if (drc->drc_next_rrd->header.drr_type == DRR_BEGIN) {
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
drc->drc_next_rrd = NULL;
return (SET_ERROR(EINVAL));
}
/*
* Note: checksum is of everything up to but not including the
* checksum itself.
*/
ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
receive_cksum(drc,
offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
&drc->drc_next_rrd->header);
zio_cksum_t cksum_orig =
drc->drc_next_rrd->header.drr_u.drr_checksum.drr_checksum;
zio_cksum_t *cksump =
&drc->drc_next_rrd->header.drr_u.drr_checksum.drr_checksum;
if (drc->drc_byteswap)
byteswap_record(&drc->drc_next_rrd->header);
if ((!ZIO_CHECKSUM_IS_ZERO(cksump)) &&
!ZIO_CHECKSUM_EQUAL(drc->drc_cksum, *cksump)) {
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
drc->drc_next_rrd = NULL;
return (SET_ERROR(ECKSUM));
}
receive_cksum(drc, sizeof (cksum_orig), &cksum_orig);
return (0);
}
/*
* Issue the prefetch reads for any necessary indirect blocks.
*
* We use the object ignore list to tell us whether or not to issue prefetches
* for a given object. We do this for both correctness (in case the blocksize
* of an object has changed) and performance (if the object doesn't exist, don't
* needlessly try to issue prefetches). We also trim the list as we go through
* the stream to prevent it from growing to an unbounded size.
*
* The object numbers within will always be in sorted order, and any write
* records we see will also be in sorted order, but they're not sorted with
* respect to each other (i.e. we can get several object records before
* receiving each object's write records). As a result, once we've reached a
* given object number, we can safely remove any reference to lower object
* numbers in the ignore list. In practice, we receive up to 32 object records
* before receiving write records, so the list can have up to 32 nodes in it.
*/
/* ARGSUSED */
static void
receive_read_prefetch(dmu_recv_cookie_t *drc, uint64_t object, uint64_t offset,
uint64_t length)
{
if (!objlist_exists(drc->drc_ignore_objlist, object)) {
dmu_prefetch(drc->drc_os, object, 1, offset, length,
ZIO_PRIORITY_SYNC_READ);
}
}
/*
* Read records off the stream, issuing any necessary prefetches.
*/
static int
receive_read_record(dmu_recv_cookie_t *drc)
{
int err;
switch (drc->drc_rrd->header.drr_type) {
case DRR_OBJECT:
{
struct drr_object *drro =
&drc->drc_rrd->header.drr_u.drr_object;
uint32_t size = DRR_OBJECT_PAYLOAD_SIZE(drro);
void *buf = NULL;
dmu_object_info_t doi;
if (size != 0)
buf = kmem_zalloc(size, KM_SLEEP);
err = receive_read_payload_and_next_header(drc, size, buf);
if (err != 0) {
kmem_free(buf, size);
return (err);
}
err = dmu_object_info(drc->drc_os, drro->drr_object, &doi);
/*
* See receive_read_prefetch for an explanation why we're
* storing this object in the ignore_obj_list.
*/
if (err == ENOENT || err == EEXIST ||
(err == 0 && doi.doi_data_block_size != drro->drr_blksz)) {
objlist_insert(drc->drc_ignore_objlist,
drro->drr_object);
err = 0;
}
return (err);
}
case DRR_FREEOBJECTS:
{
err = receive_read_payload_and_next_header(drc, 0, NULL);
return (err);
}
case DRR_WRITE:
{
struct drr_write *drrw = &drc->drc_rrd->header.drr_u.drr_write;
int size = DRR_WRITE_PAYLOAD_SIZE(drrw);
abd_t *abd = abd_alloc_linear(size, B_FALSE);
err = receive_read_payload_and_next_header(drc, size,
abd_to_buf(abd));
if (err != 0) {
abd_free(abd);
return (err);
}
drc->drc_rrd->abd = abd;
receive_read_prefetch(drc, drrw->drr_object, drrw->drr_offset,
drrw->drr_logical_size);
return (err);
}
case DRR_WRITE_EMBEDDED:
{
struct drr_write_embedded *drrwe =
&drc->drc_rrd->header.drr_u.drr_write_embedded;
uint32_t size = P2ROUNDUP(drrwe->drr_psize, 8);
void *buf = kmem_zalloc(size, KM_SLEEP);
err = receive_read_payload_and_next_header(drc, size, buf);
if (err != 0) {
kmem_free(buf, size);
return (err);
}
receive_read_prefetch(drc, drrwe->drr_object, drrwe->drr_offset,
drrwe->drr_length);
return (err);
}
case DRR_FREE:
case DRR_REDACT:
{
/*
* It might be beneficial to prefetch indirect blocks here, but
* we don't really have the data to decide for sure.
*/
err = receive_read_payload_and_next_header(drc, 0, NULL);
return (err);
}
case DRR_END:
{
struct drr_end *drre = &drc->drc_rrd->header.drr_u.drr_end;
if (!ZIO_CHECKSUM_EQUAL(drc->drc_prev_cksum,
drre->drr_checksum))
return (SET_ERROR(ECKSUM));
return (0);
}
case DRR_SPILL:
{
struct drr_spill *drrs = &drc->drc_rrd->header.drr_u.drr_spill;
int size = DRR_SPILL_PAYLOAD_SIZE(drrs);
abd_t *abd = abd_alloc_linear(size, B_FALSE);
err = receive_read_payload_and_next_header(drc, size,
abd_to_buf(abd));
if (err != 0)
abd_free(abd);
else
drc->drc_rrd->abd = abd;
return (err);
}
case DRR_OBJECT_RANGE:
{
err = receive_read_payload_and_next_header(drc, 0, NULL);
return (err);
}
default:
return (SET_ERROR(EINVAL));
}
}
static void
dprintf_drr(struct receive_record_arg *rrd, int err)
{
#ifdef ZFS_DEBUG
switch (rrd->header.drr_type) {
case DRR_OBJECT:
{
struct drr_object *drro = &rrd->header.drr_u.drr_object;
dprintf("drr_type = OBJECT obj = %llu type = %u "
"bonustype = %u blksz = %u bonuslen = %u cksumtype = %u "
"compress = %u dn_slots = %u err = %d\n",
(u_longlong_t)drro->drr_object, drro->drr_type,
drro->drr_bonustype, drro->drr_blksz, drro->drr_bonuslen,
drro->drr_checksumtype, drro->drr_compress,
drro->drr_dn_slots, err);
break;
}
case DRR_FREEOBJECTS:
{
struct drr_freeobjects *drrfo =
&rrd->header.drr_u.drr_freeobjects;
dprintf("drr_type = FREEOBJECTS firstobj = %llu "
"numobjs = %llu err = %d\n",
(u_longlong_t)drrfo->drr_firstobj,
(u_longlong_t)drrfo->drr_numobjs, err);
break;
}
case DRR_WRITE:
{
struct drr_write *drrw = &rrd->header.drr_u.drr_write;
dprintf("drr_type = WRITE obj = %llu type = %u offset = %llu "
"lsize = %llu cksumtype = %u flags = %u "
"compress = %u psize = %llu err = %d\n",
(u_longlong_t)drrw->drr_object, drrw->drr_type,
(u_longlong_t)drrw->drr_offset,
(u_longlong_t)drrw->drr_logical_size,
drrw->drr_checksumtype, drrw->drr_flags,
drrw->drr_compressiontype,
(u_longlong_t)drrw->drr_compressed_size, err);
break;
}
case DRR_WRITE_BYREF:
{
struct drr_write_byref *drrwbr =
&rrd->header.drr_u.drr_write_byref;
dprintf("drr_type = WRITE_BYREF obj = %llu offset = %llu "
"length = %llu toguid = %llx refguid = %llx "
"refobject = %llu refoffset = %llu cksumtype = %u "
"flags = %u err = %d\n",
(u_longlong_t)drrwbr->drr_object,
(u_longlong_t)drrwbr->drr_offset,
(u_longlong_t)drrwbr->drr_length,
(u_longlong_t)drrwbr->drr_toguid,
(u_longlong_t)drrwbr->drr_refguid,
(u_longlong_t)drrwbr->drr_refobject,
(u_longlong_t)drrwbr->drr_refoffset,
drrwbr->drr_checksumtype, drrwbr->drr_flags, err);
break;
}
case DRR_WRITE_EMBEDDED:
{
struct drr_write_embedded *drrwe =
&rrd->header.drr_u.drr_write_embedded;
dprintf("drr_type = WRITE_EMBEDDED obj = %llu offset = %llu "
"length = %llu compress = %u etype = %u lsize = %u "
"psize = %u err = %d\n",
(u_longlong_t)drrwe->drr_object,
(u_longlong_t)drrwe->drr_offset,
(u_longlong_t)drrwe->drr_length,
drrwe->drr_compression, drrwe->drr_etype,
drrwe->drr_lsize, drrwe->drr_psize, err);
break;
}
case DRR_FREE:
{
struct drr_free *drrf = &rrd->header.drr_u.drr_free;
dprintf("drr_type = FREE obj = %llu offset = %llu "
"length = %lld err = %d\n",
(u_longlong_t)drrf->drr_object,
(u_longlong_t)drrf->drr_offset,
(longlong_t)drrf->drr_length,
err);
break;
}
case DRR_SPILL:
{
struct drr_spill *drrs = &rrd->header.drr_u.drr_spill;
dprintf("drr_type = SPILL obj = %llu length = %llu "
"err = %d\n", (u_longlong_t)drrs->drr_object,
(u_longlong_t)drrs->drr_length, err);
break;
}
case DRR_OBJECT_RANGE:
{
struct drr_object_range *drror =
&rrd->header.drr_u.drr_object_range;
dprintf("drr_type = OBJECT_RANGE firstobj = %llu "
"numslots = %llu flags = %u err = %d\n",
(u_longlong_t)drror->drr_firstobj,
(u_longlong_t)drror->drr_numslots,
drror->drr_flags, err);
break;
}
default:
return;
}
#endif
}
/*
* Commit the records to the pool.
*/
static int
receive_process_record(struct receive_writer_arg *rwa,
struct receive_record_arg *rrd)
{
int err;
/* Processing in order, therefore bytes_read should be increasing. */
ASSERT3U(rrd->bytes_read, >=, rwa->bytes_read);
rwa->bytes_read = rrd->bytes_read;
if (rrd->header.drr_type != DRR_WRITE) {
err = flush_write_batch(rwa);
if (err != 0) {
if (rrd->abd != NULL) {
abd_free(rrd->abd);
rrd->abd = NULL;
rrd->payload = NULL;
} else if (rrd->payload != NULL) {
kmem_free(rrd->payload, rrd->payload_size);
rrd->payload = NULL;
}
return (err);
}
}
switch (rrd->header.drr_type) {
case DRR_OBJECT:
{
struct drr_object *drro = &rrd->header.drr_u.drr_object;
err = receive_object(rwa, drro, rrd->payload);
kmem_free(rrd->payload, rrd->payload_size);
rrd->payload = NULL;
break;
}
case DRR_FREEOBJECTS:
{
struct drr_freeobjects *drrfo =
&rrd->header.drr_u.drr_freeobjects;
err = receive_freeobjects(rwa, drrfo);
break;
}
case DRR_WRITE:
{
err = receive_process_write_record(rwa, rrd);
if (err != EAGAIN) {
/*
* On success, receive_process_write_record() returns
* EAGAIN to indicate that we do not want to free
* the rrd or arc_buf.
*/
ASSERT(err != 0);
abd_free(rrd->abd);
rrd->abd = NULL;
}
break;
}
case DRR_WRITE_EMBEDDED:
{
struct drr_write_embedded *drrwe =
&rrd->header.drr_u.drr_write_embedded;
err = receive_write_embedded(rwa, drrwe, rrd->payload);
kmem_free(rrd->payload, rrd->payload_size);
rrd->payload = NULL;
break;
}
case DRR_FREE:
{
struct drr_free *drrf = &rrd->header.drr_u.drr_free;
err = receive_free(rwa, drrf);
break;
}
case DRR_SPILL:
{
struct drr_spill *drrs = &rrd->header.drr_u.drr_spill;
err = receive_spill(rwa, drrs, rrd->abd);
if (err != 0)
abd_free(rrd->abd);
rrd->abd = NULL;
rrd->payload = NULL;
break;
}
case DRR_OBJECT_RANGE:
{
struct drr_object_range *drror =
&rrd->header.drr_u.drr_object_range;
err = receive_object_range(rwa, drror);
break;
}
case DRR_REDACT:
{
struct drr_redact *drrr = &rrd->header.drr_u.drr_redact;
err = receive_redact(rwa, drrr);
break;
}
default:
err = (SET_ERROR(EINVAL));
}
if (err != 0)
dprintf_drr(rrd, err);
return (err);
}
/*
* dmu_recv_stream's worker thread; pull records off the queue, and then call
* receive_process_record When we're done, signal the main thread and exit.
*/
static void
receive_writer_thread(void *arg)
{
struct receive_writer_arg *rwa = arg;
struct receive_record_arg *rrd;
fstrans_cookie_t cookie = spl_fstrans_mark();
for (rrd = bqueue_dequeue(&rwa->q); !rrd->eos_marker;
rrd = bqueue_dequeue(&rwa->q)) {
/*
* If there's an error, the main thread will stop putting things
* on the queue, but we need to clear everything in it before we
* can exit.
*/
int err = 0;
if (rwa->err == 0) {
err = receive_process_record(rwa, rrd);
} else if (rrd->abd != NULL) {
abd_free(rrd->abd);
rrd->abd = NULL;
rrd->payload = NULL;
} else if (rrd->payload != NULL) {
kmem_free(rrd->payload, rrd->payload_size);
rrd->payload = NULL;
}
/*
* EAGAIN indicates that this record has been saved (on
* raw->write_batch), and will be used again, so we don't
* free it.
*/
if (err != EAGAIN) {
if (rwa->err == 0)
rwa->err = err;
kmem_free(rrd, sizeof (*rrd));
}
}
kmem_free(rrd, sizeof (*rrd));
int err = flush_write_batch(rwa);
if (rwa->err == 0)
rwa->err = err;
mutex_enter(&rwa->mutex);
rwa->done = B_TRUE;
cv_signal(&rwa->cv);
mutex_exit(&rwa->mutex);
spl_fstrans_unmark(cookie);
thread_exit();
}
static int
resume_check(dmu_recv_cookie_t *drc, nvlist_t *begin_nvl)
{
uint64_t val;
objset_t *mos = dmu_objset_pool(drc->drc_os)->dp_meta_objset;
uint64_t dsobj = dmu_objset_id(drc->drc_os);
uint64_t resume_obj, resume_off;
if (nvlist_lookup_uint64(begin_nvl,
"resume_object", &resume_obj) != 0 ||
nvlist_lookup_uint64(begin_nvl,
"resume_offset", &resume_off) != 0) {
return (SET_ERROR(EINVAL));
}
VERIFY0(zap_lookup(mos, dsobj,
DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val));
if (resume_obj != val)
return (SET_ERROR(EINVAL));
VERIFY0(zap_lookup(mos, dsobj,
DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val));
if (resume_off != val)
return (SET_ERROR(EINVAL));
return (0);
}
/*
* Read in the stream's records, one by one, and apply them to the pool. There
* are two threads involved; the thread that calls this function will spin up a
* worker thread, read the records off the stream one by one, and issue
* prefetches for any necessary indirect blocks. It will then push the records
* onto an internal blocking queue. The worker thread will pull the records off
* the queue, and actually write the data into the DMU. This way, the worker
* thread doesn't have to wait for reads to complete, since everything it needs
* (the indirect blocks) will be prefetched.
*
* NB: callers *must* call dmu_recv_end() if this succeeds.
*/
int
dmu_recv_stream(dmu_recv_cookie_t *drc, offset_t *voffp)
{
int err = 0;
struct receive_writer_arg *rwa = kmem_zalloc(sizeof (*rwa), KM_SLEEP);
if (dsl_dataset_has_resume_receive_state(drc->drc_ds)) {
uint64_t bytes = 0;
(void) zap_lookup(drc->drc_ds->ds_dir->dd_pool->dp_meta_objset,
drc->drc_ds->ds_object, DS_FIELD_RESUME_BYTES,
sizeof (bytes), 1, &bytes);
drc->drc_bytes_read += bytes;
}
drc->drc_ignore_objlist = objlist_create();
/* these were verified in dmu_recv_begin */
ASSERT3U(DMU_GET_STREAM_HDRTYPE(drc->drc_drrb->drr_versioninfo), ==,
DMU_SUBSTREAM);
ASSERT3U(drc->drc_drrb->drr_type, <, DMU_OST_NUMTYPES);
ASSERT(dsl_dataset_phys(drc->drc_ds)->ds_flags & DS_FLAG_INCONSISTENT);
ASSERT0(drc->drc_os->os_encrypted &&
(drc->drc_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA));
/* handle DSL encryption key payload */
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) {
nvlist_t *keynvl = NULL;
ASSERT(drc->drc_os->os_encrypted);
ASSERT(drc->drc_raw);
err = nvlist_lookup_nvlist(drc->drc_begin_nvl, "crypt_keydata",
&keynvl);
if (err != 0)
goto out;
/*
* If this is a new dataset we set the key immediately.
* Otherwise we don't want to change the key until we
* are sure the rest of the receive succeeded so we stash
* the keynvl away until then.
*/
err = dsl_crypto_recv_raw(spa_name(drc->drc_os->os_spa),
drc->drc_ds->ds_object, drc->drc_fromsnapobj,
drc->drc_drrb->drr_type, keynvl, drc->drc_newfs);
if (err != 0)
goto out;
/* see comment in dmu_recv_end_sync() */
drc->drc_ivset_guid = 0;
(void) nvlist_lookup_uint64(keynvl, "to_ivset_guid",
&drc->drc_ivset_guid);
if (!drc->drc_newfs)
drc->drc_keynvl = fnvlist_dup(keynvl);
}
if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING) {
err = resume_check(drc, drc->drc_begin_nvl);
if (err != 0)
goto out;
}
/*
* If we failed before this point we will clean up any new resume
* state that was created. Now that we've gotten past the initial
* checks we are ok to retain that resume state.
*/
drc->drc_should_save = B_TRUE;
(void) bqueue_init(&rwa->q, zfs_recv_queue_ff,
MAX(zfs_recv_queue_length, 2 * zfs_max_recordsize),
offsetof(struct receive_record_arg, node));
cv_init(&rwa->cv, NULL, CV_DEFAULT, NULL);
mutex_init(&rwa->mutex, NULL, MUTEX_DEFAULT, NULL);
rwa->os = drc->drc_os;
rwa->byteswap = drc->drc_byteswap;
rwa->resumable = drc->drc_resumable;
rwa->raw = drc->drc_raw;
rwa->spill = drc->drc_spill;
rwa->full = (drc->drc_drr_begin->drr_u.drr_begin.drr_fromguid == 0);
rwa->os->os_raw_receive = drc->drc_raw;
list_create(&rwa->write_batch, sizeof (struct receive_record_arg),
offsetof(struct receive_record_arg, node.bqn_node));
(void) thread_create(NULL, 0, receive_writer_thread, rwa, 0, curproc,
TS_RUN, minclsyspri);
/*
* We're reading rwa->err without locks, which is safe since we are the
* only reader, and the worker thread is the only writer. It's ok if we
* miss a write for an iteration or two of the loop, since the writer
* thread will keep freeing records we send it until we send it an eos
* marker.
*
* We can leave this loop in 3 ways: First, if rwa->err is
* non-zero. In that case, the writer thread will free the rrd we just
* pushed. Second, if we're interrupted; in that case, either it's the
* first loop and drc->drc_rrd was never allocated, or it's later, and
* drc->drc_rrd has been handed off to the writer thread who will free
* it. Finally, if receive_read_record fails or we're at the end of the
* stream, then we free drc->drc_rrd and exit.
*/
while (rwa->err == 0) {
if (issig(JUSTLOOKING) && issig(FORREAL)) {
err = SET_ERROR(EINTR);
break;
}
ASSERT3P(drc->drc_rrd, ==, NULL);
drc->drc_rrd = drc->drc_next_rrd;
drc->drc_next_rrd = NULL;
/* Allocates and loads header into drc->drc_next_rrd */
err = receive_read_record(drc);
if (drc->drc_rrd->header.drr_type == DRR_END || err != 0) {
kmem_free(drc->drc_rrd, sizeof (*drc->drc_rrd));
drc->drc_rrd = NULL;
break;
}
bqueue_enqueue(&rwa->q, drc->drc_rrd,
sizeof (struct receive_record_arg) +
drc->drc_rrd->payload_size);
drc->drc_rrd = NULL;
}
ASSERT3P(drc->drc_rrd, ==, NULL);
drc->drc_rrd = kmem_zalloc(sizeof (*drc->drc_rrd), KM_SLEEP);
drc->drc_rrd->eos_marker = B_TRUE;
bqueue_enqueue_flush(&rwa->q, drc->drc_rrd, 1);
mutex_enter(&rwa->mutex);
while (!rwa->done) {
/*
* We need to use cv_wait_sig() so that any process that may
* be sleeping here can still fork.
*/
(void) cv_wait_sig(&rwa->cv, &rwa->mutex);
}
mutex_exit(&rwa->mutex);
/*
* If we are receiving a full stream as a clone, all object IDs which
* are greater than the maximum ID referenced in the stream are
* by definition unused and must be freed.
*/
if (drc->drc_clone && drc->drc_drrb->drr_fromguid == 0) {
uint64_t obj = rwa->max_object + 1;
int free_err = 0;
int next_err = 0;
while (next_err == 0) {
free_err = dmu_free_long_object(rwa->os, obj);
if (free_err != 0 && free_err != ENOENT)
break;
next_err = dmu_object_next(rwa->os, &obj, FALSE, 0);
}
if (err == 0) {
if (free_err != 0 && free_err != ENOENT)
err = free_err;
else if (next_err != ESRCH)
err = next_err;
}
}
cv_destroy(&rwa->cv);
mutex_destroy(&rwa->mutex);
bqueue_destroy(&rwa->q);
list_destroy(&rwa->write_batch);
if (err == 0)
err = rwa->err;
out:
/*
* If we hit an error before we started the receive_writer_thread
* we need to clean up the next_rrd we create by processing the
* DRR_BEGIN record.
*/
if (drc->drc_next_rrd != NULL)
kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
/*
* The objset will be invalidated by dmu_recv_end() when we do
* dsl_dataset_clone_swap_sync_impl().
*/
drc->drc_os = NULL;
kmem_free(rwa, sizeof (*rwa));
nvlist_free(drc->drc_begin_nvl);
if (err != 0) {
/*
* Clean up references. If receive is not resumable,
* destroy what we created, so we don't leave it in
* the inconsistent state.
*/
dmu_recv_cleanup_ds(drc);
nvlist_free(drc->drc_keynvl);
}
objlist_destroy(drc->drc_ignore_objlist);
drc->drc_ignore_objlist = NULL;
*voffp = drc->drc_voff;
return (err);
}
static int
dmu_recv_end_check(void *arg, dmu_tx_t *tx)
{
dmu_recv_cookie_t *drc = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
int error;
ASSERT3P(drc->drc_ds->ds_owner, ==, dmu_recv_tag);
if (!drc->drc_newfs) {
dsl_dataset_t *origin_head;
error = dsl_dataset_hold(dp, drc->drc_tofs, FTAG, &origin_head);
if (error != 0)
return (error);
if (drc->drc_force) {
/*
* We will destroy any snapshots in tofs (i.e. before
* origin_head) that are after the origin (which is
* the snap before drc_ds, because drc_ds can not
* have any snaps of its own).
*/
uint64_t obj;
obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
while (obj !=
dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
dsl_dataset_t *snap;
error = dsl_dataset_hold_obj(dp, obj, FTAG,
&snap);
if (error != 0)
break;
if (snap->ds_dir != origin_head->ds_dir)
error = SET_ERROR(EINVAL);
if (error == 0) {
error = dsl_destroy_snapshot_check_impl(
snap, B_FALSE);
}
obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
dsl_dataset_rele(snap, FTAG);
if (error != 0)
break;
}
if (error != 0) {
dsl_dataset_rele(origin_head, FTAG);
return (error);
}
}
if (drc->drc_keynvl != NULL) {
error = dsl_crypto_recv_raw_key_check(drc->drc_ds,
drc->drc_keynvl, tx);
if (error != 0) {
dsl_dataset_rele(origin_head, FTAG);
return (error);
}
}
error = dsl_dataset_clone_swap_check_impl(drc->drc_ds,
origin_head, drc->drc_force, drc->drc_owner, tx);
if (error != 0) {
dsl_dataset_rele(origin_head, FTAG);
return (error);
}
error = dsl_dataset_snapshot_check_impl(origin_head,
drc->drc_tosnap, tx, B_TRUE, 1,
drc->drc_cred, drc->drc_proc);
dsl_dataset_rele(origin_head, FTAG);
if (error != 0)
return (error);
error = dsl_destroy_head_check_impl(drc->drc_ds, 1);
} else {
error = dsl_dataset_snapshot_check_impl(drc->drc_ds,
drc->drc_tosnap, tx, B_TRUE, 1,
drc->drc_cred, drc->drc_proc);
}
return (error);
}
static void
dmu_recv_end_sync(void *arg, dmu_tx_t *tx)
{
dmu_recv_cookie_t *drc = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
boolean_t encrypted = drc->drc_ds->ds_dir->dd_crypto_obj != 0;
uint64_t newsnapobj;
spa_history_log_internal_ds(drc->drc_ds, "finish receiving",
tx, "snap=%s", drc->drc_tosnap);
drc->drc_ds->ds_objset->os_raw_receive = B_FALSE;
if (!drc->drc_newfs) {
dsl_dataset_t *origin_head;
VERIFY0(dsl_dataset_hold(dp, drc->drc_tofs, FTAG,
&origin_head));
if (drc->drc_force) {
/*
* Destroy any snapshots of drc_tofs (origin_head)
* after the origin (the snap before drc_ds).
*/
uint64_t obj;
obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
while (obj !=
dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
dsl_dataset_t *snap;
VERIFY0(dsl_dataset_hold_obj(dp, obj, FTAG,
&snap));
ASSERT3P(snap->ds_dir, ==, origin_head->ds_dir);
obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
dsl_destroy_snapshot_sync_impl(snap,
B_FALSE, tx);
dsl_dataset_rele(snap, FTAG);
}
}
if (drc->drc_keynvl != NULL) {
dsl_crypto_recv_raw_key_sync(drc->drc_ds,
drc->drc_keynvl, tx);
nvlist_free(drc->drc_keynvl);
drc->drc_keynvl = NULL;
}
VERIFY3P(drc->drc_ds->ds_prev, ==,
origin_head->ds_prev);
dsl_dataset_clone_swap_sync_impl(drc->drc_ds,
origin_head, tx);
/*
* The objset was evicted by dsl_dataset_clone_swap_sync_impl,
* so drc_os is no longer valid.
*/
drc->drc_os = NULL;
dsl_dataset_snapshot_sync_impl(origin_head,
drc->drc_tosnap, tx);
/* set snapshot's creation time and guid */
dmu_buf_will_dirty(origin_head->ds_prev->ds_dbuf, tx);
dsl_dataset_phys(origin_head->ds_prev)->ds_creation_time =
drc->drc_drrb->drr_creation_time;
dsl_dataset_phys(origin_head->ds_prev)->ds_guid =
drc->drc_drrb->drr_toguid;
dsl_dataset_phys(origin_head->ds_prev)->ds_flags &=
~DS_FLAG_INCONSISTENT;
dmu_buf_will_dirty(origin_head->ds_dbuf, tx);
dsl_dataset_phys(origin_head)->ds_flags &=
~DS_FLAG_INCONSISTENT;
newsnapobj =
dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
dsl_dataset_rele(origin_head, FTAG);
dsl_destroy_head_sync_impl(drc->drc_ds, tx);
if (drc->drc_owner != NULL)
VERIFY3P(origin_head->ds_owner, ==, drc->drc_owner);
} else {
dsl_dataset_t *ds = drc->drc_ds;
dsl_dataset_snapshot_sync_impl(ds, drc->drc_tosnap, tx);
/* set snapshot's creation time and guid */
dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
dsl_dataset_phys(ds->ds_prev)->ds_creation_time =
drc->drc_drrb->drr_creation_time;
dsl_dataset_phys(ds->ds_prev)->ds_guid =
drc->drc_drrb->drr_toguid;
dsl_dataset_phys(ds->ds_prev)->ds_flags &=
~DS_FLAG_INCONSISTENT;
dmu_buf_will_dirty(ds->ds_dbuf, tx);
dsl_dataset_phys(ds)->ds_flags &= ~DS_FLAG_INCONSISTENT;
if (dsl_dataset_has_resume_receive_state(ds)) {
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_FROMGUID, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_OBJECT, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_OFFSET, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_BYTES, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_TOGUID, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_TONAME, tx);
(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS, tx);
}
newsnapobj =
dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj;
}
/*
* If this is a raw receive, the crypt_keydata nvlist will include
* a to_ivset_guid for us to set on the new snapshot. This value
* will override the value generated by the snapshot code. However,
* this value may not be present, because older implementations of
* the raw send code did not include this value, and we are still
* allowed to receive them if the zfs_disable_ivset_guid_check
* tunable is set, in which case we will leave the newly-generated
* value.
*/
if (drc->drc_raw && drc->drc_ivset_guid != 0) {
dmu_object_zapify(dp->dp_meta_objset, newsnapobj,
DMU_OT_DSL_DATASET, tx);
VERIFY0(zap_update(dp->dp_meta_objset, newsnapobj,
DS_FIELD_IVSET_GUID, sizeof (uint64_t), 1,
&drc->drc_ivset_guid, tx));
}
/*
* Release the hold from dmu_recv_begin. This must be done before
* we return to open context, so that when we free the dataset's dnode
* we can evict its bonus buffer. Since the dataset may be destroyed
* at this point (and therefore won't have a valid pointer to the spa)
* we release the key mapping manually here while we do have a valid
* pointer, if it exists.
*/
if (!drc->drc_raw && encrypted) {
(void) spa_keystore_remove_mapping(dmu_tx_pool(tx)->dp_spa,
drc->drc_ds->ds_object, drc->drc_ds);
}
dsl_dataset_disown(drc->drc_ds, 0, dmu_recv_tag);
drc->drc_ds = NULL;
}
static int dmu_recv_end_modified_blocks = 3;
static int
dmu_recv_existing_end(dmu_recv_cookie_t *drc)
{
#ifdef _KERNEL
/*
* We will be destroying the ds; make sure its origin is unmounted if
* necessary.
*/
char name[ZFS_MAX_DATASET_NAME_LEN];
dsl_dataset_name(drc->drc_ds, name);
zfs_destroy_unmount_origin(name);
#endif
return (dsl_sync_task(drc->drc_tofs,
dmu_recv_end_check, dmu_recv_end_sync, drc,
dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL));
}
static int
dmu_recv_new_end(dmu_recv_cookie_t *drc)
{
return (dsl_sync_task(drc->drc_tofs,
dmu_recv_end_check, dmu_recv_end_sync, drc,
dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL));
}
int
dmu_recv_end(dmu_recv_cookie_t *drc, void *owner)
{
int error;
drc->drc_owner = owner;
if (drc->drc_newfs)
error = dmu_recv_new_end(drc);
else
error = dmu_recv_existing_end(drc);
if (error != 0) {
dmu_recv_cleanup_ds(drc);
nvlist_free(drc->drc_keynvl);
} else {
if (drc->drc_newfs) {
zvol_create_minor(drc->drc_tofs);
}
char *snapname = kmem_asprintf("%s@%s",
drc->drc_tofs, drc->drc_tosnap);
zvol_create_minor(snapname);
kmem_strfree(snapname);
}
return (error);
}
/*
* Return TRUE if this objset is currently being received into.
*/
boolean_t
dmu_objset_is_receiving(objset_t *os)
{
return (os->os_dsl_dataset != NULL &&
os->os_dsl_dataset->ds_owner == dmu_recv_tag);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_length, INT, ZMOD_RW,
"Maximum receive queue length");
ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_ff, INT, ZMOD_RW,
"Receive queue fill fraction");
ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, write_batch_size, INT, ZMOD_RW,
"Maximum amount of writes to batch into one transaction");
/* END CSTYLED */
diff --git a/module/zfs/dmu_redact.c b/module/zfs/dmu_redact.c
index 7efe423d35f0..46f4982894b5 100644
--- a/module/zfs/dmu_redact.c
+++ b/module/zfs/dmu_redact.c
@@ -1,1201 +1,1202 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2017, 2018 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/txg.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_traverse.h>
#include <sys/dmu_redact.h>
#include <sys/bqueue.h>
#include <sys/objlist.h>
#include <sys/dmu_tx.h>
#ifdef _KERNEL
#include <sys/zfs_vfsops.h>
#include <sys/zap.h>
#include <sys/zfs_znode.h>
#endif
/*
* This controls the number of entries in the buffer the redaction_list_update
* synctask uses to buffer writes to the redaction list.
*/
-int redact_sync_bufsize = 1024;
+static const int redact_sync_bufsize = 1024;
/*
* Controls how often to update the redaction list when creating a redaction
* list.
*/
-uint64_t redaction_list_update_interval_ns = 1000 * 1000 * 1000ULL; /* NS */
+static const uint64_t redaction_list_update_interval_ns =
+ 1000 * 1000 * 1000ULL; /* 1s */
/*
* This tunable controls the length of the queues that zfs redact worker threads
* use to communicate. If the dmu_redact_snap thread is blocking on these
* queues, this variable may need to be increased. If there is a significant
* slowdown at the start of a redact operation as these threads consume all the
* available IO resources, or the queues are consuming too much memory, this
* variable may need to be decreased.
*/
-int zfs_redact_queue_length = 1024 * 1024;
+static const int zfs_redact_queue_length = 1024 * 1024;
/*
* These tunables control the fill fraction of the queues by zfs redact. The
* fill fraction controls the frequency with which threads have to be
* cv_signaled. If a lot of cpu time is being spent on cv_signal, then these
* should be tuned down. If the queues empty before the signalled thread can
* catch up, then these should be tuned up.
*/
-uint64_t zfs_redact_queue_ff = 20;
+static const uint64_t zfs_redact_queue_ff = 20;
struct redact_record {
bqueue_node_t ln;
boolean_t eos_marker; /* Marks the end of the stream */
uint64_t start_object;
uint64_t start_blkid;
uint64_t end_object;
uint64_t end_blkid;
uint8_t indblkshift;
uint32_t datablksz;
};
struct redact_thread_arg {
bqueue_t q;
objset_t *os; /* Objset to traverse */
dsl_dataset_t *ds; /* Dataset to traverse */
struct redact_record *current_record;
int error_code;
boolean_t cancel;
zbookmark_phys_t resume;
objlist_t *deleted_objs;
uint64_t *num_blocks_visited;
uint64_t ignore_object; /* ignore further callbacks on this */
uint64_t txg; /* txg to traverse since */
};
/*
* The redaction node is a wrapper around the redaction record that is used
* by the redaction merging thread to sort the records and determine overlaps.
*
* It contains two nodes; one sorts the records by their start_zb, and the other
* sorts the records by their end_zb.
*/
struct redact_node {
avl_node_t avl_node_start;
avl_node_t avl_node_end;
struct redact_record *record;
struct redact_thread_arg *rt_arg;
uint32_t thread_num;
};
struct merge_data {
list_t md_redact_block_pending;
redact_block_phys_t md_coalesce_block;
uint64_t md_last_time;
redact_block_phys_t md_furthest[TXG_SIZE];
/* Lists of struct redact_block_list_node. */
list_t md_blocks[TXG_SIZE];
boolean_t md_synctask_txg[TXG_SIZE];
uint64_t md_latest_synctask_txg;
redaction_list_t *md_redaction_list;
};
/*
* A wrapper around struct redact_block so it can be stored in a list_t.
*/
struct redact_block_list_node {
redact_block_phys_t block;
list_node_t node;
};
/*
* We've found a new redaction candidate. In order to improve performance, we
* coalesce these blocks when they're adjacent to each other. This function
* handles that. If the new candidate block range is immediately after the
* range we're building, coalesce it into the range we're building. Otherwise,
* put the record we're building on the queue, and update the build pointer to
* point to the new record.
*/
static void
record_merge_enqueue(bqueue_t *q, struct redact_record **build,
struct redact_record *new)
{
if (new->eos_marker) {
if (*build != NULL)
bqueue_enqueue(q, *build, sizeof (*build));
bqueue_enqueue_flush(q, new, sizeof (*new));
return;
}
if (*build == NULL) {
*build = new;
return;
}
struct redact_record *curbuild = *build;
if ((curbuild->end_object == new->start_object &&
curbuild->end_blkid + 1 == new->start_blkid &&
curbuild->end_blkid != UINT64_MAX) ||
(curbuild->end_object + 1 == new->start_object &&
curbuild->end_blkid == UINT64_MAX && new->start_blkid == 0)) {
curbuild->end_object = new->end_object;
curbuild->end_blkid = new->end_blkid;
kmem_free(new, sizeof (*new));
} else {
bqueue_enqueue(q, curbuild, sizeof (*curbuild));
*build = new;
}
}
#ifdef _KERNEL
struct objnode {
avl_node_t node;
uint64_t obj;
};
static int
objnode_compare(const void *o1, const void *o2)
{
const struct objnode *obj1 = o1;
const struct objnode *obj2 = o2;
if (obj1->obj < obj2->obj)
return (-1);
if (obj1->obj > obj2->obj)
return (1);
return (0);
}
static objlist_t *
zfs_get_deleteq(objset_t *os)
{
objlist_t *deleteq_objlist = objlist_create();
uint64_t deleteq_obj;
zap_cursor_t zc;
zap_attribute_t za;
dmu_object_info_t doi;
ASSERT3U(os->os_phys->os_type, ==, DMU_OST_ZFS);
VERIFY0(dmu_object_info(os, MASTER_NODE_OBJ, &doi));
ASSERT3U(doi.doi_type, ==, DMU_OT_MASTER_NODE);
VERIFY0(zap_lookup(os, MASTER_NODE_OBJ,
ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj));
/*
* In order to insert objects into the objlist, they must be in sorted
* order. We don't know what order we'll get them out of the ZAP in, so
* we insert them into and remove them from an avl_tree_t to sort them.
*/
avl_tree_t at;
avl_create(&at, objnode_compare, sizeof (struct objnode),
offsetof(struct objnode, node));
for (zap_cursor_init(&zc, os, deleteq_obj);
zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) {
struct objnode *obj = kmem_zalloc(sizeof (*obj), KM_SLEEP);
obj->obj = za.za_first_integer;
avl_add(&at, obj);
}
zap_cursor_fini(&zc);
struct objnode *next, *found = avl_first(&at);
while (found != NULL) {
next = AVL_NEXT(&at, found);
objlist_insert(deleteq_objlist, found->obj);
found = next;
}
void *cookie = NULL;
while ((found = avl_destroy_nodes(&at, &cookie)) != NULL)
kmem_free(found, sizeof (*found));
avl_destroy(&at);
return (deleteq_objlist);
}
#endif
/*
* This is the callback function to traverse_dataset for the redaction threads
* for dmu_redact_snap. This thread is responsible for creating redaction
* records for all the data that is modified by the snapshots we're redacting
* with respect to. Redaction records represent ranges of data that have been
* modified by one of the redaction snapshots, and are stored in the
* redact_record struct. We need to create redaction records for three
* cases:
*
* First, if there's a normal write, we need to create a redaction record for
* that block.
*
* Second, if there's a hole, we need to create a redaction record that covers
* the whole range of the hole. If the hole is in the meta-dnode, it must cover
* every block in all of the objects in the hole.
*
* Third, if there is a deleted object, we need to create a redaction record for
* all of the blocks in that object.
*/
static int
redact_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
const zbookmark_phys_t *zb, const struct dnode_phys *dnp, void *arg)
{
(void) spa, (void) zilog;
struct redact_thread_arg *rta = arg;
struct redact_record *record;
ASSERT(zb->zb_object == DMU_META_DNODE_OBJECT ||
zb->zb_object >= rta->resume.zb_object);
if (rta->cancel)
return (SET_ERROR(EINTR));
if (rta->ignore_object == zb->zb_object)
return (0);
/*
* If we're visiting a dnode, we need to handle the case where the
* object has been deleted.
*/
if (zb->zb_level == ZB_DNODE_LEVEL) {
ASSERT3U(zb->zb_level, ==, ZB_DNODE_LEVEL);
if (zb->zb_object == 0)
return (0);
/*
* If the object has been deleted, redact all of the blocks in
* it.
*/
if (dnp->dn_type == DMU_OT_NONE ||
objlist_exists(rta->deleted_objs, zb->zb_object)) {
rta->ignore_object = zb->zb_object;
record = kmem_zalloc(sizeof (struct redact_record),
KM_SLEEP);
record->eos_marker = B_FALSE;
record->start_object = record->end_object =
zb->zb_object;
record->start_blkid = 0;
record->end_blkid = UINT64_MAX;
record_merge_enqueue(&rta->q,
&rta->current_record, record);
}
return (0);
} else if (zb->zb_level < 0) {
return (0);
} else if (zb->zb_level > 0 && !BP_IS_HOLE(bp)) {
/*
* If this is an indirect block, but not a hole, it doesn't
* provide any useful information for redaction, so ignore it.
*/
return (0);
}
/*
* At this point, there are two options left for the type of block we're
* looking at. Either this is a hole (which could be in the dnode or
* the meta-dnode), or it's a level 0 block of some sort. If it's a
* hole, we create a redaction record that covers the whole range. If
* the hole is in a dnode, we need to redact all the blocks in that
* hole. If the hole is in the meta-dnode, we instead need to redact
* all blocks in every object covered by that hole. If it's a level 0
* block, we only need to redact that single block.
*/
record = kmem_zalloc(sizeof (struct redact_record), KM_SLEEP);
record->eos_marker = B_FALSE;
record->start_object = record->end_object = zb->zb_object;
if (BP_IS_HOLE(bp)) {
record->start_blkid = zb->zb_blkid *
bp_span_in_blocks(dnp->dn_indblkshift, zb->zb_level);
record->end_blkid = ((zb->zb_blkid + 1) *
bp_span_in_blocks(dnp->dn_indblkshift, zb->zb_level)) - 1;
if (zb->zb_object == DMU_META_DNODE_OBJECT) {
record->start_object = record->start_blkid *
((SPA_MINBLOCKSIZE * dnp->dn_datablkszsec) /
sizeof (dnode_phys_t));
record->start_blkid = 0;
record->end_object = ((record->end_blkid +
1) * ((SPA_MINBLOCKSIZE * dnp->dn_datablkszsec) /
sizeof (dnode_phys_t))) - 1;
record->end_blkid = UINT64_MAX;
}
} else if (zb->zb_level != 0 ||
zb->zb_object == DMU_META_DNODE_OBJECT) {
kmem_free(record, sizeof (*record));
return (0);
} else {
record->start_blkid = record->end_blkid = zb->zb_blkid;
}
record->indblkshift = dnp->dn_indblkshift;
record->datablksz = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT;
record_merge_enqueue(&rta->q, &rta->current_record, record);
return (0);
}
static void
redact_traverse_thread(void *arg)
{
struct redact_thread_arg *rt_arg = arg;
int err;
struct redact_record *data;
#ifdef _KERNEL
if (rt_arg->os->os_phys->os_type == DMU_OST_ZFS)
rt_arg->deleted_objs = zfs_get_deleteq(rt_arg->os);
else
rt_arg->deleted_objs = objlist_create();
#else
rt_arg->deleted_objs = objlist_create();
#endif
err = traverse_dataset_resume(rt_arg->ds, rt_arg->txg,
&rt_arg->resume, TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA,
redact_cb, rt_arg);
if (err != EINTR)
rt_arg->error_code = err;
objlist_destroy(rt_arg->deleted_objs);
data = kmem_zalloc(sizeof (*data), KM_SLEEP);
data->eos_marker = B_TRUE;
record_merge_enqueue(&rt_arg->q, &rt_arg->current_record, data);
thread_exit();
}
static inline void
create_zbookmark_from_obj_off(zbookmark_phys_t *zb, uint64_t object,
uint64_t blkid)
{
zb->zb_object = object;
zb->zb_level = 0;
zb->zb_blkid = blkid;
}
/*
* This is a utility function that can do the comparison for the start or ends
* of the ranges in a redact_record.
*/
static int
redact_range_compare(uint64_t obj1, uint64_t off1, uint32_t dbss1,
uint64_t obj2, uint64_t off2, uint32_t dbss2)
{
zbookmark_phys_t z1, z2;
create_zbookmark_from_obj_off(&z1, obj1, off1);
create_zbookmark_from_obj_off(&z2, obj2, off2);
return (zbookmark_compare(dbss1 >> SPA_MINBLOCKSHIFT, 0,
dbss2 >> SPA_MINBLOCKSHIFT, 0, &z1, &z2));
}
/*
* Compare two redaction records by their range's start location. Also makes
* eos records always compare last. We use the thread number in the redact_node
* to ensure that records do not compare equal (which is not allowed in our avl
* trees).
*/
static int
redact_node_compare_start(const void *arg1, const void *arg2)
{
const struct redact_node *rn1 = arg1;
const struct redact_node *rn2 = arg2;
const struct redact_record *rr1 = rn1->record;
const struct redact_record *rr2 = rn2->record;
if (rr1->eos_marker)
return (1);
if (rr2->eos_marker)
return (-1);
int cmp = redact_range_compare(rr1->start_object, rr1->start_blkid,
rr1->datablksz, rr2->start_object, rr2->start_blkid,
rr2->datablksz);
if (cmp == 0)
cmp = (rn1->thread_num < rn2->thread_num ? -1 : 1);
return (cmp);
}
/*
* Compare two redaction records by their range's end location. Also makes
* eos records always compare last. We use the thread number in the redact_node
* to ensure that records do not compare equal (which is not allowed in our avl
* trees).
*/
static int
redact_node_compare_end(const void *arg1, const void *arg2)
{
const struct redact_node *rn1 = arg1;
const struct redact_node *rn2 = arg2;
const struct redact_record *srr1 = rn1->record;
const struct redact_record *srr2 = rn2->record;
if (srr1->eos_marker)
return (1);
if (srr2->eos_marker)
return (-1);
int cmp = redact_range_compare(srr1->end_object, srr1->end_blkid,
srr1->datablksz, srr2->end_object, srr2->end_blkid,
srr2->datablksz);
if (cmp == 0)
cmp = (rn1->thread_num < rn2->thread_num ? -1 : 1);
return (cmp);
}
/*
* Utility function that compares two redaction records to determine if any part
* of the "from" record is before any part of the "to" record. Also causes End
* of Stream redaction records to compare after all others, so that the
* redaction merging logic can stay simple.
*/
static boolean_t
redact_record_before(const struct redact_record *from,
const struct redact_record *to)
{
if (from->eos_marker == B_TRUE)
return (B_FALSE);
else if (to->eos_marker == B_TRUE)
return (B_TRUE);
return (redact_range_compare(from->start_object, from->start_blkid,
from->datablksz, to->end_object, to->end_blkid,
to->datablksz) <= 0);
}
/*
* Pop a new redaction record off the queue, check that the records are in the
* right order, and free the old data.
*/
static struct redact_record *
get_next_redact_record(bqueue_t *bq, struct redact_record *prev)
{
struct redact_record *next = bqueue_dequeue(bq);
ASSERT(redact_record_before(prev, next));
kmem_free(prev, sizeof (*prev));
return (next);
}
/*
* Remove the given redaction node from both trees, pull a new redaction record
* off the queue, free the old redaction record, update the redaction node, and
* reinsert the node into the trees.
*/
static int
update_avl_trees(avl_tree_t *start_tree, avl_tree_t *end_tree,
struct redact_node *redact_node)
{
avl_remove(start_tree, redact_node);
avl_remove(end_tree, redact_node);
redact_node->record = get_next_redact_record(&redact_node->rt_arg->q,
redact_node->record);
avl_add(end_tree, redact_node);
avl_add(start_tree, redact_node);
return (redact_node->rt_arg->error_code);
}
/*
* Synctask for updating redaction lists. We first take this txg's list of
* redacted blocks and append those to the redaction list. We then update the
* redaction list's bonus buffer. We store the furthest blocks we visited and
* the list of snapshots that we're redacting with respect to. We need these so
* that redacted sends and receives can be correctly resumed.
*/
static void
redaction_list_update_sync(void *arg, dmu_tx_t *tx)
{
struct merge_data *md = arg;
uint64_t txg = dmu_tx_get_txg(tx);
list_t *list = &md->md_blocks[txg & TXG_MASK];
redact_block_phys_t *furthest_visited =
&md->md_furthest[txg & TXG_MASK];
objset_t *mos = tx->tx_pool->dp_meta_objset;
redaction_list_t *rl = md->md_redaction_list;
int bufsize = redact_sync_bufsize;
redact_block_phys_t *buf = kmem_alloc(bufsize * sizeof (*buf),
KM_SLEEP);
int index = 0;
dmu_buf_will_dirty(rl->rl_dbuf, tx);
for (struct redact_block_list_node *rbln = list_remove_head(list);
rbln != NULL; rbln = list_remove_head(list)) {
ASSERT3U(rbln->block.rbp_object, <=,
furthest_visited->rbp_object);
ASSERT(rbln->block.rbp_object < furthest_visited->rbp_object ||
rbln->block.rbp_blkid <= furthest_visited->rbp_blkid);
buf[index] = rbln->block;
index++;
if (index == bufsize) {
dmu_write(mos, rl->rl_object,
rl->rl_phys->rlp_num_entries * sizeof (*buf),
bufsize * sizeof (*buf), buf, tx);
rl->rl_phys->rlp_num_entries += bufsize;
index = 0;
}
kmem_free(rbln, sizeof (*rbln));
}
if (index > 0) {
dmu_write(mos, rl->rl_object, rl->rl_phys->rlp_num_entries *
sizeof (*buf), index * sizeof (*buf), buf, tx);
rl->rl_phys->rlp_num_entries += index;
}
kmem_free(buf, bufsize * sizeof (*buf));
md->md_synctask_txg[txg & TXG_MASK] = B_FALSE;
rl->rl_phys->rlp_last_object = furthest_visited->rbp_object;
rl->rl_phys->rlp_last_blkid = furthest_visited->rbp_blkid;
}
static void
commit_rl_updates(objset_t *os, struct merge_data *md, uint64_t object,
uint64_t blkid)
{
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(os->os_spa)->dp_mos_dir);
dmu_tx_hold_space(tx, sizeof (struct redact_block_list_node));
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
uint64_t txg = dmu_tx_get_txg(tx);
if (!md->md_synctask_txg[txg & TXG_MASK]) {
dsl_sync_task_nowait(dmu_tx_pool(tx),
redaction_list_update_sync, md, tx);
md->md_synctask_txg[txg & TXG_MASK] = B_TRUE;
md->md_latest_synctask_txg = txg;
}
md->md_furthest[txg & TXG_MASK].rbp_object = object;
md->md_furthest[txg & TXG_MASK].rbp_blkid = blkid;
list_move_tail(&md->md_blocks[txg & TXG_MASK],
&md->md_redact_block_pending);
dmu_tx_commit(tx);
md->md_last_time = gethrtime();
}
/*
* We want to store the list of blocks that we're redacting in the bookmark's
* redaction list. However, this list is stored in the MOS, which means it can
* only be written to in syncing context. To get around this, we create a
* synctask that will write to the mos for us. We tell it what to write by
* a linked list for each current transaction group; every time we decide to
* redact a block, we append it to the transaction group that is currently in
* open context. We also update some progress information that the synctask
* will store to enable resumable redacted sends.
*/
static void
update_redaction_list(struct merge_data *md, objset_t *os,
uint64_t object, uint64_t blkid, uint64_t endblkid, uint32_t blksz)
{
boolean_t enqueue = B_FALSE;
redact_block_phys_t cur = {0};
uint64_t count = endblkid - blkid + 1;
while (count > REDACT_BLOCK_MAX_COUNT) {
update_redaction_list(md, os, object, blkid,
blkid + REDACT_BLOCK_MAX_COUNT - 1, blksz);
blkid += REDACT_BLOCK_MAX_COUNT;
count -= REDACT_BLOCK_MAX_COUNT;
}
redact_block_phys_t *coalesce = &md->md_coalesce_block;
boolean_t new;
if (coalesce->rbp_size_count == 0) {
new = B_TRUE;
enqueue = B_FALSE;
} else {
uint64_t old_count = redact_block_get_count(coalesce);
if (coalesce->rbp_object == object &&
coalesce->rbp_blkid + old_count == blkid &&
old_count + count <= REDACT_BLOCK_MAX_COUNT) {
ASSERT3U(redact_block_get_size(coalesce), ==, blksz);
redact_block_set_count(coalesce, old_count + count);
new = B_FALSE;
enqueue = B_FALSE;
} else {
new = B_TRUE;
enqueue = B_TRUE;
}
}
if (new) {
cur = *coalesce;
coalesce->rbp_blkid = blkid;
coalesce->rbp_object = object;
redact_block_set_count(coalesce, count);
redact_block_set_size(coalesce, blksz);
}
if (enqueue && redact_block_get_size(&cur) != 0) {
struct redact_block_list_node *rbln =
kmem_alloc(sizeof (struct redact_block_list_node),
KM_SLEEP);
rbln->block = cur;
list_insert_tail(&md->md_redact_block_pending, rbln);
}
if (gethrtime() > md->md_last_time +
redaction_list_update_interval_ns) {
commit_rl_updates(os, md, object, blkid);
}
}
/*
* This thread merges all the redaction records provided by the worker threads,
* and determines which blocks are redacted by all the snapshots. The algorithm
* for doing so is similar to performing a merge in mergesort with n sub-lists
* instead of 2, with some added complexity due to the fact that the entries are
* ranges, not just single blocks. This algorithm relies on the fact that the
* queues are sorted, which is ensured by the fact that traverse_dataset
* traverses the dataset in a consistent order. We pull one entry off the front
* of the queues of each secure dataset traversal thread. Then we repeat the
* following: each record represents a range of blocks modified by one of the
* redaction snapshots, and each block in that range may need to be redacted in
* the send stream. Find the record with the latest start of its range, and the
* record with the earliest end of its range. If the last start is before the
* first end, then we know that the blocks in the range [last_start, first_end]
* are covered by all of the ranges at the front of the queues, which means
* every thread redacts that whole range. For example, let's say the ranges on
* each queue look like this:
*
* Block Id 1 2 3 4 5 6 7 8 9 10 11
* Thread 1 | [====================]
* Thread 2 | [========]
* Thread 3 | [=================]
*
* Thread 3 has the last start (5), and the thread 2 has the last end (6). All
* three threads modified the range [5,6], so that data should not be sent over
* the wire. After we've determined whether or not to redact anything, we take
* the record with the first end. We discard that record, and pull a new one
* off the front of the queue it came from. In the above example, we would
* discard Thread 2's record, and pull a new one. Let's say the next record we
* pulled from Thread 2 covered range [10,11]. The new layout would look like
* this:
*
* Block Id 1 2 3 4 5 6 7 8 9 10 11
* Thread 1 | [====================]
* Thread 2 | [==]
* Thread 3 | [=================]
*
* When we compare the last start (10, from Thread 2) and the first end (9, from
* Thread 1), we see that the last start is greater than the first end.
* Therefore, we do not redact anything from these records. We'll iterate by
* replacing the record from Thread 1.
*
* We iterate by replacing the record with the lowest end because we know
* that the record with the lowest end has helped us as much as it can. All the
* ranges before it that we will ever redact have been redacted. In addition,
* by replacing the one with the lowest end, we guarantee we catch all ranges
* that need to be redacted. For example, if in the case above we had replaced
* the record from Thread 1 instead, we might have ended up with the following:
*
* Block Id 1 2 3 4 5 6 7 8 9 10 11 12
* Thread 1 | [==]
* Thread 2 | [========]
* Thread 3 | [=================]
*
* If the next record from Thread 2 had been [8,10], for example, we should have
* redacted part of that range, but because we updated Thread 1's record, we
* missed it.
*
* We implement this algorithm by using two trees. The first sorts the
* redaction records by their start_zb, and the second sorts them by their
* end_zb. We use these to find the record with the last start and the record
* with the first end. We create a record with that start and end, and send it
* on. The overall runtime of this implementation is O(n log m), where n is the
* total number of redaction records from all the different redaction snapshots,
* and m is the number of redaction snapshots.
*
* If we redact with respect to zero snapshots, we create a redaction
* record with the start object and blkid to 0, and the end object and blkid to
* UINT64_MAX. This will result in us redacting every block.
*/
static int
perform_thread_merge(bqueue_t *q, uint32_t num_threads,
struct redact_thread_arg *thread_args, boolean_t *cancel)
{
struct redact_node *redact_nodes = NULL;
avl_tree_t start_tree, end_tree;
struct redact_record *record;
struct redact_record *current_record = NULL;
int err = 0;
struct merge_data md = { {0} };
list_create(&md.md_redact_block_pending,
sizeof (struct redact_block_list_node),
offsetof(struct redact_block_list_node, node));
/*
* If we're redacting with respect to zero snapshots, then no data is
* permitted to be sent. We enqueue a record that redacts all blocks,
* and an eos marker.
*/
if (num_threads == 0) {
record = kmem_zalloc(sizeof (struct redact_record),
KM_SLEEP);
// We can't redact object 0, so don't try.
record->start_object = 1;
record->start_blkid = 0;
record->end_object = record->end_blkid = UINT64_MAX;
bqueue_enqueue(q, record, sizeof (*record));
return (0);
}
if (num_threads > 0) {
redact_nodes = kmem_zalloc(num_threads *
sizeof (*redact_nodes), KM_SLEEP);
}
avl_create(&start_tree, redact_node_compare_start,
sizeof (struct redact_node),
offsetof(struct redact_node, avl_node_start));
avl_create(&end_tree, redact_node_compare_end,
sizeof (struct redact_node),
offsetof(struct redact_node, avl_node_end));
for (int i = 0; i < num_threads; i++) {
struct redact_node *node = &redact_nodes[i];
struct redact_thread_arg *targ = &thread_args[i];
node->record = bqueue_dequeue(&targ->q);
node->rt_arg = targ;
node->thread_num = i;
avl_add(&start_tree, node);
avl_add(&end_tree, node);
}
/*
* Once the first record in the end tree has returned EOS, every record
* must be an EOS record, so we should stop.
*/
while (err == 0 && !((struct redact_node *)avl_first(&end_tree))->
record->eos_marker) {
if (*cancel) {
err = EINTR;
break;
}
struct redact_node *last_start = avl_last(&start_tree);
struct redact_node *first_end = avl_first(&end_tree);
/*
* If the last start record is before the first end record,
* then we have blocks that are redacted by all threads.
* Therefore, we should redact them. Copy the record, and send
* it to the main thread.
*/
if (redact_record_before(last_start->record,
first_end->record)) {
record = kmem_zalloc(sizeof (struct redact_record),
KM_SLEEP);
*record = *first_end->record;
record->start_object = last_start->record->start_object;
record->start_blkid = last_start->record->start_blkid;
record_merge_enqueue(q, &current_record,
record);
}
err = update_avl_trees(&start_tree, &end_tree, first_end);
}
/*
* We're done; if we were cancelled, we need to cancel our workers and
* clear out their queues. Either way, we need to remove every thread's
* redact_node struct from the avl trees.
*/
for (int i = 0; i < num_threads; i++) {
if (err != 0) {
thread_args[i].cancel = B_TRUE;
while (!redact_nodes[i].record->eos_marker) {
(void) update_avl_trees(&start_tree, &end_tree,
&redact_nodes[i]);
}
}
avl_remove(&start_tree, &redact_nodes[i]);
avl_remove(&end_tree, &redact_nodes[i]);
kmem_free(redact_nodes[i].record,
sizeof (struct redact_record));
bqueue_destroy(&thread_args[i].q);
}
avl_destroy(&start_tree);
avl_destroy(&end_tree);
kmem_free(redact_nodes, num_threads * sizeof (*redact_nodes));
if (current_record != NULL)
bqueue_enqueue(q, current_record, sizeof (current_record));
return (err);
}
struct redact_merge_thread_arg {
bqueue_t q;
spa_t *spa;
int numsnaps;
struct redact_thread_arg *thr_args;
boolean_t cancel;
int error_code;
};
static void
redact_merge_thread(void *arg)
{
struct redact_merge_thread_arg *rmta = arg;
rmta->error_code = perform_thread_merge(&rmta->q,
rmta->numsnaps, rmta->thr_args, &rmta->cancel);
struct redact_record *rec = kmem_zalloc(sizeof (*rec), KM_SLEEP);
rec->eos_marker = B_TRUE;
bqueue_enqueue_flush(&rmta->q, rec, 1);
thread_exit();
}
/*
* Find the next object in or after the redaction range passed in, and hold
* its dnode with the provided tag. Also update *object to contain the new
* object number.
*/
static int
hold_next_object(objset_t *os, struct redact_record *rec, void *tag,
uint64_t *object, dnode_t **dn)
{
int err = 0;
if (*dn != NULL)
dnode_rele(*dn, tag);
*dn = NULL;
if (*object < rec->start_object) {
*object = rec->start_object - 1;
}
err = dmu_object_next(os, object, B_FALSE, 0);
if (err != 0)
return (err);
err = dnode_hold(os, *object, tag, dn);
while (err == 0 && (*object < rec->start_object ||
DMU_OT_IS_METADATA((*dn)->dn_type))) {
dnode_rele(*dn, tag);
*dn = NULL;
err = dmu_object_next(os, object, B_FALSE, 0);
if (err != 0)
break;
err = dnode_hold(os, *object, tag, dn);
}
return (err);
}
static int
perform_redaction(objset_t *os, redaction_list_t *rl,
struct redact_merge_thread_arg *rmta)
{
int err = 0;
bqueue_t *q = &rmta->q;
struct redact_record *rec = NULL;
struct merge_data md = { {0} };
list_create(&md.md_redact_block_pending,
sizeof (struct redact_block_list_node),
offsetof(struct redact_block_list_node, node));
md.md_redaction_list = rl;
for (int i = 0; i < TXG_SIZE; i++) {
list_create(&md.md_blocks[i],
sizeof (struct redact_block_list_node),
offsetof(struct redact_block_list_node, node));
}
dnode_t *dn = NULL;
uint64_t prev_obj = 0;
for (rec = bqueue_dequeue(q); !rec->eos_marker && err == 0;
rec = get_next_redact_record(q, rec)) {
ASSERT3U(rec->start_object, !=, 0);
uint64_t object;
if (prev_obj != rec->start_object) {
object = rec->start_object - 1;
err = hold_next_object(os, rec, FTAG, &object, &dn);
} else {
object = prev_obj;
}
while (err == 0 && object <= rec->end_object) {
if (issig(JUSTLOOKING) && issig(FORREAL)) {
err = EINTR;
break;
}
/*
* Part of the current object is contained somewhere in
* the range covered by rec.
*/
uint64_t startblkid;
uint64_t endblkid;
uint64_t maxblkid = dn->dn_phys->dn_maxblkid;
if (rec->start_object < object)
startblkid = 0;
else if (rec->start_blkid > maxblkid)
break;
else
startblkid = rec->start_blkid;
if (rec->end_object > object || rec->end_blkid >
maxblkid) {
endblkid = maxblkid;
} else {
endblkid = rec->end_blkid;
}
update_redaction_list(&md, os, object, startblkid,
endblkid, dn->dn_datablksz);
if (object == rec->end_object)
break;
err = hold_next_object(os, rec, FTAG, &object, &dn);
}
if (err == ESRCH)
err = 0;
if (dn != NULL)
prev_obj = object;
}
if (err == 0 && dn != NULL)
dnode_rele(dn, FTAG);
if (err == ESRCH)
err = 0;
rmta->cancel = B_TRUE;
while (!rec->eos_marker)
rec = get_next_redact_record(q, rec);
kmem_free(rec, sizeof (*rec));
/*
* There may be a block that's being coalesced, sync that out before we
* return.
*/
if (err == 0 && md.md_coalesce_block.rbp_size_count != 0) {
struct redact_block_list_node *rbln =
kmem_alloc(sizeof (struct redact_block_list_node),
KM_SLEEP);
rbln->block = md.md_coalesce_block;
list_insert_tail(&md.md_redact_block_pending, rbln);
}
commit_rl_updates(os, &md, UINT64_MAX, UINT64_MAX);
/*
* Wait for all the redaction info to sync out before we return, so that
* anyone who attempts to resume this redaction will have all the data
* they need.
*/
dsl_pool_t *dp = spa_get_dsl(os->os_spa);
if (md.md_latest_synctask_txg != 0)
txg_wait_synced(dp, md.md_latest_synctask_txg);
for (int i = 0; i < TXG_SIZE; i++)
list_destroy(&md.md_blocks[i]);
return (err);
}
static boolean_t
redact_snaps_contains(uint64_t *snaps, uint64_t num_snaps, uint64_t guid)
{
for (int i = 0; i < num_snaps; i++) {
if (snaps[i] == guid)
return (B_TRUE);
}
return (B_FALSE);
}
int
dmu_redact_snap(const char *snapname, nvlist_t *redactnvl,
const char *redactbook)
{
int err = 0;
dsl_pool_t *dp = NULL;
dsl_dataset_t *ds = NULL;
int numsnaps = 0;
objset_t *os;
struct redact_thread_arg *args = NULL;
redaction_list_t *new_rl = NULL;
char *newredactbook;
if ((err = dsl_pool_hold(snapname, FTAG, &dp)) != 0)
return (err);
newredactbook = kmem_zalloc(sizeof (char) * ZFS_MAX_DATASET_NAME_LEN,
KM_SLEEP);
if ((err = dsl_dataset_hold_flags(dp, snapname, DS_HOLD_FLAG_DECRYPT,
FTAG, &ds)) != 0) {
goto out;
}
dsl_dataset_long_hold(ds, FTAG);
if (!ds->ds_is_snapshot || dmu_objset_from_ds(ds, &os) != 0) {
err = EINVAL;
goto out;
}
if (dsl_dataset_feature_is_active(ds, SPA_FEATURE_REDACTED_DATASETS)) {
err = EALREADY;
goto out;
}
numsnaps = fnvlist_num_pairs(redactnvl);
if (numsnaps > 0)
args = kmem_zalloc(numsnaps * sizeof (*args), KM_SLEEP);
nvpair_t *pair = NULL;
for (int i = 0; i < numsnaps; i++) {
pair = nvlist_next_nvpair(redactnvl, pair);
const char *name = nvpair_name(pair);
struct redact_thread_arg *rta = &args[i];
err = dsl_dataset_hold_flags(dp, name, DS_HOLD_FLAG_DECRYPT,
FTAG, &rta->ds);
if (err != 0)
break;
/*
* We want to do the long hold before we can get any other
* errors, because the cleanup code will release the long
* hold if rta->ds is filled in.
*/
dsl_dataset_long_hold(rta->ds, FTAG);
err = dmu_objset_from_ds(rta->ds, &rta->os);
if (err != 0)
break;
if (!dsl_dataset_is_before(rta->ds, ds, 0)) {
err = EINVAL;
break;
}
if (dsl_dataset_feature_is_active(rta->ds,
SPA_FEATURE_REDACTED_DATASETS)) {
err = EALREADY;
break;
}
}
if (err != 0)
goto out;
VERIFY3P(nvlist_next_nvpair(redactnvl, pair), ==, NULL);
boolean_t resuming = B_FALSE;
zfs_bookmark_phys_t bookmark;
(void) strlcpy(newredactbook, snapname, ZFS_MAX_DATASET_NAME_LEN);
char *c = strchr(newredactbook, '@');
ASSERT3P(c, !=, NULL);
int n = snprintf(c, ZFS_MAX_DATASET_NAME_LEN - (c - newredactbook),
"#%s", redactbook);
if (n >= ZFS_MAX_DATASET_NAME_LEN - (c - newredactbook)) {
dsl_pool_rele(dp, FTAG);
kmem_free(newredactbook,
sizeof (char) * ZFS_MAX_DATASET_NAME_LEN);
if (args != NULL)
kmem_free(args, numsnaps * sizeof (*args));
return (SET_ERROR(ENAMETOOLONG));
}
err = dsl_bookmark_lookup(dp, newredactbook, NULL, &bookmark);
if (err == 0) {
resuming = B_TRUE;
if (bookmark.zbm_redaction_obj == 0) {
err = EEXIST;
goto out;
}
err = dsl_redaction_list_hold_obj(dp,
bookmark.zbm_redaction_obj, FTAG, &new_rl);
if (err != 0) {
err = EIO;
goto out;
}
dsl_redaction_list_long_hold(dp, new_rl, FTAG);
if (new_rl->rl_phys->rlp_num_snaps != numsnaps) {
err = ESRCH;
goto out;
}
for (int i = 0; i < numsnaps; i++) {
struct redact_thread_arg *rta = &args[i];
if (!redact_snaps_contains(new_rl->rl_phys->rlp_snaps,
new_rl->rl_phys->rlp_num_snaps,
dsl_dataset_phys(rta->ds)->ds_guid)) {
err = ESRCH;
goto out;
}
}
if (new_rl->rl_phys->rlp_last_blkid == UINT64_MAX &&
new_rl->rl_phys->rlp_last_object == UINT64_MAX) {
err = EEXIST;
goto out;
}
dsl_pool_rele(dp, FTAG);
dp = NULL;
} else {
uint64_t *guids = NULL;
if (numsnaps > 0) {
guids = kmem_zalloc(numsnaps * sizeof (uint64_t),
KM_SLEEP);
}
for (int i = 0; i < numsnaps; i++) {
struct redact_thread_arg *rta = &args[i];
guids[i] = dsl_dataset_phys(rta->ds)->ds_guid;
}
dsl_pool_rele(dp, FTAG);
dp = NULL;
err = dsl_bookmark_create_redacted(newredactbook, snapname,
numsnaps, guids, FTAG, &new_rl);
kmem_free(guids, numsnaps * sizeof (uint64_t));
if (err != 0) {
goto out;
}
}
for (int i = 0; i < numsnaps; i++) {
struct redact_thread_arg *rta = &args[i];
(void) bqueue_init(&rta->q, zfs_redact_queue_ff,
zfs_redact_queue_length,
offsetof(struct redact_record, ln));
if (resuming) {
rta->resume.zb_blkid =
new_rl->rl_phys->rlp_last_blkid;
rta->resume.zb_object =
new_rl->rl_phys->rlp_last_object;
}
rta->txg = dsl_dataset_phys(ds)->ds_creation_txg;
(void) thread_create(NULL, 0, redact_traverse_thread, rta,
0, curproc, TS_RUN, minclsyspri);
}
struct redact_merge_thread_arg *rmta;
rmta = kmem_zalloc(sizeof (struct redact_merge_thread_arg), KM_SLEEP);
(void) bqueue_init(&rmta->q, zfs_redact_queue_ff,
zfs_redact_queue_length, offsetof(struct redact_record, ln));
rmta->numsnaps = numsnaps;
rmta->spa = os->os_spa;
rmta->thr_args = args;
(void) thread_create(NULL, 0, redact_merge_thread, rmta, 0, curproc,
TS_RUN, minclsyspri);
err = perform_redaction(os, new_rl, rmta);
bqueue_destroy(&rmta->q);
kmem_free(rmta, sizeof (struct redact_merge_thread_arg));
out:
kmem_free(newredactbook, sizeof (char) * ZFS_MAX_DATASET_NAME_LEN);
if (new_rl != NULL) {
dsl_redaction_list_long_rele(new_rl, FTAG);
dsl_redaction_list_rele(new_rl, FTAG);
}
for (int i = 0; i < numsnaps; i++) {
struct redact_thread_arg *rta = &args[i];
/*
* rta->ds may be NULL if we got an error while filling
* it in.
*/
if (rta->ds != NULL) {
dsl_dataset_long_rele(rta->ds, FTAG);
dsl_dataset_rele_flags(rta->ds,
DS_HOLD_FLAG_DECRYPT, FTAG);
}
}
if (args != NULL)
kmem_free(args, numsnaps * sizeof (*args));
if (dp != NULL)
dsl_pool_rele(dp, FTAG);
if (ds != NULL) {
dsl_dataset_long_rele(ds, FTAG);
dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG);
}
return (SET_ERROR(err));
}
diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c
index 6cff7fd58046..fbb1947a5dd3 100644
--- a/module/zfs/dmu_send.c
+++ b/module/zfs/dmu_send.c
@@ -1,3108 +1,3108 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
* Copyright (c) 2014, Joyent, Inc. All rights reserved.
* Copyright 2014 HybridCluster. All rights reserved.
* Copyright 2016 RackTop Systems.
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
*/
#include <sys/dmu.h>
#include <sys/dmu_impl.h>
#include <sys/dmu_tx.h>
#include <sys/dbuf.h>
#include <sys/dnode.h>
#include <sys/zfs_context.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_traverse.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_synctask.h>
#include <sys/spa_impl.h>
#include <sys/zfs_ioctl.h>
#include <sys/zap.h>
#include <sys/zio_checksum.h>
#include <sys/zfs_znode.h>
#include <zfs_fletcher.h>
#include <sys/avl.h>
#include <sys/ddt.h>
#include <sys/zfs_onexit.h>
#include <sys/dmu_send.h>
#include <sys/dmu_recv.h>
#include <sys/dsl_destroy.h>
#include <sys/blkptr.h>
#include <sys/dsl_bookmark.h>
#include <sys/zfeature.h>
#include <sys/bqueue.h>
#include <sys/zvol.h>
#include <sys/policy.h>
#include <sys/objlist.h>
#ifdef _KERNEL
#include <sys/zfs_vfsops.h>
#endif
/* Set this tunable to TRUE to replace corrupt data with 0x2f5baddb10c */
-int zfs_send_corrupt_data = B_FALSE;
+static int zfs_send_corrupt_data = B_FALSE;
/*
* This tunable controls the amount of data (measured in bytes) that will be
* prefetched by zfs send. If the main thread is blocking on reads that haven't
* completed, this variable might need to be increased. If instead the main
* thread is issuing new reads because the prefetches have fallen out of the
* cache, this may need to be decreased.
*/
-int zfs_send_queue_length = SPA_MAXBLOCKSIZE;
+static int zfs_send_queue_length = SPA_MAXBLOCKSIZE;
/*
* This tunable controls the length of the queues that zfs send worker threads
* use to communicate. If the send_main_thread is blocking on these queues,
* this variable may need to be increased. If there is a significant slowdown
* at the start of a send as these threads consume all the available IO
* resources, this variable may need to be decreased.
*/
-int zfs_send_no_prefetch_queue_length = 1024 * 1024;
+static int zfs_send_no_prefetch_queue_length = 1024 * 1024;
/*
* These tunables control the fill fraction of the queues by zfs send. The fill
* fraction controls the frequency with which threads have to be cv_signaled.
* If a lot of cpu time is being spent on cv_signal, then these should be tuned
* down. If the queues empty before the signalled thread can catch up, then
* these should be tuned up.
*/
-int zfs_send_queue_ff = 20;
-int zfs_send_no_prefetch_queue_ff = 20;
+static int zfs_send_queue_ff = 20;
+static int zfs_send_no_prefetch_queue_ff = 20;
/*
* Use this to override the recordsize calculation for fast zfs send estimates.
*/
-int zfs_override_estimate_recordsize = 0;
+static int zfs_override_estimate_recordsize = 0;
/* Set this tunable to FALSE to disable setting of DRR_FLAG_FREERECORDS */
-int zfs_send_set_freerecords_bit = B_TRUE;
+static const boolean_t zfs_send_set_freerecords_bit = B_TRUE;
/* Set this tunable to FALSE is disable sending unmodified spill blocks. */
-int zfs_send_unmodified_spill_blocks = B_TRUE;
+static int zfs_send_unmodified_spill_blocks = B_TRUE;
static inline boolean_t
overflow_multiply(uint64_t a, uint64_t b, uint64_t *c)
{
uint64_t temp = a * b;
if (b != 0 && temp / b != a)
return (B_FALSE);
*c = temp;
return (B_TRUE);
}
struct send_thread_arg {
bqueue_t q;
objset_t *os; /* Objset to traverse */
uint64_t fromtxg; /* Traverse from this txg */
int flags; /* flags to pass to traverse_dataset */
int error_code;
boolean_t cancel;
zbookmark_phys_t resume;
uint64_t *num_blocks_visited;
};
struct redact_list_thread_arg {
boolean_t cancel;
bqueue_t q;
zbookmark_phys_t resume;
redaction_list_t *rl;
boolean_t mark_redact;
int error_code;
uint64_t *num_blocks_visited;
};
struct send_merge_thread_arg {
bqueue_t q;
objset_t *os;
struct redact_list_thread_arg *from_arg;
struct send_thread_arg *to_arg;
struct redact_list_thread_arg *redact_arg;
int error;
boolean_t cancel;
};
struct send_range {
boolean_t eos_marker; /* Marks the end of the stream */
uint64_t object;
uint64_t start_blkid;
uint64_t end_blkid;
bqueue_node_t ln;
enum type {DATA, HOLE, OBJECT, OBJECT_RANGE, REDACT,
PREVIOUSLY_REDACTED} type;
union {
struct srd {
dmu_object_type_t obj_type;
uint32_t datablksz; // logical size
uint32_t datasz; // payload size
blkptr_t bp;
arc_buf_t *abuf;
abd_t *abd;
kmutex_t lock;
kcondvar_t cv;
boolean_t io_outstanding;
boolean_t io_compressed;
int io_err;
} data;
struct srh {
uint32_t datablksz;
} hole;
struct sro {
/*
* This is a pointer because embedding it in the
* struct causes these structures to be massively larger
* for all range types; this makes the code much less
* memory efficient.
*/
dnode_phys_t *dnp;
blkptr_t bp;
} object;
struct srr {
uint32_t datablksz;
} redact;
struct sror {
blkptr_t bp;
} object_range;
} sru;
};
/*
* The list of data whose inclusion in a send stream can be pending from
* one call to backup_cb to another. Multiple calls to dump_free(),
* dump_freeobjects(), and dump_redact() can be aggregated into a single
* DRR_FREE, DRR_FREEOBJECTS, or DRR_REDACT replay record.
*/
typedef enum {
PENDING_NONE,
PENDING_FREE,
PENDING_FREEOBJECTS,
PENDING_REDACT
} dmu_pendop_t;
typedef struct dmu_send_cookie {
dmu_replay_record_t *dsc_drr;
dmu_send_outparams_t *dsc_dso;
offset_t *dsc_off;
objset_t *dsc_os;
zio_cksum_t dsc_zc;
uint64_t dsc_toguid;
uint64_t dsc_fromtxg;
int dsc_err;
dmu_pendop_t dsc_pending_op;
uint64_t dsc_featureflags;
uint64_t dsc_last_data_object;
uint64_t dsc_last_data_offset;
uint64_t dsc_resume_object;
uint64_t dsc_resume_offset;
boolean_t dsc_sent_begin;
boolean_t dsc_sent_end;
} dmu_send_cookie_t;
static int do_dump(dmu_send_cookie_t *dscp, struct send_range *range);
static void
range_free(struct send_range *range)
{
if (range->type == OBJECT) {
size_t size = sizeof (dnode_phys_t) *
(range->sru.object.dnp->dn_extra_slots + 1);
kmem_free(range->sru.object.dnp, size);
} else if (range->type == DATA) {
mutex_enter(&range->sru.data.lock);
while (range->sru.data.io_outstanding)
cv_wait(&range->sru.data.cv, &range->sru.data.lock);
if (range->sru.data.abd != NULL)
abd_free(range->sru.data.abd);
if (range->sru.data.abuf != NULL) {
arc_buf_destroy(range->sru.data.abuf,
&range->sru.data.abuf);
}
mutex_exit(&range->sru.data.lock);
cv_destroy(&range->sru.data.cv);
mutex_destroy(&range->sru.data.lock);
}
kmem_free(range, sizeof (*range));
}
/*
* For all record types except BEGIN, fill in the checksum (overlaid in
* drr_u.drr_checksum.drr_checksum). The checksum verifies everything
* up to the start of the checksum itself.
*/
static int
dump_record(dmu_send_cookie_t *dscp, void *payload, int payload_len)
{
dmu_send_outparams_t *dso = dscp->dsc_dso;
ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
(void) fletcher_4_incremental_native(dscp->dsc_drr,
offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
&dscp->dsc_zc);
if (dscp->dsc_drr->drr_type == DRR_BEGIN) {
dscp->dsc_sent_begin = B_TRUE;
} else {
ASSERT(ZIO_CHECKSUM_IS_ZERO(&dscp->dsc_drr->drr_u.
drr_checksum.drr_checksum));
dscp->dsc_drr->drr_u.drr_checksum.drr_checksum = dscp->dsc_zc;
}
if (dscp->dsc_drr->drr_type == DRR_END) {
dscp->dsc_sent_end = B_TRUE;
}
(void) fletcher_4_incremental_native(&dscp->dsc_drr->
drr_u.drr_checksum.drr_checksum,
sizeof (zio_cksum_t), &dscp->dsc_zc);
*dscp->dsc_off += sizeof (dmu_replay_record_t);
dscp->dsc_err = dso->dso_outfunc(dscp->dsc_os, dscp->dsc_drr,
sizeof (dmu_replay_record_t), dso->dso_arg);
if (dscp->dsc_err != 0)
return (SET_ERROR(EINTR));
if (payload_len != 0) {
*dscp->dsc_off += payload_len;
/*
* payload is null when dso_dryrun == B_TRUE (i.e. when we're
* doing a send size calculation)
*/
if (payload != NULL) {
(void) fletcher_4_incremental_native(
payload, payload_len, &dscp->dsc_zc);
}
/*
* The code does not rely on this (len being a multiple of 8).
* We keep this assertion because of the corresponding assertion
* in receive_read(). Keeping this assertion ensures that we do
* not inadvertently break backwards compatibility (causing the
* assertion in receive_read() to trigger on old software).
*
* Raw sends cannot be received on old software, and so can
* bypass this assertion.
*/
ASSERT((payload_len % 8 == 0) ||
(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW));
dscp->dsc_err = dso->dso_outfunc(dscp->dsc_os, payload,
payload_len, dso->dso_arg);
if (dscp->dsc_err != 0)
return (SET_ERROR(EINTR));
}
return (0);
}
/*
* Fill in the drr_free struct, or perform aggregation if the previous record is
* also a free record, and the two are adjacent.
*
* Note that we send free records even for a full send, because we want to be
* able to receive a full send as a clone, which requires a list of all the free
* and freeobject records that were generated on the source.
*/
static int
dump_free(dmu_send_cookie_t *dscp, uint64_t object, uint64_t offset,
uint64_t length)
{
struct drr_free *drrf = &(dscp->dsc_drr->drr_u.drr_free);
/*
* When we receive a free record, dbuf_free_range() assumes
* that the receiving system doesn't have any dbufs in the range
* being freed. This is always true because there is a one-record
* constraint: we only send one WRITE record for any given
* object,offset. We know that the one-record constraint is
* true because we always send data in increasing order by
* object,offset.
*
* If the increasing-order constraint ever changes, we should find
* another way to assert that the one-record constraint is still
* satisfied.
*/
ASSERT(object > dscp->dsc_last_data_object ||
(object == dscp->dsc_last_data_object &&
offset > dscp->dsc_last_data_offset));
/*
* If there is a pending op, but it's not PENDING_FREE, push it out,
* since free block aggregation can only be done for blocks of the
* same type (i.e., DRR_FREE records can only be aggregated with
* other DRR_FREE records. DRR_FREEOBJECTS records can only be
* aggregated with other DRR_FREEOBJECTS records).
*/
if (dscp->dsc_pending_op != PENDING_NONE &&
dscp->dsc_pending_op != PENDING_FREE) {
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
if (dscp->dsc_pending_op == PENDING_FREE) {
/*
* Check to see whether this free block can be aggregated
* with pending one.
*/
if (drrf->drr_object == object && drrf->drr_offset +
drrf->drr_length == offset) {
if (offset + length < offset || length == UINT64_MAX)
drrf->drr_length = UINT64_MAX;
else
drrf->drr_length += length;
return (0);
} else {
/* not a continuation. Push out pending record */
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
}
/* create a FREE record and make it pending */
bzero(dscp->dsc_drr, sizeof (dmu_replay_record_t));
dscp->dsc_drr->drr_type = DRR_FREE;
drrf->drr_object = object;
drrf->drr_offset = offset;
if (offset + length < offset)
drrf->drr_length = DMU_OBJECT_END;
else
drrf->drr_length = length;
drrf->drr_toguid = dscp->dsc_toguid;
if (length == DMU_OBJECT_END) {
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
} else {
dscp->dsc_pending_op = PENDING_FREE;
}
return (0);
}
/*
* Fill in the drr_redact struct, or perform aggregation if the previous record
* is also a redaction record, and the two are adjacent.
*/
static int
dump_redact(dmu_send_cookie_t *dscp, uint64_t object, uint64_t offset,
uint64_t length)
{
struct drr_redact *drrr = &dscp->dsc_drr->drr_u.drr_redact;
/*
* If there is a pending op, but it's not PENDING_REDACT, push it out,
* since free block aggregation can only be done for blocks of the
* same type (i.e., DRR_REDACT records can only be aggregated with
* other DRR_REDACT records).
*/
if (dscp->dsc_pending_op != PENDING_NONE &&
dscp->dsc_pending_op != PENDING_REDACT) {
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
if (dscp->dsc_pending_op == PENDING_REDACT) {
/*
* Check to see whether this redacted block can be aggregated
* with pending one.
*/
if (drrr->drr_object == object && drrr->drr_offset +
drrr->drr_length == offset) {
drrr->drr_length += length;
return (0);
} else {
/* not a continuation. Push out pending record */
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
}
/* create a REDACT record and make it pending */
bzero(dscp->dsc_drr, sizeof (dmu_replay_record_t));
dscp->dsc_drr->drr_type = DRR_REDACT;
drrr->drr_object = object;
drrr->drr_offset = offset;
drrr->drr_length = length;
drrr->drr_toguid = dscp->dsc_toguid;
dscp->dsc_pending_op = PENDING_REDACT;
return (0);
}
static int
dmu_dump_write(dmu_send_cookie_t *dscp, dmu_object_type_t type, uint64_t object,
uint64_t offset, int lsize, int psize, const blkptr_t *bp,
boolean_t io_compressed, void *data)
{
uint64_t payload_size;
boolean_t raw = (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW);
struct drr_write *drrw = &(dscp->dsc_drr->drr_u.drr_write);
/*
* We send data in increasing object, offset order.
* See comment in dump_free() for details.
*/
ASSERT(object > dscp->dsc_last_data_object ||
(object == dscp->dsc_last_data_object &&
offset > dscp->dsc_last_data_offset));
dscp->dsc_last_data_object = object;
dscp->dsc_last_data_offset = offset + lsize - 1;
/*
* If there is any kind of pending aggregation (currently either
* a grouping of free objects or free blocks), push it out to
* the stream, since aggregation can't be done across operations
* of different types.
*/
if (dscp->dsc_pending_op != PENDING_NONE) {
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
/* write a WRITE record */
bzero(dscp->dsc_drr, sizeof (dmu_replay_record_t));
dscp->dsc_drr->drr_type = DRR_WRITE;
drrw->drr_object = object;
drrw->drr_type = type;
drrw->drr_offset = offset;
drrw->drr_toguid = dscp->dsc_toguid;
drrw->drr_logical_size = lsize;
/* only set the compression fields if the buf is compressed or raw */
boolean_t compressed =
(bp != NULL ? BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF &&
io_compressed : lsize != psize);
if (raw || compressed) {
ASSERT(raw || dscp->dsc_featureflags &
DMU_BACKUP_FEATURE_COMPRESSED);
ASSERT(!BP_IS_EMBEDDED(bp));
ASSERT3S(psize, >, 0);
if (raw) {
ASSERT(BP_IS_PROTECTED(bp));
/*
* This is a raw protected block so we need to pass
* along everything the receiving side will need to
* interpret this block, including the byteswap, salt,
* IV, and MAC.
*/
if (BP_SHOULD_BYTESWAP(bp))
drrw->drr_flags |= DRR_RAW_BYTESWAP;
zio_crypt_decode_params_bp(bp, drrw->drr_salt,
drrw->drr_iv);
zio_crypt_decode_mac_bp(bp, drrw->drr_mac);
} else {
/* this is a compressed block */
ASSERT(dscp->dsc_featureflags &
DMU_BACKUP_FEATURE_COMPRESSED);
ASSERT(!BP_SHOULD_BYTESWAP(bp));
ASSERT(!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)));
ASSERT3U(BP_GET_COMPRESS(bp), !=, ZIO_COMPRESS_OFF);
ASSERT3S(lsize, >=, psize);
}
/* set fields common to compressed and raw sends */
drrw->drr_compressiontype = BP_GET_COMPRESS(bp);
drrw->drr_compressed_size = psize;
payload_size = drrw->drr_compressed_size;
} else {
payload_size = drrw->drr_logical_size;
}
if (bp == NULL || BP_IS_EMBEDDED(bp) || (BP_IS_PROTECTED(bp) && !raw)) {
/*
* There's no pre-computed checksum for partial-block writes,
* embedded BP's, or encrypted BP's that are being sent as
* plaintext, so (like fletcher4-checksummed blocks) userland
* will have to compute a dedup-capable checksum itself.
*/
drrw->drr_checksumtype = ZIO_CHECKSUM_OFF;
} else {
drrw->drr_checksumtype = BP_GET_CHECKSUM(bp);
if (zio_checksum_table[drrw->drr_checksumtype].ci_flags &
ZCHECKSUM_FLAG_DEDUP)
drrw->drr_flags |= DRR_CHECKSUM_DEDUP;
DDK_SET_LSIZE(&drrw->drr_key, BP_GET_LSIZE(bp));
DDK_SET_PSIZE(&drrw->drr_key, BP_GET_PSIZE(bp));
DDK_SET_COMPRESS(&drrw->drr_key, BP_GET_COMPRESS(bp));
DDK_SET_CRYPT(&drrw->drr_key, BP_IS_PROTECTED(bp));
drrw->drr_key.ddk_cksum = bp->blk_cksum;
}
if (dump_record(dscp, data, payload_size) != 0)
return (SET_ERROR(EINTR));
return (0);
}
static int
dump_write_embedded(dmu_send_cookie_t *dscp, uint64_t object, uint64_t offset,
int blksz, const blkptr_t *bp)
{
char buf[BPE_PAYLOAD_SIZE];
struct drr_write_embedded *drrw =
&(dscp->dsc_drr->drr_u.drr_write_embedded);
if (dscp->dsc_pending_op != PENDING_NONE) {
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
ASSERT(BP_IS_EMBEDDED(bp));
bzero(dscp->dsc_drr, sizeof (dmu_replay_record_t));
dscp->dsc_drr->drr_type = DRR_WRITE_EMBEDDED;
drrw->drr_object = object;
drrw->drr_offset = offset;
drrw->drr_length = blksz;
drrw->drr_toguid = dscp->dsc_toguid;
drrw->drr_compression = BP_GET_COMPRESS(bp);
drrw->drr_etype = BPE_GET_ETYPE(bp);
drrw->drr_lsize = BPE_GET_LSIZE(bp);
drrw->drr_psize = BPE_GET_PSIZE(bp);
decode_embedded_bp_compressed(bp, buf);
if (dump_record(dscp, buf, P2ROUNDUP(drrw->drr_psize, 8)) != 0)
return (SET_ERROR(EINTR));
return (0);
}
static int
dump_spill(dmu_send_cookie_t *dscp, const blkptr_t *bp, uint64_t object,
void *data)
{
struct drr_spill *drrs = &(dscp->dsc_drr->drr_u.drr_spill);
uint64_t blksz = BP_GET_LSIZE(bp);
uint64_t payload_size = blksz;
if (dscp->dsc_pending_op != PENDING_NONE) {
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
/* write a SPILL record */
bzero(dscp->dsc_drr, sizeof (dmu_replay_record_t));
dscp->dsc_drr->drr_type = DRR_SPILL;
drrs->drr_object = object;
drrs->drr_length = blksz;
drrs->drr_toguid = dscp->dsc_toguid;
/* See comment in dump_dnode() for full details */
if (zfs_send_unmodified_spill_blocks &&
(bp->blk_birth <= dscp->dsc_fromtxg)) {
drrs->drr_flags |= DRR_SPILL_UNMODIFIED;
}
/* handle raw send fields */
if (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW) {
ASSERT(BP_IS_PROTECTED(bp));
if (BP_SHOULD_BYTESWAP(bp))
drrs->drr_flags |= DRR_RAW_BYTESWAP;
drrs->drr_compressiontype = BP_GET_COMPRESS(bp);
drrs->drr_compressed_size = BP_GET_PSIZE(bp);
zio_crypt_decode_params_bp(bp, drrs->drr_salt, drrs->drr_iv);
zio_crypt_decode_mac_bp(bp, drrs->drr_mac);
payload_size = drrs->drr_compressed_size;
}
if (dump_record(dscp, data, payload_size) != 0)
return (SET_ERROR(EINTR));
return (0);
}
static int
dump_freeobjects(dmu_send_cookie_t *dscp, uint64_t firstobj, uint64_t numobjs)
{
struct drr_freeobjects *drrfo = &(dscp->dsc_drr->drr_u.drr_freeobjects);
uint64_t maxobj = DNODES_PER_BLOCK *
(DMU_META_DNODE(dscp->dsc_os)->dn_maxblkid + 1);
/*
* ZoL < 0.7 does not handle large FREEOBJECTS records correctly,
* leading to zfs recv never completing. to avoid this issue, don't
* send FREEOBJECTS records for object IDs which cannot exist on the
* receiving side.
*/
if (maxobj > 0) {
if (maxobj <= firstobj)
return (0);
if (maxobj < firstobj + numobjs)
numobjs = maxobj - firstobj;
}
/*
* If there is a pending op, but it's not PENDING_FREEOBJECTS,
* push it out, since free block aggregation can only be done for
* blocks of the same type (i.e., DRR_FREE records can only be
* aggregated with other DRR_FREE records. DRR_FREEOBJECTS records
* can only be aggregated with other DRR_FREEOBJECTS records).
*/
if (dscp->dsc_pending_op != PENDING_NONE &&
dscp->dsc_pending_op != PENDING_FREEOBJECTS) {
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
if (dscp->dsc_pending_op == PENDING_FREEOBJECTS) {
/*
* See whether this free object array can be aggregated
* with pending one
*/
if (drrfo->drr_firstobj + drrfo->drr_numobjs == firstobj) {
drrfo->drr_numobjs += numobjs;
return (0);
} else {
/* can't be aggregated. Push out pending record */
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
}
/* write a FREEOBJECTS record */
bzero(dscp->dsc_drr, sizeof (dmu_replay_record_t));
dscp->dsc_drr->drr_type = DRR_FREEOBJECTS;
drrfo->drr_firstobj = firstobj;
drrfo->drr_numobjs = numobjs;
drrfo->drr_toguid = dscp->dsc_toguid;
dscp->dsc_pending_op = PENDING_FREEOBJECTS;
return (0);
}
static int
dump_dnode(dmu_send_cookie_t *dscp, const blkptr_t *bp, uint64_t object,
dnode_phys_t *dnp)
{
struct drr_object *drro = &(dscp->dsc_drr->drr_u.drr_object);
int bonuslen;
if (object < dscp->dsc_resume_object) {
/*
* Note: when resuming, we will visit all the dnodes in
* the block of dnodes that we are resuming from. In
* this case it's unnecessary to send the dnodes prior to
* the one we are resuming from. We should be at most one
* block's worth of dnodes behind the resume point.
*/
ASSERT3U(dscp->dsc_resume_object - object, <,
1 << (DNODE_BLOCK_SHIFT - DNODE_SHIFT));
return (0);
}
if (dnp == NULL || dnp->dn_type == DMU_OT_NONE)
return (dump_freeobjects(dscp, object, 1));
if (dscp->dsc_pending_op != PENDING_NONE) {
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
/* write an OBJECT record */
bzero(dscp->dsc_drr, sizeof (dmu_replay_record_t));
dscp->dsc_drr->drr_type = DRR_OBJECT;
drro->drr_object = object;
drro->drr_type = dnp->dn_type;
drro->drr_bonustype = dnp->dn_bonustype;
drro->drr_blksz = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT;
drro->drr_bonuslen = dnp->dn_bonuslen;
drro->drr_dn_slots = dnp->dn_extra_slots + 1;
drro->drr_checksumtype = dnp->dn_checksum;
drro->drr_compress = dnp->dn_compress;
drro->drr_toguid = dscp->dsc_toguid;
if (!(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
drro->drr_blksz > SPA_OLD_MAXBLOCKSIZE)
drro->drr_blksz = SPA_OLD_MAXBLOCKSIZE;
bonuslen = P2ROUNDUP(dnp->dn_bonuslen, 8);
if ((dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW)) {
ASSERT(BP_IS_ENCRYPTED(bp));
if (BP_SHOULD_BYTESWAP(bp))
drro->drr_flags |= DRR_RAW_BYTESWAP;
/* needed for reconstructing dnp on recv side */
drro->drr_maxblkid = dnp->dn_maxblkid;
drro->drr_indblkshift = dnp->dn_indblkshift;
drro->drr_nlevels = dnp->dn_nlevels;
drro->drr_nblkptr = dnp->dn_nblkptr;
/*
* Since we encrypt the entire bonus area, the (raw) part
* beyond the bonuslen is actually nonzero, so we need
* to send it.
*/
if (bonuslen != 0) {
drro->drr_raw_bonuslen = DN_MAX_BONUS_LEN(dnp);
bonuslen = drro->drr_raw_bonuslen;
}
}
/*
* DRR_OBJECT_SPILL is set for every dnode which references a
* spill block. This allows the receiving pool to definitively
* determine when a spill block should be kept or freed.
*/
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR)
drro->drr_flags |= DRR_OBJECT_SPILL;
if (dump_record(dscp, DN_BONUS(dnp), bonuslen) != 0)
return (SET_ERROR(EINTR));
/* Free anything past the end of the file. */
if (dump_free(dscp, object, (dnp->dn_maxblkid + 1) *
(dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT), DMU_OBJECT_END) != 0)
return (SET_ERROR(EINTR));
/*
* Send DRR_SPILL records for unmodified spill blocks. This is useful
* because changing certain attributes of the object (e.g. blocksize)
* can cause old versions of ZFS to incorrectly remove a spill block.
* Including these records in the stream forces an up to date version
* to always be written ensuring they're never lost. Current versions
* of the code which understand the DRR_FLAG_SPILL_BLOCK feature can
* ignore these unmodified spill blocks.
*/
if (zfs_send_unmodified_spill_blocks &&
(dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) &&
(DN_SPILL_BLKPTR(dnp)->blk_birth <= dscp->dsc_fromtxg)) {
struct send_range record;
blkptr_t *bp = DN_SPILL_BLKPTR(dnp);
bzero(&record, sizeof (struct send_range));
record.type = DATA;
record.object = object;
record.eos_marker = B_FALSE;
record.start_blkid = DMU_SPILL_BLKID;
record.end_blkid = record.start_blkid + 1;
record.sru.data.bp = *bp;
record.sru.data.obj_type = dnp->dn_type;
record.sru.data.datablksz = BP_GET_LSIZE(bp);
if (do_dump(dscp, &record) != 0)
return (SET_ERROR(EINTR));
}
if (dscp->dsc_err != 0)
return (SET_ERROR(EINTR));
return (0);
}
static int
dump_object_range(dmu_send_cookie_t *dscp, const blkptr_t *bp,
uint64_t firstobj, uint64_t numslots)
{
struct drr_object_range *drror =
&(dscp->dsc_drr->drr_u.drr_object_range);
/* we only use this record type for raw sends */
ASSERT(BP_IS_PROTECTED(bp));
ASSERT(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW);
ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_DNODE);
ASSERT0(BP_GET_LEVEL(bp));
if (dscp->dsc_pending_op != PENDING_NONE) {
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
dscp->dsc_pending_op = PENDING_NONE;
}
bzero(dscp->dsc_drr, sizeof (dmu_replay_record_t));
dscp->dsc_drr->drr_type = DRR_OBJECT_RANGE;
drror->drr_firstobj = firstobj;
drror->drr_numslots = numslots;
drror->drr_toguid = dscp->dsc_toguid;
if (BP_SHOULD_BYTESWAP(bp))
drror->drr_flags |= DRR_RAW_BYTESWAP;
zio_crypt_decode_params_bp(bp, drror->drr_salt, drror->drr_iv);
zio_crypt_decode_mac_bp(bp, drror->drr_mac);
if (dump_record(dscp, NULL, 0) != 0)
return (SET_ERROR(EINTR));
return (0);
}
static boolean_t
send_do_embed(const blkptr_t *bp, uint64_t featureflags)
{
if (!BP_IS_EMBEDDED(bp))
return (B_FALSE);
/*
* Compression function must be legacy, or explicitly enabled.
*/
if ((BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_LEGACY_FUNCTIONS &&
!(featureflags & DMU_BACKUP_FEATURE_LZ4)))
return (B_FALSE);
/*
* If we have not set the ZSTD feature flag, we can't send ZSTD
* compressed embedded blocks, as the receiver may not support them.
*/
if ((BP_GET_COMPRESS(bp) == ZIO_COMPRESS_ZSTD &&
!(featureflags & DMU_BACKUP_FEATURE_ZSTD)))
return (B_FALSE);
/*
* Embed type must be explicitly enabled.
*/
switch (BPE_GET_ETYPE(bp)) {
case BP_EMBEDDED_TYPE_DATA:
if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)
return (B_TRUE);
break;
default:
return (B_FALSE);
}
return (B_FALSE);
}
/*
* This function actually handles figuring out what kind of record needs to be
* dumped, and calling the appropriate helper function. In most cases,
* the data has already been read by send_reader_thread().
*/
static int
do_dump(dmu_send_cookie_t *dscp, struct send_range *range)
{
int err = 0;
switch (range->type) {
case OBJECT:
err = dump_dnode(dscp, &range->sru.object.bp, range->object,
range->sru.object.dnp);
return (err);
case OBJECT_RANGE: {
ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
if (!(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW)) {
return (0);
}
uint64_t epb = BP_GET_LSIZE(&range->sru.object_range.bp) >>
DNODE_SHIFT;
uint64_t firstobj = range->start_blkid * epb;
err = dump_object_range(dscp, &range->sru.object_range.bp,
firstobj, epb);
break;
}
case REDACT: {
struct srr *srrp = &range->sru.redact;
err = dump_redact(dscp, range->object, range->start_blkid *
srrp->datablksz, (range->end_blkid - range->start_blkid) *
srrp->datablksz);
return (err);
}
case DATA: {
struct srd *srdp = &range->sru.data;
blkptr_t *bp = &srdp->bp;
spa_t *spa =
dmu_objset_spa(dscp->dsc_os);
ASSERT3U(srdp->datablksz, ==, BP_GET_LSIZE(bp));
ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
if (BP_GET_TYPE(bp) == DMU_OT_SA) {
arc_flags_t aflags = ARC_FLAG_WAIT;
enum zio_flag zioflags = ZIO_FLAG_CANFAIL;
if (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW) {
ASSERT(BP_IS_PROTECTED(bp));
zioflags |= ZIO_FLAG_RAW;
}
zbookmark_phys_t zb;
ASSERT3U(range->start_blkid, ==, DMU_SPILL_BLKID);
zb.zb_objset = dmu_objset_id(dscp->dsc_os);
zb.zb_object = range->object;
zb.zb_level = 0;
zb.zb_blkid = range->start_blkid;
arc_buf_t *abuf = NULL;
if (!dscp->dsc_dso->dso_dryrun && arc_read(NULL, spa,
bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ,
zioflags, &aflags, &zb) != 0)
return (SET_ERROR(EIO));
err = dump_spill(dscp, bp, zb.zb_object,
(abuf == NULL ? NULL : abuf->b_data));
if (abuf != NULL)
arc_buf_destroy(abuf, &abuf);
return (err);
}
if (send_do_embed(bp, dscp->dsc_featureflags)) {
err = dump_write_embedded(dscp, range->object,
range->start_blkid * srdp->datablksz,
srdp->datablksz, bp);
return (err);
}
ASSERT(range->object > dscp->dsc_resume_object ||
(range->object == dscp->dsc_resume_object &&
range->start_blkid * srdp->datablksz >=
dscp->dsc_resume_offset));
/* it's a level-0 block of a regular object */
mutex_enter(&srdp->lock);
while (srdp->io_outstanding)
cv_wait(&srdp->cv, &srdp->lock);
err = srdp->io_err;
mutex_exit(&srdp->lock);
if (err != 0) {
if (zfs_send_corrupt_data &&
!dscp->dsc_dso->dso_dryrun) {
/*
* Send a block filled with 0x"zfs badd bloc"
*/
srdp->abuf = arc_alloc_buf(spa, &srdp->abuf,
ARC_BUFC_DATA, srdp->datablksz);
uint64_t *ptr;
for (ptr = srdp->abuf->b_data;
(char *)ptr < (char *)srdp->abuf->b_data +
srdp->datablksz; ptr++)
*ptr = 0x2f5baddb10cULL;
} else {
return (SET_ERROR(EIO));
}
}
ASSERT(dscp->dsc_dso->dso_dryrun ||
srdp->abuf != NULL || srdp->abd != NULL);
uint64_t offset = range->start_blkid * srdp->datablksz;
char *data = NULL;
if (srdp->abd != NULL) {
data = abd_to_buf(srdp->abd);
ASSERT3P(srdp->abuf, ==, NULL);
} else if (srdp->abuf != NULL) {
data = srdp->abuf->b_data;
}
/*
* If we have large blocks stored on disk but the send flags
* don't allow us to send large blocks, we split the data from
* the arc buf into chunks.
*/
if (srdp->datablksz > SPA_OLD_MAXBLOCKSIZE &&
!(dscp->dsc_featureflags &
DMU_BACKUP_FEATURE_LARGE_BLOCKS)) {
while (srdp->datablksz > 0 && err == 0) {
int n = MIN(srdp->datablksz,
SPA_OLD_MAXBLOCKSIZE);
err = dmu_dump_write(dscp, srdp->obj_type,
range->object, offset, n, n, NULL, B_FALSE,
data);
offset += n;
/*
* When doing dry run, data==NULL is used as a
* sentinel value by
* dmu_dump_write()->dump_record().
*/
if (data != NULL)
data += n;
srdp->datablksz -= n;
}
} else {
err = dmu_dump_write(dscp, srdp->obj_type,
range->object, offset,
srdp->datablksz, srdp->datasz, bp,
srdp->io_compressed, data);
}
return (err);
}
case HOLE: {
struct srh *srhp = &range->sru.hole;
if (range->object == DMU_META_DNODE_OBJECT) {
uint32_t span = srhp->datablksz >> DNODE_SHIFT;
uint64_t first_obj = range->start_blkid * span;
uint64_t numobj = range->end_blkid * span - first_obj;
return (dump_freeobjects(dscp, first_obj, numobj));
}
uint64_t offset = 0;
/*
* If this multiply overflows, we don't need to send this block.
* Even if it has a birth time, it can never not be a hole, so
* we don't need to send records for it.
*/
if (!overflow_multiply(range->start_blkid, srhp->datablksz,
&offset)) {
return (0);
}
uint64_t len = 0;
if (!overflow_multiply(range->end_blkid, srhp->datablksz, &len))
len = UINT64_MAX;
len = len - offset;
return (dump_free(dscp, range->object, offset, len));
}
default:
panic("Invalid range type in do_dump: %d", range->type);
}
return (err);
}
static struct send_range *
range_alloc(enum type type, uint64_t object, uint64_t start_blkid,
uint64_t end_blkid, boolean_t eos)
{
struct send_range *range = kmem_alloc(sizeof (*range), KM_SLEEP);
range->type = type;
range->object = object;
range->start_blkid = start_blkid;
range->end_blkid = end_blkid;
range->eos_marker = eos;
if (type == DATA) {
range->sru.data.abd = NULL;
range->sru.data.abuf = NULL;
mutex_init(&range->sru.data.lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&range->sru.data.cv, NULL, CV_DEFAULT, NULL);
range->sru.data.io_outstanding = 0;
range->sru.data.io_err = 0;
range->sru.data.io_compressed = B_FALSE;
}
return (range);
}
/*
* This is the callback function to traverse_dataset that acts as a worker
* thread for dmu_send_impl.
*/
static int
send_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
const zbookmark_phys_t *zb, const struct dnode_phys *dnp, void *arg)
{
(void) zilog;
struct send_thread_arg *sta = arg;
struct send_range *record;
ASSERT(zb->zb_object == DMU_META_DNODE_OBJECT ||
zb->zb_object >= sta->resume.zb_object);
/*
* All bps of an encrypted os should have the encryption bit set.
* If this is not true it indicates tampering and we report an error.
*/
if (sta->os->os_encrypted &&
!BP_IS_HOLE(bp) && !BP_USES_CRYPT(bp)) {
spa_log_error(spa, zb);
zfs_panic_recover("unencrypted block in encrypted "
"object set %llu", dmu_objset_id(sta->os));
return (SET_ERROR(EIO));
}
if (sta->cancel)
return (SET_ERROR(EINTR));
if (zb->zb_object != DMU_META_DNODE_OBJECT &&
DMU_OBJECT_IS_SPECIAL(zb->zb_object))
return (0);
atomic_inc_64(sta->num_blocks_visited);
if (zb->zb_level == ZB_DNODE_LEVEL) {
if (zb->zb_object == DMU_META_DNODE_OBJECT)
return (0);
record = range_alloc(OBJECT, zb->zb_object, 0, 0, B_FALSE);
record->sru.object.bp = *bp;
size_t size = sizeof (*dnp) * (dnp->dn_extra_slots + 1);
record->sru.object.dnp = kmem_alloc(size, KM_SLEEP);
bcopy(dnp, record->sru.object.dnp, size);
bqueue_enqueue(&sta->q, record, sizeof (*record));
return (0);
}
if (zb->zb_level == 0 && zb->zb_object == DMU_META_DNODE_OBJECT &&
!BP_IS_HOLE(bp)) {
record = range_alloc(OBJECT_RANGE, 0, zb->zb_blkid,
zb->zb_blkid + 1, B_FALSE);
record->sru.object_range.bp = *bp;
bqueue_enqueue(&sta->q, record, sizeof (*record));
return (0);
}
if (zb->zb_level < 0 || (zb->zb_level > 0 && !BP_IS_HOLE(bp)))
return (0);
if (zb->zb_object == DMU_META_DNODE_OBJECT && !BP_IS_HOLE(bp))
return (0);
uint64_t span = bp_span_in_blocks(dnp->dn_indblkshift, zb->zb_level);
uint64_t start;
/*
* If this multiply overflows, we don't need to send this block.
* Even if it has a birth time, it can never not be a hole, so
* we don't need to send records for it.
*/
if (!overflow_multiply(span, zb->zb_blkid, &start) || (!(zb->zb_blkid ==
DMU_SPILL_BLKID || DMU_OT_IS_METADATA(dnp->dn_type)) &&
span * zb->zb_blkid > dnp->dn_maxblkid)) {
ASSERT(BP_IS_HOLE(bp));
return (0);
}
if (zb->zb_blkid == DMU_SPILL_BLKID)
ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_SA);
enum type record_type = DATA;
if (BP_IS_HOLE(bp))
record_type = HOLE;
else if (BP_IS_REDACTED(bp))
record_type = REDACT;
else
record_type = DATA;
record = range_alloc(record_type, zb->zb_object, start,
(start + span < start ? 0 : start + span), B_FALSE);
uint64_t datablksz = (zb->zb_blkid == DMU_SPILL_BLKID ?
BP_GET_LSIZE(bp) : dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);
if (BP_IS_HOLE(bp)) {
record->sru.hole.datablksz = datablksz;
} else if (BP_IS_REDACTED(bp)) {
record->sru.redact.datablksz = datablksz;
} else {
record->sru.data.datablksz = datablksz;
record->sru.data.obj_type = dnp->dn_type;
record->sru.data.bp = *bp;
}
bqueue_enqueue(&sta->q, record, sizeof (*record));
return (0);
}
struct redact_list_cb_arg {
uint64_t *num_blocks_visited;
bqueue_t *q;
boolean_t *cancel;
boolean_t mark_redact;
};
static int
redact_list_cb(redact_block_phys_t *rb, void *arg)
{
struct redact_list_cb_arg *rlcap = arg;
atomic_inc_64(rlcap->num_blocks_visited);
if (*rlcap->cancel)
return (-1);
struct send_range *data = range_alloc(REDACT, rb->rbp_object,
rb->rbp_blkid, rb->rbp_blkid + redact_block_get_count(rb), B_FALSE);
ASSERT3U(data->end_blkid, >, rb->rbp_blkid);
if (rlcap->mark_redact) {
data->type = REDACT;
data->sru.redact.datablksz = redact_block_get_size(rb);
} else {
data->type = PREVIOUSLY_REDACTED;
}
bqueue_enqueue(rlcap->q, data, sizeof (*data));
return (0);
}
/*
* This function kicks off the traverse_dataset. It also handles setting the
* error code of the thread in case something goes wrong, and pushes the End of
* Stream record when the traverse_dataset call has finished.
*/
static void
send_traverse_thread(void *arg)
{
struct send_thread_arg *st_arg = arg;
int err = 0;
struct send_range *data;
fstrans_cookie_t cookie = spl_fstrans_mark();
err = traverse_dataset_resume(st_arg->os->os_dsl_dataset,
st_arg->fromtxg, &st_arg->resume,
st_arg->flags, send_cb, st_arg);
if (err != EINTR)
st_arg->error_code = err;
data = range_alloc(DATA, 0, 0, 0, B_TRUE);
bqueue_enqueue_flush(&st_arg->q, data, sizeof (*data));
spl_fstrans_unmark(cookie);
thread_exit();
}
/*
* Utility function that causes End of Stream records to compare after of all
* others, so that other threads' comparison logic can stay simple.
*/
static int __attribute__((unused))
send_range_after(const struct send_range *from, const struct send_range *to)
{
if (from->eos_marker == B_TRUE)
return (1);
if (to->eos_marker == B_TRUE)
return (-1);
uint64_t from_obj = from->object;
uint64_t from_end_obj = from->object + 1;
uint64_t to_obj = to->object;
uint64_t to_end_obj = to->object + 1;
if (from_obj == 0) {
ASSERT(from->type == HOLE || from->type == OBJECT_RANGE);
from_obj = from->start_blkid << DNODES_PER_BLOCK_SHIFT;
from_end_obj = from->end_blkid << DNODES_PER_BLOCK_SHIFT;
}
if (to_obj == 0) {
ASSERT(to->type == HOLE || to->type == OBJECT_RANGE);
to_obj = to->start_blkid << DNODES_PER_BLOCK_SHIFT;
to_end_obj = to->end_blkid << DNODES_PER_BLOCK_SHIFT;
}
if (from_end_obj <= to_obj)
return (-1);
if (from_obj >= to_end_obj)
return (1);
int64_t cmp = TREE_CMP(to->type == OBJECT_RANGE, from->type ==
OBJECT_RANGE);
if (unlikely(cmp))
return (cmp);
cmp = TREE_CMP(to->type == OBJECT, from->type == OBJECT);
if (unlikely(cmp))
return (cmp);
if (from->end_blkid <= to->start_blkid)
return (-1);
if (from->start_blkid >= to->end_blkid)
return (1);
return (0);
}
/*
* Pop the new data off the queue, check that the records we receive are in
* the right order, but do not free the old data. This is used so that the
* records can be sent on to the main thread without copying the data.
*/
static struct send_range *
get_next_range_nofree(bqueue_t *bq, struct send_range *prev)
{
struct send_range *next = bqueue_dequeue(bq);
ASSERT3S(send_range_after(prev, next), ==, -1);
return (next);
}
/*
* Pop the new data off the queue, check that the records we receive are in
* the right order, and free the old data.
*/
static struct send_range *
get_next_range(bqueue_t *bq, struct send_range *prev)
{
struct send_range *next = get_next_range_nofree(bq, prev);
range_free(prev);
return (next);
}
static void
redact_list_thread(void *arg)
{
struct redact_list_thread_arg *rlt_arg = arg;
struct send_range *record;
fstrans_cookie_t cookie = spl_fstrans_mark();
if (rlt_arg->rl != NULL) {
struct redact_list_cb_arg rlcba = {0};
rlcba.cancel = &rlt_arg->cancel;
rlcba.q = &rlt_arg->q;
rlcba.num_blocks_visited = rlt_arg->num_blocks_visited;
rlcba.mark_redact = rlt_arg->mark_redact;
int err = dsl_redaction_list_traverse(rlt_arg->rl,
&rlt_arg->resume, redact_list_cb, &rlcba);
if (err != EINTR)
rlt_arg->error_code = err;
}
record = range_alloc(DATA, 0, 0, 0, B_TRUE);
bqueue_enqueue_flush(&rlt_arg->q, record, sizeof (*record));
spl_fstrans_unmark(cookie);
thread_exit();
}
/*
* Compare the start point of the two provided ranges. End of stream ranges
* compare last, objects compare before any data or hole inside that object and
* multi-object holes that start at the same object.
*/
static int
send_range_start_compare(struct send_range *r1, struct send_range *r2)
{
uint64_t r1_objequiv = r1->object;
uint64_t r1_l0equiv = r1->start_blkid;
uint64_t r2_objequiv = r2->object;
uint64_t r2_l0equiv = r2->start_blkid;
int64_t cmp = TREE_CMP(r1->eos_marker, r2->eos_marker);
if (unlikely(cmp))
return (cmp);
if (r1->object == 0) {
r1_objequiv = r1->start_blkid * DNODES_PER_BLOCK;
r1_l0equiv = 0;
}
if (r2->object == 0) {
r2_objequiv = r2->start_blkid * DNODES_PER_BLOCK;
r2_l0equiv = 0;
}
cmp = TREE_CMP(r1_objequiv, r2_objequiv);
if (likely(cmp))
return (cmp);
cmp = TREE_CMP(r2->type == OBJECT_RANGE, r1->type == OBJECT_RANGE);
if (unlikely(cmp))
return (cmp);
cmp = TREE_CMP(r2->type == OBJECT, r1->type == OBJECT);
if (unlikely(cmp))
return (cmp);
return (TREE_CMP(r1_l0equiv, r2_l0equiv));
}
enum q_idx {
REDACT_IDX = 0,
TO_IDX,
FROM_IDX,
NUM_THREADS
};
/*
* This function returns the next range the send_merge_thread should operate on.
* The inputs are two arrays; the first one stores the range at the front of the
* queues stored in the second one. The ranges are sorted in descending
* priority order; the metadata from earlier ranges overrules metadata from
* later ranges. out_mask is used to return which threads the ranges came from;
* bit i is set if ranges[i] started at the same place as the returned range.
*
* This code is not hardcoded to compare a specific number of threads; it could
* be used with any number, just by changing the q_idx enum.
*
* The "next range" is the one with the earliest start; if two starts are equal,
* the highest-priority range is the next to operate on. If a higher-priority
* range starts in the middle of the first range, then the first range will be
* truncated to end where the higher-priority range starts, and we will operate
* on that one next time. In this way, we make sure that each block covered by
* some range gets covered by a returned range, and each block covered is
* returned using the metadata of the highest-priority range it appears in.
*
* For example, if the three ranges at the front of the queues were [2,4),
* [3,5), and [1,3), then the ranges returned would be [1,2) with the metadata
* from the third range, [2,4) with the metadata from the first range, and then
* [4,5) with the metadata from the second.
*/
static struct send_range *
find_next_range(struct send_range **ranges, bqueue_t **qs, uint64_t *out_mask)
{
int idx = 0; // index of the range with the earliest start
int i;
uint64_t bmask = 0;
for (i = 1; i < NUM_THREADS; i++) {
if (send_range_start_compare(ranges[i], ranges[idx]) < 0)
idx = i;
}
if (ranges[idx]->eos_marker) {
struct send_range *ret = range_alloc(DATA, 0, 0, 0, B_TRUE);
*out_mask = 0;
return (ret);
}
/*
* Find all the ranges that start at that same point.
*/
for (i = 0; i < NUM_THREADS; i++) {
if (send_range_start_compare(ranges[i], ranges[idx]) == 0)
bmask |= 1 << i;
}
*out_mask = bmask;
/*
* OBJECT_RANGE records only come from the TO thread, and should always
* be treated as overlapping with nothing and sent on immediately. They
* are only used in raw sends, and are never redacted.
*/
if (ranges[idx]->type == OBJECT_RANGE) {
ASSERT3U(idx, ==, TO_IDX);
ASSERT3U(*out_mask, ==, 1 << TO_IDX);
struct send_range *ret = ranges[idx];
ranges[idx] = get_next_range_nofree(qs[idx], ranges[idx]);
return (ret);
}
/*
* Find the first start or end point after the start of the first range.
*/
uint64_t first_change = ranges[idx]->end_blkid;
for (i = 0; i < NUM_THREADS; i++) {
if (i == idx || ranges[i]->eos_marker ||
ranges[i]->object > ranges[idx]->object ||
ranges[i]->object == DMU_META_DNODE_OBJECT)
continue;
ASSERT3U(ranges[i]->object, ==, ranges[idx]->object);
if (first_change > ranges[i]->start_blkid &&
(bmask & (1 << i)) == 0)
first_change = ranges[i]->start_blkid;
else if (first_change > ranges[i]->end_blkid)
first_change = ranges[i]->end_blkid;
}
/*
* Update all ranges to no longer overlap with the range we're
* returning. All such ranges must start at the same place as the range
* being returned, and end at or after first_change. Thus we update
* their start to first_change. If that makes them size 0, then free
* them and pull a new range from that thread.
*/
for (i = 0; i < NUM_THREADS; i++) {
if (i == idx || (bmask & (1 << i)) == 0)
continue;
ASSERT3U(first_change, >, ranges[i]->start_blkid);
ranges[i]->start_blkid = first_change;
ASSERT3U(ranges[i]->start_blkid, <=, ranges[i]->end_blkid);
if (ranges[i]->start_blkid == ranges[i]->end_blkid)
ranges[i] = get_next_range(qs[i], ranges[i]);
}
/*
* Short-circuit the simple case; if the range doesn't overlap with
* anything else, or it only overlaps with things that start at the same
* place and are longer, send it on.
*/
if (first_change == ranges[idx]->end_blkid) {
struct send_range *ret = ranges[idx];
ranges[idx] = get_next_range_nofree(qs[idx], ranges[idx]);
return (ret);
}
/*
* Otherwise, return a truncated copy of ranges[idx] and move the start
* of ranges[idx] back to first_change.
*/
struct send_range *ret = kmem_alloc(sizeof (*ret), KM_SLEEP);
*ret = *ranges[idx];
ret->end_blkid = first_change;
ranges[idx]->start_blkid = first_change;
return (ret);
}
#define FROM_AND_REDACT_BITS ((1 << REDACT_IDX) | (1 << FROM_IDX))
/*
* Merge the results from the from thread and the to thread, and then hand the
* records off to send_prefetch_thread to prefetch them. If this is not a
* send from a redaction bookmark, the from thread will push an end of stream
* record and stop, and we'll just send everything that was changed in the
* to_ds since the ancestor's creation txg. If it is, then since
* traverse_dataset has a canonical order, we can compare each change as
* they're pulled off the queues. That will give us a stream that is
* appropriately sorted, and covers all records. In addition, we pull the
* data from the redact_list_thread and use that to determine which blocks
* should be redacted.
*/
static void
send_merge_thread(void *arg)
{
struct send_merge_thread_arg *smt_arg = arg;
struct send_range *front_ranges[NUM_THREADS];
bqueue_t *queues[NUM_THREADS];
int err = 0;
fstrans_cookie_t cookie = spl_fstrans_mark();
if (smt_arg->redact_arg == NULL) {
front_ranges[REDACT_IDX] =
kmem_zalloc(sizeof (struct send_range), KM_SLEEP);
front_ranges[REDACT_IDX]->eos_marker = B_TRUE;
front_ranges[REDACT_IDX]->type = REDACT;
queues[REDACT_IDX] = NULL;
} else {
front_ranges[REDACT_IDX] =
bqueue_dequeue(&smt_arg->redact_arg->q);
queues[REDACT_IDX] = &smt_arg->redact_arg->q;
}
front_ranges[TO_IDX] = bqueue_dequeue(&smt_arg->to_arg->q);
queues[TO_IDX] = &smt_arg->to_arg->q;
front_ranges[FROM_IDX] = bqueue_dequeue(&smt_arg->from_arg->q);
queues[FROM_IDX] = &smt_arg->from_arg->q;
uint64_t mask = 0;
struct send_range *range;
for (range = find_next_range(front_ranges, queues, &mask);
!range->eos_marker && err == 0 && !smt_arg->cancel;
range = find_next_range(front_ranges, queues, &mask)) {
/*
* If the range in question was in both the from redact bookmark
* and the bookmark we're using to redact, then don't send it.
* It's already redacted on the receiving system, so a redaction
* record would be redundant.
*/
if ((mask & FROM_AND_REDACT_BITS) == FROM_AND_REDACT_BITS) {
ASSERT3U(range->type, ==, REDACT);
range_free(range);
continue;
}
bqueue_enqueue(&smt_arg->q, range, sizeof (*range));
if (smt_arg->to_arg->error_code != 0) {
err = smt_arg->to_arg->error_code;
} else if (smt_arg->from_arg->error_code != 0) {
err = smt_arg->from_arg->error_code;
} else if (smt_arg->redact_arg != NULL &&
smt_arg->redact_arg->error_code != 0) {
err = smt_arg->redact_arg->error_code;
}
}
if (smt_arg->cancel && err == 0)
err = SET_ERROR(EINTR);
smt_arg->error = err;
if (smt_arg->error != 0) {
smt_arg->to_arg->cancel = B_TRUE;
smt_arg->from_arg->cancel = B_TRUE;
if (smt_arg->redact_arg != NULL)
smt_arg->redact_arg->cancel = B_TRUE;
}
for (int i = 0; i < NUM_THREADS; i++) {
while (!front_ranges[i]->eos_marker) {
front_ranges[i] = get_next_range(queues[i],
front_ranges[i]);
}
range_free(front_ranges[i]);
}
if (range == NULL)
range = kmem_zalloc(sizeof (*range), KM_SLEEP);
range->eos_marker = B_TRUE;
bqueue_enqueue_flush(&smt_arg->q, range, 1);
spl_fstrans_unmark(cookie);
thread_exit();
}
struct send_reader_thread_arg {
struct send_merge_thread_arg *smta;
bqueue_t q;
boolean_t cancel;
boolean_t issue_reads;
uint64_t featureflags;
int error;
};
static void
dmu_send_read_done(zio_t *zio)
{
struct send_range *range = zio->io_private;
mutex_enter(&range->sru.data.lock);
if (zio->io_error != 0) {
abd_free(range->sru.data.abd);
range->sru.data.abd = NULL;
range->sru.data.io_err = zio->io_error;
}
ASSERT(range->sru.data.io_outstanding);
range->sru.data.io_outstanding = B_FALSE;
cv_broadcast(&range->sru.data.cv);
mutex_exit(&range->sru.data.lock);
}
static void
issue_data_read(struct send_reader_thread_arg *srta, struct send_range *range)
{
struct srd *srdp = &range->sru.data;
blkptr_t *bp = &srdp->bp;
objset_t *os = srta->smta->os;
ASSERT3U(range->type, ==, DATA);
ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
/*
* If we have large blocks stored on disk but
* the send flags don't allow us to send large
* blocks, we split the data from the arc buf
* into chunks.
*/
boolean_t split_large_blocks =
srdp->datablksz > SPA_OLD_MAXBLOCKSIZE &&
!(srta->featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS);
/*
* We should only request compressed data from the ARC if all
* the following are true:
* - stream compression was requested
* - we aren't splitting large blocks into smaller chunks
* - the data won't need to be byteswapped before sending
* - this isn't an embedded block
* - this isn't metadata (if receiving on a different endian
* system it can be byteswapped more easily)
*/
boolean_t request_compressed =
(srta->featureflags & DMU_BACKUP_FEATURE_COMPRESSED) &&
!split_large_blocks && !BP_SHOULD_BYTESWAP(bp) &&
!BP_IS_EMBEDDED(bp) && !DMU_OT_IS_METADATA(BP_GET_TYPE(bp));
enum zio_flag zioflags = ZIO_FLAG_CANFAIL;
if (srta->featureflags & DMU_BACKUP_FEATURE_RAW) {
zioflags |= ZIO_FLAG_RAW;
srdp->io_compressed = B_TRUE;
} else if (request_compressed) {
zioflags |= ZIO_FLAG_RAW_COMPRESS;
srdp->io_compressed = B_TRUE;
}
srdp->datasz = (zioflags & ZIO_FLAG_RAW_COMPRESS) ?
BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp);
if (!srta->issue_reads)
return;
if (BP_IS_REDACTED(bp))
return;
if (send_do_embed(bp, srta->featureflags))
return;
zbookmark_phys_t zb = {
.zb_objset = dmu_objset_id(os),
.zb_object = range->object,
.zb_level = 0,
.zb_blkid = range->start_blkid,
};
arc_flags_t aflags = ARC_FLAG_CACHED_ONLY;
int arc_err = arc_read(NULL, os->os_spa, bp,
arc_getbuf_func, &srdp->abuf, ZIO_PRIORITY_ASYNC_READ,
zioflags, &aflags, &zb);
/*
* If the data is not already cached in the ARC, we read directly
* from zio. This avoids the performance overhead of adding a new
* entry to the ARC, and we also avoid polluting the ARC cache with
* data that is not likely to be used in the future.
*/
if (arc_err != 0) {
srdp->abd = abd_alloc_linear(srdp->datasz, B_FALSE);
srdp->io_outstanding = B_TRUE;
zio_nowait(zio_read(NULL, os->os_spa, bp, srdp->abd,
srdp->datasz, dmu_send_read_done, range,
ZIO_PRIORITY_ASYNC_READ, zioflags, &zb));
}
}
/*
* Create a new record with the given values.
*/
static void
enqueue_range(struct send_reader_thread_arg *srta, bqueue_t *q, dnode_t *dn,
uint64_t blkid, uint64_t count, const blkptr_t *bp, uint32_t datablksz)
{
enum type range_type = (bp == NULL || BP_IS_HOLE(bp) ? HOLE :
(BP_IS_REDACTED(bp) ? REDACT : DATA));
struct send_range *range = range_alloc(range_type, dn->dn_object,
blkid, blkid + count, B_FALSE);
if (blkid == DMU_SPILL_BLKID)
ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_SA);
switch (range_type) {
case HOLE:
range->sru.hole.datablksz = datablksz;
break;
case DATA:
ASSERT3U(count, ==, 1);
range->sru.data.datablksz = datablksz;
range->sru.data.obj_type = dn->dn_type;
range->sru.data.bp = *bp;
issue_data_read(srta, range);
break;
case REDACT:
range->sru.redact.datablksz = datablksz;
break;
default:
break;
}
bqueue_enqueue(q, range, datablksz);
}
/*
* This thread is responsible for two things: First, it retrieves the correct
* blkptr in the to ds if we need to send the data because of something from
* the from thread. As a result of this, we're the first ones to discover that
* some indirect blocks can be discarded because they're not holes. Second,
* it issues prefetches for the data we need to send.
*/
static void
send_reader_thread(void *arg)
{
struct send_reader_thread_arg *srta = arg;
struct send_merge_thread_arg *smta = srta->smta;
bqueue_t *inq = &smta->q;
bqueue_t *outq = &srta->q;
objset_t *os = smta->os;
fstrans_cookie_t cookie = spl_fstrans_mark();
struct send_range *range = bqueue_dequeue(inq);
int err = 0;
/*
* If the record we're analyzing is from a redaction bookmark from the
* fromds, then we need to know whether or not it exists in the tods so
* we know whether to create records for it or not. If it does, we need
* the datablksz so we can generate an appropriate record for it.
* Finally, if it isn't redacted, we need the blkptr so that we can send
* a WRITE record containing the actual data.
*/
uint64_t last_obj = UINT64_MAX;
uint64_t last_obj_exists = B_TRUE;
while (!range->eos_marker && !srta->cancel && smta->error == 0 &&
err == 0) {
switch (range->type) {
case DATA:
issue_data_read(srta, range);
bqueue_enqueue(outq, range, range->sru.data.datablksz);
range = get_next_range_nofree(inq, range);
break;
case HOLE:
case OBJECT:
case OBJECT_RANGE:
case REDACT: // Redacted blocks must exist
bqueue_enqueue(outq, range, sizeof (*range));
range = get_next_range_nofree(inq, range);
break;
case PREVIOUSLY_REDACTED: {
/*
* This entry came from the "from bookmark" when
* sending from a bookmark that has a redaction
* list. We need to check if this object/blkid
* exists in the target ("to") dataset, and if
* not then we drop this entry. We also need
* to fill in the block pointer so that we know
* what to prefetch.
*
* To accomplish the above, we first cache whether or
* not the last object we examined exists. If it
* doesn't, we can drop this record. If it does, we hold
* the dnode and use it to call dbuf_dnode_findbp. We do
* this instead of dbuf_bookmark_findbp because we will
* often operate on large ranges, and holding the dnode
* once is more efficient.
*/
boolean_t object_exists = B_TRUE;
/*
* If the data is redacted, we only care if it exists,
* so that we don't send records for objects that have
* been deleted.
*/
dnode_t *dn;
if (range->object == last_obj && !last_obj_exists) {
/*
* If we're still examining the same object as
* previously, and it doesn't exist, we don't
* need to call dbuf_bookmark_findbp.
*/
object_exists = B_FALSE;
} else {
err = dnode_hold(os, range->object, FTAG, &dn);
if (err == ENOENT) {
object_exists = B_FALSE;
err = 0;
}
last_obj = range->object;
last_obj_exists = object_exists;
}
if (err != 0) {
break;
} else if (!object_exists) {
/*
* The block was modified, but doesn't
* exist in the to dataset; if it was
* deleted in the to dataset, then we'll
* visit the hole bp for it at some point.
*/
range = get_next_range(inq, range);
continue;
}
uint64_t file_max =
(dn->dn_maxblkid < range->end_blkid ?
dn->dn_maxblkid : range->end_blkid);
/*
* The object exists, so we need to try to find the
* blkptr for each block in the range we're processing.
*/
rw_enter(&dn->dn_struct_rwlock, RW_READER);
for (uint64_t blkid = range->start_blkid;
blkid < file_max; blkid++) {
blkptr_t bp;
uint32_t datablksz =
dn->dn_phys->dn_datablkszsec <<
SPA_MINBLOCKSHIFT;
uint64_t offset = blkid * datablksz;
/*
* This call finds the next non-hole block in
* the object. This is to prevent a
* performance problem where we're unredacting
* a large hole. Using dnode_next_offset to
* skip over the large hole avoids iterating
* over every block in it.
*/
err = dnode_next_offset(dn, DNODE_FIND_HAVELOCK,
&offset, 1, 1, 0);
if (err == ESRCH) {
offset = UINT64_MAX;
err = 0;
} else if (err != 0) {
break;
}
if (offset != blkid * datablksz) {
/*
* if there is a hole from here
* (blkid) to offset
*/
offset = MIN(offset, file_max *
datablksz);
uint64_t nblks = (offset / datablksz) -
blkid;
enqueue_range(srta, outq, dn, blkid,
nblks, NULL, datablksz);
blkid += nblks;
}
if (blkid >= file_max)
break;
err = dbuf_dnode_findbp(dn, 0, blkid, &bp,
NULL, NULL);
if (err != 0)
break;
ASSERT(!BP_IS_HOLE(&bp));
enqueue_range(srta, outq, dn, blkid, 1, &bp,
datablksz);
}
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
range = get_next_range(inq, range);
}
}
}
if (srta->cancel || err != 0) {
smta->cancel = B_TRUE;
srta->error = err;
} else if (smta->error != 0) {
srta->error = smta->error;
}
while (!range->eos_marker)
range = get_next_range(inq, range);
bqueue_enqueue_flush(outq, range, 1);
spl_fstrans_unmark(cookie);
thread_exit();
}
#define NUM_SNAPS_NOT_REDACTED UINT64_MAX
struct dmu_send_params {
/* Pool args */
void *tag; // Tag that dp was held with, will be used to release dp.
dsl_pool_t *dp;
/* To snapshot args */
const char *tosnap;
dsl_dataset_t *to_ds;
/* From snapshot args */
zfs_bookmark_phys_t ancestor_zb;
uint64_t *fromredactsnaps;
/* NUM_SNAPS_NOT_REDACTED if not sending from redaction bookmark */
uint64_t numfromredactsnaps;
/* Stream params */
boolean_t is_clone;
boolean_t embedok;
boolean_t large_block_ok;
boolean_t compressok;
boolean_t rawok;
boolean_t savedok;
uint64_t resumeobj;
uint64_t resumeoff;
uint64_t saved_guid;
zfs_bookmark_phys_t *redactbook;
/* Stream output params */
dmu_send_outparams_t *dso;
/* Stream progress params */
offset_t *off;
int outfd;
char saved_toname[MAXNAMELEN];
};
static int
setup_featureflags(struct dmu_send_params *dspp, objset_t *os,
uint64_t *featureflags)
{
dsl_dataset_t *to_ds = dspp->to_ds;
dsl_pool_t *dp = dspp->dp;
#ifdef _KERNEL
if (dmu_objset_type(os) == DMU_OST_ZFS) {
uint64_t version;
if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &version) != 0)
return (SET_ERROR(EINVAL));
if (version >= ZPL_VERSION_SA)
*featureflags |= DMU_BACKUP_FEATURE_SA_SPILL;
}
#endif
/* raw sends imply large_block_ok */
if ((dspp->rawok || dspp->large_block_ok) &&
dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_LARGE_BLOCKS)) {
*featureflags |= DMU_BACKUP_FEATURE_LARGE_BLOCKS;
}
/* encrypted datasets will not have embedded blocks */
if ((dspp->embedok || dspp->rawok) && !os->os_encrypted &&
spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) {
*featureflags |= DMU_BACKUP_FEATURE_EMBED_DATA;
}
/* raw send implies compressok */
if (dspp->compressok || dspp->rawok)
*featureflags |= DMU_BACKUP_FEATURE_COMPRESSED;
if (dspp->rawok && os->os_encrypted)
*featureflags |= DMU_BACKUP_FEATURE_RAW;
if ((*featureflags &
(DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_COMPRESSED |
DMU_BACKUP_FEATURE_RAW)) != 0 &&
spa_feature_is_active(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS)) {
*featureflags |= DMU_BACKUP_FEATURE_LZ4;
}
/*
* We specifically do not include DMU_BACKUP_FEATURE_EMBED_DATA here to
* allow sending ZSTD compressed datasets to a receiver that does not
* support ZSTD
*/
if ((*featureflags &
(DMU_BACKUP_FEATURE_COMPRESSED | DMU_BACKUP_FEATURE_RAW)) != 0 &&
dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_ZSTD_COMPRESS)) {
*featureflags |= DMU_BACKUP_FEATURE_ZSTD;
}
if (dspp->resumeobj != 0 || dspp->resumeoff != 0) {
*featureflags |= DMU_BACKUP_FEATURE_RESUMING;
}
if (dspp->redactbook != NULL) {
*featureflags |= DMU_BACKUP_FEATURE_REDACTED;
}
if (dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_LARGE_DNODE)) {
*featureflags |= DMU_BACKUP_FEATURE_LARGE_DNODE;
}
return (0);
}
static dmu_replay_record_t *
create_begin_record(struct dmu_send_params *dspp, objset_t *os,
uint64_t featureflags)
{
dmu_replay_record_t *drr = kmem_zalloc(sizeof (dmu_replay_record_t),
KM_SLEEP);
drr->drr_type = DRR_BEGIN;
struct drr_begin *drrb = &drr->drr_u.drr_begin;
dsl_dataset_t *to_ds = dspp->to_ds;
drrb->drr_magic = DMU_BACKUP_MAGIC;
drrb->drr_creation_time = dsl_dataset_phys(to_ds)->ds_creation_time;
drrb->drr_type = dmu_objset_type(os);
drrb->drr_toguid = dsl_dataset_phys(to_ds)->ds_guid;
drrb->drr_fromguid = dspp->ancestor_zb.zbm_guid;
DMU_SET_STREAM_HDRTYPE(drrb->drr_versioninfo, DMU_SUBSTREAM);
DMU_SET_FEATUREFLAGS(drrb->drr_versioninfo, featureflags);
if (dspp->is_clone)
drrb->drr_flags |= DRR_FLAG_CLONE;
if (dsl_dataset_phys(dspp->to_ds)->ds_flags & DS_FLAG_CI_DATASET)
drrb->drr_flags |= DRR_FLAG_CI_DATA;
if (zfs_send_set_freerecords_bit)
drrb->drr_flags |= DRR_FLAG_FREERECORDS;
drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_SPILL_BLOCK;
if (dspp->savedok) {
drrb->drr_toguid = dspp->saved_guid;
strlcpy(drrb->drr_toname, dspp->saved_toname,
sizeof (drrb->drr_toname));
} else {
dsl_dataset_name(to_ds, drrb->drr_toname);
if (!to_ds->ds_is_snapshot) {
(void) strlcat(drrb->drr_toname, "@--head--",
sizeof (drrb->drr_toname));
}
}
return (drr);
}
static void
setup_to_thread(struct send_thread_arg *to_arg, objset_t *to_os,
dmu_sendstatus_t *dssp, uint64_t fromtxg, boolean_t rawok)
{
VERIFY0(bqueue_init(&to_arg->q, zfs_send_no_prefetch_queue_ff,
MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
offsetof(struct send_range, ln)));
to_arg->error_code = 0;
to_arg->cancel = B_FALSE;
to_arg->os = to_os;
to_arg->fromtxg = fromtxg;
to_arg->flags = TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA;
if (rawok)
to_arg->flags |= TRAVERSE_NO_DECRYPT;
if (zfs_send_corrupt_data)
to_arg->flags |= TRAVERSE_HARD;
to_arg->num_blocks_visited = &dssp->dss_blocks;
(void) thread_create(NULL, 0, send_traverse_thread, to_arg, 0,
curproc, TS_RUN, minclsyspri);
}
static void
setup_from_thread(struct redact_list_thread_arg *from_arg,
redaction_list_t *from_rl, dmu_sendstatus_t *dssp)
{
VERIFY0(bqueue_init(&from_arg->q, zfs_send_no_prefetch_queue_ff,
MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
offsetof(struct send_range, ln)));
from_arg->error_code = 0;
from_arg->cancel = B_FALSE;
from_arg->rl = from_rl;
from_arg->mark_redact = B_FALSE;
from_arg->num_blocks_visited = &dssp->dss_blocks;
/*
* If from_ds is null, send_traverse_thread just returns success and
* enqueues an eos marker.
*/
(void) thread_create(NULL, 0, redact_list_thread, from_arg, 0,
curproc, TS_RUN, minclsyspri);
}
static void
setup_redact_list_thread(struct redact_list_thread_arg *rlt_arg,
struct dmu_send_params *dspp, redaction_list_t *rl, dmu_sendstatus_t *dssp)
{
if (dspp->redactbook == NULL)
return;
rlt_arg->cancel = B_FALSE;
VERIFY0(bqueue_init(&rlt_arg->q, zfs_send_no_prefetch_queue_ff,
MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
offsetof(struct send_range, ln)));
rlt_arg->error_code = 0;
rlt_arg->mark_redact = B_TRUE;
rlt_arg->rl = rl;
rlt_arg->num_blocks_visited = &dssp->dss_blocks;
(void) thread_create(NULL, 0, redact_list_thread, rlt_arg, 0,
curproc, TS_RUN, minclsyspri);
}
static void
setup_merge_thread(struct send_merge_thread_arg *smt_arg,
struct dmu_send_params *dspp, struct redact_list_thread_arg *from_arg,
struct send_thread_arg *to_arg, struct redact_list_thread_arg *rlt_arg,
objset_t *os)
{
VERIFY0(bqueue_init(&smt_arg->q, zfs_send_no_prefetch_queue_ff,
MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
offsetof(struct send_range, ln)));
smt_arg->cancel = B_FALSE;
smt_arg->error = 0;
smt_arg->from_arg = from_arg;
smt_arg->to_arg = to_arg;
if (dspp->redactbook != NULL)
smt_arg->redact_arg = rlt_arg;
smt_arg->os = os;
(void) thread_create(NULL, 0, send_merge_thread, smt_arg, 0, curproc,
TS_RUN, minclsyspri);
}
static void
setup_reader_thread(struct send_reader_thread_arg *srt_arg,
struct dmu_send_params *dspp, struct send_merge_thread_arg *smt_arg,
uint64_t featureflags)
{
VERIFY0(bqueue_init(&srt_arg->q, zfs_send_queue_ff,
MAX(zfs_send_queue_length, 2 * zfs_max_recordsize),
offsetof(struct send_range, ln)));
srt_arg->smta = smt_arg;
srt_arg->issue_reads = !dspp->dso->dso_dryrun;
srt_arg->featureflags = featureflags;
(void) thread_create(NULL, 0, send_reader_thread, srt_arg, 0,
curproc, TS_RUN, minclsyspri);
}
static int
setup_resume_points(struct dmu_send_params *dspp,
struct send_thread_arg *to_arg, struct redact_list_thread_arg *from_arg,
struct redact_list_thread_arg *rlt_arg,
struct send_merge_thread_arg *smt_arg, boolean_t resuming, objset_t *os,
redaction_list_t *redact_rl, nvlist_t *nvl)
{
(void) smt_arg;
dsl_dataset_t *to_ds = dspp->to_ds;
int err = 0;
uint64_t obj = 0;
uint64_t blkid = 0;
if (resuming) {
obj = dspp->resumeobj;
dmu_object_info_t to_doi;
err = dmu_object_info(os, obj, &to_doi);
if (err != 0)
return (err);
blkid = dspp->resumeoff / to_doi.doi_data_block_size;
}
/*
* If we're resuming a redacted send, we can skip to the appropriate
* point in the redaction bookmark by binary searching through it.
*/
if (redact_rl != NULL) {
SET_BOOKMARK(&rlt_arg->resume, to_ds->ds_object, obj, 0, blkid);
}
SET_BOOKMARK(&to_arg->resume, to_ds->ds_object, obj, 0, blkid);
if (nvlist_exists(nvl, BEGINNV_REDACT_FROM_SNAPS)) {
uint64_t objset = dspp->ancestor_zb.zbm_redaction_obj;
/*
* Note: If the resume point is in an object whose
* blocksize is different in the from vs to snapshots,
* we will have divided by the "wrong" blocksize.
* However, in this case fromsnap's send_cb() will
* detect that the blocksize has changed and therefore
* ignore this object.
*
* If we're resuming a send from a redaction bookmark,
* we still cannot accidentally suggest blocks behind
* the to_ds. In addition, we know that any blocks in
* the object in the to_ds will have to be sent, since
* the size changed. Therefore, we can't cause any harm
* this way either.
*/
SET_BOOKMARK(&from_arg->resume, objset, obj, 0, blkid);
}
if (resuming) {
fnvlist_add_uint64(nvl, BEGINNV_RESUME_OBJECT, dspp->resumeobj);
fnvlist_add_uint64(nvl, BEGINNV_RESUME_OFFSET, dspp->resumeoff);
}
return (0);
}
static dmu_sendstatus_t *
setup_send_progress(struct dmu_send_params *dspp)
{
dmu_sendstatus_t *dssp = kmem_zalloc(sizeof (*dssp), KM_SLEEP);
dssp->dss_outfd = dspp->outfd;
dssp->dss_off = dspp->off;
dssp->dss_proc = curproc;
mutex_enter(&dspp->to_ds->ds_sendstream_lock);
list_insert_head(&dspp->to_ds->ds_sendstreams, dssp);
mutex_exit(&dspp->to_ds->ds_sendstream_lock);
return (dssp);
}
/*
* Actually do the bulk of the work in a zfs send.
*
* The idea is that we want to do a send from ancestor_zb to to_ds. We also
* want to not send any data that has been modified by all the datasets in
* redactsnaparr, and store the list of blocks that are redacted in this way in
* a bookmark named redactbook, created on the to_ds. We do this by creating
* several worker threads, whose function is described below.
*
* There are three cases.
* The first case is a redacted zfs send. In this case there are 5 threads.
* The first thread is the to_ds traversal thread: it calls dataset_traverse on
* the to_ds and finds all the blocks that have changed since ancestor_zb (if
* it's a full send, that's all blocks in the dataset). It then sends those
* blocks on to the send merge thread. The redact list thread takes the data
* from the redaction bookmark and sends those blocks on to the send merge
* thread. The send merge thread takes the data from the to_ds traversal
* thread, and combines it with the redaction records from the redact list
* thread. If a block appears in both the to_ds's data and the redaction data,
* the send merge thread will mark it as redacted and send it on to the prefetch
* thread. Otherwise, the send merge thread will send the block on to the
* prefetch thread unchanged. The prefetch thread will issue prefetch reads for
* any data that isn't redacted, and then send the data on to the main thread.
* The main thread behaves the same as in a normal send case, issuing demand
* reads for data blocks and sending out records over the network
*
* The graphic below diagrams the flow of data in the case of a redacted zfs
* send. Each box represents a thread, and each line represents the flow of
* data.
*
* Records from the |
* redaction bookmark |
* +--------------------+ | +---------------------------+
* | | v | Send Merge Thread |
* | Redact List Thread +----------> Apply redaction marks to |
* | | | records as specified by |
* +--------------------+ | redaction ranges |
* +----^---------------+------+
* | | Merged data
* | |
* | +------------v--------+
* | | Prefetch Thread |
* +--------------------+ | | Issues prefetch |
* | to_ds Traversal | | | reads of data blocks|
* | Thread (finds +---------------+ +------------+--------+
* | candidate blocks) | Blocks modified | Prefetched data
* +--------------------+ by to_ds since |
* ancestor_zb +------------v----+
* | Main Thread | File Descriptor
* | Sends data over +->(to zfs receive)
* | wire |
* +-----------------+
*
* The second case is an incremental send from a redaction bookmark. The to_ds
* traversal thread and the main thread behave the same as in the redacted
* send case. The new thread is the from bookmark traversal thread. It
* iterates over the redaction list in the redaction bookmark, and enqueues
* records for each block that was redacted in the original send. The send
* merge thread now has to merge the data from the two threads. For details
* about that process, see the header comment of send_merge_thread(). Any data
* it decides to send on will be prefetched by the prefetch thread. Note that
* you can perform a redacted send from a redaction bookmark; in that case,
* the data flow behaves very similarly to the flow in the redacted send case,
* except with the addition of the bookmark traversal thread iterating over the
* redaction bookmark. The send_merge_thread also has to take on the
* responsibility of merging the redact list thread's records, the bookmark
* traversal thread's records, and the to_ds records.
*
* +---------------------+
* | |
* | Redact List Thread +--------------+
* | | |
* +---------------------+ |
* Blocks in redaction list | Ranges modified by every secure snap
* of from bookmark | (or EOS if not readcted)
* |
* +---------------------+ | +----v----------------------+
* | bookmark Traversal | v | Send Merge Thread |
* | Thread (finds +---------> Merges bookmark, rlt, and |
* | candidate blocks) | | to_ds send records |
* +---------------------+ +----^---------------+------+
* | | Merged data
* | +------------v--------+
* | | Prefetch Thread |
* +--------------------+ | | Issues prefetch |
* | to_ds Traversal | | | reads of data blocks|
* | Thread (finds +---------------+ +------------+--------+
* | candidate blocks) | Blocks modified | Prefetched data
* +--------------------+ by to_ds since +------------v----+
* ancestor_zb | Main Thread | File Descriptor
* | Sends data over +->(to zfs receive)
* | wire |
* +-----------------+
*
* The final case is a simple zfs full or incremental send. The to_ds traversal
* thread behaves the same as always. The redact list thread is never started.
* The send merge thread takes all the blocks that the to_ds traversal thread
* sends it, prefetches the data, and sends the blocks on to the main thread.
* The main thread sends the data over the wire.
*
* To keep performance acceptable, we want to prefetch the data in the worker
* threads. While the to_ds thread could simply use the TRAVERSE_PREFETCH
* feature built into traverse_dataset, the combining and deletion of records
* due to redaction and sends from redaction bookmarks mean that we could
* issue many unnecessary prefetches. As a result, we only prefetch data
* after we've determined that the record is not going to be redacted. To
* prevent the prefetching from getting too far ahead of the main thread, the
* blocking queues that are used for communication are capped not by the
* number of entries in the queue, but by the sum of the size of the
* prefetches associated with them. The limit on the amount of data that the
* thread can prefetch beyond what the main thread has reached is controlled
* by the global variable zfs_send_queue_length. In addition, to prevent poor
* performance in the beginning of a send, we also limit the distance ahead
* that the traversal threads can be. That distance is controlled by the
* zfs_send_no_prefetch_queue_length tunable.
*
* Note: Releases dp using the specified tag.
*/
static int
dmu_send_impl(struct dmu_send_params *dspp)
{
objset_t *os;
dmu_replay_record_t *drr;
dmu_sendstatus_t *dssp;
dmu_send_cookie_t dsc = {0};
int err;
uint64_t fromtxg = dspp->ancestor_zb.zbm_creation_txg;
uint64_t featureflags = 0;
struct redact_list_thread_arg *from_arg;
struct send_thread_arg *to_arg;
struct redact_list_thread_arg *rlt_arg;
struct send_merge_thread_arg *smt_arg;
struct send_reader_thread_arg *srt_arg;
struct send_range *range;
redaction_list_t *from_rl = NULL;
redaction_list_t *redact_rl = NULL;
boolean_t resuming = (dspp->resumeobj != 0 || dspp->resumeoff != 0);
boolean_t book_resuming = resuming;
dsl_dataset_t *to_ds = dspp->to_ds;
zfs_bookmark_phys_t *ancestor_zb = &dspp->ancestor_zb;
dsl_pool_t *dp = dspp->dp;
void *tag = dspp->tag;
err = dmu_objset_from_ds(to_ds, &os);
if (err != 0) {
dsl_pool_rele(dp, tag);
return (err);
}
/*
* If this is a non-raw send of an encrypted ds, we can ensure that
* the objset_phys_t is authenticated. This is safe because this is
* either a snapshot or we have owned the dataset, ensuring that
* it can't be modified.
*/
if (!dspp->rawok && os->os_encrypted &&
arc_is_unauthenticated(os->os_phys_buf)) {
zbookmark_phys_t zb;
SET_BOOKMARK(&zb, to_ds->ds_object, ZB_ROOT_OBJECT,
ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
err = arc_untransform(os->os_phys_buf, os->os_spa,
&zb, B_FALSE);
if (err != 0) {
dsl_pool_rele(dp, tag);
return (err);
}
ASSERT0(arc_is_unauthenticated(os->os_phys_buf));
}
if ((err = setup_featureflags(dspp, os, &featureflags)) != 0) {
dsl_pool_rele(dp, tag);
return (err);
}
/*
* If we're doing a redacted send, hold the bookmark's redaction list.
*/
if (dspp->redactbook != NULL) {
err = dsl_redaction_list_hold_obj(dp,
dspp->redactbook->zbm_redaction_obj, FTAG,
&redact_rl);
if (err != 0) {
dsl_pool_rele(dp, tag);
return (SET_ERROR(EINVAL));
}
dsl_redaction_list_long_hold(dp, redact_rl, FTAG);
}
/*
* If we're sending from a redaction bookmark, hold the redaction list
* so that we can consider sending the redacted blocks.
*/
if (ancestor_zb->zbm_redaction_obj != 0) {
err = dsl_redaction_list_hold_obj(dp,
ancestor_zb->zbm_redaction_obj, FTAG, &from_rl);
if (err != 0) {
if (redact_rl != NULL) {
dsl_redaction_list_long_rele(redact_rl, FTAG);
dsl_redaction_list_rele(redact_rl, FTAG);
}
dsl_pool_rele(dp, tag);
return (SET_ERROR(EINVAL));
}
dsl_redaction_list_long_hold(dp, from_rl, FTAG);
}
dsl_dataset_long_hold(to_ds, FTAG);
from_arg = kmem_zalloc(sizeof (*from_arg), KM_SLEEP);
to_arg = kmem_zalloc(sizeof (*to_arg), KM_SLEEP);
rlt_arg = kmem_zalloc(sizeof (*rlt_arg), KM_SLEEP);
smt_arg = kmem_zalloc(sizeof (*smt_arg), KM_SLEEP);
srt_arg = kmem_zalloc(sizeof (*srt_arg), KM_SLEEP);
drr = create_begin_record(dspp, os, featureflags);
dssp = setup_send_progress(dspp);
dsc.dsc_drr = drr;
dsc.dsc_dso = dspp->dso;
dsc.dsc_os = os;
dsc.dsc_off = dspp->off;
dsc.dsc_toguid = dsl_dataset_phys(to_ds)->ds_guid;
dsc.dsc_fromtxg = fromtxg;
dsc.dsc_pending_op = PENDING_NONE;
dsc.dsc_featureflags = featureflags;
dsc.dsc_resume_object = dspp->resumeobj;
dsc.dsc_resume_offset = dspp->resumeoff;
dsl_pool_rele(dp, tag);
void *payload = NULL;
size_t payload_len = 0;
nvlist_t *nvl = fnvlist_alloc();
/*
* If we're doing a redacted send, we include the snapshots we're
* redacted with respect to so that the target system knows what send
* streams can be correctly received on top of this dataset. If we're
* instead sending a redacted dataset, we include the snapshots that the
* dataset was created with respect to.
*/
if (dspp->redactbook != NULL) {
fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_SNAPS,
redact_rl->rl_phys->rlp_snaps,
redact_rl->rl_phys->rlp_num_snaps);
} else if (dsl_dataset_feature_is_active(to_ds,
SPA_FEATURE_REDACTED_DATASETS)) {
uint64_t *tods_guids;
uint64_t length;
VERIFY(dsl_dataset_get_uint64_array_feature(to_ds,
SPA_FEATURE_REDACTED_DATASETS, &length, &tods_guids));
fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_SNAPS, tods_guids,
length);
}
/*
* If we're sending from a redaction bookmark, then we should retrieve
* the guids of that bookmark so we can send them over the wire.
*/
if (from_rl != NULL) {
fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_FROM_SNAPS,
from_rl->rl_phys->rlp_snaps,
from_rl->rl_phys->rlp_num_snaps);
}
/*
* If the snapshot we're sending from is redacted, include the redaction
* list in the stream.
*/
if (dspp->numfromredactsnaps != NUM_SNAPS_NOT_REDACTED) {
ASSERT3P(from_rl, ==, NULL);
fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_FROM_SNAPS,
dspp->fromredactsnaps, (uint_t)dspp->numfromredactsnaps);
if (dspp->numfromredactsnaps > 0) {
kmem_free(dspp->fromredactsnaps,
dspp->numfromredactsnaps * sizeof (uint64_t));
dspp->fromredactsnaps = NULL;
}
}
if (resuming || book_resuming) {
err = setup_resume_points(dspp, to_arg, from_arg,
rlt_arg, smt_arg, resuming, os, redact_rl, nvl);
if (err != 0)
goto out;
}
if (featureflags & DMU_BACKUP_FEATURE_RAW) {
uint64_t ivset_guid = (ancestor_zb != NULL) ?
ancestor_zb->zbm_ivset_guid : 0;
nvlist_t *keynvl = NULL;
ASSERT(os->os_encrypted);
err = dsl_crypto_populate_key_nvlist(os, ivset_guid,
&keynvl);
if (err != 0) {
fnvlist_free(nvl);
goto out;
}
fnvlist_add_nvlist(nvl, "crypt_keydata", keynvl);
fnvlist_free(keynvl);
}
if (!nvlist_empty(nvl)) {
payload = fnvlist_pack(nvl, &payload_len);
drr->drr_payloadlen = payload_len;
}
fnvlist_free(nvl);
err = dump_record(&dsc, payload, payload_len);
fnvlist_pack_free(payload, payload_len);
if (err != 0) {
err = dsc.dsc_err;
goto out;
}
setup_to_thread(to_arg, os, dssp, fromtxg, dspp->rawok);
setup_from_thread(from_arg, from_rl, dssp);
setup_redact_list_thread(rlt_arg, dspp, redact_rl, dssp);
setup_merge_thread(smt_arg, dspp, from_arg, to_arg, rlt_arg, os);
setup_reader_thread(srt_arg, dspp, smt_arg, featureflags);
range = bqueue_dequeue(&srt_arg->q);
while (err == 0 && !range->eos_marker) {
err = do_dump(&dsc, range);
range = get_next_range(&srt_arg->q, range);
if (issig(JUSTLOOKING) && issig(FORREAL))
err = SET_ERROR(EINTR);
}
/*
* If we hit an error or are interrupted, cancel our worker threads and
* clear the queue of any pending records. The threads will pass the
* cancel up the tree of worker threads, and each one will clean up any
* pending records before exiting.
*/
if (err != 0) {
srt_arg->cancel = B_TRUE;
while (!range->eos_marker) {
range = get_next_range(&srt_arg->q, range);
}
}
range_free(range);
bqueue_destroy(&srt_arg->q);
bqueue_destroy(&smt_arg->q);
if (dspp->redactbook != NULL)
bqueue_destroy(&rlt_arg->q);
bqueue_destroy(&to_arg->q);
bqueue_destroy(&from_arg->q);
if (err == 0 && srt_arg->error != 0)
err = srt_arg->error;
if (err != 0)
goto out;
if (dsc.dsc_pending_op != PENDING_NONE)
if (dump_record(&dsc, NULL, 0) != 0)
err = SET_ERROR(EINTR);
if (err != 0) {
if (err == EINTR && dsc.dsc_err != 0)
err = dsc.dsc_err;
goto out;
}
/*
* Send the DRR_END record if this is not a saved stream.
* Otherwise, the omitted DRR_END record will signal to
* the receive side that the stream is incomplete.
*/
if (!dspp->savedok) {
bzero(drr, sizeof (dmu_replay_record_t));
drr->drr_type = DRR_END;
drr->drr_u.drr_end.drr_checksum = dsc.dsc_zc;
drr->drr_u.drr_end.drr_toguid = dsc.dsc_toguid;
if (dump_record(&dsc, NULL, 0) != 0)
err = dsc.dsc_err;
}
out:
mutex_enter(&to_ds->ds_sendstream_lock);
list_remove(&to_ds->ds_sendstreams, dssp);
mutex_exit(&to_ds->ds_sendstream_lock);
VERIFY(err != 0 || (dsc.dsc_sent_begin &&
(dsc.dsc_sent_end || dspp->savedok)));
kmem_free(drr, sizeof (dmu_replay_record_t));
kmem_free(dssp, sizeof (dmu_sendstatus_t));
kmem_free(from_arg, sizeof (*from_arg));
kmem_free(to_arg, sizeof (*to_arg));
kmem_free(rlt_arg, sizeof (*rlt_arg));
kmem_free(smt_arg, sizeof (*smt_arg));
kmem_free(srt_arg, sizeof (*srt_arg));
dsl_dataset_long_rele(to_ds, FTAG);
if (from_rl != NULL) {
dsl_redaction_list_long_rele(from_rl, FTAG);
dsl_redaction_list_rele(from_rl, FTAG);
}
if (redact_rl != NULL) {
dsl_redaction_list_long_rele(redact_rl, FTAG);
dsl_redaction_list_rele(redact_rl, FTAG);
}
return (err);
}
int
dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
boolean_t embedok, boolean_t large_block_ok, boolean_t compressok,
boolean_t rawok, boolean_t savedok, int outfd, offset_t *off,
dmu_send_outparams_t *dsop)
{
int err;
dsl_dataset_t *fromds;
ds_hold_flags_t dsflags;
struct dmu_send_params dspp = {0};
dspp.embedok = embedok;
dspp.large_block_ok = large_block_ok;
dspp.compressok = compressok;
dspp.outfd = outfd;
dspp.off = off;
dspp.dso = dsop;
dspp.tag = FTAG;
dspp.rawok = rawok;
dspp.savedok = savedok;
dsflags = (rawok) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
err = dsl_pool_hold(pool, FTAG, &dspp.dp);
if (err != 0)
return (err);
err = dsl_dataset_hold_obj_flags(dspp.dp, tosnap, dsflags, FTAG,
&dspp.to_ds);
if (err != 0) {
dsl_pool_rele(dspp.dp, FTAG);
return (err);
}
if (fromsnap != 0) {
err = dsl_dataset_hold_obj_flags(dspp.dp, fromsnap, dsflags,
FTAG, &fromds);
if (err != 0) {
dsl_dataset_rele_flags(dspp.to_ds, dsflags, FTAG);
dsl_pool_rele(dspp.dp, FTAG);
return (err);
}
dspp.ancestor_zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid;
dspp.ancestor_zb.zbm_creation_txg =
dsl_dataset_phys(fromds)->ds_creation_txg;
dspp.ancestor_zb.zbm_creation_time =
dsl_dataset_phys(fromds)->ds_creation_time;
if (dsl_dataset_is_zapified(fromds)) {
(void) zap_lookup(dspp.dp->dp_meta_objset,
fromds->ds_object, DS_FIELD_IVSET_GUID, 8, 1,
&dspp.ancestor_zb.zbm_ivset_guid);
}
/* See dmu_send for the reasons behind this. */
uint64_t *fromredact;
if (!dsl_dataset_get_uint64_array_feature(fromds,
SPA_FEATURE_REDACTED_DATASETS,
&dspp.numfromredactsnaps,
&fromredact)) {
dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
} else if (dspp.numfromredactsnaps > 0) {
uint64_t size = dspp.numfromredactsnaps *
sizeof (uint64_t);
dspp.fromredactsnaps = kmem_zalloc(size, KM_SLEEP);
bcopy(fromredact, dspp.fromredactsnaps, size);
}
boolean_t is_before =
dsl_dataset_is_before(dspp.to_ds, fromds, 0);
dspp.is_clone = (dspp.to_ds->ds_dir !=
fromds->ds_dir);
dsl_dataset_rele(fromds, FTAG);
if (!is_before) {
dsl_pool_rele(dspp.dp, FTAG);
err = SET_ERROR(EXDEV);
} else {
err = dmu_send_impl(&dspp);
}
} else {
dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
err = dmu_send_impl(&dspp);
}
dsl_dataset_rele(dspp.to_ds, FTAG);
return (err);
}
int
dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok,
boolean_t large_block_ok, boolean_t compressok, boolean_t rawok,
boolean_t savedok, uint64_t resumeobj, uint64_t resumeoff,
const char *redactbook, int outfd, offset_t *off,
dmu_send_outparams_t *dsop)
{
int err = 0;
ds_hold_flags_t dsflags;
boolean_t owned = B_FALSE;
dsl_dataset_t *fromds = NULL;
zfs_bookmark_phys_t book = {0};
struct dmu_send_params dspp = {0};
dsflags = (rawok) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
dspp.tosnap = tosnap;
dspp.embedok = embedok;
dspp.large_block_ok = large_block_ok;
dspp.compressok = compressok;
dspp.outfd = outfd;
dspp.off = off;
dspp.dso = dsop;
dspp.tag = FTAG;
dspp.resumeobj = resumeobj;
dspp.resumeoff = resumeoff;
dspp.rawok = rawok;
dspp.savedok = savedok;
if (fromsnap != NULL && strpbrk(fromsnap, "@#") == NULL)
return (SET_ERROR(EINVAL));
err = dsl_pool_hold(tosnap, FTAG, &dspp.dp);
if (err != 0)
return (err);
if (strchr(tosnap, '@') == NULL && spa_writeable(dspp.dp->dp_spa)) {
/*
* We are sending a filesystem or volume. Ensure
* that it doesn't change by owning the dataset.
*/
if (savedok) {
/*
* We are looking for the dataset that represents the
* partially received send stream. If this stream was
* received as a new snapshot of an existing dataset,
* this will be saved in a hidden clone named
* "<pool>/<dataset>/%recv". Otherwise, the stream
* will be saved in the live dataset itself. In
* either case we need to use dsl_dataset_own_force()
* because the stream is marked as inconsistent,
* which would normally make it unavailable to be
* owned.
*/
char *name = kmem_asprintf("%s/%s", tosnap,
recv_clone_name);
err = dsl_dataset_own_force(dspp.dp, name, dsflags,
FTAG, &dspp.to_ds);
if (err == ENOENT) {
err = dsl_dataset_own_force(dspp.dp, tosnap,
dsflags, FTAG, &dspp.to_ds);
}
if (err == 0) {
err = zap_lookup(dspp.dp->dp_meta_objset,
dspp.to_ds->ds_object,
DS_FIELD_RESUME_TOGUID, 8, 1,
&dspp.saved_guid);
}
if (err == 0) {
err = zap_lookup(dspp.dp->dp_meta_objset,
dspp.to_ds->ds_object,
DS_FIELD_RESUME_TONAME, 1,
sizeof (dspp.saved_toname),
dspp.saved_toname);
}
if (err != 0)
dsl_dataset_disown(dspp.to_ds, dsflags, FTAG);
kmem_strfree(name);
} else {
err = dsl_dataset_own(dspp.dp, tosnap, dsflags,
FTAG, &dspp.to_ds);
}
owned = B_TRUE;
} else {
err = dsl_dataset_hold_flags(dspp.dp, tosnap, dsflags, FTAG,
&dspp.to_ds);
}
if (err != 0) {
dsl_pool_rele(dspp.dp, FTAG);
return (err);
}
if (redactbook != NULL) {
char path[ZFS_MAX_DATASET_NAME_LEN];
(void) strlcpy(path, tosnap, sizeof (path));
char *at = strchr(path, '@');
if (at == NULL) {
err = EINVAL;
} else {
(void) snprintf(at, sizeof (path) - (at - path), "#%s",
redactbook);
err = dsl_bookmark_lookup(dspp.dp, path,
NULL, &book);
dspp.redactbook = &book;
}
}
if (err != 0) {
dsl_pool_rele(dspp.dp, FTAG);
if (owned)
dsl_dataset_disown(dspp.to_ds, dsflags, FTAG);
else
dsl_dataset_rele_flags(dspp.to_ds, dsflags, FTAG);
return (err);
}
if (fromsnap != NULL) {
zfs_bookmark_phys_t *zb = &dspp.ancestor_zb;
int fsnamelen;
if (strpbrk(tosnap, "@#") != NULL)
fsnamelen = strpbrk(tosnap, "@#") - tosnap;
else
fsnamelen = strlen(tosnap);
/*
* If the fromsnap is in a different filesystem, then
* mark the send stream as a clone.
*/
if (strncmp(tosnap, fromsnap, fsnamelen) != 0 ||
(fromsnap[fsnamelen] != '@' &&
fromsnap[fsnamelen] != '#')) {
dspp.is_clone = B_TRUE;
}
if (strchr(fromsnap, '@') != NULL) {
err = dsl_dataset_hold(dspp.dp, fromsnap, FTAG,
&fromds);
if (err != 0) {
ASSERT3P(fromds, ==, NULL);
} else {
/*
* We need to make a deep copy of the redact
* snapshots of the from snapshot, because the
* array will be freed when we evict from_ds.
*/
uint64_t *fromredact;
if (!dsl_dataset_get_uint64_array_feature(
fromds, SPA_FEATURE_REDACTED_DATASETS,
&dspp.numfromredactsnaps,
&fromredact)) {
dspp.numfromredactsnaps =
NUM_SNAPS_NOT_REDACTED;
} else if (dspp.numfromredactsnaps > 0) {
uint64_t size =
dspp.numfromredactsnaps *
sizeof (uint64_t);
dspp.fromredactsnaps = kmem_zalloc(size,
KM_SLEEP);
bcopy(fromredact, dspp.fromredactsnaps,
size);
}
if (!dsl_dataset_is_before(dspp.to_ds, fromds,
0)) {
err = SET_ERROR(EXDEV);
} else {
zb->zbm_creation_txg =
dsl_dataset_phys(fromds)->
ds_creation_txg;
zb->zbm_creation_time =
dsl_dataset_phys(fromds)->
ds_creation_time;
zb->zbm_guid =
dsl_dataset_phys(fromds)->ds_guid;
zb->zbm_redaction_obj = 0;
if (dsl_dataset_is_zapified(fromds)) {
(void) zap_lookup(
dspp.dp->dp_meta_objset,
fromds->ds_object,
DS_FIELD_IVSET_GUID, 8, 1,
&zb->zbm_ivset_guid);
}
}
dsl_dataset_rele(fromds, FTAG);
}
} else {
dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
err = dsl_bookmark_lookup(dspp.dp, fromsnap, dspp.to_ds,
zb);
if (err == EXDEV && zb->zbm_redaction_obj != 0 &&
zb->zbm_guid ==
dsl_dataset_phys(dspp.to_ds)->ds_guid)
err = 0;
}
if (err == 0) {
/* dmu_send_impl will call dsl_pool_rele for us. */
err = dmu_send_impl(&dspp);
} else {
dsl_pool_rele(dspp.dp, FTAG);
}
} else {
dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
err = dmu_send_impl(&dspp);
}
if (owned)
dsl_dataset_disown(dspp.to_ds, dsflags, FTAG);
else
dsl_dataset_rele_flags(dspp.to_ds, dsflags, FTAG);
return (err);
}
static int
dmu_adjust_send_estimate_for_indirects(dsl_dataset_t *ds, uint64_t uncompressed,
uint64_t compressed, boolean_t stream_compressed, uint64_t *sizep)
{
int err = 0;
uint64_t size;
/*
* Assume that space (both on-disk and in-stream) is dominated by
* data. We will adjust for indirect blocks and the copies property,
* but ignore per-object space used (eg, dnodes and DRR_OBJECT records).
*/
uint64_t recordsize;
uint64_t record_count;
objset_t *os;
VERIFY0(dmu_objset_from_ds(ds, &os));
/* Assume all (uncompressed) blocks are recordsize. */
if (zfs_override_estimate_recordsize != 0) {
recordsize = zfs_override_estimate_recordsize;
} else if (os->os_phys->os_type == DMU_OST_ZVOL) {
err = dsl_prop_get_int_ds(ds,
zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &recordsize);
} else {
err = dsl_prop_get_int_ds(ds,
zfs_prop_to_name(ZFS_PROP_RECORDSIZE), &recordsize);
}
if (err != 0)
return (err);
record_count = uncompressed / recordsize;
/*
* If we're estimating a send size for a compressed stream, use the
* compressed data size to estimate the stream size. Otherwise, use the
* uncompressed data size.
*/
size = stream_compressed ? compressed : uncompressed;
/*
* Subtract out approximate space used by indirect blocks.
* Assume most space is used by data blocks (non-indirect, non-dnode).
* Assume no ditto blocks or internal fragmentation.
*
* Therefore, space used by indirect blocks is sizeof(blkptr_t) per
* block.
*/
size -= record_count * sizeof (blkptr_t);
/* Add in the space for the record associated with each block. */
size += record_count * sizeof (dmu_replay_record_t);
*sizep = size;
return (0);
}
int
dmu_send_estimate_fast(dsl_dataset_t *origds, dsl_dataset_t *fromds,
zfs_bookmark_phys_t *frombook, boolean_t stream_compressed,
boolean_t saved, uint64_t *sizep)
{
int err;
dsl_dataset_t *ds = origds;
uint64_t uncomp, comp;
ASSERT(dsl_pool_config_held(origds->ds_dir->dd_pool));
ASSERT(fromds == NULL || frombook == NULL);
/*
* If this is a saved send we may actually be sending
* from the %recv clone used for resuming.
*/
if (saved) {
objset_t *mos = origds->ds_dir->dd_pool->dp_meta_objset;
uint64_t guid;
char dsname[ZFS_MAX_DATASET_NAME_LEN + 6];
dsl_dataset_name(origds, dsname);
(void) strcat(dsname, "/");
(void) strcat(dsname, recv_clone_name);
err = dsl_dataset_hold(origds->ds_dir->dd_pool,
dsname, FTAG, &ds);
if (err != ENOENT && err != 0) {
return (err);
} else if (err == ENOENT) {
ds = origds;
}
/* check that this dataset has partially received data */
err = zap_lookup(mos, ds->ds_object,
DS_FIELD_RESUME_TOGUID, 8, 1, &guid);
if (err != 0) {
err = SET_ERROR(err == ENOENT ? EINVAL : err);
goto out;
}
err = zap_lookup(mos, ds->ds_object,
DS_FIELD_RESUME_TONAME, 1, sizeof (dsname), dsname);
if (err != 0) {
err = SET_ERROR(err == ENOENT ? EINVAL : err);
goto out;
}
}
/* tosnap must be a snapshot or the target of a saved send */
if (!ds->ds_is_snapshot && ds == origds)
return (SET_ERROR(EINVAL));
if (fromds != NULL) {
uint64_t used;
if (!fromds->ds_is_snapshot) {
err = SET_ERROR(EINVAL);
goto out;
}
if (!dsl_dataset_is_before(ds, fromds, 0)) {
err = SET_ERROR(EXDEV);
goto out;
}
err = dsl_dataset_space_written(fromds, ds, &used, &comp,
&uncomp);
if (err != 0)
goto out;
} else if (frombook != NULL) {
uint64_t used;
err = dsl_dataset_space_written_bookmark(frombook, ds, &used,
&comp, &uncomp);
if (err != 0)
goto out;
} else {
uncomp = dsl_dataset_phys(ds)->ds_uncompressed_bytes;
comp = dsl_dataset_phys(ds)->ds_compressed_bytes;
}
err = dmu_adjust_send_estimate_for_indirects(ds, uncomp, comp,
stream_compressed, sizep);
/*
* Add the size of the BEGIN and END records to the estimate.
*/
*sizep += 2 * sizeof (dmu_replay_record_t);
out:
if (ds != origds)
dsl_dataset_rele(ds, FTAG);
return (err);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_send, zfs_send_, corrupt_data, INT, ZMOD_RW,
"Allow sending corrupt data");
ZFS_MODULE_PARAM(zfs_send, zfs_send_, queue_length, INT, ZMOD_RW,
"Maximum send queue length");
ZFS_MODULE_PARAM(zfs_send, zfs_send_, unmodified_spill_blocks, INT, ZMOD_RW,
"Send unmodified spill blocks");
ZFS_MODULE_PARAM(zfs_send, zfs_send_, no_prefetch_queue_length, INT, ZMOD_RW,
"Maximum send queue length for non-prefetch queues");
ZFS_MODULE_PARAM(zfs_send, zfs_send_, queue_ff, INT, ZMOD_RW,
"Send queue fill fraction");
ZFS_MODULE_PARAM(zfs_send, zfs_send_, no_prefetch_queue_ff, INT, ZMOD_RW,
"Send queue fill fraction for non-prefetch queues");
ZFS_MODULE_PARAM(zfs_send, zfs_, override_estimate_recordsize, INT, ZMOD_RW,
"Override block size estimate with fixed size");
/* END CSTYLED */
diff --git a/module/zfs/dmu_traverse.c b/module/zfs/dmu_traverse.c
index 2f1c2978b3be..3763c17de820 100644
--- a/module/zfs/dmu_traverse.c
+++ b/module/zfs/dmu_traverse.c
@@ -1,827 +1,827 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_traverse.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_pool.h>
#include <sys/dnode.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/zio.h>
#include <sys/dmu_impl.h>
#include <sys/sa.h>
#include <sys/sa_impl.h>
#include <sys/callb.h>
#include <sys/zfeature.h>
-int32_t zfs_pd_bytes_max = 50 * 1024 * 1024; /* 50MB */
-int32_t send_holes_without_birth_time = 1;
-int32_t zfs_traverse_indirect_prefetch_limit = 32;
+static int32_t zfs_pd_bytes_max = 50 * 1024 * 1024; /* 50MB */
+static int32_t send_holes_without_birth_time = 1;
+static int32_t zfs_traverse_indirect_prefetch_limit = 32;
typedef struct prefetch_data {
kmutex_t pd_mtx;
kcondvar_t pd_cv;
int32_t pd_bytes_fetched;
int pd_flags;
boolean_t pd_cancel;
boolean_t pd_exited;
zbookmark_phys_t pd_resume;
} prefetch_data_t;
typedef struct traverse_data {
spa_t *td_spa;
uint64_t td_objset;
blkptr_t *td_rootbp;
uint64_t td_min_txg;
zbookmark_phys_t *td_resume;
int td_flags;
prefetch_data_t *td_pfd;
boolean_t td_paused;
uint64_t td_hole_birth_enabled_txg;
blkptr_cb_t *td_func;
void *td_arg;
boolean_t td_realloc_possible;
} traverse_data_t;
static int traverse_dnode(traverse_data_t *td, const blkptr_t *bp,
const dnode_phys_t *dnp, uint64_t objset, uint64_t object);
static void prefetch_dnode_metadata(traverse_data_t *td, const dnode_phys_t *,
uint64_t objset, uint64_t object);
static int
traverse_zil_block(zilog_t *zilog, const blkptr_t *bp, void *arg,
uint64_t claim_txg)
{
traverse_data_t *td = arg;
zbookmark_phys_t zb;
if (BP_IS_HOLE(bp))
return (0);
if (claim_txg == 0 && bp->blk_birth >= spa_min_claim_txg(td->td_spa))
return (-1);
SET_BOOKMARK(&zb, td->td_objset, ZB_ZIL_OBJECT, ZB_ZIL_LEVEL,
bp->blk_cksum.zc_word[ZIL_ZC_SEQ]);
(void) td->td_func(td->td_spa, zilog, bp, &zb, NULL, td->td_arg);
return (0);
}
static int
traverse_zil_record(zilog_t *zilog, const lr_t *lrc, void *arg,
uint64_t claim_txg)
{
traverse_data_t *td = arg;
if (lrc->lrc_txtype == TX_WRITE) {
lr_write_t *lr = (lr_write_t *)lrc;
blkptr_t *bp = &lr->lr_blkptr;
zbookmark_phys_t zb;
if (BP_IS_HOLE(bp))
return (0);
if (claim_txg == 0 || bp->blk_birth < claim_txg)
return (0);
SET_BOOKMARK(&zb, td->td_objset, lr->lr_foid,
ZB_ZIL_LEVEL, lr->lr_offset / BP_GET_LSIZE(bp));
(void) td->td_func(td->td_spa, zilog, bp, &zb, NULL,
td->td_arg);
}
return (0);
}
static void
traverse_zil(traverse_data_t *td, zil_header_t *zh)
{
uint64_t claim_txg = zh->zh_claim_txg;
/*
* We only want to visit blocks that have been claimed but not yet
* replayed; plus blocks that are already stable in read-only mode.
*/
if (claim_txg == 0 && spa_writeable(td->td_spa))
return;
zilog_t *zilog = zil_alloc(spa_get_dsl(td->td_spa)->dp_meta_objset, zh);
(void) zil_parse(zilog, traverse_zil_block, traverse_zil_record, td,
claim_txg, !(td->td_flags & TRAVERSE_NO_DECRYPT));
zil_free(zilog);
}
typedef enum resume_skip {
RESUME_SKIP_ALL,
RESUME_SKIP_NONE,
RESUME_SKIP_CHILDREN
} resume_skip_t;
/*
* Returns RESUME_SKIP_ALL if td indicates that we are resuming a traversal and
* the block indicated by zb does not need to be visited at all. Returns
* RESUME_SKIP_CHILDREN if we are resuming a post traversal and we reach the
* resume point. This indicates that this block should be visited but not its
* children (since they must have been visited in a previous traversal).
* Otherwise returns RESUME_SKIP_NONE.
*/
static resume_skip_t
resume_skip_check(traverse_data_t *td, const dnode_phys_t *dnp,
const zbookmark_phys_t *zb)
{
if (td->td_resume != NULL && !ZB_IS_ZERO(td->td_resume)) {
/*
* If we already visited this bp & everything below,
* don't bother doing it again.
*/
if (zbookmark_subtree_completed(dnp, zb, td->td_resume))
return (RESUME_SKIP_ALL);
/*
* If we found the block we're trying to resume from, zero
* the bookmark out to indicate that we have resumed.
*/
if (bcmp(zb, td->td_resume, sizeof (*zb)) == 0) {
bzero(td->td_resume, sizeof (*zb));
if (td->td_flags & TRAVERSE_POST)
return (RESUME_SKIP_CHILDREN);
}
}
return (RESUME_SKIP_NONE);
}
/*
* Returns B_TRUE, if prefetch read is issued, otherwise B_FALSE.
*/
static boolean_t
traverse_prefetch_metadata(traverse_data_t *td,
const blkptr_t *bp, const zbookmark_phys_t *zb)
{
arc_flags_t flags = ARC_FLAG_NOWAIT | ARC_FLAG_PREFETCH;
int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE;
if (!(td->td_flags & TRAVERSE_PREFETCH_METADATA))
return (B_FALSE);
/*
* If we are in the process of resuming, don't prefetch, because
* some children will not be needed (and in fact may have already
* been freed).
*/
if (td->td_resume != NULL && !ZB_IS_ZERO(td->td_resume))
return (B_FALSE);
if (BP_IS_HOLE(bp) || bp->blk_birth <= td->td_min_txg)
return (B_FALSE);
if (BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_DNODE)
return (B_FALSE);
ASSERT(!BP_IS_REDACTED(bp));
if ((td->td_flags & TRAVERSE_NO_DECRYPT) && BP_IS_PROTECTED(bp))
zio_flags |= ZIO_FLAG_RAW;
(void) arc_read(NULL, td->td_spa, bp, NULL, NULL,
ZIO_PRIORITY_ASYNC_READ, zio_flags, &flags, zb);
return (B_TRUE);
}
static boolean_t
prefetch_needed(prefetch_data_t *pfd, const blkptr_t *bp)
{
ASSERT(pfd->pd_flags & TRAVERSE_PREFETCH_DATA);
if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp) ||
BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG || BP_IS_REDACTED(bp))
return (B_FALSE);
return (B_TRUE);
}
static int
traverse_visitbp(traverse_data_t *td, const dnode_phys_t *dnp,
const blkptr_t *bp, const zbookmark_phys_t *zb)
{
int err = 0;
arc_buf_t *buf = NULL;
prefetch_data_t *pd = td->td_pfd;
switch (resume_skip_check(td, dnp, zb)) {
case RESUME_SKIP_ALL:
return (0);
case RESUME_SKIP_CHILDREN:
goto post;
case RESUME_SKIP_NONE:
break;
default:
ASSERT(0);
}
if (bp->blk_birth == 0) {
/*
* Since this block has a birth time of 0 it must be one of
* two things: a hole created before the
* SPA_FEATURE_HOLE_BIRTH feature was enabled, or a hole
* which has always been a hole in an object.
*
* If a file is written sparsely, then the unwritten parts of
* the file were "always holes" -- that is, they have been
* holes since this object was allocated. However, we (and
* our callers) can not necessarily tell when an object was
* allocated. Therefore, if it's possible that this object
* was freed and then its object number reused, we need to
* visit all the holes with birth==0.
*
* If it isn't possible that the object number was reused,
* then if SPA_FEATURE_HOLE_BIRTH was enabled before we wrote
* all the blocks we will visit as part of this traversal,
* then this hole must have always existed, so we can skip
* it. We visit blocks born after (exclusive) td_min_txg.
*
* Note that the meta-dnode cannot be reallocated.
*/
if (!send_holes_without_birth_time &&
(!td->td_realloc_possible ||
zb->zb_object == DMU_META_DNODE_OBJECT) &&
td->td_hole_birth_enabled_txg <= td->td_min_txg)
return (0);
} else if (bp->blk_birth <= td->td_min_txg) {
return (0);
}
if (pd != NULL && !pd->pd_exited && prefetch_needed(pd, bp)) {
uint64_t size = BP_GET_LSIZE(bp);
mutex_enter(&pd->pd_mtx);
ASSERT(pd->pd_bytes_fetched >= 0);
while (pd->pd_bytes_fetched < size && !pd->pd_exited)
cv_wait_sig(&pd->pd_cv, &pd->pd_mtx);
pd->pd_bytes_fetched -= size;
cv_broadcast(&pd->pd_cv);
mutex_exit(&pd->pd_mtx);
}
if (BP_IS_HOLE(bp) || BP_IS_REDACTED(bp)) {
err = td->td_func(td->td_spa, NULL, bp, zb, dnp, td->td_arg);
if (err != 0)
goto post;
return (0);
}
if (td->td_flags & TRAVERSE_PRE) {
err = td->td_func(td->td_spa, NULL, bp, zb, dnp,
td->td_arg);
if (err == TRAVERSE_VISIT_NO_CHILDREN)
return (0);
if (err != 0)
goto post;
}
if (BP_GET_LEVEL(bp) > 0) {
uint32_t flags = ARC_FLAG_WAIT;
int32_t i, ptidx, pidx;
uint32_t prefetchlimit;
int32_t epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
zbookmark_phys_t *czb;
ASSERT(!BP_IS_PROTECTED(bp));
err = arc_read(NULL, td->td_spa, bp, arc_getbuf_func, &buf,
ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
if (err != 0)
goto post;
czb = kmem_alloc(sizeof (zbookmark_phys_t), KM_SLEEP);
/*
* When performing a traversal it is beneficial to
* asynchronously read-ahead the upcoming indirect
* blocks since they will be needed shortly. However,
* since a 128k indirect (non-L0) block may contain up
* to 1024 128-byte block pointers, its preferable to not
* prefetch them all at once. Issuing a large number of
* async reads may effect performance, and the earlier
* the indirect blocks are prefetched the less likely
* they are to still be resident in the ARC when needed.
* Therefore, prefetching indirect blocks is limited to
* zfs_traverse_indirect_prefetch_limit=32 blocks by
* default.
*
* pidx: Index for which next prefetch to be issued.
* ptidx: Index at which next prefetch to be triggered.
*/
ptidx = 0;
pidx = 1;
prefetchlimit = zfs_traverse_indirect_prefetch_limit;
for (i = 0; i < epb; i++) {
if (prefetchlimit && i == ptidx) {
ASSERT3S(ptidx, <=, pidx);
for (uint32_t prefetched = 0; pidx < epb &&
prefetched < prefetchlimit; pidx++) {
SET_BOOKMARK(czb, zb->zb_objset,
zb->zb_object, zb->zb_level - 1,
zb->zb_blkid * epb + pidx);
if (traverse_prefetch_metadata(td,
&((blkptr_t *)buf->b_data)[pidx],
czb) == B_TRUE) {
prefetched++;
if (prefetched ==
MAX(prefetchlimit / 2, 1))
ptidx = pidx;
}
}
}
/* recursively visitbp() blocks below this */
SET_BOOKMARK(czb, zb->zb_objset, zb->zb_object,
zb->zb_level - 1,
zb->zb_blkid * epb + i);
err = traverse_visitbp(td, dnp,
&((blkptr_t *)buf->b_data)[i], czb);
if (err != 0)
break;
}
kmem_free(czb, sizeof (zbookmark_phys_t));
} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
uint32_t flags = ARC_FLAG_WAIT;
uint32_t zio_flags = ZIO_FLAG_CANFAIL;
int32_t i;
int32_t epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
dnode_phys_t *child_dnp;
/*
* dnode blocks might have their bonus buffers encrypted, so
* we must be careful to honor TRAVERSE_NO_DECRYPT
*/
if ((td->td_flags & TRAVERSE_NO_DECRYPT) && BP_IS_PROTECTED(bp))
zio_flags |= ZIO_FLAG_RAW;
err = arc_read(NULL, td->td_spa, bp, arc_getbuf_func, &buf,
ZIO_PRIORITY_ASYNC_READ, zio_flags, &flags, zb);
if (err != 0)
goto post;
child_dnp = buf->b_data;
for (i = 0; i < epb; i += child_dnp[i].dn_extra_slots + 1) {
prefetch_dnode_metadata(td, &child_dnp[i],
zb->zb_objset, zb->zb_blkid * epb + i);
}
/* recursively visitbp() blocks below this */
for (i = 0; i < epb; i += child_dnp[i].dn_extra_slots + 1) {
err = traverse_dnode(td, bp, &child_dnp[i],
zb->zb_objset, zb->zb_blkid * epb + i);
if (err != 0)
break;
}
} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
uint32_t zio_flags = ZIO_FLAG_CANFAIL;
arc_flags_t flags = ARC_FLAG_WAIT;
objset_phys_t *osp;
if ((td->td_flags & TRAVERSE_NO_DECRYPT) && BP_IS_PROTECTED(bp))
zio_flags |= ZIO_FLAG_RAW;
err = arc_read(NULL, td->td_spa, bp, arc_getbuf_func, &buf,
ZIO_PRIORITY_ASYNC_READ, zio_flags, &flags, zb);
if (err != 0)
goto post;
osp = buf->b_data;
prefetch_dnode_metadata(td, &osp->os_meta_dnode, zb->zb_objset,
DMU_META_DNODE_OBJECT);
/*
* See the block comment above for the goal of this variable.
* If the maxblkid of the meta-dnode is 0, then we know that
* we've never had more than DNODES_PER_BLOCK objects in the
* dataset, which means we can't have reused any object ids.
*/
if (osp->os_meta_dnode.dn_maxblkid == 0)
td->td_realloc_possible = B_FALSE;
if (OBJSET_BUF_HAS_USERUSED(buf)) {
if (OBJSET_BUF_HAS_PROJECTUSED(buf))
prefetch_dnode_metadata(td,
&osp->os_projectused_dnode,
zb->zb_objset, DMU_PROJECTUSED_OBJECT);
prefetch_dnode_metadata(td, &osp->os_groupused_dnode,
zb->zb_objset, DMU_GROUPUSED_OBJECT);
prefetch_dnode_metadata(td, &osp->os_userused_dnode,
zb->zb_objset, DMU_USERUSED_OBJECT);
}
err = traverse_dnode(td, bp, &osp->os_meta_dnode, zb->zb_objset,
DMU_META_DNODE_OBJECT);
if (err == 0 && OBJSET_BUF_HAS_USERUSED(buf)) {
if (OBJSET_BUF_HAS_PROJECTUSED(buf))
err = traverse_dnode(td, bp,
&osp->os_projectused_dnode, zb->zb_objset,
DMU_PROJECTUSED_OBJECT);
if (err == 0)
err = traverse_dnode(td, bp,
&osp->os_groupused_dnode, zb->zb_objset,
DMU_GROUPUSED_OBJECT);
if (err == 0)
err = traverse_dnode(td, bp,
&osp->os_userused_dnode, zb->zb_objset,
DMU_USERUSED_OBJECT);
}
}
if (buf)
arc_buf_destroy(buf, &buf);
post:
if (err == 0 && (td->td_flags & TRAVERSE_POST))
err = td->td_func(td->td_spa, NULL, bp, zb, dnp, td->td_arg);
if ((td->td_flags & TRAVERSE_HARD) && (err == EIO || err == ECKSUM)) {
/*
* Ignore this disk error as requested by the HARD flag,
* and continue traversal.
*/
err = 0;
}
/*
* If we are stopping here, set td_resume.
*/
if (td->td_resume != NULL && err != 0 && !td->td_paused) {
td->td_resume->zb_objset = zb->zb_objset;
td->td_resume->zb_object = zb->zb_object;
td->td_resume->zb_level = 0;
/*
* If we have stopped on an indirect block (e.g. due to
* i/o error), we have not visited anything below it.
* Set the bookmark to the first level-0 block that we need
* to visit. This way, the resuming code does not need to
* deal with resuming from indirect blocks.
*
* Note, if zb_level <= 0, dnp may be NULL, so we don't want
* to dereference it.
*/
td->td_resume->zb_blkid = zb->zb_blkid;
if (zb->zb_level > 0) {
td->td_resume->zb_blkid <<= zb->zb_level *
(dnp->dn_indblkshift - SPA_BLKPTRSHIFT);
}
td->td_paused = B_TRUE;
}
return (err);
}
static void
prefetch_dnode_metadata(traverse_data_t *td, const dnode_phys_t *dnp,
uint64_t objset, uint64_t object)
{
int j;
zbookmark_phys_t czb;
for (j = 0; j < dnp->dn_nblkptr; j++) {
SET_BOOKMARK(&czb, objset, object, dnp->dn_nlevels - 1, j);
traverse_prefetch_metadata(td, &dnp->dn_blkptr[j], &czb);
}
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
SET_BOOKMARK(&czb, objset, object, 0, DMU_SPILL_BLKID);
traverse_prefetch_metadata(td, DN_SPILL_BLKPTR(dnp), &czb);
}
}
static int
traverse_dnode(traverse_data_t *td, const blkptr_t *bp, const dnode_phys_t *dnp,
uint64_t objset, uint64_t object)
{
int j, err = 0;
zbookmark_phys_t czb;
if (object != DMU_META_DNODE_OBJECT && td->td_resume != NULL &&
object < td->td_resume->zb_object)
return (0);
if (td->td_flags & TRAVERSE_PRE) {
SET_BOOKMARK(&czb, objset, object, ZB_DNODE_LEVEL,
ZB_DNODE_BLKID);
err = td->td_func(td->td_spa, NULL, bp, &czb, dnp,
td->td_arg);
if (err == TRAVERSE_VISIT_NO_CHILDREN)
return (0);
if (err != 0)
return (err);
}
for (j = 0; j < dnp->dn_nblkptr; j++) {
SET_BOOKMARK(&czb, objset, object, dnp->dn_nlevels - 1, j);
err = traverse_visitbp(td, dnp, &dnp->dn_blkptr[j], &czb);
if (err != 0)
break;
}
if (err == 0 && (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) {
SET_BOOKMARK(&czb, objset, object, 0, DMU_SPILL_BLKID);
err = traverse_visitbp(td, dnp, DN_SPILL_BLKPTR(dnp), &czb);
}
if (err == 0 && (td->td_flags & TRAVERSE_POST)) {
SET_BOOKMARK(&czb, objset, object, ZB_DNODE_LEVEL,
ZB_DNODE_BLKID);
err = td->td_func(td->td_spa, NULL, bp, &czb, dnp,
td->td_arg);
if (err == TRAVERSE_VISIT_NO_CHILDREN)
return (0);
if (err != 0)
return (err);
}
return (err);
}
static int
traverse_prefetcher(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
const zbookmark_phys_t *zb, const dnode_phys_t *dnp, void *arg)
{
(void) zilog, (void) dnp;
prefetch_data_t *pfd = arg;
int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE;
arc_flags_t aflags = ARC_FLAG_NOWAIT | ARC_FLAG_PREFETCH |
ARC_FLAG_PRESCIENT_PREFETCH;
ASSERT(pfd->pd_bytes_fetched >= 0);
if (zb->zb_level == ZB_DNODE_LEVEL)
return (0);
if (pfd->pd_cancel)
return (SET_ERROR(EINTR));
if (!prefetch_needed(pfd, bp))
return (0);
mutex_enter(&pfd->pd_mtx);
while (!pfd->pd_cancel && pfd->pd_bytes_fetched >= zfs_pd_bytes_max)
cv_wait_sig(&pfd->pd_cv, &pfd->pd_mtx);
pfd->pd_bytes_fetched += BP_GET_LSIZE(bp);
cv_broadcast(&pfd->pd_cv);
mutex_exit(&pfd->pd_mtx);
if ((pfd->pd_flags & TRAVERSE_NO_DECRYPT) && BP_IS_PROTECTED(bp))
zio_flags |= ZIO_FLAG_RAW;
(void) arc_read(NULL, spa, bp, NULL, NULL, ZIO_PRIORITY_ASYNC_READ,
zio_flags, &aflags, zb);
return (0);
}
static void
traverse_prefetch_thread(void *arg)
{
traverse_data_t *td_main = arg;
traverse_data_t td = *td_main;
zbookmark_phys_t czb;
fstrans_cookie_t cookie = spl_fstrans_mark();
td.td_func = traverse_prefetcher;
td.td_arg = td_main->td_pfd;
td.td_pfd = NULL;
td.td_resume = &td_main->td_pfd->pd_resume;
SET_BOOKMARK(&czb, td.td_objset,
ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
(void) traverse_visitbp(&td, NULL, td.td_rootbp, &czb);
mutex_enter(&td_main->td_pfd->pd_mtx);
td_main->td_pfd->pd_exited = B_TRUE;
cv_broadcast(&td_main->td_pfd->pd_cv);
mutex_exit(&td_main->td_pfd->pd_mtx);
spl_fstrans_unmark(cookie);
}
/*
* NB: dataset must not be changing on-disk (eg, is a snapshot or we are
* in syncing context).
*/
static int
traverse_impl(spa_t *spa, dsl_dataset_t *ds, uint64_t objset, blkptr_t *rootbp,
uint64_t txg_start, zbookmark_phys_t *resume, int flags,
blkptr_cb_t func, void *arg)
{
traverse_data_t *td;
prefetch_data_t *pd;
zbookmark_phys_t *czb;
int err;
ASSERT(ds == NULL || objset == ds->ds_object);
ASSERT(!(flags & TRAVERSE_PRE) || !(flags & TRAVERSE_POST));
td = kmem_alloc(sizeof (traverse_data_t), KM_SLEEP);
pd = kmem_zalloc(sizeof (prefetch_data_t), KM_SLEEP);
czb = kmem_alloc(sizeof (zbookmark_phys_t), KM_SLEEP);
td->td_spa = spa;
td->td_objset = objset;
td->td_rootbp = rootbp;
td->td_min_txg = txg_start;
td->td_resume = resume;
td->td_func = func;
td->td_arg = arg;
td->td_pfd = pd;
td->td_flags = flags;
td->td_paused = B_FALSE;
td->td_realloc_possible = (txg_start == 0 ? B_FALSE : B_TRUE);
if (spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) {
VERIFY(spa_feature_enabled_txg(spa,
SPA_FEATURE_HOLE_BIRTH, &td->td_hole_birth_enabled_txg));
} else {
td->td_hole_birth_enabled_txg = UINT64_MAX;
}
pd->pd_flags = flags;
if (resume != NULL)
pd->pd_resume = *resume;
mutex_init(&pd->pd_mtx, NULL, MUTEX_DEFAULT, NULL);
cv_init(&pd->pd_cv, NULL, CV_DEFAULT, NULL);
SET_BOOKMARK(czb, td->td_objset,
ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
/* See comment on ZIL traversal in dsl_scan_visitds. */
if (ds != NULL && !ds->ds_is_snapshot && !BP_IS_HOLE(rootbp)) {
enum zio_flag zio_flags = ZIO_FLAG_CANFAIL;
uint32_t flags = ARC_FLAG_WAIT;
objset_phys_t *osp;
arc_buf_t *buf;
ASSERT(!BP_IS_REDACTED(rootbp));
if ((td->td_flags & TRAVERSE_NO_DECRYPT) &&
BP_IS_PROTECTED(rootbp))
zio_flags |= ZIO_FLAG_RAW;
err = arc_read(NULL, td->td_spa, rootbp, arc_getbuf_func,
&buf, ZIO_PRIORITY_ASYNC_READ, zio_flags, &flags, czb);
if (err != 0) {
/*
* If both TRAVERSE_HARD and TRAVERSE_PRE are set,
* continue to visitbp so that td_func can be called
* in pre stage, and err will reset to zero.
*/
if (!(td->td_flags & TRAVERSE_HARD) ||
!(td->td_flags & TRAVERSE_PRE))
goto out;
} else {
osp = buf->b_data;
traverse_zil(td, &osp->os_zil_header);
arc_buf_destroy(buf, &buf);
}
}
if (!(flags & TRAVERSE_PREFETCH_DATA) ||
taskq_dispatch(spa->spa_prefetch_taskq, traverse_prefetch_thread,
td, TQ_NOQUEUE) == TASKQID_INVALID)
pd->pd_exited = B_TRUE;
err = traverse_visitbp(td, NULL, rootbp, czb);
mutex_enter(&pd->pd_mtx);
pd->pd_cancel = B_TRUE;
cv_broadcast(&pd->pd_cv);
while (!pd->pd_exited)
cv_wait_sig(&pd->pd_cv, &pd->pd_mtx);
mutex_exit(&pd->pd_mtx);
out:
mutex_destroy(&pd->pd_mtx);
cv_destroy(&pd->pd_cv);
kmem_free(czb, sizeof (zbookmark_phys_t));
kmem_free(pd, sizeof (struct prefetch_data));
kmem_free(td, sizeof (struct traverse_data));
return (err);
}
/*
* NB: dataset must not be changing on-disk (eg, is a snapshot or we are
* in syncing context).
*/
int
traverse_dataset_resume(dsl_dataset_t *ds, uint64_t txg_start,
zbookmark_phys_t *resume,
int flags, blkptr_cb_t func, void *arg)
{
return (traverse_impl(ds->ds_dir->dd_pool->dp_spa, ds, ds->ds_object,
&dsl_dataset_phys(ds)->ds_bp, txg_start, resume, flags, func, arg));
}
int
traverse_dataset(dsl_dataset_t *ds, uint64_t txg_start,
int flags, blkptr_cb_t func, void *arg)
{
return (traverse_dataset_resume(ds, txg_start, NULL, flags, func, arg));
}
int
traverse_dataset_destroyed(spa_t *spa, blkptr_t *blkptr,
uint64_t txg_start, zbookmark_phys_t *resume, int flags,
blkptr_cb_t func, void *arg)
{
return (traverse_impl(spa, NULL, ZB_DESTROYED_OBJSET,
blkptr, txg_start, resume, flags, func, arg));
}
/*
* NB: pool must not be changing on-disk (eg, from zdb or sync context).
*/
int
traverse_pool(spa_t *spa, uint64_t txg_start, int flags,
blkptr_cb_t func, void *arg)
{
int err;
dsl_pool_t *dp = spa_get_dsl(spa);
objset_t *mos = dp->dp_meta_objset;
boolean_t hard = (flags & TRAVERSE_HARD);
/* visit the MOS */
err = traverse_impl(spa, NULL, 0, spa_get_rootblkptr(spa),
txg_start, NULL, flags, func, arg);
if (err != 0)
return (err);
/* visit each dataset */
for (uint64_t obj = 1; err == 0;
err = dmu_object_next(mos, &obj, B_FALSE, txg_start)) {
dmu_object_info_t doi;
err = dmu_object_info(mos, obj, &doi);
if (err != 0) {
if (hard)
continue;
break;
}
if (doi.doi_bonus_type == DMU_OT_DSL_DATASET) {
dsl_dataset_t *ds;
uint64_t txg = txg_start;
dsl_pool_config_enter(dp, FTAG);
err = dsl_dataset_hold_obj(dp, obj, FTAG, &ds);
dsl_pool_config_exit(dp, FTAG);
if (err != 0) {
if (hard)
continue;
break;
}
if (dsl_dataset_phys(ds)->ds_prev_snap_txg > txg)
txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;
err = traverse_dataset(ds, txg, flags, func, arg);
dsl_dataset_rele(ds, FTAG);
if (err != 0)
break;
}
}
if (err == ESRCH)
err = 0;
return (err);
}
EXPORT_SYMBOL(traverse_dataset);
EXPORT_SYMBOL(traverse_pool);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, pd_bytes_max, INT, ZMOD_RW,
"Max number of bytes to prefetch");
ZFS_MODULE_PARAM(zfs, zfs_, traverse_indirect_prefetch_limit, INT, ZMOD_RW,
"Traverse prefetch number of blocks pointed by indirect block");
#if defined(_KERNEL)
module_param_named(ignore_hole_birth, send_holes_without_birth_time, int, 0644);
MODULE_PARM_DESC(ignore_hole_birth,
"Alias for send_holes_without_birth_time");
#endif
ZFS_MODULE_PARAM(zfs, , send_holes_without_birth_time, INT, ZMOD_RW,
"Ignore hole_birth txg for zfs send");
/* END CSTYLED */
diff --git a/module/zfs/dmu_tx.c b/module/zfs/dmu_tx.c
index 5fa516866668..b4735bb7ff54 100644
--- a/module/zfs/dmu_tx.c
+++ b/module/zfs/dmu_tx.c
@@ -1,1425 +1,1424 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2012, 2017 by Delphix. All rights reserved.
*/
#include <sys/dmu.h>
#include <sys/dmu_impl.h>
#include <sys/dbuf.h>
#include <sys/dmu_tx.h>
#include <sys/dmu_objset.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_pool.h>
#include <sys/zap_impl.h>
#include <sys/spa.h>
#include <sys/sa.h>
#include <sys/sa_impl.h>
#include <sys/zfs_context.h>
#include <sys/trace_zfs.h>
typedef void (*dmu_tx_hold_func_t)(dmu_tx_t *tx, struct dnode *dn,
uint64_t arg1, uint64_t arg2);
dmu_tx_stats_t dmu_tx_stats = {
{ "dmu_tx_assigned", KSTAT_DATA_UINT64 },
{ "dmu_tx_delay", KSTAT_DATA_UINT64 },
{ "dmu_tx_error", KSTAT_DATA_UINT64 },
{ "dmu_tx_suspended", KSTAT_DATA_UINT64 },
{ "dmu_tx_group", KSTAT_DATA_UINT64 },
{ "dmu_tx_memory_reserve", KSTAT_DATA_UINT64 },
{ "dmu_tx_memory_reclaim", KSTAT_DATA_UINT64 },
{ "dmu_tx_dirty_throttle", KSTAT_DATA_UINT64 },
{ "dmu_tx_dirty_delay", KSTAT_DATA_UINT64 },
{ "dmu_tx_dirty_over_max", KSTAT_DATA_UINT64 },
{ "dmu_tx_wrlog_over_max", KSTAT_DATA_UINT64 },
{ "dmu_tx_dirty_frees_delay", KSTAT_DATA_UINT64 },
{ "dmu_tx_quota", KSTAT_DATA_UINT64 },
};
static kstat_t *dmu_tx_ksp;
dmu_tx_t *
dmu_tx_create_dd(dsl_dir_t *dd)
{
dmu_tx_t *tx = kmem_zalloc(sizeof (dmu_tx_t), KM_SLEEP);
tx->tx_dir = dd;
if (dd != NULL)
tx->tx_pool = dd->dd_pool;
list_create(&tx->tx_holds, sizeof (dmu_tx_hold_t),
offsetof(dmu_tx_hold_t, txh_node));
list_create(&tx->tx_callbacks, sizeof (dmu_tx_callback_t),
offsetof(dmu_tx_callback_t, dcb_node));
tx->tx_start = gethrtime();
return (tx);
}
dmu_tx_t *
dmu_tx_create(objset_t *os)
{
dmu_tx_t *tx = dmu_tx_create_dd(os->os_dsl_dataset->ds_dir);
tx->tx_objset = os;
return (tx);
}
dmu_tx_t *
dmu_tx_create_assigned(struct dsl_pool *dp, uint64_t txg)
{
dmu_tx_t *tx = dmu_tx_create_dd(NULL);
TXG_VERIFY(dp->dp_spa, txg);
tx->tx_pool = dp;
tx->tx_txg = txg;
tx->tx_anyobj = TRUE;
return (tx);
}
int
dmu_tx_is_syncing(dmu_tx_t *tx)
{
return (tx->tx_anyobj);
}
int
dmu_tx_private_ok(dmu_tx_t *tx)
{
return (tx->tx_anyobj);
}
static dmu_tx_hold_t *
dmu_tx_hold_dnode_impl(dmu_tx_t *tx, dnode_t *dn, enum dmu_tx_hold_type type,
uint64_t arg1, uint64_t arg2)
{
dmu_tx_hold_t *txh;
if (dn != NULL) {
(void) zfs_refcount_add(&dn->dn_holds, tx);
if (tx->tx_txg != 0) {
mutex_enter(&dn->dn_mtx);
/*
* dn->dn_assigned_txg == tx->tx_txg doesn't pose a
* problem, but there's no way for it to happen (for
* now, at least).
*/
ASSERT(dn->dn_assigned_txg == 0);
dn->dn_assigned_txg = tx->tx_txg;
(void) zfs_refcount_add(&dn->dn_tx_holds, tx);
mutex_exit(&dn->dn_mtx);
}
}
txh = kmem_zalloc(sizeof (dmu_tx_hold_t), KM_SLEEP);
txh->txh_tx = tx;
txh->txh_dnode = dn;
zfs_refcount_create(&txh->txh_space_towrite);
zfs_refcount_create(&txh->txh_memory_tohold);
txh->txh_type = type;
txh->txh_arg1 = arg1;
txh->txh_arg2 = arg2;
list_insert_tail(&tx->tx_holds, txh);
return (txh);
}
static dmu_tx_hold_t *
dmu_tx_hold_object_impl(dmu_tx_t *tx, objset_t *os, uint64_t object,
enum dmu_tx_hold_type type, uint64_t arg1, uint64_t arg2)
{
dnode_t *dn = NULL;
dmu_tx_hold_t *txh;
int err;
if (object != DMU_NEW_OBJECT) {
err = dnode_hold(os, object, FTAG, &dn);
if (err != 0) {
tx->tx_err = err;
return (NULL);
}
}
txh = dmu_tx_hold_dnode_impl(tx, dn, type, arg1, arg2);
if (dn != NULL)
dnode_rele(dn, FTAG);
return (txh);
}
void
dmu_tx_add_new_object(dmu_tx_t *tx, dnode_t *dn)
{
/*
* If we're syncing, they can manipulate any object anyhow, and
* the hold on the dnode_t can cause problems.
*/
if (!dmu_tx_is_syncing(tx))
(void) dmu_tx_hold_dnode_impl(tx, dn, THT_NEWOBJECT, 0, 0);
}
/*
* This function reads specified data from disk. The specified data will
* be needed to perform the transaction -- i.e, it will be read after
* we do dmu_tx_assign(). There are two reasons that we read the data now
* (before dmu_tx_assign()):
*
* 1. Reading it now has potentially better performance. The transaction
* has not yet been assigned, so the TXG is not held open, and also the
* caller typically has less locks held when calling dmu_tx_hold_*() than
* after the transaction has been assigned. This reduces the lock (and txg)
* hold times, thus reducing lock contention.
*
* 2. It is easier for callers (primarily the ZPL) to handle i/o errors
* that are detected before they start making changes to the DMU state
* (i.e. now). Once the transaction has been assigned, and some DMU
* state has been changed, it can be difficult to recover from an i/o
* error (e.g. to undo the changes already made in memory at the DMU
* layer). Typically code to do so does not exist in the caller -- it
* assumes that the data has already been cached and thus i/o errors are
* not possible.
*
* It has been observed that the i/o initiated here can be a performance
* problem, and it appears to be optional, because we don't look at the
* data which is read. However, removing this read would only serve to
* move the work elsewhere (after the dmu_tx_assign()), where it may
* have a greater impact on performance (in addition to the impact on
* fault tolerance noted above).
*/
static int
dmu_tx_check_ioerr(zio_t *zio, dnode_t *dn, int level, uint64_t blkid)
{
int err;
dmu_buf_impl_t *db;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
db = dbuf_hold_level(dn, level, blkid, FTAG);
rw_exit(&dn->dn_struct_rwlock);
if (db == NULL)
return (SET_ERROR(EIO));
err = dbuf_read(db, zio, DB_RF_CANFAIL | DB_RF_NOPREFETCH);
dbuf_rele(db, FTAG);
return (err);
}
/* ARGSUSED */
static void
dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
{
dnode_t *dn = txh->txh_dnode;
int err = 0;
if (len == 0)
return;
(void) zfs_refcount_add_many(&txh->txh_space_towrite, len, FTAG);
if (dn == NULL)
return;
/*
* For i/o error checking, read the blocks that will be needed
* to perform the write: the first and last level-0 blocks (if
* they are not aligned, i.e. if they are partial-block writes),
* and all the level-1 blocks.
*/
if (dn->dn_maxblkid == 0) {
if (off < dn->dn_datablksz &&
(off > 0 || len < dn->dn_datablksz)) {
err = dmu_tx_check_ioerr(NULL, dn, 0, 0);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
} else {
zio_t *zio = zio_root(dn->dn_objset->os_spa,
NULL, NULL, ZIO_FLAG_CANFAIL);
/* first level-0 block */
uint64_t start = off >> dn->dn_datablkshift;
if (P2PHASE(off, dn->dn_datablksz) || len < dn->dn_datablksz) {
err = dmu_tx_check_ioerr(zio, dn, 0, start);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
/* last level-0 block */
uint64_t end = (off + len - 1) >> dn->dn_datablkshift;
if (end != start && end <= dn->dn_maxblkid &&
P2PHASE(off + len, dn->dn_datablksz)) {
err = dmu_tx_check_ioerr(zio, dn, 0, end);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
/* level-1 blocks */
if (dn->dn_nlevels > 1) {
int shft = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
for (uint64_t i = (start >> shft) + 1;
i < end >> shft; i++) {
err = dmu_tx_check_ioerr(zio, dn, 1, i);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
}
err = zio_wait(zio);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
}
static void
dmu_tx_count_dnode(dmu_tx_hold_t *txh)
{
(void) zfs_refcount_add_many(&txh->txh_space_towrite,
DNODE_MIN_SIZE, FTAG);
}
void
dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len)
{
dmu_tx_hold_t *txh;
ASSERT0(tx->tx_txg);
ASSERT3U(len, <=, DMU_MAX_ACCESS);
ASSERT(len == 0 || UINT64_MAX - off >= len - 1);
txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
object, THT_WRITE, off, len);
if (txh != NULL) {
dmu_tx_count_write(txh, off, len);
dmu_tx_count_dnode(txh);
}
}
void
dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, int len)
{
dmu_tx_hold_t *txh;
ASSERT0(tx->tx_txg);
ASSERT3U(len, <=, DMU_MAX_ACCESS);
ASSERT(len == 0 || UINT64_MAX - off >= len - 1);
txh = dmu_tx_hold_dnode_impl(tx, dn, THT_WRITE, off, len);
if (txh != NULL) {
dmu_tx_count_write(txh, off, len);
dmu_tx_count_dnode(txh);
}
}
/*
* This function marks the transaction as being a "net free". The end
* result is that refquotas will be disabled for this transaction, and
* this transaction will be able to use half of the pool space overhead
* (see dsl_pool_adjustedsize()). Therefore this function should only
* be called for transactions that we expect will not cause a net increase
* in the amount of space used (but it's OK if that is occasionally not true).
*/
void
dmu_tx_mark_netfree(dmu_tx_t *tx)
{
tx->tx_netfree = B_TRUE;
}
static void
dmu_tx_hold_free_impl(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
{
dmu_tx_t *tx = txh->txh_tx;
dnode_t *dn = txh->txh_dnode;
int err;
ASSERT(tx->tx_txg == 0);
dmu_tx_count_dnode(txh);
if (off >= (dn->dn_maxblkid + 1) * dn->dn_datablksz)
return;
if (len == DMU_OBJECT_END)
len = (dn->dn_maxblkid + 1) * dn->dn_datablksz - off;
dmu_tx_count_dnode(txh);
/*
* For i/o error checking, we read the first and last level-0
* blocks if they are not aligned, and all the level-1 blocks.
*
* Note: dbuf_free_range() assumes that we have not instantiated
* any level-0 dbufs that will be completely freed. Therefore we must
* exercise care to not read or count the first and last blocks
* if they are blocksize-aligned.
*/
if (dn->dn_datablkshift == 0) {
if (off != 0 || len < dn->dn_datablksz)
dmu_tx_count_write(txh, 0, dn->dn_datablksz);
} else {
/* first block will be modified if it is not aligned */
if (!IS_P2ALIGNED(off, 1 << dn->dn_datablkshift))
dmu_tx_count_write(txh, off, 1);
/* last block will be modified if it is not aligned */
if (!IS_P2ALIGNED(off + len, 1 << dn->dn_datablkshift))
dmu_tx_count_write(txh, off + len, 1);
}
/*
* Check level-1 blocks.
*/
if (dn->dn_nlevels > 1) {
int shift = dn->dn_datablkshift + dn->dn_indblkshift -
SPA_BLKPTRSHIFT;
uint64_t start = off >> shift;
uint64_t end = (off + len) >> shift;
ASSERT(dn->dn_indblkshift != 0);
/*
* dnode_reallocate() can result in an object with indirect
* blocks having an odd data block size. In this case,
* just check the single block.
*/
if (dn->dn_datablkshift == 0)
start = end = 0;
zio_t *zio = zio_root(tx->tx_pool->dp_spa,
NULL, NULL, ZIO_FLAG_CANFAIL);
for (uint64_t i = start; i <= end; i++) {
uint64_t ibyte = i << shift;
err = dnode_next_offset(dn, 0, &ibyte, 2, 1, 0);
i = ibyte >> shift;
if (err == ESRCH || i > end)
break;
if (err != 0) {
tx->tx_err = err;
(void) zio_wait(zio);
return;
}
(void) zfs_refcount_add_many(&txh->txh_memory_tohold,
1 << dn->dn_indblkshift, FTAG);
err = dmu_tx_check_ioerr(zio, dn, 1, i);
if (err != 0) {
tx->tx_err = err;
(void) zio_wait(zio);
return;
}
}
err = zio_wait(zio);
if (err != 0) {
tx->tx_err = err;
return;
}
}
}
void
dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, uint64_t len)
{
dmu_tx_hold_t *txh;
txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
object, THT_FREE, off, len);
if (txh != NULL)
(void) dmu_tx_hold_free_impl(txh, off, len);
}
void
dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, uint64_t len)
{
dmu_tx_hold_t *txh;
txh = dmu_tx_hold_dnode_impl(tx, dn, THT_FREE, off, len);
if (txh != NULL)
(void) dmu_tx_hold_free_impl(txh, off, len);
}
static void
dmu_tx_hold_zap_impl(dmu_tx_hold_t *txh, const char *name)
{
dmu_tx_t *tx = txh->txh_tx;
dnode_t *dn = txh->txh_dnode;
int err;
ASSERT(tx->tx_txg == 0);
dmu_tx_count_dnode(txh);
/*
* Modifying a almost-full microzap is around the worst case (128KB)
*
* If it is a fat zap, the worst case would be 7*16KB=112KB:
* - 3 blocks overwritten: target leaf, ptrtbl block, header block
* - 4 new blocks written if adding:
* - 2 blocks for possibly split leaves,
* - 2 grown ptrtbl blocks
*/
(void) zfs_refcount_add_many(&txh->txh_space_towrite,
MZAP_MAX_BLKSZ, FTAG);
if (dn == NULL)
return;
ASSERT3U(DMU_OT_BYTESWAP(dn->dn_type), ==, DMU_BSWAP_ZAP);
if (dn->dn_maxblkid == 0 || name == NULL) {
/*
* This is a microzap (only one block), or we don't know
* the name. Check the first block for i/o errors.
*/
err = dmu_tx_check_ioerr(NULL, dn, 0, 0);
if (err != 0) {
tx->tx_err = err;
}
} else {
/*
* Access the name so that we'll check for i/o errors to
* the leaf blocks, etc. We ignore ENOENT, as this name
* may not yet exist.
*/
err = zap_lookup_by_dnode(dn, name, 8, 0, NULL);
if (err == EIO || err == ECKSUM || err == ENXIO) {
tx->tx_err = err;
}
}
}
void
dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name)
{
dmu_tx_hold_t *txh;
ASSERT0(tx->tx_txg);
txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
object, THT_ZAP, add, (uintptr_t)name);
if (txh != NULL)
dmu_tx_hold_zap_impl(txh, name);
}
void
dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add, const char *name)
{
dmu_tx_hold_t *txh;
ASSERT0(tx->tx_txg);
ASSERT(dn != NULL);
txh = dmu_tx_hold_dnode_impl(tx, dn, THT_ZAP, add, (uintptr_t)name);
if (txh != NULL)
dmu_tx_hold_zap_impl(txh, name);
}
void
dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object)
{
dmu_tx_hold_t *txh;
ASSERT(tx->tx_txg == 0);
txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
object, THT_BONUS, 0, 0);
if (txh)
dmu_tx_count_dnode(txh);
}
void
dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn)
{
dmu_tx_hold_t *txh;
ASSERT0(tx->tx_txg);
txh = dmu_tx_hold_dnode_impl(tx, dn, THT_BONUS, 0, 0);
if (txh)
dmu_tx_count_dnode(txh);
}
void
dmu_tx_hold_space(dmu_tx_t *tx, uint64_t space)
{
dmu_tx_hold_t *txh;
ASSERT(tx->tx_txg == 0);
txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
DMU_NEW_OBJECT, THT_SPACE, space, 0);
if (txh) {
(void) zfs_refcount_add_many(
&txh->txh_space_towrite, space, FTAG);
}
}
#ifdef ZFS_DEBUG
void
dmu_tx_dirty_buf(dmu_tx_t *tx, dmu_buf_impl_t *db)
{
boolean_t match_object = B_FALSE;
boolean_t match_offset = B_FALSE;
DB_DNODE_ENTER(db);
dnode_t *dn = DB_DNODE(db);
ASSERT(tx->tx_txg != 0);
ASSERT(tx->tx_objset == NULL || dn->dn_objset == tx->tx_objset);
ASSERT3U(dn->dn_object, ==, db->db.db_object);
if (tx->tx_anyobj) {
DB_DNODE_EXIT(db);
return;
}
/* XXX No checking on the meta dnode for now */
if (db->db.db_object == DMU_META_DNODE_OBJECT) {
DB_DNODE_EXIT(db);
return;
}
for (dmu_tx_hold_t *txh = list_head(&tx->tx_holds); txh != NULL;
txh = list_next(&tx->tx_holds, txh)) {
ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg);
if (txh->txh_dnode == dn && txh->txh_type != THT_NEWOBJECT)
match_object = TRUE;
if (txh->txh_dnode == NULL || txh->txh_dnode == dn) {
int datablkshift = dn->dn_datablkshift ?
dn->dn_datablkshift : SPA_MAXBLOCKSHIFT;
int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
int shift = datablkshift + epbs * db->db_level;
uint64_t beginblk = shift >= 64 ? 0 :
(txh->txh_arg1 >> shift);
uint64_t endblk = shift >= 64 ? 0 :
((txh->txh_arg1 + txh->txh_arg2 - 1) >> shift);
uint64_t blkid = db->db_blkid;
/* XXX txh_arg2 better not be zero... */
dprintf("found txh type %x beginblk=%llx endblk=%llx\n",
txh->txh_type, (u_longlong_t)beginblk,
(u_longlong_t)endblk);
switch (txh->txh_type) {
case THT_WRITE:
if (blkid >= beginblk && blkid <= endblk)
match_offset = TRUE;
/*
* We will let this hold work for the bonus
* or spill buffer so that we don't need to
* hold it when creating a new object.
*/
if (blkid == DMU_BONUS_BLKID ||
blkid == DMU_SPILL_BLKID)
match_offset = TRUE;
/*
* They might have to increase nlevels,
* thus dirtying the new TLIBs. Or the
* might have to change the block size,
* thus dirying the new lvl=0 blk=0.
*/
if (blkid == 0)
match_offset = TRUE;
break;
case THT_FREE:
/*
* We will dirty all the level 1 blocks in
* the free range and perhaps the first and
* last level 0 block.
*/
if (blkid >= beginblk && (blkid <= endblk ||
txh->txh_arg2 == DMU_OBJECT_END))
match_offset = TRUE;
break;
case THT_SPILL:
if (blkid == DMU_SPILL_BLKID)
match_offset = TRUE;
break;
case THT_BONUS:
if (blkid == DMU_BONUS_BLKID)
match_offset = TRUE;
break;
case THT_ZAP:
match_offset = TRUE;
break;
case THT_NEWOBJECT:
match_object = TRUE;
break;
default:
cmn_err(CE_PANIC, "bad txh_type %d",
txh->txh_type);
}
}
if (match_object && match_offset) {
DB_DNODE_EXIT(db);
return;
}
}
DB_DNODE_EXIT(db);
panic("dirtying dbuf obj=%llx lvl=%u blkid=%llx but not tx_held\n",
(u_longlong_t)db->db.db_object, db->db_level,
(u_longlong_t)db->db_blkid);
}
#endif
/*
* If we can't do 10 iops, something is wrong. Let us go ahead
* and hit zfs_dirty_data_max.
*/
-hrtime_t zfs_delay_max_ns = 100 * MICROSEC; /* 100 milliseconds */
-int zfs_delay_resolution_ns = 100 * 1000; /* 100 microseconds */
+static const hrtime_t zfs_delay_max_ns = 100 * MICROSEC; /* 100 milliseconds */
/*
* We delay transactions when we've determined that the backend storage
* isn't able to accommodate the rate of incoming writes.
*
* If there is already a transaction waiting, we delay relative to when
* that transaction finishes waiting. This way the calculated min_time
* is independent of the number of threads concurrently executing
* transactions.
*
* If we are the only waiter, wait relative to when the transaction
* started, rather than the current time. This credits the transaction for
* "time already served", e.g. reading indirect blocks.
*
* The minimum time for a transaction to take is calculated as:
* min_time = scale * (dirty - min) / (max - dirty)
* min_time is then capped at zfs_delay_max_ns.
*
* The delay has two degrees of freedom that can be adjusted via tunables.
* The percentage of dirty data at which we start to delay is defined by
* zfs_delay_min_dirty_percent. This should typically be at or above
* zfs_vdev_async_write_active_max_dirty_percent so that we only start to
* delay after writing at full speed has failed to keep up with the incoming
* write rate. The scale of the curve is defined by zfs_delay_scale. Roughly
* speaking, this variable determines the amount of delay at the midpoint of
* the curve.
*
* delay
* 10ms +-------------------------------------------------------------*+
* | *|
* 9ms + *+
* | *|
* 8ms + *+
* | * |
* 7ms + * +
* | * |
* 6ms + * +
* | * |
* 5ms + * +
* | * |
* 4ms + * +
* | * |
* 3ms + * +
* | * |
* 2ms + (midpoint) * +
* | | ** |
* 1ms + v *** +
* | zfs_delay_scale ----------> ******** |
* 0 +-------------------------------------*********----------------+
* 0% <- zfs_dirty_data_max -> 100%
*
* Note that since the delay is added to the outstanding time remaining on the
* most recent transaction, the delay is effectively the inverse of IOPS.
* Here the midpoint of 500us translates to 2000 IOPS. The shape of the curve
* was chosen such that small changes in the amount of accumulated dirty data
* in the first 3/4 of the curve yield relatively small differences in the
* amount of delay.
*
* The effects can be easier to understand when the amount of delay is
* represented on a log scale:
*
* delay
* 100ms +-------------------------------------------------------------++
* + +
* | |
* + *+
* 10ms + *+
* + ** +
* | (midpoint) ** |
* + | ** +
* 1ms + v **** +
* + zfs_delay_scale ----------> ***** +
* | **** |
* + **** +
* 100us + ** +
* + * +
* | * |
* + * +
* 10us + * +
* + +
* | |
* + +
* +--------------------------------------------------------------+
* 0% <- zfs_dirty_data_max -> 100%
*
* Note here that only as the amount of dirty data approaches its limit does
* the delay start to increase rapidly. The goal of a properly tuned system
* should be to keep the amount of dirty data out of that range by first
* ensuring that the appropriate limits are set for the I/O scheduler to reach
* optimal throughput on the backend storage, and then by changing the value
* of zfs_delay_scale to increase the steepness of the curve.
*/
static void
dmu_tx_delay(dmu_tx_t *tx, uint64_t dirty)
{
dsl_pool_t *dp = tx->tx_pool;
uint64_t delay_min_bytes =
zfs_dirty_data_max * zfs_delay_min_dirty_percent / 100;
hrtime_t wakeup, min_tx_time, now;
if (dirty <= delay_min_bytes)
return;
/*
* The caller has already waited until we are under the max.
* We make them pass us the amount of dirty data so we don't
* have to handle the case of it being >= the max, which could
* cause a divide-by-zero if it's == the max.
*/
ASSERT3U(dirty, <, zfs_dirty_data_max);
now = gethrtime();
min_tx_time = zfs_delay_scale *
(dirty - delay_min_bytes) / (zfs_dirty_data_max - dirty);
min_tx_time = MIN(min_tx_time, zfs_delay_max_ns);
if (now > tx->tx_start + min_tx_time)
return;
DTRACE_PROBE3(delay__mintime, dmu_tx_t *, tx, uint64_t, dirty,
uint64_t, min_tx_time);
mutex_enter(&dp->dp_lock);
wakeup = MAX(tx->tx_start + min_tx_time,
dp->dp_last_wakeup + min_tx_time);
dp->dp_last_wakeup = wakeup;
mutex_exit(&dp->dp_lock);
zfs_sleep_until(wakeup);
}
/*
* This routine attempts to assign the transaction to a transaction group.
* To do so, we must determine if there is sufficient free space on disk.
*
* If this is a "netfree" transaction (i.e. we called dmu_tx_mark_netfree()
* on it), then it is assumed that there is sufficient free space,
* unless there's insufficient slop space in the pool (see the comment
* above spa_slop_shift in spa_misc.c).
*
* If it is not a "netfree" transaction, then if the data already on disk
* is over the allowed usage (e.g. quota), this will fail with EDQUOT or
* ENOSPC. Otherwise, if the current rough estimate of pending changes,
* plus the rough estimate of this transaction's changes, may exceed the
* allowed usage, then this will fail with ERESTART, which will cause the
* caller to wait for the pending changes to be written to disk (by waiting
* for the next TXG to open), and then check the space usage again.
*
* The rough estimate of pending changes is comprised of the sum of:
*
* - this transaction's holds' txh_space_towrite
*
* - dd_tempreserved[], which is the sum of in-flight transactions'
* holds' txh_space_towrite (i.e. those transactions that have called
* dmu_tx_assign() but not yet called dmu_tx_commit()).
*
* - dd_space_towrite[], which is the amount of dirtied dbufs.
*
* Note that all of these values are inflated by spa_get_worst_case_asize(),
* which means that we may get ERESTART well before we are actually in danger
* of running out of space, but this also mitigates any small inaccuracies
* in the rough estimate (e.g. txh_space_towrite doesn't take into account
* indirect blocks, and dd_space_towrite[] doesn't take into account changes
* to the MOS).
*
* Note that due to this algorithm, it is possible to exceed the allowed
* usage by one transaction. Also, as we approach the allowed usage,
* we will allow a very limited amount of changes into each TXG, thus
* decreasing performance.
*/
static int
dmu_tx_try_assign(dmu_tx_t *tx, uint64_t txg_how)
{
spa_t *spa = tx->tx_pool->dp_spa;
ASSERT0(tx->tx_txg);
if (tx->tx_err) {
DMU_TX_STAT_BUMP(dmu_tx_error);
return (tx->tx_err);
}
if (spa_suspended(spa)) {
DMU_TX_STAT_BUMP(dmu_tx_suspended);
/*
* If the user has indicated a blocking failure mode
* then return ERESTART which will block in dmu_tx_wait().
* Otherwise, return EIO so that an error can get
* propagated back to the VOP calls.
*
* Note that we always honor the txg_how flag regardless
* of the failuremode setting.
*/
if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE &&
!(txg_how & TXG_WAIT))
return (SET_ERROR(EIO));
return (SET_ERROR(ERESTART));
}
if (!tx->tx_dirty_delayed &&
dsl_pool_wrlog_over_max(tx->tx_pool)) {
DMU_TX_STAT_BUMP(dmu_tx_wrlog_over_max);
return (SET_ERROR(ERESTART));
}
if (!tx->tx_dirty_delayed &&
dsl_pool_need_dirty_delay(tx->tx_pool)) {
tx->tx_wait_dirty = B_TRUE;
DMU_TX_STAT_BUMP(dmu_tx_dirty_delay);
return (SET_ERROR(ERESTART));
}
tx->tx_txg = txg_hold_open(tx->tx_pool, &tx->tx_txgh);
tx->tx_needassign_txh = NULL;
/*
* NB: No error returns are allowed after txg_hold_open, but
* before processing the dnode holds, due to the
* dmu_tx_unassign() logic.
*/
uint64_t towrite = 0;
uint64_t tohold = 0;
for (dmu_tx_hold_t *txh = list_head(&tx->tx_holds); txh != NULL;
txh = list_next(&tx->tx_holds, txh)) {
dnode_t *dn = txh->txh_dnode;
if (dn != NULL) {
/*
* This thread can't hold the dn_struct_rwlock
* while assigning the tx, because this can lead to
* deadlock. Specifically, if this dnode is already
* assigned to an earlier txg, this thread may need
* to wait for that txg to sync (the ERESTART case
* below). The other thread that has assigned this
* dnode to an earlier txg prevents this txg from
* syncing until its tx can complete (calling
* dmu_tx_commit()), but it may need to acquire the
* dn_struct_rwlock to do so (e.g. via
* dmu_buf_hold*()).
*
* Note that this thread can't hold the lock for
* read either, but the rwlock doesn't record
* enough information to make that assertion.
*/
ASSERT(!RW_WRITE_HELD(&dn->dn_struct_rwlock));
mutex_enter(&dn->dn_mtx);
if (dn->dn_assigned_txg == tx->tx_txg - 1) {
mutex_exit(&dn->dn_mtx);
tx->tx_needassign_txh = txh;
DMU_TX_STAT_BUMP(dmu_tx_group);
return (SET_ERROR(ERESTART));
}
if (dn->dn_assigned_txg == 0)
dn->dn_assigned_txg = tx->tx_txg;
ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg);
(void) zfs_refcount_add(&dn->dn_tx_holds, tx);
mutex_exit(&dn->dn_mtx);
}
towrite += zfs_refcount_count(&txh->txh_space_towrite);
tohold += zfs_refcount_count(&txh->txh_memory_tohold);
}
/* needed allocation: worst-case estimate of write space */
uint64_t asize = spa_get_worst_case_asize(tx->tx_pool->dp_spa, towrite);
/* calculate memory footprint estimate */
uint64_t memory = towrite + tohold;
if (tx->tx_dir != NULL && asize != 0) {
int err = dsl_dir_tempreserve_space(tx->tx_dir, memory,
asize, tx->tx_netfree, &tx->tx_tempreserve_cookie, tx);
if (err != 0)
return (err);
}
DMU_TX_STAT_BUMP(dmu_tx_assigned);
return (0);
}
static void
dmu_tx_unassign(dmu_tx_t *tx)
{
if (tx->tx_txg == 0)
return;
txg_rele_to_quiesce(&tx->tx_txgh);
/*
* Walk the transaction's hold list, removing the hold on the
* associated dnode, and notifying waiters if the refcount drops to 0.
*/
for (dmu_tx_hold_t *txh = list_head(&tx->tx_holds);
txh && txh != tx->tx_needassign_txh;
txh = list_next(&tx->tx_holds, txh)) {
dnode_t *dn = txh->txh_dnode;
if (dn == NULL)
continue;
mutex_enter(&dn->dn_mtx);
ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg);
if (zfs_refcount_remove(&dn->dn_tx_holds, tx) == 0) {
dn->dn_assigned_txg = 0;
cv_broadcast(&dn->dn_notxholds);
}
mutex_exit(&dn->dn_mtx);
}
txg_rele_to_sync(&tx->tx_txgh);
tx->tx_lasttried_txg = tx->tx_txg;
tx->tx_txg = 0;
}
/*
* Assign tx to a transaction group; txg_how is a bitmask:
*
* If TXG_WAIT is set and the currently open txg is full, this function
* will wait until there's a new txg. This should be used when no locks
* are being held. With this bit set, this function will only fail if
* we're truly out of space (or over quota).
*
* If TXG_WAIT is *not* set and we can't assign into the currently open
* txg without blocking, this function will return immediately with
* ERESTART. This should be used whenever locks are being held. On an
* ERESTART error, the caller should drop all locks, call dmu_tx_wait(),
* and try again.
*
* If TXG_NOTHROTTLE is set, this indicates that this tx should not be
* delayed due on the ZFS Write Throttle (see comments in dsl_pool.c for
* details on the throttle). This is used by the VFS operations, after
* they have already called dmu_tx_wait() (though most likely on a
* different tx).
*
* It is guaranteed that subsequent successful calls to dmu_tx_assign()
* will assign the tx to monotonically increasing txgs. Of course this is
* not strong monotonicity, because the same txg can be returned multiple
* times in a row. This guarantee holds both for subsequent calls from
* one thread and for multiple threads. For example, it is impossible to
* observe the following sequence of events:
*
* Thread 1 Thread 2
*
* dmu_tx_assign(T1, ...)
* 1 <- dmu_tx_get_txg(T1)
* dmu_tx_assign(T2, ...)
* 2 <- dmu_tx_get_txg(T2)
* dmu_tx_assign(T3, ...)
* 1 <- dmu_tx_get_txg(T3)
*/
int
dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how)
{
int err;
ASSERT(tx->tx_txg == 0);
ASSERT0(txg_how & ~(TXG_WAIT | TXG_NOTHROTTLE));
ASSERT(!dsl_pool_sync_context(tx->tx_pool));
/* If we might wait, we must not hold the config lock. */
IMPLY((txg_how & TXG_WAIT), !dsl_pool_config_held(tx->tx_pool));
if ((txg_how & TXG_NOTHROTTLE))
tx->tx_dirty_delayed = B_TRUE;
while ((err = dmu_tx_try_assign(tx, txg_how)) != 0) {
dmu_tx_unassign(tx);
if (err != ERESTART || !(txg_how & TXG_WAIT))
return (err);
dmu_tx_wait(tx);
}
txg_rele_to_quiesce(&tx->tx_txgh);
return (0);
}
void
dmu_tx_wait(dmu_tx_t *tx)
{
spa_t *spa = tx->tx_pool->dp_spa;
dsl_pool_t *dp = tx->tx_pool;
hrtime_t before;
ASSERT(tx->tx_txg == 0);
ASSERT(!dsl_pool_config_held(tx->tx_pool));
before = gethrtime();
if (tx->tx_wait_dirty) {
uint64_t dirty;
/*
* dmu_tx_try_assign() has determined that we need to wait
* because we've consumed much or all of the dirty buffer
* space.
*/
mutex_enter(&dp->dp_lock);
if (dp->dp_dirty_total >= zfs_dirty_data_max)
DMU_TX_STAT_BUMP(dmu_tx_dirty_over_max);
while (dp->dp_dirty_total >= zfs_dirty_data_max)
cv_wait(&dp->dp_spaceavail_cv, &dp->dp_lock);
dirty = dp->dp_dirty_total;
mutex_exit(&dp->dp_lock);
dmu_tx_delay(tx, dirty);
tx->tx_wait_dirty = B_FALSE;
/*
* Note: setting tx_dirty_delayed only has effect if the
* caller used TX_WAIT. Otherwise they are going to
* destroy this tx and try again. The common case,
* zfs_write(), uses TX_WAIT.
*/
tx->tx_dirty_delayed = B_TRUE;
} else if (spa_suspended(spa) || tx->tx_lasttried_txg == 0) {
/*
* If the pool is suspended we need to wait until it
* is resumed. Note that it's possible that the pool
* has become active after this thread has tried to
* obtain a tx. If that's the case then tx_lasttried_txg
* would not have been set.
*/
txg_wait_synced(dp, spa_last_synced_txg(spa) + 1);
} else if (tx->tx_needassign_txh) {
dnode_t *dn = tx->tx_needassign_txh->txh_dnode;
mutex_enter(&dn->dn_mtx);
while (dn->dn_assigned_txg == tx->tx_lasttried_txg - 1)
cv_wait(&dn->dn_notxholds, &dn->dn_mtx);
mutex_exit(&dn->dn_mtx);
tx->tx_needassign_txh = NULL;
} else {
/*
* If we have a lot of dirty data just wait until we sync
* out a TXG at which point we'll hopefully have synced
* a portion of the changes.
*/
txg_wait_synced(dp, spa_last_synced_txg(spa) + 1);
}
spa_tx_assign_add_nsecs(spa, gethrtime() - before);
}
static void
dmu_tx_destroy(dmu_tx_t *tx)
{
dmu_tx_hold_t *txh;
while ((txh = list_head(&tx->tx_holds)) != NULL) {
dnode_t *dn = txh->txh_dnode;
list_remove(&tx->tx_holds, txh);
zfs_refcount_destroy_many(&txh->txh_space_towrite,
zfs_refcount_count(&txh->txh_space_towrite));
zfs_refcount_destroy_many(&txh->txh_memory_tohold,
zfs_refcount_count(&txh->txh_memory_tohold));
kmem_free(txh, sizeof (dmu_tx_hold_t));
if (dn != NULL)
dnode_rele(dn, tx);
}
list_destroy(&tx->tx_callbacks);
list_destroy(&tx->tx_holds);
kmem_free(tx, sizeof (dmu_tx_t));
}
void
dmu_tx_commit(dmu_tx_t *tx)
{
ASSERT(tx->tx_txg != 0);
/*
* Go through the transaction's hold list and remove holds on
* associated dnodes, notifying waiters if no holds remain.
*/
for (dmu_tx_hold_t *txh = list_head(&tx->tx_holds); txh != NULL;
txh = list_next(&tx->tx_holds, txh)) {
dnode_t *dn = txh->txh_dnode;
if (dn == NULL)
continue;
mutex_enter(&dn->dn_mtx);
ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg);
if (zfs_refcount_remove(&dn->dn_tx_holds, tx) == 0) {
dn->dn_assigned_txg = 0;
cv_broadcast(&dn->dn_notxholds);
}
mutex_exit(&dn->dn_mtx);
}
if (tx->tx_tempreserve_cookie)
dsl_dir_tempreserve_clear(tx->tx_tempreserve_cookie, tx);
if (!list_is_empty(&tx->tx_callbacks))
txg_register_callbacks(&tx->tx_txgh, &tx->tx_callbacks);
if (tx->tx_anyobj == FALSE)
txg_rele_to_sync(&tx->tx_txgh);
dmu_tx_destroy(tx);
}
void
dmu_tx_abort(dmu_tx_t *tx)
{
ASSERT(tx->tx_txg == 0);
/*
* Call any registered callbacks with an error code.
*/
if (!list_is_empty(&tx->tx_callbacks))
dmu_tx_do_callbacks(&tx->tx_callbacks, SET_ERROR(ECANCELED));
dmu_tx_destroy(tx);
}
uint64_t
dmu_tx_get_txg(dmu_tx_t *tx)
{
ASSERT(tx->tx_txg != 0);
return (tx->tx_txg);
}
dsl_pool_t *
dmu_tx_pool(dmu_tx_t *tx)
{
ASSERT(tx->tx_pool != NULL);
return (tx->tx_pool);
}
void
dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *func, void *data)
{
dmu_tx_callback_t *dcb;
dcb = kmem_alloc(sizeof (dmu_tx_callback_t), KM_SLEEP);
dcb->dcb_func = func;
dcb->dcb_data = data;
list_insert_tail(&tx->tx_callbacks, dcb);
}
/*
* Call all the commit callbacks on a list, with a given error code.
*/
void
dmu_tx_do_callbacks(list_t *cb_list, int error)
{
dmu_tx_callback_t *dcb;
while ((dcb = list_tail(cb_list)) != NULL) {
list_remove(cb_list, dcb);
dcb->dcb_func(dcb->dcb_data, error);
kmem_free(dcb, sizeof (dmu_tx_callback_t));
}
}
/*
* Interface to hold a bunch of attributes.
* used for creating new files.
* attrsize is the total size of all attributes
* to be added during object creation
*
* For updating/adding a single attribute dmu_tx_hold_sa() should be used.
*/
/*
* hold necessary attribute name for attribute registration.
* should be a very rare case where this is needed. If it does
* happen it would only happen on the first write to the file system.
*/
static void
dmu_tx_sa_registration_hold(sa_os_t *sa, dmu_tx_t *tx)
{
if (!sa->sa_need_attr_registration)
return;
for (int i = 0; i != sa->sa_num_attrs; i++) {
if (!sa->sa_attr_table[i].sa_registered) {
if (sa->sa_reg_attr_obj)
dmu_tx_hold_zap(tx, sa->sa_reg_attr_obj,
B_TRUE, sa->sa_attr_table[i].sa_name);
else
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT,
B_TRUE, sa->sa_attr_table[i].sa_name);
}
}
}
void
dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object)
{
dmu_tx_hold_t *txh;
txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, object,
THT_SPILL, 0, 0);
if (txh != NULL)
(void) zfs_refcount_add_many(&txh->txh_space_towrite,
SPA_OLD_MAXBLOCKSIZE, FTAG);
}
void
dmu_tx_hold_sa_create(dmu_tx_t *tx, int attrsize)
{
sa_os_t *sa = tx->tx_objset->os_sa;
dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
if (tx->tx_objset->os_sa->sa_master_obj == 0)
return;
if (tx->tx_objset->os_sa->sa_layout_attr_obj) {
dmu_tx_hold_zap(tx, sa->sa_layout_attr_obj, B_TRUE, NULL);
} else {
dmu_tx_hold_zap(tx, sa->sa_master_obj, B_TRUE, SA_LAYOUTS);
dmu_tx_hold_zap(tx, sa->sa_master_obj, B_TRUE, SA_REGISTRY);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL);
}
dmu_tx_sa_registration_hold(sa, tx);
if (attrsize <= DN_OLD_MAX_BONUSLEN && !sa->sa_force_spill)
return;
(void) dmu_tx_hold_object_impl(tx, tx->tx_objset, DMU_NEW_OBJECT,
THT_SPILL, 0, 0);
}
/*
* Hold SA attribute
*
* dmu_tx_hold_sa(dmu_tx_t *tx, sa_handle_t *, attribute, add, size)
*
* variable_size is the total size of all variable sized attributes
* passed to this function. It is not the total size of all
* variable size attributes that *may* exist on this object.
*/
void
dmu_tx_hold_sa(dmu_tx_t *tx, sa_handle_t *hdl, boolean_t may_grow)
{
uint64_t object;
sa_os_t *sa = tx->tx_objset->os_sa;
ASSERT(hdl != NULL);
object = sa_handle_object(hdl);
dmu_buf_impl_t *db = (dmu_buf_impl_t *)hdl->sa_bonus;
DB_DNODE_ENTER(db);
dmu_tx_hold_bonus_by_dnode(tx, DB_DNODE(db));
DB_DNODE_EXIT(db);
if (tx->tx_objset->os_sa->sa_master_obj == 0)
return;
if (tx->tx_objset->os_sa->sa_reg_attr_obj == 0 ||
tx->tx_objset->os_sa->sa_layout_attr_obj == 0) {
dmu_tx_hold_zap(tx, sa->sa_master_obj, B_TRUE, SA_LAYOUTS);
dmu_tx_hold_zap(tx, sa->sa_master_obj, B_TRUE, SA_REGISTRY);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL);
}
dmu_tx_sa_registration_hold(sa, tx);
if (may_grow && tx->tx_objset->os_sa->sa_layout_attr_obj)
dmu_tx_hold_zap(tx, sa->sa_layout_attr_obj, B_TRUE, NULL);
if (sa->sa_force_spill || may_grow || hdl->sa_spill) {
ASSERT(tx->tx_txg == 0);
dmu_tx_hold_spill(tx, object);
} else {
dnode_t *dn;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
if (dn->dn_have_spill) {
ASSERT(tx->tx_txg == 0);
dmu_tx_hold_spill(tx, object);
}
DB_DNODE_EXIT(db);
}
}
void
dmu_tx_init(void)
{
dmu_tx_ksp = kstat_create("zfs", 0, "dmu_tx", "misc",
KSTAT_TYPE_NAMED, sizeof (dmu_tx_stats) / sizeof (kstat_named_t),
KSTAT_FLAG_VIRTUAL);
if (dmu_tx_ksp != NULL) {
dmu_tx_ksp->ks_data = &dmu_tx_stats;
kstat_install(dmu_tx_ksp);
}
}
void
dmu_tx_fini(void)
{
if (dmu_tx_ksp != NULL) {
kstat_delete(dmu_tx_ksp);
dmu_tx_ksp = NULL;
}
}
#if defined(_KERNEL)
EXPORT_SYMBOL(dmu_tx_create);
EXPORT_SYMBOL(dmu_tx_hold_write);
EXPORT_SYMBOL(dmu_tx_hold_write_by_dnode);
EXPORT_SYMBOL(dmu_tx_hold_free);
EXPORT_SYMBOL(dmu_tx_hold_free_by_dnode);
EXPORT_SYMBOL(dmu_tx_hold_zap);
EXPORT_SYMBOL(dmu_tx_hold_zap_by_dnode);
EXPORT_SYMBOL(dmu_tx_hold_bonus);
EXPORT_SYMBOL(dmu_tx_hold_bonus_by_dnode);
EXPORT_SYMBOL(dmu_tx_abort);
EXPORT_SYMBOL(dmu_tx_assign);
EXPORT_SYMBOL(dmu_tx_wait);
EXPORT_SYMBOL(dmu_tx_commit);
EXPORT_SYMBOL(dmu_tx_mark_netfree);
EXPORT_SYMBOL(dmu_tx_get_txg);
EXPORT_SYMBOL(dmu_tx_callback_register);
EXPORT_SYMBOL(dmu_tx_do_callbacks);
EXPORT_SYMBOL(dmu_tx_hold_spill);
EXPORT_SYMBOL(dmu_tx_hold_sa_create);
EXPORT_SYMBOL(dmu_tx_hold_sa);
#endif
diff --git a/module/zfs/dmu_zfetch.c b/module/zfs/dmu_zfetch.c
index f31728eae0f2..fdf0a1759716 100644
--- a/module/zfs/dmu_zfetch.c
+++ b/module/zfs/dmu_zfetch.c
@@ -1,552 +1,552 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright (c) 2013, 2017 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/dnode.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_zfetch.h>
#include <sys/dmu.h>
#include <sys/dbuf.h>
#include <sys/kstat.h>
#include <sys/wmsum.h>
/*
* This tunable disables predictive prefetch. Note that it leaves "prescient"
* prefetch (e.g. prefetch for zfs send) intact. Unlike predictive prefetch,
* prescient prefetch never issues i/os that end up not being needed,
* so it can't hurt performance.
*/
-int zfs_prefetch_disable = B_FALSE;
+static int zfs_prefetch_disable = B_FALSE;
/* max # of streams per zfetch */
-unsigned int zfetch_max_streams = 8;
+static unsigned int zfetch_max_streams = 8;
/* min time before stream reclaim */
-unsigned int zfetch_min_sec_reap = 2;
+static unsigned int zfetch_min_sec_reap = 2;
/* max bytes to prefetch per stream (default 8MB) */
unsigned int zfetch_max_distance = 8 * 1024 * 1024;
/* max bytes to prefetch indirects for per stream (default 64MB) */
unsigned int zfetch_max_idistance = 64 * 1024 * 1024;
/* max number of bytes in an array_read in which we allow prefetching (1MB) */
unsigned long zfetch_array_rd_sz = 1024 * 1024;
typedef struct zfetch_stats {
kstat_named_t zfetchstat_hits;
kstat_named_t zfetchstat_misses;
kstat_named_t zfetchstat_max_streams;
kstat_named_t zfetchstat_io_issued;
} zfetch_stats_t;
static zfetch_stats_t zfetch_stats = {
{ "hits", KSTAT_DATA_UINT64 },
{ "misses", KSTAT_DATA_UINT64 },
{ "max_streams", KSTAT_DATA_UINT64 },
{ "io_issued", KSTAT_DATA_UINT64 },
};
struct {
wmsum_t zfetchstat_hits;
wmsum_t zfetchstat_misses;
wmsum_t zfetchstat_max_streams;
wmsum_t zfetchstat_io_issued;
} zfetch_sums;
#define ZFETCHSTAT_BUMP(stat) \
wmsum_add(&zfetch_sums.stat, 1)
#define ZFETCHSTAT_ADD(stat, val) \
wmsum_add(&zfetch_sums.stat, val)
-kstat_t *zfetch_ksp;
+static kstat_t *zfetch_ksp;
static int
zfetch_kstats_update(kstat_t *ksp, int rw)
{
zfetch_stats_t *zs = ksp->ks_data;
if (rw == KSTAT_WRITE)
return (EACCES);
zs->zfetchstat_hits.value.ui64 =
wmsum_value(&zfetch_sums.zfetchstat_hits);
zs->zfetchstat_misses.value.ui64 =
wmsum_value(&zfetch_sums.zfetchstat_misses);
zs->zfetchstat_max_streams.value.ui64 =
wmsum_value(&zfetch_sums.zfetchstat_max_streams);
zs->zfetchstat_io_issued.value.ui64 =
wmsum_value(&zfetch_sums.zfetchstat_io_issued);
return (0);
}
void
zfetch_init(void)
{
wmsum_init(&zfetch_sums.zfetchstat_hits, 0);
wmsum_init(&zfetch_sums.zfetchstat_misses, 0);
wmsum_init(&zfetch_sums.zfetchstat_max_streams, 0);
wmsum_init(&zfetch_sums.zfetchstat_io_issued, 0);
zfetch_ksp = kstat_create("zfs", 0, "zfetchstats", "misc",
KSTAT_TYPE_NAMED, sizeof (zfetch_stats) / sizeof (kstat_named_t),
KSTAT_FLAG_VIRTUAL);
if (zfetch_ksp != NULL) {
zfetch_ksp->ks_data = &zfetch_stats;
zfetch_ksp->ks_update = zfetch_kstats_update;
kstat_install(zfetch_ksp);
}
}
void
zfetch_fini(void)
{
if (zfetch_ksp != NULL) {
kstat_delete(zfetch_ksp);
zfetch_ksp = NULL;
}
wmsum_fini(&zfetch_sums.zfetchstat_hits);
wmsum_fini(&zfetch_sums.zfetchstat_misses);
wmsum_fini(&zfetch_sums.zfetchstat_max_streams);
wmsum_fini(&zfetch_sums.zfetchstat_io_issued);
}
/*
* This takes a pointer to a zfetch structure and a dnode. It performs the
* necessary setup for the zfetch structure, grokking data from the
* associated dnode.
*/
void
dmu_zfetch_init(zfetch_t *zf, dnode_t *dno)
{
if (zf == NULL)
return;
zf->zf_dnode = dno;
zf->zf_numstreams = 0;
list_create(&zf->zf_stream, sizeof (zstream_t),
offsetof(zstream_t, zs_node));
mutex_init(&zf->zf_lock, NULL, MUTEX_DEFAULT, NULL);
}
static void
dmu_zfetch_stream_fini(zstream_t *zs)
{
ASSERT(!list_link_active(&zs->zs_node));
zfs_refcount_destroy(&zs->zs_callers);
zfs_refcount_destroy(&zs->zs_refs);
kmem_free(zs, sizeof (*zs));
}
static void
dmu_zfetch_stream_remove(zfetch_t *zf, zstream_t *zs)
{
ASSERT(MUTEX_HELD(&zf->zf_lock));
list_remove(&zf->zf_stream, zs);
zf->zf_numstreams--;
membar_producer();
if (zfs_refcount_remove(&zs->zs_refs, NULL) == 0)
dmu_zfetch_stream_fini(zs);
}
/*
* Clean-up state associated with a zfetch structure (e.g. destroy the
* streams). This doesn't free the zfetch_t itself, that's left to the caller.
*/
void
dmu_zfetch_fini(zfetch_t *zf)
{
zstream_t *zs;
mutex_enter(&zf->zf_lock);
while ((zs = list_head(&zf->zf_stream)) != NULL)
dmu_zfetch_stream_remove(zf, zs);
mutex_exit(&zf->zf_lock);
list_destroy(&zf->zf_stream);
mutex_destroy(&zf->zf_lock);
zf->zf_dnode = NULL;
}
/*
* If there aren't too many streams already, create a new stream.
* The "blkid" argument is the next block that we expect this stream to access.
* While we're here, clean up old streams (which haven't been
* accessed for at least zfetch_min_sec_reap seconds).
*/
static void
dmu_zfetch_stream_create(zfetch_t *zf, uint64_t blkid)
{
zstream_t *zs_next;
hrtime_t now = gethrtime();
ASSERT(MUTEX_HELD(&zf->zf_lock));
/*
* Clean up old streams.
*/
for (zstream_t *zs = list_head(&zf->zf_stream);
zs != NULL; zs = zs_next) {
zs_next = list_next(&zf->zf_stream, zs);
/*
* Skip if still active. 1 -- zf_stream reference.
*/
if (zfs_refcount_count(&zs->zs_refs) != 1)
continue;
if (((now - zs->zs_atime) / NANOSEC) >
zfetch_min_sec_reap)
dmu_zfetch_stream_remove(zf, zs);
}
/*
* The maximum number of streams is normally zfetch_max_streams,
* but for small files we lower it such that it's at least possible
* for all the streams to be non-overlapping.
*
* If we are already at the maximum number of streams for this file,
* even after removing old streams, then don't create this stream.
*/
uint32_t max_streams = MAX(1, MIN(zfetch_max_streams,
zf->zf_dnode->dn_maxblkid * zf->zf_dnode->dn_datablksz /
zfetch_max_distance));
if (zf->zf_numstreams >= max_streams) {
ZFETCHSTAT_BUMP(zfetchstat_max_streams);
return;
}
zstream_t *zs = kmem_zalloc(sizeof (*zs), KM_SLEEP);
zs->zs_blkid = blkid;
zs->zs_pf_blkid1 = blkid;
zs->zs_pf_blkid = blkid;
zs->zs_ipf_blkid1 = blkid;
zs->zs_ipf_blkid = blkid;
zs->zs_atime = now;
zs->zs_fetch = zf;
zs->zs_missed = B_FALSE;
zfs_refcount_create(&zs->zs_callers);
zfs_refcount_create(&zs->zs_refs);
/* One reference for zf_stream. */
zfs_refcount_add(&zs->zs_refs, NULL);
zf->zf_numstreams++;
list_insert_head(&zf->zf_stream, zs);
}
static void
dmu_zfetch_stream_done(void *arg, boolean_t io_issued)
{
(void) io_issued;
zstream_t *zs = arg;
if (zfs_refcount_remove(&zs->zs_refs, NULL) == 0)
dmu_zfetch_stream_fini(zs);
}
/*
* This is the predictive prefetch entry point. dmu_zfetch_prepare()
* associates dnode access specified with blkid and nblks arguments with
* prefetch stream, predicts further accesses based on that stats and returns
* the stream pointer on success. That pointer must later be passed to
* dmu_zfetch_run() to initiate the speculative prefetch for the stream and
* release it. dmu_zfetch() is a wrapper for simple cases when window between
* prediction and prefetch initiation is not needed.
* fetch_data argument specifies whether actual data blocks should be fetched:
* FALSE -- prefetch only indirect blocks for predicted data blocks;
* TRUE -- prefetch predicted data blocks plus following indirect blocks.
*/
zstream_t *
dmu_zfetch_prepare(zfetch_t *zf, uint64_t blkid, uint64_t nblks,
boolean_t fetch_data, boolean_t have_lock)
{
zstream_t *zs;
int64_t pf_start, ipf_start;
int64_t pf_ahead_blks, max_blks;
int max_dist_blks, pf_nblks, ipf_nblks;
uint64_t end_of_access_blkid, maxblkid;
end_of_access_blkid = blkid + nblks;
spa_t *spa = zf->zf_dnode->dn_objset->os_spa;
if (zfs_prefetch_disable)
return (NULL);
/*
* If we haven't yet loaded the indirect vdevs' mappings, we
* can only read from blocks that we carefully ensure are on
* concrete vdevs (or previously-loaded indirect vdevs). So we
* can't allow the predictive prefetcher to attempt reads of other
* blocks (e.g. of the MOS's dnode object).
*/
if (!spa_indirect_vdevs_loaded(spa))
return (NULL);
/*
* As a fast path for small (single-block) files, ignore access
* to the first block.
*/
if (!have_lock && blkid == 0)
return (NULL);
if (!have_lock)
rw_enter(&zf->zf_dnode->dn_struct_rwlock, RW_READER);
/*
* A fast path for small files for which no prefetch will
* happen.
*/
maxblkid = zf->zf_dnode->dn_maxblkid;
if (maxblkid < 2) {
if (!have_lock)
rw_exit(&zf->zf_dnode->dn_struct_rwlock);
return (NULL);
}
mutex_enter(&zf->zf_lock);
/*
* Find matching prefetch stream. Depending on whether the accesses
* are block-aligned, first block of the new access may either follow
* the last block of the previous access, or be equal to it.
*/
for (zs = list_head(&zf->zf_stream); zs != NULL;
zs = list_next(&zf->zf_stream, zs)) {
if (blkid == zs->zs_blkid) {
break;
} else if (blkid + 1 == zs->zs_blkid) {
blkid++;
nblks--;
break;
}
}
/*
* If the file is ending, remove the matching stream if found.
* If not found then it is too late to create a new one now.
*/
if (end_of_access_blkid >= maxblkid) {
if (zs != NULL)
dmu_zfetch_stream_remove(zf, zs);
mutex_exit(&zf->zf_lock);
if (!have_lock)
rw_exit(&zf->zf_dnode->dn_struct_rwlock);
return (NULL);
}
/* Exit if we already prefetched this block before. */
if (nblks == 0) {
mutex_exit(&zf->zf_lock);
if (!have_lock)
rw_exit(&zf->zf_dnode->dn_struct_rwlock);
return (NULL);
}
if (zs == NULL) {
/*
* This access is not part of any existing stream. Create
* a new stream for it.
*/
dmu_zfetch_stream_create(zf, end_of_access_blkid);
mutex_exit(&zf->zf_lock);
if (!have_lock)
rw_exit(&zf->zf_dnode->dn_struct_rwlock);
ZFETCHSTAT_BUMP(zfetchstat_misses);
return (NULL);
}
/*
* This access was to a block that we issued a prefetch for on
* behalf of this stream. Issue further prefetches for this stream.
*
* Normally, we start prefetching where we stopped
* prefetching last (zs_pf_blkid). But when we get our first
* hit on this stream, zs_pf_blkid == zs_blkid, we don't
* want to prefetch the block we just accessed. In this case,
* start just after the block we just accessed.
*/
pf_start = MAX(zs->zs_pf_blkid, end_of_access_blkid);
if (zs->zs_pf_blkid1 < end_of_access_blkid)
zs->zs_pf_blkid1 = end_of_access_blkid;
if (zs->zs_ipf_blkid1 < end_of_access_blkid)
zs->zs_ipf_blkid1 = end_of_access_blkid;
/*
* Double our amount of prefetched data, but don't let the
* prefetch get further ahead than zfetch_max_distance.
*/
if (fetch_data) {
max_dist_blks =
zfetch_max_distance >> zf->zf_dnode->dn_datablkshift;
/*
* Previously, we were (zs_pf_blkid - blkid) ahead. We
* want to now be double that, so read that amount again,
* plus the amount we are catching up by (i.e. the amount
* read just now).
*/
pf_ahead_blks = zs->zs_pf_blkid - blkid + nblks;
max_blks = max_dist_blks - (pf_start - end_of_access_blkid);
pf_nblks = MIN(pf_ahead_blks, max_blks);
} else {
pf_nblks = 0;
}
zs->zs_pf_blkid = pf_start + pf_nblks;
/*
* Do the same for indirects, starting from where we stopped last,
* or where we will stop reading data blocks (and the indirects
* that point to them).
*/
ipf_start = MAX(zs->zs_ipf_blkid, zs->zs_pf_blkid);
max_dist_blks = zfetch_max_idistance >> zf->zf_dnode->dn_datablkshift;
/*
* We want to double our distance ahead of the data prefetch
* (or reader, if we are not prefetching data). Previously, we
* were (zs_ipf_blkid - blkid) ahead. To double that, we read
* that amount again, plus the amount we are catching up by
* (i.e. the amount read now + the amount of data prefetched now).
*/
pf_ahead_blks = zs->zs_ipf_blkid - blkid + nblks + pf_nblks;
max_blks = max_dist_blks - (ipf_start - zs->zs_pf_blkid);
ipf_nblks = MIN(pf_ahead_blks, max_blks);
zs->zs_ipf_blkid = ipf_start + ipf_nblks;
zs->zs_blkid = end_of_access_blkid;
/* Protect the stream from reclamation. */
zs->zs_atime = gethrtime();
zfs_refcount_add(&zs->zs_refs, NULL);
/* Count concurrent callers. */
zfs_refcount_add(&zs->zs_callers, NULL);
mutex_exit(&zf->zf_lock);
if (!have_lock)
rw_exit(&zf->zf_dnode->dn_struct_rwlock);
ZFETCHSTAT_BUMP(zfetchstat_hits);
return (zs);
}
void
dmu_zfetch_run(zstream_t *zs, boolean_t missed, boolean_t have_lock)
{
zfetch_t *zf = zs->zs_fetch;
int64_t pf_start, pf_end, ipf_start, ipf_end;
int epbs, issued;
if (missed)
zs->zs_missed = missed;
/*
* Postpone the prefetch if there are more concurrent callers.
* It happens when multiple requests are waiting for the same
* indirect block. The last one will run the prefetch for all.
*/
if (zfs_refcount_remove(&zs->zs_callers, NULL) != 0) {
/* Drop reference taken in dmu_zfetch_prepare(). */
if (zfs_refcount_remove(&zs->zs_refs, NULL) == 0)
dmu_zfetch_stream_fini(zs);
return;
}
mutex_enter(&zf->zf_lock);
if (zs->zs_missed) {
pf_start = zs->zs_pf_blkid1;
pf_end = zs->zs_pf_blkid1 = zs->zs_pf_blkid;
} else {
pf_start = pf_end = 0;
}
ipf_start = MAX(zs->zs_pf_blkid1, zs->zs_ipf_blkid1);
ipf_end = zs->zs_ipf_blkid1 = zs->zs_ipf_blkid;
mutex_exit(&zf->zf_lock);
ASSERT3S(pf_start, <=, pf_end);
ASSERT3S(ipf_start, <=, ipf_end);
epbs = zf->zf_dnode->dn_indblkshift - SPA_BLKPTRSHIFT;
ipf_start = P2ROUNDUP(ipf_start, 1 << epbs) >> epbs;
ipf_end = P2ROUNDUP(ipf_end, 1 << epbs) >> epbs;
ASSERT3S(ipf_start, <=, ipf_end);
issued = pf_end - pf_start + ipf_end - ipf_start;
if (issued > 1) {
/* More references on top of taken in dmu_zfetch_prepare(). */
for (int i = 0; i < issued - 1; i++)
zfs_refcount_add(&zs->zs_refs, NULL);
} else if (issued == 0) {
/* Some other thread has done our work, so drop the ref. */
if (zfs_refcount_remove(&zs->zs_refs, NULL) == 0)
dmu_zfetch_stream_fini(zs);
return;
}
if (!have_lock)
rw_enter(&zf->zf_dnode->dn_struct_rwlock, RW_READER);
issued = 0;
for (int64_t blk = pf_start; blk < pf_end; blk++) {
issued += dbuf_prefetch_impl(zf->zf_dnode, 0, blk,
ZIO_PRIORITY_ASYNC_READ, ARC_FLAG_PREDICTIVE_PREFETCH,
dmu_zfetch_stream_done, zs);
}
for (int64_t iblk = ipf_start; iblk < ipf_end; iblk++) {
issued += dbuf_prefetch_impl(zf->zf_dnode, 1, iblk,
ZIO_PRIORITY_ASYNC_READ, ARC_FLAG_PREDICTIVE_PREFETCH,
dmu_zfetch_stream_done, zs);
}
if (!have_lock)
rw_exit(&zf->zf_dnode->dn_struct_rwlock);
if (issued)
ZFETCHSTAT_ADD(zfetchstat_io_issued, issued);
}
void
dmu_zfetch(zfetch_t *zf, uint64_t blkid, uint64_t nblks, boolean_t fetch_data,
boolean_t missed, boolean_t have_lock)
{
zstream_t *zs;
zs = dmu_zfetch_prepare(zf, blkid, nblks, fetch_data, have_lock);
if (zs)
dmu_zfetch_run(zs, missed, have_lock);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_prefetch, zfs_prefetch_, disable, INT, ZMOD_RW,
"Disable all ZFS prefetching");
ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, max_streams, UINT, ZMOD_RW,
"Max number of streams per zfetch");
ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, min_sec_reap, UINT, ZMOD_RW,
"Min time before stream reclaim");
ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, max_distance, UINT, ZMOD_RW,
"Max bytes to prefetch per stream");
ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, max_idistance, UINT, ZMOD_RW,
"Max bytes to prefetch indirects for per stream");
ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, array_rd_sz, ULONG, ZMOD_RW,
"Number of bytes in a array_read");
/* END CSTYLED */
diff --git a/module/zfs/dsl_dataset.c b/module/zfs/dsl_dataset.c
index 4623bcec4ea8..115f3df5d539 100644
--- a/module/zfs/dsl_dataset.c
+++ b/module/zfs/dsl_dataset.c
@@ -1,5009 +1,5009 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright (c) 2014, Joyent, Inc. All rights reserved.
* Copyright (c) 2014 RackTop Systems.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
* Copyright 2016, OmniTI Computer Consulting, Inc. All rights reserved.
* Copyright 2017 Nexenta Systems, Inc.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
* Copyright (c) 2020 The FreeBSD Foundation [1]
*
* [1] Portions of this software were developed by Allan Jude
* under sponsorship from the FreeBSD Foundation.
*/
#include <sys/dmu_objset.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_synctask.h>
#include <sys/dmu_traverse.h>
#include <sys/dmu_impl.h>
#include <sys/dmu_tx.h>
#include <sys/arc.h>
#include <sys/zio.h>
#include <sys/zap.h>
#include <sys/zfeature.h>
#include <sys/unique.h>
#include <sys/zfs_context.h>
#include <sys/zfs_ioctl.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/vdev.h>
#include <sys/zfs_znode.h>
#include <sys/zfs_onexit.h>
#include <sys/zvol.h>
#include <sys/dsl_scan.h>
#include <sys/dsl_deadlist.h>
#include <sys/dsl_destroy.h>
#include <sys/dsl_userhold.h>
#include <sys/dsl_bookmark.h>
#include <sys/policy.h>
#include <sys/dmu_send.h>
#include <sys/dmu_recv.h>
#include <sys/zio_compress.h>
#include <zfs_fletcher.h>
#include <sys/zio_checksum.h>
/*
* The SPA supports block sizes up to 16MB. However, very large blocks
* can have an impact on i/o latency (e.g. tying up a spinning disk for
* ~300ms), and also potentially on the memory allocator. Therefore,
* we do not allow the recordsize to be set larger than zfs_max_recordsize
* (default 1MB). Larger blocks can be created by changing this tunable,
* and pools with larger blocks can always be imported and used, regardless
* of this setting.
*/
int zfs_max_recordsize = 1 * 1024 * 1024;
-int zfs_allow_redacted_dataset_mount = 0;
+static int zfs_allow_redacted_dataset_mount = 0;
#define SWITCH64(x, y) \
{ \
uint64_t __tmp = (x); \
(x) = (y); \
(y) = __tmp; \
}
#define DS_REF_MAX (1ULL << 62)
extern inline dsl_dataset_phys_t *dsl_dataset_phys(dsl_dataset_t *ds);
static void dsl_dataset_set_remap_deadlist_object(dsl_dataset_t *ds,
uint64_t obj, dmu_tx_t *tx);
static void dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t *ds,
dmu_tx_t *tx);
static void unload_zfeature(dsl_dataset_t *ds, spa_feature_t f);
extern int spa_asize_inflation;
static zil_header_t zero_zil;
/*
* Figure out how much of this delta should be propagated to the dsl_dir
* layer. If there's a refreservation, that space has already been
* partially accounted for in our ancestors.
*/
static int64_t
parent_delta(dsl_dataset_t *ds, int64_t delta)
{
dsl_dataset_phys_t *ds_phys;
uint64_t old_bytes, new_bytes;
if (ds->ds_reserved == 0)
return (delta);
ds_phys = dsl_dataset_phys(ds);
old_bytes = MAX(ds_phys->ds_unique_bytes, ds->ds_reserved);
new_bytes = MAX(ds_phys->ds_unique_bytes + delta, ds->ds_reserved);
ASSERT3U(ABS((int64_t)(new_bytes - old_bytes)), <=, ABS(delta));
return (new_bytes - old_bytes);
}
void
dsl_dataset_block_born(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx)
{
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
int used = bp_get_dsize_sync(spa, bp);
int compressed = BP_GET_PSIZE(bp);
int uncompressed = BP_GET_UCSIZE(bp);
int64_t delta;
spa_feature_t f;
dprintf_bp(bp, "ds=%p", ds);
ASSERT(dmu_tx_is_syncing(tx));
/* It could have been compressed away to nothing */
if (BP_IS_HOLE(bp) || BP_IS_REDACTED(bp))
return;
ASSERT(BP_GET_TYPE(bp) != DMU_OT_NONE);
ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp)));
if (ds == NULL) {
dsl_pool_mos_diduse_space(tx->tx_pool,
used, compressed, uncompressed);
return;
}
ASSERT3U(bp->blk_birth, >, dsl_dataset_phys(ds)->ds_prev_snap_txg);
dmu_buf_will_dirty(ds->ds_dbuf, tx);
mutex_enter(&ds->ds_lock);
delta = parent_delta(ds, used);
dsl_dataset_phys(ds)->ds_referenced_bytes += used;
dsl_dataset_phys(ds)->ds_compressed_bytes += compressed;
dsl_dataset_phys(ds)->ds_uncompressed_bytes += uncompressed;
dsl_dataset_phys(ds)->ds_unique_bytes += used;
if (BP_GET_LSIZE(bp) > SPA_OLD_MAXBLOCKSIZE) {
ds->ds_feature_activation[SPA_FEATURE_LARGE_BLOCKS] =
(void *)B_TRUE;
}
f = zio_checksum_to_feature(BP_GET_CHECKSUM(bp));
if (f != SPA_FEATURE_NONE) {
ASSERT3S(spa_feature_table[f].fi_type, ==,
ZFEATURE_TYPE_BOOLEAN);
ds->ds_feature_activation[f] = (void *)B_TRUE;
}
f = zio_compress_to_feature(BP_GET_COMPRESS(bp));
if (f != SPA_FEATURE_NONE) {
ASSERT3S(spa_feature_table[f].fi_type, ==,
ZFEATURE_TYPE_BOOLEAN);
ds->ds_feature_activation[f] = (void *)B_TRUE;
}
/*
* Track block for livelist, but ignore embedded blocks because
* they do not need to be freed.
*/
if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) &&
bp->blk_birth > ds->ds_dir->dd_origin_txg &&
!(BP_IS_EMBEDDED(bp))) {
ASSERT(dsl_dir_is_clone(ds->ds_dir));
ASSERT(spa_feature_is_enabled(spa,
SPA_FEATURE_LIVELIST));
bplist_append(&ds->ds_dir->dd_pending_allocs, bp);
}
mutex_exit(&ds->ds_lock);
dsl_dir_diduse_transfer_space(ds->ds_dir, delta,
compressed, uncompressed, used,
DD_USED_REFRSRV, DD_USED_HEAD, tx);
}
/*
* Called when the specified segment has been remapped, and is thus no
* longer referenced in the head dataset. The vdev must be indirect.
*
* If the segment is referenced by a snapshot, put it on the remap deadlist.
* Otherwise, add this segment to the obsolete spacemap.
*/
void
dsl_dataset_block_remapped(dsl_dataset_t *ds, uint64_t vdev, uint64_t offset,
uint64_t size, uint64_t birth, dmu_tx_t *tx)
{
spa_t *spa = ds->ds_dir->dd_pool->dp_spa;
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(birth <= tx->tx_txg);
ASSERT(!ds->ds_is_snapshot);
if (birth > dsl_dataset_phys(ds)->ds_prev_snap_txg) {
spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx);
} else {
blkptr_t fakebp;
dva_t *dva = &fakebp.blk_dva[0];
ASSERT(ds != NULL);
mutex_enter(&ds->ds_remap_deadlist_lock);
if (!dsl_dataset_remap_deadlist_exists(ds)) {
dsl_dataset_create_remap_deadlist(ds, tx);
}
mutex_exit(&ds->ds_remap_deadlist_lock);
BP_ZERO(&fakebp);
fakebp.blk_birth = birth;
DVA_SET_VDEV(dva, vdev);
DVA_SET_OFFSET(dva, offset);
DVA_SET_ASIZE(dva, size);
dsl_deadlist_insert(&ds->ds_remap_deadlist, &fakebp, B_FALSE,
tx);
}
}
int
dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx,
boolean_t async)
{
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
int used = bp_get_dsize_sync(spa, bp);
int compressed = BP_GET_PSIZE(bp);
int uncompressed = BP_GET_UCSIZE(bp);
if (BP_IS_HOLE(bp) || BP_IS_REDACTED(bp))
return (0);
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(bp->blk_birth <= tx->tx_txg);
if (ds == NULL) {
dsl_free(tx->tx_pool, tx->tx_txg, bp);
dsl_pool_mos_diduse_space(tx->tx_pool,
-used, -compressed, -uncompressed);
return (used);
}
ASSERT3P(tx->tx_pool, ==, ds->ds_dir->dd_pool);
ASSERT(!ds->ds_is_snapshot);
dmu_buf_will_dirty(ds->ds_dbuf, tx);
/*
* Track block for livelist, but ignore embedded blocks because
* they do not need to be freed.
*/
if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) &&
bp->blk_birth > ds->ds_dir->dd_origin_txg &&
!(BP_IS_EMBEDDED(bp))) {
ASSERT(dsl_dir_is_clone(ds->ds_dir));
ASSERT(spa_feature_is_enabled(spa,
SPA_FEATURE_LIVELIST));
bplist_append(&ds->ds_dir->dd_pending_frees, bp);
}
if (bp->blk_birth > dsl_dataset_phys(ds)->ds_prev_snap_txg) {
int64_t delta;
dprintf_bp(bp, "freeing ds=%llu", (u_longlong_t)ds->ds_object);
dsl_free(tx->tx_pool, tx->tx_txg, bp);
mutex_enter(&ds->ds_lock);
ASSERT(dsl_dataset_phys(ds)->ds_unique_bytes >= used ||
!DS_UNIQUE_IS_ACCURATE(ds));
delta = parent_delta(ds, -used);
dsl_dataset_phys(ds)->ds_unique_bytes -= used;
mutex_exit(&ds->ds_lock);
dsl_dir_diduse_transfer_space(ds->ds_dir,
delta, -compressed, -uncompressed, -used,
DD_USED_REFRSRV, DD_USED_HEAD, tx);
} else {
dprintf_bp(bp, "putting on dead list: %s", "");
if (async) {
/*
* We are here as part of zio's write done callback,
* which means we're a zio interrupt thread. We can't
* call dsl_deadlist_insert() now because it may block
* waiting for I/O. Instead, put bp on the deferred
* queue and let dsl_pool_sync() finish the job.
*/
bplist_append(&ds->ds_pending_deadlist, bp);
} else {
dsl_deadlist_insert(&ds->ds_deadlist, bp, B_FALSE, tx);
}
ASSERT3U(ds->ds_prev->ds_object, ==,
dsl_dataset_phys(ds)->ds_prev_snap_obj);
ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_num_children > 0);
/* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj ==
ds->ds_object && bp->blk_birth >
dsl_dataset_phys(ds->ds_prev)->ds_prev_snap_txg) {
dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
mutex_enter(&ds->ds_prev->ds_lock);
dsl_dataset_phys(ds->ds_prev)->ds_unique_bytes += used;
mutex_exit(&ds->ds_prev->ds_lock);
}
if (bp->blk_birth > ds->ds_dir->dd_origin_txg) {
dsl_dir_transfer_space(ds->ds_dir, used,
DD_USED_HEAD, DD_USED_SNAP, tx);
}
}
dsl_bookmark_block_killed(ds, bp, tx);
mutex_enter(&ds->ds_lock);
ASSERT3U(dsl_dataset_phys(ds)->ds_referenced_bytes, >=, used);
dsl_dataset_phys(ds)->ds_referenced_bytes -= used;
ASSERT3U(dsl_dataset_phys(ds)->ds_compressed_bytes, >=, compressed);
dsl_dataset_phys(ds)->ds_compressed_bytes -= compressed;
ASSERT3U(dsl_dataset_phys(ds)->ds_uncompressed_bytes, >=, uncompressed);
dsl_dataset_phys(ds)->ds_uncompressed_bytes -= uncompressed;
mutex_exit(&ds->ds_lock);
return (used);
}
struct feature_type_uint64_array_arg {
uint64_t length;
uint64_t *array;
};
static void
unload_zfeature(dsl_dataset_t *ds, spa_feature_t f)
{
switch (spa_feature_table[f].fi_type) {
case ZFEATURE_TYPE_BOOLEAN:
break;
case ZFEATURE_TYPE_UINT64_ARRAY:
{
struct feature_type_uint64_array_arg *ftuaa = ds->ds_feature[f];
kmem_free(ftuaa->array, ftuaa->length * sizeof (uint64_t));
kmem_free(ftuaa, sizeof (*ftuaa));
break;
}
default:
panic("Invalid zfeature type %d", spa_feature_table[f].fi_type);
}
}
static int
load_zfeature(objset_t *mos, dsl_dataset_t *ds, spa_feature_t f)
{
int err = 0;
switch (spa_feature_table[f].fi_type) {
case ZFEATURE_TYPE_BOOLEAN:
err = zap_contains(mos, ds->ds_object,
spa_feature_table[f].fi_guid);
if (err == 0) {
ds->ds_feature[f] = (void *)B_TRUE;
} else {
ASSERT3U(err, ==, ENOENT);
err = 0;
}
break;
case ZFEATURE_TYPE_UINT64_ARRAY:
{
uint64_t int_size, num_int;
uint64_t *data;
err = zap_length(mos, ds->ds_object,
spa_feature_table[f].fi_guid, &int_size, &num_int);
if (err != 0) {
ASSERT3U(err, ==, ENOENT);
err = 0;
break;
}
ASSERT3U(int_size, ==, sizeof (uint64_t));
data = kmem_alloc(int_size * num_int, KM_SLEEP);
VERIFY0(zap_lookup(mos, ds->ds_object,
spa_feature_table[f].fi_guid, int_size, num_int, data));
struct feature_type_uint64_array_arg *ftuaa =
kmem_alloc(sizeof (*ftuaa), KM_SLEEP);
ftuaa->length = num_int;
ftuaa->array = data;
ds->ds_feature[f] = ftuaa;
break;
}
default:
panic("Invalid zfeature type %d", spa_feature_table[f].fi_type);
}
return (err);
}
/*
* We have to release the fsid synchronously or we risk that a subsequent
* mount of the same dataset will fail to unique_insert the fsid. This
* failure would manifest itself as the fsid of this dataset changing
* between mounts which makes NFS clients quite unhappy.
*/
static void
dsl_dataset_evict_sync(void *dbu)
{
dsl_dataset_t *ds = dbu;
ASSERT(ds->ds_owner == NULL);
unique_remove(ds->ds_fsid_guid);
}
static void
dsl_dataset_evict_async(void *dbu)
{
dsl_dataset_t *ds = dbu;
ASSERT(ds->ds_owner == NULL);
ds->ds_dbuf = NULL;
if (ds->ds_objset != NULL)
dmu_objset_evict(ds->ds_objset);
if (ds->ds_prev) {
dsl_dataset_rele(ds->ds_prev, ds);
ds->ds_prev = NULL;
}
dsl_bookmark_fini_ds(ds);
bplist_destroy(&ds->ds_pending_deadlist);
if (dsl_deadlist_is_open(&ds->ds_deadlist))
dsl_deadlist_close(&ds->ds_deadlist);
if (dsl_deadlist_is_open(&ds->ds_remap_deadlist))
dsl_deadlist_close(&ds->ds_remap_deadlist);
if (ds->ds_dir)
dsl_dir_async_rele(ds->ds_dir, ds);
ASSERT(!list_link_active(&ds->ds_synced_link));
for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
if (dsl_dataset_feature_is_active(ds, f))
unload_zfeature(ds, f);
}
list_destroy(&ds->ds_prop_cbs);
mutex_destroy(&ds->ds_lock);
mutex_destroy(&ds->ds_opening_lock);
mutex_destroy(&ds->ds_sendstream_lock);
mutex_destroy(&ds->ds_remap_deadlist_lock);
zfs_refcount_destroy(&ds->ds_longholds);
rrw_destroy(&ds->ds_bp_rwlock);
kmem_free(ds, sizeof (dsl_dataset_t));
}
int
dsl_dataset_get_snapname(dsl_dataset_t *ds)
{
dsl_dataset_phys_t *headphys;
int err;
dmu_buf_t *headdbuf;
dsl_pool_t *dp = ds->ds_dir->dd_pool;
objset_t *mos = dp->dp_meta_objset;
if (ds->ds_snapname[0])
return (0);
if (dsl_dataset_phys(ds)->ds_next_snap_obj == 0)
return (0);
err = dmu_bonus_hold(mos, dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj,
FTAG, &headdbuf);
if (err != 0)
return (err);
headphys = headdbuf->db_data;
err = zap_value_search(dp->dp_meta_objset,
headphys->ds_snapnames_zapobj, ds->ds_object, 0, ds->ds_snapname);
if (err != 0 && zfs_recover == B_TRUE) {
err = 0;
(void) snprintf(ds->ds_snapname, sizeof (ds->ds_snapname),
"SNAPOBJ=%llu-ERR=%d",
(unsigned long long)ds->ds_object, err);
}
dmu_buf_rele(headdbuf, FTAG);
return (err);
}
int
dsl_dataset_snap_lookup(dsl_dataset_t *ds, const char *name, uint64_t *value)
{
objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj;
matchtype_t mt = 0;
int err;
if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET)
mt = MT_NORMALIZE;
err = zap_lookup_norm(mos, snapobj, name, 8, 1,
value, mt, NULL, 0, NULL);
if (err == ENOTSUP && (mt & MT_NORMALIZE))
err = zap_lookup(mos, snapobj, name, 8, 1, value);
return (err);
}
int
dsl_dataset_snap_remove(dsl_dataset_t *ds, const char *name, dmu_tx_t *tx,
boolean_t adj_cnt)
{
objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj;
matchtype_t mt = 0;
int err;
dsl_dir_snap_cmtime_update(ds->ds_dir);
if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET)
mt = MT_NORMALIZE;
err = zap_remove_norm(mos, snapobj, name, mt, tx);
if (err == ENOTSUP && (mt & MT_NORMALIZE))
err = zap_remove(mos, snapobj, name, tx);
if (err == 0 && adj_cnt)
dsl_fs_ss_count_adjust(ds->ds_dir, -1,
DD_FIELD_SNAPSHOT_COUNT, tx);
return (err);
}
boolean_t
dsl_dataset_try_add_ref(dsl_pool_t *dp, dsl_dataset_t *ds, void *tag)
{
dmu_buf_t *dbuf = ds->ds_dbuf;
boolean_t result = B_FALSE;
if (dbuf != NULL && dmu_buf_try_add_ref(dbuf, dp->dp_meta_objset,
ds->ds_object, DMU_BONUS_BLKID, tag)) {
if (ds == dmu_buf_get_user(dbuf))
result = B_TRUE;
else
dmu_buf_rele(dbuf, tag);
}
return (result);
}
int
dsl_dataset_hold_obj(dsl_pool_t *dp, uint64_t dsobj, void *tag,
dsl_dataset_t **dsp)
{
objset_t *mos = dp->dp_meta_objset;
dmu_buf_t *dbuf;
dsl_dataset_t *ds;
int err;
dmu_object_info_t doi;
ASSERT(dsl_pool_config_held(dp));
err = dmu_bonus_hold(mos, dsobj, tag, &dbuf);
if (err != 0)
return (err);
/* Make sure dsobj has the correct object type. */
dmu_object_info_from_db(dbuf, &doi);
if (doi.doi_bonus_type != DMU_OT_DSL_DATASET) {
dmu_buf_rele(dbuf, tag);
return (SET_ERROR(EINVAL));
}
ds = dmu_buf_get_user(dbuf);
if (ds == NULL) {
dsl_dataset_t *winner = NULL;
ds = kmem_zalloc(sizeof (dsl_dataset_t), KM_SLEEP);
ds->ds_dbuf = dbuf;
ds->ds_object = dsobj;
ds->ds_is_snapshot = dsl_dataset_phys(ds)->ds_num_children != 0;
list_link_init(&ds->ds_synced_link);
err = dsl_dir_hold_obj(dp, dsl_dataset_phys(ds)->ds_dir_obj,
NULL, ds, &ds->ds_dir);
if (err != 0) {
kmem_free(ds, sizeof (dsl_dataset_t));
dmu_buf_rele(dbuf, tag);
return (err);
}
mutex_init(&ds->ds_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&ds->ds_opening_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&ds->ds_sendstream_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&ds->ds_remap_deadlist_lock,
NULL, MUTEX_DEFAULT, NULL);
rrw_init(&ds->ds_bp_rwlock, B_FALSE);
zfs_refcount_create(&ds->ds_longholds);
bplist_create(&ds->ds_pending_deadlist);
list_create(&ds->ds_sendstreams, sizeof (dmu_sendstatus_t),
offsetof(dmu_sendstatus_t, dss_link));
list_create(&ds->ds_prop_cbs, sizeof (dsl_prop_cb_record_t),
offsetof(dsl_prop_cb_record_t, cbr_ds_node));
if (doi.doi_type == DMU_OTN_ZAP_METADATA) {
spa_feature_t f;
for (f = 0; f < SPA_FEATURES; f++) {
if (!(spa_feature_table[f].fi_flags &
ZFEATURE_FLAG_PER_DATASET))
continue;
err = load_zfeature(mos, ds, f);
}
}
if (!ds->ds_is_snapshot) {
ds->ds_snapname[0] = '\0';
if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
err = dsl_dataset_hold_obj(dp,
dsl_dataset_phys(ds)->ds_prev_snap_obj,
ds, &ds->ds_prev);
}
err = dsl_bookmark_init_ds(ds);
} else {
if (zfs_flags & ZFS_DEBUG_SNAPNAMES)
err = dsl_dataset_get_snapname(ds);
if (err == 0 &&
dsl_dataset_phys(ds)->ds_userrefs_obj != 0) {
err = zap_count(
ds->ds_dir->dd_pool->dp_meta_objset,
dsl_dataset_phys(ds)->ds_userrefs_obj,
&ds->ds_userrefs);
}
}
if (err == 0 && !ds->ds_is_snapshot) {
err = dsl_prop_get_int_ds(ds,
zfs_prop_to_name(ZFS_PROP_REFRESERVATION),
&ds->ds_reserved);
if (err == 0) {
err = dsl_prop_get_int_ds(ds,
zfs_prop_to_name(ZFS_PROP_REFQUOTA),
&ds->ds_quota);
}
} else {
ds->ds_reserved = ds->ds_quota = 0;
}
if (err == 0 && ds->ds_dir->dd_crypto_obj != 0 &&
ds->ds_is_snapshot &&
zap_contains(mos, dsobj, DS_FIELD_IVSET_GUID) != 0) {
dp->dp_spa->spa_errata =
ZPOOL_ERRATA_ZOL_8308_ENCRYPTION;
}
dsl_deadlist_open(&ds->ds_deadlist,
mos, dsl_dataset_phys(ds)->ds_deadlist_obj);
uint64_t remap_deadlist_obj =
dsl_dataset_get_remap_deadlist_object(ds);
if (remap_deadlist_obj != 0) {
dsl_deadlist_open(&ds->ds_remap_deadlist, mos,
remap_deadlist_obj);
}
dmu_buf_init_user(&ds->ds_dbu, dsl_dataset_evict_sync,
dsl_dataset_evict_async, &ds->ds_dbuf);
if (err == 0)
winner = dmu_buf_set_user_ie(dbuf, &ds->ds_dbu);
if (err != 0 || winner != NULL) {
bplist_destroy(&ds->ds_pending_deadlist);
dsl_deadlist_close(&ds->ds_deadlist);
if (dsl_deadlist_is_open(&ds->ds_remap_deadlist))
dsl_deadlist_close(&ds->ds_remap_deadlist);
dsl_bookmark_fini_ds(ds);
if (ds->ds_prev)
dsl_dataset_rele(ds->ds_prev, ds);
dsl_dir_rele(ds->ds_dir, ds);
for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
if (dsl_dataset_feature_is_active(ds, f))
unload_zfeature(ds, f);
}
list_destroy(&ds->ds_prop_cbs);
list_destroy(&ds->ds_sendstreams);
mutex_destroy(&ds->ds_lock);
mutex_destroy(&ds->ds_opening_lock);
mutex_destroy(&ds->ds_sendstream_lock);
mutex_destroy(&ds->ds_remap_deadlist_lock);
zfs_refcount_destroy(&ds->ds_longholds);
rrw_destroy(&ds->ds_bp_rwlock);
kmem_free(ds, sizeof (dsl_dataset_t));
if (err != 0) {
dmu_buf_rele(dbuf, tag);
return (err);
}
ds = winner;
} else {
ds->ds_fsid_guid =
unique_insert(dsl_dataset_phys(ds)->ds_fsid_guid);
if (ds->ds_fsid_guid !=
dsl_dataset_phys(ds)->ds_fsid_guid) {
zfs_dbgmsg("ds_fsid_guid changed from "
"%llx to %llx for pool %s dataset id %llu",
(long long)
dsl_dataset_phys(ds)->ds_fsid_guid,
(long long)ds->ds_fsid_guid,
spa_name(dp->dp_spa),
(u_longlong_t)dsobj);
}
}
}
ASSERT3P(ds->ds_dbuf, ==, dbuf);
ASSERT3P(dsl_dataset_phys(ds), ==, dbuf->db_data);
ASSERT(dsl_dataset_phys(ds)->ds_prev_snap_obj != 0 ||
spa_version(dp->dp_spa) < SPA_VERSION_ORIGIN ||
dp->dp_origin_snap == NULL || ds == dp->dp_origin_snap);
*dsp = ds;
return (0);
}
int
dsl_dataset_create_key_mapping(dsl_dataset_t *ds)
{
dsl_dir_t *dd = ds->ds_dir;
if (dd->dd_crypto_obj == 0)
return (0);
return (spa_keystore_create_mapping(dd->dd_pool->dp_spa,
ds, ds, &ds->ds_key_mapping));
}
int
dsl_dataset_hold_obj_flags(dsl_pool_t *dp, uint64_t dsobj,
ds_hold_flags_t flags, void *tag, dsl_dataset_t **dsp)
{
int err;
err = dsl_dataset_hold_obj(dp, dsobj, tag, dsp);
if (err != 0)
return (err);
ASSERT3P(*dsp, !=, NULL);
if (flags & DS_HOLD_FLAG_DECRYPT) {
err = dsl_dataset_create_key_mapping(*dsp);
if (err != 0)
dsl_dataset_rele(*dsp, tag);
}
return (err);
}
int
dsl_dataset_hold_flags(dsl_pool_t *dp, const char *name, ds_hold_flags_t flags,
void *tag, dsl_dataset_t **dsp)
{
dsl_dir_t *dd;
const char *snapname;
uint64_t obj;
int err = 0;
dsl_dataset_t *ds;
err = dsl_dir_hold(dp, name, FTAG, &dd, &snapname);
if (err != 0)
return (err);
ASSERT(dsl_pool_config_held(dp));
obj = dsl_dir_phys(dd)->dd_head_dataset_obj;
if (obj != 0)
err = dsl_dataset_hold_obj_flags(dp, obj, flags, tag, &ds);
else
err = SET_ERROR(ENOENT);
/* we may be looking for a snapshot */
if (err == 0 && snapname != NULL) {
dsl_dataset_t *snap_ds;
if (*snapname++ != '@') {
dsl_dataset_rele_flags(ds, flags, tag);
dsl_dir_rele(dd, FTAG);
return (SET_ERROR(ENOENT));
}
dprintf("looking for snapshot '%s'\n", snapname);
err = dsl_dataset_snap_lookup(ds, snapname, &obj);
if (err == 0) {
err = dsl_dataset_hold_obj_flags(dp, obj, flags, tag,
&snap_ds);
}
dsl_dataset_rele_flags(ds, flags, tag);
if (err == 0) {
mutex_enter(&snap_ds->ds_lock);
if (snap_ds->ds_snapname[0] == 0)
(void) strlcpy(snap_ds->ds_snapname, snapname,
sizeof (snap_ds->ds_snapname));
mutex_exit(&snap_ds->ds_lock);
ds = snap_ds;
}
}
if (err == 0)
*dsp = ds;
dsl_dir_rele(dd, FTAG);
return (err);
}
int
dsl_dataset_hold(dsl_pool_t *dp, const char *name, void *tag,
dsl_dataset_t **dsp)
{
return (dsl_dataset_hold_flags(dp, name, 0, tag, dsp));
}
static int
dsl_dataset_own_obj_impl(dsl_pool_t *dp, uint64_t dsobj, ds_hold_flags_t flags,
void *tag, boolean_t override, dsl_dataset_t **dsp)
{
int err = dsl_dataset_hold_obj_flags(dp, dsobj, flags, tag, dsp);
if (err != 0)
return (err);
if (!dsl_dataset_tryown(*dsp, tag, override)) {
dsl_dataset_rele_flags(*dsp, flags, tag);
*dsp = NULL;
return (SET_ERROR(EBUSY));
}
return (0);
}
int
dsl_dataset_own_obj(dsl_pool_t *dp, uint64_t dsobj, ds_hold_flags_t flags,
void *tag, dsl_dataset_t **dsp)
{
return (dsl_dataset_own_obj_impl(dp, dsobj, flags, tag, B_FALSE, dsp));
}
int
dsl_dataset_own_obj_force(dsl_pool_t *dp, uint64_t dsobj,
ds_hold_flags_t flags, void *tag, dsl_dataset_t **dsp)
{
return (dsl_dataset_own_obj_impl(dp, dsobj, flags, tag, B_TRUE, dsp));
}
static int
dsl_dataset_own_impl(dsl_pool_t *dp, const char *name, ds_hold_flags_t flags,
void *tag, boolean_t override, dsl_dataset_t **dsp)
{
int err = dsl_dataset_hold_flags(dp, name, flags, tag, dsp);
if (err != 0)
return (err);
if (!dsl_dataset_tryown(*dsp, tag, override)) {
dsl_dataset_rele_flags(*dsp, flags, tag);
return (SET_ERROR(EBUSY));
}
return (0);
}
int
dsl_dataset_own_force(dsl_pool_t *dp, const char *name, ds_hold_flags_t flags,
void *tag, dsl_dataset_t **dsp)
{
return (dsl_dataset_own_impl(dp, name, flags, tag, B_TRUE, dsp));
}
int
dsl_dataset_own(dsl_pool_t *dp, const char *name, ds_hold_flags_t flags,
void *tag, dsl_dataset_t **dsp)
{
return (dsl_dataset_own_impl(dp, name, flags, tag, B_FALSE, dsp));
}
/*
* See the comment above dsl_pool_hold() for details. In summary, a long
* hold is used to prevent destruction of a dataset while the pool hold
* is dropped, allowing other concurrent operations (e.g. spa_sync()).
*
* The dataset and pool must be held when this function is called. After it
* is called, the pool hold may be released while the dataset is still held
* and accessed.
*/
void
-dsl_dataset_long_hold(dsl_dataset_t *ds, void *tag)
+dsl_dataset_long_hold(dsl_dataset_t *ds, const void *tag)
{
ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
(void) zfs_refcount_add(&ds->ds_longholds, tag);
}
void
-dsl_dataset_long_rele(dsl_dataset_t *ds, void *tag)
+dsl_dataset_long_rele(dsl_dataset_t *ds, const void *tag)
{
(void) zfs_refcount_remove(&ds->ds_longholds, tag);
}
/* Return B_TRUE if there are any long holds on this dataset. */
boolean_t
dsl_dataset_long_held(dsl_dataset_t *ds)
{
return (!zfs_refcount_is_zero(&ds->ds_longholds));
}
void
dsl_dataset_name(dsl_dataset_t *ds, char *name)
{
if (ds == NULL) {
(void) strlcpy(name, "mos", ZFS_MAX_DATASET_NAME_LEN);
} else {
dsl_dir_name(ds->ds_dir, name);
VERIFY0(dsl_dataset_get_snapname(ds));
if (ds->ds_snapname[0]) {
VERIFY3U(strlcat(name, "@", ZFS_MAX_DATASET_NAME_LEN),
<, ZFS_MAX_DATASET_NAME_LEN);
/*
* We use a "recursive" mutex so that we
* can call dprintf_ds() with ds_lock held.
*/
if (!MUTEX_HELD(&ds->ds_lock)) {
mutex_enter(&ds->ds_lock);
VERIFY3U(strlcat(name, ds->ds_snapname,
ZFS_MAX_DATASET_NAME_LEN), <,
ZFS_MAX_DATASET_NAME_LEN);
mutex_exit(&ds->ds_lock);
} else {
VERIFY3U(strlcat(name, ds->ds_snapname,
ZFS_MAX_DATASET_NAME_LEN), <,
ZFS_MAX_DATASET_NAME_LEN);
}
}
}
}
int
dsl_dataset_namelen(dsl_dataset_t *ds)
{
VERIFY0(dsl_dataset_get_snapname(ds));
mutex_enter(&ds->ds_lock);
int len = strlen(ds->ds_snapname);
mutex_exit(&ds->ds_lock);
/* add '@' if ds is a snap */
if (len > 0)
len++;
len += dsl_dir_namelen(ds->ds_dir);
return (len);
}
void
dsl_dataset_rele(dsl_dataset_t *ds, void *tag)
{
dmu_buf_rele(ds->ds_dbuf, tag);
}
void
dsl_dataset_remove_key_mapping(dsl_dataset_t *ds)
{
dsl_dir_t *dd = ds->ds_dir;
if (dd == NULL || dd->dd_crypto_obj == 0)
return;
(void) spa_keystore_remove_mapping(dd->dd_pool->dp_spa,
ds->ds_object, ds);
}
void
dsl_dataset_rele_flags(dsl_dataset_t *ds, ds_hold_flags_t flags, void *tag)
{
if (flags & DS_HOLD_FLAG_DECRYPT)
dsl_dataset_remove_key_mapping(ds);
dsl_dataset_rele(ds, tag);
}
void
dsl_dataset_disown(dsl_dataset_t *ds, ds_hold_flags_t flags, void *tag)
{
ASSERT3P(ds->ds_owner, ==, tag);
ASSERT(ds->ds_dbuf != NULL);
mutex_enter(&ds->ds_lock);
ds->ds_owner = NULL;
mutex_exit(&ds->ds_lock);
dsl_dataset_long_rele(ds, tag);
dsl_dataset_rele_flags(ds, flags, tag);
}
boolean_t
dsl_dataset_tryown(dsl_dataset_t *ds, void *tag, boolean_t override)
{
boolean_t gotit = FALSE;
ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
mutex_enter(&ds->ds_lock);
if (ds->ds_owner == NULL && (override || !(DS_IS_INCONSISTENT(ds) ||
(dsl_dataset_feature_is_active(ds,
SPA_FEATURE_REDACTED_DATASETS) &&
!zfs_allow_redacted_dataset_mount)))) {
ds->ds_owner = tag;
dsl_dataset_long_hold(ds, tag);
gotit = TRUE;
}
mutex_exit(&ds->ds_lock);
return (gotit);
}
boolean_t
dsl_dataset_has_owner(dsl_dataset_t *ds)
{
boolean_t rv;
mutex_enter(&ds->ds_lock);
rv = (ds->ds_owner != NULL);
mutex_exit(&ds->ds_lock);
return (rv);
}
static boolean_t
zfeature_active(spa_feature_t f, void *arg)
{
switch (spa_feature_table[f].fi_type) {
case ZFEATURE_TYPE_BOOLEAN: {
boolean_t val = (boolean_t)(uintptr_t)arg;
ASSERT(val == B_FALSE || val == B_TRUE);
return (val);
}
case ZFEATURE_TYPE_UINT64_ARRAY:
/*
* In this case, arg is a uint64_t array. The feature is active
* if the array is non-null.
*/
return (arg != NULL);
default:
panic("Invalid zfeature type %d", spa_feature_table[f].fi_type);
return (B_FALSE);
}
}
boolean_t
dsl_dataset_feature_is_active(dsl_dataset_t *ds, spa_feature_t f)
{
return (zfeature_active(f, ds->ds_feature[f]));
}
/*
* The buffers passed out by this function are references to internal buffers;
* they should not be freed by callers of this function, and they should not be
* used after the dataset has been released.
*/
boolean_t
dsl_dataset_get_uint64_array_feature(dsl_dataset_t *ds, spa_feature_t f,
uint64_t *outlength, uint64_t **outp)
{
VERIFY(spa_feature_table[f].fi_type & ZFEATURE_TYPE_UINT64_ARRAY);
if (!dsl_dataset_feature_is_active(ds, f)) {
return (B_FALSE);
}
struct feature_type_uint64_array_arg *ftuaa = ds->ds_feature[f];
*outp = ftuaa->array;
*outlength = ftuaa->length;
return (B_TRUE);
}
void
dsl_dataset_activate_feature(uint64_t dsobj, spa_feature_t f, void *arg,
dmu_tx_t *tx)
{
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset;
uint64_t zero = 0;
VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET);
spa_feature_incr(spa, f, tx);
dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx);
switch (spa_feature_table[f].fi_type) {
case ZFEATURE_TYPE_BOOLEAN:
ASSERT3S((boolean_t)(uintptr_t)arg, ==, B_TRUE);
VERIFY0(zap_add(mos, dsobj, spa_feature_table[f].fi_guid,
sizeof (zero), 1, &zero, tx));
break;
case ZFEATURE_TYPE_UINT64_ARRAY:
{
struct feature_type_uint64_array_arg *ftuaa = arg;
VERIFY0(zap_add(mos, dsobj, spa_feature_table[f].fi_guid,
sizeof (uint64_t), ftuaa->length, ftuaa->array, tx));
break;
}
default:
panic("Invalid zfeature type %d", spa_feature_table[f].fi_type);
}
}
static void
dsl_dataset_deactivate_feature_impl(dsl_dataset_t *ds, spa_feature_t f,
dmu_tx_t *tx)
{
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset;
uint64_t dsobj = ds->ds_object;
VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET);
VERIFY0(zap_remove(mos, dsobj, spa_feature_table[f].fi_guid, tx));
spa_feature_decr(spa, f, tx);
ds->ds_feature[f] = NULL;
}
void
dsl_dataset_deactivate_feature(dsl_dataset_t *ds, spa_feature_t f, dmu_tx_t *tx)
{
unload_zfeature(ds, f);
dsl_dataset_deactivate_feature_impl(ds, f, tx);
}
uint64_t
dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin,
dsl_crypto_params_t *dcp, uint64_t flags, dmu_tx_t *tx)
{
dsl_pool_t *dp = dd->dd_pool;
dmu_buf_t *dbuf;
dsl_dataset_phys_t *dsphys;
uint64_t dsobj;
objset_t *mos = dp->dp_meta_objset;
if (origin == NULL)
origin = dp->dp_origin_snap;
ASSERT(origin == NULL || origin->ds_dir->dd_pool == dp);
ASSERT(origin == NULL || dsl_dataset_phys(origin)->ds_num_children > 0);
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(dsl_dir_phys(dd)->dd_head_dataset_obj == 0);
dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0,
DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx);
VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf));
dmu_buf_will_dirty(dbuf, tx);
dsphys = dbuf->db_data;
bzero(dsphys, sizeof (dsl_dataset_phys_t));
dsphys->ds_dir_obj = dd->dd_object;
dsphys->ds_flags = flags;
dsphys->ds_fsid_guid = unique_create();
(void) random_get_pseudo_bytes((void*)&dsphys->ds_guid,
sizeof (dsphys->ds_guid));
dsphys->ds_snapnames_zapobj =
zap_create_norm(mos, U8_TEXTPREP_TOUPPER, DMU_OT_DSL_DS_SNAP_MAP,
DMU_OT_NONE, 0, tx);
dsphys->ds_creation_time = gethrestime_sec();
dsphys->ds_creation_txg = tx->tx_txg == TXG_INITIAL ? 1 : tx->tx_txg;
if (origin == NULL) {
dsphys->ds_deadlist_obj = dsl_deadlist_alloc(mos, tx);
} else {
dsl_dataset_t *ohds; /* head of the origin snapshot */
dsphys->ds_prev_snap_obj = origin->ds_object;
dsphys->ds_prev_snap_txg =
dsl_dataset_phys(origin)->ds_creation_txg;
dsphys->ds_referenced_bytes =
dsl_dataset_phys(origin)->ds_referenced_bytes;
dsphys->ds_compressed_bytes =
dsl_dataset_phys(origin)->ds_compressed_bytes;
dsphys->ds_uncompressed_bytes =
dsl_dataset_phys(origin)->ds_uncompressed_bytes;
rrw_enter(&origin->ds_bp_rwlock, RW_READER, FTAG);
dsphys->ds_bp = dsl_dataset_phys(origin)->ds_bp;
rrw_exit(&origin->ds_bp_rwlock, FTAG);
/*
* Inherit flags that describe the dataset's contents
* (INCONSISTENT) or properties (Case Insensitive).
*/
dsphys->ds_flags |= dsl_dataset_phys(origin)->ds_flags &
(DS_FLAG_INCONSISTENT | DS_FLAG_CI_DATASET);
for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
if (zfeature_active(f, origin->ds_feature[f])) {
dsl_dataset_activate_feature(dsobj, f,
origin->ds_feature[f], tx);
}
}
dmu_buf_will_dirty(origin->ds_dbuf, tx);
dsl_dataset_phys(origin)->ds_num_children++;
VERIFY0(dsl_dataset_hold_obj(dp,
dsl_dir_phys(origin->ds_dir)->dd_head_dataset_obj,
FTAG, &ohds));
dsphys->ds_deadlist_obj = dsl_deadlist_clone(&ohds->ds_deadlist,
dsphys->ds_prev_snap_txg, dsphys->ds_prev_snap_obj, tx);
dsl_dataset_rele(ohds, FTAG);
if (spa_version(dp->dp_spa) >= SPA_VERSION_NEXT_CLONES) {
if (dsl_dataset_phys(origin)->ds_next_clones_obj == 0) {
dsl_dataset_phys(origin)->ds_next_clones_obj =
zap_create(mos,
DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx);
}
VERIFY0(zap_add_int(mos,
dsl_dataset_phys(origin)->ds_next_clones_obj,
dsobj, tx));
}
dmu_buf_will_dirty(dd->dd_dbuf, tx);
dsl_dir_phys(dd)->dd_origin_obj = origin->ds_object;
if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) {
if (dsl_dir_phys(origin->ds_dir)->dd_clones == 0) {
dmu_buf_will_dirty(origin->ds_dir->dd_dbuf, tx);
dsl_dir_phys(origin->ds_dir)->dd_clones =
zap_create(mos,
DMU_OT_DSL_CLONES, DMU_OT_NONE, 0, tx);
}
VERIFY0(zap_add_int(mos,
dsl_dir_phys(origin->ds_dir)->dd_clones,
dsobj, tx));
}
}
/* handle encryption */
dsl_dataset_create_crypt_sync(dsobj, dd, origin, dcp, tx);
if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE)
dsphys->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
dmu_buf_rele(dbuf, FTAG);
dmu_buf_will_dirty(dd->dd_dbuf, tx);
dsl_dir_phys(dd)->dd_head_dataset_obj = dsobj;
return (dsobj);
}
static void
dsl_dataset_zero_zil(dsl_dataset_t *ds, dmu_tx_t *tx)
{
objset_t *os;
VERIFY0(dmu_objset_from_ds(ds, &os));
if (bcmp(&os->os_zil_header, &zero_zil, sizeof (zero_zil)) != 0) {
dsl_pool_t *dp = ds->ds_dir->dd_pool;
zio_t *zio;
bzero(&os->os_zil_header, sizeof (os->os_zil_header));
if (os->os_encrypted)
os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_TRUE;
zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
dsl_dataset_sync(ds, zio, tx);
VERIFY0(zio_wait(zio));
/* dsl_dataset_sync_done will drop this reference. */
dmu_buf_add_ref(ds->ds_dbuf, ds);
dsl_dataset_sync_done(ds, tx);
}
}
uint64_t
dsl_dataset_create_sync(dsl_dir_t *pdd, const char *lastname,
dsl_dataset_t *origin, uint64_t flags, cred_t *cr,
dsl_crypto_params_t *dcp, dmu_tx_t *tx)
{
dsl_pool_t *dp = pdd->dd_pool;
uint64_t dsobj, ddobj;
dsl_dir_t *dd;
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(lastname[0] != '@');
/*
* Filesystems will eventually have their origin set to dp_origin_snap,
* but that's taken care of in dsl_dataset_create_sync_dd. When
* creating a filesystem, this function is called with origin equal to
* NULL.
*/
if (origin != NULL)
ASSERT3P(origin, !=, dp->dp_origin_snap);
ddobj = dsl_dir_create_sync(dp, pdd, lastname, tx);
VERIFY0(dsl_dir_hold_obj(dp, ddobj, lastname, FTAG, &dd));
dsobj = dsl_dataset_create_sync_dd(dd, origin, dcp,
flags & ~DS_CREATE_FLAG_NODIRTY, tx);
dsl_deleg_set_create_perms(dd, tx, cr);
/*
* If we are creating a clone and the livelist feature is enabled,
* add the entry DD_FIELD_LIVELIST to ZAP.
*/
if (origin != NULL &&
spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LIVELIST)) {
objset_t *mos = dd->dd_pool->dp_meta_objset;
dsl_dir_zapify(dd, tx);
uint64_t obj = dsl_deadlist_alloc(mos, tx);
VERIFY0(zap_add(mos, dd->dd_object, DD_FIELD_LIVELIST,
sizeof (uint64_t), 1, &obj, tx));
spa_feature_incr(dp->dp_spa, SPA_FEATURE_LIVELIST, tx);
}
/*
* Since we're creating a new node we know it's a leaf, so we can
* initialize the counts if the limit feature is active.
*/
if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_FS_SS_LIMIT)) {
uint64_t cnt = 0;
objset_t *os = dd->dd_pool->dp_meta_objset;
dsl_dir_zapify(dd, tx);
VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_FILESYSTEM_COUNT,
sizeof (cnt), 1, &cnt, tx));
VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_SNAPSHOT_COUNT,
sizeof (cnt), 1, &cnt, tx));
}
dsl_dir_rele(dd, FTAG);
/*
* If we are creating a clone, make sure we zero out any stale
* data from the origin snapshots zil header.
*/
if (origin != NULL && !(flags & DS_CREATE_FLAG_NODIRTY)) {
dsl_dataset_t *ds;
VERIFY0(dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
dsl_dataset_zero_zil(ds, tx);
dsl_dataset_rele(ds, FTAG);
}
return (dsobj);
}
/*
* The unique space in the head dataset can be calculated by subtracting
* the space used in the most recent snapshot, that is still being used
* in this file system, from the space currently in use. To figure out
* the space in the most recent snapshot still in use, we need to take
* the total space used in the snapshot and subtract out the space that
* has been freed up since the snapshot was taken.
*/
void
dsl_dataset_recalc_head_uniq(dsl_dataset_t *ds)
{
uint64_t mrs_used;
uint64_t dlused, dlcomp, dluncomp;
ASSERT(!ds->ds_is_snapshot);
if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0)
mrs_used = dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes;
else
mrs_used = 0;
dsl_deadlist_space(&ds->ds_deadlist, &dlused, &dlcomp, &dluncomp);
ASSERT3U(dlused, <=, mrs_used);
dsl_dataset_phys(ds)->ds_unique_bytes =
dsl_dataset_phys(ds)->ds_referenced_bytes - (mrs_used - dlused);
if (spa_version(ds->ds_dir->dd_pool->dp_spa) >=
SPA_VERSION_UNIQUE_ACCURATE)
dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
}
void
dsl_dataset_remove_from_next_clones(dsl_dataset_t *ds, uint64_t obj,
dmu_tx_t *tx)
{
objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
uint64_t count __maybe_unused;
int err;
ASSERT(dsl_dataset_phys(ds)->ds_num_children >= 2);
err = zap_remove_int(mos, dsl_dataset_phys(ds)->ds_next_clones_obj,
obj, tx);
/*
* The err should not be ENOENT, but a bug in a previous version
* of the code could cause upgrade_clones_cb() to not set
* ds_next_snap_obj when it should, leading to a missing entry.
* If we knew that the pool was created after
* SPA_VERSION_NEXT_CLONES, we could assert that it isn't
* ENOENT. However, at least we can check that we don't have
* too many entries in the next_clones_obj even after failing to
* remove this one.
*/
if (err != ENOENT)
VERIFY0(err);
ASSERT0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj,
&count));
ASSERT3U(count, <=, dsl_dataset_phys(ds)->ds_num_children - 2);
}
blkptr_t *
dsl_dataset_get_blkptr(dsl_dataset_t *ds)
{
return (&dsl_dataset_phys(ds)->ds_bp);
}
spa_t *
dsl_dataset_get_spa(dsl_dataset_t *ds)
{
return (ds->ds_dir->dd_pool->dp_spa);
}
void
dsl_dataset_dirty(dsl_dataset_t *ds, dmu_tx_t *tx)
{
dsl_pool_t *dp;
if (ds == NULL) /* this is the meta-objset */
return;
ASSERT(ds->ds_objset != NULL);
if (dsl_dataset_phys(ds)->ds_next_snap_obj != 0)
panic("dirtying snapshot!");
/* Must not dirty a dataset in the same txg where it got snapshotted. */
ASSERT3U(tx->tx_txg, >, dsl_dataset_phys(ds)->ds_prev_snap_txg);
dp = ds->ds_dir->dd_pool;
if (txg_list_add(&dp->dp_dirty_datasets, ds, tx->tx_txg)) {
objset_t *os = ds->ds_objset;
/* up the hold count until we can be written out */
dmu_buf_add_ref(ds->ds_dbuf, ds);
/* if this dataset is encrypted, grab a reference to the DCK */
if (ds->ds_dir->dd_crypto_obj != 0 &&
!os->os_raw_receive &&
!os->os_next_write_raw[tx->tx_txg & TXG_MASK]) {
ASSERT3P(ds->ds_key_mapping, !=, NULL);
key_mapping_add_ref(ds->ds_key_mapping, ds);
}
}
}
static int
dsl_dataset_snapshot_reserve_space(dsl_dataset_t *ds, dmu_tx_t *tx)
{
uint64_t asize;
if (!dmu_tx_is_syncing(tx))
return (0);
/*
* If there's an fs-only reservation, any blocks that might become
* owned by the snapshot dataset must be accommodated by space
* outside of the reservation.
*/
ASSERT(ds->ds_reserved == 0 || DS_UNIQUE_IS_ACCURATE(ds));
asize = MIN(dsl_dataset_phys(ds)->ds_unique_bytes, ds->ds_reserved);
if (asize > dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE))
return (SET_ERROR(ENOSPC));
/*
* Propagate any reserved space for this snapshot to other
* snapshot checks in this sync group.
*/
if (asize > 0)
dsl_dir_willuse_space(ds->ds_dir, asize, tx);
return (0);
}
int
dsl_dataset_snapshot_check_impl(dsl_dataset_t *ds, const char *snapname,
dmu_tx_t *tx, boolean_t recv, uint64_t cnt, cred_t *cr, proc_t *proc)
{
int error;
uint64_t value;
ds->ds_trysnap_txg = tx->tx_txg;
if (!dmu_tx_is_syncing(tx))
return (0);
/*
* We don't allow multiple snapshots of the same txg. If there
* is already one, try again.
*/
if (dsl_dataset_phys(ds)->ds_prev_snap_txg >= tx->tx_txg)
return (SET_ERROR(EAGAIN));
/*
* Check for conflicting snapshot name.
*/
error = dsl_dataset_snap_lookup(ds, snapname, &value);
if (error == 0)
return (SET_ERROR(EEXIST));
if (error != ENOENT)
return (error);
/*
* We don't allow taking snapshots of inconsistent datasets, such as
* those into which we are currently receiving. However, if we are
* creating this snapshot as part of a receive, this check will be
* executed atomically with respect to the completion of the receive
* itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
* case we ignore this, knowing it will be fixed up for us shortly in
* dmu_recv_end_sync().
*/
if (!recv && DS_IS_INCONSISTENT(ds))
return (SET_ERROR(EBUSY));
/*
* Skip the check for temporary snapshots or if we have already checked
* the counts in dsl_dataset_snapshot_check. This means we really only
* check the count here when we're receiving a stream.
*/
if (cnt != 0 && cr != NULL) {
error = dsl_fs_ss_limit_check(ds->ds_dir, cnt,
ZFS_PROP_SNAPSHOT_LIMIT, NULL, cr, proc);
if (error != 0)
return (error);
}
error = dsl_dataset_snapshot_reserve_space(ds, tx);
if (error != 0)
return (error);
return (0);
}
int
dsl_dataset_snapshot_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_snapshot_arg_t *ddsa = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
nvpair_t *pair;
int rv = 0;
/*
* Pre-compute how many total new snapshots will be created for each
* level in the tree and below. This is needed for validating the
* snapshot limit when either taking a recursive snapshot or when
* taking multiple snapshots.
*
* The problem is that the counts are not actually adjusted when
* we are checking, only when we finally sync. For a single snapshot,
* this is easy, the count will increase by 1 at each node up the tree,
* but its more complicated for the recursive/multiple snapshot case.
*
* The dsl_fs_ss_limit_check function does recursively check the count
* at each level up the tree but since it is validating each snapshot
* independently we need to be sure that we are validating the complete
* count for the entire set of snapshots. We do this by rolling up the
* counts for each component of the name into an nvlist and then
* checking each of those cases with the aggregated count.
*
* This approach properly handles not only the recursive snapshot
* case (where we get all of those on the ddsa_snaps list) but also
* the sibling case (e.g. snapshot a/b and a/c so that we will also
* validate the limit on 'a' using a count of 2).
*
* We validate the snapshot names in the third loop and only report
* name errors once.
*/
if (dmu_tx_is_syncing(tx)) {
char *nm;
nvlist_t *cnt_track = NULL;
cnt_track = fnvlist_alloc();
nm = kmem_alloc(MAXPATHLEN, KM_SLEEP);
/* Rollup aggregated counts into the cnt_track list */
for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL);
pair != NULL;
pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) {
char *pdelim;
uint64_t val;
(void) strlcpy(nm, nvpair_name(pair), MAXPATHLEN);
pdelim = strchr(nm, '@');
if (pdelim == NULL)
continue;
*pdelim = '\0';
do {
if (nvlist_lookup_uint64(cnt_track, nm,
&val) == 0) {
/* update existing entry */
fnvlist_add_uint64(cnt_track, nm,
val + 1);
} else {
/* add to list */
fnvlist_add_uint64(cnt_track, nm, 1);
}
pdelim = strrchr(nm, '/');
if (pdelim != NULL)
*pdelim = '\0';
} while (pdelim != NULL);
}
kmem_free(nm, MAXPATHLEN);
/* Check aggregated counts at each level */
for (pair = nvlist_next_nvpair(cnt_track, NULL);
pair != NULL; pair = nvlist_next_nvpair(cnt_track, pair)) {
int error = 0;
char *name;
uint64_t cnt = 0;
dsl_dataset_t *ds;
name = nvpair_name(pair);
cnt = fnvpair_value_uint64(pair);
ASSERT(cnt > 0);
error = dsl_dataset_hold(dp, name, FTAG, &ds);
if (error == 0) {
error = dsl_fs_ss_limit_check(ds->ds_dir, cnt,
ZFS_PROP_SNAPSHOT_LIMIT, NULL,
ddsa->ddsa_cr, ddsa->ddsa_proc);
dsl_dataset_rele(ds, FTAG);
}
if (error != 0) {
if (ddsa->ddsa_errors != NULL)
fnvlist_add_int32(ddsa->ddsa_errors,
name, error);
rv = error;
/* only report one error for this check */
break;
}
}
nvlist_free(cnt_track);
}
for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL);
pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) {
int error = 0;
dsl_dataset_t *ds;
char *name, *atp = NULL;
char dsname[ZFS_MAX_DATASET_NAME_LEN];
name = nvpair_name(pair);
if (strlen(name) >= ZFS_MAX_DATASET_NAME_LEN)
error = SET_ERROR(ENAMETOOLONG);
if (error == 0) {
atp = strchr(name, '@');
if (atp == NULL)
error = SET_ERROR(EINVAL);
if (error == 0)
(void) strlcpy(dsname, name, atp - name + 1);
}
if (error == 0)
error = dsl_dataset_hold(dp, dsname, FTAG, &ds);
if (error == 0) {
/* passing 0/NULL skips dsl_fs_ss_limit_check */
error = dsl_dataset_snapshot_check_impl(ds,
atp + 1, tx, B_FALSE, 0, NULL, NULL);
dsl_dataset_rele(ds, FTAG);
}
if (error != 0) {
if (ddsa->ddsa_errors != NULL) {
fnvlist_add_int32(ddsa->ddsa_errors,
name, error);
}
rv = error;
}
}
return (rv);
}
void
dsl_dataset_snapshot_sync_impl(dsl_dataset_t *ds, const char *snapname,
dmu_tx_t *tx)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
dmu_buf_t *dbuf;
dsl_dataset_phys_t *dsphys;
uint64_t dsobj, crtxg;
objset_t *mos = dp->dp_meta_objset;
static zil_header_t zero_zil __maybe_unused;
objset_t *os __maybe_unused;
ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));
/*
* If we are on an old pool, the zil must not be active, in which
* case it will be zeroed. Usually zil_suspend() accomplishes this.
*/
ASSERT(spa_version(dmu_tx_pool(tx)->dp_spa) >= SPA_VERSION_FAST_SNAP ||
dmu_objset_from_ds(ds, &os) != 0 ||
bcmp(&os->os_phys->os_zil_header, &zero_zil,
sizeof (zero_zil)) == 0);
/* Should not snapshot a dirty dataset. */
ASSERT(!txg_list_member(&ds->ds_dir->dd_pool->dp_dirty_datasets,
ds, tx->tx_txg));
dsl_fs_ss_count_adjust(ds->ds_dir, 1, DD_FIELD_SNAPSHOT_COUNT, tx);
/*
* The origin's ds_creation_txg has to be < TXG_INITIAL
*/
if (strcmp(snapname, ORIGIN_DIR_NAME) == 0)
crtxg = 1;
else
crtxg = tx->tx_txg;
dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0,
DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx);
VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf));
dmu_buf_will_dirty(dbuf, tx);
dsphys = dbuf->db_data;
bzero(dsphys, sizeof (dsl_dataset_phys_t));
dsphys->ds_dir_obj = ds->ds_dir->dd_object;
dsphys->ds_fsid_guid = unique_create();
(void) random_get_pseudo_bytes((void*)&dsphys->ds_guid,
sizeof (dsphys->ds_guid));
dsphys->ds_prev_snap_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
dsphys->ds_prev_snap_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;
dsphys->ds_next_snap_obj = ds->ds_object;
dsphys->ds_num_children = 1;
dsphys->ds_creation_time = gethrestime_sec();
dsphys->ds_creation_txg = crtxg;
dsphys->ds_deadlist_obj = dsl_dataset_phys(ds)->ds_deadlist_obj;
dsphys->ds_referenced_bytes = dsl_dataset_phys(ds)->ds_referenced_bytes;
dsphys->ds_compressed_bytes = dsl_dataset_phys(ds)->ds_compressed_bytes;
dsphys->ds_uncompressed_bytes =
dsl_dataset_phys(ds)->ds_uncompressed_bytes;
dsphys->ds_flags = dsl_dataset_phys(ds)->ds_flags;
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
dsphys->ds_bp = dsl_dataset_phys(ds)->ds_bp;
rrw_exit(&ds->ds_bp_rwlock, FTAG);
dmu_buf_rele(dbuf, FTAG);
for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
if (zfeature_active(f, ds->ds_feature[f])) {
dsl_dataset_activate_feature(dsobj, f,
ds->ds_feature[f], tx);
}
}
ASSERT3U(ds->ds_prev != 0, ==,
dsl_dataset_phys(ds)->ds_prev_snap_obj != 0);
if (ds->ds_prev) {
uint64_t next_clones_obj =
dsl_dataset_phys(ds->ds_prev)->ds_next_clones_obj;
ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj ==
ds->ds_object ||
dsl_dataset_phys(ds->ds_prev)->ds_num_children > 1);
if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj ==
ds->ds_object) {
dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, ==,
dsl_dataset_phys(ds->ds_prev)->ds_creation_txg);
dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj = dsobj;
} else if (next_clones_obj != 0) {
dsl_dataset_remove_from_next_clones(ds->ds_prev,
dsphys->ds_next_snap_obj, tx);
VERIFY0(zap_add_int(mos,
next_clones_obj, dsobj, tx));
}
}
/*
* If we have a reference-reservation on this dataset, we will
* need to increase the amount of refreservation being charged
* since our unique space is going to zero.
*/
if (ds->ds_reserved) {
int64_t delta;
ASSERT(DS_UNIQUE_IS_ACCURATE(ds));
delta = MIN(dsl_dataset_phys(ds)->ds_unique_bytes,
ds->ds_reserved);
dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV,
delta, 0, 0, tx);
}
dmu_buf_will_dirty(ds->ds_dbuf, tx);
dsl_dataset_phys(ds)->ds_deadlist_obj =
dsl_deadlist_clone(&ds->ds_deadlist, UINT64_MAX,
dsl_dataset_phys(ds)->ds_prev_snap_obj, tx);
dsl_deadlist_close(&ds->ds_deadlist);
dsl_deadlist_open(&ds->ds_deadlist, mos,
dsl_dataset_phys(ds)->ds_deadlist_obj);
dsl_deadlist_add_key(&ds->ds_deadlist,
dsl_dataset_phys(ds)->ds_prev_snap_txg, tx);
dsl_bookmark_snapshotted(ds, tx);
if (dsl_dataset_remap_deadlist_exists(ds)) {
uint64_t remap_deadlist_obj =
dsl_dataset_get_remap_deadlist_object(ds);
/*
* Move the remap_deadlist to the snapshot. The head
* will create a new remap deadlist on demand, from
* dsl_dataset_block_remapped().
*/
dsl_dataset_unset_remap_deadlist_object(ds, tx);
dsl_deadlist_close(&ds->ds_remap_deadlist);
dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx);
VERIFY0(zap_add(mos, dsobj, DS_FIELD_REMAP_DEADLIST,
sizeof (remap_deadlist_obj), 1, &remap_deadlist_obj, tx));
}
/*
* Create a ivset guid for this snapshot if the dataset is
* encrypted. This may be overridden by a raw receive. A
* previous implementation of this code did not have this
* field as part of the on-disk format for ZFS encryption
* (see errata #4). As part of the remediation for this
* issue, we ask the user to enable the bookmark_v2 feature
* which is now a dependency of the encryption feature. We
* use this as a heuristic to determine when the user has
* elected to correct any datasets created with the old code.
* As a result, we only do this step if the bookmark_v2
* feature is enabled, which limits the number of states a
* given pool / dataset can be in with regards to terms of
* correcting the issue.
*/
if (ds->ds_dir->dd_crypto_obj != 0 &&
spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_BOOKMARK_V2)) {
uint64_t ivset_guid = unique_create();
dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx);
VERIFY0(zap_add(mos, dsobj, DS_FIELD_IVSET_GUID,
sizeof (ivset_guid), 1, &ivset_guid, tx));
}
ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, <, tx->tx_txg);
dsl_dataset_phys(ds)->ds_prev_snap_obj = dsobj;
dsl_dataset_phys(ds)->ds_prev_snap_txg = crtxg;
dsl_dataset_phys(ds)->ds_unique_bytes = 0;
if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE)
dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_UNIQUE_ACCURATE;
VERIFY0(zap_add(mos, dsl_dataset_phys(ds)->ds_snapnames_zapobj,
snapname, 8, 1, &dsobj, tx));
if (ds->ds_prev)
dsl_dataset_rele(ds->ds_prev, ds);
VERIFY0(dsl_dataset_hold_obj(dp,
dsl_dataset_phys(ds)->ds_prev_snap_obj, ds, &ds->ds_prev));
dsl_scan_ds_snapshotted(ds, tx);
dsl_dir_snap_cmtime_update(ds->ds_dir);
spa_history_log_internal_ds(ds->ds_prev, "snapshot", tx, " ");
}
void
dsl_dataset_snapshot_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_snapshot_arg_t *ddsa = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
nvpair_t *pair;
for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL);
pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) {
dsl_dataset_t *ds;
char *name, *atp;
char dsname[ZFS_MAX_DATASET_NAME_LEN];
name = nvpair_name(pair);
atp = strchr(name, '@');
(void) strlcpy(dsname, name, atp - name + 1);
VERIFY0(dsl_dataset_hold(dp, dsname, FTAG, &ds));
dsl_dataset_snapshot_sync_impl(ds, atp + 1, tx);
if (ddsa->ddsa_props != NULL) {
dsl_props_set_sync_impl(ds->ds_prev,
ZPROP_SRC_LOCAL, ddsa->ddsa_props, tx);
}
dsl_dataset_rele(ds, FTAG);
}
}
/*
* The snapshots must all be in the same pool.
* All-or-nothing: if there are any failures, nothing will be modified.
*/
int
dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors)
{
dsl_dataset_snapshot_arg_t ddsa;
nvpair_t *pair;
boolean_t needsuspend;
int error;
spa_t *spa;
char *firstname;
nvlist_t *suspended = NULL;
pair = nvlist_next_nvpair(snaps, NULL);
if (pair == NULL)
return (0);
firstname = nvpair_name(pair);
error = spa_open(firstname, &spa, FTAG);
if (error != 0)
return (error);
needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP);
spa_close(spa, FTAG);
if (needsuspend) {
suspended = fnvlist_alloc();
for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
pair = nvlist_next_nvpair(snaps, pair)) {
char fsname[ZFS_MAX_DATASET_NAME_LEN];
char *snapname = nvpair_name(pair);
char *atp;
void *cookie;
atp = strchr(snapname, '@');
if (atp == NULL) {
error = SET_ERROR(EINVAL);
break;
}
(void) strlcpy(fsname, snapname, atp - snapname + 1);
error = zil_suspend(fsname, &cookie);
if (error != 0)
break;
fnvlist_add_uint64(suspended, fsname,
(uintptr_t)cookie);
}
}
ddsa.ddsa_snaps = snaps;
ddsa.ddsa_props = props;
ddsa.ddsa_errors = errors;
ddsa.ddsa_cr = CRED();
ddsa.ddsa_proc = curproc;
if (error == 0) {
error = dsl_sync_task(firstname, dsl_dataset_snapshot_check,
dsl_dataset_snapshot_sync, &ddsa,
fnvlist_num_pairs(snaps) * 3, ZFS_SPACE_CHECK_NORMAL);
}
if (suspended != NULL) {
for (pair = nvlist_next_nvpair(suspended, NULL); pair != NULL;
pair = nvlist_next_nvpair(suspended, pair)) {
zil_resume((void *)(uintptr_t)
fnvpair_value_uint64(pair));
}
fnvlist_free(suspended);
}
if (error == 0) {
for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
pair = nvlist_next_nvpair(snaps, pair)) {
zvol_create_minor(nvpair_name(pair));
}
}
return (error);
}
typedef struct dsl_dataset_snapshot_tmp_arg {
const char *ddsta_fsname;
const char *ddsta_snapname;
minor_t ddsta_cleanup_minor;
const char *ddsta_htag;
} dsl_dataset_snapshot_tmp_arg_t;
static int
dsl_dataset_snapshot_tmp_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_snapshot_tmp_arg_t *ddsta = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *ds;
int error;
error = dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds);
if (error != 0)
return (error);
/* NULL cred means no limit check for tmp snapshot */
error = dsl_dataset_snapshot_check_impl(ds, ddsta->ddsta_snapname,
tx, B_FALSE, 0, NULL, NULL);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
if (spa_version(dp->dp_spa) < SPA_VERSION_USERREFS) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(ENOTSUP));
}
error = dsl_dataset_user_hold_check_one(NULL, ddsta->ddsta_htag,
B_TRUE, tx);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
dsl_dataset_rele(ds, FTAG);
return (0);
}
static void
dsl_dataset_snapshot_tmp_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_snapshot_tmp_arg_t *ddsta = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *ds = NULL;
VERIFY0(dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds));
dsl_dataset_snapshot_sync_impl(ds, ddsta->ddsta_snapname, tx);
dsl_dataset_user_hold_sync_one(ds->ds_prev, ddsta->ddsta_htag,
ddsta->ddsta_cleanup_minor, gethrestime_sec(), tx);
dsl_destroy_snapshot_sync_impl(ds->ds_prev, B_TRUE, tx);
dsl_dataset_rele(ds, FTAG);
}
int
dsl_dataset_snapshot_tmp(const char *fsname, const char *snapname,
minor_t cleanup_minor, const char *htag)
{
dsl_dataset_snapshot_tmp_arg_t ddsta;
int error;
spa_t *spa;
boolean_t needsuspend;
void *cookie;
ddsta.ddsta_fsname = fsname;
ddsta.ddsta_snapname = snapname;
ddsta.ddsta_cleanup_minor = cleanup_minor;
ddsta.ddsta_htag = htag;
error = spa_open(fsname, &spa, FTAG);
if (error != 0)
return (error);
needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP);
spa_close(spa, FTAG);
if (needsuspend) {
error = zil_suspend(fsname, &cookie);
if (error != 0)
return (error);
}
error = dsl_sync_task(fsname, dsl_dataset_snapshot_tmp_check,
dsl_dataset_snapshot_tmp_sync, &ddsta, 3, ZFS_SPACE_CHECK_RESERVED);
if (needsuspend)
zil_resume(cookie);
return (error);
}
void
dsl_dataset_sync(dsl_dataset_t *ds, zio_t *zio, dmu_tx_t *tx)
{
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(ds->ds_objset != NULL);
ASSERT(dsl_dataset_phys(ds)->ds_next_snap_obj == 0);
/*
* in case we had to change ds_fsid_guid when we opened it,
* sync it out now.
*/
dmu_buf_will_dirty(ds->ds_dbuf, tx);
dsl_dataset_phys(ds)->ds_fsid_guid = ds->ds_fsid_guid;
if (ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] != 0) {
VERIFY0(zap_update(tx->tx_pool->dp_meta_objset,
ds->ds_object, DS_FIELD_RESUME_OBJECT, 8, 1,
&ds->ds_resume_object[tx->tx_txg & TXG_MASK], tx));
VERIFY0(zap_update(tx->tx_pool->dp_meta_objset,
ds->ds_object, DS_FIELD_RESUME_OFFSET, 8, 1,
&ds->ds_resume_offset[tx->tx_txg & TXG_MASK], tx));
VERIFY0(zap_update(tx->tx_pool->dp_meta_objset,
ds->ds_object, DS_FIELD_RESUME_BYTES, 8, 1,
&ds->ds_resume_bytes[tx->tx_txg & TXG_MASK], tx));
ds->ds_resume_object[tx->tx_txg & TXG_MASK] = 0;
ds->ds_resume_offset[tx->tx_txg & TXG_MASK] = 0;
ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] = 0;
}
dmu_objset_sync(ds->ds_objset, zio, tx);
for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
if (zfeature_active(f, ds->ds_feature_activation[f])) {
if (zfeature_active(f, ds->ds_feature[f]))
continue;
dsl_dataset_activate_feature(ds->ds_object, f,
ds->ds_feature_activation[f], tx);
ds->ds_feature[f] = ds->ds_feature_activation[f];
}
}
}
/*
* Check if the percentage of blocks shared between the clone and the
* snapshot (as opposed to those that are clone only) is below a certain
* threshold
*/
static boolean_t
dsl_livelist_should_disable(dsl_dataset_t *ds)
{
uint64_t used, referenced;
int percent_shared;
used = dsl_dir_get_usedds(ds->ds_dir);
referenced = dsl_get_referenced(ds);
ASSERT3U(referenced, >=, 0);
ASSERT3U(used, >=, 0);
if (referenced == 0)
return (B_FALSE);
percent_shared = (100 * (referenced - used)) / referenced;
if (percent_shared <= zfs_livelist_min_percent_shared)
return (B_TRUE);
return (B_FALSE);
}
/*
* Check if it is possible to combine two livelist entries into one.
* This is the case if the combined number of 'live' blkptrs (ALLOCs that
* don't have a matching FREE) is under the maximum sublist size.
* We check this by subtracting twice the total number of frees from the total
* number of blkptrs. FREEs are counted twice because each FREE blkptr
* will cancel out an ALLOC blkptr when the livelist is processed.
*/
static boolean_t
dsl_livelist_should_condense(dsl_deadlist_entry_t *first,
dsl_deadlist_entry_t *next)
{
uint64_t total_free = first->dle_bpobj.bpo_phys->bpo_num_freed +
next->dle_bpobj.bpo_phys->bpo_num_freed;
uint64_t total_entries = first->dle_bpobj.bpo_phys->bpo_num_blkptrs +
next->dle_bpobj.bpo_phys->bpo_num_blkptrs;
if ((total_entries - (2 * total_free)) < zfs_livelist_max_entries)
return (B_TRUE);
return (B_FALSE);
}
typedef struct try_condense_arg {
spa_t *spa;
dsl_dataset_t *ds;
} try_condense_arg_t;
/*
* Iterate over the livelist entries, searching for a pair to condense.
* A nonzero return value means stop, 0 means keep looking.
*/
static int
dsl_livelist_try_condense(void *arg, dsl_deadlist_entry_t *first)
{
try_condense_arg_t *tca = arg;
spa_t *spa = tca->spa;
dsl_dataset_t *ds = tca->ds;
dsl_deadlist_t *ll = &ds->ds_dir->dd_livelist;
dsl_deadlist_entry_t *next;
/* The condense thread has not yet been created at import */
if (spa->spa_livelist_condense_zthr == NULL)
return (1);
/* A condense is already in progress */
if (spa->spa_to_condense.ds != NULL)
return (1);
next = AVL_NEXT(&ll->dl_tree, &first->dle_node);
/* The livelist has only one entry - don't condense it */
if (next == NULL)
return (1);
/* Next is the newest entry - don't condense it */
if (AVL_NEXT(&ll->dl_tree, &next->dle_node) == NULL)
return (1);
/* This pair is not ready to condense but keep looking */
if (!dsl_livelist_should_condense(first, next))
return (0);
/*
* Add a ref to prevent the dataset from being evicted while
* the condense zthr or synctask are running. Ref will be
* released at the end of the condense synctask
*/
dmu_buf_add_ref(ds->ds_dbuf, spa);
spa->spa_to_condense.ds = ds;
spa->spa_to_condense.first = first;
spa->spa_to_condense.next = next;
spa->spa_to_condense.syncing = B_FALSE;
spa->spa_to_condense.cancelled = B_FALSE;
zthr_wakeup(spa->spa_livelist_condense_zthr);
return (1);
}
static void
dsl_flush_pending_livelist(dsl_dataset_t *ds, dmu_tx_t *tx)
{
dsl_dir_t *dd = ds->ds_dir;
spa_t *spa = ds->ds_dir->dd_pool->dp_spa;
dsl_deadlist_entry_t *last = dsl_deadlist_last(&dd->dd_livelist);
/* Check if we need to add a new sub-livelist */
if (last == NULL) {
/* The livelist is empty */
dsl_deadlist_add_key(&dd->dd_livelist,
tx->tx_txg - 1, tx);
} else if (spa_sync_pass(spa) == 1) {
/*
* Check if the newest entry is full. If it is, make a new one.
* We only do this once per sync because we could overfill a
* sublist in one sync pass and don't want to add another entry
* for a txg that is already represented. This ensures that
* blkptrs born in the same txg are stored in the same sublist.
*/
bpobj_t bpobj = last->dle_bpobj;
uint64_t all = bpobj.bpo_phys->bpo_num_blkptrs;
uint64_t free = bpobj.bpo_phys->bpo_num_freed;
uint64_t alloc = all - free;
if (alloc > zfs_livelist_max_entries) {
dsl_deadlist_add_key(&dd->dd_livelist,
tx->tx_txg - 1, tx);
}
}
/* Insert each entry into the on-disk livelist */
bplist_iterate(&dd->dd_pending_allocs,
dsl_deadlist_insert_alloc_cb, &dd->dd_livelist, tx);
bplist_iterate(&dd->dd_pending_frees,
dsl_deadlist_insert_free_cb, &dd->dd_livelist, tx);
/* Attempt to condense every pair of adjacent entries */
try_condense_arg_t arg = {
.spa = spa,
.ds = ds
};
dsl_deadlist_iterate(&dd->dd_livelist, dsl_livelist_try_condense,
&arg);
}
void
dsl_dataset_sync_done(dsl_dataset_t *ds, dmu_tx_t *tx)
{
objset_t *os = ds->ds_objset;
bplist_iterate(&ds->ds_pending_deadlist,
dsl_deadlist_insert_alloc_cb, &ds->ds_deadlist, tx);
if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist)) {
dsl_flush_pending_livelist(ds, tx);
if (dsl_livelist_should_disable(ds)) {
dsl_dir_remove_livelist(ds->ds_dir, tx, B_TRUE);
}
}
dsl_bookmark_sync_done(ds, tx);
multilist_destroy(&os->os_synced_dnodes);
if (os->os_encrypted)
os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_FALSE;
else
ASSERT0(os->os_next_write_raw[tx->tx_txg & TXG_MASK]);
ASSERT(!dmu_objset_is_dirty(os, dmu_tx_get_txg(tx)));
dmu_buf_rele(ds->ds_dbuf, ds);
}
int
get_clones_stat_impl(dsl_dataset_t *ds, nvlist_t *val)
{
uint64_t count = 0;
objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
zap_cursor_t zc;
zap_attribute_t za;
ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
/*
* There may be missing entries in ds_next_clones_obj
* due to a bug in a previous version of the code.
* Only trust it if it has the right number of entries.
*/
if (dsl_dataset_phys(ds)->ds_next_clones_obj != 0) {
VERIFY0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj,
&count));
}
if (count != dsl_dataset_phys(ds)->ds_num_children - 1) {
return (SET_ERROR(ENOENT));
}
for (zap_cursor_init(&zc, mos,
dsl_dataset_phys(ds)->ds_next_clones_obj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
dsl_dataset_t *clone;
char buf[ZFS_MAX_DATASET_NAME_LEN];
VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool,
za.za_first_integer, FTAG, &clone));
dsl_dir_name(clone->ds_dir, buf);
fnvlist_add_boolean(val, buf);
dsl_dataset_rele(clone, FTAG);
}
zap_cursor_fini(&zc);
return (0);
}
void
get_clones_stat(dsl_dataset_t *ds, nvlist_t *nv)
{
nvlist_t *propval = fnvlist_alloc();
nvlist_t *val = fnvlist_alloc();
if (get_clones_stat_impl(ds, val) == 0) {
fnvlist_add_nvlist(propval, ZPROP_VALUE, val);
fnvlist_add_nvlist(nv, zfs_prop_to_name(ZFS_PROP_CLONES),
propval);
}
nvlist_free(val);
nvlist_free(propval);
}
/*
* Returns a string that represents the receive resume stats token. It should
* be freed with strfree().
*/
char *
get_receive_resume_stats_impl(dsl_dataset_t *ds)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
if (dsl_dataset_has_resume_receive_state(ds)) {
char *str;
void *packed;
uint8_t *compressed;
uint64_t val;
nvlist_t *token_nv = fnvlist_alloc();
size_t packed_size, compressed_size;
if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val) == 0) {
fnvlist_add_uint64(token_nv, "fromguid", val);
}
if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val) == 0) {
fnvlist_add_uint64(token_nv, "object", val);
}
if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val) == 0) {
fnvlist_add_uint64(token_nv, "offset", val);
}
if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_BYTES, sizeof (val), 1, &val) == 0) {
fnvlist_add_uint64(token_nv, "bytes", val);
}
if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val) == 0) {
fnvlist_add_uint64(token_nv, "toguid", val);
}
char buf[MAXNAMELEN];
if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_TONAME, 1, sizeof (buf), buf) == 0) {
fnvlist_add_string(token_nv, "toname", buf);
}
if (zap_contains(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_LARGEBLOCK) == 0) {
fnvlist_add_boolean(token_nv, "largeblockok");
}
if (zap_contains(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_EMBEDOK) == 0) {
fnvlist_add_boolean(token_nv, "embedok");
}
if (zap_contains(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_COMPRESSOK) == 0) {
fnvlist_add_boolean(token_nv, "compressok");
}
if (zap_contains(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_RAWOK) == 0) {
fnvlist_add_boolean(token_nv, "rawok");
}
if (dsl_dataset_feature_is_active(ds,
SPA_FEATURE_REDACTED_DATASETS)) {
uint64_t num_redact_snaps;
uint64_t *redact_snaps;
VERIFY(dsl_dataset_get_uint64_array_feature(ds,
SPA_FEATURE_REDACTED_DATASETS, &num_redact_snaps,
&redact_snaps));
fnvlist_add_uint64_array(token_nv, "redact_snaps",
redact_snaps, num_redact_snaps);
}
if (zap_contains(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS) == 0) {
uint64_t num_redact_snaps, int_size;
uint64_t *redact_snaps;
VERIFY0(zap_length(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS, &int_size,
&num_redact_snaps));
ASSERT3U(int_size, ==, sizeof (uint64_t));
redact_snaps = kmem_alloc(int_size * num_redact_snaps,
KM_SLEEP);
VERIFY0(zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS, int_size,
num_redact_snaps, redact_snaps));
fnvlist_add_uint64_array(token_nv, "book_redact_snaps",
redact_snaps, num_redact_snaps);
kmem_free(redact_snaps, int_size * num_redact_snaps);
}
packed = fnvlist_pack(token_nv, &packed_size);
fnvlist_free(token_nv);
compressed = kmem_alloc(packed_size, KM_SLEEP);
compressed_size = gzip_compress(packed, compressed,
packed_size, packed_size, 6);
zio_cksum_t cksum;
fletcher_4_native_varsize(compressed, compressed_size, &cksum);
size_t alloc_size = compressed_size * 2 + 1;
str = kmem_alloc(alloc_size, KM_SLEEP);
for (int i = 0; i < compressed_size; i++) {
size_t offset = i * 2;
(void) snprintf(str + offset, alloc_size - offset,
"%02x", compressed[i]);
}
str[compressed_size * 2] = '\0';
char *propval = kmem_asprintf("%u-%llx-%llx-%s",
ZFS_SEND_RESUME_TOKEN_VERSION,
(longlong_t)cksum.zc_word[0],
(longlong_t)packed_size, str);
kmem_free(packed, packed_size);
kmem_free(str, alloc_size);
kmem_free(compressed, packed_size);
return (propval);
}
return (kmem_strdup(""));
}
/*
* Returns a string that represents the receive resume stats token of the
* dataset's child. It should be freed with strfree().
*/
char *
get_child_receive_stats(dsl_dataset_t *ds)
{
char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];
dsl_dataset_t *recv_ds;
dsl_dataset_name(ds, recvname);
if (strlcat(recvname, "/", sizeof (recvname)) <
sizeof (recvname) &&
strlcat(recvname, recv_clone_name, sizeof (recvname)) <
sizeof (recvname) &&
dsl_dataset_hold(ds->ds_dir->dd_pool, recvname, FTAG,
&recv_ds) == 0) {
char *propval = get_receive_resume_stats_impl(recv_ds);
dsl_dataset_rele(recv_ds, FTAG);
return (propval);
}
return (kmem_strdup(""));
}
static void
get_receive_resume_stats(dsl_dataset_t *ds, nvlist_t *nv)
{
char *propval = get_receive_resume_stats_impl(ds);
if (strcmp(propval, "") != 0) {
dsl_prop_nvlist_add_string(nv,
ZFS_PROP_RECEIVE_RESUME_TOKEN, propval);
} else {
char *childval = get_child_receive_stats(ds);
if (strcmp(childval, "") != 0) {
dsl_prop_nvlist_add_string(nv,
ZFS_PROP_RECEIVE_RESUME_TOKEN, childval);
}
kmem_strfree(childval);
}
kmem_strfree(propval);
}
uint64_t
dsl_get_refratio(dsl_dataset_t *ds)
{
uint64_t ratio = dsl_dataset_phys(ds)->ds_compressed_bytes == 0 ? 100 :
(dsl_dataset_phys(ds)->ds_uncompressed_bytes * 100 /
dsl_dataset_phys(ds)->ds_compressed_bytes);
return (ratio);
}
uint64_t
dsl_get_logicalreferenced(dsl_dataset_t *ds)
{
return (dsl_dataset_phys(ds)->ds_uncompressed_bytes);
}
uint64_t
dsl_get_compressratio(dsl_dataset_t *ds)
{
if (ds->ds_is_snapshot) {
return (dsl_get_refratio(ds));
} else {
dsl_dir_t *dd = ds->ds_dir;
mutex_enter(&dd->dd_lock);
uint64_t val = dsl_dir_get_compressratio(dd);
mutex_exit(&dd->dd_lock);
return (val);
}
}
uint64_t
dsl_get_used(dsl_dataset_t *ds)
{
if (ds->ds_is_snapshot) {
return (dsl_dataset_phys(ds)->ds_unique_bytes);
} else {
dsl_dir_t *dd = ds->ds_dir;
mutex_enter(&dd->dd_lock);
uint64_t val = dsl_dir_get_used(dd);
mutex_exit(&dd->dd_lock);
return (val);
}
}
uint64_t
dsl_get_creation(dsl_dataset_t *ds)
{
return (dsl_dataset_phys(ds)->ds_creation_time);
}
uint64_t
dsl_get_creationtxg(dsl_dataset_t *ds)
{
return (dsl_dataset_phys(ds)->ds_creation_txg);
}
uint64_t
dsl_get_refquota(dsl_dataset_t *ds)
{
return (ds->ds_quota);
}
uint64_t
dsl_get_refreservation(dsl_dataset_t *ds)
{
return (ds->ds_reserved);
}
uint64_t
dsl_get_guid(dsl_dataset_t *ds)
{
return (dsl_dataset_phys(ds)->ds_guid);
}
uint64_t
dsl_get_unique(dsl_dataset_t *ds)
{
return (dsl_dataset_phys(ds)->ds_unique_bytes);
}
uint64_t
dsl_get_objsetid(dsl_dataset_t *ds)
{
return (ds->ds_object);
}
uint64_t
dsl_get_userrefs(dsl_dataset_t *ds)
{
return (ds->ds_userrefs);
}
uint64_t
dsl_get_defer_destroy(dsl_dataset_t *ds)
{
return (DS_IS_DEFER_DESTROY(ds) ? 1 : 0);
}
uint64_t
dsl_get_referenced(dsl_dataset_t *ds)
{
return (dsl_dataset_phys(ds)->ds_referenced_bytes);
}
uint64_t
dsl_get_numclones(dsl_dataset_t *ds)
{
ASSERT(ds->ds_is_snapshot);
return (dsl_dataset_phys(ds)->ds_num_children - 1);
}
uint64_t
dsl_get_inconsistent(dsl_dataset_t *ds)
{
return ((dsl_dataset_phys(ds)->ds_flags & DS_FLAG_INCONSISTENT) ?
1 : 0);
}
uint64_t
dsl_get_redacted(dsl_dataset_t *ds)
{
return (dsl_dataset_feature_is_active(ds,
SPA_FEATURE_REDACTED_DATASETS));
}
uint64_t
dsl_get_available(dsl_dataset_t *ds)
{
uint64_t refdbytes = dsl_get_referenced(ds);
uint64_t availbytes = dsl_dir_space_available(ds->ds_dir,
NULL, 0, TRUE);
if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) {
availbytes +=
ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes;
}
if (ds->ds_quota != 0) {
/*
* Adjust available bytes according to refquota
*/
if (refdbytes < ds->ds_quota) {
availbytes = MIN(availbytes,
ds->ds_quota - refdbytes);
} else {
availbytes = 0;
}
}
return (availbytes);
}
int
dsl_get_written(dsl_dataset_t *ds, uint64_t *written)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
dsl_dataset_t *prev;
int err = dsl_dataset_hold_obj(dp,
dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);
if (err == 0) {
uint64_t comp, uncomp;
err = dsl_dataset_space_written(prev, ds, written,
&comp, &uncomp);
dsl_dataset_rele(prev, FTAG);
}
return (err);
}
/*
* 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN.
*/
int
dsl_get_prev_snap(dsl_dataset_t *ds, char *snap)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
if (ds->ds_prev != NULL && ds->ds_prev != dp->dp_origin_snap) {
dsl_dataset_name(ds->ds_prev, snap);
return (0);
} else {
return (SET_ERROR(ENOENT));
}
}
void
dsl_get_redact_snaps(dsl_dataset_t *ds, nvlist_t *propval)
{
uint64_t nsnaps;
uint64_t *snaps;
if (dsl_dataset_get_uint64_array_feature(ds,
SPA_FEATURE_REDACTED_DATASETS, &nsnaps, &snaps)) {
fnvlist_add_uint64_array(propval, ZPROP_VALUE, snaps,
nsnaps);
}
}
/*
* Returns the mountpoint property and source for the given dataset in the value
* and source buffers. The value buffer must be at least as large as MAXPATHLEN
* and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN.
* Returns 0 on success and an error on failure.
*/
int
dsl_get_mountpoint(dsl_dataset_t *ds, const char *dsname, char *value,
char *source)
{
int error;
dsl_pool_t *dp = ds->ds_dir->dd_pool;
/* Retrieve the mountpoint value stored in the zap object */
error = dsl_prop_get_ds(ds, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT), 1,
ZAP_MAXVALUELEN, value, source);
if (error != 0) {
return (error);
}
/*
* Process the dsname and source to find the full mountpoint string.
* Can be skipped for 'legacy' or 'none'.
*/
if (value[0] == '/') {
char *buf = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP);
char *root = buf;
const char *relpath;
/*
* If we inherit the mountpoint, even from a dataset
* with a received value, the source will be the path of
* the dataset we inherit from. If source is
* ZPROP_SOURCE_VAL_RECVD, the received value is not
* inherited.
*/
if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0) {
relpath = "";
} else {
ASSERT0(strncmp(dsname, source, strlen(source)));
relpath = dsname + strlen(source);
if (relpath[0] == '/')
relpath++;
}
spa_altroot(dp->dp_spa, root, ZAP_MAXVALUELEN);
/*
* Special case an alternate root of '/'. This will
* avoid having multiple leading slashes in the
* mountpoint path.
*/
if (strcmp(root, "/") == 0)
root++;
/*
* If the mountpoint is '/' then skip over this
* if we are obtaining either an alternate root or
* an inherited mountpoint.
*/
char *mnt = value;
if (value[1] == '\0' && (root[0] != '\0' ||
relpath[0] != '\0'))
mnt = value + 1;
if (relpath[0] == '\0') {
(void) snprintf(value, ZAP_MAXVALUELEN, "%s%s",
root, mnt);
} else {
(void) snprintf(value, ZAP_MAXVALUELEN, "%s%s%s%s",
root, mnt, relpath[0] == '@' ? "" : "/",
relpath);
}
kmem_free(buf, ZAP_MAXVALUELEN);
}
return (0);
}
void
dsl_dataset_stats(dsl_dataset_t *ds, nvlist_t *nv)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
ASSERT(dsl_pool_config_held(dp));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRATIO,
dsl_get_refratio(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_LOGICALREFERENCED,
dsl_get_logicalreferenced(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_COMPRESSRATIO,
dsl_get_compressratio(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USED,
dsl_get_used(ds));
if (ds->ds_is_snapshot) {
get_clones_stat(ds, nv);
} else {
char buf[ZFS_MAX_DATASET_NAME_LEN];
if (dsl_get_prev_snap(ds, buf) == 0)
dsl_prop_nvlist_add_string(nv, ZFS_PROP_PREV_SNAP,
buf);
dsl_dir_stats(ds->ds_dir, nv);
}
nvlist_t *propval = fnvlist_alloc();
dsl_get_redact_snaps(ds, propval);
fnvlist_add_nvlist(nv, zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS),
propval);
nvlist_free(propval);
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_AVAILABLE,
dsl_get_available(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFERENCED,
dsl_get_referenced(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATION,
dsl_get_creation(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATETXG,
dsl_get_creationtxg(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFQUOTA,
dsl_get_refquota(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRESERVATION,
dsl_get_refreservation(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_GUID,
dsl_get_guid(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_UNIQUE,
dsl_get_unique(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_OBJSETID,
dsl_get_objsetid(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERREFS,
dsl_get_userrefs(ds));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_DEFER_DESTROY,
dsl_get_defer_destroy(ds));
dsl_dataset_crypt_stats(ds, nv);
if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
uint64_t written;
if (dsl_get_written(ds, &written) == 0) {
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_WRITTEN,
written);
}
}
if (!dsl_dataset_is_snapshot(ds)) {
/*
* A failed "newfs" (e.g. full) resumable receive leaves
* the stats set on this dataset. Check here for the prop.
*/
get_receive_resume_stats(ds, nv);
/*
* A failed incremental resumable receive leaves the
* stats set on our child named "%recv". Check the child
* for the prop.
*/
/* 6 extra bytes for /%recv */
char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];
dsl_dataset_t *recv_ds;
dsl_dataset_name(ds, recvname);
if (strlcat(recvname, "/", sizeof (recvname)) <
sizeof (recvname) &&
strlcat(recvname, recv_clone_name, sizeof (recvname)) <
sizeof (recvname) &&
dsl_dataset_hold(dp, recvname, FTAG, &recv_ds) == 0) {
get_receive_resume_stats(recv_ds, nv);
dsl_dataset_rele(recv_ds, FTAG);
}
}
}
void
dsl_dataset_fast_stat(dsl_dataset_t *ds, dmu_objset_stats_t *stat)
{
dsl_pool_t *dp __maybe_unused = ds->ds_dir->dd_pool;
ASSERT(dsl_pool_config_held(dp));
stat->dds_creation_txg = dsl_get_creationtxg(ds);
stat->dds_inconsistent = dsl_get_inconsistent(ds);
stat->dds_guid = dsl_get_guid(ds);
stat->dds_redacted = dsl_get_redacted(ds);
stat->dds_origin[0] = '\0';
if (ds->ds_is_snapshot) {
stat->dds_is_snapshot = B_TRUE;
stat->dds_num_clones = dsl_get_numclones(ds);
} else {
stat->dds_is_snapshot = B_FALSE;
stat->dds_num_clones = 0;
if (dsl_dir_is_clone(ds->ds_dir)) {
dsl_dir_get_origin(ds->ds_dir, stat->dds_origin);
}
}
}
uint64_t
dsl_dataset_fsid_guid(dsl_dataset_t *ds)
{
return (ds->ds_fsid_guid);
}
void
dsl_dataset_space(dsl_dataset_t *ds,
uint64_t *refdbytesp, uint64_t *availbytesp,
uint64_t *usedobjsp, uint64_t *availobjsp)
{
*refdbytesp = dsl_dataset_phys(ds)->ds_referenced_bytes;
*availbytesp = dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE);
if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes)
*availbytesp +=
ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes;
if (ds->ds_quota != 0) {
/*
* Adjust available bytes according to refquota
*/
if (*refdbytesp < ds->ds_quota)
*availbytesp = MIN(*availbytesp,
ds->ds_quota - *refdbytesp);
else
*availbytesp = 0;
}
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
*usedobjsp = BP_GET_FILL(&dsl_dataset_phys(ds)->ds_bp);
rrw_exit(&ds->ds_bp_rwlock, FTAG);
*availobjsp = DN_MAX_OBJECT - *usedobjsp;
}
boolean_t
dsl_dataset_modified_since_snap(dsl_dataset_t *ds, dsl_dataset_t *snap)
{
dsl_pool_t *dp __maybe_unused = ds->ds_dir->dd_pool;
uint64_t birth;
ASSERT(dsl_pool_config_held(dp));
if (snap == NULL)
return (B_FALSE);
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
birth = dsl_dataset_get_blkptr(ds)->blk_birth;
rrw_exit(&ds->ds_bp_rwlock, FTAG);
if (birth > dsl_dataset_phys(snap)->ds_creation_txg) {
objset_t *os, *os_snap;
/*
* It may be that only the ZIL differs, because it was
* reset in the head. Don't count that as being
* modified.
*/
if (dmu_objset_from_ds(ds, &os) != 0)
return (B_TRUE);
if (dmu_objset_from_ds(snap, &os_snap) != 0)
return (B_TRUE);
return (bcmp(&os->os_phys->os_meta_dnode,
&os_snap->os_phys->os_meta_dnode,
sizeof (os->os_phys->os_meta_dnode)) != 0);
}
return (B_FALSE);
}
typedef struct dsl_dataset_rename_snapshot_arg {
const char *ddrsa_fsname;
const char *ddrsa_oldsnapname;
const char *ddrsa_newsnapname;
boolean_t ddrsa_recursive;
dmu_tx_t *ddrsa_tx;
} dsl_dataset_rename_snapshot_arg_t;
static int
dsl_dataset_rename_snapshot_check_impl(dsl_pool_t *dp,
dsl_dataset_t *hds, void *arg)
{
(void) dp;
dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
int error;
uint64_t val;
error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val);
if (error != 0) {
/* ignore nonexistent snapshots */
return (error == ENOENT ? 0 : error);
}
/* new name should not exist */
error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_newsnapname, &val);
if (error == 0)
error = SET_ERROR(EEXIST);
else if (error == ENOENT)
error = 0;
/* dataset name + 1 for the "@" + the new snapshot name must fit */
if (dsl_dir_namelen(hds->ds_dir) + 1 +
strlen(ddrsa->ddrsa_newsnapname) >= ZFS_MAX_DATASET_NAME_LEN)
error = SET_ERROR(ENAMETOOLONG);
return (error);
}
static int
dsl_dataset_rename_snapshot_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *hds;
int error;
error = dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds);
if (error != 0)
return (error);
if (ddrsa->ddrsa_recursive) {
error = dmu_objset_find_dp(dp, hds->ds_dir->dd_object,
dsl_dataset_rename_snapshot_check_impl, ddrsa,
DS_FIND_CHILDREN);
} else {
error = dsl_dataset_rename_snapshot_check_impl(dp, hds, ddrsa);
}
dsl_dataset_rele(hds, FTAG);
return (error);
}
static int
dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t *dp,
dsl_dataset_t *hds, void *arg)
{
dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
dsl_dataset_t *ds;
uint64_t val;
dmu_tx_t *tx = ddrsa->ddrsa_tx;
int error;
error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val);
ASSERT(error == 0 || error == ENOENT);
if (error == ENOENT) {
/* ignore nonexistent snapshots */
return (0);
}
VERIFY0(dsl_dataset_hold_obj(dp, val, FTAG, &ds));
/* log before we change the name */
spa_history_log_internal_ds(ds, "rename", tx,
"-> @%s", ddrsa->ddrsa_newsnapname);
VERIFY0(dsl_dataset_snap_remove(hds, ddrsa->ddrsa_oldsnapname, tx,
B_FALSE));
mutex_enter(&ds->ds_lock);
(void) strlcpy(ds->ds_snapname, ddrsa->ddrsa_newsnapname,
sizeof (ds->ds_snapname));
mutex_exit(&ds->ds_lock);
VERIFY0(zap_add(dp->dp_meta_objset,
dsl_dataset_phys(hds)->ds_snapnames_zapobj,
ds->ds_snapname, 8, 1, &ds->ds_object, tx));
zvol_rename_minors(dp->dp_spa, ddrsa->ddrsa_oldsnapname,
ddrsa->ddrsa_newsnapname, B_TRUE);
dsl_dataset_rele(ds, FTAG);
return (0);
}
static void
dsl_dataset_rename_snapshot_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *hds = NULL;
VERIFY0(dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds));
ddrsa->ddrsa_tx = tx;
if (ddrsa->ddrsa_recursive) {
VERIFY0(dmu_objset_find_dp(dp, hds->ds_dir->dd_object,
dsl_dataset_rename_snapshot_sync_impl, ddrsa,
DS_FIND_CHILDREN));
} else {
VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp, hds, ddrsa));
}
dsl_dataset_rele(hds, FTAG);
}
int
dsl_dataset_rename_snapshot(const char *fsname,
const char *oldsnapname, const char *newsnapname, boolean_t recursive)
{
dsl_dataset_rename_snapshot_arg_t ddrsa;
ddrsa.ddrsa_fsname = fsname;
ddrsa.ddrsa_oldsnapname = oldsnapname;
ddrsa.ddrsa_newsnapname = newsnapname;
ddrsa.ddrsa_recursive = recursive;
return (dsl_sync_task(fsname, dsl_dataset_rename_snapshot_check,
dsl_dataset_rename_snapshot_sync, &ddrsa,
1, ZFS_SPACE_CHECK_RESERVED));
}
/*
* If we're doing an ownership handoff, we need to make sure that there is
* only one long hold on the dataset. We're not allowed to change anything here
* so we don't permanently release the long hold or regular hold here. We want
* to do this only when syncing to avoid the dataset unexpectedly going away
* when we release the long hold.
*/
static int
dsl_dataset_handoff_check(dsl_dataset_t *ds, void *owner, dmu_tx_t *tx)
{
boolean_t held = B_FALSE;
if (!dmu_tx_is_syncing(tx))
return (0);
dsl_dir_t *dd = ds->ds_dir;
mutex_enter(&dd->dd_activity_lock);
uint64_t holds = zfs_refcount_count(&ds->ds_longholds) -
(owner != NULL ? 1 : 0);
/*
* The value of dd_activity_waiters can chance as soon as we drop the
* lock, but we're fine with that; new waiters coming in or old
* waiters leaving doesn't cause problems, since we're going to cancel
* waiters later anyway. The goal of this check is to verify that no
* non-waiters have long-holds, and all new long-holds will be
* prevented because we're holding the pool config as writer.
*/
if (holds != dd->dd_activity_waiters)
held = B_TRUE;
mutex_exit(&dd->dd_activity_lock);
if (held)
return (SET_ERROR(EBUSY));
return (0);
}
int
dsl_dataset_rollback_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_rollback_arg_t *ddra = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *ds;
int64_t unused_refres_delta;
int error;
error = dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds);
if (error != 0)
return (error);
/* must not be a snapshot */
if (ds->ds_is_snapshot) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EINVAL));
}
/* must have a most recent snapshot */
if (dsl_dataset_phys(ds)->ds_prev_snap_txg < TXG_INITIAL) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(ESRCH));
}
/*
* No rollback to a snapshot created in the current txg, because
* the rollback may dirty the dataset and create blocks that are
* not reachable from the rootbp while having a birth txg that
* falls into the snapshot's range.
*/
if (dmu_tx_is_syncing(tx) &&
dsl_dataset_phys(ds)->ds_prev_snap_txg >= tx->tx_txg) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EAGAIN));
}
/*
* If the expected target snapshot is specified, then check that
* the latest snapshot is it.
*/
if (ddra->ddra_tosnap != NULL) {
dsl_dataset_t *snapds;
/* Check if the target snapshot exists at all. */
error = dsl_dataset_hold(dp, ddra->ddra_tosnap, FTAG, &snapds);
if (error != 0) {
/*
* ESRCH is used to signal that the target snapshot does
* not exist, while ENOENT is used to report that
* the rolled back dataset does not exist.
* ESRCH is also used to cover other cases where the
* target snapshot is not related to the dataset being
* rolled back such as being in a different pool.
*/
if (error == ENOENT || error == EXDEV)
error = SET_ERROR(ESRCH);
dsl_dataset_rele(ds, FTAG);
return (error);
}
ASSERT(snapds->ds_is_snapshot);
/* Check if the snapshot is the latest snapshot indeed. */
if (snapds != ds->ds_prev) {
/*
* Distinguish between the case where the only problem
* is intervening snapshots (EEXIST) vs the snapshot
* not being a valid target for rollback (ESRCH).
*/
if (snapds->ds_dir == ds->ds_dir ||
(dsl_dir_is_clone(ds->ds_dir) &&
dsl_dir_phys(ds->ds_dir)->dd_origin_obj ==
snapds->ds_object)) {
error = SET_ERROR(EEXIST);
} else {
error = SET_ERROR(ESRCH);
}
dsl_dataset_rele(snapds, FTAG);
dsl_dataset_rele(ds, FTAG);
return (error);
}
dsl_dataset_rele(snapds, FTAG);
}
/* must not have any bookmarks after the most recent snapshot */
if (dsl_bookmark_latest_txg(ds) >
dsl_dataset_phys(ds)->ds_prev_snap_txg) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EEXIST));
}
error = dsl_dataset_handoff_check(ds, ddra->ddra_owner, tx);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
/*
* Check if the snap we are rolling back to uses more than
* the refquota.
*/
if (ds->ds_quota != 0 &&
dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes > ds->ds_quota) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EDQUOT));
}
/*
* When we do the clone swap, we will temporarily use more space
* due to the refreservation (the head will no longer have any
* unique space, so the entire amount of the refreservation will need
* to be free). We will immediately destroy the clone, freeing
* this space, but the freeing happens over many txg's.
*/
unused_refres_delta = (int64_t)MIN(ds->ds_reserved,
dsl_dataset_phys(ds)->ds_unique_bytes);
if (unused_refres_delta > 0 &&
unused_refres_delta >
dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE)) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(ENOSPC));
}
dsl_dataset_rele(ds, FTAG);
return (0);
}
void
dsl_dataset_rollback_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_rollback_arg_t *ddra = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *ds, *clone;
uint64_t cloneobj;
char namebuf[ZFS_MAX_DATASET_NAME_LEN];
VERIFY0(dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds));
dsl_dataset_name(ds->ds_prev, namebuf);
fnvlist_add_string(ddra->ddra_result, "target", namebuf);
cloneobj = dsl_dataset_create_sync(ds->ds_dir, "%rollback",
ds->ds_prev, DS_CREATE_FLAG_NODIRTY, kcred, NULL, tx);
VERIFY0(dsl_dataset_hold_obj(dp, cloneobj, FTAG, &clone));
dsl_dataset_clone_swap_sync_impl(clone, ds, tx);
dsl_dataset_zero_zil(ds, tx);
dsl_destroy_head_sync_impl(clone, tx);
dsl_dataset_rele(clone, FTAG);
dsl_dataset_rele(ds, FTAG);
}
/*
* Rolls back the given filesystem or volume to the most recent snapshot.
* The name of the most recent snapshot will be returned under key "target"
* in the result nvlist.
*
* If owner != NULL:
* - The existing dataset MUST be owned by the specified owner at entry
* - Upon return, dataset will still be held by the same owner, whether we
* succeed or not.
*
* This mode is required any time the existing filesystem is mounted. See
* notes above zfs_suspend_fs() for further details.
*/
int
dsl_dataset_rollback(const char *fsname, const char *tosnap, void *owner,
nvlist_t *result)
{
dsl_dataset_rollback_arg_t ddra;
ddra.ddra_fsname = fsname;
ddra.ddra_tosnap = tosnap;
ddra.ddra_owner = owner;
ddra.ddra_result = result;
return (dsl_sync_task(fsname, dsl_dataset_rollback_check,
dsl_dataset_rollback_sync, &ddra,
1, ZFS_SPACE_CHECK_RESERVED));
}
struct promotenode {
list_node_t link;
dsl_dataset_t *ds;
};
static int snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep);
static int promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp,
void *tag);
static void promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag);
int
dsl_dataset_promote_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_promote_arg_t *ddpa = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *hds;
struct promotenode *snap;
dsl_dataset_t *origin_ds, *origin_head;
int err;
uint64_t unused;
uint64_t ss_mv_cnt;
size_t max_snap_len;
boolean_t conflicting_snaps;
err = promote_hold(ddpa, dp, FTAG);
if (err != 0)
return (err);
hds = ddpa->ddpa_clone;
max_snap_len = MAXNAMELEN - strlen(ddpa->ddpa_clonename) - 1;
if (dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE) {
promote_rele(ddpa, FTAG);
return (SET_ERROR(EXDEV));
}
snap = list_head(&ddpa->shared_snaps);
origin_head = snap->ds;
if (snap == NULL) {
err = SET_ERROR(ENOENT);
goto out;
}
origin_ds = snap->ds;
/*
* Encrypted clones share a DSL Crypto Key with their origin's dsl dir.
* When doing a promote we must make sure the encryption root for
* both the target and the target's origin does not change to avoid
* needing to rewrap encryption keys
*/
err = dsl_dataset_promote_crypt_check(hds->ds_dir, origin_ds->ds_dir);
if (err != 0)
goto out;
/*
* Compute and check the amount of space to transfer. Since this is
* so expensive, don't do the preliminary check.
*/
if (!dmu_tx_is_syncing(tx)) {
promote_rele(ddpa, FTAG);
return (0);
}
/* compute origin's new unique space */
snap = list_tail(&ddpa->clone_snaps);
ASSERT(snap != NULL);
ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==,
origin_ds->ds_object);
dsl_deadlist_space_range(&snap->ds->ds_deadlist,
dsl_dataset_phys(origin_ds)->ds_prev_snap_txg, UINT64_MAX,
&ddpa->unique, &unused, &unused);
/*
* Walk the snapshots that we are moving
*
* Compute space to transfer. Consider the incremental changes
* to used by each snapshot:
* (my used) = (prev's used) + (blocks born) - (blocks killed)
* So each snapshot gave birth to:
* (blocks born) = (my used) - (prev's used) + (blocks killed)
* So a sequence would look like:
* (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
* Which simplifies to:
* uN + kN + kN-1 + ... + k1 + k0
* Note however, if we stop before we reach the ORIGIN we get:
* uN + kN + kN-1 + ... + kM - uM-1
*/
conflicting_snaps = B_FALSE;
ss_mv_cnt = 0;
ddpa->used = dsl_dataset_phys(origin_ds)->ds_referenced_bytes;
ddpa->comp = dsl_dataset_phys(origin_ds)->ds_compressed_bytes;
ddpa->uncomp = dsl_dataset_phys(origin_ds)->ds_uncompressed_bytes;
for (snap = list_head(&ddpa->shared_snaps); snap;
snap = list_next(&ddpa->shared_snaps, snap)) {
uint64_t val, dlused, dlcomp, dluncomp;
dsl_dataset_t *ds = snap->ds;
ss_mv_cnt++;
/*
* If there are long holds, we won't be able to evict
* the objset.
*/
if (dsl_dataset_long_held(ds)) {
err = SET_ERROR(EBUSY);
goto out;
}
/* Check that the snapshot name does not conflict */
VERIFY0(dsl_dataset_get_snapname(ds));
if (strlen(ds->ds_snapname) >= max_snap_len) {
err = SET_ERROR(ENAMETOOLONG);
goto out;
}
err = dsl_dataset_snap_lookup(hds, ds->ds_snapname, &val);
if (err == 0) {
fnvlist_add_boolean(ddpa->err_ds,
snap->ds->ds_snapname);
conflicting_snaps = B_TRUE;
} else if (err != ENOENT) {
goto out;
}
/* The very first snapshot does not have a deadlist */
if (dsl_dataset_phys(ds)->ds_prev_snap_obj == 0)
continue;
dsl_deadlist_space(&ds->ds_deadlist,
&dlused, &dlcomp, &dluncomp);
ddpa->used += dlused;
ddpa->comp += dlcomp;
ddpa->uncomp += dluncomp;
}
/*
* Check that bookmarks that are being transferred don't have
* name conflicts.
*/
for (dsl_bookmark_node_t *dbn = avl_first(&origin_head->ds_bookmarks);
dbn != NULL && dbn->dbn_phys.zbm_creation_txg <=
dsl_dataset_phys(origin_ds)->ds_creation_txg;
dbn = AVL_NEXT(&origin_head->ds_bookmarks, dbn)) {
if (strlen(dbn->dbn_name) >= max_snap_len) {
err = SET_ERROR(ENAMETOOLONG);
goto out;
}
zfs_bookmark_phys_t bm;
err = dsl_bookmark_lookup_impl(ddpa->ddpa_clone,
dbn->dbn_name, &bm);
if (err == 0) {
fnvlist_add_boolean(ddpa->err_ds, dbn->dbn_name);
conflicting_snaps = B_TRUE;
} else if (err == ESRCH) {
err = 0;
} else if (err != 0) {
goto out;
}
}
/*
* In order to return the full list of conflicting snapshots, we check
* whether there was a conflict after traversing all of them.
*/
if (conflicting_snaps) {
err = SET_ERROR(EEXIST);
goto out;
}
/*
* If we are a clone of a clone then we never reached ORIGIN,
* so we need to subtract out the clone origin's used space.
*/
if (ddpa->origin_origin) {
ddpa->used -=
dsl_dataset_phys(ddpa->origin_origin)->ds_referenced_bytes;
ddpa->comp -=
dsl_dataset_phys(ddpa->origin_origin)->ds_compressed_bytes;
ddpa->uncomp -=
dsl_dataset_phys(ddpa->origin_origin)->
ds_uncompressed_bytes;
}
/* Check that there is enough space and limit headroom here */
err = dsl_dir_transfer_possible(origin_ds->ds_dir, hds->ds_dir,
0, ss_mv_cnt, ddpa->used, ddpa->cr, ddpa->proc);
if (err != 0)
goto out;
/*
* Compute the amounts of space that will be used by snapshots
* after the promotion (for both origin and clone). For each,
* it is the amount of space that will be on all of their
* deadlists (that was not born before their new origin).
*/
if (dsl_dir_phys(hds->ds_dir)->dd_flags & DD_FLAG_USED_BREAKDOWN) {
uint64_t space;
/*
* Note, typically this will not be a clone of a clone,
* so dd_origin_txg will be < TXG_INITIAL, so
* these snaplist_space() -> dsl_deadlist_space_range()
* calls will be fast because they do not have to
* iterate over all bps.
*/
snap = list_head(&ddpa->origin_snaps);
if (snap == NULL) {
err = SET_ERROR(ENOENT);
goto out;
}
err = snaplist_space(&ddpa->shared_snaps,
snap->ds->ds_dir->dd_origin_txg, &ddpa->cloneusedsnap);
if (err != 0)
goto out;
err = snaplist_space(&ddpa->clone_snaps,
snap->ds->ds_dir->dd_origin_txg, &space);
if (err != 0)
goto out;
ddpa->cloneusedsnap += space;
}
if (dsl_dir_phys(origin_ds->ds_dir)->dd_flags &
DD_FLAG_USED_BREAKDOWN) {
err = snaplist_space(&ddpa->origin_snaps,
dsl_dataset_phys(origin_ds)->ds_creation_txg,
&ddpa->originusedsnap);
if (err != 0)
goto out;
}
out:
promote_rele(ddpa, FTAG);
return (err);
}
void
dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_promote_arg_t *ddpa = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *hds;
struct promotenode *snap;
dsl_dataset_t *origin_ds;
dsl_dataset_t *origin_head;
dsl_dir_t *dd;
dsl_dir_t *odd = NULL;
uint64_t oldnext_obj;
int64_t delta;
ASSERT(nvlist_empty(ddpa->err_ds));
VERIFY0(promote_hold(ddpa, dp, FTAG));
hds = ddpa->ddpa_clone;
ASSERT0(dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE);
snap = list_head(&ddpa->shared_snaps);
origin_ds = snap->ds;
dd = hds->ds_dir;
snap = list_head(&ddpa->origin_snaps);
origin_head = snap->ds;
/*
* We need to explicitly open odd, since origin_ds's dd will be
* changing.
*/
VERIFY0(dsl_dir_hold_obj(dp, origin_ds->ds_dir->dd_object,
NULL, FTAG, &odd));
dsl_dataset_promote_crypt_sync(hds->ds_dir, odd, tx);
/* change origin's next snap */
dmu_buf_will_dirty(origin_ds->ds_dbuf, tx);
oldnext_obj = dsl_dataset_phys(origin_ds)->ds_next_snap_obj;
snap = list_tail(&ddpa->clone_snaps);
ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==,
origin_ds->ds_object);
dsl_dataset_phys(origin_ds)->ds_next_snap_obj = snap->ds->ds_object;
/* change the origin's next clone */
if (dsl_dataset_phys(origin_ds)->ds_next_clones_obj) {
dsl_dataset_remove_from_next_clones(origin_ds,
snap->ds->ds_object, tx);
VERIFY0(zap_add_int(dp->dp_meta_objset,
dsl_dataset_phys(origin_ds)->ds_next_clones_obj,
oldnext_obj, tx));
}
/* change origin */
dmu_buf_will_dirty(dd->dd_dbuf, tx);
ASSERT3U(dsl_dir_phys(dd)->dd_origin_obj, ==, origin_ds->ds_object);
dsl_dir_phys(dd)->dd_origin_obj = dsl_dir_phys(odd)->dd_origin_obj;
dd->dd_origin_txg = origin_head->ds_dir->dd_origin_txg;
dmu_buf_will_dirty(odd->dd_dbuf, tx);
dsl_dir_phys(odd)->dd_origin_obj = origin_ds->ds_object;
origin_head->ds_dir->dd_origin_txg =
dsl_dataset_phys(origin_ds)->ds_creation_txg;
/* change dd_clone entries */
if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) {
VERIFY0(zap_remove_int(dp->dp_meta_objset,
dsl_dir_phys(odd)->dd_clones, hds->ds_object, tx));
VERIFY0(zap_add_int(dp->dp_meta_objset,
dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones,
hds->ds_object, tx));
VERIFY0(zap_remove_int(dp->dp_meta_objset,
dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones,
origin_head->ds_object, tx));
if (dsl_dir_phys(dd)->dd_clones == 0) {
dsl_dir_phys(dd)->dd_clones =
zap_create(dp->dp_meta_objset, DMU_OT_DSL_CLONES,
DMU_OT_NONE, 0, tx);
}
VERIFY0(zap_add_int(dp->dp_meta_objset,
dsl_dir_phys(dd)->dd_clones, origin_head->ds_object, tx));
}
/*
* Move bookmarks to this dir.
*/
dsl_bookmark_node_t *dbn_next;
for (dsl_bookmark_node_t *dbn = avl_first(&origin_head->ds_bookmarks);
dbn != NULL && dbn->dbn_phys.zbm_creation_txg <=
dsl_dataset_phys(origin_ds)->ds_creation_txg;
dbn = dbn_next) {
dbn_next = AVL_NEXT(&origin_head->ds_bookmarks, dbn);
avl_remove(&origin_head->ds_bookmarks, dbn);
VERIFY0(zap_remove(dp->dp_meta_objset,
origin_head->ds_bookmarks_obj, dbn->dbn_name, tx));
dsl_bookmark_node_add(hds, dbn, tx);
}
dsl_bookmark_next_changed(hds, origin_ds, tx);
/* move snapshots to this dir */
for (snap = list_head(&ddpa->shared_snaps); snap;
snap = list_next(&ddpa->shared_snaps, snap)) {
dsl_dataset_t *ds = snap->ds;
/*
* Property callbacks are registered to a particular
* dsl_dir. Since ours is changing, evict the objset
* so that they will be unregistered from the old dsl_dir.
*/
if (ds->ds_objset) {
dmu_objset_evict(ds->ds_objset);
ds->ds_objset = NULL;
}
/* move snap name entry */
VERIFY0(dsl_dataset_get_snapname(ds));
VERIFY0(dsl_dataset_snap_remove(origin_head,
ds->ds_snapname, tx, B_TRUE));
VERIFY0(zap_add(dp->dp_meta_objset,
dsl_dataset_phys(hds)->ds_snapnames_zapobj, ds->ds_snapname,
8, 1, &ds->ds_object, tx));
dsl_fs_ss_count_adjust(hds->ds_dir, 1,
DD_FIELD_SNAPSHOT_COUNT, tx);
/* change containing dsl_dir */
dmu_buf_will_dirty(ds->ds_dbuf, tx);
ASSERT3U(dsl_dataset_phys(ds)->ds_dir_obj, ==, odd->dd_object);
dsl_dataset_phys(ds)->ds_dir_obj = dd->dd_object;
ASSERT3P(ds->ds_dir, ==, odd);
dsl_dir_rele(ds->ds_dir, ds);
VERIFY0(dsl_dir_hold_obj(dp, dd->dd_object,
NULL, ds, &ds->ds_dir));
/* move any clone references */
if (dsl_dataset_phys(ds)->ds_next_clones_obj &&
spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) {
zap_cursor_t zc;
zap_attribute_t za;
for (zap_cursor_init(&zc, dp->dp_meta_objset,
dsl_dataset_phys(ds)->ds_next_clones_obj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
dsl_dataset_t *cnds;
uint64_t o;
if (za.za_first_integer == oldnext_obj) {
/*
* We've already moved the
* origin's reference.
*/
continue;
}
VERIFY0(dsl_dataset_hold_obj(dp,
za.za_first_integer, FTAG, &cnds));
o = dsl_dir_phys(cnds->ds_dir)->
dd_head_dataset_obj;
VERIFY0(zap_remove_int(dp->dp_meta_objset,
dsl_dir_phys(odd)->dd_clones, o, tx));
VERIFY0(zap_add_int(dp->dp_meta_objset,
dsl_dir_phys(dd)->dd_clones, o, tx));
dsl_dataset_rele(cnds, FTAG);
}
zap_cursor_fini(&zc);
}
ASSERT(!dsl_prop_hascb(ds));
}
/*
* Change space accounting.
* Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
* both be valid, or both be 0 (resulting in delta == 0). This
* is true for each of {clone,origin} independently.
*/
delta = ddpa->cloneusedsnap -
dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_SNAP];
ASSERT3S(delta, >=, 0);
ASSERT3U(ddpa->used, >=, delta);
dsl_dir_diduse_space(dd, DD_USED_SNAP, delta, 0, 0, tx);
dsl_dir_diduse_space(dd, DD_USED_HEAD,
ddpa->used - delta, ddpa->comp, ddpa->uncomp, tx);
delta = ddpa->originusedsnap -
dsl_dir_phys(odd)->dd_used_breakdown[DD_USED_SNAP];
ASSERT3S(delta, <=, 0);
ASSERT3U(ddpa->used, >=, -delta);
dsl_dir_diduse_space(odd, DD_USED_SNAP, delta, 0, 0, tx);
dsl_dir_diduse_space(odd, DD_USED_HEAD,
-ddpa->used - delta, -ddpa->comp, -ddpa->uncomp, tx);
dsl_dataset_phys(origin_ds)->ds_unique_bytes = ddpa->unique;
/*
* Since livelists are specific to a clone's origin txg, they
* are no longer accurate. Destroy the livelist from the clone being
* promoted. If the origin dataset is a clone, destroy its livelist
* as well.
*/
dsl_dir_remove_livelist(dd, tx, B_TRUE);
dsl_dir_remove_livelist(odd, tx, B_TRUE);
/* log history record */
spa_history_log_internal_ds(hds, "promote", tx, " ");
dsl_dir_rele(odd, FTAG);
promote_rele(ddpa, FTAG);
}
/*
* Make a list of dsl_dataset_t's for the snapshots between first_obj
* (exclusive) and last_obj (inclusive). The list will be in reverse
* order (last_obj will be the list_head()). If first_obj == 0, do all
* snapshots back to this dataset's origin.
*/
static int
snaplist_make(dsl_pool_t *dp,
uint64_t first_obj, uint64_t last_obj, list_t *l, void *tag)
{
uint64_t obj = last_obj;
list_create(l, sizeof (struct promotenode),
offsetof(struct promotenode, link));
while (obj != first_obj) {
dsl_dataset_t *ds;
struct promotenode *snap;
int err;
err = dsl_dataset_hold_obj(dp, obj, tag, &ds);
ASSERT(err != ENOENT);
if (err != 0)
return (err);
if (first_obj == 0)
first_obj = dsl_dir_phys(ds->ds_dir)->dd_origin_obj;
snap = kmem_alloc(sizeof (*snap), KM_SLEEP);
snap->ds = ds;
list_insert_tail(l, snap);
obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
}
return (0);
}
static int
snaplist_space(list_t *l, uint64_t mintxg, uint64_t *spacep)
{
struct promotenode *snap;
*spacep = 0;
for (snap = list_head(l); snap; snap = list_next(l, snap)) {
uint64_t used, comp, uncomp;
dsl_deadlist_space_range(&snap->ds->ds_deadlist,
mintxg, UINT64_MAX, &used, &comp, &uncomp);
*spacep += used;
}
return (0);
}
static void
snaplist_destroy(list_t *l, void *tag)
{
struct promotenode *snap;
if (l == NULL || !list_link_active(&l->list_head))
return;
while ((snap = list_tail(l)) != NULL) {
list_remove(l, snap);
dsl_dataset_rele(snap->ds, tag);
kmem_free(snap, sizeof (*snap));
}
list_destroy(l);
}
static int
promote_hold(dsl_dataset_promote_arg_t *ddpa, dsl_pool_t *dp, void *tag)
{
int error;
dsl_dir_t *dd;
struct promotenode *snap;
error = dsl_dataset_hold(dp, ddpa->ddpa_clonename, tag,
&ddpa->ddpa_clone);
if (error != 0)
return (error);
dd = ddpa->ddpa_clone->ds_dir;
if (ddpa->ddpa_clone->ds_is_snapshot ||
!dsl_dir_is_clone(dd)) {
dsl_dataset_rele(ddpa->ddpa_clone, tag);
return (SET_ERROR(EINVAL));
}
error = snaplist_make(dp, 0, dsl_dir_phys(dd)->dd_origin_obj,
&ddpa->shared_snaps, tag);
if (error != 0)
goto out;
error = snaplist_make(dp, 0, ddpa->ddpa_clone->ds_object,
&ddpa->clone_snaps, tag);
if (error != 0)
goto out;
snap = list_head(&ddpa->shared_snaps);
ASSERT3U(snap->ds->ds_object, ==, dsl_dir_phys(dd)->dd_origin_obj);
error = snaplist_make(dp, dsl_dir_phys(dd)->dd_origin_obj,
dsl_dir_phys(snap->ds->ds_dir)->dd_head_dataset_obj,
&ddpa->origin_snaps, tag);
if (error != 0)
goto out;
if (dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj != 0) {
error = dsl_dataset_hold_obj(dp,
dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj,
tag, &ddpa->origin_origin);
if (error != 0)
goto out;
}
out:
if (error != 0)
promote_rele(ddpa, tag);
return (error);
}
static void
promote_rele(dsl_dataset_promote_arg_t *ddpa, void *tag)
{
snaplist_destroy(&ddpa->shared_snaps, tag);
snaplist_destroy(&ddpa->clone_snaps, tag);
snaplist_destroy(&ddpa->origin_snaps, tag);
if (ddpa->origin_origin != NULL)
dsl_dataset_rele(ddpa->origin_origin, tag);
dsl_dataset_rele(ddpa->ddpa_clone, tag);
}
/*
* Promote a clone.
*
* If it fails due to a conflicting snapshot name, "conflsnap" will be filled
* in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
*/
int
dsl_dataset_promote(const char *name, char *conflsnap)
{
dsl_dataset_promote_arg_t ddpa = { 0 };
uint64_t numsnaps;
int error;
nvpair_t *snap_pair;
objset_t *os;
/*
* We will modify space proportional to the number of
* snapshots. Compute numsnaps.
*/
error = dmu_objset_hold(name, FTAG, &os);
if (error != 0)
return (error);
error = zap_count(dmu_objset_pool(os)->dp_meta_objset,
dsl_dataset_phys(dmu_objset_ds(os))->ds_snapnames_zapobj,
&numsnaps);
dmu_objset_rele(os, FTAG);
if (error != 0)
return (error);
ddpa.ddpa_clonename = name;
ddpa.err_ds = fnvlist_alloc();
ddpa.cr = CRED();
ddpa.proc = curproc;
error = dsl_sync_task(name, dsl_dataset_promote_check,
dsl_dataset_promote_sync, &ddpa,
2 + numsnaps, ZFS_SPACE_CHECK_RESERVED);
/*
* Return the first conflicting snapshot found.
*/
snap_pair = nvlist_next_nvpair(ddpa.err_ds, NULL);
if (snap_pair != NULL && conflsnap != NULL)
(void) strlcpy(conflsnap, nvpair_name(snap_pair),
ZFS_MAX_DATASET_NAME_LEN);
fnvlist_free(ddpa.err_ds);
return (error);
}
int
dsl_dataset_clone_swap_check_impl(dsl_dataset_t *clone,
dsl_dataset_t *origin_head, boolean_t force, void *owner, dmu_tx_t *tx)
{
/*
* "slack" factor for received datasets with refquota set on them.
* See the bottom of this function for details on its use.
*/
uint64_t refquota_slack = (uint64_t)DMU_MAX_ACCESS *
spa_asize_inflation;
int64_t unused_refres_delta;
/* they should both be heads */
if (clone->ds_is_snapshot ||
origin_head->ds_is_snapshot)
return (SET_ERROR(EINVAL));
/* if we are not forcing, the branch point should be just before them */
if (!force && clone->ds_prev != origin_head->ds_prev)
return (SET_ERROR(EINVAL));
/* clone should be the clone (unless they are unrelated) */
if (clone->ds_prev != NULL &&
clone->ds_prev != clone->ds_dir->dd_pool->dp_origin_snap &&
origin_head->ds_dir != clone->ds_prev->ds_dir)
return (SET_ERROR(EINVAL));
/* the clone should be a child of the origin */
if (clone->ds_dir->dd_parent != origin_head->ds_dir)
return (SET_ERROR(EINVAL));
/* origin_head shouldn't be modified unless 'force' */
if (!force &&
dsl_dataset_modified_since_snap(origin_head, origin_head->ds_prev))
return (SET_ERROR(ETXTBSY));
/* origin_head should have no long holds (e.g. is not mounted) */
if (dsl_dataset_handoff_check(origin_head, owner, tx))
return (SET_ERROR(EBUSY));
/* check amount of any unconsumed refreservation */
unused_refres_delta =
(int64_t)MIN(origin_head->ds_reserved,
dsl_dataset_phys(origin_head)->ds_unique_bytes) -
(int64_t)MIN(origin_head->ds_reserved,
dsl_dataset_phys(clone)->ds_unique_bytes);
if (unused_refres_delta > 0 &&
unused_refres_delta >
dsl_dir_space_available(origin_head->ds_dir, NULL, 0, TRUE))
return (SET_ERROR(ENOSPC));
/*
* The clone can't be too much over the head's refquota.
*
* To ensure that the entire refquota can be used, we allow one
* transaction to exceed the refquota. Therefore, this check
* needs to also allow for the space referenced to be more than the
* refquota. The maximum amount of space that one transaction can use
* on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
* overage ensures that we are able to receive a filesystem that
* exceeds the refquota on the source system.
*
* So that overage is the refquota_slack we use below.
*/
if (origin_head->ds_quota != 0 &&
dsl_dataset_phys(clone)->ds_referenced_bytes >
origin_head->ds_quota + refquota_slack)
return (SET_ERROR(EDQUOT));
return (0);
}
static void
dsl_dataset_swap_remap_deadlists(dsl_dataset_t *clone,
dsl_dataset_t *origin, dmu_tx_t *tx)
{
uint64_t clone_remap_dl_obj, origin_remap_dl_obj;
dsl_pool_t *dp = dmu_tx_pool(tx);
ASSERT(dsl_pool_sync_context(dp));
clone_remap_dl_obj = dsl_dataset_get_remap_deadlist_object(clone);
origin_remap_dl_obj = dsl_dataset_get_remap_deadlist_object(origin);
if (clone_remap_dl_obj != 0) {
dsl_deadlist_close(&clone->ds_remap_deadlist);
dsl_dataset_unset_remap_deadlist_object(clone, tx);
}
if (origin_remap_dl_obj != 0) {
dsl_deadlist_close(&origin->ds_remap_deadlist);
dsl_dataset_unset_remap_deadlist_object(origin, tx);
}
if (clone_remap_dl_obj != 0) {
dsl_dataset_set_remap_deadlist_object(origin,
clone_remap_dl_obj, tx);
dsl_deadlist_open(&origin->ds_remap_deadlist,
dp->dp_meta_objset, clone_remap_dl_obj);
}
if (origin_remap_dl_obj != 0) {
dsl_dataset_set_remap_deadlist_object(clone,
origin_remap_dl_obj, tx);
dsl_deadlist_open(&clone->ds_remap_deadlist,
dp->dp_meta_objset, origin_remap_dl_obj);
}
}
void
dsl_dataset_clone_swap_sync_impl(dsl_dataset_t *clone,
dsl_dataset_t *origin_head, dmu_tx_t *tx)
{
dsl_pool_t *dp = dmu_tx_pool(tx);
int64_t unused_refres_delta;
ASSERT(clone->ds_reserved == 0);
/*
* NOTE: On DEBUG kernels there could be a race between this and
* the check function if spa_asize_inflation is adjusted...
*/
ASSERT(origin_head->ds_quota == 0 ||
dsl_dataset_phys(clone)->ds_unique_bytes <= origin_head->ds_quota +
DMU_MAX_ACCESS * spa_asize_inflation);
ASSERT3P(clone->ds_prev, ==, origin_head->ds_prev);
dsl_dir_cancel_waiters(origin_head->ds_dir);
/*
* Swap per-dataset feature flags.
*/
for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
if (!(spa_feature_table[f].fi_flags &
ZFEATURE_FLAG_PER_DATASET)) {
ASSERT(!dsl_dataset_feature_is_active(clone, f));
ASSERT(!dsl_dataset_feature_is_active(origin_head, f));
continue;
}
boolean_t clone_inuse = dsl_dataset_feature_is_active(clone, f);
void *clone_feature = clone->ds_feature[f];
boolean_t origin_head_inuse =
dsl_dataset_feature_is_active(origin_head, f);
void *origin_head_feature = origin_head->ds_feature[f];
if (clone_inuse)
dsl_dataset_deactivate_feature_impl(clone, f, tx);
if (origin_head_inuse)
dsl_dataset_deactivate_feature_impl(origin_head, f, tx);
if (clone_inuse) {
dsl_dataset_activate_feature(origin_head->ds_object, f,
clone_feature, tx);
origin_head->ds_feature[f] = clone_feature;
}
if (origin_head_inuse) {
dsl_dataset_activate_feature(clone->ds_object, f,
origin_head_feature, tx);
clone->ds_feature[f] = origin_head_feature;
}
}
dmu_buf_will_dirty(clone->ds_dbuf, tx);
dmu_buf_will_dirty(origin_head->ds_dbuf, tx);
if (clone->ds_objset != NULL) {
dmu_objset_evict(clone->ds_objset);
clone->ds_objset = NULL;
}
if (origin_head->ds_objset != NULL) {
dmu_objset_evict(origin_head->ds_objset);
origin_head->ds_objset = NULL;
}
unused_refres_delta =
(int64_t)MIN(origin_head->ds_reserved,
dsl_dataset_phys(origin_head)->ds_unique_bytes) -
(int64_t)MIN(origin_head->ds_reserved,
dsl_dataset_phys(clone)->ds_unique_bytes);
/*
* Reset origin's unique bytes.
*/
{
dsl_dataset_t *origin = clone->ds_prev;
uint64_t comp, uncomp;
dmu_buf_will_dirty(origin->ds_dbuf, tx);
dsl_deadlist_space_range(&clone->ds_deadlist,
dsl_dataset_phys(origin)->ds_prev_snap_txg, UINT64_MAX,
&dsl_dataset_phys(origin)->ds_unique_bytes, &comp, &uncomp);
}
/* swap blkptrs */
{
rrw_enter(&clone->ds_bp_rwlock, RW_WRITER, FTAG);
rrw_enter(&origin_head->ds_bp_rwlock, RW_WRITER, FTAG);
blkptr_t tmp;
tmp = dsl_dataset_phys(origin_head)->ds_bp;
dsl_dataset_phys(origin_head)->ds_bp =
dsl_dataset_phys(clone)->ds_bp;
dsl_dataset_phys(clone)->ds_bp = tmp;
rrw_exit(&origin_head->ds_bp_rwlock, FTAG);
rrw_exit(&clone->ds_bp_rwlock, FTAG);
}
/* set dd_*_bytes */
{
int64_t dused, dcomp, duncomp;
uint64_t cdl_used, cdl_comp, cdl_uncomp;
uint64_t odl_used, odl_comp, odl_uncomp;
ASSERT3U(dsl_dir_phys(clone->ds_dir)->
dd_used_breakdown[DD_USED_SNAP], ==, 0);
dsl_deadlist_space(&clone->ds_deadlist,
&cdl_used, &cdl_comp, &cdl_uncomp);
dsl_deadlist_space(&origin_head->ds_deadlist,
&odl_used, &odl_comp, &odl_uncomp);
dused = dsl_dataset_phys(clone)->ds_referenced_bytes +
cdl_used -
(dsl_dataset_phys(origin_head)->ds_referenced_bytes +
odl_used);
dcomp = dsl_dataset_phys(clone)->ds_compressed_bytes +
cdl_comp -
(dsl_dataset_phys(origin_head)->ds_compressed_bytes +
odl_comp);
duncomp = dsl_dataset_phys(clone)->ds_uncompressed_bytes +
cdl_uncomp -
(dsl_dataset_phys(origin_head)->ds_uncompressed_bytes +
odl_uncomp);
dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_HEAD,
dused, dcomp, duncomp, tx);
dsl_dir_diduse_space(clone->ds_dir, DD_USED_HEAD,
-dused, -dcomp, -duncomp, tx);
/*
* The difference in the space used by snapshots is the
* difference in snapshot space due to the head's
* deadlist (since that's the only thing that's
* changing that affects the snapused).
*/
dsl_deadlist_space_range(&clone->ds_deadlist,
origin_head->ds_dir->dd_origin_txg, UINT64_MAX,
&cdl_used, &cdl_comp, &cdl_uncomp);
dsl_deadlist_space_range(&origin_head->ds_deadlist,
origin_head->ds_dir->dd_origin_txg, UINT64_MAX,
&odl_used, &odl_comp, &odl_uncomp);
dsl_dir_transfer_space(origin_head->ds_dir, cdl_used - odl_used,
DD_USED_HEAD, DD_USED_SNAP, tx);
}
/* swap ds_*_bytes */
SWITCH64(dsl_dataset_phys(origin_head)->ds_referenced_bytes,
dsl_dataset_phys(clone)->ds_referenced_bytes);
SWITCH64(dsl_dataset_phys(origin_head)->ds_compressed_bytes,
dsl_dataset_phys(clone)->ds_compressed_bytes);
SWITCH64(dsl_dataset_phys(origin_head)->ds_uncompressed_bytes,
dsl_dataset_phys(clone)->ds_uncompressed_bytes);
SWITCH64(dsl_dataset_phys(origin_head)->ds_unique_bytes,
dsl_dataset_phys(clone)->ds_unique_bytes);
/* apply any parent delta for change in unconsumed refreservation */
dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_REFRSRV,
unused_refres_delta, 0, 0, tx);
/*
* Swap deadlists.
*/
dsl_deadlist_close(&clone->ds_deadlist);
dsl_deadlist_close(&origin_head->ds_deadlist);
SWITCH64(dsl_dataset_phys(origin_head)->ds_deadlist_obj,
dsl_dataset_phys(clone)->ds_deadlist_obj);
dsl_deadlist_open(&clone->ds_deadlist, dp->dp_meta_objset,
dsl_dataset_phys(clone)->ds_deadlist_obj);
dsl_deadlist_open(&origin_head->ds_deadlist, dp->dp_meta_objset,
dsl_dataset_phys(origin_head)->ds_deadlist_obj);
dsl_dataset_swap_remap_deadlists(clone, origin_head, tx);
/*
* If there is a bookmark at the origin, its "next dataset" is
* changing, so we need to reset its FBN.
*/
dsl_bookmark_next_changed(origin_head, origin_head->ds_prev, tx);
dsl_scan_ds_clone_swapped(origin_head, clone, tx);
/*
* Destroy any livelists associated with the clone or the origin,
* since after the swap the corresponding livelists are no longer
* valid.
*/
dsl_dir_remove_livelist(clone->ds_dir, tx, B_TRUE);
dsl_dir_remove_livelist(origin_head->ds_dir, tx, B_TRUE);
spa_history_log_internal_ds(clone, "clone swap", tx,
"parent=%s", origin_head->ds_dir->dd_myname);
}
/*
* Given a pool name and a dataset object number in that pool,
* return the name of that dataset.
*/
int
dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
int error;
error = dsl_pool_hold(pname, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold_obj(dp, obj, FTAG, &ds);
if (error == 0) {
dsl_dataset_name(ds, buf);
dsl_dataset_rele(ds, FTAG);
}
dsl_pool_rele(dp, FTAG);
return (error);
}
int
dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota,
uint64_t asize, uint64_t inflight, uint64_t *used, uint64_t *ref_rsrv)
{
int error = 0;
ASSERT3S(asize, >, 0);
/*
* *ref_rsrv is the portion of asize that will come from any
* unconsumed refreservation space.
*/
*ref_rsrv = 0;
mutex_enter(&ds->ds_lock);
/*
* Make a space adjustment for reserved bytes.
*/
if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) {
ASSERT3U(*used, >=,
ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes);
*used -=
(ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes);
*ref_rsrv =
asize - MIN(asize, parent_delta(ds, asize + inflight));
}
if (!check_quota || ds->ds_quota == 0) {
mutex_exit(&ds->ds_lock);
return (0);
}
/*
* If they are requesting more space, and our current estimate
* is over quota, they get to try again unless the actual
* on-disk is over quota and there are no pending changes (which
* may free up space for us).
*/
if (dsl_dataset_phys(ds)->ds_referenced_bytes + inflight >=
ds->ds_quota) {
if (inflight > 0 ||
dsl_dataset_phys(ds)->ds_referenced_bytes < ds->ds_quota)
error = SET_ERROR(ERESTART);
else
error = SET_ERROR(EDQUOT);
}
mutex_exit(&ds->ds_lock);
return (error);
}
typedef struct dsl_dataset_set_qr_arg {
const char *ddsqra_name;
zprop_source_t ddsqra_source;
uint64_t ddsqra_value;
} dsl_dataset_set_qr_arg_t;
static int
dsl_dataset_set_refquota_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_set_qr_arg_t *ddsqra = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *ds;
int error;
uint64_t newval;
if (spa_version(dp->dp_spa) < SPA_VERSION_REFQUOTA)
return (SET_ERROR(ENOTSUP));
error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds);
if (error != 0)
return (error);
if (ds->ds_is_snapshot) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EINVAL));
}
error = dsl_prop_predict(ds->ds_dir,
zfs_prop_to_name(ZFS_PROP_REFQUOTA),
ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
if (newval == 0) {
dsl_dataset_rele(ds, FTAG);
return (0);
}
if (newval < dsl_dataset_phys(ds)->ds_referenced_bytes ||
newval < ds->ds_reserved) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(ENOSPC));
}
dsl_dataset_rele(ds, FTAG);
return (0);
}
static void
dsl_dataset_set_refquota_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_set_qr_arg_t *ddsqra = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *ds = NULL;
uint64_t newval;
VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds));
dsl_prop_set_sync_impl(ds,
zfs_prop_to_name(ZFS_PROP_REFQUOTA),
ddsqra->ddsqra_source, sizeof (ddsqra->ddsqra_value), 1,
&ddsqra->ddsqra_value, tx);
VERIFY0(dsl_prop_get_int_ds(ds,
zfs_prop_to_name(ZFS_PROP_REFQUOTA), &newval));
if (ds->ds_quota != newval) {
dmu_buf_will_dirty(ds->ds_dbuf, tx);
ds->ds_quota = newval;
}
dsl_dataset_rele(ds, FTAG);
}
int
dsl_dataset_set_refquota(const char *dsname, zprop_source_t source,
uint64_t refquota)
{
dsl_dataset_set_qr_arg_t ddsqra;
ddsqra.ddsqra_name = dsname;
ddsqra.ddsqra_source = source;
ddsqra.ddsqra_value = refquota;
return (dsl_sync_task(dsname, dsl_dataset_set_refquota_check,
dsl_dataset_set_refquota_sync, &ddsqra, 0,
ZFS_SPACE_CHECK_EXTRA_RESERVED));
}
static int
dsl_dataset_set_refreservation_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_set_qr_arg_t *ddsqra = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *ds;
int error;
uint64_t newval, unique;
if (spa_version(dp->dp_spa) < SPA_VERSION_REFRESERVATION)
return (SET_ERROR(ENOTSUP));
error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds);
if (error != 0)
return (error);
if (ds->ds_is_snapshot) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EINVAL));
}
error = dsl_prop_predict(ds->ds_dir,
zfs_prop_to_name(ZFS_PROP_REFRESERVATION),
ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
return (error);
}
/*
* If we are doing the preliminary check in open context, the
* space estimates may be inaccurate.
*/
if (!dmu_tx_is_syncing(tx)) {
dsl_dataset_rele(ds, FTAG);
return (0);
}
mutex_enter(&ds->ds_lock);
if (!DS_UNIQUE_IS_ACCURATE(ds))
dsl_dataset_recalc_head_uniq(ds);
unique = dsl_dataset_phys(ds)->ds_unique_bytes;
mutex_exit(&ds->ds_lock);
if (MAX(unique, newval) > MAX(unique, ds->ds_reserved)) {
uint64_t delta = MAX(unique, newval) -
MAX(unique, ds->ds_reserved);
if (delta >
dsl_dir_space_available(ds->ds_dir, NULL, 0, B_TRUE) ||
(ds->ds_quota > 0 && newval > ds->ds_quota)) {
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(ENOSPC));
}
}
dsl_dataset_rele(ds, FTAG);
return (0);
}
void
dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t *ds,
zprop_source_t source, uint64_t value, dmu_tx_t *tx)
{
uint64_t newval;
uint64_t unique;
int64_t delta;
dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_REFRESERVATION),
source, sizeof (value), 1, &value, tx);
VERIFY0(dsl_prop_get_int_ds(ds,
zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &newval));
dmu_buf_will_dirty(ds->ds_dbuf, tx);
mutex_enter(&ds->ds_dir->dd_lock);
mutex_enter(&ds->ds_lock);
ASSERT(DS_UNIQUE_IS_ACCURATE(ds));
unique = dsl_dataset_phys(ds)->ds_unique_bytes;
delta = MAX(0, (int64_t)(newval - unique)) -
MAX(0, (int64_t)(ds->ds_reserved - unique));
ds->ds_reserved = newval;
mutex_exit(&ds->ds_lock);
dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, delta, 0, 0, tx);
mutex_exit(&ds->ds_dir->dd_lock);
}
static void
dsl_dataset_set_refreservation_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_set_qr_arg_t *ddsqra = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *ds = NULL;
VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds));
dsl_dataset_set_refreservation_sync_impl(ds,
ddsqra->ddsqra_source, ddsqra->ddsqra_value, tx);
dsl_dataset_rele(ds, FTAG);
}
int
dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source,
uint64_t refreservation)
{
dsl_dataset_set_qr_arg_t ddsqra;
ddsqra.ddsqra_name = dsname;
ddsqra.ddsqra_source = source;
ddsqra.ddsqra_value = refreservation;
return (dsl_sync_task(dsname, dsl_dataset_set_refreservation_check,
dsl_dataset_set_refreservation_sync, &ddsqra, 0,
ZFS_SPACE_CHECK_EXTRA_RESERVED));
}
typedef struct dsl_dataset_set_compression_arg {
const char *ddsca_name;
zprop_source_t ddsca_source;
uint64_t ddsca_value;
} dsl_dataset_set_compression_arg_t;
static int
dsl_dataset_set_compression_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_set_compression_arg_t *ddsca = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
uint64_t compval = ZIO_COMPRESS_ALGO(ddsca->ddsca_value);
spa_feature_t f = zio_compress_to_feature(compval);
if (f == SPA_FEATURE_NONE)
return (SET_ERROR(EINVAL));
if (!spa_feature_is_enabled(dp->dp_spa, f))
return (SET_ERROR(ENOTSUP));
return (0);
}
static void
dsl_dataset_set_compression_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_set_compression_arg_t *ddsca = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *ds = NULL;
uint64_t compval = ZIO_COMPRESS_ALGO(ddsca->ddsca_value);
spa_feature_t f = zio_compress_to_feature(compval);
ASSERT3S(spa_feature_table[f].fi_type, ==, ZFEATURE_TYPE_BOOLEAN);
VERIFY0(dsl_dataset_hold(dp, ddsca->ddsca_name, FTAG, &ds));
if (zfeature_active(f, ds->ds_feature[f]) != B_TRUE) {
ds->ds_feature_activation[f] = (void *)B_TRUE;
dsl_dataset_activate_feature(ds->ds_object, f,
ds->ds_feature_activation[f], tx);
ds->ds_feature[f] = ds->ds_feature_activation[f];
}
dsl_dataset_rele(ds, FTAG);
}
int
dsl_dataset_set_compression(const char *dsname, zprop_source_t source,
uint64_t compression)
{
dsl_dataset_set_compression_arg_t ddsca;
/*
* The sync task is only required for zstd in order to activate
* the feature flag when the property is first set.
*/
if (ZIO_COMPRESS_ALGO(compression) != ZIO_COMPRESS_ZSTD)
return (0);
ddsca.ddsca_name = dsname;
ddsca.ddsca_source = source;
ddsca.ddsca_value = compression;
return (dsl_sync_task(dsname, dsl_dataset_set_compression_check,
dsl_dataset_set_compression_sync, &ddsca, 0,
ZFS_SPACE_CHECK_EXTRA_RESERVED));
}
/*
* Return (in *usedp) the amount of space referenced by "new" that was not
* referenced at the time the bookmark corresponds to. "New" may be a
* snapshot or a head. The bookmark must be before new, in
* new's filesystem (or its origin) -- caller verifies this.
*
* The written space is calculated by considering two components: First, we
* ignore any freed space, and calculate the written as new's used space
* minus old's used space. Next, we add in the amount of space that was freed
* between the two time points, thus reducing new's used space relative to
* old's. Specifically, this is the space that was born before
* zbm_creation_txg, and freed before new (ie. on new's deadlist or a
* previous deadlist).
*
* space freed [---------------------]
* snapshots ---O-------O--------O-------O------
* bookmark new
*
* Note, the bookmark's zbm_*_bytes_refd must be valid, but if the HAS_FBN
* flag is not set, we will calculate the freed_before_next based on the
* next snapshot's deadlist, rather than using zbm_*_freed_before_next_snap.
*/
static int
dsl_dataset_space_written_impl(zfs_bookmark_phys_t *bmp,
dsl_dataset_t *new, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
{
int err = 0;
dsl_pool_t *dp = new->ds_dir->dd_pool;
ASSERT(dsl_pool_config_held(dp));
if (dsl_dataset_is_snapshot(new)) {
ASSERT3U(bmp->zbm_creation_txg, <,
dsl_dataset_phys(new)->ds_creation_txg);
}
*usedp = 0;
*usedp += dsl_dataset_phys(new)->ds_referenced_bytes;
*usedp -= bmp->zbm_referenced_bytes_refd;
*compp = 0;
*compp += dsl_dataset_phys(new)->ds_compressed_bytes;
*compp -= bmp->zbm_compressed_bytes_refd;
*uncompp = 0;
*uncompp += dsl_dataset_phys(new)->ds_uncompressed_bytes;
*uncompp -= bmp->zbm_uncompressed_bytes_refd;
dsl_dataset_t *snap = new;
while (dsl_dataset_phys(snap)->ds_prev_snap_txg >
bmp->zbm_creation_txg) {
uint64_t used, comp, uncomp;
dsl_deadlist_space_range(&snap->ds_deadlist,
0, bmp->zbm_creation_txg,
&used, &comp, &uncomp);
*usedp += used;
*compp += comp;
*uncompp += uncomp;
uint64_t snapobj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
if (snap != new)
dsl_dataset_rele(snap, FTAG);
err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &snap);
if (err != 0)
break;
}
/*
* We might not have the FBN if we are calculating written from
* a snapshot (because we didn't know the correct "next" snapshot
* until now).
*/
if (bmp->zbm_flags & ZBM_FLAG_HAS_FBN) {
*usedp += bmp->zbm_referenced_freed_before_next_snap;
*compp += bmp->zbm_compressed_freed_before_next_snap;
*uncompp += bmp->zbm_uncompressed_freed_before_next_snap;
} else {
ASSERT3U(dsl_dataset_phys(snap)->ds_prev_snap_txg, ==,
bmp->zbm_creation_txg);
uint64_t used, comp, uncomp;
dsl_deadlist_space(&snap->ds_deadlist, &used, &comp, &uncomp);
*usedp += used;
*compp += comp;
*uncompp += uncomp;
}
if (snap != new)
dsl_dataset_rele(snap, FTAG);
return (err);
}
/*
* Return (in *usedp) the amount of space written in new that was not
* present at the time the bookmark corresponds to. New may be a
* snapshot or the head. Old must be a bookmark before new, in
* new's filesystem (or its origin) -- caller verifies this.
*/
int
dsl_dataset_space_written_bookmark(zfs_bookmark_phys_t *bmp,
dsl_dataset_t *new, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
{
if (!(bmp->zbm_flags & ZBM_FLAG_HAS_FBN))
return (SET_ERROR(ENOTSUP));
return (dsl_dataset_space_written_impl(bmp, new,
usedp, compp, uncompp));
}
/*
* Return (in *usedp) the amount of space written in new that is not
* present in oldsnap. New may be a snapshot or the head. Old must be
* a snapshot before new, in new's filesystem (or its origin). If not then
* fail and return EINVAL.
*/
int
dsl_dataset_space_written(dsl_dataset_t *oldsnap, dsl_dataset_t *new,
uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
{
if (!dsl_dataset_is_before(new, oldsnap, 0))
return (SET_ERROR(EINVAL));
zfs_bookmark_phys_t zbm = { 0 };
dsl_dataset_phys_t *dsp = dsl_dataset_phys(oldsnap);
zbm.zbm_guid = dsp->ds_guid;
zbm.zbm_creation_txg = dsp->ds_creation_txg;
zbm.zbm_creation_time = dsp->ds_creation_time;
zbm.zbm_referenced_bytes_refd = dsp->ds_referenced_bytes;
zbm.zbm_compressed_bytes_refd = dsp->ds_compressed_bytes;
zbm.zbm_uncompressed_bytes_refd = dsp->ds_uncompressed_bytes;
/*
* If oldsnap is the origin (or origin's origin, ...) of new,
* we can't easily calculate the effective FBN. Therefore,
* we do not set ZBM_FLAG_HAS_FBN, so that the _impl will calculate
* it relative to the correct "next": the next snapshot towards "new",
* rather than the next snapshot in oldsnap's dsl_dir.
*/
return (dsl_dataset_space_written_impl(&zbm, new,
usedp, compp, uncompp));
}
/*
* Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
* lastsnap, and all snapshots in between are deleted.
*
* blocks that would be freed [---------------------------]
* snapshots ---O-------O--------O-------O--------O
* firstsnap lastsnap
*
* This is the set of blocks that were born after the snap before firstsnap,
* (birth > firstsnap->prev_snap_txg) and died before the snap after the
* last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
* We calculate this by iterating over the relevant deadlists (from the snap
* after lastsnap, backward to the snap after firstsnap), summing up the
* space on the deadlist that was born after the snap before firstsnap.
*/
int
dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap,
dsl_dataset_t *lastsnap,
uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
{
int err = 0;
uint64_t snapobj;
dsl_pool_t *dp = firstsnap->ds_dir->dd_pool;
ASSERT(firstsnap->ds_is_snapshot);
ASSERT(lastsnap->ds_is_snapshot);
/*
* Check that the snapshots are in the same dsl_dir, and firstsnap
* is before lastsnap.
*/
if (firstsnap->ds_dir != lastsnap->ds_dir ||
dsl_dataset_phys(firstsnap)->ds_creation_txg >
dsl_dataset_phys(lastsnap)->ds_creation_txg)
return (SET_ERROR(EINVAL));
*usedp = *compp = *uncompp = 0;
snapobj = dsl_dataset_phys(lastsnap)->ds_next_snap_obj;
while (snapobj != firstsnap->ds_object) {
dsl_dataset_t *ds;
uint64_t used, comp, uncomp;
err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &ds);
if (err != 0)
break;
dsl_deadlist_space_range(&ds->ds_deadlist,
dsl_dataset_phys(firstsnap)->ds_prev_snap_txg, UINT64_MAX,
&used, &comp, &uncomp);
*usedp += used;
*compp += comp;
*uncompp += uncomp;
snapobj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
ASSERT3U(snapobj, !=, 0);
dsl_dataset_rele(ds, FTAG);
}
return (err);
}
/*
* Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
* For example, they could both be snapshots of the same filesystem, and
* 'earlier' is before 'later'. Or 'earlier' could be the origin of
* 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
* filesystem. Or 'earlier' could be the origin's origin.
*
* If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
*/
boolean_t
dsl_dataset_is_before(dsl_dataset_t *later, dsl_dataset_t *earlier,
uint64_t earlier_txg)
{
dsl_pool_t *dp = later->ds_dir->dd_pool;
int error;
boolean_t ret;
ASSERT(dsl_pool_config_held(dp));
ASSERT(earlier->ds_is_snapshot || earlier_txg != 0);
if (earlier_txg == 0)
earlier_txg = dsl_dataset_phys(earlier)->ds_creation_txg;
if (later->ds_is_snapshot &&
earlier_txg >= dsl_dataset_phys(later)->ds_creation_txg)
return (B_FALSE);
if (later->ds_dir == earlier->ds_dir)
return (B_TRUE);
/*
* We check dd_origin_obj explicitly here rather than using
* dsl_dir_is_clone() so that we will return TRUE if "earlier"
* is $ORIGIN@$ORIGIN. dsl_dataset_space_written() depends on
* this behavior.
*/
if (dsl_dir_phys(later->ds_dir)->dd_origin_obj == 0)
return (B_FALSE);
dsl_dataset_t *origin;
error = dsl_dataset_hold_obj(dp,
dsl_dir_phys(later->ds_dir)->dd_origin_obj, FTAG, &origin);
if (error != 0)
return (B_FALSE);
if (dsl_dataset_phys(origin)->ds_creation_txg == earlier_txg &&
origin->ds_dir == earlier->ds_dir) {
dsl_dataset_rele(origin, FTAG);
return (B_TRUE);
}
ret = dsl_dataset_is_before(origin, earlier, earlier_txg);
dsl_dataset_rele(origin, FTAG);
return (ret);
}
void
dsl_dataset_zapify(dsl_dataset_t *ds, dmu_tx_t *tx)
{
objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
dmu_object_zapify(mos, ds->ds_object, DMU_OT_DSL_DATASET, tx);
}
boolean_t
dsl_dataset_is_zapified(dsl_dataset_t *ds)
{
dmu_object_info_t doi;
dmu_object_info_from_db(ds->ds_dbuf, &doi);
return (doi.doi_type == DMU_OTN_ZAP_METADATA);
}
boolean_t
dsl_dataset_has_resume_receive_state(dsl_dataset_t *ds)
{
return (dsl_dataset_is_zapified(ds) &&
zap_contains(ds->ds_dir->dd_pool->dp_meta_objset,
ds->ds_object, DS_FIELD_RESUME_TOGUID) == 0);
}
uint64_t
dsl_dataset_get_remap_deadlist_object(dsl_dataset_t *ds)
{
uint64_t remap_deadlist_obj;
int err;
if (!dsl_dataset_is_zapified(ds))
return (0);
err = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset, ds->ds_object,
DS_FIELD_REMAP_DEADLIST, sizeof (remap_deadlist_obj), 1,
&remap_deadlist_obj);
if (err != 0) {
VERIFY3S(err, ==, ENOENT);
return (0);
}
ASSERT(remap_deadlist_obj != 0);
return (remap_deadlist_obj);
}
boolean_t
dsl_dataset_remap_deadlist_exists(dsl_dataset_t *ds)
{
EQUIV(dsl_deadlist_is_open(&ds->ds_remap_deadlist),
dsl_dataset_get_remap_deadlist_object(ds) != 0);
return (dsl_deadlist_is_open(&ds->ds_remap_deadlist));
}
static void
dsl_dataset_set_remap_deadlist_object(dsl_dataset_t *ds, uint64_t obj,
dmu_tx_t *tx)
{
ASSERT(obj != 0);
dsl_dataset_zapify(ds, tx);
VERIFY0(zap_add(ds->ds_dir->dd_pool->dp_meta_objset, ds->ds_object,
DS_FIELD_REMAP_DEADLIST, sizeof (obj), 1, &obj, tx));
}
static void
dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t *ds, dmu_tx_t *tx)
{
VERIFY0(zap_remove(ds->ds_dir->dd_pool->dp_meta_objset,
ds->ds_object, DS_FIELD_REMAP_DEADLIST, tx));
}
void
dsl_dataset_destroy_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx)
{
uint64_t remap_deadlist_object;
spa_t *spa = ds->ds_dir->dd_pool->dp_spa;
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(dsl_dataset_remap_deadlist_exists(ds));
remap_deadlist_object = ds->ds_remap_deadlist.dl_object;
dsl_deadlist_close(&ds->ds_remap_deadlist);
dsl_deadlist_free(spa_meta_objset(spa), remap_deadlist_object, tx);
dsl_dataset_unset_remap_deadlist_object(ds, tx);
spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
}
void
dsl_dataset_create_remap_deadlist(dsl_dataset_t *ds, dmu_tx_t *tx)
{
uint64_t remap_deadlist_obj;
spa_t *spa = ds->ds_dir->dd_pool->dp_spa;
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(MUTEX_HELD(&ds->ds_remap_deadlist_lock));
/*
* Currently we only create remap deadlists when there are indirect
* vdevs with referenced mappings.
*/
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL));
remap_deadlist_obj = dsl_deadlist_clone(
&ds->ds_deadlist, UINT64_MAX,
dsl_dataset_phys(ds)->ds_prev_snap_obj, tx);
dsl_dataset_set_remap_deadlist_object(ds,
remap_deadlist_obj, tx);
dsl_deadlist_open(&ds->ds_remap_deadlist, spa_meta_objset(spa),
remap_deadlist_obj);
spa_feature_incr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
}
void
dsl_dataset_activate_redaction(dsl_dataset_t *ds, uint64_t *redact_snaps,
uint64_t num_redact_snaps, dmu_tx_t *tx)
{
uint64_t dsobj = ds->ds_object;
struct feature_type_uint64_array_arg *ftuaa =
kmem_zalloc(sizeof (*ftuaa), KM_SLEEP);
ftuaa->length = (int64_t)num_redact_snaps;
if (num_redact_snaps > 0) {
ftuaa->array = kmem_alloc(num_redact_snaps * sizeof (uint64_t),
KM_SLEEP);
bcopy(redact_snaps, ftuaa->array, num_redact_snaps *
sizeof (uint64_t));
}
dsl_dataset_activate_feature(dsobj, SPA_FEATURE_REDACTED_DATASETS,
ftuaa, tx);
ds->ds_feature[SPA_FEATURE_REDACTED_DATASETS] = ftuaa;
}
/* BEGIN CSTYLED */
#if defined(_LP64)
#define RECORDSIZE_PERM ZMOD_RW
#else
/* Limited to 1M on 32-bit platforms due to lack of virtual address space */
#define RECORDSIZE_PERM ZMOD_RD
#endif
ZFS_MODULE_PARAM(zfs, zfs_, max_recordsize, INT, RECORDSIZE_PERM,
"Max allowed record size");
ZFS_MODULE_PARAM(zfs, zfs_, allow_redacted_dataset_mount, INT, ZMOD_RW,
"Allow mounting of redacted datasets");
/* END CSTYLED */
EXPORT_SYMBOL(dsl_dataset_hold);
EXPORT_SYMBOL(dsl_dataset_hold_flags);
EXPORT_SYMBOL(dsl_dataset_hold_obj);
EXPORT_SYMBOL(dsl_dataset_hold_obj_flags);
EXPORT_SYMBOL(dsl_dataset_own);
EXPORT_SYMBOL(dsl_dataset_own_obj);
EXPORT_SYMBOL(dsl_dataset_name);
EXPORT_SYMBOL(dsl_dataset_rele);
EXPORT_SYMBOL(dsl_dataset_rele_flags);
EXPORT_SYMBOL(dsl_dataset_disown);
EXPORT_SYMBOL(dsl_dataset_tryown);
EXPORT_SYMBOL(dsl_dataset_create_sync);
EXPORT_SYMBOL(dsl_dataset_create_sync_dd);
EXPORT_SYMBOL(dsl_dataset_snapshot_check);
EXPORT_SYMBOL(dsl_dataset_snapshot_sync);
EXPORT_SYMBOL(dsl_dataset_promote);
EXPORT_SYMBOL(dsl_dataset_user_hold);
EXPORT_SYMBOL(dsl_dataset_user_release);
EXPORT_SYMBOL(dsl_dataset_get_holds);
EXPORT_SYMBOL(dsl_dataset_get_blkptr);
EXPORT_SYMBOL(dsl_dataset_get_spa);
EXPORT_SYMBOL(dsl_dataset_modified_since_snap);
EXPORT_SYMBOL(dsl_dataset_space_written);
EXPORT_SYMBOL(dsl_dataset_space_wouldfree);
EXPORT_SYMBOL(dsl_dataset_sync);
EXPORT_SYMBOL(dsl_dataset_block_born);
EXPORT_SYMBOL(dsl_dataset_block_kill);
EXPORT_SYMBOL(dsl_dataset_dirty);
EXPORT_SYMBOL(dsl_dataset_stats);
EXPORT_SYMBOL(dsl_dataset_fast_stat);
EXPORT_SYMBOL(dsl_dataset_space);
EXPORT_SYMBOL(dsl_dataset_fsid_guid);
EXPORT_SYMBOL(dsl_dsobj_to_dsname);
EXPORT_SYMBOL(dsl_dataset_check_quota);
EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl);
EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl);
diff --git a/module/zfs/dsl_pool.c b/module/zfs/dsl_pool.c
index 5d1522a7bfd0..ef7e9d2d248a 100644
--- a/module/zfs/dsl_pool.c
+++ b/module/zfs/dsl_pool.c
@@ -1,1489 +1,1489 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright (c) 2013 Steven Hartland. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
*/
#include <sys/dsl_pool.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_synctask.h>
#include <sys/dsl_scan.h>
#include <sys/dnode.h>
#include <sys/dmu_tx.h>
#include <sys/dmu_objset.h>
#include <sys/arc.h>
#include <sys/zap.h>
#include <sys/zio.h>
#include <sys/zfs_context.h>
#include <sys/fs/zfs.h>
#include <sys/zfs_znode.h>
#include <sys/spa_impl.h>
#include <sys/vdev_impl.h>
#include <sys/metaslab_impl.h>
#include <sys/bptree.h>
#include <sys/zfeature.h>
#include <sys/zil_impl.h>
#include <sys/dsl_userhold.h>
#include <sys/trace_zfs.h>
#include <sys/mmp.h>
/*
* ZFS Write Throttle
* ------------------
*
* ZFS must limit the rate of incoming writes to the rate at which it is able
* to sync data modifications to the backend storage. Throttling by too much
* creates an artificial limit; throttling by too little can only be sustained
* for short periods and would lead to highly lumpy performance. On a per-pool
* basis, ZFS tracks the amount of modified (dirty) data. As operations change
* data, the amount of dirty data increases; as ZFS syncs out data, the amount
* of dirty data decreases. When the amount of dirty data exceeds a
* predetermined threshold further modifications are blocked until the amount
* of dirty data decreases (as data is synced out).
*
* The limit on dirty data is tunable, and should be adjusted according to
* both the IO capacity and available memory of the system. The larger the
* window, the more ZFS is able to aggregate and amortize metadata (and data)
* changes. However, memory is a limited resource, and allowing for more dirty
* data comes at the cost of keeping other useful data in memory (for example
* ZFS data cached by the ARC).
*
* Implementation
*
* As buffers are modified dsl_pool_willuse_space() increments both the per-
* txg (dp_dirty_pertxg[]) and poolwide (dp_dirty_total) accounting of
* dirty space used; dsl_pool_dirty_space() decrements those values as data
* is synced out from dsl_pool_sync(). While only the poolwide value is
* relevant, the per-txg value is useful for debugging. The tunable
* zfs_dirty_data_max determines the dirty space limit. Once that value is
* exceeded, new writes are halted until space frees up.
*
* The zfs_dirty_data_sync_percent tunable dictates the threshold at which we
* ensure that there is a txg syncing (see the comment in txg.c for a full
* description of transaction group stages).
*
* The IO scheduler uses both the dirty space limit and current amount of
* dirty data as inputs. Those values affect the number of concurrent IOs ZFS
* issues. See the comment in vdev_queue.c for details of the IO scheduler.
*
* The delay is also calculated based on the amount of dirty data. See the
* comment above dmu_tx_delay() for details.
*/
/*
* zfs_dirty_data_max will be set to zfs_dirty_data_max_percent% of all memory,
* capped at zfs_dirty_data_max_max. It can also be overridden with a module
* parameter.
*/
unsigned long zfs_dirty_data_max = 0;
unsigned long zfs_dirty_data_max_max = 0;
int zfs_dirty_data_max_percent = 10;
int zfs_dirty_data_max_max_percent = 25;
/*
* zfs_wrlog_data_max, the upper limit of TX_WRITE log data.
* Once it is reached, write operation is blocked,
* until log data is cleared out after txg sync.
* It only counts TX_WRITE log with WR_COPIED or WR_NEED_COPY.
*/
unsigned long zfs_wrlog_data_max = 0;
/*
* If there's at least this much dirty data (as a percentage of
* zfs_dirty_data_max), push out a txg. This should be less than
* zfs_vdev_async_write_active_min_dirty_percent.
*/
-int zfs_dirty_data_sync_percent = 20;
+static int zfs_dirty_data_sync_percent = 20;
/*
* Once there is this amount of dirty data, the dmu_tx_delay() will kick in
* and delay each transaction.
* This value should be >= zfs_vdev_async_write_active_max_dirty_percent.
*/
int zfs_delay_min_dirty_percent = 60;
/*
* This controls how quickly the delay approaches infinity.
* Larger values cause it to delay more for a given amount of dirty data.
* Therefore larger values will cause there to be less dirty data for a
* given throughput.
*
* For the smoothest delay, this value should be about 1 billion divided
* by the maximum number of operations per second. This will smoothly
* handle between 10x and 1/10th this number.
*
* Note: zfs_delay_scale * zfs_dirty_data_max must be < 2^64, due to the
* multiply in dmu_tx_delay().
*/
unsigned long zfs_delay_scale = 1000 * 1000 * 1000 / 2000;
/*
* This determines the number of threads used by the dp_sync_taskq.
*/
-int zfs_sync_taskq_batch_pct = 75;
+static int zfs_sync_taskq_batch_pct = 75;
/*
* These tunables determine the behavior of how zil_itxg_clean() is
* called via zil_clean() in the context of spa_sync(). When an itxg
* list needs to be cleaned, TQ_NOSLEEP will be used when dispatching.
* If the dispatch fails, the call to zil_itxg_clean() will occur
* synchronously in the context of spa_sync(), which can negatively
* impact the performance of spa_sync() (e.g. in the case of the itxg
* list having a large number of itxs that needs to be cleaned).
*
* Thus, these tunables can be used to manipulate the behavior of the
* taskq used by zil_clean(); they determine the number of taskq entries
* that are pre-populated when the taskq is first created (via the
* "zfs_zil_clean_taskq_minalloc" tunable) and the maximum number of
* taskq entries that are cached after an on-demand allocation (via the
* "zfs_zil_clean_taskq_maxalloc").
*
* The idea being, we want to try reasonably hard to ensure there will
* already be a taskq entry pre-allocated by the time that it is needed
* by zil_clean(). This way, we can avoid the possibility of an
* on-demand allocation of a new taskq entry from failing, which would
* result in zil_itxg_clean() being called synchronously from zil_clean()
* (which can adversely affect performance of spa_sync()).
*
* Additionally, the number of threads used by the taskq can be
* configured via the "zfs_zil_clean_taskq_nthr_pct" tunable.
*/
-int zfs_zil_clean_taskq_nthr_pct = 100;
-int zfs_zil_clean_taskq_minalloc = 1024;
-int zfs_zil_clean_taskq_maxalloc = 1024 * 1024;
+static int zfs_zil_clean_taskq_nthr_pct = 100;
+static int zfs_zil_clean_taskq_minalloc = 1024;
+static int zfs_zil_clean_taskq_maxalloc = 1024 * 1024;
int
dsl_pool_open_special_dir(dsl_pool_t *dp, const char *name, dsl_dir_t **ddp)
{
uint64_t obj;
int err;
err = zap_lookup(dp->dp_meta_objset,
dsl_dir_phys(dp->dp_root_dir)->dd_child_dir_zapobj,
name, sizeof (obj), 1, &obj);
if (err)
return (err);
return (dsl_dir_hold_obj(dp, obj, name, dp, ddp));
}
static dsl_pool_t *
dsl_pool_open_impl(spa_t *spa, uint64_t txg)
{
dsl_pool_t *dp;
blkptr_t *bp = spa_get_rootblkptr(spa);
dp = kmem_zalloc(sizeof (dsl_pool_t), KM_SLEEP);
dp->dp_spa = spa;
dp->dp_meta_rootbp = *bp;
rrw_init(&dp->dp_config_rwlock, B_TRUE);
txg_init(dp, txg);
mmp_init(spa);
txg_list_create(&dp->dp_dirty_datasets, spa,
offsetof(dsl_dataset_t, ds_dirty_link));
txg_list_create(&dp->dp_dirty_zilogs, spa,
offsetof(zilog_t, zl_dirty_link));
txg_list_create(&dp->dp_dirty_dirs, spa,
offsetof(dsl_dir_t, dd_dirty_link));
txg_list_create(&dp->dp_sync_tasks, spa,
offsetof(dsl_sync_task_t, dst_node));
txg_list_create(&dp->dp_early_sync_tasks, spa,
offsetof(dsl_sync_task_t, dst_node));
dp->dp_sync_taskq = taskq_create("dp_sync_taskq",
zfs_sync_taskq_batch_pct, minclsyspri, 1, INT_MAX,
TASKQ_THREADS_CPU_PCT);
dp->dp_zil_clean_taskq = taskq_create("dp_zil_clean_taskq",
zfs_zil_clean_taskq_nthr_pct, minclsyspri,
zfs_zil_clean_taskq_minalloc,
zfs_zil_clean_taskq_maxalloc,
TASKQ_PREPOPULATE | TASKQ_THREADS_CPU_PCT);
mutex_init(&dp->dp_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&dp->dp_spaceavail_cv, NULL, CV_DEFAULT, NULL);
aggsum_init(&dp->dp_wrlog_total, 0);
for (int i = 0; i < TXG_SIZE; i++) {
aggsum_init(&dp->dp_wrlog_pertxg[i], 0);
}
dp->dp_zrele_taskq = taskq_create("z_zrele", 100, defclsyspri,
boot_ncpus * 8, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC |
TASKQ_THREADS_CPU_PCT);
dp->dp_unlinked_drain_taskq = taskq_create("z_unlinked_drain",
100, defclsyspri, boot_ncpus, INT_MAX,
TASKQ_PREPOPULATE | TASKQ_DYNAMIC | TASKQ_THREADS_CPU_PCT);
return (dp);
}
int
dsl_pool_init(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
{
int err;
dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
/*
* Initialize the caller's dsl_pool_t structure before we actually open
* the meta objset. This is done because a self-healing write zio may
* be issued as part of dmu_objset_open_impl() and the spa needs its
* dsl_pool_t initialized in order to handle the write.
*/
*dpp = dp;
err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp,
&dp->dp_meta_objset);
if (err != 0) {
dsl_pool_close(dp);
*dpp = NULL;
}
return (err);
}
int
dsl_pool_open(dsl_pool_t *dp)
{
int err;
dsl_dir_t *dd;
dsl_dataset_t *ds;
uint64_t obj;
rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1,
&dp->dp_root_dir_obj);
if (err)
goto out;
err = dsl_dir_hold_obj(dp, dp->dp_root_dir_obj,
NULL, dp, &dp->dp_root_dir);
if (err)
goto out;
err = dsl_pool_open_special_dir(dp, MOS_DIR_NAME, &dp->dp_mos_dir);
if (err)
goto out;
if (spa_version(dp->dp_spa) >= SPA_VERSION_ORIGIN) {
err = dsl_pool_open_special_dir(dp, ORIGIN_DIR_NAME, &dd);
if (err)
goto out;
err = dsl_dataset_hold_obj(dp,
dsl_dir_phys(dd)->dd_head_dataset_obj, FTAG, &ds);
if (err == 0) {
err = dsl_dataset_hold_obj(dp,
dsl_dataset_phys(ds)->ds_prev_snap_obj, dp,
&dp->dp_origin_snap);
dsl_dataset_rele(ds, FTAG);
}
dsl_dir_rele(dd, dp);
if (err)
goto out;
}
if (spa_version(dp->dp_spa) >= SPA_VERSION_DEADLISTS) {
err = dsl_pool_open_special_dir(dp, FREE_DIR_NAME,
&dp->dp_free_dir);
if (err)
goto out;
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj);
if (err)
goto out;
VERIFY0(bpobj_open(&dp->dp_free_bpobj,
dp->dp_meta_objset, obj));
}
if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_OBSOLETE_COUNTS)) {
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_OBSOLETE_BPOBJ, sizeof (uint64_t), 1, &obj);
if (err == 0) {
VERIFY0(bpobj_open(&dp->dp_obsolete_bpobj,
dp->dp_meta_objset, obj));
} else if (err == ENOENT) {
/*
* We might not have created the remap bpobj yet.
*/
err = 0;
} else {
goto out;
}
}
/*
* Note: errors ignored, because the these special dirs, used for
* space accounting, are only created on demand.
*/
(void) dsl_pool_open_special_dir(dp, LEAK_DIR_NAME,
&dp->dp_leak_dir);
if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_ASYNC_DESTROY)) {
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_BPTREE_OBJ, sizeof (uint64_t), 1,
&dp->dp_bptree_obj);
if (err != 0)
goto out;
}
if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMPTY_BPOBJ)) {
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
&dp->dp_empty_bpobj);
if (err != 0)
goto out;
}
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_TMP_USERREFS, sizeof (uint64_t), 1,
&dp->dp_tmp_userrefs_obj);
if (err == ENOENT)
err = 0;
if (err)
goto out;
err = dsl_scan_init(dp, dp->dp_tx.tx_open_txg);
out:
rrw_exit(&dp->dp_config_rwlock, FTAG);
return (err);
}
void
dsl_pool_close(dsl_pool_t *dp)
{
/*
* Drop our references from dsl_pool_open().
*
* Since we held the origin_snap from "syncing" context (which
* includes pool-opening context), it actually only got a "ref"
* and not a hold, so just drop that here.
*/
if (dp->dp_origin_snap != NULL)
dsl_dataset_rele(dp->dp_origin_snap, dp);
if (dp->dp_mos_dir != NULL)
dsl_dir_rele(dp->dp_mos_dir, dp);
if (dp->dp_free_dir != NULL)
dsl_dir_rele(dp->dp_free_dir, dp);
if (dp->dp_leak_dir != NULL)
dsl_dir_rele(dp->dp_leak_dir, dp);
if (dp->dp_root_dir != NULL)
dsl_dir_rele(dp->dp_root_dir, dp);
bpobj_close(&dp->dp_free_bpobj);
bpobj_close(&dp->dp_obsolete_bpobj);
/* undo the dmu_objset_open_impl(mos) from dsl_pool_open() */
if (dp->dp_meta_objset != NULL)
dmu_objset_evict(dp->dp_meta_objset);
txg_list_destroy(&dp->dp_dirty_datasets);
txg_list_destroy(&dp->dp_dirty_zilogs);
txg_list_destroy(&dp->dp_sync_tasks);
txg_list_destroy(&dp->dp_early_sync_tasks);
txg_list_destroy(&dp->dp_dirty_dirs);
taskq_destroy(dp->dp_zil_clean_taskq);
taskq_destroy(dp->dp_sync_taskq);
/*
* We can't set retry to TRUE since we're explicitly specifying
* a spa to flush. This is good enough; any missed buffers for
* this spa won't cause trouble, and they'll eventually fall
* out of the ARC just like any other unused buffer.
*/
arc_flush(dp->dp_spa, FALSE);
mmp_fini(dp->dp_spa);
txg_fini(dp);
dsl_scan_fini(dp);
dmu_buf_user_evict_wait();
rrw_destroy(&dp->dp_config_rwlock);
mutex_destroy(&dp->dp_lock);
cv_destroy(&dp->dp_spaceavail_cv);
ASSERT0(aggsum_value(&dp->dp_wrlog_total));
aggsum_fini(&dp->dp_wrlog_total);
for (int i = 0; i < TXG_SIZE; i++) {
ASSERT0(aggsum_value(&dp->dp_wrlog_pertxg[i]));
aggsum_fini(&dp->dp_wrlog_pertxg[i]);
}
taskq_destroy(dp->dp_unlinked_drain_taskq);
taskq_destroy(dp->dp_zrele_taskq);
if (dp->dp_blkstats != NULL) {
mutex_destroy(&dp->dp_blkstats->zab_lock);
vmem_free(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
}
kmem_free(dp, sizeof (dsl_pool_t));
}
void
dsl_pool_create_obsolete_bpobj(dsl_pool_t *dp, dmu_tx_t *tx)
{
uint64_t obj;
/*
* Currently, we only create the obsolete_bpobj where there are
* indirect vdevs with referenced mappings.
*/
ASSERT(spa_feature_is_active(dp->dp_spa, SPA_FEATURE_DEVICE_REMOVAL));
/* create and open the obsolete_bpobj */
obj = bpobj_alloc(dp->dp_meta_objset, SPA_OLD_MAXBLOCKSIZE, tx);
VERIFY0(bpobj_open(&dp->dp_obsolete_bpobj, dp->dp_meta_objset, obj));
VERIFY0(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_OBSOLETE_BPOBJ, sizeof (uint64_t), 1, &obj, tx));
spa_feature_incr(dp->dp_spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
}
void
dsl_pool_destroy_obsolete_bpobj(dsl_pool_t *dp, dmu_tx_t *tx)
{
spa_feature_decr(dp->dp_spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
VERIFY0(zap_remove(dp->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_OBSOLETE_BPOBJ, tx));
bpobj_free(dp->dp_meta_objset,
dp->dp_obsolete_bpobj.bpo_object, tx);
bpobj_close(&dp->dp_obsolete_bpobj);
}
dsl_pool_t *
dsl_pool_create(spa_t *spa, nvlist_t *zplprops __attribute__((unused)),
dsl_crypto_params_t *dcp, uint64_t txg)
{
int err;
dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg);
#ifdef _KERNEL
objset_t *os;
#else
objset_t *os __attribute__((unused));
#endif
dsl_dataset_t *ds;
uint64_t obj;
rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
/* create and open the MOS (meta-objset) */
dp->dp_meta_objset = dmu_objset_create_impl(spa,
NULL, &dp->dp_meta_rootbp, DMU_OST_META, tx);
spa->spa_meta_objset = dp->dp_meta_objset;
/* create the pool directory */
err = zap_create_claim(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_OT_OBJECT_DIRECTORY, DMU_OT_NONE, 0, tx);
ASSERT0(err);
/* Initialize scan structures */
VERIFY0(dsl_scan_init(dp, txg));
/* create and open the root dir */
dp->dp_root_dir_obj = dsl_dir_create_sync(dp, NULL, NULL, tx);
VERIFY0(dsl_dir_hold_obj(dp, dp->dp_root_dir_obj,
NULL, dp, &dp->dp_root_dir));
/* create and open the meta-objset dir */
(void) dsl_dir_create_sync(dp, dp->dp_root_dir, MOS_DIR_NAME, tx);
VERIFY0(dsl_pool_open_special_dir(dp,
MOS_DIR_NAME, &dp->dp_mos_dir));
if (spa_version(spa) >= SPA_VERSION_DEADLISTS) {
/* create and open the free dir */
(void) dsl_dir_create_sync(dp, dp->dp_root_dir,
FREE_DIR_NAME, tx);
VERIFY0(dsl_pool_open_special_dir(dp,
FREE_DIR_NAME, &dp->dp_free_dir));
/* create and open the free_bplist */
obj = bpobj_alloc(dp->dp_meta_objset, SPA_OLD_MAXBLOCKSIZE, tx);
VERIFY(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx) == 0);
VERIFY0(bpobj_open(&dp->dp_free_bpobj,
dp->dp_meta_objset, obj));
}
if (spa_version(spa) >= SPA_VERSION_DSL_SCRUB)
dsl_pool_create_origin(dp, tx);
/*
* Some features may be needed when creating the root dataset, so we
* create the feature objects here.
*/
if (spa_version(spa) >= SPA_VERSION_FEATURES)
spa_feature_create_zap_objects(spa, tx);
if (dcp != NULL && dcp->cp_crypt != ZIO_CRYPT_OFF &&
dcp->cp_crypt != ZIO_CRYPT_INHERIT)
spa_feature_enable(spa, SPA_FEATURE_ENCRYPTION, tx);
/* create the root dataset */
obj = dsl_dataset_create_sync_dd(dp->dp_root_dir, NULL, dcp, 0, tx);
/* create the root objset */
VERIFY0(dsl_dataset_hold_obj_flags(dp, obj,
DS_HOLD_FLAG_DECRYPT, FTAG, &ds));
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
os = dmu_objset_create_impl(dp->dp_spa, ds,
dsl_dataset_get_blkptr(ds), DMU_OST_ZFS, tx);
rrw_exit(&ds->ds_bp_rwlock, FTAG);
#ifdef _KERNEL
zfs_create_fs(os, kcred, zplprops, tx);
#endif
dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG);
dmu_tx_commit(tx);
rrw_exit(&dp->dp_config_rwlock, FTAG);
return (dp);
}
/*
* Account for the meta-objset space in its placeholder dsl_dir.
*/
void
dsl_pool_mos_diduse_space(dsl_pool_t *dp,
int64_t used, int64_t comp, int64_t uncomp)
{
ASSERT3U(comp, ==, uncomp); /* it's all metadata */
mutex_enter(&dp->dp_lock);
dp->dp_mos_used_delta += used;
dp->dp_mos_compressed_delta += comp;
dp->dp_mos_uncompressed_delta += uncomp;
mutex_exit(&dp->dp_lock);
}
static void
dsl_pool_sync_mos(dsl_pool_t *dp, dmu_tx_t *tx)
{
zio_t *zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
dmu_objset_sync(dp->dp_meta_objset, zio, tx);
VERIFY0(zio_wait(zio));
dmu_objset_sync_done(dp->dp_meta_objset, tx);
taskq_wait(dp->dp_sync_taskq);
multilist_destroy(&dp->dp_meta_objset->os_synced_dnodes);
dprintf_bp(&dp->dp_meta_rootbp, "meta objset rootbp is %s", "");
spa_set_rootblkptr(dp->dp_spa, &dp->dp_meta_rootbp);
}
static void
dsl_pool_dirty_delta(dsl_pool_t *dp, int64_t delta)
{
ASSERT(MUTEX_HELD(&dp->dp_lock));
if (delta < 0)
ASSERT3U(-delta, <=, dp->dp_dirty_total);
dp->dp_dirty_total += delta;
/*
* Note: we signal even when increasing dp_dirty_total.
* This ensures forward progress -- each thread wakes the next waiter.
*/
if (dp->dp_dirty_total < zfs_dirty_data_max)
cv_signal(&dp->dp_spaceavail_cv);
}
void
dsl_pool_wrlog_count(dsl_pool_t *dp, int64_t size, uint64_t txg)
{
ASSERT3S(size, >=, 0);
aggsum_add(&dp->dp_wrlog_pertxg[txg & TXG_MASK], size);
aggsum_add(&dp->dp_wrlog_total, size);
/* Choose a value slightly bigger than min dirty sync bytes */
uint64_t sync_min =
zfs_dirty_data_max * (zfs_dirty_data_sync_percent + 10) / 100;
if (aggsum_compare(&dp->dp_wrlog_pertxg[txg & TXG_MASK], sync_min) > 0)
txg_kick(dp, txg);
}
boolean_t
dsl_pool_wrlog_over_max(dsl_pool_t *dp)
{
return (aggsum_compare(&dp->dp_wrlog_total, zfs_wrlog_data_max) > 0);
}
static void
dsl_pool_wrlog_clear(dsl_pool_t *dp, uint64_t txg)
{
int64_t delta;
delta = -(int64_t)aggsum_value(&dp->dp_wrlog_pertxg[txg & TXG_MASK]);
aggsum_add(&dp->dp_wrlog_pertxg[txg & TXG_MASK], delta);
aggsum_add(&dp->dp_wrlog_total, delta);
}
#ifdef ZFS_DEBUG
static boolean_t
dsl_early_sync_task_verify(dsl_pool_t *dp, uint64_t txg)
{
spa_t *spa = dp->dp_spa;
vdev_t *rvd = spa->spa_root_vdev;
for (uint64_t c = 0; c < rvd->vdev_children; c++) {
vdev_t *vd = rvd->vdev_child[c];
txg_list_t *tl = &vd->vdev_ms_list;
metaslab_t *ms;
for (ms = txg_list_head(tl, TXG_CLEAN(txg)); ms;
ms = txg_list_next(tl, ms, TXG_CLEAN(txg))) {
VERIFY(range_tree_is_empty(ms->ms_freeing));
VERIFY(range_tree_is_empty(ms->ms_checkpointing));
}
}
return (B_TRUE);
}
#else
#define dsl_early_sync_task_verify(dp, txg) \
((void) sizeof (dp), (void) sizeof (txg), B_TRUE)
#endif
void
dsl_pool_sync(dsl_pool_t *dp, uint64_t txg)
{
zio_t *zio;
dmu_tx_t *tx;
dsl_dir_t *dd;
dsl_dataset_t *ds;
objset_t *mos = dp->dp_meta_objset;
list_t synced_datasets;
list_create(&synced_datasets, sizeof (dsl_dataset_t),
offsetof(dsl_dataset_t, ds_synced_link));
tx = dmu_tx_create_assigned(dp, txg);
/*
* Run all early sync tasks before writing out any dirty blocks.
* For more info on early sync tasks see block comment in
* dsl_early_sync_task().
*/
if (!txg_list_empty(&dp->dp_early_sync_tasks, txg)) {
dsl_sync_task_t *dst;
ASSERT3U(spa_sync_pass(dp->dp_spa), ==, 1);
while ((dst =
txg_list_remove(&dp->dp_early_sync_tasks, txg)) != NULL) {
ASSERT(dsl_early_sync_task_verify(dp, txg));
dsl_sync_task_sync(dst, tx);
}
ASSERT(dsl_early_sync_task_verify(dp, txg));
}
/*
* Write out all dirty blocks of dirty datasets.
*/
zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
while ((ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) != NULL) {
/*
* We must not sync any non-MOS datasets twice, because
* we may have taken a snapshot of them. However, we
* may sync newly-created datasets on pass 2.
*/
ASSERT(!list_link_active(&ds->ds_synced_link));
list_insert_tail(&synced_datasets, ds);
dsl_dataset_sync(ds, zio, tx);
}
VERIFY0(zio_wait(zio));
/*
* Update the long range free counter after
* we're done syncing user data
*/
mutex_enter(&dp->dp_lock);
ASSERT(spa_sync_pass(dp->dp_spa) == 1 ||
dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] == 0);
dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] = 0;
mutex_exit(&dp->dp_lock);
/*
* After the data blocks have been written (ensured by the zio_wait()
* above), update the user/group/project space accounting. This happens
* in tasks dispatched to dp_sync_taskq, so wait for them before
* continuing.
*/
for (ds = list_head(&synced_datasets); ds != NULL;
ds = list_next(&synced_datasets, ds)) {
dmu_objset_sync_done(ds->ds_objset, tx);
}
taskq_wait(dp->dp_sync_taskq);
/*
* Sync the datasets again to push out the changes due to
* userspace updates. This must be done before we process the
* sync tasks, so that any snapshots will have the correct
* user accounting information (and we won't get confused
* about which blocks are part of the snapshot).
*/
zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
while ((ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) != NULL) {
objset_t *os = ds->ds_objset;
ASSERT(list_link_active(&ds->ds_synced_link));
dmu_buf_rele(ds->ds_dbuf, ds);
dsl_dataset_sync(ds, zio, tx);
/*
* Release any key mappings created by calls to
* dsl_dataset_dirty() from the userquota accounting
* code paths.
*/
if (os->os_encrypted && !os->os_raw_receive &&
!os->os_next_write_raw[txg & TXG_MASK]) {
ASSERT3P(ds->ds_key_mapping, !=, NULL);
key_mapping_rele(dp->dp_spa, ds->ds_key_mapping, ds);
}
}
VERIFY0(zio_wait(zio));
/*
* Now that the datasets have been completely synced, we can
* clean up our in-memory structures accumulated while syncing:
*
* - move dead blocks from the pending deadlist and livelists
* to the on-disk versions
* - release hold from dsl_dataset_dirty()
* - release key mapping hold from dsl_dataset_dirty()
*/
while ((ds = list_remove_head(&synced_datasets)) != NULL) {
objset_t *os = ds->ds_objset;
if (os->os_encrypted && !os->os_raw_receive &&
!os->os_next_write_raw[txg & TXG_MASK]) {
ASSERT3P(ds->ds_key_mapping, !=, NULL);
key_mapping_rele(dp->dp_spa, ds->ds_key_mapping, ds);
}
dsl_dataset_sync_done(ds, tx);
}
while ((dd = txg_list_remove(&dp->dp_dirty_dirs, txg)) != NULL) {
dsl_dir_sync(dd, tx);
}
/*
* The MOS's space is accounted for in the pool/$MOS
* (dp_mos_dir). We can't modify the mos while we're syncing
* it, so we remember the deltas and apply them here.
*/
if (dp->dp_mos_used_delta != 0 || dp->dp_mos_compressed_delta != 0 ||
dp->dp_mos_uncompressed_delta != 0) {
dsl_dir_diduse_space(dp->dp_mos_dir, DD_USED_HEAD,
dp->dp_mos_used_delta,
dp->dp_mos_compressed_delta,
dp->dp_mos_uncompressed_delta, tx);
dp->dp_mos_used_delta = 0;
dp->dp_mos_compressed_delta = 0;
dp->dp_mos_uncompressed_delta = 0;
}
if (dmu_objset_is_dirty(mos, txg)) {
dsl_pool_sync_mos(dp, tx);
}
/*
* We have written all of the accounted dirty data, so our
* dp_space_towrite should now be zero. However, some seldom-used
* code paths do not adhere to this (e.g. dbuf_undirty()). Shore up
* the accounting of any dirtied space now.
*
* Note that, besides any dirty data from datasets, the amount of
* dirty data in the MOS is also accounted by the pool. Therefore,
* we want to do this cleanup after dsl_pool_sync_mos() so we don't
* attempt to update the accounting for the same dirty data twice.
* (i.e. at this point we only update the accounting for the space
* that we know that we "leaked").
*/
dsl_pool_undirty_space(dp, dp->dp_dirty_pertxg[txg & TXG_MASK], txg);
/*
* If we modify a dataset in the same txg that we want to destroy it,
* its dsl_dir's dd_dbuf will be dirty, and thus have a hold on it.
* dsl_dir_destroy_check() will fail if there are unexpected holds.
* Therefore, we want to sync the MOS (thus syncing the dd_dbuf
* and clearing the hold on it) before we process the sync_tasks.
* The MOS data dirtied by the sync_tasks will be synced on the next
* pass.
*/
if (!txg_list_empty(&dp->dp_sync_tasks, txg)) {
dsl_sync_task_t *dst;
/*
* No more sync tasks should have been added while we
* were syncing.
*/
ASSERT3U(spa_sync_pass(dp->dp_spa), ==, 1);
while ((dst = txg_list_remove(&dp->dp_sync_tasks, txg)) != NULL)
dsl_sync_task_sync(dst, tx);
}
dmu_tx_commit(tx);
DTRACE_PROBE2(dsl_pool_sync__done, dsl_pool_t *dp, dp, uint64_t, txg);
}
void
dsl_pool_sync_done(dsl_pool_t *dp, uint64_t txg)
{
zilog_t *zilog;
while ((zilog = txg_list_head(&dp->dp_dirty_zilogs, txg))) {
dsl_dataset_t *ds = dmu_objset_ds(zilog->zl_os);
/*
* We don't remove the zilog from the dp_dirty_zilogs
* list until after we've cleaned it. This ensures that
* callers of zilog_is_dirty() receive an accurate
* answer when they are racing with the spa sync thread.
*/
zil_clean(zilog, txg);
(void) txg_list_remove_this(&dp->dp_dirty_zilogs, zilog, txg);
ASSERT(!dmu_objset_is_dirty(zilog->zl_os, txg));
dmu_buf_rele(ds->ds_dbuf, zilog);
}
dsl_pool_wrlog_clear(dp, txg);
ASSERT(!dmu_objset_is_dirty(dp->dp_meta_objset, txg));
}
/*
* TRUE if the current thread is the tx_sync_thread or if we
* are being called from SPA context during pool initialization.
*/
int
dsl_pool_sync_context(dsl_pool_t *dp)
{
return (curthread == dp->dp_tx.tx_sync_thread ||
spa_is_initializing(dp->dp_spa) ||
taskq_member(dp->dp_sync_taskq, curthread));
}
/*
* This function returns the amount of allocatable space in the pool
* minus whatever space is currently reserved by ZFS for specific
* purposes. Specifically:
*
* 1] Any reserved SLOP space
* 2] Any space used by the checkpoint
* 3] Any space used for deferred frees
*
* The latter 2 are especially important because they are needed to
* rectify the SPA's and DMU's different understanding of how much space
* is used. Now the DMU is aware of that extra space tracked by the SPA
* without having to maintain a separate special dir (e.g similar to
* $MOS, $FREEING, and $LEAKED).
*
* Note: By deferred frees here, we mean the frees that were deferred
* in spa_sync() after sync pass 1 (spa_deferred_bpobj), and not the
* segments placed in ms_defer trees during metaslab_sync_done().
*/
uint64_t
dsl_pool_adjustedsize(dsl_pool_t *dp, zfs_space_check_t slop_policy)
{
spa_t *spa = dp->dp_spa;
uint64_t space, resv, adjustedsize;
uint64_t spa_deferred_frees =
spa->spa_deferred_bpobj.bpo_phys->bpo_bytes;
space = spa_get_dspace(spa)
- spa_get_checkpoint_space(spa) - spa_deferred_frees;
resv = spa_get_slop_space(spa);
switch (slop_policy) {
case ZFS_SPACE_CHECK_NORMAL:
break;
case ZFS_SPACE_CHECK_RESERVED:
resv >>= 1;
break;
case ZFS_SPACE_CHECK_EXTRA_RESERVED:
resv >>= 2;
break;
case ZFS_SPACE_CHECK_NONE:
resv = 0;
break;
default:
panic("invalid slop policy value: %d", slop_policy);
break;
}
adjustedsize = (space >= resv) ? (space - resv) : 0;
return (adjustedsize);
}
uint64_t
dsl_pool_unreserved_space(dsl_pool_t *dp, zfs_space_check_t slop_policy)
{
uint64_t poolsize = dsl_pool_adjustedsize(dp, slop_policy);
uint64_t deferred =
metaslab_class_get_deferred(spa_normal_class(dp->dp_spa));
uint64_t quota = (poolsize >= deferred) ? (poolsize - deferred) : 0;
return (quota);
}
boolean_t
dsl_pool_need_dirty_delay(dsl_pool_t *dp)
{
uint64_t delay_min_bytes =
zfs_dirty_data_max * zfs_delay_min_dirty_percent / 100;
mutex_enter(&dp->dp_lock);
uint64_t dirty = dp->dp_dirty_total;
mutex_exit(&dp->dp_lock);
return (dirty > delay_min_bytes);
}
static boolean_t
dsl_pool_need_dirty_sync(dsl_pool_t *dp, uint64_t txg)
{
ASSERT(MUTEX_HELD(&dp->dp_lock));
uint64_t dirty_min_bytes =
zfs_dirty_data_max * zfs_dirty_data_sync_percent / 100;
uint64_t dirty = dp->dp_dirty_pertxg[txg & TXG_MASK];
return (dirty > dirty_min_bytes);
}
void
dsl_pool_dirty_space(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx)
{
if (space > 0) {
mutex_enter(&dp->dp_lock);
dp->dp_dirty_pertxg[tx->tx_txg & TXG_MASK] += space;
dsl_pool_dirty_delta(dp, space);
boolean_t needsync = !dmu_tx_is_syncing(tx) &&
dsl_pool_need_dirty_sync(dp, tx->tx_txg);
mutex_exit(&dp->dp_lock);
if (needsync)
txg_kick(dp, tx->tx_txg);
}
}
void
dsl_pool_undirty_space(dsl_pool_t *dp, int64_t space, uint64_t txg)
{
ASSERT3S(space, >=, 0);
if (space == 0)
return;
mutex_enter(&dp->dp_lock);
if (dp->dp_dirty_pertxg[txg & TXG_MASK] < space) {
/* XXX writing something we didn't dirty? */
space = dp->dp_dirty_pertxg[txg & TXG_MASK];
}
ASSERT3U(dp->dp_dirty_pertxg[txg & TXG_MASK], >=, space);
dp->dp_dirty_pertxg[txg & TXG_MASK] -= space;
ASSERT3U(dp->dp_dirty_total, >=, space);
dsl_pool_dirty_delta(dp, -space);
mutex_exit(&dp->dp_lock);
}
/* ARGSUSED */
static int
upgrade_clones_cb(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg)
{
dmu_tx_t *tx = arg;
dsl_dataset_t *ds, *prev = NULL;
int err;
err = dsl_dataset_hold_obj(dp, hds->ds_object, FTAG, &ds);
if (err)
return (err);
while (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
err = dsl_dataset_hold_obj(dp,
dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);
if (err) {
dsl_dataset_rele(ds, FTAG);
return (err);
}
if (dsl_dataset_phys(prev)->ds_next_snap_obj != ds->ds_object)
break;
dsl_dataset_rele(ds, FTAG);
ds = prev;
prev = NULL;
}
if (prev == NULL) {
prev = dp->dp_origin_snap;
/*
* The $ORIGIN can't have any data, or the accounting
* will be wrong.
*/
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
ASSERT0(dsl_dataset_phys(prev)->ds_bp.blk_birth);
rrw_exit(&ds->ds_bp_rwlock, FTAG);
/* The origin doesn't get attached to itself */
if (ds->ds_object == prev->ds_object) {
dsl_dataset_rele(ds, FTAG);
return (0);
}
dmu_buf_will_dirty(ds->ds_dbuf, tx);
dsl_dataset_phys(ds)->ds_prev_snap_obj = prev->ds_object;
dsl_dataset_phys(ds)->ds_prev_snap_txg =
dsl_dataset_phys(prev)->ds_creation_txg;
dmu_buf_will_dirty(ds->ds_dir->dd_dbuf, tx);
dsl_dir_phys(ds->ds_dir)->dd_origin_obj = prev->ds_object;
dmu_buf_will_dirty(prev->ds_dbuf, tx);
dsl_dataset_phys(prev)->ds_num_children++;
if (dsl_dataset_phys(ds)->ds_next_snap_obj == 0) {
ASSERT(ds->ds_prev == NULL);
VERIFY0(dsl_dataset_hold_obj(dp,
dsl_dataset_phys(ds)->ds_prev_snap_obj,
ds, &ds->ds_prev));
}
}
ASSERT3U(dsl_dir_phys(ds->ds_dir)->dd_origin_obj, ==, prev->ds_object);
ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_obj, ==, prev->ds_object);
if (dsl_dataset_phys(prev)->ds_next_clones_obj == 0) {
dmu_buf_will_dirty(prev->ds_dbuf, tx);
dsl_dataset_phys(prev)->ds_next_clones_obj =
zap_create(dp->dp_meta_objset,
DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx);
}
VERIFY0(zap_add_int(dp->dp_meta_objset,
dsl_dataset_phys(prev)->ds_next_clones_obj, ds->ds_object, tx));
dsl_dataset_rele(ds, FTAG);
if (prev != dp->dp_origin_snap)
dsl_dataset_rele(prev, FTAG);
return (0);
}
void
dsl_pool_upgrade_clones(dsl_pool_t *dp, dmu_tx_t *tx)
{
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(dp->dp_origin_snap != NULL);
VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj, upgrade_clones_cb,
tx, DS_FIND_CHILDREN | DS_FIND_SERIALIZE));
}
/* ARGSUSED */
static int
upgrade_dir_clones_cb(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg)
{
dmu_tx_t *tx = arg;
objset_t *mos = dp->dp_meta_objset;
if (dsl_dir_phys(ds->ds_dir)->dd_origin_obj != 0) {
dsl_dataset_t *origin;
VERIFY0(dsl_dataset_hold_obj(dp,
dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &origin));
if (dsl_dir_phys(origin->ds_dir)->dd_clones == 0) {
dmu_buf_will_dirty(origin->ds_dir->dd_dbuf, tx);
dsl_dir_phys(origin->ds_dir)->dd_clones =
zap_create(mos, DMU_OT_DSL_CLONES, DMU_OT_NONE,
0, tx);
}
VERIFY0(zap_add_int(dp->dp_meta_objset,
dsl_dir_phys(origin->ds_dir)->dd_clones,
ds->ds_object, tx));
dsl_dataset_rele(origin, FTAG);
}
return (0);
}
void
dsl_pool_upgrade_dir_clones(dsl_pool_t *dp, dmu_tx_t *tx)
{
uint64_t obj;
ASSERT(dmu_tx_is_syncing(tx));
(void) dsl_dir_create_sync(dp, dp->dp_root_dir, FREE_DIR_NAME, tx);
VERIFY0(dsl_pool_open_special_dir(dp,
FREE_DIR_NAME, &dp->dp_free_dir));
/*
* We can't use bpobj_alloc(), because spa_version() still
* returns the old version, and we need a new-version bpobj with
* subobj support. So call dmu_object_alloc() directly.
*/
obj = dmu_object_alloc(dp->dp_meta_objset, DMU_OT_BPOBJ,
SPA_OLD_MAXBLOCKSIZE, DMU_OT_BPOBJ_HDR, sizeof (bpobj_phys_t), tx);
VERIFY0(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx));
VERIFY0(bpobj_open(&dp->dp_free_bpobj, dp->dp_meta_objset, obj));
VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
upgrade_dir_clones_cb, tx, DS_FIND_CHILDREN | DS_FIND_SERIALIZE));
}
void
dsl_pool_create_origin(dsl_pool_t *dp, dmu_tx_t *tx)
{
uint64_t dsobj;
dsl_dataset_t *ds;
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(dp->dp_origin_snap == NULL);
ASSERT(rrw_held(&dp->dp_config_rwlock, RW_WRITER));
/* create the origin dir, ds, & snap-ds */
dsobj = dsl_dataset_create_sync(dp->dp_root_dir, ORIGIN_DIR_NAME,
NULL, 0, kcred, NULL, tx);
VERIFY0(dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
dsl_dataset_snapshot_sync_impl(ds, ORIGIN_DIR_NAME, tx);
VERIFY0(dsl_dataset_hold_obj(dp, dsl_dataset_phys(ds)->ds_prev_snap_obj,
dp, &dp->dp_origin_snap));
dsl_dataset_rele(ds, FTAG);
}
taskq_t *
dsl_pool_zrele_taskq(dsl_pool_t *dp)
{
return (dp->dp_zrele_taskq);
}
taskq_t *
dsl_pool_unlinked_drain_taskq(dsl_pool_t *dp)
{
return (dp->dp_unlinked_drain_taskq);
}
/*
* Walk through the pool-wide zap object of temporary snapshot user holds
* and release them.
*/
void
dsl_pool_clean_tmp_userrefs(dsl_pool_t *dp)
{
zap_attribute_t za;
zap_cursor_t zc;
objset_t *mos = dp->dp_meta_objset;
uint64_t zapobj = dp->dp_tmp_userrefs_obj;
nvlist_t *holds;
if (zapobj == 0)
return;
ASSERT(spa_version(dp->dp_spa) >= SPA_VERSION_USERREFS);
holds = fnvlist_alloc();
for (zap_cursor_init(&zc, mos, zapobj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
char *htag;
nvlist_t *tags;
htag = strchr(za.za_name, '-');
*htag = '\0';
++htag;
if (nvlist_lookup_nvlist(holds, za.za_name, &tags) != 0) {
tags = fnvlist_alloc();
fnvlist_add_boolean(tags, htag);
fnvlist_add_nvlist(holds, za.za_name, tags);
fnvlist_free(tags);
} else {
fnvlist_add_boolean(tags, htag);
}
}
dsl_dataset_user_release_tmp(dp, holds);
fnvlist_free(holds);
zap_cursor_fini(&zc);
}
/*
* Create the pool-wide zap object for storing temporary snapshot holds.
*/
static void
dsl_pool_user_hold_create_obj(dsl_pool_t *dp, dmu_tx_t *tx)
{
objset_t *mos = dp->dp_meta_objset;
ASSERT(dp->dp_tmp_userrefs_obj == 0);
ASSERT(dmu_tx_is_syncing(tx));
dp->dp_tmp_userrefs_obj = zap_create_link(mos, DMU_OT_USERREFS,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_TMP_USERREFS, tx);
}
static int
dsl_pool_user_hold_rele_impl(dsl_pool_t *dp, uint64_t dsobj,
const char *tag, uint64_t now, dmu_tx_t *tx, boolean_t holding)
{
objset_t *mos = dp->dp_meta_objset;
uint64_t zapobj = dp->dp_tmp_userrefs_obj;
char *name;
int error;
ASSERT(spa_version(dp->dp_spa) >= SPA_VERSION_USERREFS);
ASSERT(dmu_tx_is_syncing(tx));
/*
* If the pool was created prior to SPA_VERSION_USERREFS, the
* zap object for temporary holds might not exist yet.
*/
if (zapobj == 0) {
if (holding) {
dsl_pool_user_hold_create_obj(dp, tx);
zapobj = dp->dp_tmp_userrefs_obj;
} else {
return (SET_ERROR(ENOENT));
}
}
name = kmem_asprintf("%llx-%s", (u_longlong_t)dsobj, tag);
if (holding)
error = zap_add(mos, zapobj, name, 8, 1, &now, tx);
else
error = zap_remove(mos, zapobj, name, tx);
kmem_strfree(name);
return (error);
}
/*
* Add a temporary hold for the given dataset object and tag.
*/
int
dsl_pool_user_hold(dsl_pool_t *dp, uint64_t dsobj, const char *tag,
uint64_t now, dmu_tx_t *tx)
{
return (dsl_pool_user_hold_rele_impl(dp, dsobj, tag, now, tx, B_TRUE));
}
/*
* Release a temporary hold for the given dataset object and tag.
*/
int
dsl_pool_user_release(dsl_pool_t *dp, uint64_t dsobj, const char *tag,
dmu_tx_t *tx)
{
return (dsl_pool_user_hold_rele_impl(dp, dsobj, tag, 0,
tx, B_FALSE));
}
/*
* DSL Pool Configuration Lock
*
* The dp_config_rwlock protects against changes to DSL state (e.g. dataset
* creation / destruction / rename / property setting). It must be held for
* read to hold a dataset or dsl_dir. I.e. you must call
* dsl_pool_config_enter() or dsl_pool_hold() before calling
* dsl_{dataset,dir}_hold{_obj}. In most circumstances, the dp_config_rwlock
* must be held continuously until all datasets and dsl_dirs are released.
*
* The only exception to this rule is that if a "long hold" is placed on
* a dataset, then the dp_config_rwlock may be dropped while the dataset
* is still held. The long hold will prevent the dataset from being
* destroyed -- the destroy will fail with EBUSY. A long hold can be
* obtained by calling dsl_dataset_long_hold(), or by "owning" a dataset
* (by calling dsl_{dataset,objset}_{try}own{_obj}).
*
* Legitimate long-holders (including owners) should be long-running, cancelable
* tasks that should cause "zfs destroy" to fail. This includes DMU
* consumers (i.e. a ZPL filesystem being mounted or ZVOL being open),
* "zfs send", and "zfs diff". There are several other long-holders whose
* uses are suboptimal (e.g. "zfs promote", and zil_suspend()).
*
* The usual formula for long-holding would be:
* dsl_pool_hold()
* dsl_dataset_hold()
* ... perform checks ...
* dsl_dataset_long_hold()
* dsl_pool_rele()
* ... perform long-running task ...
* dsl_dataset_long_rele()
* dsl_dataset_rele()
*
* Note that when the long hold is released, the dataset is still held but
* the pool is not held. The dataset may change arbitrarily during this time
* (e.g. it could be destroyed). Therefore you shouldn't do anything to the
* dataset except release it.
*
* Operations generally fall somewhere into the following taxonomy:
*
* Read-Only Modifying
*
* Dataset Layer / MOS zfs get zfs destroy
*
* Individual Dataset read() write()
*
*
* Dataset Layer Operations
*
* Modifying operations should generally use dsl_sync_task(). The synctask
* infrastructure enforces proper locking strategy with respect to the
* dp_config_rwlock. See the comment above dsl_sync_task() for details.
*
* Read-only operations will manually hold the pool, then the dataset, obtain
* information from the dataset, then release the pool and dataset.
* dmu_objset_{hold,rele}() are convenience routines that also do the pool
* hold/rele.
*
*
* Operations On Individual Datasets
*
* Objects _within_ an objset should only be modified by the current 'owner'
* of the objset to prevent incorrect concurrent modification. Thus, use
* {dmu_objset,dsl_dataset}_own to mark some entity as the current owner,
* and fail with EBUSY if there is already an owner. The owner can then
* implement its own locking strategy, independent of the dataset layer's
* locking infrastructure.
* (E.g., the ZPL has its own set of locks to control concurrency. A regular
* vnop will not reach into the dataset layer).
*
* Ideally, objects would also only be read by the objset’s owner, so that we
* don’t observe state mid-modification.
* (E.g. the ZPL is creating a new object and linking it into a directory; if
* you don’t coordinate with the ZPL to hold ZPL-level locks, you could see an
* intermediate state. The ioctl level violates this but in pretty benign
* ways, e.g. reading the zpl props object.)
*/
int
dsl_pool_hold(const char *name, void *tag, dsl_pool_t **dp)
{
spa_t *spa;
int error;
error = spa_open(name, &spa, tag);
if (error == 0) {
*dp = spa_get_dsl(spa);
dsl_pool_config_enter(*dp, tag);
}
return (error);
}
void
dsl_pool_rele(dsl_pool_t *dp, void *tag)
{
dsl_pool_config_exit(dp, tag);
spa_close(dp->dp_spa, tag);
}
void
dsl_pool_config_enter(dsl_pool_t *dp, void *tag)
{
/*
* We use a "reentrant" reader-writer lock, but not reentrantly.
*
* The rrwlock can (with the track_all flag) track all reading threads,
* which is very useful for debugging which code path failed to release
* the lock, and for verifying that the *current* thread does hold
* the lock.
*
* (Unlike a rwlock, which knows that N threads hold it for
* read, but not *which* threads, so rw_held(RW_READER) returns TRUE
* if any thread holds it for read, even if this thread doesn't).
*/
ASSERT(!rrw_held(&dp->dp_config_rwlock, RW_READER));
rrw_enter(&dp->dp_config_rwlock, RW_READER, tag);
}
void
dsl_pool_config_enter_prio(dsl_pool_t *dp, void *tag)
{
ASSERT(!rrw_held(&dp->dp_config_rwlock, RW_READER));
rrw_enter_read_prio(&dp->dp_config_rwlock, tag);
}
void
dsl_pool_config_exit(dsl_pool_t *dp, void *tag)
{
rrw_exit(&dp->dp_config_rwlock, tag);
}
boolean_t
dsl_pool_config_held(dsl_pool_t *dp)
{
return (RRW_LOCK_HELD(&dp->dp_config_rwlock));
}
boolean_t
dsl_pool_config_held_writer(dsl_pool_t *dp)
{
return (RRW_WRITE_HELD(&dp->dp_config_rwlock));
}
EXPORT_SYMBOL(dsl_pool_config_enter);
EXPORT_SYMBOL(dsl_pool_config_exit);
/* BEGIN CSTYLED */
/* zfs_dirty_data_max_percent only applied at module load in arc_init(). */
ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max_percent, INT, ZMOD_RD,
"Max percent of RAM allowed to be dirty");
/* zfs_dirty_data_max_max_percent only applied at module load in arc_init(). */
ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max_max_percent, INT, ZMOD_RD,
"zfs_dirty_data_max upper bound as % of RAM");
ZFS_MODULE_PARAM(zfs, zfs_, delay_min_dirty_percent, INT, ZMOD_RW,
"Transaction delay threshold");
ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max, ULONG, ZMOD_RW,
"Determines the dirty space limit");
ZFS_MODULE_PARAM(zfs, zfs_, wrlog_data_max, ULONG, ZMOD_RW,
"The size limit of write-transaction zil log data");
/* zfs_dirty_data_max_max only applied at module load in arc_init(). */
ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max_max, ULONG, ZMOD_RD,
"zfs_dirty_data_max upper bound in bytes");
ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_sync_percent, INT, ZMOD_RW,
"Dirty data txg sync threshold as a percentage of zfs_dirty_data_max");
ZFS_MODULE_PARAM(zfs, zfs_, delay_scale, ULONG, ZMOD_RW,
"How quickly delay approaches infinity");
ZFS_MODULE_PARAM(zfs, zfs_, sync_taskq_batch_pct, INT, ZMOD_RW,
"Max percent of CPUs that are used to sync dirty data");
ZFS_MODULE_PARAM(zfs_zil, zfs_zil_, clean_taskq_nthr_pct, INT, ZMOD_RW,
"Max percent of CPUs that are used per dp_sync_taskq");
ZFS_MODULE_PARAM(zfs_zil, zfs_zil_, clean_taskq_minalloc, INT, ZMOD_RW,
"Number of taskq entries that are pre-populated");
ZFS_MODULE_PARAM(zfs_zil, zfs_zil_, clean_taskq_maxalloc, INT, ZMOD_RW,
"Max number of taskq entries that are cached");
/* END CSTYLED */
diff --git a/module/zfs/dsl_scan.c b/module/zfs/dsl_scan.c
index 859a865dd726..677d320eddba 100644
--- a/module/zfs/dsl_scan.c
+++ b/module/zfs/dsl_scan.c
@@ -1,4440 +1,4441 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2021 by Delphix. All rights reserved.
* Copyright 2016 Gary Mills
* Copyright (c) 2017, 2019, Datto Inc. All rights reserved.
* Copyright (c) 2015, Nexenta Systems, Inc. All rights reserved.
* Copyright 2019 Joyent, Inc.
*/
#include <sys/dsl_scan.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_synctask.h>
#include <sys/dnode.h>
#include <sys/dmu_tx.h>
#include <sys/dmu_objset.h>
#include <sys/arc.h>
#include <sys/zap.h>
#include <sys/zio.h>
#include <sys/zfs_context.h>
#include <sys/fs/zfs.h>
#include <sys/zfs_znode.h>
#include <sys/spa_impl.h>
#include <sys/vdev_impl.h>
#include <sys/zil_impl.h>
#include <sys/zio_checksum.h>
#include <sys/ddt.h>
#include <sys/sa.h>
#include <sys/sa_impl.h>
#include <sys/zfeature.h>
#include <sys/abd.h>
#include <sys/range_tree.h>
#ifdef _KERNEL
#include <sys/zfs_vfsops.h>
#endif
/*
* Grand theory statement on scan queue sorting
*
* Scanning is implemented by recursively traversing all indirection levels
* in an object and reading all blocks referenced from said objects. This
* results in us approximately traversing the object from lowest logical
* offset to the highest. For best performance, we would want the logical
* blocks to be physically contiguous. However, this is frequently not the
* case with pools given the allocation patterns of copy-on-write filesystems.
* So instead, we put the I/Os into a reordering queue and issue them in a
* way that will most benefit physical disks (LBA-order).
*
* Queue management:
*
* Ideally, we would want to scan all metadata and queue up all block I/O
* prior to starting to issue it, because that allows us to do an optimal
* sorting job. This can however consume large amounts of memory. Therefore
* we continuously monitor the size of the queues and constrain them to 5%
* (zfs_scan_mem_lim_fact) of physmem. If the queues grow larger than this
* limit, we clear out a few of the largest extents at the head of the queues
* to make room for more scanning. Hopefully, these extents will be fairly
* large and contiguous, allowing us to approach sequential I/O throughput
* even without a fully sorted tree.
*
* Metadata scanning takes place in dsl_scan_visit(), which is called from
* dsl_scan_sync() every spa_sync(). If we have either fully scanned all
* metadata on the pool, or we need to make room in memory because our
* queues are too large, dsl_scan_visit() is postponed and
* scan_io_queues_run() is called from dsl_scan_sync() instead. This implies
* that metadata scanning and queued I/O issuing are mutually exclusive. This
* allows us to provide maximum sequential I/O throughput for the majority of
* I/O's issued since sequential I/O performance is significantly negatively
* impacted if it is interleaved with random I/O.
*
* Implementation Notes
*
* One side effect of the queued scanning algorithm is that the scanning code
* needs to be notified whenever a block is freed. This is needed to allow
* the scanning code to remove these I/Os from the issuing queue. Additionally,
* we do not attempt to queue gang blocks to be issued sequentially since this
* is very hard to do and would have an extremely limited performance benefit.
* Instead, we simply issue gang I/Os as soon as we find them using the legacy
* algorithm.
*
* Backwards compatibility
*
* This new algorithm is backwards compatible with the legacy on-disk data
* structures (and therefore does not require a new feature flag).
* Periodically during scanning (see zfs_scan_checkpoint_intval), the scan
* will stop scanning metadata (in logical order) and wait for all outstanding
* sorted I/O to complete. Once this is done, we write out a checkpoint
* bookmark, indicating that we have scanned everything logically before it.
* If the pool is imported on a machine without the new sorting algorithm,
* the scan simply resumes from the last checkpoint using the legacy algorithm.
*/
typedef int (scan_cb_t)(dsl_pool_t *, const blkptr_t *,
const zbookmark_phys_t *);
static scan_cb_t dsl_scan_scrub_cb;
static int scan_ds_queue_compare(const void *a, const void *b);
static int scan_prefetch_queue_compare(const void *a, const void *b);
static void scan_ds_queue_clear(dsl_scan_t *scn);
static void scan_ds_prefetch_queue_clear(dsl_scan_t *scn);
static boolean_t scan_ds_queue_contains(dsl_scan_t *scn, uint64_t dsobj,
uint64_t *txg);
static void scan_ds_queue_insert(dsl_scan_t *scn, uint64_t dsobj, uint64_t txg);
static void scan_ds_queue_remove(dsl_scan_t *scn, uint64_t dsobj);
static void scan_ds_queue_sync(dsl_scan_t *scn, dmu_tx_t *tx);
static uint64_t dsl_scan_count_data_disks(vdev_t *vd);
extern int zfs_vdev_async_write_active_min_dirty_percent;
/*
* By default zfs will check to ensure it is not over the hard memory
* limit before each txg. If finer-grained control of this is needed
* this value can be set to 1 to enable checking before scanning each
* block.
*/
-int zfs_scan_strict_mem_lim = B_FALSE;
+static int zfs_scan_strict_mem_lim = B_FALSE;
/*
* Maximum number of parallelly executed bytes per leaf vdev. We attempt
* to strike a balance here between keeping the vdev queues full of I/Os
* at all times and not overflowing the queues to cause long latency,
* which would cause long txg sync times. No matter what, we will not
* overload the drives with I/O, since that is protected by
* zfs_vdev_scrub_max_active.
*/
-unsigned long zfs_scan_vdev_limit = 4 << 20;
+static unsigned long zfs_scan_vdev_limit = 4 << 20;
-int zfs_scan_issue_strategy = 0;
-int zfs_scan_legacy = B_FALSE; /* don't queue & sort zios, go direct */
-unsigned long zfs_scan_max_ext_gap = 2 << 20; /* in bytes */
+static int zfs_scan_issue_strategy = 0;
+static int zfs_scan_legacy = B_FALSE; /* don't queue & sort zios, go direct */
+static unsigned long zfs_scan_max_ext_gap = 2 << 20; /* in bytes */
/*
* fill_weight is non-tunable at runtime, so we copy it at module init from
* zfs_scan_fill_weight. Runtime adjustments to zfs_scan_fill_weight would
* break queue sorting.
*/
-int zfs_scan_fill_weight = 3;
+static int zfs_scan_fill_weight = 3;
static uint64_t fill_weight;
/* See dsl_scan_should_clear() for details on the memory limit tunables */
-uint64_t zfs_scan_mem_lim_min = 16 << 20; /* bytes */
-uint64_t zfs_scan_mem_lim_soft_max = 128 << 20; /* bytes */
-int zfs_scan_mem_lim_fact = 20; /* fraction of physmem */
-int zfs_scan_mem_lim_soft_fact = 20; /* fraction of mem lim above */
-
-int zfs_scrub_min_time_ms = 1000; /* min millisecs to scrub per txg */
-int zfs_obsolete_min_time_ms = 500; /* min millisecs to obsolete per txg */
-int zfs_free_min_time_ms = 1000; /* min millisecs to free per txg */
-int zfs_resilver_min_time_ms = 3000; /* min millisecs to resilver per txg */
-int zfs_scan_checkpoint_intval = 7200; /* in seconds */
+static const uint64_t zfs_scan_mem_lim_min = 16 << 20; /* bytes */
+static const uint64_t zfs_scan_mem_lim_soft_max = 128 << 20; /* bytes */
+static int zfs_scan_mem_lim_fact = 20; /* fraction of physmem */
+static int zfs_scan_mem_lim_soft_fact = 20; /* fraction of mem lim above */
+
+static int zfs_scrub_min_time_ms = 1000; /* min millis to scrub per txg */
+static int zfs_obsolete_min_time_ms = 500; /* min millis to obsolete per txg */
+static int zfs_free_min_time_ms = 1000; /* min millis to free per txg */
+static int zfs_resilver_min_time_ms = 3000; /* min millis to resilver per txg */
+static int zfs_scan_checkpoint_intval = 7200; /* in seconds */
int zfs_scan_suspend_progress = 0; /* set to prevent scans from progressing */
-int zfs_no_scrub_io = B_FALSE; /* set to disable scrub i/o */
-int zfs_no_scrub_prefetch = B_FALSE; /* set to disable scrub prefetch */
-enum ddt_class zfs_scrub_ddt_class_max = DDT_CLASS_DUPLICATE;
+static int zfs_no_scrub_io = B_FALSE; /* set to disable scrub i/o */
+static int zfs_no_scrub_prefetch = B_FALSE; /* set to disable scrub prefetch */
+static const enum ddt_class zfs_scrub_ddt_class_max = DDT_CLASS_DUPLICATE;
/* max number of blocks to free in a single TXG */
-unsigned long zfs_async_block_max_blocks = ULONG_MAX;
+static unsigned long zfs_async_block_max_blocks = ULONG_MAX;
/* max number of dedup blocks to free in a single TXG */
-unsigned long zfs_max_async_dedup_frees = 100000;
+static unsigned long zfs_max_async_dedup_frees = 100000;
-int zfs_resilver_disable_defer = 0; /* set to disable resilver deferring */
+/* set to disable resilver deferring */
+static int zfs_resilver_disable_defer = B_FALSE;
/*
* We wait a few txgs after importing a pool to begin scanning so that
* the import / mounting code isn't held up by scrub / resilver IO.
* Unfortunately, it is a bit difficult to determine exactly how long
* this will take since userspace will trigger fs mounts asynchronously
* and the kernel will create zvol minors asynchronously. As a result,
* the value provided here is a bit arbitrary, but represents a
* reasonable estimate of how many txgs it will take to finish fully
* importing a pool
*/
#define SCAN_IMPORT_WAIT_TXGS 5
#define DSL_SCAN_IS_SCRUB_RESILVER(scn) \
((scn)->scn_phys.scn_func == POOL_SCAN_SCRUB || \
(scn)->scn_phys.scn_func == POOL_SCAN_RESILVER)
/*
* Enable/disable the processing of the free_bpobj object.
*/
-int zfs_free_bpobj_enabled = 1;
+static int zfs_free_bpobj_enabled = 1;
/* the order has to match pool_scan_type */
static scan_cb_t *scan_funcs[POOL_SCAN_FUNCS] = {
NULL,
dsl_scan_scrub_cb, /* POOL_SCAN_SCRUB */
dsl_scan_scrub_cb, /* POOL_SCAN_RESILVER */
};
/* In core node for the scn->scn_queue. Represents a dataset to be scanned */
typedef struct {
uint64_t sds_dsobj;
uint64_t sds_txg;
avl_node_t sds_node;
} scan_ds_t;
/*
* This controls what conditions are placed on dsl_scan_sync_state():
* SYNC_OPTIONAL) write out scn_phys iff scn_bytes_pending == 0
* SYNC_MANDATORY) write out scn_phys always. scn_bytes_pending must be 0.
* SYNC_CACHED) if scn_bytes_pending == 0, write out scn_phys. Otherwise
* write out the scn_phys_cached version.
* See dsl_scan_sync_state for details.
*/
typedef enum {
SYNC_OPTIONAL,
SYNC_MANDATORY,
SYNC_CACHED
} state_sync_type_t;
/*
* This struct represents the minimum information needed to reconstruct a
* zio for sequential scanning. This is useful because many of these will
* accumulate in the sequential IO queues before being issued, so saving
* memory matters here.
*/
typedef struct scan_io {
/* fields from blkptr_t */
uint64_t sio_blk_prop;
uint64_t sio_phys_birth;
uint64_t sio_birth;
zio_cksum_t sio_cksum;
uint32_t sio_nr_dvas;
/* fields from zio_t */
uint32_t sio_flags;
zbookmark_phys_t sio_zb;
/* members for queue sorting */
union {
avl_node_t sio_addr_node; /* link into issuing queue */
list_node_t sio_list_node; /* link for issuing to disk */
} sio_nodes;
/*
* There may be up to SPA_DVAS_PER_BP DVAs here from the bp,
* depending on how many were in the original bp. Only the
* first DVA is really used for sorting and issuing purposes.
* The other DVAs (if provided) simply exist so that the zio
* layer can find additional copies to repair from in the
* event of an error. This array must go at the end of the
* struct to allow this for the variable number of elements.
*/
dva_t sio_dva[0];
} scan_io_t;
#define SIO_SET_OFFSET(sio, x) DVA_SET_OFFSET(&(sio)->sio_dva[0], x)
#define SIO_SET_ASIZE(sio, x) DVA_SET_ASIZE(&(sio)->sio_dva[0], x)
#define SIO_GET_OFFSET(sio) DVA_GET_OFFSET(&(sio)->sio_dva[0])
#define SIO_GET_ASIZE(sio) DVA_GET_ASIZE(&(sio)->sio_dva[0])
#define SIO_GET_END_OFFSET(sio) \
(SIO_GET_OFFSET(sio) + SIO_GET_ASIZE(sio))
#define SIO_GET_MUSED(sio) \
(sizeof (scan_io_t) + ((sio)->sio_nr_dvas * sizeof (dva_t)))
struct dsl_scan_io_queue {
dsl_scan_t *q_scn; /* associated dsl_scan_t */
vdev_t *q_vd; /* top-level vdev that this queue represents */
/* trees used for sorting I/Os and extents of I/Os */
range_tree_t *q_exts_by_addr;
zfs_btree_t q_exts_by_size;
avl_tree_t q_sios_by_addr;
uint64_t q_sio_memused;
/* members for zio rate limiting */
uint64_t q_maxinflight_bytes;
uint64_t q_inflight_bytes;
kcondvar_t q_zio_cv; /* used under vd->vdev_scan_io_queue_lock */
/* per txg statistics */
uint64_t q_total_seg_size_this_txg;
uint64_t q_segs_this_txg;
uint64_t q_total_zio_size_this_txg;
uint64_t q_zios_this_txg;
};
/* private data for dsl_scan_prefetch_cb() */
typedef struct scan_prefetch_ctx {
zfs_refcount_t spc_refcnt; /* refcount for memory management */
dsl_scan_t *spc_scn; /* dsl_scan_t for the pool */
boolean_t spc_root; /* is this prefetch for an objset? */
uint8_t spc_indblkshift; /* dn_indblkshift of current dnode */
uint16_t spc_datablkszsec; /* dn_idatablkszsec of current dnode */
} scan_prefetch_ctx_t;
/* private data for dsl_scan_prefetch() */
typedef struct scan_prefetch_issue_ctx {
avl_node_t spic_avl_node; /* link into scn->scn_prefetch_queue */
scan_prefetch_ctx_t *spic_spc; /* spc for the callback */
blkptr_t spic_bp; /* bp to prefetch */
zbookmark_phys_t spic_zb; /* bookmark to prefetch */
} scan_prefetch_issue_ctx_t;
static void scan_exec_io(dsl_pool_t *dp, const blkptr_t *bp, int zio_flags,
const zbookmark_phys_t *zb, dsl_scan_io_queue_t *queue);
static void scan_io_queue_insert_impl(dsl_scan_io_queue_t *queue,
scan_io_t *sio);
static dsl_scan_io_queue_t *scan_io_queue_create(vdev_t *vd);
static void scan_io_queues_destroy(dsl_scan_t *scn);
static kmem_cache_t *sio_cache[SPA_DVAS_PER_BP];
/* sio->sio_nr_dvas must be set so we know which cache to free from */
static void
sio_free(scan_io_t *sio)
{
ASSERT3U(sio->sio_nr_dvas, >, 0);
ASSERT3U(sio->sio_nr_dvas, <=, SPA_DVAS_PER_BP);
kmem_cache_free(sio_cache[sio->sio_nr_dvas - 1], sio);
}
/* It is up to the caller to set sio->sio_nr_dvas for freeing */
static scan_io_t *
sio_alloc(unsigned short nr_dvas)
{
ASSERT3U(nr_dvas, >, 0);
ASSERT3U(nr_dvas, <=, SPA_DVAS_PER_BP);
return (kmem_cache_alloc(sio_cache[nr_dvas - 1], KM_SLEEP));
}
void
scan_init(void)
{
/*
* This is used in ext_size_compare() to weight segments
* based on how sparse they are. This cannot be changed
* mid-scan and the tree comparison functions don't currently
* have a mechanism for passing additional context to the
* compare functions. Thus we store this value globally and
* we only allow it to be set at module initialization time
*/
fill_weight = zfs_scan_fill_weight;
for (int i = 0; i < SPA_DVAS_PER_BP; i++) {
char name[36];
(void) snprintf(name, sizeof (name), "sio_cache_%d", i);
sio_cache[i] = kmem_cache_create(name,
(sizeof (scan_io_t) + ((i + 1) * sizeof (dva_t))),
0, NULL, NULL, NULL, NULL, NULL, 0);
}
}
void
scan_fini(void)
{
for (int i = 0; i < SPA_DVAS_PER_BP; i++) {
kmem_cache_destroy(sio_cache[i]);
}
}
static inline boolean_t
dsl_scan_is_running(const dsl_scan_t *scn)
{
return (scn->scn_phys.scn_state == DSS_SCANNING);
}
boolean_t
dsl_scan_resilvering(dsl_pool_t *dp)
{
return (dsl_scan_is_running(dp->dp_scan) &&
dp->dp_scan->scn_phys.scn_func == POOL_SCAN_RESILVER);
}
static inline void
sio2bp(const scan_io_t *sio, blkptr_t *bp)
{
bzero(bp, sizeof (*bp));
bp->blk_prop = sio->sio_blk_prop;
bp->blk_phys_birth = sio->sio_phys_birth;
bp->blk_birth = sio->sio_birth;
bp->blk_fill = 1; /* we always only work with data pointers */
bp->blk_cksum = sio->sio_cksum;
ASSERT3U(sio->sio_nr_dvas, >, 0);
ASSERT3U(sio->sio_nr_dvas, <=, SPA_DVAS_PER_BP);
bcopy(sio->sio_dva, bp->blk_dva, sio->sio_nr_dvas * sizeof (dva_t));
}
static inline void
bp2sio(const blkptr_t *bp, scan_io_t *sio, int dva_i)
{
sio->sio_blk_prop = bp->blk_prop;
sio->sio_phys_birth = bp->blk_phys_birth;
sio->sio_birth = bp->blk_birth;
sio->sio_cksum = bp->blk_cksum;
sio->sio_nr_dvas = BP_GET_NDVAS(bp);
/*
* Copy the DVAs to the sio. We need all copies of the block so
* that the self healing code can use the alternate copies if the
* first is corrupted. We want the DVA at index dva_i to be first
* in the sio since this is the primary one that we want to issue.
*/
for (int i = 0, j = dva_i; i < sio->sio_nr_dvas; i++, j++) {
sio->sio_dva[i] = bp->blk_dva[j % sio->sio_nr_dvas];
}
}
int
dsl_scan_init(dsl_pool_t *dp, uint64_t txg)
{
int err;
dsl_scan_t *scn;
spa_t *spa = dp->dp_spa;
uint64_t f;
scn = dp->dp_scan = kmem_zalloc(sizeof (dsl_scan_t), KM_SLEEP);
scn->scn_dp = dp;
/*
* It's possible that we're resuming a scan after a reboot so
* make sure that the scan_async_destroying flag is initialized
* appropriately.
*/
ASSERT(!scn->scn_async_destroying);
scn->scn_async_destroying = spa_feature_is_active(dp->dp_spa,
SPA_FEATURE_ASYNC_DESTROY);
/*
* Calculate the max number of in-flight bytes for pool-wide
* scanning operations (minimum 1MB). Limits for the issuing
* phase are done per top-level vdev and are handled separately.
*/
scn->scn_maxinflight_bytes = MAX(zfs_scan_vdev_limit *
dsl_scan_count_data_disks(spa->spa_root_vdev), 1ULL << 20);
avl_create(&scn->scn_queue, scan_ds_queue_compare, sizeof (scan_ds_t),
offsetof(scan_ds_t, sds_node));
avl_create(&scn->scn_prefetch_queue, scan_prefetch_queue_compare,
sizeof (scan_prefetch_issue_ctx_t),
offsetof(scan_prefetch_issue_ctx_t, spic_avl_node));
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
"scrub_func", sizeof (uint64_t), 1, &f);
if (err == 0) {
/*
* There was an old-style scrub in progress. Restart a
* new-style scrub from the beginning.
*/
scn->scn_restart_txg = txg;
zfs_dbgmsg("old-style scrub was in progress for %s; "
"restarting new-style scrub in txg %llu",
spa->spa_name,
(longlong_t)scn->scn_restart_txg);
/*
* Load the queue obj from the old location so that it
* can be freed by dsl_scan_done().
*/
(void) zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
"scrub_queue", sizeof (uint64_t), 1,
&scn->scn_phys.scn_queue_obj);
} else {
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
&scn->scn_phys);
/*
* Detect if the pool contains the signature of #2094. If it
* does properly update the scn->scn_phys structure and notify
* the administrator by setting an errata for the pool.
*/
if (err == EOVERFLOW) {
uint64_t zaptmp[SCAN_PHYS_NUMINTS + 1];
VERIFY3S(SCAN_PHYS_NUMINTS, ==, 24);
VERIFY3S(offsetof(dsl_scan_phys_t, scn_flags), ==,
(23 * sizeof (uint64_t)));
err = zap_lookup(dp->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SCAN,
sizeof (uint64_t), SCAN_PHYS_NUMINTS + 1, &zaptmp);
if (err == 0) {
uint64_t overflow = zaptmp[SCAN_PHYS_NUMINTS];
if (overflow & ~DSL_SCAN_FLAGS_MASK ||
scn->scn_async_destroying) {
spa->spa_errata =
ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY;
return (EOVERFLOW);
}
bcopy(zaptmp, &scn->scn_phys,
SCAN_PHYS_NUMINTS * sizeof (uint64_t));
scn->scn_phys.scn_flags = overflow;
/* Required scrub already in progress. */
if (scn->scn_phys.scn_state == DSS_FINISHED ||
scn->scn_phys.scn_state == DSS_CANCELED)
spa->spa_errata =
ZPOOL_ERRATA_ZOL_2094_SCRUB;
}
}
if (err == ENOENT)
return (0);
else if (err)
return (err);
/*
* We might be restarting after a reboot, so jump the issued
* counter to how far we've scanned. We know we're consistent
* up to here.
*/
scn->scn_issued_before_pass = scn->scn_phys.scn_examined;
if (dsl_scan_is_running(scn) &&
spa_prev_software_version(dp->dp_spa) < SPA_VERSION_SCAN) {
/*
* A new-type scrub was in progress on an old
* pool, and the pool was accessed by old
* software. Restart from the beginning, since
* the old software may have changed the pool in
* the meantime.
*/
scn->scn_restart_txg = txg;
zfs_dbgmsg("new-style scrub for %s was modified "
"by old software; restarting in txg %llu",
spa->spa_name,
(longlong_t)scn->scn_restart_txg);
} else if (dsl_scan_resilvering(dp)) {
/*
* If a resilver is in progress and there are already
* errors, restart it instead of finishing this scan and
* then restarting it. If there haven't been any errors
* then remember that the incore DTL is valid.
*/
if (scn->scn_phys.scn_errors > 0) {
scn->scn_restart_txg = txg;
zfs_dbgmsg("resilver can't excise DTL_MISSING "
"when finished; restarting on %s in txg "
"%llu",
spa->spa_name,
(u_longlong_t)scn->scn_restart_txg);
} else {
/* it's safe to excise DTL when finished */
spa->spa_scrub_started = B_TRUE;
}
}
}
bcopy(&scn->scn_phys, &scn->scn_phys_cached, sizeof (scn->scn_phys));
/* reload the queue into the in-core state */
if (scn->scn_phys.scn_queue_obj != 0) {
zap_cursor_t zc;
zap_attribute_t za;
for (zap_cursor_init(&zc, dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj);
zap_cursor_retrieve(&zc, &za) == 0;
(void) zap_cursor_advance(&zc)) {
scan_ds_queue_insert(scn,
zfs_strtonum(za.za_name, NULL),
za.za_first_integer);
}
zap_cursor_fini(&zc);
}
spa_scan_stat_init(spa);
return (0);
}
void
dsl_scan_fini(dsl_pool_t *dp)
{
if (dp->dp_scan != NULL) {
dsl_scan_t *scn = dp->dp_scan;
if (scn->scn_taskq != NULL)
taskq_destroy(scn->scn_taskq);
scan_ds_queue_clear(scn);
avl_destroy(&scn->scn_queue);
scan_ds_prefetch_queue_clear(scn);
avl_destroy(&scn->scn_prefetch_queue);
kmem_free(dp->dp_scan, sizeof (dsl_scan_t));
dp->dp_scan = NULL;
}
}
static boolean_t
dsl_scan_restarting(dsl_scan_t *scn, dmu_tx_t *tx)
{
return (scn->scn_restart_txg != 0 &&
scn->scn_restart_txg <= tx->tx_txg);
}
boolean_t
dsl_scan_resilver_scheduled(dsl_pool_t *dp)
{
return ((dp->dp_scan && dp->dp_scan->scn_restart_txg != 0) ||
(spa_async_tasks(dp->dp_spa) & SPA_ASYNC_RESILVER));
}
boolean_t
dsl_scan_scrubbing(const dsl_pool_t *dp)
{
dsl_scan_phys_t *scn_phys = &dp->dp_scan->scn_phys;
return (scn_phys->scn_state == DSS_SCANNING &&
scn_phys->scn_func == POOL_SCAN_SCRUB);
}
boolean_t
dsl_scan_is_paused_scrub(const dsl_scan_t *scn)
{
return (dsl_scan_scrubbing(scn->scn_dp) &&
scn->scn_phys.scn_flags & DSF_SCRUB_PAUSED);
}
/*
* Writes out a persistent dsl_scan_phys_t record to the pool directory.
* Because we can be running in the block sorting algorithm, we do not always
* want to write out the record, only when it is "safe" to do so. This safety
* condition is achieved by making sure that the sorting queues are empty
* (scn_bytes_pending == 0). When this condition is not true, the sync'd state
* is inconsistent with how much actual scanning progress has been made. The
* kind of sync to be performed is specified by the sync_type argument. If the
* sync is optional, we only sync if the queues are empty. If the sync is
* mandatory, we do a hard ASSERT to make sure that the queues are empty. The
* third possible state is a "cached" sync. This is done in response to:
* 1) The dataset that was in the last sync'd dsl_scan_phys_t having been
* destroyed, so we wouldn't be able to restart scanning from it.
* 2) The snapshot that was in the last sync'd dsl_scan_phys_t having been
* superseded by a newer snapshot.
* 3) The dataset that was in the last sync'd dsl_scan_phys_t having been
* swapped with its clone.
* In all cases, a cached sync simply rewrites the last record we've written,
* just slightly modified. For the modifications that are performed to the
* last written dsl_scan_phys_t, see dsl_scan_ds_destroyed,
* dsl_scan_ds_snapshotted and dsl_scan_ds_clone_swapped.
*/
static void
dsl_scan_sync_state(dsl_scan_t *scn, dmu_tx_t *tx, state_sync_type_t sync_type)
{
int i;
spa_t *spa = scn->scn_dp->dp_spa;
ASSERT(sync_type != SYNC_MANDATORY || scn->scn_bytes_pending == 0);
if (scn->scn_bytes_pending == 0) {
for (i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
vdev_t *vd = spa->spa_root_vdev->vdev_child[i];
dsl_scan_io_queue_t *q = vd->vdev_scan_io_queue;
if (q == NULL)
continue;
mutex_enter(&vd->vdev_scan_io_queue_lock);
ASSERT3P(avl_first(&q->q_sios_by_addr), ==, NULL);
ASSERT3P(zfs_btree_first(&q->q_exts_by_size, NULL), ==,
NULL);
ASSERT3P(range_tree_first(q->q_exts_by_addr), ==, NULL);
mutex_exit(&vd->vdev_scan_io_queue_lock);
}
if (scn->scn_phys.scn_queue_obj != 0)
scan_ds_queue_sync(scn, tx);
VERIFY0(zap_update(scn->scn_dp->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
&scn->scn_phys, tx));
bcopy(&scn->scn_phys, &scn->scn_phys_cached,
sizeof (scn->scn_phys));
if (scn->scn_checkpointing)
zfs_dbgmsg("finish scan checkpoint for %s",
spa->spa_name);
scn->scn_checkpointing = B_FALSE;
scn->scn_last_checkpoint = ddi_get_lbolt();
} else if (sync_type == SYNC_CACHED) {
VERIFY0(zap_update(scn->scn_dp->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
&scn->scn_phys_cached, tx));
}
}
int
dsl_scan_setup_check(void *arg, dmu_tx_t *tx)
{
(void) arg;
dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;
vdev_t *rvd = scn->scn_dp->dp_spa->spa_root_vdev;
if (dsl_scan_is_running(scn) || vdev_rebuild_active(rvd))
return (SET_ERROR(EBUSY));
return (0);
}
void
dsl_scan_setup_sync(void *arg, dmu_tx_t *tx)
{
dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;
pool_scan_func_t *funcp = arg;
dmu_object_type_t ot = 0;
dsl_pool_t *dp = scn->scn_dp;
spa_t *spa = dp->dp_spa;
ASSERT(!dsl_scan_is_running(scn));
ASSERT(*funcp > POOL_SCAN_NONE && *funcp < POOL_SCAN_FUNCS);
bzero(&scn->scn_phys, sizeof (scn->scn_phys));
scn->scn_phys.scn_func = *funcp;
scn->scn_phys.scn_state = DSS_SCANNING;
scn->scn_phys.scn_min_txg = 0;
scn->scn_phys.scn_max_txg = tx->tx_txg;
scn->scn_phys.scn_ddt_class_max = DDT_CLASSES - 1; /* the entire DDT */
scn->scn_phys.scn_start_time = gethrestime_sec();
scn->scn_phys.scn_errors = 0;
scn->scn_phys.scn_to_examine = spa->spa_root_vdev->vdev_stat.vs_alloc;
scn->scn_issued_before_pass = 0;
scn->scn_restart_txg = 0;
scn->scn_done_txg = 0;
scn->scn_last_checkpoint = 0;
scn->scn_checkpointing = B_FALSE;
spa_scan_stat_init(spa);
if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
scn->scn_phys.scn_ddt_class_max = zfs_scrub_ddt_class_max;
/* rewrite all disk labels */
vdev_config_dirty(spa->spa_root_vdev);
if (vdev_resilver_needed(spa->spa_root_vdev,
&scn->scn_phys.scn_min_txg, &scn->scn_phys.scn_max_txg)) {
nvlist_t *aux = fnvlist_alloc();
fnvlist_add_string(aux, ZFS_EV_RESILVER_TYPE,
"healing");
spa_event_notify(spa, NULL, aux,
ESC_ZFS_RESILVER_START);
nvlist_free(aux);
} else {
spa_event_notify(spa, NULL, NULL, ESC_ZFS_SCRUB_START);
}
spa->spa_scrub_started = B_TRUE;
/*
* If this is an incremental scrub, limit the DDT scrub phase
* to just the auto-ditto class (for correctness); the rest
* of the scrub should go faster using top-down pruning.
*/
if (scn->scn_phys.scn_min_txg > TXG_INITIAL)
scn->scn_phys.scn_ddt_class_max = DDT_CLASS_DITTO;
/*
* When starting a resilver clear any existing rebuild state.
* This is required to prevent stale rebuild status from
* being reported when a rebuild is run, then a resilver and
* finally a scrub. In which case only the scrub status
* should be reported by 'zpool status'.
*/
if (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) {
vdev_t *rvd = spa->spa_root_vdev;
for (uint64_t i = 0; i < rvd->vdev_children; i++) {
vdev_t *vd = rvd->vdev_child[i];
vdev_rebuild_clear_sync(
(void *)(uintptr_t)vd->vdev_id, tx);
}
}
}
/* back to the generic stuff */
if (dp->dp_blkstats == NULL) {
dp->dp_blkstats =
vmem_alloc(sizeof (zfs_all_blkstats_t), KM_SLEEP);
mutex_init(&dp->dp_blkstats->zab_lock, NULL,
MUTEX_DEFAULT, NULL);
}
bzero(&dp->dp_blkstats->zab_type, sizeof (dp->dp_blkstats->zab_type));
if (spa_version(spa) < SPA_VERSION_DSL_SCRUB)
ot = DMU_OT_ZAP_OTHER;
scn->scn_phys.scn_queue_obj = zap_create(dp->dp_meta_objset,
ot ? ot : DMU_OT_SCAN_QUEUE, DMU_OT_NONE, 0, tx);
bcopy(&scn->scn_phys, &scn->scn_phys_cached, sizeof (scn->scn_phys));
dsl_scan_sync_state(scn, tx, SYNC_MANDATORY);
spa_history_log_internal(spa, "scan setup", tx,
"func=%u mintxg=%llu maxtxg=%llu",
*funcp, (u_longlong_t)scn->scn_phys.scn_min_txg,
(u_longlong_t)scn->scn_phys.scn_max_txg);
}
/*
* Called by the ZFS_IOC_POOL_SCAN ioctl to start a scrub or resilver.
* Can also be called to resume a paused scrub.
*/
int
dsl_scan(dsl_pool_t *dp, pool_scan_func_t func)
{
spa_t *spa = dp->dp_spa;
dsl_scan_t *scn = dp->dp_scan;
/*
* Purge all vdev caches and probe all devices. We do this here
* rather than in sync context because this requires a writer lock
* on the spa_config lock, which we can't do from sync context. The
* spa_scrub_reopen flag indicates that vdev_open() should not
* attempt to start another scrub.
*/
spa_vdev_state_enter(spa, SCL_NONE);
spa->spa_scrub_reopen = B_TRUE;
vdev_reopen(spa->spa_root_vdev);
spa->spa_scrub_reopen = B_FALSE;
(void) spa_vdev_state_exit(spa, NULL, 0);
if (func == POOL_SCAN_RESILVER) {
dsl_scan_restart_resilver(spa->spa_dsl_pool, 0);
return (0);
}
if (func == POOL_SCAN_SCRUB && dsl_scan_is_paused_scrub(scn)) {
/* got scrub start cmd, resume paused scrub */
int err = dsl_scrub_set_pause_resume(scn->scn_dp,
POOL_SCRUB_NORMAL);
if (err == 0) {
spa_event_notify(spa, NULL, NULL, ESC_ZFS_SCRUB_RESUME);
return (SET_ERROR(ECANCELED));
}
return (SET_ERROR(err));
}
return (dsl_sync_task(spa_name(spa), dsl_scan_setup_check,
dsl_scan_setup_sync, &func, 0, ZFS_SPACE_CHECK_EXTRA_RESERVED));
}
static void
dsl_scan_done(dsl_scan_t *scn, boolean_t complete, dmu_tx_t *tx)
{
static const char *old_names[] = {
"scrub_bookmark",
"scrub_ddt_bookmark",
"scrub_ddt_class_max",
"scrub_queue",
"scrub_min_txg",
"scrub_max_txg",
"scrub_func",
"scrub_errors",
NULL
};
dsl_pool_t *dp = scn->scn_dp;
spa_t *spa = dp->dp_spa;
int i;
/* Remove any remnants of an old-style scrub. */
for (i = 0; old_names[i]; i++) {
(void) zap_remove(dp->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, old_names[i], tx);
}
if (scn->scn_phys.scn_queue_obj != 0) {
VERIFY0(dmu_object_free(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, tx));
scn->scn_phys.scn_queue_obj = 0;
}
scan_ds_queue_clear(scn);
scan_ds_prefetch_queue_clear(scn);
scn->scn_phys.scn_flags &= ~DSF_SCRUB_PAUSED;
/*
* If we were "restarted" from a stopped state, don't bother
* with anything else.
*/
if (!dsl_scan_is_running(scn)) {
ASSERT(!scn->scn_is_sorted);
return;
}
if (scn->scn_is_sorted) {
scan_io_queues_destroy(scn);
scn->scn_is_sorted = B_FALSE;
if (scn->scn_taskq != NULL) {
taskq_destroy(scn->scn_taskq);
scn->scn_taskq = NULL;
}
}
scn->scn_phys.scn_state = complete ? DSS_FINISHED : DSS_CANCELED;
spa_notify_waiters(spa);
if (dsl_scan_restarting(scn, tx))
spa_history_log_internal(spa, "scan aborted, restarting", tx,
"errors=%llu", (u_longlong_t)spa_get_errlog_size(spa));
else if (!complete)
spa_history_log_internal(spa, "scan cancelled", tx,
"errors=%llu", (u_longlong_t)spa_get_errlog_size(spa));
else
spa_history_log_internal(spa, "scan done", tx,
"errors=%llu", (u_longlong_t)spa_get_errlog_size(spa));
if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
spa->spa_scrub_active = B_FALSE;
/*
* If the scrub/resilver completed, update all DTLs to
* reflect this. Whether it succeeded or not, vacate
* all temporary scrub DTLs.
*
* As the scrub does not currently support traversing
* data that have been freed but are part of a checkpoint,
* we don't mark the scrub as done in the DTLs as faults
* may still exist in those vdevs.
*/
if (complete &&
!spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
vdev_dtl_reassess(spa->spa_root_vdev, tx->tx_txg,
scn->scn_phys.scn_max_txg, B_TRUE, B_FALSE);
if (scn->scn_phys.scn_min_txg) {
nvlist_t *aux = fnvlist_alloc();
fnvlist_add_string(aux, ZFS_EV_RESILVER_TYPE,
"healing");
spa_event_notify(spa, NULL, aux,
ESC_ZFS_RESILVER_FINISH);
nvlist_free(aux);
} else {
spa_event_notify(spa, NULL, NULL,
ESC_ZFS_SCRUB_FINISH);
}
} else {
vdev_dtl_reassess(spa->spa_root_vdev, tx->tx_txg,
0, B_TRUE, B_FALSE);
}
spa_errlog_rotate(spa);
/*
* Don't clear flag until after vdev_dtl_reassess to ensure that
* DTL_MISSING will get updated when possible.
*/
spa->spa_scrub_started = B_FALSE;
/*
* We may have finished replacing a device.
* Let the async thread assess this and handle the detach.
*/
spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);
/*
* Clear any resilver_deferred flags in the config.
* If there are drives that need resilvering, kick
* off an asynchronous request to start resilver.
* vdev_clear_resilver_deferred() may update the config
* before the resilver can restart. In the event of
* a crash during this period, the spa loading code
* will find the drives that need to be resilvered
* and start the resilver then.
*/
if (spa_feature_is_enabled(spa, SPA_FEATURE_RESILVER_DEFER) &&
vdev_clear_resilver_deferred(spa->spa_root_vdev, tx)) {
spa_history_log_internal(spa,
"starting deferred resilver", tx, "errors=%llu",
(u_longlong_t)spa_get_errlog_size(spa));
spa_async_request(spa, SPA_ASYNC_RESILVER);
}
/* Clear recent error events (i.e. duplicate events tracking) */
if (complete)
zfs_ereport_clear(spa, NULL);
}
scn->scn_phys.scn_end_time = gethrestime_sec();
if (spa->spa_errata == ZPOOL_ERRATA_ZOL_2094_SCRUB)
spa->spa_errata = 0;
ASSERT(!dsl_scan_is_running(scn));
}
static int
dsl_scan_cancel_check(void *arg, dmu_tx_t *tx)
{
(void) arg;
dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;
if (!dsl_scan_is_running(scn))
return (SET_ERROR(ENOENT));
return (0);
}
static void
dsl_scan_cancel_sync(void *arg, dmu_tx_t *tx)
{
(void) arg;
dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;
dsl_scan_done(scn, B_FALSE, tx);
dsl_scan_sync_state(scn, tx, SYNC_MANDATORY);
spa_event_notify(scn->scn_dp->dp_spa, NULL, NULL, ESC_ZFS_SCRUB_ABORT);
}
int
dsl_scan_cancel(dsl_pool_t *dp)
{
return (dsl_sync_task(spa_name(dp->dp_spa), dsl_scan_cancel_check,
dsl_scan_cancel_sync, NULL, 3, ZFS_SPACE_CHECK_RESERVED));
}
static int
dsl_scrub_pause_resume_check(void *arg, dmu_tx_t *tx)
{
pool_scrub_cmd_t *cmd = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_scan_t *scn = dp->dp_scan;
if (*cmd == POOL_SCRUB_PAUSE) {
/* can't pause a scrub when there is no in-progress scrub */
if (!dsl_scan_scrubbing(dp))
return (SET_ERROR(ENOENT));
/* can't pause a paused scrub */
if (dsl_scan_is_paused_scrub(scn))
return (SET_ERROR(EBUSY));
} else if (*cmd != POOL_SCRUB_NORMAL) {
return (SET_ERROR(ENOTSUP));
}
return (0);
}
static void
dsl_scrub_pause_resume_sync(void *arg, dmu_tx_t *tx)
{
pool_scrub_cmd_t *cmd = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
spa_t *spa = dp->dp_spa;
dsl_scan_t *scn = dp->dp_scan;
if (*cmd == POOL_SCRUB_PAUSE) {
/* can't pause a scrub when there is no in-progress scrub */
spa->spa_scan_pass_scrub_pause = gethrestime_sec();
scn->scn_phys.scn_flags |= DSF_SCRUB_PAUSED;
scn->scn_phys_cached.scn_flags |= DSF_SCRUB_PAUSED;
dsl_scan_sync_state(scn, tx, SYNC_CACHED);
spa_event_notify(spa, NULL, NULL, ESC_ZFS_SCRUB_PAUSED);
spa_notify_waiters(spa);
} else {
ASSERT3U(*cmd, ==, POOL_SCRUB_NORMAL);
if (dsl_scan_is_paused_scrub(scn)) {
/*
* We need to keep track of how much time we spend
* paused per pass so that we can adjust the scrub rate
* shown in the output of 'zpool status'
*/
spa->spa_scan_pass_scrub_spent_paused +=
gethrestime_sec() - spa->spa_scan_pass_scrub_pause;
spa->spa_scan_pass_scrub_pause = 0;
scn->scn_phys.scn_flags &= ~DSF_SCRUB_PAUSED;
scn->scn_phys_cached.scn_flags &= ~DSF_SCRUB_PAUSED;
dsl_scan_sync_state(scn, tx, SYNC_CACHED);
}
}
}
/*
* Set scrub pause/resume state if it makes sense to do so
*/
int
dsl_scrub_set_pause_resume(const dsl_pool_t *dp, pool_scrub_cmd_t cmd)
{
return (dsl_sync_task(spa_name(dp->dp_spa),
dsl_scrub_pause_resume_check, dsl_scrub_pause_resume_sync, &cmd, 3,
ZFS_SPACE_CHECK_RESERVED));
}
/* start a new scan, or restart an existing one. */
void
dsl_scan_restart_resilver(dsl_pool_t *dp, uint64_t txg)
{
if (txg == 0) {
dmu_tx_t *tx;
tx = dmu_tx_create_dd(dp->dp_mos_dir);
VERIFY(0 == dmu_tx_assign(tx, TXG_WAIT));
txg = dmu_tx_get_txg(tx);
dp->dp_scan->scn_restart_txg = txg;
dmu_tx_commit(tx);
} else {
dp->dp_scan->scn_restart_txg = txg;
}
zfs_dbgmsg("restarting resilver for %s at txg=%llu",
dp->dp_spa->spa_name, (longlong_t)txg);
}
void
dsl_free(dsl_pool_t *dp, uint64_t txg, const blkptr_t *bp)
{
zio_free(dp->dp_spa, txg, bp);
}
void
dsl_free_sync(zio_t *pio, dsl_pool_t *dp, uint64_t txg, const blkptr_t *bpp)
{
ASSERT(dsl_pool_sync_context(dp));
zio_nowait(zio_free_sync(pio, dp->dp_spa, txg, bpp, pio->io_flags));
}
static int
scan_ds_queue_compare(const void *a, const void *b)
{
const scan_ds_t *sds_a = a, *sds_b = b;
if (sds_a->sds_dsobj < sds_b->sds_dsobj)
return (-1);
if (sds_a->sds_dsobj == sds_b->sds_dsobj)
return (0);
return (1);
}
static void
scan_ds_queue_clear(dsl_scan_t *scn)
{
void *cookie = NULL;
scan_ds_t *sds;
while ((sds = avl_destroy_nodes(&scn->scn_queue, &cookie)) != NULL) {
kmem_free(sds, sizeof (*sds));
}
}
static boolean_t
scan_ds_queue_contains(dsl_scan_t *scn, uint64_t dsobj, uint64_t *txg)
{
scan_ds_t srch, *sds;
srch.sds_dsobj = dsobj;
sds = avl_find(&scn->scn_queue, &srch, NULL);
if (sds != NULL && txg != NULL)
*txg = sds->sds_txg;
return (sds != NULL);
}
static void
scan_ds_queue_insert(dsl_scan_t *scn, uint64_t dsobj, uint64_t txg)
{
scan_ds_t *sds;
avl_index_t where;
sds = kmem_zalloc(sizeof (*sds), KM_SLEEP);
sds->sds_dsobj = dsobj;
sds->sds_txg = txg;
VERIFY3P(avl_find(&scn->scn_queue, sds, &where), ==, NULL);
avl_insert(&scn->scn_queue, sds, where);
}
static void
scan_ds_queue_remove(dsl_scan_t *scn, uint64_t dsobj)
{
scan_ds_t srch, *sds;
srch.sds_dsobj = dsobj;
sds = avl_find(&scn->scn_queue, &srch, NULL);
VERIFY(sds != NULL);
avl_remove(&scn->scn_queue, sds);
kmem_free(sds, sizeof (*sds));
}
static void
scan_ds_queue_sync(dsl_scan_t *scn, dmu_tx_t *tx)
{
dsl_pool_t *dp = scn->scn_dp;
spa_t *spa = dp->dp_spa;
dmu_object_type_t ot = (spa_version(spa) >= SPA_VERSION_DSL_SCRUB) ?
DMU_OT_SCAN_QUEUE : DMU_OT_ZAP_OTHER;
ASSERT0(scn->scn_bytes_pending);
ASSERT(scn->scn_phys.scn_queue_obj != 0);
VERIFY0(dmu_object_free(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, tx));
scn->scn_phys.scn_queue_obj = zap_create(dp->dp_meta_objset, ot,
DMU_OT_NONE, 0, tx);
for (scan_ds_t *sds = avl_first(&scn->scn_queue);
sds != NULL; sds = AVL_NEXT(&scn->scn_queue, sds)) {
VERIFY0(zap_add_int_key(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, sds->sds_dsobj,
sds->sds_txg, tx));
}
}
/*
* Computes the memory limit state that we're currently in. A sorted scan
* needs quite a bit of memory to hold the sorting queue, so we need to
* reasonably constrain the size so it doesn't impact overall system
* performance. We compute two limits:
* 1) Hard memory limit: if the amount of memory used by the sorting
* queues on a pool gets above this value, we stop the metadata
* scanning portion and start issuing the queued up and sorted
* I/Os to reduce memory usage.
* This limit is calculated as a fraction of physmem (by default 5%).
* We constrain the lower bound of the hard limit to an absolute
* minimum of zfs_scan_mem_lim_min (default: 16 MiB). We also constrain
* the upper bound to 5% of the total pool size - no chance we'll
* ever need that much memory, but just to keep the value in check.
* 2) Soft memory limit: once we hit the hard memory limit, we start
* issuing I/O to reduce queue memory usage, but we don't want to
* completely empty out the queues, since we might be able to find I/Os
* that will fill in the gaps of our non-sequential IOs at some point
* in the future. So we stop the issuing of I/Os once the amount of
* memory used drops below the soft limit (at which point we stop issuing
* I/O and start scanning metadata again).
*
* This limit is calculated by subtracting a fraction of the hard
* limit from the hard limit. By default this fraction is 5%, so
* the soft limit is 95% of the hard limit. We cap the size of the
* difference between the hard and soft limits at an absolute
* maximum of zfs_scan_mem_lim_soft_max (default: 128 MiB) - this is
* sufficient to not cause too frequent switching between the
* metadata scan and I/O issue (even at 2k recordsize, 128 MiB's
* worth of queues is about 1.2 GiB of on-pool data, so scanning
* that should take at least a decent fraction of a second).
*/
static boolean_t
dsl_scan_should_clear(dsl_scan_t *scn)
{
spa_t *spa = scn->scn_dp->dp_spa;
vdev_t *rvd = scn->scn_dp->dp_spa->spa_root_vdev;
uint64_t alloc, mlim_hard, mlim_soft, mused;
alloc = metaslab_class_get_alloc(spa_normal_class(spa));
alloc += metaslab_class_get_alloc(spa_special_class(spa));
alloc += metaslab_class_get_alloc(spa_dedup_class(spa));
mlim_hard = MAX((physmem / zfs_scan_mem_lim_fact) * PAGESIZE,
zfs_scan_mem_lim_min);
mlim_hard = MIN(mlim_hard, alloc / 20);
mlim_soft = mlim_hard - MIN(mlim_hard / zfs_scan_mem_lim_soft_fact,
zfs_scan_mem_lim_soft_max);
mused = 0;
for (uint64_t i = 0; i < rvd->vdev_children; i++) {
vdev_t *tvd = rvd->vdev_child[i];
dsl_scan_io_queue_t *queue;
mutex_enter(&tvd->vdev_scan_io_queue_lock);
queue = tvd->vdev_scan_io_queue;
if (queue != NULL) {
/* # extents in exts_by_size = # in exts_by_addr */
mused += zfs_btree_numnodes(&queue->q_exts_by_size) *
sizeof (range_seg_gap_t) + queue->q_sio_memused;
}
mutex_exit(&tvd->vdev_scan_io_queue_lock);
}
dprintf("current scan memory usage: %llu bytes\n", (longlong_t)mused);
if (mused == 0)
ASSERT0(scn->scn_bytes_pending);
/*
* If we are above our hard limit, we need to clear out memory.
* If we are below our soft limit, we need to accumulate sequential IOs.
* Otherwise, we should keep doing whatever we are currently doing.
*/
if (mused >= mlim_hard)
return (B_TRUE);
else if (mused < mlim_soft)
return (B_FALSE);
else
return (scn->scn_clearing);
}
static boolean_t
dsl_scan_check_suspend(dsl_scan_t *scn, const zbookmark_phys_t *zb)
{
/* we never skip user/group accounting objects */
if (zb && (int64_t)zb->zb_object < 0)
return (B_FALSE);
if (scn->scn_suspending)
return (B_TRUE); /* we're already suspending */
if (!ZB_IS_ZERO(&scn->scn_phys.scn_bookmark))
return (B_FALSE); /* we're resuming */
/* We only know how to resume from level-0 and objset blocks. */
if (zb && (zb->zb_level != 0 && zb->zb_level != ZB_ROOT_LEVEL))
return (B_FALSE);
/*
* We suspend if:
* - we have scanned for at least the minimum time (default 1 sec
* for scrub, 3 sec for resilver), and either we have sufficient
* dirty data that we are starting to write more quickly
* (default 30%), someone is explicitly waiting for this txg
* to complete, or we have used up all of the time in the txg
* timeout (default 5 sec).
* or
* - the spa is shutting down because this pool is being exported
* or the machine is rebooting.
* or
* - the scan queue has reached its memory use limit
*/
uint64_t curr_time_ns = gethrtime();
uint64_t scan_time_ns = curr_time_ns - scn->scn_sync_start_time;
uint64_t sync_time_ns = curr_time_ns -
scn->scn_dp->dp_spa->spa_sync_starttime;
int dirty_pct = scn->scn_dp->dp_dirty_total * 100 / zfs_dirty_data_max;
int mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
zfs_resilver_min_time_ms : zfs_scrub_min_time_ms;
if ((NSEC2MSEC(scan_time_ns) > mintime &&
(dirty_pct >= zfs_vdev_async_write_active_min_dirty_percent ||
txg_sync_waiting(scn->scn_dp) ||
NSEC2SEC(sync_time_ns) >= zfs_txg_timeout)) ||
spa_shutting_down(scn->scn_dp->dp_spa) ||
(zfs_scan_strict_mem_lim && dsl_scan_should_clear(scn))) {
if (zb && zb->zb_level == ZB_ROOT_LEVEL) {
dprintf("suspending at first available bookmark "
"%llx/%llx/%llx/%llx\n",
(longlong_t)zb->zb_objset,
(longlong_t)zb->zb_object,
(longlong_t)zb->zb_level,
(longlong_t)zb->zb_blkid);
SET_BOOKMARK(&scn->scn_phys.scn_bookmark,
zb->zb_objset, 0, 0, 0);
} else if (zb != NULL) {
dprintf("suspending at bookmark %llx/%llx/%llx/%llx\n",
(longlong_t)zb->zb_objset,
(longlong_t)zb->zb_object,
(longlong_t)zb->zb_level,
(longlong_t)zb->zb_blkid);
scn->scn_phys.scn_bookmark = *zb;
} else {
#ifdef ZFS_DEBUG
dsl_scan_phys_t *scnp = &scn->scn_phys;
dprintf("suspending at at DDT bookmark "
"%llx/%llx/%llx/%llx\n",
(longlong_t)scnp->scn_ddt_bookmark.ddb_class,
(longlong_t)scnp->scn_ddt_bookmark.ddb_type,
(longlong_t)scnp->scn_ddt_bookmark.ddb_checksum,
(longlong_t)scnp->scn_ddt_bookmark.ddb_cursor);
#endif
}
scn->scn_suspending = B_TRUE;
return (B_TRUE);
}
return (B_FALSE);
}
typedef struct zil_scan_arg {
dsl_pool_t *zsa_dp;
zil_header_t *zsa_zh;
} zil_scan_arg_t;
static int
dsl_scan_zil_block(zilog_t *zilog, const blkptr_t *bp, void *arg,
uint64_t claim_txg)
{
(void) zilog;
zil_scan_arg_t *zsa = arg;
dsl_pool_t *dp = zsa->zsa_dp;
dsl_scan_t *scn = dp->dp_scan;
zil_header_t *zh = zsa->zsa_zh;
zbookmark_phys_t zb;
ASSERT(!BP_IS_REDACTED(bp));
if (BP_IS_HOLE(bp) || bp->blk_birth <= scn->scn_phys.scn_cur_min_txg)
return (0);
/*
* One block ("stubby") can be allocated a long time ago; we
* want to visit that one because it has been allocated
* (on-disk) even if it hasn't been claimed (even though for
* scrub there's nothing to do to it).
*/
if (claim_txg == 0 && bp->blk_birth >= spa_min_claim_txg(dp->dp_spa))
return (0);
SET_BOOKMARK(&zb, zh->zh_log.blk_cksum.zc_word[ZIL_ZC_OBJSET],
ZB_ZIL_OBJECT, ZB_ZIL_LEVEL, bp->blk_cksum.zc_word[ZIL_ZC_SEQ]);
VERIFY(0 == scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb));
return (0);
}
static int
dsl_scan_zil_record(zilog_t *zilog, const lr_t *lrc, void *arg,
uint64_t claim_txg)
{
(void) zilog;
if (lrc->lrc_txtype == TX_WRITE) {
zil_scan_arg_t *zsa = arg;
dsl_pool_t *dp = zsa->zsa_dp;
dsl_scan_t *scn = dp->dp_scan;
zil_header_t *zh = zsa->zsa_zh;
const lr_write_t *lr = (const lr_write_t *)lrc;
const blkptr_t *bp = &lr->lr_blkptr;
zbookmark_phys_t zb;
ASSERT(!BP_IS_REDACTED(bp));
if (BP_IS_HOLE(bp) ||
bp->blk_birth <= scn->scn_phys.scn_cur_min_txg)
return (0);
/*
* birth can be < claim_txg if this record's txg is
* already txg sync'ed (but this log block contains
* other records that are not synced)
*/
if (claim_txg == 0 || bp->blk_birth < claim_txg)
return (0);
SET_BOOKMARK(&zb, zh->zh_log.blk_cksum.zc_word[ZIL_ZC_OBJSET],
lr->lr_foid, ZB_ZIL_LEVEL,
lr->lr_offset / BP_GET_LSIZE(bp));
VERIFY(0 == scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb));
}
return (0);
}
static void
dsl_scan_zil(dsl_pool_t *dp, zil_header_t *zh)
{
uint64_t claim_txg = zh->zh_claim_txg;
zil_scan_arg_t zsa = { dp, zh };
zilog_t *zilog;
ASSERT(spa_writeable(dp->dp_spa));
/*
* We only want to visit blocks that have been claimed but not yet
* replayed (or, in read-only mode, blocks that *would* be claimed).
*/
if (claim_txg == 0)
return;
zilog = zil_alloc(dp->dp_meta_objset, zh);
(void) zil_parse(zilog, dsl_scan_zil_block, dsl_scan_zil_record, &zsa,
claim_txg, B_FALSE);
zil_free(zilog);
}
/*
* We compare scan_prefetch_issue_ctx_t's based on their bookmarks. The idea
* here is to sort the AVL tree by the order each block will be needed.
*/
static int
scan_prefetch_queue_compare(const void *a, const void *b)
{
const scan_prefetch_issue_ctx_t *spic_a = a, *spic_b = b;
const scan_prefetch_ctx_t *spc_a = spic_a->spic_spc;
const scan_prefetch_ctx_t *spc_b = spic_b->spic_spc;
return (zbookmark_compare(spc_a->spc_datablkszsec,
spc_a->spc_indblkshift, spc_b->spc_datablkszsec,
spc_b->spc_indblkshift, &spic_a->spic_zb, &spic_b->spic_zb));
}
static void
scan_prefetch_ctx_rele(scan_prefetch_ctx_t *spc, void *tag)
{
if (zfs_refcount_remove(&spc->spc_refcnt, tag) == 0) {
zfs_refcount_destroy(&spc->spc_refcnt);
kmem_free(spc, sizeof (scan_prefetch_ctx_t));
}
}
static scan_prefetch_ctx_t *
scan_prefetch_ctx_create(dsl_scan_t *scn, dnode_phys_t *dnp, void *tag)
{
scan_prefetch_ctx_t *spc;
spc = kmem_alloc(sizeof (scan_prefetch_ctx_t), KM_SLEEP);
zfs_refcount_create(&spc->spc_refcnt);
zfs_refcount_add(&spc->spc_refcnt, tag);
spc->spc_scn = scn;
if (dnp != NULL) {
spc->spc_datablkszsec = dnp->dn_datablkszsec;
spc->spc_indblkshift = dnp->dn_indblkshift;
spc->spc_root = B_FALSE;
} else {
spc->spc_datablkszsec = 0;
spc->spc_indblkshift = 0;
spc->spc_root = B_TRUE;
}
return (spc);
}
static void
scan_prefetch_ctx_add_ref(scan_prefetch_ctx_t *spc, void *tag)
{
zfs_refcount_add(&spc->spc_refcnt, tag);
}
static void
scan_ds_prefetch_queue_clear(dsl_scan_t *scn)
{
spa_t *spa = scn->scn_dp->dp_spa;
void *cookie = NULL;
scan_prefetch_issue_ctx_t *spic = NULL;
mutex_enter(&spa->spa_scrub_lock);
while ((spic = avl_destroy_nodes(&scn->scn_prefetch_queue,
&cookie)) != NULL) {
scan_prefetch_ctx_rele(spic->spic_spc, scn);
kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
}
mutex_exit(&spa->spa_scrub_lock);
}
static boolean_t
dsl_scan_check_prefetch_resume(scan_prefetch_ctx_t *spc,
const zbookmark_phys_t *zb)
{
zbookmark_phys_t *last_zb = &spc->spc_scn->scn_prefetch_bookmark;
dnode_phys_t tmp_dnp;
dnode_phys_t *dnp = (spc->spc_root) ? NULL : &tmp_dnp;
if (zb->zb_objset != last_zb->zb_objset)
return (B_TRUE);
if ((int64_t)zb->zb_object < 0)
return (B_FALSE);
tmp_dnp.dn_datablkszsec = spc->spc_datablkszsec;
tmp_dnp.dn_indblkshift = spc->spc_indblkshift;
if (zbookmark_subtree_completed(dnp, zb, last_zb))
return (B_TRUE);
return (B_FALSE);
}
static void
dsl_scan_prefetch(scan_prefetch_ctx_t *spc, blkptr_t *bp, zbookmark_phys_t *zb)
{
avl_index_t idx;
dsl_scan_t *scn = spc->spc_scn;
spa_t *spa = scn->scn_dp->dp_spa;
scan_prefetch_issue_ctx_t *spic;
if (zfs_no_scrub_prefetch || BP_IS_REDACTED(bp))
return;
if (BP_IS_HOLE(bp) || bp->blk_birth <= scn->scn_phys.scn_cur_min_txg ||
(BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_DNODE &&
BP_GET_TYPE(bp) != DMU_OT_OBJSET))
return;
if (dsl_scan_check_prefetch_resume(spc, zb))
return;
scan_prefetch_ctx_add_ref(spc, scn);
spic = kmem_alloc(sizeof (scan_prefetch_issue_ctx_t), KM_SLEEP);
spic->spic_spc = spc;
spic->spic_bp = *bp;
spic->spic_zb = *zb;
/*
* Add the IO to the queue of blocks to prefetch. This allows us to
* prioritize blocks that we will need first for the main traversal
* thread.
*/
mutex_enter(&spa->spa_scrub_lock);
if (avl_find(&scn->scn_prefetch_queue, spic, &idx) != NULL) {
/* this block is already queued for prefetch */
kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
scan_prefetch_ctx_rele(spc, scn);
mutex_exit(&spa->spa_scrub_lock);
return;
}
avl_insert(&scn->scn_prefetch_queue, spic, idx);
cv_broadcast(&spa->spa_scrub_io_cv);
mutex_exit(&spa->spa_scrub_lock);
}
static void
dsl_scan_prefetch_dnode(dsl_scan_t *scn, dnode_phys_t *dnp,
uint64_t objset, uint64_t object)
{
int i;
zbookmark_phys_t zb;
scan_prefetch_ctx_t *spc;
if (dnp->dn_nblkptr == 0 && !(dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR))
return;
SET_BOOKMARK(&zb, objset, object, 0, 0);
spc = scan_prefetch_ctx_create(scn, dnp, FTAG);
for (i = 0; i < dnp->dn_nblkptr; i++) {
zb.zb_level = BP_GET_LEVEL(&dnp->dn_blkptr[i]);
zb.zb_blkid = i;
dsl_scan_prefetch(spc, &dnp->dn_blkptr[i], &zb);
}
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
zb.zb_level = 0;
zb.zb_blkid = DMU_SPILL_BLKID;
dsl_scan_prefetch(spc, DN_SPILL_BLKPTR(dnp), &zb);
}
scan_prefetch_ctx_rele(spc, FTAG);
}
static void
dsl_scan_prefetch_cb(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp,
arc_buf_t *buf, void *private)
{
(void) zio;
scan_prefetch_ctx_t *spc = private;
dsl_scan_t *scn = spc->spc_scn;
spa_t *spa = scn->scn_dp->dp_spa;
/* broadcast that the IO has completed for rate limiting purposes */
mutex_enter(&spa->spa_scrub_lock);
ASSERT3U(spa->spa_scrub_inflight, >=, BP_GET_PSIZE(bp));
spa->spa_scrub_inflight -= BP_GET_PSIZE(bp);
cv_broadcast(&spa->spa_scrub_io_cv);
mutex_exit(&spa->spa_scrub_lock);
/* if there was an error or we are done prefetching, just cleanup */
if (buf == NULL || scn->scn_prefetch_stop)
goto out;
if (BP_GET_LEVEL(bp) > 0) {
int i;
blkptr_t *cbp;
int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
zbookmark_phys_t czb;
for (i = 0, cbp = buf->b_data; i < epb; i++, cbp++) {
SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
zb->zb_level - 1, zb->zb_blkid * epb + i);
dsl_scan_prefetch(spc, cbp, &czb);
}
} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
dnode_phys_t *cdnp;
int i;
int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
for (i = 0, cdnp = buf->b_data; i < epb;
i += cdnp->dn_extra_slots + 1,
cdnp += cdnp->dn_extra_slots + 1) {
dsl_scan_prefetch_dnode(scn, cdnp,
zb->zb_objset, zb->zb_blkid * epb + i);
}
} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
objset_phys_t *osp = buf->b_data;
dsl_scan_prefetch_dnode(scn, &osp->os_meta_dnode,
zb->zb_objset, DMU_META_DNODE_OBJECT);
if (OBJSET_BUF_HAS_USERUSED(buf)) {
dsl_scan_prefetch_dnode(scn,
&osp->os_groupused_dnode, zb->zb_objset,
DMU_GROUPUSED_OBJECT);
dsl_scan_prefetch_dnode(scn,
&osp->os_userused_dnode, zb->zb_objset,
DMU_USERUSED_OBJECT);
}
}
out:
if (buf != NULL)
arc_buf_destroy(buf, private);
scan_prefetch_ctx_rele(spc, scn);
}
static void
dsl_scan_prefetch_thread(void *arg)
{
dsl_scan_t *scn = arg;
spa_t *spa = scn->scn_dp->dp_spa;
scan_prefetch_issue_ctx_t *spic;
/* loop until we are told to stop */
while (!scn->scn_prefetch_stop) {
arc_flags_t flags = ARC_FLAG_NOWAIT |
ARC_FLAG_PRESCIENT_PREFETCH | ARC_FLAG_PREFETCH;
int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SCAN_THREAD;
mutex_enter(&spa->spa_scrub_lock);
/*
* Wait until we have an IO to issue and are not above our
* maximum in flight limit.
*/
while (!scn->scn_prefetch_stop &&
(avl_numnodes(&scn->scn_prefetch_queue) == 0 ||
spa->spa_scrub_inflight >= scn->scn_maxinflight_bytes)) {
cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
}
/* recheck if we should stop since we waited for the cv */
if (scn->scn_prefetch_stop) {
mutex_exit(&spa->spa_scrub_lock);
break;
}
/* remove the prefetch IO from the tree */
spic = avl_first(&scn->scn_prefetch_queue);
spa->spa_scrub_inflight += BP_GET_PSIZE(&spic->spic_bp);
avl_remove(&scn->scn_prefetch_queue, spic);
mutex_exit(&spa->spa_scrub_lock);
if (BP_IS_PROTECTED(&spic->spic_bp)) {
ASSERT(BP_GET_TYPE(&spic->spic_bp) == DMU_OT_DNODE ||
BP_GET_TYPE(&spic->spic_bp) == DMU_OT_OBJSET);
ASSERT3U(BP_GET_LEVEL(&spic->spic_bp), ==, 0);
zio_flags |= ZIO_FLAG_RAW;
}
/* issue the prefetch asynchronously */
(void) arc_read(scn->scn_zio_root, scn->scn_dp->dp_spa,
&spic->spic_bp, dsl_scan_prefetch_cb, spic->spic_spc,
ZIO_PRIORITY_SCRUB, zio_flags, &flags, &spic->spic_zb);
kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
}
ASSERT(scn->scn_prefetch_stop);
/* free any prefetches we didn't get to complete */
mutex_enter(&spa->spa_scrub_lock);
while ((spic = avl_first(&scn->scn_prefetch_queue)) != NULL) {
avl_remove(&scn->scn_prefetch_queue, spic);
scan_prefetch_ctx_rele(spic->spic_spc, scn);
kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
}
ASSERT0(avl_numnodes(&scn->scn_prefetch_queue));
mutex_exit(&spa->spa_scrub_lock);
}
static boolean_t
dsl_scan_check_resume(dsl_scan_t *scn, const dnode_phys_t *dnp,
const zbookmark_phys_t *zb)
{
/*
* We never skip over user/group accounting objects (obj<0)
*/
if (!ZB_IS_ZERO(&scn->scn_phys.scn_bookmark) &&
(int64_t)zb->zb_object >= 0) {
/*
* If we already visited this bp & everything below (in
* a prior txg sync), don't bother doing it again.
*/
if (zbookmark_subtree_completed(dnp, zb,
&scn->scn_phys.scn_bookmark))
return (B_TRUE);
/*
* If we found the block we're trying to resume from, or
* we went past it to a different object, zero it out to
* indicate that it's OK to start checking for suspending
* again.
*/
if (bcmp(zb, &scn->scn_phys.scn_bookmark, sizeof (*zb)) == 0 ||
zb->zb_object > scn->scn_phys.scn_bookmark.zb_object) {
dprintf("resuming at %llx/%llx/%llx/%llx\n",
(longlong_t)zb->zb_objset,
(longlong_t)zb->zb_object,
(longlong_t)zb->zb_level,
(longlong_t)zb->zb_blkid);
bzero(&scn->scn_phys.scn_bookmark, sizeof (*zb));
}
}
return (B_FALSE);
}
static void dsl_scan_visitbp(blkptr_t *bp, const zbookmark_phys_t *zb,
dnode_phys_t *dnp, dsl_dataset_t *ds, dsl_scan_t *scn,
dmu_objset_type_t ostype, dmu_tx_t *tx);
inline __attribute__((always_inline)) static void dsl_scan_visitdnode(
dsl_scan_t *, dsl_dataset_t *ds, dmu_objset_type_t ostype,
dnode_phys_t *dnp, uint64_t object, dmu_tx_t *tx);
/*
* Return nonzero on i/o error.
* Return new buf to write out in *bufp.
*/
inline __attribute__((always_inline)) static int
dsl_scan_recurse(dsl_scan_t *scn, dsl_dataset_t *ds, dmu_objset_type_t ostype,
dnode_phys_t *dnp, const blkptr_t *bp,
const zbookmark_phys_t *zb, dmu_tx_t *tx)
{
dsl_pool_t *dp = scn->scn_dp;
int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SCAN_THREAD;
int err;
ASSERT(!BP_IS_REDACTED(bp));
if (BP_GET_LEVEL(bp) > 0) {
arc_flags_t flags = ARC_FLAG_WAIT;
int i;
blkptr_t *cbp;
int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
arc_buf_t *buf;
err = arc_read(NULL, dp->dp_spa, bp, arc_getbuf_func, &buf,
ZIO_PRIORITY_SCRUB, zio_flags, &flags, zb);
if (err) {
scn->scn_phys.scn_errors++;
return (err);
}
for (i = 0, cbp = buf->b_data; i < epb; i++, cbp++) {
zbookmark_phys_t czb;
SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
zb->zb_level - 1,
zb->zb_blkid * epb + i);
dsl_scan_visitbp(cbp, &czb, dnp,
ds, scn, ostype, tx);
}
arc_buf_destroy(buf, &buf);
} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
arc_flags_t flags = ARC_FLAG_WAIT;
dnode_phys_t *cdnp;
int i;
int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
arc_buf_t *buf;
if (BP_IS_PROTECTED(bp)) {
ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
zio_flags |= ZIO_FLAG_RAW;
}
err = arc_read(NULL, dp->dp_spa, bp, arc_getbuf_func, &buf,
ZIO_PRIORITY_SCRUB, zio_flags, &flags, zb);
if (err) {
scn->scn_phys.scn_errors++;
return (err);
}
for (i = 0, cdnp = buf->b_data; i < epb;
i += cdnp->dn_extra_slots + 1,
cdnp += cdnp->dn_extra_slots + 1) {
dsl_scan_visitdnode(scn, ds, ostype,
cdnp, zb->zb_blkid * epb + i, tx);
}
arc_buf_destroy(buf, &buf);
} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
arc_flags_t flags = ARC_FLAG_WAIT;
objset_phys_t *osp;
arc_buf_t *buf;
err = arc_read(NULL, dp->dp_spa, bp, arc_getbuf_func, &buf,
ZIO_PRIORITY_SCRUB, zio_flags, &flags, zb);
if (err) {
scn->scn_phys.scn_errors++;
return (err);
}
osp = buf->b_data;
dsl_scan_visitdnode(scn, ds, osp->os_type,
&osp->os_meta_dnode, DMU_META_DNODE_OBJECT, tx);
if (OBJSET_BUF_HAS_USERUSED(buf)) {
/*
* We also always visit user/group/project accounting
* objects, and never skip them, even if we are
* suspending. This is necessary so that the
* space deltas from this txg get integrated.
*/
if (OBJSET_BUF_HAS_PROJECTUSED(buf))
dsl_scan_visitdnode(scn, ds, osp->os_type,
&osp->os_projectused_dnode,
DMU_PROJECTUSED_OBJECT, tx);
dsl_scan_visitdnode(scn, ds, osp->os_type,
&osp->os_groupused_dnode,
DMU_GROUPUSED_OBJECT, tx);
dsl_scan_visitdnode(scn, ds, osp->os_type,
&osp->os_userused_dnode,
DMU_USERUSED_OBJECT, tx);
}
arc_buf_destroy(buf, &buf);
}
return (0);
}
inline __attribute__((always_inline)) static void
dsl_scan_visitdnode(dsl_scan_t *scn, dsl_dataset_t *ds,
dmu_objset_type_t ostype, dnode_phys_t *dnp,
uint64_t object, dmu_tx_t *tx)
{
int j;
for (j = 0; j < dnp->dn_nblkptr; j++) {
zbookmark_phys_t czb;
SET_BOOKMARK(&czb, ds ? ds->ds_object : 0, object,
dnp->dn_nlevels - 1, j);
dsl_scan_visitbp(&dnp->dn_blkptr[j],
&czb, dnp, ds, scn, ostype, tx);
}
if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
zbookmark_phys_t czb;
SET_BOOKMARK(&czb, ds ? ds->ds_object : 0, object,
0, DMU_SPILL_BLKID);
dsl_scan_visitbp(DN_SPILL_BLKPTR(dnp),
&czb, dnp, ds, scn, ostype, tx);
}
}
/*
* The arguments are in this order because mdb can only print the
* first 5; we want them to be useful.
*/
static void
dsl_scan_visitbp(blkptr_t *bp, const zbookmark_phys_t *zb,
dnode_phys_t *dnp, dsl_dataset_t *ds, dsl_scan_t *scn,
dmu_objset_type_t ostype, dmu_tx_t *tx)
{
dsl_pool_t *dp = scn->scn_dp;
blkptr_t *bp_toread = NULL;
if (dsl_scan_check_suspend(scn, zb))
return;
if (dsl_scan_check_resume(scn, dnp, zb))
return;
scn->scn_visited_this_txg++;
/*
* This debugging is commented out to conserve stack space. This
* function is called recursively and the debugging adds several
* bytes to the stack for each call. It can be commented back in
* if required to debug an issue in dsl_scan_visitbp().
*
* dprintf_bp(bp,
* "visiting ds=%p/%llu zb=%llx/%llx/%llx/%llx bp=%p",
* ds, ds ? ds->ds_object : 0,
* zb->zb_objset, zb->zb_object, zb->zb_level, zb->zb_blkid,
* bp);
*/
if (BP_IS_HOLE(bp)) {
scn->scn_holes_this_txg++;
return;
}
if (BP_IS_REDACTED(bp)) {
ASSERT(dsl_dataset_feature_is_active(ds,
SPA_FEATURE_REDACTED_DATASETS));
return;
}
if (bp->blk_birth <= scn->scn_phys.scn_cur_min_txg) {
scn->scn_lt_min_this_txg++;
return;
}
bp_toread = kmem_alloc(sizeof (blkptr_t), KM_SLEEP);
*bp_toread = *bp;
if (dsl_scan_recurse(scn, ds, ostype, dnp, bp_toread, zb, tx) != 0)
goto out;
/*
* If dsl_scan_ddt() has already visited this block, it will have
* already done any translations or scrubbing, so don't call the
* callback again.
*/
if (ddt_class_contains(dp->dp_spa,
scn->scn_phys.scn_ddt_class_max, bp)) {
scn->scn_ddt_contained_this_txg++;
goto out;
}
/*
* If this block is from the future (after cur_max_txg), then we
* are doing this on behalf of a deleted snapshot, and we will
* revisit the future block on the next pass of this dataset.
* Don't scan it now unless we need to because something
* under it was modified.
*/
if (BP_PHYSICAL_BIRTH(bp) > scn->scn_phys.scn_cur_max_txg) {
scn->scn_gt_max_this_txg++;
goto out;
}
scan_funcs[scn->scn_phys.scn_func](dp, bp, zb);
out:
kmem_free(bp_toread, sizeof (blkptr_t));
}
static void
dsl_scan_visit_rootbp(dsl_scan_t *scn, dsl_dataset_t *ds, blkptr_t *bp,
dmu_tx_t *tx)
{
zbookmark_phys_t zb;
scan_prefetch_ctx_t *spc;
SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET,
ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
if (ZB_IS_ZERO(&scn->scn_phys.scn_bookmark)) {
SET_BOOKMARK(&scn->scn_prefetch_bookmark,
zb.zb_objset, 0, 0, 0);
} else {
scn->scn_prefetch_bookmark = scn->scn_phys.scn_bookmark;
}
scn->scn_objsets_visited_this_txg++;
spc = scan_prefetch_ctx_create(scn, NULL, FTAG);
dsl_scan_prefetch(spc, bp, &zb);
scan_prefetch_ctx_rele(spc, FTAG);
dsl_scan_visitbp(bp, &zb, NULL, ds, scn, DMU_OST_NONE, tx);
dprintf_ds(ds, "finished scan%s", "");
}
static void
ds_destroyed_scn_phys(dsl_dataset_t *ds, dsl_scan_phys_t *scn_phys)
{
if (scn_phys->scn_bookmark.zb_objset == ds->ds_object) {
if (ds->ds_is_snapshot) {
/*
* Note:
* - scn_cur_{min,max}_txg stays the same.
* - Setting the flag is not really necessary if
* scn_cur_max_txg == scn_max_txg, because there
* is nothing after this snapshot that we care
* about. However, we set it anyway and then
* ignore it when we retraverse it in
* dsl_scan_visitds().
*/
scn_phys->scn_bookmark.zb_objset =
dsl_dataset_phys(ds)->ds_next_snap_obj;
zfs_dbgmsg("destroying ds %llu on %s; currently "
"traversing; reset zb_objset to %llu",
(u_longlong_t)ds->ds_object,
ds->ds_dir->dd_pool->dp_spa->spa_name,
(u_longlong_t)dsl_dataset_phys(ds)->
ds_next_snap_obj);
scn_phys->scn_flags |= DSF_VISIT_DS_AGAIN;
} else {
SET_BOOKMARK(&scn_phys->scn_bookmark,
ZB_DESTROYED_OBJSET, 0, 0, 0);
zfs_dbgmsg("destroying ds %llu on %s; currently "
"traversing; reset bookmark to -1,0,0,0",
(u_longlong_t)ds->ds_object,
ds->ds_dir->dd_pool->dp_spa->spa_name);
}
}
}
/*
* Invoked when a dataset is destroyed. We need to make sure that:
*
* 1) If it is the dataset that was currently being scanned, we write
* a new dsl_scan_phys_t and marking the objset reference in it
* as destroyed.
* 2) Remove it from the work queue, if it was present.
*
* If the dataset was actually a snapshot, instead of marking the dataset
* as destroyed, we instead substitute the next snapshot in line.
*/
void
dsl_scan_ds_destroyed(dsl_dataset_t *ds, dmu_tx_t *tx)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
dsl_scan_t *scn = dp->dp_scan;
uint64_t mintxg;
if (!dsl_scan_is_running(scn))
return;
ds_destroyed_scn_phys(ds, &scn->scn_phys);
ds_destroyed_scn_phys(ds, &scn->scn_phys_cached);
if (scan_ds_queue_contains(scn, ds->ds_object, &mintxg)) {
scan_ds_queue_remove(scn, ds->ds_object);
if (ds->ds_is_snapshot)
scan_ds_queue_insert(scn,
dsl_dataset_phys(ds)->ds_next_snap_obj, mintxg);
}
if (zap_lookup_int_key(dp->dp_meta_objset, scn->scn_phys.scn_queue_obj,
ds->ds_object, &mintxg) == 0) {
ASSERT3U(dsl_dataset_phys(ds)->ds_num_children, <=, 1);
VERIFY3U(0, ==, zap_remove_int(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, ds->ds_object, tx));
if (ds->ds_is_snapshot) {
/*
* We keep the same mintxg; it could be >
* ds_creation_txg if the previous snapshot was
* deleted too.
*/
VERIFY(zap_add_int_key(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj,
dsl_dataset_phys(ds)->ds_next_snap_obj,
mintxg, tx) == 0);
zfs_dbgmsg("destroying ds %llu on %s; in queue; "
"replacing with %llu",
(u_longlong_t)ds->ds_object,
dp->dp_spa->spa_name,
(u_longlong_t)dsl_dataset_phys(ds)->
ds_next_snap_obj);
} else {
zfs_dbgmsg("destroying ds %llu on %s; in queue; "
"removing",
(u_longlong_t)ds->ds_object,
dp->dp_spa->spa_name);
}
}
/*
* dsl_scan_sync() should be called after this, and should sync
* out our changed state, but just to be safe, do it here.
*/
dsl_scan_sync_state(scn, tx, SYNC_CACHED);
}
static void
ds_snapshotted_bookmark(dsl_dataset_t *ds, zbookmark_phys_t *scn_bookmark)
{
if (scn_bookmark->zb_objset == ds->ds_object) {
scn_bookmark->zb_objset =
dsl_dataset_phys(ds)->ds_prev_snap_obj;
zfs_dbgmsg("snapshotting ds %llu on %s; currently traversing; "
"reset zb_objset to %llu",
(u_longlong_t)ds->ds_object,
ds->ds_dir->dd_pool->dp_spa->spa_name,
(u_longlong_t)dsl_dataset_phys(ds)->ds_prev_snap_obj);
}
}
/*
* Called when a dataset is snapshotted. If we were currently traversing
* this snapshot, we reset our bookmark to point at the newly created
* snapshot. We also modify our work queue to remove the old snapshot and
* replace with the new one.
*/
void
dsl_scan_ds_snapshotted(dsl_dataset_t *ds, dmu_tx_t *tx)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
dsl_scan_t *scn = dp->dp_scan;
uint64_t mintxg;
if (!dsl_scan_is_running(scn))
return;
ASSERT(dsl_dataset_phys(ds)->ds_prev_snap_obj != 0);
ds_snapshotted_bookmark(ds, &scn->scn_phys.scn_bookmark);
ds_snapshotted_bookmark(ds, &scn->scn_phys_cached.scn_bookmark);
if (scan_ds_queue_contains(scn, ds->ds_object, &mintxg)) {
scan_ds_queue_remove(scn, ds->ds_object);
scan_ds_queue_insert(scn,
dsl_dataset_phys(ds)->ds_prev_snap_obj, mintxg);
}
if (zap_lookup_int_key(dp->dp_meta_objset, scn->scn_phys.scn_queue_obj,
ds->ds_object, &mintxg) == 0) {
VERIFY3U(0, ==, zap_remove_int(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, ds->ds_object, tx));
VERIFY(zap_add_int_key(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj,
dsl_dataset_phys(ds)->ds_prev_snap_obj, mintxg, tx) == 0);
zfs_dbgmsg("snapshotting ds %llu on %s; in queue; "
"replacing with %llu",
(u_longlong_t)ds->ds_object,
dp->dp_spa->spa_name,
(u_longlong_t)dsl_dataset_phys(ds)->ds_prev_snap_obj);
}
dsl_scan_sync_state(scn, tx, SYNC_CACHED);
}
static void
ds_clone_swapped_bookmark(dsl_dataset_t *ds1, dsl_dataset_t *ds2,
zbookmark_phys_t *scn_bookmark)
{
if (scn_bookmark->zb_objset == ds1->ds_object) {
scn_bookmark->zb_objset = ds2->ds_object;
zfs_dbgmsg("clone_swap ds %llu on %s; currently traversing; "
"reset zb_objset to %llu",
(u_longlong_t)ds1->ds_object,
ds1->ds_dir->dd_pool->dp_spa->spa_name,
(u_longlong_t)ds2->ds_object);
} else if (scn_bookmark->zb_objset == ds2->ds_object) {
scn_bookmark->zb_objset = ds1->ds_object;
zfs_dbgmsg("clone_swap ds %llu on %s; currently traversing; "
"reset zb_objset to %llu",
(u_longlong_t)ds2->ds_object,
ds2->ds_dir->dd_pool->dp_spa->spa_name,
(u_longlong_t)ds1->ds_object);
}
}
/*
* Called when an origin dataset and its clone are swapped. If we were
* currently traversing the dataset, we need to switch to traversing the
* newly promoted clone.
*/
void
dsl_scan_ds_clone_swapped(dsl_dataset_t *ds1, dsl_dataset_t *ds2, dmu_tx_t *tx)
{
dsl_pool_t *dp = ds1->ds_dir->dd_pool;
dsl_scan_t *scn = dp->dp_scan;
uint64_t mintxg1, mintxg2;
boolean_t ds1_queued, ds2_queued;
if (!dsl_scan_is_running(scn))
return;
ds_clone_swapped_bookmark(ds1, ds2, &scn->scn_phys.scn_bookmark);
ds_clone_swapped_bookmark(ds1, ds2, &scn->scn_phys_cached.scn_bookmark);
/*
* Handle the in-memory scan queue.
*/
ds1_queued = scan_ds_queue_contains(scn, ds1->ds_object, &mintxg1);
ds2_queued = scan_ds_queue_contains(scn, ds2->ds_object, &mintxg2);
/* Sanity checking. */
if (ds1_queued) {
ASSERT3U(mintxg1, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
ASSERT3U(mintxg1, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
}
if (ds2_queued) {
ASSERT3U(mintxg2, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
ASSERT3U(mintxg2, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
}
if (ds1_queued && ds2_queued) {
/*
* If both are queued, we don't need to do anything.
* The swapping code below would not handle this case correctly,
* since we can't insert ds2 if it is already there. That's
* because scan_ds_queue_insert() prohibits a duplicate insert
* and panics.
*/
} else if (ds1_queued) {
scan_ds_queue_remove(scn, ds1->ds_object);
scan_ds_queue_insert(scn, ds2->ds_object, mintxg1);
} else if (ds2_queued) {
scan_ds_queue_remove(scn, ds2->ds_object);
scan_ds_queue_insert(scn, ds1->ds_object, mintxg2);
}
/*
* Handle the on-disk scan queue.
* The on-disk state is an out-of-date version of the in-memory state,
* so the in-memory and on-disk values for ds1_queued and ds2_queued may
* be different. Therefore we need to apply the swap logic to the
* on-disk state independently of the in-memory state.
*/
ds1_queued = zap_lookup_int_key(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, ds1->ds_object, &mintxg1) == 0;
ds2_queued = zap_lookup_int_key(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, ds2->ds_object, &mintxg2) == 0;
/* Sanity checking. */
if (ds1_queued) {
ASSERT3U(mintxg1, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
ASSERT3U(mintxg1, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
}
if (ds2_queued) {
ASSERT3U(mintxg2, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
ASSERT3U(mintxg2, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
}
if (ds1_queued && ds2_queued) {
/*
* If both are queued, we don't need to do anything.
* Alternatively, we could check for EEXIST from
* zap_add_int_key() and back out to the original state, but
* that would be more work than checking for this case upfront.
*/
} else if (ds1_queued) {
VERIFY3S(0, ==, zap_remove_int(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, ds1->ds_object, tx));
VERIFY3S(0, ==, zap_add_int_key(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, ds2->ds_object, mintxg1, tx));
zfs_dbgmsg("clone_swap ds %llu on %s; in queue; "
"replacing with %llu",
(u_longlong_t)ds1->ds_object,
dp->dp_spa->spa_name,
(u_longlong_t)ds2->ds_object);
} else if (ds2_queued) {
VERIFY3S(0, ==, zap_remove_int(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, ds2->ds_object, tx));
VERIFY3S(0, ==, zap_add_int_key(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, ds1->ds_object, mintxg2, tx));
zfs_dbgmsg("clone_swap ds %llu on %s; in queue; "
"replacing with %llu",
(u_longlong_t)ds2->ds_object,
dp->dp_spa->spa_name,
(u_longlong_t)ds1->ds_object);
}
dsl_scan_sync_state(scn, tx, SYNC_CACHED);
}
static int
enqueue_clones_cb(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg)
{
uint64_t originobj = *(uint64_t *)arg;
dsl_dataset_t *ds;
int err;
dsl_scan_t *scn = dp->dp_scan;
if (dsl_dir_phys(hds->ds_dir)->dd_origin_obj != originobj)
return (0);
err = dsl_dataset_hold_obj(dp, hds->ds_object, FTAG, &ds);
if (err)
return (err);
while (dsl_dataset_phys(ds)->ds_prev_snap_obj != originobj) {
dsl_dataset_t *prev;
err = dsl_dataset_hold_obj(dp,
dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);
dsl_dataset_rele(ds, FTAG);
if (err)
return (err);
ds = prev;
}
scan_ds_queue_insert(scn, ds->ds_object,
dsl_dataset_phys(ds)->ds_prev_snap_txg);
dsl_dataset_rele(ds, FTAG);
return (0);
}
static void
dsl_scan_visitds(dsl_scan_t *scn, uint64_t dsobj, dmu_tx_t *tx)
{
dsl_pool_t *dp = scn->scn_dp;
dsl_dataset_t *ds;
VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
if (scn->scn_phys.scn_cur_min_txg >=
scn->scn_phys.scn_max_txg) {
/*
* This can happen if this snapshot was created after the
* scan started, and we already completed a previous snapshot
* that was created after the scan started. This snapshot
* only references blocks with:
*
* birth < our ds_creation_txg
* cur_min_txg is no less than ds_creation_txg.
* We have already visited these blocks.
* or
* birth > scn_max_txg
* The scan requested not to visit these blocks.
*
* Subsequent snapshots (and clones) can reference our
* blocks, or blocks with even higher birth times.
* Therefore we do not need to visit them either,
* so we do not add them to the work queue.
*
* Note that checking for cur_min_txg >= cur_max_txg
* is not sufficient, because in that case we may need to
* visit subsequent snapshots. This happens when min_txg > 0,
* which raises cur_min_txg. In this case we will visit
* this dataset but skip all of its blocks, because the
* rootbp's birth time is < cur_min_txg. Then we will
* add the next snapshots/clones to the work queue.
*/
char *dsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
dsl_dataset_name(ds, dsname);
zfs_dbgmsg("scanning dataset %llu (%s) is unnecessary because "
"cur_min_txg (%llu) >= max_txg (%llu)",
(longlong_t)dsobj, dsname,
(longlong_t)scn->scn_phys.scn_cur_min_txg,
(longlong_t)scn->scn_phys.scn_max_txg);
kmem_free(dsname, MAXNAMELEN);
goto out;
}
/*
* Only the ZIL in the head (non-snapshot) is valid. Even though
* snapshots can have ZIL block pointers (which may be the same
* BP as in the head), they must be ignored. In addition, $ORIGIN
* doesn't have a objset (i.e. its ds_bp is a hole) so we don't
* need to look for a ZIL in it either. So we traverse the ZIL here,
* rather than in scan_recurse(), because the regular snapshot
* block-sharing rules don't apply to it.
*/
if (!dsl_dataset_is_snapshot(ds) &&
(dp->dp_origin_snap == NULL ||
ds->ds_dir != dp->dp_origin_snap->ds_dir)) {
objset_t *os;
if (dmu_objset_from_ds(ds, &os) != 0) {
goto out;
}
dsl_scan_zil(dp, &os->os_zil_header);
}
/*
* Iterate over the bps in this ds.
*/
dmu_buf_will_dirty(ds->ds_dbuf, tx);
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
dsl_scan_visit_rootbp(scn, ds, &dsl_dataset_phys(ds)->ds_bp, tx);
rrw_exit(&ds->ds_bp_rwlock, FTAG);
char *dsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
dsl_dataset_name(ds, dsname);
zfs_dbgmsg("scanned dataset %llu (%s) with min=%llu max=%llu; "
"suspending=%u",
(longlong_t)dsobj, dsname,
(longlong_t)scn->scn_phys.scn_cur_min_txg,
(longlong_t)scn->scn_phys.scn_cur_max_txg,
(int)scn->scn_suspending);
kmem_free(dsname, ZFS_MAX_DATASET_NAME_LEN);
if (scn->scn_suspending)
goto out;
/*
* We've finished this pass over this dataset.
*/
/*
* If we did not completely visit this dataset, do another pass.
*/
if (scn->scn_phys.scn_flags & DSF_VISIT_DS_AGAIN) {
zfs_dbgmsg("incomplete pass on %s; visiting again",
dp->dp_spa->spa_name);
scn->scn_phys.scn_flags &= ~DSF_VISIT_DS_AGAIN;
scan_ds_queue_insert(scn, ds->ds_object,
scn->scn_phys.scn_cur_max_txg);
goto out;
}
/*
* Add descendant datasets to work queue.
*/
if (dsl_dataset_phys(ds)->ds_next_snap_obj != 0) {
scan_ds_queue_insert(scn,
dsl_dataset_phys(ds)->ds_next_snap_obj,
dsl_dataset_phys(ds)->ds_creation_txg);
}
if (dsl_dataset_phys(ds)->ds_num_children > 1) {
boolean_t usenext = B_FALSE;
if (dsl_dataset_phys(ds)->ds_next_clones_obj != 0) {
uint64_t count;
/*
* A bug in a previous version of the code could
* cause upgrade_clones_cb() to not set
* ds_next_snap_obj when it should, leading to a
* missing entry. Therefore we can only use the
* next_clones_obj when its count is correct.
*/
int err = zap_count(dp->dp_meta_objset,
dsl_dataset_phys(ds)->ds_next_clones_obj, &count);
if (err == 0 &&
count == dsl_dataset_phys(ds)->ds_num_children - 1)
usenext = B_TRUE;
}
if (usenext) {
zap_cursor_t zc;
zap_attribute_t za;
for (zap_cursor_init(&zc, dp->dp_meta_objset,
dsl_dataset_phys(ds)->ds_next_clones_obj);
zap_cursor_retrieve(&zc, &za) == 0;
(void) zap_cursor_advance(&zc)) {
scan_ds_queue_insert(scn,
zfs_strtonum(za.za_name, NULL),
dsl_dataset_phys(ds)->ds_creation_txg);
}
zap_cursor_fini(&zc);
} else {
VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
enqueue_clones_cb, &ds->ds_object,
DS_FIND_CHILDREN));
}
}
out:
dsl_dataset_rele(ds, FTAG);
}
static int
enqueue_cb(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg)
{
(void) arg;
dsl_dataset_t *ds;
int err;
dsl_scan_t *scn = dp->dp_scan;
err = dsl_dataset_hold_obj(dp, hds->ds_object, FTAG, &ds);
if (err)
return (err);
while (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
dsl_dataset_t *prev;
err = dsl_dataset_hold_obj(dp,
dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);
if (err) {
dsl_dataset_rele(ds, FTAG);
return (err);
}
/*
* If this is a clone, we don't need to worry about it for now.
*/
if (dsl_dataset_phys(prev)->ds_next_snap_obj != ds->ds_object) {
dsl_dataset_rele(ds, FTAG);
dsl_dataset_rele(prev, FTAG);
return (0);
}
dsl_dataset_rele(ds, FTAG);
ds = prev;
}
scan_ds_queue_insert(scn, ds->ds_object,
dsl_dataset_phys(ds)->ds_prev_snap_txg);
dsl_dataset_rele(ds, FTAG);
return (0);
}
void
dsl_scan_ddt_entry(dsl_scan_t *scn, enum zio_checksum checksum,
ddt_entry_t *dde, dmu_tx_t *tx)
{
(void) tx;
const ddt_key_t *ddk = &dde->dde_key;
ddt_phys_t *ddp = dde->dde_phys;
blkptr_t bp;
zbookmark_phys_t zb = { 0 };
if (!dsl_scan_is_running(scn))
return;
/*
* This function is special because it is the only thing
* that can add scan_io_t's to the vdev scan queues from
* outside dsl_scan_sync(). For the most part this is ok
* as long as it is called from within syncing context.
* However, dsl_scan_sync() expects that no new sio's will
* be added between when all the work for a scan is done
* and the next txg when the scan is actually marked as
* completed. This check ensures we do not issue new sio's
* during this period.
*/
if (scn->scn_done_txg != 0)
return;
for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
if (ddp->ddp_phys_birth == 0 ||
ddp->ddp_phys_birth > scn->scn_phys.scn_max_txg)
continue;
ddt_bp_create(checksum, ddk, ddp, &bp);
scn->scn_visited_this_txg++;
scan_funcs[scn->scn_phys.scn_func](scn->scn_dp, &bp, &zb);
}
}
/*
* Scrub/dedup interaction.
*
* If there are N references to a deduped block, we don't want to scrub it
* N times -- ideally, we should scrub it exactly once.
*
* We leverage the fact that the dde's replication class (enum ddt_class)
* is ordered from highest replication class (DDT_CLASS_DITTO) to lowest
* (DDT_CLASS_UNIQUE) so that we may walk the DDT in that order.
*
* To prevent excess scrubbing, the scrub begins by walking the DDT
* to find all blocks with refcnt > 1, and scrubs each of these once.
* Since there are two replication classes which contain blocks with
* refcnt > 1, we scrub the highest replication class (DDT_CLASS_DITTO) first.
* Finally the top-down scrub begins, only visiting blocks with refcnt == 1.
*
* There would be nothing more to say if a block's refcnt couldn't change
* during a scrub, but of course it can so we must account for changes
* in a block's replication class.
*
* Here's an example of what can occur:
*
* If a block has refcnt > 1 during the DDT scrub phase, but has refcnt == 1
* when visited during the top-down scrub phase, it will be scrubbed twice.
* This negates our scrub optimization, but is otherwise harmless.
*
* If a block has refcnt == 1 during the DDT scrub phase, but has refcnt > 1
* on each visit during the top-down scrub phase, it will never be scrubbed.
* To catch this, ddt_sync_entry() notifies the scrub code whenever a block's
* reference class transitions to a higher level (i.e DDT_CLASS_UNIQUE to
* DDT_CLASS_DUPLICATE); if it transitions from refcnt == 1 to refcnt > 1
* while a scrub is in progress, it scrubs the block right then.
*/
static void
dsl_scan_ddt(dsl_scan_t *scn, dmu_tx_t *tx)
{
ddt_bookmark_t *ddb = &scn->scn_phys.scn_ddt_bookmark;
ddt_entry_t dde;
int error;
uint64_t n = 0;
bzero(&dde, sizeof (ddt_entry_t));
while ((error = ddt_walk(scn->scn_dp->dp_spa, ddb, &dde)) == 0) {
ddt_t *ddt;
if (ddb->ddb_class > scn->scn_phys.scn_ddt_class_max)
break;
dprintf("visiting ddb=%llu/%llu/%llu/%llx\n",
(longlong_t)ddb->ddb_class,
(longlong_t)ddb->ddb_type,
(longlong_t)ddb->ddb_checksum,
(longlong_t)ddb->ddb_cursor);
/* There should be no pending changes to the dedup table */
ddt = scn->scn_dp->dp_spa->spa_ddt[ddb->ddb_checksum];
ASSERT(avl_first(&ddt->ddt_tree) == NULL);
dsl_scan_ddt_entry(scn, ddb->ddb_checksum, &dde, tx);
n++;
if (dsl_scan_check_suspend(scn, NULL))
break;
}
zfs_dbgmsg("scanned %llu ddt entries on %s with class_max = %u; "
"suspending=%u", (longlong_t)n, scn->scn_dp->dp_spa->spa_name,
(int)scn->scn_phys.scn_ddt_class_max, (int)scn->scn_suspending);
ASSERT(error == 0 || error == ENOENT);
ASSERT(error != ENOENT ||
ddb->ddb_class > scn->scn_phys.scn_ddt_class_max);
}
static uint64_t
dsl_scan_ds_maxtxg(dsl_dataset_t *ds)
{
uint64_t smt = ds->ds_dir->dd_pool->dp_scan->scn_phys.scn_max_txg;
if (ds->ds_is_snapshot)
return (MIN(smt, dsl_dataset_phys(ds)->ds_creation_txg));
return (smt);
}
static void
dsl_scan_visit(dsl_scan_t *scn, dmu_tx_t *tx)
{
scan_ds_t *sds;
dsl_pool_t *dp = scn->scn_dp;
if (scn->scn_phys.scn_ddt_bookmark.ddb_class <=
scn->scn_phys.scn_ddt_class_max) {
scn->scn_phys.scn_cur_min_txg = scn->scn_phys.scn_min_txg;
scn->scn_phys.scn_cur_max_txg = scn->scn_phys.scn_max_txg;
dsl_scan_ddt(scn, tx);
if (scn->scn_suspending)
return;
}
if (scn->scn_phys.scn_bookmark.zb_objset == DMU_META_OBJSET) {
/* First do the MOS & ORIGIN */
scn->scn_phys.scn_cur_min_txg = scn->scn_phys.scn_min_txg;
scn->scn_phys.scn_cur_max_txg = scn->scn_phys.scn_max_txg;
dsl_scan_visit_rootbp(scn, NULL,
&dp->dp_meta_rootbp, tx);
spa_set_rootblkptr(dp->dp_spa, &dp->dp_meta_rootbp);
if (scn->scn_suspending)
return;
if (spa_version(dp->dp_spa) < SPA_VERSION_DSL_SCRUB) {
VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
enqueue_cb, NULL, DS_FIND_CHILDREN));
} else {
dsl_scan_visitds(scn,
dp->dp_origin_snap->ds_object, tx);
}
ASSERT(!scn->scn_suspending);
} else if (scn->scn_phys.scn_bookmark.zb_objset !=
ZB_DESTROYED_OBJSET) {
uint64_t dsobj = scn->scn_phys.scn_bookmark.zb_objset;
/*
* If we were suspended, continue from here. Note if the
* ds we were suspended on was deleted, the zb_objset may
* be -1, so we will skip this and find a new objset
* below.
*/
dsl_scan_visitds(scn, dsobj, tx);
if (scn->scn_suspending)
return;
}
/*
* In case we suspended right at the end of the ds, zero the
* bookmark so we don't think that we're still trying to resume.
*/
bzero(&scn->scn_phys.scn_bookmark, sizeof (zbookmark_phys_t));
/*
* Keep pulling things out of the dataset avl queue. Updates to the
* persistent zap-object-as-queue happen only at checkpoints.
*/
while ((sds = avl_first(&scn->scn_queue)) != NULL) {
dsl_dataset_t *ds;
uint64_t dsobj = sds->sds_dsobj;
uint64_t txg = sds->sds_txg;
/* dequeue and free the ds from the queue */
scan_ds_queue_remove(scn, dsobj);
sds = NULL;
/* set up min / max txg */
VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
if (txg != 0) {
scn->scn_phys.scn_cur_min_txg =
MAX(scn->scn_phys.scn_min_txg, txg);
} else {
scn->scn_phys.scn_cur_min_txg =
MAX(scn->scn_phys.scn_min_txg,
dsl_dataset_phys(ds)->ds_prev_snap_txg);
}
scn->scn_phys.scn_cur_max_txg = dsl_scan_ds_maxtxg(ds);
dsl_dataset_rele(ds, FTAG);
dsl_scan_visitds(scn, dsobj, tx);
if (scn->scn_suspending)
return;
}
/* No more objsets to fetch, we're done */
scn->scn_phys.scn_bookmark.zb_objset = ZB_DESTROYED_OBJSET;
ASSERT0(scn->scn_suspending);
}
static uint64_t
dsl_scan_count_data_disks(vdev_t *rvd)
{
uint64_t i, leaves = 0;
for (i = 0; i < rvd->vdev_children; i++) {
vdev_t *vd = rvd->vdev_child[i];
if (vd->vdev_islog || vd->vdev_isspare || vd->vdev_isl2cache)
continue;
leaves += vdev_get_ndisks(vd) - vdev_get_nparity(vd);
}
return (leaves);
}
static void
scan_io_queues_update_zio_stats(dsl_scan_io_queue_t *q, const blkptr_t *bp)
{
int i;
uint64_t cur_size = 0;
for (i = 0; i < BP_GET_NDVAS(bp); i++) {
cur_size += DVA_GET_ASIZE(&bp->blk_dva[i]);
}
q->q_total_zio_size_this_txg += cur_size;
q->q_zios_this_txg++;
}
static void
scan_io_queues_update_seg_stats(dsl_scan_io_queue_t *q, uint64_t start,
uint64_t end)
{
q->q_total_seg_size_this_txg += end - start;
q->q_segs_this_txg++;
}
static boolean_t
scan_io_queue_check_suspend(dsl_scan_t *scn)
{
/* See comment in dsl_scan_check_suspend() */
uint64_t curr_time_ns = gethrtime();
uint64_t scan_time_ns = curr_time_ns - scn->scn_sync_start_time;
uint64_t sync_time_ns = curr_time_ns -
scn->scn_dp->dp_spa->spa_sync_starttime;
int dirty_pct = scn->scn_dp->dp_dirty_total * 100 / zfs_dirty_data_max;
int mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
zfs_resilver_min_time_ms : zfs_scrub_min_time_ms;
return ((NSEC2MSEC(scan_time_ns) > mintime &&
(dirty_pct >= zfs_vdev_async_write_active_min_dirty_percent ||
txg_sync_waiting(scn->scn_dp) ||
NSEC2SEC(sync_time_ns) >= zfs_txg_timeout)) ||
spa_shutting_down(scn->scn_dp->dp_spa));
}
/*
* Given a list of scan_io_t's in io_list, this issues the I/Os out to
* disk. This consumes the io_list and frees the scan_io_t's. This is
* called when emptying queues, either when we're up against the memory
* limit or when we have finished scanning. Returns B_TRUE if we stopped
* processing the list before we finished. Any sios that were not issued
* will remain in the io_list.
*/
static boolean_t
scan_io_queue_issue(dsl_scan_io_queue_t *queue, list_t *io_list)
{
dsl_scan_t *scn = queue->q_scn;
scan_io_t *sio;
int64_t bytes_issued = 0;
boolean_t suspended = B_FALSE;
while ((sio = list_head(io_list)) != NULL) {
blkptr_t bp;
if (scan_io_queue_check_suspend(scn)) {
suspended = B_TRUE;
break;
}
sio2bp(sio, &bp);
bytes_issued += SIO_GET_ASIZE(sio);
scan_exec_io(scn->scn_dp, &bp, sio->sio_flags,
&sio->sio_zb, queue);
(void) list_remove_head(io_list);
scan_io_queues_update_zio_stats(queue, &bp);
sio_free(sio);
}
atomic_add_64(&scn->scn_bytes_pending, -bytes_issued);
return (suspended);
}
/*
* This function removes sios from an IO queue which reside within a given
* range_seg_t and inserts them (in offset order) into a list. Note that
* we only ever return a maximum of 32 sios at once. If there are more sios
* to process within this segment that did not make it onto the list we
* return B_TRUE and otherwise B_FALSE.
*/
static boolean_t
scan_io_queue_gather(dsl_scan_io_queue_t *queue, range_seg_t *rs, list_t *list)
{
scan_io_t *srch_sio, *sio, *next_sio;
avl_index_t idx;
uint_t num_sios = 0;
int64_t bytes_issued = 0;
ASSERT(rs != NULL);
ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));
srch_sio = sio_alloc(1);
srch_sio->sio_nr_dvas = 1;
SIO_SET_OFFSET(srch_sio, rs_get_start(rs, queue->q_exts_by_addr));
/*
* The exact start of the extent might not contain any matching zios,
* so if that's the case, examine the next one in the tree.
*/
sio = avl_find(&queue->q_sios_by_addr, srch_sio, &idx);
sio_free(srch_sio);
if (sio == NULL)
sio = avl_nearest(&queue->q_sios_by_addr, idx, AVL_AFTER);
while (sio != NULL && SIO_GET_OFFSET(sio) < rs_get_end(rs,
queue->q_exts_by_addr) && num_sios <= 32) {
ASSERT3U(SIO_GET_OFFSET(sio), >=, rs_get_start(rs,
queue->q_exts_by_addr));
ASSERT3U(SIO_GET_END_OFFSET(sio), <=, rs_get_end(rs,
queue->q_exts_by_addr));
next_sio = AVL_NEXT(&queue->q_sios_by_addr, sio);
avl_remove(&queue->q_sios_by_addr, sio);
queue->q_sio_memused -= SIO_GET_MUSED(sio);
bytes_issued += SIO_GET_ASIZE(sio);
num_sios++;
list_insert_tail(list, sio);
sio = next_sio;
}
/*
* We limit the number of sios we process at once to 32 to avoid
* biting off more than we can chew. If we didn't take everything
* in the segment we update it to reflect the work we were able to
* complete. Otherwise, we remove it from the range tree entirely.
*/
if (sio != NULL && SIO_GET_OFFSET(sio) < rs_get_end(rs,
queue->q_exts_by_addr)) {
range_tree_adjust_fill(queue->q_exts_by_addr, rs,
-bytes_issued);
range_tree_resize_segment(queue->q_exts_by_addr, rs,
SIO_GET_OFFSET(sio), rs_get_end(rs,
queue->q_exts_by_addr) - SIO_GET_OFFSET(sio));
return (B_TRUE);
} else {
uint64_t rstart = rs_get_start(rs, queue->q_exts_by_addr);
uint64_t rend = rs_get_end(rs, queue->q_exts_by_addr);
range_tree_remove(queue->q_exts_by_addr, rstart, rend - rstart);
return (B_FALSE);
}
}
/*
* This is called from the queue emptying thread and selects the next
* extent from which we are to issue I/Os. The behavior of this function
* depends on the state of the scan, the current memory consumption and
* whether or not we are performing a scan shutdown.
* 1) We select extents in an elevator algorithm (LBA-order) if the scan
* needs to perform a checkpoint
* 2) We select the largest available extent if we are up against the
* memory limit.
* 3) Otherwise we don't select any extents.
*/
static range_seg_t *
scan_io_queue_fetch_ext(dsl_scan_io_queue_t *queue)
{
dsl_scan_t *scn = queue->q_scn;
range_tree_t *rt = queue->q_exts_by_addr;
ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));
ASSERT(scn->scn_is_sorted);
/* handle tunable overrides */
if (scn->scn_checkpointing || scn->scn_clearing) {
if (zfs_scan_issue_strategy == 1) {
return (range_tree_first(rt));
} else if (zfs_scan_issue_strategy == 2) {
/*
* We need to get the original entry in the by_addr
* tree so we can modify it.
*/
range_seg_t *size_rs =
zfs_btree_first(&queue->q_exts_by_size, NULL);
if (size_rs == NULL)
return (NULL);
uint64_t start = rs_get_start(size_rs, rt);
uint64_t size = rs_get_end(size_rs, rt) - start;
range_seg_t *addr_rs = range_tree_find(rt, start,
size);
ASSERT3P(addr_rs, !=, NULL);
ASSERT3U(rs_get_start(size_rs, rt), ==,
rs_get_start(addr_rs, rt));
ASSERT3U(rs_get_end(size_rs, rt), ==,
rs_get_end(addr_rs, rt));
return (addr_rs);
}
}
/*
* During normal clearing, we want to issue our largest segments
* first, keeping IO as sequential as possible, and leaving the
* smaller extents for later with the hope that they might eventually
* grow to larger sequential segments. However, when the scan is
* checkpointing, no new extents will be added to the sorting queue,
* so the way we are sorted now is as good as it will ever get.
* In this case, we instead switch to issuing extents in LBA order.
*/
if (scn->scn_checkpointing) {
return (range_tree_first(rt));
} else if (scn->scn_clearing) {
/*
* We need to get the original entry in the by_addr
* tree so we can modify it.
*/
range_seg_t *size_rs = zfs_btree_first(&queue->q_exts_by_size,
NULL);
if (size_rs == NULL)
return (NULL);
uint64_t start = rs_get_start(size_rs, rt);
uint64_t size = rs_get_end(size_rs, rt) - start;
range_seg_t *addr_rs = range_tree_find(rt, start, size);
ASSERT3P(addr_rs, !=, NULL);
ASSERT3U(rs_get_start(size_rs, rt), ==, rs_get_start(addr_rs,
rt));
ASSERT3U(rs_get_end(size_rs, rt), ==, rs_get_end(addr_rs, rt));
return (addr_rs);
} else {
return (NULL);
}
}
static void
scan_io_queues_run_one(void *arg)
{
dsl_scan_io_queue_t *queue = arg;
kmutex_t *q_lock = &queue->q_vd->vdev_scan_io_queue_lock;
boolean_t suspended = B_FALSE;
range_seg_t *rs = NULL;
scan_io_t *sio = NULL;
list_t sio_list;
ASSERT(queue->q_scn->scn_is_sorted);
list_create(&sio_list, sizeof (scan_io_t),
offsetof(scan_io_t, sio_nodes.sio_list_node));
mutex_enter(q_lock);
/* Calculate maximum in-flight bytes for this vdev. */
queue->q_maxinflight_bytes = MAX(1, zfs_scan_vdev_limit *
(vdev_get_ndisks(queue->q_vd) - vdev_get_nparity(queue->q_vd)));
/* reset per-queue scan statistics for this txg */
queue->q_total_seg_size_this_txg = 0;
queue->q_segs_this_txg = 0;
queue->q_total_zio_size_this_txg = 0;
queue->q_zios_this_txg = 0;
/* loop until we run out of time or sios */
while ((rs = scan_io_queue_fetch_ext(queue)) != NULL) {
uint64_t seg_start = 0, seg_end = 0;
boolean_t more_left = B_TRUE;
ASSERT(list_is_empty(&sio_list));
/* loop while we still have sios left to process in this rs */
while (more_left) {
scan_io_t *first_sio, *last_sio;
/*
* We have selected which extent needs to be
* processed next. Gather up the corresponding sios.
*/
more_left = scan_io_queue_gather(queue, rs, &sio_list);
ASSERT(!list_is_empty(&sio_list));
first_sio = list_head(&sio_list);
last_sio = list_tail(&sio_list);
seg_end = SIO_GET_END_OFFSET(last_sio);
if (seg_start == 0)
seg_start = SIO_GET_OFFSET(first_sio);
/*
* Issuing sios can take a long time so drop the
* queue lock. The sio queue won't be updated by
* other threads since we're in syncing context so
* we can be sure that our trees will remain exactly
* as we left them.
*/
mutex_exit(q_lock);
suspended = scan_io_queue_issue(queue, &sio_list);
mutex_enter(q_lock);
if (suspended)
break;
}
/* update statistics for debugging purposes */
scan_io_queues_update_seg_stats(queue, seg_start, seg_end);
if (suspended)
break;
}
/*
* If we were suspended in the middle of processing,
* requeue any unfinished sios and exit.
*/
while ((sio = list_head(&sio_list)) != NULL) {
list_remove(&sio_list, sio);
scan_io_queue_insert_impl(queue, sio);
}
mutex_exit(q_lock);
list_destroy(&sio_list);
}
/*
* Performs an emptying run on all scan queues in the pool. This just
* punches out one thread per top-level vdev, each of which processes
* only that vdev's scan queue. We can parallelize the I/O here because
* we know that each queue's I/Os only affect its own top-level vdev.
*
* This function waits for the queue runs to complete, and must be
* called from dsl_scan_sync (or in general, syncing context).
*/
static void
scan_io_queues_run(dsl_scan_t *scn)
{
spa_t *spa = scn->scn_dp->dp_spa;
ASSERT(scn->scn_is_sorted);
ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));
if (scn->scn_bytes_pending == 0)
return;
if (scn->scn_taskq == NULL) {
int nthreads = spa->spa_root_vdev->vdev_children;
/*
* We need to make this taskq *always* execute as many
* threads in parallel as we have top-level vdevs and no
* less, otherwise strange serialization of the calls to
* scan_io_queues_run_one can occur during spa_sync runs
* and that significantly impacts performance.
*/
scn->scn_taskq = taskq_create("dsl_scan_iss", nthreads,
minclsyspri, nthreads, nthreads, TASKQ_PREPOPULATE);
}
for (uint64_t i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
vdev_t *vd = spa->spa_root_vdev->vdev_child[i];
mutex_enter(&vd->vdev_scan_io_queue_lock);
if (vd->vdev_scan_io_queue != NULL) {
VERIFY(taskq_dispatch(scn->scn_taskq,
scan_io_queues_run_one, vd->vdev_scan_io_queue,
TQ_SLEEP) != TASKQID_INVALID);
}
mutex_exit(&vd->vdev_scan_io_queue_lock);
}
/*
* Wait for the queues to finish issuing their IOs for this run
* before we return. There may still be IOs in flight at this
* point.
*/
taskq_wait(scn->scn_taskq);
}
static boolean_t
dsl_scan_async_block_should_pause(dsl_scan_t *scn)
{
uint64_t elapsed_nanosecs;
if (zfs_recover)
return (B_FALSE);
if (zfs_async_block_max_blocks != 0 &&
scn->scn_visited_this_txg >= zfs_async_block_max_blocks) {
return (B_TRUE);
}
if (zfs_max_async_dedup_frees != 0 &&
scn->scn_dedup_frees_this_txg >= zfs_max_async_dedup_frees) {
return (B_TRUE);
}
elapsed_nanosecs = gethrtime() - scn->scn_sync_start_time;
return (elapsed_nanosecs / NANOSEC > zfs_txg_timeout ||
(NSEC2MSEC(elapsed_nanosecs) > scn->scn_async_block_min_time_ms &&
txg_sync_waiting(scn->scn_dp)) ||
spa_shutting_down(scn->scn_dp->dp_spa));
}
static int
dsl_scan_free_block_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
{
dsl_scan_t *scn = arg;
if (!scn->scn_is_bptree ||
(BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_OBJSET)) {
if (dsl_scan_async_block_should_pause(scn))
return (SET_ERROR(ERESTART));
}
zio_nowait(zio_free_sync(scn->scn_zio_root, scn->scn_dp->dp_spa,
dmu_tx_get_txg(tx), bp, 0));
dsl_dir_diduse_space(tx->tx_pool->dp_free_dir, DD_USED_HEAD,
-bp_get_dsize_sync(scn->scn_dp->dp_spa, bp),
-BP_GET_PSIZE(bp), -BP_GET_UCSIZE(bp), tx);
scn->scn_visited_this_txg++;
if (BP_GET_DEDUP(bp))
scn->scn_dedup_frees_this_txg++;
return (0);
}
static void
dsl_scan_update_stats(dsl_scan_t *scn)
{
spa_t *spa = scn->scn_dp->dp_spa;
uint64_t i;
uint64_t seg_size_total = 0, zio_size_total = 0;
uint64_t seg_count_total = 0, zio_count_total = 0;
for (i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
vdev_t *vd = spa->spa_root_vdev->vdev_child[i];
dsl_scan_io_queue_t *queue = vd->vdev_scan_io_queue;
if (queue == NULL)
continue;
seg_size_total += queue->q_total_seg_size_this_txg;
zio_size_total += queue->q_total_zio_size_this_txg;
seg_count_total += queue->q_segs_this_txg;
zio_count_total += queue->q_zios_this_txg;
}
if (seg_count_total == 0 || zio_count_total == 0) {
scn->scn_avg_seg_size_this_txg = 0;
scn->scn_avg_zio_size_this_txg = 0;
scn->scn_segs_this_txg = 0;
scn->scn_zios_this_txg = 0;
return;
}
scn->scn_avg_seg_size_this_txg = seg_size_total / seg_count_total;
scn->scn_avg_zio_size_this_txg = zio_size_total / zio_count_total;
scn->scn_segs_this_txg = seg_count_total;
scn->scn_zios_this_txg = zio_count_total;
}
static int
bpobj_dsl_scan_free_block_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed,
dmu_tx_t *tx)
{
ASSERT(!bp_freed);
return (dsl_scan_free_block_cb(arg, bp, tx));
}
static int
dsl_scan_obsolete_block_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed,
dmu_tx_t *tx)
{
ASSERT(!bp_freed);
dsl_scan_t *scn = arg;
const dva_t *dva = &bp->blk_dva[0];
if (dsl_scan_async_block_should_pause(scn))
return (SET_ERROR(ERESTART));
spa_vdev_indirect_mark_obsolete(scn->scn_dp->dp_spa,
DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva),
DVA_GET_ASIZE(dva), tx);
scn->scn_visited_this_txg++;
return (0);
}
boolean_t
dsl_scan_active(dsl_scan_t *scn)
{
spa_t *spa = scn->scn_dp->dp_spa;
uint64_t used = 0, comp, uncomp;
boolean_t clones_left;
if (spa->spa_load_state != SPA_LOAD_NONE)
return (B_FALSE);
if (spa_shutting_down(spa))
return (B_FALSE);
if ((dsl_scan_is_running(scn) && !dsl_scan_is_paused_scrub(scn)) ||
(scn->scn_async_destroying && !scn->scn_async_stalled))
return (B_TRUE);
if (spa_version(scn->scn_dp->dp_spa) >= SPA_VERSION_DEADLISTS) {
(void) bpobj_space(&scn->scn_dp->dp_free_bpobj,
&used, &comp, &uncomp);
}
clones_left = spa_livelist_delete_check(spa);
return ((used != 0) || (clones_left));
}
static boolean_t
dsl_scan_check_deferred(vdev_t *vd)
{
boolean_t need_resilver = B_FALSE;
for (int c = 0; c < vd->vdev_children; c++) {
need_resilver |=
dsl_scan_check_deferred(vd->vdev_child[c]);
}
if (!vdev_is_concrete(vd) || vd->vdev_aux ||
!vd->vdev_ops->vdev_op_leaf)
return (need_resilver);
if (!vd->vdev_resilver_deferred)
need_resilver = B_TRUE;
return (need_resilver);
}
static boolean_t
dsl_scan_need_resilver(spa_t *spa, const dva_t *dva, size_t psize,
uint64_t phys_birth)
{
vdev_t *vd;
vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));
if (vd->vdev_ops == &vdev_indirect_ops) {
/*
* The indirect vdev can point to multiple
* vdevs. For simplicity, always create
* the resilver zio_t. zio_vdev_io_start()
* will bypass the child resilver i/o's if
* they are on vdevs that don't have DTL's.
*/
return (B_TRUE);
}
if (DVA_GET_GANG(dva)) {
/*
* Gang members may be spread across multiple
* vdevs, so the best estimate we have is the
* scrub range, which has already been checked.
* XXX -- it would be better to change our
* allocation policy to ensure that all
* gang members reside on the same vdev.
*/
return (B_TRUE);
}
/*
* Check if the top-level vdev must resilver this offset.
* When the offset does not intersect with a dirty leaf DTL
* then it may be possible to skip the resilver IO. The psize
* is provided instead of asize to simplify the check for RAIDZ.
*/
if (!vdev_dtl_need_resilver(vd, dva, psize, phys_birth))
return (B_FALSE);
/*
* Check that this top-level vdev has a device under it which
* is resilvering and is not deferred.
*/
if (!dsl_scan_check_deferred(vd))
return (B_FALSE);
return (B_TRUE);
}
static int
dsl_process_async_destroys(dsl_pool_t *dp, dmu_tx_t *tx)
{
dsl_scan_t *scn = dp->dp_scan;
spa_t *spa = dp->dp_spa;
int err = 0;
if (spa_suspend_async_destroy(spa))
return (0);
if (zfs_free_bpobj_enabled &&
spa_version(spa) >= SPA_VERSION_DEADLISTS) {
scn->scn_is_bptree = B_FALSE;
scn->scn_async_block_min_time_ms = zfs_free_min_time_ms;
scn->scn_zio_root = zio_root(spa, NULL,
NULL, ZIO_FLAG_MUSTSUCCEED);
err = bpobj_iterate(&dp->dp_free_bpobj,
bpobj_dsl_scan_free_block_cb, scn, tx);
VERIFY0(zio_wait(scn->scn_zio_root));
scn->scn_zio_root = NULL;
if (err != 0 && err != ERESTART)
zfs_panic_recover("error %u from bpobj_iterate()", err);
}
if (err == 0 && spa_feature_is_active(spa, SPA_FEATURE_ASYNC_DESTROY)) {
ASSERT(scn->scn_async_destroying);
scn->scn_is_bptree = B_TRUE;
scn->scn_zio_root = zio_root(spa, NULL,
NULL, ZIO_FLAG_MUSTSUCCEED);
err = bptree_iterate(dp->dp_meta_objset,
dp->dp_bptree_obj, B_TRUE, dsl_scan_free_block_cb, scn, tx);
VERIFY0(zio_wait(scn->scn_zio_root));
scn->scn_zio_root = NULL;
if (err == EIO || err == ECKSUM) {
err = 0;
} else if (err != 0 && err != ERESTART) {
zfs_panic_recover("error %u from "
"traverse_dataset_destroyed()", err);
}
if (bptree_is_empty(dp->dp_meta_objset, dp->dp_bptree_obj)) {
/* finished; deactivate async destroy feature */
spa_feature_decr(spa, SPA_FEATURE_ASYNC_DESTROY, tx);
ASSERT(!spa_feature_is_active(spa,
SPA_FEATURE_ASYNC_DESTROY));
VERIFY0(zap_remove(dp->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_BPTREE_OBJ, tx));
VERIFY0(bptree_free(dp->dp_meta_objset,
dp->dp_bptree_obj, tx));
dp->dp_bptree_obj = 0;
scn->scn_async_destroying = B_FALSE;
scn->scn_async_stalled = B_FALSE;
} else {
/*
* If we didn't make progress, mark the async
* destroy as stalled, so that we will not initiate
* a spa_sync() on its behalf. Note that we only
* check this if we are not finished, because if the
* bptree had no blocks for us to visit, we can
* finish without "making progress".
*/
scn->scn_async_stalled =
(scn->scn_visited_this_txg == 0);
}
}
if (scn->scn_visited_this_txg) {
zfs_dbgmsg("freed %llu blocks in %llums from "
"free_bpobj/bptree on %s in txg %llu; err=%u",
(longlong_t)scn->scn_visited_this_txg,
(longlong_t)
NSEC2MSEC(gethrtime() - scn->scn_sync_start_time),
spa->spa_name, (longlong_t)tx->tx_txg, err);
scn->scn_visited_this_txg = 0;
scn->scn_dedup_frees_this_txg = 0;
/*
* Write out changes to the DDT that may be required as a
* result of the blocks freed. This ensures that the DDT
* is clean when a scrub/resilver runs.
*/
ddt_sync(spa, tx->tx_txg);
}
if (err != 0)
return (err);
if (dp->dp_free_dir != NULL && !scn->scn_async_destroying &&
zfs_free_leak_on_eio &&
(dsl_dir_phys(dp->dp_free_dir)->dd_used_bytes != 0 ||
dsl_dir_phys(dp->dp_free_dir)->dd_compressed_bytes != 0 ||
dsl_dir_phys(dp->dp_free_dir)->dd_uncompressed_bytes != 0)) {
/*
* We have finished background destroying, but there is still
* some space left in the dp_free_dir. Transfer this leaked
* space to the dp_leak_dir.
*/
if (dp->dp_leak_dir == NULL) {
rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
(void) dsl_dir_create_sync(dp, dp->dp_root_dir,
LEAK_DIR_NAME, tx);
VERIFY0(dsl_pool_open_special_dir(dp,
LEAK_DIR_NAME, &dp->dp_leak_dir));
rrw_exit(&dp->dp_config_rwlock, FTAG);
}
dsl_dir_diduse_space(dp->dp_leak_dir, DD_USED_HEAD,
dsl_dir_phys(dp->dp_free_dir)->dd_used_bytes,
dsl_dir_phys(dp->dp_free_dir)->dd_compressed_bytes,
dsl_dir_phys(dp->dp_free_dir)->dd_uncompressed_bytes, tx);
dsl_dir_diduse_space(dp->dp_free_dir, DD_USED_HEAD,
-dsl_dir_phys(dp->dp_free_dir)->dd_used_bytes,
-dsl_dir_phys(dp->dp_free_dir)->dd_compressed_bytes,
-dsl_dir_phys(dp->dp_free_dir)->dd_uncompressed_bytes, tx);
}
if (dp->dp_free_dir != NULL && !scn->scn_async_destroying &&
!spa_livelist_delete_check(spa)) {
/* finished; verify that space accounting went to zero */
ASSERT0(dsl_dir_phys(dp->dp_free_dir)->dd_used_bytes);
ASSERT0(dsl_dir_phys(dp->dp_free_dir)->dd_compressed_bytes);
ASSERT0(dsl_dir_phys(dp->dp_free_dir)->dd_uncompressed_bytes);
}
spa_notify_waiters(spa);
EQUIV(bpobj_is_open(&dp->dp_obsolete_bpobj),
0 == zap_contains(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_OBSOLETE_BPOBJ));
if (err == 0 && bpobj_is_open(&dp->dp_obsolete_bpobj)) {
ASSERT(spa_feature_is_active(dp->dp_spa,
SPA_FEATURE_OBSOLETE_COUNTS));
scn->scn_is_bptree = B_FALSE;
scn->scn_async_block_min_time_ms = zfs_obsolete_min_time_ms;
err = bpobj_iterate(&dp->dp_obsolete_bpobj,
dsl_scan_obsolete_block_cb, scn, tx);
if (err != 0 && err != ERESTART)
zfs_panic_recover("error %u from bpobj_iterate()", err);
if (bpobj_is_empty(&dp->dp_obsolete_bpobj))
dsl_pool_destroy_obsolete_bpobj(dp, tx);
}
return (0);
}
/*
* This is the primary entry point for scans that is called from syncing
* context. Scans must happen entirely during syncing context so that we
* can guarantee that blocks we are currently scanning will not change out
* from under us. While a scan is active, this function controls how quickly
* transaction groups proceed, instead of the normal handling provided by
* txg_sync_thread().
*/
void
dsl_scan_sync(dsl_pool_t *dp, dmu_tx_t *tx)
{
int err = 0;
dsl_scan_t *scn = dp->dp_scan;
spa_t *spa = dp->dp_spa;
state_sync_type_t sync_type = SYNC_OPTIONAL;
if (spa->spa_resilver_deferred &&
!spa_feature_is_active(dp->dp_spa, SPA_FEATURE_RESILVER_DEFER))
spa_feature_incr(spa, SPA_FEATURE_RESILVER_DEFER, tx);
/*
* Check for scn_restart_txg before checking spa_load_state, so
* that we can restart an old-style scan while the pool is being
* imported (see dsl_scan_init). We also restart scans if there
* is a deferred resilver and the user has manually disabled
* deferred resilvers via the tunable.
*/
if (dsl_scan_restarting(scn, tx) ||
(spa->spa_resilver_deferred && zfs_resilver_disable_defer)) {
pool_scan_func_t func = POOL_SCAN_SCRUB;
dsl_scan_done(scn, B_FALSE, tx);
if (vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL))
func = POOL_SCAN_RESILVER;
zfs_dbgmsg("restarting scan func=%u on %s txg=%llu",
func, dp->dp_spa->spa_name, (longlong_t)tx->tx_txg);
dsl_scan_setup_sync(&func, tx);
}
/*
* Only process scans in sync pass 1.
*/
if (spa_sync_pass(spa) > 1)
return;
/*
* If the spa is shutting down, then stop scanning. This will
* ensure that the scan does not dirty any new data during the
* shutdown phase.
*/
if (spa_shutting_down(spa))
return;
/*
* If the scan is inactive due to a stalled async destroy, try again.
*/
if (!scn->scn_async_stalled && !dsl_scan_active(scn))
return;
/* reset scan statistics */
scn->scn_visited_this_txg = 0;
scn->scn_dedup_frees_this_txg = 0;
scn->scn_holes_this_txg = 0;
scn->scn_lt_min_this_txg = 0;
scn->scn_gt_max_this_txg = 0;
scn->scn_ddt_contained_this_txg = 0;
scn->scn_objsets_visited_this_txg = 0;
scn->scn_avg_seg_size_this_txg = 0;
scn->scn_segs_this_txg = 0;
scn->scn_avg_zio_size_this_txg = 0;
scn->scn_zios_this_txg = 0;
scn->scn_suspending = B_FALSE;
scn->scn_sync_start_time = gethrtime();
spa->spa_scrub_active = B_TRUE;
/*
* First process the async destroys. If we suspend, don't do
* any scrubbing or resilvering. This ensures that there are no
* async destroys while we are scanning, so the scan code doesn't
* have to worry about traversing it. It is also faster to free the
* blocks than to scrub them.
*/
err = dsl_process_async_destroys(dp, tx);
if (err != 0)
return;
if (!dsl_scan_is_running(scn) || dsl_scan_is_paused_scrub(scn))
return;
/*
* Wait a few txgs after importing to begin scanning so that
* we can get the pool imported quickly.
*/
if (spa->spa_syncing_txg < spa->spa_first_txg + SCAN_IMPORT_WAIT_TXGS)
return;
/*
* zfs_scan_suspend_progress can be set to disable scan progress.
* We don't want to spin the txg_sync thread, so we add a delay
* here to simulate the time spent doing a scan. This is mostly
* useful for testing and debugging.
*/
if (zfs_scan_suspend_progress) {
uint64_t scan_time_ns = gethrtime() - scn->scn_sync_start_time;
int mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
zfs_resilver_min_time_ms : zfs_scrub_min_time_ms;
while (zfs_scan_suspend_progress &&
!txg_sync_waiting(scn->scn_dp) &&
!spa_shutting_down(scn->scn_dp->dp_spa) &&
NSEC2MSEC(scan_time_ns) < mintime) {
delay(hz);
scan_time_ns = gethrtime() - scn->scn_sync_start_time;
}
return;
}
/*
* It is possible to switch from unsorted to sorted at any time,
* but afterwards the scan will remain sorted unless reloaded from
* a checkpoint after a reboot.
*/
if (!zfs_scan_legacy) {
scn->scn_is_sorted = B_TRUE;
if (scn->scn_last_checkpoint == 0)
scn->scn_last_checkpoint = ddi_get_lbolt();
}
/*
* For sorted scans, determine what kind of work we will be doing
* this txg based on our memory limitations and whether or not we
* need to perform a checkpoint.
*/
if (scn->scn_is_sorted) {
/*
* If we are over our checkpoint interval, set scn_clearing
* so that we can begin checkpointing immediately. The
* checkpoint allows us to save a consistent bookmark
* representing how much data we have scrubbed so far.
* Otherwise, use the memory limit to determine if we should
* scan for metadata or start issue scrub IOs. We accumulate
* metadata until we hit our hard memory limit at which point
* we issue scrub IOs until we are at our soft memory limit.
*/
if (scn->scn_checkpointing ||
ddi_get_lbolt() - scn->scn_last_checkpoint >
SEC_TO_TICK(zfs_scan_checkpoint_intval)) {
if (!scn->scn_checkpointing)
zfs_dbgmsg("begin scan checkpoint for %s",
spa->spa_name);
scn->scn_checkpointing = B_TRUE;
scn->scn_clearing = B_TRUE;
} else {
boolean_t should_clear = dsl_scan_should_clear(scn);
if (should_clear && !scn->scn_clearing) {
zfs_dbgmsg("begin scan clearing for %s",
spa->spa_name);
scn->scn_clearing = B_TRUE;
} else if (!should_clear && scn->scn_clearing) {
zfs_dbgmsg("finish scan clearing for %s",
spa->spa_name);
scn->scn_clearing = B_FALSE;
}
}
} else {
ASSERT0(scn->scn_checkpointing);
ASSERT0(scn->scn_clearing);
}
if (!scn->scn_clearing && scn->scn_done_txg == 0) {
/* Need to scan metadata for more blocks to scrub */
dsl_scan_phys_t *scnp = &scn->scn_phys;
taskqid_t prefetch_tqid;
/*
* Recalculate the max number of in-flight bytes for pool-wide
* scanning operations (minimum 1MB). Limits for the issuing
* phase are done per top-level vdev and are handled separately.
*/
scn->scn_maxinflight_bytes = MAX(zfs_scan_vdev_limit *
dsl_scan_count_data_disks(spa->spa_root_vdev), 1ULL << 20);
if (scnp->scn_ddt_bookmark.ddb_class <=
scnp->scn_ddt_class_max) {
ASSERT(ZB_IS_ZERO(&scnp->scn_bookmark));
zfs_dbgmsg("doing scan sync for %s txg %llu; "
"ddt bm=%llu/%llu/%llu/%llx",
spa->spa_name,
(longlong_t)tx->tx_txg,
(longlong_t)scnp->scn_ddt_bookmark.ddb_class,
(longlong_t)scnp->scn_ddt_bookmark.ddb_type,
(longlong_t)scnp->scn_ddt_bookmark.ddb_checksum,
(longlong_t)scnp->scn_ddt_bookmark.ddb_cursor);
} else {
zfs_dbgmsg("doing scan sync for %s txg %llu; "
"bm=%llu/%llu/%llu/%llu",
spa->spa_name,
(longlong_t)tx->tx_txg,
(longlong_t)scnp->scn_bookmark.zb_objset,
(longlong_t)scnp->scn_bookmark.zb_object,
(longlong_t)scnp->scn_bookmark.zb_level,
(longlong_t)scnp->scn_bookmark.zb_blkid);
}
scn->scn_zio_root = zio_root(dp->dp_spa, NULL,
NULL, ZIO_FLAG_CANFAIL);
scn->scn_prefetch_stop = B_FALSE;
prefetch_tqid = taskq_dispatch(dp->dp_sync_taskq,
dsl_scan_prefetch_thread, scn, TQ_SLEEP);
ASSERT(prefetch_tqid != TASKQID_INVALID);
dsl_pool_config_enter(dp, FTAG);
dsl_scan_visit(scn, tx);
dsl_pool_config_exit(dp, FTAG);
mutex_enter(&dp->dp_spa->spa_scrub_lock);
scn->scn_prefetch_stop = B_TRUE;
cv_broadcast(&spa->spa_scrub_io_cv);
mutex_exit(&dp->dp_spa->spa_scrub_lock);
taskq_wait_id(dp->dp_sync_taskq, prefetch_tqid);
(void) zio_wait(scn->scn_zio_root);
scn->scn_zio_root = NULL;
zfs_dbgmsg("scan visited %llu blocks of %s in %llums "
"(%llu os's, %llu holes, %llu < mintxg, "
"%llu in ddt, %llu > maxtxg)",
(longlong_t)scn->scn_visited_this_txg,
spa->spa_name,
(longlong_t)NSEC2MSEC(gethrtime() -
scn->scn_sync_start_time),
(longlong_t)scn->scn_objsets_visited_this_txg,
(longlong_t)scn->scn_holes_this_txg,
(longlong_t)scn->scn_lt_min_this_txg,
(longlong_t)scn->scn_ddt_contained_this_txg,
(longlong_t)scn->scn_gt_max_this_txg);
if (!scn->scn_suspending) {
ASSERT0(avl_numnodes(&scn->scn_queue));
scn->scn_done_txg = tx->tx_txg + 1;
if (scn->scn_is_sorted) {
scn->scn_checkpointing = B_TRUE;
scn->scn_clearing = B_TRUE;
}
zfs_dbgmsg("scan complete for %s txg %llu",
spa->spa_name,
(longlong_t)tx->tx_txg);
}
} else if (scn->scn_is_sorted && scn->scn_bytes_pending != 0) {
ASSERT(scn->scn_clearing);
/* need to issue scrubbing IOs from per-vdev queues */
scn->scn_zio_root = zio_root(dp->dp_spa, NULL,
NULL, ZIO_FLAG_CANFAIL);
scan_io_queues_run(scn);
(void) zio_wait(scn->scn_zio_root);
scn->scn_zio_root = NULL;
/* calculate and dprintf the current memory usage */
(void) dsl_scan_should_clear(scn);
dsl_scan_update_stats(scn);
zfs_dbgmsg("scan issued %llu blocks for %s (%llu segs) "
"in %llums (avg_block_size = %llu, avg_seg_size = %llu)",
(longlong_t)scn->scn_zios_this_txg,
spa->spa_name,
(longlong_t)scn->scn_segs_this_txg,
(longlong_t)NSEC2MSEC(gethrtime() -
scn->scn_sync_start_time),
(longlong_t)scn->scn_avg_zio_size_this_txg,
(longlong_t)scn->scn_avg_seg_size_this_txg);
} else if (scn->scn_done_txg != 0 && scn->scn_done_txg <= tx->tx_txg) {
/* Finished with everything. Mark the scrub as complete */
zfs_dbgmsg("scan issuing complete txg %llu for %s",
(longlong_t)tx->tx_txg,
spa->spa_name);
ASSERT3U(scn->scn_done_txg, !=, 0);
ASSERT0(spa->spa_scrub_inflight);
ASSERT0(scn->scn_bytes_pending);
dsl_scan_done(scn, B_TRUE, tx);
sync_type = SYNC_MANDATORY;
}
dsl_scan_sync_state(scn, tx, sync_type);
}
static void
count_block(dsl_scan_t *scn, zfs_all_blkstats_t *zab, const blkptr_t *bp)
{
int i;
/*
* Don't count embedded bp's, since we already did the work of
* scanning these when we scanned the containing block.
*/
if (BP_IS_EMBEDDED(bp))
return;
/*
* Update the spa's stats on how many bytes we have issued.
* Sequential scrubs create a zio for each DVA of the bp. Each
* of these will include all DVAs for repair purposes, but the
* zio code will only try the first one unless there is an issue.
* Therefore, we should only count the first DVA for these IOs.
*/
if (scn->scn_is_sorted) {
atomic_add_64(&scn->scn_dp->dp_spa->spa_scan_pass_issued,
DVA_GET_ASIZE(&bp->blk_dva[0]));
} else {
spa_t *spa = scn->scn_dp->dp_spa;
for (i = 0; i < BP_GET_NDVAS(bp); i++) {
atomic_add_64(&spa->spa_scan_pass_issued,
DVA_GET_ASIZE(&bp->blk_dva[i]));
}
}
/*
* If we resume after a reboot, zab will be NULL; don't record
* incomplete stats in that case.
*/
if (zab == NULL)
return;
mutex_enter(&zab->zab_lock);
for (i = 0; i < 4; i++) {
int l = (i < 2) ? BP_GET_LEVEL(bp) : DN_MAX_LEVELS;
int t = (i & 1) ? BP_GET_TYPE(bp) : DMU_OT_TOTAL;
if (t & DMU_OT_NEWTYPE)
t = DMU_OT_OTHER;
zfs_blkstat_t *zb = &zab->zab_type[l][t];
int equal;
zb->zb_count++;
zb->zb_asize += BP_GET_ASIZE(bp);
zb->zb_lsize += BP_GET_LSIZE(bp);
zb->zb_psize += BP_GET_PSIZE(bp);
zb->zb_gangs += BP_COUNT_GANG(bp);
switch (BP_GET_NDVAS(bp)) {
case 2:
if (DVA_GET_VDEV(&bp->blk_dva[0]) ==
DVA_GET_VDEV(&bp->blk_dva[1]))
zb->zb_ditto_2_of_2_samevdev++;
break;
case 3:
equal = (DVA_GET_VDEV(&bp->blk_dva[0]) ==
DVA_GET_VDEV(&bp->blk_dva[1])) +
(DVA_GET_VDEV(&bp->blk_dva[0]) ==
DVA_GET_VDEV(&bp->blk_dva[2])) +
(DVA_GET_VDEV(&bp->blk_dva[1]) ==
DVA_GET_VDEV(&bp->blk_dva[2]));
if (equal == 1)
zb->zb_ditto_2_of_3_samevdev++;
else if (equal == 3)
zb->zb_ditto_3_of_3_samevdev++;
break;
}
}
mutex_exit(&zab->zab_lock);
}
static void
scan_io_queue_insert_impl(dsl_scan_io_queue_t *queue, scan_io_t *sio)
{
avl_index_t idx;
int64_t asize = SIO_GET_ASIZE(sio);
dsl_scan_t *scn = queue->q_scn;
ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));
if (avl_find(&queue->q_sios_by_addr, sio, &idx) != NULL) {
/* block is already scheduled for reading */
atomic_add_64(&scn->scn_bytes_pending, -asize);
sio_free(sio);
return;
}
avl_insert(&queue->q_sios_by_addr, sio, idx);
queue->q_sio_memused += SIO_GET_MUSED(sio);
range_tree_add(queue->q_exts_by_addr, SIO_GET_OFFSET(sio), asize);
}
/*
* Given all the info we got from our metadata scanning process, we
* construct a scan_io_t and insert it into the scan sorting queue. The
* I/O must already be suitable for us to process. This is controlled
* by dsl_scan_enqueue().
*/
static void
scan_io_queue_insert(dsl_scan_io_queue_t *queue, const blkptr_t *bp, int dva_i,
int zio_flags, const zbookmark_phys_t *zb)
{
dsl_scan_t *scn = queue->q_scn;
scan_io_t *sio = sio_alloc(BP_GET_NDVAS(bp));
ASSERT0(BP_IS_GANG(bp));
ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));
bp2sio(bp, sio, dva_i);
sio->sio_flags = zio_flags;
sio->sio_zb = *zb;
/*
* Increment the bytes pending counter now so that we can't
* get an integer underflow in case the worker processes the
* zio before we get to incrementing this counter.
*/
atomic_add_64(&scn->scn_bytes_pending, SIO_GET_ASIZE(sio));
scan_io_queue_insert_impl(queue, sio);
}
/*
* Given a set of I/O parameters as discovered by the metadata traversal
* process, attempts to place the I/O into the sorted queues (if allowed),
* or immediately executes the I/O.
*/
static void
dsl_scan_enqueue(dsl_pool_t *dp, const blkptr_t *bp, int zio_flags,
const zbookmark_phys_t *zb)
{
spa_t *spa = dp->dp_spa;
ASSERT(!BP_IS_EMBEDDED(bp));
/*
* Gang blocks are hard to issue sequentially, so we just issue them
* here immediately instead of queuing them.
*/
if (!dp->dp_scan->scn_is_sorted || BP_IS_GANG(bp)) {
scan_exec_io(dp, bp, zio_flags, zb, NULL);
return;
}
for (int i = 0; i < BP_GET_NDVAS(bp); i++) {
dva_t dva;
vdev_t *vdev;
dva = bp->blk_dva[i];
vdev = vdev_lookup_top(spa, DVA_GET_VDEV(&dva));
ASSERT(vdev != NULL);
mutex_enter(&vdev->vdev_scan_io_queue_lock);
if (vdev->vdev_scan_io_queue == NULL)
vdev->vdev_scan_io_queue = scan_io_queue_create(vdev);
ASSERT(dp->dp_scan != NULL);
scan_io_queue_insert(vdev->vdev_scan_io_queue, bp,
i, zio_flags, zb);
mutex_exit(&vdev->vdev_scan_io_queue_lock);
}
}
static int
dsl_scan_scrub_cb(dsl_pool_t *dp,
const blkptr_t *bp, const zbookmark_phys_t *zb)
{
dsl_scan_t *scn = dp->dp_scan;
spa_t *spa = dp->dp_spa;
uint64_t phys_birth = BP_PHYSICAL_BIRTH(bp);
size_t psize = BP_GET_PSIZE(bp);
boolean_t needs_io = B_FALSE;
int zio_flags = ZIO_FLAG_SCAN_THREAD | ZIO_FLAG_RAW | ZIO_FLAG_CANFAIL;
if (phys_birth <= scn->scn_phys.scn_min_txg ||
phys_birth >= scn->scn_phys.scn_max_txg) {
count_block(scn, dp->dp_blkstats, bp);
return (0);
}
/* Embedded BP's have phys_birth==0, so we reject them above. */
ASSERT(!BP_IS_EMBEDDED(bp));
ASSERT(DSL_SCAN_IS_SCRUB_RESILVER(scn));
if (scn->scn_phys.scn_func == POOL_SCAN_SCRUB) {
zio_flags |= ZIO_FLAG_SCRUB;
needs_io = B_TRUE;
} else {
ASSERT3U(scn->scn_phys.scn_func, ==, POOL_SCAN_RESILVER);
zio_flags |= ZIO_FLAG_RESILVER;
needs_io = B_FALSE;
}
/* If it's an intent log block, failure is expected. */
if (zb->zb_level == ZB_ZIL_LEVEL)
zio_flags |= ZIO_FLAG_SPECULATIVE;
for (int d = 0; d < BP_GET_NDVAS(bp); d++) {
const dva_t *dva = &bp->blk_dva[d];
/*
* Keep track of how much data we've examined so that
* zpool(8) status can make useful progress reports.
*/
scn->scn_phys.scn_examined += DVA_GET_ASIZE(dva);
spa->spa_scan_pass_exam += DVA_GET_ASIZE(dva);
/* if it's a resilver, this may not be in the target range */
if (!needs_io)
needs_io = dsl_scan_need_resilver(spa, dva, psize,
phys_birth);
}
if (needs_io && !zfs_no_scrub_io) {
dsl_scan_enqueue(dp, bp, zio_flags, zb);
} else {
count_block(scn, dp->dp_blkstats, bp);
}
/* do not relocate this block */
return (0);
}
static void
dsl_scan_scrub_done(zio_t *zio)
{
spa_t *spa = zio->io_spa;
blkptr_t *bp = zio->io_bp;
dsl_scan_io_queue_t *queue = zio->io_private;
abd_free(zio->io_abd);
if (queue == NULL) {
mutex_enter(&spa->spa_scrub_lock);
ASSERT3U(spa->spa_scrub_inflight, >=, BP_GET_PSIZE(bp));
spa->spa_scrub_inflight -= BP_GET_PSIZE(bp);
cv_broadcast(&spa->spa_scrub_io_cv);
mutex_exit(&spa->spa_scrub_lock);
} else {
mutex_enter(&queue->q_vd->vdev_scan_io_queue_lock);
ASSERT3U(queue->q_inflight_bytes, >=, BP_GET_PSIZE(bp));
queue->q_inflight_bytes -= BP_GET_PSIZE(bp);
cv_broadcast(&queue->q_zio_cv);
mutex_exit(&queue->q_vd->vdev_scan_io_queue_lock);
}
if (zio->io_error && (zio->io_error != ECKSUM ||
!(zio->io_flags & ZIO_FLAG_SPECULATIVE))) {
atomic_inc_64(&spa->spa_dsl_pool->dp_scan->scn_phys.scn_errors);
}
}
/*
* Given a scanning zio's information, executes the zio. The zio need
* not necessarily be only sortable, this function simply executes the
* zio, no matter what it is. The optional queue argument allows the
* caller to specify that they want per top level vdev IO rate limiting
* instead of the legacy global limiting.
*/
static void
scan_exec_io(dsl_pool_t *dp, const blkptr_t *bp, int zio_flags,
const zbookmark_phys_t *zb, dsl_scan_io_queue_t *queue)
{
spa_t *spa = dp->dp_spa;
dsl_scan_t *scn = dp->dp_scan;
size_t size = BP_GET_PSIZE(bp);
abd_t *data = abd_alloc_for_io(size, B_FALSE);
if (queue == NULL) {
ASSERT3U(scn->scn_maxinflight_bytes, >, 0);
mutex_enter(&spa->spa_scrub_lock);
while (spa->spa_scrub_inflight >= scn->scn_maxinflight_bytes)
cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
spa->spa_scrub_inflight += BP_GET_PSIZE(bp);
mutex_exit(&spa->spa_scrub_lock);
} else {
kmutex_t *q_lock = &queue->q_vd->vdev_scan_io_queue_lock;
ASSERT3U(queue->q_maxinflight_bytes, >, 0);
mutex_enter(q_lock);
while (queue->q_inflight_bytes >= queue->q_maxinflight_bytes)
cv_wait(&queue->q_zio_cv, q_lock);
queue->q_inflight_bytes += BP_GET_PSIZE(bp);
mutex_exit(q_lock);
}
count_block(scn, dp->dp_blkstats, bp);
zio_nowait(zio_read(scn->scn_zio_root, spa, bp, data, size,
dsl_scan_scrub_done, queue, ZIO_PRIORITY_SCRUB, zio_flags, zb));
}
/*
* This is the primary extent sorting algorithm. We balance two parameters:
* 1) how many bytes of I/O are in an extent
* 2) how well the extent is filled with I/O (as a fraction of its total size)
* Since we allow extents to have gaps between their constituent I/Os, it's
* possible to have a fairly large extent that contains the same amount of
* I/O bytes than a much smaller extent, which just packs the I/O more tightly.
* The algorithm sorts based on a score calculated from the extent's size,
* the relative fill volume (in %) and a "fill weight" parameter that controls
* the split between whether we prefer larger extents or more well populated
* extents:
*
* SCORE = FILL_IN_BYTES + (FILL_IN_PERCENT * FILL_IN_BYTES * FILL_WEIGHT)
*
* Example:
* 1) assume extsz = 64 MiB
* 2) assume fill = 32 MiB (extent is half full)
* 3) assume fill_weight = 3
* 4) SCORE = 32M + (((32M * 100) / 64M) * 3 * 32M) / 100
* SCORE = 32M + (50 * 3 * 32M) / 100
* SCORE = 32M + (4800M / 100)
* SCORE = 32M + 48M
* ^ ^
* | +--- final total relative fill-based score
* +--------- final total fill-based score
* SCORE = 80M
*
* As can be seen, at fill_ratio=3, the algorithm is slightly biased towards
* extents that are more completely filled (in a 3:2 ratio) vs just larger.
* Note that as an optimization, we replace multiplication and division by
* 100 with bitshifting by 7 (which effectively multiplies and divides by 128).
*/
static int
ext_size_compare(const void *x, const void *y)
{
const range_seg_gap_t *rsa = x, *rsb = y;
uint64_t sa = rsa->rs_end - rsa->rs_start;
uint64_t sb = rsb->rs_end - rsb->rs_start;
uint64_t score_a, score_b;
score_a = rsa->rs_fill + ((((rsa->rs_fill << 7) / sa) *
fill_weight * rsa->rs_fill) >> 7);
score_b = rsb->rs_fill + ((((rsb->rs_fill << 7) / sb) *
fill_weight * rsb->rs_fill) >> 7);
if (score_a > score_b)
return (-1);
if (score_a == score_b) {
if (rsa->rs_start < rsb->rs_start)
return (-1);
if (rsa->rs_start == rsb->rs_start)
return (0);
return (1);
}
return (1);
}
/*
* Comparator for the q_sios_by_addr tree. Sorting is simply performed
* based on LBA-order (from lowest to highest).
*/
static int
sio_addr_compare(const void *x, const void *y)
{
const scan_io_t *a = x, *b = y;
return (TREE_CMP(SIO_GET_OFFSET(a), SIO_GET_OFFSET(b)));
}
/* IO queues are created on demand when they are needed. */
static dsl_scan_io_queue_t *
scan_io_queue_create(vdev_t *vd)
{
dsl_scan_t *scn = vd->vdev_spa->spa_dsl_pool->dp_scan;
dsl_scan_io_queue_t *q = kmem_zalloc(sizeof (*q), KM_SLEEP);
q->q_scn = scn;
q->q_vd = vd;
q->q_sio_memused = 0;
cv_init(&q->q_zio_cv, NULL, CV_DEFAULT, NULL);
q->q_exts_by_addr = range_tree_create_impl(&rt_btree_ops, RANGE_SEG_GAP,
&q->q_exts_by_size, 0, 0, ext_size_compare, zfs_scan_max_ext_gap);
avl_create(&q->q_sios_by_addr, sio_addr_compare,
sizeof (scan_io_t), offsetof(scan_io_t, sio_nodes.sio_addr_node));
return (q);
}
/*
* Destroys a scan queue and all segments and scan_io_t's contained in it.
* No further execution of I/O occurs, anything pending in the queue is
* simply freed without being executed.
*/
void
dsl_scan_io_queue_destroy(dsl_scan_io_queue_t *queue)
{
dsl_scan_t *scn = queue->q_scn;
scan_io_t *sio;
void *cookie = NULL;
int64_t bytes_dequeued = 0;
ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));
while ((sio = avl_destroy_nodes(&queue->q_sios_by_addr, &cookie)) !=
NULL) {
ASSERT(range_tree_contains(queue->q_exts_by_addr,
SIO_GET_OFFSET(sio), SIO_GET_ASIZE(sio)));
bytes_dequeued += SIO_GET_ASIZE(sio);
queue->q_sio_memused -= SIO_GET_MUSED(sio);
sio_free(sio);
}
ASSERT0(queue->q_sio_memused);
atomic_add_64(&scn->scn_bytes_pending, -bytes_dequeued);
range_tree_vacate(queue->q_exts_by_addr, NULL, queue);
range_tree_destroy(queue->q_exts_by_addr);
avl_destroy(&queue->q_sios_by_addr);
cv_destroy(&queue->q_zio_cv);
kmem_free(queue, sizeof (*queue));
}
/*
* Properly transfers a dsl_scan_queue_t from `svd' to `tvd'. This is
* called on behalf of vdev_top_transfer when creating or destroying
* a mirror vdev due to zpool attach/detach.
*/
void
dsl_scan_io_queue_vdev_xfer(vdev_t *svd, vdev_t *tvd)
{
mutex_enter(&svd->vdev_scan_io_queue_lock);
mutex_enter(&tvd->vdev_scan_io_queue_lock);
VERIFY3P(tvd->vdev_scan_io_queue, ==, NULL);
tvd->vdev_scan_io_queue = svd->vdev_scan_io_queue;
svd->vdev_scan_io_queue = NULL;
if (tvd->vdev_scan_io_queue != NULL)
tvd->vdev_scan_io_queue->q_vd = tvd;
mutex_exit(&tvd->vdev_scan_io_queue_lock);
mutex_exit(&svd->vdev_scan_io_queue_lock);
}
static void
scan_io_queues_destroy(dsl_scan_t *scn)
{
vdev_t *rvd = scn->scn_dp->dp_spa->spa_root_vdev;
for (uint64_t i = 0; i < rvd->vdev_children; i++) {
vdev_t *tvd = rvd->vdev_child[i];
mutex_enter(&tvd->vdev_scan_io_queue_lock);
if (tvd->vdev_scan_io_queue != NULL)
dsl_scan_io_queue_destroy(tvd->vdev_scan_io_queue);
tvd->vdev_scan_io_queue = NULL;
mutex_exit(&tvd->vdev_scan_io_queue_lock);
}
}
static void
dsl_scan_freed_dva(spa_t *spa, const blkptr_t *bp, int dva_i)
{
dsl_pool_t *dp = spa->spa_dsl_pool;
dsl_scan_t *scn = dp->dp_scan;
vdev_t *vdev;
kmutex_t *q_lock;
dsl_scan_io_queue_t *queue;
scan_io_t *srch_sio, *sio;
avl_index_t idx;
uint64_t start, size;
vdev = vdev_lookup_top(spa, DVA_GET_VDEV(&bp->blk_dva[dva_i]));
ASSERT(vdev != NULL);
q_lock = &vdev->vdev_scan_io_queue_lock;
queue = vdev->vdev_scan_io_queue;
mutex_enter(q_lock);
if (queue == NULL) {
mutex_exit(q_lock);
return;
}
srch_sio = sio_alloc(BP_GET_NDVAS(bp));
bp2sio(bp, srch_sio, dva_i);
start = SIO_GET_OFFSET(srch_sio);
size = SIO_GET_ASIZE(srch_sio);
/*
* We can find the zio in two states:
* 1) Cold, just sitting in the queue of zio's to be issued at
* some point in the future. In this case, all we do is
* remove the zio from the q_sios_by_addr tree, decrement
* its data volume from the containing range_seg_t and
* resort the q_exts_by_size tree to reflect that the
* range_seg_t has lost some of its 'fill'. We don't shorten
* the range_seg_t - this is usually rare enough not to be
* worth the extra hassle of trying keep track of precise
* extent boundaries.
* 2) Hot, where the zio is currently in-flight in
* dsl_scan_issue_ios. In this case, we can't simply
* reach in and stop the in-flight zio's, so we instead
* block the caller. Eventually, dsl_scan_issue_ios will
* be done with issuing the zio's it gathered and will
* signal us.
*/
sio = avl_find(&queue->q_sios_by_addr, srch_sio, &idx);
sio_free(srch_sio);
if (sio != NULL) {
int64_t asize = SIO_GET_ASIZE(sio);
blkptr_t tmpbp;
/* Got it while it was cold in the queue */
ASSERT3U(start, ==, SIO_GET_OFFSET(sio));
ASSERT3U(size, ==, asize);
avl_remove(&queue->q_sios_by_addr, sio);
queue->q_sio_memused -= SIO_GET_MUSED(sio);
ASSERT(range_tree_contains(queue->q_exts_by_addr, start, size));
range_tree_remove_fill(queue->q_exts_by_addr, start, size);
/*
* We only update scn_bytes_pending in the cold path,
* otherwise it will already have been accounted for as
* part of the zio's execution.
*/
atomic_add_64(&scn->scn_bytes_pending, -asize);
/* count the block as though we issued it */
sio2bp(sio, &tmpbp);
count_block(scn, dp->dp_blkstats, &tmpbp);
sio_free(sio);
}
mutex_exit(q_lock);
}
/*
* Callback invoked when a zio_free() zio is executing. This needs to be
* intercepted to prevent the zio from deallocating a particular portion
* of disk space and it then getting reallocated and written to, while we
* still have it queued up for processing.
*/
void
dsl_scan_freed(spa_t *spa, const blkptr_t *bp)
{
dsl_pool_t *dp = spa->spa_dsl_pool;
dsl_scan_t *scn = dp->dp_scan;
ASSERT(!BP_IS_EMBEDDED(bp));
ASSERT(scn != NULL);
if (!dsl_scan_is_running(scn))
return;
for (int i = 0; i < BP_GET_NDVAS(bp); i++)
dsl_scan_freed_dva(spa, bp, i);
}
/*
* Check if a vdev needs resilvering (non-empty DTL), if so, and resilver has
* not started, start it. Otherwise, only restart if max txg in DTL range is
* greater than the max txg in the current scan. If the DTL max is less than
* the scan max, then the vdev has not missed any new data since the resilver
* started, so a restart is not needed.
*/
void
dsl_scan_assess_vdev(dsl_pool_t *dp, vdev_t *vd)
{
uint64_t min, max;
if (!vdev_resilver_needed(vd, &min, &max))
return;
if (!dsl_scan_resilvering(dp)) {
spa_async_request(dp->dp_spa, SPA_ASYNC_RESILVER);
return;
}
if (max <= dp->dp_scan->scn_phys.scn_max_txg)
return;
/* restart is needed, check if it can be deferred */
if (spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_RESILVER_DEFER))
vdev_defer_resilver(vd);
else
spa_async_request(dp->dp_spa, SPA_ASYNC_RESILVER);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, scan_vdev_limit, ULONG, ZMOD_RW,
"Max bytes in flight per leaf vdev for scrubs and resilvers");
ZFS_MODULE_PARAM(zfs, zfs_, scrub_min_time_ms, INT, ZMOD_RW,
"Min millisecs to scrub per txg");
ZFS_MODULE_PARAM(zfs, zfs_, obsolete_min_time_ms, INT, ZMOD_RW,
"Min millisecs to obsolete per txg");
ZFS_MODULE_PARAM(zfs, zfs_, free_min_time_ms, INT, ZMOD_RW,
"Min millisecs to free per txg");
ZFS_MODULE_PARAM(zfs, zfs_, resilver_min_time_ms, INT, ZMOD_RW,
"Min millisecs to resilver per txg");
ZFS_MODULE_PARAM(zfs, zfs_, scan_suspend_progress, INT, ZMOD_RW,
"Set to prevent scans from progressing");
ZFS_MODULE_PARAM(zfs, zfs_, no_scrub_io, INT, ZMOD_RW,
"Set to disable scrub I/O");
ZFS_MODULE_PARAM(zfs, zfs_, no_scrub_prefetch, INT, ZMOD_RW,
"Set to disable scrub prefetching");
ZFS_MODULE_PARAM(zfs, zfs_, async_block_max_blocks, ULONG, ZMOD_RW,
"Max number of blocks freed in one txg");
ZFS_MODULE_PARAM(zfs, zfs_, max_async_dedup_frees, ULONG, ZMOD_RW,
"Max number of dedup blocks freed in one txg");
ZFS_MODULE_PARAM(zfs, zfs_, free_bpobj_enabled, INT, ZMOD_RW,
"Enable processing of the free_bpobj");
ZFS_MODULE_PARAM(zfs, zfs_, scan_mem_lim_fact, INT, ZMOD_RW,
"Fraction of RAM for scan hard limit");
ZFS_MODULE_PARAM(zfs, zfs_, scan_issue_strategy, INT, ZMOD_RW,
"IO issuing strategy during scrubbing. "
"0 = default, 1 = LBA, 2 = size");
ZFS_MODULE_PARAM(zfs, zfs_, scan_legacy, INT, ZMOD_RW,
"Scrub using legacy non-sequential method");
ZFS_MODULE_PARAM(zfs, zfs_, scan_checkpoint_intval, INT, ZMOD_RW,
"Scan progress on-disk checkpointing interval");
ZFS_MODULE_PARAM(zfs, zfs_, scan_max_ext_gap, ULONG, ZMOD_RW,
"Max gap in bytes between sequential scrub / resilver I/Os");
ZFS_MODULE_PARAM(zfs, zfs_, scan_mem_lim_soft_fact, INT, ZMOD_RW,
"Fraction of hard limit used as soft limit");
ZFS_MODULE_PARAM(zfs, zfs_, scan_strict_mem_lim, INT, ZMOD_RW,
"Tunable to attempt to reduce lock contention");
ZFS_MODULE_PARAM(zfs, zfs_, scan_fill_weight, INT, ZMOD_RW,
"Tunable to adjust bias towards more filled segments during scans");
ZFS_MODULE_PARAM(zfs, zfs_, resilver_disable_defer, INT, ZMOD_RW,
"Process all resilvers immediately");
/* END CSTYLED */
diff --git a/module/zfs/fm.c b/module/zfs/fm.c
index f4b001b2787d..5a52a881c687 100644
--- a/module/zfs/fm.c
+++ b/module/zfs/fm.c
@@ -1,1374 +1,1374 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
*/
/*
* Fault Management Architecture (FMA) Resource and Protocol Support
*
* The routines contained herein provide services to support kernel subsystems
* in publishing fault management telemetry (see PSARC 2002/412 and 2003/089).
*
* Name-Value Pair Lists
*
* The embodiment of an FMA protocol element (event, fmri or authority) is a
* name-value pair list (nvlist_t). FMA-specific nvlist constructor and
* destructor functions, fm_nvlist_create() and fm_nvlist_destroy(), are used
* to create an nvpair list using custom allocators. Callers may choose to
* allocate either from the kernel memory allocator, or from a preallocated
* buffer, useful in constrained contexts like high-level interrupt routines.
*
* Protocol Event and FMRI Construction
*
* Convenience routines are provided to construct nvlist events according to
* the FMA Event Protocol and Naming Schema specification for ereports and
* FMRIs for the dev, cpu, hc, mem, legacy hc and de schemes.
*
* ENA Manipulation
*
* Routines to generate ENA formats 0, 1 and 2 are available as well as
* routines to increment formats 1 and 2. Individual fields within the
* ENA are extractable via fm_ena_time_get(), fm_ena_id_get(),
* fm_ena_format_get() and fm_ena_gen_get().
*/
#include <sys/types.h>
#include <sys/time.h>
#include <sys/list.h>
#include <sys/nvpair.h>
#include <sys/cmn_err.h>
#include <sys/sysmacros.h>
#include <sys/sunddi.h>
#include <sys/systeminfo.h>
#include <sys/fm/util.h>
#include <sys/fm/protocol.h>
#include <sys/kstat.h>
#include <sys/zfs_context.h>
#ifdef _KERNEL
#include <sys/atomic.h>
#include <sys/condvar.h>
#include <sys/zfs_ioctl.h>
-int zfs_zevent_len_max = 512;
+static int zfs_zevent_len_max = 512;
static int zevent_len_cur = 0;
static int zevent_waiters = 0;
static int zevent_flags = 0;
/* Num events rate limited since the last time zfs_zevent_next() was called */
static uint64_t ratelimit_dropped = 0;
/*
* The EID (Event IDentifier) is used to uniquely tag a zevent when it is
* posted. The posted EIDs are monotonically increasing but not persistent.
* They will be reset to the initial value (1) each time the kernel module is
* loaded.
*/
static uint64_t zevent_eid = 0;
static kmutex_t zevent_lock;
static list_t zevent_list;
static kcondvar_t zevent_cv;
#endif /* _KERNEL */
/*
* Common fault management kstats to record event generation failures
*/
struct erpt_kstat {
kstat_named_t erpt_dropped; /* num erpts dropped on post */
kstat_named_t erpt_set_failed; /* num erpt set failures */
kstat_named_t fmri_set_failed; /* num fmri set failures */
kstat_named_t payload_set_failed; /* num payload set failures */
kstat_named_t erpt_duplicates; /* num duplicate erpts */
};
static struct erpt_kstat erpt_kstat_data = {
{ "erpt-dropped", KSTAT_DATA_UINT64 },
{ "erpt-set-failed", KSTAT_DATA_UINT64 },
{ "fmri-set-failed", KSTAT_DATA_UINT64 },
{ "payload-set-failed", KSTAT_DATA_UINT64 },
{ "erpt-duplicates", KSTAT_DATA_UINT64 }
};
kstat_t *fm_ksp;
#ifdef _KERNEL
static zevent_t *
zfs_zevent_alloc(void)
{
zevent_t *ev;
ev = kmem_zalloc(sizeof (zevent_t), KM_SLEEP);
list_create(&ev->ev_ze_list, sizeof (zfs_zevent_t),
offsetof(zfs_zevent_t, ze_node));
list_link_init(&ev->ev_node);
return (ev);
}
static void
zfs_zevent_free(zevent_t *ev)
{
/* Run provided cleanup callback */
ev->ev_cb(ev->ev_nvl, ev->ev_detector);
list_destroy(&ev->ev_ze_list);
kmem_free(ev, sizeof (zevent_t));
}
static void
zfs_zevent_drain(zevent_t *ev)
{
zfs_zevent_t *ze;
ASSERT(MUTEX_HELD(&zevent_lock));
list_remove(&zevent_list, ev);
/* Remove references to this event in all private file data */
while ((ze = list_head(&ev->ev_ze_list)) != NULL) {
list_remove(&ev->ev_ze_list, ze);
ze->ze_zevent = NULL;
ze->ze_dropped++;
}
zfs_zevent_free(ev);
}
void
zfs_zevent_drain_all(int *count)
{
zevent_t *ev;
mutex_enter(&zevent_lock);
while ((ev = list_head(&zevent_list)) != NULL)
zfs_zevent_drain(ev);
*count = zevent_len_cur;
zevent_len_cur = 0;
mutex_exit(&zevent_lock);
}
/*
* New zevents are inserted at the head. If the maximum queue
* length is exceeded a zevent will be drained from the tail.
* As part of this any user space processes which currently have
* a reference to this zevent_t in their private data will have
* this reference set to NULL.
*/
static void
zfs_zevent_insert(zevent_t *ev)
{
ASSERT(MUTEX_HELD(&zevent_lock));
list_insert_head(&zevent_list, ev);
if (zevent_len_cur >= zfs_zevent_len_max)
zfs_zevent_drain(list_tail(&zevent_list));
else
zevent_len_cur++;
}
/*
* Post a zevent. The cb will be called when nvl and detector are no longer
* needed, i.e.:
* - An error happened and a zevent can't be posted. In this case, cb is called
* before zfs_zevent_post() returns.
* - The event is being drained and freed.
*/
int
zfs_zevent_post(nvlist_t *nvl, nvlist_t *detector, zevent_cb_t *cb)
{
inode_timespec_t tv;
int64_t tv_array[2];
uint64_t eid;
size_t nvl_size = 0;
zevent_t *ev;
int error;
ASSERT(cb != NULL);
gethrestime(&tv);
tv_array[0] = tv.tv_sec;
tv_array[1] = tv.tv_nsec;
error = nvlist_add_int64_array(nvl, FM_EREPORT_TIME, tv_array, 2);
if (error) {
atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
goto out;
}
eid = atomic_inc_64_nv(&zevent_eid);
error = nvlist_add_uint64(nvl, FM_EREPORT_EID, eid);
if (error) {
atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
goto out;
}
error = nvlist_size(nvl, &nvl_size, NV_ENCODE_NATIVE);
if (error) {
atomic_inc_64(&erpt_kstat_data.erpt_dropped.value.ui64);
goto out;
}
if (nvl_size > ERPT_DATA_SZ || nvl_size == 0) {
atomic_inc_64(&erpt_kstat_data.erpt_dropped.value.ui64);
error = EOVERFLOW;
goto out;
}
ev = zfs_zevent_alloc();
if (ev == NULL) {
atomic_inc_64(&erpt_kstat_data.erpt_dropped.value.ui64);
error = ENOMEM;
goto out;
}
ev->ev_nvl = nvl;
ev->ev_detector = detector;
ev->ev_cb = cb;
ev->ev_eid = eid;
mutex_enter(&zevent_lock);
zfs_zevent_insert(ev);
cv_broadcast(&zevent_cv);
mutex_exit(&zevent_lock);
out:
if (error)
cb(nvl, detector);
return (error);
}
void
zfs_zevent_track_duplicate(void)
{
atomic_inc_64(&erpt_kstat_data.erpt_duplicates.value.ui64);
}
static int
zfs_zevent_minor_to_state(minor_t minor, zfs_zevent_t **ze)
{
*ze = zfsdev_get_state(minor, ZST_ZEVENT);
if (*ze == NULL)
return (SET_ERROR(EBADF));
return (0);
}
zfs_file_t *
zfs_zevent_fd_hold(int fd, minor_t *minorp, zfs_zevent_t **ze)
{
zfs_file_t *fp = zfs_file_get(fd);
if (fp == NULL)
return (NULL);
int error = zfsdev_getminor(fp, minorp);
if (error == 0)
error = zfs_zevent_minor_to_state(*minorp, ze);
if (error) {
zfs_zevent_fd_rele(fp);
fp = NULL;
}
return (fp);
}
void
zfs_zevent_fd_rele(zfs_file_t *fp)
{
zfs_file_put(fp);
}
/*
* Get the next zevent in the stream and place a copy in 'event'. This
* may fail with ENOMEM if the encoded nvlist size exceeds the passed
* 'event_size'. In this case the stream pointer is not advanced and
* and 'event_size' is set to the minimum required buffer size.
*/
int
zfs_zevent_next(zfs_zevent_t *ze, nvlist_t **event, uint64_t *event_size,
uint64_t *dropped)
{
zevent_t *ev;
size_t size;
int error = 0;
mutex_enter(&zevent_lock);
if (ze->ze_zevent == NULL) {
/* New stream start at the beginning/tail */
ev = list_tail(&zevent_list);
if (ev == NULL) {
error = ENOENT;
goto out;
}
} else {
/*
* Existing stream continue with the next element and remove
* ourselves from the wait queue for the previous element
*/
ev = list_prev(&zevent_list, ze->ze_zevent);
if (ev == NULL) {
error = ENOENT;
goto out;
}
}
VERIFY(nvlist_size(ev->ev_nvl, &size, NV_ENCODE_NATIVE) == 0);
if (size > *event_size) {
*event_size = size;
error = ENOMEM;
goto out;
}
if (ze->ze_zevent)
list_remove(&ze->ze_zevent->ev_ze_list, ze);
ze->ze_zevent = ev;
list_insert_head(&ev->ev_ze_list, ze);
(void) nvlist_dup(ev->ev_nvl, event, KM_SLEEP);
*dropped = ze->ze_dropped;
#ifdef _KERNEL
/* Include events dropped due to rate limiting */
*dropped += atomic_swap_64(&ratelimit_dropped, 0);
#endif
ze->ze_dropped = 0;
out:
mutex_exit(&zevent_lock);
return (error);
}
/*
* Wait in an interruptible state for any new events.
*/
int
zfs_zevent_wait(zfs_zevent_t *ze)
{
int error = EAGAIN;
mutex_enter(&zevent_lock);
zevent_waiters++;
while (error == EAGAIN) {
if (zevent_flags & ZEVENT_SHUTDOWN) {
error = SET_ERROR(ESHUTDOWN);
break;
}
error = cv_wait_sig(&zevent_cv, &zevent_lock);
if (signal_pending(current)) {
error = SET_ERROR(EINTR);
break;
} else if (!list_is_empty(&zevent_list)) {
error = 0;
continue;
} else {
error = EAGAIN;
}
}
zevent_waiters--;
mutex_exit(&zevent_lock);
return (error);
}
/*
* The caller may seek to a specific EID by passing that EID. If the EID
* is still available in the posted list of events the cursor is positioned
* there. Otherwise ENOENT is returned and the cursor is not moved.
*
* There are two reserved EIDs which may be passed and will never fail.
* ZEVENT_SEEK_START positions the cursor at the start of the list, and
* ZEVENT_SEEK_END positions the cursor at the end of the list.
*/
int
zfs_zevent_seek(zfs_zevent_t *ze, uint64_t eid)
{
zevent_t *ev;
int error = 0;
mutex_enter(&zevent_lock);
if (eid == ZEVENT_SEEK_START) {
if (ze->ze_zevent)
list_remove(&ze->ze_zevent->ev_ze_list, ze);
ze->ze_zevent = NULL;
goto out;
}
if (eid == ZEVENT_SEEK_END) {
if (ze->ze_zevent)
list_remove(&ze->ze_zevent->ev_ze_list, ze);
ev = list_head(&zevent_list);
if (ev) {
ze->ze_zevent = ev;
list_insert_head(&ev->ev_ze_list, ze);
} else {
ze->ze_zevent = NULL;
}
goto out;
}
for (ev = list_tail(&zevent_list); ev != NULL;
ev = list_prev(&zevent_list, ev)) {
if (ev->ev_eid == eid) {
if (ze->ze_zevent)
list_remove(&ze->ze_zevent->ev_ze_list, ze);
ze->ze_zevent = ev;
list_insert_head(&ev->ev_ze_list, ze);
break;
}
}
if (ev == NULL)
error = ENOENT;
out:
mutex_exit(&zevent_lock);
return (error);
}
void
zfs_zevent_init(zfs_zevent_t **zep)
{
zfs_zevent_t *ze;
ze = *zep = kmem_zalloc(sizeof (zfs_zevent_t), KM_SLEEP);
list_link_init(&ze->ze_node);
}
void
zfs_zevent_destroy(zfs_zevent_t *ze)
{
mutex_enter(&zevent_lock);
if (ze->ze_zevent)
list_remove(&ze->ze_zevent->ev_ze_list, ze);
mutex_exit(&zevent_lock);
kmem_free(ze, sizeof (zfs_zevent_t));
}
#endif /* _KERNEL */
/*
* Wrappers for FM nvlist allocators
*/
static void *
i_fm_alloc(nv_alloc_t *nva, size_t size)
{
(void) nva;
return (kmem_zalloc(size, KM_SLEEP));
}
static void
i_fm_free(nv_alloc_t *nva, void *buf, size_t size)
{
(void) nva;
kmem_free(buf, size);
}
-const nv_alloc_ops_t fm_mem_alloc_ops = {
+static const nv_alloc_ops_t fm_mem_alloc_ops = {
.nv_ao_init = NULL,
.nv_ao_fini = NULL,
.nv_ao_alloc = i_fm_alloc,
.nv_ao_free = i_fm_free,
.nv_ao_reset = NULL
};
/*
* Create and initialize a new nv_alloc_t for a fixed buffer, buf. A pointer
* to the newly allocated nv_alloc_t structure is returned upon success or NULL
* is returned to indicate that the nv_alloc structure could not be created.
*/
nv_alloc_t *
fm_nva_xcreate(char *buf, size_t bufsz)
{
nv_alloc_t *nvhdl = kmem_zalloc(sizeof (nv_alloc_t), KM_SLEEP);
if (bufsz == 0 || nv_alloc_init(nvhdl, nv_fixed_ops, buf, bufsz) != 0) {
kmem_free(nvhdl, sizeof (nv_alloc_t));
return (NULL);
}
return (nvhdl);
}
/*
* Destroy a previously allocated nv_alloc structure. The fixed buffer
* associated with nva must be freed by the caller.
*/
void
fm_nva_xdestroy(nv_alloc_t *nva)
{
nv_alloc_fini(nva);
kmem_free(nva, sizeof (nv_alloc_t));
}
/*
* Create a new nv list. A pointer to a new nv list structure is returned
* upon success or NULL is returned to indicate that the structure could
* not be created. The newly created nv list is created and managed by the
* operations installed in nva. If nva is NULL, the default FMA nva
* operations are installed and used.
*
* When called from the kernel and nva == NULL, this function must be called
* from passive kernel context with no locks held that can prevent a
* sleeping memory allocation from occurring. Otherwise, this function may
* be called from other kernel contexts as long a valid nva created via
* fm_nva_create() is supplied.
*/
nvlist_t *
fm_nvlist_create(nv_alloc_t *nva)
{
int hdl_alloced = 0;
nvlist_t *nvl;
nv_alloc_t *nvhdl;
if (nva == NULL) {
nvhdl = kmem_zalloc(sizeof (nv_alloc_t), KM_SLEEP);
if (nv_alloc_init(nvhdl, &fm_mem_alloc_ops, NULL, 0) != 0) {
kmem_free(nvhdl, sizeof (nv_alloc_t));
return (NULL);
}
hdl_alloced = 1;
} else {
nvhdl = nva;
}
if (nvlist_xalloc(&nvl, NV_UNIQUE_NAME, nvhdl) != 0) {
if (hdl_alloced) {
nv_alloc_fini(nvhdl);
kmem_free(nvhdl, sizeof (nv_alloc_t));
}
return (NULL);
}
return (nvl);
}
/*
* Destroy a previously allocated nvlist structure. flag indicates whether
* or not the associated nva structure should be freed (FM_NVA_FREE) or
* retained (FM_NVA_RETAIN). Retaining the nv alloc structure allows
* it to be re-used for future nvlist creation operations.
*/
void
fm_nvlist_destroy(nvlist_t *nvl, int flag)
{
nv_alloc_t *nva = nvlist_lookup_nv_alloc(nvl);
nvlist_free(nvl);
if (nva != NULL) {
if (flag == FM_NVA_FREE)
fm_nva_xdestroy(nva);
}
}
int
i_fm_payload_set(nvlist_t *payload, const char *name, va_list ap)
{
int nelem, ret = 0;
data_type_t type;
while (ret == 0 && name != NULL) {
type = va_arg(ap, data_type_t);
switch (type) {
case DATA_TYPE_BYTE:
ret = nvlist_add_byte(payload, name,
va_arg(ap, uint_t));
break;
case DATA_TYPE_BYTE_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_byte_array(payload, name,
va_arg(ap, uchar_t *), nelem);
break;
case DATA_TYPE_BOOLEAN_VALUE:
ret = nvlist_add_boolean_value(payload, name,
va_arg(ap, boolean_t));
break;
case DATA_TYPE_BOOLEAN_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_boolean_array(payload, name,
va_arg(ap, boolean_t *), nelem);
break;
case DATA_TYPE_INT8:
ret = nvlist_add_int8(payload, name,
va_arg(ap, int));
break;
case DATA_TYPE_INT8_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_int8_array(payload, name,
va_arg(ap, int8_t *), nelem);
break;
case DATA_TYPE_UINT8:
ret = nvlist_add_uint8(payload, name,
va_arg(ap, uint_t));
break;
case DATA_TYPE_UINT8_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_uint8_array(payload, name,
va_arg(ap, uint8_t *), nelem);
break;
case DATA_TYPE_INT16:
ret = nvlist_add_int16(payload, name,
va_arg(ap, int));
break;
case DATA_TYPE_INT16_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_int16_array(payload, name,
va_arg(ap, int16_t *), nelem);
break;
case DATA_TYPE_UINT16:
ret = nvlist_add_uint16(payload, name,
va_arg(ap, uint_t));
break;
case DATA_TYPE_UINT16_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_uint16_array(payload, name,
va_arg(ap, uint16_t *), nelem);
break;
case DATA_TYPE_INT32:
ret = nvlist_add_int32(payload, name,
va_arg(ap, int32_t));
break;
case DATA_TYPE_INT32_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_int32_array(payload, name,
va_arg(ap, int32_t *), nelem);
break;
case DATA_TYPE_UINT32:
ret = nvlist_add_uint32(payload, name,
va_arg(ap, uint32_t));
break;
case DATA_TYPE_UINT32_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_uint32_array(payload, name,
va_arg(ap, uint32_t *), nelem);
break;
case DATA_TYPE_INT64:
ret = nvlist_add_int64(payload, name,
va_arg(ap, int64_t));
break;
case DATA_TYPE_INT64_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_int64_array(payload, name,
va_arg(ap, int64_t *), nelem);
break;
case DATA_TYPE_UINT64:
ret = nvlist_add_uint64(payload, name,
va_arg(ap, uint64_t));
break;
case DATA_TYPE_UINT64_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_uint64_array(payload, name,
va_arg(ap, uint64_t *), nelem);
break;
case DATA_TYPE_STRING:
ret = nvlist_add_string(payload, name,
va_arg(ap, char *));
break;
case DATA_TYPE_STRING_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_string_array(payload, name,
va_arg(ap, const char **), nelem);
break;
case DATA_TYPE_NVLIST:
ret = nvlist_add_nvlist(payload, name,
va_arg(ap, nvlist_t *));
break;
case DATA_TYPE_NVLIST_ARRAY:
nelem = va_arg(ap, int);
ret = nvlist_add_nvlist_array(payload, name,
va_arg(ap, const nvlist_t **), nelem);
break;
default:
ret = EINVAL;
}
name = va_arg(ap, char *);
}
return (ret);
}
void
fm_payload_set(nvlist_t *payload, ...)
{
int ret;
const char *name;
va_list ap;
va_start(ap, payload);
name = va_arg(ap, char *);
ret = i_fm_payload_set(payload, name, ap);
va_end(ap);
if (ret)
atomic_inc_64(&erpt_kstat_data.payload_set_failed.value.ui64);
}
/*
* Set-up and validate the members of an ereport event according to:
*
* Member name Type Value
* ====================================================
* class string ereport
* version uint8_t 0
* ena uint64_t <ena>
* detector nvlist_t <detector>
* ereport-payload nvlist_t <var args>
*
* We don't actually add a 'version' member to the payload. Really,
* the version quoted to us by our caller is that of the category 1
* "ereport" event class (and we require FM_EREPORT_VERS0) but
* the payload version of the actual leaf class event under construction
* may be something else. Callers should supply a version in the varargs,
* or (better) we could take two version arguments - one for the
* ereport category 1 classification (expect FM_EREPORT_VERS0) and one
* for the leaf class.
*/
void
fm_ereport_set(nvlist_t *ereport, int version, const char *erpt_class,
uint64_t ena, const nvlist_t *detector, ...)
{
char ereport_class[FM_MAX_CLASS];
const char *name;
va_list ap;
int ret;
if (version != FM_EREPORT_VERS0) {
atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
return;
}
(void) snprintf(ereport_class, FM_MAX_CLASS, "%s.%s",
FM_EREPORT_CLASS, erpt_class);
if (nvlist_add_string(ereport, FM_CLASS, ereport_class) != 0) {
atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
return;
}
if (nvlist_add_uint64(ereport, FM_EREPORT_ENA, ena)) {
atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
}
if (nvlist_add_nvlist(ereport, FM_EREPORT_DETECTOR,
(nvlist_t *)detector) != 0) {
atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
}
va_start(ap, detector);
name = va_arg(ap, const char *);
ret = i_fm_payload_set(ereport, name, ap);
va_end(ap);
if (ret)
atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
}
/*
* Set-up and validate the members of an hc fmri according to;
*
* Member name Type Value
* ===================================================
* version uint8_t 0
* auth nvlist_t <auth>
* hc-name string <name>
* hc-id string <id>
*
* Note that auth and hc-id are optional members.
*/
#define HC_MAXPAIRS 20
#define HC_MAXNAMELEN 50
static int
fm_fmri_hc_set_common(nvlist_t *fmri, int version, const nvlist_t *auth)
{
if (version != FM_HC_SCHEME_VERSION) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return (0);
}
if (nvlist_add_uint8(fmri, FM_VERSION, version) != 0 ||
nvlist_add_string(fmri, FM_FMRI_SCHEME, FM_FMRI_SCHEME_HC) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return (0);
}
if (auth != NULL && nvlist_add_nvlist(fmri, FM_FMRI_AUTHORITY,
(nvlist_t *)auth) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return (0);
}
return (1);
}
void
fm_fmri_hc_set(nvlist_t *fmri, int version, const nvlist_t *auth,
nvlist_t *snvl, int npairs, ...)
{
nv_alloc_t *nva = nvlist_lookup_nv_alloc(fmri);
nvlist_t *pairs[HC_MAXPAIRS];
va_list ap;
int i;
if (!fm_fmri_hc_set_common(fmri, version, auth))
return;
npairs = MIN(npairs, HC_MAXPAIRS);
va_start(ap, npairs);
for (i = 0; i < npairs; i++) {
const char *name = va_arg(ap, const char *);
uint32_t id = va_arg(ap, uint32_t);
char idstr[11];
(void) snprintf(idstr, sizeof (idstr), "%u", id);
pairs[i] = fm_nvlist_create(nva);
if (nvlist_add_string(pairs[i], FM_FMRI_HC_NAME, name) != 0 ||
nvlist_add_string(pairs[i], FM_FMRI_HC_ID, idstr) != 0) {
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
}
}
va_end(ap);
if (nvlist_add_nvlist_array(fmri, FM_FMRI_HC_LIST,
(const nvlist_t **)pairs, npairs) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
}
for (i = 0; i < npairs; i++)
fm_nvlist_destroy(pairs[i], FM_NVA_RETAIN);
if (snvl != NULL) {
if (nvlist_add_nvlist(fmri, FM_FMRI_HC_SPECIFIC, snvl) != 0) {
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
}
}
}
void
fm_fmri_hc_create(nvlist_t *fmri, int version, const nvlist_t *auth,
nvlist_t *snvl, nvlist_t *bboard, int npairs, ...)
{
nv_alloc_t *nva = nvlist_lookup_nv_alloc(fmri);
nvlist_t *pairs[HC_MAXPAIRS];
nvlist_t **hcl;
uint_t n;
int i, j;
va_list ap;
char *hcname, *hcid;
if (!fm_fmri_hc_set_common(fmri, version, auth))
return;
/*
* copy the bboard nvpairs to the pairs array
*/
if (nvlist_lookup_nvlist_array(bboard, FM_FMRI_HC_LIST, &hcl, &n)
!= 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
for (i = 0; i < n; i++) {
if (nvlist_lookup_string(hcl[i], FM_FMRI_HC_NAME,
&hcname) != 0) {
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
if (nvlist_lookup_string(hcl[i], FM_FMRI_HC_ID, &hcid) != 0) {
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
pairs[i] = fm_nvlist_create(nva);
if (nvlist_add_string(pairs[i], FM_FMRI_HC_NAME, hcname) != 0 ||
nvlist_add_string(pairs[i], FM_FMRI_HC_ID, hcid) != 0) {
for (j = 0; j <= i; j++) {
if (pairs[j] != NULL)
fm_nvlist_destroy(pairs[j],
FM_NVA_RETAIN);
}
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
}
/*
* create the pairs from passed in pairs
*/
npairs = MIN(npairs, HC_MAXPAIRS);
va_start(ap, npairs);
for (i = n; i < npairs + n; i++) {
const char *name = va_arg(ap, const char *);
uint32_t id = va_arg(ap, uint32_t);
char idstr[11];
(void) snprintf(idstr, sizeof (idstr), "%u", id);
pairs[i] = fm_nvlist_create(nva);
if (nvlist_add_string(pairs[i], FM_FMRI_HC_NAME, name) != 0 ||
nvlist_add_string(pairs[i], FM_FMRI_HC_ID, idstr) != 0) {
for (j = 0; j <= i; j++) {
if (pairs[j] != NULL)
fm_nvlist_destroy(pairs[j],
FM_NVA_RETAIN);
}
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
}
va_end(ap);
/*
* Create the fmri hc list
*/
if (nvlist_add_nvlist_array(fmri, FM_FMRI_HC_LIST,
(const nvlist_t **)pairs, npairs + n) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
for (i = 0; i < npairs + n; i++) {
fm_nvlist_destroy(pairs[i], FM_NVA_RETAIN);
}
if (snvl != NULL) {
if (nvlist_add_nvlist(fmri, FM_FMRI_HC_SPECIFIC, snvl) != 0) {
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
}
}
/*
* Set-up and validate the members of an dev fmri according to:
*
* Member name Type Value
* ====================================================
* version uint8_t 0
* auth nvlist_t <auth>
* devpath string <devpath>
* [devid] string <devid>
* [target-port-l0id] string <target-port-lun0-id>
*
* Note that auth and devid are optional members.
*/
void
fm_fmri_dev_set(nvlist_t *fmri_dev, int version, const nvlist_t *auth,
const char *devpath, const char *devid, const char *tpl0)
{
int err = 0;
if (version != DEV_SCHEME_VERSION0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
err |= nvlist_add_uint8(fmri_dev, FM_VERSION, version);
err |= nvlist_add_string(fmri_dev, FM_FMRI_SCHEME, FM_FMRI_SCHEME_DEV);
if (auth != NULL) {
err |= nvlist_add_nvlist(fmri_dev, FM_FMRI_AUTHORITY,
(nvlist_t *)auth);
}
err |= nvlist_add_string(fmri_dev, FM_FMRI_DEV_PATH, devpath);
if (devid != NULL)
err |= nvlist_add_string(fmri_dev, FM_FMRI_DEV_ID, devid);
if (tpl0 != NULL)
err |= nvlist_add_string(fmri_dev, FM_FMRI_DEV_TGTPTLUN0, tpl0);
if (err)
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
}
/*
* Set-up and validate the members of an cpu fmri according to:
*
* Member name Type Value
* ====================================================
* version uint8_t 0
* auth nvlist_t <auth>
* cpuid uint32_t <cpu_id>
* cpumask uint8_t <cpu_mask>
* serial uint64_t <serial_id>
*
* Note that auth, cpumask, serial are optional members.
*
*/
void
fm_fmri_cpu_set(nvlist_t *fmri_cpu, int version, const nvlist_t *auth,
uint32_t cpu_id, uint8_t *cpu_maskp, const char *serial_idp)
{
uint64_t *failedp = &erpt_kstat_data.fmri_set_failed.value.ui64;
if (version < CPU_SCHEME_VERSION1) {
atomic_inc_64(failedp);
return;
}
if (nvlist_add_uint8(fmri_cpu, FM_VERSION, version) != 0) {
atomic_inc_64(failedp);
return;
}
if (nvlist_add_string(fmri_cpu, FM_FMRI_SCHEME,
FM_FMRI_SCHEME_CPU) != 0) {
atomic_inc_64(failedp);
return;
}
if (auth != NULL && nvlist_add_nvlist(fmri_cpu, FM_FMRI_AUTHORITY,
(nvlist_t *)auth) != 0)
atomic_inc_64(failedp);
if (nvlist_add_uint32(fmri_cpu, FM_FMRI_CPU_ID, cpu_id) != 0)
atomic_inc_64(failedp);
if (cpu_maskp != NULL && nvlist_add_uint8(fmri_cpu, FM_FMRI_CPU_MASK,
*cpu_maskp) != 0)
atomic_inc_64(failedp);
if (serial_idp == NULL || nvlist_add_string(fmri_cpu,
FM_FMRI_CPU_SERIAL_ID, (char *)serial_idp) != 0)
atomic_inc_64(failedp);
}
/*
* Set-up and validate the members of a mem according to:
*
* Member name Type Value
* ====================================================
* version uint8_t 0
* auth nvlist_t <auth> [optional]
* unum string <unum>
* serial string <serial> [optional*]
* offset uint64_t <offset> [optional]
*
* * serial is required if offset is present
*/
void
fm_fmri_mem_set(nvlist_t *fmri, int version, const nvlist_t *auth,
const char *unum, const char *serial, uint64_t offset)
{
if (version != MEM_SCHEME_VERSION0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
if (!serial && (offset != (uint64_t)-1)) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
if (nvlist_add_uint8(fmri, FM_VERSION, version) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
if (nvlist_add_string(fmri, FM_FMRI_SCHEME, FM_FMRI_SCHEME_MEM) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
if (auth != NULL) {
if (nvlist_add_nvlist(fmri, FM_FMRI_AUTHORITY,
(nvlist_t *)auth) != 0) {
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
}
}
if (nvlist_add_string(fmri, FM_FMRI_MEM_UNUM, unum) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
}
if (serial != NULL) {
if (nvlist_add_string_array(fmri, FM_FMRI_MEM_SERIAL_ID,
(const char **)&serial, 1) != 0) {
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
}
if (offset != (uint64_t)-1 && nvlist_add_uint64(fmri,
FM_FMRI_MEM_OFFSET, offset) != 0) {
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
}
}
}
void
fm_fmri_zfs_set(nvlist_t *fmri, int version, uint64_t pool_guid,
uint64_t vdev_guid)
{
if (version != ZFS_SCHEME_VERSION0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
if (nvlist_add_uint8(fmri, FM_VERSION, version) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
if (nvlist_add_string(fmri, FM_FMRI_SCHEME, FM_FMRI_SCHEME_ZFS) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
return;
}
if (nvlist_add_uint64(fmri, FM_FMRI_ZFS_POOL, pool_guid) != 0) {
atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
}
if (vdev_guid != 0) {
if (nvlist_add_uint64(fmri, FM_FMRI_ZFS_VDEV, vdev_guid) != 0) {
atomic_inc_64(
&erpt_kstat_data.fmri_set_failed.value.ui64);
}
}
}
uint64_t
fm_ena_increment(uint64_t ena)
{
uint64_t new_ena;
switch (ENA_FORMAT(ena)) {
case FM_ENA_FMT1:
new_ena = ena + (1 << ENA_FMT1_GEN_SHFT);
break;
case FM_ENA_FMT2:
new_ena = ena + (1 << ENA_FMT2_GEN_SHFT);
break;
default:
new_ena = 0;
}
return (new_ena);
}
uint64_t
fm_ena_generate_cpu(uint64_t timestamp, processorid_t cpuid, uchar_t format)
{
uint64_t ena = 0;
switch (format) {
case FM_ENA_FMT1:
if (timestamp) {
ena = (uint64_t)((format & ENA_FORMAT_MASK) |
((cpuid << ENA_FMT1_CPUID_SHFT) &
ENA_FMT1_CPUID_MASK) |
((timestamp << ENA_FMT1_TIME_SHFT) &
ENA_FMT1_TIME_MASK));
} else {
ena = (uint64_t)((format & ENA_FORMAT_MASK) |
((cpuid << ENA_FMT1_CPUID_SHFT) &
ENA_FMT1_CPUID_MASK) |
((gethrtime() << ENA_FMT1_TIME_SHFT) &
ENA_FMT1_TIME_MASK));
}
break;
case FM_ENA_FMT2:
ena = (uint64_t)((format & ENA_FORMAT_MASK) |
((timestamp << ENA_FMT2_TIME_SHFT) & ENA_FMT2_TIME_MASK));
break;
default:
break;
}
return (ena);
}
uint64_t
fm_ena_generate(uint64_t timestamp, uchar_t format)
{
uint64_t ena;
kpreempt_disable();
ena = fm_ena_generate_cpu(timestamp, getcpuid(), format);
kpreempt_enable();
return (ena);
}
uint64_t
fm_ena_generation_get(uint64_t ena)
{
uint64_t gen;
switch (ENA_FORMAT(ena)) {
case FM_ENA_FMT1:
gen = (ena & ENA_FMT1_GEN_MASK) >> ENA_FMT1_GEN_SHFT;
break;
case FM_ENA_FMT2:
gen = (ena & ENA_FMT2_GEN_MASK) >> ENA_FMT2_GEN_SHFT;
break;
default:
gen = 0;
break;
}
return (gen);
}
uchar_t
fm_ena_format_get(uint64_t ena)
{
return (ENA_FORMAT(ena));
}
uint64_t
fm_ena_id_get(uint64_t ena)
{
uint64_t id;
switch (ENA_FORMAT(ena)) {
case FM_ENA_FMT1:
id = (ena & ENA_FMT1_ID_MASK) >> ENA_FMT1_ID_SHFT;
break;
case FM_ENA_FMT2:
id = (ena & ENA_FMT2_ID_MASK) >> ENA_FMT2_ID_SHFT;
break;
default:
id = 0;
}
return (id);
}
uint64_t
fm_ena_time_get(uint64_t ena)
{
uint64_t time;
switch (ENA_FORMAT(ena)) {
case FM_ENA_FMT1:
time = (ena & ENA_FMT1_TIME_MASK) >> ENA_FMT1_TIME_SHFT;
break;
case FM_ENA_FMT2:
time = (ena & ENA_FMT2_TIME_MASK) >> ENA_FMT2_TIME_SHFT;
break;
default:
time = 0;
}
return (time);
}
#ifdef _KERNEL
/*
* Helper function to increment ereport dropped count. Used by the event
* rate limiting code to give feedback to the user about how many events were
* rate limited by including them in the 'dropped' count.
*/
void
fm_erpt_dropped_increment(void)
{
atomic_inc_64(&ratelimit_dropped);
}
void
fm_init(void)
{
zevent_len_cur = 0;
zevent_flags = 0;
/* Initialize zevent allocation and generation kstats */
fm_ksp = kstat_create("zfs", 0, "fm", "misc", KSTAT_TYPE_NAMED,
sizeof (struct erpt_kstat) / sizeof (kstat_named_t),
KSTAT_FLAG_VIRTUAL);
if (fm_ksp != NULL) {
fm_ksp->ks_data = &erpt_kstat_data;
kstat_install(fm_ksp);
} else {
cmn_err(CE_NOTE, "failed to create fm/misc kstat\n");
}
mutex_init(&zevent_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&zevent_list, sizeof (zevent_t),
offsetof(zevent_t, ev_node));
cv_init(&zevent_cv, NULL, CV_DEFAULT, NULL);
zfs_ereport_init();
}
void
fm_fini(void)
{
int count;
zfs_ereport_fini();
zfs_zevent_drain_all(&count);
mutex_enter(&zevent_lock);
cv_broadcast(&zevent_cv);
zevent_flags |= ZEVENT_SHUTDOWN;
while (zevent_waiters > 0) {
mutex_exit(&zevent_lock);
schedule();
mutex_enter(&zevent_lock);
}
mutex_exit(&zevent_lock);
cv_destroy(&zevent_cv);
list_destroy(&zevent_list);
mutex_destroy(&zevent_lock);
if (fm_ksp != NULL) {
kstat_delete(fm_ksp);
fm_ksp = NULL;
}
}
#endif /* _KERNEL */
ZFS_MODULE_PARAM(zfs_zevent, zfs_zevent_, len_max, INT, ZMOD_RW,
"Max event queue length");
diff --git a/module/zfs/metaslab.c b/module/zfs/metaslab.c
index 9e216c38d954..145f79fae3df 100644
--- a/module/zfs/metaslab.c
+++ b/module/zfs/metaslab.c
@@ -1,6257 +1,6251 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright (c) 2015, Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2017, Intel Corporation.
*/
#include <sys/zfs_context.h>
#include <sys/dmu.h>
#include <sys/dmu_tx.h>
#include <sys/space_map.h>
#include <sys/metaslab_impl.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_draid.h>
#include <sys/zio.h>
#include <sys/spa_impl.h>
#include <sys/zfeature.h>
#include <sys/vdev_indirect_mapping.h>
#include <sys/zap.h>
#include <sys/btree.h>
#define WITH_DF_BLOCK_ALLOCATOR
#define GANG_ALLOCATION(flags) \
((flags) & (METASLAB_GANG_CHILD | METASLAB_GANG_HEADER))
/*
* Metaslab granularity, in bytes. This is roughly similar to what would be
* referred to as the "stripe size" in traditional RAID arrays. In normal
* operation, we will try to write this amount of data to a top-level vdev
* before moving on to the next one.
*/
-unsigned long metaslab_aliquot = 512 << 10;
+static unsigned long metaslab_aliquot = 512 << 10;
/*
* For testing, make some blocks above a certain size be gang blocks.
*/
unsigned long metaslab_force_ganging = SPA_MAXBLOCKSIZE + 1;
/*
* In pools where the log space map feature is not enabled we touch
* multiple metaslabs (and their respective space maps) with each
* transaction group. Thus, we benefit from having a small space map
* block size since it allows us to issue more I/O operations scattered
* around the disk. So a sane default for the space map block size
* is 8~16K.
*/
int zfs_metaslab_sm_blksz_no_log = (1 << 14);
/*
* When the log space map feature is enabled, we accumulate a lot of
* changes per metaslab that are flushed once in a while so we benefit
* from a bigger block size like 128K for the metaslab space maps.
*/
int zfs_metaslab_sm_blksz_with_log = (1 << 17);
/*
* The in-core space map representation is more compact than its on-disk form.
* The zfs_condense_pct determines how much more compact the in-core
* space map representation must be before we compact it on-disk.
* Values should be greater than or equal to 100.
*/
int zfs_condense_pct = 200;
/*
* Condensing a metaslab is not guaranteed to actually reduce the amount of
* space used on disk. In particular, a space map uses data in increments of
* MAX(1 << ashift, space_map_blksz), so a metaslab might use the
* same number of blocks after condensing. Since the goal of condensing is to
* reduce the number of IOPs required to read the space map, we only want to
* condense when we can be sure we will reduce the number of blocks used by the
* space map. Unfortunately, we cannot precisely compute whether or not this is
* the case in metaslab_should_condense since we are holding ms_lock. Instead,
* we apply the following heuristic: do not condense a spacemap unless the
* uncondensed size consumes greater than zfs_metaslab_condense_block_threshold
* blocks.
*/
-int zfs_metaslab_condense_block_threshold = 4;
+static const int zfs_metaslab_condense_block_threshold = 4;
/*
* The zfs_mg_noalloc_threshold defines which metaslab groups should
* be eligible for allocation. The value is defined as a percentage of
* free space. Metaslab groups that have more free space than
* zfs_mg_noalloc_threshold are always eligible for allocations. Once
* a metaslab group's free space is less than or equal to the
* zfs_mg_noalloc_threshold the allocator will avoid allocating to that
* group unless all groups in the pool have reached zfs_mg_noalloc_threshold.
* Once all groups in the pool reach zfs_mg_noalloc_threshold then all
* groups are allowed to accept allocations. Gang blocks are always
* eligible to allocate on any metaslab group. The default value of 0 means
* no metaslab group will be excluded based on this criterion.
*/
-int zfs_mg_noalloc_threshold = 0;
+static int zfs_mg_noalloc_threshold = 0;
/*
* Metaslab groups are considered eligible for allocations if their
* fragmentation metric (measured as a percentage) is less than or
* equal to zfs_mg_fragmentation_threshold. If a metaslab group
* exceeds this threshold then it will be skipped unless all metaslab
* groups within the metaslab class have also crossed this threshold.
*
* This tunable was introduced to avoid edge cases where we continue
* allocating from very fragmented disks in our pool while other, less
* fragmented disks, exists. On the other hand, if all disks in the
* pool are uniformly approaching the threshold, the threshold can
* be a speed bump in performance, where we keep switching the disks
* that we allocate from (e.g. we allocate some segments from disk A
* making it bypassing the threshold while freeing segments from disk
* B getting its fragmentation below the threshold).
*
* Empirically, we've seen that our vdev selection for allocations is
* good enough that fragmentation increases uniformly across all vdevs
* the majority of the time. Thus we set the threshold percentage high
* enough to avoid hitting the speed bump on pools that are being pushed
* to the edge.
*/
-int zfs_mg_fragmentation_threshold = 95;
+static int zfs_mg_fragmentation_threshold = 95;
/*
* Allow metaslabs to keep their active state as long as their fragmentation
* percentage is less than or equal to zfs_metaslab_fragmentation_threshold. An
* active metaslab that exceeds this threshold will no longer keep its active
* status allowing better metaslabs to be selected.
*/
-int zfs_metaslab_fragmentation_threshold = 70;
+static int zfs_metaslab_fragmentation_threshold = 70;
/*
* When set will load all metaslabs when pool is first opened.
*/
-int metaslab_debug_load = 0;
+int metaslab_debug_load = B_FALSE;
/*
* When set will prevent metaslabs from being unloaded.
*/
-int metaslab_debug_unload = 0;
+static int metaslab_debug_unload = B_FALSE;
/*
* Minimum size which forces the dynamic allocator to change
* it's allocation strategy. Once the space map cannot satisfy
* an allocation of this size then it switches to using more
* aggressive strategy (i.e search by size rather than offset).
*/
uint64_t metaslab_df_alloc_threshold = SPA_OLD_MAXBLOCKSIZE;
/*
* The minimum free space, in percent, which must be available
* in a space map to continue allocations in a first-fit fashion.
* Once the space map's free space drops below this level we dynamically
* switch to using best-fit allocations.
*/
int metaslab_df_free_pct = 4;
/*
* Maximum distance to search forward from the last offset. Without this
* limit, fragmented pools can see >100,000 iterations and
* metaslab_block_picker() becomes the performance limiting factor on
* high-performance storage.
*
* With the default setting of 16MB, we typically see less than 500
* iterations, even with very fragmented, ashift=9 pools. The maximum number
* of iterations possible is:
* metaslab_df_max_search / (2 * (1<<ashift))
* With the default setting of 16MB this is 16*1024 (with ashift=9) or
* 2048 (with ashift=12).
*/
-int metaslab_df_max_search = 16 * 1024 * 1024;
+static int metaslab_df_max_search = 16 * 1024 * 1024;
/*
* Forces the metaslab_block_picker function to search for at least this many
* segments forwards until giving up on finding a segment that the allocation
* will fit into.
*/
-uint32_t metaslab_min_search_count = 100;
+static const uint32_t metaslab_min_search_count = 100;
/*
* If we are not searching forward (due to metaslab_df_max_search,
* metaslab_df_free_pct, or metaslab_df_alloc_threshold), this tunable
* controls what segment is used. If it is set, we will use the largest free
* segment. If it is not set, we will use a segment of exactly the requested
* size (or larger).
*/
-int metaslab_df_use_largest_segment = B_FALSE;
+static int metaslab_df_use_largest_segment = B_FALSE;
/*
* Percentage of all cpus that can be used by the metaslab taskq.
*/
int metaslab_load_pct = 50;
/*
* These tunables control how long a metaslab will remain loaded after the
* last allocation from it. A metaslab can't be unloaded until at least
* metaslab_unload_delay TXG's and metaslab_unload_delay_ms milliseconds
* have elapsed. However, zfs_metaslab_mem_limit may cause it to be
* unloaded sooner. These settings are intended to be generous -- to keep
* metaslabs loaded for a long time, reducing the rate of metaslab loading.
*/
-int metaslab_unload_delay = 32;
-int metaslab_unload_delay_ms = 10 * 60 * 1000; /* ten minutes */
+static int metaslab_unload_delay = 32;
+static int metaslab_unload_delay_ms = 10 * 60 * 1000; /* ten minutes */
/*
* Max number of metaslabs per group to preload.
*/
int metaslab_preload_limit = 10;
/*
* Enable/disable preloading of metaslab.
*/
-int metaslab_preload_enabled = B_TRUE;
+static int metaslab_preload_enabled = B_TRUE;
/*
* Enable/disable fragmentation weighting on metaslabs.
*/
-int metaslab_fragmentation_factor_enabled = B_TRUE;
+static int metaslab_fragmentation_factor_enabled = B_TRUE;
/*
* Enable/disable lba weighting (i.e. outer tracks are given preference).
*/
-int metaslab_lba_weighting_enabled = B_TRUE;
+static int metaslab_lba_weighting_enabled = B_TRUE;
/*
* Enable/disable metaslab group biasing.
*/
-int metaslab_bias_enabled = B_TRUE;
+static int metaslab_bias_enabled = B_TRUE;
/*
* Enable/disable remapping of indirect DVAs to their concrete vdevs.
*/
-boolean_t zfs_remap_blkptr_enable = B_TRUE;
+static const boolean_t zfs_remap_blkptr_enable = B_TRUE;
/*
* Enable/disable segment-based metaslab selection.
*/
-int zfs_metaslab_segment_weight_enabled = B_TRUE;
+static int zfs_metaslab_segment_weight_enabled = B_TRUE;
/*
* When using segment-based metaslab selection, we will continue
* allocating from the active metaslab until we have exhausted
* zfs_metaslab_switch_threshold of its buckets.
*/
-int zfs_metaslab_switch_threshold = 2;
+static int zfs_metaslab_switch_threshold = 2;
/*
* Internal switch to enable/disable the metaslab allocation tracing
* facility.
*/
-boolean_t metaslab_trace_enabled = B_FALSE;
+static const boolean_t metaslab_trace_enabled = B_FALSE;
/*
* Maximum entries that the metaslab allocation tracing facility will keep
* in a given list when running in non-debug mode. We limit the number
* of entries in non-debug mode to prevent us from using up too much memory.
* The limit should be sufficiently large that we don't expect any allocation
* to every exceed this value. In debug mode, the system will panic if this
* limit is ever reached allowing for further investigation.
*/
-uint64_t metaslab_trace_max_entries = 5000;
+static const uint64_t metaslab_trace_max_entries = 5000;
/*
* Maximum number of metaslabs per group that can be disabled
* simultaneously.
*/
-int max_disabled_ms = 3;
+static const int max_disabled_ms = 3;
/*
* Time (in seconds) to respect ms_max_size when the metaslab is not loaded.
* To avoid 64-bit overflow, don't set above UINT32_MAX.
*/
-unsigned long zfs_metaslab_max_size_cache_sec = 3600; /* 1 hour */
+static unsigned long zfs_metaslab_max_size_cache_sec = 1 * 60 * 60; /* 1 hour */
/*
* Maximum percentage of memory to use on storing loaded metaslabs. If loading
* a metaslab would take it over this percentage, the oldest selected metaslab
* is automatically unloaded.
*/
-int zfs_metaslab_mem_limit = 25;
+static int zfs_metaslab_mem_limit = 25;
/*
* Force the per-metaslab range trees to use 64-bit integers to store
* segments. Used for debugging purposes.
*/
-boolean_t zfs_metaslab_force_large_segs = B_FALSE;
+static const boolean_t zfs_metaslab_force_large_segs = B_FALSE;
/*
* By default we only store segments over a certain size in the size-sorted
* metaslab trees (ms_allocatable_by_size and
* ms_unflushed_frees_by_size). This dramatically reduces memory usage and
* improves load and unload times at the cost of causing us to use slightly
* larger segments than we would otherwise in some cases.
*/
-uint32_t metaslab_by_size_min_shift = 14;
+static const uint32_t metaslab_by_size_min_shift = 14;
/*
* If not set, we will first try normal allocation. If that fails then
* we will do a gang allocation. If that fails then we will do a "try hard"
* gang allocation. If that fails then we will have a multi-layer gang
* block.
*
* If set, we will first try normal allocation. If that fails then
* we will do a "try hard" allocation. If that fails we will do a gang
* allocation. If that fails we will do a "try hard" gang allocation. If
* that fails then we will have a multi-layer gang block.
*/
-int zfs_metaslab_try_hard_before_gang = B_FALSE;
+static int zfs_metaslab_try_hard_before_gang = B_FALSE;
/*
* When not trying hard, we only consider the best zfs_metaslab_find_max_tries
* metaslabs. This improves performance, especially when there are many
* metaslabs per vdev and the allocation can't actually be satisfied (so we
* would otherwise iterate all the metaslabs). If there is a metaslab with a
* worse weight but it can actually satisfy the allocation, we won't find it
* until trying hard. This may happen if the worse metaslab is not loaded
* (and the true weight is better than we have calculated), or due to weight
* bucketization. E.g. we are looking for a 60K segment, and the best
* metaslabs all have free segments in the 32-63K bucket, but the best
* zfs_metaslab_find_max_tries metaslabs have ms_max_size <60KB, and a
* subsequent metaslab has ms_max_size >60KB (but fewer segments in this
* bucket, and therefore a lower weight).
*/
-int zfs_metaslab_find_max_tries = 100;
+static int zfs_metaslab_find_max_tries = 100;
static uint64_t metaslab_weight(metaslab_t *, boolean_t);
static void metaslab_set_fragmentation(metaslab_t *, boolean_t);
static void metaslab_free_impl(vdev_t *, uint64_t, uint64_t, boolean_t);
static void metaslab_check_free_impl(vdev_t *, uint64_t, uint64_t);
static void metaslab_passivate(metaslab_t *msp, uint64_t weight);
static uint64_t metaslab_weight_from_range_tree(metaslab_t *msp);
static void metaslab_flush_update(metaslab_t *, dmu_tx_t *);
static unsigned int metaslab_idx_func(multilist_t *, void *);
static void metaslab_evict(metaslab_t *, uint64_t);
static void metaslab_rt_add(range_tree_t *rt, range_seg_t *rs, void *arg);
kmem_cache_t *metaslab_alloc_trace_cache;
typedef struct metaslab_stats {
kstat_named_t metaslabstat_trace_over_limit;
kstat_named_t metaslabstat_reload_tree;
kstat_named_t metaslabstat_too_many_tries;
kstat_named_t metaslabstat_try_hard;
} metaslab_stats_t;
static metaslab_stats_t metaslab_stats = {
{ "trace_over_limit", KSTAT_DATA_UINT64 },
{ "reload_tree", KSTAT_DATA_UINT64 },
{ "too_many_tries", KSTAT_DATA_UINT64 },
{ "try_hard", KSTAT_DATA_UINT64 },
};
#define METASLABSTAT_BUMP(stat) \
atomic_inc_64(&metaslab_stats.stat.value.ui64);
-kstat_t *metaslab_ksp;
+static kstat_t *metaslab_ksp;
void
metaslab_stat_init(void)
{
ASSERT(metaslab_alloc_trace_cache == NULL);
metaslab_alloc_trace_cache = kmem_cache_create(
"metaslab_alloc_trace_cache", sizeof (metaslab_alloc_trace_t),
0, NULL, NULL, NULL, NULL, NULL, 0);
metaslab_ksp = kstat_create("zfs", 0, "metaslab_stats",
"misc", KSTAT_TYPE_NAMED, sizeof (metaslab_stats) /
sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
if (metaslab_ksp != NULL) {
metaslab_ksp->ks_data = &metaslab_stats;
kstat_install(metaslab_ksp);
}
}
void
metaslab_stat_fini(void)
{
if (metaslab_ksp != NULL) {
kstat_delete(metaslab_ksp);
metaslab_ksp = NULL;
}
kmem_cache_destroy(metaslab_alloc_trace_cache);
metaslab_alloc_trace_cache = NULL;
}
/*
* ==========================================================================
* Metaslab classes
* ==========================================================================
*/
metaslab_class_t *
-metaslab_class_create(spa_t *spa, metaslab_ops_t *ops)
+metaslab_class_create(spa_t *spa, const metaslab_ops_t *ops)
{
metaslab_class_t *mc;
mc = kmem_zalloc(offsetof(metaslab_class_t,
mc_allocator[spa->spa_alloc_count]), KM_SLEEP);
mc->mc_spa = spa;
mc->mc_ops = ops;
mutex_init(&mc->mc_lock, NULL, MUTEX_DEFAULT, NULL);
multilist_create(&mc->mc_metaslab_txg_list, sizeof (metaslab_t),
offsetof(metaslab_t, ms_class_txg_node), metaslab_idx_func);
for (int i = 0; i < spa->spa_alloc_count; i++) {
metaslab_class_allocator_t *mca = &mc->mc_allocator[i];
mca->mca_rotor = NULL;
zfs_refcount_create_tracked(&mca->mca_alloc_slots);
}
return (mc);
}
void
metaslab_class_destroy(metaslab_class_t *mc)
{
spa_t *spa = mc->mc_spa;
ASSERT(mc->mc_alloc == 0);
ASSERT(mc->mc_deferred == 0);
ASSERT(mc->mc_space == 0);
ASSERT(mc->mc_dspace == 0);
for (int i = 0; i < spa->spa_alloc_count; i++) {
metaslab_class_allocator_t *mca = &mc->mc_allocator[i];
ASSERT(mca->mca_rotor == NULL);
zfs_refcount_destroy(&mca->mca_alloc_slots);
}
mutex_destroy(&mc->mc_lock);
multilist_destroy(&mc->mc_metaslab_txg_list);
kmem_free(mc, offsetof(metaslab_class_t,
mc_allocator[spa->spa_alloc_count]));
}
int
metaslab_class_validate(metaslab_class_t *mc)
{
metaslab_group_t *mg;
vdev_t *vd;
/*
* Must hold one of the spa_config locks.
*/
ASSERT(spa_config_held(mc->mc_spa, SCL_ALL, RW_READER) ||
spa_config_held(mc->mc_spa, SCL_ALL, RW_WRITER));
if ((mg = mc->mc_allocator[0].mca_rotor) == NULL)
return (0);
do {
vd = mg->mg_vd;
ASSERT(vd->vdev_mg != NULL);
ASSERT3P(vd->vdev_top, ==, vd);
ASSERT3P(mg->mg_class, ==, mc);
ASSERT3P(vd->vdev_ops, !=, &vdev_hole_ops);
} while ((mg = mg->mg_next) != mc->mc_allocator[0].mca_rotor);
return (0);
}
static void
metaslab_class_space_update(metaslab_class_t *mc, int64_t alloc_delta,
int64_t defer_delta, int64_t space_delta, int64_t dspace_delta)
{
atomic_add_64(&mc->mc_alloc, alloc_delta);
atomic_add_64(&mc->mc_deferred, defer_delta);
atomic_add_64(&mc->mc_space, space_delta);
atomic_add_64(&mc->mc_dspace, dspace_delta);
}
uint64_t
metaslab_class_get_alloc(metaslab_class_t *mc)
{
return (mc->mc_alloc);
}
uint64_t
metaslab_class_get_deferred(metaslab_class_t *mc)
{
return (mc->mc_deferred);
}
uint64_t
metaslab_class_get_space(metaslab_class_t *mc)
{
return (mc->mc_space);
}
uint64_t
metaslab_class_get_dspace(metaslab_class_t *mc)
{
return (spa_deflate(mc->mc_spa) ? mc->mc_dspace : mc->mc_space);
}
void
metaslab_class_histogram_verify(metaslab_class_t *mc)
{
spa_t *spa = mc->mc_spa;
vdev_t *rvd = spa->spa_root_vdev;
uint64_t *mc_hist;
int i;
if ((zfs_flags & ZFS_DEBUG_HISTOGRAM_VERIFY) == 0)
return;
mc_hist = kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE,
KM_SLEEP);
mutex_enter(&mc->mc_lock);
for (int c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
metaslab_group_t *mg = vdev_get_mg(tvd, mc);
/*
* Skip any holes, uninitialized top-levels, or
* vdevs that are not in this metalab class.
*/
if (!vdev_is_concrete(tvd) || tvd->vdev_ms_shift == 0 ||
mg->mg_class != mc) {
continue;
}
IMPLY(mg == mg->mg_vd->vdev_log_mg,
mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));
for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++)
mc_hist[i] += mg->mg_histogram[i];
}
for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
VERIFY3U(mc_hist[i], ==, mc->mc_histogram[i]);
}
mutex_exit(&mc->mc_lock);
kmem_free(mc_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
}
/*
* Calculate the metaslab class's fragmentation metric. The metric
* is weighted based on the space contribution of each metaslab group.
* The return value will be a number between 0 and 100 (inclusive), or
* ZFS_FRAG_INVALID if the metric has not been set. See comment above the
* zfs_frag_table for more information about the metric.
*/
uint64_t
metaslab_class_fragmentation(metaslab_class_t *mc)
{
vdev_t *rvd = mc->mc_spa->spa_root_vdev;
uint64_t fragmentation = 0;
spa_config_enter(mc->mc_spa, SCL_VDEV, FTAG, RW_READER);
for (int c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
metaslab_group_t *mg = tvd->vdev_mg;
/*
* Skip any holes, uninitialized top-levels,
* or vdevs that are not in this metalab class.
*/
if (!vdev_is_concrete(tvd) || tvd->vdev_ms_shift == 0 ||
mg->mg_class != mc) {
continue;
}
/*
* If a metaslab group does not contain a fragmentation
* metric then just bail out.
*/
if (mg->mg_fragmentation == ZFS_FRAG_INVALID) {
spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
return (ZFS_FRAG_INVALID);
}
/*
* Determine how much this metaslab_group is contributing
* to the overall pool fragmentation metric.
*/
fragmentation += mg->mg_fragmentation *
metaslab_group_get_space(mg);
}
fragmentation /= metaslab_class_get_space(mc);
ASSERT3U(fragmentation, <=, 100);
spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
return (fragmentation);
}
/*
* Calculate the amount of expandable space that is available in
* this metaslab class. If a device is expanded then its expandable
* space will be the amount of allocatable space that is currently not
* part of this metaslab class.
*/
uint64_t
metaslab_class_expandable_space(metaslab_class_t *mc)
{
vdev_t *rvd = mc->mc_spa->spa_root_vdev;
uint64_t space = 0;
spa_config_enter(mc->mc_spa, SCL_VDEV, FTAG, RW_READER);
for (int c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
metaslab_group_t *mg = tvd->vdev_mg;
if (!vdev_is_concrete(tvd) || tvd->vdev_ms_shift == 0 ||
mg->mg_class != mc) {
continue;
}
/*
* Calculate if we have enough space to add additional
* metaslabs. We report the expandable space in terms
* of the metaslab size since that's the unit of expansion.
*/
space += P2ALIGN(tvd->vdev_max_asize - tvd->vdev_asize,
1ULL << tvd->vdev_ms_shift);
}
spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
return (space);
}
void
metaslab_class_evict_old(metaslab_class_t *mc, uint64_t txg)
{
multilist_t *ml = &mc->mc_metaslab_txg_list;
for (int i = 0; i < multilist_get_num_sublists(ml); i++) {
multilist_sublist_t *mls = multilist_sublist_lock(ml, i);
metaslab_t *msp = multilist_sublist_head(mls);
multilist_sublist_unlock(mls);
while (msp != NULL) {
mutex_enter(&msp->ms_lock);
/*
* If the metaslab has been removed from the list
* (which could happen if we were at the memory limit
* and it was evicted during this loop), then we can't
* proceed and we should restart the sublist.
*/
if (!multilist_link_active(&msp->ms_class_txg_node)) {
mutex_exit(&msp->ms_lock);
i--;
break;
}
mls = multilist_sublist_lock(ml, i);
metaslab_t *next_msp = multilist_sublist_next(mls, msp);
multilist_sublist_unlock(mls);
if (txg >
msp->ms_selected_txg + metaslab_unload_delay &&
gethrtime() > msp->ms_selected_time +
(uint64_t)MSEC2NSEC(metaslab_unload_delay_ms)) {
metaslab_evict(msp, txg);
} else {
/*
* Once we've hit a metaslab selected too
* recently to evict, we're done evicting for
* now.
*/
mutex_exit(&msp->ms_lock);
break;
}
mutex_exit(&msp->ms_lock);
msp = next_msp;
}
}
}
static int
metaslab_compare(const void *x1, const void *x2)
{
const metaslab_t *m1 = (const metaslab_t *)x1;
const metaslab_t *m2 = (const metaslab_t *)x2;
int sort1 = 0;
int sort2 = 0;
if (m1->ms_allocator != -1 && m1->ms_primary)
sort1 = 1;
else if (m1->ms_allocator != -1 && !m1->ms_primary)
sort1 = 2;
if (m2->ms_allocator != -1 && m2->ms_primary)
sort2 = 1;
else if (m2->ms_allocator != -1 && !m2->ms_primary)
sort2 = 2;
/*
* Sort inactive metaslabs first, then primaries, then secondaries. When
* selecting a metaslab to allocate from, an allocator first tries its
* primary, then secondary active metaslab. If it doesn't have active
* metaslabs, or can't allocate from them, it searches for an inactive
* metaslab to activate. If it can't find a suitable one, it will steal
* a primary or secondary metaslab from another allocator.
*/
if (sort1 < sort2)
return (-1);
if (sort1 > sort2)
return (1);
int cmp = TREE_CMP(m2->ms_weight, m1->ms_weight);
if (likely(cmp))
return (cmp);
IMPLY(TREE_CMP(m1->ms_start, m2->ms_start) == 0, m1 == m2);
return (TREE_CMP(m1->ms_start, m2->ms_start));
}
/*
* ==========================================================================
* Metaslab groups
* ==========================================================================
*/
/*
* Update the allocatable flag and the metaslab group's capacity.
* The allocatable flag is set to true if the capacity is below
* the zfs_mg_noalloc_threshold or has a fragmentation value that is
* greater than zfs_mg_fragmentation_threshold. If a metaslab group
* transitions from allocatable to non-allocatable or vice versa then the
* metaslab group's class is updated to reflect the transition.
*/
static void
metaslab_group_alloc_update(metaslab_group_t *mg)
{
vdev_t *vd = mg->mg_vd;
metaslab_class_t *mc = mg->mg_class;
vdev_stat_t *vs = &vd->vdev_stat;
boolean_t was_allocatable;
boolean_t was_initialized;
ASSERT(vd == vd->vdev_top);
ASSERT3U(spa_config_held(mc->mc_spa, SCL_ALLOC, RW_READER), ==,
SCL_ALLOC);
mutex_enter(&mg->mg_lock);
was_allocatable = mg->mg_allocatable;
was_initialized = mg->mg_initialized;
mg->mg_free_capacity = ((vs->vs_space - vs->vs_alloc) * 100) /
(vs->vs_space + 1);
mutex_enter(&mc->mc_lock);
/*
* If the metaslab group was just added then it won't
* have any space until we finish syncing out this txg.
* At that point we will consider it initialized and available
* for allocations. We also don't consider non-activated
* metaslab groups (e.g. vdevs that are in the middle of being removed)
* to be initialized, because they can't be used for allocation.
*/
mg->mg_initialized = metaslab_group_initialized(mg);
if (!was_initialized && mg->mg_initialized) {
mc->mc_groups++;
} else if (was_initialized && !mg->mg_initialized) {
ASSERT3U(mc->mc_groups, >, 0);
mc->mc_groups--;
}
if (mg->mg_initialized)
mg->mg_no_free_space = B_FALSE;
/*
* A metaslab group is considered allocatable if it has plenty
* of free space or is not heavily fragmented. We only take
* fragmentation into account if the metaslab group has a valid
* fragmentation metric (i.e. a value between 0 and 100).
*/
mg->mg_allocatable = (mg->mg_activation_count > 0 &&
mg->mg_free_capacity > zfs_mg_noalloc_threshold &&
(mg->mg_fragmentation == ZFS_FRAG_INVALID ||
mg->mg_fragmentation <= zfs_mg_fragmentation_threshold));
/*
* The mc_alloc_groups maintains a count of the number of
* groups in this metaslab class that are still above the
* zfs_mg_noalloc_threshold. This is used by the allocating
* threads to determine if they should avoid allocations to
* a given group. The allocator will avoid allocations to a group
* if that group has reached or is below the zfs_mg_noalloc_threshold
* and there are still other groups that are above the threshold.
* When a group transitions from allocatable to non-allocatable or
* vice versa we update the metaslab class to reflect that change.
* When the mc_alloc_groups value drops to 0 that means that all
* groups have reached the zfs_mg_noalloc_threshold making all groups
* eligible for allocations. This effectively means that all devices
* are balanced again.
*/
if (was_allocatable && !mg->mg_allocatable)
mc->mc_alloc_groups--;
else if (!was_allocatable && mg->mg_allocatable)
mc->mc_alloc_groups++;
mutex_exit(&mc->mc_lock);
mutex_exit(&mg->mg_lock);
}
int
metaslab_sort_by_flushed(const void *va, const void *vb)
{
const metaslab_t *a = va;
const metaslab_t *b = vb;
int cmp = TREE_CMP(a->ms_unflushed_txg, b->ms_unflushed_txg);
if (likely(cmp))
return (cmp);
uint64_t a_vdev_id = a->ms_group->mg_vd->vdev_id;
uint64_t b_vdev_id = b->ms_group->mg_vd->vdev_id;
cmp = TREE_CMP(a_vdev_id, b_vdev_id);
if (cmp)
return (cmp);
return (TREE_CMP(a->ms_id, b->ms_id));
}
metaslab_group_t *
metaslab_group_create(metaslab_class_t *mc, vdev_t *vd, int allocators)
{
metaslab_group_t *mg;
mg = kmem_zalloc(offsetof(metaslab_group_t,
mg_allocator[allocators]), KM_SLEEP);
mutex_init(&mg->mg_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&mg->mg_ms_disabled_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&mg->mg_ms_disabled_cv, NULL, CV_DEFAULT, NULL);
avl_create(&mg->mg_metaslab_tree, metaslab_compare,
sizeof (metaslab_t), offsetof(metaslab_t, ms_group_node));
mg->mg_vd = vd;
mg->mg_class = mc;
mg->mg_activation_count = 0;
mg->mg_initialized = B_FALSE;
mg->mg_no_free_space = B_TRUE;
mg->mg_allocators = allocators;
for (int i = 0; i < allocators; i++) {
metaslab_group_allocator_t *mga = &mg->mg_allocator[i];
zfs_refcount_create_tracked(&mga->mga_alloc_queue_depth);
}
mg->mg_taskq = taskq_create("metaslab_group_taskq", metaslab_load_pct,
maxclsyspri, 10, INT_MAX, TASKQ_THREADS_CPU_PCT | TASKQ_DYNAMIC);
return (mg);
}
void
metaslab_group_destroy(metaslab_group_t *mg)
{
ASSERT(mg->mg_prev == NULL);
ASSERT(mg->mg_next == NULL);
/*
* We may have gone below zero with the activation count
* either because we never activated in the first place or
* because we're done, and possibly removing the vdev.
*/
ASSERT(mg->mg_activation_count <= 0);
taskq_destroy(mg->mg_taskq);
avl_destroy(&mg->mg_metaslab_tree);
mutex_destroy(&mg->mg_lock);
mutex_destroy(&mg->mg_ms_disabled_lock);
cv_destroy(&mg->mg_ms_disabled_cv);
for (int i = 0; i < mg->mg_allocators; i++) {
metaslab_group_allocator_t *mga = &mg->mg_allocator[i];
zfs_refcount_destroy(&mga->mga_alloc_queue_depth);
}
kmem_free(mg, offsetof(metaslab_group_t,
mg_allocator[mg->mg_allocators]));
}
void
metaslab_group_activate(metaslab_group_t *mg)
{
metaslab_class_t *mc = mg->mg_class;
spa_t *spa = mc->mc_spa;
metaslab_group_t *mgprev, *mgnext;
ASSERT3U(spa_config_held(spa, SCL_ALLOC, RW_WRITER), !=, 0);
ASSERT(mg->mg_prev == NULL);
ASSERT(mg->mg_next == NULL);
ASSERT(mg->mg_activation_count <= 0);
if (++mg->mg_activation_count <= 0)
return;
mg->mg_aliquot = metaslab_aliquot * MAX(1, mg->mg_vd->vdev_children);
metaslab_group_alloc_update(mg);
if ((mgprev = mc->mc_allocator[0].mca_rotor) == NULL) {
mg->mg_prev = mg;
mg->mg_next = mg;
} else {
mgnext = mgprev->mg_next;
mg->mg_prev = mgprev;
mg->mg_next = mgnext;
mgprev->mg_next = mg;
mgnext->mg_prev = mg;
}
for (int i = 0; i < spa->spa_alloc_count; i++) {
mc->mc_allocator[i].mca_rotor = mg;
mg = mg->mg_next;
}
}
/*
* Passivate a metaslab group and remove it from the allocation rotor.
* Callers must hold both the SCL_ALLOC and SCL_ZIO lock prior to passivating
* a metaslab group. This function will momentarily drop spa_config_locks
* that are lower than the SCL_ALLOC lock (see comment below).
*/
void
metaslab_group_passivate(metaslab_group_t *mg)
{
metaslab_class_t *mc = mg->mg_class;
spa_t *spa = mc->mc_spa;
metaslab_group_t *mgprev, *mgnext;
int locks = spa_config_held(spa, SCL_ALL, RW_WRITER);
ASSERT3U(spa_config_held(spa, SCL_ALLOC | SCL_ZIO, RW_WRITER), ==,
(SCL_ALLOC | SCL_ZIO));
if (--mg->mg_activation_count != 0) {
for (int i = 0; i < spa->spa_alloc_count; i++)
ASSERT(mc->mc_allocator[i].mca_rotor != mg);
ASSERT(mg->mg_prev == NULL);
ASSERT(mg->mg_next == NULL);
ASSERT(mg->mg_activation_count < 0);
return;
}
/*
* The spa_config_lock is an array of rwlocks, ordered as
* follows (from highest to lowest):
* SCL_CONFIG > SCL_STATE > SCL_L2ARC > SCL_ALLOC >
* SCL_ZIO > SCL_FREE > SCL_VDEV
* (For more information about the spa_config_lock see spa_misc.c)
* The higher the lock, the broader its coverage. When we passivate
* a metaslab group, we must hold both the SCL_ALLOC and the SCL_ZIO
* config locks. However, the metaslab group's taskq might be trying
* to preload metaslabs so we must drop the SCL_ZIO lock and any
* lower locks to allow the I/O to complete. At a minimum,
* we continue to hold the SCL_ALLOC lock, which prevents any future
* allocations from taking place and any changes to the vdev tree.
*/
spa_config_exit(spa, locks & ~(SCL_ZIO - 1), spa);
taskq_wait_outstanding(mg->mg_taskq, 0);
spa_config_enter(spa, locks & ~(SCL_ZIO - 1), spa, RW_WRITER);
metaslab_group_alloc_update(mg);
for (int i = 0; i < mg->mg_allocators; i++) {
metaslab_group_allocator_t *mga = &mg->mg_allocator[i];
metaslab_t *msp = mga->mga_primary;
if (msp != NULL) {
mutex_enter(&msp->ms_lock);
metaslab_passivate(msp,
metaslab_weight_from_range_tree(msp));
mutex_exit(&msp->ms_lock);
}
msp = mga->mga_secondary;
if (msp != NULL) {
mutex_enter(&msp->ms_lock);
metaslab_passivate(msp,
metaslab_weight_from_range_tree(msp));
mutex_exit(&msp->ms_lock);
}
}
mgprev = mg->mg_prev;
mgnext = mg->mg_next;
if (mg == mgnext) {
mgnext = NULL;
} else {
mgprev->mg_next = mgnext;
mgnext->mg_prev = mgprev;
}
for (int i = 0; i < spa->spa_alloc_count; i++) {
if (mc->mc_allocator[i].mca_rotor == mg)
mc->mc_allocator[i].mca_rotor = mgnext;
}
mg->mg_prev = NULL;
mg->mg_next = NULL;
}
boolean_t
metaslab_group_initialized(metaslab_group_t *mg)
{
vdev_t *vd = mg->mg_vd;
vdev_stat_t *vs = &vd->vdev_stat;
return (vs->vs_space != 0 && mg->mg_activation_count > 0);
}
uint64_t
metaslab_group_get_space(metaslab_group_t *mg)
{
/*
* Note that the number of nodes in mg_metaslab_tree may be one less
* than vdev_ms_count, due to the embedded log metaslab.
*/
mutex_enter(&mg->mg_lock);
uint64_t ms_count = avl_numnodes(&mg->mg_metaslab_tree);
mutex_exit(&mg->mg_lock);
return ((1ULL << mg->mg_vd->vdev_ms_shift) * ms_count);
}
void
metaslab_group_histogram_verify(metaslab_group_t *mg)
{
uint64_t *mg_hist;
avl_tree_t *t = &mg->mg_metaslab_tree;
uint64_t ashift = mg->mg_vd->vdev_ashift;
if ((zfs_flags & ZFS_DEBUG_HISTOGRAM_VERIFY) == 0)
return;
mg_hist = kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE,
KM_SLEEP);
ASSERT3U(RANGE_TREE_HISTOGRAM_SIZE, >=,
SPACE_MAP_HISTOGRAM_SIZE + ashift);
mutex_enter(&mg->mg_lock);
for (metaslab_t *msp = avl_first(t);
msp != NULL; msp = AVL_NEXT(t, msp)) {
VERIFY3P(msp->ms_group, ==, mg);
/* skip if not active */
if (msp->ms_sm == NULL)
continue;
for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
mg_hist[i + ashift] +=
msp->ms_sm->sm_phys->smp_histogram[i];
}
}
for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i ++)
VERIFY3U(mg_hist[i], ==, mg->mg_histogram[i]);
mutex_exit(&mg->mg_lock);
kmem_free(mg_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
}
static void
metaslab_group_histogram_add(metaslab_group_t *mg, metaslab_t *msp)
{
metaslab_class_t *mc = mg->mg_class;
uint64_t ashift = mg->mg_vd->vdev_ashift;
ASSERT(MUTEX_HELD(&msp->ms_lock));
if (msp->ms_sm == NULL)
return;
mutex_enter(&mg->mg_lock);
mutex_enter(&mc->mc_lock);
for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
IMPLY(mg == mg->mg_vd->vdev_log_mg,
mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));
mg->mg_histogram[i + ashift] +=
msp->ms_sm->sm_phys->smp_histogram[i];
mc->mc_histogram[i + ashift] +=
msp->ms_sm->sm_phys->smp_histogram[i];
}
mutex_exit(&mc->mc_lock);
mutex_exit(&mg->mg_lock);
}
void
metaslab_group_histogram_remove(metaslab_group_t *mg, metaslab_t *msp)
{
metaslab_class_t *mc = mg->mg_class;
uint64_t ashift = mg->mg_vd->vdev_ashift;
ASSERT(MUTEX_HELD(&msp->ms_lock));
if (msp->ms_sm == NULL)
return;
mutex_enter(&mg->mg_lock);
mutex_enter(&mc->mc_lock);
for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
ASSERT3U(mg->mg_histogram[i + ashift], >=,
msp->ms_sm->sm_phys->smp_histogram[i]);
ASSERT3U(mc->mc_histogram[i + ashift], >=,
msp->ms_sm->sm_phys->smp_histogram[i]);
IMPLY(mg == mg->mg_vd->vdev_log_mg,
mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));
mg->mg_histogram[i + ashift] -=
msp->ms_sm->sm_phys->smp_histogram[i];
mc->mc_histogram[i + ashift] -=
msp->ms_sm->sm_phys->smp_histogram[i];
}
mutex_exit(&mc->mc_lock);
mutex_exit(&mg->mg_lock);
}
static void
metaslab_group_add(metaslab_group_t *mg, metaslab_t *msp)
{
ASSERT(msp->ms_group == NULL);
mutex_enter(&mg->mg_lock);
msp->ms_group = mg;
msp->ms_weight = 0;
avl_add(&mg->mg_metaslab_tree, msp);
mutex_exit(&mg->mg_lock);
mutex_enter(&msp->ms_lock);
metaslab_group_histogram_add(mg, msp);
mutex_exit(&msp->ms_lock);
}
static void
metaslab_group_remove(metaslab_group_t *mg, metaslab_t *msp)
{
mutex_enter(&msp->ms_lock);
metaslab_group_histogram_remove(mg, msp);
mutex_exit(&msp->ms_lock);
mutex_enter(&mg->mg_lock);
ASSERT(msp->ms_group == mg);
avl_remove(&mg->mg_metaslab_tree, msp);
metaslab_class_t *mc = msp->ms_group->mg_class;
multilist_sublist_t *mls =
multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
if (multilist_link_active(&msp->ms_class_txg_node))
multilist_sublist_remove(mls, msp);
multilist_sublist_unlock(mls);
msp->ms_group = NULL;
mutex_exit(&mg->mg_lock);
}
static void
metaslab_group_sort_impl(metaslab_group_t *mg, metaslab_t *msp, uint64_t weight)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
ASSERT(MUTEX_HELD(&mg->mg_lock));
ASSERT(msp->ms_group == mg);
avl_remove(&mg->mg_metaslab_tree, msp);
msp->ms_weight = weight;
avl_add(&mg->mg_metaslab_tree, msp);
}
static void
metaslab_group_sort(metaslab_group_t *mg, metaslab_t *msp, uint64_t weight)
{
/*
* Although in principle the weight can be any value, in
* practice we do not use values in the range [1, 511].
*/
ASSERT(weight >= SPA_MINBLOCKSIZE || weight == 0);
ASSERT(MUTEX_HELD(&msp->ms_lock));
mutex_enter(&mg->mg_lock);
metaslab_group_sort_impl(mg, msp, weight);
mutex_exit(&mg->mg_lock);
}
/*
* Calculate the fragmentation for a given metaslab group. We can use
* a simple average here since all metaslabs within the group must have
* the same size. The return value will be a value between 0 and 100
* (inclusive), or ZFS_FRAG_INVALID if less than half of the metaslab in this
* group have a fragmentation metric.
*/
uint64_t
metaslab_group_fragmentation(metaslab_group_t *mg)
{
vdev_t *vd = mg->mg_vd;
uint64_t fragmentation = 0;
uint64_t valid_ms = 0;
for (int m = 0; m < vd->vdev_ms_count; m++) {
metaslab_t *msp = vd->vdev_ms[m];
if (msp->ms_fragmentation == ZFS_FRAG_INVALID)
continue;
if (msp->ms_group != mg)
continue;
valid_ms++;
fragmentation += msp->ms_fragmentation;
}
if (valid_ms <= mg->mg_vd->vdev_ms_count / 2)
return (ZFS_FRAG_INVALID);
fragmentation /= valid_ms;
ASSERT3U(fragmentation, <=, 100);
return (fragmentation);
}
/*
* Determine if a given metaslab group should skip allocations. A metaslab
* group should avoid allocations if its free capacity is less than the
* zfs_mg_noalloc_threshold or its fragmentation metric is greater than
* zfs_mg_fragmentation_threshold and there is at least one metaslab group
* that can still handle allocations. If the allocation throttle is enabled
* then we skip allocations to devices that have reached their maximum
* allocation queue depth unless the selected metaslab group is the only
* eligible group remaining.
*/
static boolean_t
metaslab_group_allocatable(metaslab_group_t *mg, metaslab_group_t *rotor,
uint64_t psize, int allocator, int d)
{
spa_t *spa = mg->mg_vd->vdev_spa;
metaslab_class_t *mc = mg->mg_class;
/*
* We can only consider skipping this metaslab group if it's
* in the normal metaslab class and there are other metaslab
* groups to select from. Otherwise, we always consider it eligible
* for allocations.
*/
if ((mc != spa_normal_class(spa) &&
mc != spa_special_class(spa) &&
mc != spa_dedup_class(spa)) ||
mc->mc_groups <= 1)
return (B_TRUE);
/*
* If the metaslab group's mg_allocatable flag is set (see comments
* in metaslab_group_alloc_update() for more information) and
* the allocation throttle is disabled then allow allocations to this
* device. However, if the allocation throttle is enabled then
* check if we have reached our allocation limit (mga_alloc_queue_depth)
* to determine if we should allow allocations to this metaslab group.
* If all metaslab groups are no longer considered allocatable
* (mc_alloc_groups == 0) or we're trying to allocate the smallest
* gang block size then we allow allocations on this metaslab group
* regardless of the mg_allocatable or throttle settings.
*/
if (mg->mg_allocatable) {
metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
int64_t qdepth;
uint64_t qmax = mga->mga_cur_max_alloc_queue_depth;
if (!mc->mc_alloc_throttle_enabled)
return (B_TRUE);
/*
* If this metaslab group does not have any free space, then
* there is no point in looking further.
*/
if (mg->mg_no_free_space)
return (B_FALSE);
/*
* Relax allocation throttling for ditto blocks. Due to
* random imbalances in allocation it tends to push copies
* to one vdev, that looks a bit better at the moment.
*/
qmax = qmax * (4 + d) / 4;
qdepth = zfs_refcount_count(&mga->mga_alloc_queue_depth);
/*
* If this metaslab group is below its qmax or it's
* the only allocatable metasable group, then attempt
* to allocate from it.
*/
if (qdepth < qmax || mc->mc_alloc_groups == 1)
return (B_TRUE);
ASSERT3U(mc->mc_alloc_groups, >, 1);
/*
* Since this metaslab group is at or over its qmax, we
* need to determine if there are metaslab groups after this
* one that might be able to handle this allocation. This is
* racy since we can't hold the locks for all metaslab
* groups at the same time when we make this check.
*/
for (metaslab_group_t *mgp = mg->mg_next;
mgp != rotor; mgp = mgp->mg_next) {
metaslab_group_allocator_t *mgap =
&mgp->mg_allocator[allocator];
qmax = mgap->mga_cur_max_alloc_queue_depth;
qmax = qmax * (4 + d) / 4;
qdepth =
zfs_refcount_count(&mgap->mga_alloc_queue_depth);
/*
* If there is another metaslab group that
* might be able to handle the allocation, then
* we return false so that we skip this group.
*/
if (qdepth < qmax && !mgp->mg_no_free_space)
return (B_FALSE);
}
/*
* We didn't find another group to handle the allocation
* so we can't skip this metaslab group even though
* we are at or over our qmax.
*/
return (B_TRUE);
} else if (mc->mc_alloc_groups == 0 || psize == SPA_MINBLOCKSIZE) {
return (B_TRUE);
}
return (B_FALSE);
}
/*
* ==========================================================================
* Range tree callbacks
* ==========================================================================
*/
/*
* Comparison function for the private size-ordered tree using 32-bit
* ranges. Tree is sorted by size, larger sizes at the end of the tree.
*/
static int
metaslab_rangesize32_compare(const void *x1, const void *x2)
{
const range_seg32_t *r1 = x1;
const range_seg32_t *r2 = x2;
uint64_t rs_size1 = r1->rs_end - r1->rs_start;
uint64_t rs_size2 = r2->rs_end - r2->rs_start;
int cmp = TREE_CMP(rs_size1, rs_size2);
if (likely(cmp))
return (cmp);
return (TREE_CMP(r1->rs_start, r2->rs_start));
}
/*
* Comparison function for the private size-ordered tree using 64-bit
* ranges. Tree is sorted by size, larger sizes at the end of the tree.
*/
static int
metaslab_rangesize64_compare(const void *x1, const void *x2)
{
const range_seg64_t *r1 = x1;
const range_seg64_t *r2 = x2;
uint64_t rs_size1 = r1->rs_end - r1->rs_start;
uint64_t rs_size2 = r2->rs_end - r2->rs_start;
int cmp = TREE_CMP(rs_size1, rs_size2);
if (likely(cmp))
return (cmp);
return (TREE_CMP(r1->rs_start, r2->rs_start));
}
typedef struct metaslab_rt_arg {
zfs_btree_t *mra_bt;
uint32_t mra_floor_shift;
} metaslab_rt_arg_t;
struct mssa_arg {
range_tree_t *rt;
metaslab_rt_arg_t *mra;
};
static void
metaslab_size_sorted_add(void *arg, uint64_t start, uint64_t size)
{
struct mssa_arg *mssap = arg;
range_tree_t *rt = mssap->rt;
metaslab_rt_arg_t *mrap = mssap->mra;
range_seg_max_t seg = {0};
rs_set_start(&seg, rt, start);
rs_set_end(&seg, rt, start + size);
metaslab_rt_add(rt, &seg, mrap);
}
static void
metaslab_size_tree_full_load(range_tree_t *rt)
{
metaslab_rt_arg_t *mrap = rt->rt_arg;
METASLABSTAT_BUMP(metaslabstat_reload_tree);
ASSERT0(zfs_btree_numnodes(mrap->mra_bt));
mrap->mra_floor_shift = 0;
struct mssa_arg arg = {0};
arg.rt = rt;
arg.mra = mrap;
range_tree_walk(rt, metaslab_size_sorted_add, &arg);
}
/*
* Create any block allocator specific components. The current allocators
* rely on using both a size-ordered range_tree_t and an array of uint64_t's.
*/
static void
metaslab_rt_create(range_tree_t *rt, void *arg)
{
metaslab_rt_arg_t *mrap = arg;
zfs_btree_t *size_tree = mrap->mra_bt;
size_t size;
int (*compare) (const void *, const void *);
switch (rt->rt_type) {
case RANGE_SEG32:
size = sizeof (range_seg32_t);
compare = metaslab_rangesize32_compare;
break;
case RANGE_SEG64:
size = sizeof (range_seg64_t);
compare = metaslab_rangesize64_compare;
break;
default:
panic("Invalid range seg type %d", rt->rt_type);
}
zfs_btree_create(size_tree, compare, size);
mrap->mra_floor_shift = metaslab_by_size_min_shift;
}
static void
metaslab_rt_destroy(range_tree_t *rt, void *arg)
{
(void) rt;
metaslab_rt_arg_t *mrap = arg;
zfs_btree_t *size_tree = mrap->mra_bt;
zfs_btree_destroy(size_tree);
kmem_free(mrap, sizeof (*mrap));
}
static void
metaslab_rt_add(range_tree_t *rt, range_seg_t *rs, void *arg)
{
metaslab_rt_arg_t *mrap = arg;
zfs_btree_t *size_tree = mrap->mra_bt;
if (rs_get_end(rs, rt) - rs_get_start(rs, rt) <
(1 << mrap->mra_floor_shift))
return;
zfs_btree_add(size_tree, rs);
}
static void
metaslab_rt_remove(range_tree_t *rt, range_seg_t *rs, void *arg)
{
metaslab_rt_arg_t *mrap = arg;
zfs_btree_t *size_tree = mrap->mra_bt;
if (rs_get_end(rs, rt) - rs_get_start(rs, rt) < (1 <<
mrap->mra_floor_shift))
return;
zfs_btree_remove(size_tree, rs);
}
static void
metaslab_rt_vacate(range_tree_t *rt, void *arg)
{
metaslab_rt_arg_t *mrap = arg;
zfs_btree_t *size_tree = mrap->mra_bt;
zfs_btree_clear(size_tree);
zfs_btree_destroy(size_tree);
metaslab_rt_create(rt, arg);
}
-static range_tree_ops_t metaslab_rt_ops = {
+static const range_tree_ops_t metaslab_rt_ops = {
.rtop_create = metaslab_rt_create,
.rtop_destroy = metaslab_rt_destroy,
.rtop_add = metaslab_rt_add,
.rtop_remove = metaslab_rt_remove,
.rtop_vacate = metaslab_rt_vacate
};
/*
* ==========================================================================
* Common allocator routines
* ==========================================================================
*/
/*
* Return the maximum contiguous segment within the metaslab.
*/
uint64_t
metaslab_largest_allocatable(metaslab_t *msp)
{
zfs_btree_t *t = &msp->ms_allocatable_by_size;
range_seg_t *rs;
if (t == NULL)
return (0);
if (zfs_btree_numnodes(t) == 0)
metaslab_size_tree_full_load(msp->ms_allocatable);
rs = zfs_btree_last(t, NULL);
if (rs == NULL)
return (0);
return (rs_get_end(rs, msp->ms_allocatable) - rs_get_start(rs,
msp->ms_allocatable));
}
/*
* Return the maximum contiguous segment within the unflushed frees of this
* metaslab.
*/
static uint64_t
metaslab_largest_unflushed_free(metaslab_t *msp)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
if (msp->ms_unflushed_frees == NULL)
return (0);
if (zfs_btree_numnodes(&msp->ms_unflushed_frees_by_size) == 0)
metaslab_size_tree_full_load(msp->ms_unflushed_frees);
range_seg_t *rs = zfs_btree_last(&msp->ms_unflushed_frees_by_size,
NULL);
if (rs == NULL)
return (0);
/*
* When a range is freed from the metaslab, that range is added to
* both the unflushed frees and the deferred frees. While the block
* will eventually be usable, if the metaslab were loaded the range
* would not be added to the ms_allocatable tree until TXG_DEFER_SIZE
* txgs had passed. As a result, when attempting to estimate an upper
* bound for the largest currently-usable free segment in the
* metaslab, we need to not consider any ranges currently in the defer
* trees. This algorithm approximates the largest available chunk in
* the largest range in the unflushed_frees tree by taking the first
* chunk. While this may be a poor estimate, it should only remain so
* briefly and should eventually self-correct as frees are no longer
* deferred. Similar logic applies to the ms_freed tree. See
* metaslab_load() for more details.
*
* There are two primary sources of inaccuracy in this estimate. Both
* are tolerated for performance reasons. The first source is that we
* only check the largest segment for overlaps. Smaller segments may
* have more favorable overlaps with the other trees, resulting in
* larger usable chunks. Second, we only look at the first chunk in
* the largest segment; there may be other usable chunks in the
* largest segment, but we ignore them.
*/
uint64_t rstart = rs_get_start(rs, msp->ms_unflushed_frees);
uint64_t rsize = rs_get_end(rs, msp->ms_unflushed_frees) - rstart;
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
uint64_t start = 0;
uint64_t size = 0;
boolean_t found = range_tree_find_in(msp->ms_defer[t], rstart,
rsize, &start, &size);
if (found) {
if (rstart == start)
return (0);
rsize = start - rstart;
}
}
uint64_t start = 0;
uint64_t size = 0;
boolean_t found = range_tree_find_in(msp->ms_freed, rstart,
rsize, &start, &size);
if (found)
rsize = start - rstart;
return (rsize);
}
static range_seg_t *
metaslab_block_find(zfs_btree_t *t, range_tree_t *rt, uint64_t start,
uint64_t size, zfs_btree_index_t *where)
{
range_seg_t *rs;
range_seg_max_t rsearch;
rs_set_start(&rsearch, rt, start);
rs_set_end(&rsearch, rt, start + size);
rs = zfs_btree_find(t, &rsearch, where);
if (rs == NULL) {
rs = zfs_btree_next(t, where, where);
}
return (rs);
}
#if defined(WITH_DF_BLOCK_ALLOCATOR) || \
defined(WITH_CF_BLOCK_ALLOCATOR)
/*
* This is a helper function that can be used by the allocator to find a
* suitable block to allocate. This will search the specified B-tree looking
* for a block that matches the specified criteria.
*/
static uint64_t
metaslab_block_picker(range_tree_t *rt, uint64_t *cursor, uint64_t size,
uint64_t max_search)
{
if (*cursor == 0)
*cursor = rt->rt_start;
zfs_btree_t *bt = &rt->rt_root;
zfs_btree_index_t where;
range_seg_t *rs = metaslab_block_find(bt, rt, *cursor, size, &where);
uint64_t first_found;
int count_searched = 0;
if (rs != NULL)
first_found = rs_get_start(rs, rt);
while (rs != NULL && (rs_get_start(rs, rt) - first_found <=
max_search || count_searched < metaslab_min_search_count)) {
uint64_t offset = rs_get_start(rs, rt);
if (offset + size <= rs_get_end(rs, rt)) {
*cursor = offset + size;
return (offset);
}
rs = zfs_btree_next(bt, &where, &where);
count_searched++;
}
*cursor = 0;
return (-1ULL);
}
#endif /* WITH_DF/CF_BLOCK_ALLOCATOR */
#if defined(WITH_DF_BLOCK_ALLOCATOR)
/*
* ==========================================================================
* Dynamic Fit (df) block allocator
*
* Search for a free chunk of at least this size, starting from the last
* offset (for this alignment of block) looking for up to
* metaslab_df_max_search bytes (16MB). If a large enough free chunk is not
* found within 16MB, then return a free chunk of exactly the requested size (or
* larger).
*
* If it seems like searching from the last offset will be unproductive, skip
* that and just return a free chunk of exactly the requested size (or larger).
* This is based on metaslab_df_alloc_threshold and metaslab_df_free_pct. This
* mechanism is probably not very useful and may be removed in the future.
*
* The behavior when not searching can be changed to return the largest free
* chunk, instead of a free chunk of exactly the requested size, by setting
* metaslab_df_use_largest_segment.
* ==========================================================================
*/
static uint64_t
metaslab_df_alloc(metaslab_t *msp, uint64_t size)
{
/*
* Find the largest power of 2 block size that evenly divides the
* requested size. This is used to try to allocate blocks with similar
* alignment from the same area of the metaslab (i.e. same cursor
* bucket) but it does not guarantee that other allocations sizes
* may exist in the same region.
*/
uint64_t align = size & -size;
uint64_t *cursor = &msp->ms_lbas[highbit64(align) - 1];
range_tree_t *rt = msp->ms_allocatable;
int free_pct = range_tree_space(rt) * 100 / msp->ms_size;
uint64_t offset;
ASSERT(MUTEX_HELD(&msp->ms_lock));
/*
* If we're running low on space, find a segment based on size,
* rather than iterating based on offset.
*/
if (metaslab_largest_allocatable(msp) < metaslab_df_alloc_threshold ||
free_pct < metaslab_df_free_pct) {
offset = -1;
} else {
offset = metaslab_block_picker(rt,
cursor, size, metaslab_df_max_search);
}
if (offset == -1) {
range_seg_t *rs;
if (zfs_btree_numnodes(&msp->ms_allocatable_by_size) == 0)
metaslab_size_tree_full_load(msp->ms_allocatable);
if (metaslab_df_use_largest_segment) {
/* use largest free segment */
rs = zfs_btree_last(&msp->ms_allocatable_by_size, NULL);
} else {
zfs_btree_index_t where;
/* use segment of this size, or next largest */
rs = metaslab_block_find(&msp->ms_allocatable_by_size,
rt, msp->ms_start, size, &where);
}
if (rs != NULL && rs_get_start(rs, rt) + size <= rs_get_end(rs,
rt)) {
offset = rs_get_start(rs, rt);
*cursor = offset + size;
}
}
return (offset);
}
-static metaslab_ops_t metaslab_df_ops = {
+const metaslab_ops_t zfs_metaslab_ops = {
metaslab_df_alloc
};
-
-metaslab_ops_t *zfs_metaslab_ops = &metaslab_df_ops;
#endif /* WITH_DF_BLOCK_ALLOCATOR */
#if defined(WITH_CF_BLOCK_ALLOCATOR)
/*
* ==========================================================================
* Cursor fit block allocator -
* Select the largest region in the metaslab, set the cursor to the beginning
* of the range and the cursor_end to the end of the range. As allocations
* are made advance the cursor. Continue allocating from the cursor until
* the range is exhausted and then find a new range.
* ==========================================================================
*/
static uint64_t
metaslab_cf_alloc(metaslab_t *msp, uint64_t size)
{
range_tree_t *rt = msp->ms_allocatable;
zfs_btree_t *t = &msp->ms_allocatable_by_size;
uint64_t *cursor = &msp->ms_lbas[0];
uint64_t *cursor_end = &msp->ms_lbas[1];
uint64_t offset = 0;
ASSERT(MUTEX_HELD(&msp->ms_lock));
ASSERT3U(*cursor_end, >=, *cursor);
if ((*cursor + size) > *cursor_end) {
range_seg_t *rs;
if (zfs_btree_numnodes(t) == 0)
metaslab_size_tree_full_load(msp->ms_allocatable);
rs = zfs_btree_last(t, NULL);
if (rs == NULL || (rs_get_end(rs, rt) - rs_get_start(rs, rt)) <
size)
return (-1ULL);
*cursor = rs_get_start(rs, rt);
*cursor_end = rs_get_end(rs, rt);
}
offset = *cursor;
*cursor += size;
return (offset);
}
-static metaslab_ops_t metaslab_cf_ops = {
+const metaslab_ops_t zfs_metaslab_ops = {
metaslab_cf_alloc
};
-
-metaslab_ops_t *zfs_metaslab_ops = &metaslab_cf_ops;
#endif /* WITH_CF_BLOCK_ALLOCATOR */
#if defined(WITH_NDF_BLOCK_ALLOCATOR)
/*
* ==========================================================================
* New dynamic fit allocator -
* Select a region that is large enough to allocate 2^metaslab_ndf_clump_shift
* contiguous blocks. If no region is found then just use the largest segment
* that remains.
* ==========================================================================
*/
/*
* Determines desired number of contiguous blocks (2^metaslab_ndf_clump_shift)
* to request from the allocator.
*/
uint64_t metaslab_ndf_clump_shift = 4;
static uint64_t
metaslab_ndf_alloc(metaslab_t *msp, uint64_t size)
{
zfs_btree_t *t = &msp->ms_allocatable->rt_root;
range_tree_t *rt = msp->ms_allocatable;
zfs_btree_index_t where;
range_seg_t *rs;
range_seg_max_t rsearch;
uint64_t hbit = highbit64(size);
uint64_t *cursor = &msp->ms_lbas[hbit - 1];
uint64_t max_size = metaslab_largest_allocatable(msp);
ASSERT(MUTEX_HELD(&msp->ms_lock));
if (max_size < size)
return (-1ULL);
rs_set_start(&rsearch, rt, *cursor);
rs_set_end(&rsearch, rt, *cursor + size);
rs = zfs_btree_find(t, &rsearch, &where);
if (rs == NULL || (rs_get_end(rs, rt) - rs_get_start(rs, rt)) < size) {
t = &msp->ms_allocatable_by_size;
rs_set_start(&rsearch, rt, 0);
rs_set_end(&rsearch, rt, MIN(max_size, 1ULL << (hbit +
metaslab_ndf_clump_shift)));
rs = zfs_btree_find(t, &rsearch, &where);
if (rs == NULL)
rs = zfs_btree_next(t, &where, &where);
ASSERT(rs != NULL);
}
if ((rs_get_end(rs, rt) - rs_get_start(rs, rt)) >= size) {
*cursor = rs_get_start(rs, rt) + size;
return (rs_get_start(rs, rt));
}
return (-1ULL);
}
-static metaslab_ops_t metaslab_ndf_ops = {
+const metaslab_ops_t zfs_metaslab_ops = {
metaslab_ndf_alloc
};
-
-metaslab_ops_t *zfs_metaslab_ops = &metaslab_ndf_ops;
#endif /* WITH_NDF_BLOCK_ALLOCATOR */
/*
* ==========================================================================
* Metaslabs
* ==========================================================================
*/
/*
* Wait for any in-progress metaslab loads to complete.
*/
static void
metaslab_load_wait(metaslab_t *msp)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
while (msp->ms_loading) {
ASSERT(!msp->ms_loaded);
cv_wait(&msp->ms_load_cv, &msp->ms_lock);
}
}
/*
* Wait for any in-progress flushing to complete.
*/
static void
metaslab_flush_wait(metaslab_t *msp)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
while (msp->ms_flushing)
cv_wait(&msp->ms_flush_cv, &msp->ms_lock);
}
static unsigned int
metaslab_idx_func(multilist_t *ml, void *arg)
{
metaslab_t *msp = arg;
/*
* ms_id values are allocated sequentially, so full 64bit
* division would be a waste of time, so limit it to 32 bits.
*/
return ((unsigned int)msp->ms_id % multilist_get_num_sublists(ml));
}
uint64_t
metaslab_allocated_space(metaslab_t *msp)
{
return (msp->ms_allocated_space);
}
/*
* Verify that the space accounting on disk matches the in-core range_trees.
*/
static void
metaslab_verify_space(metaslab_t *msp, uint64_t txg)
{
spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
uint64_t allocating = 0;
uint64_t sm_free_space, msp_free_space;
ASSERT(MUTEX_HELD(&msp->ms_lock));
ASSERT(!msp->ms_condensing);
if ((zfs_flags & ZFS_DEBUG_METASLAB_VERIFY) == 0)
return;
/*
* We can only verify the metaslab space when we're called
* from syncing context with a loaded metaslab that has an
* allocated space map. Calling this in non-syncing context
* does not provide a consistent view of the metaslab since
* we're performing allocations in the future.
*/
if (txg != spa_syncing_txg(spa) || msp->ms_sm == NULL ||
!msp->ms_loaded)
return;
/*
* Even though the smp_alloc field can get negative,
* when it comes to a metaslab's space map, that should
* never be the case.
*/
ASSERT3S(space_map_allocated(msp->ms_sm), >=, 0);
ASSERT3U(space_map_allocated(msp->ms_sm), >=,
range_tree_space(msp->ms_unflushed_frees));
ASSERT3U(metaslab_allocated_space(msp), ==,
space_map_allocated(msp->ms_sm) +
range_tree_space(msp->ms_unflushed_allocs) -
range_tree_space(msp->ms_unflushed_frees));
sm_free_space = msp->ms_size - metaslab_allocated_space(msp);
/*
* Account for future allocations since we would have
* already deducted that space from the ms_allocatable.
*/
for (int t = 0; t < TXG_CONCURRENT_STATES; t++) {
allocating +=
range_tree_space(msp->ms_allocating[(txg + t) & TXG_MASK]);
}
ASSERT3U(allocating + msp->ms_allocated_this_txg, ==,
msp->ms_allocating_total);
ASSERT3U(msp->ms_deferspace, ==,
range_tree_space(msp->ms_defer[0]) +
range_tree_space(msp->ms_defer[1]));
msp_free_space = range_tree_space(msp->ms_allocatable) + allocating +
msp->ms_deferspace + range_tree_space(msp->ms_freed);
VERIFY3U(sm_free_space, ==, msp_free_space);
}
static void
metaslab_aux_histograms_clear(metaslab_t *msp)
{
/*
* Auxiliary histograms are only cleared when resetting them,
* which can only happen while the metaslab is loaded.
*/
ASSERT(msp->ms_loaded);
bzero(msp->ms_synchist, sizeof (msp->ms_synchist));
for (int t = 0; t < TXG_DEFER_SIZE; t++)
bzero(msp->ms_deferhist[t], sizeof (msp->ms_deferhist[t]));
}
static void
metaslab_aux_histogram_add(uint64_t *histogram, uint64_t shift,
range_tree_t *rt)
{
/*
* This is modeled after space_map_histogram_add(), so refer to that
* function for implementation details. We want this to work like
* the space map histogram, and not the range tree histogram, as we
* are essentially constructing a delta that will be later subtracted
* from the space map histogram.
*/
int idx = 0;
for (int i = shift; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
ASSERT3U(i, >=, idx + shift);
histogram[idx] += rt->rt_histogram[i] << (i - idx - shift);
if (idx < SPACE_MAP_HISTOGRAM_SIZE - 1) {
ASSERT3U(idx + shift, ==, i);
idx++;
ASSERT3U(idx, <, SPACE_MAP_HISTOGRAM_SIZE);
}
}
}
/*
* Called at every sync pass that the metaslab gets synced.
*
* The reason is that we want our auxiliary histograms to be updated
* wherever the metaslab's space map histogram is updated. This way
* we stay consistent on which parts of the metaslab space map's
* histogram are currently not available for allocations (e.g because
* they are in the defer, freed, and freeing trees).
*/
static void
metaslab_aux_histograms_update(metaslab_t *msp)
{
space_map_t *sm = msp->ms_sm;
ASSERT(sm != NULL);
/*
* This is similar to the metaslab's space map histogram updates
* that take place in metaslab_sync(). The only difference is that
* we only care about segments that haven't made it into the
* ms_allocatable tree yet.
*/
if (msp->ms_loaded) {
metaslab_aux_histograms_clear(msp);
metaslab_aux_histogram_add(msp->ms_synchist,
sm->sm_shift, msp->ms_freed);
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
metaslab_aux_histogram_add(msp->ms_deferhist[t],
sm->sm_shift, msp->ms_defer[t]);
}
}
metaslab_aux_histogram_add(msp->ms_synchist,
sm->sm_shift, msp->ms_freeing);
}
/*
* Called every time we are done syncing (writing to) the metaslab,
* i.e. at the end of each sync pass.
* [see the comment in metaslab_impl.h for ms_synchist, ms_deferhist]
*/
static void
metaslab_aux_histograms_update_done(metaslab_t *msp, boolean_t defer_allowed)
{
spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
space_map_t *sm = msp->ms_sm;
if (sm == NULL) {
/*
* We came here from metaslab_init() when creating/opening a
* pool, looking at a metaslab that hasn't had any allocations
* yet.
*/
return;
}
/*
* This is similar to the actions that we take for the ms_freed
* and ms_defer trees in metaslab_sync_done().
*/
uint64_t hist_index = spa_syncing_txg(spa) % TXG_DEFER_SIZE;
if (defer_allowed) {
bcopy(msp->ms_synchist, msp->ms_deferhist[hist_index],
sizeof (msp->ms_synchist));
} else {
bzero(msp->ms_deferhist[hist_index],
sizeof (msp->ms_deferhist[hist_index]));
}
bzero(msp->ms_synchist, sizeof (msp->ms_synchist));
}
/*
* Ensure that the metaslab's weight and fragmentation are consistent
* with the contents of the histogram (either the range tree's histogram
* or the space map's depending whether the metaslab is loaded).
*/
static void
metaslab_verify_weight_and_frag(metaslab_t *msp)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
if ((zfs_flags & ZFS_DEBUG_METASLAB_VERIFY) == 0)
return;
/*
* We can end up here from vdev_remove_complete(), in which case we
* cannot do these assertions because we hold spa config locks and
* thus we are not allowed to read from the DMU.
*
* We check if the metaslab group has been removed and if that's
* the case we return immediately as that would mean that we are
* here from the aforementioned code path.
*/
if (msp->ms_group == NULL)
return;
/*
* Devices being removed always return a weight of 0 and leave
* fragmentation and ms_max_size as is - there is nothing for
* us to verify here.
*/
vdev_t *vd = msp->ms_group->mg_vd;
if (vd->vdev_removing)
return;
/*
* If the metaslab is dirty it probably means that we've done
* some allocations or frees that have changed our histograms
* and thus the weight.
*/
for (int t = 0; t < TXG_SIZE; t++) {
if (txg_list_member(&vd->vdev_ms_list, msp, t))
return;
}
/*
* This verification checks that our in-memory state is consistent
* with what's on disk. If the pool is read-only then there aren't
* any changes and we just have the initially-loaded state.
*/
if (!spa_writeable(msp->ms_group->mg_vd->vdev_spa))
return;
/* some extra verification for in-core tree if you can */
if (msp->ms_loaded) {
range_tree_stat_verify(msp->ms_allocatable);
VERIFY(space_map_histogram_verify(msp->ms_sm,
msp->ms_allocatable));
}
uint64_t weight = msp->ms_weight;
uint64_t was_active = msp->ms_weight & METASLAB_ACTIVE_MASK;
boolean_t space_based = WEIGHT_IS_SPACEBASED(msp->ms_weight);
uint64_t frag = msp->ms_fragmentation;
uint64_t max_segsize = msp->ms_max_size;
msp->ms_weight = 0;
msp->ms_fragmentation = 0;
/*
* This function is used for verification purposes and thus should
* not introduce any side-effects/mutations on the system's state.
*
* Regardless of whether metaslab_weight() thinks this metaslab
* should be active or not, we want to ensure that the actual weight
* (and therefore the value of ms_weight) would be the same if it
* was to be recalculated at this point.
*
* In addition we set the nodirty flag so metaslab_weight() does
* not dirty the metaslab for future TXGs (e.g. when trying to
* force condensing to upgrade the metaslab spacemaps).
*/
msp->ms_weight = metaslab_weight(msp, B_TRUE) | was_active;
VERIFY3U(max_segsize, ==, msp->ms_max_size);
/*
* If the weight type changed then there is no point in doing
* verification. Revert fields to their original values.
*/
if ((space_based && !WEIGHT_IS_SPACEBASED(msp->ms_weight)) ||
(!space_based && WEIGHT_IS_SPACEBASED(msp->ms_weight))) {
msp->ms_fragmentation = frag;
msp->ms_weight = weight;
return;
}
VERIFY3U(msp->ms_fragmentation, ==, frag);
VERIFY3U(msp->ms_weight, ==, weight);
}
/*
* If we're over the zfs_metaslab_mem_limit, select the loaded metaslab from
* this class that was used longest ago, and attempt to unload it. We don't
* want to spend too much time in this loop to prevent performance
* degradation, and we expect that most of the time this operation will
* succeed. Between that and the normal unloading processing during txg sync,
* we expect this to keep the metaslab memory usage under control.
*/
static void
metaslab_potentially_evict(metaslab_class_t *mc)
{
#ifdef _KERNEL
uint64_t allmem = arc_all_memory();
uint64_t inuse = spl_kmem_cache_inuse(zfs_btree_leaf_cache);
uint64_t size = spl_kmem_cache_entry_size(zfs_btree_leaf_cache);
int tries = 0;
for (; allmem * zfs_metaslab_mem_limit / 100 < inuse * size &&
tries < multilist_get_num_sublists(&mc->mc_metaslab_txg_list) * 2;
tries++) {
unsigned int idx = multilist_get_random_index(
&mc->mc_metaslab_txg_list);
multilist_sublist_t *mls =
multilist_sublist_lock(&mc->mc_metaslab_txg_list, idx);
metaslab_t *msp = multilist_sublist_head(mls);
multilist_sublist_unlock(mls);
while (msp != NULL && allmem * zfs_metaslab_mem_limit / 100 <
inuse * size) {
VERIFY3P(mls, ==, multilist_sublist_lock(
&mc->mc_metaslab_txg_list, idx));
ASSERT3U(idx, ==,
metaslab_idx_func(&mc->mc_metaslab_txg_list, msp));
if (!multilist_link_active(&msp->ms_class_txg_node)) {
multilist_sublist_unlock(mls);
break;
}
metaslab_t *next_msp = multilist_sublist_next(mls, msp);
multilist_sublist_unlock(mls);
/*
* If the metaslab is currently loading there are two
* cases. If it's the metaslab we're evicting, we
* can't continue on or we'll panic when we attempt to
* recursively lock the mutex. If it's another
* metaslab that's loading, it can be safely skipped,
* since we know it's very new and therefore not a
* good eviction candidate. We check later once the
* lock is held that the metaslab is fully loaded
* before actually unloading it.
*/
if (msp->ms_loading) {
msp = next_msp;
inuse =
spl_kmem_cache_inuse(zfs_btree_leaf_cache);
continue;
}
/*
* We can't unload metaslabs with no spacemap because
* they're not ready to be unloaded yet. We can't
* unload metaslabs with outstanding allocations
* because doing so could cause the metaslab's weight
* to decrease while it's unloaded, which violates an
* invariant that we use to prevent unnecessary
* loading. We also don't unload metaslabs that are
* currently active because they are high-weight
* metaslabs that are likely to be used in the near
* future.
*/
mutex_enter(&msp->ms_lock);
if (msp->ms_allocator == -1 && msp->ms_sm != NULL &&
msp->ms_allocating_total == 0) {
metaslab_unload(msp);
}
mutex_exit(&msp->ms_lock);
msp = next_msp;
inuse = spl_kmem_cache_inuse(zfs_btree_leaf_cache);
}
}
#else
- (void) mc;
+ (void) mc, (void) zfs_metaslab_mem_limit;
#endif
}
static int
metaslab_load_impl(metaslab_t *msp)
{
int error = 0;
ASSERT(MUTEX_HELD(&msp->ms_lock));
ASSERT(msp->ms_loading);
ASSERT(!msp->ms_condensing);
/*
* We temporarily drop the lock to unblock other operations while we
* are reading the space map. Therefore, metaslab_sync() and
* metaslab_sync_done() can run at the same time as we do.
*
* If we are using the log space maps, metaslab_sync() can't write to
* the metaslab's space map while we are loading as we only write to
* it when we are flushing the metaslab, and that can't happen while
* we are loading it.
*
* If we are not using log space maps though, metaslab_sync() can
* append to the space map while we are loading. Therefore we load
* only entries that existed when we started the load. Additionally,
* metaslab_sync_done() has to wait for the load to complete because
* there are potential races like metaslab_load() loading parts of the
* space map that are currently being appended by metaslab_sync(). If
* we didn't, the ms_allocatable would have entries that
* metaslab_sync_done() would try to re-add later.
*
* That's why before dropping the lock we remember the synced length
* of the metaslab and read up to that point of the space map,
* ignoring entries appended by metaslab_sync() that happen after we
* drop the lock.
*/
uint64_t length = msp->ms_synced_length;
mutex_exit(&msp->ms_lock);
hrtime_t load_start = gethrtime();
metaslab_rt_arg_t *mrap;
if (msp->ms_allocatable->rt_arg == NULL) {
mrap = kmem_zalloc(sizeof (*mrap), KM_SLEEP);
} else {
mrap = msp->ms_allocatable->rt_arg;
msp->ms_allocatable->rt_ops = NULL;
msp->ms_allocatable->rt_arg = NULL;
}
mrap->mra_bt = &msp->ms_allocatable_by_size;
mrap->mra_floor_shift = metaslab_by_size_min_shift;
if (msp->ms_sm != NULL) {
error = space_map_load_length(msp->ms_sm, msp->ms_allocatable,
SM_FREE, length);
/* Now, populate the size-sorted tree. */
metaslab_rt_create(msp->ms_allocatable, mrap);
msp->ms_allocatable->rt_ops = &metaslab_rt_ops;
msp->ms_allocatable->rt_arg = mrap;
struct mssa_arg arg = {0};
arg.rt = msp->ms_allocatable;
arg.mra = mrap;
range_tree_walk(msp->ms_allocatable, metaslab_size_sorted_add,
&arg);
} else {
/*
* Add the size-sorted tree first, since we don't need to load
* the metaslab from the spacemap.
*/
metaslab_rt_create(msp->ms_allocatable, mrap);
msp->ms_allocatable->rt_ops = &metaslab_rt_ops;
msp->ms_allocatable->rt_arg = mrap;
/*
* The space map has not been allocated yet, so treat
* all the space in the metaslab as free and add it to the
* ms_allocatable tree.
*/
range_tree_add(msp->ms_allocatable,
msp->ms_start, msp->ms_size);
if (msp->ms_new) {
/*
* If the ms_sm doesn't exist, this means that this
* metaslab hasn't gone through metaslab_sync() and
* thus has never been dirtied. So we shouldn't
* expect any unflushed allocs or frees from previous
* TXGs.
*/
ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));
}
}
/*
* We need to grab the ms_sync_lock to prevent metaslab_sync() from
* changing the ms_sm (or log_sm) and the metaslab's range trees
* while we are about to use them and populate the ms_allocatable.
* The ms_lock is insufficient for this because metaslab_sync() doesn't
* hold the ms_lock while writing the ms_checkpointing tree to disk.
*/
mutex_enter(&msp->ms_sync_lock);
mutex_enter(&msp->ms_lock);
ASSERT(!msp->ms_condensing);
ASSERT(!msp->ms_flushing);
if (error != 0) {
mutex_exit(&msp->ms_sync_lock);
return (error);
}
ASSERT3P(msp->ms_group, !=, NULL);
msp->ms_loaded = B_TRUE;
/*
* Apply all the unflushed changes to ms_allocatable right
* away so any manipulations we do below have a clear view
* of what is allocated and what is free.
*/
range_tree_walk(msp->ms_unflushed_allocs,
range_tree_remove, msp->ms_allocatable);
range_tree_walk(msp->ms_unflushed_frees,
range_tree_add, msp->ms_allocatable);
ASSERT3P(msp->ms_group, !=, NULL);
spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
if (spa_syncing_log_sm(spa) != NULL) {
ASSERT(spa_feature_is_enabled(spa,
SPA_FEATURE_LOG_SPACEMAP));
/*
* If we use a log space map we add all the segments
* that are in ms_unflushed_frees so they are available
* for allocation.
*
* ms_allocatable needs to contain all free segments
* that are ready for allocations (thus not segments
* from ms_freeing, ms_freed, and the ms_defer trees).
* But if we grab the lock in this code path at a sync
* pass later that 1, then it also contains the
* segments of ms_freed (they were added to it earlier
* in this path through ms_unflushed_frees). So we
* need to remove all the segments that exist in
* ms_freed from ms_allocatable as they will be added
* later in metaslab_sync_done().
*
* When there's no log space map, the ms_allocatable
* correctly doesn't contain any segments that exist
* in ms_freed [see ms_synced_length].
*/
range_tree_walk(msp->ms_freed,
range_tree_remove, msp->ms_allocatable);
}
/*
* If we are not using the log space map, ms_allocatable
* contains the segments that exist in the ms_defer trees
* [see ms_synced_length]. Thus we need to remove them
* from ms_allocatable as they will be added again in
* metaslab_sync_done().
*
* If we are using the log space map, ms_allocatable still
* contains the segments that exist in the ms_defer trees.
* Not because it read them through the ms_sm though. But
* because these segments are part of ms_unflushed_frees
* whose segments we add to ms_allocatable earlier in this
* code path.
*/
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
range_tree_walk(msp->ms_defer[t],
range_tree_remove, msp->ms_allocatable);
}
/*
* Call metaslab_recalculate_weight_and_sort() now that the
* metaslab is loaded so we get the metaslab's real weight.
*
* Unless this metaslab was created with older software and
* has not yet been converted to use segment-based weight, we
* expect the new weight to be better or equal to the weight
* that the metaslab had while it was not loaded. This is
* because the old weight does not take into account the
* consolidation of adjacent segments between TXGs. [see
* comment for ms_synchist and ms_deferhist[] for more info]
*/
uint64_t weight = msp->ms_weight;
uint64_t max_size = msp->ms_max_size;
metaslab_recalculate_weight_and_sort(msp);
if (!WEIGHT_IS_SPACEBASED(weight))
ASSERT3U(weight, <=, msp->ms_weight);
msp->ms_max_size = metaslab_largest_allocatable(msp);
ASSERT3U(max_size, <=, msp->ms_max_size);
hrtime_t load_end = gethrtime();
msp->ms_load_time = load_end;
zfs_dbgmsg("metaslab_load: txg %llu, spa %s, vdev_id %llu, "
"ms_id %llu, smp_length %llu, "
"unflushed_allocs %llu, unflushed_frees %llu, "
"freed %llu, defer %llu + %llu, unloaded time %llu ms, "
"loading_time %lld ms, ms_max_size %llu, "
"max size error %lld, "
"old_weight %llx, new_weight %llx",
(u_longlong_t)spa_syncing_txg(spa), spa_name(spa),
(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
(u_longlong_t)msp->ms_id,
(u_longlong_t)space_map_length(msp->ms_sm),
(u_longlong_t)range_tree_space(msp->ms_unflushed_allocs),
(u_longlong_t)range_tree_space(msp->ms_unflushed_frees),
(u_longlong_t)range_tree_space(msp->ms_freed),
(u_longlong_t)range_tree_space(msp->ms_defer[0]),
(u_longlong_t)range_tree_space(msp->ms_defer[1]),
(longlong_t)((load_start - msp->ms_unload_time) / 1000000),
(longlong_t)((load_end - load_start) / 1000000),
(u_longlong_t)msp->ms_max_size,
(u_longlong_t)msp->ms_max_size - max_size,
(u_longlong_t)weight, (u_longlong_t)msp->ms_weight);
metaslab_verify_space(msp, spa_syncing_txg(spa));
mutex_exit(&msp->ms_sync_lock);
return (0);
}
int
metaslab_load(metaslab_t *msp)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
/*
* There may be another thread loading the same metaslab, if that's
* the case just wait until the other thread is done and return.
*/
metaslab_load_wait(msp);
if (msp->ms_loaded)
return (0);
VERIFY(!msp->ms_loading);
ASSERT(!msp->ms_condensing);
/*
* We set the loading flag BEFORE potentially dropping the lock to
* wait for an ongoing flush (see ms_flushing below). This way other
* threads know that there is already a thread that is loading this
* metaslab.
*/
msp->ms_loading = B_TRUE;
/*
* Wait for any in-progress flushing to finish as we drop the ms_lock
* both here (during space_map_load()) and in metaslab_flush() (when
* we flush our changes to the ms_sm).
*/
if (msp->ms_flushing)
metaslab_flush_wait(msp);
/*
* In the possibility that we were waiting for the metaslab to be
* flushed (where we temporarily dropped the ms_lock), ensure that
* no one else loaded the metaslab somehow.
*/
ASSERT(!msp->ms_loaded);
/*
* If we're loading a metaslab in the normal class, consider evicting
* another one to keep our memory usage under the limit defined by the
* zfs_metaslab_mem_limit tunable.
*/
if (spa_normal_class(msp->ms_group->mg_class->mc_spa) ==
msp->ms_group->mg_class) {
metaslab_potentially_evict(msp->ms_group->mg_class);
}
int error = metaslab_load_impl(msp);
ASSERT(MUTEX_HELD(&msp->ms_lock));
msp->ms_loading = B_FALSE;
cv_broadcast(&msp->ms_load_cv);
return (error);
}
void
metaslab_unload(metaslab_t *msp)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
/*
* This can happen if a metaslab is selected for eviction (in
* metaslab_potentially_evict) and then unloaded during spa_sync (via
* metaslab_class_evict_old).
*/
if (!msp->ms_loaded)
return;
range_tree_vacate(msp->ms_allocatable, NULL, NULL);
msp->ms_loaded = B_FALSE;
msp->ms_unload_time = gethrtime();
msp->ms_activation_weight = 0;
msp->ms_weight &= ~METASLAB_ACTIVE_MASK;
if (msp->ms_group != NULL) {
metaslab_class_t *mc = msp->ms_group->mg_class;
multilist_sublist_t *mls =
multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
if (multilist_link_active(&msp->ms_class_txg_node))
multilist_sublist_remove(mls, msp);
multilist_sublist_unlock(mls);
spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
zfs_dbgmsg("metaslab_unload: txg %llu, spa %s, vdev_id %llu, "
"ms_id %llu, weight %llx, "
"selected txg %llu (%llu ms ago), alloc_txg %llu, "
"loaded %llu ms ago, max_size %llu",
(u_longlong_t)spa_syncing_txg(spa), spa_name(spa),
(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
(u_longlong_t)msp->ms_id,
(u_longlong_t)msp->ms_weight,
(u_longlong_t)msp->ms_selected_txg,
(u_longlong_t)(msp->ms_unload_time -
msp->ms_selected_time) / 1000 / 1000,
(u_longlong_t)msp->ms_alloc_txg,
(u_longlong_t)(msp->ms_unload_time -
msp->ms_load_time) / 1000 / 1000,
(u_longlong_t)msp->ms_max_size);
}
/*
* We explicitly recalculate the metaslab's weight based on its space
* map (as it is now not loaded). We want unload metaslabs to always
* have their weights calculated from the space map histograms, while
* loaded ones have it calculated from their in-core range tree
* [see metaslab_load()]. This way, the weight reflects the information
* available in-core, whether it is loaded or not.
*
* If ms_group == NULL means that we came here from metaslab_fini(),
* at which point it doesn't make sense for us to do the recalculation
* and the sorting.
*/
if (msp->ms_group != NULL)
metaslab_recalculate_weight_and_sort(msp);
}
/*
* We want to optimize the memory use of the per-metaslab range
* trees. To do this, we store the segments in the range trees in
* units of sectors, zero-indexing from the start of the metaslab. If
* the vdev_ms_shift - the vdev_ashift is less than 32, we can store
* the ranges using two uint32_ts, rather than two uint64_ts.
*/
range_seg_type_t
metaslab_calculate_range_tree_type(vdev_t *vdev, metaslab_t *msp,
uint64_t *start, uint64_t *shift)
{
if (vdev->vdev_ms_shift - vdev->vdev_ashift < 32 &&
!zfs_metaslab_force_large_segs) {
*shift = vdev->vdev_ashift;
*start = msp->ms_start;
return (RANGE_SEG32);
} else {
*shift = 0;
*start = 0;
return (RANGE_SEG64);
}
}
void
metaslab_set_selected_txg(metaslab_t *msp, uint64_t txg)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
metaslab_class_t *mc = msp->ms_group->mg_class;
multilist_sublist_t *mls =
multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
if (multilist_link_active(&msp->ms_class_txg_node))
multilist_sublist_remove(mls, msp);
msp->ms_selected_txg = txg;
msp->ms_selected_time = gethrtime();
multilist_sublist_insert_tail(mls, msp);
multilist_sublist_unlock(mls);
}
void
metaslab_space_update(vdev_t *vd, metaslab_class_t *mc, int64_t alloc_delta,
int64_t defer_delta, int64_t space_delta)
{
vdev_space_update(vd, alloc_delta, defer_delta, space_delta);
ASSERT3P(vd->vdev_spa->spa_root_vdev, ==, vd->vdev_parent);
ASSERT(vd->vdev_ms_count != 0);
metaslab_class_space_update(mc, alloc_delta, defer_delta, space_delta,
vdev_deflated_space(vd, space_delta));
}
int
metaslab_init(metaslab_group_t *mg, uint64_t id, uint64_t object,
uint64_t txg, metaslab_t **msp)
{
vdev_t *vd = mg->mg_vd;
spa_t *spa = vd->vdev_spa;
objset_t *mos = spa->spa_meta_objset;
metaslab_t *ms;
int error;
ms = kmem_zalloc(sizeof (metaslab_t), KM_SLEEP);
mutex_init(&ms->ms_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&ms->ms_sync_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&ms->ms_load_cv, NULL, CV_DEFAULT, NULL);
cv_init(&ms->ms_flush_cv, NULL, CV_DEFAULT, NULL);
multilist_link_init(&ms->ms_class_txg_node);
ms->ms_id = id;
ms->ms_start = id << vd->vdev_ms_shift;
ms->ms_size = 1ULL << vd->vdev_ms_shift;
ms->ms_allocator = -1;
ms->ms_new = B_TRUE;
vdev_ops_t *ops = vd->vdev_ops;
if (ops->vdev_op_metaslab_init != NULL)
ops->vdev_op_metaslab_init(vd, &ms->ms_start, &ms->ms_size);
/*
* We only open space map objects that already exist. All others
* will be opened when we finally allocate an object for it. For
* readonly pools there is no need to open the space map object.
*
* Note:
* When called from vdev_expand(), we can't call into the DMU as
* we are holding the spa_config_lock as a writer and we would
* deadlock [see relevant comment in vdev_metaslab_init()]. in
* that case, the object parameter is zero though, so we won't
* call into the DMU.
*/
if (object != 0 && !(spa->spa_mode == SPA_MODE_READ &&
!spa->spa_read_spacemaps)) {
error = space_map_open(&ms->ms_sm, mos, object, ms->ms_start,
ms->ms_size, vd->vdev_ashift);
if (error != 0) {
kmem_free(ms, sizeof (metaslab_t));
return (error);
}
ASSERT(ms->ms_sm != NULL);
ms->ms_allocated_space = space_map_allocated(ms->ms_sm);
}
uint64_t shift, start;
range_seg_type_t type =
metaslab_calculate_range_tree_type(vd, ms, &start, &shift);
ms->ms_allocatable = range_tree_create(NULL, type, NULL, start, shift);
for (int t = 0; t < TXG_SIZE; t++) {
ms->ms_allocating[t] = range_tree_create(NULL, type,
NULL, start, shift);
}
ms->ms_freeing = range_tree_create(NULL, type, NULL, start, shift);
ms->ms_freed = range_tree_create(NULL, type, NULL, start, shift);
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
ms->ms_defer[t] = range_tree_create(NULL, type, NULL,
start, shift);
}
ms->ms_checkpointing =
range_tree_create(NULL, type, NULL, start, shift);
ms->ms_unflushed_allocs =
range_tree_create(NULL, type, NULL, start, shift);
metaslab_rt_arg_t *mrap = kmem_zalloc(sizeof (*mrap), KM_SLEEP);
mrap->mra_bt = &ms->ms_unflushed_frees_by_size;
mrap->mra_floor_shift = metaslab_by_size_min_shift;
ms->ms_unflushed_frees = range_tree_create(&metaslab_rt_ops,
type, mrap, start, shift);
ms->ms_trim = range_tree_create(NULL, type, NULL, start, shift);
metaslab_group_add(mg, ms);
metaslab_set_fragmentation(ms, B_FALSE);
/*
* If we're opening an existing pool (txg == 0) or creating
* a new one (txg == TXG_INITIAL), all space is available now.
* If we're adding space to an existing pool, the new space
* does not become available until after this txg has synced.
* The metaslab's weight will also be initialized when we sync
* out this txg. This ensures that we don't attempt to allocate
* from it before we have initialized it completely.
*/
if (txg <= TXG_INITIAL) {
metaslab_sync_done(ms, 0);
metaslab_space_update(vd, mg->mg_class,
metaslab_allocated_space(ms), 0, 0);
}
if (txg != 0) {
vdev_dirty(vd, 0, NULL, txg);
vdev_dirty(vd, VDD_METASLAB, ms, txg);
}
*msp = ms;
return (0);
}
static void
metaslab_fini_flush_data(metaslab_t *msp)
{
spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
if (metaslab_unflushed_txg(msp) == 0) {
ASSERT3P(avl_find(&spa->spa_metaslabs_by_flushed, msp, NULL),
==, NULL);
return;
}
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));
mutex_enter(&spa->spa_flushed_ms_lock);
avl_remove(&spa->spa_metaslabs_by_flushed, msp);
mutex_exit(&spa->spa_flushed_ms_lock);
spa_log_sm_decrement_mscount(spa, metaslab_unflushed_txg(msp));
spa_log_summary_decrement_mscount(spa, metaslab_unflushed_txg(msp));
}
uint64_t
metaslab_unflushed_changes_memused(metaslab_t *ms)
{
return ((range_tree_numsegs(ms->ms_unflushed_allocs) +
range_tree_numsegs(ms->ms_unflushed_frees)) *
ms->ms_unflushed_allocs->rt_root.bt_elem_size);
}
void
metaslab_fini(metaslab_t *msp)
{
metaslab_group_t *mg = msp->ms_group;
vdev_t *vd = mg->mg_vd;
spa_t *spa = vd->vdev_spa;
metaslab_fini_flush_data(msp);
metaslab_group_remove(mg, msp);
mutex_enter(&msp->ms_lock);
VERIFY(msp->ms_group == NULL);
/*
* If this metaslab hasn't been through metaslab_sync_done() yet its
* space hasn't been accounted for in its vdev and doesn't need to be
* subtracted.
*/
if (!msp->ms_new) {
metaslab_space_update(vd, mg->mg_class,
-metaslab_allocated_space(msp), 0, -msp->ms_size);
}
space_map_close(msp->ms_sm);
msp->ms_sm = NULL;
metaslab_unload(msp);
range_tree_destroy(msp->ms_allocatable);
range_tree_destroy(msp->ms_freeing);
range_tree_destroy(msp->ms_freed);
ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
metaslab_unflushed_changes_memused(msp));
spa->spa_unflushed_stats.sus_memused -=
metaslab_unflushed_changes_memused(msp);
range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
range_tree_destroy(msp->ms_unflushed_allocs);
range_tree_destroy(msp->ms_checkpointing);
range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);
range_tree_destroy(msp->ms_unflushed_frees);
for (int t = 0; t < TXG_SIZE; t++) {
range_tree_destroy(msp->ms_allocating[t]);
}
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
range_tree_destroy(msp->ms_defer[t]);
}
ASSERT0(msp->ms_deferspace);
for (int t = 0; t < TXG_SIZE; t++)
ASSERT(!txg_list_member(&vd->vdev_ms_list, msp, t));
range_tree_vacate(msp->ms_trim, NULL, NULL);
range_tree_destroy(msp->ms_trim);
mutex_exit(&msp->ms_lock);
cv_destroy(&msp->ms_load_cv);
cv_destroy(&msp->ms_flush_cv);
mutex_destroy(&msp->ms_lock);
mutex_destroy(&msp->ms_sync_lock);
ASSERT3U(msp->ms_allocator, ==, -1);
kmem_free(msp, sizeof (metaslab_t));
}
#define FRAGMENTATION_TABLE_SIZE 17
/*
* This table defines a segment size based fragmentation metric that will
* allow each metaslab to derive its own fragmentation value. This is done
* by calculating the space in each bucket of the spacemap histogram and
* multiplying that by the fragmentation metric in this table. Doing
* this for all buckets and dividing it by the total amount of free
* space in this metaslab (i.e. the total free space in all buckets) gives
* us the fragmentation metric. This means that a high fragmentation metric
* equates to most of the free space being comprised of small segments.
* Conversely, if the metric is low, then most of the free space is in
* large segments. A 10% change in fragmentation equates to approximately
* double the number of segments.
*
* This table defines 0% fragmented space using 16MB segments. Testing has
* shown that segments that are greater than or equal to 16MB do not suffer
* from drastic performance problems. Using this value, we derive the rest
* of the table. Since the fragmentation value is never stored on disk, it
* is possible to change these calculations in the future.
*/
-int zfs_frag_table[FRAGMENTATION_TABLE_SIZE] = {
+static const int zfs_frag_table[FRAGMENTATION_TABLE_SIZE] = {
100, /* 512B */
100, /* 1K */
98, /* 2K */
95, /* 4K */
90, /* 8K */
80, /* 16K */
70, /* 32K */
60, /* 64K */
50, /* 128K */
40, /* 256K */
30, /* 512K */
20, /* 1M */
15, /* 2M */
10, /* 4M */
5, /* 8M */
0 /* 16M */
};
/*
* Calculate the metaslab's fragmentation metric and set ms_fragmentation.
* Setting this value to ZFS_FRAG_INVALID means that the metaslab has not
* been upgraded and does not support this metric. Otherwise, the return
* value should be in the range [0, 100].
*/
static void
metaslab_set_fragmentation(metaslab_t *msp, boolean_t nodirty)
{
spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
uint64_t fragmentation = 0;
uint64_t total = 0;
boolean_t feature_enabled = spa_feature_is_enabled(spa,
SPA_FEATURE_SPACEMAP_HISTOGRAM);
if (!feature_enabled) {
msp->ms_fragmentation = ZFS_FRAG_INVALID;
return;
}
/*
* A null space map means that the entire metaslab is free
* and thus is not fragmented.
*/
if (msp->ms_sm == NULL) {
msp->ms_fragmentation = 0;
return;
}
/*
* If this metaslab's space map has not been upgraded, flag it
* so that we upgrade next time we encounter it.
*/
if (msp->ms_sm->sm_dbuf->db_size != sizeof (space_map_phys_t)) {
uint64_t txg = spa_syncing_txg(spa);
vdev_t *vd = msp->ms_group->mg_vd;
/*
* If we've reached the final dirty txg, then we must
* be shutting down the pool. We don't want to dirty
* any data past this point so skip setting the condense
* flag. We can retry this action the next time the pool
* is imported. We also skip marking this metaslab for
* condensing if the caller has explicitly set nodirty.
*/
if (!nodirty &&
spa_writeable(spa) && txg < spa_final_dirty_txg(spa)) {
msp->ms_condense_wanted = B_TRUE;
vdev_dirty(vd, VDD_METASLAB, msp, txg + 1);
zfs_dbgmsg("txg %llu, requesting force condense: "
"ms_id %llu, vdev_id %llu", (u_longlong_t)txg,
(u_longlong_t)msp->ms_id,
(u_longlong_t)vd->vdev_id);
}
msp->ms_fragmentation = ZFS_FRAG_INVALID;
return;
}
for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
uint64_t space = 0;
uint8_t shift = msp->ms_sm->sm_shift;
int idx = MIN(shift - SPA_MINBLOCKSHIFT + i,
FRAGMENTATION_TABLE_SIZE - 1);
if (msp->ms_sm->sm_phys->smp_histogram[i] == 0)
continue;
space = msp->ms_sm->sm_phys->smp_histogram[i] << (i + shift);
total += space;
ASSERT3U(idx, <, FRAGMENTATION_TABLE_SIZE);
fragmentation += space * zfs_frag_table[idx];
}
if (total > 0)
fragmentation /= total;
ASSERT3U(fragmentation, <=, 100);
msp->ms_fragmentation = fragmentation;
}
/*
* Compute a weight -- a selection preference value -- for the given metaslab.
* This is based on the amount of free space, the level of fragmentation,
* the LBA range, and whether the metaslab is loaded.
*/
static uint64_t
metaslab_space_weight(metaslab_t *msp)
{
metaslab_group_t *mg = msp->ms_group;
vdev_t *vd = mg->mg_vd;
uint64_t weight, space;
ASSERT(MUTEX_HELD(&msp->ms_lock));
/*
* The baseline weight is the metaslab's free space.
*/
space = msp->ms_size - metaslab_allocated_space(msp);
if (metaslab_fragmentation_factor_enabled &&
msp->ms_fragmentation != ZFS_FRAG_INVALID) {
/*
* Use the fragmentation information to inversely scale
* down the baseline weight. We need to ensure that we
* don't exclude this metaslab completely when it's 100%
* fragmented. To avoid this we reduce the fragmented value
* by 1.
*/
space = (space * (100 - (msp->ms_fragmentation - 1))) / 100;
/*
* If space < SPA_MINBLOCKSIZE, then we will not allocate from
* this metaslab again. The fragmentation metric may have
* decreased the space to something smaller than
* SPA_MINBLOCKSIZE, so reset the space to SPA_MINBLOCKSIZE
* so that we can consume any remaining space.
*/
if (space > 0 && space < SPA_MINBLOCKSIZE)
space = SPA_MINBLOCKSIZE;
}
weight = space;
/*
* Modern disks have uniform bit density and constant angular velocity.
* Therefore, the outer recording zones are faster (higher bandwidth)
* than the inner zones by the ratio of outer to inner track diameter,
* which is typically around 2:1. We account for this by assigning
* higher weight to lower metaslabs (multiplier ranging from 2x to 1x).
* In effect, this means that we'll select the metaslab with the most
* free bandwidth rather than simply the one with the most free space.
*/
if (!vd->vdev_nonrot && metaslab_lba_weighting_enabled) {
weight = 2 * weight - (msp->ms_id * weight) / vd->vdev_ms_count;
ASSERT(weight >= space && weight <= 2 * space);
}
/*
* If this metaslab is one we're actively using, adjust its
* weight to make it preferable to any inactive metaslab so
* we'll polish it off. If the fragmentation on this metaslab
* has exceed our threshold, then don't mark it active.
*/
if (msp->ms_loaded && msp->ms_fragmentation != ZFS_FRAG_INVALID &&
msp->ms_fragmentation <= zfs_metaslab_fragmentation_threshold) {
weight |= (msp->ms_weight & METASLAB_ACTIVE_MASK);
}
WEIGHT_SET_SPACEBASED(weight);
return (weight);
}
/*
* Return the weight of the specified metaslab, according to the segment-based
* weighting algorithm. The metaslab must be loaded. This function can
* be called within a sync pass since it relies only on the metaslab's
* range tree which is always accurate when the metaslab is loaded.
*/
static uint64_t
metaslab_weight_from_range_tree(metaslab_t *msp)
{
uint64_t weight = 0;
uint32_t segments = 0;
ASSERT(msp->ms_loaded);
for (int i = RANGE_TREE_HISTOGRAM_SIZE - 1; i >= SPA_MINBLOCKSHIFT;
i--) {
uint8_t shift = msp->ms_group->mg_vd->vdev_ashift;
int max_idx = SPACE_MAP_HISTOGRAM_SIZE + shift - 1;
segments <<= 1;
segments += msp->ms_allocatable->rt_histogram[i];
/*
* The range tree provides more precision than the space map
* and must be downgraded so that all values fit within the
* space map's histogram. This allows us to compare loaded
* vs. unloaded metaslabs to determine which metaslab is
* considered "best".
*/
if (i > max_idx)
continue;
if (segments != 0) {
WEIGHT_SET_COUNT(weight, segments);
WEIGHT_SET_INDEX(weight, i);
WEIGHT_SET_ACTIVE(weight, 0);
break;
}
}
return (weight);
}
/*
* Calculate the weight based on the on-disk histogram. Should be applied
* only to unloaded metaslabs (i.e no incoming allocations) in-order to
* give results consistent with the on-disk state
*/
static uint64_t
metaslab_weight_from_spacemap(metaslab_t *msp)
{
space_map_t *sm = msp->ms_sm;
ASSERT(!msp->ms_loaded);
ASSERT(sm != NULL);
ASSERT3U(space_map_object(sm), !=, 0);
ASSERT3U(sm->sm_dbuf->db_size, ==, sizeof (space_map_phys_t));
/*
* Create a joint histogram from all the segments that have made
* it to the metaslab's space map histogram, that are not yet
* available for allocation because they are still in the freeing
* pipeline (e.g. freeing, freed, and defer trees). Then subtract
* these segments from the space map's histogram to get a more
* accurate weight.
*/
uint64_t deferspace_histogram[SPACE_MAP_HISTOGRAM_SIZE] = {0};
for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++)
deferspace_histogram[i] += msp->ms_synchist[i];
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
deferspace_histogram[i] += msp->ms_deferhist[t][i];
}
}
uint64_t weight = 0;
for (int i = SPACE_MAP_HISTOGRAM_SIZE - 1; i >= 0; i--) {
ASSERT3U(sm->sm_phys->smp_histogram[i], >=,
deferspace_histogram[i]);
uint64_t count =
sm->sm_phys->smp_histogram[i] - deferspace_histogram[i];
if (count != 0) {
WEIGHT_SET_COUNT(weight, count);
WEIGHT_SET_INDEX(weight, i + sm->sm_shift);
WEIGHT_SET_ACTIVE(weight, 0);
break;
}
}
return (weight);
}
/*
* Compute a segment-based weight for the specified metaslab. The weight
* is determined by highest bucket in the histogram. The information
* for the highest bucket is encoded into the weight value.
*/
static uint64_t
metaslab_segment_weight(metaslab_t *msp)
{
metaslab_group_t *mg = msp->ms_group;
uint64_t weight = 0;
uint8_t shift = mg->mg_vd->vdev_ashift;
ASSERT(MUTEX_HELD(&msp->ms_lock));
/*
* The metaslab is completely free.
*/
if (metaslab_allocated_space(msp) == 0) {
int idx = highbit64(msp->ms_size) - 1;
int max_idx = SPACE_MAP_HISTOGRAM_SIZE + shift - 1;
if (idx < max_idx) {
WEIGHT_SET_COUNT(weight, 1ULL);
WEIGHT_SET_INDEX(weight, idx);
} else {
WEIGHT_SET_COUNT(weight, 1ULL << (idx - max_idx));
WEIGHT_SET_INDEX(weight, max_idx);
}
WEIGHT_SET_ACTIVE(weight, 0);
ASSERT(!WEIGHT_IS_SPACEBASED(weight));
return (weight);
}
ASSERT3U(msp->ms_sm->sm_dbuf->db_size, ==, sizeof (space_map_phys_t));
/*
* If the metaslab is fully allocated then just make the weight 0.
*/
if (metaslab_allocated_space(msp) == msp->ms_size)
return (0);
/*
* If the metaslab is already loaded, then use the range tree to
* determine the weight. Otherwise, we rely on the space map information
* to generate the weight.
*/
if (msp->ms_loaded) {
weight = metaslab_weight_from_range_tree(msp);
} else {
weight = metaslab_weight_from_spacemap(msp);
}
/*
* If the metaslab was active the last time we calculated its weight
* then keep it active. We want to consume the entire region that
* is associated with this weight.
*/
if (msp->ms_activation_weight != 0 && weight != 0)
WEIGHT_SET_ACTIVE(weight, WEIGHT_GET_ACTIVE(msp->ms_weight));
return (weight);
}
/*
* Determine if we should attempt to allocate from this metaslab. If the
* metaslab is loaded, then we can determine if the desired allocation
* can be satisfied by looking at the size of the maximum free segment
* on that metaslab. Otherwise, we make our decision based on the metaslab's
* weight. For segment-based weighting we can determine the maximum
* allocation based on the index encoded in its value. For space-based
* weights we rely on the entire weight (excluding the weight-type bit).
*/
static boolean_t
metaslab_should_allocate(metaslab_t *msp, uint64_t asize, boolean_t try_hard)
{
/*
* If the metaslab is loaded, ms_max_size is definitive and we can use
* the fast check. If it's not, the ms_max_size is a lower bound (once
* set), and we should use the fast check as long as we're not in
* try_hard and it's been less than zfs_metaslab_max_size_cache_sec
* seconds since the metaslab was unloaded.
*/
if (msp->ms_loaded ||
(msp->ms_max_size != 0 && !try_hard && gethrtime() <
msp->ms_unload_time + SEC2NSEC(zfs_metaslab_max_size_cache_sec)))
return (msp->ms_max_size >= asize);
boolean_t should_allocate;
if (!WEIGHT_IS_SPACEBASED(msp->ms_weight)) {
/*
* The metaslab segment weight indicates segments in the
* range [2^i, 2^(i+1)), where i is the index in the weight.
* Since the asize might be in the middle of the range, we
* should attempt the allocation if asize < 2^(i+1).
*/
should_allocate = (asize <
1ULL << (WEIGHT_GET_INDEX(msp->ms_weight) + 1));
} else {
should_allocate = (asize <=
(msp->ms_weight & ~METASLAB_WEIGHT_TYPE));
}
return (should_allocate);
}
static uint64_t
metaslab_weight(metaslab_t *msp, boolean_t nodirty)
{
vdev_t *vd = msp->ms_group->mg_vd;
spa_t *spa = vd->vdev_spa;
uint64_t weight;
ASSERT(MUTEX_HELD(&msp->ms_lock));
metaslab_set_fragmentation(msp, nodirty);
/*
* Update the maximum size. If the metaslab is loaded, this will
* ensure that we get an accurate maximum size if newly freed space
* has been added back into the free tree. If the metaslab is
* unloaded, we check if there's a larger free segment in the
* unflushed frees. This is a lower bound on the largest allocatable
* segment size. Coalescing of adjacent entries may reveal larger
* allocatable segments, but we aren't aware of those until loading
* the space map into a range tree.
*/
if (msp->ms_loaded) {
msp->ms_max_size = metaslab_largest_allocatable(msp);
} else {
msp->ms_max_size = MAX(msp->ms_max_size,
metaslab_largest_unflushed_free(msp));
}
/*
* Segment-based weighting requires space map histogram support.
*/
if (zfs_metaslab_segment_weight_enabled &&
spa_feature_is_enabled(spa, SPA_FEATURE_SPACEMAP_HISTOGRAM) &&
(msp->ms_sm == NULL || msp->ms_sm->sm_dbuf->db_size ==
sizeof (space_map_phys_t))) {
weight = metaslab_segment_weight(msp);
} else {
weight = metaslab_space_weight(msp);
}
return (weight);
}
void
metaslab_recalculate_weight_and_sort(metaslab_t *msp)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
/* note: we preserve the mask (e.g. indication of primary, etc..) */
uint64_t was_active = msp->ms_weight & METASLAB_ACTIVE_MASK;
metaslab_group_sort(msp->ms_group, msp,
metaslab_weight(msp, B_FALSE) | was_active);
}
static int
metaslab_activate_allocator(metaslab_group_t *mg, metaslab_t *msp,
int allocator, uint64_t activation_weight)
{
metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
ASSERT(MUTEX_HELD(&msp->ms_lock));
/*
* If we're activating for the claim code, we don't want to actually
* set the metaslab up for a specific allocator.
*/
if (activation_weight == METASLAB_WEIGHT_CLAIM) {
ASSERT0(msp->ms_activation_weight);
msp->ms_activation_weight = msp->ms_weight;
metaslab_group_sort(mg, msp, msp->ms_weight |
activation_weight);
return (0);
}
metaslab_t **mspp = (activation_weight == METASLAB_WEIGHT_PRIMARY ?
&mga->mga_primary : &mga->mga_secondary);
mutex_enter(&mg->mg_lock);
if (*mspp != NULL) {
mutex_exit(&mg->mg_lock);
return (EEXIST);
}
*mspp = msp;
ASSERT3S(msp->ms_allocator, ==, -1);
msp->ms_allocator = allocator;
msp->ms_primary = (activation_weight == METASLAB_WEIGHT_PRIMARY);
ASSERT0(msp->ms_activation_weight);
msp->ms_activation_weight = msp->ms_weight;
metaslab_group_sort_impl(mg, msp,
msp->ms_weight | activation_weight);
mutex_exit(&mg->mg_lock);
return (0);
}
static int
metaslab_activate(metaslab_t *msp, int allocator, uint64_t activation_weight)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
/*
* The current metaslab is already activated for us so there
* is nothing to do. Already activated though, doesn't mean
* that this metaslab is activated for our allocator nor our
* requested activation weight. The metaslab could have started
* as an active one for our allocator but changed allocators
* while we were waiting to grab its ms_lock or we stole it
* [see find_valid_metaslab()]. This means that there is a
* possibility of passivating a metaslab of another allocator
* or from a different activation mask, from this thread.
*/
if ((msp->ms_weight & METASLAB_ACTIVE_MASK) != 0) {
ASSERT(msp->ms_loaded);
return (0);
}
int error = metaslab_load(msp);
if (error != 0) {
metaslab_group_sort(msp->ms_group, msp, 0);
return (error);
}
/*
* When entering metaslab_load() we may have dropped the
* ms_lock because we were loading this metaslab, or we
* were waiting for another thread to load it for us. In
* that scenario, we recheck the weight of the metaslab
* to see if it was activated by another thread.
*
* If the metaslab was activated for another allocator or
* it was activated with a different activation weight (e.g.
* we wanted to make it a primary but it was activated as
* secondary) we return error (EBUSY).
*
* If the metaslab was activated for the same allocator
* and requested activation mask, skip activating it.
*/
if ((msp->ms_weight & METASLAB_ACTIVE_MASK) != 0) {
if (msp->ms_allocator != allocator)
return (EBUSY);
if ((msp->ms_weight & activation_weight) == 0)
return (SET_ERROR(EBUSY));
EQUIV((activation_weight == METASLAB_WEIGHT_PRIMARY),
msp->ms_primary);
return (0);
}
/*
* If the metaslab has literally 0 space, it will have weight 0. In
* that case, don't bother activating it. This can happen if the
* metaslab had space during find_valid_metaslab, but another thread
* loaded it and used all that space while we were waiting to grab the
* lock.
*/
if (msp->ms_weight == 0) {
ASSERT0(range_tree_space(msp->ms_allocatable));
return (SET_ERROR(ENOSPC));
}
if ((error = metaslab_activate_allocator(msp->ms_group, msp,
allocator, activation_weight)) != 0) {
return (error);
}
ASSERT(msp->ms_loaded);
ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
return (0);
}
static void
metaslab_passivate_allocator(metaslab_group_t *mg, metaslab_t *msp,
uint64_t weight)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
ASSERT(msp->ms_loaded);
if (msp->ms_weight & METASLAB_WEIGHT_CLAIM) {
metaslab_group_sort(mg, msp, weight);
return;
}
mutex_enter(&mg->mg_lock);
ASSERT3P(msp->ms_group, ==, mg);
ASSERT3S(0, <=, msp->ms_allocator);
ASSERT3U(msp->ms_allocator, <, mg->mg_allocators);
metaslab_group_allocator_t *mga = &mg->mg_allocator[msp->ms_allocator];
if (msp->ms_primary) {
ASSERT3P(mga->mga_primary, ==, msp);
ASSERT(msp->ms_weight & METASLAB_WEIGHT_PRIMARY);
mga->mga_primary = NULL;
} else {
ASSERT3P(mga->mga_secondary, ==, msp);
ASSERT(msp->ms_weight & METASLAB_WEIGHT_SECONDARY);
mga->mga_secondary = NULL;
}
msp->ms_allocator = -1;
metaslab_group_sort_impl(mg, msp, weight);
mutex_exit(&mg->mg_lock);
}
static void
metaslab_passivate(metaslab_t *msp, uint64_t weight)
{
uint64_t size __maybe_unused = weight & ~METASLAB_WEIGHT_TYPE;
/*
* If size < SPA_MINBLOCKSIZE, then we will not allocate from
* this metaslab again. In that case, it had better be empty,
* or we would be leaving space on the table.
*/
ASSERT(!WEIGHT_IS_SPACEBASED(msp->ms_weight) ||
size >= SPA_MINBLOCKSIZE ||
range_tree_space(msp->ms_allocatable) == 0);
ASSERT0(weight & METASLAB_ACTIVE_MASK);
ASSERT(msp->ms_activation_weight != 0);
msp->ms_activation_weight = 0;
metaslab_passivate_allocator(msp->ms_group, msp, weight);
ASSERT0(msp->ms_weight & METASLAB_ACTIVE_MASK);
}
/*
* Segment-based metaslabs are activated once and remain active until
* we either fail an allocation attempt (similar to space-based metaslabs)
* or have exhausted the free space in zfs_metaslab_switch_threshold
* buckets since the metaslab was activated. This function checks to see
* if we've exhausted the zfs_metaslab_switch_threshold buckets in the
* metaslab and passivates it proactively. This will allow us to select a
* metaslab with a larger contiguous region, if any, remaining within this
* metaslab group. If we're in sync pass > 1, then we continue using this
* metaslab so that we don't dirty more block and cause more sync passes.
*/
static void
metaslab_segment_may_passivate(metaslab_t *msp)
{
spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
if (WEIGHT_IS_SPACEBASED(msp->ms_weight) || spa_sync_pass(spa) > 1)
return;
/*
* Since we are in the middle of a sync pass, the most accurate
* information that is accessible to us is the in-core range tree
* histogram; calculate the new weight based on that information.
*/
uint64_t weight = metaslab_weight_from_range_tree(msp);
int activation_idx = WEIGHT_GET_INDEX(msp->ms_activation_weight);
int current_idx = WEIGHT_GET_INDEX(weight);
if (current_idx <= activation_idx - zfs_metaslab_switch_threshold)
metaslab_passivate(msp, weight);
}
static void
metaslab_preload(void *arg)
{
metaslab_t *msp = arg;
metaslab_class_t *mc = msp->ms_group->mg_class;
spa_t *spa = mc->mc_spa;
fstrans_cookie_t cookie = spl_fstrans_mark();
ASSERT(!MUTEX_HELD(&msp->ms_group->mg_lock));
mutex_enter(&msp->ms_lock);
(void) metaslab_load(msp);
metaslab_set_selected_txg(msp, spa_syncing_txg(spa));
mutex_exit(&msp->ms_lock);
spl_fstrans_unmark(cookie);
}
static void
metaslab_group_preload(metaslab_group_t *mg)
{
spa_t *spa = mg->mg_vd->vdev_spa;
metaslab_t *msp;
avl_tree_t *t = &mg->mg_metaslab_tree;
int m = 0;
if (spa_shutting_down(spa) || !metaslab_preload_enabled) {
taskq_wait_outstanding(mg->mg_taskq, 0);
return;
}
mutex_enter(&mg->mg_lock);
/*
* Load the next potential metaslabs
*/
for (msp = avl_first(t); msp != NULL; msp = AVL_NEXT(t, msp)) {
ASSERT3P(msp->ms_group, ==, mg);
/*
* We preload only the maximum number of metaslabs specified
* by metaslab_preload_limit. If a metaslab is being forced
* to condense then we preload it too. This will ensure
* that force condensing happens in the next txg.
*/
if (++m > metaslab_preload_limit && !msp->ms_condense_wanted) {
continue;
}
VERIFY(taskq_dispatch(mg->mg_taskq, metaslab_preload,
msp, TQ_SLEEP) != TASKQID_INVALID);
}
mutex_exit(&mg->mg_lock);
}
/*
* Determine if the space map's on-disk footprint is past our tolerance for
* inefficiency. We would like to use the following criteria to make our
* decision:
*
* 1. Do not condense if the size of the space map object would dramatically
* increase as a result of writing out the free space range tree.
*
* 2. Condense if the on on-disk space map representation is at least
* zfs_condense_pct/100 times the size of the optimal representation
* (i.e. zfs_condense_pct = 110 and in-core = 1MB, optimal = 1.1MB).
*
* 3. Do not condense if the on-disk size of the space map does not actually
* decrease.
*
* Unfortunately, we cannot compute the on-disk size of the space map in this
* context because we cannot accurately compute the effects of compression, etc.
* Instead, we apply the heuristic described in the block comment for
* zfs_metaslab_condense_block_threshold - we only condense if the space used
* is greater than a threshold number of blocks.
*/
static boolean_t
metaslab_should_condense(metaslab_t *msp)
{
space_map_t *sm = msp->ms_sm;
vdev_t *vd = msp->ms_group->mg_vd;
uint64_t vdev_blocksize = 1 << vd->vdev_ashift;
ASSERT(MUTEX_HELD(&msp->ms_lock));
ASSERT(msp->ms_loaded);
ASSERT(sm != NULL);
ASSERT3U(spa_sync_pass(vd->vdev_spa), ==, 1);
/*
* We always condense metaslabs that are empty and metaslabs for
* which a condense request has been made.
*/
if (range_tree_numsegs(msp->ms_allocatable) == 0 ||
msp->ms_condense_wanted)
return (B_TRUE);
uint64_t record_size = MAX(sm->sm_blksz, vdev_blocksize);
uint64_t object_size = space_map_length(sm);
uint64_t optimal_size = space_map_estimate_optimal_size(sm,
msp->ms_allocatable, SM_NO_VDEVID);
return (object_size >= (optimal_size * zfs_condense_pct / 100) &&
object_size > zfs_metaslab_condense_block_threshold * record_size);
}
/*
* Condense the on-disk space map representation to its minimized form.
* The minimized form consists of a small number of allocations followed
* by the entries of the free range tree (ms_allocatable). The condensed
* spacemap contains all the entries of previous TXGs (including those in
* the pool-wide log spacemaps; thus this is effectively a superset of
* metaslab_flush()), but this TXG's entries still need to be written.
*/
static void
metaslab_condense(metaslab_t *msp, dmu_tx_t *tx)
{
range_tree_t *condense_tree;
space_map_t *sm = msp->ms_sm;
uint64_t txg = dmu_tx_get_txg(tx);
spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
ASSERT(MUTEX_HELD(&msp->ms_lock));
ASSERT(msp->ms_loaded);
ASSERT(msp->ms_sm != NULL);
/*
* In order to condense the space map, we need to change it so it
* only describes which segments are currently allocated and free.
*
* All the current free space resides in the ms_allocatable, all
* the ms_defer trees, and all the ms_allocating trees. We ignore
* ms_freed because it is empty because we're in sync pass 1. We
* ignore ms_freeing because these changes are not yet reflected
* in the spacemap (they will be written later this txg).
*
* So to truncate the space map to represent all the entries of
* previous TXGs we do the following:
*
* 1] We create a range tree (condense tree) that is 100% empty.
* 2] We add to it all segments found in the ms_defer trees
* as those segments are marked as free in the original space
* map. We do the same with the ms_allocating trees for the same
* reason. Adding these segments should be a relatively
* inexpensive operation since we expect these trees to have a
* small number of nodes.
* 3] We vacate any unflushed allocs, since they are not frees we
* need to add to the condense tree. Then we vacate any
* unflushed frees as they should already be part of ms_allocatable.
* 4] At this point, we would ideally like to add all segments
* in the ms_allocatable tree from the condense tree. This way
* we would write all the entries of the condense tree as the
* condensed space map, which would only contain freed
* segments with everything else assumed to be allocated.
*
* Doing so can be prohibitively expensive as ms_allocatable can
* be large, and therefore computationally expensive to add to
* the condense_tree. Instead we first sync out an entry marking
* everything as allocated, then the condense_tree and then the
* ms_allocatable, in the condensed space map. While this is not
* optimal, it is typically close to optimal and more importantly
* much cheaper to compute.
*
* 5] Finally, as both of the unflushed trees were written to our
* new and condensed metaslab space map, we basically flushed
* all the unflushed changes to disk, thus we call
* metaslab_flush_update().
*/
ASSERT3U(spa_sync_pass(spa), ==, 1);
ASSERT(range_tree_is_empty(msp->ms_freed)); /* since it is pass 1 */
zfs_dbgmsg("condensing: txg %llu, msp[%llu] %px, vdev id %llu, "
"spa %s, smp size %llu, segments %llu, forcing condense=%s",
(u_longlong_t)txg, (u_longlong_t)msp->ms_id, msp,
(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
spa->spa_name, (u_longlong_t)space_map_length(msp->ms_sm),
(u_longlong_t)range_tree_numsegs(msp->ms_allocatable),
msp->ms_condense_wanted ? "TRUE" : "FALSE");
msp->ms_condense_wanted = B_FALSE;
range_seg_type_t type;
uint64_t shift, start;
type = metaslab_calculate_range_tree_type(msp->ms_group->mg_vd, msp,
&start, &shift);
condense_tree = range_tree_create(NULL, type, NULL, start, shift);
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
range_tree_walk(msp->ms_defer[t],
range_tree_add, condense_tree);
}
for (int t = 0; t < TXG_CONCURRENT_STATES; t++) {
range_tree_walk(msp->ms_allocating[(txg + t) & TXG_MASK],
range_tree_add, condense_tree);
}
ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
metaslab_unflushed_changes_memused(msp));
spa->spa_unflushed_stats.sus_memused -=
metaslab_unflushed_changes_memused(msp);
range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);
/*
* We're about to drop the metaslab's lock thus allowing other
* consumers to change it's content. Set the metaslab's ms_condensing
* flag to ensure that allocations on this metaslab do not occur
* while we're in the middle of committing it to disk. This is only
* critical for ms_allocatable as all other range trees use per TXG
* views of their content.
*/
msp->ms_condensing = B_TRUE;
mutex_exit(&msp->ms_lock);
uint64_t object = space_map_object(msp->ms_sm);
space_map_truncate(sm,
spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP) ?
zfs_metaslab_sm_blksz_with_log : zfs_metaslab_sm_blksz_no_log, tx);
/*
* space_map_truncate() may have reallocated the spacemap object.
* If so, update the vdev_ms_array.
*/
if (space_map_object(msp->ms_sm) != object) {
object = space_map_object(msp->ms_sm);
dmu_write(spa->spa_meta_objset,
msp->ms_group->mg_vd->vdev_ms_array, sizeof (uint64_t) *
msp->ms_id, sizeof (uint64_t), &object, tx);
}
/*
* Note:
* When the log space map feature is enabled, each space map will
* always have ALLOCS followed by FREES for each sync pass. This is
* typically true even when the log space map feature is disabled,
* except from the case where a metaslab goes through metaslab_sync()
* and gets condensed. In that case the metaslab's space map will have
* ALLOCS followed by FREES (due to condensing) followed by ALLOCS
* followed by FREES (due to space_map_write() in metaslab_sync()) for
* sync pass 1.
*/
range_tree_t *tmp_tree = range_tree_create(NULL, type, NULL, start,
shift);
range_tree_add(tmp_tree, msp->ms_start, msp->ms_size);
space_map_write(sm, tmp_tree, SM_ALLOC, SM_NO_VDEVID, tx);
space_map_write(sm, msp->ms_allocatable, SM_FREE, SM_NO_VDEVID, tx);
space_map_write(sm, condense_tree, SM_FREE, SM_NO_VDEVID, tx);
range_tree_vacate(condense_tree, NULL, NULL);
range_tree_destroy(condense_tree);
range_tree_vacate(tmp_tree, NULL, NULL);
range_tree_destroy(tmp_tree);
mutex_enter(&msp->ms_lock);
msp->ms_condensing = B_FALSE;
metaslab_flush_update(msp, tx);
}
/*
* Called when the metaslab has been flushed (its own spacemap now reflects
* all the contents of the pool-wide spacemap log). Updates the metaslab's
* metadata and any pool-wide related log space map data (e.g. summary,
* obsolete logs, etc..) to reflect that.
*/
static void
metaslab_flush_update(metaslab_t *msp, dmu_tx_t *tx)
{
metaslab_group_t *mg = msp->ms_group;
spa_t *spa = mg->mg_vd->vdev_spa;
ASSERT(MUTEX_HELD(&msp->ms_lock));
ASSERT3U(spa_sync_pass(spa), ==, 1);
ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));
/*
* Just because a metaslab got flushed, that doesn't mean that
* it will pass through metaslab_sync_done(). Thus, make sure to
* update ms_synced_length here in case it doesn't.
*/
msp->ms_synced_length = space_map_length(msp->ms_sm);
/*
* We may end up here from metaslab_condense() without the
* feature being active. In that case this is a no-op.
*/
if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))
return;
ASSERT(spa_syncing_log_sm(spa) != NULL);
ASSERT(msp->ms_sm != NULL);
ASSERT(metaslab_unflushed_txg(msp) != 0);
ASSERT3P(avl_find(&spa->spa_metaslabs_by_flushed, msp, NULL), ==, msp);
VERIFY3U(tx->tx_txg, <=, spa_final_dirty_txg(spa));
/* update metaslab's position in our flushing tree */
uint64_t ms_prev_flushed_txg = metaslab_unflushed_txg(msp);
mutex_enter(&spa->spa_flushed_ms_lock);
avl_remove(&spa->spa_metaslabs_by_flushed, msp);
metaslab_set_unflushed_txg(msp, spa_syncing_txg(spa), tx);
avl_add(&spa->spa_metaslabs_by_flushed, msp);
mutex_exit(&spa->spa_flushed_ms_lock);
/* update metaslab counts of spa_log_sm_t nodes */
spa_log_sm_decrement_mscount(spa, ms_prev_flushed_txg);
spa_log_sm_increment_current_mscount(spa);
/* cleanup obsolete logs if any */
uint64_t log_blocks_before = spa_log_sm_nblocks(spa);
spa_cleanup_old_sm_logs(spa, tx);
uint64_t log_blocks_after = spa_log_sm_nblocks(spa);
VERIFY3U(log_blocks_after, <=, log_blocks_before);
/* update log space map summary */
uint64_t blocks_gone = log_blocks_before - log_blocks_after;
spa_log_summary_add_flushed_metaslab(spa);
spa_log_summary_decrement_mscount(spa, ms_prev_flushed_txg);
spa_log_summary_decrement_blkcount(spa, blocks_gone);
}
boolean_t
metaslab_flush(metaslab_t *msp, dmu_tx_t *tx)
{
spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
ASSERT(MUTEX_HELD(&msp->ms_lock));
ASSERT3U(spa_sync_pass(spa), ==, 1);
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));
ASSERT(msp->ms_sm != NULL);
ASSERT(metaslab_unflushed_txg(msp) != 0);
ASSERT(avl_find(&spa->spa_metaslabs_by_flushed, msp, NULL) != NULL);
/*
* There is nothing wrong with flushing the same metaslab twice, as
* this codepath should work on that case. However, the current
* flushing scheme makes sure to avoid this situation as we would be
* making all these calls without having anything meaningful to write
* to disk. We assert this behavior here.
*/
ASSERT3U(metaslab_unflushed_txg(msp), <, dmu_tx_get_txg(tx));
/*
* We can not flush while loading, because then we would
* not load the ms_unflushed_{allocs,frees}.
*/
if (msp->ms_loading)
return (B_FALSE);
metaslab_verify_space(msp, dmu_tx_get_txg(tx));
metaslab_verify_weight_and_frag(msp);
/*
* Metaslab condensing is effectively flushing. Therefore if the
* metaslab can be condensed we can just condense it instead of
* flushing it.
*
* Note that metaslab_condense() does call metaslab_flush_update()
* so we can just return immediately after condensing. We also
* don't need to care about setting ms_flushing or broadcasting
* ms_flush_cv, even if we temporarily drop the ms_lock in
* metaslab_condense(), as the metaslab is already loaded.
*/
if (msp->ms_loaded && metaslab_should_condense(msp)) {
metaslab_group_t *mg = msp->ms_group;
/*
* For all histogram operations below refer to the
* comments of metaslab_sync() where we follow a
* similar procedure.
*/
metaslab_group_histogram_verify(mg);
metaslab_class_histogram_verify(mg->mg_class);
metaslab_group_histogram_remove(mg, msp);
metaslab_condense(msp, tx);
space_map_histogram_clear(msp->ms_sm);
space_map_histogram_add(msp->ms_sm, msp->ms_allocatable, tx);
ASSERT(range_tree_is_empty(msp->ms_freed));
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
space_map_histogram_add(msp->ms_sm,
msp->ms_defer[t], tx);
}
metaslab_aux_histograms_update(msp);
metaslab_group_histogram_add(mg, msp);
metaslab_group_histogram_verify(mg);
metaslab_class_histogram_verify(mg->mg_class);
metaslab_verify_space(msp, dmu_tx_get_txg(tx));
/*
* Since we recreated the histogram (and potentially
* the ms_sm too while condensing) ensure that the
* weight is updated too because we are not guaranteed
* that this metaslab is dirty and will go through
* metaslab_sync_done().
*/
metaslab_recalculate_weight_and_sort(msp);
return (B_TRUE);
}
msp->ms_flushing = B_TRUE;
uint64_t sm_len_before = space_map_length(msp->ms_sm);
mutex_exit(&msp->ms_lock);
space_map_write(msp->ms_sm, msp->ms_unflushed_allocs, SM_ALLOC,
SM_NO_VDEVID, tx);
space_map_write(msp->ms_sm, msp->ms_unflushed_frees, SM_FREE,
SM_NO_VDEVID, tx);
mutex_enter(&msp->ms_lock);
uint64_t sm_len_after = space_map_length(msp->ms_sm);
if (zfs_flags & ZFS_DEBUG_LOG_SPACEMAP) {
zfs_dbgmsg("flushing: txg %llu, spa %s, vdev_id %llu, "
"ms_id %llu, unflushed_allocs %llu, unflushed_frees %llu, "
"appended %llu bytes", (u_longlong_t)dmu_tx_get_txg(tx),
spa_name(spa),
(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
(u_longlong_t)msp->ms_id,
(u_longlong_t)range_tree_space(msp->ms_unflushed_allocs),
(u_longlong_t)range_tree_space(msp->ms_unflushed_frees),
(u_longlong_t)(sm_len_after - sm_len_before));
}
ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
metaslab_unflushed_changes_memused(msp));
spa->spa_unflushed_stats.sus_memused -=
metaslab_unflushed_changes_memused(msp);
range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);
metaslab_verify_space(msp, dmu_tx_get_txg(tx));
metaslab_verify_weight_and_frag(msp);
metaslab_flush_update(msp, tx);
metaslab_verify_space(msp, dmu_tx_get_txg(tx));
metaslab_verify_weight_and_frag(msp);
msp->ms_flushing = B_FALSE;
cv_broadcast(&msp->ms_flush_cv);
return (B_TRUE);
}
/*
* Write a metaslab to disk in the context of the specified transaction group.
*/
void
metaslab_sync(metaslab_t *msp, uint64_t txg)
{
metaslab_group_t *mg = msp->ms_group;
vdev_t *vd = mg->mg_vd;
spa_t *spa = vd->vdev_spa;
objset_t *mos = spa_meta_objset(spa);
range_tree_t *alloctree = msp->ms_allocating[txg & TXG_MASK];
dmu_tx_t *tx;
ASSERT(!vd->vdev_ishole);
/*
* This metaslab has just been added so there's no work to do now.
*/
if (msp->ms_new) {
ASSERT0(range_tree_space(alloctree));
ASSERT0(range_tree_space(msp->ms_freeing));
ASSERT0(range_tree_space(msp->ms_freed));
ASSERT0(range_tree_space(msp->ms_checkpointing));
ASSERT0(range_tree_space(msp->ms_trim));
return;
}
/*
* Normally, we don't want to process a metaslab if there are no
* allocations or frees to perform. However, if the metaslab is being
* forced to condense, it's loaded and we're not beyond the final
* dirty txg, we need to let it through. Not condensing beyond the
* final dirty txg prevents an issue where metaslabs that need to be
* condensed but were loaded for other reasons could cause a panic
* here. By only checking the txg in that branch of the conditional,
* we preserve the utility of the VERIFY statements in all other
* cases.
*/
if (range_tree_is_empty(alloctree) &&
range_tree_is_empty(msp->ms_freeing) &&
range_tree_is_empty(msp->ms_checkpointing) &&
!(msp->ms_loaded && msp->ms_condense_wanted &&
txg <= spa_final_dirty_txg(spa)))
return;
VERIFY3U(txg, <=, spa_final_dirty_txg(spa));
/*
* The only state that can actually be changing concurrently
* with metaslab_sync() is the metaslab's ms_allocatable. No
* other thread can be modifying this txg's alloc, freeing,
* freed, or space_map_phys_t. We drop ms_lock whenever we
* could call into the DMU, because the DMU can call down to
* us (e.g. via zio_free()) at any time.
*
* The spa_vdev_remove_thread() can be reading metaslab state
* concurrently, and it is locked out by the ms_sync_lock.
* Note that the ms_lock is insufficient for this, because it
* is dropped by space_map_write().
*/
tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);
/*
* Generate a log space map if one doesn't exist already.
*/
spa_generate_syncing_log_sm(spa, tx);
if (msp->ms_sm == NULL) {
uint64_t new_object = space_map_alloc(mos,
spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP) ?
zfs_metaslab_sm_blksz_with_log :
zfs_metaslab_sm_blksz_no_log, tx);
VERIFY3U(new_object, !=, 0);
dmu_write(mos, vd->vdev_ms_array, sizeof (uint64_t) *
msp->ms_id, sizeof (uint64_t), &new_object, tx);
VERIFY0(space_map_open(&msp->ms_sm, mos, new_object,
msp->ms_start, msp->ms_size, vd->vdev_ashift));
ASSERT(msp->ms_sm != NULL);
ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));
ASSERT0(metaslab_allocated_space(msp));
}
if (metaslab_unflushed_txg(msp) == 0 &&
spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) {
ASSERT(spa_syncing_log_sm(spa) != NULL);
metaslab_set_unflushed_txg(msp, spa_syncing_txg(spa), tx);
spa_log_sm_increment_current_mscount(spa);
spa_log_summary_add_flushed_metaslab(spa);
ASSERT(msp->ms_sm != NULL);
mutex_enter(&spa->spa_flushed_ms_lock);
avl_add(&spa->spa_metaslabs_by_flushed, msp);
mutex_exit(&spa->spa_flushed_ms_lock);
ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));
}
if (!range_tree_is_empty(msp->ms_checkpointing) &&
vd->vdev_checkpoint_sm == NULL) {
ASSERT(spa_has_checkpoint(spa));
uint64_t new_object = space_map_alloc(mos,
zfs_vdev_standard_sm_blksz, tx);
VERIFY3U(new_object, !=, 0);
VERIFY0(space_map_open(&vd->vdev_checkpoint_sm,
mos, new_object, 0, vd->vdev_asize, vd->vdev_ashift));
ASSERT3P(vd->vdev_checkpoint_sm, !=, NULL);
/*
* We save the space map object as an entry in vdev_top_zap
* so it can be retrieved when the pool is reopened after an
* export or through zdb.
*/
VERIFY0(zap_add(vd->vdev_spa->spa_meta_objset,
vd->vdev_top_zap, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM,
sizeof (new_object), 1, &new_object, tx));
}
mutex_enter(&msp->ms_sync_lock);
mutex_enter(&msp->ms_lock);
/*
* Note: metaslab_condense() clears the space map's histogram.
* Therefore we must verify and remove this histogram before
* condensing.
*/
metaslab_group_histogram_verify(mg);
metaslab_class_histogram_verify(mg->mg_class);
metaslab_group_histogram_remove(mg, msp);
if (spa->spa_sync_pass == 1 && msp->ms_loaded &&
metaslab_should_condense(msp))
metaslab_condense(msp, tx);
/*
* We'll be going to disk to sync our space accounting, thus we
* drop the ms_lock during that time so allocations coming from
* open-context (ZIL) for future TXGs do not block.
*/
mutex_exit(&msp->ms_lock);
space_map_t *log_sm = spa_syncing_log_sm(spa);
if (log_sm != NULL) {
ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP));
space_map_write(log_sm, alloctree, SM_ALLOC,
vd->vdev_id, tx);
space_map_write(log_sm, msp->ms_freeing, SM_FREE,
vd->vdev_id, tx);
mutex_enter(&msp->ms_lock);
ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
metaslab_unflushed_changes_memused(msp));
spa->spa_unflushed_stats.sus_memused -=
metaslab_unflushed_changes_memused(msp);
range_tree_remove_xor_add(alloctree,
msp->ms_unflushed_frees, msp->ms_unflushed_allocs);
range_tree_remove_xor_add(msp->ms_freeing,
msp->ms_unflushed_allocs, msp->ms_unflushed_frees);
spa->spa_unflushed_stats.sus_memused +=
metaslab_unflushed_changes_memused(msp);
} else {
ASSERT(!spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP));
space_map_write(msp->ms_sm, alloctree, SM_ALLOC,
SM_NO_VDEVID, tx);
space_map_write(msp->ms_sm, msp->ms_freeing, SM_FREE,
SM_NO_VDEVID, tx);
mutex_enter(&msp->ms_lock);
}
msp->ms_allocated_space += range_tree_space(alloctree);
ASSERT3U(msp->ms_allocated_space, >=,
range_tree_space(msp->ms_freeing));
msp->ms_allocated_space -= range_tree_space(msp->ms_freeing);
if (!range_tree_is_empty(msp->ms_checkpointing)) {
ASSERT(spa_has_checkpoint(spa));
ASSERT3P(vd->vdev_checkpoint_sm, !=, NULL);
/*
* Since we are doing writes to disk and the ms_checkpointing
* tree won't be changing during that time, we drop the
* ms_lock while writing to the checkpoint space map, for the
* same reason mentioned above.
*/
mutex_exit(&msp->ms_lock);
space_map_write(vd->vdev_checkpoint_sm,
msp->ms_checkpointing, SM_FREE, SM_NO_VDEVID, tx);
mutex_enter(&msp->ms_lock);
spa->spa_checkpoint_info.sci_dspace +=
range_tree_space(msp->ms_checkpointing);
vd->vdev_stat.vs_checkpoint_space +=
range_tree_space(msp->ms_checkpointing);
ASSERT3U(vd->vdev_stat.vs_checkpoint_space, ==,
-space_map_allocated(vd->vdev_checkpoint_sm));
range_tree_vacate(msp->ms_checkpointing, NULL, NULL);
}
if (msp->ms_loaded) {
/*
* When the space map is loaded, we have an accurate
* histogram in the range tree. This gives us an opportunity
* to bring the space map's histogram up-to-date so we clear
* it first before updating it.
*/
space_map_histogram_clear(msp->ms_sm);
space_map_histogram_add(msp->ms_sm, msp->ms_allocatable, tx);
/*
* Since we've cleared the histogram we need to add back
* any free space that has already been processed, plus
* any deferred space. This allows the on-disk histogram
* to accurately reflect all free space even if some space
* is not yet available for allocation (i.e. deferred).
*/
space_map_histogram_add(msp->ms_sm, msp->ms_freed, tx);
/*
* Add back any deferred free space that has not been
* added back into the in-core free tree yet. This will
* ensure that we don't end up with a space map histogram
* that is completely empty unless the metaslab is fully
* allocated.
*/
for (int t = 0; t < TXG_DEFER_SIZE; t++) {
space_map_histogram_add(msp->ms_sm,
msp->ms_defer[t], tx);
}
}
/*
* Always add the free space from this sync pass to the space
* map histogram. We want to make sure that the on-disk histogram
* accounts for all free space. If the space map is not loaded,
* then we will lose some accuracy but will correct it the next
* time we load the space map.
*/
space_map_histogram_add(msp->ms_sm, msp->ms_freeing, tx);
metaslab_aux_histograms_update(msp);
metaslab_group_histogram_add(mg, msp);
metaslab_group_histogram_verify(mg);
metaslab_class_histogram_verify(mg->mg_class);
/*
* For sync pass 1, we avoid traversing this txg's free range tree
* and instead will just swap the pointers for freeing and freed.
* We can safely do this since the freed_tree is guaranteed to be
* empty on the initial pass.
*
* Keep in mind that even if we are currently using a log spacemap
* we want current frees to end up in the ms_allocatable (but not
* get appended to the ms_sm) so their ranges can be reused as usual.
*/
if (spa_sync_pass(spa) == 1) {
range_tree_swap(&msp->ms_freeing, &msp->ms_freed);
ASSERT0(msp->ms_allocated_this_txg);
} else {
range_tree_vacate(msp->ms_freeing,
range_tree_add, msp->ms_freed);
}
msp->ms_allocated_this_txg += range_tree_space(alloctree);
range_tree_vacate(alloctree, NULL, NULL);
ASSERT0(range_tree_space(msp->ms_allocating[txg & TXG_MASK]));
ASSERT0(range_tree_space(msp->ms_allocating[TXG_CLEAN(txg)
& TXG_MASK]));
ASSERT0(range_tree_space(msp->ms_freeing));
ASSERT0(range_tree_space(msp->ms_checkpointing));
mutex_exit(&msp->ms_lock);
/*
* Verify that the space map object ID has been recorded in the
* vdev_ms_array.
*/
uint64_t object;
VERIFY0(dmu_read(mos, vd->vdev_ms_array,
msp->ms_id * sizeof (uint64_t), sizeof (uint64_t), &object, 0));
VERIFY3U(object, ==, space_map_object(msp->ms_sm));
mutex_exit(&msp->ms_sync_lock);
dmu_tx_commit(tx);
}
static void
metaslab_evict(metaslab_t *msp, uint64_t txg)
{
if (!msp->ms_loaded || msp->ms_disabled != 0)
return;
for (int t = 1; t < TXG_CONCURRENT_STATES; t++) {
VERIFY0(range_tree_space(
msp->ms_allocating[(txg + t) & TXG_MASK]));
}
if (msp->ms_allocator != -1)
metaslab_passivate(msp, msp->ms_weight & ~METASLAB_ACTIVE_MASK);
if (!metaslab_debug_unload)
metaslab_unload(msp);
}
/*
* Called after a transaction group has completely synced to mark
* all of the metaslab's free space as usable.
*/
void
metaslab_sync_done(metaslab_t *msp, uint64_t txg)
{
metaslab_group_t *mg = msp->ms_group;
vdev_t *vd = mg->mg_vd;
spa_t *spa = vd->vdev_spa;
range_tree_t **defer_tree;
int64_t alloc_delta, defer_delta;
boolean_t defer_allowed = B_TRUE;
ASSERT(!vd->vdev_ishole);
mutex_enter(&msp->ms_lock);
if (msp->ms_new) {
/* this is a new metaslab, add its capacity to the vdev */
metaslab_space_update(vd, mg->mg_class, 0, 0, msp->ms_size);
/* there should be no allocations nor frees at this point */
VERIFY0(msp->ms_allocated_this_txg);
VERIFY0(range_tree_space(msp->ms_freed));
}
ASSERT0(range_tree_space(msp->ms_freeing));
ASSERT0(range_tree_space(msp->ms_checkpointing));
defer_tree = &msp->ms_defer[txg % TXG_DEFER_SIZE];
uint64_t free_space = metaslab_class_get_space(spa_normal_class(spa)) -
metaslab_class_get_alloc(spa_normal_class(spa));
if (free_space <= spa_get_slop_space(spa) || vd->vdev_removing) {
defer_allowed = B_FALSE;
}
defer_delta = 0;
alloc_delta = msp->ms_allocated_this_txg -
range_tree_space(msp->ms_freed);
if (defer_allowed) {
defer_delta = range_tree_space(msp->ms_freed) -
range_tree_space(*defer_tree);
} else {
defer_delta -= range_tree_space(*defer_tree);
}
metaslab_space_update(vd, mg->mg_class, alloc_delta + defer_delta,
defer_delta, 0);
if (spa_syncing_log_sm(spa) == NULL) {
/*
* If there's a metaslab_load() in progress and we don't have
* a log space map, it means that we probably wrote to the
* metaslab's space map. If this is the case, we need to
* make sure that we wait for the load to complete so that we
* have a consistent view at the in-core side of the metaslab.
*/
metaslab_load_wait(msp);
} else {
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));
}
/*
* When auto-trimming is enabled, free ranges which are added to
* ms_allocatable are also be added to ms_trim. The ms_trim tree is
* periodically consumed by the vdev_autotrim_thread() which issues
* trims for all ranges and then vacates the tree. The ms_trim tree
* can be discarded at any time with the sole consequence of recent
* frees not being trimmed.
*/
if (spa_get_autotrim(spa) == SPA_AUTOTRIM_ON) {
range_tree_walk(*defer_tree, range_tree_add, msp->ms_trim);
if (!defer_allowed) {
range_tree_walk(msp->ms_freed, range_tree_add,
msp->ms_trim);
}
} else {
range_tree_vacate(msp->ms_trim, NULL, NULL);
}
/*
* Move the frees from the defer_tree back to the free
* range tree (if it's loaded). Swap the freed_tree and
* the defer_tree -- this is safe to do because we've
* just emptied out the defer_tree.
*/
range_tree_vacate(*defer_tree,
msp->ms_loaded ? range_tree_add : NULL, msp->ms_allocatable);
if (defer_allowed) {
range_tree_swap(&msp->ms_freed, defer_tree);
} else {
range_tree_vacate(msp->ms_freed,
msp->ms_loaded ? range_tree_add : NULL,
msp->ms_allocatable);
}
msp->ms_synced_length = space_map_length(msp->ms_sm);
msp->ms_deferspace += defer_delta;
ASSERT3S(msp->ms_deferspace, >=, 0);
ASSERT3S(msp->ms_deferspace, <=, msp->ms_size);
if (msp->ms_deferspace != 0) {
/*
* Keep syncing this metaslab until all deferred frees
* are back in circulation.
*/
vdev_dirty(vd, VDD_METASLAB, msp, txg + 1);
}
metaslab_aux_histograms_update_done(msp, defer_allowed);
if (msp->ms_new) {
msp->ms_new = B_FALSE;
mutex_enter(&mg->mg_lock);
mg->mg_ms_ready++;
mutex_exit(&mg->mg_lock);
}
/*
* Re-sort metaslab within its group now that we've adjusted
* its allocatable space.
*/
metaslab_recalculate_weight_and_sort(msp);
ASSERT0(range_tree_space(msp->ms_allocating[txg & TXG_MASK]));
ASSERT0(range_tree_space(msp->ms_freeing));
ASSERT0(range_tree_space(msp->ms_freed));
ASSERT0(range_tree_space(msp->ms_checkpointing));
msp->ms_allocating_total -= msp->ms_allocated_this_txg;
msp->ms_allocated_this_txg = 0;
mutex_exit(&msp->ms_lock);
}
void
metaslab_sync_reassess(metaslab_group_t *mg)
{
spa_t *spa = mg->mg_class->mc_spa;
spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);
metaslab_group_alloc_update(mg);
mg->mg_fragmentation = metaslab_group_fragmentation(mg);
/*
* Preload the next potential metaslabs but only on active
* metaslab groups. We can get into a state where the metaslab
* is no longer active since we dirty metaslabs as we remove a
* a device, thus potentially making the metaslab group eligible
* for preloading.
*/
if (mg->mg_activation_count > 0) {
metaslab_group_preload(mg);
}
spa_config_exit(spa, SCL_ALLOC, FTAG);
}
/*
* When writing a ditto block (i.e. more than one DVA for a given BP) on
* the same vdev as an existing DVA of this BP, then try to allocate it
* on a different metaslab than existing DVAs (i.e. a unique metaslab).
*/
static boolean_t
metaslab_is_unique(metaslab_t *msp, dva_t *dva)
{
uint64_t dva_ms_id;
if (DVA_GET_ASIZE(dva) == 0)
return (B_TRUE);
if (msp->ms_group->mg_vd->vdev_id != DVA_GET_VDEV(dva))
return (B_TRUE);
dva_ms_id = DVA_GET_OFFSET(dva) >> msp->ms_group->mg_vd->vdev_ms_shift;
return (msp->ms_id != dva_ms_id);
}
/*
* ==========================================================================
* Metaslab allocation tracing facility
* ==========================================================================
*/
/*
* Add an allocation trace element to the allocation tracing list.
*/
static void
metaslab_trace_add(zio_alloc_list_t *zal, metaslab_group_t *mg,
metaslab_t *msp, uint64_t psize, uint32_t dva_id, uint64_t offset,
int allocator)
{
metaslab_alloc_trace_t *mat;
if (!metaslab_trace_enabled)
return;
/*
* When the tracing list reaches its maximum we remove
* the second element in the list before adding a new one.
* By removing the second element we preserve the original
* entry as a clue to what allocations steps have already been
* performed.
*/
if (zal->zal_size == metaslab_trace_max_entries) {
metaslab_alloc_trace_t *mat_next;
#ifdef ZFS_DEBUG
panic("too many entries in allocation list");
#endif
METASLABSTAT_BUMP(metaslabstat_trace_over_limit);
zal->zal_size--;
mat_next = list_next(&zal->zal_list, list_head(&zal->zal_list));
list_remove(&zal->zal_list, mat_next);
kmem_cache_free(metaslab_alloc_trace_cache, mat_next);
}
mat = kmem_cache_alloc(metaslab_alloc_trace_cache, KM_SLEEP);
list_link_init(&mat->mat_list_node);
mat->mat_mg = mg;
mat->mat_msp = msp;
mat->mat_size = psize;
mat->mat_dva_id = dva_id;
mat->mat_offset = offset;
mat->mat_weight = 0;
mat->mat_allocator = allocator;
if (msp != NULL)
mat->mat_weight = msp->ms_weight;
/*
* The list is part of the zio so locking is not required. Only
* a single thread will perform allocations for a given zio.
*/
list_insert_tail(&zal->zal_list, mat);
zal->zal_size++;
ASSERT3U(zal->zal_size, <=, metaslab_trace_max_entries);
}
void
metaslab_trace_init(zio_alloc_list_t *zal)
{
list_create(&zal->zal_list, sizeof (metaslab_alloc_trace_t),
offsetof(metaslab_alloc_trace_t, mat_list_node));
zal->zal_size = 0;
}
void
metaslab_trace_fini(zio_alloc_list_t *zal)
{
metaslab_alloc_trace_t *mat;
while ((mat = list_remove_head(&zal->zal_list)) != NULL)
kmem_cache_free(metaslab_alloc_trace_cache, mat);
list_destroy(&zal->zal_list);
zal->zal_size = 0;
}
/*
* ==========================================================================
* Metaslab block operations
* ==========================================================================
*/
static void
metaslab_group_alloc_increment(spa_t *spa, uint64_t vdev, void *tag, int flags,
int allocator)
{
if (!(flags & METASLAB_ASYNC_ALLOC) ||
(flags & METASLAB_DONT_THROTTLE))
return;
metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
if (!mg->mg_class->mc_alloc_throttle_enabled)
return;
metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
(void) zfs_refcount_add(&mga->mga_alloc_queue_depth, tag);
}
static void
metaslab_group_increment_qdepth(metaslab_group_t *mg, int allocator)
{
metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
metaslab_class_allocator_t *mca =
&mg->mg_class->mc_allocator[allocator];
uint64_t max = mg->mg_max_alloc_queue_depth;
uint64_t cur = mga->mga_cur_max_alloc_queue_depth;
while (cur < max) {
if (atomic_cas_64(&mga->mga_cur_max_alloc_queue_depth,
cur, cur + 1) == cur) {
atomic_inc_64(&mca->mca_alloc_max_slots);
return;
}
cur = mga->mga_cur_max_alloc_queue_depth;
}
}
void
metaslab_group_alloc_decrement(spa_t *spa, uint64_t vdev, void *tag, int flags,
int allocator, boolean_t io_complete)
{
if (!(flags & METASLAB_ASYNC_ALLOC) ||
(flags & METASLAB_DONT_THROTTLE))
return;
metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
if (!mg->mg_class->mc_alloc_throttle_enabled)
return;
metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
(void) zfs_refcount_remove(&mga->mga_alloc_queue_depth, tag);
if (io_complete)
metaslab_group_increment_qdepth(mg, allocator);
}
void
metaslab_group_alloc_verify(spa_t *spa, const blkptr_t *bp, void *tag,
int allocator)
{
#ifdef ZFS_DEBUG
const dva_t *dva = bp->blk_dva;
int ndvas = BP_GET_NDVAS(bp);
for (int d = 0; d < ndvas; d++) {
uint64_t vdev = DVA_GET_VDEV(&dva[d]);
metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
VERIFY(zfs_refcount_not_held(&mga->mga_alloc_queue_depth, tag));
}
#endif
}
static uint64_t
metaslab_block_alloc(metaslab_t *msp, uint64_t size, uint64_t txg)
{
uint64_t start;
range_tree_t *rt = msp->ms_allocatable;
metaslab_class_t *mc = msp->ms_group->mg_class;
ASSERT(MUTEX_HELD(&msp->ms_lock));
VERIFY(!msp->ms_condensing);
VERIFY0(msp->ms_disabled);
start = mc->mc_ops->msop_alloc(msp, size);
if (start != -1ULL) {
metaslab_group_t *mg = msp->ms_group;
vdev_t *vd = mg->mg_vd;
VERIFY0(P2PHASE(start, 1ULL << vd->vdev_ashift));
VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
VERIFY3U(range_tree_space(rt) - size, <=, msp->ms_size);
range_tree_remove(rt, start, size);
range_tree_clear(msp->ms_trim, start, size);
if (range_tree_is_empty(msp->ms_allocating[txg & TXG_MASK]))
vdev_dirty(mg->mg_vd, VDD_METASLAB, msp, txg);
range_tree_add(msp->ms_allocating[txg & TXG_MASK], start, size);
msp->ms_allocating_total += size;
/* Track the last successful allocation */
msp->ms_alloc_txg = txg;
metaslab_verify_space(msp, txg);
}
/*
* Now that we've attempted the allocation we need to update the
* metaslab's maximum block size since it may have changed.
*/
msp->ms_max_size = metaslab_largest_allocatable(msp);
return (start);
}
/*
* Find the metaslab with the highest weight that is less than what we've
* already tried. In the common case, this means that we will examine each
* metaslab at most once. Note that concurrent callers could reorder metaslabs
* by activation/passivation once we have dropped the mg_lock. If a metaslab is
* activated by another thread, and we fail to allocate from the metaslab we
* have selected, we may not try the newly-activated metaslab, and instead
* activate another metaslab. This is not optimal, but generally does not cause
* any problems (a possible exception being if every metaslab is completely full
* except for the newly-activated metaslab which we fail to examine).
*/
static metaslab_t *
find_valid_metaslab(metaslab_group_t *mg, uint64_t activation_weight,
dva_t *dva, int d, boolean_t want_unique, uint64_t asize, int allocator,
boolean_t try_hard, zio_alloc_list_t *zal, metaslab_t *search,
boolean_t *was_active)
{
avl_index_t idx;
avl_tree_t *t = &mg->mg_metaslab_tree;
metaslab_t *msp = avl_find(t, search, &idx);
if (msp == NULL)
msp = avl_nearest(t, idx, AVL_AFTER);
int tries = 0;
for (; msp != NULL; msp = AVL_NEXT(t, msp)) {
int i;
if (!try_hard && tries > zfs_metaslab_find_max_tries) {
METASLABSTAT_BUMP(metaslabstat_too_many_tries);
return (NULL);
}
tries++;
if (!metaslab_should_allocate(msp, asize, try_hard)) {
metaslab_trace_add(zal, mg, msp, asize, d,
TRACE_TOO_SMALL, allocator);
continue;
}
/*
* If the selected metaslab is condensing or disabled,
* skip it.
*/
if (msp->ms_condensing || msp->ms_disabled > 0)
continue;
*was_active = msp->ms_allocator != -1;
/*
* If we're activating as primary, this is our first allocation
* from this disk, so we don't need to check how close we are.
* If the metaslab under consideration was already active,
* we're getting desperate enough to steal another allocator's
* metaslab, so we still don't care about distances.
*/
if (activation_weight == METASLAB_WEIGHT_PRIMARY || *was_active)
break;
for (i = 0; i < d; i++) {
if (want_unique &&
!metaslab_is_unique(msp, &dva[i]))
break; /* try another metaslab */
}
if (i == d)
break;
}
if (msp != NULL) {
search->ms_weight = msp->ms_weight;
search->ms_start = msp->ms_start + 1;
search->ms_allocator = msp->ms_allocator;
search->ms_primary = msp->ms_primary;
}
return (msp);
}
static void
metaslab_active_mask_verify(metaslab_t *msp)
{
ASSERT(MUTEX_HELD(&msp->ms_lock));
if ((zfs_flags & ZFS_DEBUG_METASLAB_VERIFY) == 0)
return;
if ((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0)
return;
if (msp->ms_weight & METASLAB_WEIGHT_PRIMARY) {
VERIFY0(msp->ms_weight & METASLAB_WEIGHT_SECONDARY);
VERIFY0(msp->ms_weight & METASLAB_WEIGHT_CLAIM);
VERIFY3S(msp->ms_allocator, !=, -1);
VERIFY(msp->ms_primary);
return;
}
if (msp->ms_weight & METASLAB_WEIGHT_SECONDARY) {
VERIFY0(msp->ms_weight & METASLAB_WEIGHT_PRIMARY);
VERIFY0(msp->ms_weight & METASLAB_WEIGHT_CLAIM);
VERIFY3S(msp->ms_allocator, !=, -1);
VERIFY(!msp->ms_primary);
return;
}
if (msp->ms_weight & METASLAB_WEIGHT_CLAIM) {
VERIFY0(msp->ms_weight & METASLAB_WEIGHT_PRIMARY);
VERIFY0(msp->ms_weight & METASLAB_WEIGHT_SECONDARY);
VERIFY3S(msp->ms_allocator, ==, -1);
return;
}
}
static uint64_t
metaslab_group_alloc_normal(metaslab_group_t *mg, zio_alloc_list_t *zal,
uint64_t asize, uint64_t txg, boolean_t want_unique, dva_t *dva, int d,
int allocator, boolean_t try_hard)
{
metaslab_t *msp = NULL;
uint64_t offset = -1ULL;
uint64_t activation_weight = METASLAB_WEIGHT_PRIMARY;
for (int i = 0; i < d; i++) {
if (activation_weight == METASLAB_WEIGHT_PRIMARY &&
DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
activation_weight = METASLAB_WEIGHT_SECONDARY;
} else if (activation_weight == METASLAB_WEIGHT_SECONDARY &&
DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
activation_weight = METASLAB_WEIGHT_CLAIM;
break;
}
}
/*
* If we don't have enough metaslabs active to fill the entire array, we
* just use the 0th slot.
*/
if (mg->mg_ms_ready < mg->mg_allocators * 3)
allocator = 0;
metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
ASSERT3U(mg->mg_vd->vdev_ms_count, >=, 2);
metaslab_t *search = kmem_alloc(sizeof (*search), KM_SLEEP);
search->ms_weight = UINT64_MAX;
search->ms_start = 0;
/*
* At the end of the metaslab tree are the already-active metaslabs,
* first the primaries, then the secondaries. When we resume searching
* through the tree, we need to consider ms_allocator and ms_primary so
* we start in the location right after where we left off, and don't
* accidentally loop forever considering the same metaslabs.
*/
search->ms_allocator = -1;
search->ms_primary = B_TRUE;
for (;;) {
boolean_t was_active = B_FALSE;
mutex_enter(&mg->mg_lock);
if (activation_weight == METASLAB_WEIGHT_PRIMARY &&
mga->mga_primary != NULL) {
msp = mga->mga_primary;
/*
* Even though we don't hold the ms_lock for the
* primary metaslab, those fields should not
* change while we hold the mg_lock. Thus it is
* safe to make assertions on them.
*/
ASSERT(msp->ms_primary);
ASSERT3S(msp->ms_allocator, ==, allocator);
ASSERT(msp->ms_loaded);
was_active = B_TRUE;
ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
} else if (activation_weight == METASLAB_WEIGHT_SECONDARY &&
mga->mga_secondary != NULL) {
msp = mga->mga_secondary;
/*
* See comment above about the similar assertions
* for the primary metaslab.
*/
ASSERT(!msp->ms_primary);
ASSERT3S(msp->ms_allocator, ==, allocator);
ASSERT(msp->ms_loaded);
was_active = B_TRUE;
ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
} else {
msp = find_valid_metaslab(mg, activation_weight, dva, d,
want_unique, asize, allocator, try_hard, zal,
search, &was_active);
}
mutex_exit(&mg->mg_lock);
if (msp == NULL) {
kmem_free(search, sizeof (*search));
return (-1ULL);
}
mutex_enter(&msp->ms_lock);
metaslab_active_mask_verify(msp);
/*
* This code is disabled out because of issues with
* tracepoints in non-gpl kernel modules.
*/
#if 0
DTRACE_PROBE3(ms__activation__attempt,
metaslab_t *, msp, uint64_t, activation_weight,
boolean_t, was_active);
#endif
/*
* Ensure that the metaslab we have selected is still
* capable of handling our request. It's possible that
* another thread may have changed the weight while we
* were blocked on the metaslab lock. We check the
* active status first to see if we need to set_selected_txg
* a new metaslab.
*/
if (was_active && !(msp->ms_weight & METASLAB_ACTIVE_MASK)) {
ASSERT3S(msp->ms_allocator, ==, -1);
mutex_exit(&msp->ms_lock);
continue;
}
/*
* If the metaslab was activated for another allocator
* while we were waiting in the ms_lock above, or it's
* a primary and we're seeking a secondary (or vice versa),
* we go back and select a new metaslab.
*/
if (!was_active && (msp->ms_weight & METASLAB_ACTIVE_MASK) &&
(msp->ms_allocator != -1) &&
(msp->ms_allocator != allocator || ((activation_weight ==
METASLAB_WEIGHT_PRIMARY) != msp->ms_primary))) {
ASSERT(msp->ms_loaded);
ASSERT((msp->ms_weight & METASLAB_WEIGHT_CLAIM) ||
msp->ms_allocator != -1);
mutex_exit(&msp->ms_lock);
continue;
}
/*
* This metaslab was used for claiming regions allocated
* by the ZIL during pool import. Once these regions are
* claimed we don't need to keep the CLAIM bit set
* anymore. Passivate this metaslab to zero its activation
* mask.
*/
if (msp->ms_weight & METASLAB_WEIGHT_CLAIM &&
activation_weight != METASLAB_WEIGHT_CLAIM) {
ASSERT(msp->ms_loaded);
ASSERT3S(msp->ms_allocator, ==, -1);
metaslab_passivate(msp, msp->ms_weight &
~METASLAB_WEIGHT_CLAIM);
mutex_exit(&msp->ms_lock);
continue;
}
metaslab_set_selected_txg(msp, txg);
int activation_error =
metaslab_activate(msp, allocator, activation_weight);
metaslab_active_mask_verify(msp);
/*
* If the metaslab was activated by another thread for
* another allocator or activation_weight (EBUSY), or it
* failed because another metaslab was assigned as primary
* for this allocator (EEXIST) we continue using this
* metaslab for our allocation, rather than going on to a
* worse metaslab (we waited for that metaslab to be loaded
* after all).
*
* If the activation failed due to an I/O error or ENOSPC we
* skip to the next metaslab.
*/
boolean_t activated;
if (activation_error == 0) {
activated = B_TRUE;
} else if (activation_error == EBUSY ||
activation_error == EEXIST) {
activated = B_FALSE;
} else {
mutex_exit(&msp->ms_lock);
continue;
}
ASSERT(msp->ms_loaded);
/*
* Now that we have the lock, recheck to see if we should
* continue to use this metaslab for this allocation. The
* the metaslab is now loaded so metaslab_should_allocate()
* can accurately determine if the allocation attempt should
* proceed.
*/
if (!metaslab_should_allocate(msp, asize, try_hard)) {
/* Passivate this metaslab and select a new one. */
metaslab_trace_add(zal, mg, msp, asize, d,
TRACE_TOO_SMALL, allocator);
goto next;
}
/*
* If this metaslab is currently condensing then pick again
* as we can't manipulate this metaslab until it's committed
* to disk. If this metaslab is being initialized, we shouldn't
* allocate from it since the allocated region might be
* overwritten after allocation.
*/
if (msp->ms_condensing) {
metaslab_trace_add(zal, mg, msp, asize, d,
TRACE_CONDENSING, allocator);
if (activated) {
metaslab_passivate(msp, msp->ms_weight &
~METASLAB_ACTIVE_MASK);
}
mutex_exit(&msp->ms_lock);
continue;
} else if (msp->ms_disabled > 0) {
metaslab_trace_add(zal, mg, msp, asize, d,
TRACE_DISABLED, allocator);
if (activated) {
metaslab_passivate(msp, msp->ms_weight &
~METASLAB_ACTIVE_MASK);
}
mutex_exit(&msp->ms_lock);
continue;
}
offset = metaslab_block_alloc(msp, asize, txg);
metaslab_trace_add(zal, mg, msp, asize, d, offset, allocator);
if (offset != -1ULL) {
/* Proactively passivate the metaslab, if needed */
if (activated)
metaslab_segment_may_passivate(msp);
break;
}
next:
ASSERT(msp->ms_loaded);
/*
* This code is disabled out because of issues with
* tracepoints in non-gpl kernel modules.
*/
#if 0
DTRACE_PROBE2(ms__alloc__failure, metaslab_t *, msp,
uint64_t, asize);
#endif
/*
* We were unable to allocate from this metaslab so determine
* a new weight for this metaslab. Now that we have loaded
* the metaslab we can provide a better hint to the metaslab
* selector.
*
* For space-based metaslabs, we use the maximum block size.
* This information is only available when the metaslab
* is loaded and is more accurate than the generic free
* space weight that was calculated by metaslab_weight().
* This information allows us to quickly compare the maximum
* available allocation in the metaslab to the allocation
* size being requested.
*
* For segment-based metaslabs, determine the new weight
* based on the highest bucket in the range tree. We
* explicitly use the loaded segment weight (i.e. the range
* tree histogram) since it contains the space that is
* currently available for allocation and is accurate
* even within a sync pass.
*/
uint64_t weight;
if (WEIGHT_IS_SPACEBASED(msp->ms_weight)) {
weight = metaslab_largest_allocatable(msp);
WEIGHT_SET_SPACEBASED(weight);
} else {
weight = metaslab_weight_from_range_tree(msp);
}
if (activated) {
metaslab_passivate(msp, weight);
} else {
/*
* For the case where we use the metaslab that is
* active for another allocator we want to make
* sure that we retain the activation mask.
*
* Note that we could attempt to use something like
* metaslab_recalculate_weight_and_sort() that
* retains the activation mask here. That function
* uses metaslab_weight() to set the weight though
* which is not as accurate as the calculations
* above.
*/
weight |= msp->ms_weight & METASLAB_ACTIVE_MASK;
metaslab_group_sort(mg, msp, weight);
}
metaslab_active_mask_verify(msp);
/*
* We have just failed an allocation attempt, check
* that metaslab_should_allocate() agrees. Otherwise,
* we may end up in an infinite loop retrying the same
* metaslab.
*/
ASSERT(!metaslab_should_allocate(msp, asize, try_hard));
mutex_exit(&msp->ms_lock);
}
mutex_exit(&msp->ms_lock);
kmem_free(search, sizeof (*search));
return (offset);
}
static uint64_t
metaslab_group_alloc(metaslab_group_t *mg, zio_alloc_list_t *zal,
uint64_t asize, uint64_t txg, boolean_t want_unique, dva_t *dva, int d,
int allocator, boolean_t try_hard)
{
uint64_t offset;
ASSERT(mg->mg_initialized);
offset = metaslab_group_alloc_normal(mg, zal, asize, txg, want_unique,
dva, d, allocator, try_hard);
mutex_enter(&mg->mg_lock);
if (offset == -1ULL) {
mg->mg_failed_allocations++;
metaslab_trace_add(zal, mg, NULL, asize, d,
TRACE_GROUP_FAILURE, allocator);
if (asize == SPA_GANGBLOCKSIZE) {
/*
* This metaslab group was unable to allocate
* the minimum gang block size so it must be out of
* space. We must notify the allocation throttle
* to start skipping allocation attempts to this
* metaslab group until more space becomes available.
* Note: this failure cannot be caused by the
* allocation throttle since the allocation throttle
* is only responsible for skipping devices and
* not failing block allocations.
*/
mg->mg_no_free_space = B_TRUE;
}
}
mg->mg_allocations++;
mutex_exit(&mg->mg_lock);
return (offset);
}
/*
* Allocate a block for the specified i/o.
*/
int
metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
dva_t *dva, int d, dva_t *hintdva, uint64_t txg, int flags,
zio_alloc_list_t *zal, int allocator)
{
metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];
metaslab_group_t *mg, *fast_mg, *rotor;
vdev_t *vd;
boolean_t try_hard = B_FALSE;
ASSERT(!DVA_IS_VALID(&dva[d]));
/*
* For testing, make some blocks above a certain size be gang blocks.
* This will result in more split blocks when using device removal,
* and a large number of split blocks coupled with ztest-induced
* damage can result in extremely long reconstruction times. This
* will also test spilling from special to normal.
*/
if (psize >= metaslab_force_ganging && (random_in_range(100) < 3)) {
metaslab_trace_add(zal, NULL, NULL, psize, d, TRACE_FORCE_GANG,
allocator);
return (SET_ERROR(ENOSPC));
}
/*
* Start at the rotor and loop through all mgs until we find something.
* Note that there's no locking on mca_rotor or mca_aliquot because
* nothing actually breaks if we miss a few updates -- we just won't
* allocate quite as evenly. It all balances out over time.
*
* If we are doing ditto or log blocks, try to spread them across
* consecutive vdevs. If we're forced to reuse a vdev before we've
* allocated all of our ditto blocks, then try and spread them out on
* that vdev as much as possible. If it turns out to not be possible,
* gradually lower our standards until anything becomes acceptable.
* Also, allocating on consecutive vdevs (as opposed to random vdevs)
* gives us hope of containing our fault domains to something we're
* able to reason about. Otherwise, any two top-level vdev failures
* will guarantee the loss of data. With consecutive allocation,
* only two adjacent top-level vdev failures will result in data loss.
*
* If we are doing gang blocks (hintdva is non-NULL), try to keep
* ourselves on the same vdev as our gang block header. That
* way, we can hope for locality in vdev_cache, plus it makes our
* fault domains something tractable.
*/
if (hintdva) {
vd = vdev_lookup_top(spa, DVA_GET_VDEV(&hintdva[d]));
/*
* It's possible the vdev we're using as the hint no
* longer exists or its mg has been closed (e.g. by
* device removal). Consult the rotor when
* all else fails.
*/
if (vd != NULL && vd->vdev_mg != NULL) {
mg = vdev_get_mg(vd, mc);
if (flags & METASLAB_HINTBP_AVOID &&
mg->mg_next != NULL)
mg = mg->mg_next;
} else {
mg = mca->mca_rotor;
}
} else if (d != 0) {
vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d - 1]));
mg = vd->vdev_mg->mg_next;
} else if (flags & METASLAB_FASTWRITE) {
mg = fast_mg = mca->mca_rotor;
do {
if (fast_mg->mg_vd->vdev_pending_fastwrite <
mg->mg_vd->vdev_pending_fastwrite)
mg = fast_mg;
} while ((fast_mg = fast_mg->mg_next) != mca->mca_rotor);
} else {
ASSERT(mca->mca_rotor != NULL);
mg = mca->mca_rotor;
}
/*
* If the hint put us into the wrong metaslab class, or into a
* metaslab group that has been passivated, just follow the rotor.
*/
if (mg->mg_class != mc || mg->mg_activation_count <= 0)
mg = mca->mca_rotor;
rotor = mg;
top:
do {
boolean_t allocatable;
ASSERT(mg->mg_activation_count == 1);
vd = mg->mg_vd;
/*
* Don't allocate from faulted devices.
*/
if (try_hard) {
spa_config_enter(spa, SCL_ZIO, FTAG, RW_READER);
allocatable = vdev_allocatable(vd);
spa_config_exit(spa, SCL_ZIO, FTAG);
} else {
allocatable = vdev_allocatable(vd);
}
/*
* Determine if the selected metaslab group is eligible
* for allocations. If we're ganging then don't allow
* this metaslab group to skip allocations since that would
* inadvertently return ENOSPC and suspend the pool
* even though space is still available.
*/
if (allocatable && !GANG_ALLOCATION(flags) && !try_hard) {
allocatable = metaslab_group_allocatable(mg, rotor,
psize, allocator, d);
}
if (!allocatable) {
metaslab_trace_add(zal, mg, NULL, psize, d,
TRACE_NOT_ALLOCATABLE, allocator);
goto next;
}
ASSERT(mg->mg_initialized);
/*
* Avoid writing single-copy data to a failing,
* non-redundant vdev, unless we've already tried all
* other vdevs.
*/
if ((vd->vdev_stat.vs_write_errors > 0 ||
vd->vdev_state < VDEV_STATE_HEALTHY) &&
d == 0 && !try_hard && vd->vdev_children == 0) {
metaslab_trace_add(zal, mg, NULL, psize, d,
TRACE_VDEV_ERROR, allocator);
goto next;
}
ASSERT(mg->mg_class == mc);
uint64_t asize = vdev_psize_to_asize(vd, psize);
ASSERT(P2PHASE(asize, 1ULL << vd->vdev_ashift) == 0);
/*
* If we don't need to try hard, then require that the
* block be on a different metaslab from any other DVAs
* in this BP (unique=true). If we are trying hard, then
* allow any metaslab to be used (unique=false).
*/
uint64_t offset = metaslab_group_alloc(mg, zal, asize, txg,
!try_hard, dva, d, allocator, try_hard);
if (offset != -1ULL) {
/*
* If we've just selected this metaslab group,
* figure out whether the corresponding vdev is
* over- or under-used relative to the pool,
* and set an allocation bias to even it out.
*
* Bias is also used to compensate for unequally
* sized vdevs so that space is allocated fairly.
*/
if (mca->mca_aliquot == 0 && metaslab_bias_enabled) {
vdev_stat_t *vs = &vd->vdev_stat;
int64_t vs_free = vs->vs_space - vs->vs_alloc;
int64_t mc_free = mc->mc_space - mc->mc_alloc;
int64_t ratio;
/*
* Calculate how much more or less we should
* try to allocate from this device during
* this iteration around the rotor.
*
* This basically introduces a zero-centered
* bias towards the devices with the most
* free space, while compensating for vdev
* size differences.
*
* Examples:
* vdev V1 = 16M/128M
* vdev V2 = 16M/128M
* ratio(V1) = 100% ratio(V2) = 100%
*
* vdev V1 = 16M/128M
* vdev V2 = 64M/128M
* ratio(V1) = 127% ratio(V2) = 72%
*
* vdev V1 = 16M/128M
* vdev V2 = 64M/512M
* ratio(V1) = 40% ratio(V2) = 160%
*/
ratio = (vs_free * mc->mc_alloc_groups * 100) /
(mc_free + 1);
mg->mg_bias = ((ratio - 100) *
(int64_t)mg->mg_aliquot) / 100;
} else if (!metaslab_bias_enabled) {
mg->mg_bias = 0;
}
if ((flags & METASLAB_FASTWRITE) ||
atomic_add_64_nv(&mca->mca_aliquot, asize) >=
mg->mg_aliquot + mg->mg_bias) {
mca->mca_rotor = mg->mg_next;
mca->mca_aliquot = 0;
}
DVA_SET_VDEV(&dva[d], vd->vdev_id);
DVA_SET_OFFSET(&dva[d], offset);
DVA_SET_GANG(&dva[d],
((flags & METASLAB_GANG_HEADER) ? 1 : 0));
DVA_SET_ASIZE(&dva[d], asize);
if (flags & METASLAB_FASTWRITE) {
atomic_add_64(&vd->vdev_pending_fastwrite,
psize);
}
return (0);
}
next:
mca->mca_rotor = mg->mg_next;
mca->mca_aliquot = 0;
} while ((mg = mg->mg_next) != rotor);
/*
* If we haven't tried hard, perhaps do so now.
*/
if (!try_hard && (zfs_metaslab_try_hard_before_gang ||
GANG_ALLOCATION(flags) || (flags & METASLAB_ZIL) != 0 ||
psize <= 1 << spa->spa_min_ashift)) {
METASLABSTAT_BUMP(metaslabstat_try_hard);
try_hard = B_TRUE;
goto top;
}
bzero(&dva[d], sizeof (dva_t));
metaslab_trace_add(zal, rotor, NULL, psize, d, TRACE_ENOSPC, allocator);
return (SET_ERROR(ENOSPC));
}
void
metaslab_free_concrete(vdev_t *vd, uint64_t offset, uint64_t asize,
boolean_t checkpoint)
{
metaslab_t *msp;
spa_t *spa = vd->vdev_spa;
ASSERT(vdev_is_concrete(vd));
ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);
ASSERT3U(offset >> vd->vdev_ms_shift, <, vd->vdev_ms_count);
msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];
VERIFY(!msp->ms_condensing);
VERIFY3U(offset, >=, msp->ms_start);
VERIFY3U(offset + asize, <=, msp->ms_start + msp->ms_size);
VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
VERIFY0(P2PHASE(asize, 1ULL << vd->vdev_ashift));
metaslab_check_free_impl(vd, offset, asize);
mutex_enter(&msp->ms_lock);
if (range_tree_is_empty(msp->ms_freeing) &&
range_tree_is_empty(msp->ms_checkpointing)) {
vdev_dirty(vd, VDD_METASLAB, msp, spa_syncing_txg(spa));
}
if (checkpoint) {
ASSERT(spa_has_checkpoint(spa));
range_tree_add(msp->ms_checkpointing, offset, asize);
} else {
range_tree_add(msp->ms_freeing, offset, asize);
}
mutex_exit(&msp->ms_lock);
}
void
metaslab_free_impl_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
uint64_t size, void *arg)
{
(void) inner_offset;
boolean_t *checkpoint = arg;
ASSERT3P(checkpoint, !=, NULL);
if (vd->vdev_ops->vdev_op_remap != NULL)
vdev_indirect_mark_obsolete(vd, offset, size);
else
metaslab_free_impl(vd, offset, size, *checkpoint);
}
static void
metaslab_free_impl(vdev_t *vd, uint64_t offset, uint64_t size,
boolean_t checkpoint)
{
spa_t *spa = vd->vdev_spa;
ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);
if (spa_syncing_txg(spa) > spa_freeze_txg(spa))
return;
if (spa->spa_vdev_removal != NULL &&
spa->spa_vdev_removal->svr_vdev_id == vd->vdev_id &&
vdev_is_concrete(vd)) {
/*
* Note: we check if the vdev is concrete because when
* we complete the removal, we first change the vdev to be
* an indirect vdev (in open context), and then (in syncing
* context) clear spa_vdev_removal.
*/
free_from_removing_vdev(vd, offset, size);
} else if (vd->vdev_ops->vdev_op_remap != NULL) {
vdev_indirect_mark_obsolete(vd, offset, size);
vd->vdev_ops->vdev_op_remap(vd, offset, size,
metaslab_free_impl_cb, &checkpoint);
} else {
metaslab_free_concrete(vd, offset, size, checkpoint);
}
}
typedef struct remap_blkptr_cb_arg {
blkptr_t *rbca_bp;
spa_remap_cb_t rbca_cb;
vdev_t *rbca_remap_vd;
uint64_t rbca_remap_offset;
void *rbca_cb_arg;
} remap_blkptr_cb_arg_t;
static void
remap_blkptr_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
uint64_t size, void *arg)
{
remap_blkptr_cb_arg_t *rbca = arg;
blkptr_t *bp = rbca->rbca_bp;
/* We can not remap split blocks. */
if (size != DVA_GET_ASIZE(&bp->blk_dva[0]))
return;
ASSERT0(inner_offset);
if (rbca->rbca_cb != NULL) {
/*
* At this point we know that we are not handling split
* blocks and we invoke the callback on the previous
* vdev which must be indirect.
*/
ASSERT3P(rbca->rbca_remap_vd->vdev_ops, ==, &vdev_indirect_ops);
rbca->rbca_cb(rbca->rbca_remap_vd->vdev_id,
rbca->rbca_remap_offset, size, rbca->rbca_cb_arg);
/* set up remap_blkptr_cb_arg for the next call */
rbca->rbca_remap_vd = vd;
rbca->rbca_remap_offset = offset;
}
/*
* The phys birth time is that of dva[0]. This ensures that we know
* when each dva was written, so that resilver can determine which
* blocks need to be scrubbed (i.e. those written during the time
* the vdev was offline). It also ensures that the key used in
* the ARC hash table is unique (i.e. dva[0] + phys_birth). If
* we didn't change the phys_birth, a lookup in the ARC for a
* remapped BP could find the data that was previously stored at
* this vdev + offset.
*/
vdev_t *oldvd = vdev_lookup_top(vd->vdev_spa,
DVA_GET_VDEV(&bp->blk_dva[0]));
vdev_indirect_births_t *vib = oldvd->vdev_indirect_births;
bp->blk_phys_birth = vdev_indirect_births_physbirth(vib,
DVA_GET_OFFSET(&bp->blk_dva[0]), DVA_GET_ASIZE(&bp->blk_dva[0]));
DVA_SET_VDEV(&bp->blk_dva[0], vd->vdev_id);
DVA_SET_OFFSET(&bp->blk_dva[0], offset);
}
/*
* If the block pointer contains any indirect DVAs, modify them to refer to
* concrete DVAs. Note that this will sometimes not be possible, leaving
* the indirect DVA in place. This happens if the indirect DVA spans multiple
* segments in the mapping (i.e. it is a "split block").
*
* If the BP was remapped, calls the callback on the original dva (note the
* callback can be called multiple times if the original indirect DVA refers
* to another indirect DVA, etc).
*
* Returns TRUE if the BP was remapped.
*/
boolean_t
spa_remap_blkptr(spa_t *spa, blkptr_t *bp, spa_remap_cb_t callback, void *arg)
{
remap_blkptr_cb_arg_t rbca;
if (!zfs_remap_blkptr_enable)
return (B_FALSE);
if (!spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS))
return (B_FALSE);
/*
* Dedup BP's can not be remapped, because ddt_phys_select() depends
* on DVA[0] being the same in the BP as in the DDT (dedup table).
*/
if (BP_GET_DEDUP(bp))
return (B_FALSE);
/*
* Gang blocks can not be remapped, because
* zio_checksum_gang_verifier() depends on the DVA[0] that's in
* the BP used to read the gang block header (GBH) being the same
* as the DVA[0] that we allocated for the GBH.
*/
if (BP_IS_GANG(bp))
return (B_FALSE);
/*
* Embedded BP's have no DVA to remap.
*/
if (BP_GET_NDVAS(bp) < 1)
return (B_FALSE);
/*
* Note: we only remap dva[0]. If we remapped other dvas, we
* would no longer know what their phys birth txg is.
*/
dva_t *dva = &bp->blk_dva[0];
uint64_t offset = DVA_GET_OFFSET(dva);
uint64_t size = DVA_GET_ASIZE(dva);
vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));
if (vd->vdev_ops->vdev_op_remap == NULL)
return (B_FALSE);
rbca.rbca_bp = bp;
rbca.rbca_cb = callback;
rbca.rbca_remap_vd = vd;
rbca.rbca_remap_offset = offset;
rbca.rbca_cb_arg = arg;
/*
* remap_blkptr_cb() will be called in order for each level of
* indirection, until a concrete vdev is reached or a split block is
* encountered. old_vd and old_offset are updated within the callback
* as we go from the one indirect vdev to the next one (either concrete
* or indirect again) in that order.
*/
vd->vdev_ops->vdev_op_remap(vd, offset, size, remap_blkptr_cb, &rbca);
/* Check if the DVA wasn't remapped because it is a split block */
if (DVA_GET_VDEV(&rbca.rbca_bp->blk_dva[0]) == vd->vdev_id)
return (B_FALSE);
return (B_TRUE);
}
/*
* Undo the allocation of a DVA which happened in the given transaction group.
*/
void
metaslab_unalloc_dva(spa_t *spa, const dva_t *dva, uint64_t txg)
{
metaslab_t *msp;
vdev_t *vd;
uint64_t vdev = DVA_GET_VDEV(dva);
uint64_t offset = DVA_GET_OFFSET(dva);
uint64_t size = DVA_GET_ASIZE(dva);
ASSERT(DVA_IS_VALID(dva));
ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);
if (txg > spa_freeze_txg(spa))
return;
if ((vd = vdev_lookup_top(spa, vdev)) == NULL || !DVA_IS_VALID(dva) ||
(offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count) {
zfs_panic_recover("metaslab_free_dva(): bad DVA %llu:%llu:%llu",
(u_longlong_t)vdev, (u_longlong_t)offset,
(u_longlong_t)size);
return;
}
ASSERT(!vd->vdev_removing);
ASSERT(vdev_is_concrete(vd));
ASSERT0(vd->vdev_indirect_config.vic_mapping_object);
ASSERT3P(vd->vdev_indirect_mapping, ==, NULL);
if (DVA_GET_GANG(dva))
size = vdev_gang_header_asize(vd);
msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];
mutex_enter(&msp->ms_lock);
range_tree_remove(msp->ms_allocating[txg & TXG_MASK],
offset, size);
msp->ms_allocating_total -= size;
VERIFY(!msp->ms_condensing);
VERIFY3U(offset, >=, msp->ms_start);
VERIFY3U(offset + size, <=, msp->ms_start + msp->ms_size);
VERIFY3U(range_tree_space(msp->ms_allocatable) + size, <=,
msp->ms_size);
VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
range_tree_add(msp->ms_allocatable, offset, size);
mutex_exit(&msp->ms_lock);
}
/*
* Free the block represented by the given DVA.
*/
void
metaslab_free_dva(spa_t *spa, const dva_t *dva, boolean_t checkpoint)
{
uint64_t vdev = DVA_GET_VDEV(dva);
uint64_t offset = DVA_GET_OFFSET(dva);
uint64_t size = DVA_GET_ASIZE(dva);
vdev_t *vd = vdev_lookup_top(spa, vdev);
ASSERT(DVA_IS_VALID(dva));
ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);
if (DVA_GET_GANG(dva)) {
size = vdev_gang_header_asize(vd);
}
metaslab_free_impl(vd, offset, size, checkpoint);
}
/*
* Reserve some allocation slots. The reservation system must be called
* before we call into the allocator. If there aren't any available slots
* then the I/O will be throttled until an I/O completes and its slots are
* freed up. The function returns true if it was successful in placing
* the reservation.
*/
boolean_t
metaslab_class_throttle_reserve(metaslab_class_t *mc, int slots, int allocator,
zio_t *zio, int flags)
{
metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];
uint64_t max = mca->mca_alloc_max_slots;
ASSERT(mc->mc_alloc_throttle_enabled);
if (GANG_ALLOCATION(flags) || (flags & METASLAB_MUST_RESERVE) ||
zfs_refcount_count(&mca->mca_alloc_slots) + slots <= max) {
/*
* The potential race between _count() and _add() is covered
* by the allocator lock in most cases, or irrelevant due to
* GANG_ALLOCATION() or METASLAB_MUST_RESERVE set in others.
* But even if we assume some other non-existing scenario, the
* worst that can happen is few more I/Os get to allocation
* earlier, that is not a problem.
*
* We reserve the slots individually so that we can unreserve
* them individually when an I/O completes.
*/
for (int d = 0; d < slots; d++)
zfs_refcount_add(&mca->mca_alloc_slots, zio);
zio->io_flags |= ZIO_FLAG_IO_ALLOCATING;
return (B_TRUE);
}
return (B_FALSE);
}
void
metaslab_class_throttle_unreserve(metaslab_class_t *mc, int slots,
int allocator, zio_t *zio)
{
metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];
ASSERT(mc->mc_alloc_throttle_enabled);
for (int d = 0; d < slots; d++)
zfs_refcount_remove(&mca->mca_alloc_slots, zio);
}
static int
metaslab_claim_concrete(vdev_t *vd, uint64_t offset, uint64_t size,
uint64_t txg)
{
metaslab_t *msp;
spa_t *spa = vd->vdev_spa;
int error = 0;
if (offset >> vd->vdev_ms_shift >= vd->vdev_ms_count)
return (SET_ERROR(ENXIO));
ASSERT3P(vd->vdev_ms, !=, NULL);
msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];
mutex_enter(&msp->ms_lock);
if ((txg != 0 && spa_writeable(spa)) || !msp->ms_loaded) {
error = metaslab_activate(msp, 0, METASLAB_WEIGHT_CLAIM);
if (error == EBUSY) {
ASSERT(msp->ms_loaded);
ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
error = 0;
}
}
if (error == 0 &&
!range_tree_contains(msp->ms_allocatable, offset, size))
error = SET_ERROR(ENOENT);
if (error || txg == 0) { /* txg == 0 indicates dry run */
mutex_exit(&msp->ms_lock);
return (error);
}
VERIFY(!msp->ms_condensing);
VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
VERIFY3U(range_tree_space(msp->ms_allocatable) - size, <=,
msp->ms_size);
range_tree_remove(msp->ms_allocatable, offset, size);
range_tree_clear(msp->ms_trim, offset, size);
if (spa_writeable(spa)) { /* don't dirty if we're zdb(8) */
metaslab_class_t *mc = msp->ms_group->mg_class;
multilist_sublist_t *mls =
multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
if (!multilist_link_active(&msp->ms_class_txg_node)) {
msp->ms_selected_txg = txg;
multilist_sublist_insert_head(mls, msp);
}
multilist_sublist_unlock(mls);
if (range_tree_is_empty(msp->ms_allocating[txg & TXG_MASK]))
vdev_dirty(vd, VDD_METASLAB, msp, txg);
range_tree_add(msp->ms_allocating[txg & TXG_MASK],
offset, size);
msp->ms_allocating_total += size;
}
mutex_exit(&msp->ms_lock);
return (0);
}
typedef struct metaslab_claim_cb_arg_t {
uint64_t mcca_txg;
int mcca_error;
} metaslab_claim_cb_arg_t;
static void
metaslab_claim_impl_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
uint64_t size, void *arg)
{
(void) inner_offset;
metaslab_claim_cb_arg_t *mcca_arg = arg;
if (mcca_arg->mcca_error == 0) {
mcca_arg->mcca_error = metaslab_claim_concrete(vd, offset,
size, mcca_arg->mcca_txg);
}
}
int
metaslab_claim_impl(vdev_t *vd, uint64_t offset, uint64_t size, uint64_t txg)
{
if (vd->vdev_ops->vdev_op_remap != NULL) {
metaslab_claim_cb_arg_t arg;
/*
* Only zdb(8) can claim on indirect vdevs. This is used
* to detect leaks of mapped space (that are not accounted
* for in the obsolete counts, spacemap, or bpobj).
*/
ASSERT(!spa_writeable(vd->vdev_spa));
arg.mcca_error = 0;
arg.mcca_txg = txg;
vd->vdev_ops->vdev_op_remap(vd, offset, size,
metaslab_claim_impl_cb, &arg);
if (arg.mcca_error == 0) {
arg.mcca_error = metaslab_claim_concrete(vd,
offset, size, txg);
}
return (arg.mcca_error);
} else {
return (metaslab_claim_concrete(vd, offset, size, txg));
}
}
/*
* Intent log support: upon opening the pool after a crash, notify the SPA
* of blocks that the intent log has allocated for immediate write, but
* which are still considered free by the SPA because the last transaction
* group didn't commit yet.
*/
static int
metaslab_claim_dva(spa_t *spa, const dva_t *dva, uint64_t txg)
{
uint64_t vdev = DVA_GET_VDEV(dva);
uint64_t offset = DVA_GET_OFFSET(dva);
uint64_t size = DVA_GET_ASIZE(dva);
vdev_t *vd;
if ((vd = vdev_lookup_top(spa, vdev)) == NULL) {
return (SET_ERROR(ENXIO));
}
ASSERT(DVA_IS_VALID(dva));
if (DVA_GET_GANG(dva))
size = vdev_gang_header_asize(vd);
return (metaslab_claim_impl(vd, offset, size, txg));
}
int
metaslab_alloc(spa_t *spa, metaslab_class_t *mc, uint64_t psize, blkptr_t *bp,
int ndvas, uint64_t txg, blkptr_t *hintbp, int flags,
zio_alloc_list_t *zal, zio_t *zio, int allocator)
{
dva_t *dva = bp->blk_dva;
dva_t *hintdva = (hintbp != NULL) ? hintbp->blk_dva : NULL;
int error = 0;
ASSERT(bp->blk_birth == 0);
ASSERT(BP_PHYSICAL_BIRTH(bp) == 0);
spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);
if (mc->mc_allocator[allocator].mca_rotor == NULL) {
/* no vdevs in this class */
spa_config_exit(spa, SCL_ALLOC, FTAG);
return (SET_ERROR(ENOSPC));
}
ASSERT(ndvas > 0 && ndvas <= spa_max_replication(spa));
ASSERT(BP_GET_NDVAS(bp) == 0);
ASSERT(hintbp == NULL || ndvas <= BP_GET_NDVAS(hintbp));
ASSERT3P(zal, !=, NULL);
for (int d = 0; d < ndvas; d++) {
error = metaslab_alloc_dva(spa, mc, psize, dva, d, hintdva,
txg, flags, zal, allocator);
if (error != 0) {
for (d--; d >= 0; d--) {
metaslab_unalloc_dva(spa, &dva[d], txg);
metaslab_group_alloc_decrement(spa,
DVA_GET_VDEV(&dva[d]), zio, flags,
allocator, B_FALSE);
bzero(&dva[d], sizeof (dva_t));
}
spa_config_exit(spa, SCL_ALLOC, FTAG);
return (error);
} else {
/*
* Update the metaslab group's queue depth
* based on the newly allocated dva.
*/
metaslab_group_alloc_increment(spa,
DVA_GET_VDEV(&dva[d]), zio, flags, allocator);
}
}
ASSERT(error == 0);
ASSERT(BP_GET_NDVAS(bp) == ndvas);
spa_config_exit(spa, SCL_ALLOC, FTAG);
BP_SET_BIRTH(bp, txg, 0);
return (0);
}
void
metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg, boolean_t now)
{
const dva_t *dva = bp->blk_dva;
int ndvas = BP_GET_NDVAS(bp);
ASSERT(!BP_IS_HOLE(bp));
ASSERT(!now || bp->blk_birth >= spa_syncing_txg(spa));
/*
* If we have a checkpoint for the pool we need to make sure that
* the blocks that we free that are part of the checkpoint won't be
* reused until the checkpoint is discarded or we revert to it.
*
* The checkpoint flag is passed down the metaslab_free code path
* and is set whenever we want to add a block to the checkpoint's
* accounting. That is, we "checkpoint" blocks that existed at the
* time the checkpoint was created and are therefore referenced by
* the checkpointed uberblock.
*
* Note that, we don't checkpoint any blocks if the current
* syncing txg <= spa_checkpoint_txg. We want these frees to sync
* normally as they will be referenced by the checkpointed uberblock.
*/
boolean_t checkpoint = B_FALSE;
if (bp->blk_birth <= spa->spa_checkpoint_txg &&
spa_syncing_txg(spa) > spa->spa_checkpoint_txg) {
/*
* At this point, if the block is part of the checkpoint
* there is no way it was created in the current txg.
*/
ASSERT(!now);
ASSERT3U(spa_syncing_txg(spa), ==, txg);
checkpoint = B_TRUE;
}
spa_config_enter(spa, SCL_FREE, FTAG, RW_READER);
for (int d = 0; d < ndvas; d++) {
if (now) {
metaslab_unalloc_dva(spa, &dva[d], txg);
} else {
ASSERT3U(txg, ==, spa_syncing_txg(spa));
metaslab_free_dva(spa, &dva[d], checkpoint);
}
}
spa_config_exit(spa, SCL_FREE, FTAG);
}
int
metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg)
{
const dva_t *dva = bp->blk_dva;
int ndvas = BP_GET_NDVAS(bp);
int error = 0;
ASSERT(!BP_IS_HOLE(bp));
if (txg != 0) {
/*
* First do a dry run to make sure all DVAs are claimable,
* so we don't have to unwind from partial failures below.
*/
if ((error = metaslab_claim(spa, bp, 0)) != 0)
return (error);
}
spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);
for (int d = 0; d < ndvas; d++) {
error = metaslab_claim_dva(spa, &dva[d], txg);
if (error != 0)
break;
}
spa_config_exit(spa, SCL_ALLOC, FTAG);
ASSERT(error == 0 || txg == 0);
return (error);
}
void
metaslab_fastwrite_mark(spa_t *spa, const blkptr_t *bp)
{
const dva_t *dva = bp->blk_dva;
int ndvas = BP_GET_NDVAS(bp);
uint64_t psize = BP_GET_PSIZE(bp);
int d;
vdev_t *vd;
ASSERT(!BP_IS_HOLE(bp));
ASSERT(!BP_IS_EMBEDDED(bp));
ASSERT(psize > 0);
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
for (d = 0; d < ndvas; d++) {
if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
continue;
atomic_add_64(&vd->vdev_pending_fastwrite, psize);
}
spa_config_exit(spa, SCL_VDEV, FTAG);
}
void
metaslab_fastwrite_unmark(spa_t *spa, const blkptr_t *bp)
{
const dva_t *dva = bp->blk_dva;
int ndvas = BP_GET_NDVAS(bp);
uint64_t psize = BP_GET_PSIZE(bp);
int d;
vdev_t *vd;
ASSERT(!BP_IS_HOLE(bp));
ASSERT(!BP_IS_EMBEDDED(bp));
ASSERT(psize > 0);
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
for (d = 0; d < ndvas; d++) {
if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
continue;
ASSERT3U(vd->vdev_pending_fastwrite, >=, psize);
atomic_sub_64(&vd->vdev_pending_fastwrite, psize);
}
spa_config_exit(spa, SCL_VDEV, FTAG);
}
static void
metaslab_check_free_impl_cb(uint64_t inner, vdev_t *vd, uint64_t offset,
uint64_t size, void *arg)
{
(void) inner, (void) arg;
if (vd->vdev_ops == &vdev_indirect_ops)
return;
metaslab_check_free_impl(vd, offset, size);
}
static void
metaslab_check_free_impl(vdev_t *vd, uint64_t offset, uint64_t size)
{
metaslab_t *msp;
spa_t *spa __maybe_unused = vd->vdev_spa;
if ((zfs_flags & ZFS_DEBUG_ZIO_FREE) == 0)
return;
if (vd->vdev_ops->vdev_op_remap != NULL) {
vd->vdev_ops->vdev_op_remap(vd, offset, size,
metaslab_check_free_impl_cb, NULL);
return;
}
ASSERT(vdev_is_concrete(vd));
ASSERT3U(offset >> vd->vdev_ms_shift, <, vd->vdev_ms_count);
ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);
msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];
mutex_enter(&msp->ms_lock);
if (msp->ms_loaded) {
range_tree_verify_not_present(msp->ms_allocatable,
offset, size);
}
/*
* Check all segments that currently exist in the freeing pipeline.
*
* It would intuitively make sense to also check the current allocating
* tree since metaslab_unalloc_dva() exists for extents that are
* allocated and freed in the same sync pass within the same txg.
* Unfortunately there are places (e.g. the ZIL) where we allocate a
* segment but then we free part of it within the same txg
* [see zil_sync()]. Thus, we don't call range_tree_verify() in the
* current allocating tree.
*/
range_tree_verify_not_present(msp->ms_freeing, offset, size);
range_tree_verify_not_present(msp->ms_checkpointing, offset, size);
range_tree_verify_not_present(msp->ms_freed, offset, size);
for (int j = 0; j < TXG_DEFER_SIZE; j++)
range_tree_verify_not_present(msp->ms_defer[j], offset, size);
range_tree_verify_not_present(msp->ms_trim, offset, size);
mutex_exit(&msp->ms_lock);
}
void
metaslab_check_free(spa_t *spa, const blkptr_t *bp)
{
if ((zfs_flags & ZFS_DEBUG_ZIO_FREE) == 0)
return;
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
for (int i = 0; i < BP_GET_NDVAS(bp); i++) {
uint64_t vdev = DVA_GET_VDEV(&bp->blk_dva[i]);
vdev_t *vd = vdev_lookup_top(spa, vdev);
uint64_t offset = DVA_GET_OFFSET(&bp->blk_dva[i]);
uint64_t size = DVA_GET_ASIZE(&bp->blk_dva[i]);
if (DVA_GET_GANG(&bp->blk_dva[i]))
size = vdev_gang_header_asize(vd);
ASSERT3P(vd, !=, NULL);
metaslab_check_free_impl(vd, offset, size);
}
spa_config_exit(spa, SCL_VDEV, FTAG);
}
static void
metaslab_group_disable_wait(metaslab_group_t *mg)
{
ASSERT(MUTEX_HELD(&mg->mg_ms_disabled_lock));
while (mg->mg_disabled_updating) {
cv_wait(&mg->mg_ms_disabled_cv, &mg->mg_ms_disabled_lock);
}
}
static void
metaslab_group_disabled_increment(metaslab_group_t *mg)
{
ASSERT(MUTEX_HELD(&mg->mg_ms_disabled_lock));
ASSERT(mg->mg_disabled_updating);
while (mg->mg_ms_disabled >= max_disabled_ms) {
cv_wait(&mg->mg_ms_disabled_cv, &mg->mg_ms_disabled_lock);
}
mg->mg_ms_disabled++;
ASSERT3U(mg->mg_ms_disabled, <=, max_disabled_ms);
}
/*
* Mark the metaslab as disabled to prevent any allocations on this metaslab.
* We must also track how many metaslabs are currently disabled within a
* metaslab group and limit them to prevent allocation failures from
* occurring because all metaslabs are disabled.
*/
void
metaslab_disable(metaslab_t *msp)
{
ASSERT(!MUTEX_HELD(&msp->ms_lock));
metaslab_group_t *mg = msp->ms_group;
mutex_enter(&mg->mg_ms_disabled_lock);
/*
* To keep an accurate count of how many threads have disabled
* a specific metaslab group, we only allow one thread to mark
* the metaslab group at a time. This ensures that the value of
* ms_disabled will be accurate when we decide to mark a metaslab
* group as disabled. To do this we force all other threads
* to wait till the metaslab's mg_disabled_updating flag is no
* longer set.
*/
metaslab_group_disable_wait(mg);
mg->mg_disabled_updating = B_TRUE;
if (msp->ms_disabled == 0) {
metaslab_group_disabled_increment(mg);
}
mutex_enter(&msp->ms_lock);
msp->ms_disabled++;
mutex_exit(&msp->ms_lock);
mg->mg_disabled_updating = B_FALSE;
cv_broadcast(&mg->mg_ms_disabled_cv);
mutex_exit(&mg->mg_ms_disabled_lock);
}
void
metaslab_enable(metaslab_t *msp, boolean_t sync, boolean_t unload)
{
metaslab_group_t *mg = msp->ms_group;
spa_t *spa = mg->mg_vd->vdev_spa;
/*
* Wait for the outstanding IO to be synced to prevent newly
* allocated blocks from being overwritten. This used by
* initialize and TRIM which are modifying unallocated space.
*/
if (sync)
txg_wait_synced(spa_get_dsl(spa), 0);
mutex_enter(&mg->mg_ms_disabled_lock);
mutex_enter(&msp->ms_lock);
if (--msp->ms_disabled == 0) {
mg->mg_ms_disabled--;
cv_broadcast(&mg->mg_ms_disabled_cv);
if (unload)
metaslab_unload(msp);
}
mutex_exit(&msp->ms_lock);
mutex_exit(&mg->mg_ms_disabled_lock);
}
static void
metaslab_update_ondisk_flush_data(metaslab_t *ms, dmu_tx_t *tx)
{
vdev_t *vd = ms->ms_group->mg_vd;
spa_t *spa = vd->vdev_spa;
objset_t *mos = spa_meta_objset(spa);
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));
metaslab_unflushed_phys_t entry = {
.msp_unflushed_txg = metaslab_unflushed_txg(ms),
};
uint64_t entry_size = sizeof (entry);
uint64_t entry_offset = ms->ms_id * entry_size;
uint64_t object = 0;
int err = zap_lookup(mos, vd->vdev_top_zap,
VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, sizeof (uint64_t), 1,
&object);
if (err == ENOENT) {
object = dmu_object_alloc(mos, DMU_OTN_UINT64_METADATA,
SPA_OLD_MAXBLOCKSIZE, DMU_OT_NONE, 0, tx);
VERIFY0(zap_add(mos, vd->vdev_top_zap,
VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, sizeof (uint64_t), 1,
&object, tx));
} else {
VERIFY0(err);
}
dmu_write(spa_meta_objset(spa), object, entry_offset, entry_size,
&entry, tx);
}
void
metaslab_set_unflushed_txg(metaslab_t *ms, uint64_t txg, dmu_tx_t *tx)
{
spa_t *spa = ms->ms_group->mg_vd->vdev_spa;
if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))
return;
ms->ms_unflushed_txg = txg;
metaslab_update_ondisk_flush_data(ms, tx);
}
uint64_t
metaslab_unflushed_txg(metaslab_t *ms)
{
return (ms->ms_unflushed_txg);
}
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, aliquot, ULONG, ZMOD_RW,
"Allocation granularity (a.k.a. stripe size)");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, debug_load, INT, ZMOD_RW,
"Load all metaslabs when pool is first opened");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, debug_unload, INT, ZMOD_RW,
"Prevent metaslabs from being unloaded");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, preload_enabled, INT, ZMOD_RW,
"Preload potential metaslabs during reassessment");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, unload_delay, INT, ZMOD_RW,
"Delay in txgs after metaslab was last used before unloading");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, unload_delay_ms, INT, ZMOD_RW,
"Delay in milliseconds after metaslab was last used before unloading");
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_mg, zfs_mg_, noalloc_threshold, INT, ZMOD_RW,
"Percentage of metaslab group size that should be free to make it "
"eligible for allocation");
ZFS_MODULE_PARAM(zfs_mg, zfs_mg_, fragmentation_threshold, INT, ZMOD_RW,
"Percentage of metaslab group size that should be considered eligible "
"for allocations unless all metaslab groups within the metaslab class "
"have also crossed this threshold");
ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, fragmentation_threshold, INT,
ZMOD_RW, "Fragmentation for metaslab to allow allocation");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, fragmentation_factor_enabled, INT, ZMOD_RW,
"Use the fragmentation metric to prefer less fragmented metaslabs");
/* END CSTYLED */
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, lba_weighting_enabled, INT, ZMOD_RW,
"Prefer metaslabs with lower LBAs");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, bias_enabled, INT, ZMOD_RW,
"Enable metaslab group biasing");
ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, segment_weight_enabled, INT,
ZMOD_RW, "Enable segment-based metaslab selection");
ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, switch_threshold, INT, ZMOD_RW,
"Segment-based metaslab selection maximum buckets before switching");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, force_ganging, ULONG, ZMOD_RW,
"Blocks larger than this size are forced to be gang blocks");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, df_max_search, INT, ZMOD_RW,
"Max distance (bytes) to search forward before using size tree");
ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, df_use_largest_segment, INT, ZMOD_RW,
"When looking in size tree, use largest segment instead of exact fit");
ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, max_size_cache_sec, ULONG,
ZMOD_RW, "How long to trust the cached max chunk size of a metaslab");
ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, mem_limit, INT, ZMOD_RW,
"Percentage of memory that can be used to store metaslab range trees");
ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, try_hard_before_gang, INT,
ZMOD_RW, "Try hard to allocate before ganging");
ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, find_max_tries, INT, ZMOD_RW,
"Normally only consider this many of the best metaslabs in each vdev");
diff --git a/module/zfs/mmp.c b/module/zfs/mmp.c
index f67a4eb22a2d..abdce3a32e6a 100644
--- a/module/zfs/mmp.c
+++ b/module/zfs/mmp.c
@@ -1,744 +1,744 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2017 by Lawrence Livermore National Security, LLC.
*/
#include <sys/abd.h>
#include <sys/mmp.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/time.h>
#include <sys/vdev.h>
#include <sys/vdev_impl.h>
#include <sys/zfs_context.h>
#include <sys/callb.h>
/*
* Multi-Modifier Protection (MMP) attempts to prevent a user from importing
* or opening a pool on more than one host at a time. In particular, it
* prevents "zpool import -f" on a host from succeeding while the pool is
* already imported on another host. There are many other ways in which a
* device could be used by two hosts for different purposes at the same time
* resulting in pool damage. This implementation does not attempt to detect
* those cases.
*
* MMP operates by ensuring there are frequent visible changes on disk (a
* "heartbeat") at all times. And by altering the import process to check
* for these changes and failing the import when they are detected. This
* functionality is enabled by setting the 'multihost' pool property to on.
*
* Uberblocks written by the txg_sync thread always go into the first
* (N-MMP_BLOCKS_PER_LABEL) slots, the remaining slots are reserved for MMP.
* They are used to hold uberblocks which are exactly the same as the last
* synced uberblock except that the ub_timestamp and mmp_config are frequently
* updated. Like all other uberblocks, the slot is written with an embedded
* checksum, and slots with invalid checksums are ignored. This provides the
* "heartbeat", with no risk of overwriting good uberblocks that must be
* preserved, e.g. previous txgs and associated block pointers.
*
* Three optional fields are added to uberblock structure; ub_mmp_magic,
* ub_mmp_config, and ub_mmp_delay. The ub_mmp_magic value allows zfs to tell
* whether the other ub_mmp_* fields are valid. The ub_mmp_config field tells
* the importing host the settings of zfs_multihost_interval and
* zfs_multihost_fail_intervals on the host which last had (or currently has)
* the pool imported. These determine how long a host must wait to detect
* activity in the pool, before concluding the pool is not in use. The
* mmp_delay field is a decaying average of the amount of time between
* completion of successive MMP writes, in nanoseconds. It indicates whether
* MMP is enabled.
*
* During import an activity test may now be performed to determine if
* the pool is in use. The activity test is typically required if the
* ZPOOL_CONFIG_HOSTID does not match the system hostid, the pool state is
* POOL_STATE_ACTIVE, and the pool is not a root pool.
*
* The activity test finds the "best" uberblock (highest txg, timestamp, and, if
* ub_mmp_magic is valid, sequence number from ub_mmp_config). It then waits
* some time, and finds the "best" uberblock again. If any of the mentioned
* fields have different values in the newly read uberblock, the pool is in use
* by another host and the import fails. In order to assure the accuracy of the
* activity test, the default values result in an activity test duration of 20x
* the mmp write interval.
*
* The duration of the "zpool import" activity test depends on the information
* available in the "best" uberblock:
*
* 1) If uberblock was written by zfs-0.8 or newer and fail_intervals > 0:
* ub_mmp_config.fail_intervals * ub_mmp_config.multihost_interval * 2
*
* In this case, a weak guarantee is provided. Since the host which last had
* the pool imported will suspend the pool if no mmp writes land within
* fail_intervals * multihost_interval ms, the absence of writes during that
* time means either the pool is not imported, or it is imported but the pool
* is suspended and no further writes will occur.
*
* Note that resuming the suspended pool on the remote host would invalidate
* this guarantee, and so it is not allowed.
*
* The factor of 2 provides a conservative safety factor and derives from
* MMP_IMPORT_SAFETY_FACTOR;
*
* 2) If uberblock was written by zfs-0.8 or newer and fail_intervals == 0:
* (ub_mmp_config.multihost_interval + ub_mmp_delay) *
* zfs_multihost_import_intervals
*
* In this case no guarantee can provided. However, as long as some devices
* are healthy and connected, it is likely that at least one write will land
* within (multihost_interval + mmp_delay) because multihost_interval is
* enough time for a write to be attempted to each leaf vdev, and mmp_delay
* is enough for one to land, based on past delays. Multiplying by
* zfs_multihost_import_intervals provides a conservative safety factor.
*
* 3) If uberblock was written by zfs-0.7:
* (zfs_multihost_interval + ub_mmp_delay) * zfs_multihost_import_intervals
*
* The same logic as case #2 applies, but we do not know remote tunables.
*
* We use the local value for zfs_multihost_interval because the original MMP
* did not record this value in the uberblock.
*
* ub_mmp_delay >= (zfs_multihost_interval / leaves), so if the other host
* has a much larger zfs_multihost_interval set, ub_mmp_delay will reflect
* that. We will have waited enough time for zfs_multihost_import_intervals
* writes to be issued and all but one to land.
*
* single device pool example delays
*
* import_delay = (1 + 1) * 20 = 40s #defaults, no I/O delay
* import_delay = (1 + 10) * 20 = 220s #defaults, 10s I/O delay
* import_delay = (10 + 10) * 20 = 400s #10s multihost_interval,
* no I/O delay
* 100 device pool example delays
*
* import_delay = (1 + .01) * 20 = 20s #defaults, no I/O delay
* import_delay = (1 + 10) * 20 = 220s #defaults, 10s I/O delay
* import_delay = (10 + .1) * 20 = 202s #10s multihost_interval,
* no I/O delay
*
* 4) Otherwise, this uberblock was written by a pre-MMP zfs:
* zfs_multihost_import_intervals * zfs_multihost_interval
*
* In this case local tunables are used. By default this product = 10s, long
* enough for a pool with any activity at all to write at least one
* uberblock. No guarantee can be provided.
*
* Additionally, the duration is then extended by a random 25% to attempt to to
* detect simultaneous imports. For example, if both partner hosts are rebooted
* at the same time and automatically attempt to import the pool.
*/
/*
* Used to control the frequency of mmp writes which are performed when the
* 'multihost' pool property is on. This is one factor used to determine the
* length of the activity check during import.
*
* On average an mmp write will be issued for each leaf vdev every
* zfs_multihost_interval milliseconds. In practice, the observed period can
* vary with the I/O load and this observed value is the ub_mmp_delay which is
* stored in the uberblock. The minimum allowed value is 100 ms.
*/
ulong_t zfs_multihost_interval = MMP_DEFAULT_INTERVAL;
/*
* Used to control the duration of the activity test on import. Smaller values
* of zfs_multihost_import_intervals will reduce the import time but increase
* the risk of failing to detect an active pool. The total activity check time
* is never allowed to drop below one second. A value of 0 is ignored and
* treated as if it was set to 1.
*/
uint_t zfs_multihost_import_intervals = MMP_DEFAULT_IMPORT_INTERVALS;
/*
* Controls the behavior of the pool when mmp write failures or delays are
* detected.
*
* When zfs_multihost_fail_intervals = 0, mmp write failures or delays are
* ignored. The failures will still be reported to the ZED which depending on
* its configuration may take action such as suspending the pool or taking a
* device offline.
*
* When zfs_multihost_fail_intervals > 0, the pool will be suspended if
* zfs_multihost_fail_intervals * zfs_multihost_interval milliseconds pass
* without a successful mmp write. This guarantees the activity test will see
* mmp writes if the pool is imported. A value of 1 is ignored and treated as
* if it was set to 2, because a single leaf vdev pool will issue a write once
* per multihost_interval and thus any variation in latency would cause the
* pool to be suspended.
*/
uint_t zfs_multihost_fail_intervals = MMP_DEFAULT_FAIL_INTERVALS;
-char *mmp_tag = "mmp_write_uberblock";
+static void *const mmp_tag = "mmp_write_uberblock";
static void mmp_thread(void *arg);
void
mmp_init(spa_t *spa)
{
mmp_thread_t *mmp = &spa->spa_mmp;
mutex_init(&mmp->mmp_thread_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&mmp->mmp_thread_cv, NULL, CV_DEFAULT, NULL);
mutex_init(&mmp->mmp_io_lock, NULL, MUTEX_DEFAULT, NULL);
mmp->mmp_kstat_id = 1;
}
void
mmp_fini(spa_t *spa)
{
mmp_thread_t *mmp = &spa->spa_mmp;
mutex_destroy(&mmp->mmp_thread_lock);
cv_destroy(&mmp->mmp_thread_cv);
mutex_destroy(&mmp->mmp_io_lock);
}
static void
mmp_thread_enter(mmp_thread_t *mmp, callb_cpr_t *cpr)
{
CALLB_CPR_INIT(cpr, &mmp->mmp_thread_lock, callb_generic_cpr, FTAG);
mutex_enter(&mmp->mmp_thread_lock);
}
static void
mmp_thread_exit(mmp_thread_t *mmp, kthread_t **mpp, callb_cpr_t *cpr)
{
ASSERT(*mpp != NULL);
*mpp = NULL;
cv_broadcast(&mmp->mmp_thread_cv);
CALLB_CPR_EXIT(cpr); /* drops &mmp->mmp_thread_lock */
thread_exit();
}
void
mmp_thread_start(spa_t *spa)
{
mmp_thread_t *mmp = &spa->spa_mmp;
if (spa_writeable(spa)) {
mutex_enter(&mmp->mmp_thread_lock);
if (!mmp->mmp_thread) {
mmp->mmp_thread = thread_create(NULL, 0, mmp_thread,
spa, 0, &p0, TS_RUN, defclsyspri);
zfs_dbgmsg("MMP thread started pool '%s' "
"gethrtime %llu", spa_name(spa), gethrtime());
}
mutex_exit(&mmp->mmp_thread_lock);
}
}
void
mmp_thread_stop(spa_t *spa)
{
mmp_thread_t *mmp = &spa->spa_mmp;
mutex_enter(&mmp->mmp_thread_lock);
mmp->mmp_thread_exiting = 1;
cv_broadcast(&mmp->mmp_thread_cv);
while (mmp->mmp_thread) {
cv_wait(&mmp->mmp_thread_cv, &mmp->mmp_thread_lock);
}
mutex_exit(&mmp->mmp_thread_lock);
zfs_dbgmsg("MMP thread stopped pool '%s' gethrtime %llu",
spa_name(spa), gethrtime());
ASSERT(mmp->mmp_thread == NULL);
mmp->mmp_thread_exiting = 0;
}
typedef enum mmp_vdev_state_flag {
MMP_FAIL_NOT_WRITABLE = (1 << 0),
MMP_FAIL_WRITE_PENDING = (1 << 1),
} mmp_vdev_state_flag_t;
/*
* Find a leaf vdev to write an MMP block to. It must not have an outstanding
* mmp write (if so a new write will also likely block). If there is no usable
* leaf, a nonzero error value is returned. The error value returned is a bit
* field.
*
* MMP_FAIL_WRITE_PENDING One or more leaf vdevs are writeable, but have an
* outstanding MMP write.
* MMP_FAIL_NOT_WRITABLE One or more leaf vdevs are not writeable.
*/
static int
mmp_next_leaf(spa_t *spa)
{
vdev_t *leaf;
vdev_t *starting_leaf;
int fail_mask = 0;
ASSERT(MUTEX_HELD(&spa->spa_mmp.mmp_io_lock));
ASSERT(spa_config_held(spa, SCL_STATE, RW_READER));
ASSERT(list_link_active(&spa->spa_leaf_list.list_head) == B_TRUE);
ASSERT(!list_is_empty(&spa->spa_leaf_list));
if (spa->spa_mmp.mmp_leaf_last_gen != spa->spa_leaf_list_gen) {
spa->spa_mmp.mmp_last_leaf = list_head(&spa->spa_leaf_list);
spa->spa_mmp.mmp_leaf_last_gen = spa->spa_leaf_list_gen;
}
leaf = spa->spa_mmp.mmp_last_leaf;
if (leaf == NULL)
leaf = list_head(&spa->spa_leaf_list);
starting_leaf = leaf;
do {
leaf = list_next(&spa->spa_leaf_list, leaf);
if (leaf == NULL)
leaf = list_head(&spa->spa_leaf_list);
/*
* We skip unwritable, offline, detached, and dRAID spare
* devices as they are either not legal targets or the write
* may fail or not be seen by other hosts. Skipped dRAID
* spares can never be written so the fail mask is not set.
*/
if (!vdev_writeable(leaf) || leaf->vdev_offline ||
leaf->vdev_detached) {
fail_mask |= MMP_FAIL_NOT_WRITABLE;
} else if (leaf->vdev_ops == &vdev_draid_spare_ops) {
continue;
} else if (leaf->vdev_mmp_pending != 0) {
fail_mask |= MMP_FAIL_WRITE_PENDING;
} else {
spa->spa_mmp.mmp_last_leaf = leaf;
return (0);
}
} while (leaf != starting_leaf);
ASSERT(fail_mask);
return (fail_mask);
}
/*
* MMP writes are issued on a fixed schedule, but may complete at variable,
* much longer, intervals. The mmp_delay captures long periods between
* successful writes for any reason, including disk latency, scheduling delays,
* etc.
*
* The mmp_delay is usually calculated as a decaying average, but if the latest
* delay is higher we do not average it, so that we do not hide sudden spikes
* which the importing host must wait for.
*
* If writes are occurring frequently, such as due to a high rate of txg syncs,
* the mmp_delay could become very small. Since those short delays depend on
* activity we cannot count on, we never allow mmp_delay to get lower than rate
* expected if only mmp_thread writes occur.
*
* If an mmp write was skipped or fails, and we have already waited longer than
* mmp_delay, we need to update it so the next write reflects the longer delay.
*
* Do not set mmp_delay if the multihost property is not on, so as not to
* trigger an activity check on import.
*/
static void
mmp_delay_update(spa_t *spa, boolean_t write_completed)
{
mmp_thread_t *mts = &spa->spa_mmp;
hrtime_t delay = gethrtime() - mts->mmp_last_write;
ASSERT(MUTEX_HELD(&mts->mmp_io_lock));
if (spa_multihost(spa) == B_FALSE) {
mts->mmp_delay = 0;
return;
}
if (delay > mts->mmp_delay)
mts->mmp_delay = delay;
if (write_completed == B_FALSE)
return;
mts->mmp_last_write = gethrtime();
/*
* strictly less than, in case delay was changed above.
*/
if (delay < mts->mmp_delay) {
hrtime_t min_delay =
MSEC2NSEC(MMP_INTERVAL_OK(zfs_multihost_interval)) /
MAX(1, vdev_count_leaves(spa));
mts->mmp_delay = MAX(((delay + mts->mmp_delay * 127) / 128),
min_delay);
}
}
static void
mmp_write_done(zio_t *zio)
{
spa_t *spa = zio->io_spa;
vdev_t *vd = zio->io_vd;
mmp_thread_t *mts = zio->io_private;
mutex_enter(&mts->mmp_io_lock);
uint64_t mmp_kstat_id = vd->vdev_mmp_kstat_id;
hrtime_t mmp_write_duration = gethrtime() - vd->vdev_mmp_pending;
mmp_delay_update(spa, (zio->io_error == 0));
vd->vdev_mmp_pending = 0;
vd->vdev_mmp_kstat_id = 0;
mutex_exit(&mts->mmp_io_lock);
spa_config_exit(spa, SCL_STATE, mmp_tag);
spa_mmp_history_set(spa, mmp_kstat_id, zio->io_error,
mmp_write_duration);
abd_free(zio->io_abd);
}
/*
* When the uberblock on-disk is updated by a spa_sync,
* creating a new "best" uberblock, update the one stored
* in the mmp thread state, used for mmp writes.
*/
void
mmp_update_uberblock(spa_t *spa, uberblock_t *ub)
{
mmp_thread_t *mmp = &spa->spa_mmp;
mutex_enter(&mmp->mmp_io_lock);
mmp->mmp_ub = *ub;
mmp->mmp_seq = 1;
mmp->mmp_ub.ub_timestamp = gethrestime_sec();
mmp_delay_update(spa, B_TRUE);
mutex_exit(&mmp->mmp_io_lock);
}
/*
* Choose a random vdev, label, and MMP block, and write over it
* with a copy of the last-synced uberblock, whose timestamp
* has been updated to reflect that the pool is in use.
*/
static void
mmp_write_uberblock(spa_t *spa)
{
int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
mmp_thread_t *mmp = &spa->spa_mmp;
uberblock_t *ub;
vdev_t *vd = NULL;
int label, error;
uint64_t offset;
hrtime_t lock_acquire_time = gethrtime();
spa_config_enter(spa, SCL_STATE, mmp_tag, RW_READER);
lock_acquire_time = gethrtime() - lock_acquire_time;
if (lock_acquire_time > (MSEC2NSEC(MMP_MIN_INTERVAL) / 10))
zfs_dbgmsg("MMP SCL_STATE acquisition pool '%s' took %llu ns "
"gethrtime %llu", spa_name(spa), lock_acquire_time,
gethrtime());
mutex_enter(&mmp->mmp_io_lock);
error = mmp_next_leaf(spa);
/*
* spa_mmp_history has two types of entries:
* Issued MMP write: records time issued, error status, etc.
* Skipped MMP write: an MMP write could not be issued because no
* suitable leaf vdev was available. See comment above struct
* spa_mmp_history for details.
*/
if (error) {
mmp_delay_update(spa, B_FALSE);
if (mmp->mmp_skip_error == error) {
spa_mmp_history_set_skip(spa, mmp->mmp_kstat_id - 1);
} else {
mmp->mmp_skip_error = error;
spa_mmp_history_add(spa, mmp->mmp_ub.ub_txg,
gethrestime_sec(), mmp->mmp_delay, NULL, 0,
mmp->mmp_kstat_id++, error);
zfs_dbgmsg("MMP error choosing leaf pool '%s' "
"gethrtime %llu fail_mask %#x", spa_name(spa),
gethrtime(), error);
}
mutex_exit(&mmp->mmp_io_lock);
spa_config_exit(spa, SCL_STATE, mmp_tag);
return;
}
vd = spa->spa_mmp.mmp_last_leaf;
if (mmp->mmp_skip_error != 0) {
mmp->mmp_skip_error = 0;
zfs_dbgmsg("MMP write after skipping due to unavailable "
"leaves, pool '%s' gethrtime %llu leaf %llu",
spa_name(spa), (u_longlong_t)gethrtime(),
(u_longlong_t)vd->vdev_guid);
}
if (mmp->mmp_zio_root == NULL)
mmp->mmp_zio_root = zio_root(spa, NULL, NULL,
flags | ZIO_FLAG_GODFATHER);
if (mmp->mmp_ub.ub_timestamp != gethrestime_sec()) {
/*
* Want to reset mmp_seq when timestamp advances because after
* an mmp_seq wrap new values will not be chosen by
* uberblock_compare() as the "best".
*/
mmp->mmp_ub.ub_timestamp = gethrestime_sec();
mmp->mmp_seq = 1;
}
ub = &mmp->mmp_ub;
ub->ub_mmp_magic = MMP_MAGIC;
ub->ub_mmp_delay = mmp->mmp_delay;
ub->ub_mmp_config = MMP_SEQ_SET(mmp->mmp_seq) |
MMP_INTERVAL_SET(MMP_INTERVAL_OK(zfs_multihost_interval)) |
MMP_FAIL_INT_SET(MMP_FAIL_INTVS_OK(
zfs_multihost_fail_intervals));
vd->vdev_mmp_pending = gethrtime();
vd->vdev_mmp_kstat_id = mmp->mmp_kstat_id;
zio_t *zio = zio_null(mmp->mmp_zio_root, spa, NULL, NULL, NULL, flags);
abd_t *ub_abd = abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd), B_TRUE);
abd_zero(ub_abd, VDEV_UBERBLOCK_SIZE(vd));
abd_copy_from_buf(ub_abd, ub, sizeof (uberblock_t));
mmp->mmp_seq++;
mmp->mmp_kstat_id++;
mutex_exit(&mmp->mmp_io_lock);
offset = VDEV_UBERBLOCK_OFFSET(vd, VDEV_UBERBLOCK_COUNT(vd) -
MMP_BLOCKS_PER_LABEL + random_in_range(MMP_BLOCKS_PER_LABEL));
label = random_in_range(VDEV_LABELS);
vdev_label_write(zio, vd, label, ub_abd, offset,
VDEV_UBERBLOCK_SIZE(vd), mmp_write_done, mmp,
flags | ZIO_FLAG_DONT_PROPAGATE);
(void) spa_mmp_history_add(spa, ub->ub_txg, ub->ub_timestamp,
ub->ub_mmp_delay, vd, label, vd->vdev_mmp_kstat_id, 0);
zio_nowait(zio);
}
static void
mmp_thread(void *arg)
{
spa_t *spa = (spa_t *)arg;
mmp_thread_t *mmp = &spa->spa_mmp;
boolean_t suspended = spa_suspended(spa);
boolean_t multihost = spa_multihost(spa);
uint64_t mmp_interval = MSEC2NSEC(MMP_INTERVAL_OK(
zfs_multihost_interval));
uint32_t mmp_fail_intervals = MMP_FAIL_INTVS_OK(
zfs_multihost_fail_intervals);
hrtime_t mmp_fail_ns = mmp_fail_intervals * mmp_interval;
boolean_t last_spa_suspended = suspended;
boolean_t last_spa_multihost = multihost;
uint64_t last_mmp_interval = mmp_interval;
uint32_t last_mmp_fail_intervals = mmp_fail_intervals;
hrtime_t last_mmp_fail_ns = mmp_fail_ns;
callb_cpr_t cpr;
int skip_wait = 0;
mmp_thread_enter(mmp, &cpr);
/*
* There have been no MMP writes yet. Setting mmp_last_write here gives
* us one mmp_fail_ns period, which is consistent with the activity
* check duration, to try to land an MMP write before MMP suspends the
* pool (if so configured).
*/
mutex_enter(&mmp->mmp_io_lock);
mmp->mmp_last_write = gethrtime();
mmp->mmp_delay = MSEC2NSEC(MMP_INTERVAL_OK(zfs_multihost_interval));
mutex_exit(&mmp->mmp_io_lock);
while (!mmp->mmp_thread_exiting) {
hrtime_t next_time = gethrtime() +
MSEC2NSEC(MMP_DEFAULT_INTERVAL);
int leaves = MAX(vdev_count_leaves(spa), 1);
/* Detect changes in tunables or state */
last_spa_suspended = suspended;
last_spa_multihost = multihost;
suspended = spa_suspended(spa);
multihost = spa_multihost(spa);
last_mmp_interval = mmp_interval;
last_mmp_fail_intervals = mmp_fail_intervals;
last_mmp_fail_ns = mmp_fail_ns;
mmp_interval = MSEC2NSEC(MMP_INTERVAL_OK(
zfs_multihost_interval));
mmp_fail_intervals = MMP_FAIL_INTVS_OK(
zfs_multihost_fail_intervals);
/* Smooth so pool is not suspended when reducing tunables */
if (mmp_fail_intervals * mmp_interval < mmp_fail_ns) {
mmp_fail_ns = (mmp_fail_ns * 31 +
mmp_fail_intervals * mmp_interval) / 32;
} else {
mmp_fail_ns = mmp_fail_intervals *
mmp_interval;
}
if (mmp_interval != last_mmp_interval ||
mmp_fail_intervals != last_mmp_fail_intervals) {
/*
* We want other hosts to see new tunables as quickly as
* possible. Write out at higher frequency than usual.
*/
skip_wait += leaves;
}
if (multihost)
next_time = gethrtime() + mmp_interval / leaves;
if (mmp_fail_ns != last_mmp_fail_ns) {
zfs_dbgmsg("MMP interval change pool '%s' "
"gethrtime %llu last_mmp_interval %llu "
"mmp_interval %llu last_mmp_fail_intervals %u "
"mmp_fail_intervals %u mmp_fail_ns %llu "
"skip_wait %d leaves %d next_time %llu",
spa_name(spa), (u_longlong_t)gethrtime(),
(u_longlong_t)last_mmp_interval,
(u_longlong_t)mmp_interval, last_mmp_fail_intervals,
mmp_fail_intervals, (u_longlong_t)mmp_fail_ns,
skip_wait, leaves, (u_longlong_t)next_time);
}
/*
* MMP off => on, or suspended => !suspended:
* No writes occurred recently. Update mmp_last_write to give
* us some time to try.
*/
if ((!last_spa_multihost && multihost) ||
(last_spa_suspended && !suspended)) {
zfs_dbgmsg("MMP state change pool '%s': gethrtime %llu "
"last_spa_multihost %u multihost %u "
"last_spa_suspended %u suspended %u",
spa_name(spa), (u_longlong_t)gethrtime(),
last_spa_multihost, multihost, last_spa_suspended,
suspended);
mutex_enter(&mmp->mmp_io_lock);
mmp->mmp_last_write = gethrtime();
mmp->mmp_delay = mmp_interval;
mutex_exit(&mmp->mmp_io_lock);
}
/*
* MMP on => off:
* mmp_delay == 0 tells importing node to skip activity check.
*/
if (last_spa_multihost && !multihost) {
mutex_enter(&mmp->mmp_io_lock);
mmp->mmp_delay = 0;
mutex_exit(&mmp->mmp_io_lock);
}
/*
* Suspend the pool if no MMP write has succeeded in over
* mmp_interval * mmp_fail_intervals nanoseconds.
*/
if (multihost && !suspended && mmp_fail_intervals &&
(gethrtime() - mmp->mmp_last_write) > mmp_fail_ns) {
zfs_dbgmsg("MMP suspending pool '%s': gethrtime %llu "
"mmp_last_write %llu mmp_interval %llu "
"mmp_fail_intervals %llu mmp_fail_ns %llu",
spa_name(spa), (u_longlong_t)gethrtime(),
(u_longlong_t)mmp->mmp_last_write,
(u_longlong_t)mmp_interval,
(u_longlong_t)mmp_fail_intervals,
(u_longlong_t)mmp_fail_ns);
cmn_err(CE_WARN, "MMP writes to pool '%s' have not "
"succeeded in over %llu ms; suspending pool. "
"Hrtime %llu",
spa_name(spa),
NSEC2MSEC(gethrtime() - mmp->mmp_last_write),
gethrtime());
zio_suspend(spa, NULL, ZIO_SUSPEND_MMP);
}
if (multihost && !suspended)
mmp_write_uberblock(spa);
if (skip_wait > 0) {
next_time = gethrtime() + MSEC2NSEC(MMP_MIN_INTERVAL) /
leaves;
skip_wait--;
}
CALLB_CPR_SAFE_BEGIN(&cpr);
(void) cv_timedwait_idle_hires(&mmp->mmp_thread_cv,
&mmp->mmp_thread_lock, next_time, USEC2NSEC(100),
CALLOUT_FLAG_ABSOLUTE);
CALLB_CPR_SAFE_END(&cpr, &mmp->mmp_thread_lock);
}
/* Outstanding writes are allowed to complete. */
zio_wait(mmp->mmp_zio_root);
mmp->mmp_zio_root = NULL;
mmp_thread_exit(mmp, &mmp->mmp_thread, &cpr);
}
/*
* Signal the MMP thread to wake it, when it is sleeping on
* its cv. Used when some module parameter has changed and
* we want the thread to know about it.
* Only signal if the pool is active and mmp thread is
* running, otherwise there is no thread to wake.
*/
static void
mmp_signal_thread(spa_t *spa)
{
mmp_thread_t *mmp = &spa->spa_mmp;
mutex_enter(&mmp->mmp_thread_lock);
if (mmp->mmp_thread)
cv_broadcast(&mmp->mmp_thread_cv);
mutex_exit(&mmp->mmp_thread_lock);
}
void
mmp_signal_all_threads(void)
{
spa_t *spa = NULL;
mutex_enter(&spa_namespace_lock);
while ((spa = spa_next(spa))) {
if (spa->spa_state == POOL_STATE_ACTIVE)
mmp_signal_thread(spa);
}
mutex_exit(&spa_namespace_lock);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM_CALL(zfs_multihost, zfs_multihost_, interval,
param_set_multihost_interval, param_get_ulong, ZMOD_RW,
"Milliseconds between mmp writes to each leaf");
/* END CSTYLED */
ZFS_MODULE_PARAM(zfs_multihost, zfs_multihost_, fail_intervals, UINT, ZMOD_RW,
"Max allowed period without a successful mmp write");
ZFS_MODULE_PARAM(zfs_multihost, zfs_multihost_, import_intervals, UINT, ZMOD_RW,
"Number of zfs_multihost_interval periods to wait for activity");
diff --git a/module/zfs/range_tree.c b/module/zfs/range_tree.c
index 67910f9ffde0..f9fb97476b64 100644
--- a/module/zfs/range_tree.c
+++ b/module/zfs/range_tree.c
@@ -1,921 +1,921 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright (c) 2013, 2019 by Delphix. All rights reserved.
* Copyright (c) 2015, Nexenta Systems, Inc. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/dmu.h>
#include <sys/dnode.h>
#include <sys/zio.h>
#include <sys/range_tree.h>
/*
* Range trees are tree-based data structures that can be used to
* track free space or generally any space allocation information.
* A range tree keeps track of individual segments and automatically
* provides facilities such as adjacent extent merging and extent
* splitting in response to range add/remove requests.
*
* A range tree starts out completely empty, with no segments in it.
* Adding an allocation via range_tree_add to the range tree can either:
* 1) create a new extent
* 2) extend an adjacent extent
* 3) merge two adjacent extents
* Conversely, removing an allocation via range_tree_remove can:
* 1) completely remove an extent
* 2) shorten an extent (if the allocation was near one of its ends)
* 3) split an extent into two extents, in effect punching a hole
*
* A range tree is also capable of 'bridging' gaps when adding
* allocations. This is useful for cases when close proximity of
* allocations is an important detail that needs to be represented
* in the range tree. See range_tree_set_gap(). The default behavior
* is not to bridge gaps (i.e. the maximum allowed gap size is 0).
*
* In order to traverse a range tree, use either the range_tree_walk()
* or range_tree_vacate() functions.
*
* To obtain more accurate information on individual segment
* operations that the range tree performs "under the hood", you can
* specify a set of callbacks by passing a range_tree_ops_t structure
* to the range_tree_create function. Any callbacks that are non-NULL
* are then called at the appropriate times.
*
* The range tree code also supports a special variant of range trees
* that can bridge small gaps between segments. This kind of tree is used
* by the dsl scanning code to group I/Os into mostly sequential chunks to
* optimize disk performance. The code here attempts to do this with as
* little memory and computational overhead as possible. One limitation of
* this implementation is that segments of range trees with gaps can only
* support removing complete segments.
*/
static inline void
rs_copy(range_seg_t *src, range_seg_t *dest, range_tree_t *rt)
{
ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
size_t size = 0;
switch (rt->rt_type) {
case RANGE_SEG32:
size = sizeof (range_seg32_t);
break;
case RANGE_SEG64:
size = sizeof (range_seg64_t);
break;
case RANGE_SEG_GAP:
size = sizeof (range_seg_gap_t);
break;
default:
VERIFY(0);
}
bcopy(src, dest, size);
}
void
range_tree_stat_verify(range_tree_t *rt)
{
range_seg_t *rs;
zfs_btree_index_t where;
uint64_t hist[RANGE_TREE_HISTOGRAM_SIZE] = { 0 };
int i;
for (rs = zfs_btree_first(&rt->rt_root, &where); rs != NULL;
rs = zfs_btree_next(&rt->rt_root, &where, &where)) {
uint64_t size = rs_get_end(rs, rt) - rs_get_start(rs, rt);
int idx = highbit64(size) - 1;
hist[idx]++;
ASSERT3U(hist[idx], !=, 0);
}
for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
if (hist[i] != rt->rt_histogram[i]) {
zfs_dbgmsg("i=%d, hist=%px, hist=%llu, rt_hist=%llu",
i, hist, (u_longlong_t)hist[i],
(u_longlong_t)rt->rt_histogram[i]);
}
VERIFY3U(hist[i], ==, rt->rt_histogram[i]);
}
}
static void
range_tree_stat_incr(range_tree_t *rt, range_seg_t *rs)
{
uint64_t size = rs_get_end(rs, rt) - rs_get_start(rs, rt);
int idx = highbit64(size) - 1;
ASSERT(size != 0);
ASSERT3U(idx, <,
sizeof (rt->rt_histogram) / sizeof (*rt->rt_histogram));
rt->rt_histogram[idx]++;
ASSERT3U(rt->rt_histogram[idx], !=, 0);
}
static void
range_tree_stat_decr(range_tree_t *rt, range_seg_t *rs)
{
uint64_t size = rs_get_end(rs, rt) - rs_get_start(rs, rt);
int idx = highbit64(size) - 1;
ASSERT(size != 0);
ASSERT3U(idx, <,
sizeof (rt->rt_histogram) / sizeof (*rt->rt_histogram));
ASSERT3U(rt->rt_histogram[idx], !=, 0);
rt->rt_histogram[idx]--;
}
static int
range_tree_seg32_compare(const void *x1, const void *x2)
{
const range_seg32_t *r1 = x1;
const range_seg32_t *r2 = x2;
ASSERT3U(r1->rs_start, <=, r1->rs_end);
ASSERT3U(r2->rs_start, <=, r2->rs_end);
return ((r1->rs_start >= r2->rs_end) - (r1->rs_end <= r2->rs_start));
}
static int
range_tree_seg64_compare(const void *x1, const void *x2)
{
const range_seg64_t *r1 = x1;
const range_seg64_t *r2 = x2;
ASSERT3U(r1->rs_start, <=, r1->rs_end);
ASSERT3U(r2->rs_start, <=, r2->rs_end);
return ((r1->rs_start >= r2->rs_end) - (r1->rs_end <= r2->rs_start));
}
static int
range_tree_seg_gap_compare(const void *x1, const void *x2)
{
const range_seg_gap_t *r1 = x1;
const range_seg_gap_t *r2 = x2;
ASSERT3U(r1->rs_start, <=, r1->rs_end);
ASSERT3U(r2->rs_start, <=, r2->rs_end);
return ((r1->rs_start >= r2->rs_end) - (r1->rs_end <= r2->rs_start));
}
range_tree_t *
-range_tree_create_impl(range_tree_ops_t *ops, range_seg_type_t type, void *arg,
- uint64_t start, uint64_t shift,
+range_tree_create_impl(const range_tree_ops_t *ops, range_seg_type_t type,
+ void *arg, uint64_t start, uint64_t shift,
int (*zfs_btree_compare) (const void *, const void *),
uint64_t gap)
{
range_tree_t *rt = kmem_zalloc(sizeof (range_tree_t), KM_SLEEP);
ASSERT3U(shift, <, 64);
ASSERT3U(type, <=, RANGE_SEG_NUM_TYPES);
size_t size;
int (*compare) (const void *, const void *);
switch (type) {
case RANGE_SEG32:
size = sizeof (range_seg32_t);
compare = range_tree_seg32_compare;
break;
case RANGE_SEG64:
size = sizeof (range_seg64_t);
compare = range_tree_seg64_compare;
break;
case RANGE_SEG_GAP:
size = sizeof (range_seg_gap_t);
compare = range_tree_seg_gap_compare;
break;
default:
panic("Invalid range seg type %d", type);
}
zfs_btree_create(&rt->rt_root, compare, size);
rt->rt_ops = ops;
rt->rt_gap = gap;
rt->rt_arg = arg;
rt->rt_type = type;
rt->rt_start = start;
rt->rt_shift = shift;
rt->rt_btree_compare = zfs_btree_compare;
if (rt->rt_ops != NULL && rt->rt_ops->rtop_create != NULL)
rt->rt_ops->rtop_create(rt, rt->rt_arg);
return (rt);
}
range_tree_t *
-range_tree_create(range_tree_ops_t *ops, range_seg_type_t type,
+range_tree_create(const range_tree_ops_t *ops, range_seg_type_t type,
void *arg, uint64_t start, uint64_t shift)
{
return (range_tree_create_impl(ops, type, arg, start, shift, NULL, 0));
}
void
range_tree_destroy(range_tree_t *rt)
{
VERIFY0(rt->rt_space);
if (rt->rt_ops != NULL && rt->rt_ops->rtop_destroy != NULL)
rt->rt_ops->rtop_destroy(rt, rt->rt_arg);
zfs_btree_destroy(&rt->rt_root);
kmem_free(rt, sizeof (*rt));
}
void
range_tree_adjust_fill(range_tree_t *rt, range_seg_t *rs, int64_t delta)
{
if (delta < 0 && delta * -1 >= rs_get_fill(rs, rt)) {
zfs_panic_recover("zfs: attempting to decrease fill to or "
"below 0; probable double remove in segment [%llx:%llx]",
(longlong_t)rs_get_start(rs, rt),
(longlong_t)rs_get_end(rs, rt));
}
if (rs_get_fill(rs, rt) + delta > rs_get_end(rs, rt) -
rs_get_start(rs, rt)) {
zfs_panic_recover("zfs: attempting to increase fill beyond "
"max; probable double add in segment [%llx:%llx]",
(longlong_t)rs_get_start(rs, rt),
(longlong_t)rs_get_end(rs, rt));
}
if (rt->rt_ops != NULL && rt->rt_ops->rtop_remove != NULL)
rt->rt_ops->rtop_remove(rt, rs, rt->rt_arg);
rs_set_fill(rs, rt, rs_get_fill(rs, rt) + delta);
if (rt->rt_ops != NULL && rt->rt_ops->rtop_add != NULL)
rt->rt_ops->rtop_add(rt, rs, rt->rt_arg);
}
static void
range_tree_add_impl(void *arg, uint64_t start, uint64_t size, uint64_t fill)
{
range_tree_t *rt = arg;
zfs_btree_index_t where;
range_seg_t *rs_before, *rs_after, *rs;
range_seg_max_t tmp, rsearch;
uint64_t end = start + size, gap = rt->rt_gap;
uint64_t bridge_size = 0;
boolean_t merge_before, merge_after;
ASSERT3U(size, !=, 0);
ASSERT3U(fill, <=, size);
ASSERT3U(start + size, >, start);
rs_set_start(&rsearch, rt, start);
rs_set_end(&rsearch, rt, end);
rs = zfs_btree_find(&rt->rt_root, &rsearch, &where);
/*
* If this is a gap-supporting range tree, it is possible that we
* are inserting into an existing segment. In this case simply
* bump the fill count and call the remove / add callbacks. If the
* new range will extend an existing segment, we remove the
* existing one, apply the new extent to it and re-insert it using
* the normal code paths.
*/
if (rs != NULL) {
if (gap == 0) {
zfs_panic_recover("zfs: adding existent segment to "
"range tree (offset=%llx size=%llx)",
(longlong_t)start, (longlong_t)size);
return;
}
uint64_t rstart = rs_get_start(rs, rt);
uint64_t rend = rs_get_end(rs, rt);
if (rstart <= start && rend >= end) {
range_tree_adjust_fill(rt, rs, fill);
return;
}
if (rt->rt_ops != NULL && rt->rt_ops->rtop_remove != NULL)
rt->rt_ops->rtop_remove(rt, rs, rt->rt_arg);
range_tree_stat_decr(rt, rs);
rt->rt_space -= rend - rstart;
fill += rs_get_fill(rs, rt);
start = MIN(start, rstart);
end = MAX(end, rend);
size = end - start;
zfs_btree_remove(&rt->rt_root, rs);
range_tree_add_impl(rt, start, size, fill);
return;
}
ASSERT3P(rs, ==, NULL);
/*
* Determine whether or not we will have to merge with our neighbors.
* If gap != 0, we might need to merge with our neighbors even if we
* aren't directly touching.
*/
zfs_btree_index_t where_before, where_after;
rs_before = zfs_btree_prev(&rt->rt_root, &where, &where_before);
rs_after = zfs_btree_next(&rt->rt_root, &where, &where_after);
merge_before = (rs_before != NULL && rs_get_end(rs_before, rt) >=
start - gap);
merge_after = (rs_after != NULL && rs_get_start(rs_after, rt) <= end +
gap);
if (merge_before && gap != 0)
bridge_size += start - rs_get_end(rs_before, rt);
if (merge_after && gap != 0)
bridge_size += rs_get_start(rs_after, rt) - end;
if (merge_before && merge_after) {
if (rt->rt_ops != NULL && rt->rt_ops->rtop_remove != NULL) {
rt->rt_ops->rtop_remove(rt, rs_before, rt->rt_arg);
rt->rt_ops->rtop_remove(rt, rs_after, rt->rt_arg);
}
range_tree_stat_decr(rt, rs_before);
range_tree_stat_decr(rt, rs_after);
rs_copy(rs_after, &tmp, rt);
uint64_t before_start = rs_get_start_raw(rs_before, rt);
uint64_t before_fill = rs_get_fill(rs_before, rt);
uint64_t after_fill = rs_get_fill(rs_after, rt);
zfs_btree_remove_idx(&rt->rt_root, &where_before);
/*
* We have to re-find the node because our old reference is
* invalid as soon as we do any mutating btree operations.
*/
rs_after = zfs_btree_find(&rt->rt_root, &tmp, &where_after);
rs_set_start_raw(rs_after, rt, before_start);
rs_set_fill(rs_after, rt, after_fill + before_fill + fill);
rs = rs_after;
} else if (merge_before) {
if (rt->rt_ops != NULL && rt->rt_ops->rtop_remove != NULL)
rt->rt_ops->rtop_remove(rt, rs_before, rt->rt_arg);
range_tree_stat_decr(rt, rs_before);
uint64_t before_fill = rs_get_fill(rs_before, rt);
rs_set_end(rs_before, rt, end);
rs_set_fill(rs_before, rt, before_fill + fill);
rs = rs_before;
} else if (merge_after) {
if (rt->rt_ops != NULL && rt->rt_ops->rtop_remove != NULL)
rt->rt_ops->rtop_remove(rt, rs_after, rt->rt_arg);
range_tree_stat_decr(rt, rs_after);
uint64_t after_fill = rs_get_fill(rs_after, rt);
rs_set_start(rs_after, rt, start);
rs_set_fill(rs_after, rt, after_fill + fill);
rs = rs_after;
} else {
rs = &tmp;
rs_set_start(rs, rt, start);
rs_set_end(rs, rt, end);
rs_set_fill(rs, rt, fill);
zfs_btree_add_idx(&rt->rt_root, rs, &where);
}
if (gap != 0) {
ASSERT3U(rs_get_fill(rs, rt), <=, rs_get_end(rs, rt) -
rs_get_start(rs, rt));
} else {
ASSERT3U(rs_get_fill(rs, rt), ==, rs_get_end(rs, rt) -
rs_get_start(rs, rt));
}
if (rt->rt_ops != NULL && rt->rt_ops->rtop_add != NULL)
rt->rt_ops->rtop_add(rt, rs, rt->rt_arg);
range_tree_stat_incr(rt, rs);
rt->rt_space += size + bridge_size;
}
void
range_tree_add(void *arg, uint64_t start, uint64_t size)
{
range_tree_add_impl(arg, start, size, size);
}
static void
range_tree_remove_impl(range_tree_t *rt, uint64_t start, uint64_t size,
boolean_t do_fill)
{
zfs_btree_index_t where;
range_seg_t *rs;
range_seg_max_t rsearch, rs_tmp;
uint64_t end = start + size;
boolean_t left_over, right_over;
VERIFY3U(size, !=, 0);
VERIFY3U(size, <=, rt->rt_space);
if (rt->rt_type == RANGE_SEG64)
ASSERT3U(start + size, >, start);
rs_set_start(&rsearch, rt, start);
rs_set_end(&rsearch, rt, end);
rs = zfs_btree_find(&rt->rt_root, &rsearch, &where);
/* Make sure we completely overlap with someone */
if (rs == NULL) {
zfs_panic_recover("zfs: removing nonexistent segment from "
"range tree (offset=%llx size=%llx)",
(longlong_t)start, (longlong_t)size);
return;
}
/*
* Range trees with gap support must only remove complete segments
* from the tree. This allows us to maintain accurate fill accounting
* and to ensure that bridged sections are not leaked. If we need to
* remove less than the full segment, we can only adjust the fill count.
*/
if (rt->rt_gap != 0) {
if (do_fill) {
if (rs_get_fill(rs, rt) == size) {
start = rs_get_start(rs, rt);
end = rs_get_end(rs, rt);
size = end - start;
} else {
range_tree_adjust_fill(rt, rs, -size);
return;
}
} else if (rs_get_start(rs, rt) != start ||
rs_get_end(rs, rt) != end) {
zfs_panic_recover("zfs: freeing partial segment of "
"gap tree (offset=%llx size=%llx) of "
"(offset=%llx size=%llx)",
(longlong_t)start, (longlong_t)size,
(longlong_t)rs_get_start(rs, rt),
(longlong_t)rs_get_end(rs, rt) - rs_get_start(rs,
rt));
return;
}
}
VERIFY3U(rs_get_start(rs, rt), <=, start);
VERIFY3U(rs_get_end(rs, rt), >=, end);
left_over = (rs_get_start(rs, rt) != start);
right_over = (rs_get_end(rs, rt) != end);
range_tree_stat_decr(rt, rs);
if (rt->rt_ops != NULL && rt->rt_ops->rtop_remove != NULL)
rt->rt_ops->rtop_remove(rt, rs, rt->rt_arg);
if (left_over && right_over) {
range_seg_max_t newseg;
rs_set_start(&newseg, rt, end);
rs_set_end_raw(&newseg, rt, rs_get_end_raw(rs, rt));
rs_set_fill(&newseg, rt, rs_get_end(rs, rt) - end);
range_tree_stat_incr(rt, &newseg);
// This modifies the buffer already inside the range tree
rs_set_end(rs, rt, start);
rs_copy(rs, &rs_tmp, rt);
if (zfs_btree_next(&rt->rt_root, &where, &where) != NULL)
zfs_btree_add_idx(&rt->rt_root, &newseg, &where);
else
zfs_btree_add(&rt->rt_root, &newseg);
if (rt->rt_ops != NULL && rt->rt_ops->rtop_add != NULL)
rt->rt_ops->rtop_add(rt, &newseg, rt->rt_arg);
} else if (left_over) {
// This modifies the buffer already inside the range tree
rs_set_end(rs, rt, start);
rs_copy(rs, &rs_tmp, rt);
} else if (right_over) {
// This modifies the buffer already inside the range tree
rs_set_start(rs, rt, end);
rs_copy(rs, &rs_tmp, rt);
} else {
zfs_btree_remove_idx(&rt->rt_root, &where);
rs = NULL;
}
if (rs != NULL) {
/*
* The fill of the leftover segment will always be equal to
* the size, since we do not support removing partial segments
* of range trees with gaps.
*/
rs_set_fill_raw(rs, rt, rs_get_end_raw(rs, rt) -
rs_get_start_raw(rs, rt));
range_tree_stat_incr(rt, &rs_tmp);
if (rt->rt_ops != NULL && rt->rt_ops->rtop_add != NULL)
rt->rt_ops->rtop_add(rt, &rs_tmp, rt->rt_arg);
}
rt->rt_space -= size;
}
void
range_tree_remove(void *arg, uint64_t start, uint64_t size)
{
range_tree_remove_impl(arg, start, size, B_FALSE);
}
void
range_tree_remove_fill(range_tree_t *rt, uint64_t start, uint64_t size)
{
range_tree_remove_impl(rt, start, size, B_TRUE);
}
void
range_tree_resize_segment(range_tree_t *rt, range_seg_t *rs,
uint64_t newstart, uint64_t newsize)
{
int64_t delta = newsize - (rs_get_end(rs, rt) - rs_get_start(rs, rt));
range_tree_stat_decr(rt, rs);
if (rt->rt_ops != NULL && rt->rt_ops->rtop_remove != NULL)
rt->rt_ops->rtop_remove(rt, rs, rt->rt_arg);
rs_set_start(rs, rt, newstart);
rs_set_end(rs, rt, newstart + newsize);
range_tree_stat_incr(rt, rs);
if (rt->rt_ops != NULL && rt->rt_ops->rtop_add != NULL)
rt->rt_ops->rtop_add(rt, rs, rt->rt_arg);
rt->rt_space += delta;
}
static range_seg_t *
range_tree_find_impl(range_tree_t *rt, uint64_t start, uint64_t size)
{
range_seg_max_t rsearch;
uint64_t end = start + size;
VERIFY(size != 0);
rs_set_start(&rsearch, rt, start);
rs_set_end(&rsearch, rt, end);
return (zfs_btree_find(&rt->rt_root, &rsearch, NULL));
}
range_seg_t *
range_tree_find(range_tree_t *rt, uint64_t start, uint64_t size)
{
if (rt->rt_type == RANGE_SEG64)
ASSERT3U(start + size, >, start);
range_seg_t *rs = range_tree_find_impl(rt, start, size);
if (rs != NULL && rs_get_start(rs, rt) <= start &&
rs_get_end(rs, rt) >= start + size) {
return (rs);
}
return (NULL);
}
void
range_tree_verify_not_present(range_tree_t *rt, uint64_t off, uint64_t size)
{
range_seg_t *rs = range_tree_find(rt, off, size);
if (rs != NULL)
panic("segment already in tree; rs=%p", (void *)rs);
}
boolean_t
range_tree_contains(range_tree_t *rt, uint64_t start, uint64_t size)
{
return (range_tree_find(rt, start, size) != NULL);
}
/*
* Returns the first subset of the given range which overlaps with the range
* tree. Returns true if there is a segment in the range, and false if there
* isn't.
*/
boolean_t
range_tree_find_in(range_tree_t *rt, uint64_t start, uint64_t size,
uint64_t *ostart, uint64_t *osize)
{
if (rt->rt_type == RANGE_SEG64)
ASSERT3U(start + size, >, start);
range_seg_max_t rsearch;
rs_set_start(&rsearch, rt, start);
rs_set_end_raw(&rsearch, rt, rs_get_start_raw(&rsearch, rt) + 1);
zfs_btree_index_t where;
range_seg_t *rs = zfs_btree_find(&rt->rt_root, &rsearch, &where);
if (rs != NULL) {
*ostart = start;
*osize = MIN(size, rs_get_end(rs, rt) - start);
return (B_TRUE);
}
rs = zfs_btree_next(&rt->rt_root, &where, &where);
if (rs == NULL || rs_get_start(rs, rt) > start + size)
return (B_FALSE);
*ostart = rs_get_start(rs, rt);
*osize = MIN(start + size, rs_get_end(rs, rt)) -
rs_get_start(rs, rt);
return (B_TRUE);
}
/*
* Ensure that this range is not in the tree, regardless of whether
* it is currently in the tree.
*/
void
range_tree_clear(range_tree_t *rt, uint64_t start, uint64_t size)
{
range_seg_t *rs;
if (size == 0)
return;
if (rt->rt_type == RANGE_SEG64)
ASSERT3U(start + size, >, start);
while ((rs = range_tree_find_impl(rt, start, size)) != NULL) {
uint64_t free_start = MAX(rs_get_start(rs, rt), start);
uint64_t free_end = MIN(rs_get_end(rs, rt), start + size);
range_tree_remove(rt, free_start, free_end - free_start);
}
}
void
range_tree_swap(range_tree_t **rtsrc, range_tree_t **rtdst)
{
range_tree_t *rt;
ASSERT0(range_tree_space(*rtdst));
ASSERT0(zfs_btree_numnodes(&(*rtdst)->rt_root));
rt = *rtsrc;
*rtsrc = *rtdst;
*rtdst = rt;
}
void
range_tree_vacate(range_tree_t *rt, range_tree_func_t *func, void *arg)
{
if (rt->rt_ops != NULL && rt->rt_ops->rtop_vacate != NULL)
rt->rt_ops->rtop_vacate(rt, rt->rt_arg);
if (func != NULL) {
range_seg_t *rs;
zfs_btree_index_t *cookie = NULL;
while ((rs = zfs_btree_destroy_nodes(&rt->rt_root, &cookie)) !=
NULL) {
func(arg, rs_get_start(rs, rt), rs_get_end(rs, rt) -
rs_get_start(rs, rt));
}
} else {
zfs_btree_clear(&rt->rt_root);
}
bzero(rt->rt_histogram, sizeof (rt->rt_histogram));
rt->rt_space = 0;
}
void
range_tree_walk(range_tree_t *rt, range_tree_func_t *func, void *arg)
{
zfs_btree_index_t where;
for (range_seg_t *rs = zfs_btree_first(&rt->rt_root, &where);
rs != NULL; rs = zfs_btree_next(&rt->rt_root, &where, &where)) {
func(arg, rs_get_start(rs, rt), rs_get_end(rs, rt) -
rs_get_start(rs, rt));
}
}
range_seg_t *
range_tree_first(range_tree_t *rt)
{
return (zfs_btree_first(&rt->rt_root, NULL));
}
uint64_t
range_tree_space(range_tree_t *rt)
{
return (rt->rt_space);
}
uint64_t
range_tree_numsegs(range_tree_t *rt)
{
return ((rt == NULL) ? 0 : zfs_btree_numnodes(&rt->rt_root));
}
boolean_t
range_tree_is_empty(range_tree_t *rt)
{
ASSERT(rt != NULL);
return (range_tree_space(rt) == 0);
}
void
rt_btree_create(range_tree_t *rt, void *arg)
{
zfs_btree_t *size_tree = arg;
size_t size;
switch (rt->rt_type) {
case RANGE_SEG32:
size = sizeof (range_seg32_t);
break;
case RANGE_SEG64:
size = sizeof (range_seg64_t);
break;
case RANGE_SEG_GAP:
size = sizeof (range_seg_gap_t);
break;
default:
panic("Invalid range seg type %d", rt->rt_type);
}
zfs_btree_create(size_tree, rt->rt_btree_compare, size);
}
void
rt_btree_destroy(range_tree_t *rt, void *arg)
{
(void) rt;
zfs_btree_t *size_tree = arg;
ASSERT0(zfs_btree_numnodes(size_tree));
zfs_btree_destroy(size_tree);
}
void
rt_btree_add(range_tree_t *rt, range_seg_t *rs, void *arg)
{
(void) rt;
zfs_btree_t *size_tree = arg;
zfs_btree_add(size_tree, rs);
}
void
rt_btree_remove(range_tree_t *rt, range_seg_t *rs, void *arg)
{
(void) rt;
zfs_btree_t *size_tree = arg;
zfs_btree_remove(size_tree, rs);
}
void
rt_btree_vacate(range_tree_t *rt, void *arg)
{
zfs_btree_t *size_tree = arg;
zfs_btree_clear(size_tree);
zfs_btree_destroy(size_tree);
rt_btree_create(rt, arg);
}
-range_tree_ops_t rt_btree_ops = {
+const range_tree_ops_t rt_btree_ops = {
.rtop_create = rt_btree_create,
.rtop_destroy = rt_btree_destroy,
.rtop_add = rt_btree_add,
.rtop_remove = rt_btree_remove,
.rtop_vacate = rt_btree_vacate
};
/*
* Remove any overlapping ranges between the given segment [start, end)
* from removefrom. Add non-overlapping leftovers to addto.
*/
void
range_tree_remove_xor_add_segment(uint64_t start, uint64_t end,
range_tree_t *removefrom, range_tree_t *addto)
{
zfs_btree_index_t where;
range_seg_max_t starting_rs;
rs_set_start(&starting_rs, removefrom, start);
rs_set_end_raw(&starting_rs, removefrom, rs_get_start_raw(&starting_rs,
removefrom) + 1);
range_seg_t *curr = zfs_btree_find(&removefrom->rt_root,
&starting_rs, &where);
if (curr == NULL)
curr = zfs_btree_next(&removefrom->rt_root, &where, &where);
range_seg_t *next;
for (; curr != NULL; curr = next) {
if (start == end)
return;
VERIFY3U(start, <, end);
/* there is no overlap */
if (end <= rs_get_start(curr, removefrom)) {
range_tree_add(addto, start, end - start);
return;
}
uint64_t overlap_start = MAX(rs_get_start(curr, removefrom),
start);
uint64_t overlap_end = MIN(rs_get_end(curr, removefrom),
end);
uint64_t overlap_size = overlap_end - overlap_start;
ASSERT3S(overlap_size, >, 0);
range_seg_max_t rs;
rs_copy(curr, &rs, removefrom);
range_tree_remove(removefrom, overlap_start, overlap_size);
if (start < overlap_start)
range_tree_add(addto, start, overlap_start - start);
start = overlap_end;
next = zfs_btree_find(&removefrom->rt_root, &rs, &where);
/*
* If we find something here, we only removed part of the
* curr segment. Either there's some left at the end
* because we've reached the end of the range we're removing,
* or there's some left at the start because we started
* partway through the range. Either way, we continue with
* the loop. If it's the former, we'll return at the start of
* the loop, and if it's the latter we'll see if there is more
* area to process.
*/
if (next != NULL) {
ASSERT(start == end || start == rs_get_end(&rs,
removefrom));
}
next = zfs_btree_next(&removefrom->rt_root, &where, &where);
}
VERIFY3P(curr, ==, NULL);
if (start != end) {
VERIFY3U(start, <, end);
range_tree_add(addto, start, end - start);
} else {
VERIFY3U(start, ==, end);
}
}
/*
* For each entry in rt, if it exists in removefrom, remove it
* from removefrom. Otherwise, add it to addto.
*/
void
range_tree_remove_xor_add(range_tree_t *rt, range_tree_t *removefrom,
range_tree_t *addto)
{
zfs_btree_index_t where;
for (range_seg_t *rs = zfs_btree_first(&rt->rt_root, &where); rs;
rs = zfs_btree_next(&rt->rt_root, &where, &where)) {
range_tree_remove_xor_add_segment(rs_get_start(rs, rt),
rs_get_end(rs, rt), removefrom, addto);
}
}
uint64_t
range_tree_min(range_tree_t *rt)
{
range_seg_t *rs = zfs_btree_first(&rt->rt_root, NULL);
return (rs != NULL ? rs_get_start(rs, rt) : 0);
}
uint64_t
range_tree_max(range_tree_t *rt)
{
range_seg_t *rs = zfs_btree_last(&rt->rt_root, NULL);
return (rs != NULL ? rs_get_end(rs, rt) : 0);
}
uint64_t
range_tree_span(range_tree_t *rt)
{
return (range_tree_max(rt) - range_tree_min(rt));
}
diff --git a/module/zfs/refcount.c b/module/zfs/refcount.c
index 35a379dded69..390b2fdedc70 100644
--- a/module/zfs/refcount.c
+++ b/module/zfs/refcount.c
@@ -1,336 +1,336 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2021 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/zfs_refcount.h>
+#ifdef ZFS_DEBUG
/*
* Reference count tracking is disabled by default. It's memory requirements
* are reasonable, however as implemented it consumes a significant amount of
* cpu time. Until its performance is improved it should be manually enabled.
*/
-int reference_tracking_enable = FALSE;
-int reference_history = 3; /* tunable */
+int reference_tracking_enable = B_FALSE;
+static int reference_history = 3; /* tunable */
-#ifdef ZFS_DEBUG
static kmem_cache_t *reference_cache;
static kmem_cache_t *reference_history_cache;
void
zfs_refcount_init(void)
{
reference_cache = kmem_cache_create("reference_cache",
sizeof (reference_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
reference_history_cache = kmem_cache_create("reference_history_cache",
sizeof (uint64_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
}
void
zfs_refcount_fini(void)
{
kmem_cache_destroy(reference_cache);
kmem_cache_destroy(reference_history_cache);
}
void
zfs_refcount_create(zfs_refcount_t *rc)
{
mutex_init(&rc->rc_mtx, NULL, MUTEX_DEFAULT, NULL);
list_create(&rc->rc_list, sizeof (reference_t),
offsetof(reference_t, ref_link));
list_create(&rc->rc_removed, sizeof (reference_t),
offsetof(reference_t, ref_link));
rc->rc_count = 0;
rc->rc_removed_count = 0;
rc->rc_tracked = reference_tracking_enable;
}
void
zfs_refcount_create_tracked(zfs_refcount_t *rc)
{
zfs_refcount_create(rc);
rc->rc_tracked = B_TRUE;
}
void
zfs_refcount_create_untracked(zfs_refcount_t *rc)
{
zfs_refcount_create(rc);
rc->rc_tracked = B_FALSE;
}
void
zfs_refcount_destroy_many(zfs_refcount_t *rc, uint64_t number)
{
reference_t *ref;
ASSERT3U(rc->rc_count, ==, number);
while ((ref = list_head(&rc->rc_list))) {
list_remove(&rc->rc_list, ref);
kmem_cache_free(reference_cache, ref);
}
list_destroy(&rc->rc_list);
while ((ref = list_head(&rc->rc_removed))) {
list_remove(&rc->rc_removed, ref);
kmem_cache_free(reference_history_cache, ref->ref_removed);
kmem_cache_free(reference_cache, ref);
}
list_destroy(&rc->rc_removed);
mutex_destroy(&rc->rc_mtx);
}
void
zfs_refcount_destroy(zfs_refcount_t *rc)
{
zfs_refcount_destroy_many(rc, 0);
}
int
zfs_refcount_is_zero(zfs_refcount_t *rc)
{
return (zfs_refcount_count(rc) == 0);
}
int64_t
zfs_refcount_count(zfs_refcount_t *rc)
{
return (atomic_load_64(&rc->rc_count));
}
int64_t
zfs_refcount_add_many(zfs_refcount_t *rc, uint64_t number, const void *holder)
{
reference_t *ref = NULL;
int64_t count;
if (!rc->rc_tracked) {
count = atomic_add_64_nv(&(rc)->rc_count, number);
ASSERT3U(count, >=, number);
return (count);
}
ref = kmem_cache_alloc(reference_cache, KM_SLEEP);
ref->ref_holder = holder;
ref->ref_number = number;
mutex_enter(&rc->rc_mtx);
ASSERT3U(rc->rc_count, >=, 0);
list_insert_head(&rc->rc_list, ref);
rc->rc_count += number;
count = rc->rc_count;
mutex_exit(&rc->rc_mtx);
return (count);
}
int64_t
zfs_refcount_add(zfs_refcount_t *rc, const void *holder)
{
return (zfs_refcount_add_many(rc, 1, holder));
}
int64_t
zfs_refcount_remove_many(zfs_refcount_t *rc, uint64_t number,
const void *holder)
{
reference_t *ref;
int64_t count;
if (!rc->rc_tracked) {
count = atomic_add_64_nv(&(rc)->rc_count, -number);
ASSERT3S(count, >=, 0);
return (count);
}
mutex_enter(&rc->rc_mtx);
ASSERT3U(rc->rc_count, >=, number);
for (ref = list_head(&rc->rc_list); ref;
ref = list_next(&rc->rc_list, ref)) {
if (ref->ref_holder == holder && ref->ref_number == number) {
list_remove(&rc->rc_list, ref);
if (reference_history > 0) {
ref->ref_removed =
kmem_cache_alloc(reference_history_cache,
KM_SLEEP);
list_insert_head(&rc->rc_removed, ref);
rc->rc_removed_count++;
if (rc->rc_removed_count > reference_history) {
ref = list_tail(&rc->rc_removed);
list_remove(&rc->rc_removed, ref);
kmem_cache_free(reference_history_cache,
ref->ref_removed);
kmem_cache_free(reference_cache, ref);
rc->rc_removed_count--;
}
} else {
kmem_cache_free(reference_cache, ref);
}
rc->rc_count -= number;
count = rc->rc_count;
mutex_exit(&rc->rc_mtx);
return (count);
}
}
panic("No such hold %p on refcount %llx", holder,
(u_longlong_t)(uintptr_t)rc);
return (-1);
}
int64_t
zfs_refcount_remove(zfs_refcount_t *rc, const void *holder)
{
return (zfs_refcount_remove_many(rc, 1, holder));
}
void
zfs_refcount_transfer(zfs_refcount_t *dst, zfs_refcount_t *src)
{
int64_t count, removed_count;
list_t list, removed;
list_create(&list, sizeof (reference_t),
offsetof(reference_t, ref_link));
list_create(&removed, sizeof (reference_t),
offsetof(reference_t, ref_link));
mutex_enter(&src->rc_mtx);
count = src->rc_count;
removed_count = src->rc_removed_count;
src->rc_count = 0;
src->rc_removed_count = 0;
list_move_tail(&list, &src->rc_list);
list_move_tail(&removed, &src->rc_removed);
mutex_exit(&src->rc_mtx);
mutex_enter(&dst->rc_mtx);
dst->rc_count += count;
dst->rc_removed_count += removed_count;
list_move_tail(&dst->rc_list, &list);
list_move_tail(&dst->rc_removed, &removed);
mutex_exit(&dst->rc_mtx);
list_destroy(&list);
list_destroy(&removed);
}
void
zfs_refcount_transfer_ownership_many(zfs_refcount_t *rc, uint64_t number,
const void *current_holder, const void *new_holder)
{
reference_t *ref;
boolean_t found = B_FALSE;
if (!rc->rc_tracked)
return;
mutex_enter(&rc->rc_mtx);
for (ref = list_head(&rc->rc_list); ref;
ref = list_next(&rc->rc_list, ref)) {
if (ref->ref_holder == current_holder &&
ref->ref_number == number) {
ref->ref_holder = new_holder;
found = B_TRUE;
break;
}
}
ASSERT(found);
mutex_exit(&rc->rc_mtx);
}
void
zfs_refcount_transfer_ownership(zfs_refcount_t *rc, const void *current_holder,
const void *new_holder)
{
return (zfs_refcount_transfer_ownership_many(rc, 1, current_holder,
new_holder));
}
/*
* If tracking is enabled, return true if a reference exists that matches
* the "holder" tag. If tracking is disabled, then return true if a reference
* might be held.
*/
boolean_t
zfs_refcount_held(zfs_refcount_t *rc, const void *holder)
{
reference_t *ref;
if (!rc->rc_tracked)
return (zfs_refcount_count(rc) > 0);
mutex_enter(&rc->rc_mtx);
for (ref = list_head(&rc->rc_list); ref;
ref = list_next(&rc->rc_list, ref)) {
if (ref->ref_holder == holder) {
mutex_exit(&rc->rc_mtx);
return (B_TRUE);
}
}
mutex_exit(&rc->rc_mtx);
return (B_FALSE);
}
/*
* If tracking is enabled, return true if a reference does not exist that
* matches the "holder" tag. If tracking is disabled, always return true
* since the reference might not be held.
*/
boolean_t
zfs_refcount_not_held(zfs_refcount_t *rc, const void *holder)
{
reference_t *ref;
if (!rc->rc_tracked)
return (B_TRUE);
mutex_enter(&rc->rc_mtx);
for (ref = list_head(&rc->rc_list); ref;
ref = list_next(&rc->rc_list, ref)) {
if (ref->ref_holder == holder) {
mutex_exit(&rc->rc_mtx);
return (B_FALSE);
}
}
mutex_exit(&rc->rc_mtx);
return (B_TRUE);
}
EXPORT_SYMBOL(zfs_refcount_create);
EXPORT_SYMBOL(zfs_refcount_destroy);
EXPORT_SYMBOL(zfs_refcount_is_zero);
EXPORT_SYMBOL(zfs_refcount_count);
EXPORT_SYMBOL(zfs_refcount_add);
EXPORT_SYMBOL(zfs_refcount_remove);
EXPORT_SYMBOL(zfs_refcount_held);
/* BEGIN CSTYLED */
-ZFS_MODULE_PARAM(zfs, ,reference_tracking_enable, INT, ZMOD_RW,
+ZFS_MODULE_PARAM(zfs, , reference_tracking_enable, INT, ZMOD_RW,
"Track reference holders to refcount_t objects");
-ZFS_MODULE_PARAM(zfs, ,reference_history, INT, ZMOD_RW,
+ZFS_MODULE_PARAM(zfs, , reference_history, INT, ZMOD_RW,
"Maximum reference holders being tracked");
/* END CSTYLED */
#endif /* ZFS_DEBUG */
diff --git a/module/zfs/sa.c b/module/zfs/sa.c
index b69b0c68f13c..a078af159c1f 100644
--- a/module/zfs/sa.c
+++ b/module/zfs/sa.c
@@ -1,2257 +1,2256 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013, 2017 by Delphix. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/types.h>
#include <sys/param.h>
#include <sys/sysmacros.h>
#include <sys/dmu.h>
#include <sys/dmu_impl.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_tx.h>
#include <sys/dbuf.h>
#include <sys/dnode.h>
#include <sys/zap.h>
#include <sys/sa.h>
#include <sys/sunddi.h>
#include <sys/sa_impl.h>
#include <sys/errno.h>
#include <sys/zfs_context.h>
#ifdef _KERNEL
#include <sys/zfs_znode.h>
#endif
/*
* ZFS System attributes:
*
* A generic mechanism to allow for arbitrary attributes
* to be stored in a dnode. The data will be stored in the bonus buffer of
* the dnode and if necessary a special "spill" block will be used to handle
* overflow situations. The spill block will be sized to fit the data
* from 512 - 128K. When a spill block is used the BP (blkptr_t) for the
* spill block is stored at the end of the current bonus buffer. Any
* attributes that would be in the way of the blkptr_t will be relocated
* into the spill block.
*
* Attribute registration:
*
* Stored persistently on a per dataset basis
* a mapping between attribute "string" names and their actual attribute
* numeric values, length, and byteswap function. The names are only used
* during registration. All attributes are known by their unique attribute
* id value. If an attribute can have a variable size then the value
* 0 will be used to indicate this.
*
* Attribute Layout:
*
* Attribute layouts are a way to compactly store multiple attributes, but
* without taking the overhead associated with managing each attribute
* individually. Since you will typically have the same set of attributes
* stored in the same order a single table will be used to represent that
* layout. The ZPL for example will usually have only about 10 different
* layouts (regular files, device files, symlinks,
* regular files + scanstamp, files/dir with extended attributes, and then
* you have the possibility of all of those minus ACL, because it would
* be kicked out into the spill block)
*
* Layouts are simply an array of the attributes and their
* ordering i.e. [0, 1, 4, 5, 2]
*
* Each distinct layout is given a unique layout number and that is what's
* stored in the header at the beginning of the SA data buffer.
*
* A layout only covers a single dbuf (bonus or spill). If a set of
* attributes is split up between the bonus buffer and a spill buffer then
* two different layouts will be used. This allows us to byteswap the
* spill without looking at the bonus buffer and keeps the on disk format of
* the bonus and spill buffer the same.
*
* Adding a single attribute will cause the entire set of attributes to
* be rewritten and could result in a new layout number being constructed
* as part of the rewrite if no such layout exists for the new set of
* attributes. The new attribute will be appended to the end of the already
* existing attributes.
*
* Both the attribute registration and attribute layout information are
* stored in normal ZAP attributes. Their should be a small number of
* known layouts and the set of attributes is assumed to typically be quite
* small.
*
* The registered attributes and layout "table" information is maintained
* in core and a special "sa_os_t" is attached to the objset_t.
*
* A special interface is provided to allow for quickly applying
* a large set of attributes at once. sa_replace_all_by_template() is
* used to set an array of attributes. This is used by the ZPL when
* creating a brand new file. The template that is passed into the function
* specifies the attribute, size for variable length attributes, location of
* data and special "data locator" function if the data isn't in a contiguous
* location.
*
* Byteswap implications:
*
* Since the SA attributes are not entirely self describing we can't do
* the normal byteswap processing. The special ZAP layout attribute and
* attribute registration attributes define the byteswap function and the
* size of the attributes, unless it is variable sized.
* The normal ZFS byteswapping infrastructure assumes you don't need
* to read any objects in order to do the necessary byteswapping. Whereas
* SA attributes can only be properly byteswapped if the dataset is opened
* and the layout/attribute ZAP attributes are available. Because of this
* the SA attributes will be byteswapped when they are first accessed by
* the SA code that will read the SA data.
*/
typedef void (sa_iterfunc_t)(void *hdr, void *addr, sa_attr_type_t,
uint16_t length, int length_idx, boolean_t, void *userp);
static int sa_build_index(sa_handle_t *hdl, sa_buf_type_t buftype);
static void sa_idx_tab_hold(objset_t *os, sa_idx_tab_t *idx_tab);
static sa_idx_tab_t *sa_find_idx_tab(objset_t *os, dmu_object_type_t bonustype,
sa_hdr_phys_t *hdr);
static void sa_idx_tab_rele(objset_t *os, void *arg);
static void sa_copy_data(sa_data_locator_t *func, void *start, void *target,
int buflen);
static int sa_modify_attrs(sa_handle_t *hdl, sa_attr_type_t newattr,
sa_data_op_t action, sa_data_locator_t *locator, void *datastart,
uint16_t buflen, dmu_tx_t *tx);
-arc_byteswap_func_t sa_bswap_table[] = {
+static const arc_byteswap_func_t sa_bswap_table[] = {
byteswap_uint64_array,
byteswap_uint32_array,
byteswap_uint16_array,
byteswap_uint8_array,
zfs_acl_byteswap,
};
#ifdef HAVE_EFFICIENT_UNALIGNED_ACCESS
#define SA_COPY_DATA(f, s, t, l) \
do { \
if (f == NULL) { \
if (l == 8) { \
*(uint64_t *)t = *(uint64_t *)s; \
} else if (l == 16) { \
*(uint64_t *)t = *(uint64_t *)s; \
*(uint64_t *)((uintptr_t)t + 8) = \
*(uint64_t *)((uintptr_t)s + 8); \
} else { \
bcopy(s, t, l); \
} \
} else { \
sa_copy_data(f, s, t, l); \
} \
} while (0)
#else
#define SA_COPY_DATA(f, s, t, l) sa_copy_data(f, s, t, l)
#endif
/*
* This table is fixed and cannot be changed. Its purpose is to
* allow the SA code to work with both old/new ZPL file systems.
* It contains the list of legacy attributes. These attributes aren't
* stored in the "attribute" registry zap objects, since older ZPL file systems
* won't have the registry. Only objsets of type ZFS_TYPE_FILESYSTEM will
* use this static table.
*/
-sa_attr_reg_t sa_legacy_attrs[] = {
+static const sa_attr_reg_t sa_legacy_attrs[] = {
{"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
{"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
{"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
{"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
{"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
{"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
{"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
{"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
{"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
{"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
{"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
{"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
{"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
{"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
{"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
{"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
};
/*
* This is only used for objects of type DMU_OT_ZNODE
*/
-sa_attr_type_t sa_legacy_zpl_layout[] = {
+static const sa_attr_type_t sa_legacy_zpl_layout[] = {
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
};
/*
* Special dummy layout used for buffers with no attributes.
*/
-sa_attr_type_t sa_dummy_zpl_layout[] = { 0 };
+static const sa_attr_type_t sa_dummy_zpl_layout[] = { 0 };
-static int sa_legacy_attr_count = ARRAY_SIZE(sa_legacy_attrs);
+static const size_t sa_legacy_attr_count = ARRAY_SIZE(sa_legacy_attrs);
static kmem_cache_t *sa_cache = NULL;
static int
sa_cache_constructor(void *buf, void *unused, int kmflag)
{
(void) unused, (void) kmflag;
sa_handle_t *hdl = buf;
mutex_init(&hdl->sa_lock, NULL, MUTEX_DEFAULT, NULL);
return (0);
}
static void
sa_cache_destructor(void *buf, void *unused)
{
(void) unused;
sa_handle_t *hdl = buf;
mutex_destroy(&hdl->sa_lock);
}
void
sa_cache_init(void)
{
sa_cache = kmem_cache_create("sa_cache",
sizeof (sa_handle_t), 0, sa_cache_constructor,
sa_cache_destructor, NULL, NULL, NULL, 0);
}
void
sa_cache_fini(void)
{
if (sa_cache)
kmem_cache_destroy(sa_cache);
}
static int
layout_num_compare(const void *arg1, const void *arg2)
{
const sa_lot_t *node1 = (const sa_lot_t *)arg1;
const sa_lot_t *node2 = (const sa_lot_t *)arg2;
return (TREE_CMP(node1->lot_num, node2->lot_num));
}
static int
layout_hash_compare(const void *arg1, const void *arg2)
{
const sa_lot_t *node1 = (const sa_lot_t *)arg1;
const sa_lot_t *node2 = (const sa_lot_t *)arg2;
int cmp = TREE_CMP(node1->lot_hash, node2->lot_hash);
if (likely(cmp))
return (cmp);
return (TREE_CMP(node1->lot_instance, node2->lot_instance));
}
static boolean_t
sa_layout_equal(sa_lot_t *tbf, sa_attr_type_t *attrs, int count)
{
int i;
if (count != tbf->lot_attr_count)
return (1);
for (i = 0; i != count; i++) {
if (attrs[i] != tbf->lot_attrs[i])
return (1);
}
return (0);
}
#define SA_ATTR_HASH(attr) (zfs_crc64_table[(-1ULL ^ attr) & 0xFF])
static uint64_t
-sa_layout_info_hash(sa_attr_type_t *attrs, int attr_count)
+sa_layout_info_hash(const sa_attr_type_t *attrs, int attr_count)
{
- int i;
uint64_t crc = -1ULL;
- for (i = 0; i != attr_count; i++)
+ for (int i = 0; i != attr_count; i++)
crc ^= SA_ATTR_HASH(attrs[i]);
return (crc);
}
static int
sa_get_spill(sa_handle_t *hdl)
{
int rc;
if (hdl->sa_spill == NULL) {
if ((rc = dmu_spill_hold_existing(hdl->sa_bonus, NULL,
&hdl->sa_spill)) == 0)
VERIFY(0 == sa_build_index(hdl, SA_SPILL));
} else {
rc = 0;
}
return (rc);
}
/*
* Main attribute lookup/update function
* returns 0 for success or non zero for failures
*
* Operates on bulk array, first failure will abort further processing
*/
static int
sa_attr_op(sa_handle_t *hdl, sa_bulk_attr_t *bulk, int count,
sa_data_op_t data_op, dmu_tx_t *tx)
{
sa_os_t *sa = hdl->sa_os->os_sa;
int i;
int error = 0;
sa_buf_type_t buftypes;
buftypes = 0;
ASSERT(count > 0);
for (i = 0; i != count; i++) {
ASSERT(bulk[i].sa_attr <= hdl->sa_os->os_sa->sa_num_attrs);
bulk[i].sa_addr = NULL;
/* First check the bonus buffer */
if (hdl->sa_bonus_tab && TOC_ATTR_PRESENT(
hdl->sa_bonus_tab->sa_idx_tab[bulk[i].sa_attr])) {
SA_ATTR_INFO(sa, hdl->sa_bonus_tab,
SA_GET_HDR(hdl, SA_BONUS),
bulk[i].sa_attr, bulk[i], SA_BONUS, hdl);
if (tx && !(buftypes & SA_BONUS)) {
dmu_buf_will_dirty(hdl->sa_bonus, tx);
buftypes |= SA_BONUS;
}
}
if (bulk[i].sa_addr == NULL &&
((error = sa_get_spill(hdl)) == 0)) {
if (TOC_ATTR_PRESENT(
hdl->sa_spill_tab->sa_idx_tab[bulk[i].sa_attr])) {
SA_ATTR_INFO(sa, hdl->sa_spill_tab,
SA_GET_HDR(hdl, SA_SPILL),
bulk[i].sa_attr, bulk[i], SA_SPILL, hdl);
if (tx && !(buftypes & SA_SPILL) &&
bulk[i].sa_size == bulk[i].sa_length) {
dmu_buf_will_dirty(hdl->sa_spill, tx);
buftypes |= SA_SPILL;
}
}
}
if (error && error != ENOENT) {
return ((error == ECKSUM) ? EIO : error);
}
switch (data_op) {
case SA_LOOKUP:
if (bulk[i].sa_addr == NULL)
return (SET_ERROR(ENOENT));
if (bulk[i].sa_data) {
SA_COPY_DATA(bulk[i].sa_data_func,
bulk[i].sa_addr, bulk[i].sa_data,
bulk[i].sa_size);
}
continue;
case SA_UPDATE:
/* existing rewrite of attr */
if (bulk[i].sa_addr &&
bulk[i].sa_size == bulk[i].sa_length) {
SA_COPY_DATA(bulk[i].sa_data_func,
bulk[i].sa_data, bulk[i].sa_addr,
bulk[i].sa_length);
continue;
} else if (bulk[i].sa_addr) { /* attr size change */
error = sa_modify_attrs(hdl, bulk[i].sa_attr,
SA_REPLACE, bulk[i].sa_data_func,
bulk[i].sa_data, bulk[i].sa_length, tx);
} else { /* adding new attribute */
error = sa_modify_attrs(hdl, bulk[i].sa_attr,
SA_ADD, bulk[i].sa_data_func,
bulk[i].sa_data, bulk[i].sa_length, tx);
}
if (error)
return (error);
break;
default:
break;
}
}
return (error);
}
static sa_lot_t *
-sa_add_layout_entry(objset_t *os, sa_attr_type_t *attrs, int attr_count,
+sa_add_layout_entry(objset_t *os, const sa_attr_type_t *attrs, int attr_count,
uint64_t lot_num, uint64_t hash, boolean_t zapadd, dmu_tx_t *tx)
{
sa_os_t *sa = os->os_sa;
sa_lot_t *tb, *findtb;
int i;
avl_index_t loc;
ASSERT(MUTEX_HELD(&sa->sa_lock));
tb = kmem_zalloc(sizeof (sa_lot_t), KM_SLEEP);
tb->lot_attr_count = attr_count;
tb->lot_attrs = kmem_alloc(sizeof (sa_attr_type_t) * attr_count,
KM_SLEEP);
bcopy(attrs, tb->lot_attrs, sizeof (sa_attr_type_t) * attr_count);
tb->lot_num = lot_num;
tb->lot_hash = hash;
tb->lot_instance = 0;
if (zapadd) {
char attr_name[8];
if (sa->sa_layout_attr_obj == 0) {
sa->sa_layout_attr_obj = zap_create_link(os,
DMU_OT_SA_ATTR_LAYOUTS,
sa->sa_master_obj, SA_LAYOUTS, tx);
}
(void) snprintf(attr_name, sizeof (attr_name),
"%d", (int)lot_num);
VERIFY(0 == zap_update(os, os->os_sa->sa_layout_attr_obj,
attr_name, 2, attr_count, attrs, tx));
}
list_create(&tb->lot_idx_tab, sizeof (sa_idx_tab_t),
offsetof(sa_idx_tab_t, sa_next));
for (i = 0; i != attr_count; i++) {
if (sa->sa_attr_table[tb->lot_attrs[i]].sa_length == 0)
tb->lot_var_sizes++;
}
avl_add(&sa->sa_layout_num_tree, tb);
/* verify we don't have a hash collision */
if ((findtb = avl_find(&sa->sa_layout_hash_tree, tb, &loc)) != NULL) {
for (; findtb && findtb->lot_hash == hash;
findtb = AVL_NEXT(&sa->sa_layout_hash_tree, findtb)) {
if (findtb->lot_instance != tb->lot_instance)
break;
tb->lot_instance++;
}
}
avl_add(&sa->sa_layout_hash_tree, tb);
return (tb);
}
static void
sa_find_layout(objset_t *os, uint64_t hash, sa_attr_type_t *attrs,
int count, dmu_tx_t *tx, sa_lot_t **lot)
{
sa_lot_t *tb, tbsearch;
avl_index_t loc;
sa_os_t *sa = os->os_sa;
boolean_t found = B_FALSE;
mutex_enter(&sa->sa_lock);
tbsearch.lot_hash = hash;
tbsearch.lot_instance = 0;
tb = avl_find(&sa->sa_layout_hash_tree, &tbsearch, &loc);
if (tb) {
for (; tb && tb->lot_hash == hash;
tb = AVL_NEXT(&sa->sa_layout_hash_tree, tb)) {
if (sa_layout_equal(tb, attrs, count) == 0) {
found = B_TRUE;
break;
}
}
}
if (!found) {
tb = sa_add_layout_entry(os, attrs, count,
avl_numnodes(&sa->sa_layout_num_tree), hash, B_TRUE, tx);
}
mutex_exit(&sa->sa_lock);
*lot = tb;
}
static int
sa_resize_spill(sa_handle_t *hdl, uint32_t size, dmu_tx_t *tx)
{
int error;
uint32_t blocksize;
if (size == 0) {
blocksize = SPA_MINBLOCKSIZE;
} else if (size > SPA_OLD_MAXBLOCKSIZE) {
ASSERT(0);
return (SET_ERROR(EFBIG));
} else {
blocksize = P2ROUNDUP_TYPED(size, SPA_MINBLOCKSIZE, uint32_t);
}
error = dbuf_spill_set_blksz(hdl->sa_spill, blocksize, tx);
ASSERT(error == 0);
return (error);
}
static void
sa_copy_data(sa_data_locator_t *func, void *datastart, void *target, int buflen)
{
if (func == NULL) {
bcopy(datastart, target, buflen);
} else {
boolean_t start;
int bytes;
void *dataptr;
void *saptr = target;
uint32_t length;
start = B_TRUE;
bytes = 0;
while (bytes < buflen) {
func(&dataptr, &length, buflen, start, datastart);
bcopy(dataptr, saptr, length);
saptr = (void *)((caddr_t)saptr + length);
bytes += length;
start = B_FALSE;
}
}
}
/*
* Determine several different values pertaining to system attribute
* buffers.
*
* Return the size of the sa_hdr_phys_t header for the buffer. Each
* variable length attribute except the first contributes two bytes to
* the header size, which is then rounded up to an 8-byte boundary.
*
* The following output parameters are also computed.
*
* index - The index of the first attribute in attr_desc that will
* spill over. Only valid if will_spill is set.
*
* total - The total number of bytes of all system attributes described
* in attr_desc.
*
* will_spill - Set when spilling is necessary. It is only set when
* the buftype is SA_BONUS.
*/
static int
sa_find_sizes(sa_os_t *sa, sa_bulk_attr_t *attr_desc, int attr_count,
dmu_buf_t *db, sa_buf_type_t buftype, int full_space, int *index,
int *total, boolean_t *will_spill)
{
int var_size_count = 0;
int i;
int hdrsize;
int extra_hdrsize;
if (buftype == SA_BONUS && sa->sa_force_spill) {
*total = 0;
*index = 0;
*will_spill = B_TRUE;
return (0);
}
*index = -1;
*total = 0;
*will_spill = B_FALSE;
extra_hdrsize = 0;
hdrsize = (SA_BONUSTYPE_FROM_DB(db) == DMU_OT_ZNODE) ? 0 :
sizeof (sa_hdr_phys_t);
ASSERT(IS_P2ALIGNED(full_space, 8));
for (i = 0; i != attr_count; i++) {
boolean_t is_var_sz, might_spill_here;
int tmp_hdrsize;
*total = P2ROUNDUP(*total, 8);
*total += attr_desc[i].sa_length;
if (*will_spill)
continue;
is_var_sz = (SA_REGISTERED_LEN(sa, attr_desc[i].sa_attr) == 0);
if (is_var_sz)
var_size_count++;
/*
* Calculate what the SA header size would be if this
* attribute doesn't spill.
*/
tmp_hdrsize = hdrsize + ((is_var_sz && var_size_count > 1) ?
sizeof (uint16_t) : 0);
/*
* Check whether this attribute spans into the space
* that would be used by the spill block pointer should
* a spill block be needed.
*/
might_spill_here =
buftype == SA_BONUS && *index == -1 &&
(*total + P2ROUNDUP(tmp_hdrsize, 8)) >
(full_space - sizeof (blkptr_t));
if (is_var_sz && var_size_count > 1) {
if (buftype == SA_SPILL ||
tmp_hdrsize + *total < full_space) {
/*
* Record the extra header size in case this
* increase needs to be reversed due to
* spill-over.
*/
hdrsize = tmp_hdrsize;
if (*index != -1 || might_spill_here)
extra_hdrsize += sizeof (uint16_t);
} else {
ASSERT(buftype == SA_BONUS);
if (*index == -1)
*index = i;
*will_spill = B_TRUE;
continue;
}
}
/*
* Store index of where spill *could* occur. Then
* continue to count the remaining attribute sizes. The
* sum is used later for sizing bonus and spill buffer.
*/
if (might_spill_here)
*index = i;
if ((*total + P2ROUNDUP(hdrsize, 8)) > full_space &&
buftype == SA_BONUS)
*will_spill = B_TRUE;
}
if (*will_spill)
hdrsize -= extra_hdrsize;
hdrsize = P2ROUNDUP(hdrsize, 8);
return (hdrsize);
}
#define BUF_SPACE_NEEDED(total, header) (total + header)
/*
* Find layout that corresponds to ordering of attributes
* If not found a new layout number is created and added to
* persistent layout tables.
*/
static int
sa_build_layouts(sa_handle_t *hdl, sa_bulk_attr_t *attr_desc, int attr_count,
dmu_tx_t *tx)
{
sa_os_t *sa = hdl->sa_os->os_sa;
uint64_t hash;
sa_buf_type_t buftype;
sa_hdr_phys_t *sahdr;
void *data_start;
sa_attr_type_t *attrs, *attrs_start;
int i, lot_count;
int dnodesize;
int spill_idx;
int hdrsize;
int spillhdrsize = 0;
int used;
dmu_object_type_t bonustype;
sa_lot_t *lot;
int len_idx;
int spill_used;
int bonuslen;
boolean_t spilling;
dmu_buf_will_dirty(hdl->sa_bonus, tx);
bonustype = SA_BONUSTYPE_FROM_DB(hdl->sa_bonus);
dmu_object_dnsize_from_db(hdl->sa_bonus, &dnodesize);
bonuslen = DN_BONUS_SIZE(dnodesize);
/* first determine bonus header size and sum of all attributes */
hdrsize = sa_find_sizes(sa, attr_desc, attr_count, hdl->sa_bonus,
SA_BONUS, bonuslen, &spill_idx, &used, &spilling);
if (used > SPA_OLD_MAXBLOCKSIZE)
return (SET_ERROR(EFBIG));
VERIFY0(dmu_set_bonus(hdl->sa_bonus, spilling ?
MIN(bonuslen - sizeof (blkptr_t), used + hdrsize) :
used + hdrsize, tx));
ASSERT((bonustype == DMU_OT_ZNODE && spilling == 0) ||
bonustype == DMU_OT_SA);
/* setup and size spill buffer when needed */
if (spilling) {
boolean_t dummy;
if (hdl->sa_spill == NULL) {
VERIFY(dmu_spill_hold_by_bonus(hdl->sa_bonus, 0, NULL,
&hdl->sa_spill) == 0);
}
dmu_buf_will_dirty(hdl->sa_spill, tx);
spillhdrsize = sa_find_sizes(sa, &attr_desc[spill_idx],
attr_count - spill_idx, hdl->sa_spill, SA_SPILL,
hdl->sa_spill->db_size, &i, &spill_used, &dummy);
if (spill_used > SPA_OLD_MAXBLOCKSIZE)
return (SET_ERROR(EFBIG));
if (BUF_SPACE_NEEDED(spill_used, spillhdrsize) >
hdl->sa_spill->db_size)
VERIFY(0 == sa_resize_spill(hdl,
BUF_SPACE_NEEDED(spill_used, spillhdrsize), tx));
}
/* setup starting pointers to lay down data */
data_start = (void *)((uintptr_t)hdl->sa_bonus->db_data + hdrsize);
sahdr = (sa_hdr_phys_t *)hdl->sa_bonus->db_data;
buftype = SA_BONUS;
attrs_start = attrs = kmem_alloc(sizeof (sa_attr_type_t) * attr_count,
KM_SLEEP);
lot_count = 0;
for (i = 0, len_idx = 0, hash = -1ULL; i != attr_count; i++) {
uint16_t length;
ASSERT(IS_P2ALIGNED(data_start, 8));
attrs[i] = attr_desc[i].sa_attr;
length = SA_REGISTERED_LEN(sa, attrs[i]);
if (length == 0)
length = attr_desc[i].sa_length;
if (spilling && i == spill_idx) { /* switch to spill buffer */
VERIFY(bonustype == DMU_OT_SA);
if (buftype == SA_BONUS && !sa->sa_force_spill) {
sa_find_layout(hdl->sa_os, hash, attrs_start,
lot_count, tx, &lot);
SA_SET_HDR(sahdr, lot->lot_num, hdrsize);
}
buftype = SA_SPILL;
hash = -1ULL;
len_idx = 0;
sahdr = (sa_hdr_phys_t *)hdl->sa_spill->db_data;
sahdr->sa_magic = SA_MAGIC;
data_start = (void *)((uintptr_t)sahdr +
spillhdrsize);
attrs_start = &attrs[i];
lot_count = 0;
}
hash ^= SA_ATTR_HASH(attrs[i]);
attr_desc[i].sa_addr = data_start;
attr_desc[i].sa_size = length;
SA_COPY_DATA(attr_desc[i].sa_data_func, attr_desc[i].sa_data,
data_start, length);
if (sa->sa_attr_table[attrs[i]].sa_length == 0) {
sahdr->sa_lengths[len_idx++] = length;
}
data_start = (void *)P2ROUNDUP(((uintptr_t)data_start +
length), 8);
lot_count++;
}
sa_find_layout(hdl->sa_os, hash, attrs_start, lot_count, tx, &lot);
/*
* Verify that old znodes always have layout number 0.
* Must be DMU_OT_SA for arbitrary layouts
*/
VERIFY((bonustype == DMU_OT_ZNODE && lot->lot_num == 0) ||
(bonustype == DMU_OT_SA && lot->lot_num > 1));
if (bonustype == DMU_OT_SA) {
SA_SET_HDR(sahdr, lot->lot_num,
buftype == SA_BONUS ? hdrsize : spillhdrsize);
}
kmem_free(attrs, sizeof (sa_attr_type_t) * attr_count);
if (hdl->sa_bonus_tab) {
sa_idx_tab_rele(hdl->sa_os, hdl->sa_bonus_tab);
hdl->sa_bonus_tab = NULL;
}
if (!sa->sa_force_spill)
VERIFY(0 == sa_build_index(hdl, SA_BONUS));
if (hdl->sa_spill) {
sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);
if (!spilling) {
/*
* remove spill block that is no longer needed.
*/
dmu_buf_rele(hdl->sa_spill, NULL);
hdl->sa_spill = NULL;
hdl->sa_spill_tab = NULL;
VERIFY(0 == dmu_rm_spill(hdl->sa_os,
sa_handle_object(hdl), tx));
} else {
VERIFY(0 == sa_build_index(hdl, SA_SPILL));
}
}
return (0);
}
static void
sa_free_attr_table(sa_os_t *sa)
{
int i;
if (sa->sa_attr_table == NULL)
return;
for (i = 0; i != sa->sa_num_attrs; i++) {
if (sa->sa_attr_table[i].sa_name)
kmem_free(sa->sa_attr_table[i].sa_name,
strlen(sa->sa_attr_table[i].sa_name) + 1);
}
kmem_free(sa->sa_attr_table,
sizeof (sa_attr_table_t) * sa->sa_num_attrs);
sa->sa_attr_table = NULL;
}
static int
-sa_attr_table_setup(objset_t *os, sa_attr_reg_t *reg_attrs, int count)
+sa_attr_table_setup(objset_t *os, const sa_attr_reg_t *reg_attrs, int count)
{
sa_os_t *sa = os->os_sa;
uint64_t sa_attr_count = 0;
uint64_t sa_reg_count = 0;
int error = 0;
uint64_t attr_value;
sa_attr_table_t *tb;
zap_cursor_t zc;
zap_attribute_t za;
int registered_count = 0;
int i;
dmu_objset_type_t ostype = dmu_objset_type(os);
sa->sa_user_table =
kmem_zalloc(count * sizeof (sa_attr_type_t), KM_SLEEP);
sa->sa_user_table_sz = count * sizeof (sa_attr_type_t);
if (sa->sa_reg_attr_obj != 0) {
error = zap_count(os, sa->sa_reg_attr_obj,
&sa_attr_count);
/*
* Make sure we retrieved a count and that it isn't zero
*/
if (error || (error == 0 && sa_attr_count == 0)) {
if (error == 0)
error = SET_ERROR(EINVAL);
goto bail;
}
sa_reg_count = sa_attr_count;
}
if (ostype == DMU_OST_ZFS && sa_attr_count == 0)
sa_attr_count += sa_legacy_attr_count;
/* Allocate attribute numbers for attributes that aren't registered */
for (i = 0; i != count; i++) {
boolean_t found = B_FALSE;
int j;
if (ostype == DMU_OST_ZFS) {
for (j = 0; j != sa_legacy_attr_count; j++) {
if (strcmp(reg_attrs[i].sa_name,
sa_legacy_attrs[j].sa_name) == 0) {
sa->sa_user_table[i] =
sa_legacy_attrs[j].sa_attr;
found = B_TRUE;
}
}
}
if (found)
continue;
if (sa->sa_reg_attr_obj)
error = zap_lookup(os, sa->sa_reg_attr_obj,
reg_attrs[i].sa_name, 8, 1, &attr_value);
else
error = SET_ERROR(ENOENT);
switch (error) {
case ENOENT:
sa->sa_user_table[i] = (sa_attr_type_t)sa_attr_count;
sa_attr_count++;
break;
case 0:
sa->sa_user_table[i] = ATTR_NUM(attr_value);
break;
default:
goto bail;
}
}
sa->sa_num_attrs = sa_attr_count;
tb = sa->sa_attr_table =
kmem_zalloc(sizeof (sa_attr_table_t) * sa_attr_count, KM_SLEEP);
/*
* Attribute table is constructed from requested attribute list,
* previously foreign registered attributes, and also the legacy
* ZPL set of attributes.
*/
if (sa->sa_reg_attr_obj) {
for (zap_cursor_init(&zc, os, sa->sa_reg_attr_obj);
(error = zap_cursor_retrieve(&zc, &za)) == 0;
zap_cursor_advance(&zc)) {
uint64_t value;
value = za.za_first_integer;
registered_count++;
tb[ATTR_NUM(value)].sa_attr = ATTR_NUM(value);
tb[ATTR_NUM(value)].sa_length = ATTR_LENGTH(value);
tb[ATTR_NUM(value)].sa_byteswap = ATTR_BSWAP(value);
tb[ATTR_NUM(value)].sa_registered = B_TRUE;
if (tb[ATTR_NUM(value)].sa_name) {
continue;
}
tb[ATTR_NUM(value)].sa_name =
kmem_zalloc(strlen(za.za_name) +1, KM_SLEEP);
(void) strlcpy(tb[ATTR_NUM(value)].sa_name, za.za_name,
strlen(za.za_name) +1);
}
zap_cursor_fini(&zc);
/*
* Make sure we processed the correct number of registered
* attributes
*/
if (registered_count != sa_reg_count) {
ASSERT(error != 0);
goto bail;
}
}
if (ostype == DMU_OST_ZFS) {
for (i = 0; i != sa_legacy_attr_count; i++) {
if (tb[i].sa_name)
continue;
tb[i].sa_attr = sa_legacy_attrs[i].sa_attr;
tb[i].sa_length = sa_legacy_attrs[i].sa_length;
tb[i].sa_byteswap = sa_legacy_attrs[i].sa_byteswap;
tb[i].sa_registered = B_FALSE;
tb[i].sa_name =
kmem_zalloc(strlen(sa_legacy_attrs[i].sa_name) +1,
KM_SLEEP);
(void) strlcpy(tb[i].sa_name,
sa_legacy_attrs[i].sa_name,
strlen(sa_legacy_attrs[i].sa_name) + 1);
}
}
for (i = 0; i != count; i++) {
sa_attr_type_t attr_id;
attr_id = sa->sa_user_table[i];
if (tb[attr_id].sa_name)
continue;
tb[attr_id].sa_length = reg_attrs[i].sa_length;
tb[attr_id].sa_byteswap = reg_attrs[i].sa_byteswap;
tb[attr_id].sa_attr = attr_id;
tb[attr_id].sa_name =
kmem_zalloc(strlen(reg_attrs[i].sa_name) + 1, KM_SLEEP);
(void) strlcpy(tb[attr_id].sa_name, reg_attrs[i].sa_name,
strlen(reg_attrs[i].sa_name) + 1);
}
sa->sa_need_attr_registration =
(sa_attr_count != registered_count);
return (0);
bail:
kmem_free(sa->sa_user_table, count * sizeof (sa_attr_type_t));
sa->sa_user_table = NULL;
sa_free_attr_table(sa);
ASSERT(error != 0);
return (error);
}
int
-sa_setup(objset_t *os, uint64_t sa_obj, sa_attr_reg_t *reg_attrs, int count,
- sa_attr_type_t **user_table)
+sa_setup(objset_t *os, uint64_t sa_obj, const sa_attr_reg_t *reg_attrs,
+ int count, sa_attr_type_t **user_table)
{
zap_cursor_t zc;
zap_attribute_t za;
sa_os_t *sa;
dmu_objset_type_t ostype = dmu_objset_type(os);
sa_attr_type_t *tb;
int error;
mutex_enter(&os->os_user_ptr_lock);
if (os->os_sa) {
mutex_enter(&os->os_sa->sa_lock);
mutex_exit(&os->os_user_ptr_lock);
tb = os->os_sa->sa_user_table;
mutex_exit(&os->os_sa->sa_lock);
*user_table = tb;
return (0);
}
sa = kmem_zalloc(sizeof (sa_os_t), KM_SLEEP);
mutex_init(&sa->sa_lock, NULL, MUTEX_NOLOCKDEP, NULL);
sa->sa_master_obj = sa_obj;
os->os_sa = sa;
mutex_enter(&sa->sa_lock);
mutex_exit(&os->os_user_ptr_lock);
avl_create(&sa->sa_layout_num_tree, layout_num_compare,
sizeof (sa_lot_t), offsetof(sa_lot_t, lot_num_node));
avl_create(&sa->sa_layout_hash_tree, layout_hash_compare,
sizeof (sa_lot_t), offsetof(sa_lot_t, lot_hash_node));
if (sa_obj) {
error = zap_lookup(os, sa_obj, SA_LAYOUTS,
8, 1, &sa->sa_layout_attr_obj);
if (error != 0 && error != ENOENT)
goto fail;
error = zap_lookup(os, sa_obj, SA_REGISTRY,
8, 1, &sa->sa_reg_attr_obj);
if (error != 0 && error != ENOENT)
goto fail;
}
if ((error = sa_attr_table_setup(os, reg_attrs, count)) != 0)
goto fail;
if (sa->sa_layout_attr_obj != 0) {
uint64_t layout_count;
error = zap_count(os, sa->sa_layout_attr_obj,
&layout_count);
/*
* Layout number count should be > 0
*/
if (error || (error == 0 && layout_count == 0)) {
if (error == 0)
error = SET_ERROR(EINVAL);
goto fail;
}
for (zap_cursor_init(&zc, os, sa->sa_layout_attr_obj);
(error = zap_cursor_retrieve(&zc, &za)) == 0;
zap_cursor_advance(&zc)) {
sa_attr_type_t *lot_attrs;
uint64_t lot_num;
lot_attrs = kmem_zalloc(sizeof (sa_attr_type_t) *
za.za_num_integers, KM_SLEEP);
if ((error = (zap_lookup(os, sa->sa_layout_attr_obj,
za.za_name, 2, za.za_num_integers,
lot_attrs))) != 0) {
kmem_free(lot_attrs, sizeof (sa_attr_type_t) *
za.za_num_integers);
break;
}
VERIFY(ddi_strtoull(za.za_name, NULL, 10,
(unsigned long long *)&lot_num) == 0);
(void) sa_add_layout_entry(os, lot_attrs,
za.za_num_integers, lot_num,
sa_layout_info_hash(lot_attrs,
za.za_num_integers), B_FALSE, NULL);
kmem_free(lot_attrs, sizeof (sa_attr_type_t) *
za.za_num_integers);
}
zap_cursor_fini(&zc);
/*
* Make sure layout count matches number of entries added
* to AVL tree
*/
if (avl_numnodes(&sa->sa_layout_num_tree) != layout_count) {
ASSERT(error != 0);
goto fail;
}
}
/* Add special layout number for old ZNODES */
if (ostype == DMU_OST_ZFS) {
(void) sa_add_layout_entry(os, sa_legacy_zpl_layout,
sa_legacy_attr_count, 0,
sa_layout_info_hash(sa_legacy_zpl_layout,
sa_legacy_attr_count), B_FALSE, NULL);
(void) sa_add_layout_entry(os, sa_dummy_zpl_layout, 0, 1,
0, B_FALSE, NULL);
}
*user_table = os->os_sa->sa_user_table;
mutex_exit(&sa->sa_lock);
return (0);
fail:
os->os_sa = NULL;
sa_free_attr_table(sa);
if (sa->sa_user_table)
kmem_free(sa->sa_user_table, sa->sa_user_table_sz);
mutex_exit(&sa->sa_lock);
avl_destroy(&sa->sa_layout_hash_tree);
avl_destroy(&sa->sa_layout_num_tree);
mutex_destroy(&sa->sa_lock);
kmem_free(sa, sizeof (sa_os_t));
return ((error == ECKSUM) ? EIO : error);
}
void
sa_tear_down(objset_t *os)
{
sa_os_t *sa = os->os_sa;
sa_lot_t *layout;
void *cookie;
kmem_free(sa->sa_user_table, sa->sa_user_table_sz);
/* Free up attr table */
sa_free_attr_table(sa);
cookie = NULL;
while ((layout =
avl_destroy_nodes(&sa->sa_layout_hash_tree, &cookie))) {
sa_idx_tab_t *tab;
while ((tab = list_head(&layout->lot_idx_tab))) {
ASSERT(zfs_refcount_count(&tab->sa_refcount));
sa_idx_tab_rele(os, tab);
}
}
cookie = NULL;
while ((layout = avl_destroy_nodes(&sa->sa_layout_num_tree, &cookie))) {
kmem_free(layout->lot_attrs,
sizeof (sa_attr_type_t) * layout->lot_attr_count);
kmem_free(layout, sizeof (sa_lot_t));
}
avl_destroy(&sa->sa_layout_hash_tree);
avl_destroy(&sa->sa_layout_num_tree);
mutex_destroy(&sa->sa_lock);
kmem_free(sa, sizeof (sa_os_t));
os->os_sa = NULL;
}
static void
sa_build_idx_tab(void *hdr, void *attr_addr, sa_attr_type_t attr,
uint16_t length, int length_idx, boolean_t var_length, void *userp)
{
sa_idx_tab_t *idx_tab = userp;
if (var_length) {
ASSERT(idx_tab->sa_variable_lengths);
idx_tab->sa_variable_lengths[length_idx] = length;
}
TOC_ATTR_ENCODE(idx_tab->sa_idx_tab[attr], length_idx,
(uint32_t)((uintptr_t)attr_addr - (uintptr_t)hdr));
}
static void
sa_attr_iter(objset_t *os, sa_hdr_phys_t *hdr, dmu_object_type_t type,
sa_iterfunc_t func, sa_lot_t *tab, void *userp)
{
void *data_start;
sa_lot_t *tb = tab;
sa_lot_t search;
avl_index_t loc;
sa_os_t *sa = os->os_sa;
int i;
uint16_t *length_start = NULL;
uint8_t length_idx = 0;
if (tab == NULL) {
search.lot_num = SA_LAYOUT_NUM(hdr, type);
tb = avl_find(&sa->sa_layout_num_tree, &search, &loc);
ASSERT(tb);
}
if (IS_SA_BONUSTYPE(type)) {
data_start = (void *)P2ROUNDUP(((uintptr_t)hdr +
offsetof(sa_hdr_phys_t, sa_lengths) +
(sizeof (uint16_t) * tb->lot_var_sizes)), 8);
length_start = hdr->sa_lengths;
} else {
data_start = hdr;
}
for (i = 0; i != tb->lot_attr_count; i++) {
int attr_length, reg_length;
uint8_t idx_len;
reg_length = sa->sa_attr_table[tb->lot_attrs[i]].sa_length;
if (reg_length) {
attr_length = reg_length;
idx_len = 0;
} else {
attr_length = length_start[length_idx];
idx_len = length_idx++;
}
func(hdr, data_start, tb->lot_attrs[i], attr_length,
idx_len, reg_length == 0 ? B_TRUE : B_FALSE, userp);
data_start = (void *)P2ROUNDUP(((uintptr_t)data_start +
attr_length), 8);
}
}
static void
sa_byteswap_cb(void *hdr, void *attr_addr, sa_attr_type_t attr,
uint16_t length, int length_idx, boolean_t variable_length, void *userp)
{
(void) hdr, (void) length_idx, (void) variable_length;
sa_handle_t *hdl = userp;
sa_os_t *sa = hdl->sa_os->os_sa;
sa_bswap_table[sa->sa_attr_table[attr].sa_byteswap](attr_addr, length);
}
static void
sa_byteswap(sa_handle_t *hdl, sa_buf_type_t buftype)
{
sa_hdr_phys_t *sa_hdr_phys = SA_GET_HDR(hdl, buftype);
dmu_buf_impl_t *db;
int num_lengths = 1;
int i;
sa_os_t *sa __maybe_unused = hdl->sa_os->os_sa;
ASSERT(MUTEX_HELD(&sa->sa_lock));
if (sa_hdr_phys->sa_magic == SA_MAGIC)
return;
db = SA_GET_DB(hdl, buftype);
if (buftype == SA_SPILL) {
arc_release(db->db_buf, NULL);
arc_buf_thaw(db->db_buf);
}
sa_hdr_phys->sa_magic = BSWAP_32(sa_hdr_phys->sa_magic);
sa_hdr_phys->sa_layout_info = BSWAP_16(sa_hdr_phys->sa_layout_info);
/*
* Determine number of variable lengths in header
* The standard 8 byte header has one for free and a
* 16 byte header would have 4 + 1;
*/
if (SA_HDR_SIZE(sa_hdr_phys) > 8)
num_lengths += (SA_HDR_SIZE(sa_hdr_phys) - 8) >> 1;
for (i = 0; i != num_lengths; i++)
sa_hdr_phys->sa_lengths[i] =
BSWAP_16(sa_hdr_phys->sa_lengths[i]);
sa_attr_iter(hdl->sa_os, sa_hdr_phys, DMU_OT_SA,
sa_byteswap_cb, NULL, hdl);
if (buftype == SA_SPILL)
arc_buf_freeze(((dmu_buf_impl_t *)hdl->sa_spill)->db_buf);
}
static int
sa_build_index(sa_handle_t *hdl, sa_buf_type_t buftype)
{
sa_hdr_phys_t *sa_hdr_phys;
dmu_buf_impl_t *db = SA_GET_DB(hdl, buftype);
dmu_object_type_t bonustype = SA_BONUSTYPE_FROM_DB(db);
sa_os_t *sa = hdl->sa_os->os_sa;
sa_idx_tab_t *idx_tab;
sa_hdr_phys = SA_GET_HDR(hdl, buftype);
mutex_enter(&sa->sa_lock);
/* Do we need to byteswap? */
/* only check if not old znode */
if (IS_SA_BONUSTYPE(bonustype) && sa_hdr_phys->sa_magic != SA_MAGIC &&
sa_hdr_phys->sa_magic != 0) {
if (BSWAP_32(sa_hdr_phys->sa_magic) != SA_MAGIC) {
mutex_exit(&sa->sa_lock);
zfs_dbgmsg("Buffer Header: %x != SA_MAGIC:%x "
"object=%#llx\n", sa_hdr_phys->sa_magic, SA_MAGIC,
(u_longlong_t)db->db.db_object);
return (SET_ERROR(EIO));
}
sa_byteswap(hdl, buftype);
}
idx_tab = sa_find_idx_tab(hdl->sa_os, bonustype, sa_hdr_phys);
if (buftype == SA_BONUS)
hdl->sa_bonus_tab = idx_tab;
else
hdl->sa_spill_tab = idx_tab;
mutex_exit(&sa->sa_lock);
return (0);
}
static void
sa_evict_sync(void *dbu)
{
(void) dbu;
panic("evicting sa dbuf\n");
}
static void
sa_idx_tab_rele(objset_t *os, void *arg)
{
sa_os_t *sa = os->os_sa;
sa_idx_tab_t *idx_tab = arg;
if (idx_tab == NULL)
return;
mutex_enter(&sa->sa_lock);
if (zfs_refcount_remove(&idx_tab->sa_refcount, NULL) == 0) {
list_remove(&idx_tab->sa_layout->lot_idx_tab, idx_tab);
if (idx_tab->sa_variable_lengths)
kmem_free(idx_tab->sa_variable_lengths,
sizeof (uint16_t) *
idx_tab->sa_layout->lot_var_sizes);
zfs_refcount_destroy(&idx_tab->sa_refcount);
kmem_free(idx_tab->sa_idx_tab,
sizeof (uint32_t) * sa->sa_num_attrs);
kmem_free(idx_tab, sizeof (sa_idx_tab_t));
}
mutex_exit(&sa->sa_lock);
}
static void
sa_idx_tab_hold(objset_t *os, sa_idx_tab_t *idx_tab)
{
sa_os_t *sa __maybe_unused = os->os_sa;
ASSERT(MUTEX_HELD(&sa->sa_lock));
(void) zfs_refcount_add(&idx_tab->sa_refcount, NULL);
}
void
sa_spill_rele(sa_handle_t *hdl)
{
mutex_enter(&hdl->sa_lock);
if (hdl->sa_spill) {
sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);
dmu_buf_rele(hdl->sa_spill, NULL);
hdl->sa_spill = NULL;
hdl->sa_spill_tab = NULL;
}
mutex_exit(&hdl->sa_lock);
}
void
sa_handle_destroy(sa_handle_t *hdl)
{
dmu_buf_t *db = hdl->sa_bonus;
mutex_enter(&hdl->sa_lock);
(void) dmu_buf_remove_user(db, &hdl->sa_dbu);
if (hdl->sa_bonus_tab)
sa_idx_tab_rele(hdl->sa_os, hdl->sa_bonus_tab);
if (hdl->sa_spill_tab)
sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);
dmu_buf_rele(hdl->sa_bonus, NULL);
if (hdl->sa_spill)
dmu_buf_rele(hdl->sa_spill, NULL);
mutex_exit(&hdl->sa_lock);
kmem_cache_free(sa_cache, hdl);
}
int
sa_handle_get_from_db(objset_t *os, dmu_buf_t *db, void *userp,
sa_handle_type_t hdl_type, sa_handle_t **handlepp)
{
int error = 0;
sa_handle_t *handle = NULL;
#ifdef ZFS_DEBUG
dmu_object_info_t doi;
dmu_object_info_from_db(db, &doi);
ASSERT(doi.doi_bonus_type == DMU_OT_SA ||
doi.doi_bonus_type == DMU_OT_ZNODE);
#endif
/* find handle, if it exists */
/* if one doesn't exist then create a new one, and initialize it */
if (hdl_type == SA_HDL_SHARED)
handle = dmu_buf_get_user(db);
if (handle == NULL) {
sa_handle_t *winner = NULL;
handle = kmem_cache_alloc(sa_cache, KM_SLEEP);
handle->sa_dbu.dbu_evict_func_sync = NULL;
handle->sa_dbu.dbu_evict_func_async = NULL;
handle->sa_userp = userp;
handle->sa_bonus = db;
handle->sa_os = os;
handle->sa_spill = NULL;
handle->sa_bonus_tab = NULL;
handle->sa_spill_tab = NULL;
error = sa_build_index(handle, SA_BONUS);
if (hdl_type == SA_HDL_SHARED) {
dmu_buf_init_user(&handle->sa_dbu, sa_evict_sync, NULL,
NULL);
winner = dmu_buf_set_user_ie(db, &handle->sa_dbu);
}
if (winner != NULL) {
kmem_cache_free(sa_cache, handle);
handle = winner;
}
}
*handlepp = handle;
return (error);
}
int
sa_handle_get(objset_t *objset, uint64_t objid, void *userp,
sa_handle_type_t hdl_type, sa_handle_t **handlepp)
{
dmu_buf_t *db;
int error;
if ((error = dmu_bonus_hold(objset, objid, NULL, &db)))
return (error);
return (sa_handle_get_from_db(objset, db, userp, hdl_type,
handlepp));
}
int
sa_buf_hold(objset_t *objset, uint64_t obj_num, void *tag, dmu_buf_t **db)
{
return (dmu_bonus_hold(objset, obj_num, tag, db));
}
void
sa_buf_rele(dmu_buf_t *db, void *tag)
{
dmu_buf_rele(db, tag);
}
static int
sa_lookup_impl(sa_handle_t *hdl, sa_bulk_attr_t *bulk, int count)
{
ASSERT(hdl);
ASSERT(MUTEX_HELD(&hdl->sa_lock));
return (sa_attr_op(hdl, bulk, count, SA_LOOKUP, NULL));
}
static int
sa_lookup_locked(sa_handle_t *hdl, sa_attr_type_t attr, void *buf,
uint32_t buflen)
{
int error;
sa_bulk_attr_t bulk;
VERIFY3U(buflen, <=, SA_ATTR_MAX_LEN);
bulk.sa_attr = attr;
bulk.sa_data = buf;
bulk.sa_length = buflen;
bulk.sa_data_func = NULL;
ASSERT(hdl);
error = sa_lookup_impl(hdl, &bulk, 1);
return (error);
}
int
sa_lookup(sa_handle_t *hdl, sa_attr_type_t attr, void *buf, uint32_t buflen)
{
int error;
mutex_enter(&hdl->sa_lock);
error = sa_lookup_locked(hdl, attr, buf, buflen);
mutex_exit(&hdl->sa_lock);
return (error);
}
#ifdef _KERNEL
int
sa_lookup_uio(sa_handle_t *hdl, sa_attr_type_t attr, zfs_uio_t *uio)
{
int error;
sa_bulk_attr_t bulk;
bulk.sa_data = NULL;
bulk.sa_attr = attr;
bulk.sa_data_func = NULL;
ASSERT(hdl);
mutex_enter(&hdl->sa_lock);
if ((error = sa_attr_op(hdl, &bulk, 1, SA_LOOKUP, NULL)) == 0) {
error = zfs_uiomove((void *)bulk.sa_addr, MIN(bulk.sa_size,
zfs_uio_resid(uio)), UIO_READ, uio);
}
mutex_exit(&hdl->sa_lock);
return (error);
}
/*
* For the existed object that is upgraded from old system, its ondisk layout
* has no slot for the project ID attribute. But quota accounting logic needs
* to access related slots by offset directly. So we need to adjust these old
* objects' layout to make the project ID to some unified and fixed offset.
*/
int
sa_add_projid(sa_handle_t *hdl, dmu_tx_t *tx, uint64_t projid)
{
znode_t *zp = sa_get_userdata(hdl);
dmu_buf_t *db = sa_get_db(hdl);
zfsvfs_t *zfsvfs = ZTOZSB(zp);
int count = 0, err = 0;
sa_bulk_attr_t *bulk, *attrs;
zfs_acl_locator_cb_t locate = { 0 };
uint64_t uid, gid, mode, rdev, xattr = 0, parent, gen, links;
uint64_t crtime[2], mtime[2], ctime[2], atime[2];
zfs_acl_phys_t znode_acl = { 0 };
char scanstamp[AV_SCANSTAMP_SZ];
if (zp->z_acl_cached == NULL) {
zfs_acl_t *aclp;
mutex_enter(&zp->z_acl_lock);
err = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
mutex_exit(&zp->z_acl_lock);
if (err != 0 && err != ENOENT)
return (err);
}
bulk = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
mutex_enter(&hdl->sa_lock);
mutex_enter(&zp->z_lock);
err = sa_lookup_locked(hdl, SA_ZPL_PROJID(zfsvfs), &projid,
sizeof (uint64_t));
if (unlikely(err == 0))
/* Someone has added project ID attr by race. */
err = EEXIST;
if (err != ENOENT)
goto out;
/* First do a bulk query of the attributes that aren't cached */
if (zp->z_is_sa) {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
&mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
&gen, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
&uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
&gid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
&parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
&atime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
&mtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
&ctime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL,
&crtime, 16);
if (Z_ISBLK(ZTOTYPE(zp)) || Z_ISCHR(ZTOTYPE(zp)))
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL,
&rdev, 8);
} else {
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
&atime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
&mtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
&ctime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL,
&crtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
&gen, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
&mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
&parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL,
&xattr, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL,
&rdev, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
&uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
&gid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
&znode_acl, 88);
}
err = sa_bulk_lookup_locked(hdl, bulk, count);
if (err != 0)
goto out;
err = sa_lookup_locked(hdl, SA_ZPL_XATTR(zfsvfs), &xattr, 8);
if (err != 0 && err != ENOENT)
goto out;
zp->z_projid = projid;
zp->z_pflags |= ZFS_PROJID;
links = ZTONLNK(zp);
count = 0;
err = 0;
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
&zp->z_size, 8);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_GEN(zfsvfs), NULL, &gen, 8);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, 8);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
&crtime, 16);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_PROJID(zfsvfs), NULL, &projid, 8);
if (Z_ISBLK(ZTOTYPE(zp)) || Z_ISCHR(ZTOTYPE(zp)))
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
&rdev, 8);
if (zp->z_acl_cached != NULL) {
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
&zp->z_acl_cached->z_acl_count, 8);
if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
zfs_acl_xform(zp, zp->z_acl_cached, CRED());
locate.cb_aclp = zp->z_acl_cached;
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
zfs_acl_data_locator, &locate,
zp->z_acl_cached->z_acl_bytes);
}
if (xattr)
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_XATTR(zfsvfs), NULL,
&xattr, 8);
if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
scanstamp, AV_SCANSTAMP_SZ);
SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_SCANSTAMP(zfsvfs), NULL,
scanstamp, AV_SCANSTAMP_SZ);
zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
}
VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
VERIFY(sa_replace_all_by_template_locked(hdl, attrs, count, tx) == 0);
if (znode_acl.z_acl_extern_obj) {
VERIFY(0 == dmu_object_free(zfsvfs->z_os,
znode_acl.z_acl_extern_obj, tx));
}
zp->z_is_sa = B_TRUE;
out:
mutex_exit(&zp->z_lock);
mutex_exit(&hdl->sa_lock);
kmem_free(attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
kmem_free(bulk, sizeof (sa_bulk_attr_t) * ZPL_END);
return (err);
}
#endif
static sa_idx_tab_t *
sa_find_idx_tab(objset_t *os, dmu_object_type_t bonustype, sa_hdr_phys_t *hdr)
{
sa_idx_tab_t *idx_tab;
sa_os_t *sa = os->os_sa;
sa_lot_t *tb, search;
avl_index_t loc;
/*
* Deterimine layout number. If SA node and header == 0 then
* force the index table to the dummy "1" empty layout.
*
* The layout number would only be zero for a newly created file
* that has not added any attributes yet, or with crypto enabled which
* doesn't write any attributes to the bonus buffer.
*/
search.lot_num = SA_LAYOUT_NUM(hdr, bonustype);
tb = avl_find(&sa->sa_layout_num_tree, &search, &loc);
/* Verify header size is consistent with layout information */
ASSERT(tb);
ASSERT((IS_SA_BONUSTYPE(bonustype) &&
SA_HDR_SIZE_MATCH_LAYOUT(hdr, tb)) || !IS_SA_BONUSTYPE(bonustype) ||
(IS_SA_BONUSTYPE(bonustype) && hdr->sa_layout_info == 0));
/*
* See if any of the already existing TOC entries can be reused?
*/
for (idx_tab = list_head(&tb->lot_idx_tab); idx_tab;
idx_tab = list_next(&tb->lot_idx_tab, idx_tab)) {
boolean_t valid_idx = B_TRUE;
int i;
if (tb->lot_var_sizes != 0 &&
idx_tab->sa_variable_lengths != NULL) {
for (i = 0; i != tb->lot_var_sizes; i++) {
if (hdr->sa_lengths[i] !=
idx_tab->sa_variable_lengths[i]) {
valid_idx = B_FALSE;
break;
}
}
}
if (valid_idx) {
sa_idx_tab_hold(os, idx_tab);
return (idx_tab);
}
}
/* No such luck, create a new entry */
idx_tab = kmem_zalloc(sizeof (sa_idx_tab_t), KM_SLEEP);
idx_tab->sa_idx_tab =
kmem_zalloc(sizeof (uint32_t) * sa->sa_num_attrs, KM_SLEEP);
idx_tab->sa_layout = tb;
zfs_refcount_create(&idx_tab->sa_refcount);
if (tb->lot_var_sizes)
idx_tab->sa_variable_lengths = kmem_alloc(sizeof (uint16_t) *
tb->lot_var_sizes, KM_SLEEP);
sa_attr_iter(os, hdr, bonustype, sa_build_idx_tab,
tb, idx_tab);
sa_idx_tab_hold(os, idx_tab); /* one hold for consumer */
sa_idx_tab_hold(os, idx_tab); /* one for layout */
list_insert_tail(&tb->lot_idx_tab, idx_tab);
return (idx_tab);
}
void
sa_default_locator(void **dataptr, uint32_t *len, uint32_t total_len,
boolean_t start, void *userdata)
{
ASSERT(start);
*dataptr = userdata;
*len = total_len;
}
static void
sa_attr_register_sync(sa_handle_t *hdl, dmu_tx_t *tx)
{
uint64_t attr_value = 0;
sa_os_t *sa = hdl->sa_os->os_sa;
sa_attr_table_t *tb = sa->sa_attr_table;
int i;
mutex_enter(&sa->sa_lock);
if (!sa->sa_need_attr_registration || sa->sa_master_obj == 0) {
mutex_exit(&sa->sa_lock);
return;
}
if (sa->sa_reg_attr_obj == 0) {
sa->sa_reg_attr_obj = zap_create_link(hdl->sa_os,
DMU_OT_SA_ATTR_REGISTRATION,
sa->sa_master_obj, SA_REGISTRY, tx);
}
for (i = 0; i != sa->sa_num_attrs; i++) {
if (sa->sa_attr_table[i].sa_registered)
continue;
ATTR_ENCODE(attr_value, tb[i].sa_attr, tb[i].sa_length,
tb[i].sa_byteswap);
VERIFY(0 == zap_update(hdl->sa_os, sa->sa_reg_attr_obj,
tb[i].sa_name, 8, 1, &attr_value, tx));
tb[i].sa_registered = B_TRUE;
}
sa->sa_need_attr_registration = B_FALSE;
mutex_exit(&sa->sa_lock);
}
/*
* Replace all attributes with attributes specified in template.
* If dnode had a spill buffer then those attributes will be
* also be replaced, possibly with just an empty spill block
*
* This interface is intended to only be used for bulk adding of
* attributes for a new file. It will also be used by the ZPL
* when converting and old formatted znode to native SA support.
*/
int
sa_replace_all_by_template_locked(sa_handle_t *hdl, sa_bulk_attr_t *attr_desc,
int attr_count, dmu_tx_t *tx)
{
sa_os_t *sa = hdl->sa_os->os_sa;
if (sa->sa_need_attr_registration)
sa_attr_register_sync(hdl, tx);
return (sa_build_layouts(hdl, attr_desc, attr_count, tx));
}
int
sa_replace_all_by_template(sa_handle_t *hdl, sa_bulk_attr_t *attr_desc,
int attr_count, dmu_tx_t *tx)
{
int error;
mutex_enter(&hdl->sa_lock);
error = sa_replace_all_by_template_locked(hdl, attr_desc,
attr_count, tx);
mutex_exit(&hdl->sa_lock);
return (error);
}
/*
* Add/remove a single attribute or replace a variable-sized attribute value
* with a value of a different size, and then rewrite the entire set
* of attributes.
* Same-length attribute value replacement (including fixed-length attributes)
* is handled more efficiently by the upper layers.
*/
static int
sa_modify_attrs(sa_handle_t *hdl, sa_attr_type_t newattr,
sa_data_op_t action, sa_data_locator_t *locator, void *datastart,
uint16_t buflen, dmu_tx_t *tx)
{
sa_os_t *sa = hdl->sa_os->os_sa;
dmu_buf_impl_t *db = (dmu_buf_impl_t *)hdl->sa_bonus;
dnode_t *dn;
sa_bulk_attr_t *attr_desc;
void *old_data[2];
int bonus_attr_count = 0;
int bonus_data_size = 0;
int spill_data_size = 0;
int spill_attr_count = 0;
int error;
uint16_t length, reg_length;
int i, j, k, length_idx;
sa_hdr_phys_t *hdr;
sa_idx_tab_t *idx_tab;
int attr_count;
int count;
ASSERT(MUTEX_HELD(&hdl->sa_lock));
/* First make of copy of the old data */
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
if (dn->dn_bonuslen != 0) {
bonus_data_size = hdl->sa_bonus->db_size;
old_data[0] = kmem_alloc(bonus_data_size, KM_SLEEP);
bcopy(hdl->sa_bonus->db_data, old_data[0],
hdl->sa_bonus->db_size);
bonus_attr_count = hdl->sa_bonus_tab->sa_layout->lot_attr_count;
} else {
old_data[0] = NULL;
}
DB_DNODE_EXIT(db);
/* Bring spill buffer online if it isn't currently */
if ((error = sa_get_spill(hdl)) == 0) {
spill_data_size = hdl->sa_spill->db_size;
old_data[1] = vmem_alloc(spill_data_size, KM_SLEEP);
bcopy(hdl->sa_spill->db_data, old_data[1],
hdl->sa_spill->db_size);
spill_attr_count =
hdl->sa_spill_tab->sa_layout->lot_attr_count;
} else if (error && error != ENOENT) {
if (old_data[0])
kmem_free(old_data[0], bonus_data_size);
return (error);
} else {
old_data[1] = NULL;
}
/* build descriptor of all attributes */
attr_count = bonus_attr_count + spill_attr_count;
if (action == SA_ADD)
attr_count++;
else if (action == SA_REMOVE)
attr_count--;
attr_desc = kmem_zalloc(sizeof (sa_bulk_attr_t) * attr_count, KM_SLEEP);
/*
* loop through bonus and spill buffer if it exists, and
* build up new attr_descriptor to reset the attributes
*/
k = j = 0;
count = bonus_attr_count;
hdr = SA_GET_HDR(hdl, SA_BONUS);
idx_tab = SA_IDX_TAB_GET(hdl, SA_BONUS);
for (; k != 2; k++) {
/*
* Iterate over each attribute in layout. Fetch the
* size of variable-length attributes needing rewrite
* from sa_lengths[].
*/
for (i = 0, length_idx = 0; i != count; i++) {
sa_attr_type_t attr;
attr = idx_tab->sa_layout->lot_attrs[i];
reg_length = SA_REGISTERED_LEN(sa, attr);
if (reg_length == 0) {
length = hdr->sa_lengths[length_idx];
length_idx++;
} else {
length = reg_length;
}
if (attr == newattr) {
/*
* There is nothing to do for SA_REMOVE,
* so it is just skipped.
*/
if (action == SA_REMOVE)
continue;
/*
* Duplicate attributes are not allowed, so the
* action can not be SA_ADD here.
*/
ASSERT3S(action, ==, SA_REPLACE);
/*
* Only a variable-sized attribute can be
* replaced here, and its size must be changing.
*/
ASSERT3U(reg_length, ==, 0);
ASSERT3U(length, !=, buflen);
SA_ADD_BULK_ATTR(attr_desc, j, attr,
locator, datastart, buflen);
} else {
SA_ADD_BULK_ATTR(attr_desc, j, attr,
NULL, (void *)
(TOC_OFF(idx_tab->sa_idx_tab[attr]) +
(uintptr_t)old_data[k]), length);
}
}
if (k == 0 && hdl->sa_spill) {
hdr = SA_GET_HDR(hdl, SA_SPILL);
idx_tab = SA_IDX_TAB_GET(hdl, SA_SPILL);
count = spill_attr_count;
} else {
break;
}
}
if (action == SA_ADD) {
reg_length = SA_REGISTERED_LEN(sa, newattr);
IMPLY(reg_length != 0, reg_length == buflen);
SA_ADD_BULK_ATTR(attr_desc, j, newattr, locator,
datastart, buflen);
}
ASSERT3U(j, ==, attr_count);
error = sa_build_layouts(hdl, attr_desc, attr_count, tx);
if (old_data[0])
kmem_free(old_data[0], bonus_data_size);
if (old_data[1])
vmem_free(old_data[1], spill_data_size);
kmem_free(attr_desc, sizeof (sa_bulk_attr_t) * attr_count);
return (error);
}
static int
sa_bulk_update_impl(sa_handle_t *hdl, sa_bulk_attr_t *bulk, int count,
dmu_tx_t *tx)
{
int error;
sa_os_t *sa = hdl->sa_os->os_sa;
dmu_object_type_t bonustype;
dmu_buf_t *saved_spill;
ASSERT(hdl);
ASSERT(MUTEX_HELD(&hdl->sa_lock));
bonustype = SA_BONUSTYPE_FROM_DB(SA_GET_DB(hdl, SA_BONUS));
saved_spill = hdl->sa_spill;
/* sync out registration table if necessary */
if (sa->sa_need_attr_registration)
sa_attr_register_sync(hdl, tx);
error = sa_attr_op(hdl, bulk, count, SA_UPDATE, tx);
if (error == 0 && !IS_SA_BONUSTYPE(bonustype) && sa->sa_update_cb)
sa->sa_update_cb(hdl, tx);
/*
* If saved_spill is NULL and current sa_spill is not NULL that
* means we increased the refcount of the spill buffer through
* sa_get_spill() or dmu_spill_hold_by_dnode(). Therefore we
* must release the hold before calling dmu_tx_commit() to avoid
* making a copy of this buffer in dbuf_sync_leaf() due to the
* reference count now being greater than 1.
*/
if (!saved_spill && hdl->sa_spill) {
if (hdl->sa_spill_tab) {
sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);
hdl->sa_spill_tab = NULL;
}
dmu_buf_rele(hdl->sa_spill, NULL);
hdl->sa_spill = NULL;
}
return (error);
}
/*
* update or add new attribute
*/
int
sa_update(sa_handle_t *hdl, sa_attr_type_t type,
void *buf, uint32_t buflen, dmu_tx_t *tx)
{
int error;
sa_bulk_attr_t bulk;
VERIFY3U(buflen, <=, SA_ATTR_MAX_LEN);
bulk.sa_attr = type;
bulk.sa_data_func = NULL;
bulk.sa_length = buflen;
bulk.sa_data = buf;
mutex_enter(&hdl->sa_lock);
error = sa_bulk_update_impl(hdl, &bulk, 1, tx);
mutex_exit(&hdl->sa_lock);
return (error);
}
/*
* Return size of an attribute
*/
int
sa_size(sa_handle_t *hdl, sa_attr_type_t attr, int *size)
{
sa_bulk_attr_t bulk;
int error;
bulk.sa_data = NULL;
bulk.sa_attr = attr;
bulk.sa_data_func = NULL;
ASSERT(hdl);
mutex_enter(&hdl->sa_lock);
if ((error = sa_attr_op(hdl, &bulk, 1, SA_LOOKUP, NULL)) != 0) {
mutex_exit(&hdl->sa_lock);
return (error);
}
*size = bulk.sa_size;
mutex_exit(&hdl->sa_lock);
return (0);
}
int
sa_bulk_lookup_locked(sa_handle_t *hdl, sa_bulk_attr_t *attrs, int count)
{
ASSERT(hdl);
ASSERT(MUTEX_HELD(&hdl->sa_lock));
return (sa_lookup_impl(hdl, attrs, count));
}
int
sa_bulk_lookup(sa_handle_t *hdl, sa_bulk_attr_t *attrs, int count)
{
int error;
ASSERT(hdl);
mutex_enter(&hdl->sa_lock);
error = sa_bulk_lookup_locked(hdl, attrs, count);
mutex_exit(&hdl->sa_lock);
return (error);
}
int
sa_bulk_update(sa_handle_t *hdl, sa_bulk_attr_t *attrs, int count, dmu_tx_t *tx)
{
int error;
ASSERT(hdl);
mutex_enter(&hdl->sa_lock);
error = sa_bulk_update_impl(hdl, attrs, count, tx);
mutex_exit(&hdl->sa_lock);
return (error);
}
int
sa_remove(sa_handle_t *hdl, sa_attr_type_t attr, dmu_tx_t *tx)
{
int error;
mutex_enter(&hdl->sa_lock);
error = sa_modify_attrs(hdl, attr, SA_REMOVE, NULL,
NULL, 0, tx);
mutex_exit(&hdl->sa_lock);
return (error);
}
void
sa_object_info(sa_handle_t *hdl, dmu_object_info_t *doi)
{
dmu_object_info_from_db(hdl->sa_bonus, doi);
}
void
sa_object_size(sa_handle_t *hdl, uint32_t *blksize, u_longlong_t *nblocks)
{
dmu_object_size_from_db(hdl->sa_bonus,
blksize, nblocks);
}
void
sa_set_userp(sa_handle_t *hdl, void *ptr)
{
hdl->sa_userp = ptr;
}
dmu_buf_t *
sa_get_db(sa_handle_t *hdl)
{
return (hdl->sa_bonus);
}
void *
sa_get_userdata(sa_handle_t *hdl)
{
return (hdl->sa_userp);
}
void
sa_register_update_callback_locked(objset_t *os, sa_update_cb_t *func)
{
ASSERT(MUTEX_HELD(&os->os_sa->sa_lock));
os->os_sa->sa_update_cb = func;
}
void
sa_register_update_callback(objset_t *os, sa_update_cb_t *func)
{
mutex_enter(&os->os_sa->sa_lock);
sa_register_update_callback_locked(os, func);
mutex_exit(&os->os_sa->sa_lock);
}
uint64_t
sa_handle_object(sa_handle_t *hdl)
{
return (hdl->sa_bonus->db_object);
}
boolean_t
sa_enabled(objset_t *os)
{
return (os->os_sa == NULL);
}
int
sa_set_sa_object(objset_t *os, uint64_t sa_object)
{
sa_os_t *sa = os->os_sa;
if (sa->sa_master_obj)
return (1);
sa->sa_master_obj = sa_object;
return (0);
}
int
sa_hdrsize(void *arg)
{
sa_hdr_phys_t *hdr = arg;
return (SA_HDR_SIZE(hdr));
}
void
sa_handle_lock(sa_handle_t *hdl)
{
ASSERT(hdl);
mutex_enter(&hdl->sa_lock);
}
void
sa_handle_unlock(sa_handle_t *hdl)
{
ASSERT(hdl);
mutex_exit(&hdl->sa_lock);
}
#ifdef _KERNEL
EXPORT_SYMBOL(sa_handle_get);
EXPORT_SYMBOL(sa_handle_get_from_db);
EXPORT_SYMBOL(sa_handle_destroy);
EXPORT_SYMBOL(sa_buf_hold);
EXPORT_SYMBOL(sa_buf_rele);
EXPORT_SYMBOL(sa_spill_rele);
EXPORT_SYMBOL(sa_lookup);
EXPORT_SYMBOL(sa_update);
EXPORT_SYMBOL(sa_remove);
EXPORT_SYMBOL(sa_bulk_lookup);
EXPORT_SYMBOL(sa_bulk_lookup_locked);
EXPORT_SYMBOL(sa_bulk_update);
EXPORT_SYMBOL(sa_size);
EXPORT_SYMBOL(sa_object_info);
EXPORT_SYMBOL(sa_object_size);
EXPORT_SYMBOL(sa_get_userdata);
EXPORT_SYMBOL(sa_set_userp);
EXPORT_SYMBOL(sa_get_db);
EXPORT_SYMBOL(sa_handle_object);
EXPORT_SYMBOL(sa_register_update_callback);
EXPORT_SYMBOL(sa_setup);
EXPORT_SYMBOL(sa_replace_all_by_template);
EXPORT_SYMBOL(sa_replace_all_by_template_locked);
EXPORT_SYMBOL(sa_enabled);
EXPORT_SYMBOL(sa_cache_init);
EXPORT_SYMBOL(sa_cache_fini);
EXPORT_SYMBOL(sa_set_sa_object);
EXPORT_SYMBOL(sa_hdrsize);
EXPORT_SYMBOL(sa_handle_lock);
EXPORT_SYMBOL(sa_handle_unlock);
EXPORT_SYMBOL(sa_lookup_uio);
EXPORT_SYMBOL(sa_add_projid);
#endif /* _KERNEL */
diff --git a/module/zfs/spa.c b/module/zfs/spa.c
index 3c0316990b31..01af91b7b18f 100644
--- a/module/zfs/spa.c
+++ b/module/zfs/spa.c
@@ -1,9958 +1,9960 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright (c) 2018, Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright 2013 Saso Kiselkov. All rights reserved.
* Copyright (c) 2014 Integros [integros.com]
* Copyright 2016 Toomas Soome <tsoome@me.com>
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
* Copyright 2018 Joyent, Inc.
* Copyright (c) 2017, 2019, Datto Inc. All rights reserved.
* Copyright 2017 Joyent, Inc.
* Copyright (c) 2017, Intel Corporation.
* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
*/
/*
* SPA: Storage Pool Allocator
*
* This file contains all the routines used when modifying on-disk SPA state.
* This includes opening, importing, destroying, exporting a pool, and syncing a
* pool.
*/
#include <sys/zfs_context.h>
#include <sys/fm/fs/zfs.h>
#include <sys/spa_impl.h>
#include <sys/zio.h>
#include <sys/zio_checksum.h>
#include <sys/dmu.h>
#include <sys/dmu_tx.h>
#include <sys/zap.h>
#include <sys/zil.h>
#include <sys/ddt.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_removal.h>
#include <sys/vdev_indirect_mapping.h>
#include <sys/vdev_indirect_births.h>
#include <sys/vdev_initialize.h>
#include <sys/vdev_rebuild.h>
#include <sys/vdev_trim.h>
#include <sys/vdev_disk.h>
#include <sys/vdev_draid.h>
#include <sys/metaslab.h>
#include <sys/metaslab_impl.h>
#include <sys/mmp.h>
#include <sys/uberblock_impl.h>
#include <sys/txg.h>
#include <sys/avl.h>
#include <sys/bpobj.h>
#include <sys/dmu_traverse.h>
#include <sys/dmu_objset.h>
#include <sys/unique.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_synctask.h>
#include <sys/fs/zfs.h>
#include <sys/arc.h>
#include <sys/callb.h>
#include <sys/systeminfo.h>
#include <sys/spa_boot.h>
#include <sys/zfs_ioctl.h>
#include <sys/dsl_scan.h>
#include <sys/zfeature.h>
#include <sys/dsl_destroy.h>
#include <sys/zvol.h>
#ifdef _KERNEL
#include <sys/fm/protocol.h>
#include <sys/fm/util.h>
#include <sys/callb.h>
#include <sys/zone.h>
#include <sys/vmsystm.h>
#endif /* _KERNEL */
#include "zfs_prop.h"
#include "zfs_comutil.h"
/*
* The interval, in seconds, at which failed configuration cache file writes
* should be retried.
*/
int zfs_ccw_retry_interval = 300;
typedef enum zti_modes {
ZTI_MODE_FIXED, /* value is # of threads (min 1) */
ZTI_MODE_BATCH, /* cpu-intensive; value is ignored */
ZTI_MODE_SCALE, /* Taskqs scale with CPUs. */
ZTI_MODE_NULL, /* don't create a taskq */
ZTI_NMODES
} zti_modes_t;
#define ZTI_P(n, q) { ZTI_MODE_FIXED, (n), (q) }
#define ZTI_PCT(n) { ZTI_MODE_ONLINE_PERCENT, (n), 1 }
#define ZTI_BATCH { ZTI_MODE_BATCH, 0, 1 }
#define ZTI_SCALE { ZTI_MODE_SCALE, 0, 1 }
#define ZTI_NULL { ZTI_MODE_NULL, 0, 0 }
#define ZTI_N(n) ZTI_P(n, 1)
#define ZTI_ONE ZTI_N(1)
typedef struct zio_taskq_info {
zti_modes_t zti_mode;
uint_t zti_value;
uint_t zti_count;
} zio_taskq_info_t;
static const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = {
"iss", "iss_h", "int", "int_h"
};
/*
* This table defines the taskq settings for each ZFS I/O type. When
* initializing a pool, we use this table to create an appropriately sized
* taskq. Some operations are low volume and therefore have a small, static
* number of threads assigned to their taskqs using the ZTI_N(#) or ZTI_ONE
* macros. Other operations process a large amount of data; the ZTI_BATCH
* macro causes us to create a taskq oriented for throughput. Some operations
* are so high frequency and short-lived that the taskq itself can become a
* point of lock contention. The ZTI_P(#, #) macro indicates that we need an
* additional degree of parallelism specified by the number of threads per-
* taskq and the number of taskqs; when dispatching an event in this case, the
* particular taskq is chosen at random. ZTI_SCALE is similar to ZTI_BATCH,
* but with number of taskqs also scaling with number of CPUs.
*
* The different taskq priorities are to handle the different contexts (issue
* and interrupt) and then to reserve threads for ZIO_PRIORITY_NOW I/Os that
* need to be handled with minimum delay.
*/
-const zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = {
+static const zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = {
/* ISSUE ISSUE_HIGH INTR INTR_HIGH */
{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* NULL */
{ ZTI_N(8), ZTI_NULL, ZTI_SCALE, ZTI_NULL }, /* READ */
{ ZTI_BATCH, ZTI_N(5), ZTI_SCALE, ZTI_N(5) }, /* WRITE */
{ ZTI_SCALE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* FREE */
{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* CLAIM */
{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* IOCTL */
{ ZTI_N(4), ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* TRIM */
};
static void spa_sync_version(void *arg, dmu_tx_t *tx);
static void spa_sync_props(void *arg, dmu_tx_t *tx);
static boolean_t spa_has_active_shared_spare(spa_t *spa);
static int spa_load_impl(spa_t *spa, spa_import_type_t type, char **ereport);
static void spa_vdev_resilver_done(spa_t *spa);
-uint_t zio_taskq_batch_pct = 80; /* 1 thread per cpu in pset */
-uint_t zio_taskq_batch_tpq; /* threads per taskq */
-boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */
-uint_t zio_taskq_basedc = 80; /* base duty cycle */
+static uint_t zio_taskq_batch_pct = 80; /* 1 thread per cpu in pset */
+static uint_t zio_taskq_batch_tpq; /* threads per taskq */
+static const boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */
+static const uint_t zio_taskq_basedc = 80; /* base duty cycle */
-boolean_t spa_create_process = B_TRUE; /* no process ==> no sysdc */
+static const boolean_t spa_create_process = B_TRUE; /* no process => no sysdc */
/*
* Report any spa_load_verify errors found, but do not fail spa_load.
* This is used by zdb to analyze non-idle pools.
*/
boolean_t spa_load_verify_dryrun = B_FALSE;
/*
* Allow read spacemaps in case of readonly import (spa_mode == SPA_MODE_READ).
* This is used by zdb for spacemaps verification.
*/
boolean_t spa_mode_readable_spacemaps = B_FALSE;
/*
* This (illegal) pool name is used when temporarily importing a spa_t in order
* to get the vdev stats associated with the imported devices.
*/
#define TRYIMPORT_NAME "$import"
/*
* For debugging purposes: print out vdev tree during pool import.
*/
-int spa_load_print_vdev_tree = B_FALSE;
+static int spa_load_print_vdev_tree = B_FALSE;
/*
* A non-zero value for zfs_max_missing_tvds means that we allow importing
* pools with missing top-level vdevs. This is strictly intended for advanced
* pool recovery cases since missing data is almost inevitable. Pools with
* missing devices can only be imported read-only for safety reasons, and their
* fail-mode will be automatically set to "continue".
*
* With 1 missing vdev we should be able to import the pool and mount all
* datasets. User data that was not modified after the missing device has been
* added should be recoverable. This means that snapshots created prior to the
* addition of that device should be completely intact.
*
* With 2 missing vdevs, some datasets may fail to mount since there are
* dataset statistics that are stored as regular metadata. Some data might be
* recoverable if those vdevs were added recently.
*
* With 3 or more missing vdevs, the pool is severely damaged and MOS entries
* may be missing entirely. Chances of data recovery are very low. Note that
* there are also risks of performing an inadvertent rewind as we might be
* missing all the vdevs with the latest uberblocks.
*/
unsigned long zfs_max_missing_tvds = 0;
/*
* The parameters below are similar to zfs_max_missing_tvds but are only
* intended for a preliminary open of the pool with an untrusted config which
* might be incomplete or out-dated.
*
* We are more tolerant for pools opened from a cachefile since we could have
* an out-dated cachefile where a device removal was not registered.
* We could have set the limit arbitrarily high but in the case where devices
* are really missing we would want to return the proper error codes; we chose
* SPA_DVAS_PER_BP - 1 so that some copies of the MOS would still be available
* and we get a chance to retrieve the trusted config.
*/
uint64_t zfs_max_missing_tvds_cachefile = SPA_DVAS_PER_BP - 1;
/*
* In the case where config was assembled by scanning device paths (/dev/dsks
* by default) we are less tolerant since all the existing devices should have
* been detected and we want spa_load to return the right error codes.
*/
uint64_t zfs_max_missing_tvds_scan = 0;
/*
* Debugging aid that pauses spa_sync() towards the end.
*/
-boolean_t zfs_pause_spa_sync = B_FALSE;
+static const boolean_t zfs_pause_spa_sync = B_FALSE;
/*
* Variables to indicate the livelist condense zthr func should wait at certain
* points for the livelist to be removed - used to test condense/destroy races
*/
-int zfs_livelist_condense_zthr_pause = 0;
-int zfs_livelist_condense_sync_pause = 0;
+static int zfs_livelist_condense_zthr_pause = 0;
+static int zfs_livelist_condense_sync_pause = 0;
/*
* Variables to track whether or not condense cancellation has been
* triggered in testing.
*/
-int zfs_livelist_condense_sync_cancel = 0;
-int zfs_livelist_condense_zthr_cancel = 0;
+static int zfs_livelist_condense_sync_cancel = 0;
+static int zfs_livelist_condense_zthr_cancel = 0;
/*
* Variable to track whether or not extra ALLOC blkptrs were added to a
* livelist entry while it was being condensed (caused by the way we track
* remapped blkptrs in dbuf_remap_impl)
*/
-int zfs_livelist_condense_new_alloc = 0;
+static int zfs_livelist_condense_new_alloc = 0;
/*
* ==========================================================================
* SPA properties routines
* ==========================================================================
*/
/*
* Add a (source=src, propname=propval) list to an nvlist.
*/
static void
spa_prop_add_list(nvlist_t *nvl, zpool_prop_t prop, char *strval,
uint64_t intval, zprop_source_t src)
{
const char *propname = zpool_prop_to_name(prop);
nvlist_t *propval;
propval = fnvlist_alloc();
fnvlist_add_uint64(propval, ZPROP_SOURCE, src);
if (strval != NULL)
fnvlist_add_string(propval, ZPROP_VALUE, strval);
else
fnvlist_add_uint64(propval, ZPROP_VALUE, intval);
fnvlist_add_nvlist(nvl, propname, propval);
nvlist_free(propval);
}
/*
* Get property values from the spa configuration.
*/
static void
spa_prop_get_config(spa_t *spa, nvlist_t **nvp)
{
vdev_t *rvd = spa->spa_root_vdev;
dsl_pool_t *pool = spa->spa_dsl_pool;
uint64_t size, alloc, cap, version;
const zprop_source_t src = ZPROP_SRC_NONE;
spa_config_dirent_t *dp;
metaslab_class_t *mc = spa_normal_class(spa);
ASSERT(MUTEX_HELD(&spa->spa_props_lock));
if (rvd != NULL) {
alloc = metaslab_class_get_alloc(mc);
alloc += metaslab_class_get_alloc(spa_special_class(spa));
alloc += metaslab_class_get_alloc(spa_dedup_class(spa));
alloc += metaslab_class_get_alloc(spa_embedded_log_class(spa));
size = metaslab_class_get_space(mc);
size += metaslab_class_get_space(spa_special_class(spa));
size += metaslab_class_get_space(spa_dedup_class(spa));
size += metaslab_class_get_space(spa_embedded_log_class(spa));
spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src);
spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src);
spa_prop_add_list(*nvp, ZPOOL_PROP_ALLOCATED, NULL, alloc, src);
spa_prop_add_list(*nvp, ZPOOL_PROP_FREE, NULL,
size - alloc, src);
spa_prop_add_list(*nvp, ZPOOL_PROP_CHECKPOINT, NULL,
spa->spa_checkpoint_info.sci_dspace, src);
spa_prop_add_list(*nvp, ZPOOL_PROP_FRAGMENTATION, NULL,
metaslab_class_fragmentation(mc), src);
spa_prop_add_list(*nvp, ZPOOL_PROP_EXPANDSZ, NULL,
metaslab_class_expandable_space(mc), src);
spa_prop_add_list(*nvp, ZPOOL_PROP_READONLY, NULL,
(spa_mode(spa) == SPA_MODE_READ), src);
cap = (size == 0) ? 0 : (alloc * 100 / size);
spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src);
spa_prop_add_list(*nvp, ZPOOL_PROP_DEDUPRATIO, NULL,
ddt_get_pool_dedup_ratio(spa), src);
spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL,
rvd->vdev_state, src);
version = spa_version(spa);
if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION)) {
spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL,
version, ZPROP_SRC_DEFAULT);
} else {
spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL,
version, ZPROP_SRC_LOCAL);
}
spa_prop_add_list(*nvp, ZPOOL_PROP_LOAD_GUID,
NULL, spa_load_guid(spa), src);
}
if (pool != NULL) {
/*
* The $FREE directory was introduced in SPA_VERSION_DEADLISTS,
* when opening pools before this version freedir will be NULL.
*/
if (pool->dp_free_dir != NULL) {
spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL,
dsl_dir_phys(pool->dp_free_dir)->dd_used_bytes,
src);
} else {
spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING,
NULL, 0, src);
}
if (pool->dp_leak_dir != NULL) {
spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED, NULL,
dsl_dir_phys(pool->dp_leak_dir)->dd_used_bytes,
src);
} else {
spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED,
NULL, 0, src);
}
}
spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src);
if (spa->spa_comment != NULL) {
spa_prop_add_list(*nvp, ZPOOL_PROP_COMMENT, spa->spa_comment,
0, ZPROP_SRC_LOCAL);
}
if (spa->spa_compatibility != NULL) {
spa_prop_add_list(*nvp, ZPOOL_PROP_COMPATIBILITY,
spa->spa_compatibility, 0, ZPROP_SRC_LOCAL);
}
if (spa->spa_root != NULL)
spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root,
0, ZPROP_SRC_LOCAL);
if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) {
spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
MIN(zfs_max_recordsize, SPA_MAXBLOCKSIZE), ZPROP_SRC_NONE);
} else {
spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
SPA_OLD_MAXBLOCKSIZE, ZPROP_SRC_NONE);
}
if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) {
spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
DNODE_MAX_SIZE, ZPROP_SRC_NONE);
} else {
spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
DNODE_MIN_SIZE, ZPROP_SRC_NONE);
}
if ((dp = list_head(&spa->spa_config_list)) != NULL) {
if (dp->scd_path == NULL) {
spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
"none", 0, ZPROP_SRC_LOCAL);
} else if (strcmp(dp->scd_path, spa_config_path) != 0) {
spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
dp->scd_path, 0, ZPROP_SRC_LOCAL);
}
}
}
/*
* Get zpool property values.
*/
int
spa_prop_get(spa_t *spa, nvlist_t **nvp)
{
objset_t *mos = spa->spa_meta_objset;
zap_cursor_t zc;
zap_attribute_t za;
dsl_pool_t *dp;
int err;
err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP);
if (err)
return (err);
dp = spa_get_dsl(spa);
dsl_pool_config_enter(dp, FTAG);
mutex_enter(&spa->spa_props_lock);
/*
* Get properties from the spa config.
*/
spa_prop_get_config(spa, nvp);
/* If no pool property object, no more prop to get. */
if (mos == NULL || spa->spa_pool_props_object == 0)
goto out;
/*
* Get properties from the MOS pool property object.
*/
for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object);
(err = zap_cursor_retrieve(&zc, &za)) == 0;
zap_cursor_advance(&zc)) {
uint64_t intval = 0;
char *strval = NULL;
zprop_source_t src = ZPROP_SRC_DEFAULT;
zpool_prop_t prop;
if ((prop = zpool_name_to_prop(za.za_name)) == ZPOOL_PROP_INVAL)
continue;
switch (za.za_integer_length) {
case 8:
/* integer property */
if (za.za_first_integer !=
zpool_prop_default_numeric(prop))
src = ZPROP_SRC_LOCAL;
if (prop == ZPOOL_PROP_BOOTFS) {
dsl_dataset_t *ds = NULL;
err = dsl_dataset_hold_obj(dp,
za.za_first_integer, FTAG, &ds);
if (err != 0)
break;
strval = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN,
KM_SLEEP);
dsl_dataset_name(ds, strval);
dsl_dataset_rele(ds, FTAG);
} else {
strval = NULL;
intval = za.za_first_integer;
}
spa_prop_add_list(*nvp, prop, strval, intval, src);
if (strval != NULL)
kmem_free(strval, ZFS_MAX_DATASET_NAME_LEN);
break;
case 1:
/* string property */
strval = kmem_alloc(za.za_num_integers, KM_SLEEP);
err = zap_lookup(mos, spa->spa_pool_props_object,
za.za_name, 1, za.za_num_integers, strval);
if (err) {
kmem_free(strval, za.za_num_integers);
break;
}
spa_prop_add_list(*nvp, prop, strval, 0, src);
kmem_free(strval, za.za_num_integers);
break;
default:
break;
}
}
zap_cursor_fini(&zc);
out:
mutex_exit(&spa->spa_props_lock);
dsl_pool_config_exit(dp, FTAG);
if (err && err != ENOENT) {
nvlist_free(*nvp);
*nvp = NULL;
return (err);
}
return (0);
}
/*
* Validate the given pool properties nvlist and modify the list
* for the property values to be set.
*/
static int
spa_prop_validate(spa_t *spa, nvlist_t *props)
{
nvpair_t *elem;
int error = 0, reset_bootfs = 0;
uint64_t objnum = 0;
boolean_t has_feature = B_FALSE;
elem = NULL;
while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
uint64_t intval;
char *strval, *slash, *check, *fname;
const char *propname = nvpair_name(elem);
zpool_prop_t prop = zpool_name_to_prop(propname);
switch (prop) {
case ZPOOL_PROP_INVAL:
if (!zpool_prop_feature(propname)) {
error = SET_ERROR(EINVAL);
break;
}
/*
* Sanitize the input.
*/
if (nvpair_type(elem) != DATA_TYPE_UINT64) {
error = SET_ERROR(EINVAL);
break;
}
if (nvpair_value_uint64(elem, &intval) != 0) {
error = SET_ERROR(EINVAL);
break;
}
if (intval != 0) {
error = SET_ERROR(EINVAL);
break;
}
fname = strchr(propname, '@') + 1;
if (zfeature_lookup_name(fname, NULL) != 0) {
error = SET_ERROR(EINVAL);
break;
}
has_feature = B_TRUE;
break;
case ZPOOL_PROP_VERSION:
error = nvpair_value_uint64(elem, &intval);
if (!error &&
(intval < spa_version(spa) ||
intval > SPA_VERSION_BEFORE_FEATURES ||
has_feature))
error = SET_ERROR(EINVAL);
break;
case ZPOOL_PROP_DELEGATION:
case ZPOOL_PROP_AUTOREPLACE:
case ZPOOL_PROP_LISTSNAPS:
case ZPOOL_PROP_AUTOEXPAND:
case ZPOOL_PROP_AUTOTRIM:
error = nvpair_value_uint64(elem, &intval);
if (!error && intval > 1)
error = SET_ERROR(EINVAL);
break;
case ZPOOL_PROP_MULTIHOST:
error = nvpair_value_uint64(elem, &intval);
if (!error && intval > 1)
error = SET_ERROR(EINVAL);
if (!error) {
uint32_t hostid = zone_get_hostid(NULL);
if (hostid)
spa->spa_hostid = hostid;
else
error = SET_ERROR(ENOTSUP);
}
break;
case ZPOOL_PROP_BOOTFS:
/*
* If the pool version is less than SPA_VERSION_BOOTFS,
* or the pool is still being created (version == 0),
* the bootfs property cannot be set.
*/
if (spa_version(spa) < SPA_VERSION_BOOTFS) {
error = SET_ERROR(ENOTSUP);
break;
}
/*
* Make sure the vdev config is bootable
*/
if (!vdev_is_bootable(spa->spa_root_vdev)) {
error = SET_ERROR(ENOTSUP);
break;
}
reset_bootfs = 1;
error = nvpair_value_string(elem, &strval);
if (!error) {
objset_t *os;
if (strval == NULL || strval[0] == '\0') {
objnum = zpool_prop_default_numeric(
ZPOOL_PROP_BOOTFS);
break;
}
error = dmu_objset_hold(strval, FTAG, &os);
if (error != 0)
break;
/* Must be ZPL. */
if (dmu_objset_type(os) != DMU_OST_ZFS) {
error = SET_ERROR(ENOTSUP);
} else {
objnum = dmu_objset_id(os);
}
dmu_objset_rele(os, FTAG);
}
break;
case ZPOOL_PROP_FAILUREMODE:
error = nvpair_value_uint64(elem, &intval);
if (!error && intval > ZIO_FAILURE_MODE_PANIC)
error = SET_ERROR(EINVAL);
/*
* This is a special case which only occurs when
* the pool has completely failed. This allows
* the user to change the in-core failmode property
* without syncing it out to disk (I/Os might
* currently be blocked). We do this by returning
* EIO to the caller (spa_prop_set) to trick it
* into thinking we encountered a property validation
* error.
*/
if (!error && spa_suspended(spa)) {
spa->spa_failmode = intval;
error = SET_ERROR(EIO);
}
break;
case ZPOOL_PROP_CACHEFILE:
if ((error = nvpair_value_string(elem, &strval)) != 0)
break;
if (strval[0] == '\0')
break;
if (strcmp(strval, "none") == 0)
break;
if (strval[0] != '/') {
error = SET_ERROR(EINVAL);
break;
}
slash = strrchr(strval, '/');
ASSERT(slash != NULL);
if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
strcmp(slash, "/..") == 0)
error = SET_ERROR(EINVAL);
break;
case ZPOOL_PROP_COMMENT:
if ((error = nvpair_value_string(elem, &strval)) != 0)
break;
for (check = strval; *check != '\0'; check++) {
if (!isprint(*check)) {
error = SET_ERROR(EINVAL);
break;
}
}
if (strlen(strval) > ZPROP_MAX_COMMENT)
error = SET_ERROR(E2BIG);
break;
default:
break;
}
if (error)
break;
}
(void) nvlist_remove_all(props,
zpool_prop_to_name(ZPOOL_PROP_DEDUPDITTO));
if (!error && reset_bootfs) {
error = nvlist_remove(props,
zpool_prop_to_name(ZPOOL_PROP_BOOTFS), DATA_TYPE_STRING);
if (!error) {
error = nvlist_add_uint64(props,
zpool_prop_to_name(ZPOOL_PROP_BOOTFS), objnum);
}
}
return (error);
}
void
spa_configfile_set(spa_t *spa, nvlist_t *nvp, boolean_t need_sync)
{
char *cachefile;
spa_config_dirent_t *dp;
if (nvlist_lookup_string(nvp, zpool_prop_to_name(ZPOOL_PROP_CACHEFILE),
&cachefile) != 0)
return;
dp = kmem_alloc(sizeof (spa_config_dirent_t),
KM_SLEEP);
if (cachefile[0] == '\0')
dp->scd_path = spa_strdup(spa_config_path);
else if (strcmp(cachefile, "none") == 0)
dp->scd_path = NULL;
else
dp->scd_path = spa_strdup(cachefile);
list_insert_head(&spa->spa_config_list, dp);
if (need_sync)
spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
}
int
spa_prop_set(spa_t *spa, nvlist_t *nvp)
{
int error;
nvpair_t *elem = NULL;
boolean_t need_sync = B_FALSE;
if ((error = spa_prop_validate(spa, nvp)) != 0)
return (error);
while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) {
zpool_prop_t prop = zpool_name_to_prop(nvpair_name(elem));
if (prop == ZPOOL_PROP_CACHEFILE ||
prop == ZPOOL_PROP_ALTROOT ||
prop == ZPOOL_PROP_READONLY)
continue;
if (prop == ZPOOL_PROP_VERSION || prop == ZPOOL_PROP_INVAL) {
uint64_t ver = 0;
if (prop == ZPOOL_PROP_VERSION) {
VERIFY(nvpair_value_uint64(elem, &ver) == 0);
} else {
ASSERT(zpool_prop_feature(nvpair_name(elem)));
ver = SPA_VERSION_FEATURES;
need_sync = B_TRUE;
}
/* Save time if the version is already set. */
if (ver == spa_version(spa))
continue;
/*
* In addition to the pool directory object, we might
* create the pool properties object, the features for
* read object, the features for write object, or the
* feature descriptions object.
*/
error = dsl_sync_task(spa->spa_name, NULL,
spa_sync_version, &ver,
6, ZFS_SPACE_CHECK_RESERVED);
if (error)
return (error);
continue;
}
need_sync = B_TRUE;
break;
}
if (need_sync) {
return (dsl_sync_task(spa->spa_name, NULL, spa_sync_props,
nvp, 6, ZFS_SPACE_CHECK_RESERVED));
}
return (0);
}
/*
* If the bootfs property value is dsobj, clear it.
*/
void
spa_prop_clear_bootfs(spa_t *spa, uint64_t dsobj, dmu_tx_t *tx)
{
if (spa->spa_bootfs == dsobj && spa->spa_pool_props_object != 0) {
VERIFY(zap_remove(spa->spa_meta_objset,
spa->spa_pool_props_object,
zpool_prop_to_name(ZPOOL_PROP_BOOTFS), tx) == 0);
spa->spa_bootfs = 0;
}
}
static int
spa_change_guid_check(void *arg, dmu_tx_t *tx)
{
uint64_t *newguid __maybe_unused = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *rvd = spa->spa_root_vdev;
uint64_t vdev_state;
if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
int error = (spa_has_checkpoint(spa)) ?
ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
return (SET_ERROR(error));
}
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
vdev_state = rvd->vdev_state;
spa_config_exit(spa, SCL_STATE, FTAG);
if (vdev_state != VDEV_STATE_HEALTHY)
return (SET_ERROR(ENXIO));
ASSERT3U(spa_guid(spa), !=, *newguid);
return (0);
}
static void
spa_change_guid_sync(void *arg, dmu_tx_t *tx)
{
uint64_t *newguid = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
uint64_t oldguid;
vdev_t *rvd = spa->spa_root_vdev;
oldguid = spa_guid(spa);
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
rvd->vdev_guid = *newguid;
rvd->vdev_guid_sum += (*newguid - oldguid);
vdev_config_dirty(rvd);
spa_config_exit(spa, SCL_STATE, FTAG);
spa_history_log_internal(spa, "guid change", tx, "old=%llu new=%llu",
(u_longlong_t)oldguid, (u_longlong_t)*newguid);
}
/*
* Change the GUID for the pool. This is done so that we can later
* re-import a pool built from a clone of our own vdevs. We will modify
* the root vdev's guid, our own pool guid, and then mark all of our
* vdevs dirty. Note that we must make sure that all our vdevs are
* online when we do this, or else any vdevs that weren't present
* would be orphaned from our pool. We are also going to issue a
* sysevent to update any watchers.
*/
int
spa_change_guid(spa_t *spa)
{
int error;
uint64_t guid;
mutex_enter(&spa->spa_vdev_top_lock);
mutex_enter(&spa_namespace_lock);
guid = spa_generate_guid(NULL);
error = dsl_sync_task(spa->spa_name, spa_change_guid_check,
spa_change_guid_sync, &guid, 5, ZFS_SPACE_CHECK_RESERVED);
if (error == 0) {
spa_write_cachefile(spa, B_FALSE, B_TRUE);
spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_REGUID);
}
mutex_exit(&spa_namespace_lock);
mutex_exit(&spa->spa_vdev_top_lock);
return (error);
}
/*
* ==========================================================================
* SPA state manipulation (open/create/destroy/import/export)
* ==========================================================================
*/
static int
spa_error_entry_compare(const void *a, const void *b)
{
const spa_error_entry_t *sa = (const spa_error_entry_t *)a;
const spa_error_entry_t *sb = (const spa_error_entry_t *)b;
int ret;
ret = memcmp(&sa->se_bookmark, &sb->se_bookmark,
sizeof (zbookmark_phys_t));
return (TREE_ISIGN(ret));
}
/*
* Utility function which retrieves copies of the current logs and
* re-initializes them in the process.
*/
void
spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub)
{
ASSERT(MUTEX_HELD(&spa->spa_errlist_lock));
bcopy(&spa->spa_errlist_last, last, sizeof (avl_tree_t));
bcopy(&spa->spa_errlist_scrub, scrub, sizeof (avl_tree_t));
avl_create(&spa->spa_errlist_scrub,
spa_error_entry_compare, sizeof (spa_error_entry_t),
offsetof(spa_error_entry_t, se_avl));
avl_create(&spa->spa_errlist_last,
spa_error_entry_compare, sizeof (spa_error_entry_t),
offsetof(spa_error_entry_t, se_avl));
}
static void
spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
{
const zio_taskq_info_t *ztip = &zio_taskqs[t][q];
enum zti_modes mode = ztip->zti_mode;
uint_t value = ztip->zti_value;
uint_t count = ztip->zti_count;
spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
uint_t cpus, flags = TASKQ_DYNAMIC;
boolean_t batch = B_FALSE;
switch (mode) {
case ZTI_MODE_FIXED:
ASSERT3U(value, >, 0);
break;
case ZTI_MODE_BATCH:
batch = B_TRUE;
flags |= TASKQ_THREADS_CPU_PCT;
value = MIN(zio_taskq_batch_pct, 100);
break;
case ZTI_MODE_SCALE:
flags |= TASKQ_THREADS_CPU_PCT;
/*
* We want more taskqs to reduce lock contention, but we want
* less for better request ordering and CPU utilization.
*/
cpus = MAX(1, boot_ncpus * zio_taskq_batch_pct / 100);
if (zio_taskq_batch_tpq > 0) {
count = MAX(1, (cpus + zio_taskq_batch_tpq / 2) /
zio_taskq_batch_tpq);
} else {
/*
* Prefer 6 threads per taskq, but no more taskqs
* than threads in them on large systems. For 80%:
*
* taskq taskq total
* cpus taskqs percent threads threads
* ------- ------- ------- ------- -------
* 1 1 80% 1 1
* 2 1 80% 1 1
* 4 1 80% 3 3
* 8 2 40% 3 6
* 16 3 27% 4 12
* 32 5 16% 5 25
* 64 7 11% 7 49
* 128 10 8% 10 100
* 256 14 6% 15 210
*/
count = 1 + cpus / 6;
while (count * count > cpus)
count--;
}
/* Limit each taskq within 100% to not trigger assertion. */
count = MAX(count, (zio_taskq_batch_pct + 99) / 100);
value = (zio_taskq_batch_pct + count / 2) / count;
break;
case ZTI_MODE_NULL:
tqs->stqs_count = 0;
tqs->stqs_taskq = NULL;
return;
default:
panic("unrecognized mode for %s_%s taskq (%u:%u) in "
"spa_activate()",
zio_type_name[t], zio_taskq_types[q], mode, value);
break;
}
ASSERT3U(count, >, 0);
tqs->stqs_count = count;
tqs->stqs_taskq = kmem_alloc(count * sizeof (taskq_t *), KM_SLEEP);
for (uint_t i = 0; i < count; i++) {
taskq_t *tq;
char name[32];
if (count > 1)
(void) snprintf(name, sizeof (name), "%s_%s_%u",
zio_type_name[t], zio_taskq_types[q], i);
else
(void) snprintf(name, sizeof (name), "%s_%s",
zio_type_name[t], zio_taskq_types[q]);
if (zio_taskq_sysdc && spa->spa_proc != &p0) {
if (batch)
flags |= TASKQ_DC_BATCH;
+ (void) zio_taskq_basedc;
tq = taskq_create_sysdc(name, value, 50, INT_MAX,
spa->spa_proc, zio_taskq_basedc, flags);
} else {
pri_t pri = maxclsyspri;
/*
* The write issue taskq can be extremely CPU
* intensive. Run it at slightly less important
* priority than the other taskqs.
*
* Under Linux and FreeBSD this means incrementing
* the priority value as opposed to platforms like
* illumos where it should be decremented.
*
* On FreeBSD, if priorities divided by four (RQ_PPQ)
* are equal then a difference between them is
* insignificant.
*/
if (t == ZIO_TYPE_WRITE && q == ZIO_TASKQ_ISSUE) {
#if defined(__linux__)
pri++;
#elif defined(__FreeBSD__)
pri += 4;
#else
#error "unknown OS"
#endif
}
tq = taskq_create_proc(name, value, pri, 50,
INT_MAX, spa->spa_proc, flags);
}
tqs->stqs_taskq[i] = tq;
}
}
static void
spa_taskqs_fini(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
{
spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
if (tqs->stqs_taskq == NULL) {
ASSERT3U(tqs->stqs_count, ==, 0);
return;
}
for (uint_t i = 0; i < tqs->stqs_count; i++) {
ASSERT3P(tqs->stqs_taskq[i], !=, NULL);
taskq_destroy(tqs->stqs_taskq[i]);
}
kmem_free(tqs->stqs_taskq, tqs->stqs_count * sizeof (taskq_t *));
tqs->stqs_taskq = NULL;
}
/*
* Dispatch a task to the appropriate taskq for the ZFS I/O type and priority.
* Note that a type may have multiple discrete taskqs to avoid lock contention
* on the taskq itself. In that case we choose which taskq at random by using
* the low bits of gethrtime().
*/
void
spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent)
{
spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
taskq_t *tq;
ASSERT3P(tqs->stqs_taskq, !=, NULL);
ASSERT3U(tqs->stqs_count, !=, 0);
if (tqs->stqs_count == 1) {
tq = tqs->stqs_taskq[0];
} else {
tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count];
}
taskq_dispatch_ent(tq, func, arg, flags, ent);
}
/*
* Same as spa_taskq_dispatch_ent() but block on the task until completion.
*/
void
spa_taskq_dispatch_sync(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
task_func_t *func, void *arg, uint_t flags)
{
spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
taskq_t *tq;
taskqid_t id;
ASSERT3P(tqs->stqs_taskq, !=, NULL);
ASSERT3U(tqs->stqs_count, !=, 0);
if (tqs->stqs_count == 1) {
tq = tqs->stqs_taskq[0];
} else {
tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count];
}
id = taskq_dispatch(tq, func, arg, flags);
if (id)
taskq_wait_id(tq, id);
}
static void
spa_create_zio_taskqs(spa_t *spa)
{
for (int t = 0; t < ZIO_TYPES; t++) {
for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
spa_taskqs_init(spa, t, q);
}
}
}
/*
* Disabled until spa_thread() can be adapted for Linux.
*/
#undef HAVE_SPA_THREAD
#if defined(_KERNEL) && defined(HAVE_SPA_THREAD)
static void
spa_thread(void *arg)
{
psetid_t zio_taskq_psrset_bind = PS_NONE;
callb_cpr_t cprinfo;
spa_t *spa = arg;
user_t *pu = PTOU(curproc);
CALLB_CPR_INIT(&cprinfo, &spa->spa_proc_lock, callb_generic_cpr,
spa->spa_name);
ASSERT(curproc != &p0);
(void) snprintf(pu->u_psargs, sizeof (pu->u_psargs),
"zpool-%s", spa->spa_name);
(void) strlcpy(pu->u_comm, pu->u_psargs, sizeof (pu->u_comm));
/* bind this thread to the requested psrset */
if (zio_taskq_psrset_bind != PS_NONE) {
pool_lock();
mutex_enter(&cpu_lock);
mutex_enter(&pidlock);
mutex_enter(&curproc->p_lock);
if (cpupart_bind_thread(curthread, zio_taskq_psrset_bind,
0, NULL, NULL) == 0) {
curthread->t_bind_pset = zio_taskq_psrset_bind;
} else {
cmn_err(CE_WARN,
"Couldn't bind process for zfs pool \"%s\" to "
"pset %d\n", spa->spa_name, zio_taskq_psrset_bind);
}
mutex_exit(&curproc->p_lock);
mutex_exit(&pidlock);
mutex_exit(&cpu_lock);
pool_unlock();
}
if (zio_taskq_sysdc) {
sysdc_thread_enter(curthread, 100, 0);
}
spa->spa_proc = curproc;
spa->spa_did = curthread->t_did;
spa_create_zio_taskqs(spa);
mutex_enter(&spa->spa_proc_lock);
ASSERT(spa->spa_proc_state == SPA_PROC_CREATED);
spa->spa_proc_state = SPA_PROC_ACTIVE;
cv_broadcast(&spa->spa_proc_cv);
CALLB_CPR_SAFE_BEGIN(&cprinfo);
while (spa->spa_proc_state == SPA_PROC_ACTIVE)
cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock);
CALLB_CPR_SAFE_END(&cprinfo, &spa->spa_proc_lock);
ASSERT(spa->spa_proc_state == SPA_PROC_DEACTIVATE);
spa->spa_proc_state = SPA_PROC_GONE;
spa->spa_proc = &p0;
cv_broadcast(&spa->spa_proc_cv);
CALLB_CPR_EXIT(&cprinfo); /* drops spa_proc_lock */
mutex_enter(&curproc->p_lock);
lwp_exit();
}
#endif
/*
* Activate an uninitialized pool.
*/
static void
spa_activate(spa_t *spa, spa_mode_t mode)
{
ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);
spa->spa_state = POOL_STATE_ACTIVE;
spa->spa_mode = mode;
spa->spa_read_spacemaps = spa_mode_readable_spacemaps;
- spa->spa_normal_class = metaslab_class_create(spa, zfs_metaslab_ops);
- spa->spa_log_class = metaslab_class_create(spa, zfs_metaslab_ops);
+ spa->spa_normal_class = metaslab_class_create(spa, &zfs_metaslab_ops);
+ spa->spa_log_class = metaslab_class_create(spa, &zfs_metaslab_ops);
spa->spa_embedded_log_class =
- metaslab_class_create(spa, zfs_metaslab_ops);
- spa->spa_special_class = metaslab_class_create(spa, zfs_metaslab_ops);
- spa->spa_dedup_class = metaslab_class_create(spa, zfs_metaslab_ops);
+ metaslab_class_create(spa, &zfs_metaslab_ops);
+ spa->spa_special_class = metaslab_class_create(spa, &zfs_metaslab_ops);
+ spa->spa_dedup_class = metaslab_class_create(spa, &zfs_metaslab_ops);
/* Try to create a covering process */
mutex_enter(&spa->spa_proc_lock);
ASSERT(spa->spa_proc_state == SPA_PROC_NONE);
ASSERT(spa->spa_proc == &p0);
spa->spa_did = 0;
+ (void) spa_create_process;
#ifdef HAVE_SPA_THREAD
/* Only create a process if we're going to be around a while. */
if (spa_create_process && strcmp(spa->spa_name, TRYIMPORT_NAME) != 0) {
if (newproc(spa_thread, (caddr_t)spa, syscid, maxclsyspri,
NULL, 0) == 0) {
spa->spa_proc_state = SPA_PROC_CREATED;
while (spa->spa_proc_state == SPA_PROC_CREATED) {
cv_wait(&spa->spa_proc_cv,
&spa->spa_proc_lock);
}
ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE);
ASSERT(spa->spa_proc != &p0);
ASSERT(spa->spa_did != 0);
} else {
#ifdef _KERNEL
cmn_err(CE_WARN,
"Couldn't create process for zfs pool \"%s\"\n",
spa->spa_name);
#endif
}
}
#endif /* HAVE_SPA_THREAD */
mutex_exit(&spa->spa_proc_lock);
/* If we didn't create a process, we need to create our taskqs. */
if (spa->spa_proc == &p0) {
spa_create_zio_taskqs(spa);
}
for (size_t i = 0; i < TXG_SIZE; i++) {
spa->spa_txg_zio[i] = zio_root(spa, NULL, NULL,
ZIO_FLAG_CANFAIL);
}
list_create(&spa->spa_config_dirty_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_config_dirty_node));
list_create(&spa->spa_evicting_os_list, sizeof (objset_t),
offsetof(objset_t, os_evicting_node));
list_create(&spa->spa_state_dirty_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_state_dirty_node));
txg_list_create(&spa->spa_vdev_txg_list, spa,
offsetof(struct vdev, vdev_txg_node));
avl_create(&spa->spa_errlist_scrub,
spa_error_entry_compare, sizeof (spa_error_entry_t),
offsetof(spa_error_entry_t, se_avl));
avl_create(&spa->spa_errlist_last,
spa_error_entry_compare, sizeof (spa_error_entry_t),
offsetof(spa_error_entry_t, se_avl));
spa_keystore_init(&spa->spa_keystore);
/*
* This taskq is used to perform zvol-minor-related tasks
* asynchronously. This has several advantages, including easy
* resolution of various deadlocks.
*
* The taskq must be single threaded to ensure tasks are always
* processed in the order in which they were dispatched.
*
* A taskq per pool allows one to keep the pools independent.
* This way if one pool is suspended, it will not impact another.
*
* The preferred location to dispatch a zvol minor task is a sync
* task. In this context, there is easy access to the spa_t and minimal
* error handling is required because the sync task must succeed.
*/
spa->spa_zvol_taskq = taskq_create("z_zvol", 1, defclsyspri,
1, INT_MAX, 0);
/*
* Taskq dedicated to prefetcher threads: this is used to prevent the
* pool traverse code from monopolizing the global (and limited)
* system_taskq by inappropriately scheduling long running tasks on it.
*/
spa->spa_prefetch_taskq = taskq_create("z_prefetch", 100,
defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC | TASKQ_THREADS_CPU_PCT);
/*
* The taskq to upgrade datasets in this pool. Currently used by
* feature SPA_FEATURE_USEROBJ_ACCOUNTING/SPA_FEATURE_PROJECT_QUOTA.
*/
spa->spa_upgrade_taskq = taskq_create("z_upgrade", 100,
defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC | TASKQ_THREADS_CPU_PCT);
}
/*
* Opposite of spa_activate().
*/
static void
spa_deactivate(spa_t *spa)
{
ASSERT(spa->spa_sync_on == B_FALSE);
ASSERT(spa->spa_dsl_pool == NULL);
ASSERT(spa->spa_root_vdev == NULL);
ASSERT(spa->spa_async_zio_root == NULL);
ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED);
spa_evicting_os_wait(spa);
if (spa->spa_zvol_taskq) {
taskq_destroy(spa->spa_zvol_taskq);
spa->spa_zvol_taskq = NULL;
}
if (spa->spa_prefetch_taskq) {
taskq_destroy(spa->spa_prefetch_taskq);
spa->spa_prefetch_taskq = NULL;
}
if (spa->spa_upgrade_taskq) {
taskq_destroy(spa->spa_upgrade_taskq);
spa->spa_upgrade_taskq = NULL;
}
txg_list_destroy(&spa->spa_vdev_txg_list);
list_destroy(&spa->spa_config_dirty_list);
list_destroy(&spa->spa_evicting_os_list);
list_destroy(&spa->spa_state_dirty_list);
taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
for (int t = 0; t < ZIO_TYPES; t++) {
for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
spa_taskqs_fini(spa, t, q);
}
}
for (size_t i = 0; i < TXG_SIZE; i++) {
ASSERT3P(spa->spa_txg_zio[i], !=, NULL);
VERIFY0(zio_wait(spa->spa_txg_zio[i]));
spa->spa_txg_zio[i] = NULL;
}
metaslab_class_destroy(spa->spa_normal_class);
spa->spa_normal_class = NULL;
metaslab_class_destroy(spa->spa_log_class);
spa->spa_log_class = NULL;
metaslab_class_destroy(spa->spa_embedded_log_class);
spa->spa_embedded_log_class = NULL;
metaslab_class_destroy(spa->spa_special_class);
spa->spa_special_class = NULL;
metaslab_class_destroy(spa->spa_dedup_class);
spa->spa_dedup_class = NULL;
/*
* If this was part of an import or the open otherwise failed, we may
* still have errors left in the queues. Empty them just in case.
*/
spa_errlog_drain(spa);
avl_destroy(&spa->spa_errlist_scrub);
avl_destroy(&spa->spa_errlist_last);
spa_keystore_fini(&spa->spa_keystore);
spa->spa_state = POOL_STATE_UNINITIALIZED;
mutex_enter(&spa->spa_proc_lock);
if (spa->spa_proc_state != SPA_PROC_NONE) {
ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE);
spa->spa_proc_state = SPA_PROC_DEACTIVATE;
cv_broadcast(&spa->spa_proc_cv);
while (spa->spa_proc_state == SPA_PROC_DEACTIVATE) {
ASSERT(spa->spa_proc != &p0);
cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock);
}
ASSERT(spa->spa_proc_state == SPA_PROC_GONE);
spa->spa_proc_state = SPA_PROC_NONE;
}
ASSERT(spa->spa_proc == &p0);
mutex_exit(&spa->spa_proc_lock);
/*
* We want to make sure spa_thread() has actually exited the ZFS
* module, so that the module can't be unloaded out from underneath
* it.
*/
if (spa->spa_did != 0) {
thread_join(spa->spa_did);
spa->spa_did = 0;
}
}
/*
* Verify a pool configuration, and construct the vdev tree appropriately. This
* will create all the necessary vdevs in the appropriate layout, with each vdev
* in the CLOSED state. This will prep the pool before open/creation/import.
* All vdev validation is done by the vdev_alloc() routine.
*/
int
spa_config_parse(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent,
uint_t id, int atype)
{
nvlist_t **child;
uint_t children;
int error;
if ((error = vdev_alloc(spa, vdp, nv, parent, id, atype)) != 0)
return (error);
if ((*vdp)->vdev_ops->vdev_op_leaf)
return (0);
error = nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children);
if (error == ENOENT)
return (0);
if (error) {
vdev_free(*vdp);
*vdp = NULL;
return (SET_ERROR(EINVAL));
}
for (int c = 0; c < children; c++) {
vdev_t *vd;
if ((error = spa_config_parse(spa, &vd, child[c], *vdp, c,
atype)) != 0) {
vdev_free(*vdp);
*vdp = NULL;
return (error);
}
}
ASSERT(*vdp != NULL);
return (0);
}
static boolean_t
spa_should_flush_logs_on_unload(spa_t *spa)
{
if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))
return (B_FALSE);
if (!spa_writeable(spa))
return (B_FALSE);
if (!spa->spa_sync_on)
return (B_FALSE);
if (spa_state(spa) != POOL_STATE_EXPORTED)
return (B_FALSE);
if (zfs_keep_log_spacemaps_at_export)
return (B_FALSE);
return (B_TRUE);
}
/*
* Opens a transaction that will set the flag that will instruct
* spa_sync to attempt to flush all the metaslabs for that txg.
*/
static void
spa_unload_log_sm_flush_all(spa_t *spa)
{
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
ASSERT3U(spa->spa_log_flushall_txg, ==, 0);
spa->spa_log_flushall_txg = dmu_tx_get_txg(tx);
dmu_tx_commit(tx);
txg_wait_synced(spa_get_dsl(spa), spa->spa_log_flushall_txg);
}
static void
spa_unload_log_sm_metadata(spa_t *spa)
{
void *cookie = NULL;
spa_log_sm_t *sls;
while ((sls = avl_destroy_nodes(&spa->spa_sm_logs_by_txg,
&cookie)) != NULL) {
VERIFY0(sls->sls_mscount);
kmem_free(sls, sizeof (spa_log_sm_t));
}
for (log_summary_entry_t *e = list_head(&spa->spa_log_summary);
e != NULL; e = list_head(&spa->spa_log_summary)) {
VERIFY0(e->lse_mscount);
list_remove(&spa->spa_log_summary, e);
kmem_free(e, sizeof (log_summary_entry_t));
}
spa->spa_unflushed_stats.sus_nblocks = 0;
spa->spa_unflushed_stats.sus_memused = 0;
spa->spa_unflushed_stats.sus_blocklimit = 0;
}
static void
spa_destroy_aux_threads(spa_t *spa)
{
if (spa->spa_condense_zthr != NULL) {
zthr_destroy(spa->spa_condense_zthr);
spa->spa_condense_zthr = NULL;
}
if (spa->spa_checkpoint_discard_zthr != NULL) {
zthr_destroy(spa->spa_checkpoint_discard_zthr);
spa->spa_checkpoint_discard_zthr = NULL;
}
if (spa->spa_livelist_delete_zthr != NULL) {
zthr_destroy(spa->spa_livelist_delete_zthr);
spa->spa_livelist_delete_zthr = NULL;
}
if (spa->spa_livelist_condense_zthr != NULL) {
zthr_destroy(spa->spa_livelist_condense_zthr);
spa->spa_livelist_condense_zthr = NULL;
}
}
/*
* Opposite of spa_load().
*/
static void
spa_unload(spa_t *spa)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
ASSERT(spa_state(spa) != POOL_STATE_UNINITIALIZED);
spa_import_progress_remove(spa_guid(spa));
spa_load_note(spa, "UNLOADING");
spa_wake_waiters(spa);
/*
* If the log space map feature is enabled and the pool is getting
* exported (but not destroyed), we want to spend some time flushing
* as many metaslabs as we can in an attempt to destroy log space
* maps and save import time.
*/
if (spa_should_flush_logs_on_unload(spa))
spa_unload_log_sm_flush_all(spa);
/*
* Stop async tasks.
*/
spa_async_suspend(spa);
if (spa->spa_root_vdev) {
vdev_t *root_vdev = spa->spa_root_vdev;
vdev_initialize_stop_all(root_vdev, VDEV_INITIALIZE_ACTIVE);
vdev_trim_stop_all(root_vdev, VDEV_TRIM_ACTIVE);
vdev_autotrim_stop_all(spa);
vdev_rebuild_stop_all(spa);
}
/*
* Stop syncing.
*/
if (spa->spa_sync_on) {
txg_sync_stop(spa->spa_dsl_pool);
spa->spa_sync_on = B_FALSE;
}
/*
* This ensures that there is no async metaslab prefetching
* while we attempt to unload the spa.
*/
if (spa->spa_root_vdev != NULL) {
for (int c = 0; c < spa->spa_root_vdev->vdev_children; c++) {
vdev_t *vc = spa->spa_root_vdev->vdev_child[c];
if (vc->vdev_mg != NULL)
taskq_wait(vc->vdev_mg->mg_taskq);
}
}
if (spa->spa_mmp.mmp_thread)
mmp_thread_stop(spa);
/*
* Wait for any outstanding async I/O to complete.
*/
if (spa->spa_async_zio_root != NULL) {
for (int i = 0; i < max_ncpus; i++)
(void) zio_wait(spa->spa_async_zio_root[i]);
kmem_free(spa->spa_async_zio_root, max_ncpus * sizeof (void *));
spa->spa_async_zio_root = NULL;
}
if (spa->spa_vdev_removal != NULL) {
spa_vdev_removal_destroy(spa->spa_vdev_removal);
spa->spa_vdev_removal = NULL;
}
spa_destroy_aux_threads(spa);
spa_condense_fini(spa);
bpobj_close(&spa->spa_deferred_bpobj);
spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);
/*
* Close all vdevs.
*/
if (spa->spa_root_vdev)
vdev_free(spa->spa_root_vdev);
ASSERT(spa->spa_root_vdev == NULL);
/*
* Close the dsl pool.
*/
if (spa->spa_dsl_pool) {
dsl_pool_close(spa->spa_dsl_pool);
spa->spa_dsl_pool = NULL;
spa->spa_meta_objset = NULL;
}
ddt_unload(spa);
spa_unload_log_sm_metadata(spa);
/*
* Drop and purge level 2 cache
*/
spa_l2cache_drop(spa);
for (int i = 0; i < spa->spa_spares.sav_count; i++)
vdev_free(spa->spa_spares.sav_vdevs[i]);
if (spa->spa_spares.sav_vdevs) {
kmem_free(spa->spa_spares.sav_vdevs,
spa->spa_spares.sav_count * sizeof (void *));
spa->spa_spares.sav_vdevs = NULL;
}
if (spa->spa_spares.sav_config) {
nvlist_free(spa->spa_spares.sav_config);
spa->spa_spares.sav_config = NULL;
}
spa->spa_spares.sav_count = 0;
for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
vdev_clear_stats(spa->spa_l2cache.sav_vdevs[i]);
vdev_free(spa->spa_l2cache.sav_vdevs[i]);
}
if (spa->spa_l2cache.sav_vdevs) {
kmem_free(spa->spa_l2cache.sav_vdevs,
spa->spa_l2cache.sav_count * sizeof (void *));
spa->spa_l2cache.sav_vdevs = NULL;
}
if (spa->spa_l2cache.sav_config) {
nvlist_free(spa->spa_l2cache.sav_config);
spa->spa_l2cache.sav_config = NULL;
}
spa->spa_l2cache.sav_count = 0;
spa->spa_async_suspended = 0;
spa->spa_indirect_vdevs_loaded = B_FALSE;
if (spa->spa_comment != NULL) {
spa_strfree(spa->spa_comment);
spa->spa_comment = NULL;
}
if (spa->spa_compatibility != NULL) {
spa_strfree(spa->spa_compatibility);
spa->spa_compatibility = NULL;
}
spa_config_exit(spa, SCL_ALL, spa);
}
/*
* Load (or re-load) the current list of vdevs describing the active spares for
* this pool. When this is called, we have some form of basic information in
* 'spa_spares.sav_config'. We parse this into vdevs, try to open them, and
* then re-generate a more complete list including status information.
*/
void
spa_load_spares(spa_t *spa)
{
nvlist_t **spares;
uint_t nspares;
int i;
vdev_t *vd, *tvd;
#ifndef _KERNEL
/*
* zdb opens both the current state of the pool and the
* checkpointed state (if present), with a different spa_t.
*
* As spare vdevs are shared among open pools, we skip loading
* them when we load the checkpointed state of the pool.
*/
if (!spa_writeable(spa))
return;
#endif
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
/*
* First, close and free any existing spare vdevs.
*/
for (i = 0; i < spa->spa_spares.sav_count; i++) {
vd = spa->spa_spares.sav_vdevs[i];
/* Undo the call to spa_activate() below */
if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid,
B_FALSE)) != NULL && tvd->vdev_isspare)
spa_spare_remove(tvd);
vdev_close(vd);
vdev_free(vd);
}
if (spa->spa_spares.sav_vdevs)
kmem_free(spa->spa_spares.sav_vdevs,
spa->spa_spares.sav_count * sizeof (void *));
if (spa->spa_spares.sav_config == NULL)
nspares = 0;
else
VERIFY0(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, &spares, &nspares));
spa->spa_spares.sav_count = (int)nspares;
spa->spa_spares.sav_vdevs = NULL;
if (nspares == 0)
return;
/*
* Construct the array of vdevs, opening them to get status in the
* process. For each spare, there is potentially two different vdev_t
* structures associated with it: one in the list of spares (used only
* for basic validation purposes) and one in the active vdev
* configuration (if it's spared in). During this phase we open and
* validate each vdev on the spare list. If the vdev also exists in the
* active configuration, then we also mark this vdev as an active spare.
*/
spa->spa_spares.sav_vdevs = kmem_zalloc(nspares * sizeof (void *),
KM_SLEEP);
for (i = 0; i < spa->spa_spares.sav_count; i++) {
VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0,
VDEV_ALLOC_SPARE) == 0);
ASSERT(vd != NULL);
spa->spa_spares.sav_vdevs[i] = vd;
if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid,
B_FALSE)) != NULL) {
if (!tvd->vdev_isspare)
spa_spare_add(tvd);
/*
* We only mark the spare active if we were successfully
* able to load the vdev. Otherwise, importing a pool
* with a bad active spare would result in strange
* behavior, because multiple pool would think the spare
* is actively in use.
*
* There is a vulnerability here to an equally bizarre
* circumstance, where a dead active spare is later
* brought back to life (onlined or otherwise). Given
* the rarity of this scenario, and the extra complexity
* it adds, we ignore the possibility.
*/
if (!vdev_is_dead(tvd))
spa_spare_activate(tvd);
}
vd->vdev_top = vd;
vd->vdev_aux = &spa->spa_spares;
if (vdev_open(vd) != 0)
continue;
if (vdev_validate_aux(vd) == 0)
spa_spare_add(vd);
}
/*
* Recompute the stashed list of spares, with status information
* this time.
*/
fnvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES);
spares = kmem_alloc(spa->spa_spares.sav_count * sizeof (void *),
KM_SLEEP);
for (i = 0; i < spa->spa_spares.sav_count; i++)
spares[i] = vdev_config_generate(spa,
spa->spa_spares.sav_vdevs[i], B_TRUE, VDEV_CONFIG_SPARE);
fnvlist_add_nvlist_array(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares,
spa->spa_spares.sav_count);
for (i = 0; i < spa->spa_spares.sav_count; i++)
nvlist_free(spares[i]);
kmem_free(spares, spa->spa_spares.sav_count * sizeof (void *));
}
/*
* Load (or re-load) the current list of vdevs describing the active l2cache for
* this pool. When this is called, we have some form of basic information in
* 'spa_l2cache.sav_config'. We parse this into vdevs, try to open them, and
* then re-generate a more complete list including status information.
* Devices which are already active have their details maintained, and are
* not re-opened.
*/
void
spa_load_l2cache(spa_t *spa)
{
nvlist_t **l2cache = NULL;
uint_t nl2cache;
int i, j, oldnvdevs;
uint64_t guid;
vdev_t *vd, **oldvdevs, **newvdevs;
spa_aux_vdev_t *sav = &spa->spa_l2cache;
#ifndef _KERNEL
/*
* zdb opens both the current state of the pool and the
* checkpointed state (if present), with a different spa_t.
*
* As L2 caches are part of the ARC which is shared among open
* pools, we skip loading them when we load the checkpointed
* state of the pool.
*/
if (!spa_writeable(spa))
return;
#endif
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
oldvdevs = sav->sav_vdevs;
oldnvdevs = sav->sav_count;
sav->sav_vdevs = NULL;
sav->sav_count = 0;
if (sav->sav_config == NULL) {
nl2cache = 0;
newvdevs = NULL;
goto out;
}
VERIFY0(nvlist_lookup_nvlist_array(sav->sav_config,
ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache));
newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_SLEEP);
/*
* Process new nvlist of vdevs.
*/
for (i = 0; i < nl2cache; i++) {
guid = fnvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID);
newvdevs[i] = NULL;
for (j = 0; j < oldnvdevs; j++) {
vd = oldvdevs[j];
if (vd != NULL && guid == vd->vdev_guid) {
/*
* Retain previous vdev for add/remove ops.
*/
newvdevs[i] = vd;
oldvdevs[j] = NULL;
break;
}
}
if (newvdevs[i] == NULL) {
/*
* Create new vdev
*/
VERIFY(spa_config_parse(spa, &vd, l2cache[i], NULL, 0,
VDEV_ALLOC_L2CACHE) == 0);
ASSERT(vd != NULL);
newvdevs[i] = vd;
/*
* Commit this vdev as an l2cache device,
* even if it fails to open.
*/
spa_l2cache_add(vd);
vd->vdev_top = vd;
vd->vdev_aux = sav;
spa_l2cache_activate(vd);
if (vdev_open(vd) != 0)
continue;
(void) vdev_validate_aux(vd);
if (!vdev_is_dead(vd))
l2arc_add_vdev(spa, vd);
/*
* Upon cache device addition to a pool or pool
* creation with a cache device or if the header
* of the device is invalid we issue an async
* TRIM command for the whole device which will
* execute if l2arc_trim_ahead > 0.
*/
spa_async_request(spa, SPA_ASYNC_L2CACHE_TRIM);
}
}
sav->sav_vdevs = newvdevs;
sav->sav_count = (int)nl2cache;
/*
* Recompute the stashed list of l2cache devices, with status
* information this time.
*/
fnvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE);
if (sav->sav_count > 0)
l2cache = kmem_alloc(sav->sav_count * sizeof (void *),
KM_SLEEP);
for (i = 0; i < sav->sav_count; i++)
l2cache[i] = vdev_config_generate(spa,
sav->sav_vdevs[i], B_TRUE, VDEV_CONFIG_L2CACHE);
fnvlist_add_nvlist_array(sav->sav_config, ZPOOL_CONFIG_L2CACHE,
(const nvlist_t * const *)l2cache, sav->sav_count);
out:
/*
* Purge vdevs that were dropped
*/
for (i = 0; i < oldnvdevs; i++) {
uint64_t pool;
vd = oldvdevs[i];
if (vd != NULL) {
ASSERT(vd->vdev_isl2cache);
if (spa_l2cache_exists(vd->vdev_guid, &pool) &&
pool != 0ULL && l2arc_vdev_present(vd))
l2arc_remove_vdev(vd);
vdev_clear_stats(vd);
vdev_free(vd);
}
}
if (oldvdevs)
kmem_free(oldvdevs, oldnvdevs * sizeof (void *));
for (i = 0; i < sav->sav_count; i++)
nvlist_free(l2cache[i]);
if (sav->sav_count)
kmem_free(l2cache, sav->sav_count * sizeof (void *));
}
static int
load_nvlist(spa_t *spa, uint64_t obj, nvlist_t **value)
{
dmu_buf_t *db;
char *packed = NULL;
size_t nvsize = 0;
int error;
*value = NULL;
error = dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db);
if (error)
return (error);
nvsize = *(uint64_t *)db->db_data;
dmu_buf_rele(db, FTAG);
packed = vmem_alloc(nvsize, KM_SLEEP);
error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed,
DMU_READ_PREFETCH);
if (error == 0)
error = nvlist_unpack(packed, nvsize, value, 0);
vmem_free(packed, nvsize);
return (error);
}
/*
* Concrete top-level vdevs that are not missing and are not logs. At every
* spa_sync we write new uberblocks to at least SPA_SYNC_MIN_VDEVS core tvds.
*/
static uint64_t
spa_healthy_core_tvds(spa_t *spa)
{
vdev_t *rvd = spa->spa_root_vdev;
uint64_t tvds = 0;
for (uint64_t i = 0; i < rvd->vdev_children; i++) {
vdev_t *vd = rvd->vdev_child[i];
if (vd->vdev_islog)
continue;
if (vdev_is_concrete(vd) && !vdev_is_dead(vd))
tvds++;
}
return (tvds);
}
/*
* Checks to see if the given vdev could not be opened, in which case we post a
* sysevent to notify the autoreplace code that the device has been removed.
*/
static void
spa_check_removed(vdev_t *vd)
{
for (uint64_t c = 0; c < vd->vdev_children; c++)
spa_check_removed(vd->vdev_child[c]);
if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd) &&
vdev_is_concrete(vd)) {
zfs_post_autoreplace(vd->vdev_spa, vd);
spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_CHECK);
}
}
static int
spa_check_for_missing_logs(spa_t *spa)
{
vdev_t *rvd = spa->spa_root_vdev;
/*
* If we're doing a normal import, then build up any additional
* diagnostic information about missing log devices.
* We'll pass this up to the user for further processing.
*/
if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG)) {
nvlist_t **child, *nv;
uint64_t idx = 0;
child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t *),
KM_SLEEP);
nv = fnvlist_alloc();
for (uint64_t c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
/*
* We consider a device as missing only if it failed
* to open (i.e. offline or faulted is not considered
* as missing).
*/
if (tvd->vdev_islog &&
tvd->vdev_state == VDEV_STATE_CANT_OPEN) {
child[idx++] = vdev_config_generate(spa, tvd,
B_FALSE, VDEV_CONFIG_MISSING);
}
}
if (idx > 0) {
fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
(const nvlist_t * const *)child, idx);
fnvlist_add_nvlist(spa->spa_load_info,
ZPOOL_CONFIG_MISSING_DEVICES, nv);
for (uint64_t i = 0; i < idx; i++)
nvlist_free(child[i]);
}
nvlist_free(nv);
kmem_free(child, rvd->vdev_children * sizeof (char **));
if (idx > 0) {
spa_load_failed(spa, "some log devices are missing");
vdev_dbgmsg_print_tree(rvd, 2);
return (SET_ERROR(ENXIO));
}
} else {
for (uint64_t c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
if (tvd->vdev_islog &&
tvd->vdev_state == VDEV_STATE_CANT_OPEN) {
spa_set_log_state(spa, SPA_LOG_CLEAR);
spa_load_note(spa, "some log devices are "
"missing, ZIL is dropped.");
vdev_dbgmsg_print_tree(rvd, 2);
break;
}
}
}
return (0);
}
/*
* Check for missing log devices
*/
static boolean_t
spa_check_logs(spa_t *spa)
{
boolean_t rv = B_FALSE;
dsl_pool_t *dp = spa_get_dsl(spa);
switch (spa->spa_log_state) {
default:
break;
case SPA_LOG_MISSING:
/* need to recheck in case slog has been restored */
case SPA_LOG_UNKNOWN:
rv = (dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
zil_check_log_chain, NULL, DS_FIND_CHILDREN) != 0);
if (rv)
spa_set_log_state(spa, SPA_LOG_MISSING);
break;
}
return (rv);
}
/*
* Passivate any log vdevs (note, does not apply to embedded log metaslabs).
*/
static boolean_t
spa_passivate_log(spa_t *spa)
{
vdev_t *rvd = spa->spa_root_vdev;
boolean_t slog_found = B_FALSE;
ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER));
for (int c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
if (tvd->vdev_islog) {
ASSERT3P(tvd->vdev_log_mg, ==, NULL);
metaslab_group_passivate(tvd->vdev_mg);
slog_found = B_TRUE;
}
}
return (slog_found);
}
/*
* Activate any log vdevs (note, does not apply to embedded log metaslabs).
*/
static void
spa_activate_log(spa_t *spa)
{
vdev_t *rvd = spa->spa_root_vdev;
ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER));
for (int c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
if (tvd->vdev_islog) {
ASSERT3P(tvd->vdev_log_mg, ==, NULL);
metaslab_group_activate(tvd->vdev_mg);
}
}
}
int
spa_reset_logs(spa_t *spa)
{
int error;
error = dmu_objset_find(spa_name(spa), zil_reset,
NULL, DS_FIND_CHILDREN);
if (error == 0) {
/*
* We successfully offlined the log device, sync out the
* current txg so that the "stubby" block can be removed
* by zil_sync().
*/
txg_wait_synced(spa->spa_dsl_pool, 0);
}
return (error);
}
static void
spa_aux_check_removed(spa_aux_vdev_t *sav)
{
for (int i = 0; i < sav->sav_count; i++)
spa_check_removed(sav->sav_vdevs[i]);
}
void
spa_claim_notify(zio_t *zio)
{
spa_t *spa = zio->io_spa;
if (zio->io_error)
return;
mutex_enter(&spa->spa_props_lock); /* any mutex will do */
if (spa->spa_claim_max_txg < zio->io_bp->blk_birth)
spa->spa_claim_max_txg = zio->io_bp->blk_birth;
mutex_exit(&spa->spa_props_lock);
}
typedef struct spa_load_error {
uint64_t sle_meta_count;
uint64_t sle_data_count;
} spa_load_error_t;
static void
spa_load_verify_done(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
spa_load_error_t *sle = zio->io_private;
dmu_object_type_t type = BP_GET_TYPE(bp);
int error = zio->io_error;
spa_t *spa = zio->io_spa;
abd_free(zio->io_abd);
if (error) {
if ((BP_GET_LEVEL(bp) != 0 || DMU_OT_IS_METADATA(type)) &&
type != DMU_OT_INTENT_LOG)
atomic_inc_64(&sle->sle_meta_count);
else
atomic_inc_64(&sle->sle_data_count);
}
mutex_enter(&spa->spa_scrub_lock);
spa->spa_load_verify_bytes -= BP_GET_PSIZE(bp);
cv_broadcast(&spa->spa_scrub_io_cv);
mutex_exit(&spa->spa_scrub_lock);
}
/*
* Maximum number of inflight bytes is the log2 fraction of the arc size.
* By default, we set it to 1/16th of the arc.
*/
-int spa_load_verify_shift = 4;
-int spa_load_verify_metadata = B_TRUE;
-int spa_load_verify_data = B_TRUE;
+static int spa_load_verify_shift = 4;
+static int spa_load_verify_metadata = B_TRUE;
+static int spa_load_verify_data = B_TRUE;
static int
spa_load_verify_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
const zbookmark_phys_t *zb, const dnode_phys_t *dnp, void *arg)
{
(void) zilog, (void) dnp;
if (zb->zb_level == ZB_DNODE_LEVEL || BP_IS_HOLE(bp) ||
BP_IS_EMBEDDED(bp) || BP_IS_REDACTED(bp))
return (0);
/*
* Note: normally this routine will not be called if
* spa_load_verify_metadata is not set. However, it may be useful
* to manually set the flag after the traversal has begun.
*/
if (!spa_load_verify_metadata)
return (0);
if (!BP_IS_METADATA(bp) && !spa_load_verify_data)
return (0);
uint64_t maxinflight_bytes =
arc_target_bytes() >> spa_load_verify_shift;
zio_t *rio = arg;
size_t size = BP_GET_PSIZE(bp);
mutex_enter(&spa->spa_scrub_lock);
while (spa->spa_load_verify_bytes >= maxinflight_bytes)
cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
spa->spa_load_verify_bytes += size;
mutex_exit(&spa->spa_scrub_lock);
zio_nowait(zio_read(rio, spa, bp, abd_alloc_for_io(size, B_FALSE), size,
spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB,
ZIO_FLAG_SPECULATIVE | ZIO_FLAG_CANFAIL |
ZIO_FLAG_SCRUB | ZIO_FLAG_RAW, zb));
return (0);
}
static int
verify_dataset_name_len(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg)
{
(void) dp, (void) arg;
if (dsl_dataset_namelen(ds) >= ZFS_MAX_DATASET_NAME_LEN)
return (SET_ERROR(ENAMETOOLONG));
return (0);
}
static int
spa_load_verify(spa_t *spa)
{
zio_t *rio;
spa_load_error_t sle = { 0 };
zpool_load_policy_t policy;
boolean_t verify_ok = B_FALSE;
int error = 0;
zpool_get_load_policy(spa->spa_config, &policy);
if (policy.zlp_rewind & ZPOOL_NEVER_REWIND)
return (0);
dsl_pool_config_enter(spa->spa_dsl_pool, FTAG);
error = dmu_objset_find_dp(spa->spa_dsl_pool,
spa->spa_dsl_pool->dp_root_dir_obj, verify_dataset_name_len, NULL,
DS_FIND_CHILDREN);
dsl_pool_config_exit(spa->spa_dsl_pool, FTAG);
if (error != 0)
return (error);
rio = zio_root(spa, NULL, &sle,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE);
if (spa_load_verify_metadata) {
if (spa->spa_extreme_rewind) {
spa_load_note(spa, "performing a complete scan of the "
"pool since extreme rewind is on. This may take "
"a very long time.\n (spa_load_verify_data=%u, "
"spa_load_verify_metadata=%u)",
spa_load_verify_data, spa_load_verify_metadata);
}
error = traverse_pool(spa, spa->spa_verify_min_txg,
TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA |
TRAVERSE_NO_DECRYPT, spa_load_verify_cb, rio);
}
(void) zio_wait(rio);
ASSERT0(spa->spa_load_verify_bytes);
spa->spa_load_meta_errors = sle.sle_meta_count;
spa->spa_load_data_errors = sle.sle_data_count;
if (sle.sle_meta_count != 0 || sle.sle_data_count != 0) {
spa_load_note(spa, "spa_load_verify found %llu metadata errors "
"and %llu data errors", (u_longlong_t)sle.sle_meta_count,
(u_longlong_t)sle.sle_data_count);
}
if (spa_load_verify_dryrun ||
(!error && sle.sle_meta_count <= policy.zlp_maxmeta &&
sle.sle_data_count <= policy.zlp_maxdata)) {
int64_t loss = 0;
verify_ok = B_TRUE;
spa->spa_load_txg = spa->spa_uberblock.ub_txg;
spa->spa_load_txg_ts = spa->spa_uberblock.ub_timestamp;
loss = spa->spa_last_ubsync_txg_ts - spa->spa_load_txg_ts;
fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_LOAD_TIME,
spa->spa_load_txg_ts);
fnvlist_add_int64(spa->spa_load_info, ZPOOL_CONFIG_REWIND_TIME,
loss);
fnvlist_add_uint64(spa->spa_load_info,
ZPOOL_CONFIG_LOAD_DATA_ERRORS, sle.sle_data_count);
} else {
spa->spa_load_max_txg = spa->spa_uberblock.ub_txg;
}
if (spa_load_verify_dryrun)
return (0);
if (error) {
if (error != ENXIO && error != EIO)
error = SET_ERROR(EIO);
return (error);
}
return (verify_ok ? 0 : EIO);
}
/*
* Find a value in the pool props object.
*/
static void
spa_prop_find(spa_t *spa, zpool_prop_t prop, uint64_t *val)
{
(void) zap_lookup(spa->spa_meta_objset, spa->spa_pool_props_object,
zpool_prop_to_name(prop), sizeof (uint64_t), 1, val);
}
/*
* Find a value in the pool directory object.
*/
static int
spa_dir_prop(spa_t *spa, const char *name, uint64_t *val, boolean_t log_enoent)
{
int error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
name, sizeof (uint64_t), 1, val);
if (error != 0 && (error != ENOENT || log_enoent)) {
spa_load_failed(spa, "couldn't get '%s' value in MOS directory "
"[error=%d]", name, error);
}
return (error);
}
static int
spa_vdev_err(vdev_t *vdev, vdev_aux_t aux, int err)
{
vdev_set_state(vdev, B_TRUE, VDEV_STATE_CANT_OPEN, aux);
return (SET_ERROR(err));
}
boolean_t
spa_livelist_delete_check(spa_t *spa)
{
return (spa->spa_livelists_to_delete != 0);
}
static boolean_t
spa_livelist_delete_cb_check(void *arg, zthr_t *z)
{
(void) z;
spa_t *spa = arg;
return (spa_livelist_delete_check(spa));
}
static int
delete_blkptr_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
{
spa_t *spa = arg;
zio_free(spa, tx->tx_txg, bp);
dsl_dir_diduse_space(tx->tx_pool->dp_free_dir, DD_USED_HEAD,
-bp_get_dsize_sync(spa, bp),
-BP_GET_PSIZE(bp), -BP_GET_UCSIZE(bp), tx);
return (0);
}
static int
dsl_get_next_livelist_obj(objset_t *os, uint64_t zap_obj, uint64_t *llp)
{
int err;
zap_cursor_t zc;
zap_attribute_t za;
zap_cursor_init(&zc, os, zap_obj);
err = zap_cursor_retrieve(&zc, &za);
zap_cursor_fini(&zc);
if (err == 0)
*llp = za.za_first_integer;
return (err);
}
/*
* Components of livelist deletion that must be performed in syncing
* context: freeing block pointers and updating the pool-wide data
* structures to indicate how much work is left to do
*/
typedef struct sublist_delete_arg {
spa_t *spa;
dsl_deadlist_t *ll;
uint64_t key;
bplist_t *to_free;
} sublist_delete_arg_t;
static void
sublist_delete_sync(void *arg, dmu_tx_t *tx)
{
sublist_delete_arg_t *sda = arg;
spa_t *spa = sda->spa;
dsl_deadlist_t *ll = sda->ll;
uint64_t key = sda->key;
bplist_t *to_free = sda->to_free;
bplist_iterate(to_free, delete_blkptr_cb, spa, tx);
dsl_deadlist_remove_entry(ll, key, tx);
}
typedef struct livelist_delete_arg {
spa_t *spa;
uint64_t ll_obj;
uint64_t zap_obj;
} livelist_delete_arg_t;
static void
livelist_delete_sync(void *arg, dmu_tx_t *tx)
{
livelist_delete_arg_t *lda = arg;
spa_t *spa = lda->spa;
uint64_t ll_obj = lda->ll_obj;
uint64_t zap_obj = lda->zap_obj;
objset_t *mos = spa->spa_meta_objset;
uint64_t count;
/* free the livelist and decrement the feature count */
VERIFY0(zap_remove_int(mos, zap_obj, ll_obj, tx));
dsl_deadlist_free(mos, ll_obj, tx);
spa_feature_decr(spa, SPA_FEATURE_LIVELIST, tx);
VERIFY0(zap_count(mos, zap_obj, &count));
if (count == 0) {
/* no more livelists to delete */
VERIFY0(zap_remove(mos, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_DELETED_CLONES, tx));
VERIFY0(zap_destroy(mos, zap_obj, tx));
spa->spa_livelists_to_delete = 0;
spa_notify_waiters(spa);
}
}
/*
* Load in the value for the livelist to be removed and open it. Then,
* load its first sublist and determine which block pointers should actually
* be freed. Then, call a synctask which performs the actual frees and updates
* the pool-wide livelist data.
*/
static void
spa_livelist_delete_cb(void *arg, zthr_t *z)
{
spa_t *spa = arg;
uint64_t ll_obj = 0, count;
objset_t *mos = spa->spa_meta_objset;
uint64_t zap_obj = spa->spa_livelists_to_delete;
/*
* Determine the next livelist to delete. This function should only
* be called if there is at least one deleted clone.
*/
VERIFY0(dsl_get_next_livelist_obj(mos, zap_obj, &ll_obj));
VERIFY0(zap_count(mos, ll_obj, &count));
if (count > 0) {
dsl_deadlist_t *ll;
dsl_deadlist_entry_t *dle;
bplist_t to_free;
ll = kmem_zalloc(sizeof (dsl_deadlist_t), KM_SLEEP);
dsl_deadlist_open(ll, mos, ll_obj);
dle = dsl_deadlist_first(ll);
ASSERT3P(dle, !=, NULL);
bplist_create(&to_free);
int err = dsl_process_sub_livelist(&dle->dle_bpobj, &to_free,
z, NULL);
if (err == 0) {
sublist_delete_arg_t sync_arg = {
.spa = spa,
.ll = ll,
.key = dle->dle_mintxg,
.to_free = &to_free
};
zfs_dbgmsg("deleting sublist (id %llu) from"
" livelist %llu, %lld remaining",
(u_longlong_t)dle->dle_bpobj.bpo_object,
(u_longlong_t)ll_obj, (longlong_t)count - 1);
VERIFY0(dsl_sync_task(spa_name(spa), NULL,
sublist_delete_sync, &sync_arg, 0,
ZFS_SPACE_CHECK_DESTROY));
} else {
VERIFY3U(err, ==, EINTR);
}
bplist_clear(&to_free);
bplist_destroy(&to_free);
dsl_deadlist_close(ll);
kmem_free(ll, sizeof (dsl_deadlist_t));
} else {
livelist_delete_arg_t sync_arg = {
.spa = spa,
.ll_obj = ll_obj,
.zap_obj = zap_obj
};
zfs_dbgmsg("deletion of livelist %llu completed",
(u_longlong_t)ll_obj);
VERIFY0(dsl_sync_task(spa_name(spa), NULL, livelist_delete_sync,
&sync_arg, 0, ZFS_SPACE_CHECK_DESTROY));
}
}
static void
spa_start_livelist_destroy_thread(spa_t *spa)
{
ASSERT3P(spa->spa_livelist_delete_zthr, ==, NULL);
spa->spa_livelist_delete_zthr =
zthr_create("z_livelist_destroy",
spa_livelist_delete_cb_check, spa_livelist_delete_cb, spa,
minclsyspri);
}
typedef struct livelist_new_arg {
bplist_t *allocs;
bplist_t *frees;
} livelist_new_arg_t;
static int
livelist_track_new_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed,
dmu_tx_t *tx)
{
ASSERT(tx == NULL);
livelist_new_arg_t *lna = arg;
if (bp_freed) {
bplist_append(lna->frees, bp);
} else {
bplist_append(lna->allocs, bp);
zfs_livelist_condense_new_alloc++;
}
return (0);
}
typedef struct livelist_condense_arg {
spa_t *spa;
bplist_t to_keep;
uint64_t first_size;
uint64_t next_size;
} livelist_condense_arg_t;
static void
spa_livelist_condense_sync(void *arg, dmu_tx_t *tx)
{
livelist_condense_arg_t *lca = arg;
spa_t *spa = lca->spa;
bplist_t new_frees;
dsl_dataset_t *ds = spa->spa_to_condense.ds;
/* Have we been cancelled? */
if (spa->spa_to_condense.cancelled) {
zfs_livelist_condense_sync_cancel++;
goto out;
}
dsl_deadlist_entry_t *first = spa->spa_to_condense.first;
dsl_deadlist_entry_t *next = spa->spa_to_condense.next;
dsl_deadlist_t *ll = &ds->ds_dir->dd_livelist;
/*
* It's possible that the livelist was changed while the zthr was
* running. Therefore, we need to check for new blkptrs in the two
* entries being condensed and continue to track them in the livelist.
* Because of the way we handle remapped blkptrs (see dbuf_remap_impl),
* it's possible that the newly added blkptrs are FREEs or ALLOCs so
* we need to sort them into two different bplists.
*/
uint64_t first_obj = first->dle_bpobj.bpo_object;
uint64_t next_obj = next->dle_bpobj.bpo_object;
uint64_t cur_first_size = first->dle_bpobj.bpo_phys->bpo_num_blkptrs;
uint64_t cur_next_size = next->dle_bpobj.bpo_phys->bpo_num_blkptrs;
bplist_create(&new_frees);
livelist_new_arg_t new_bps = {
.allocs = &lca->to_keep,
.frees = &new_frees,
};
if (cur_first_size > lca->first_size) {
VERIFY0(livelist_bpobj_iterate_from_nofree(&first->dle_bpobj,
livelist_track_new_cb, &new_bps, lca->first_size));
}
if (cur_next_size > lca->next_size) {
VERIFY0(livelist_bpobj_iterate_from_nofree(&next->dle_bpobj,
livelist_track_new_cb, &new_bps, lca->next_size));
}
dsl_deadlist_clear_entry(first, ll, tx);
ASSERT(bpobj_is_empty(&first->dle_bpobj));
dsl_deadlist_remove_entry(ll, next->dle_mintxg, tx);
bplist_iterate(&lca->to_keep, dsl_deadlist_insert_alloc_cb, ll, tx);
bplist_iterate(&new_frees, dsl_deadlist_insert_free_cb, ll, tx);
bplist_destroy(&new_frees);
char dsname[ZFS_MAX_DATASET_NAME_LEN];
dsl_dataset_name(ds, dsname);
zfs_dbgmsg("txg %llu condensing livelist of %s (id %llu), bpobj %llu "
"(%llu blkptrs) and bpobj %llu (%llu blkptrs) -> bpobj %llu "
"(%llu blkptrs)", (u_longlong_t)tx->tx_txg, dsname,
(u_longlong_t)ds->ds_object, (u_longlong_t)first_obj,
(u_longlong_t)cur_first_size, (u_longlong_t)next_obj,
(u_longlong_t)cur_next_size,
(u_longlong_t)first->dle_bpobj.bpo_object,
(u_longlong_t)first->dle_bpobj.bpo_phys->bpo_num_blkptrs);
out:
dmu_buf_rele(ds->ds_dbuf, spa);
spa->spa_to_condense.ds = NULL;
bplist_clear(&lca->to_keep);
bplist_destroy(&lca->to_keep);
kmem_free(lca, sizeof (livelist_condense_arg_t));
spa->spa_to_condense.syncing = B_FALSE;
}
static void
spa_livelist_condense_cb(void *arg, zthr_t *t)
{
while (zfs_livelist_condense_zthr_pause &&
!(zthr_has_waiters(t) || zthr_iscancelled(t)))
delay(1);
spa_t *spa = arg;
dsl_deadlist_entry_t *first = spa->spa_to_condense.first;
dsl_deadlist_entry_t *next = spa->spa_to_condense.next;
uint64_t first_size, next_size;
livelist_condense_arg_t *lca =
kmem_alloc(sizeof (livelist_condense_arg_t), KM_SLEEP);
bplist_create(&lca->to_keep);
/*
* Process the livelists (matching FREEs and ALLOCs) in open context
* so we have minimal work in syncing context to condense.
*
* We save bpobj sizes (first_size and next_size) to use later in
* syncing context to determine if entries were added to these sublists
* while in open context. This is possible because the clone is still
* active and open for normal writes and we want to make sure the new,
* unprocessed blockpointers are inserted into the livelist normally.
*
* Note that dsl_process_sub_livelist() both stores the size number of
* blockpointers and iterates over them while the bpobj's lock held, so
* the sizes returned to us are consistent which what was actually
* processed.
*/
int err = dsl_process_sub_livelist(&first->dle_bpobj, &lca->to_keep, t,
&first_size);
if (err == 0)
err = dsl_process_sub_livelist(&next->dle_bpobj, &lca->to_keep,
t, &next_size);
if (err == 0) {
while (zfs_livelist_condense_sync_pause &&
!(zthr_has_waiters(t) || zthr_iscancelled(t)))
delay(1);
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
dmu_tx_mark_netfree(tx);
dmu_tx_hold_space(tx, 1);
err = dmu_tx_assign(tx, TXG_NOWAIT | TXG_NOTHROTTLE);
if (err == 0) {
/*
* Prevent the condense zthr restarting before
* the synctask completes.
*/
spa->spa_to_condense.syncing = B_TRUE;
lca->spa = spa;
lca->first_size = first_size;
lca->next_size = next_size;
dsl_sync_task_nowait(spa_get_dsl(spa),
spa_livelist_condense_sync, lca, tx);
dmu_tx_commit(tx);
return;
}
}
/*
* Condensing can not continue: either it was externally stopped or
* we were unable to assign to a tx because the pool has run out of
* space. In the second case, we'll just end up trying to condense
* again in a later txg.
*/
ASSERT(err != 0);
bplist_clear(&lca->to_keep);
bplist_destroy(&lca->to_keep);
kmem_free(lca, sizeof (livelist_condense_arg_t));
dmu_buf_rele(spa->spa_to_condense.ds->ds_dbuf, spa);
spa->spa_to_condense.ds = NULL;
if (err == EINTR)
zfs_livelist_condense_zthr_cancel++;
}
/*
* Check that there is something to condense but that a condense is not
* already in progress and that condensing has not been cancelled.
*/
static boolean_t
spa_livelist_condense_cb_check(void *arg, zthr_t *z)
{
(void) z;
spa_t *spa = arg;
if ((spa->spa_to_condense.ds != NULL) &&
(spa->spa_to_condense.syncing == B_FALSE) &&
(spa->spa_to_condense.cancelled == B_FALSE)) {
return (B_TRUE);
}
return (B_FALSE);
}
static void
spa_start_livelist_condensing_thread(spa_t *spa)
{
spa->spa_to_condense.ds = NULL;
spa->spa_to_condense.first = NULL;
spa->spa_to_condense.next = NULL;
spa->spa_to_condense.syncing = B_FALSE;
spa->spa_to_condense.cancelled = B_FALSE;
ASSERT3P(spa->spa_livelist_condense_zthr, ==, NULL);
spa->spa_livelist_condense_zthr =
zthr_create("z_livelist_condense",
spa_livelist_condense_cb_check,
spa_livelist_condense_cb, spa, minclsyspri);
}
static void
spa_spawn_aux_threads(spa_t *spa)
{
ASSERT(spa_writeable(spa));
ASSERT(MUTEX_HELD(&spa_namespace_lock));
spa_start_indirect_condensing_thread(spa);
spa_start_livelist_destroy_thread(spa);
spa_start_livelist_condensing_thread(spa);
ASSERT3P(spa->spa_checkpoint_discard_zthr, ==, NULL);
spa->spa_checkpoint_discard_zthr =
zthr_create("z_checkpoint_discard",
spa_checkpoint_discard_thread_check,
spa_checkpoint_discard_thread, spa, minclsyspri);
}
/*
* Fix up config after a partly-completed split. This is done with the
* ZPOOL_CONFIG_SPLIT nvlist. Both the splitting pool and the split-off
* pool have that entry in their config, but only the splitting one contains
* a list of all the guids of the vdevs that are being split off.
*
* This function determines what to do with that list: either rejoin
* all the disks to the pool, or complete the splitting process. To attempt
* the rejoin, each disk that is offlined is marked online again, and
* we do a reopen() call. If the vdev label for every disk that was
* marked online indicates it was successfully split off (VDEV_AUX_SPLIT_POOL)
* then we call vdev_split() on each disk, and complete the split.
*
* Otherwise we leave the config alone, with all the vdevs in place in
* the original pool.
*/
static void
spa_try_repair(spa_t *spa, nvlist_t *config)
{
uint_t extracted;
uint64_t *glist;
uint_t i, gcount;
nvlist_t *nvl;
vdev_t **vd;
boolean_t attempt_reopen;
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) != 0)
return;
/* check that the config is complete */
if (nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST,
&glist, &gcount) != 0)
return;
vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_SLEEP);
/* attempt to online all the vdevs & validate */
attempt_reopen = B_TRUE;
for (i = 0; i < gcount; i++) {
if (glist[i] == 0) /* vdev is hole */
continue;
vd[i] = spa_lookup_by_guid(spa, glist[i], B_FALSE);
if (vd[i] == NULL) {
/*
* Don't bother attempting to reopen the disks;
* just do the split.
*/
attempt_reopen = B_FALSE;
} else {
/* attempt to re-online it */
vd[i]->vdev_offline = B_FALSE;
}
}
if (attempt_reopen) {
vdev_reopen(spa->spa_root_vdev);
/* check each device to see what state it's in */
for (extracted = 0, i = 0; i < gcount; i++) {
if (vd[i] != NULL &&
vd[i]->vdev_stat.vs_aux != VDEV_AUX_SPLIT_POOL)
break;
++extracted;
}
}
/*
* If every disk has been moved to the new pool, or if we never
* even attempted to look at them, then we split them off for
* good.
*/
if (!attempt_reopen || gcount == extracted) {
for (i = 0; i < gcount; i++)
if (vd[i] != NULL)
vdev_split(vd[i]);
vdev_reopen(spa->spa_root_vdev);
}
kmem_free(vd, gcount * sizeof (vdev_t *));
}
static int
spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type)
{
char *ereport = FM_EREPORT_ZFS_POOL;
int error;
spa->spa_load_state = state;
(void) spa_import_progress_set_state(spa_guid(spa),
spa_load_state(spa));
gethrestime(&spa->spa_loaded_ts);
error = spa_load_impl(spa, type, &ereport);
/*
* Don't count references from objsets that are already closed
* and are making their way through the eviction process.
*/
spa_evicting_os_wait(spa);
spa->spa_minref = zfs_refcount_count(&spa->spa_refcount);
if (error) {
if (error != EEXIST) {
spa->spa_loaded_ts.tv_sec = 0;
spa->spa_loaded_ts.tv_nsec = 0;
}
if (error != EBADF) {
(void) zfs_ereport_post(ereport, spa,
NULL, NULL, NULL, 0);
}
}
spa->spa_load_state = error ? SPA_LOAD_ERROR : SPA_LOAD_NONE;
spa->spa_ena = 0;
(void) spa_import_progress_set_state(spa_guid(spa),
spa_load_state(spa));
return (error);
}
#ifdef ZFS_DEBUG
/*
* Count the number of per-vdev ZAPs associated with all of the vdevs in the
* vdev tree rooted in the given vd, and ensure that each ZAP is present in the
* spa's per-vdev ZAP list.
*/
static uint64_t
vdev_count_verify_zaps(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
uint64_t total = 0;
if (vd->vdev_top_zap != 0) {
total++;
ASSERT0(zap_lookup_int(spa->spa_meta_objset,
spa->spa_all_vdev_zaps, vd->vdev_top_zap));
}
if (vd->vdev_leaf_zap != 0) {
total++;
ASSERT0(zap_lookup_int(spa->spa_meta_objset,
spa->spa_all_vdev_zaps, vd->vdev_leaf_zap));
}
for (uint64_t i = 0; i < vd->vdev_children; i++) {
total += vdev_count_verify_zaps(vd->vdev_child[i]);
}
return (total);
}
#else
#define vdev_count_verify_zaps(vd) ((void) sizeof (vd), 0)
#endif
/*
* Determine whether the activity check is required.
*/
static boolean_t
spa_activity_check_required(spa_t *spa, uberblock_t *ub, nvlist_t *label,
nvlist_t *config)
{
uint64_t state = 0;
uint64_t hostid = 0;
uint64_t tryconfig_txg = 0;
uint64_t tryconfig_timestamp = 0;
uint16_t tryconfig_mmp_seq = 0;
nvlist_t *nvinfo;
if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
(void) nvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG,
&tryconfig_txg);
(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
&tryconfig_timestamp);
(void) nvlist_lookup_uint16(nvinfo, ZPOOL_CONFIG_MMP_SEQ,
&tryconfig_mmp_seq);
}
(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, &state);
/*
* Disable the MMP activity check - This is used by zdb which
* is intended to be used on potentially active pools.
*/
if (spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP)
return (B_FALSE);
/*
* Skip the activity check when the MMP feature is disabled.
*/
if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay == 0)
return (B_FALSE);
/*
* If the tryconfig_ values are nonzero, they are the results of an
* earlier tryimport. If they all match the uberblock we just found,
* then the pool has not changed and we return false so we do not test
* a second time.
*/
if (tryconfig_txg && tryconfig_txg == ub->ub_txg &&
tryconfig_timestamp && tryconfig_timestamp == ub->ub_timestamp &&
tryconfig_mmp_seq && tryconfig_mmp_seq ==
(MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0))
return (B_FALSE);
/*
* Allow the activity check to be skipped when importing the pool
* on the same host which last imported it. Since the hostid from
* configuration may be stale use the one read from the label.
*/
if (nvlist_exists(label, ZPOOL_CONFIG_HOSTID))
hostid = fnvlist_lookup_uint64(label, ZPOOL_CONFIG_HOSTID);
if (hostid == spa_get_hostid(spa))
return (B_FALSE);
/*
* Skip the activity test when the pool was cleanly exported.
*/
if (state != POOL_STATE_ACTIVE)
return (B_FALSE);
return (B_TRUE);
}
/*
* Nanoseconds the activity check must watch for changes on-disk.
*/
static uint64_t
spa_activity_check_duration(spa_t *spa, uberblock_t *ub)
{
uint64_t import_intervals = MAX(zfs_multihost_import_intervals, 1);
uint64_t multihost_interval = MSEC2NSEC(
MMP_INTERVAL_OK(zfs_multihost_interval));
uint64_t import_delay = MAX(NANOSEC, import_intervals *
multihost_interval);
/*
* Local tunables determine a minimum duration except for the case
* where we know when the remote host will suspend the pool if MMP
* writes do not land.
*
* See Big Theory comment at the top of mmp.c for the reasoning behind
* these cases and times.
*/
ASSERT(MMP_IMPORT_SAFETY_FACTOR >= 100);
if (MMP_INTERVAL_VALID(ub) && MMP_FAIL_INT_VALID(ub) &&
MMP_FAIL_INT(ub) > 0) {
/* MMP on remote host will suspend pool after failed writes */
import_delay = MMP_FAIL_INT(ub) * MSEC2NSEC(MMP_INTERVAL(ub)) *
MMP_IMPORT_SAFETY_FACTOR / 100;
zfs_dbgmsg("fail_intvals>0 import_delay=%llu ub_mmp "
"mmp_fails=%llu ub_mmp mmp_interval=%llu "
"import_intervals=%llu", (u_longlong_t)import_delay,
(u_longlong_t)MMP_FAIL_INT(ub),
(u_longlong_t)MMP_INTERVAL(ub),
(u_longlong_t)import_intervals);
} else if (MMP_INTERVAL_VALID(ub) && MMP_FAIL_INT_VALID(ub) &&
MMP_FAIL_INT(ub) == 0) {
/* MMP on remote host will never suspend pool */
import_delay = MAX(import_delay, (MSEC2NSEC(MMP_INTERVAL(ub)) +
ub->ub_mmp_delay) * import_intervals);
zfs_dbgmsg("fail_intvals=0 import_delay=%llu ub_mmp "
"mmp_interval=%llu ub_mmp_delay=%llu "
"import_intervals=%llu", (u_longlong_t)import_delay,
(u_longlong_t)MMP_INTERVAL(ub),
(u_longlong_t)ub->ub_mmp_delay,
(u_longlong_t)import_intervals);
} else if (MMP_VALID(ub)) {
/*
* zfs-0.7 compatibility case
*/
import_delay = MAX(import_delay, (multihost_interval +
ub->ub_mmp_delay) * import_intervals);
zfs_dbgmsg("import_delay=%llu ub_mmp_delay=%llu "
"import_intervals=%llu leaves=%u",
(u_longlong_t)import_delay,
(u_longlong_t)ub->ub_mmp_delay,
(u_longlong_t)import_intervals,
vdev_count_leaves(spa));
} else {
/* Using local tunings is the only reasonable option */
zfs_dbgmsg("pool last imported on non-MMP aware "
"host using import_delay=%llu multihost_interval=%llu "
"import_intervals=%llu", (u_longlong_t)import_delay,
(u_longlong_t)multihost_interval,
(u_longlong_t)import_intervals);
}
return (import_delay);
}
/*
* Perform the import activity check. If the user canceled the import or
* we detected activity then fail.
*/
static int
spa_activity_check(spa_t *spa, uberblock_t *ub, nvlist_t *config)
{
uint64_t txg = ub->ub_txg;
uint64_t timestamp = ub->ub_timestamp;
uint64_t mmp_config = ub->ub_mmp_config;
uint16_t mmp_seq = MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0;
uint64_t import_delay;
hrtime_t import_expire;
nvlist_t *mmp_label = NULL;
vdev_t *rvd = spa->spa_root_vdev;
kcondvar_t cv;
kmutex_t mtx;
int error = 0;
cv_init(&cv, NULL, CV_DEFAULT, NULL);
mutex_init(&mtx, NULL, MUTEX_DEFAULT, NULL);
mutex_enter(&mtx);
/*
* If ZPOOL_CONFIG_MMP_TXG is present an activity check was performed
* during the earlier tryimport. If the txg recorded there is 0 then
* the pool is known to be active on another host.
*
* Otherwise, the pool might be in use on another host. Check for
* changes in the uberblocks on disk if necessary.
*/
if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
nvlist_t *nvinfo = fnvlist_lookup_nvlist(config,
ZPOOL_CONFIG_LOAD_INFO);
if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_TXG) &&
fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG) == 0) {
vdev_uberblock_load(rvd, ub, &mmp_label);
error = SET_ERROR(EREMOTEIO);
goto out;
}
}
import_delay = spa_activity_check_duration(spa, ub);
/* Add a small random factor in case of simultaneous imports (0-25%) */
import_delay += import_delay * random_in_range(250) / 1000;
import_expire = gethrtime() + import_delay;
while (gethrtime() < import_expire) {
(void) spa_import_progress_set_mmp_check(spa_guid(spa),
NSEC2SEC(import_expire - gethrtime()));
vdev_uberblock_load(rvd, ub, &mmp_label);
if (txg != ub->ub_txg || timestamp != ub->ub_timestamp ||
mmp_seq != (MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0)) {
zfs_dbgmsg("multihost activity detected "
"txg %llu ub_txg %llu "
"timestamp %llu ub_timestamp %llu "
"mmp_config %#llx ub_mmp_config %#llx",
(u_longlong_t)txg, (u_longlong_t)ub->ub_txg,
(u_longlong_t)timestamp,
(u_longlong_t)ub->ub_timestamp,
(u_longlong_t)mmp_config,
(u_longlong_t)ub->ub_mmp_config);
error = SET_ERROR(EREMOTEIO);
break;
}
if (mmp_label) {
nvlist_free(mmp_label);
mmp_label = NULL;
}
error = cv_timedwait_sig(&cv, &mtx, ddi_get_lbolt() + hz);
if (error != -1) {
error = SET_ERROR(EINTR);
break;
}
error = 0;
}
out:
mutex_exit(&mtx);
mutex_destroy(&mtx);
cv_destroy(&cv);
/*
* If the pool is determined to be active store the status in the
* spa->spa_load_info nvlist. If the remote hostname or hostid are
* available from configuration read from disk store them as well.
* This allows 'zpool import' to generate a more useful message.
*
* ZPOOL_CONFIG_MMP_STATE - observed pool status (mandatory)
* ZPOOL_CONFIG_MMP_HOSTNAME - hostname from the active pool
* ZPOOL_CONFIG_MMP_HOSTID - hostid from the active pool
*/
if (error == EREMOTEIO) {
char *hostname = "<unknown>";
uint64_t hostid = 0;
if (mmp_label) {
if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTNAME)) {
hostname = fnvlist_lookup_string(mmp_label,
ZPOOL_CONFIG_HOSTNAME);
fnvlist_add_string(spa->spa_load_info,
ZPOOL_CONFIG_MMP_HOSTNAME, hostname);
}
if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTID)) {
hostid = fnvlist_lookup_uint64(mmp_label,
ZPOOL_CONFIG_HOSTID);
fnvlist_add_uint64(spa->spa_load_info,
ZPOOL_CONFIG_MMP_HOSTID, hostid);
}
}
fnvlist_add_uint64(spa->spa_load_info,
ZPOOL_CONFIG_MMP_STATE, MMP_STATE_ACTIVE);
fnvlist_add_uint64(spa->spa_load_info,
ZPOOL_CONFIG_MMP_TXG, 0);
error = spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO);
}
if (mmp_label)
nvlist_free(mmp_label);
return (error);
}
static int
spa_verify_host(spa_t *spa, nvlist_t *mos_config)
{
uint64_t hostid;
char *hostname;
uint64_t myhostid = 0;
if (!spa_is_root(spa) && nvlist_lookup_uint64(mos_config,
ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
hostname = fnvlist_lookup_string(mos_config,
ZPOOL_CONFIG_HOSTNAME);
myhostid = zone_get_hostid(NULL);
if (hostid != 0 && myhostid != 0 && hostid != myhostid) {
cmn_err(CE_WARN, "pool '%s' could not be "
"loaded as it was last accessed by "
"another system (host: %s hostid: 0x%llx). "
"See: https://openzfs.github.io/openzfs-docs/msg/"
"ZFS-8000-EY",
spa_name(spa), hostname, (u_longlong_t)hostid);
spa_load_failed(spa, "hostid verification failed: pool "
"last accessed by host: %s (hostid: 0x%llx)",
hostname, (u_longlong_t)hostid);
return (SET_ERROR(EBADF));
}
}
return (0);
}
static int
spa_ld_parse_config(spa_t *spa, spa_import_type_t type)
{
int error = 0;
nvlist_t *nvtree, *nvl, *config = spa->spa_config;
int parse;
vdev_t *rvd;
uint64_t pool_guid;
char *comment;
char *compatibility;
/*
* Versioning wasn't explicitly added to the label until later, so if
* it's not present treat it as the initial version.
*/
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
&spa->spa_ubsync.ub_version) != 0)
spa->spa_ubsync.ub_version = SPA_VERSION_INITIAL;
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) {
spa_load_failed(spa, "invalid config provided: '%s' missing",
ZPOOL_CONFIG_POOL_GUID);
return (SET_ERROR(EINVAL));
}
/*
* If we are doing an import, ensure that the pool is not already
* imported by checking if its pool guid already exists in the
* spa namespace.
*
* The only case that we allow an already imported pool to be
* imported again, is when the pool is checkpointed and we want to
* look at its checkpointed state from userland tools like zdb.
*/
#ifdef _KERNEL
if ((spa->spa_load_state == SPA_LOAD_IMPORT ||
spa->spa_load_state == SPA_LOAD_TRYIMPORT) &&
spa_guid_exists(pool_guid, 0)) {
#else
if ((spa->spa_load_state == SPA_LOAD_IMPORT ||
spa->spa_load_state == SPA_LOAD_TRYIMPORT) &&
spa_guid_exists(pool_guid, 0) &&
!spa_importing_readonly_checkpoint(spa)) {
#endif
spa_load_failed(spa, "a pool with guid %llu is already open",
(u_longlong_t)pool_guid);
return (SET_ERROR(EEXIST));
}
spa->spa_config_guid = pool_guid;
nvlist_free(spa->spa_load_info);
spa->spa_load_info = fnvlist_alloc();
ASSERT(spa->spa_comment == NULL);
if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
spa->spa_comment = spa_strdup(comment);
ASSERT(spa->spa_compatibility == NULL);
if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMPATIBILITY,
&compatibility) == 0)
spa->spa_compatibility = spa_strdup(compatibility);
(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
&spa->spa_config_txg);
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) == 0)
spa->spa_config_splitting = fnvlist_dup(nvl);
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvtree)) {
spa_load_failed(spa, "invalid config provided: '%s' missing",
ZPOOL_CONFIG_VDEV_TREE);
return (SET_ERROR(EINVAL));
}
/*
* Create "The Godfather" zio to hold all async IOs
*/
spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *),
KM_SLEEP);
for (int i = 0; i < max_ncpus; i++) {
spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE |
ZIO_FLAG_GODFATHER);
}
/*
* Parse the configuration into a vdev tree. We explicitly set the
* value that will be returned by spa_version() since parsing the
* configuration requires knowing the version number.
*/
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
parse = (type == SPA_IMPORT_EXISTING ?
VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT);
error = spa_config_parse(spa, &rvd, nvtree, NULL, 0, parse);
spa_config_exit(spa, SCL_ALL, FTAG);
if (error != 0) {
spa_load_failed(spa, "unable to parse config [error=%d]",
error);
return (error);
}
ASSERT(spa->spa_root_vdev == rvd);
ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT);
ASSERT3U(spa->spa_max_ashift, <=, SPA_MAXBLOCKSHIFT);
if (type != SPA_IMPORT_ASSEMBLE) {
ASSERT(spa_guid(spa) == pool_guid);
}
return (0);
}
/*
* Recursively open all vdevs in the vdev tree. This function is called twice:
* first with the untrusted config, then with the trusted config.
*/
static int
spa_ld_open_vdevs(spa_t *spa)
{
int error = 0;
/*
* spa_missing_tvds_allowed defines how many top-level vdevs can be
* missing/unopenable for the root vdev to be still considered openable.
*/
if (spa->spa_trust_config) {
spa->spa_missing_tvds_allowed = zfs_max_missing_tvds;
} else if (spa->spa_config_source == SPA_CONFIG_SRC_CACHEFILE) {
spa->spa_missing_tvds_allowed = zfs_max_missing_tvds_cachefile;
} else if (spa->spa_config_source == SPA_CONFIG_SRC_SCAN) {
spa->spa_missing_tvds_allowed = zfs_max_missing_tvds_scan;
} else {
spa->spa_missing_tvds_allowed = 0;
}
spa->spa_missing_tvds_allowed =
MAX(zfs_max_missing_tvds, spa->spa_missing_tvds_allowed);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
error = vdev_open(spa->spa_root_vdev);
spa_config_exit(spa, SCL_ALL, FTAG);
if (spa->spa_missing_tvds != 0) {
spa_load_note(spa, "vdev tree has %lld missing top-level "
"vdevs.", (u_longlong_t)spa->spa_missing_tvds);
if (spa->spa_trust_config && (spa->spa_mode & SPA_MODE_WRITE)) {
/*
* Although theoretically we could allow users to open
* incomplete pools in RW mode, we'd need to add a lot
* of extra logic (e.g. adjust pool space to account
* for missing vdevs).
* This limitation also prevents users from accidentally
* opening the pool in RW mode during data recovery and
* damaging it further.
*/
spa_load_note(spa, "pools with missing top-level "
"vdevs can only be opened in read-only mode.");
error = SET_ERROR(ENXIO);
} else {
spa_load_note(spa, "current settings allow for maximum "
"%lld missing top-level vdevs at this stage.",
(u_longlong_t)spa->spa_missing_tvds_allowed);
}
}
if (error != 0) {
spa_load_failed(spa, "unable to open vdev tree [error=%d]",
error);
}
if (spa->spa_missing_tvds != 0 || error != 0)
vdev_dbgmsg_print_tree(spa->spa_root_vdev, 2);
return (error);
}
/*
* We need to validate the vdev labels against the configuration that
* we have in hand. This function is called twice: first with an untrusted
* config, then with a trusted config. The validation is more strict when the
* config is trusted.
*/
static int
spa_ld_validate_vdevs(spa_t *spa)
{
int error = 0;
vdev_t *rvd = spa->spa_root_vdev;
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
error = vdev_validate(rvd);
spa_config_exit(spa, SCL_ALL, FTAG);
if (error != 0) {
spa_load_failed(spa, "vdev_validate failed [error=%d]", error);
return (error);
}
if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) {
spa_load_failed(spa, "cannot open vdev tree after invalidating "
"some vdevs");
vdev_dbgmsg_print_tree(rvd, 2);
return (SET_ERROR(ENXIO));
}
return (0);
}
static void
spa_ld_select_uberblock_done(spa_t *spa, uberblock_t *ub)
{
spa->spa_state = POOL_STATE_ACTIVE;
spa->spa_ubsync = spa->spa_uberblock;
spa->spa_verify_min_txg = spa->spa_extreme_rewind ?
TXG_INITIAL - 1 : spa_last_synced_txg(spa) - TXG_DEFER_SIZE - 1;
spa->spa_first_txg = spa->spa_last_ubsync_txg ?
spa->spa_last_ubsync_txg : spa_last_synced_txg(spa) + 1;
spa->spa_claim_max_txg = spa->spa_first_txg;
spa->spa_prev_software_version = ub->ub_software_version;
}
static int
spa_ld_select_uberblock(spa_t *spa, spa_import_type_t type)
{
vdev_t *rvd = spa->spa_root_vdev;
nvlist_t *label;
uberblock_t *ub = &spa->spa_uberblock;
boolean_t activity_check = B_FALSE;
/*
* If we are opening the checkpointed state of the pool by
* rewinding to it, at this point we will have written the
* checkpointed uberblock to the vdev labels, so searching
* the labels will find the right uberblock. However, if
* we are opening the checkpointed state read-only, we have
* not modified the labels. Therefore, we must ignore the
* labels and continue using the spa_uberblock that was set
* by spa_ld_checkpoint_rewind.
*
* Note that it would be fine to ignore the labels when
* rewinding (opening writeable) as well. However, if we
* crash just after writing the labels, we will end up
* searching the labels. Doing so in the common case means
* that this code path gets exercised normally, rather than
* just in the edge case.
*/
if (ub->ub_checkpoint_txg != 0 &&
spa_importing_readonly_checkpoint(spa)) {
spa_ld_select_uberblock_done(spa, ub);
return (0);
}
/*
* Find the best uberblock.
*/
vdev_uberblock_load(rvd, ub, &label);
/*
* If we weren't able to find a single valid uberblock, return failure.
*/
if (ub->ub_txg == 0) {
nvlist_free(label);
spa_load_failed(spa, "no valid uberblock found");
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO));
}
if (spa->spa_load_max_txg != UINT64_MAX) {
(void) spa_import_progress_set_max_txg(spa_guid(spa),
(u_longlong_t)spa->spa_load_max_txg);
}
spa_load_note(spa, "using uberblock with txg=%llu",
(u_longlong_t)ub->ub_txg);
/*
* For pools which have the multihost property on determine if the
* pool is truly inactive and can be safely imported. Prevent
* hosts which don't have a hostid set from importing the pool.
*/
activity_check = spa_activity_check_required(spa, ub, label,
spa->spa_config);
if (activity_check) {
if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay &&
spa_get_hostid(spa) == 0) {
nvlist_free(label);
fnvlist_add_uint64(spa->spa_load_info,
ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
}
int error = spa_activity_check(spa, ub, spa->spa_config);
if (error) {
nvlist_free(label);
return (error);
}
fnvlist_add_uint64(spa->spa_load_info,
ZPOOL_CONFIG_MMP_STATE, MMP_STATE_INACTIVE);
fnvlist_add_uint64(spa->spa_load_info,
ZPOOL_CONFIG_MMP_TXG, ub->ub_txg);
fnvlist_add_uint16(spa->spa_load_info,
ZPOOL_CONFIG_MMP_SEQ,
(MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0));
}
/*
* If the pool has an unsupported version we can't open it.
*/
if (!SPA_VERSION_IS_SUPPORTED(ub->ub_version)) {
nvlist_free(label);
spa_load_failed(spa, "version %llu is not supported",
(u_longlong_t)ub->ub_version);
return (spa_vdev_err(rvd, VDEV_AUX_VERSION_NEWER, ENOTSUP));
}
if (ub->ub_version >= SPA_VERSION_FEATURES) {
nvlist_t *features;
/*
* If we weren't able to find what's necessary for reading the
* MOS in the label, return failure.
*/
if (label == NULL) {
spa_load_failed(spa, "label config unavailable");
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
ENXIO));
}
if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_FEATURES_FOR_READ,
&features) != 0) {
nvlist_free(label);
spa_load_failed(spa, "invalid label: '%s' missing",
ZPOOL_CONFIG_FEATURES_FOR_READ);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
ENXIO));
}
/*
* Update our in-core representation with the definitive values
* from the label.
*/
nvlist_free(spa->spa_label_features);
spa->spa_label_features = fnvlist_dup(features);
}
nvlist_free(label);
/*
* Look through entries in the label nvlist's features_for_read. If
* there is a feature listed there which we don't understand then we
* cannot open a pool.
*/
if (ub->ub_version >= SPA_VERSION_FEATURES) {
nvlist_t *unsup_feat;
unsup_feat = fnvlist_alloc();
for (nvpair_t *nvp = nvlist_next_nvpair(spa->spa_label_features,
NULL); nvp != NULL;
nvp = nvlist_next_nvpair(spa->spa_label_features, nvp)) {
if (!zfeature_is_supported(nvpair_name(nvp))) {
fnvlist_add_string(unsup_feat,
nvpair_name(nvp), "");
}
}
if (!nvlist_empty(unsup_feat)) {
fnvlist_add_nvlist(spa->spa_load_info,
ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat);
nvlist_free(unsup_feat);
spa_load_failed(spa, "some features are unsupported");
return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
ENOTSUP));
}
nvlist_free(unsup_feat);
}
if (type != SPA_IMPORT_ASSEMBLE && spa->spa_config_splitting) {
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa_try_repair(spa, spa->spa_config);
spa_config_exit(spa, SCL_ALL, FTAG);
nvlist_free(spa->spa_config_splitting);
spa->spa_config_splitting = NULL;
}
/*
* Initialize internal SPA structures.
*/
spa_ld_select_uberblock_done(spa, ub);
return (0);
}
static int
spa_ld_open_rootbp(spa_t *spa)
{
int error = 0;
vdev_t *rvd = spa->spa_root_vdev;
error = dsl_pool_init(spa, spa->spa_first_txg, &spa->spa_dsl_pool);
if (error != 0) {
spa_load_failed(spa, "unable to open rootbp in dsl_pool_init "
"[error=%d]", error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset;
return (0);
}
static int
spa_ld_trusted_config(spa_t *spa, spa_import_type_t type,
boolean_t reloading)
{
vdev_t *mrvd, *rvd = spa->spa_root_vdev;
nvlist_t *nv, *mos_config, *policy;
int error = 0, copy_error;
uint64_t healthy_tvds, healthy_tvds_mos;
uint64_t mos_config_txg;
if (spa_dir_prop(spa, DMU_POOL_CONFIG, &spa->spa_config_object, B_TRUE)
!= 0)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
/*
* If we're assembling a pool from a split, the config provided is
* already trusted so there is nothing to do.
*/
if (type == SPA_IMPORT_ASSEMBLE)
return (0);
healthy_tvds = spa_healthy_core_tvds(spa);
if (load_nvlist(spa, spa->spa_config_object, &mos_config)
!= 0) {
spa_load_failed(spa, "unable to retrieve MOS config");
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
/*
* If we are doing an open, pool owner wasn't verified yet, thus do
* the verification here.
*/
if (spa->spa_load_state == SPA_LOAD_OPEN) {
error = spa_verify_host(spa, mos_config);
if (error != 0) {
nvlist_free(mos_config);
return (error);
}
}
nv = fnvlist_lookup_nvlist(mos_config, ZPOOL_CONFIG_VDEV_TREE);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
/*
* Build a new vdev tree from the trusted config
*/
error = spa_config_parse(spa, &mrvd, nv, NULL, 0, VDEV_ALLOC_LOAD);
if (error != 0) {
nvlist_free(mos_config);
spa_config_exit(spa, SCL_ALL, FTAG);
spa_load_failed(spa, "spa_config_parse failed [error=%d]",
error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
}
/*
* Vdev paths in the MOS may be obsolete. If the untrusted config was
* obtained by scanning /dev/dsk, then it will have the right vdev
* paths. We update the trusted MOS config with this information.
* We first try to copy the paths with vdev_copy_path_strict, which
* succeeds only when both configs have exactly the same vdev tree.
* If that fails, we fall back to a more flexible method that has a
* best effort policy.
*/
copy_error = vdev_copy_path_strict(rvd, mrvd);
if (copy_error != 0 || spa_load_print_vdev_tree) {
spa_load_note(spa, "provided vdev tree:");
vdev_dbgmsg_print_tree(rvd, 2);
spa_load_note(spa, "MOS vdev tree:");
vdev_dbgmsg_print_tree(mrvd, 2);
}
if (copy_error != 0) {
spa_load_note(spa, "vdev_copy_path_strict failed, falling "
"back to vdev_copy_path_relaxed");
vdev_copy_path_relaxed(rvd, mrvd);
}
vdev_close(rvd);
vdev_free(rvd);
spa->spa_root_vdev = mrvd;
rvd = mrvd;
spa_config_exit(spa, SCL_ALL, FTAG);
/*
* We will use spa_config if we decide to reload the spa or if spa_load
* fails and we rewind. We must thus regenerate the config using the
* MOS information with the updated paths. ZPOOL_LOAD_POLICY is used to
* pass settings on how to load the pool and is not stored in the MOS.
* We copy it over to our new, trusted config.
*/
mos_config_txg = fnvlist_lookup_uint64(mos_config,
ZPOOL_CONFIG_POOL_TXG);
nvlist_free(mos_config);
mos_config = spa_config_generate(spa, NULL, mos_config_txg, B_FALSE);
if (nvlist_lookup_nvlist(spa->spa_config, ZPOOL_LOAD_POLICY,
&policy) == 0)
fnvlist_add_nvlist(mos_config, ZPOOL_LOAD_POLICY, policy);
spa_config_set(spa, mos_config);
spa->spa_config_source = SPA_CONFIG_SRC_MOS;
/*
* Now that we got the config from the MOS, we should be more strict
* in checking blkptrs and can make assumptions about the consistency
* of the vdev tree. spa_trust_config must be set to true before opening
* vdevs in order for them to be writeable.
*/
spa->spa_trust_config = B_TRUE;
/*
* Open and validate the new vdev tree
*/
error = spa_ld_open_vdevs(spa);
if (error != 0)
return (error);
error = spa_ld_validate_vdevs(spa);
if (error != 0)
return (error);
if (copy_error != 0 || spa_load_print_vdev_tree) {
spa_load_note(spa, "final vdev tree:");
vdev_dbgmsg_print_tree(rvd, 2);
}
if (spa->spa_load_state != SPA_LOAD_TRYIMPORT &&
!spa->spa_extreme_rewind && zfs_max_missing_tvds == 0) {
/*
* Sanity check to make sure that we are indeed loading the
* latest uberblock. If we missed SPA_SYNC_MIN_VDEVS tvds
* in the config provided and they happened to be the only ones
* to have the latest uberblock, we could involuntarily perform
* an extreme rewind.
*/
healthy_tvds_mos = spa_healthy_core_tvds(spa);
if (healthy_tvds_mos - healthy_tvds >=
SPA_SYNC_MIN_VDEVS) {
spa_load_note(spa, "config provided misses too many "
"top-level vdevs compared to MOS (%lld vs %lld). ",
(u_longlong_t)healthy_tvds,
(u_longlong_t)healthy_tvds_mos);
spa_load_note(spa, "vdev tree:");
vdev_dbgmsg_print_tree(rvd, 2);
if (reloading) {
spa_load_failed(spa, "config was already "
"provided from MOS. Aborting.");
return (spa_vdev_err(rvd,
VDEV_AUX_CORRUPT_DATA, EIO));
}
spa_load_note(spa, "spa must be reloaded using MOS "
"config");
return (SET_ERROR(EAGAIN));
}
}
error = spa_check_for_missing_logs(spa);
if (error != 0)
return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO));
if (rvd->vdev_guid_sum != spa->spa_uberblock.ub_guid_sum) {
spa_load_failed(spa, "uberblock guid sum doesn't match MOS "
"guid sum (%llu != %llu)",
(u_longlong_t)spa->spa_uberblock.ub_guid_sum,
(u_longlong_t)rvd->vdev_guid_sum);
return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM,
ENXIO));
}
return (0);
}
static int
spa_ld_open_indirect_vdev_metadata(spa_t *spa)
{
int error = 0;
vdev_t *rvd = spa->spa_root_vdev;
/*
* Everything that we read before spa_remove_init() must be stored
* on concreted vdevs. Therefore we do this as early as possible.
*/
error = spa_remove_init(spa);
if (error != 0) {
spa_load_failed(spa, "spa_remove_init failed [error=%d]",
error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
/*
* Retrieve information needed to condense indirect vdev mappings.
*/
error = spa_condense_init(spa);
if (error != 0) {
spa_load_failed(spa, "spa_condense_init failed [error=%d]",
error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
}
return (0);
}
static int
spa_ld_check_features(spa_t *spa, boolean_t *missing_feat_writep)
{
int error = 0;
vdev_t *rvd = spa->spa_root_vdev;
if (spa_version(spa) >= SPA_VERSION_FEATURES) {
boolean_t missing_feat_read = B_FALSE;
nvlist_t *unsup_feat, *enabled_feat;
if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_READ,
&spa->spa_feat_for_read_obj, B_TRUE) != 0) {
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_WRITE,
&spa->spa_feat_for_write_obj, B_TRUE) != 0) {
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
if (spa_dir_prop(spa, DMU_POOL_FEATURE_DESCRIPTIONS,
&spa->spa_feat_desc_obj, B_TRUE) != 0) {
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
enabled_feat = fnvlist_alloc();
unsup_feat = fnvlist_alloc();
if (!spa_features_check(spa, B_FALSE,
unsup_feat, enabled_feat))
missing_feat_read = B_TRUE;
if (spa_writeable(spa) ||
spa->spa_load_state == SPA_LOAD_TRYIMPORT) {
if (!spa_features_check(spa, B_TRUE,
unsup_feat, enabled_feat)) {
*missing_feat_writep = B_TRUE;
}
}
fnvlist_add_nvlist(spa->spa_load_info,
ZPOOL_CONFIG_ENABLED_FEAT, enabled_feat);
if (!nvlist_empty(unsup_feat)) {
fnvlist_add_nvlist(spa->spa_load_info,
ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat);
}
fnvlist_free(enabled_feat);
fnvlist_free(unsup_feat);
if (!missing_feat_read) {
fnvlist_add_boolean(spa->spa_load_info,
ZPOOL_CONFIG_CAN_RDONLY);
}
/*
* If the state is SPA_LOAD_TRYIMPORT, our objective is
* twofold: to determine whether the pool is available for
* import in read-write mode and (if it is not) whether the
* pool is available for import in read-only mode. If the pool
* is available for import in read-write mode, it is displayed
* as available in userland; if it is not available for import
* in read-only mode, it is displayed as unavailable in
* userland. If the pool is available for import in read-only
* mode but not read-write mode, it is displayed as unavailable
* in userland with a special note that the pool is actually
* available for open in read-only mode.
*
* As a result, if the state is SPA_LOAD_TRYIMPORT and we are
* missing a feature for write, we must first determine whether
* the pool can be opened read-only before returning to
* userland in order to know whether to display the
* abovementioned note.
*/
if (missing_feat_read || (*missing_feat_writep &&
spa_writeable(spa))) {
spa_load_failed(spa, "pool uses unsupported features");
return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
ENOTSUP));
}
/*
* Load refcounts for ZFS features from disk into an in-memory
* cache during SPA initialization.
*/
for (spa_feature_t i = 0; i < SPA_FEATURES; i++) {
uint64_t refcount;
error = feature_get_refcount_from_disk(spa,
&spa_feature_table[i], &refcount);
if (error == 0) {
spa->spa_feat_refcount_cache[i] = refcount;
} else if (error == ENOTSUP) {
spa->spa_feat_refcount_cache[i] =
SPA_FEATURE_DISABLED;
} else {
spa_load_failed(spa, "error getting refcount "
"for feature %s [error=%d]",
spa_feature_table[i].fi_guid, error);
return (spa_vdev_err(rvd,
VDEV_AUX_CORRUPT_DATA, EIO));
}
}
}
if (spa_feature_is_active(spa, SPA_FEATURE_ENABLED_TXG)) {
if (spa_dir_prop(spa, DMU_POOL_FEATURE_ENABLED_TXG,
&spa->spa_feat_enabled_txg_obj, B_TRUE) != 0)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
/*
* Encryption was added before bookmark_v2, even though bookmark_v2
* is now a dependency. If this pool has encryption enabled without
* bookmark_v2, trigger an errata message.
*/
if (spa_feature_is_enabled(spa, SPA_FEATURE_ENCRYPTION) &&
!spa_feature_is_enabled(spa, SPA_FEATURE_BOOKMARK_V2)) {
spa->spa_errata = ZPOOL_ERRATA_ZOL_8308_ENCRYPTION;
}
return (0);
}
static int
spa_ld_load_special_directories(spa_t *spa)
{
int error = 0;
vdev_t *rvd = spa->spa_root_vdev;
spa->spa_is_initializing = B_TRUE;
error = dsl_pool_open(spa->spa_dsl_pool);
spa->spa_is_initializing = B_FALSE;
if (error != 0) {
spa_load_failed(spa, "dsl_pool_open failed [error=%d]", error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
return (0);
}
static int
spa_ld_get_props(spa_t *spa)
{
int error = 0;
uint64_t obj;
vdev_t *rvd = spa->spa_root_vdev;
/* Grab the checksum salt from the MOS. */
error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_CHECKSUM_SALT, 1,
sizeof (spa->spa_cksum_salt.zcs_bytes),
spa->spa_cksum_salt.zcs_bytes);
if (error == ENOENT) {
/* Generate a new salt for subsequent use */
(void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes,
sizeof (spa->spa_cksum_salt.zcs_bytes));
} else if (error != 0) {
spa_load_failed(spa, "unable to retrieve checksum salt from "
"MOS [error=%d]", error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
if (spa_dir_prop(spa, DMU_POOL_SYNC_BPOBJ, &obj, B_TRUE) != 0)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
error = bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj);
if (error != 0) {
spa_load_failed(spa, "error opening deferred-frees bpobj "
"[error=%d]", error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
/*
* Load the bit that tells us to use the new accounting function
* (raid-z deflation). If we have an older pool, this will not
* be present.
*/
error = spa_dir_prop(spa, DMU_POOL_DEFLATE, &spa->spa_deflate, B_FALSE);
if (error != 0 && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
error = spa_dir_prop(spa, DMU_POOL_CREATION_VERSION,
&spa->spa_creation_version, B_FALSE);
if (error != 0 && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
/*
* Load the persistent error log. If we have an older pool, this will
* not be present.
*/
error = spa_dir_prop(spa, DMU_POOL_ERRLOG_LAST, &spa->spa_errlog_last,
B_FALSE);
if (error != 0 && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
error = spa_dir_prop(spa, DMU_POOL_ERRLOG_SCRUB,
&spa->spa_errlog_scrub, B_FALSE);
if (error != 0 && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
/*
* Load the livelist deletion field. If a livelist is queued for
* deletion, indicate that in the spa
*/
error = spa_dir_prop(spa, DMU_POOL_DELETED_CLONES,
&spa->spa_livelists_to_delete, B_FALSE);
if (error != 0 && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
/*
* Load the history object. If we have an older pool, this
* will not be present.
*/
error = spa_dir_prop(spa, DMU_POOL_HISTORY, &spa->spa_history, B_FALSE);
if (error != 0 && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
/*
* Load the per-vdev ZAP map. If we have an older pool, this will not
* be present; in this case, defer its creation to a later time to
* avoid dirtying the MOS this early / out of sync context. See
* spa_sync_config_object.
*/
/* The sentinel is only available in the MOS config. */
nvlist_t *mos_config;
if (load_nvlist(spa, spa->spa_config_object, &mos_config) != 0) {
spa_load_failed(spa, "unable to retrieve MOS config");
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
error = spa_dir_prop(spa, DMU_POOL_VDEV_ZAP_MAP,
&spa->spa_all_vdev_zaps, B_FALSE);
if (error == ENOENT) {
VERIFY(!nvlist_exists(mos_config,
ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
spa->spa_avz_action = AVZ_ACTION_INITIALIZE;
ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
} else if (error != 0) {
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
} else if (!nvlist_exists(mos_config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) {
/*
* An older version of ZFS overwrote the sentinel value, so
* we have orphaned per-vdev ZAPs in the MOS. Defer their
* destruction to later; see spa_sync_config_object.
*/
spa->spa_avz_action = AVZ_ACTION_DESTROY;
/*
* We're assuming that no vdevs have had their ZAPs created
* before this. Better be sure of it.
*/
ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
}
nvlist_free(mos_config);
spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
error = spa_dir_prop(spa, DMU_POOL_PROPS, &spa->spa_pool_props_object,
B_FALSE);
if (error && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
if (error == 0) {
uint64_t autoreplace = 0;
spa_prop_find(spa, ZPOOL_PROP_BOOTFS, &spa->spa_bootfs);
spa_prop_find(spa, ZPOOL_PROP_AUTOREPLACE, &autoreplace);
spa_prop_find(spa, ZPOOL_PROP_DELEGATION, &spa->spa_delegation);
spa_prop_find(spa, ZPOOL_PROP_FAILUREMODE, &spa->spa_failmode);
spa_prop_find(spa, ZPOOL_PROP_AUTOEXPAND, &spa->spa_autoexpand);
spa_prop_find(spa, ZPOOL_PROP_MULTIHOST, &spa->spa_multihost);
spa_prop_find(spa, ZPOOL_PROP_AUTOTRIM, &spa->spa_autotrim);
spa->spa_autoreplace = (autoreplace != 0);
}
/*
* If we are importing a pool with missing top-level vdevs,
* we enforce that the pool doesn't panic or get suspended on
* error since the likelihood of missing data is extremely high.
*/
if (spa->spa_missing_tvds > 0 &&
spa->spa_failmode != ZIO_FAILURE_MODE_CONTINUE &&
spa->spa_load_state != SPA_LOAD_TRYIMPORT) {
spa_load_note(spa, "forcing failmode to 'continue' "
"as some top level vdevs are missing");
spa->spa_failmode = ZIO_FAILURE_MODE_CONTINUE;
}
return (0);
}
static int
spa_ld_open_aux_vdevs(spa_t *spa, spa_import_type_t type)
{
int error = 0;
vdev_t *rvd = spa->spa_root_vdev;
/*
* If we're assembling the pool from the split-off vdevs of
* an existing pool, we don't want to attach the spares & cache
* devices.
*/
/*
* Load any hot spares for this pool.
*/
error = spa_dir_prop(spa, DMU_POOL_SPARES, &spa->spa_spares.sav_object,
B_FALSE);
if (error != 0 && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
if (error == 0 && type != SPA_IMPORT_ASSEMBLE) {
ASSERT(spa_version(spa) >= SPA_VERSION_SPARES);
if (load_nvlist(spa, spa->spa_spares.sav_object,
&spa->spa_spares.sav_config) != 0) {
spa_load_failed(spa, "error loading spares nvlist");
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa_load_spares(spa);
spa_config_exit(spa, SCL_ALL, FTAG);
} else if (error == 0) {
spa->spa_spares.sav_sync = B_TRUE;
}
/*
* Load any level 2 ARC devices for this pool.
*/
error = spa_dir_prop(spa, DMU_POOL_L2CACHE,
&spa->spa_l2cache.sav_object, B_FALSE);
if (error != 0 && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
if (error == 0 && type != SPA_IMPORT_ASSEMBLE) {
ASSERT(spa_version(spa) >= SPA_VERSION_L2CACHE);
if (load_nvlist(spa, spa->spa_l2cache.sav_object,
&spa->spa_l2cache.sav_config) != 0) {
spa_load_failed(spa, "error loading l2cache nvlist");
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa_load_l2cache(spa);
spa_config_exit(spa, SCL_ALL, FTAG);
} else if (error == 0) {
spa->spa_l2cache.sav_sync = B_TRUE;
}
return (0);
}
static int
spa_ld_load_vdev_metadata(spa_t *spa)
{
int error = 0;
vdev_t *rvd = spa->spa_root_vdev;
/*
* If the 'multihost' property is set, then never allow a pool to
* be imported when the system hostid is zero. The exception to
* this rule is zdb which is always allowed to access pools.
*/
if (spa_multihost(spa) && spa_get_hostid(spa) == 0 &&
(spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP) == 0) {
fnvlist_add_uint64(spa->spa_load_info,
ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
}
/*
* If the 'autoreplace' property is set, then post a resource notifying
* the ZFS DE that it should not issue any faults for unopenable
* devices. We also iterate over the vdevs, and post a sysevent for any
* unopenable vdevs so that the normal autoreplace handler can take
* over.
*/
if (spa->spa_autoreplace && spa->spa_load_state != SPA_LOAD_TRYIMPORT) {
spa_check_removed(spa->spa_root_vdev);
/*
* For the import case, this is done in spa_import(), because
* at this point we're using the spare definitions from
* the MOS config, not necessarily from the userland config.
*/
if (spa->spa_load_state != SPA_LOAD_IMPORT) {
spa_aux_check_removed(&spa->spa_spares);
spa_aux_check_removed(&spa->spa_l2cache);
}
}
/*
* Load the vdev metadata such as metaslabs, DTLs, spacemap object, etc.
*/
error = vdev_load(rvd);
if (error != 0) {
spa_load_failed(spa, "vdev_load failed [error=%d]", error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
}
error = spa_ld_log_spacemaps(spa);
if (error != 0) {
spa_load_failed(spa, "spa_ld_log_sm_data failed [error=%d]",
error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
}
/*
* Propagate the leaf DTLs we just loaded all the way up the vdev tree.
*/
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
vdev_dtl_reassess(rvd, 0, 0, B_FALSE, B_FALSE);
spa_config_exit(spa, SCL_ALL, FTAG);
return (0);
}
static int
spa_ld_load_dedup_tables(spa_t *spa)
{
int error = 0;
vdev_t *rvd = spa->spa_root_vdev;
error = ddt_load(spa);
if (error != 0) {
spa_load_failed(spa, "ddt_load failed [error=%d]", error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
return (0);
}
static int
spa_ld_verify_logs(spa_t *spa, spa_import_type_t type, char **ereport)
{
vdev_t *rvd = spa->spa_root_vdev;
if (type != SPA_IMPORT_ASSEMBLE && spa_writeable(spa)) {
boolean_t missing = spa_check_logs(spa);
if (missing) {
if (spa->spa_missing_tvds != 0) {
spa_load_note(spa, "spa_check_logs failed "
"so dropping the logs");
} else {
*ereport = FM_EREPORT_ZFS_LOG_REPLAY;
spa_load_failed(spa, "spa_check_logs failed");
return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG,
ENXIO));
}
}
}
return (0);
}
static int
spa_ld_verify_pool_data(spa_t *spa)
{
int error = 0;
vdev_t *rvd = spa->spa_root_vdev;
/*
* We've successfully opened the pool, verify that we're ready
* to start pushing transactions.
*/
if (spa->spa_load_state != SPA_LOAD_TRYIMPORT) {
error = spa_load_verify(spa);
if (error != 0) {
spa_load_failed(spa, "spa_load_verify failed "
"[error=%d]", error);
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
error));
}
}
return (0);
}
static void
spa_ld_claim_log_blocks(spa_t *spa)
{
dmu_tx_t *tx;
dsl_pool_t *dp = spa_get_dsl(spa);
/*
* Claim log blocks that haven't been committed yet.
* This must all happen in a single txg.
* Note: spa_claim_max_txg is updated by spa_claim_notify(),
* invoked from zil_claim_log_block()'s i/o done callback.
* Price of rollback is that we abandon the log.
*/
spa->spa_claiming = B_TRUE;
tx = dmu_tx_create_assigned(dp, spa_first_txg(spa));
(void) dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
zil_claim, tx, DS_FIND_CHILDREN);
dmu_tx_commit(tx);
spa->spa_claiming = B_FALSE;
spa_set_log_state(spa, SPA_LOG_GOOD);
}
static void
spa_ld_check_for_config_update(spa_t *spa, uint64_t config_cache_txg,
boolean_t update_config_cache)
{
vdev_t *rvd = spa->spa_root_vdev;
int need_update = B_FALSE;
/*
* If the config cache is stale, or we have uninitialized
* metaslabs (see spa_vdev_add()), then update the config.
*
* If this is a verbatim import, trust the current
* in-core spa_config and update the disk labels.
*/
if (update_config_cache || config_cache_txg != spa->spa_config_txg ||
spa->spa_load_state == SPA_LOAD_IMPORT ||
spa->spa_load_state == SPA_LOAD_RECOVER ||
(spa->spa_import_flags & ZFS_IMPORT_VERBATIM))
need_update = B_TRUE;
for (int c = 0; c < rvd->vdev_children; c++)
if (rvd->vdev_child[c]->vdev_ms_array == 0)
need_update = B_TRUE;
/*
* Update the config cache asynchronously in case we're the
* root pool, in which case the config cache isn't writable yet.
*/
if (need_update)
spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
}
static void
spa_ld_prepare_for_reload(spa_t *spa)
{
spa_mode_t mode = spa->spa_mode;
int async_suspended = spa->spa_async_suspended;
spa_unload(spa);
spa_deactivate(spa);
spa_activate(spa, mode);
/*
* We save the value of spa_async_suspended as it gets reset to 0 by
* spa_unload(). We want to restore it back to the original value before
* returning as we might be calling spa_async_resume() later.
*/
spa->spa_async_suspended = async_suspended;
}
static int
spa_ld_read_checkpoint_txg(spa_t *spa)
{
uberblock_t checkpoint;
int error = 0;
ASSERT0(spa->spa_checkpoint_txg);
ASSERT(MUTEX_HELD(&spa_namespace_lock));
error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t),
sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint);
if (error == ENOENT)
return (0);
if (error != 0)
return (error);
ASSERT3U(checkpoint.ub_txg, !=, 0);
ASSERT3U(checkpoint.ub_checkpoint_txg, !=, 0);
ASSERT3U(checkpoint.ub_timestamp, !=, 0);
spa->spa_checkpoint_txg = checkpoint.ub_txg;
spa->spa_checkpoint_info.sci_timestamp = checkpoint.ub_timestamp;
return (0);
}
static int
spa_ld_mos_init(spa_t *spa, spa_import_type_t type)
{
int error = 0;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
ASSERT(spa->spa_config_source != SPA_CONFIG_SRC_NONE);
/*
* Never trust the config that is provided unless we are assembling
* a pool following a split.
* This means don't trust blkptrs and the vdev tree in general. This
* also effectively puts the spa in read-only mode since
* spa_writeable() checks for spa_trust_config to be true.
* We will later load a trusted config from the MOS.
*/
if (type != SPA_IMPORT_ASSEMBLE)
spa->spa_trust_config = B_FALSE;
/*
* Parse the config provided to create a vdev tree.
*/
error = spa_ld_parse_config(spa, type);
if (error != 0)
return (error);
spa_import_progress_add(spa);
/*
* Now that we have the vdev tree, try to open each vdev. This involves
* opening the underlying physical device, retrieving its geometry and
* probing the vdev with a dummy I/O. The state of each vdev will be set
* based on the success of those operations. After this we'll be ready
* to read from the vdevs.
*/
error = spa_ld_open_vdevs(spa);
if (error != 0)
return (error);
/*
* Read the label of each vdev and make sure that the GUIDs stored
* there match the GUIDs in the config provided.
* If we're assembling a new pool that's been split off from an
* existing pool, the labels haven't yet been updated so we skip
* validation for now.
*/
if (type != SPA_IMPORT_ASSEMBLE) {
error = spa_ld_validate_vdevs(spa);
if (error != 0)
return (error);
}
/*
* Read all vdev labels to find the best uberblock (i.e. latest,
* unless spa_load_max_txg is set) and store it in spa_uberblock. We
* get the list of features required to read blkptrs in the MOS from
* the vdev label with the best uberblock and verify that our version
* of zfs supports them all.
*/
error = spa_ld_select_uberblock(spa, type);
if (error != 0)
return (error);
/*
* Pass that uberblock to the dsl_pool layer which will open the root
* blkptr. This blkptr points to the latest version of the MOS and will
* allow us to read its contents.
*/
error = spa_ld_open_rootbp(spa);
if (error != 0)
return (error);
return (0);
}
static int
spa_ld_checkpoint_rewind(spa_t *spa)
{
uberblock_t checkpoint;
int error = 0;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
ASSERT(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT);
error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t),
sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint);
if (error != 0) {
spa_load_failed(spa, "unable to retrieve checkpointed "
"uberblock from the MOS config [error=%d]", error);
if (error == ENOENT)
error = ZFS_ERR_NO_CHECKPOINT;
return (error);
}
ASSERT3U(checkpoint.ub_txg, <, spa->spa_uberblock.ub_txg);
ASSERT3U(checkpoint.ub_txg, ==, checkpoint.ub_checkpoint_txg);
/*
* We need to update the txg and timestamp of the checkpointed
* uberblock to be higher than the latest one. This ensures that
* the checkpointed uberblock is selected if we were to close and
* reopen the pool right after we've written it in the vdev labels.
* (also see block comment in vdev_uberblock_compare)
*/
checkpoint.ub_txg = spa->spa_uberblock.ub_txg + 1;
checkpoint.ub_timestamp = gethrestime_sec();
/*
* Set current uberblock to be the checkpointed uberblock.
*/
spa->spa_uberblock = checkpoint;
/*
* If we are doing a normal rewind, then the pool is open for
* writing and we sync the "updated" checkpointed uberblock to
* disk. Once this is done, we've basically rewound the whole
* pool and there is no way back.
*
* There are cases when we don't want to attempt and sync the
* checkpointed uberblock to disk because we are opening a
* pool as read-only. Specifically, verifying the checkpointed
* state with zdb, and importing the checkpointed state to get
* a "preview" of its content.
*/
if (spa_writeable(spa)) {
vdev_t *rvd = spa->spa_root_vdev;
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
vdev_t *svd[SPA_SYNC_MIN_VDEVS] = { NULL };
int svdcount = 0;
int children = rvd->vdev_children;
int c0 = random_in_range(children);
for (int c = 0; c < children; c++) {
vdev_t *vd = rvd->vdev_child[(c0 + c) % children];
/* Stop when revisiting the first vdev */
if (c > 0 && svd[0] == vd)
break;
if (vd->vdev_ms_array == 0 || vd->vdev_islog ||
!vdev_is_concrete(vd))
continue;
svd[svdcount++] = vd;
if (svdcount == SPA_SYNC_MIN_VDEVS)
break;
}
error = vdev_config_sync(svd, svdcount, spa->spa_first_txg);
if (error == 0)
spa->spa_last_synced_guid = rvd->vdev_guid;
spa_config_exit(spa, SCL_ALL, FTAG);
if (error != 0) {
spa_load_failed(spa, "failed to write checkpointed "
"uberblock to the vdev labels [error=%d]", error);
return (error);
}
}
return (0);
}
static int
spa_ld_mos_with_trusted_config(spa_t *spa, spa_import_type_t type,
boolean_t *update_config_cache)
{
int error;
/*
* Parse the config for pool, open and validate vdevs,
* select an uberblock, and use that uberblock to open
* the MOS.
*/
error = spa_ld_mos_init(spa, type);
if (error != 0)
return (error);
/*
* Retrieve the trusted config stored in the MOS and use it to create
* a new, exact version of the vdev tree, then reopen all vdevs.
*/
error = spa_ld_trusted_config(spa, type, B_FALSE);
if (error == EAGAIN) {
if (update_config_cache != NULL)
*update_config_cache = B_TRUE;
/*
* Redo the loading process with the trusted config if it is
* too different from the untrusted config.
*/
spa_ld_prepare_for_reload(spa);
spa_load_note(spa, "RELOADING");
error = spa_ld_mos_init(spa, type);
if (error != 0)
return (error);
error = spa_ld_trusted_config(spa, type, B_TRUE);
if (error != 0)
return (error);
} else if (error != 0) {
return (error);
}
return (0);
}
/*
* Load an existing storage pool, using the config provided. This config
* describes which vdevs are part of the pool and is later validated against
* partial configs present in each vdev's label and an entire copy of the
* config stored in the MOS.
*/
static int
spa_load_impl(spa_t *spa, spa_import_type_t type, char **ereport)
{
int error = 0;
boolean_t missing_feat_write = B_FALSE;
boolean_t checkpoint_rewind =
(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT);
boolean_t update_config_cache = B_FALSE;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
ASSERT(spa->spa_config_source != SPA_CONFIG_SRC_NONE);
spa_load_note(spa, "LOADING");
error = spa_ld_mos_with_trusted_config(spa, type, &update_config_cache);
if (error != 0)
return (error);
/*
* If we are rewinding to the checkpoint then we need to repeat
* everything we've done so far in this function but this time
* selecting the checkpointed uberblock and using that to open
* the MOS.
*/
if (checkpoint_rewind) {
/*
* If we are rewinding to the checkpoint update config cache
* anyway.
*/
update_config_cache = B_TRUE;
/*
* Extract the checkpointed uberblock from the current MOS
* and use this as the pool's uberblock from now on. If the
* pool is imported as writeable we also write the checkpoint
* uberblock to the labels, making the rewind permanent.
*/
error = spa_ld_checkpoint_rewind(spa);
if (error != 0)
return (error);
/*
* Redo the loading process again with the
* checkpointed uberblock.
*/
spa_ld_prepare_for_reload(spa);
spa_load_note(spa, "LOADING checkpointed uberblock");
error = spa_ld_mos_with_trusted_config(spa, type, NULL);
if (error != 0)
return (error);
}
/*
* Retrieve the checkpoint txg if the pool has a checkpoint.
*/
error = spa_ld_read_checkpoint_txg(spa);
if (error != 0)
return (error);
/*
* Retrieve the mapping of indirect vdevs. Those vdevs were removed
* from the pool and their contents were re-mapped to other vdevs. Note
* that everything that we read before this step must have been
* rewritten on concrete vdevs after the last device removal was
* initiated. Otherwise we could be reading from indirect vdevs before
* we have loaded their mappings.
*/
error = spa_ld_open_indirect_vdev_metadata(spa);
if (error != 0)
return (error);
/*
* Retrieve the full list of active features from the MOS and check if
* they are all supported.
*/
error = spa_ld_check_features(spa, &missing_feat_write);
if (error != 0)
return (error);
/*
* Load several special directories from the MOS needed by the dsl_pool
* layer.
*/
error = spa_ld_load_special_directories(spa);
if (error != 0)
return (error);
/*
* Retrieve pool properties from the MOS.
*/
error = spa_ld_get_props(spa);
if (error != 0)
return (error);
/*
* Retrieve the list of auxiliary devices - cache devices and spares -
* and open them.
*/
error = spa_ld_open_aux_vdevs(spa, type);
if (error != 0)
return (error);
/*
* Load the metadata for all vdevs. Also check if unopenable devices
* should be autoreplaced.
*/
error = spa_ld_load_vdev_metadata(spa);
if (error != 0)
return (error);
error = spa_ld_load_dedup_tables(spa);
if (error != 0)
return (error);
/*
* Verify the logs now to make sure we don't have any unexpected errors
* when we claim log blocks later.
*/
error = spa_ld_verify_logs(spa, type, ereport);
if (error != 0)
return (error);
if (missing_feat_write) {
ASSERT(spa->spa_load_state == SPA_LOAD_TRYIMPORT);
/*
* At this point, we know that we can open the pool in
* read-only mode but not read-write mode. We now have enough
* information and can return to userland.
*/
return (spa_vdev_err(spa->spa_root_vdev, VDEV_AUX_UNSUP_FEAT,
ENOTSUP));
}
/*
* Traverse the last txgs to make sure the pool was left off in a safe
* state. When performing an extreme rewind, we verify the whole pool,
* which can take a very long time.
*/
error = spa_ld_verify_pool_data(spa);
if (error != 0)
return (error);
/*
* Calculate the deflated space for the pool. This must be done before
* we write anything to the pool because we'd need to update the space
* accounting using the deflated sizes.
*/
spa_update_dspace(spa);
/*
* We have now retrieved all the information we needed to open the
* pool. If we are importing the pool in read-write mode, a few
* additional steps must be performed to finish the import.
*/
if (spa_writeable(spa) && (spa->spa_load_state == SPA_LOAD_RECOVER ||
spa->spa_load_max_txg == UINT64_MAX)) {
uint64_t config_cache_txg = spa->spa_config_txg;
ASSERT(spa->spa_load_state != SPA_LOAD_TRYIMPORT);
/*
* In case of a checkpoint rewind, log the original txg
* of the checkpointed uberblock.
*/
if (checkpoint_rewind) {
spa_history_log_internal(spa, "checkpoint rewind",
NULL, "rewound state to txg=%llu",
(u_longlong_t)spa->spa_uberblock.ub_checkpoint_txg);
}
/*
* Traverse the ZIL and claim all blocks.
*/
spa_ld_claim_log_blocks(spa);
/*
* Kick-off the syncing thread.
*/
spa->spa_sync_on = B_TRUE;
txg_sync_start(spa->spa_dsl_pool);
mmp_thread_start(spa);
/*
* Wait for all claims to sync. We sync up to the highest
* claimed log block birth time so that claimed log blocks
* don't appear to be from the future. spa_claim_max_txg
* will have been set for us by ZIL traversal operations
* performed above.
*/
txg_wait_synced(spa->spa_dsl_pool, spa->spa_claim_max_txg);
/*
* Check if we need to request an update of the config. On the
* next sync, we would update the config stored in vdev labels
* and the cachefile (by default /etc/zfs/zpool.cache).
*/
spa_ld_check_for_config_update(spa, config_cache_txg,
update_config_cache);
/*
* Check if a rebuild was in progress and if so resume it.
* Then check all DTLs to see if anything needs resilvering.
* The resilver will be deferred if a rebuild was started.
*/
if (vdev_rebuild_active(spa->spa_root_vdev)) {
vdev_rebuild_restart(spa);
} else if (!dsl_scan_resilvering(spa->spa_dsl_pool) &&
vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) {
spa_async_request(spa, SPA_ASYNC_RESILVER);
}
/*
* Log the fact that we booted up (so that we can detect if
* we rebooted in the middle of an operation).
*/
spa_history_log_version(spa, "open", NULL);
spa_restart_removal(spa);
spa_spawn_aux_threads(spa);
/*
* Delete any inconsistent datasets.
*
* Note:
* Since we may be issuing deletes for clones here,
* we make sure to do so after we've spawned all the
* auxiliary threads above (from which the livelist
* deletion zthr is part of).
*/
(void) dmu_objset_find(spa_name(spa),
dsl_destroy_inconsistent, NULL, DS_FIND_CHILDREN);
/*
* Clean up any stale temporary dataset userrefs.
*/
dsl_pool_clean_tmp_userrefs(spa->spa_dsl_pool);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vdev_initialize_restart(spa->spa_root_vdev);
vdev_trim_restart(spa->spa_root_vdev);
vdev_autotrim_restart(spa);
spa_config_exit(spa, SCL_CONFIG, FTAG);
}
spa_import_progress_remove(spa_guid(spa));
spa_async_request(spa, SPA_ASYNC_L2CACHE_REBUILD);
spa_load_note(spa, "LOADED");
return (0);
}
static int
spa_load_retry(spa_t *spa, spa_load_state_t state)
{
spa_mode_t mode = spa->spa_mode;
spa_unload(spa);
spa_deactivate(spa);
spa->spa_load_max_txg = spa->spa_uberblock.ub_txg - 1;
spa_activate(spa, mode);
spa_async_suspend(spa);
spa_load_note(spa, "spa_load_retry: rewind, max txg: %llu",
(u_longlong_t)spa->spa_load_max_txg);
return (spa_load(spa, state, SPA_IMPORT_EXISTING));
}
/*
* If spa_load() fails this function will try loading prior txg's. If
* 'state' is SPA_LOAD_RECOVER and one of these loads succeeds the pool
* will be rewound to that txg. If 'state' is not SPA_LOAD_RECOVER this
* function will not rewind the pool and will return the same error as
* spa_load().
*/
static int
spa_load_best(spa_t *spa, spa_load_state_t state, uint64_t max_request,
int rewind_flags)
{
nvlist_t *loadinfo = NULL;
nvlist_t *config = NULL;
int load_error, rewind_error;
uint64_t safe_rewind_txg;
uint64_t min_txg;
if (spa->spa_load_txg && state == SPA_LOAD_RECOVER) {
spa->spa_load_max_txg = spa->spa_load_txg;
spa_set_log_state(spa, SPA_LOG_CLEAR);
} else {
spa->spa_load_max_txg = max_request;
if (max_request != UINT64_MAX)
spa->spa_extreme_rewind = B_TRUE;
}
load_error = rewind_error = spa_load(spa, state, SPA_IMPORT_EXISTING);
if (load_error == 0)
return (0);
if (load_error == ZFS_ERR_NO_CHECKPOINT) {
/*
* When attempting checkpoint-rewind on a pool with no
* checkpoint, we should not attempt to load uberblocks
* from previous txgs when spa_load fails.
*/
ASSERT(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT);
spa_import_progress_remove(spa_guid(spa));
return (load_error);
}
if (spa->spa_root_vdev != NULL)
config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
spa->spa_last_ubsync_txg = spa->spa_uberblock.ub_txg;
spa->spa_last_ubsync_txg_ts = spa->spa_uberblock.ub_timestamp;
if (rewind_flags & ZPOOL_NEVER_REWIND) {
nvlist_free(config);
spa_import_progress_remove(spa_guid(spa));
return (load_error);
}
if (state == SPA_LOAD_RECOVER) {
/* Price of rolling back is discarding txgs, including log */
spa_set_log_state(spa, SPA_LOG_CLEAR);
} else {
/*
* If we aren't rolling back save the load info from our first
* import attempt so that we can restore it after attempting
* to rewind.
*/
loadinfo = spa->spa_load_info;
spa->spa_load_info = fnvlist_alloc();
}
spa->spa_load_max_txg = spa->spa_last_ubsync_txg;
safe_rewind_txg = spa->spa_last_ubsync_txg - TXG_DEFER_SIZE;
min_txg = (rewind_flags & ZPOOL_EXTREME_REWIND) ?
TXG_INITIAL : safe_rewind_txg;
/*
* Continue as long as we're finding errors, we're still within
* the acceptable rewind range, and we're still finding uberblocks
*/
while (rewind_error && spa->spa_uberblock.ub_txg >= min_txg &&
spa->spa_uberblock.ub_txg <= spa->spa_load_max_txg) {
if (spa->spa_load_max_txg < safe_rewind_txg)
spa->spa_extreme_rewind = B_TRUE;
rewind_error = spa_load_retry(spa, state);
}
spa->spa_extreme_rewind = B_FALSE;
spa->spa_load_max_txg = UINT64_MAX;
if (config && (rewind_error || state != SPA_LOAD_RECOVER))
spa_config_set(spa, config);
else
nvlist_free(config);
if (state == SPA_LOAD_RECOVER) {
ASSERT3P(loadinfo, ==, NULL);
spa_import_progress_remove(spa_guid(spa));
return (rewind_error);
} else {
/* Store the rewind info as part of the initial load info */
fnvlist_add_nvlist(loadinfo, ZPOOL_CONFIG_REWIND_INFO,
spa->spa_load_info);
/* Restore the initial load info */
fnvlist_free(spa->spa_load_info);
spa->spa_load_info = loadinfo;
spa_import_progress_remove(spa_guid(spa));
return (load_error);
}
}
/*
* Pool Open/Import
*
* The import case is identical to an open except that the configuration is sent
* down from userland, instead of grabbed from the configuration cache. For the
* case of an open, the pool configuration will exist in the
* POOL_STATE_UNINITIALIZED state.
*
* The stats information (gen/count/ustats) is used to gather vdev statistics at
* the same time open the pool, without having to keep around the spa_t in some
* ambiguous state.
*/
static int
spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t *nvpolicy,
nvlist_t **config)
{
spa_t *spa;
spa_load_state_t state = SPA_LOAD_OPEN;
int error;
int locked = B_FALSE;
int firstopen = B_FALSE;
*spapp = NULL;
/*
* As disgusting as this is, we need to support recursive calls to this
* function because dsl_dir_open() is called during spa_load(), and ends
* up calling spa_open() again. The real fix is to figure out how to
* avoid dsl_dir_open() calling this in the first place.
*/
if (MUTEX_NOT_HELD(&spa_namespace_lock)) {
mutex_enter(&spa_namespace_lock);
locked = B_TRUE;
}
if ((spa = spa_lookup(pool)) == NULL) {
if (locked)
mutex_exit(&spa_namespace_lock);
return (SET_ERROR(ENOENT));
}
if (spa->spa_state == POOL_STATE_UNINITIALIZED) {
zpool_load_policy_t policy;
firstopen = B_TRUE;
zpool_get_load_policy(nvpolicy ? nvpolicy : spa->spa_config,
&policy);
if (policy.zlp_rewind & ZPOOL_DO_REWIND)
state = SPA_LOAD_RECOVER;
spa_activate(spa, spa_mode_global);
if (state != SPA_LOAD_RECOVER)
spa->spa_last_ubsync_txg = spa->spa_load_txg = 0;
spa->spa_config_source = SPA_CONFIG_SRC_CACHEFILE;
zfs_dbgmsg("spa_open_common: opening %s", pool);
error = spa_load_best(spa, state, policy.zlp_txg,
policy.zlp_rewind);
if (error == EBADF) {
/*
* If vdev_validate() returns failure (indicated by
* EBADF), it indicates that one of the vdevs indicates
* that the pool has been exported or destroyed. If
* this is the case, the config cache is out of sync and
* we should remove the pool from the namespace.
*/
spa_unload(spa);
spa_deactivate(spa);
spa_write_cachefile(spa, B_TRUE, B_TRUE);
spa_remove(spa);
if (locked)
mutex_exit(&spa_namespace_lock);
return (SET_ERROR(ENOENT));
}
if (error) {
/*
* We can't open the pool, but we still have useful
* information: the state of each vdev after the
* attempted vdev_open(). Return this to the user.
*/
if (config != NULL && spa->spa_config) {
*config = fnvlist_dup(spa->spa_config);
fnvlist_add_nvlist(*config,
ZPOOL_CONFIG_LOAD_INFO,
spa->spa_load_info);
}
spa_unload(spa);
spa_deactivate(spa);
spa->spa_last_open_failed = error;
if (locked)
mutex_exit(&spa_namespace_lock);
*spapp = NULL;
return (error);
}
}
spa_open_ref(spa, tag);
if (config != NULL)
*config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
/*
* If we've recovered the pool, pass back any information we
* gathered while doing the load.
*/
if (state == SPA_LOAD_RECOVER) {
fnvlist_add_nvlist(*config, ZPOOL_CONFIG_LOAD_INFO,
spa->spa_load_info);
}
if (locked) {
spa->spa_last_open_failed = 0;
spa->spa_last_ubsync_txg = 0;
spa->spa_load_txg = 0;
mutex_exit(&spa_namespace_lock);
}
if (firstopen)
zvol_create_minors_recursive(spa_name(spa));
*spapp = spa;
return (0);
}
int
spa_open_rewind(const char *name, spa_t **spapp, void *tag, nvlist_t *policy,
nvlist_t **config)
{
return (spa_open_common(name, spapp, tag, policy, config));
}
int
spa_open(const char *name, spa_t **spapp, void *tag)
{
return (spa_open_common(name, spapp, tag, NULL, NULL));
}
/*
* Lookup the given spa_t, incrementing the inject count in the process,
* preventing it from being exported or destroyed.
*/
spa_t *
spa_inject_addref(char *name)
{
spa_t *spa;
mutex_enter(&spa_namespace_lock);
if ((spa = spa_lookup(name)) == NULL) {
mutex_exit(&spa_namespace_lock);
return (NULL);
}
spa->spa_inject_ref++;
mutex_exit(&spa_namespace_lock);
return (spa);
}
void
spa_inject_delref(spa_t *spa)
{
mutex_enter(&spa_namespace_lock);
spa->spa_inject_ref--;
mutex_exit(&spa_namespace_lock);
}
/*
* Add spares device information to the nvlist.
*/
static void
spa_add_spares(spa_t *spa, nvlist_t *config)
{
nvlist_t **spares;
uint_t i, nspares;
nvlist_t *nvroot;
uint64_t guid;
vdev_stat_t *vs;
uint_t vsc;
uint64_t pool;
ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));
if (spa->spa_spares.sav_count == 0)
return;
nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
VERIFY0(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, &spares, &nspares));
if (nspares != 0) {
fnvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
(const nvlist_t * const *)spares, nspares);
VERIFY0(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
&spares, &nspares));
/*
* Go through and find any spares which have since been
* repurposed as an active spare. If this is the case, update
* their status appropriately.
*/
for (i = 0; i < nspares; i++) {
guid = fnvlist_lookup_uint64(spares[i],
ZPOOL_CONFIG_GUID);
if (spa_spare_exists(guid, &pool, NULL) &&
pool != 0ULL) {
VERIFY0(nvlist_lookup_uint64_array(spares[i],
ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs,
&vsc));
vs->vs_state = VDEV_STATE_CANT_OPEN;
vs->vs_aux = VDEV_AUX_SPARED;
}
}
}
}
/*
* Add l2cache device information to the nvlist, including vdev stats.
*/
static void
spa_add_l2cache(spa_t *spa, nvlist_t *config)
{
nvlist_t **l2cache;
uint_t i, j, nl2cache;
nvlist_t *nvroot;
uint64_t guid;
vdev_t *vd;
vdev_stat_t *vs;
uint_t vsc;
ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));
if (spa->spa_l2cache.sav_count == 0)
return;
nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
VERIFY0(nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache));
if (nl2cache != 0) {
fnvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
(const nvlist_t * const *)l2cache, nl2cache);
VERIFY0(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
&l2cache, &nl2cache));
/*
* Update level 2 cache device stats.
*/
for (i = 0; i < nl2cache; i++) {
guid = fnvlist_lookup_uint64(l2cache[i],
ZPOOL_CONFIG_GUID);
vd = NULL;
for (j = 0; j < spa->spa_l2cache.sav_count; j++) {
if (guid ==
spa->spa_l2cache.sav_vdevs[j]->vdev_guid) {
vd = spa->spa_l2cache.sav_vdevs[j];
break;
}
}
ASSERT(vd != NULL);
VERIFY0(nvlist_lookup_uint64_array(l2cache[i],
ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc));
vdev_get_stats(vd, vs);
vdev_config_generate_stats(vd, l2cache[i]);
}
}
}
static void
spa_feature_stats_from_disk(spa_t *spa, nvlist_t *features)
{
zap_cursor_t zc;
zap_attribute_t za;
if (spa->spa_feat_for_read_obj != 0) {
for (zap_cursor_init(&zc, spa->spa_meta_objset,
spa->spa_feat_for_read_obj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
ASSERT(za.za_integer_length == sizeof (uint64_t) &&
za.za_num_integers == 1);
VERIFY0(nvlist_add_uint64(features, za.za_name,
za.za_first_integer));
}
zap_cursor_fini(&zc);
}
if (spa->spa_feat_for_write_obj != 0) {
for (zap_cursor_init(&zc, spa->spa_meta_objset,
spa->spa_feat_for_write_obj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
ASSERT(za.za_integer_length == sizeof (uint64_t) &&
za.za_num_integers == 1);
VERIFY0(nvlist_add_uint64(features, za.za_name,
za.za_first_integer));
}
zap_cursor_fini(&zc);
}
}
static void
spa_feature_stats_from_cache(spa_t *spa, nvlist_t *features)
{
int i;
for (i = 0; i < SPA_FEATURES; i++) {
zfeature_info_t feature = spa_feature_table[i];
uint64_t refcount;
if (feature_get_refcount(spa, &feature, &refcount) != 0)
continue;
VERIFY0(nvlist_add_uint64(features, feature.fi_guid, refcount));
}
}
/*
* Store a list of pool features and their reference counts in the
* config.
*
* The first time this is called on a spa, allocate a new nvlist, fetch
* the pool features and reference counts from disk, then save the list
* in the spa. In subsequent calls on the same spa use the saved nvlist
* and refresh its values from the cached reference counts. This
* ensures we don't block here on I/O on a suspended pool so 'zpool
* clear' can resume the pool.
*/
static void
spa_add_feature_stats(spa_t *spa, nvlist_t *config)
{
nvlist_t *features;
ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));
mutex_enter(&spa->spa_feat_stats_lock);
features = spa->spa_feat_stats;
if (features != NULL) {
spa_feature_stats_from_cache(spa, features);
} else {
VERIFY0(nvlist_alloc(&features, NV_UNIQUE_NAME, KM_SLEEP));
spa->spa_feat_stats = features;
spa_feature_stats_from_disk(spa, features);
}
VERIFY0(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS,
features));
mutex_exit(&spa->spa_feat_stats_lock);
}
int
spa_get_stats(const char *name, nvlist_t **config,
char *altroot, size_t buflen)
{
int error;
spa_t *spa;
*config = NULL;
error = spa_open_common(name, &spa, FTAG, NULL, config);
if (spa != NULL) {
/*
* This still leaves a window of inconsistency where the spares
* or l2cache devices could change and the config would be
* self-inconsistent.
*/
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
if (*config != NULL) {
uint64_t loadtimes[2];
loadtimes[0] = spa->spa_loaded_ts.tv_sec;
loadtimes[1] = spa->spa_loaded_ts.tv_nsec;
fnvlist_add_uint64_array(*config,
ZPOOL_CONFIG_LOADED_TIME, loadtimes, 2);
fnvlist_add_uint64(*config,
ZPOOL_CONFIG_ERRCOUNT,
spa_get_errlog_size(spa));
if (spa_suspended(spa)) {
fnvlist_add_uint64(*config,
ZPOOL_CONFIG_SUSPENDED,
spa->spa_failmode);
fnvlist_add_uint64(*config,
ZPOOL_CONFIG_SUSPENDED_REASON,
spa->spa_suspended);
}
spa_add_spares(spa, *config);
spa_add_l2cache(spa, *config);
spa_add_feature_stats(spa, *config);
}
}
/*
* We want to get the alternate root even for faulted pools, so we cheat
* and call spa_lookup() directly.
*/
if (altroot) {
if (spa == NULL) {
mutex_enter(&spa_namespace_lock);
spa = spa_lookup(name);
if (spa)
spa_altroot(spa, altroot, buflen);
else
altroot[0] = '\0';
spa = NULL;
mutex_exit(&spa_namespace_lock);
} else {
spa_altroot(spa, altroot, buflen);
}
}
if (spa != NULL) {
spa_config_exit(spa, SCL_CONFIG, FTAG);
spa_close(spa, FTAG);
}
return (error);
}
/*
* Validate that the auxiliary device array is well formed. We must have an
* array of nvlists, each which describes a valid leaf vdev. If this is an
* import (mode is VDEV_ALLOC_SPARE), then we allow corrupted spares to be
* specified, as long as they are well-formed.
*/
static int
spa_validate_aux_devs(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode,
spa_aux_vdev_t *sav, const char *config, uint64_t version,
vdev_labeltype_t label)
{
nvlist_t **dev;
uint_t i, ndev;
vdev_t *vd;
int error;
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
/*
* It's acceptable to have no devs specified.
*/
if (nvlist_lookup_nvlist_array(nvroot, config, &dev, &ndev) != 0)
return (0);
if (ndev == 0)
return (SET_ERROR(EINVAL));
/*
* Make sure the pool is formatted with a version that supports this
* device type.
*/
if (spa_version(spa) < version)
return (SET_ERROR(ENOTSUP));
/*
* Set the pending device list so we correctly handle device in-use
* checking.
*/
sav->sav_pending = dev;
sav->sav_npending = ndev;
for (i = 0; i < ndev; i++) {
if ((error = spa_config_parse(spa, &vd, dev[i], NULL, 0,
mode)) != 0)
goto out;
if (!vd->vdev_ops->vdev_op_leaf) {
vdev_free(vd);
error = SET_ERROR(EINVAL);
goto out;
}
vd->vdev_top = vd;
if ((error = vdev_open(vd)) == 0 &&
(error = vdev_label_init(vd, crtxg, label)) == 0) {
fnvlist_add_uint64(dev[i], ZPOOL_CONFIG_GUID,
vd->vdev_guid);
}
vdev_free(vd);
if (error &&
(mode != VDEV_ALLOC_SPARE && mode != VDEV_ALLOC_L2CACHE))
goto out;
else
error = 0;
}
out:
sav->sav_pending = NULL;
sav->sav_npending = 0;
return (error);
}
static int
spa_validate_aux(spa_t *spa, nvlist_t *nvroot, uint64_t crtxg, int mode)
{
int error;
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
if ((error = spa_validate_aux_devs(spa, nvroot, crtxg, mode,
&spa->spa_spares, ZPOOL_CONFIG_SPARES, SPA_VERSION_SPARES,
VDEV_LABEL_SPARE)) != 0) {
return (error);
}
return (spa_validate_aux_devs(spa, nvroot, crtxg, mode,
&spa->spa_l2cache, ZPOOL_CONFIG_L2CACHE, SPA_VERSION_L2CACHE,
VDEV_LABEL_L2CACHE));
}
static void
spa_set_aux_vdevs(spa_aux_vdev_t *sav, nvlist_t **devs, int ndevs,
const char *config)
{
int i;
if (sav->sav_config != NULL) {
nvlist_t **olddevs;
uint_t oldndevs;
nvlist_t **newdevs;
/*
* Generate new dev list by concatenating with the
* current dev list.
*/
VERIFY0(nvlist_lookup_nvlist_array(sav->sav_config, config,
&olddevs, &oldndevs));
newdevs = kmem_alloc(sizeof (void *) *
(ndevs + oldndevs), KM_SLEEP);
for (i = 0; i < oldndevs; i++)
newdevs[i] = fnvlist_dup(olddevs[i]);
for (i = 0; i < ndevs; i++)
newdevs[i + oldndevs] = fnvlist_dup(devs[i]);
fnvlist_remove(sav->sav_config, config);
fnvlist_add_nvlist_array(sav->sav_config, config,
(const nvlist_t * const *)newdevs, ndevs + oldndevs);
for (i = 0; i < oldndevs + ndevs; i++)
nvlist_free(newdevs[i]);
kmem_free(newdevs, (oldndevs + ndevs) * sizeof (void *));
} else {
/*
* Generate a new dev list.
*/
sav->sav_config = fnvlist_alloc();
fnvlist_add_nvlist_array(sav->sav_config, config,
(const nvlist_t * const *)devs, ndevs);
}
}
/*
* Stop and drop level 2 ARC devices
*/
void
spa_l2cache_drop(spa_t *spa)
{
vdev_t *vd;
int i;
spa_aux_vdev_t *sav = &spa->spa_l2cache;
for (i = 0; i < sav->sav_count; i++) {
uint64_t pool;
vd = sav->sav_vdevs[i];
ASSERT(vd != NULL);
if (spa_l2cache_exists(vd->vdev_guid, &pool) &&
pool != 0ULL && l2arc_vdev_present(vd))
l2arc_remove_vdev(vd);
}
}
/*
* Verify encryption parameters for spa creation. If we are encrypting, we must
* have the encryption feature flag enabled.
*/
static int
spa_create_check_encryption_params(dsl_crypto_params_t *dcp,
boolean_t has_encryption)
{
if (dcp->cp_crypt != ZIO_CRYPT_OFF &&
dcp->cp_crypt != ZIO_CRYPT_INHERIT &&
!has_encryption)
return (SET_ERROR(ENOTSUP));
return (dmu_objset_create_crypt_check(NULL, dcp, NULL));
}
/*
* Pool Creation
*/
int
spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
nvlist_t *zplprops, dsl_crypto_params_t *dcp)
{
spa_t *spa;
char *altroot = NULL;
vdev_t *rvd;
dsl_pool_t *dp;
dmu_tx_t *tx;
int error = 0;
uint64_t txg = TXG_INITIAL;
nvlist_t **spares, **l2cache;
uint_t nspares, nl2cache;
uint64_t version, obj, ndraid = 0;
boolean_t has_features;
boolean_t has_encryption;
boolean_t has_allocclass;
spa_feature_t feat;
char *feat_name;
char *poolname;
nvlist_t *nvl;
if (props == NULL ||
nvlist_lookup_string(props, "tname", &poolname) != 0)
poolname = (char *)pool;
/*
* If this pool already exists, return failure.
*/
mutex_enter(&spa_namespace_lock);
if (spa_lookup(poolname) != NULL) {
mutex_exit(&spa_namespace_lock);
return (SET_ERROR(EEXIST));
}
/*
* Allocate a new spa_t structure.
*/
nvl = fnvlist_alloc();
fnvlist_add_string(nvl, ZPOOL_CONFIG_POOL_NAME, pool);
(void) nvlist_lookup_string(props,
zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
spa = spa_add(poolname, nvl, altroot);
fnvlist_free(nvl);
spa_activate(spa, spa_mode_global);
if (props && (error = spa_prop_validate(spa, props))) {
spa_deactivate(spa);
spa_remove(spa);
mutex_exit(&spa_namespace_lock);
return (error);
}
/*
* Temporary pool names should never be written to disk.
*/
if (poolname != pool)
spa->spa_import_flags |= ZFS_IMPORT_TEMP_NAME;
has_features = B_FALSE;
has_encryption = B_FALSE;
has_allocclass = B_FALSE;
for (nvpair_t *elem = nvlist_next_nvpair(props, NULL);
elem != NULL; elem = nvlist_next_nvpair(props, elem)) {
if (zpool_prop_feature(nvpair_name(elem))) {
has_features = B_TRUE;
feat_name = strchr(nvpair_name(elem), '@') + 1;
VERIFY0(zfeature_lookup_name(feat_name, &feat));
if (feat == SPA_FEATURE_ENCRYPTION)
has_encryption = B_TRUE;
if (feat == SPA_FEATURE_ALLOCATION_CLASSES)
has_allocclass = B_TRUE;
}
}
/* verify encryption params, if they were provided */
if (dcp != NULL) {
error = spa_create_check_encryption_params(dcp, has_encryption);
if (error != 0) {
spa_deactivate(spa);
spa_remove(spa);
mutex_exit(&spa_namespace_lock);
return (error);
}
}
if (!has_allocclass && zfs_special_devs(nvroot, NULL)) {
spa_deactivate(spa);
spa_remove(spa);
mutex_exit(&spa_namespace_lock);
return (ENOTSUP);
}
if (has_features || nvlist_lookup_uint64(props,
zpool_prop_to_name(ZPOOL_PROP_VERSION), &version) != 0) {
version = SPA_VERSION;
}
ASSERT(SPA_VERSION_IS_SUPPORTED(version));
spa->spa_first_txg = txg;
spa->spa_uberblock.ub_txg = txg - 1;
spa->spa_uberblock.ub_version = version;
spa->spa_ubsync = spa->spa_uberblock;
spa->spa_load_state = SPA_LOAD_CREATE;
spa->spa_removing_phys.sr_state = DSS_NONE;
spa->spa_removing_phys.sr_removing_vdev = -1;
spa->spa_removing_phys.sr_prev_indirect_vdev = -1;
spa->spa_indirect_vdevs_loaded = B_TRUE;
/*
* Create "The Godfather" zio to hold all async IOs
*/
spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *),
KM_SLEEP);
for (int i = 0; i < max_ncpus; i++) {
spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE |
ZIO_FLAG_GODFATHER);
}
/*
* Create the root vdev.
*/
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_ADD);
ASSERT(error != 0 || rvd != NULL);
ASSERT(error != 0 || spa->spa_root_vdev == rvd);
if (error == 0 && !zfs_allocatable_devs(nvroot))
error = SET_ERROR(EINVAL);
if (error == 0 &&
(error = vdev_create(rvd, txg, B_FALSE)) == 0 &&
(error = vdev_draid_spare_create(nvroot, rvd, &ndraid, 0)) == 0 &&
(error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) == 0) {
/*
* instantiate the metaslab groups (this will dirty the vdevs)
* we can no longer error exit past this point
*/
for (int c = 0; error == 0 && c < rvd->vdev_children; c++) {
vdev_t *vd = rvd->vdev_child[c];
vdev_metaslab_set_size(vd);
vdev_expand(vd, txg);
}
}
spa_config_exit(spa, SCL_ALL, FTAG);
if (error != 0) {
spa_unload(spa);
spa_deactivate(spa);
spa_remove(spa);
mutex_exit(&spa_namespace_lock);
return (error);
}
/*
* Get the list of spares, if specified.
*/
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
&spares, &nspares) == 0) {
spa->spa_spares.sav_config = fnvlist_alloc();
fnvlist_add_nvlist_array(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares,
nspares);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa_load_spares(spa);
spa_config_exit(spa, SCL_ALL, FTAG);
spa->spa_spares.sav_sync = B_TRUE;
}
/*
* Get the list of level 2 cache devices, if specified.
*/
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
&l2cache, &nl2cache) == 0) {
VERIFY0(nvlist_alloc(&spa->spa_l2cache.sav_config,
NV_UNIQUE_NAME, KM_SLEEP));
fnvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache,
nl2cache);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa_load_l2cache(spa);
spa_config_exit(spa, SCL_ALL, FTAG);
spa->spa_l2cache.sav_sync = B_TRUE;
}
spa->spa_is_initializing = B_TRUE;
spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, dcp, txg);
spa->spa_is_initializing = B_FALSE;
/*
* Create DDTs (dedup tables).
*/
ddt_create(spa);
spa_update_dspace(spa);
tx = dmu_tx_create_assigned(dp, txg);
/*
* Create the pool's history object.
*/
if (version >= SPA_VERSION_ZPOOL_HISTORY && !spa->spa_history)
spa_history_create_obj(spa, tx);
spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_CREATE);
spa_history_log_version(spa, "create", tx);
/*
* Create the pool config object.
*/
spa->spa_config_object = dmu_object_alloc(spa->spa_meta_objset,
DMU_OT_PACKED_NVLIST, SPA_CONFIG_BLOCKSIZE,
DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx);
if (zap_add(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG,
sizeof (uint64_t), 1, &spa->spa_config_object, tx) != 0) {
cmn_err(CE_PANIC, "failed to add pool config");
}
if (zap_add(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION,
sizeof (uint64_t), 1, &version, tx) != 0) {
cmn_err(CE_PANIC, "failed to add pool version");
}
/* Newly created pools with the right version are always deflated. */
if (version >= SPA_VERSION_RAIDZ_DEFLATE) {
spa->spa_deflate = TRUE;
if (zap_add(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) {
cmn_err(CE_PANIC, "failed to add deflate");
}
}
/*
* Create the deferred-free bpobj. Turn off compression
* because sync-to-convergence takes longer if the blocksize
* keeps changing.
*/
obj = bpobj_alloc(spa->spa_meta_objset, 1 << 14, tx);
dmu_object_set_compress(spa->spa_meta_objset, obj,
ZIO_COMPRESS_OFF, tx);
if (zap_add(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPOBJ,
sizeof (uint64_t), 1, &obj, tx) != 0) {
cmn_err(CE_PANIC, "failed to add bpobj");
}
VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj,
spa->spa_meta_objset, obj));
/*
* Generate some random noise for salted checksums to operate on.
*/
(void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes,
sizeof (spa->spa_cksum_salt.zcs_bytes));
/*
* Set pool properties.
*/
spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS);
spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE);
spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND);
spa->spa_multihost = zpool_prop_default_numeric(ZPOOL_PROP_MULTIHOST);
spa->spa_autotrim = zpool_prop_default_numeric(ZPOOL_PROP_AUTOTRIM);
if (props != NULL) {
spa_configfile_set(spa, props, B_FALSE);
spa_sync_props(props, tx);
}
for (int i = 0; i < ndraid; i++)
spa_feature_incr(spa, SPA_FEATURE_DRAID, tx);
dmu_tx_commit(tx);
spa->spa_sync_on = B_TRUE;
txg_sync_start(dp);
mmp_thread_start(spa);
txg_wait_synced(dp, txg);
spa_spawn_aux_threads(spa);
spa_write_cachefile(spa, B_FALSE, B_TRUE);
/*
* Don't count references from objsets that are already closed
* and are making their way through the eviction process.
*/
spa_evicting_os_wait(spa);
spa->spa_minref = zfs_refcount_count(&spa->spa_refcount);
spa->spa_load_state = SPA_LOAD_NONE;
mutex_exit(&spa_namespace_lock);
return (0);
}
/*
* Import a non-root pool into the system.
*/
int
spa_import(char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags)
{
spa_t *spa;
char *altroot = NULL;
spa_load_state_t state = SPA_LOAD_IMPORT;
zpool_load_policy_t policy;
spa_mode_t mode = spa_mode_global;
uint64_t readonly = B_FALSE;
int error;
nvlist_t *nvroot;
nvlist_t **spares, **l2cache;
uint_t nspares, nl2cache;
/*
* If a pool with this name exists, return failure.
*/
mutex_enter(&spa_namespace_lock);
if (spa_lookup(pool) != NULL) {
mutex_exit(&spa_namespace_lock);
return (SET_ERROR(EEXIST));
}
/*
* Create and initialize the spa structure.
*/
(void) nvlist_lookup_string(props,
zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
(void) nvlist_lookup_uint64(props,
zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
if (readonly)
mode = SPA_MODE_READ;
spa = spa_add(pool, config, altroot);
spa->spa_import_flags = flags;
/*
* Verbatim import - Take a pool and insert it into the namespace
* as if it had been loaded at boot.
*/
if (spa->spa_import_flags & ZFS_IMPORT_VERBATIM) {
if (props != NULL)
spa_configfile_set(spa, props, B_FALSE);
spa_write_cachefile(spa, B_FALSE, B_TRUE);
spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);
zfs_dbgmsg("spa_import: verbatim import of %s", pool);
mutex_exit(&spa_namespace_lock);
return (0);
}
spa_activate(spa, mode);
/*
* Don't start async tasks until we know everything is healthy.
*/
spa_async_suspend(spa);
zpool_get_load_policy(config, &policy);
if (policy.zlp_rewind & ZPOOL_DO_REWIND)
state = SPA_LOAD_RECOVER;
spa->spa_config_source = SPA_CONFIG_SRC_TRYIMPORT;
if (state != SPA_LOAD_RECOVER) {
spa->spa_last_ubsync_txg = spa->spa_load_txg = 0;
zfs_dbgmsg("spa_import: importing %s", pool);
} else {
zfs_dbgmsg("spa_import: importing %s, max_txg=%lld "
"(RECOVERY MODE)", pool, (longlong_t)policy.zlp_txg);
}
error = spa_load_best(spa, state, policy.zlp_txg, policy.zlp_rewind);
/*
* Propagate anything learned while loading the pool and pass it
* back to caller (i.e. rewind info, missing devices, etc).
*/
fnvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
/*
* Toss any existing sparelist, as it doesn't have any validity
* anymore, and conflicts with spa_has_spare().
*/
if (spa->spa_spares.sav_config) {
nvlist_free(spa->spa_spares.sav_config);
spa->spa_spares.sav_config = NULL;
spa_load_spares(spa);
}
if (spa->spa_l2cache.sav_config) {
nvlist_free(spa->spa_l2cache.sav_config);
spa->spa_l2cache.sav_config = NULL;
spa_load_l2cache(spa);
}
nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
spa_config_exit(spa, SCL_ALL, FTAG);
if (props != NULL)
spa_configfile_set(spa, props, B_FALSE);
if (error != 0 || (props && spa_writeable(spa) &&
(error = spa_prop_set(spa, props)))) {
spa_unload(spa);
spa_deactivate(spa);
spa_remove(spa);
mutex_exit(&spa_namespace_lock);
return (error);
}
spa_async_resume(spa);
/*
* Override any spares and level 2 cache devices as specified by
* the user, as these may have correct device names/devids, etc.
*/
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
&spares, &nspares) == 0) {
if (spa->spa_spares.sav_config)
fnvlist_remove(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES);
else
spa->spa_spares.sav_config = fnvlist_alloc();
fnvlist_add_nvlist_array(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares,
nspares);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa_load_spares(spa);
spa_config_exit(spa, SCL_ALL, FTAG);
spa->spa_spares.sav_sync = B_TRUE;
}
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
&l2cache, &nl2cache) == 0) {
if (spa->spa_l2cache.sav_config)
fnvlist_remove(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE);
else
spa->spa_l2cache.sav_config = fnvlist_alloc();
fnvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache,
nl2cache);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa_load_l2cache(spa);
spa_config_exit(spa, SCL_ALL, FTAG);
spa->spa_l2cache.sav_sync = B_TRUE;
}
/*
* Check for any removed devices.
*/
if (spa->spa_autoreplace) {
spa_aux_check_removed(&spa->spa_spares);
spa_aux_check_removed(&spa->spa_l2cache);
}
if (spa_writeable(spa)) {
/*
* Update the config cache to include the newly-imported pool.
*/
spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
}
/*
* It's possible that the pool was expanded while it was exported.
* We kick off an async task to handle this for us.
*/
spa_async_request(spa, SPA_ASYNC_AUTOEXPAND);
spa_history_log_version(spa, "import", NULL);
spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);
mutex_exit(&spa_namespace_lock);
zvol_create_minors_recursive(pool);
return (0);
}
nvlist_t *
spa_tryimport(nvlist_t *tryconfig)
{
nvlist_t *config = NULL;
char *poolname, *cachefile;
spa_t *spa;
uint64_t state;
int error;
zpool_load_policy_t policy;
if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_POOL_NAME, &poolname))
return (NULL);
if (nvlist_lookup_uint64(tryconfig, ZPOOL_CONFIG_POOL_STATE, &state))
return (NULL);
/*
* Create and initialize the spa structure.
*/
mutex_enter(&spa_namespace_lock);
spa = spa_add(TRYIMPORT_NAME, tryconfig, NULL);
spa_activate(spa, SPA_MODE_READ);
/*
* Rewind pool if a max txg was provided.
*/
zpool_get_load_policy(spa->spa_config, &policy);
if (policy.zlp_txg != UINT64_MAX) {
spa->spa_load_max_txg = policy.zlp_txg;
spa->spa_extreme_rewind = B_TRUE;
zfs_dbgmsg("spa_tryimport: importing %s, max_txg=%lld",
poolname, (longlong_t)policy.zlp_txg);
} else {
zfs_dbgmsg("spa_tryimport: importing %s", poolname);
}
if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_CACHEFILE, &cachefile)
== 0) {
zfs_dbgmsg("spa_tryimport: using cachefile '%s'", cachefile);
spa->spa_config_source = SPA_CONFIG_SRC_CACHEFILE;
} else {
spa->spa_config_source = SPA_CONFIG_SRC_SCAN;
}
error = spa_load(spa, SPA_LOAD_TRYIMPORT, SPA_IMPORT_EXISTING);
/*
* If 'tryconfig' was at least parsable, return the current config.
*/
if (spa->spa_root_vdev != NULL) {
config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, poolname);
fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, state);
fnvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
spa->spa_uberblock.ub_timestamp);
fnvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO,
spa->spa_load_info);
fnvlist_add_uint64(config, ZPOOL_CONFIG_ERRATA,
spa->spa_errata);
/*
* If the bootfs property exists on this pool then we
* copy it out so that external consumers can tell which
* pools are bootable.
*/
if ((!error || error == EEXIST) && spa->spa_bootfs) {
char *tmpname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
/*
* We have to play games with the name since the
* pool was opened as TRYIMPORT_NAME.
*/
if (dsl_dsobj_to_dsname(spa_name(spa),
spa->spa_bootfs, tmpname) == 0) {
char *cp;
char *dsname;
dsname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
cp = strchr(tmpname, '/');
if (cp == NULL) {
(void) strlcpy(dsname, tmpname,
MAXPATHLEN);
} else {
(void) snprintf(dsname, MAXPATHLEN,
"%s/%s", poolname, ++cp);
}
fnvlist_add_string(config, ZPOOL_CONFIG_BOOTFS,
dsname);
kmem_free(dsname, MAXPATHLEN);
}
kmem_free(tmpname, MAXPATHLEN);
}
/*
* Add the list of hot spares and level 2 cache devices.
*/
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
spa_add_spares(spa, config);
spa_add_l2cache(spa, config);
spa_config_exit(spa, SCL_CONFIG, FTAG);
}
spa_unload(spa);
spa_deactivate(spa);
spa_remove(spa);
mutex_exit(&spa_namespace_lock);
return (config);
}
/*
* Pool export/destroy
*
* The act of destroying or exporting a pool is very simple. We make sure there
* is no more pending I/O and any references to the pool are gone. Then, we
* update the pool state and sync all the labels to disk, removing the
* configuration from the cache afterwards. If the 'hardforce' flag is set, then
* we don't sync the labels or remove the configuration cache.
*/
static int
spa_export_common(const char *pool, int new_state, nvlist_t **oldconfig,
boolean_t force, boolean_t hardforce)
{
int error;
spa_t *spa;
if (oldconfig)
*oldconfig = NULL;
if (!(spa_mode_global & SPA_MODE_WRITE))
return (SET_ERROR(EROFS));
mutex_enter(&spa_namespace_lock);
if ((spa = spa_lookup(pool)) == NULL) {
mutex_exit(&spa_namespace_lock);
return (SET_ERROR(ENOENT));
}
if (spa->spa_is_exporting) {
/* the pool is being exported by another thread */
mutex_exit(&spa_namespace_lock);
return (SET_ERROR(ZFS_ERR_EXPORT_IN_PROGRESS));
}
spa->spa_is_exporting = B_TRUE;
/*
* Put a hold on the pool, drop the namespace lock, stop async tasks,
* reacquire the namespace lock, and see if we can export.
*/
spa_open_ref(spa, FTAG);
mutex_exit(&spa_namespace_lock);
spa_async_suspend(spa);
if (spa->spa_zvol_taskq) {
zvol_remove_minors(spa, spa_name(spa), B_TRUE);
taskq_wait(spa->spa_zvol_taskq);
}
mutex_enter(&spa_namespace_lock);
spa_close(spa, FTAG);
if (spa->spa_state == POOL_STATE_UNINITIALIZED)
goto export_spa;
/*
* The pool will be in core if it's openable, in which case we can
* modify its state. Objsets may be open only because they're dirty,
* so we have to force it to sync before checking spa_refcnt.
*/
if (spa->spa_sync_on) {
txg_wait_synced(spa->spa_dsl_pool, 0);
spa_evicting_os_wait(spa);
}
/*
* A pool cannot be exported or destroyed if there are active
* references. If we are resetting a pool, allow references by
* fault injection handlers.
*/
if (!spa_refcount_zero(spa) || (spa->spa_inject_ref != 0)) {
error = SET_ERROR(EBUSY);
goto fail;
}
if (spa->spa_sync_on) {
/*
* A pool cannot be exported if it has an active shared spare.
* This is to prevent other pools stealing the active spare
* from an exported pool. At user's own will, such pool can
* be forcedly exported.
*/
if (!force && new_state == POOL_STATE_EXPORTED &&
spa_has_active_shared_spare(spa)) {
error = SET_ERROR(EXDEV);
goto fail;
}
/*
* We're about to export or destroy this pool. Make sure
* we stop all initialization and trim activity here before
* we set the spa_final_txg. This will ensure that all
* dirty data resulting from the initialization is
* committed to disk before we unload the pool.
*/
if (spa->spa_root_vdev != NULL) {
vdev_t *rvd = spa->spa_root_vdev;
vdev_initialize_stop_all(rvd, VDEV_INITIALIZE_ACTIVE);
vdev_trim_stop_all(rvd, VDEV_TRIM_ACTIVE);
vdev_autotrim_stop_all(spa);
vdev_rebuild_stop_all(spa);
}
/*
* We want this to be reflected on every label,
* so mark them all dirty. spa_unload() will do the
* final sync that pushes these changes out.
*/
if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) {
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa->spa_state = new_state;
spa->spa_final_txg = spa_last_synced_txg(spa) +
TXG_DEFER_SIZE + 1;
vdev_config_dirty(spa->spa_root_vdev);
spa_config_exit(spa, SCL_ALL, FTAG);
}
}
export_spa:
if (new_state == POOL_STATE_DESTROYED)
spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_DESTROY);
else if (new_state == POOL_STATE_EXPORTED)
spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_EXPORT);
if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
spa_unload(spa);
spa_deactivate(spa);
}
if (oldconfig && spa->spa_config)
*oldconfig = fnvlist_dup(spa->spa_config);
if (new_state != POOL_STATE_UNINITIALIZED) {
if (!hardforce)
spa_write_cachefile(spa, B_TRUE, B_TRUE);
spa_remove(spa);
} else {
/*
* If spa_remove() is not called for this spa_t and
* there is any possibility that it can be reused,
* we make sure to reset the exporting flag.
*/
spa->spa_is_exporting = B_FALSE;
}
mutex_exit(&spa_namespace_lock);
return (0);
fail:
spa->spa_is_exporting = B_FALSE;
spa_async_resume(spa);
mutex_exit(&spa_namespace_lock);
return (error);
}
/*
* Destroy a storage pool.
*/
int
spa_destroy(const char *pool)
{
return (spa_export_common(pool, POOL_STATE_DESTROYED, NULL,
B_FALSE, B_FALSE));
}
/*
* Export a storage pool.
*/
int
spa_export(const char *pool, nvlist_t **oldconfig, boolean_t force,
boolean_t hardforce)
{
return (spa_export_common(pool, POOL_STATE_EXPORTED, oldconfig,
force, hardforce));
}
/*
* Similar to spa_export(), this unloads the spa_t without actually removing it
* from the namespace in any way.
*/
int
spa_reset(const char *pool)
{
return (spa_export_common(pool, POOL_STATE_UNINITIALIZED, NULL,
B_FALSE, B_FALSE));
}
/*
* ==========================================================================
* Device manipulation
* ==========================================================================
*/
/*
* This is called as a synctask to increment the draid feature flag
*/
static void
spa_draid_feature_incr(void *arg, dmu_tx_t *tx)
{
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
int draid = (int)(uintptr_t)arg;
for (int c = 0; c < draid; c++)
spa_feature_incr(spa, SPA_FEATURE_DRAID, tx);
}
/*
* Add a device to a storage pool.
*/
int
spa_vdev_add(spa_t *spa, nvlist_t *nvroot)
{
uint64_t txg, ndraid = 0;
int error;
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *vd, *tvd;
nvlist_t **spares, **l2cache;
uint_t nspares, nl2cache;
ASSERT(spa_writeable(spa));
txg = spa_vdev_enter(spa);
if ((error = spa_config_parse(spa, &vd, nvroot, NULL, 0,
VDEV_ALLOC_ADD)) != 0)
return (spa_vdev_exit(spa, NULL, txg, error));
spa->spa_pending_vdev = vd; /* spa_vdev_exit() will clear this */
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares,
&nspares) != 0)
nspares = 0;
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache,
&nl2cache) != 0)
nl2cache = 0;
if (vd->vdev_children == 0 && nspares == 0 && nl2cache == 0)
return (spa_vdev_exit(spa, vd, txg, EINVAL));
if (vd->vdev_children != 0 &&
(error = vdev_create(vd, txg, B_FALSE)) != 0) {
return (spa_vdev_exit(spa, vd, txg, error));
}
/*
* The virtual dRAID spares must be added after vdev tree is created
* and the vdev guids are generated. The guid of their associated
* dRAID is stored in the config and used when opening the spare.
*/
if ((error = vdev_draid_spare_create(nvroot, vd, &ndraid,
rvd->vdev_children)) == 0) {
if (ndraid > 0 && nvlist_lookup_nvlist_array(nvroot,
ZPOOL_CONFIG_SPARES, &spares, &nspares) != 0)
nspares = 0;
} else {
return (spa_vdev_exit(spa, vd, txg, error));
}
/*
* We must validate the spares and l2cache devices after checking the
* children. Otherwise, vdev_inuse() will blindly overwrite the spare.
*/
if ((error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) != 0)
return (spa_vdev_exit(spa, vd, txg, error));
/*
* If we are in the middle of a device removal, we can only add
* devices which match the existing devices in the pool.
* If we are in the middle of a removal, or have some indirect
* vdevs, we can not add raidz or dRAID top levels.
*/
if (spa->spa_vdev_removal != NULL ||
spa->spa_removing_phys.sr_prev_indirect_vdev != -1) {
for (int c = 0; c < vd->vdev_children; c++) {
tvd = vd->vdev_child[c];
if (spa->spa_vdev_removal != NULL &&
tvd->vdev_ashift != spa->spa_max_ashift) {
return (spa_vdev_exit(spa, vd, txg, EINVAL));
}
/* Fail if top level vdev is raidz or a dRAID */
if (vdev_get_nparity(tvd) != 0)
return (spa_vdev_exit(spa, vd, txg, EINVAL));
/*
* Need the top level mirror to be
* a mirror of leaf vdevs only
*/
if (tvd->vdev_ops == &vdev_mirror_ops) {
for (uint64_t cid = 0;
cid < tvd->vdev_children; cid++) {
vdev_t *cvd = tvd->vdev_child[cid];
if (!cvd->vdev_ops->vdev_op_leaf) {
return (spa_vdev_exit(spa, vd,
txg, EINVAL));
}
}
}
}
}
for (int c = 0; c < vd->vdev_children; c++) {
tvd = vd->vdev_child[c];
vdev_remove_child(vd, tvd);
tvd->vdev_id = rvd->vdev_children;
vdev_add_child(rvd, tvd);
vdev_config_dirty(tvd);
}
if (nspares != 0) {
spa_set_aux_vdevs(&spa->spa_spares, spares, nspares,
ZPOOL_CONFIG_SPARES);
spa_load_spares(spa);
spa->spa_spares.sav_sync = B_TRUE;
}
if (nl2cache != 0) {
spa_set_aux_vdevs(&spa->spa_l2cache, l2cache, nl2cache,
ZPOOL_CONFIG_L2CACHE);
spa_load_l2cache(spa);
spa->spa_l2cache.sav_sync = B_TRUE;
}
/*
* We can't increment a feature while holding spa_vdev so we
* have to do it in a synctask.
*/
if (ndraid != 0) {
dmu_tx_t *tx;
tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
dsl_sync_task_nowait(spa->spa_dsl_pool, spa_draid_feature_incr,
(void *)(uintptr_t)ndraid, tx);
dmu_tx_commit(tx);
}
/*
* We have to be careful when adding new vdevs to an existing pool.
* If other threads start allocating from these vdevs before we
* sync the config cache, and we lose power, then upon reboot we may
* fail to open the pool because there are DVAs that the config cache
* can't translate. Therefore, we first add the vdevs without
* initializing metaslabs; sync the config cache (via spa_vdev_exit());
* and then let spa_config_update() initialize the new metaslabs.
*
* spa_load() checks for added-but-not-initialized vdevs, so that
* if we lose power at any point in this sequence, the remaining
* steps will be completed the next time we load the pool.
*/
(void) spa_vdev_exit(spa, vd, txg, 0);
mutex_enter(&spa_namespace_lock);
spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
spa_event_notify(spa, NULL, NULL, ESC_ZFS_VDEV_ADD);
mutex_exit(&spa_namespace_lock);
return (0);
}
/*
* Attach a device to a mirror. The arguments are the path to any device
* in the mirror, and the nvroot for the new device. If the path specifies
* a device that is not mirrored, we automatically insert the mirror vdev.
*
* If 'replacing' is specified, the new device is intended to replace the
* existing device; in this case the two devices are made into their own
* mirror using the 'replacing' vdev, which is functionally identical to
* the mirror vdev (it actually reuses all the same ops) but has a few
* extra rules: you can't attach to it after it's been created, and upon
* completion of resilvering, the first disk (the one being replaced)
* is automatically detached.
*
* If 'rebuild' is specified, then sequential reconstruction (a.ka. rebuild)
* should be performed instead of traditional healing reconstruction. From
* an administrators perspective these are both resilver operations.
*/
int
spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing,
int rebuild)
{
uint64_t txg, dtl_max_txg;
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *oldvd, *newvd, *newrootvd, *pvd, *tvd;
vdev_ops_t *pvops;
char *oldvdpath, *newvdpath;
int newvd_isspare;
int error;
ASSERT(spa_writeable(spa));
txg = spa_vdev_enter(spa);
oldvd = spa_lookup_by_guid(spa, guid, B_FALSE);
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
error = (spa_has_checkpoint(spa)) ?
ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
return (spa_vdev_exit(spa, NULL, txg, error));
}
if (rebuild) {
if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD))
return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
if (dsl_scan_resilvering(spa_get_dsl(spa)))
return (spa_vdev_exit(spa, NULL, txg,
ZFS_ERR_RESILVER_IN_PROGRESS));
} else {
if (vdev_rebuild_active(rvd))
return (spa_vdev_exit(spa, NULL, txg,
ZFS_ERR_REBUILD_IN_PROGRESS));
}
if (spa->spa_vdev_removal != NULL)
return (spa_vdev_exit(spa, NULL, txg, EBUSY));
if (oldvd == NULL)
return (spa_vdev_exit(spa, NULL, txg, ENODEV));
if (!oldvd->vdev_ops->vdev_op_leaf)
return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
pvd = oldvd->vdev_parent;
if ((error = spa_config_parse(spa, &newrootvd, nvroot, NULL, 0,
VDEV_ALLOC_ATTACH)) != 0)
return (spa_vdev_exit(spa, NULL, txg, EINVAL));
if (newrootvd->vdev_children != 1)
return (spa_vdev_exit(spa, newrootvd, txg, EINVAL));
newvd = newrootvd->vdev_child[0];
if (!newvd->vdev_ops->vdev_op_leaf)
return (spa_vdev_exit(spa, newrootvd, txg, EINVAL));
if ((error = vdev_create(newrootvd, txg, replacing)) != 0)
return (spa_vdev_exit(spa, newrootvd, txg, error));
/*
* Spares can't replace logs
*/
if (oldvd->vdev_top->vdev_islog && newvd->vdev_isspare)
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
/*
* A dRAID spare can only replace a child of its parent dRAID vdev.
*/
if (newvd->vdev_ops == &vdev_draid_spare_ops &&
oldvd->vdev_top != vdev_draid_spare_get_parent(newvd)) {
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
}
if (rebuild) {
/*
* For rebuilds, the top vdev must support reconstruction
* using only space maps. This means the only allowable
* vdevs types are the root vdev, a mirror, or dRAID.
*/
tvd = pvd;
if (pvd->vdev_top != NULL)
tvd = pvd->vdev_top;
if (tvd->vdev_ops != &vdev_mirror_ops &&
tvd->vdev_ops != &vdev_root_ops &&
tvd->vdev_ops != &vdev_draid_ops) {
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
}
}
if (!replacing) {
/*
* For attach, the only allowable parent is a mirror or the root
* vdev.
*/
if (pvd->vdev_ops != &vdev_mirror_ops &&
pvd->vdev_ops != &vdev_root_ops)
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
pvops = &vdev_mirror_ops;
} else {
/*
* Active hot spares can only be replaced by inactive hot
* spares.
*/
if (pvd->vdev_ops == &vdev_spare_ops &&
oldvd->vdev_isspare &&
!spa_has_spare(spa, newvd->vdev_guid))
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
/*
* If the source is a hot spare, and the parent isn't already a
* spare, then we want to create a new hot spare. Otherwise, we
* want to create a replacing vdev. The user is not allowed to
* attach to a spared vdev child unless the 'isspare' state is
* the same (spare replaces spare, non-spare replaces
* non-spare).
*/
if (pvd->vdev_ops == &vdev_replacing_ops &&
spa_version(spa) < SPA_VERSION_MULTI_REPLACE) {
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
} else if (pvd->vdev_ops == &vdev_spare_ops &&
newvd->vdev_isspare != oldvd->vdev_isspare) {
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
}
if (newvd->vdev_isspare)
pvops = &vdev_spare_ops;
else
pvops = &vdev_replacing_ops;
}
/*
* Make sure the new device is big enough.
*/
if (newvd->vdev_asize < vdev_get_min_asize(oldvd))
return (spa_vdev_exit(spa, newrootvd, txg, EOVERFLOW));
/*
* The new device cannot have a higher alignment requirement
* than the top-level vdev.
*/
if (newvd->vdev_ashift > oldvd->vdev_top->vdev_ashift)
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
/*
* If this is an in-place replacement, update oldvd's path and devid
* to make it distinguishable from newvd, and unopenable from now on.
*/
if (strcmp(oldvd->vdev_path, newvd->vdev_path) == 0) {
spa_strfree(oldvd->vdev_path);
oldvd->vdev_path = kmem_alloc(strlen(newvd->vdev_path) + 5,
KM_SLEEP);
(void) snprintf(oldvd->vdev_path, strlen(newvd->vdev_path) + 5,
"%s/%s", newvd->vdev_path, "old");
if (oldvd->vdev_devid != NULL) {
spa_strfree(oldvd->vdev_devid);
oldvd->vdev_devid = NULL;
}
}
/*
* If the parent is not a mirror, or if we're replacing, insert the new
* mirror/replacing/spare vdev above oldvd.
*/
if (pvd->vdev_ops != pvops)
pvd = vdev_add_parent(oldvd, pvops);
ASSERT(pvd->vdev_top->vdev_parent == rvd);
ASSERT(pvd->vdev_ops == pvops);
ASSERT(oldvd->vdev_parent == pvd);
/*
* Extract the new device from its root and add it to pvd.
*/
vdev_remove_child(newrootvd, newvd);
newvd->vdev_id = pvd->vdev_children;
newvd->vdev_crtxg = oldvd->vdev_crtxg;
vdev_add_child(pvd, newvd);
/*
* Reevaluate the parent vdev state.
*/
vdev_propagate_state(pvd);
tvd = newvd->vdev_top;
ASSERT(pvd->vdev_top == tvd);
ASSERT(tvd->vdev_parent == rvd);
vdev_config_dirty(tvd);
/*
* Set newvd's DTL to [TXG_INITIAL, dtl_max_txg) so that we account
* for any dmu_sync-ed blocks. It will propagate upward when
* spa_vdev_exit() calls vdev_dtl_reassess().
*/
dtl_max_txg = txg + TXG_CONCURRENT_STATES;
vdev_dtl_dirty(newvd, DTL_MISSING,
TXG_INITIAL, dtl_max_txg - TXG_INITIAL);
if (newvd->vdev_isspare) {
spa_spare_activate(newvd);
spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_SPARE);
}
oldvdpath = spa_strdup(oldvd->vdev_path);
newvdpath = spa_strdup(newvd->vdev_path);
newvd_isspare = newvd->vdev_isspare;
/*
* Mark newvd's DTL dirty in this txg.
*/
vdev_dirty(tvd, VDD_DTL, newvd, txg);
/*
* Schedule the resilver or rebuild to restart in the future. We do
* this to ensure that dmu_sync-ed blocks have been stitched into the
* respective datasets.
*/
if (rebuild) {
newvd->vdev_rebuild_txg = txg;
vdev_rebuild(tvd);
} else {
newvd->vdev_resilver_txg = txg;
if (dsl_scan_resilvering(spa_get_dsl(spa)) &&
spa_feature_is_enabled(spa, SPA_FEATURE_RESILVER_DEFER)) {
vdev_defer_resilver(newvd);
} else {
dsl_scan_restart_resilver(spa->spa_dsl_pool,
dtl_max_txg);
}
}
if (spa->spa_bootfs)
spa_event_notify(spa, newvd, NULL, ESC_ZFS_BOOTFS_VDEV_ATTACH);
spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_ATTACH);
/*
* Commit the config
*/
(void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0);
spa_history_log_internal(spa, "vdev attach", NULL,
"%s vdev=%s %s vdev=%s",
replacing && newvd_isspare ? "spare in" :
replacing ? "replace" : "attach", newvdpath,
replacing ? "for" : "to", oldvdpath);
spa_strfree(oldvdpath);
spa_strfree(newvdpath);
return (0);
}
/*
* Detach a device from a mirror or replacing vdev.
*
* If 'replace_done' is specified, only detach if the parent
* is a replacing vdev.
*/
int
spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done)
{
uint64_t txg;
int error;
vdev_t *rvd __maybe_unused = spa->spa_root_vdev;
vdev_t *vd, *pvd, *cvd, *tvd;
boolean_t unspare = B_FALSE;
uint64_t unspare_guid = 0;
char *vdpath;
ASSERT(spa_writeable(spa));
txg = spa_vdev_detach_enter(spa, guid);
vd = spa_lookup_by_guid(spa, guid, B_FALSE);
/*
* Besides being called directly from the userland through the
* ioctl interface, spa_vdev_detach() can be potentially called
* at the end of spa_vdev_resilver_done().
*
* In the regular case, when we have a checkpoint this shouldn't
* happen as we never empty the DTLs of a vdev during the scrub
* [see comment in dsl_scan_done()]. Thus spa_vdev_resilvering_done()
* should never get here when we have a checkpoint.
*
* That said, even in a case when we checkpoint the pool exactly
* as spa_vdev_resilver_done() calls this function everything
* should be fine as the resilver will return right away.
*/
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
error = (spa_has_checkpoint(spa)) ?
ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
return (spa_vdev_exit(spa, NULL, txg, error));
}
if (vd == NULL)
return (spa_vdev_exit(spa, NULL, txg, ENODEV));
if (!vd->vdev_ops->vdev_op_leaf)
return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
pvd = vd->vdev_parent;
/*
* If the parent/child relationship is not as expected, don't do it.
* Consider M(A,R(B,C)) -- that is, a mirror of A with a replacing
* vdev that's replacing B with C. The user's intent in replacing
* is to go from M(A,B) to M(A,C). If the user decides to cancel
* the replace by detaching C, the expected behavior is to end up
* M(A,B). But suppose that right after deciding to detach C,
* the replacement of B completes. We would have M(A,C), and then
* ask to detach C, which would leave us with just A -- not what
* the user wanted. To prevent this, we make sure that the
* parent/child relationship hasn't changed -- in this example,
* that C's parent is still the replacing vdev R.
*/
if (pvd->vdev_guid != pguid && pguid != 0)
return (spa_vdev_exit(spa, NULL, txg, EBUSY));
/*
* Only 'replacing' or 'spare' vdevs can be replaced.
*/
if (replace_done && pvd->vdev_ops != &vdev_replacing_ops &&
pvd->vdev_ops != &vdev_spare_ops)
return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
ASSERT(pvd->vdev_ops != &vdev_spare_ops ||
spa_version(spa) >= SPA_VERSION_SPARES);
/*
* Only mirror, replacing, and spare vdevs support detach.
*/
if (pvd->vdev_ops != &vdev_replacing_ops &&
pvd->vdev_ops != &vdev_mirror_ops &&
pvd->vdev_ops != &vdev_spare_ops)
return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
/*
* If this device has the only valid copy of some data,
* we cannot safely detach it.
*/
if (vdev_dtl_required(vd))
return (spa_vdev_exit(spa, NULL, txg, EBUSY));
ASSERT(pvd->vdev_children >= 2);
/*
* If we are detaching the second disk from a replacing vdev, then
* check to see if we changed the original vdev's path to have "/old"
* at the end in spa_vdev_attach(). If so, undo that change now.
*/
if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id > 0 &&
vd->vdev_path != NULL) {
size_t len = strlen(vd->vdev_path);
for (int c = 0; c < pvd->vdev_children; c++) {
cvd = pvd->vdev_child[c];
if (cvd == vd || cvd->vdev_path == NULL)
continue;
if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 &&
strcmp(cvd->vdev_path + len, "/old") == 0) {
spa_strfree(cvd->vdev_path);
cvd->vdev_path = spa_strdup(vd->vdev_path);
break;
}
}
}
/*
* If we are detaching the original disk from a normal spare, then it
* implies that the spare should become a real disk, and be removed
* from the active spare list for the pool. dRAID spares on the
* other hand are coupled to the pool and thus should never be removed
* from the spares list.
*/
if (pvd->vdev_ops == &vdev_spare_ops && vd->vdev_id == 0) {
vdev_t *last_cvd = pvd->vdev_child[pvd->vdev_children - 1];
if (last_cvd->vdev_isspare &&
last_cvd->vdev_ops != &vdev_draid_spare_ops) {
unspare = B_TRUE;
}
}
/*
* Erase the disk labels so the disk can be used for other things.
* This must be done after all other error cases are handled,
* but before we disembowel vd (so we can still do I/O to it).
* But if we can't do it, don't treat the error as fatal --
* it may be that the unwritability of the disk is the reason
* it's being detached!
*/
error = vdev_label_init(vd, 0, VDEV_LABEL_REMOVE);
/*
* Remove vd from its parent and compact the parent's children.
*/
vdev_remove_child(pvd, vd);
vdev_compact_children(pvd);
/*
* Remember one of the remaining children so we can get tvd below.
*/
cvd = pvd->vdev_child[pvd->vdev_children - 1];
/*
* If we need to remove the remaining child from the list of hot spares,
* do it now, marking the vdev as no longer a spare in the process.
* We must do this before vdev_remove_parent(), because that can
* change the GUID if it creates a new toplevel GUID. For a similar
* reason, we must remove the spare now, in the same txg as the detach;
* otherwise someone could attach a new sibling, change the GUID, and
* the subsequent attempt to spa_vdev_remove(unspare_guid) would fail.
*/
if (unspare) {
ASSERT(cvd->vdev_isspare);
spa_spare_remove(cvd);
unspare_guid = cvd->vdev_guid;
(void) spa_vdev_remove(spa, unspare_guid, B_TRUE);
cvd->vdev_unspare = B_TRUE;
}
/*
* If the parent mirror/replacing vdev only has one child,
* the parent is no longer needed. Remove it from the tree.
*/
if (pvd->vdev_children == 1) {
if (pvd->vdev_ops == &vdev_spare_ops)
cvd->vdev_unspare = B_FALSE;
vdev_remove_parent(cvd);
}
/*
* We don't set tvd until now because the parent we just removed
* may have been the previous top-level vdev.
*/
tvd = cvd->vdev_top;
ASSERT(tvd->vdev_parent == rvd);
/*
* Reevaluate the parent vdev state.
*/
vdev_propagate_state(cvd);
/*
* If the 'autoexpand' property is set on the pool then automatically
* try to expand the size of the pool. For example if the device we
* just detached was smaller than the others, it may be possible to
* add metaslabs (i.e. grow the pool). We need to reopen the vdev
* first so that we can obtain the updated sizes of the leaf vdevs.
*/
if (spa->spa_autoexpand) {
vdev_reopen(tvd);
vdev_expand(tvd, txg);
}
vdev_config_dirty(tvd);
/*
* Mark vd's DTL as dirty in this txg. vdev_dtl_sync() will see that
* vd->vdev_detached is set and free vd's DTL object in syncing context.
* But first make sure we're not on any *other* txg's DTL list, to
* prevent vd from being accessed after it's freed.
*/
vdpath = spa_strdup(vd->vdev_path ? vd->vdev_path : "none");
for (int t = 0; t < TXG_SIZE; t++)
(void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t);
vd->vdev_detached = B_TRUE;
vdev_dirty(tvd, VDD_DTL, vd, txg);
spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE);
spa_notify_waiters(spa);
/* hang on to the spa before we release the lock */
spa_open_ref(spa, FTAG);
error = spa_vdev_exit(spa, vd, txg, 0);
spa_history_log_internal(spa, "detach", NULL,
"vdev=%s", vdpath);
spa_strfree(vdpath);
/*
* If this was the removal of the original device in a hot spare vdev,
* then we want to go through and remove the device from the hot spare
* list of every other pool.
*/
if (unspare) {
spa_t *altspa = NULL;
mutex_enter(&spa_namespace_lock);
while ((altspa = spa_next(altspa)) != NULL) {
if (altspa->spa_state != POOL_STATE_ACTIVE ||
altspa == spa)
continue;
spa_open_ref(altspa, FTAG);
mutex_exit(&spa_namespace_lock);
(void) spa_vdev_remove(altspa, unspare_guid, B_TRUE);
mutex_enter(&spa_namespace_lock);
spa_close(altspa, FTAG);
}
mutex_exit(&spa_namespace_lock);
/* search the rest of the vdevs for spares to remove */
spa_vdev_resilver_done(spa);
}
/* all done with the spa; OK to release */
mutex_enter(&spa_namespace_lock);
spa_close(spa, FTAG);
mutex_exit(&spa_namespace_lock);
return (error);
}
static int
spa_vdev_initialize_impl(spa_t *spa, uint64_t guid, uint64_t cmd_type,
list_t *vd_list)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
/* Look up vdev and ensure it's a leaf. */
vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE);
if (vd == NULL || vd->vdev_detached) {
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (SET_ERROR(ENODEV));
} else if (!vd->vdev_ops->vdev_op_leaf || !vdev_is_concrete(vd)) {
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (SET_ERROR(EINVAL));
} else if (!vdev_writeable(vd)) {
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (SET_ERROR(EROFS));
}
mutex_enter(&vd->vdev_initialize_lock);
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
/*
* When we activate an initialize action we check to see
* if the vdev_initialize_thread is NULL. We do this instead
* of using the vdev_initialize_state since there might be
* a previous initialization process which has completed but
* the thread is not exited.
*/
if (cmd_type == POOL_INITIALIZE_START &&
(vd->vdev_initialize_thread != NULL ||
vd->vdev_top->vdev_removing)) {
mutex_exit(&vd->vdev_initialize_lock);
return (SET_ERROR(EBUSY));
} else if (cmd_type == POOL_INITIALIZE_CANCEL &&
(vd->vdev_initialize_state != VDEV_INITIALIZE_ACTIVE &&
vd->vdev_initialize_state != VDEV_INITIALIZE_SUSPENDED)) {
mutex_exit(&vd->vdev_initialize_lock);
return (SET_ERROR(ESRCH));
} else if (cmd_type == POOL_INITIALIZE_SUSPEND &&
vd->vdev_initialize_state != VDEV_INITIALIZE_ACTIVE) {
mutex_exit(&vd->vdev_initialize_lock);
return (SET_ERROR(ESRCH));
}
switch (cmd_type) {
case POOL_INITIALIZE_START:
vdev_initialize(vd);
break;
case POOL_INITIALIZE_CANCEL:
vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED, vd_list);
break;
case POOL_INITIALIZE_SUSPEND:
vdev_initialize_stop(vd, VDEV_INITIALIZE_SUSPENDED, vd_list);
break;
default:
panic("invalid cmd_type %llu", (unsigned long long)cmd_type);
}
mutex_exit(&vd->vdev_initialize_lock);
return (0);
}
int
spa_vdev_initialize(spa_t *spa, nvlist_t *nv, uint64_t cmd_type,
nvlist_t *vdev_errlist)
{
int total_errors = 0;
list_t vd_list;
list_create(&vd_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_initialize_node));
/*
* We hold the namespace lock through the whole function
* to prevent any changes to the pool while we're starting or
* stopping initialization. The config and state locks are held so that
* we can properly assess the vdev state before we commit to
* the initializing operation.
*/
mutex_enter(&spa_namespace_lock);
for (nvpair_t *pair = nvlist_next_nvpair(nv, NULL);
pair != NULL; pair = nvlist_next_nvpair(nv, pair)) {
uint64_t vdev_guid = fnvpair_value_uint64(pair);
int error = spa_vdev_initialize_impl(spa, vdev_guid, cmd_type,
&vd_list);
if (error != 0) {
char guid_as_str[MAXNAMELEN];
(void) snprintf(guid_as_str, sizeof (guid_as_str),
"%llu", (unsigned long long)vdev_guid);
fnvlist_add_int64(vdev_errlist, guid_as_str, error);
total_errors++;
}
}
/* Wait for all initialize threads to stop. */
vdev_initialize_stop_wait(spa, &vd_list);
/* Sync out the initializing state */
txg_wait_synced(spa->spa_dsl_pool, 0);
mutex_exit(&spa_namespace_lock);
list_destroy(&vd_list);
return (total_errors);
}
static int
spa_vdev_trim_impl(spa_t *spa, uint64_t guid, uint64_t cmd_type,
uint64_t rate, boolean_t partial, boolean_t secure, list_t *vd_list)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
/* Look up vdev and ensure it's a leaf. */
vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE);
if (vd == NULL || vd->vdev_detached) {
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (SET_ERROR(ENODEV));
} else if (!vd->vdev_ops->vdev_op_leaf || !vdev_is_concrete(vd)) {
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (SET_ERROR(EINVAL));
} else if (!vdev_writeable(vd)) {
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (SET_ERROR(EROFS));
} else if (!vd->vdev_has_trim) {
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (SET_ERROR(EOPNOTSUPP));
} else if (secure && !vd->vdev_has_securetrim) {
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (SET_ERROR(EOPNOTSUPP));
}
mutex_enter(&vd->vdev_trim_lock);
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
/*
* When we activate a TRIM action we check to see if the
* vdev_trim_thread is NULL. We do this instead of using the
* vdev_trim_state since there might be a previous TRIM process
* which has completed but the thread is not exited.
*/
if (cmd_type == POOL_TRIM_START &&
(vd->vdev_trim_thread != NULL || vd->vdev_top->vdev_removing)) {
mutex_exit(&vd->vdev_trim_lock);
return (SET_ERROR(EBUSY));
} else if (cmd_type == POOL_TRIM_CANCEL &&
(vd->vdev_trim_state != VDEV_TRIM_ACTIVE &&
vd->vdev_trim_state != VDEV_TRIM_SUSPENDED)) {
mutex_exit(&vd->vdev_trim_lock);
return (SET_ERROR(ESRCH));
} else if (cmd_type == POOL_TRIM_SUSPEND &&
vd->vdev_trim_state != VDEV_TRIM_ACTIVE) {
mutex_exit(&vd->vdev_trim_lock);
return (SET_ERROR(ESRCH));
}
switch (cmd_type) {
case POOL_TRIM_START:
vdev_trim(vd, rate, partial, secure);
break;
case POOL_TRIM_CANCEL:
vdev_trim_stop(vd, VDEV_TRIM_CANCELED, vd_list);
break;
case POOL_TRIM_SUSPEND:
vdev_trim_stop(vd, VDEV_TRIM_SUSPENDED, vd_list);
break;
default:
panic("invalid cmd_type %llu", (unsigned long long)cmd_type);
}
mutex_exit(&vd->vdev_trim_lock);
return (0);
}
/*
* Initiates a manual TRIM for the requested vdevs. This kicks off individual
* TRIM threads for each child vdev. These threads pass over all of the free
* space in the vdev's metaslabs and issues TRIM commands for that space.
*/
int
spa_vdev_trim(spa_t *spa, nvlist_t *nv, uint64_t cmd_type, uint64_t rate,
boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist)
{
int total_errors = 0;
list_t vd_list;
list_create(&vd_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_trim_node));
/*
* We hold the namespace lock through the whole function
* to prevent any changes to the pool while we're starting or
* stopping TRIM. The config and state locks are held so that
* we can properly assess the vdev state before we commit to
* the TRIM operation.
*/
mutex_enter(&spa_namespace_lock);
for (nvpair_t *pair = nvlist_next_nvpair(nv, NULL);
pair != NULL; pair = nvlist_next_nvpair(nv, pair)) {
uint64_t vdev_guid = fnvpair_value_uint64(pair);
int error = spa_vdev_trim_impl(spa, vdev_guid, cmd_type,
rate, partial, secure, &vd_list);
if (error != 0) {
char guid_as_str[MAXNAMELEN];
(void) snprintf(guid_as_str, sizeof (guid_as_str),
"%llu", (unsigned long long)vdev_guid);
fnvlist_add_int64(vdev_errlist, guid_as_str, error);
total_errors++;
}
}
/* Wait for all TRIM threads to stop. */
vdev_trim_stop_wait(spa, &vd_list);
/* Sync out the TRIM state */
txg_wait_synced(spa->spa_dsl_pool, 0);
mutex_exit(&spa_namespace_lock);
list_destroy(&vd_list);
return (total_errors);
}
/*
* Split a set of devices from their mirrors, and create a new pool from them.
*/
int
spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
nvlist_t *props, boolean_t exp)
{
int error = 0;
uint64_t txg, *glist;
spa_t *newspa;
uint_t c, children, lastlog;
nvlist_t **child, *nvl, *tmp;
dmu_tx_t *tx;
char *altroot = NULL;
vdev_t *rvd, **vml = NULL; /* vdev modify list */
boolean_t activate_slog;
ASSERT(spa_writeable(spa));
txg = spa_vdev_enter(spa);
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
error = (spa_has_checkpoint(spa)) ?
ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
return (spa_vdev_exit(spa, NULL, txg, error));
}
/* clear the log and flush everything up to now */
activate_slog = spa_passivate_log(spa);
(void) spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);
error = spa_reset_logs(spa);
txg = spa_vdev_config_enter(spa);
if (activate_slog)
spa_activate_log(spa);
if (error != 0)
return (spa_vdev_exit(spa, NULL, txg, error));
/* check new spa name before going any further */
if (spa_lookup(newname) != NULL)
return (spa_vdev_exit(spa, NULL, txg, EEXIST));
/*
* scan through all the children to ensure they're all mirrors
*/
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvl) != 0 ||
nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN, &child,
&children) != 0)
return (spa_vdev_exit(spa, NULL, txg, EINVAL));
/* first, check to ensure we've got the right child count */
rvd = spa->spa_root_vdev;
lastlog = 0;
for (c = 0; c < rvd->vdev_children; c++) {
vdev_t *vd = rvd->vdev_child[c];
/* don't count the holes & logs as children */
if (vd->vdev_islog || (vd->vdev_ops != &vdev_indirect_ops &&
!vdev_is_concrete(vd))) {
if (lastlog == 0)
lastlog = c;
continue;
}
lastlog = 0;
}
if (children != (lastlog != 0 ? lastlog : rvd->vdev_children))
return (spa_vdev_exit(spa, NULL, txg, EINVAL));
/* next, ensure no spare or cache devices are part of the split */
if (nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_SPARES, &tmp) == 0 ||
nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0)
return (spa_vdev_exit(spa, NULL, txg, EINVAL));
vml = kmem_zalloc(children * sizeof (vdev_t *), KM_SLEEP);
glist = kmem_zalloc(children * sizeof (uint64_t), KM_SLEEP);
/* then, loop over each vdev and validate it */
for (c = 0; c < children; c++) {
uint64_t is_hole = 0;
(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
&is_hole);
if (is_hole != 0) {
if (spa->spa_root_vdev->vdev_child[c]->vdev_ishole ||
spa->spa_root_vdev->vdev_child[c]->vdev_islog) {
continue;
} else {
error = SET_ERROR(EINVAL);
break;
}
}
/* deal with indirect vdevs */
if (spa->spa_root_vdev->vdev_child[c]->vdev_ops ==
&vdev_indirect_ops)
continue;
/* which disk is going to be split? */
if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID,
&glist[c]) != 0) {
error = SET_ERROR(EINVAL);
break;
}
/* look it up in the spa */
vml[c] = spa_lookup_by_guid(spa, glist[c], B_FALSE);
if (vml[c] == NULL) {
error = SET_ERROR(ENODEV);
break;
}
/* make sure there's nothing stopping the split */
if (vml[c]->vdev_parent->vdev_ops != &vdev_mirror_ops ||
vml[c]->vdev_islog ||
!vdev_is_concrete(vml[c]) ||
vml[c]->vdev_isspare ||
vml[c]->vdev_isl2cache ||
!vdev_writeable(vml[c]) ||
vml[c]->vdev_children != 0 ||
vml[c]->vdev_state != VDEV_STATE_HEALTHY ||
c != spa->spa_root_vdev->vdev_child[c]->vdev_id) {
error = SET_ERROR(EINVAL);
break;
}
if (vdev_dtl_required(vml[c]) ||
vdev_resilver_needed(vml[c], NULL, NULL)) {
error = SET_ERROR(EBUSY);
break;
}
/* we need certain info from the top level */
fnvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_ARRAY,
vml[c]->vdev_top->vdev_ms_array);
fnvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_SHIFT,
vml[c]->vdev_top->vdev_ms_shift);
fnvlist_add_uint64(child[c], ZPOOL_CONFIG_ASIZE,
vml[c]->vdev_top->vdev_asize);
fnvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT,
vml[c]->vdev_top->vdev_ashift);
/* transfer per-vdev ZAPs */
ASSERT3U(vml[c]->vdev_leaf_zap, !=, 0);
VERIFY0(nvlist_add_uint64(child[c],
ZPOOL_CONFIG_VDEV_LEAF_ZAP, vml[c]->vdev_leaf_zap));
ASSERT3U(vml[c]->vdev_top->vdev_top_zap, !=, 0);
VERIFY0(nvlist_add_uint64(child[c],
ZPOOL_CONFIG_VDEV_TOP_ZAP,
vml[c]->vdev_parent->vdev_top_zap));
}
if (error != 0) {
kmem_free(vml, children * sizeof (vdev_t *));
kmem_free(glist, children * sizeof (uint64_t));
return (spa_vdev_exit(spa, NULL, txg, error));
}
/* stop writers from using the disks */
for (c = 0; c < children; c++) {
if (vml[c] != NULL)
vml[c]->vdev_offline = B_TRUE;
}
vdev_reopen(spa->spa_root_vdev);
/*
* Temporarily record the splitting vdevs in the spa config. This
* will disappear once the config is regenerated.
*/
nvl = fnvlist_alloc();
fnvlist_add_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, glist, children);
kmem_free(glist, children * sizeof (uint64_t));
mutex_enter(&spa->spa_props_lock);
fnvlist_add_nvlist(spa->spa_config, ZPOOL_CONFIG_SPLIT, nvl);
mutex_exit(&spa->spa_props_lock);
spa->spa_config_splitting = nvl;
vdev_config_dirty(spa->spa_root_vdev);
/* configure and create the new pool */
fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, newname);
fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
exp ? POOL_STATE_EXPORTED : POOL_STATE_ACTIVE);
fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa_version(spa));
fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, spa->spa_config_txg);
fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
spa_generate_guid(NULL));
VERIFY0(nvlist_add_boolean(config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
(void) nvlist_lookup_string(props,
zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
/* add the new pool to the namespace */
newspa = spa_add(newname, config, altroot);
newspa->spa_avz_action = AVZ_ACTION_REBUILD;
newspa->spa_config_txg = spa->spa_config_txg;
spa_set_log_state(newspa, SPA_LOG_CLEAR);
/* release the spa config lock, retaining the namespace lock */
spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);
if (zio_injection_enabled)
zio_handle_panic_injection(spa, FTAG, 1);
spa_activate(newspa, spa_mode_global);
spa_async_suspend(newspa);
/*
* Temporarily stop the initializing and TRIM activity. We set the
* state to ACTIVE so that we know to resume initializing or TRIM
* once the split has completed.
*/
list_t vd_initialize_list;
list_create(&vd_initialize_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_initialize_node));
list_t vd_trim_list;
list_create(&vd_trim_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_trim_node));
for (c = 0; c < children; c++) {
if (vml[c] != NULL && vml[c]->vdev_ops != &vdev_indirect_ops) {
mutex_enter(&vml[c]->vdev_initialize_lock);
vdev_initialize_stop(vml[c],
VDEV_INITIALIZE_ACTIVE, &vd_initialize_list);
mutex_exit(&vml[c]->vdev_initialize_lock);
mutex_enter(&vml[c]->vdev_trim_lock);
vdev_trim_stop(vml[c], VDEV_TRIM_ACTIVE, &vd_trim_list);
mutex_exit(&vml[c]->vdev_trim_lock);
}
}
vdev_initialize_stop_wait(spa, &vd_initialize_list);
vdev_trim_stop_wait(spa, &vd_trim_list);
list_destroy(&vd_initialize_list);
list_destroy(&vd_trim_list);
newspa->spa_config_source = SPA_CONFIG_SRC_SPLIT;
newspa->spa_is_splitting = B_TRUE;
/* create the new pool from the disks of the original pool */
error = spa_load(newspa, SPA_LOAD_IMPORT, SPA_IMPORT_ASSEMBLE);
if (error)
goto out;
/* if that worked, generate a real config for the new pool */
if (newspa->spa_root_vdev != NULL) {
newspa->spa_config_splitting = fnvlist_alloc();
fnvlist_add_uint64(newspa->spa_config_splitting,
ZPOOL_CONFIG_SPLIT_GUID, spa_guid(spa));
spa_config_set(newspa, spa_config_generate(newspa, NULL, -1ULL,
B_TRUE));
}
/* set the props */
if (props != NULL) {
spa_configfile_set(newspa, props, B_FALSE);
error = spa_prop_set(newspa, props);
if (error)
goto out;
}
/* flush everything */
txg = spa_vdev_config_enter(newspa);
vdev_config_dirty(newspa->spa_root_vdev);
(void) spa_vdev_config_exit(newspa, NULL, txg, 0, FTAG);
if (zio_injection_enabled)
zio_handle_panic_injection(spa, FTAG, 2);
spa_async_resume(newspa);
/* finally, update the original pool's config */
txg = spa_vdev_config_enter(spa);
tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error != 0)
dmu_tx_abort(tx);
for (c = 0; c < children; c++) {
if (vml[c] != NULL && vml[c]->vdev_ops != &vdev_indirect_ops) {
vdev_t *tvd = vml[c]->vdev_top;
/*
* Need to be sure the detachable VDEV is not
* on any *other* txg's DTL list to prevent it
* from being accessed after it's freed.
*/
for (int t = 0; t < TXG_SIZE; t++) {
(void) txg_list_remove_this(
&tvd->vdev_dtl_list, vml[c], t);
}
vdev_split(vml[c]);
if (error == 0)
spa_history_log_internal(spa, "detach", tx,
"vdev=%s", vml[c]->vdev_path);
vdev_free(vml[c]);
}
}
spa->spa_avz_action = AVZ_ACTION_REBUILD;
vdev_config_dirty(spa->spa_root_vdev);
spa->spa_config_splitting = NULL;
nvlist_free(nvl);
if (error == 0)
dmu_tx_commit(tx);
(void) spa_vdev_exit(spa, NULL, txg, 0);
if (zio_injection_enabled)
zio_handle_panic_injection(spa, FTAG, 3);
/* split is complete; log a history record */
spa_history_log_internal(newspa, "split", NULL,
"from pool %s", spa_name(spa));
newspa->spa_is_splitting = B_FALSE;
kmem_free(vml, children * sizeof (vdev_t *));
/* if we're not going to mount the filesystems in userland, export */
if (exp)
error = spa_export_common(newname, POOL_STATE_EXPORTED, NULL,
B_FALSE, B_FALSE);
return (error);
out:
spa_unload(newspa);
spa_deactivate(newspa);
spa_remove(newspa);
txg = spa_vdev_config_enter(spa);
/* re-online all offlined disks */
for (c = 0; c < children; c++) {
if (vml[c] != NULL)
vml[c]->vdev_offline = B_FALSE;
}
/* restart initializing or trimming disks as necessary */
spa_async_request(spa, SPA_ASYNC_INITIALIZE_RESTART);
spa_async_request(spa, SPA_ASYNC_TRIM_RESTART);
spa_async_request(spa, SPA_ASYNC_AUTOTRIM_RESTART);
vdev_reopen(spa->spa_root_vdev);
nvlist_free(spa->spa_config_splitting);
spa->spa_config_splitting = NULL;
(void) spa_vdev_exit(spa, NULL, txg, error);
kmem_free(vml, children * sizeof (vdev_t *));
return (error);
}
/*
* Find any device that's done replacing, or a vdev marked 'unspare' that's
* currently spared, so we can detach it.
*/
static vdev_t *
spa_vdev_resilver_done_hunt(vdev_t *vd)
{
vdev_t *newvd, *oldvd;
for (int c = 0; c < vd->vdev_children; c++) {
oldvd = spa_vdev_resilver_done_hunt(vd->vdev_child[c]);
if (oldvd != NULL)
return (oldvd);
}
/*
* Check for a completed replacement. We always consider the first
* vdev in the list to be the oldest vdev, and the last one to be
* the newest (see spa_vdev_attach() for how that works). In
* the case where the newest vdev is faulted, we will not automatically
* remove it after a resilver completes. This is OK as it will require
* user intervention to determine which disk the admin wishes to keep.
*/
if (vd->vdev_ops == &vdev_replacing_ops) {
ASSERT(vd->vdev_children > 1);
newvd = vd->vdev_child[vd->vdev_children - 1];
oldvd = vd->vdev_child[0];
if (vdev_dtl_empty(newvd, DTL_MISSING) &&
vdev_dtl_empty(newvd, DTL_OUTAGE) &&
!vdev_dtl_required(oldvd))
return (oldvd);
}
/*
* Check for a completed resilver with the 'unspare' flag set.
* Also potentially update faulted state.
*/
if (vd->vdev_ops == &vdev_spare_ops) {
vdev_t *first = vd->vdev_child[0];
vdev_t *last = vd->vdev_child[vd->vdev_children - 1];
if (last->vdev_unspare) {
oldvd = first;
newvd = last;
} else if (first->vdev_unspare) {
oldvd = last;
newvd = first;
} else {
oldvd = NULL;
}
if (oldvd != NULL &&
vdev_dtl_empty(newvd, DTL_MISSING) &&
vdev_dtl_empty(newvd, DTL_OUTAGE) &&
!vdev_dtl_required(oldvd))
return (oldvd);
vdev_propagate_state(vd);
/*
* If there are more than two spares attached to a disk,
* and those spares are not required, then we want to
* attempt to free them up now so that they can be used
* by other pools. Once we're back down to a single
* disk+spare, we stop removing them.
*/
if (vd->vdev_children > 2) {
newvd = vd->vdev_child[1];
if (newvd->vdev_isspare && last->vdev_isspare &&
vdev_dtl_empty(last, DTL_MISSING) &&
vdev_dtl_empty(last, DTL_OUTAGE) &&
!vdev_dtl_required(newvd))
return (newvd);
}
}
return (NULL);
}
static void
spa_vdev_resilver_done(spa_t *spa)
{
vdev_t *vd, *pvd, *ppvd;
uint64_t guid, sguid, pguid, ppguid;
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
while ((vd = spa_vdev_resilver_done_hunt(spa->spa_root_vdev)) != NULL) {
pvd = vd->vdev_parent;
ppvd = pvd->vdev_parent;
guid = vd->vdev_guid;
pguid = pvd->vdev_guid;
ppguid = ppvd->vdev_guid;
sguid = 0;
/*
* If we have just finished replacing a hot spared device, then
* we need to detach the parent's first child (the original hot
* spare) as well.
*/
if (ppvd->vdev_ops == &vdev_spare_ops && pvd->vdev_id == 0 &&
ppvd->vdev_children == 2) {
ASSERT(pvd->vdev_ops == &vdev_replacing_ops);
sguid = ppvd->vdev_child[1]->vdev_guid;
}
ASSERT(vd->vdev_resilver_txg == 0 || !vdev_dtl_required(vd));
spa_config_exit(spa, SCL_ALL, FTAG);
if (spa_vdev_detach(spa, guid, pguid, B_TRUE) != 0)
return;
if (sguid && spa_vdev_detach(spa, sguid, ppguid, B_TRUE) != 0)
return;
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
}
spa_config_exit(spa, SCL_ALL, FTAG);
/*
* If a detach was not performed above replace waiters will not have
* been notified. In which case we must do so now.
*/
spa_notify_waiters(spa);
}
/*
* Update the stored path or FRU for this vdev.
*/
static int
spa_vdev_set_common(spa_t *spa, uint64_t guid, const char *value,
boolean_t ispath)
{
vdev_t *vd;
boolean_t sync = B_FALSE;
ASSERT(spa_writeable(spa));
spa_vdev_state_enter(spa, SCL_ALL);
if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
return (spa_vdev_state_exit(spa, NULL, ENOENT));
if (!vd->vdev_ops->vdev_op_leaf)
return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
if (ispath) {
if (strcmp(value, vd->vdev_path) != 0) {
spa_strfree(vd->vdev_path);
vd->vdev_path = spa_strdup(value);
sync = B_TRUE;
}
} else {
if (vd->vdev_fru == NULL) {
vd->vdev_fru = spa_strdup(value);
sync = B_TRUE;
} else if (strcmp(value, vd->vdev_fru) != 0) {
spa_strfree(vd->vdev_fru);
vd->vdev_fru = spa_strdup(value);
sync = B_TRUE;
}
}
return (spa_vdev_state_exit(spa, sync ? vd : NULL, 0));
}
int
spa_vdev_setpath(spa_t *spa, uint64_t guid, const char *newpath)
{
return (spa_vdev_set_common(spa, guid, newpath, B_TRUE));
}
int
spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru)
{
return (spa_vdev_set_common(spa, guid, newfru, B_FALSE));
}
/*
* ==========================================================================
* SPA Scanning
* ==========================================================================
*/
int
spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t cmd)
{
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
if (dsl_scan_resilvering(spa->spa_dsl_pool))
return (SET_ERROR(EBUSY));
return (dsl_scrub_set_pause_resume(spa->spa_dsl_pool, cmd));
}
int
spa_scan_stop(spa_t *spa)
{
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
if (dsl_scan_resilvering(spa->spa_dsl_pool))
return (SET_ERROR(EBUSY));
return (dsl_scan_cancel(spa->spa_dsl_pool));
}
int
spa_scan(spa_t *spa, pool_scan_func_t func)
{
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
if (func >= POOL_SCAN_FUNCS || func == POOL_SCAN_NONE)
return (SET_ERROR(ENOTSUP));
if (func == POOL_SCAN_RESILVER &&
!spa_feature_is_enabled(spa, SPA_FEATURE_RESILVER_DEFER))
return (SET_ERROR(ENOTSUP));
/*
* If a resilver was requested, but there is no DTL on a
* writeable leaf device, we have nothing to do.
*/
if (func == POOL_SCAN_RESILVER &&
!vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) {
spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);
return (0);
}
return (dsl_scan(spa->spa_dsl_pool, func));
}
/*
* ==========================================================================
* SPA async task processing
* ==========================================================================
*/
static void
spa_async_remove(spa_t *spa, vdev_t *vd)
{
if (vd->vdev_remove_wanted) {
vd->vdev_remove_wanted = B_FALSE;
vd->vdev_delayed_close = B_FALSE;
vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE);
/*
* We want to clear the stats, but we don't want to do a full
* vdev_clear() as that will cause us to throw away
* degraded/faulted state as well as attempt to reopen the
* device, all of which is a waste.
*/
vd->vdev_stat.vs_read_errors = 0;
vd->vdev_stat.vs_write_errors = 0;
vd->vdev_stat.vs_checksum_errors = 0;
vdev_state_dirty(vd->vdev_top);
/* Tell userspace that the vdev is gone. */
zfs_post_remove(spa, vd);
}
for (int c = 0; c < vd->vdev_children; c++)
spa_async_remove(spa, vd->vdev_child[c]);
}
static void
spa_async_probe(spa_t *spa, vdev_t *vd)
{
if (vd->vdev_probe_wanted) {
vd->vdev_probe_wanted = B_FALSE;
vdev_reopen(vd); /* vdev_open() does the actual probe */
}
for (int c = 0; c < vd->vdev_children; c++)
spa_async_probe(spa, vd->vdev_child[c]);
}
static void
spa_async_autoexpand(spa_t *spa, vdev_t *vd)
{
if (!spa->spa_autoexpand)
return;
for (int c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
spa_async_autoexpand(spa, cvd);
}
if (!vd->vdev_ops->vdev_op_leaf || vd->vdev_physpath == NULL)
return;
spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_AUTOEXPAND);
}
static void
spa_async_thread(void *arg)
{
spa_t *spa = (spa_t *)arg;
dsl_pool_t *dp = spa->spa_dsl_pool;
int tasks;
ASSERT(spa->spa_sync_on);
mutex_enter(&spa->spa_async_lock);
tasks = spa->spa_async_tasks;
spa->spa_async_tasks = 0;
mutex_exit(&spa->spa_async_lock);
/*
* See if the config needs to be updated.
*/
if (tasks & SPA_ASYNC_CONFIG_UPDATE) {
uint64_t old_space, new_space;
mutex_enter(&spa_namespace_lock);
old_space = metaslab_class_get_space(spa_normal_class(spa));
old_space += metaslab_class_get_space(spa_special_class(spa));
old_space += metaslab_class_get_space(spa_dedup_class(spa));
old_space += metaslab_class_get_space(
spa_embedded_log_class(spa));
spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
new_space = metaslab_class_get_space(spa_normal_class(spa));
new_space += metaslab_class_get_space(spa_special_class(spa));
new_space += metaslab_class_get_space(spa_dedup_class(spa));
new_space += metaslab_class_get_space(
spa_embedded_log_class(spa));
mutex_exit(&spa_namespace_lock);
/*
* If the pool grew as a result of the config update,
* then log an internal history event.
*/
if (new_space != old_space) {
spa_history_log_internal(spa, "vdev online", NULL,
"pool '%s' size: %llu(+%llu)",
spa_name(spa), (u_longlong_t)new_space,
(u_longlong_t)(new_space - old_space));
}
}
/*
* See if any devices need to be marked REMOVED.
*/
if (tasks & SPA_ASYNC_REMOVE) {
spa_vdev_state_enter(spa, SCL_NONE);
spa_async_remove(spa, spa->spa_root_vdev);
for (int i = 0; i < spa->spa_l2cache.sav_count; i++)
spa_async_remove(spa, spa->spa_l2cache.sav_vdevs[i]);
for (int i = 0; i < spa->spa_spares.sav_count; i++)
spa_async_remove(spa, spa->spa_spares.sav_vdevs[i]);
(void) spa_vdev_state_exit(spa, NULL, 0);
}
if ((tasks & SPA_ASYNC_AUTOEXPAND) && !spa_suspended(spa)) {
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
spa_async_autoexpand(spa, spa->spa_root_vdev);
spa_config_exit(spa, SCL_CONFIG, FTAG);
}
/*
* See if any devices need to be probed.
*/
if (tasks & SPA_ASYNC_PROBE) {
spa_vdev_state_enter(spa, SCL_NONE);
spa_async_probe(spa, spa->spa_root_vdev);
(void) spa_vdev_state_exit(spa, NULL, 0);
}
/*
* If any devices are done replacing, detach them.
*/
if (tasks & SPA_ASYNC_RESILVER_DONE ||
tasks & SPA_ASYNC_REBUILD_DONE) {
spa_vdev_resilver_done(spa);
}
/*
* Kick off a resilver.
*/
if (tasks & SPA_ASYNC_RESILVER &&
!vdev_rebuild_active(spa->spa_root_vdev) &&
(!dsl_scan_resilvering(dp) ||
!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_RESILVER_DEFER)))
dsl_scan_restart_resilver(dp, 0);
if (tasks & SPA_ASYNC_INITIALIZE_RESTART) {
mutex_enter(&spa_namespace_lock);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vdev_initialize_restart(spa->spa_root_vdev);
spa_config_exit(spa, SCL_CONFIG, FTAG);
mutex_exit(&spa_namespace_lock);
}
if (tasks & SPA_ASYNC_TRIM_RESTART) {
mutex_enter(&spa_namespace_lock);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vdev_trim_restart(spa->spa_root_vdev);
spa_config_exit(spa, SCL_CONFIG, FTAG);
mutex_exit(&spa_namespace_lock);
}
if (tasks & SPA_ASYNC_AUTOTRIM_RESTART) {
mutex_enter(&spa_namespace_lock);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vdev_autotrim_restart(spa);
spa_config_exit(spa, SCL_CONFIG, FTAG);
mutex_exit(&spa_namespace_lock);
}
/*
* Kick off L2 cache whole device TRIM.
*/
if (tasks & SPA_ASYNC_L2CACHE_TRIM) {
mutex_enter(&spa_namespace_lock);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vdev_trim_l2arc(spa);
spa_config_exit(spa, SCL_CONFIG, FTAG);
mutex_exit(&spa_namespace_lock);
}
/*
* Kick off L2 cache rebuilding.
*/
if (tasks & SPA_ASYNC_L2CACHE_REBUILD) {
mutex_enter(&spa_namespace_lock);
spa_config_enter(spa, SCL_L2ARC, FTAG, RW_READER);
l2arc_spa_rebuild_start(spa);
spa_config_exit(spa, SCL_L2ARC, FTAG);
mutex_exit(&spa_namespace_lock);
}
/*
* Let the world know that we're done.
*/
mutex_enter(&spa->spa_async_lock);
spa->spa_async_thread = NULL;
cv_broadcast(&spa->spa_async_cv);
mutex_exit(&spa->spa_async_lock);
thread_exit();
}
void
spa_async_suspend(spa_t *spa)
{
mutex_enter(&spa->spa_async_lock);
spa->spa_async_suspended++;
while (spa->spa_async_thread != NULL)
cv_wait(&spa->spa_async_cv, &spa->spa_async_lock);
mutex_exit(&spa->spa_async_lock);
spa_vdev_remove_suspend(spa);
zthr_t *condense_thread = spa->spa_condense_zthr;
if (condense_thread != NULL)
zthr_cancel(condense_thread);
zthr_t *discard_thread = spa->spa_checkpoint_discard_zthr;
if (discard_thread != NULL)
zthr_cancel(discard_thread);
zthr_t *ll_delete_thread = spa->spa_livelist_delete_zthr;
if (ll_delete_thread != NULL)
zthr_cancel(ll_delete_thread);
zthr_t *ll_condense_thread = spa->spa_livelist_condense_zthr;
if (ll_condense_thread != NULL)
zthr_cancel(ll_condense_thread);
}
void
spa_async_resume(spa_t *spa)
{
mutex_enter(&spa->spa_async_lock);
ASSERT(spa->spa_async_suspended != 0);
spa->spa_async_suspended--;
mutex_exit(&spa->spa_async_lock);
spa_restart_removal(spa);
zthr_t *condense_thread = spa->spa_condense_zthr;
if (condense_thread != NULL)
zthr_resume(condense_thread);
zthr_t *discard_thread = spa->spa_checkpoint_discard_zthr;
if (discard_thread != NULL)
zthr_resume(discard_thread);
zthr_t *ll_delete_thread = spa->spa_livelist_delete_zthr;
if (ll_delete_thread != NULL)
zthr_resume(ll_delete_thread);
zthr_t *ll_condense_thread = spa->spa_livelist_condense_zthr;
if (ll_condense_thread != NULL)
zthr_resume(ll_condense_thread);
}
static boolean_t
spa_async_tasks_pending(spa_t *spa)
{
uint_t non_config_tasks;
uint_t config_task;
boolean_t config_task_suspended;
non_config_tasks = spa->spa_async_tasks & ~SPA_ASYNC_CONFIG_UPDATE;
config_task = spa->spa_async_tasks & SPA_ASYNC_CONFIG_UPDATE;
if (spa->spa_ccw_fail_time == 0) {
config_task_suspended = B_FALSE;
} else {
config_task_suspended =
(gethrtime() - spa->spa_ccw_fail_time) <
((hrtime_t)zfs_ccw_retry_interval * NANOSEC);
}
return (non_config_tasks || (config_task && !config_task_suspended));
}
static void
spa_async_dispatch(spa_t *spa)
{
mutex_enter(&spa->spa_async_lock);
if (spa_async_tasks_pending(spa) &&
!spa->spa_async_suspended &&
spa->spa_async_thread == NULL)
spa->spa_async_thread = thread_create(NULL, 0,
spa_async_thread, spa, 0, &p0, TS_RUN, maxclsyspri);
mutex_exit(&spa->spa_async_lock);
}
void
spa_async_request(spa_t *spa, int task)
{
zfs_dbgmsg("spa=%s async request task=%u", spa->spa_name, task);
mutex_enter(&spa->spa_async_lock);
spa->spa_async_tasks |= task;
mutex_exit(&spa->spa_async_lock);
}
int
spa_async_tasks(spa_t *spa)
{
return (spa->spa_async_tasks);
}
/*
* ==========================================================================
* SPA syncing routines
* ==========================================================================
*/
static int
bpobj_enqueue_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed,
dmu_tx_t *tx)
{
bpobj_t *bpo = arg;
bpobj_enqueue(bpo, bp, bp_freed, tx);
return (0);
}
int
bpobj_enqueue_alloc_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
{
return (bpobj_enqueue_cb(arg, bp, B_FALSE, tx));
}
int
bpobj_enqueue_free_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
{
return (bpobj_enqueue_cb(arg, bp, B_TRUE, tx));
}
static int
spa_free_sync_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
{
zio_t *pio = arg;
zio_nowait(zio_free_sync(pio, pio->io_spa, dmu_tx_get_txg(tx), bp,
pio->io_flags));
return (0);
}
static int
bpobj_spa_free_sync_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed,
dmu_tx_t *tx)
{
ASSERT(!bp_freed);
return (spa_free_sync_cb(arg, bp, tx));
}
/*
* Note: this simple function is not inlined to make it easier to dtrace the
* amount of time spent syncing frees.
*/
static void
spa_sync_frees(spa_t *spa, bplist_t *bpl, dmu_tx_t *tx)
{
zio_t *zio = zio_root(spa, NULL, NULL, 0);
bplist_iterate(bpl, spa_free_sync_cb, zio, tx);
VERIFY(zio_wait(zio) == 0);
}
/*
* Note: this simple function is not inlined to make it easier to dtrace the
* amount of time spent syncing deferred frees.
*/
static void
spa_sync_deferred_frees(spa_t *spa, dmu_tx_t *tx)
{
if (spa_sync_pass(spa) != 1)
return;
/*
* Note:
* If the log space map feature is active, we stop deferring
* frees to the next TXG and therefore running this function
* would be considered a no-op as spa_deferred_bpobj should
* not have any entries.
*
* That said we run this function anyway (instead of returning
* immediately) for the edge-case scenario where we just
* activated the log space map feature in this TXG but we have
* deferred frees from the previous TXG.
*/
zio_t *zio = zio_root(spa, NULL, NULL, 0);
VERIFY3U(bpobj_iterate(&spa->spa_deferred_bpobj,
bpobj_spa_free_sync_cb, zio, tx), ==, 0);
VERIFY0(zio_wait(zio));
}
static void
spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx)
{
char *packed = NULL;
size_t bufsize;
size_t nvsize = 0;
dmu_buf_t *db;
VERIFY(nvlist_size(nv, &nvsize, NV_ENCODE_XDR) == 0);
/*
* Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration
* information. This avoids the dmu_buf_will_dirty() path and
* saves us a pre-read to get data we don't actually care about.
*/
bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE);
packed = vmem_alloc(bufsize, KM_SLEEP);
VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR,
KM_SLEEP) == 0);
bzero(packed + nvsize, bufsize - nvsize);
dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx);
vmem_free(packed, bufsize);
VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db));
dmu_buf_will_dirty(db, tx);
*(uint64_t *)db->db_data = nvsize;
dmu_buf_rele(db, FTAG);
}
static void
spa_sync_aux_dev(spa_t *spa, spa_aux_vdev_t *sav, dmu_tx_t *tx,
const char *config, const char *entry)
{
nvlist_t *nvroot;
nvlist_t **list;
int i;
if (!sav->sav_sync)
return;
/*
* Update the MOS nvlist describing the list of available devices.
* spa_validate_aux() will have already made sure this nvlist is
* valid and the vdevs are labeled appropriately.
*/
if (sav->sav_object == 0) {
sav->sav_object = dmu_object_alloc(spa->spa_meta_objset,
DMU_OT_PACKED_NVLIST, 1 << 14, DMU_OT_PACKED_NVLIST_SIZE,
sizeof (uint64_t), tx);
VERIFY(zap_update(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, entry, sizeof (uint64_t), 1,
&sav->sav_object, tx) == 0);
}
nvroot = fnvlist_alloc();
if (sav->sav_count == 0) {
fnvlist_add_nvlist_array(nvroot, config,
(const nvlist_t * const *)NULL, 0);
} else {
list = kmem_alloc(sav->sav_count*sizeof (void *), KM_SLEEP);
for (i = 0; i < sav->sav_count; i++)
list[i] = vdev_config_generate(spa, sav->sav_vdevs[i],
B_FALSE, VDEV_CONFIG_L2CACHE);
fnvlist_add_nvlist_array(nvroot, config,
(const nvlist_t * const *)list, sav->sav_count);
for (i = 0; i < sav->sav_count; i++)
nvlist_free(list[i]);
kmem_free(list, sav->sav_count * sizeof (void *));
}
spa_sync_nvlist(spa, sav->sav_object, nvroot, tx);
nvlist_free(nvroot);
sav->sav_sync = B_FALSE;
}
/*
* Rebuild spa's all-vdev ZAP from the vdev ZAPs indicated in each vdev_t.
* The all-vdev ZAP must be empty.
*/
static void
spa_avz_build(vdev_t *vd, uint64_t avz, dmu_tx_t *tx)
{
spa_t *spa = vd->vdev_spa;
if (vd->vdev_top_zap != 0) {
VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
vd->vdev_top_zap, tx));
}
if (vd->vdev_leaf_zap != 0) {
VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
vd->vdev_leaf_zap, tx));
}
for (uint64_t i = 0; i < vd->vdev_children; i++) {
spa_avz_build(vd->vdev_child[i], avz, tx);
}
}
static void
spa_sync_config_object(spa_t *spa, dmu_tx_t *tx)
{
nvlist_t *config;
/*
* If the pool is being imported from a pre-per-vdev-ZAP version of ZFS,
* its config may not be dirty but we still need to build per-vdev ZAPs.
* Similarly, if the pool is being assembled (e.g. after a split), we
* need to rebuild the AVZ although the config may not be dirty.
*/
if (list_is_empty(&spa->spa_config_dirty_list) &&
spa->spa_avz_action == AVZ_ACTION_NONE)
return;
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
ASSERT(spa->spa_avz_action == AVZ_ACTION_NONE ||
spa->spa_avz_action == AVZ_ACTION_INITIALIZE ||
spa->spa_all_vdev_zaps != 0);
if (spa->spa_avz_action == AVZ_ACTION_REBUILD) {
/* Make and build the new AVZ */
uint64_t new_avz = zap_create(spa->spa_meta_objset,
DMU_OTN_ZAP_METADATA, DMU_OT_NONE, 0, tx);
spa_avz_build(spa->spa_root_vdev, new_avz, tx);
/* Diff old AVZ with new one */
zap_cursor_t zc;
zap_attribute_t za;
for (zap_cursor_init(&zc, spa->spa_meta_objset,
spa->spa_all_vdev_zaps);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
uint64_t vdzap = za.za_first_integer;
if (zap_lookup_int(spa->spa_meta_objset, new_avz,
vdzap) == ENOENT) {
/*
* ZAP is listed in old AVZ but not in new one;
* destroy it
*/
VERIFY0(zap_destroy(spa->spa_meta_objset, vdzap,
tx));
}
}
zap_cursor_fini(&zc);
/* Destroy the old AVZ */
VERIFY0(zap_destroy(spa->spa_meta_objset,
spa->spa_all_vdev_zaps, tx));
/* Replace the old AVZ in the dir obj with the new one */
VERIFY0(zap_update(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP,
sizeof (new_avz), 1, &new_avz, tx));
spa->spa_all_vdev_zaps = new_avz;
} else if (spa->spa_avz_action == AVZ_ACTION_DESTROY) {
zap_cursor_t zc;
zap_attribute_t za;
/* Walk through the AVZ and destroy all listed ZAPs */
for (zap_cursor_init(&zc, spa->spa_meta_objset,
spa->spa_all_vdev_zaps);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
uint64_t zap = za.za_first_integer;
VERIFY0(zap_destroy(spa->spa_meta_objset, zap, tx));
}
zap_cursor_fini(&zc);
/* Destroy and unlink the AVZ itself */
VERIFY0(zap_destroy(spa->spa_meta_objset,
spa->spa_all_vdev_zaps, tx));
VERIFY0(zap_remove(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP, tx));
spa->spa_all_vdev_zaps = 0;
}
if (spa->spa_all_vdev_zaps == 0) {
spa->spa_all_vdev_zaps = zap_create_link(spa->spa_meta_objset,
DMU_OTN_ZAP_METADATA, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_VDEV_ZAP_MAP, tx);
}
spa->spa_avz_action = AVZ_ACTION_NONE;
/* Create ZAPs for vdevs that don't have them. */
vdev_construct_zaps(spa->spa_root_vdev, tx);
config = spa_config_generate(spa, spa->spa_root_vdev,
dmu_tx_get_txg(tx), B_FALSE);
/*
* If we're upgrading the spa version then make sure that
* the config object gets updated with the correct version.
*/
if (spa->spa_ubsync.ub_version < spa->spa_uberblock.ub_version)
fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
spa->spa_uberblock.ub_version);
spa_config_exit(spa, SCL_STATE, FTAG);
nvlist_free(spa->spa_config_syncing);
spa->spa_config_syncing = config;
spa_sync_nvlist(spa, spa->spa_config_object, config, tx);
}
static void
spa_sync_version(void *arg, dmu_tx_t *tx)
{
uint64_t *versionp = arg;
uint64_t version = *versionp;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
/*
* Setting the version is special cased when first creating the pool.
*/
ASSERT(tx->tx_txg != TXG_INITIAL);
ASSERT(SPA_VERSION_IS_SUPPORTED(version));
ASSERT(version >= spa_version(spa));
spa->spa_uberblock.ub_version = version;
vdev_config_dirty(spa->spa_root_vdev);
spa_history_log_internal(spa, "set", tx, "version=%lld",
(longlong_t)version);
}
/*
* Set zpool properties.
*/
static void
spa_sync_props(void *arg, dmu_tx_t *tx)
{
nvlist_t *nvp = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
objset_t *mos = spa->spa_meta_objset;
nvpair_t *elem = NULL;
mutex_enter(&spa->spa_props_lock);
while ((elem = nvlist_next_nvpair(nvp, elem))) {
uint64_t intval;
char *strval, *fname;
zpool_prop_t prop;
const char *propname;
zprop_type_t proptype;
spa_feature_t fid;
switch (prop = zpool_name_to_prop(nvpair_name(elem))) {
case ZPOOL_PROP_INVAL:
/*
* We checked this earlier in spa_prop_validate().
*/
ASSERT(zpool_prop_feature(nvpair_name(elem)));
fname = strchr(nvpair_name(elem), '@') + 1;
VERIFY0(zfeature_lookup_name(fname, &fid));
spa_feature_enable(spa, fid, tx);
spa_history_log_internal(spa, "set", tx,
"%s=enabled", nvpair_name(elem));
break;
case ZPOOL_PROP_VERSION:
intval = fnvpair_value_uint64(elem);
/*
* The version is synced separately before other
* properties and should be correct by now.
*/
ASSERT3U(spa_version(spa), >=, intval);
break;
case ZPOOL_PROP_ALTROOT:
/*
* 'altroot' is a non-persistent property. It should
* have been set temporarily at creation or import time.
*/
ASSERT(spa->spa_root != NULL);
break;
case ZPOOL_PROP_READONLY:
case ZPOOL_PROP_CACHEFILE:
/*
* 'readonly' and 'cachefile' are also non-persistent
* properties.
*/
break;
case ZPOOL_PROP_COMMENT:
strval = fnvpair_value_string(elem);
if (spa->spa_comment != NULL)
spa_strfree(spa->spa_comment);
spa->spa_comment = spa_strdup(strval);
/*
* We need to dirty the configuration on all the vdevs
* so that their labels get updated. We also need to
* update the cache file to keep it in sync with the
* MOS version. It's unnecessary to do this for pool
* creation since the vdev's configuration has already
* been dirtied.
*/
if (tx->tx_txg != TXG_INITIAL) {
vdev_config_dirty(spa->spa_root_vdev);
spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
}
spa_history_log_internal(spa, "set", tx,
"%s=%s", nvpair_name(elem), strval);
break;
case ZPOOL_PROP_COMPATIBILITY:
strval = fnvpair_value_string(elem);
if (spa->spa_compatibility != NULL)
spa_strfree(spa->spa_compatibility);
spa->spa_compatibility = spa_strdup(strval);
/*
* Dirty the configuration on vdevs as above.
*/
if (tx->tx_txg != TXG_INITIAL) {
vdev_config_dirty(spa->spa_root_vdev);
spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
}
spa_history_log_internal(spa, "set", tx,
"%s=%s", nvpair_name(elem), strval);
break;
default:
/*
* Set pool property values in the poolprops mos object.
*/
if (spa->spa_pool_props_object == 0) {
spa->spa_pool_props_object =
zap_create_link(mos, DMU_OT_POOL_PROPS,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS,
tx);
}
/* normalize the property name */
propname = zpool_prop_to_name(prop);
proptype = zpool_prop_get_type(prop);
if (nvpair_type(elem) == DATA_TYPE_STRING) {
ASSERT(proptype == PROP_TYPE_STRING);
strval = fnvpair_value_string(elem);
VERIFY0(zap_update(mos,
spa->spa_pool_props_object, propname,
1, strlen(strval) + 1, strval, tx));
spa_history_log_internal(spa, "set", tx,
"%s=%s", nvpair_name(elem), strval);
} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
intval = fnvpair_value_uint64(elem);
if (proptype == PROP_TYPE_INDEX) {
const char *unused;
VERIFY0(zpool_prop_index_to_string(
prop, intval, &unused));
}
VERIFY0(zap_update(mos,
spa->spa_pool_props_object, propname,
8, 1, &intval, tx));
spa_history_log_internal(spa, "set", tx,
"%s=%lld", nvpair_name(elem),
(longlong_t)intval);
} else {
ASSERT(0); /* not allowed */
}
switch (prop) {
case ZPOOL_PROP_DELEGATION:
spa->spa_delegation = intval;
break;
case ZPOOL_PROP_BOOTFS:
spa->spa_bootfs = intval;
break;
case ZPOOL_PROP_FAILUREMODE:
spa->spa_failmode = intval;
break;
case ZPOOL_PROP_AUTOTRIM:
spa->spa_autotrim = intval;
spa_async_request(spa,
SPA_ASYNC_AUTOTRIM_RESTART);
break;
case ZPOOL_PROP_AUTOEXPAND:
spa->spa_autoexpand = intval;
if (tx->tx_txg != TXG_INITIAL)
spa_async_request(spa,
SPA_ASYNC_AUTOEXPAND);
break;
case ZPOOL_PROP_MULTIHOST:
spa->spa_multihost = intval;
break;
default:
break;
}
}
}
mutex_exit(&spa->spa_props_lock);
}
/*
* Perform one-time upgrade on-disk changes. spa_version() does not
* reflect the new version this txg, so there must be no changes this
* txg to anything that the upgrade code depends on after it executes.
* Therefore this must be called after dsl_pool_sync() does the sync
* tasks.
*/
static void
spa_sync_upgrades(spa_t *spa, dmu_tx_t *tx)
{
if (spa_sync_pass(spa) != 1)
return;
dsl_pool_t *dp = spa->spa_dsl_pool;
rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN &&
spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) {
dsl_pool_create_origin(dp, tx);
/* Keeping the origin open increases spa_minref */
spa->spa_minref += 3;
}
if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES &&
spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) {
dsl_pool_upgrade_clones(dp, tx);
}
if (spa->spa_ubsync.ub_version < SPA_VERSION_DIR_CLONES &&
spa->spa_uberblock.ub_version >= SPA_VERSION_DIR_CLONES) {
dsl_pool_upgrade_dir_clones(dp, tx);
/* Keeping the freedir open increases spa_minref */
spa->spa_minref += 3;
}
if (spa->spa_ubsync.ub_version < SPA_VERSION_FEATURES &&
spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
spa_feature_create_zap_objects(spa, tx);
}
/*
* LZ4_COMPRESS feature's behaviour was changed to activate_on_enable
* when possibility to use lz4 compression for metadata was added
* Old pools that have this feature enabled must be upgraded to have
* this feature active
*/
if (spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
boolean_t lz4_en = spa_feature_is_enabled(spa,
SPA_FEATURE_LZ4_COMPRESS);
boolean_t lz4_ac = spa_feature_is_active(spa,
SPA_FEATURE_LZ4_COMPRESS);
if (lz4_en && !lz4_ac)
spa_feature_incr(spa, SPA_FEATURE_LZ4_COMPRESS, tx);
}
/*
* If we haven't written the salt, do so now. Note that the
* feature may not be activated yet, but that's fine since
* the presence of this ZAP entry is backwards compatible.
*/
if (zap_contains(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_CHECKSUM_SALT) == ENOENT) {
VERIFY0(zap_add(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CHECKSUM_SALT, 1,
sizeof (spa->spa_cksum_salt.zcs_bytes),
spa->spa_cksum_salt.zcs_bytes, tx));
}
rrw_exit(&dp->dp_config_rwlock, FTAG);
}
static void
vdev_indirect_state_sync_verify(vdev_t *vd)
{
vdev_indirect_mapping_t *vim __maybe_unused = vd->vdev_indirect_mapping;
vdev_indirect_births_t *vib __maybe_unused = vd->vdev_indirect_births;
if (vd->vdev_ops == &vdev_indirect_ops) {
ASSERT(vim != NULL);
ASSERT(vib != NULL);
}
uint64_t obsolete_sm_object = 0;
ASSERT0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
if (obsolete_sm_object != 0) {
ASSERT(vd->vdev_obsolete_sm != NULL);
ASSERT(vd->vdev_removing ||
vd->vdev_ops == &vdev_indirect_ops);
ASSERT(vdev_indirect_mapping_num_entries(vim) > 0);
ASSERT(vdev_indirect_mapping_bytes_mapped(vim) > 0);
ASSERT3U(obsolete_sm_object, ==,
space_map_object(vd->vdev_obsolete_sm));
ASSERT3U(vdev_indirect_mapping_bytes_mapped(vim), >=,
space_map_allocated(vd->vdev_obsolete_sm));
}
ASSERT(vd->vdev_obsolete_segments != NULL);
/*
* Since frees / remaps to an indirect vdev can only
* happen in syncing context, the obsolete segments
* tree must be empty when we start syncing.
*/
ASSERT0(range_tree_space(vd->vdev_obsolete_segments));
}
/*
* Set the top-level vdev's max queue depth. Evaluate each top-level's
* async write queue depth in case it changed. The max queue depth will
* not change in the middle of syncing out this txg.
*/
static void
spa_sync_adjust_vdev_max_queue_depth(spa_t *spa)
{
ASSERT(spa_writeable(spa));
vdev_t *rvd = spa->spa_root_vdev;
uint32_t max_queue_depth = zfs_vdev_async_write_max_active *
zfs_vdev_queue_depth_pct / 100;
metaslab_class_t *normal = spa_normal_class(spa);
metaslab_class_t *special = spa_special_class(spa);
metaslab_class_t *dedup = spa_dedup_class(spa);
uint64_t slots_per_allocator = 0;
for (int c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
metaslab_group_t *mg = tvd->vdev_mg;
if (mg == NULL || !metaslab_group_initialized(mg))
continue;
metaslab_class_t *mc = mg->mg_class;
if (mc != normal && mc != special && mc != dedup)
continue;
/*
* It is safe to do a lock-free check here because only async
* allocations look at mg_max_alloc_queue_depth, and async
* allocations all happen from spa_sync().
*/
for (int i = 0; i < mg->mg_allocators; i++) {
ASSERT0(zfs_refcount_count(
&(mg->mg_allocator[i].mga_alloc_queue_depth)));
}
mg->mg_max_alloc_queue_depth = max_queue_depth;
for (int i = 0; i < mg->mg_allocators; i++) {
mg->mg_allocator[i].mga_cur_max_alloc_queue_depth =
zfs_vdev_def_queue_depth;
}
slots_per_allocator += zfs_vdev_def_queue_depth;
}
for (int i = 0; i < spa->spa_alloc_count; i++) {
ASSERT0(zfs_refcount_count(&normal->mc_allocator[i].
mca_alloc_slots));
ASSERT0(zfs_refcount_count(&special->mc_allocator[i].
mca_alloc_slots));
ASSERT0(zfs_refcount_count(&dedup->mc_allocator[i].
mca_alloc_slots));
normal->mc_allocator[i].mca_alloc_max_slots =
slots_per_allocator;
special->mc_allocator[i].mca_alloc_max_slots =
slots_per_allocator;
dedup->mc_allocator[i].mca_alloc_max_slots =
slots_per_allocator;
}
normal->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
special->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
dedup->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
}
static void
spa_sync_condense_indirect(spa_t *spa, dmu_tx_t *tx)
{
ASSERT(spa_writeable(spa));
vdev_t *rvd = spa->spa_root_vdev;
for (int c = 0; c < rvd->vdev_children; c++) {
vdev_t *vd = rvd->vdev_child[c];
vdev_indirect_state_sync_verify(vd);
if (vdev_indirect_should_condense(vd)) {
spa_condense_indirect_start_sync(vd, tx);
break;
}
}
}
static void
spa_sync_iterate_to_convergence(spa_t *spa, dmu_tx_t *tx)
{
objset_t *mos = spa->spa_meta_objset;
dsl_pool_t *dp = spa->spa_dsl_pool;
uint64_t txg = tx->tx_txg;
bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK];
do {
int pass = ++spa->spa_sync_pass;
spa_sync_config_object(spa, tx);
spa_sync_aux_dev(spa, &spa->spa_spares, tx,
ZPOOL_CONFIG_SPARES, DMU_POOL_SPARES);
spa_sync_aux_dev(spa, &spa->spa_l2cache, tx,
ZPOOL_CONFIG_L2CACHE, DMU_POOL_L2CACHE);
spa_errlog_sync(spa, txg);
dsl_pool_sync(dp, txg);
if (pass < zfs_sync_pass_deferred_free ||
spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) {
/*
* If the log space map feature is active we don't
* care about deferred frees and the deferred bpobj
* as the log space map should effectively have the
* same results (i.e. appending only to one object).
*/
spa_sync_frees(spa, free_bpl, tx);
} else {
/*
* We can not defer frees in pass 1, because
* we sync the deferred frees later in pass 1.
*/
ASSERT3U(pass, >, 1);
bplist_iterate(free_bpl, bpobj_enqueue_alloc_cb,
&spa->spa_deferred_bpobj, tx);
}
ddt_sync(spa, txg);
dsl_scan_sync(dp, tx);
svr_sync(spa, tx);
spa_sync_upgrades(spa, tx);
spa_flush_metaslabs(spa, tx);
vdev_t *vd = NULL;
while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, txg))
!= NULL)
vdev_sync(vd, txg);
/*
* Note: We need to check if the MOS is dirty because we could
* have marked the MOS dirty without updating the uberblock
* (e.g. if we have sync tasks but no dirty user data). We need
* to check the uberblock's rootbp because it is updated if we
* have synced out dirty data (though in this case the MOS will
* most likely also be dirty due to second order effects, we
* don't want to rely on that here).
*/
if (pass == 1 &&
spa->spa_uberblock.ub_rootbp.blk_birth < txg &&
!dmu_objset_is_dirty(mos, txg)) {
/*
* Nothing changed on the first pass, therefore this
* TXG is a no-op. Avoid syncing deferred frees, so
* that we can keep this TXG as a no-op.
*/
ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg));
ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg));
ASSERT(txg_list_empty(&dp->dp_sync_tasks, txg));
ASSERT(txg_list_empty(&dp->dp_early_sync_tasks, txg));
break;
}
spa_sync_deferred_frees(spa, tx);
} while (dmu_objset_is_dirty(mos, txg));
}
/*
* Rewrite the vdev configuration (which includes the uberblock) to
* commit the transaction group.
*
* If there are no dirty vdevs, we sync the uberblock to a few random
* top-level vdevs that are known to be visible in the config cache
* (see spa_vdev_add() for a complete description). If there *are* dirty
* vdevs, sync the uberblock to all vdevs.
*/
static void
spa_sync_rewrite_vdev_config(spa_t *spa, dmu_tx_t *tx)
{
vdev_t *rvd = spa->spa_root_vdev;
uint64_t txg = tx->tx_txg;
for (;;) {
int error = 0;
/*
* We hold SCL_STATE to prevent vdev open/close/etc.
* while we're attempting to write the vdev labels.
*/
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
if (list_is_empty(&spa->spa_config_dirty_list)) {
vdev_t *svd[SPA_SYNC_MIN_VDEVS] = { NULL };
int svdcount = 0;
int children = rvd->vdev_children;
int c0 = random_in_range(children);
for (int c = 0; c < children; c++) {
vdev_t *vd =
rvd->vdev_child[(c0 + c) % children];
/* Stop when revisiting the first vdev */
if (c > 0 && svd[0] == vd)
break;
if (vd->vdev_ms_array == 0 ||
vd->vdev_islog ||
!vdev_is_concrete(vd))
continue;
svd[svdcount++] = vd;
if (svdcount == SPA_SYNC_MIN_VDEVS)
break;
}
error = vdev_config_sync(svd, svdcount, txg);
} else {
error = vdev_config_sync(rvd->vdev_child,
rvd->vdev_children, txg);
}
if (error == 0)
spa->spa_last_synced_guid = rvd->vdev_guid;
spa_config_exit(spa, SCL_STATE, FTAG);
if (error == 0)
break;
zio_suspend(spa, NULL, ZIO_SUSPEND_IOERR);
zio_resume_wait(spa);
}
}
/*
* Sync the specified transaction group. New blocks may be dirtied as
* part of the process, so we iterate until it converges.
*/
void
spa_sync(spa_t *spa, uint64_t txg)
{
vdev_t *vd = NULL;
VERIFY(spa_writeable(spa));
/*
* Wait for i/os issued in open context that need to complete
* before this txg syncs.
*/
(void) zio_wait(spa->spa_txg_zio[txg & TXG_MASK]);
spa->spa_txg_zio[txg & TXG_MASK] = zio_root(spa, NULL, NULL,
ZIO_FLAG_CANFAIL);
/*
* Lock out configuration changes.
*/
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
spa->spa_syncing_txg = txg;
spa->spa_sync_pass = 0;
for (int i = 0; i < spa->spa_alloc_count; i++) {
mutex_enter(&spa->spa_allocs[i].spaa_lock);
VERIFY0(avl_numnodes(&spa->spa_allocs[i].spaa_tree));
mutex_exit(&spa->spa_allocs[i].spaa_lock);
}
/*
* If there are any pending vdev state changes, convert them
* into config changes that go out with this transaction group.
*/
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
while (list_head(&spa->spa_state_dirty_list) != NULL) {
/*
* We need the write lock here because, for aux vdevs,
* calling vdev_config_dirty() modifies sav_config.
* This is ugly and will become unnecessary when we
* eliminate the aux vdev wart by integrating all vdevs
* into the root vdev tree.
*/
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_WRITER);
while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) {
vdev_state_clean(vd);
vdev_config_dirty(vd);
}
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
}
spa_config_exit(spa, SCL_STATE, FTAG);
dsl_pool_t *dp = spa->spa_dsl_pool;
dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg);
spa->spa_sync_starttime = gethrtime();
taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq,
spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() +
NSEC_TO_TICK(spa->spa_deadman_synctime));
/*
* If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg,
* set spa_deflate if we have no raid-z vdevs.
*/
if (spa->spa_ubsync.ub_version < SPA_VERSION_RAIDZ_DEFLATE &&
spa->spa_uberblock.ub_version >= SPA_VERSION_RAIDZ_DEFLATE) {
vdev_t *rvd = spa->spa_root_vdev;
int i;
for (i = 0; i < rvd->vdev_children; i++) {
vd = rvd->vdev_child[i];
if (vd->vdev_deflate_ratio != SPA_MINBLOCKSIZE)
break;
}
if (i == rvd->vdev_children) {
spa->spa_deflate = TRUE;
VERIFY0(zap_add(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
sizeof (uint64_t), 1, &spa->spa_deflate, tx));
}
}
spa_sync_adjust_vdev_max_queue_depth(spa);
spa_sync_condense_indirect(spa, tx);
spa_sync_iterate_to_convergence(spa, tx);
#ifdef ZFS_DEBUG
if (!list_is_empty(&spa->spa_config_dirty_list)) {
/*
* Make sure that the number of ZAPs for all the vdevs matches
* the number of ZAPs in the per-vdev ZAP list. This only gets
* called if the config is dirty; otherwise there may be
* outstanding AVZ operations that weren't completed in
* spa_sync_config_object.
*/
uint64_t all_vdev_zap_entry_count;
ASSERT0(zap_count(spa->spa_meta_objset,
spa->spa_all_vdev_zaps, &all_vdev_zap_entry_count));
ASSERT3U(vdev_count_verify_zaps(spa->spa_root_vdev), ==,
all_vdev_zap_entry_count);
}
#endif
if (spa->spa_vdev_removal != NULL) {
ASSERT0(spa->spa_vdev_removal->svr_bytes_done[txg & TXG_MASK]);
}
spa_sync_rewrite_vdev_config(spa, tx);
dmu_tx_commit(tx);
taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
spa->spa_deadman_tqid = 0;
/*
* Clear the dirty config list.
*/
while ((vd = list_head(&spa->spa_config_dirty_list)) != NULL)
vdev_config_clean(vd);
/*
* Now that the new config has synced transactionally,
* let it become visible to the config cache.
*/
if (spa->spa_config_syncing != NULL) {
spa_config_set(spa, spa->spa_config_syncing);
spa->spa_config_txg = txg;
spa->spa_config_syncing = NULL;
}
dsl_pool_sync_done(dp, txg);
for (int i = 0; i < spa->spa_alloc_count; i++) {
mutex_enter(&spa->spa_allocs[i].spaa_lock);
VERIFY0(avl_numnodes(&spa->spa_allocs[i].spaa_tree));
mutex_exit(&spa->spa_allocs[i].spaa_lock);
}
/*
* Update usable space statistics.
*/
while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)))
!= NULL)
vdev_sync_done(vd, txg);
metaslab_class_evict_old(spa->spa_normal_class, txg);
metaslab_class_evict_old(spa->spa_log_class, txg);
spa_sync_close_syncing_log_sm(spa);
spa_update_dspace(spa);
/*
* It had better be the case that we didn't dirty anything
* since vdev_config_sync().
*/
ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg));
ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg));
ASSERT(txg_list_empty(&spa->spa_vdev_txg_list, txg));
while (zfs_pause_spa_sync)
delay(1);
spa->spa_sync_pass = 0;
/*
* Update the last synced uberblock here. We want to do this at
* the end of spa_sync() so that consumers of spa_last_synced_txg()
* will be guaranteed that all the processing associated with
* that txg has been completed.
*/
spa->spa_ubsync = spa->spa_uberblock;
spa_config_exit(spa, SCL_CONFIG, FTAG);
spa_handle_ignored_writes(spa);
/*
* If any async tasks have been requested, kick them off.
*/
spa_async_dispatch(spa);
}
/*
* Sync all pools. We don't want to hold the namespace lock across these
* operations, so we take a reference on the spa_t and drop the lock during the
* sync.
*/
void
spa_sync_allpools(void)
{
spa_t *spa = NULL;
mutex_enter(&spa_namespace_lock);
while ((spa = spa_next(spa)) != NULL) {
if (spa_state(spa) != POOL_STATE_ACTIVE ||
!spa_writeable(spa) || spa_suspended(spa))
continue;
spa_open_ref(spa, FTAG);
mutex_exit(&spa_namespace_lock);
txg_wait_synced(spa_get_dsl(spa), 0);
mutex_enter(&spa_namespace_lock);
spa_close(spa, FTAG);
}
mutex_exit(&spa_namespace_lock);
}
/*
* ==========================================================================
* Miscellaneous routines
* ==========================================================================
*/
/*
* Remove all pools in the system.
*/
void
spa_evict_all(void)
{
spa_t *spa;
/*
* Remove all cached state. All pools should be closed now,
* so every spa in the AVL tree should be unreferenced.
*/
mutex_enter(&spa_namespace_lock);
while ((spa = spa_next(NULL)) != NULL) {
/*
* Stop async tasks. The async thread may need to detach
* a device that's been replaced, which requires grabbing
* spa_namespace_lock, so we must drop it here.
*/
spa_open_ref(spa, FTAG);
mutex_exit(&spa_namespace_lock);
spa_async_suspend(spa);
mutex_enter(&spa_namespace_lock);
spa_close(spa, FTAG);
if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
spa_unload(spa);
spa_deactivate(spa);
}
spa_remove(spa);
}
mutex_exit(&spa_namespace_lock);
}
vdev_t *
spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t aux)
{
vdev_t *vd;
int i;
if ((vd = vdev_lookup_by_guid(spa->spa_root_vdev, guid)) != NULL)
return (vd);
if (aux) {
for (i = 0; i < spa->spa_l2cache.sav_count; i++) {
vd = spa->spa_l2cache.sav_vdevs[i];
if (vd->vdev_guid == guid)
return (vd);
}
for (i = 0; i < spa->spa_spares.sav_count; i++) {
vd = spa->spa_spares.sav_vdevs[i];
if (vd->vdev_guid == guid)
return (vd);
}
}
return (NULL);
}
void
spa_upgrade(spa_t *spa, uint64_t version)
{
ASSERT(spa_writeable(spa));
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
/*
* This should only be called for a non-faulted pool, and since a
* future version would result in an unopenable pool, this shouldn't be
* possible.
*/
ASSERT(SPA_VERSION_IS_SUPPORTED(spa->spa_uberblock.ub_version));
ASSERT3U(version, >=, spa->spa_uberblock.ub_version);
spa->spa_uberblock.ub_version = version;
vdev_config_dirty(spa->spa_root_vdev);
spa_config_exit(spa, SCL_ALL, FTAG);
txg_wait_synced(spa_get_dsl(spa), 0);
}
static boolean_t
spa_has_aux_vdev(spa_t *spa, uint64_t guid, spa_aux_vdev_t *sav)
{
(void) spa;
int i;
uint64_t vdev_guid;
for (i = 0; i < sav->sav_count; i++)
if (sav->sav_vdevs[i]->vdev_guid == guid)
return (B_TRUE);
for (i = 0; i < sav->sav_npending; i++) {
if (nvlist_lookup_uint64(sav->sav_pending[i], ZPOOL_CONFIG_GUID,
&vdev_guid) == 0 && vdev_guid == guid)
return (B_TRUE);
}
return (B_FALSE);
}
boolean_t
spa_has_l2cache(spa_t *spa, uint64_t guid)
{
return (spa_has_aux_vdev(spa, guid, &spa->spa_l2cache));
}
boolean_t
spa_has_spare(spa_t *spa, uint64_t guid)
{
return (spa_has_aux_vdev(spa, guid, &spa->spa_spares));
}
/*
* Check if a pool has an active shared spare device.
* Note: reference count of an active spare is 2, as a spare and as a replace
*/
static boolean_t
spa_has_active_shared_spare(spa_t *spa)
{
int i, refcnt;
uint64_t pool;
spa_aux_vdev_t *sav = &spa->spa_spares;
for (i = 0; i < sav->sav_count; i++) {
if (spa_spare_exists(sav->sav_vdevs[i]->vdev_guid, &pool,
&refcnt) && pool != 0ULL && pool == spa_guid(spa) &&
refcnt > 2)
return (B_TRUE);
}
return (B_FALSE);
}
uint64_t
spa_total_metaslabs(spa_t *spa)
{
vdev_t *rvd = spa->spa_root_vdev;
uint64_t m = 0;
for (uint64_t c = 0; c < rvd->vdev_children; c++) {
vdev_t *vd = rvd->vdev_child[c];
if (!vdev_is_concrete(vd))
continue;
m += vd->vdev_ms_count;
}
return (m);
}
/*
* Notify any waiting threads that some activity has switched from being in-
* progress to not-in-progress so that the thread can wake up and determine
* whether it is finished waiting.
*/
void
spa_notify_waiters(spa_t *spa)
{
/*
* Acquiring spa_activities_lock here prevents the cv_broadcast from
* happening between the waiting thread's check and cv_wait.
*/
mutex_enter(&spa->spa_activities_lock);
cv_broadcast(&spa->spa_activities_cv);
mutex_exit(&spa->spa_activities_lock);
}
/*
* Notify any waiting threads that the pool is exporting, and then block until
* they are finished using the spa_t.
*/
void
spa_wake_waiters(spa_t *spa)
{
mutex_enter(&spa->spa_activities_lock);
spa->spa_waiters_cancel = B_TRUE;
cv_broadcast(&spa->spa_activities_cv);
while (spa->spa_waiters != 0)
cv_wait(&spa->spa_waiters_cv, &spa->spa_activities_lock);
spa->spa_waiters_cancel = B_FALSE;
mutex_exit(&spa->spa_activities_lock);
}
/* Whether the vdev or any of its descendants are being initialized/trimmed. */
static boolean_t
spa_vdev_activity_in_progress_impl(vdev_t *vd, zpool_wait_activity_t activity)
{
spa_t *spa = vd->vdev_spa;
ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER));
ASSERT(MUTEX_HELD(&spa->spa_activities_lock));
ASSERT(activity == ZPOOL_WAIT_INITIALIZE ||
activity == ZPOOL_WAIT_TRIM);
kmutex_t *lock = activity == ZPOOL_WAIT_INITIALIZE ?
&vd->vdev_initialize_lock : &vd->vdev_trim_lock;
mutex_exit(&spa->spa_activities_lock);
mutex_enter(lock);
mutex_enter(&spa->spa_activities_lock);
boolean_t in_progress = (activity == ZPOOL_WAIT_INITIALIZE) ?
(vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE) :
(vd->vdev_trim_state == VDEV_TRIM_ACTIVE);
mutex_exit(lock);
if (in_progress)
return (B_TRUE);
for (int i = 0; i < vd->vdev_children; i++) {
if (spa_vdev_activity_in_progress_impl(vd->vdev_child[i],
activity))
return (B_TRUE);
}
return (B_FALSE);
}
/*
* If use_guid is true, this checks whether the vdev specified by guid is
* being initialized/trimmed. Otherwise, it checks whether any vdev in the pool
* is being initialized/trimmed. The caller must hold the config lock and
* spa_activities_lock.
*/
static int
spa_vdev_activity_in_progress(spa_t *spa, boolean_t use_guid, uint64_t guid,
zpool_wait_activity_t activity, boolean_t *in_progress)
{
mutex_exit(&spa->spa_activities_lock);
spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
mutex_enter(&spa->spa_activities_lock);
vdev_t *vd;
if (use_guid) {
vd = spa_lookup_by_guid(spa, guid, B_FALSE);
if (vd == NULL || !vd->vdev_ops->vdev_op_leaf) {
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (EINVAL);
}
} else {
vd = spa->spa_root_vdev;
}
*in_progress = spa_vdev_activity_in_progress_impl(vd, activity);
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (0);
}
/*
* Locking for waiting threads
* ---------------------------
*
* Waiting threads need a way to check whether a given activity is in progress,
* and then, if it is, wait for it to complete. Each activity will have some
* in-memory representation of the relevant on-disk state which can be used to
* determine whether or not the activity is in progress. The in-memory state and
* the locking used to protect it will be different for each activity, and may
* not be suitable for use with a cvar (e.g., some state is protected by the
* config lock). To allow waiting threads to wait without any races, another
* lock, spa_activities_lock, is used.
*
* When the state is checked, both the activity-specific lock (if there is one)
* and spa_activities_lock are held. In some cases, the activity-specific lock
* is acquired explicitly (e.g. the config lock). In others, the locking is
* internal to some check (e.g. bpobj_is_empty). After checking, the waiting
* thread releases the activity-specific lock and, if the activity is in
* progress, then cv_waits using spa_activities_lock.
*
* The waiting thread is woken when another thread, one completing some
* activity, updates the state of the activity and then calls
* spa_notify_waiters, which will cv_broadcast. This 'completing' thread only
* needs to hold its activity-specific lock when updating the state, and this
* lock can (but doesn't have to) be dropped before calling spa_notify_waiters.
*
* Because spa_notify_waiters acquires spa_activities_lock before broadcasting,
* and because it is held when the waiting thread checks the state of the
* activity, it can never be the case that the completing thread both updates
* the activity state and cv_broadcasts in between the waiting thread's check
* and cv_wait. Thus, a waiting thread can never miss a wakeup.
*
* In order to prevent deadlock, when the waiting thread does its check, in some
* cases it will temporarily drop spa_activities_lock in order to acquire the
* activity-specific lock. The order in which spa_activities_lock and the
* activity specific lock are acquired in the waiting thread is determined by
* the order in which they are acquired in the completing thread; if the
* completing thread calls spa_notify_waiters with the activity-specific lock
* held, then the waiting thread must also acquire the activity-specific lock
* first.
*/
static int
spa_activity_in_progress(spa_t *spa, zpool_wait_activity_t activity,
boolean_t use_tag, uint64_t tag, boolean_t *in_progress)
{
int error = 0;
ASSERT(MUTEX_HELD(&spa->spa_activities_lock));
switch (activity) {
case ZPOOL_WAIT_CKPT_DISCARD:
*in_progress =
(spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT) &&
zap_contains(spa_meta_objset(spa),
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ZPOOL_CHECKPOINT) ==
ENOENT);
break;
case ZPOOL_WAIT_FREE:
*in_progress = ((spa_version(spa) >= SPA_VERSION_DEADLISTS &&
!bpobj_is_empty(&spa->spa_dsl_pool->dp_free_bpobj)) ||
spa_feature_is_active(spa, SPA_FEATURE_ASYNC_DESTROY) ||
spa_livelist_delete_check(spa));
break;
case ZPOOL_WAIT_INITIALIZE:
case ZPOOL_WAIT_TRIM:
error = spa_vdev_activity_in_progress(spa, use_tag, tag,
activity, in_progress);
break;
case ZPOOL_WAIT_REPLACE:
mutex_exit(&spa->spa_activities_lock);
spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
mutex_enter(&spa->spa_activities_lock);
*in_progress = vdev_replace_in_progress(spa->spa_root_vdev);
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
break;
case ZPOOL_WAIT_REMOVE:
*in_progress = (spa->spa_removing_phys.sr_state ==
DSS_SCANNING);
break;
case ZPOOL_WAIT_RESILVER:
if ((*in_progress = vdev_rebuild_active(spa->spa_root_vdev)))
break;
fallthrough;
case ZPOOL_WAIT_SCRUB:
{
boolean_t scanning, paused, is_scrub;
dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;
is_scrub = (scn->scn_phys.scn_func == POOL_SCAN_SCRUB);
scanning = (scn->scn_phys.scn_state == DSS_SCANNING);
paused = dsl_scan_is_paused_scrub(scn);
*in_progress = (scanning && !paused &&
is_scrub == (activity == ZPOOL_WAIT_SCRUB));
break;
}
default:
panic("unrecognized value for activity %d", activity);
}
return (error);
}
static int
spa_wait_common(const char *pool, zpool_wait_activity_t activity,
boolean_t use_tag, uint64_t tag, boolean_t *waited)
{
/*
* The tag is used to distinguish between instances of an activity.
* 'initialize' and 'trim' are the only activities that we use this for.
* The other activities can only have a single instance in progress in a
* pool at one time, making the tag unnecessary.
*
* There can be multiple devices being replaced at once, but since they
* all finish once resilvering finishes, we don't bother keeping track
* of them individually, we just wait for them all to finish.
*/
if (use_tag && activity != ZPOOL_WAIT_INITIALIZE &&
activity != ZPOOL_WAIT_TRIM)
return (EINVAL);
if (activity < 0 || activity >= ZPOOL_WAIT_NUM_ACTIVITIES)
return (EINVAL);
spa_t *spa;
int error = spa_open(pool, &spa, FTAG);
if (error != 0)
return (error);
/*
* Increment the spa's waiter count so that we can call spa_close and
* still ensure that the spa_t doesn't get freed before this thread is
* finished with it when the pool is exported. We want to call spa_close
* before we start waiting because otherwise the additional ref would
* prevent the pool from being exported or destroyed throughout the
* potentially long wait.
*/
mutex_enter(&spa->spa_activities_lock);
spa->spa_waiters++;
spa_close(spa, FTAG);
*waited = B_FALSE;
for (;;) {
boolean_t in_progress;
error = spa_activity_in_progress(spa, activity, use_tag, tag,
&in_progress);
if (error || !in_progress || spa->spa_waiters_cancel)
break;
*waited = B_TRUE;
if (cv_wait_sig(&spa->spa_activities_cv,
&spa->spa_activities_lock) == 0) {
error = EINTR;
break;
}
}
spa->spa_waiters--;
cv_signal(&spa->spa_waiters_cv);
mutex_exit(&spa->spa_activities_lock);
return (error);
}
/*
* Wait for a particular instance of the specified activity to complete, where
* the instance is identified by 'tag'
*/
int
spa_wait_tag(const char *pool, zpool_wait_activity_t activity, uint64_t tag,
boolean_t *waited)
{
return (spa_wait_common(pool, activity, B_TRUE, tag, waited));
}
/*
* Wait for all instances of the specified activity complete
*/
int
spa_wait(const char *pool, zpool_wait_activity_t activity, boolean_t *waited)
{
return (spa_wait_common(pool, activity, B_FALSE, 0, waited));
}
sysevent_t *
spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, const char *name)
{
sysevent_t *ev = NULL;
#ifdef _KERNEL
nvlist_t *resource;
resource = zfs_event_create(spa, vd, FM_SYSEVENT_CLASS, name, hist_nvl);
if (resource) {
ev = kmem_alloc(sizeof (sysevent_t), KM_SLEEP);
ev->resource = resource;
}
#else
(void) spa, (void) vd, (void) hist_nvl, (void) name;
#endif
return (ev);
}
void
spa_event_post(sysevent_t *ev)
{
#ifdef _KERNEL
if (ev) {
zfs_zevent_post(ev->resource, NULL, zfs_zevent_post_cb);
kmem_free(ev, sizeof (*ev));
}
#else
(void) ev;
#endif
}
/*
* Post a zevent corresponding to the given sysevent. The 'name' must be one
* of the event definitions in sys/sysevent/eventdefs.h. The payload will be
* filled in from the spa and (optionally) the vdev. This doesn't do anything
* in the userland libzpool, as we don't want consumers to misinterpret ztest
* or zdb as real changes.
*/
void
spa_event_notify(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, const char *name)
{
spa_event_post(spa_event_create(spa, vd, hist_nvl, name));
}
/* state manipulation functions */
EXPORT_SYMBOL(spa_open);
EXPORT_SYMBOL(spa_open_rewind);
EXPORT_SYMBOL(spa_get_stats);
EXPORT_SYMBOL(spa_create);
EXPORT_SYMBOL(spa_import);
EXPORT_SYMBOL(spa_tryimport);
EXPORT_SYMBOL(spa_destroy);
EXPORT_SYMBOL(spa_export);
EXPORT_SYMBOL(spa_reset);
EXPORT_SYMBOL(spa_async_request);
EXPORT_SYMBOL(spa_async_suspend);
EXPORT_SYMBOL(spa_async_resume);
EXPORT_SYMBOL(spa_inject_addref);
EXPORT_SYMBOL(spa_inject_delref);
EXPORT_SYMBOL(spa_scan_stat_init);
EXPORT_SYMBOL(spa_scan_get_stats);
/* device manipulation */
EXPORT_SYMBOL(spa_vdev_add);
EXPORT_SYMBOL(spa_vdev_attach);
EXPORT_SYMBOL(spa_vdev_detach);
EXPORT_SYMBOL(spa_vdev_setpath);
EXPORT_SYMBOL(spa_vdev_setfru);
EXPORT_SYMBOL(spa_vdev_split_mirror);
/* spare statech is global across all pools) */
EXPORT_SYMBOL(spa_spare_add);
EXPORT_SYMBOL(spa_spare_remove);
EXPORT_SYMBOL(spa_spare_exists);
EXPORT_SYMBOL(spa_spare_activate);
/* L2ARC statech is global across all pools) */
EXPORT_SYMBOL(spa_l2cache_add);
EXPORT_SYMBOL(spa_l2cache_remove);
EXPORT_SYMBOL(spa_l2cache_exists);
EXPORT_SYMBOL(spa_l2cache_activate);
EXPORT_SYMBOL(spa_l2cache_drop);
/* scanning */
EXPORT_SYMBOL(spa_scan);
EXPORT_SYMBOL(spa_scan_stop);
/* spa syncing */
EXPORT_SYMBOL(spa_sync); /* only for DMU use */
EXPORT_SYMBOL(spa_sync_allpools);
/* properties */
EXPORT_SYMBOL(spa_prop_set);
EXPORT_SYMBOL(spa_prop_get);
EXPORT_SYMBOL(spa_prop_clear_bootfs);
/* asynchronous event notification */
EXPORT_SYMBOL(spa_event_notify);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_shift, INT, ZMOD_RW,
"log2 fraction of arc that can be used by inflight I/Os when "
"verifying pool during import");
ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_metadata, INT, ZMOD_RW,
"Set to traverse metadata on pool import");
ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_data, INT, ZMOD_RW,
"Set to traverse data on pool import");
ZFS_MODULE_PARAM(zfs_spa, spa_, load_print_vdev_tree, INT, ZMOD_RW,
"Print vdev tree to zfs_dbgmsg during pool import");
ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_batch_pct, UINT, ZMOD_RD,
"Percentage of CPUs to run an IO worker thread");
ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_batch_tpq, UINT, ZMOD_RD,
"Number of threads per IO worker taskqueue");
ZFS_MODULE_PARAM(zfs, zfs_, max_missing_tvds, ULONG, ZMOD_RW,
"Allow importing pool with up to this number of missing top-level "
"vdevs (in read-only mode)");
ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, zthr_pause, INT, ZMOD_RW,
"Set the livelist condense zthr to pause");
ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, sync_pause, INT, ZMOD_RW,
"Set the livelist condense synctask to pause");
ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, sync_cancel, INT, ZMOD_RW,
"Whether livelist condensing was canceled in the synctask");
ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, zthr_cancel, INT, ZMOD_RW,
"Whether livelist condensing was canceled in the zthr function");
ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, new_alloc, INT, ZMOD_RW,
"Whether extra ALLOC blkptrs were added to a livelist entry while it "
"was being condensed");
/* END CSTYLED */
diff --git a/module/zfs/spa_checkpoint.c b/module/zfs/spa_checkpoint.c
index ddcdb6801053..68c3ae2e0c31 100644
--- a/module/zfs/spa_checkpoint.c
+++ b/module/zfs/spa_checkpoint.c
@@ -1,637 +1,637 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2017 by Delphix. All rights reserved.
*/
/*
* Storage Pool Checkpoint
*
* A storage pool checkpoint can be thought of as a pool-wide snapshot or
* a stable version of extreme rewind that guarantees no blocks from the
* checkpointed state will have been overwritten. It remembers the entire
* state of the storage pool (e.g. snapshots, dataset names, etc..) from the
* point that it was taken and the user can rewind back to that point even if
* they applied destructive operations on their datasets or even enabled new
* zpool on-disk features. If a pool has a checkpoint that is no longer
* needed, the user can discard it.
*
* == On disk data structures used ==
*
* - The pool has a new feature flag and a new entry in the MOS. The feature
* flag is set to active when we create the checkpoint and remains active
* until the checkpoint is fully discarded. The entry in the MOS config
* (DMU_POOL_ZPOOL_CHECKPOINT) is populated with the uberblock that
* references the state of the pool when we take the checkpoint. The entry
* remains populated until we start discarding the checkpoint or we rewind
* back to it.
*
* - Each vdev contains a vdev-wide space map while the pool has a checkpoint,
* which persists until the checkpoint is fully discarded. The space map
* contains entries that have been freed in the current state of the pool
* but we want to keep around in case we decide to rewind to the checkpoint.
* [see vdev_checkpoint_sm]
*
* - Each metaslab's ms_sm space map behaves the same as without the
* checkpoint, with the only exception being the scenario when we free
* blocks that belong to the checkpoint. In this case, these blocks remain
* ALLOCATED in the metaslab's space map and they are added as FREE in the
* vdev's checkpoint space map.
*
* - Each uberblock has a field (ub_checkpoint_txg) which holds the txg that
* the uberblock was checkpointed. For normal uberblocks this field is 0.
*
* == Overview of operations ==
*
* - To create a checkpoint, we first wait for the current TXG to be synced,
* so we can use the most recently synced uberblock (spa_ubsync) as the
* checkpointed uberblock. Then we use an early synctask to place that
* uberblock in MOS config, increment the feature flag for the checkpoint
* (marking it active), and setting spa_checkpoint_txg (see its use below)
* to the TXG of the checkpointed uberblock. We use an early synctask for
* the aforementioned operations to ensure that no blocks were dirtied
* between the current TXG and the TXG of the checkpointed uberblock
* (e.g the previous txg).
*
* - When a checkpoint exists, we need to ensure that the blocks that
* belong to the checkpoint are freed but never reused. This means that
* these blocks should never end up in the ms_allocatable or the ms_freeing
* trees of a metaslab. Therefore, whenever there is a checkpoint the new
* ms_checkpointing tree is used in addition to the aforementioned ones.
*
* Whenever a block is freed and we find out that it is referenced by the
* checkpoint (we find out by comparing its birth to spa_checkpoint_txg),
* we place it in the ms_checkpointing tree instead of the ms_freeingtree.
* This way, we divide the blocks that are being freed into checkpointed
* and not-checkpointed blocks.
*
* In order to persist these frees, we write the extents from the
* ms_freeingtree to the ms_sm as usual, and the extents from the
* ms_checkpointing tree to the vdev_checkpoint_sm. This way, these
* checkpointed extents will remain allocated in the metaslab's ms_sm space
* map, and therefore won't be reused [see metaslab_sync()]. In addition,
* when we discard the checkpoint, we can find the entries that have
* actually been freed in vdev_checkpoint_sm.
* [see spa_checkpoint_discard_thread_sync()]
*
* - To discard the checkpoint we use an early synctask to delete the
* checkpointed uberblock from the MOS config, set spa_checkpoint_txg to 0,
* and wakeup the discarding zthr thread (an open-context async thread).
* We use an early synctask to ensure that the operation happens before any
* new data end up in the checkpoint's data structures.
*
* Once the synctask is done and the discarding zthr is awake, we discard
* the checkpointed data over multiple TXGs by having the zthr prefetching
* entries from vdev_checkpoint_sm and then starting a synctask that places
* them as free blocks into their respective ms_allocatable and ms_sm
* structures.
* [see spa_checkpoint_discard_thread()]
*
* When there are no entries left in the vdev_checkpoint_sm of all
* top-level vdevs, a final synctask runs that decrements the feature flag.
*
* - To rewind to the checkpoint, we first use the current uberblock and
* open the MOS so we can access the checkpointed uberblock from the MOS
* config. After we retrieve the checkpointed uberblock, we use it as the
* current uberblock for the pool by writing it to disk with an updated
* TXG, opening its version of the MOS, and moving on as usual from there.
* [see spa_ld_checkpoint_rewind()]
*
* An important note on rewinding to the checkpoint has to do with how we
* handle ZIL blocks. In the scenario of a rewind, we clear out any ZIL
* blocks that have not been claimed by the time we took the checkpoint
* as they should no longer be valid.
* [see comment in zil_claim()]
*
* == Miscellaneous information ==
*
* - In the hypothetical event that we take a checkpoint, remove a vdev,
* and attempt to rewind, the rewind would fail as the checkpointed
* uberblock would reference data in the removed device. For this reason
* and others of similar nature, we disallow the following operations that
* can change the config:
* vdev removal and attach/detach, mirror splitting, and pool reguid.
*
* - As most of the checkpoint logic is implemented in the SPA and doesn't
* distinguish datasets when it comes to space accounting, having a
* checkpoint can potentially break the boundaries set by dataset
* reservations.
*/
#include <sys/dmu_tx.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_synctask.h>
#include <sys/metaslab_impl.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/spa_checkpoint.h>
#include <sys/vdev_impl.h>
#include <sys/zap.h>
#include <sys/zfeature.h>
/*
* The following parameter limits the amount of memory to be used for the
* prefetching of the checkpoint space map done on each vdev while
* discarding the checkpoint.
*
* The reason it exists is because top-level vdevs with long checkpoint
* space maps can potentially take up a lot of memory depending on the
* amount of checkpointed data that has been freed within them while
* the pool had a checkpoint.
*/
-unsigned long zfs_spa_discard_memory_limit = 16 * 1024 * 1024;
+static unsigned long zfs_spa_discard_memory_limit = 16 * 1024 * 1024;
int
spa_checkpoint_get_stats(spa_t *spa, pool_checkpoint_stat_t *pcs)
{
if (!spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT))
return (SET_ERROR(ZFS_ERR_NO_CHECKPOINT));
bzero(pcs, sizeof (pool_checkpoint_stat_t));
int error = zap_contains(spa_meta_objset(spa),
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ZPOOL_CHECKPOINT);
ASSERT(error == 0 || error == ENOENT);
if (error == ENOENT)
pcs->pcs_state = CS_CHECKPOINT_DISCARDING;
else
pcs->pcs_state = CS_CHECKPOINT_EXISTS;
pcs->pcs_space = spa->spa_checkpoint_info.sci_dspace;
pcs->pcs_start_time = spa->spa_checkpoint_info.sci_timestamp;
return (0);
}
static void
spa_checkpoint_discard_complete_sync(void *arg, dmu_tx_t *tx)
{
spa_t *spa = arg;
spa->spa_checkpoint_info.sci_timestamp = 0;
spa_feature_decr(spa, SPA_FEATURE_POOL_CHECKPOINT, tx);
spa_notify_waiters(spa);
spa_history_log_internal(spa, "spa discard checkpoint", tx,
"finished discarding checkpointed state from the pool");
}
typedef struct spa_checkpoint_discard_sync_callback_arg {
vdev_t *sdc_vd;
uint64_t sdc_txg;
uint64_t sdc_entry_limit;
} spa_checkpoint_discard_sync_callback_arg_t;
static int
spa_checkpoint_discard_sync_callback(space_map_entry_t *sme, void *arg)
{
spa_checkpoint_discard_sync_callback_arg_t *sdc = arg;
vdev_t *vd = sdc->sdc_vd;
metaslab_t *ms = vd->vdev_ms[sme->sme_offset >> vd->vdev_ms_shift];
uint64_t end = sme->sme_offset + sme->sme_run;
if (sdc->sdc_entry_limit == 0)
return (SET_ERROR(EINTR));
/*
* Since the space map is not condensed, we know that
* none of its entries is crossing the boundaries of
* its respective metaslab.
*
* That said, there is no fundamental requirement that
* the checkpoint's space map entries should not cross
* metaslab boundaries. So if needed we could add code
* that handles metaslab-crossing segments in the future.
*/
VERIFY3U(sme->sme_type, ==, SM_FREE);
VERIFY3U(sme->sme_offset, >=, ms->ms_start);
VERIFY3U(end, <=, ms->ms_start + ms->ms_size);
/*
* At this point we should not be processing any
* other frees concurrently, so the lock is technically
* unnecessary. We use the lock anyway though to
* potentially save ourselves from future headaches.
*/
mutex_enter(&ms->ms_lock);
if (range_tree_is_empty(ms->ms_freeing))
vdev_dirty(vd, VDD_METASLAB, ms, sdc->sdc_txg);
range_tree_add(ms->ms_freeing, sme->sme_offset, sme->sme_run);
mutex_exit(&ms->ms_lock);
ASSERT3U(vd->vdev_spa->spa_checkpoint_info.sci_dspace, >=,
sme->sme_run);
ASSERT3U(vd->vdev_stat.vs_checkpoint_space, >=, sme->sme_run);
vd->vdev_spa->spa_checkpoint_info.sci_dspace -= sme->sme_run;
vd->vdev_stat.vs_checkpoint_space -= sme->sme_run;
sdc->sdc_entry_limit--;
return (0);
}
#ifdef ZFS_DEBUG
static void
spa_checkpoint_accounting_verify(spa_t *spa)
{
vdev_t *rvd = spa->spa_root_vdev;
uint64_t ckpoint_sm_space_sum = 0;
uint64_t vs_ckpoint_space_sum = 0;
for (uint64_t c = 0; c < rvd->vdev_children; c++) {
vdev_t *vd = rvd->vdev_child[c];
if (vd->vdev_checkpoint_sm != NULL) {
ckpoint_sm_space_sum +=
-space_map_allocated(vd->vdev_checkpoint_sm);
vs_ckpoint_space_sum +=
vd->vdev_stat.vs_checkpoint_space;
ASSERT3U(ckpoint_sm_space_sum, ==,
vs_ckpoint_space_sum);
} else {
ASSERT0(vd->vdev_stat.vs_checkpoint_space);
}
}
ASSERT3U(spa->spa_checkpoint_info.sci_dspace, ==, ckpoint_sm_space_sum);
}
#endif
static void
spa_checkpoint_discard_thread_sync(void *arg, dmu_tx_t *tx)
{
vdev_t *vd = arg;
int error;
/*
* The space map callback is applied only to non-debug entries.
* Because the number of debug entries is less or equal to the
* number of non-debug entries, we want to ensure that we only
* read what we prefetched from open-context.
*
* Thus, we set the maximum entries that the space map callback
* will be applied to be half the entries that could fit in the
* imposed memory limit.
*
* Note that since this is a conservative estimate we also
* assume the worst case scenario in our computation where each
* entry is two-word.
*/
uint64_t max_entry_limit =
(zfs_spa_discard_memory_limit / (2 * sizeof (uint64_t))) >> 1;
/*
* Iterate from the end of the space map towards the beginning,
* placing its entries on ms_freeing and removing them from the
* space map. The iteration stops if one of the following
* conditions is true:
*
* 1] We reached the beginning of the space map. At this point
* the space map should be completely empty and
* space_map_incremental_destroy should have returned 0.
* The next step would be to free and close the space map
* and remove its entry from its vdev's top zap. This allows
* spa_checkpoint_discard_thread() to move on to the next vdev.
*
* 2] We reached the memory limit (amount of memory used to hold
* space map entries in memory) and space_map_incremental_destroy
* returned EINTR. This means that there are entries remaining
* in the space map that will be cleared in a future invocation
* of this function by spa_checkpoint_discard_thread().
*/
spa_checkpoint_discard_sync_callback_arg_t sdc;
sdc.sdc_vd = vd;
sdc.sdc_txg = tx->tx_txg;
sdc.sdc_entry_limit = max_entry_limit;
uint64_t words_before =
space_map_length(vd->vdev_checkpoint_sm) / sizeof (uint64_t);
error = space_map_incremental_destroy(vd->vdev_checkpoint_sm,
spa_checkpoint_discard_sync_callback, &sdc, tx);
uint64_t words_after =
space_map_length(vd->vdev_checkpoint_sm) / sizeof (uint64_t);
#ifdef ZFS_DEBUG
spa_checkpoint_accounting_verify(vd->vdev_spa);
#endif
zfs_dbgmsg("discarding checkpoint: txg %llu, vdev id %lld, "
"deleted %llu words - %llu words are left",
(u_longlong_t)tx->tx_txg, (longlong_t)vd->vdev_id,
(u_longlong_t)(words_before - words_after),
(u_longlong_t)words_after);
if (error != EINTR) {
if (error != 0) {
zfs_panic_recover("zfs: error %lld was returned "
"while incrementally destroying the checkpoint "
"space map of vdev %u\n",
(longlong_t)error, vd->vdev_id);
}
ASSERT0(words_after);
ASSERT0(space_map_allocated(vd->vdev_checkpoint_sm));
ASSERT0(space_map_length(vd->vdev_checkpoint_sm));
space_map_free(vd->vdev_checkpoint_sm, tx);
space_map_close(vd->vdev_checkpoint_sm);
vd->vdev_checkpoint_sm = NULL;
VERIFY0(zap_remove(spa_meta_objset(vd->vdev_spa),
vd->vdev_top_zap, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM, tx));
}
}
static boolean_t
spa_checkpoint_discard_is_done(spa_t *spa)
{
vdev_t *rvd = spa->spa_root_vdev;
ASSERT(!spa_has_checkpoint(spa));
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT));
for (uint64_t c = 0; c < rvd->vdev_children; c++) {
if (rvd->vdev_child[c]->vdev_checkpoint_sm != NULL)
return (B_FALSE);
ASSERT0(rvd->vdev_child[c]->vdev_stat.vs_checkpoint_space);
}
return (B_TRUE);
}
boolean_t
spa_checkpoint_discard_thread_check(void *arg, zthr_t *zthr)
{
(void) zthr;
spa_t *spa = arg;
if (!spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT))
return (B_FALSE);
if (spa_has_checkpoint(spa))
return (B_FALSE);
return (B_TRUE);
}
void
spa_checkpoint_discard_thread(void *arg, zthr_t *zthr)
{
spa_t *spa = arg;
vdev_t *rvd = spa->spa_root_vdev;
for (uint64_t c = 0; c < rvd->vdev_children; c++) {
vdev_t *vd = rvd->vdev_child[c];
while (vd->vdev_checkpoint_sm != NULL) {
space_map_t *checkpoint_sm = vd->vdev_checkpoint_sm;
int numbufs;
dmu_buf_t **dbp;
if (zthr_iscancelled(zthr))
return;
ASSERT3P(vd->vdev_ops, !=, &vdev_indirect_ops);
uint64_t size = MIN(space_map_length(checkpoint_sm),
zfs_spa_discard_memory_limit);
uint64_t offset =
space_map_length(checkpoint_sm) - size;
/*
* Ensure that the part of the space map that will
* be destroyed by the synctask, is prefetched in
* memory before the synctask runs.
*/
int error = dmu_buf_hold_array_by_bonus(
checkpoint_sm->sm_dbuf, offset, size,
B_TRUE, FTAG, &numbufs, &dbp);
if (error != 0) {
zfs_panic_recover("zfs: error %d was returned "
"while prefetching checkpoint space map "
"entries of vdev %llu\n",
error, vd->vdev_id);
}
VERIFY0(dsl_sync_task(spa->spa_name, NULL,
spa_checkpoint_discard_thread_sync, vd,
0, ZFS_SPACE_CHECK_NONE));
dmu_buf_rele_array(dbp, numbufs, FTAG);
}
}
VERIFY(spa_checkpoint_discard_is_done(spa));
VERIFY0(spa->spa_checkpoint_info.sci_dspace);
VERIFY0(dsl_sync_task(spa->spa_name, NULL,
spa_checkpoint_discard_complete_sync, spa,
0, ZFS_SPACE_CHECK_NONE));
}
static int
spa_checkpoint_check(void *arg, dmu_tx_t *tx)
{
(void) arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
if (!spa_feature_is_enabled(spa, SPA_FEATURE_POOL_CHECKPOINT))
return (SET_ERROR(ENOTSUP));
if (!spa_top_vdevs_spacemap_addressable(spa))
return (SET_ERROR(ZFS_ERR_VDEV_TOO_BIG));
if (spa->spa_removing_phys.sr_state == DSS_SCANNING)
return (SET_ERROR(ZFS_ERR_DEVRM_IN_PROGRESS));
if (spa->spa_checkpoint_txg != 0)
return (SET_ERROR(ZFS_ERR_CHECKPOINT_EXISTS));
if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT))
return (SET_ERROR(ZFS_ERR_DISCARDING_CHECKPOINT));
return (0);
}
static void
spa_checkpoint_sync(void *arg, dmu_tx_t *tx)
{
(void) arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
spa_t *spa = dp->dp_spa;
uberblock_t checkpoint = spa->spa_ubsync;
/*
* At this point, there should not be a checkpoint in the MOS.
*/
ASSERT3U(zap_contains(spa_meta_objset(spa), DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_ZPOOL_CHECKPOINT), ==, ENOENT);
ASSERT0(spa->spa_checkpoint_info.sci_timestamp);
ASSERT0(spa->spa_checkpoint_info.sci_dspace);
/*
* Since the checkpointed uberblock is the one that just got synced
* (we use spa_ubsync), its txg must be equal to the txg number of
* the txg we are syncing, minus 1.
*/
ASSERT3U(checkpoint.ub_txg, ==, spa->spa_syncing_txg - 1);
/*
* Once the checkpoint is in place, we need to ensure that none of
* its blocks will be marked for reuse after it has been freed.
* When there is a checkpoint and a block is freed, we compare its
* birth txg to the txg of the checkpointed uberblock to see if the
* block is part of the checkpoint or not. Therefore, we have to set
* spa_checkpoint_txg before any frees happen in this txg (which is
* why this is done as an early_synctask as explained in the comment
* in spa_checkpoint()).
*/
spa->spa_checkpoint_txg = checkpoint.ub_txg;
spa->spa_checkpoint_info.sci_timestamp = checkpoint.ub_timestamp;
checkpoint.ub_checkpoint_txg = checkpoint.ub_txg;
VERIFY0(zap_add(spa->spa_dsl_pool->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ZPOOL_CHECKPOINT,
sizeof (uint64_t), sizeof (uberblock_t) / sizeof (uint64_t),
&checkpoint, tx));
/*
* Increment the feature refcount and thus activate the feature.
* Note that the feature will be deactivated when we've
* completely discarded all checkpointed state (both vdev
* space maps and uberblock).
*/
spa_feature_incr(spa, SPA_FEATURE_POOL_CHECKPOINT, tx);
spa_history_log_internal(spa, "spa checkpoint", tx,
"checkpointed uberblock txg=%llu", (u_longlong_t)checkpoint.ub_txg);
}
/*
* Create a checkpoint for the pool.
*/
int
spa_checkpoint(const char *pool)
{
int error;
spa_t *spa;
error = spa_open(pool, &spa, FTAG);
if (error != 0)
return (error);
mutex_enter(&spa->spa_vdev_top_lock);
/*
* Wait for current syncing txg to finish so the latest synced
* uberblock (spa_ubsync) has all the changes that we expect
* to see if we were to revert later to the checkpoint. In other
* words we want the checkpointed uberblock to include/reference
* all the changes that were pending at the time that we issued
* the checkpoint command.
*/
txg_wait_synced(spa_get_dsl(spa), 0);
/*
* As the checkpointed uberblock references blocks from the previous
* txg (spa_ubsync) we want to ensure that are not freeing any of
* these blocks in the same txg that the following synctask will
* run. Thus, we run it as an early synctask, so the dirty changes
* that are synced to disk afterwards during zios and other synctasks
* do not reuse checkpointed blocks.
*/
error = dsl_early_sync_task(pool, spa_checkpoint_check,
spa_checkpoint_sync, NULL, 0, ZFS_SPACE_CHECK_NORMAL);
mutex_exit(&spa->spa_vdev_top_lock);
spa_close(spa, FTAG);
return (error);
}
static int
spa_checkpoint_discard_check(void *arg, dmu_tx_t *tx)
{
(void) arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
if (!spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT))
return (SET_ERROR(ZFS_ERR_NO_CHECKPOINT));
if (spa->spa_checkpoint_txg == 0)
return (SET_ERROR(ZFS_ERR_DISCARDING_CHECKPOINT));
VERIFY0(zap_contains(spa_meta_objset(spa),
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ZPOOL_CHECKPOINT));
return (0);
}
static void
spa_checkpoint_discard_sync(void *arg, dmu_tx_t *tx)
{
(void) arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
VERIFY0(zap_remove(spa_meta_objset(spa), DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_ZPOOL_CHECKPOINT, tx));
spa->spa_checkpoint_txg = 0;
zthr_wakeup(spa->spa_checkpoint_discard_zthr);
spa_history_log_internal(spa, "spa discard checkpoint", tx,
"started discarding checkpointed state from the pool");
}
/*
* Discard the checkpoint from a pool.
*/
int
spa_checkpoint_discard(const char *pool)
{
/*
* Similarly to spa_checkpoint(), we want our synctask to run
* before any pending dirty data are written to disk so they
* won't end up in the checkpoint's data structures (e.g.
* ms_checkpointing and vdev_checkpoint_sm) and re-create any
* space maps that the discarding open-context thread has
* deleted.
* [see spa_discard_checkpoint_sync and spa_discard_checkpoint_thread]
*/
return (dsl_early_sync_task(pool, spa_checkpoint_discard_check,
spa_checkpoint_discard_sync, NULL, 0,
ZFS_SPACE_CHECK_DISCARD_CHECKPOINT));
}
EXPORT_SYMBOL(spa_checkpoint_get_stats);
EXPORT_SYMBOL(spa_checkpoint_discard_thread);
EXPORT_SYMBOL(spa_checkpoint_discard_thread_check);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_spa, zfs_spa_, discard_memory_limit, ULONG, ZMOD_RW,
"Limit for memory used in prefetching the checkpoint space map done "
"on each vdev while discarding the checkpoint");
/* END CSTYLED */
diff --git a/module/zfs/spa_config.c b/module/zfs/spa_config.c
index ad82932ce567..b9c6ac042f1f 100644
--- a/module/zfs/spa_config.c
+++ b/module/zfs/spa_config.c
@@ -1,623 +1,625 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright 2017 Joyent, Inc.
* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
*/
#include <sys/spa.h>
#include <sys/file.h>
#include <sys/fm/fs/zfs.h>
#include <sys/spa_impl.h>
#include <sys/nvpair.h>
#include <sys/fs/zfs.h>
#include <sys/vdev_impl.h>
#include <sys/zfs_ioctl.h>
#include <sys/systeminfo.h>
#include <sys/sunddi.h>
#include <sys/zfeature.h>
#include <sys/zfs_file.h>
#include <sys/zfs_context.h>
#ifdef _KERNEL
#include <sys/zone.h>
#endif
/*
* Pool configuration repository.
*
* Pool configuration is stored as a packed nvlist on the filesystem. By
* default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot
* (when the ZFS module is loaded). Pools can also have the 'cachefile'
* property set that allows them to be stored in an alternate location until
* the control of external software.
*
* For each cache file, we have a single nvlist which holds all the
* configuration information. When the module loads, we read this information
* from /etc/zfs/zpool.cache and populate the SPA namespace. This namespace is
* maintained independently in spa.c. Whenever the namespace is modified, or
* the configuration of a pool is changed, we call spa_write_cachefile(), which
* walks through all the active pools and writes the configuration to disk.
*/
static uint64_t spa_config_generation = 1;
/*
* This can be overridden in userland to preserve an alternate namespace for
* userland pools when doing testing.
*/
char *spa_config_path = ZPOOL_CACHE;
-int zfs_autoimport_disable = 1;
+#ifdef _KERNEL
+static int zfs_autoimport_disable = B_TRUE;
+#endif
/*
* Called when the module is first loaded, this routine loads the configuration
* file into the SPA namespace. It does not actually open or load the pools; it
* only populates the namespace.
*/
void
spa_config_load(void)
{
void *buf = NULL;
nvlist_t *nvlist, *child;
nvpair_t *nvpair;
char *pathname;
zfs_file_t *fp;
zfs_file_attr_t zfa;
uint64_t fsize;
int err;
#ifdef _KERNEL
if (zfs_autoimport_disable)
return;
#endif
/*
* Open the configuration file.
*/
pathname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
(void) snprintf(pathname, MAXPATHLEN, "%s", spa_config_path);
err = zfs_file_open(pathname, O_RDONLY, 0, &fp);
#ifdef __FreeBSD__
if (err)
err = zfs_file_open(ZPOOL_CACHE_BOOT, O_RDONLY, 0, &fp);
#endif
kmem_free(pathname, MAXPATHLEN);
if (err)
return;
if (zfs_file_getattr(fp, &zfa))
goto out;
fsize = zfa.zfa_size;
buf = kmem_alloc(fsize, KM_SLEEP);
/*
* Read the nvlist from the file.
*/
if (zfs_file_read(fp, buf, fsize, NULL) < 0)
goto out;
/*
* Unpack the nvlist.
*/
if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0)
goto out;
/*
* Iterate over all elements in the nvlist, creating a new spa_t for
* each one with the specified configuration.
*/
mutex_enter(&spa_namespace_lock);
nvpair = NULL;
while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) {
if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
continue;
child = fnvpair_value_nvlist(nvpair);
if (spa_lookup(nvpair_name(nvpair)) != NULL)
continue;
(void) spa_add(nvpair_name(nvpair), child, NULL);
}
mutex_exit(&spa_namespace_lock);
nvlist_free(nvlist);
out:
if (buf != NULL)
kmem_free(buf, fsize);
zfs_file_close(fp);
}
static int
spa_config_remove(spa_config_dirent_t *dp)
{
int error = 0;
/*
* Remove the cache file. If zfs_file_unlink() in not supported by the
* platform fallback to truncating the file which is functionally
* equivalent.
*/
error = zfs_file_unlink(dp->scd_path);
if (error == EOPNOTSUPP) {
int flags = O_RDWR | O_TRUNC;
zfs_file_t *fp;
error = zfs_file_open(dp->scd_path, flags, 0644, &fp);
if (error == 0) {
(void) zfs_file_fsync(fp, O_SYNC);
(void) zfs_file_close(fp);
}
}
return (error);
}
static int
spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
{
size_t buflen;
char *buf;
int oflags = O_RDWR | O_TRUNC | O_CREAT | O_LARGEFILE;
char *temp;
int err;
zfs_file_t *fp;
/*
* If the nvlist is empty (NULL), then remove the old cachefile.
*/
if (nvl == NULL) {
err = spa_config_remove(dp);
if (err == ENOENT)
err = 0;
return (err);
}
/*
* Pack the configuration into a buffer.
*/
buf = fnvlist_pack(nvl, &buflen);
temp = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
/*
* Write the configuration to disk. Due to the complexity involved
* in performing a rename and remove from within the kernel the file
* is instead truncated and overwritten in place. This way we always
* have a consistent view of the data or a zero length file.
*/
err = zfs_file_open(dp->scd_path, oflags, 0644, &fp);
if (err == 0) {
err = zfs_file_write(fp, buf, buflen, NULL);
if (err == 0)
err = zfs_file_fsync(fp, O_SYNC);
zfs_file_close(fp);
if (err)
(void) spa_config_remove(dp);
}
fnvlist_pack_free(buf, buflen);
kmem_free(temp, MAXPATHLEN);
return (err);
}
/*
* Synchronize pool configuration to disk. This must be called with the
* namespace lock held. Synchronizing the pool cache is typically done after
* the configuration has been synced to the MOS. This exposes a window where
* the MOS config will have been updated but the cache file has not. If
* the system were to crash at that instant then the cached config may not
* contain the correct information to open the pool and an explicit import
* would be required.
*/
void
spa_write_cachefile(spa_t *target, boolean_t removing, boolean_t postsysevent)
{
spa_config_dirent_t *dp, *tdp;
nvlist_t *nvl;
char *pool_name;
boolean_t ccw_failure;
int error = 0;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (!(spa_mode_global & SPA_MODE_WRITE))
return;
/*
* Iterate over all cachefiles for the pool, past or present. When the
* cachefile is changed, the new one is pushed onto this list, allowing
* us to update previous cachefiles that no longer contain this pool.
*/
ccw_failure = B_FALSE;
for (dp = list_head(&target->spa_config_list); dp != NULL;
dp = list_next(&target->spa_config_list, dp)) {
spa_t *spa = NULL;
if (dp->scd_path == NULL)
continue;
/*
* Iterate over all pools, adding any matching pools to 'nvl'.
*/
nvl = NULL;
while ((spa = spa_next(spa)) != NULL) {
/*
* Skip over our own pool if we're about to remove
* ourselves from the spa namespace or any pool that
* is readonly. Since we cannot guarantee that a
* readonly pool would successfully import upon reboot,
* we don't allow them to be written to the cache file.
*/
if ((spa == target && removing) ||
!spa_writeable(spa))
continue;
mutex_enter(&spa->spa_props_lock);
tdp = list_head(&spa->spa_config_list);
if (spa->spa_config == NULL ||
tdp == NULL ||
tdp->scd_path == NULL ||
strcmp(tdp->scd_path, dp->scd_path) != 0) {
mutex_exit(&spa->spa_props_lock);
continue;
}
if (nvl == NULL)
nvl = fnvlist_alloc();
if (spa->spa_import_flags & ZFS_IMPORT_TEMP_NAME)
pool_name = fnvlist_lookup_string(
spa->spa_config, ZPOOL_CONFIG_POOL_NAME);
else
pool_name = spa_name(spa);
fnvlist_add_nvlist(nvl, pool_name, spa->spa_config);
mutex_exit(&spa->spa_props_lock);
}
error = spa_config_write(dp, nvl);
if (error != 0)
ccw_failure = B_TRUE;
nvlist_free(nvl);
}
if (ccw_failure) {
/*
* Keep trying so that configuration data is
* written if/when any temporary filesystem
* resource issues are resolved.
*/
if (target->spa_ccw_fail_time == 0) {
(void) zfs_ereport_post(
FM_EREPORT_ZFS_CONFIG_CACHE_WRITE,
target, NULL, NULL, NULL, 0);
}
target->spa_ccw_fail_time = gethrtime();
spa_async_request(target, SPA_ASYNC_CONFIG_UPDATE);
} else {
/*
* Do not rate limit future attempts to update
* the config cache.
*/
target->spa_ccw_fail_time = 0;
}
/*
* Remove any config entries older than the current one.
*/
dp = list_head(&target->spa_config_list);
while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
list_remove(&target->spa_config_list, tdp);
if (tdp->scd_path != NULL)
spa_strfree(tdp->scd_path);
kmem_free(tdp, sizeof (spa_config_dirent_t));
}
spa_config_generation++;
if (postsysevent)
spa_event_notify(target, NULL, NULL, ESC_ZFS_CONFIG_SYNC);
}
/*
* Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache,
* and we don't want to allow the local zone to see all the pools anyway.
* So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration
* information for all pool visible within the zone.
*/
nvlist_t *
spa_all_configs(uint64_t *generation)
{
nvlist_t *pools;
spa_t *spa = NULL;
if (*generation == spa_config_generation)
return (NULL);
pools = fnvlist_alloc();
mutex_enter(&spa_namespace_lock);
while ((spa = spa_next(spa)) != NULL) {
if (INGLOBALZONE(curproc) ||
zone_dataset_visible(spa_name(spa), NULL)) {
mutex_enter(&spa->spa_props_lock);
fnvlist_add_nvlist(pools, spa_name(spa),
spa->spa_config);
mutex_exit(&spa->spa_props_lock);
}
}
*generation = spa_config_generation;
mutex_exit(&spa_namespace_lock);
return (pools);
}
void
spa_config_set(spa_t *spa, nvlist_t *config)
{
mutex_enter(&spa->spa_props_lock);
if (spa->spa_config != NULL && spa->spa_config != config)
nvlist_free(spa->spa_config);
spa->spa_config = config;
mutex_exit(&spa->spa_props_lock);
}
/*
* Generate the pool's configuration based on the current in-core state.
*
* We infer whether to generate a complete config or just one top-level config
* based on whether vd is the root vdev.
*/
nvlist_t *
spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
{
nvlist_t *config, *nvroot;
vdev_t *rvd = spa->spa_root_vdev;
unsigned long hostid = 0;
boolean_t locked = B_FALSE;
uint64_t split_guid;
char *pool_name;
if (vd == NULL) {
vd = rvd;
locked = B_TRUE;
spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
}
ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
(SCL_CONFIG | SCL_STATE));
/*
* If txg is -1, report the current value of spa->spa_config_txg.
*/
if (txg == -1ULL)
txg = spa->spa_config_txg;
/*
* Originally, users had to handle spa namespace collisions by either
* exporting the already imported pool or by specifying a new name for
* the pool with a conflicting name. In the case of root pools from
* virtual guests, neither approach to collision resolution is
* reasonable. This is addressed by extending the new name syntax with
* an option to specify that the new name is temporary. When specified,
* ZFS_IMPORT_TEMP_NAME will be set in spa->spa_import_flags to tell us
* to use the previous name, which we do below.
*/
if (spa->spa_import_flags & ZFS_IMPORT_TEMP_NAME) {
VERIFY0(nvlist_lookup_string(spa->spa_config,
ZPOOL_CONFIG_POOL_NAME, &pool_name));
} else
pool_name = spa_name(spa);
config = fnvlist_alloc();
fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa_version(spa));
fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, pool_name);
fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, spa_state(spa));
fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, txg);
fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, spa_guid(spa));
fnvlist_add_uint64(config, ZPOOL_CONFIG_ERRATA, spa->spa_errata);
if (spa->spa_comment != NULL)
fnvlist_add_string(config, ZPOOL_CONFIG_COMMENT,
spa->spa_comment);
if (spa->spa_compatibility != NULL)
fnvlist_add_string(config, ZPOOL_CONFIG_COMPATIBILITY,
spa->spa_compatibility);
hostid = spa_get_hostid(spa);
if (hostid != 0)
fnvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, hostid);
fnvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, utsname()->nodename);
int config_gen_flags = 0;
if (vd != rvd) {
fnvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
vd->vdev_top->vdev_guid);
fnvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
vd->vdev_guid);
if (vd->vdev_isspare)
fnvlist_add_uint64(config,
ZPOOL_CONFIG_IS_SPARE, 1ULL);
if (vd->vdev_islog)
fnvlist_add_uint64(config,
ZPOOL_CONFIG_IS_LOG, 1ULL);
vd = vd->vdev_top; /* label contains top config */
} else {
/*
* Only add the (potentially large) split information
* in the mos config, and not in the vdev labels
*/
if (spa->spa_config_splitting != NULL)
fnvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT,
spa->spa_config_splitting);
fnvlist_add_boolean(config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS);
config_gen_flags |= VDEV_CONFIG_MOS;
}
/*
* Add the top-level config. We even add this on pools which
* don't support holes in the namespace.
*/
vdev_top_config_generate(spa, config);
/*
* If we're splitting, record the original pool's guid.
*/
if (spa->spa_config_splitting != NULL &&
nvlist_lookup_uint64(spa->spa_config_splitting,
ZPOOL_CONFIG_SPLIT_GUID, &split_guid) == 0) {
fnvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID, split_guid);
}
nvroot = vdev_config_generate(spa, vd, getstats, config_gen_flags);
fnvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot);
nvlist_free(nvroot);
/*
* Store what's necessary for reading the MOS in the label.
*/
fnvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
spa->spa_label_features);
if (getstats && spa_load_state(spa) == SPA_LOAD_NONE) {
ddt_histogram_t *ddh;
ddt_stat_t *dds;
ddt_object_t *ddo;
ddh = kmem_zalloc(sizeof (ddt_histogram_t), KM_SLEEP);
ddt_get_dedup_histogram(spa, ddh);
fnvlist_add_uint64_array(config,
ZPOOL_CONFIG_DDT_HISTOGRAM,
(uint64_t *)ddh, sizeof (*ddh) / sizeof (uint64_t));
kmem_free(ddh, sizeof (ddt_histogram_t));
ddo = kmem_zalloc(sizeof (ddt_object_t), KM_SLEEP);
ddt_get_dedup_object_stats(spa, ddo);
fnvlist_add_uint64_array(config,
ZPOOL_CONFIG_DDT_OBJ_STATS,
(uint64_t *)ddo, sizeof (*ddo) / sizeof (uint64_t));
kmem_free(ddo, sizeof (ddt_object_t));
dds = kmem_zalloc(sizeof (ddt_stat_t), KM_SLEEP);
ddt_get_dedup_stats(spa, dds);
fnvlist_add_uint64_array(config,
ZPOOL_CONFIG_DDT_STATS,
(uint64_t *)dds, sizeof (*dds) / sizeof (uint64_t));
kmem_free(dds, sizeof (ddt_stat_t));
}
if (locked)
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (config);
}
/*
* Update all disk labels, generate a fresh config based on the current
* in-core state, and sync the global config cache (do not sync the config
* cache if this is a booting rootpool).
*/
void
spa_config_update(spa_t *spa, int what)
{
vdev_t *rvd = spa->spa_root_vdev;
uint64_t txg;
int c;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
txg = spa_last_synced_txg(spa) + 1;
if (what == SPA_CONFIG_UPDATE_POOL) {
vdev_config_dirty(rvd);
} else {
/*
* If we have top-level vdevs that were added but have
* not yet been prepared for allocation, do that now.
* (It's safe now because the config cache is up to date,
* so it will be able to translate the new DVAs.)
* See comments in spa_vdev_add() for full details.
*/
for (c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
/*
* Explicitly skip vdevs that are indirect or
* log vdevs that are being removed. The reason
* is that both of those can have vdev_ms_array
* set to 0 and we wouldn't want to change their
* metaslab size nor call vdev_expand() on them.
*/
if (!vdev_is_concrete(tvd) ||
(tvd->vdev_islog && tvd->vdev_removing))
continue;
if (tvd->vdev_ms_array == 0)
vdev_metaslab_set_size(tvd);
vdev_expand(tvd, txg);
}
}
spa_config_exit(spa, SCL_ALL, FTAG);
/*
* Wait for the mosconfig to be regenerated and synced.
*/
txg_wait_synced(spa->spa_dsl_pool, txg);
/*
* Update the global config cache to reflect the new mosconfig.
*/
if (!spa->spa_is_root) {
spa_write_cachefile(spa, B_FALSE,
what != SPA_CONFIG_UPDATE_POOL);
}
if (what == SPA_CONFIG_UPDATE_POOL)
spa_config_update(spa, SPA_CONFIG_UPDATE_VDEVS);
}
EXPORT_SYMBOL(spa_config_load);
EXPORT_SYMBOL(spa_all_configs);
EXPORT_SYMBOL(spa_config_set);
EXPORT_SYMBOL(spa_config_generate);
EXPORT_SYMBOL(spa_config_update);
/* BEGIN CSTYLED */
#ifdef __linux__
/* string sysctls require a char array on FreeBSD */
ZFS_MODULE_PARAM(zfs_spa, spa_, config_path, STRING, ZMOD_RD,
"SPA config file (/etc/zfs/zpool.cache)");
#endif
ZFS_MODULE_PARAM(zfs, zfs_, autoimport_disable, INT, ZMOD_RW,
"Disable pool import at module load");
/* END CSTYLED */
diff --git a/module/zfs/spa_log_spacemap.c b/module/zfs/spa_log_spacemap.c
index 6fd302b8df34..341917a6d402 100644
--- a/module/zfs/spa_log_spacemap.c
+++ b/module/zfs/spa_log_spacemap.c
@@ -1,1322 +1,1322 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2018, 2019 by Delphix. All rights reserved.
*/
#include <sys/dmu_objset.h>
#include <sys/metaslab.h>
#include <sys/metaslab_impl.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/spa_log_spacemap.h>
#include <sys/vdev_impl.h>
#include <sys/zap.h>
/*
* Log Space Maps
*
* Log space maps are an optimization in ZFS metadata allocations for pools
* whose workloads are primarily random-writes. Random-write workloads are also
* typically random-free, meaning that they are freeing from locations scattered
* throughout the pool. This means that each TXG we will have to append some
* FREE records to almost every metaslab. With log space maps, we hold their
* changes in memory and log them altogether in one pool-wide space map on-disk
* for persistence. As more blocks are accumulated in the log space maps and
* more unflushed changes are accounted in memory, we flush a selected group
* of metaslabs every TXG to relieve memory pressure and potential overheads
* when loading the pool. Flushing a metaslab to disk relieves memory as we
* flush any unflushed changes from memory to disk (i.e. the metaslab's space
* map) and saves import time by making old log space maps obsolete and
* eventually destroying them. [A log space map is said to be obsolete when all
* its entries have made it to their corresponding metaslab space maps].
*
* == On disk data structures used ==
*
* - The pool has a new feature flag and a new entry in the MOS. The feature
* is activated when we create the first log space map and remains active
* for the lifetime of the pool. The new entry in the MOS Directory [refer
* to DMU_POOL_LOG_SPACEMAP_ZAP] is populated with a ZAP whose key-value
* pairs are of the form <key: txg, value: log space map object for that txg>.
* This entry is our on-disk reference of the log space maps that exist in
* the pool for each TXG and it is used during import to load all the
* metaslab unflushed changes in memory. To see how this structure is first
* created and later populated refer to spa_generate_syncing_log_sm(). To see
* how it is used during import time refer to spa_ld_log_sm_metadata().
*
* - Each vdev has a new entry in its vdev_top_zap (see field
* VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS) which holds the msp_unflushed_txg of
* each metaslab in this vdev. This field is the on-disk counterpart of the
* in-memory field ms_unflushed_txg which tells us from which TXG and onwards
* the metaslab haven't had its changes flushed. During import, we use this
* to ignore any entries in the space map log that are for this metaslab but
* from a TXG before msp_unflushed_txg. At that point, we also populate its
* in-memory counterpart and from there both fields are updated every time
* we flush that metaslab.
*
* - A space map is created every TXG and, during that TXG, it is used to log
* all incoming changes (the log space map). When created, the log space map
* is referenced in memory by spa_syncing_log_sm and its object ID is inserted
* to the space map ZAP mentioned above. The log space map is closed at the
* end of the TXG and will be destroyed when it becomes fully obsolete. We
* know when a log space map has become obsolete by looking at the oldest
* (and smallest) ms_unflushed_txg in the pool. If the value of that is bigger
* than the log space map's TXG, then it means that there is no metaslab who
* doesn't have the changes from that log and we can therefore destroy it.
* [see spa_cleanup_old_sm_logs()].
*
* == Important in-memory structures ==
*
* - The per-spa field spa_metaslabs_by_flushed sorts all the metaslabs in
* the pool by their ms_unflushed_txg field. It is primarily used for three
* reasons. First of all, it is used during flushing where we try to flush
* metaslabs in-order from the oldest-flushed to the most recently flushed
* every TXG. Secondly, it helps us to lookup the ms_unflushed_txg of the
* oldest flushed metaslab to distinguish which log space maps have become
* obsolete and which ones are still relevant. Finally it tells us which
* metaslabs have unflushed changes in a pool where this feature was just
* enabled, as we don't immediately add all of the pool's metaslabs but we
* add them over time as they go through metaslab_sync(). The reason that
* we do that is to ease these pools into the behavior of the flushing
* algorithm (described later on).
*
* - The per-spa field spa_sm_logs_by_txg can be thought as the in-memory
* counterpart of the space map ZAP mentioned above. It's an AVL tree whose
* nodes represent the log space maps in the pool. This in-memory
* representation of log space maps in the pool sorts the log space maps by
* the TXG that they were created (which is also the TXG of their unflushed
* changes). It also contains the following extra information for each
* space map:
* [1] The number of metaslabs that were last flushed on that TXG. This is
* important because if that counter is zero and this is the oldest
* log then it means that it is also obsolete.
* [2] The number of blocks of that space map. This field is used by the
* block heuristic of our flushing algorithm (described later on).
* It represents how many blocks of metadata changes ZFS had to write
* to disk for that TXG.
*
* - The per-spa field spa_log_summary is a list of entries that summarizes
* the metaslab and block counts of all the nodes of the spa_sm_logs_by_txg
* AVL tree mentioned above. The reason this exists is that our flushing
* algorithm (described later) tries to estimate how many metaslabs to flush
* in each TXG by iterating over all the log space maps and looking at their
* block counts. Summarizing that information means that don't have to
* iterate through each space map, minimizing the runtime overhead of the
* flushing algorithm which would be induced in syncing context. In terms of
* implementation the log summary is used as a queue:
* * we modify or pop entries from its head when we flush metaslabs
* * we modify or append entries to its tail when we sync changes.
*
* - Each metaslab has two new range trees that hold its unflushed changes,
* ms_unflushed_allocs and ms_unflushed_frees. These are always disjoint.
*
* == Flushing algorithm ==
*
* The decision of how many metaslabs to flush on a give TXG is guided by
* two heuristics:
*
* [1] The memory heuristic -
* We keep track of the memory used by the unflushed trees from all the
* metaslabs [see sus_memused of spa_unflushed_stats] and we ensure that it
* stays below a certain threshold which is determined by an arbitrary hard
* limit and an arbitrary percentage of the system's memory [see
* spa_log_exceeds_memlimit()]. When we see that the memory usage of the
* unflushed changes are passing that threshold, we flush metaslabs, which
* empties their unflushed range trees, reducing the memory used.
*
* [2] The block heuristic -
* We try to keep the total number of blocks in the log space maps in check
* so the log doesn't grow indefinitely and we don't induce a lot of overhead
* when loading the pool. At the same time we don't want to flush a lot of
* metaslabs too often as this would defeat the purpose of the log space map.
* As a result we set a limit in the amount of blocks that we think it's
* acceptable for the log space maps to have and try not to cross it.
* [see sus_blocklimit from spa_unflushed_stats].
*
* In order to stay below the block limit every TXG we have to estimate how
* many metaslabs we need to flush based on the current rate of incoming blocks
* and our history of log space map blocks. The main idea here is to answer
* the question of how many metaslabs do we need to flush in order to get rid
* at least an X amount of log space map blocks. We can answer this question
* by iterating backwards from the oldest log space map to the newest one
* and looking at their metaslab and block counts. At this point the log summary
* mentioned above comes handy as it reduces the amount of things that we have
* to iterate (even though it may reduce the preciseness of our estimates due
* to its aggregation of data). So with that in mind, we project the incoming
* rate of the current TXG into the future and attempt to approximate how many
* metaslabs would we need to flush from now in order to avoid exceeding our
* block limit in different points in the future (granted that we would keep
* flushing the same number of metaslabs for every TXG). Then we take the
* maximum number from all these estimates to be on the safe side. For the
* exact implementation details of algorithm refer to
* spa_estimate_metaslabs_to_flush.
*/
/*
* This is used as the block size for the space maps used for the
* log space map feature. These space maps benefit from a bigger
* block size as we expect to be writing a lot of data to them at
* once.
*/
-unsigned long zfs_log_sm_blksz = 1ULL << 17;
+static const unsigned long zfs_log_sm_blksz = 1ULL << 17;
/*
* Percentage of the overall system's memory that ZFS allows to be
* used for unflushed changes (e.g. the sum of size of all the nodes
* in the unflushed trees).
*
* Note that this value is calculated over 1000000 for finer granularity
* (thus the _ppm suffix; reads as "parts per million"). As an example,
* the default of 1000 allows 0.1% of memory to be used.
*/
-unsigned long zfs_unflushed_max_mem_ppm = 1000;
+static unsigned long zfs_unflushed_max_mem_ppm = 1000;
/*
* Specific hard-limit in memory that ZFS allows to be used for
* unflushed changes.
*/
-unsigned long zfs_unflushed_max_mem_amt = 1ULL << 30;
+static unsigned long zfs_unflushed_max_mem_amt = 1ULL << 30;
/*
* The following tunable determines the number of blocks that can be used for
* the log space maps. It is expressed as a percentage of the total number of
* metaslabs in the pool (i.e. the default of 400 means that the number of log
* blocks is capped at 4 times the number of metaslabs).
*
* This value exists to tune our flushing algorithm, with higher values
* flushing metaslabs less often (doing less I/Os) per TXG versus lower values
* flushing metaslabs more aggressively with the upside of saving overheads
* when loading the pool. Another factor in this tradeoff is that flushing
* less often can potentially lead to better utilization of the metaslab space
* map's block size as we accumulate more changes per flush.
*
* Given that this tunable indirectly controls the flush rate (metaslabs
* flushed per txg) and that's why making it a percentage in terms of the
* number of metaslabs in the pool makes sense here.
*
* As a rule of thumb we default this tunable to 400% based on the following:
*
* 1] Assuming a constant flush rate and a constant incoming rate of log blocks
* it is reasonable to expect that the amount of obsolete entries changes
* linearly from txg to txg (e.g. the oldest log should have the most
* obsolete entries, and the most recent one the least). With this we could
* say that, at any given time, about half of the entries in the whole space
* map log are obsolete. Thus for every two entries for a metaslab in the
* log space map, only one of them is valid and actually makes it to the
* metaslab's space map.
* [factor of 2]
* 2] Each entry in the log space map is guaranteed to be two words while
* entries in metaslab space maps are generally single-word.
* [an extra factor of 2 - 400% overall]
* 3] Even if [1] and [2] are slightly less than 2 each, we haven't taken into
* account any consolidation of segments from the log space map to the
* unflushed range trees nor their history (e.g. a segment being allocated,
* then freed, then allocated again means 3 log space map entries but 0
* metaslab space map entries). Depending on the workload, we've seen ~1.8
* non-obsolete log space map entries per metaslab entry, for a total of
* ~600%. Since most of these estimates though are workload dependent, we
* default on 400% to be conservative.
*
* Thus we could say that even in the worst
* case of [1] and [2], the factor should end up being 4.
*
* That said, regardless of the number of metaslabs in the pool we need to
* provide upper and lower bounds for the log block limit.
* [see zfs_unflushed_log_block_{min,max}]
*/
-unsigned long zfs_unflushed_log_block_pct = 400;
+static unsigned long zfs_unflushed_log_block_pct = 400;
/*
* If the number of metaslabs is small and our incoming rate is high, we could
* get into a situation that we are flushing all our metaslabs every TXG. Thus
* we always allow at least this many log blocks.
*/
-unsigned long zfs_unflushed_log_block_min = 1000;
+static unsigned long zfs_unflushed_log_block_min = 1000;
/*
* If the log becomes too big, the import time of the pool can take a hit in
* terms of performance. Thus we have a hard limit in the size of the log in
* terms of blocks.
*/
-unsigned long zfs_unflushed_log_block_max = (1ULL << 18);
+static unsigned long zfs_unflushed_log_block_max = (1ULL << 18);
/*
* Max # of rows allowed for the log_summary. The tradeoff here is accuracy and
* stability of the flushing algorithm (longer summary) vs its runtime overhead
* (smaller summary is faster to traverse).
*/
-unsigned long zfs_max_logsm_summary_length = 10;
+static unsigned long zfs_max_logsm_summary_length = 10;
/*
* Tunable that sets the lower bound on the metaslabs to flush every TXG.
*
* Setting this to 0 has no effect since if the pool is idle we won't even be
* creating log space maps and therefore we won't be flushing. On the other
* hand if the pool has any incoming workload our block heuristic will start
* flushing metaslabs anyway.
*
* The point of this tunable is to be used in extreme cases where we really
* want to flush more metaslabs than our adaptable heuristic plans to flush.
*/
-unsigned long zfs_min_metaslabs_to_flush = 1;
+static unsigned long zfs_min_metaslabs_to_flush = 1;
/*
* Tunable that specifies how far in the past do we want to look when trying to
* estimate the incoming log blocks for the current TXG.
*
* Setting this too high may not only increase runtime but also minimize the
* effect of the incoming rates from the most recent TXGs as we take the
* average over all the blocks that we walk
* [see spa_estimate_incoming_log_blocks].
*/
-unsigned long zfs_max_log_walking = 5;
+static unsigned long zfs_max_log_walking = 5;
/*
* This tunable exists solely for testing purposes. It ensures that the log
* spacemaps are not flushed and destroyed during export in order for the
* relevant log spacemap import code paths to be tested (effectively simulating
* a crash).
*/
int zfs_keep_log_spacemaps_at_export = 0;
static uint64_t
spa_estimate_incoming_log_blocks(spa_t *spa)
{
ASSERT3U(spa_sync_pass(spa), ==, 1);
uint64_t steps = 0, sum = 0;
for (spa_log_sm_t *sls = avl_last(&spa->spa_sm_logs_by_txg);
sls != NULL && steps < zfs_max_log_walking;
sls = AVL_PREV(&spa->spa_sm_logs_by_txg, sls)) {
if (sls->sls_txg == spa_syncing_txg(spa)) {
/*
* skip the log created in this TXG as this would
* make our estimations inaccurate.
*/
continue;
}
sum += sls->sls_nblocks;
steps++;
}
return ((steps > 0) ? DIV_ROUND_UP(sum, steps) : 0);
}
uint64_t
spa_log_sm_blocklimit(spa_t *spa)
{
return (spa->spa_unflushed_stats.sus_blocklimit);
}
void
spa_log_sm_set_blocklimit(spa_t *spa)
{
if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) {
ASSERT0(spa_log_sm_blocklimit(spa));
return;
}
uint64_t calculated_limit =
(spa_total_metaslabs(spa) * zfs_unflushed_log_block_pct) / 100;
spa->spa_unflushed_stats.sus_blocklimit = MIN(MAX(calculated_limit,
zfs_unflushed_log_block_min), zfs_unflushed_log_block_max);
}
uint64_t
spa_log_sm_nblocks(spa_t *spa)
{
return (spa->spa_unflushed_stats.sus_nblocks);
}
/*
* Ensure that the in-memory log space map structures and the summary
* have the same block and metaslab counts.
*/
static void
spa_log_summary_verify_counts(spa_t *spa)
{
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));
if ((zfs_flags & ZFS_DEBUG_LOG_SPACEMAP) == 0)
return;
uint64_t ms_in_avl = avl_numnodes(&spa->spa_metaslabs_by_flushed);
uint64_t ms_in_summary = 0, blk_in_summary = 0;
for (log_summary_entry_t *e = list_head(&spa->spa_log_summary);
e; e = list_next(&spa->spa_log_summary, e)) {
ms_in_summary += e->lse_mscount;
blk_in_summary += e->lse_blkcount;
}
uint64_t ms_in_logs = 0, blk_in_logs = 0;
for (spa_log_sm_t *sls = avl_first(&spa->spa_sm_logs_by_txg);
sls; sls = AVL_NEXT(&spa->spa_sm_logs_by_txg, sls)) {
ms_in_logs += sls->sls_mscount;
blk_in_logs += sls->sls_nblocks;
}
VERIFY3U(ms_in_logs, ==, ms_in_summary);
VERIFY3U(ms_in_logs, ==, ms_in_avl);
VERIFY3U(blk_in_logs, ==, blk_in_summary);
VERIFY3U(blk_in_logs, ==, spa_log_sm_nblocks(spa));
}
static boolean_t
summary_entry_is_full(spa_t *spa, log_summary_entry_t *e)
{
uint64_t blocks_per_row = MAX(1,
DIV_ROUND_UP(spa_log_sm_blocklimit(spa),
zfs_max_logsm_summary_length));
return (blocks_per_row <= e->lse_blkcount);
}
/*
* Update the log summary information to reflect the fact that a metaslab
* was flushed or destroyed (e.g due to device removal or pool export/destroy).
*
* We typically flush the oldest flushed metaslab so the first (and oldest)
* entry of the summary is updated. However if that metaslab is getting loaded
* we may flush the second oldest one which may be part of an entry later in
* the summary. Moreover, if we call into this function from metaslab_fini()
* the metaslabs probably won't be ordered by ms_unflushed_txg. Thus we ask
* for a txg as an argument so we can locate the appropriate summary entry for
* the metaslab.
*/
void
spa_log_summary_decrement_mscount(spa_t *spa, uint64_t txg)
{
/*
* We don't track summary data for read-only pools and this function
* can be called from metaslab_fini(). In that case return immediately.
*/
if (!spa_writeable(spa))
return;
log_summary_entry_t *target = NULL;
for (log_summary_entry_t *e = list_head(&spa->spa_log_summary);
e != NULL; e = list_next(&spa->spa_log_summary, e)) {
if (e->lse_start > txg)
break;
target = e;
}
if (target == NULL || target->lse_mscount == 0) {
/*
* We didn't find a summary entry for this metaslab. We must be
* at the teardown of a spa_load() attempt that got an error
* while reading the log space maps.
*/
VERIFY3S(spa_load_state(spa), ==, SPA_LOAD_ERROR);
return;
}
target->lse_mscount--;
}
/*
* Update the log summary information to reflect the fact that we destroyed
* old log space maps. Since we can only destroy the oldest log space maps,
* we decrement the block count of the oldest summary entry and potentially
* destroy it when that count hits 0.
*
* This function is called after a metaslab is flushed and typically that
* metaslab is the oldest flushed, which means that this function will
* typically decrement the block count of the first entry of the summary and
* potentially free it if the block count gets to zero (its metaslab count
* should be zero too at that point).
*
* There are certain scenarios though that don't work exactly like that so we
* need to account for them:
*
* Scenario [1]: It is possible that after we flushed the oldest flushed
* metaslab and we destroyed the oldest log space map, more recent logs had 0
* metaslabs pointing to them so we got rid of them too. This can happen due
* to metaslabs being destroyed through device removal, or because the oldest
* flushed metaslab was loading but we kept flushing more recently flushed
* metaslabs due to the memory pressure of unflushed changes. Because of that,
* we always iterate from the beginning of the summary and if blocks_gone is
* bigger than the block_count of the current entry we free that entry (we
* expect its metaslab count to be zero), we decrement blocks_gone and on to
* the next entry repeating this procedure until blocks_gone gets decremented
* to 0. Doing this also works for the typical case mentioned above.
*
* Scenario [2]: The oldest flushed metaslab isn't necessarily accounted by
* the first (and oldest) entry in the summary. If the first few entries of
* the summary were only accounting metaslabs from a device that was just
* removed, then the current oldest flushed metaslab could be accounted by an
* entry somewhere in the middle of the summary. Moreover flushing that
* metaslab will destroy all the log space maps older than its ms_unflushed_txg
* because they became obsolete after the removal. Thus, iterating as we did
* for scenario [1] works out for this case too.
*
* Scenario [3]: At times we decide to flush all the metaslabs in the pool
* in one TXG (either because we are exporting the pool or because our flushing
* heuristics decided to do so). When that happens all the log space maps get
* destroyed except the one created for the current TXG which doesn't have
* any log blocks yet. As log space maps get destroyed with every metaslab that
* we flush, entries in the summary are also destroyed. This brings a weird
* corner-case when we flush the last metaslab and the log space map of the
* current TXG is in the same summary entry with other log space maps that
* are older. When that happens we are eventually left with this one last
* summary entry whose blocks are gone (blocks_gone equals the entry's block
* count) but its metaslab count is non-zero (because it accounts all the
* metaslabs in the pool as they all got flushed). Under this scenario we can't
* free this last summary entry as it's referencing all the metaslabs in the
* pool and its block count will get incremented at the end of this sync (when
* we close the syncing log space map). Thus we just decrement its current
* block count and leave it alone. In the case that the pool gets exported,
* its metaslab count will be decremented over time as we call metaslab_fini()
* for all the metaslabs in the pool and the entry will be freed at
* spa_unload_log_sm_metadata().
*/
void
spa_log_summary_decrement_blkcount(spa_t *spa, uint64_t blocks_gone)
{
for (log_summary_entry_t *e = list_head(&spa->spa_log_summary);
e != NULL; e = list_head(&spa->spa_log_summary)) {
if (e->lse_blkcount > blocks_gone) {
/*
* Assert that we stopped at an entry that is not
* obsolete.
*/
ASSERT(e->lse_mscount != 0);
e->lse_blkcount -= blocks_gone;
blocks_gone = 0;
break;
} else if (e->lse_mscount == 0) {
/* remove obsolete entry */
blocks_gone -= e->lse_blkcount;
list_remove(&spa->spa_log_summary, e);
kmem_free(e, sizeof (log_summary_entry_t));
} else {
/* Verify that this is scenario [3] mentioned above. */
VERIFY3U(blocks_gone, ==, e->lse_blkcount);
/*
* Assert that this is scenario [3] further by ensuring
* that this is the only entry in the summary.
*/
VERIFY3P(e, ==, list_tail(&spa->spa_log_summary));
ASSERT3P(e, ==, list_head(&spa->spa_log_summary));
blocks_gone = e->lse_blkcount = 0;
break;
}
}
/*
* Ensure that there is no way we are trying to remove more blocks
* than the # of blocks in the summary.
*/
ASSERT0(blocks_gone);
}
void
spa_log_sm_decrement_mscount(spa_t *spa, uint64_t txg)
{
spa_log_sm_t target = { .sls_txg = txg };
spa_log_sm_t *sls = avl_find(&spa->spa_sm_logs_by_txg,
&target, NULL);
if (sls == NULL) {
/*
* We must be at the teardown of a spa_load() attempt that
* got an error while reading the log space maps.
*/
VERIFY3S(spa_load_state(spa), ==, SPA_LOAD_ERROR);
return;
}
ASSERT(sls->sls_mscount > 0);
sls->sls_mscount--;
}
void
spa_log_sm_increment_current_mscount(spa_t *spa)
{
spa_log_sm_t *last_sls = avl_last(&spa->spa_sm_logs_by_txg);
ASSERT3U(last_sls->sls_txg, ==, spa_syncing_txg(spa));
last_sls->sls_mscount++;
}
static void
summary_add_data(spa_t *spa, uint64_t txg, uint64_t metaslabs_flushed,
uint64_t nblocks)
{
log_summary_entry_t *e = list_tail(&spa->spa_log_summary);
if (e == NULL || summary_entry_is_full(spa, e)) {
e = kmem_zalloc(sizeof (log_summary_entry_t), KM_SLEEP);
e->lse_start = txg;
list_insert_tail(&spa->spa_log_summary, e);
}
ASSERT3U(e->lse_start, <=, txg);
e->lse_mscount += metaslabs_flushed;
e->lse_blkcount += nblocks;
}
static void
spa_log_summary_add_incoming_blocks(spa_t *spa, uint64_t nblocks)
{
summary_add_data(spa, spa_syncing_txg(spa), 0, nblocks);
}
void
spa_log_summary_add_flushed_metaslab(spa_t *spa)
{
summary_add_data(spa, spa_syncing_txg(spa), 1, 0);
}
/*
* This function attempts to estimate how many metaslabs should
* we flush to satisfy our block heuristic for the log spacemap
* for the upcoming TXGs.
*
* Specifically, it first tries to estimate the number of incoming
* blocks in this TXG. Then by projecting that incoming rate to
* future TXGs and using the log summary, it figures out how many
* flushes we would need to do for future TXGs individually to
* stay below our block limit and returns the maximum number of
* flushes from those estimates.
*/
static uint64_t
spa_estimate_metaslabs_to_flush(spa_t *spa)
{
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));
ASSERT3U(spa_sync_pass(spa), ==, 1);
ASSERT(spa_log_sm_blocklimit(spa) != 0);
/*
* This variable contains the incoming rate that will be projected
* and used for our flushing estimates in the future.
*/
uint64_t incoming = spa_estimate_incoming_log_blocks(spa);
/*
* At any point in time this variable tells us how many
* TXGs in the future we are so we can make our estimations.
*/
uint64_t txgs_in_future = 1;
/*
* This variable tells us how much room do we have until we hit
* our limit. When it goes negative, it means that we've exceeded
* our limit and we need to flush.
*
* Note that since we start at the first TXG in the future (i.e.
* txgs_in_future starts from 1) we already decrement this
* variable by the incoming rate.
*/
int64_t available_blocks =
spa_log_sm_blocklimit(spa) - spa_log_sm_nblocks(spa) - incoming;
/*
* This variable tells us the total number of flushes needed to
* keep the log size within the limit when we reach txgs_in_future.
*/
uint64_t total_flushes = 0;
/* Holds the current maximum of our estimates so far. */
uint64_t max_flushes_pertxg =
MIN(avl_numnodes(&spa->spa_metaslabs_by_flushed),
zfs_min_metaslabs_to_flush);
/*
* For our estimations we only look as far in the future
* as the summary allows us.
*/
for (log_summary_entry_t *e = list_head(&spa->spa_log_summary);
e; e = list_next(&spa->spa_log_summary, e)) {
/*
* If there is still room before we exceed our limit
* then keep skipping TXGs accumulating more blocks
* based on the incoming rate until we exceed it.
*/
if (available_blocks >= 0) {
uint64_t skip_txgs = (available_blocks / incoming) + 1;
available_blocks -= (skip_txgs * incoming);
txgs_in_future += skip_txgs;
ASSERT3S(available_blocks, >=, -incoming);
}
/*
* At this point we're far enough into the future where
* the limit was just exceeded and we flush metaslabs
* based on the current entry in the summary, updating
* our available_blocks.
*/
ASSERT3S(available_blocks, <, 0);
available_blocks += e->lse_blkcount;
total_flushes += e->lse_mscount;
/*
* Keep the running maximum of the total_flushes that
* we've done so far over the number of TXGs in the
* future that we are. The idea here is to estimate
* the average number of flushes that we should do
* every TXG so that when we are that many TXGs in the
* future we stay under the limit.
*/
max_flushes_pertxg = MAX(max_flushes_pertxg,
DIV_ROUND_UP(total_flushes, txgs_in_future));
ASSERT3U(avl_numnodes(&spa->spa_metaslabs_by_flushed), >=,
max_flushes_pertxg);
}
return (max_flushes_pertxg);
}
uint64_t
spa_log_sm_memused(spa_t *spa)
{
return (spa->spa_unflushed_stats.sus_memused);
}
static boolean_t
spa_log_exceeds_memlimit(spa_t *spa)
{
if (spa_log_sm_memused(spa) > zfs_unflushed_max_mem_amt)
return (B_TRUE);
uint64_t system_mem_allowed = ((physmem * PAGESIZE) *
zfs_unflushed_max_mem_ppm) / 1000000;
if (spa_log_sm_memused(spa) > system_mem_allowed)
return (B_TRUE);
return (B_FALSE);
}
boolean_t
spa_flush_all_logs_requested(spa_t *spa)
{
return (spa->spa_log_flushall_txg != 0);
}
void
spa_flush_metaslabs(spa_t *spa, dmu_tx_t *tx)
{
uint64_t txg = dmu_tx_get_txg(tx);
if (spa_sync_pass(spa) != 1)
return;
if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))
return;
/*
* If we don't have any metaslabs with unflushed changes
* return immediately.
*/
if (avl_numnodes(&spa->spa_metaslabs_by_flushed) == 0)
return;
/*
* During SPA export we leave a few empty TXGs to go by [see
* spa_final_dirty_txg() to understand why]. For this specific
* case, it is important to not flush any metaslabs as that
* would dirty this TXG.
*
* That said, during one of these dirty TXGs that is less or
* equal to spa_final_dirty(), spa_unload() will request that
* we try to flush all the metaslabs for that TXG before
* exporting the pool, thus we ensure that we didn't get a
* request of flushing everything before we attempt to return
* immediately.
*/
if (spa->spa_uberblock.ub_rootbp.blk_birth < txg &&
!dmu_objset_is_dirty(spa_meta_objset(spa), txg) &&
!spa_flush_all_logs_requested(spa))
return;
/*
* We need to generate a log space map before flushing because this
* will set up the in-memory data (i.e. node in spa_sm_logs_by_txg)
* for this TXG's flushed metaslab count (aka sls_mscount which is
* manipulated in many ways down the metaslab_flush() codepath).
*
* That is not to say that we may generate a log space map when we
* don't need it. If we are flushing metaslabs, that means that we
* were going to write changes to disk anyway, so even if we were
* not flushing, a log space map would have been created anyway in
* metaslab_sync().
*/
spa_generate_syncing_log_sm(spa, tx);
/*
* This variable tells us how many metaslabs we want to flush based
* on the block-heuristic of our flushing algorithm (see block comment
* of log space map feature). We also decrement this as we flush
* metaslabs and attempt to destroy old log space maps.
*/
uint64_t want_to_flush;
if (spa_flush_all_logs_requested(spa)) {
ASSERT3S(spa_state(spa), ==, POOL_STATE_EXPORTED);
want_to_flush = avl_numnodes(&spa->spa_metaslabs_by_flushed);
} else {
want_to_flush = spa_estimate_metaslabs_to_flush(spa);
}
ASSERT3U(avl_numnodes(&spa->spa_metaslabs_by_flushed), >=,
want_to_flush);
/* Used purely for verification purposes */
uint64_t visited = 0;
/*
* Ideally we would only iterate through spa_metaslabs_by_flushed
* using only one variable (curr). We can't do that because
* metaslab_flush() mutates position of curr in the AVL when
* it flushes that metaslab by moving it to the end of the tree.
* Thus we always keep track of the original next node of the
* current node (curr) in another variable (next).
*/
metaslab_t *next = NULL;
for (metaslab_t *curr = avl_first(&spa->spa_metaslabs_by_flushed);
curr != NULL; curr = next) {
next = AVL_NEXT(&spa->spa_metaslabs_by_flushed, curr);
/*
* If this metaslab has been flushed this txg then we've done
* a full circle over the metaslabs.
*/
if (metaslab_unflushed_txg(curr) == txg)
break;
/*
* If we are done flushing for the block heuristic and the
* unflushed changes don't exceed the memory limit just stop.
*/
if (want_to_flush == 0 && !spa_log_exceeds_memlimit(spa))
break;
mutex_enter(&curr->ms_sync_lock);
mutex_enter(&curr->ms_lock);
boolean_t flushed = metaslab_flush(curr, tx);
mutex_exit(&curr->ms_lock);
mutex_exit(&curr->ms_sync_lock);
/*
* If we failed to flush a metaslab (because it was loading),
* then we are done with the block heuristic as it's not
* possible to destroy any log space maps once you've skipped
* a metaslab. In that case we just set our counter to 0 but
* we continue looping in case there is still memory pressure
* due to unflushed changes. Note that, flushing a metaslab
* that is not the oldest flushed in the pool, will never
* destroy any log space maps [see spa_cleanup_old_sm_logs()].
*/
if (!flushed) {
want_to_flush = 0;
} else if (want_to_flush > 0) {
want_to_flush--;
}
visited++;
}
ASSERT3U(avl_numnodes(&spa->spa_metaslabs_by_flushed), >=, visited);
}
/*
* Close the log space map for this TXG and update the block counts
* for the log's in-memory structure and the summary.
*/
void
spa_sync_close_syncing_log_sm(spa_t *spa)
{
if (spa_syncing_log_sm(spa) == NULL)
return;
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));
spa_log_sm_t *sls = avl_last(&spa->spa_sm_logs_by_txg);
ASSERT3U(sls->sls_txg, ==, spa_syncing_txg(spa));
sls->sls_nblocks = space_map_nblocks(spa_syncing_log_sm(spa));
spa->spa_unflushed_stats.sus_nblocks += sls->sls_nblocks;
/*
* Note that we can't assert that sls_mscount is not 0,
* because there is the case where the first metaslab
* in spa_metaslabs_by_flushed is loading and we were
* not able to flush any metaslabs the current TXG.
*/
ASSERT(sls->sls_nblocks != 0);
spa_log_summary_add_incoming_blocks(spa, sls->sls_nblocks);
spa_log_summary_verify_counts(spa);
space_map_close(spa->spa_syncing_log_sm);
spa->spa_syncing_log_sm = NULL;
/*
* At this point we tried to flush as many metaslabs as we
* can as the pool is getting exported. Reset the "flush all"
* so the last few TXGs before closing the pool can be empty
* (e.g. not dirty).
*/
if (spa_flush_all_logs_requested(spa)) {
ASSERT3S(spa_state(spa), ==, POOL_STATE_EXPORTED);
spa->spa_log_flushall_txg = 0;
}
}
void
spa_cleanup_old_sm_logs(spa_t *spa, dmu_tx_t *tx)
{
objset_t *mos = spa_meta_objset(spa);
uint64_t spacemap_zap;
int error = zap_lookup(mos, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_LOG_SPACEMAP_ZAP, sizeof (spacemap_zap), 1, &spacemap_zap);
if (error == ENOENT) {
ASSERT(avl_is_empty(&spa->spa_sm_logs_by_txg));
return;
}
VERIFY0(error);
metaslab_t *oldest = avl_first(&spa->spa_metaslabs_by_flushed);
uint64_t oldest_flushed_txg = metaslab_unflushed_txg(oldest);
/* Free all log space maps older than the oldest_flushed_txg. */
for (spa_log_sm_t *sls = avl_first(&spa->spa_sm_logs_by_txg);
sls && sls->sls_txg < oldest_flushed_txg;
sls = avl_first(&spa->spa_sm_logs_by_txg)) {
ASSERT0(sls->sls_mscount);
avl_remove(&spa->spa_sm_logs_by_txg, sls);
space_map_free_obj(mos, sls->sls_sm_obj, tx);
VERIFY0(zap_remove_int(mos, spacemap_zap, sls->sls_txg, tx));
spa->spa_unflushed_stats.sus_nblocks -= sls->sls_nblocks;
kmem_free(sls, sizeof (spa_log_sm_t));
}
}
static spa_log_sm_t *
spa_log_sm_alloc(uint64_t sm_obj, uint64_t txg)
{
spa_log_sm_t *sls = kmem_zalloc(sizeof (*sls), KM_SLEEP);
sls->sls_sm_obj = sm_obj;
sls->sls_txg = txg;
return (sls);
}
void
spa_generate_syncing_log_sm(spa_t *spa, dmu_tx_t *tx)
{
uint64_t txg = dmu_tx_get_txg(tx);
objset_t *mos = spa_meta_objset(spa);
if (spa_syncing_log_sm(spa) != NULL)
return;
if (!spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP))
return;
uint64_t spacemap_zap;
int error = zap_lookup(mos, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_LOG_SPACEMAP_ZAP, sizeof (spacemap_zap), 1, &spacemap_zap);
if (error == ENOENT) {
ASSERT(avl_is_empty(&spa->spa_sm_logs_by_txg));
error = 0;
spacemap_zap = zap_create(mos,
DMU_OTN_ZAP_METADATA, DMU_OT_NONE, 0, tx);
VERIFY0(zap_add(mos, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_LOG_SPACEMAP_ZAP, sizeof (spacemap_zap), 1,
&spacemap_zap, tx));
spa_feature_incr(spa, SPA_FEATURE_LOG_SPACEMAP, tx);
}
VERIFY0(error);
uint64_t sm_obj;
ASSERT3U(zap_lookup_int_key(mos, spacemap_zap, txg, &sm_obj),
==, ENOENT);
sm_obj = space_map_alloc(mos, zfs_log_sm_blksz, tx);
VERIFY0(zap_add_int_key(mos, spacemap_zap, txg, sm_obj, tx));
avl_add(&spa->spa_sm_logs_by_txg, spa_log_sm_alloc(sm_obj, txg));
/*
* We pass UINT64_MAX as the space map's representation size
* and SPA_MINBLOCKSHIFT as the shift, to make the space map
* accept any sorts of segments since there's no real advantage
* to being more restrictive (given that we're already going
* to be using 2-word entries).
*/
VERIFY0(space_map_open(&spa->spa_syncing_log_sm, mos, sm_obj,
0, UINT64_MAX, SPA_MINBLOCKSHIFT));
/*
* If the log space map feature was just enabled, the blocklimit
* has not yet been set.
*/
if (spa_log_sm_blocklimit(spa) == 0)
spa_log_sm_set_blocklimit(spa);
}
/*
* Find all the log space maps stored in the space map ZAP and sort
* them by their TXG in spa_sm_logs_by_txg.
*/
static int
spa_ld_log_sm_metadata(spa_t *spa)
{
int error;
uint64_t spacemap_zap;
ASSERT(avl_is_empty(&spa->spa_sm_logs_by_txg));
error = zap_lookup(spa_meta_objset(spa), DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_LOG_SPACEMAP_ZAP, sizeof (spacemap_zap), 1, &spacemap_zap);
if (error == ENOENT) {
/* the space map ZAP doesn't exist yet */
return (0);
} else if (error != 0) {
spa_load_failed(spa, "spa_ld_log_sm_metadata(): failed at "
"zap_lookup(DMU_POOL_DIRECTORY_OBJECT) [error %d]",
error);
return (error);
}
zap_cursor_t zc;
zap_attribute_t za;
for (zap_cursor_init(&zc, spa_meta_objset(spa), spacemap_zap);
(error = zap_cursor_retrieve(&zc, &za)) == 0;
zap_cursor_advance(&zc)) {
uint64_t log_txg = zfs_strtonum(za.za_name, NULL);
spa_log_sm_t *sls =
spa_log_sm_alloc(za.za_first_integer, log_txg);
avl_add(&spa->spa_sm_logs_by_txg, sls);
}
zap_cursor_fini(&zc);
if (error != ENOENT) {
spa_load_failed(spa, "spa_ld_log_sm_metadata(): failed at "
"zap_cursor_retrieve(spacemap_zap) [error %d]",
error);
return (error);
}
for (metaslab_t *m = avl_first(&spa->spa_metaslabs_by_flushed);
m; m = AVL_NEXT(&spa->spa_metaslabs_by_flushed, m)) {
spa_log_sm_t target = { .sls_txg = metaslab_unflushed_txg(m) };
spa_log_sm_t *sls = avl_find(&spa->spa_sm_logs_by_txg,
&target, NULL);
/*
* At this point if sls is zero it means that a bug occurred
* in ZFS the last time the pool was open or earlier in the
* import code path. In general, we would have placed a
* VERIFY() here or in this case just let the kernel panic
* with NULL pointer dereference when incrementing sls_mscount,
* but since this is the import code path we can be a bit more
* lenient. Thus, for DEBUG bits we always cause a panic, while
* in production we log the error and just fail the import.
*/
ASSERT(sls != NULL);
if (sls == NULL) {
spa_load_failed(spa, "spa_ld_log_sm_metadata(): bug "
"encountered: could not find log spacemap for "
"TXG %llu [error %d]",
(u_longlong_t)metaslab_unflushed_txg(m), ENOENT);
return (ENOENT);
}
sls->sls_mscount++;
}
return (0);
}
typedef struct spa_ld_log_sm_arg {
spa_t *slls_spa;
uint64_t slls_txg;
} spa_ld_log_sm_arg_t;
static int
spa_ld_log_sm_cb(space_map_entry_t *sme, void *arg)
{
uint64_t offset = sme->sme_offset;
uint64_t size = sme->sme_run;
uint32_t vdev_id = sme->sme_vdev;
spa_ld_log_sm_arg_t *slls = arg;
spa_t *spa = slls->slls_spa;
vdev_t *vd = vdev_lookup_top(spa, vdev_id);
/*
* If the vdev has been removed (i.e. it is indirect or a hole)
* skip this entry. The contents of this vdev have already moved
* elsewhere.
*/
if (!vdev_is_concrete(vd))
return (0);
metaslab_t *ms = vd->vdev_ms[offset >> vd->vdev_ms_shift];
ASSERT(!ms->ms_loaded);
/*
* If we have already flushed entries for this TXG to this
* metaslab's space map, then ignore it. Note that we flush
* before processing any allocations/frees for that TXG, so
* the metaslab's space map only has entries from *before*
* the unflushed TXG.
*/
if (slls->slls_txg < metaslab_unflushed_txg(ms))
return (0);
switch (sme->sme_type) {
case SM_ALLOC:
range_tree_remove_xor_add_segment(offset, offset + size,
ms->ms_unflushed_frees, ms->ms_unflushed_allocs);
break;
case SM_FREE:
range_tree_remove_xor_add_segment(offset, offset + size,
ms->ms_unflushed_allocs, ms->ms_unflushed_frees);
break;
default:
panic("invalid maptype_t");
break;
}
return (0);
}
static int
spa_ld_log_sm_data(spa_t *spa)
{
int error = 0;
/*
* If we are not going to do any writes there is no need
* to read the log space maps.
*/
if (!spa_writeable(spa))
return (0);
ASSERT0(spa->spa_unflushed_stats.sus_nblocks);
ASSERT0(spa->spa_unflushed_stats.sus_memused);
hrtime_t read_logs_starttime = gethrtime();
/* this is a no-op when we don't have space map logs */
for (spa_log_sm_t *sls = avl_first(&spa->spa_sm_logs_by_txg);
sls; sls = AVL_NEXT(&spa->spa_sm_logs_by_txg, sls)) {
space_map_t *sm = NULL;
error = space_map_open(&sm, spa_meta_objset(spa),
sls->sls_sm_obj, 0, UINT64_MAX, SPA_MINBLOCKSHIFT);
if (error != 0) {
spa_load_failed(spa, "spa_ld_log_sm_data(): failed at "
"space_map_open(obj=%llu) [error %d]",
(u_longlong_t)sls->sls_sm_obj, error);
goto out;
}
struct spa_ld_log_sm_arg vla = {
.slls_spa = spa,
.slls_txg = sls->sls_txg
};
error = space_map_iterate(sm, space_map_length(sm),
spa_ld_log_sm_cb, &vla);
if (error != 0) {
space_map_close(sm);
spa_load_failed(spa, "spa_ld_log_sm_data(): failed "
"at space_map_iterate(obj=%llu) [error %d]",
(u_longlong_t)sls->sls_sm_obj, error);
goto out;
}
ASSERT0(sls->sls_nblocks);
sls->sls_nblocks = space_map_nblocks(sm);
spa->spa_unflushed_stats.sus_nblocks += sls->sls_nblocks;
summary_add_data(spa, sls->sls_txg,
sls->sls_mscount, sls->sls_nblocks);
space_map_close(sm);
}
hrtime_t read_logs_endtime = gethrtime();
spa_load_note(spa,
"read %llu log space maps (%llu total blocks - blksz = %llu bytes) "
"in %lld ms", (u_longlong_t)avl_numnodes(&spa->spa_sm_logs_by_txg),
(u_longlong_t)spa_log_sm_nblocks(spa),
(u_longlong_t)zfs_log_sm_blksz,
(longlong_t)((read_logs_endtime - read_logs_starttime) / 1000000));
out:
/*
* Now that the metaslabs contain their unflushed changes:
* [1] recalculate their actual allocated space
* [2] recalculate their weights
* [3] sum up the memory usage of their unflushed range trees
* [4] optionally load them, if debug_load is set
*
* Note that even in the case where we get here because of an
* error (e.g. error != 0), we still want to update the fields
* below in order to have a proper teardown in spa_unload().
*/
for (metaslab_t *m = avl_first(&spa->spa_metaslabs_by_flushed);
m != NULL; m = AVL_NEXT(&spa->spa_metaslabs_by_flushed, m)) {
mutex_enter(&m->ms_lock);
m->ms_allocated_space = space_map_allocated(m->ms_sm) +
range_tree_space(m->ms_unflushed_allocs) -
range_tree_space(m->ms_unflushed_frees);
vdev_t *vd = m->ms_group->mg_vd;
metaslab_space_update(vd, m->ms_group->mg_class,
range_tree_space(m->ms_unflushed_allocs), 0, 0);
metaslab_space_update(vd, m->ms_group->mg_class,
-range_tree_space(m->ms_unflushed_frees), 0, 0);
ASSERT0(m->ms_weight & METASLAB_ACTIVE_MASK);
metaslab_recalculate_weight_and_sort(m);
spa->spa_unflushed_stats.sus_memused +=
metaslab_unflushed_changes_memused(m);
if (metaslab_debug_load && m->ms_sm != NULL) {
VERIFY0(metaslab_load(m));
metaslab_set_selected_txg(m, 0);
}
mutex_exit(&m->ms_lock);
}
return (error);
}
static int
spa_ld_unflushed_txgs(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
objset_t *mos = spa_meta_objset(spa);
if (vd->vdev_top_zap == 0)
return (0);
uint64_t object = 0;
int error = zap_lookup(mos, vd->vdev_top_zap,
VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS,
sizeof (uint64_t), 1, &object);
if (error == ENOENT)
return (0);
else if (error != 0) {
spa_load_failed(spa, "spa_ld_unflushed_txgs(): failed at "
"zap_lookup(vdev_top_zap=%llu) [error %d]",
(u_longlong_t)vd->vdev_top_zap, error);
return (error);
}
for (uint64_t m = 0; m < vd->vdev_ms_count; m++) {
metaslab_t *ms = vd->vdev_ms[m];
ASSERT(ms != NULL);
metaslab_unflushed_phys_t entry;
uint64_t entry_size = sizeof (entry);
uint64_t entry_offset = ms->ms_id * entry_size;
error = dmu_read(mos, object,
entry_offset, entry_size, &entry, 0);
if (error != 0) {
spa_load_failed(spa, "spa_ld_unflushed_txgs(): "
"failed at dmu_read(obj=%llu) [error %d]",
(u_longlong_t)object, error);
return (error);
}
ms->ms_unflushed_txg = entry.msp_unflushed_txg;
if (ms->ms_unflushed_txg != 0) {
mutex_enter(&spa->spa_flushed_ms_lock);
avl_add(&spa->spa_metaslabs_by_flushed, ms);
mutex_exit(&spa->spa_flushed_ms_lock);
}
}
return (0);
}
/*
* Read all the log space map entries into their respective
* metaslab unflushed trees and keep them sorted by TXG in the
* SPA's metadata. In addition, setup all the metadata for the
* memory and the block heuristics.
*/
int
spa_ld_log_spacemaps(spa_t *spa)
{
int error;
spa_log_sm_set_blocklimit(spa);
for (uint64_t c = 0; c < spa->spa_root_vdev->vdev_children; c++) {
vdev_t *vd = spa->spa_root_vdev->vdev_child[c];
error = spa_ld_unflushed_txgs(vd);
if (error != 0)
return (error);
}
error = spa_ld_log_sm_metadata(spa);
if (error != 0)
return (error);
/*
* Note: we don't actually expect anything to change at this point
* but we grab the config lock so we don't fail any assertions
* when using vdev_lookup_top().
*/
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
error = spa_ld_log_sm_data(spa);
spa_config_exit(spa, SCL_CONFIG, FTAG);
return (error);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, unflushed_max_mem_amt, ULONG, ZMOD_RW,
"Specific hard-limit in memory that ZFS allows to be used for "
"unflushed changes");
ZFS_MODULE_PARAM(zfs, zfs_, unflushed_max_mem_ppm, ULONG, ZMOD_RW,
"Percentage of the overall system memory that ZFS allows to be "
"used for unflushed changes (value is calculated over 1000000 for "
"finer granularity)");
ZFS_MODULE_PARAM(zfs, zfs_, unflushed_log_block_max, ULONG, ZMOD_RW,
"Hard limit (upper-bound) in the size of the space map log "
"in terms of blocks.");
ZFS_MODULE_PARAM(zfs, zfs_, unflushed_log_block_min, ULONG, ZMOD_RW,
"Lower-bound limit for the maximum amount of blocks allowed in "
"log spacemap (see zfs_unflushed_log_block_max)");
ZFS_MODULE_PARAM(zfs, zfs_, unflushed_log_block_pct, ULONG, ZMOD_RW,
"Tunable used to determine the number of blocks that can be used for "
"the spacemap log, expressed as a percentage of the total number of "
"metaslabs in the pool (e.g. 400 means the number of log blocks is "
"capped at 4 times the number of metaslabs)");
ZFS_MODULE_PARAM(zfs, zfs_, max_log_walking, ULONG, ZMOD_RW,
"The number of past TXGs that the flushing algorithm of the log "
"spacemap feature uses to estimate incoming log blocks");
ZFS_MODULE_PARAM(zfs, zfs_, max_logsm_summary_length, ULONG, ZMOD_RW,
"Maximum number of rows allowed in the summary of the spacemap log");
ZFS_MODULE_PARAM(zfs, zfs_, min_metaslabs_to_flush, ULONG, ZMOD_RW,
"Minimum number of metaslabs to flush per dirty TXG");
ZFS_MODULE_PARAM(zfs, zfs_, keep_log_spacemaps_at_export, INT, ZMOD_RW,
"Prevent the log spacemaps from being flushed and destroyed "
"during pool export/destroy");
/* END CSTYLED */
diff --git a/module/zfs/spa_misc.c b/module/zfs/spa_misc.c
index 6ab602debc66..a04766e7e33b 100644
--- a/module/zfs/spa_misc.c
+++ b/module/zfs/spa_misc.c
@@ -1,2957 +1,2956 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright 2013 Saso Kiselkov. All rights reserved.
* Copyright (c) 2017 Datto Inc.
* Copyright (c) 2017, Intel Corporation.
* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa_impl.h>
#include <sys/zio.h>
#include <sys/zio_checksum.h>
#include <sys/zio_compress.h>
#include <sys/dmu.h>
#include <sys/dmu_tx.h>
#include <sys/zap.h>
#include <sys/zil.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_initialize.h>
#include <sys/vdev_trim.h>
#include <sys/vdev_file.h>
#include <sys/vdev_raidz.h>
#include <sys/metaslab.h>
#include <sys/uberblock_impl.h>
#include <sys/txg.h>
#include <sys/avl.h>
#include <sys/unique.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_prop.h>
#include <sys/fm/util.h>
#include <sys/dsl_scan.h>
#include <sys/fs/zfs.h>
#include <sys/metaslab_impl.h>
#include <sys/arc.h>
#include <sys/ddt.h>
#include <sys/kstat.h>
#include "zfs_prop.h"
#include <sys/btree.h>
#include <sys/zfeature.h>
#include <sys/qat.h>
#include <sys/zstd/zstd.h>
/*
* SPA locking
*
* There are three basic locks for managing spa_t structures:
*
* spa_namespace_lock (global mutex)
*
* This lock must be acquired to do any of the following:
*
* - Lookup a spa_t by name
* - Add or remove a spa_t from the namespace
* - Increase spa_refcount from non-zero
* - Check if spa_refcount is zero
* - Rename a spa_t
* - add/remove/attach/detach devices
* - Held for the duration of create/destroy/import/export
*
* It does not need to handle recursion. A create or destroy may
* reference objects (files or zvols) in other pools, but by
* definition they must have an existing reference, and will never need
* to lookup a spa_t by name.
*
* spa_refcount (per-spa zfs_refcount_t protected by mutex)
*
* This reference count keep track of any active users of the spa_t. The
* spa_t cannot be destroyed or freed while this is non-zero. Internally,
* the refcount is never really 'zero' - opening a pool implicitly keeps
* some references in the DMU. Internally we check against spa_minref, but
* present the image of a zero/non-zero value to consumers.
*
* spa_config_lock[] (per-spa array of rwlocks)
*
* This protects the spa_t from config changes, and must be held in
* the following circumstances:
*
* - RW_READER to perform I/O to the spa
* - RW_WRITER to change the vdev config
*
* The locking order is fairly straightforward:
*
* spa_namespace_lock -> spa_refcount
*
* The namespace lock must be acquired to increase the refcount from 0
* or to check if it is zero.
*
* spa_refcount -> spa_config_lock[]
*
* There must be at least one valid reference on the spa_t to acquire
* the config lock.
*
* spa_namespace_lock -> spa_config_lock[]
*
* The namespace lock must always be taken before the config lock.
*
*
* The spa_namespace_lock can be acquired directly and is globally visible.
*
* The namespace is manipulated using the following functions, all of which
* require the spa_namespace_lock to be held.
*
* spa_lookup() Lookup a spa_t by name.
*
* spa_add() Create a new spa_t in the namespace.
*
* spa_remove() Remove a spa_t from the namespace. This also
* frees up any memory associated with the spa_t.
*
* spa_next() Returns the next spa_t in the system, or the
* first if NULL is passed.
*
* spa_evict_all() Shutdown and remove all spa_t structures in
* the system.
*
* spa_guid_exists() Determine whether a pool/device guid exists.
*
* The spa_refcount is manipulated using the following functions:
*
* spa_open_ref() Adds a reference to the given spa_t. Must be
* called with spa_namespace_lock held if the
* refcount is currently zero.
*
* spa_close() Remove a reference from the spa_t. This will
* not free the spa_t or remove it from the
* namespace. No locking is required.
*
* spa_refcount_zero() Returns true if the refcount is currently
* zero. Must be called with spa_namespace_lock
* held.
*
* The spa_config_lock[] is an array of rwlocks, ordered as follows:
* SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV.
* spa_config_lock[] is manipulated with spa_config_{enter,exit,held}().
*
* To read the configuration, it suffices to hold one of these locks as reader.
* To modify the configuration, you must hold all locks as writer. To modify
* vdev state without altering the vdev tree's topology (e.g. online/offline),
* you must hold SCL_STATE and SCL_ZIO as writer.
*
* We use these distinct config locks to avoid recursive lock entry.
* For example, spa_sync() (which holds SCL_CONFIG as reader) induces
* block allocations (SCL_ALLOC), which may require reading space maps
* from disk (dmu_read() -> zio_read() -> SCL_ZIO).
*
* The spa config locks cannot be normal rwlocks because we need the
* ability to hand off ownership. For example, SCL_ZIO is acquired
* by the issuing thread and later released by an interrupt thread.
* They do, however, obey the usual write-wanted semantics to prevent
* writer (i.e. system administrator) starvation.
*
* The lock acquisition rules are as follows:
*
* SCL_CONFIG
* Protects changes to the vdev tree topology, such as vdev
* add/remove/attach/detach. Protects the dirty config list
* (spa_config_dirty_list) and the set of spares and l2arc devices.
*
* SCL_STATE
* Protects changes to pool state and vdev state, such as vdev
* online/offline/fault/degrade/clear. Protects the dirty state list
* (spa_state_dirty_list) and global pool state (spa_state).
*
* SCL_ALLOC
* Protects changes to metaslab groups and classes.
* Held as reader by metaslab_alloc() and metaslab_claim().
*
* SCL_ZIO
* Held by bp-level zios (those which have no io_vd upon entry)
* to prevent changes to the vdev tree. The bp-level zio implicitly
* protects all of its vdev child zios, which do not hold SCL_ZIO.
*
* SCL_FREE
* Protects changes to metaslab groups and classes.
* Held as reader by metaslab_free(). SCL_FREE is distinct from
* SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free
* blocks in zio_done() while another i/o that holds either
* SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete.
*
* SCL_VDEV
* Held as reader to prevent changes to the vdev tree during trivial
* inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the
* other locks, and lower than all of them, to ensure that it's safe
* to acquire regardless of caller context.
*
* In addition, the following rules apply:
*
* (a) spa_props_lock protects pool properties, spa_config and spa_config_list.
* The lock ordering is SCL_CONFIG > spa_props_lock.
*
* (b) I/O operations on leaf vdevs. For any zio operation that takes
* an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(),
* or zio_write_phys() -- the caller must ensure that the config cannot
* cannot change in the interim, and that the vdev cannot be reopened.
* SCL_STATE as reader suffices for both.
*
* The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit().
*
* spa_vdev_enter() Acquire the namespace lock and the config lock
* for writing.
*
* spa_vdev_exit() Release the config lock, wait for all I/O
* to complete, sync the updated configs to the
* cache, and release the namespace lock.
*
* vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit().
* Like spa_vdev_enter/exit, these are convenience wrappers -- the actual
* locking is, always, based on spa_namespace_lock and spa_config_lock[].
*/
static avl_tree_t spa_namespace_avl;
kmutex_t spa_namespace_lock;
static kcondvar_t spa_namespace_cv;
-int spa_max_replication_override = SPA_DVAS_PER_BP;
+static const int spa_max_replication_override = SPA_DVAS_PER_BP;
static kmutex_t spa_spare_lock;
static avl_tree_t spa_spare_avl;
static kmutex_t spa_l2cache_lock;
static avl_tree_t spa_l2cache_avl;
-kmem_cache_t *spa_buffer_pool;
spa_mode_t spa_mode_global = SPA_MODE_UNINIT;
#ifdef ZFS_DEBUG
/*
* Everything except dprintf, set_error, spa, and indirect_remap is on
* by default in debug builds.
*/
int zfs_flags = ~(ZFS_DEBUG_DPRINTF | ZFS_DEBUG_SET_ERROR |
ZFS_DEBUG_INDIRECT_REMAP);
#else
int zfs_flags = 0;
#endif
/*
* zfs_recover can be set to nonzero to attempt to recover from
* otherwise-fatal errors, typically caused by on-disk corruption. When
* set, calls to zfs_panic_recover() will turn into warning messages.
* This should only be used as a last resort, as it typically results
* in leaked space, or worse.
*/
int zfs_recover = B_FALSE;
/*
* If destroy encounters an EIO while reading metadata (e.g. indirect
* blocks), space referenced by the missing metadata can not be freed.
* Normally this causes the background destroy to become "stalled", as
* it is unable to make forward progress. While in this stalled state,
* all remaining space to free from the error-encountering filesystem is
* "temporarily leaked". Set this flag to cause it to ignore the EIO,
* permanently leak the space from indirect blocks that can not be read,
* and continue to free everything else that it can.
*
* The default, "stalling" behavior is useful if the storage partially
* fails (i.e. some but not all i/os fail), and then later recovers. In
* this case, we will be able to continue pool operations while it is
* partially failed, and when it recovers, we can continue to free the
* space, with no leaks. However, note that this case is actually
* fairly rare.
*
* Typically pools either (a) fail completely (but perhaps temporarily,
* e.g. a top-level vdev going offline), or (b) have localized,
* permanent errors (e.g. disk returns the wrong data due to bit flip or
* firmware bug). In case (a), this setting does not matter because the
* pool will be suspended and the sync thread will not be able to make
* forward progress regardless. In case (b), because the error is
* permanent, the best we can do is leak the minimum amount of space,
* which is what setting this flag will do. Therefore, it is reasonable
* for this flag to normally be set, but we chose the more conservative
* approach of not setting it, so that there is no possibility of
* leaking space in the "partial temporary" failure case.
*/
int zfs_free_leak_on_eio = B_FALSE;
/*
* Expiration time in milliseconds. This value has two meanings. First it is
* used to determine when the spa_deadman() logic should fire. By default the
* spa_deadman() will fire if spa_sync() has not completed in 600 seconds.
* Secondly, the value determines if an I/O is considered "hung". Any I/O that
* has not completed in zfs_deadman_synctime_ms is considered "hung" resulting
* in one of three behaviors controlled by zfs_deadman_failmode.
*/
-unsigned long zfs_deadman_synctime_ms = 600000UL;
+unsigned long zfs_deadman_synctime_ms = 600000UL; /* 10 min. */
/*
* This value controls the maximum amount of time zio_wait() will block for an
* outstanding IO. By default this is 300 seconds at which point the "hung"
* behavior will be applied as described for zfs_deadman_synctime_ms.
*/
-unsigned long zfs_deadman_ziotime_ms = 300000UL;
+unsigned long zfs_deadman_ziotime_ms = 300000UL; /* 5 min. */
/*
* Check time in milliseconds. This defines the frequency at which we check
* for hung I/O.
*/
-unsigned long zfs_deadman_checktime_ms = 60000UL;
+unsigned long zfs_deadman_checktime_ms = 60000UL; /* 1 min. */
/*
* By default the deadman is enabled.
*/
-int zfs_deadman_enabled = 1;
+int zfs_deadman_enabled = B_TRUE;
/*
* Controls the behavior of the deadman when it detects a "hung" I/O.
* Valid values are zfs_deadman_failmode=<wait|continue|panic>.
*
* wait - Wait for the "hung" I/O (default)
* continue - Attempt to recover from a "hung" I/O
* panic - Panic the system
*/
-char *zfs_deadman_failmode = "wait";
+const char *zfs_deadman_failmode = "wait";
/*
* The worst case is single-sector max-parity RAID-Z blocks, in which
* case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1)
* times the size; so just assume that. Add to this the fact that
* we can have up to 3 DVAs per bp, and one more factor of 2 because
* the block may be dittoed with up to 3 DVAs by ddt_sync(). All together,
* the worst case is:
* (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24
*/
int spa_asize_inflation = 24;
/*
* Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in
* the pool to be consumed (bounded by spa_max_slop). This ensures that we
* don't run the pool completely out of space, due to unaccounted changes (e.g.
* to the MOS). It also limits the worst-case time to allocate space. If we
* have less than this amount of free space, most ZPL operations (e.g. write,
* create) will return ENOSPC. The ZIL metaslabs (spa_embedded_log_class) are
* also part of this 3.2% of space which can't be consumed by normal writes;
* the slop space "proper" (spa_get_slop_space()) is decreased by the embedded
* log space.
*
* Certain operations (e.g. file removal, most administrative actions) can
* use half the slop space. They will only return ENOSPC if less than half
* the slop space is free. Typically, once the pool has less than the slop
* space free, the user will use these operations to free up space in the pool.
* These are the operations that call dsl_pool_adjustedsize() with the netfree
* argument set to TRUE.
*
* Operations that are almost guaranteed to free up space in the absence of
* a pool checkpoint can use up to three quarters of the slop space
* (e.g zfs destroy).
*
* A very restricted set of operations are always permitted, regardless of
* the amount of free space. These are the operations that call
* dsl_sync_task(ZFS_SPACE_CHECK_NONE). If these operations result in a net
* increase in the amount of space used, it is possible to run the pool
* completely out of space, causing it to be permanently read-only.
*
* Note that on very small pools, the slop space will be larger than
* 3.2%, in an effort to have it be at least spa_min_slop (128MB),
* but we never allow it to be more than half the pool size.
*
* Further, on very large pools, the slop space will be smaller than
* 3.2%, to avoid reserving much more space than we actually need; bounded
* by spa_max_slop (128GB).
*
* See also the comments in zfs_space_check_t.
*/
int spa_slop_shift = 5;
-uint64_t spa_min_slop = 128ULL * 1024 * 1024;
-uint64_t spa_max_slop = 128ULL * 1024 * 1024 * 1024;
-int spa_allocators = 4;
+static const uint64_t spa_min_slop = 128ULL * 1024 * 1024;
+static const uint64_t spa_max_slop = 128ULL * 1024 * 1024 * 1024;
+static const int spa_allocators = 4;
void
spa_load_failed(spa_t *spa, const char *fmt, ...)
{
va_list adx;
char buf[256];
va_start(adx, fmt);
(void) vsnprintf(buf, sizeof (buf), fmt, adx);
va_end(adx);
zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa->spa_name,
spa->spa_trust_config ? "trusted" : "untrusted", buf);
}
void
spa_load_note(spa_t *spa, const char *fmt, ...)
{
va_list adx;
char buf[256];
va_start(adx, fmt);
(void) vsnprintf(buf, sizeof (buf), fmt, adx);
va_end(adx);
zfs_dbgmsg("spa_load(%s, config %s): %s", spa->spa_name,
spa->spa_trust_config ? "trusted" : "untrusted", buf);
}
/*
* By default dedup and user data indirects land in the special class
*/
-int zfs_ddt_data_is_special = B_TRUE;
-int zfs_user_indirect_is_special = B_TRUE;
+static int zfs_ddt_data_is_special = B_TRUE;
+static int zfs_user_indirect_is_special = B_TRUE;
/*
* The percentage of special class final space reserved for metadata only.
* Once we allocate 100 - zfs_special_class_metadata_reserve_pct we only
* let metadata into the class.
*/
-int zfs_special_class_metadata_reserve_pct = 25;
+static int zfs_special_class_metadata_reserve_pct = 25;
/*
* ==========================================================================
* SPA config locking
* ==========================================================================
*/
static void
spa_config_lock_init(spa_t *spa)
{
for (int i = 0; i < SCL_LOCKS; i++) {
spa_config_lock_t *scl = &spa->spa_config_lock[i];
mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL);
scl->scl_writer = NULL;
scl->scl_write_wanted = 0;
scl->scl_count = 0;
}
}
static void
spa_config_lock_destroy(spa_t *spa)
{
for (int i = 0; i < SCL_LOCKS; i++) {
spa_config_lock_t *scl = &spa->spa_config_lock[i];
mutex_destroy(&scl->scl_lock);
cv_destroy(&scl->scl_cv);
ASSERT(scl->scl_writer == NULL);
ASSERT(scl->scl_write_wanted == 0);
ASSERT(scl->scl_count == 0);
}
}
int
spa_config_tryenter(spa_t *spa, int locks, void *tag, krw_t rw)
{
for (int i = 0; i < SCL_LOCKS; i++) {
spa_config_lock_t *scl = &spa->spa_config_lock[i];
if (!(locks & (1 << i)))
continue;
mutex_enter(&scl->scl_lock);
if (rw == RW_READER) {
if (scl->scl_writer || scl->scl_write_wanted) {
mutex_exit(&scl->scl_lock);
spa_config_exit(spa, locks & ((1 << i) - 1),
tag);
return (0);
}
} else {
ASSERT(scl->scl_writer != curthread);
if (scl->scl_count != 0) {
mutex_exit(&scl->scl_lock);
spa_config_exit(spa, locks & ((1 << i) - 1),
tag);
return (0);
}
scl->scl_writer = curthread;
}
scl->scl_count++;
mutex_exit(&scl->scl_lock);
}
return (1);
}
void
spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw)
{
(void) tag;
int wlocks_held = 0;
ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY);
for (int i = 0; i < SCL_LOCKS; i++) {
spa_config_lock_t *scl = &spa->spa_config_lock[i];
if (scl->scl_writer == curthread)
wlocks_held |= (1 << i);
if (!(locks & (1 << i)))
continue;
mutex_enter(&scl->scl_lock);
if (rw == RW_READER) {
while (scl->scl_writer || scl->scl_write_wanted) {
cv_wait(&scl->scl_cv, &scl->scl_lock);
}
} else {
ASSERT(scl->scl_writer != curthread);
while (scl->scl_count != 0) {
scl->scl_write_wanted++;
cv_wait(&scl->scl_cv, &scl->scl_lock);
scl->scl_write_wanted--;
}
scl->scl_writer = curthread;
}
scl->scl_count++;
mutex_exit(&scl->scl_lock);
}
ASSERT3U(wlocks_held, <=, locks);
}
void
spa_config_exit(spa_t *spa, int locks, const void *tag)
{
(void) tag;
for (int i = SCL_LOCKS - 1; i >= 0; i--) {
spa_config_lock_t *scl = &spa->spa_config_lock[i];
if (!(locks & (1 << i)))
continue;
mutex_enter(&scl->scl_lock);
ASSERT(scl->scl_count > 0);
if (--scl->scl_count == 0) {
ASSERT(scl->scl_writer == NULL ||
scl->scl_writer == curthread);
scl->scl_writer = NULL; /* OK in either case */
cv_broadcast(&scl->scl_cv);
}
mutex_exit(&scl->scl_lock);
}
}
int
spa_config_held(spa_t *spa, int locks, krw_t rw)
{
int locks_held = 0;
for (int i = 0; i < SCL_LOCKS; i++) {
spa_config_lock_t *scl = &spa->spa_config_lock[i];
if (!(locks & (1 << i)))
continue;
if ((rw == RW_READER && scl->scl_count != 0) ||
(rw == RW_WRITER && scl->scl_writer == curthread))
locks_held |= 1 << i;
}
return (locks_held);
}
/*
* ==========================================================================
* SPA namespace functions
* ==========================================================================
*/
/*
* Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held.
* Returns NULL if no matching spa_t is found.
*/
spa_t *
spa_lookup(const char *name)
{
static spa_t search; /* spa_t is large; don't allocate on stack */
spa_t *spa;
avl_index_t where;
char *cp;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
(void) strlcpy(search.spa_name, name, sizeof (search.spa_name));
/*
* If it's a full dataset name, figure out the pool name and
* just use that.
*/
cp = strpbrk(search.spa_name, "/@#");
if (cp != NULL)
*cp = '\0';
spa = avl_find(&spa_namespace_avl, &search, &where);
return (spa);
}
/*
* Fires when spa_sync has not completed within zfs_deadman_synctime_ms.
* If the zfs_deadman_enabled flag is set then it inspects all vdev queues
* looking for potentially hung I/Os.
*/
void
spa_deadman(void *arg)
{
spa_t *spa = arg;
/* Disable the deadman if the pool is suspended. */
if (spa_suspended(spa))
return;
zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu",
(gethrtime() - spa->spa_sync_starttime) / NANOSEC,
(u_longlong_t)++spa->spa_deadman_calls);
if (zfs_deadman_enabled)
vdev_deadman(spa->spa_root_vdev, FTAG);
spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq,
spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() +
MSEC_TO_TICK(zfs_deadman_checktime_ms));
}
static int
spa_log_sm_sort_by_txg(const void *va, const void *vb)
{
const spa_log_sm_t *a = va;
const spa_log_sm_t *b = vb;
return (TREE_CMP(a->sls_txg, b->sls_txg));
}
/*
* Create an uninitialized spa_t with the given name. Requires
* spa_namespace_lock. The caller must ensure that the spa_t doesn't already
* exist by calling spa_lookup() first.
*/
spa_t *
spa_add(const char *name, nvlist_t *config, const char *altroot)
{
spa_t *spa;
spa_config_dirent_t *dp;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP);
mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_feat_stats_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_flushed_ms_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa->spa_activities_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL);
cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL);
cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL);
cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL);
cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL);
cv_init(&spa->spa_activities_cv, NULL, CV_DEFAULT, NULL);
cv_init(&spa->spa_waiters_cv, NULL, CV_DEFAULT, NULL);
for (int t = 0; t < TXG_SIZE; t++)
bplist_create(&spa->spa_free_bplist[t]);
(void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name));
spa->spa_state = POOL_STATE_UNINITIALIZED;
spa->spa_freeze_txg = UINT64_MAX;
spa->spa_final_txg = UINT64_MAX;
spa->spa_load_max_txg = UINT64_MAX;
spa->spa_proc = &p0;
spa->spa_proc_state = SPA_PROC_NONE;
spa->spa_trust_config = B_TRUE;
spa->spa_hostid = zone_get_hostid(NULL);
spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms);
spa->spa_deadman_ziotime = MSEC2NSEC(zfs_deadman_ziotime_ms);
spa_set_deadman_failmode(spa, zfs_deadman_failmode);
zfs_refcount_create(&spa->spa_refcount);
spa_config_lock_init(spa);
spa_stats_init(spa);
avl_add(&spa_namespace_avl, spa);
/*
* Set the alternate root, if there is one.
*/
if (altroot)
spa->spa_root = spa_strdup(altroot);
spa->spa_alloc_count = spa_allocators;
spa->spa_allocs = kmem_zalloc(spa->spa_alloc_count *
sizeof (spa_alloc_t), KM_SLEEP);
for (int i = 0; i < spa->spa_alloc_count; i++) {
mutex_init(&spa->spa_allocs[i].spaa_lock, NULL, MUTEX_DEFAULT,
NULL);
avl_create(&spa->spa_allocs[i].spaa_tree, zio_bookmark_compare,
sizeof (zio_t), offsetof(zio_t, io_alloc_node));
}
avl_create(&spa->spa_metaslabs_by_flushed, metaslab_sort_by_flushed,
sizeof (metaslab_t), offsetof(metaslab_t, ms_spa_txg_node));
avl_create(&spa->spa_sm_logs_by_txg, spa_log_sm_sort_by_txg,
sizeof (spa_log_sm_t), offsetof(spa_log_sm_t, sls_node));
list_create(&spa->spa_log_summary, sizeof (log_summary_entry_t),
offsetof(log_summary_entry_t, lse_node));
/*
* Every pool starts with the default cachefile
*/
list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t),
offsetof(spa_config_dirent_t, scd_link));
dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP);
dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path);
list_insert_head(&spa->spa_config_list, dp);
VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME,
KM_SLEEP) == 0);
if (config != NULL) {
nvlist_t *features;
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
&features) == 0) {
VERIFY(nvlist_dup(features, &spa->spa_label_features,
0) == 0);
}
VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0);
}
if (spa->spa_label_features == NULL) {
VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME,
KM_SLEEP) == 0);
}
spa->spa_min_ashift = INT_MAX;
spa->spa_max_ashift = 0;
spa->spa_min_alloc = INT_MAX;
/* Reset cached value */
spa->spa_dedup_dspace = ~0ULL;
/*
* As a pool is being created, treat all features as disabled by
* setting SPA_FEATURE_DISABLED for all entries in the feature
* refcount cache.
*/
for (int i = 0; i < SPA_FEATURES; i++) {
spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED;
}
list_create(&spa->spa_leaf_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_leaf_node));
return (spa);
}
/*
* Removes a spa_t from the namespace, freeing up any memory used. Requires
* spa_namespace_lock. This is called only after the spa_t has been closed and
* deactivated.
*/
void
spa_remove(spa_t *spa)
{
spa_config_dirent_t *dp;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
ASSERT(spa_state(spa) == POOL_STATE_UNINITIALIZED);
ASSERT3U(zfs_refcount_count(&spa->spa_refcount), ==, 0);
ASSERT0(spa->spa_waiters);
nvlist_free(spa->spa_config_splitting);
avl_remove(&spa_namespace_avl, spa);
cv_broadcast(&spa_namespace_cv);
if (spa->spa_root)
spa_strfree(spa->spa_root);
while ((dp = list_head(&spa->spa_config_list)) != NULL) {
list_remove(&spa->spa_config_list, dp);
if (dp->scd_path != NULL)
spa_strfree(dp->scd_path);
kmem_free(dp, sizeof (spa_config_dirent_t));
}
for (int i = 0; i < spa->spa_alloc_count; i++) {
avl_destroy(&spa->spa_allocs[i].spaa_tree);
mutex_destroy(&spa->spa_allocs[i].spaa_lock);
}
kmem_free(spa->spa_allocs, spa->spa_alloc_count *
sizeof (spa_alloc_t));
avl_destroy(&spa->spa_metaslabs_by_flushed);
avl_destroy(&spa->spa_sm_logs_by_txg);
list_destroy(&spa->spa_log_summary);
list_destroy(&spa->spa_config_list);
list_destroy(&spa->spa_leaf_list);
nvlist_free(spa->spa_label_features);
nvlist_free(spa->spa_load_info);
nvlist_free(spa->spa_feat_stats);
spa_config_set(spa, NULL);
zfs_refcount_destroy(&spa->spa_refcount);
spa_stats_destroy(spa);
spa_config_lock_destroy(spa);
for (int t = 0; t < TXG_SIZE; t++)
bplist_destroy(&spa->spa_free_bplist[t]);
zio_checksum_templates_free(spa);
cv_destroy(&spa->spa_async_cv);
cv_destroy(&spa->spa_evicting_os_cv);
cv_destroy(&spa->spa_proc_cv);
cv_destroy(&spa->spa_scrub_io_cv);
cv_destroy(&spa->spa_suspend_cv);
cv_destroy(&spa->spa_activities_cv);
cv_destroy(&spa->spa_waiters_cv);
mutex_destroy(&spa->spa_flushed_ms_lock);
mutex_destroy(&spa->spa_async_lock);
mutex_destroy(&spa->spa_errlist_lock);
mutex_destroy(&spa->spa_errlog_lock);
mutex_destroy(&spa->spa_evicting_os_lock);
mutex_destroy(&spa->spa_history_lock);
mutex_destroy(&spa->spa_proc_lock);
mutex_destroy(&spa->spa_props_lock);
mutex_destroy(&spa->spa_cksum_tmpls_lock);
mutex_destroy(&spa->spa_scrub_lock);
mutex_destroy(&spa->spa_suspend_lock);
mutex_destroy(&spa->spa_vdev_top_lock);
mutex_destroy(&spa->spa_feat_stats_lock);
mutex_destroy(&spa->spa_activities_lock);
kmem_free(spa, sizeof (spa_t));
}
/*
* Given a pool, return the next pool in the namespace, or NULL if there is
* none. If 'prev' is NULL, return the first pool.
*/
spa_t *
spa_next(spa_t *prev)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (prev)
return (AVL_NEXT(&spa_namespace_avl, prev));
else
return (avl_first(&spa_namespace_avl));
}
/*
* ==========================================================================
* SPA refcount functions
* ==========================================================================
*/
/*
* Add a reference to the given spa_t. Must have at least one reference, or
* have the namespace lock held.
*/
void
spa_open_ref(spa_t *spa, void *tag)
{
ASSERT(zfs_refcount_count(&spa->spa_refcount) >= spa->spa_minref ||
MUTEX_HELD(&spa_namespace_lock));
(void) zfs_refcount_add(&spa->spa_refcount, tag);
}
/*
* Remove a reference to the given spa_t. Must have at least one reference, or
* have the namespace lock held.
*/
void
spa_close(spa_t *spa, void *tag)
{
ASSERT(zfs_refcount_count(&spa->spa_refcount) > spa->spa_minref ||
MUTEX_HELD(&spa_namespace_lock));
(void) zfs_refcount_remove(&spa->spa_refcount, tag);
}
/*
* Remove a reference to the given spa_t held by a dsl dir that is
* being asynchronously released. Async releases occur from a taskq
* performing eviction of dsl datasets and dirs. The namespace lock
* isn't held and the hold by the object being evicted may contribute to
* spa_minref (e.g. dataset or directory released during pool export),
* so the asserts in spa_close() do not apply.
*/
void
spa_async_close(spa_t *spa, void *tag)
{
(void) zfs_refcount_remove(&spa->spa_refcount, tag);
}
/*
* Check to see if the spa refcount is zero. Must be called with
* spa_namespace_lock held. We really compare against spa_minref, which is the
* number of references acquired when opening a pool
*/
boolean_t
spa_refcount_zero(spa_t *spa)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
return (zfs_refcount_count(&spa->spa_refcount) == spa->spa_minref);
}
/*
* ==========================================================================
* SPA spare and l2cache tracking
* ==========================================================================
*/
/*
* Hot spares and cache devices are tracked using the same code below,
* for 'auxiliary' devices.
*/
typedef struct spa_aux {
uint64_t aux_guid;
uint64_t aux_pool;
avl_node_t aux_avl;
int aux_count;
} spa_aux_t;
static inline int
spa_aux_compare(const void *a, const void *b)
{
const spa_aux_t *sa = (const spa_aux_t *)a;
const spa_aux_t *sb = (const spa_aux_t *)b;
return (TREE_CMP(sa->aux_guid, sb->aux_guid));
}
static void
spa_aux_add(vdev_t *vd, avl_tree_t *avl)
{
avl_index_t where;
spa_aux_t search;
spa_aux_t *aux;
search.aux_guid = vd->vdev_guid;
if ((aux = avl_find(avl, &search, &where)) != NULL) {
aux->aux_count++;
} else {
aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP);
aux->aux_guid = vd->vdev_guid;
aux->aux_count = 1;
avl_insert(avl, aux, where);
}
}
static void
spa_aux_remove(vdev_t *vd, avl_tree_t *avl)
{
spa_aux_t search;
spa_aux_t *aux;
avl_index_t where;
search.aux_guid = vd->vdev_guid;
aux = avl_find(avl, &search, &where);
ASSERT(aux != NULL);
if (--aux->aux_count == 0) {
avl_remove(avl, aux);
kmem_free(aux, sizeof (spa_aux_t));
} else if (aux->aux_pool == spa_guid(vd->vdev_spa)) {
aux->aux_pool = 0ULL;
}
}
static boolean_t
spa_aux_exists(uint64_t guid, uint64_t *pool, int *refcnt, avl_tree_t *avl)
{
spa_aux_t search, *found;
search.aux_guid = guid;
found = avl_find(avl, &search, NULL);
if (pool) {
if (found)
*pool = found->aux_pool;
else
*pool = 0ULL;
}
if (refcnt) {
if (found)
*refcnt = found->aux_count;
else
*refcnt = 0;
}
return (found != NULL);
}
static void
spa_aux_activate(vdev_t *vd, avl_tree_t *avl)
{
spa_aux_t search, *found;
avl_index_t where;
search.aux_guid = vd->vdev_guid;
found = avl_find(avl, &search, &where);
ASSERT(found != NULL);
ASSERT(found->aux_pool == 0ULL);
found->aux_pool = spa_guid(vd->vdev_spa);
}
/*
* Spares are tracked globally due to the following constraints:
*
* - A spare may be part of multiple pools.
* - A spare may be added to a pool even if it's actively in use within
* another pool.
* - A spare in use in any pool can only be the source of a replacement if
* the target is a spare in the same pool.
*
* We keep track of all spares on the system through the use of a reference
* counted AVL tree. When a vdev is added as a spare, or used as a replacement
* spare, then we bump the reference count in the AVL tree. In addition, we set
* the 'vdev_isspare' member to indicate that the device is a spare (active or
* inactive). When a spare is made active (used to replace a device in the
* pool), we also keep track of which pool its been made a part of.
*
* The 'spa_spare_lock' protects the AVL tree. These functions are normally
* called under the spa_namespace lock as part of vdev reconfiguration. The
* separate spare lock exists for the status query path, which does not need to
* be completely consistent with respect to other vdev configuration changes.
*/
static int
spa_spare_compare(const void *a, const void *b)
{
return (spa_aux_compare(a, b));
}
void
spa_spare_add(vdev_t *vd)
{
mutex_enter(&spa_spare_lock);
ASSERT(!vd->vdev_isspare);
spa_aux_add(vd, &spa_spare_avl);
vd->vdev_isspare = B_TRUE;
mutex_exit(&spa_spare_lock);
}
void
spa_spare_remove(vdev_t *vd)
{
mutex_enter(&spa_spare_lock);
ASSERT(vd->vdev_isspare);
spa_aux_remove(vd, &spa_spare_avl);
vd->vdev_isspare = B_FALSE;
mutex_exit(&spa_spare_lock);
}
boolean_t
spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt)
{
boolean_t found;
mutex_enter(&spa_spare_lock);
found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl);
mutex_exit(&spa_spare_lock);
return (found);
}
void
spa_spare_activate(vdev_t *vd)
{
mutex_enter(&spa_spare_lock);
ASSERT(vd->vdev_isspare);
spa_aux_activate(vd, &spa_spare_avl);
mutex_exit(&spa_spare_lock);
}
/*
* Level 2 ARC devices are tracked globally for the same reasons as spares.
* Cache devices currently only support one pool per cache device, and so
* for these devices the aux reference count is currently unused beyond 1.
*/
static int
spa_l2cache_compare(const void *a, const void *b)
{
return (spa_aux_compare(a, b));
}
void
spa_l2cache_add(vdev_t *vd)
{
mutex_enter(&spa_l2cache_lock);
ASSERT(!vd->vdev_isl2cache);
spa_aux_add(vd, &spa_l2cache_avl);
vd->vdev_isl2cache = B_TRUE;
mutex_exit(&spa_l2cache_lock);
}
void
spa_l2cache_remove(vdev_t *vd)
{
mutex_enter(&spa_l2cache_lock);
ASSERT(vd->vdev_isl2cache);
spa_aux_remove(vd, &spa_l2cache_avl);
vd->vdev_isl2cache = B_FALSE;
mutex_exit(&spa_l2cache_lock);
}
boolean_t
spa_l2cache_exists(uint64_t guid, uint64_t *pool)
{
boolean_t found;
mutex_enter(&spa_l2cache_lock);
found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl);
mutex_exit(&spa_l2cache_lock);
return (found);
}
void
spa_l2cache_activate(vdev_t *vd)
{
mutex_enter(&spa_l2cache_lock);
ASSERT(vd->vdev_isl2cache);
spa_aux_activate(vd, &spa_l2cache_avl);
mutex_exit(&spa_l2cache_lock);
}
/*
* ==========================================================================
* SPA vdev locking
* ==========================================================================
*/
/*
* Lock the given spa_t for the purpose of adding or removing a vdev.
* Grabs the global spa_namespace_lock plus the spa config lock for writing.
* It returns the next transaction group for the spa_t.
*/
uint64_t
spa_vdev_enter(spa_t *spa)
{
mutex_enter(&spa->spa_vdev_top_lock);
mutex_enter(&spa_namespace_lock);
vdev_autotrim_stop_all(spa);
return (spa_vdev_config_enter(spa));
}
/*
* The same as spa_vdev_enter() above but additionally takes the guid of
* the vdev being detached. When there is a rebuild in process it will be
* suspended while the vdev tree is modified then resumed by spa_vdev_exit().
* The rebuild is canceled if only a single child remains after the detach.
*/
uint64_t
spa_vdev_detach_enter(spa_t *spa, uint64_t guid)
{
mutex_enter(&spa->spa_vdev_top_lock);
mutex_enter(&spa_namespace_lock);
vdev_autotrim_stop_all(spa);
if (guid != 0) {
vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE);
if (vd) {
vdev_rebuild_stop_wait(vd->vdev_top);
}
}
return (spa_vdev_config_enter(spa));
}
/*
* Internal implementation for spa_vdev_enter(). Used when a vdev
* operation requires multiple syncs (i.e. removing a device) while
* keeping the spa_namespace_lock held.
*/
uint64_t
spa_vdev_config_enter(spa_t *spa)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);
return (spa_last_synced_txg(spa) + 1);
}
/*
* Used in combination with spa_vdev_config_enter() to allow the syncing
* of multiple transactions without releasing the spa_namespace_lock.
*/
void
spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, char *tag)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
int config_changed = B_FALSE;
ASSERT(txg > spa_last_synced_txg(spa));
spa->spa_pending_vdev = NULL;
/*
* Reassess the DTLs.
*/
vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE, B_FALSE);
if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) {
config_changed = B_TRUE;
spa->spa_config_generation++;
}
/*
* Verify the metaslab classes.
*/
ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0);
ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0);
ASSERT(metaslab_class_validate(spa_embedded_log_class(spa)) == 0);
ASSERT(metaslab_class_validate(spa_special_class(spa)) == 0);
ASSERT(metaslab_class_validate(spa_dedup_class(spa)) == 0);
spa_config_exit(spa, SCL_ALL, spa);
/*
* Panic the system if the specified tag requires it. This
* is useful for ensuring that configurations are updated
* transactionally.
*/
if (zio_injection_enabled)
zio_handle_panic_injection(spa, tag, 0);
/*
* Note: this txg_wait_synced() is important because it ensures
* that there won't be more than one config change per txg.
* This allows us to use the txg as the generation number.
*/
if (error == 0)
txg_wait_synced(spa->spa_dsl_pool, txg);
if (vd != NULL) {
ASSERT(!vd->vdev_detached || vd->vdev_dtl_sm == NULL);
if (vd->vdev_ops->vdev_op_leaf) {
mutex_enter(&vd->vdev_initialize_lock);
vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED,
NULL);
mutex_exit(&vd->vdev_initialize_lock);
mutex_enter(&vd->vdev_trim_lock);
vdev_trim_stop(vd, VDEV_TRIM_CANCELED, NULL);
mutex_exit(&vd->vdev_trim_lock);
}
/*
* The vdev may be both a leaf and top-level device.
*/
vdev_autotrim_stop_wait(vd);
spa_config_enter(spa, SCL_STATE_ALL, spa, RW_WRITER);
vdev_free(vd);
spa_config_exit(spa, SCL_STATE_ALL, spa);
}
/*
* If the config changed, update the config cache.
*/
if (config_changed)
spa_write_cachefile(spa, B_FALSE, B_TRUE);
}
/*
* Unlock the spa_t after adding or removing a vdev. Besides undoing the
* locking of spa_vdev_enter(), we also want make sure the transactions have
* synced to disk, and then update the global configuration cache with the new
* information.
*/
int
spa_vdev_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error)
{
vdev_autotrim_restart(spa);
vdev_rebuild_restart(spa);
spa_vdev_config_exit(spa, vd, txg, error, FTAG);
mutex_exit(&spa_namespace_lock);
mutex_exit(&spa->spa_vdev_top_lock);
return (error);
}
/*
* Lock the given spa_t for the purpose of changing vdev state.
*/
void
spa_vdev_state_enter(spa_t *spa, int oplocks)
{
int locks = SCL_STATE_ALL | oplocks;
/*
* Root pools may need to read of the underlying devfs filesystem
* when opening up a vdev. Unfortunately if we're holding the
* SCL_ZIO lock it will result in a deadlock when we try to issue
* the read from the root filesystem. Instead we "prefetch"
* the associated vnodes that we need prior to opening the
* underlying devices and cache them so that we can prevent
* any I/O when we are doing the actual open.
*/
if (spa_is_root(spa)) {
int low = locks & ~(SCL_ZIO - 1);
int high = locks & ~low;
spa_config_enter(spa, high, spa, RW_WRITER);
vdev_hold(spa->spa_root_vdev);
spa_config_enter(spa, low, spa, RW_WRITER);
} else {
spa_config_enter(spa, locks, spa, RW_WRITER);
}
spa->spa_vdev_locks = locks;
}
int
spa_vdev_state_exit(spa_t *spa, vdev_t *vd, int error)
{
boolean_t config_changed = B_FALSE;
vdev_t *vdev_top;
if (vd == NULL || vd == spa->spa_root_vdev) {
vdev_top = spa->spa_root_vdev;
} else {
vdev_top = vd->vdev_top;
}
if (vd != NULL || error == 0)
vdev_dtl_reassess(vdev_top, 0, 0, B_FALSE, B_FALSE);
if (vd != NULL) {
if (vd != spa->spa_root_vdev)
vdev_state_dirty(vdev_top);
config_changed = B_TRUE;
spa->spa_config_generation++;
}
if (spa_is_root(spa))
vdev_rele(spa->spa_root_vdev);
ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL);
spa_config_exit(spa, spa->spa_vdev_locks, spa);
/*
* If anything changed, wait for it to sync. This ensures that,
* from the system administrator's perspective, zpool(8) commands
* are synchronous. This is important for things like zpool offline:
* when the command completes, you expect no further I/O from ZFS.
*/
if (vd != NULL)
txg_wait_synced(spa->spa_dsl_pool, 0);
/*
* If the config changed, update the config cache.
*/
if (config_changed) {
mutex_enter(&spa_namespace_lock);
spa_write_cachefile(spa, B_FALSE, B_TRUE);
mutex_exit(&spa_namespace_lock);
}
return (error);
}
/*
* ==========================================================================
* Miscellaneous functions
* ==========================================================================
*/
void
spa_activate_mos_feature(spa_t *spa, const char *feature, dmu_tx_t *tx)
{
if (!nvlist_exists(spa->spa_label_features, feature)) {
fnvlist_add_boolean(spa->spa_label_features, feature);
/*
* When we are creating the pool (tx_txg==TXG_INITIAL), we can't
* dirty the vdev config because lock SCL_CONFIG is not held.
* Thankfully, in this case we don't need to dirty the config
* because it will be written out anyway when we finish
* creating the pool.
*/
if (tx->tx_txg != TXG_INITIAL)
vdev_config_dirty(spa->spa_root_vdev);
}
}
void
spa_deactivate_mos_feature(spa_t *spa, const char *feature)
{
if (nvlist_remove_all(spa->spa_label_features, feature) == 0)
vdev_config_dirty(spa->spa_root_vdev);
}
/*
* Return the spa_t associated with given pool_guid, if it exists. If
* device_guid is non-zero, determine whether the pool exists *and* contains
* a device with the specified device_guid.
*/
spa_t *
spa_by_guid(uint64_t pool_guid, uint64_t device_guid)
{
spa_t *spa;
avl_tree_t *t = &spa_namespace_avl;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) {
if (spa->spa_state == POOL_STATE_UNINITIALIZED)
continue;
if (spa->spa_root_vdev == NULL)
continue;
if (spa_guid(spa) == pool_guid) {
if (device_guid == 0)
break;
if (vdev_lookup_by_guid(spa->spa_root_vdev,
device_guid) != NULL)
break;
/*
* Check any devices we may be in the process of adding.
*/
if (spa->spa_pending_vdev) {
if (vdev_lookup_by_guid(spa->spa_pending_vdev,
device_guid) != NULL)
break;
}
}
}
return (spa);
}
/*
* Determine whether a pool with the given pool_guid exists.
*/
boolean_t
spa_guid_exists(uint64_t pool_guid, uint64_t device_guid)
{
return (spa_by_guid(pool_guid, device_guid) != NULL);
}
char *
spa_strdup(const char *s)
{
size_t len;
char *new;
len = strlen(s);
new = kmem_alloc(len + 1, KM_SLEEP);
bcopy(s, new, len);
new[len] = '\0';
return (new);
}
void
spa_strfree(char *s)
{
kmem_free(s, strlen(s) + 1);
}
uint64_t
spa_generate_guid(spa_t *spa)
{
uint64_t guid;
if (spa != NULL) {
do {
(void) random_get_pseudo_bytes((void *)&guid,
sizeof (guid));
} while (guid == 0 || spa_guid_exists(spa_guid(spa), guid));
} else {
do {
(void) random_get_pseudo_bytes((void *)&guid,
sizeof (guid));
} while (guid == 0 || spa_guid_exists(guid, 0));
}
return (guid);
}
void
snprintf_blkptr(char *buf, size_t buflen, const blkptr_t *bp)
{
char type[256];
char *checksum = NULL;
char *compress = NULL;
if (bp != NULL) {
if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) {
dmu_object_byteswap_t bswap =
DMU_OT_BYTESWAP(BP_GET_TYPE(bp));
(void) snprintf(type, sizeof (type), "bswap %s %s",
DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ?
"metadata" : "data",
dmu_ot_byteswap[bswap].ob_name);
} else {
(void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name,
sizeof (type));
}
if (!BP_IS_EMBEDDED(bp)) {
checksum =
zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name;
}
compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name;
}
SNPRINTF_BLKPTR(snprintf, ' ', buf, buflen, bp, type, checksum,
compress);
}
void
spa_freeze(spa_t *spa)
{
uint64_t freeze_txg = 0;
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
if (spa->spa_freeze_txg == UINT64_MAX) {
freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE;
spa->spa_freeze_txg = freeze_txg;
}
spa_config_exit(spa, SCL_ALL, FTAG);
if (freeze_txg != 0)
txg_wait_synced(spa_get_dsl(spa), freeze_txg);
}
void
zfs_panic_recover(const char *fmt, ...)
{
va_list adx;
va_start(adx, fmt);
vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx);
va_end(adx);
}
/*
* This is a stripped-down version of strtoull, suitable only for converting
* lowercase hexadecimal numbers that don't overflow.
*/
uint64_t
zfs_strtonum(const char *str, char **nptr)
{
uint64_t val = 0;
char c;
int digit;
while ((c = *str) != '\0') {
if (c >= '0' && c <= '9')
digit = c - '0';
else if (c >= 'a' && c <= 'f')
digit = 10 + c - 'a';
else
break;
val *= 16;
val += digit;
str++;
}
if (nptr)
*nptr = (char *)str;
return (val);
}
void
spa_activate_allocation_classes(spa_t *spa, dmu_tx_t *tx)
{
/*
* We bump the feature refcount for each special vdev added to the pool
*/
ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES));
spa_feature_incr(spa, SPA_FEATURE_ALLOCATION_CLASSES, tx);
}
/*
* ==========================================================================
* Accessor functions
* ==========================================================================
*/
boolean_t
spa_shutting_down(spa_t *spa)
{
return (spa->spa_async_suspended);
}
dsl_pool_t *
spa_get_dsl(spa_t *spa)
{
return (spa->spa_dsl_pool);
}
boolean_t
spa_is_initializing(spa_t *spa)
{
return (spa->spa_is_initializing);
}
boolean_t
spa_indirect_vdevs_loaded(spa_t *spa)
{
return (spa->spa_indirect_vdevs_loaded);
}
blkptr_t *
spa_get_rootblkptr(spa_t *spa)
{
return (&spa->spa_ubsync.ub_rootbp);
}
void
spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp)
{
spa->spa_uberblock.ub_rootbp = *bp;
}
void
spa_altroot(spa_t *spa, char *buf, size_t buflen)
{
if (spa->spa_root == NULL)
buf[0] = '\0';
else
(void) strncpy(buf, spa->spa_root, buflen);
}
int
spa_sync_pass(spa_t *spa)
{
return (spa->spa_sync_pass);
}
char *
spa_name(spa_t *spa)
{
return (spa->spa_name);
}
uint64_t
spa_guid(spa_t *spa)
{
dsl_pool_t *dp = spa_get_dsl(spa);
uint64_t guid;
/*
* If we fail to parse the config during spa_load(), we can go through
* the error path (which posts an ereport) and end up here with no root
* vdev. We stash the original pool guid in 'spa_config_guid' to handle
* this case.
*/
if (spa->spa_root_vdev == NULL)
return (spa->spa_config_guid);
guid = spa->spa_last_synced_guid != 0 ?
spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid;
/*
* Return the most recently synced out guid unless we're
* in syncing context.
*/
if (dp && dsl_pool_sync_context(dp))
return (spa->spa_root_vdev->vdev_guid);
else
return (guid);
}
uint64_t
spa_load_guid(spa_t *spa)
{
/*
* This is a GUID that exists solely as a reference for the
* purposes of the arc. It is generated at load time, and
* is never written to persistent storage.
*/
return (spa->spa_load_guid);
}
uint64_t
spa_last_synced_txg(spa_t *spa)
{
return (spa->spa_ubsync.ub_txg);
}
uint64_t
spa_first_txg(spa_t *spa)
{
return (spa->spa_first_txg);
}
uint64_t
spa_syncing_txg(spa_t *spa)
{
return (spa->spa_syncing_txg);
}
/*
* Return the last txg where data can be dirtied. The final txgs
* will be used to just clear out any deferred frees that remain.
*/
uint64_t
spa_final_dirty_txg(spa_t *spa)
{
return (spa->spa_final_txg - TXG_DEFER_SIZE);
}
pool_state_t
spa_state(spa_t *spa)
{
return (spa->spa_state);
}
spa_load_state_t
spa_load_state(spa_t *spa)
{
return (spa->spa_load_state);
}
uint64_t
spa_freeze_txg(spa_t *spa)
{
return (spa->spa_freeze_txg);
}
/*
* Return the inflated asize for a logical write in bytes. This is used by the
* DMU to calculate the space a logical write will require on disk.
* If lsize is smaller than the largest physical block size allocatable on this
* pool we use its value instead, since the write will end up using the whole
* block anyway.
*/
uint64_t
spa_get_worst_case_asize(spa_t *spa, uint64_t lsize)
{
if (lsize == 0)
return (0); /* No inflation needed */
return (MAX(lsize, 1 << spa->spa_max_ashift) * spa_asize_inflation);
}
/*
* Return the amount of slop space in bytes. It is typically 1/32 of the pool
* (3.2%), minus the embedded log space. On very small pools, it may be
* slightly larger than this. On very large pools, it will be capped to
* the value of spa_max_slop. The embedded log space is not included in
* spa_dspace. By subtracting it, the usable space (per "zfs list") is a
* constant 97% of the total space, regardless of metaslab size (assuming the
* default spa_slop_shift=5 and a non-tiny pool).
*
* See the comment above spa_slop_shift for more details.
*/
uint64_t
spa_get_slop_space(spa_t *spa)
{
uint64_t space = 0;
uint64_t slop = 0;
/*
* Make sure spa_dedup_dspace has been set.
*/
if (spa->spa_dedup_dspace == ~0ULL)
spa_update_dspace(spa);
/*
* spa_get_dspace() includes the space only logically "used" by
* deduplicated data, so since it's not useful to reserve more
* space with more deduplicated data, we subtract that out here.
*/
space = spa_get_dspace(spa) - spa->spa_dedup_dspace;
slop = MIN(space >> spa_slop_shift, spa_max_slop);
/*
* Subtract the embedded log space, but no more than half the (3.2%)
* unusable space. Note, the "no more than half" is only relevant if
* zfs_embedded_slog_min_ms >> spa_slop_shift < 2, which is not true by
* default.
*/
uint64_t embedded_log =
metaslab_class_get_dspace(spa_embedded_log_class(spa));
slop -= MIN(embedded_log, slop >> 1);
/*
* Slop space should be at least spa_min_slop, but no more than half
* the entire pool.
*/
slop = MAX(slop, MIN(space >> 1, spa_min_slop));
return (slop);
}
uint64_t
spa_get_dspace(spa_t *spa)
{
return (spa->spa_dspace);
}
uint64_t
spa_get_checkpoint_space(spa_t *spa)
{
return (spa->spa_checkpoint_info.sci_dspace);
}
void
spa_update_dspace(spa_t *spa)
{
spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) +
ddt_get_dedup_dspace(spa);
if (spa->spa_nonallocating_dspace > 0) {
/*
* Subtract the space provided by all non-allocating vdevs that
* contribute to dspace. If a file is overwritten, its old
* blocks are freed and new blocks are allocated. If there are
* no snapshots of the file, the available space should remain
* the same. The old blocks could be freed from the
* non-allocating vdev, but the new blocks must be allocated on
* other (allocating) vdevs. By reserving the entire size of
* the non-allocating vdevs (including allocated space), we
* ensure that there will be enough space on the allocating
* vdevs for this file overwrite to succeed.
*
* Note that the DMU/DSL doesn't actually know or care
* how much space is allocated (it does its own tracking
* of how much space has been logically used). So it
* doesn't matter that the data we are moving may be
* allocated twice (on the old device and the new device).
*/
ASSERT3U(spa->spa_dspace, >=, spa->spa_nonallocating_dspace);
spa->spa_dspace -= spa->spa_nonallocating_dspace;
}
}
/*
* Return the failure mode that has been set to this pool. The default
* behavior will be to block all I/Os when a complete failure occurs.
*/
uint64_t
spa_get_failmode(spa_t *spa)
{
return (spa->spa_failmode);
}
boolean_t
spa_suspended(spa_t *spa)
{
return (spa->spa_suspended != ZIO_SUSPEND_NONE);
}
uint64_t
spa_version(spa_t *spa)
{
return (spa->spa_ubsync.ub_version);
}
boolean_t
spa_deflate(spa_t *spa)
{
return (spa->spa_deflate);
}
metaslab_class_t *
spa_normal_class(spa_t *spa)
{
return (spa->spa_normal_class);
}
metaslab_class_t *
spa_log_class(spa_t *spa)
{
return (spa->spa_log_class);
}
metaslab_class_t *
spa_embedded_log_class(spa_t *spa)
{
return (spa->spa_embedded_log_class);
}
metaslab_class_t *
spa_special_class(spa_t *spa)
{
return (spa->spa_special_class);
}
metaslab_class_t *
spa_dedup_class(spa_t *spa)
{
return (spa->spa_dedup_class);
}
/*
* Locate an appropriate allocation class
*/
metaslab_class_t *
spa_preferred_class(spa_t *spa, uint64_t size, dmu_object_type_t objtype,
uint_t level, uint_t special_smallblk)
{
/*
* ZIL allocations determine their class in zio_alloc_zil().
*/
ASSERT(objtype != DMU_OT_INTENT_LOG);
boolean_t has_special_class = spa->spa_special_class->mc_groups != 0;
if (DMU_OT_IS_DDT(objtype)) {
if (spa->spa_dedup_class->mc_groups != 0)
return (spa_dedup_class(spa));
else if (has_special_class && zfs_ddt_data_is_special)
return (spa_special_class(spa));
else
return (spa_normal_class(spa));
}
/* Indirect blocks for user data can land in special if allowed */
if (level > 0 && (DMU_OT_IS_FILE(objtype) || objtype == DMU_OT_ZVOL)) {
if (has_special_class && zfs_user_indirect_is_special)
return (spa_special_class(spa));
else
return (spa_normal_class(spa));
}
if (DMU_OT_IS_METADATA(objtype) || level > 0) {
if (has_special_class)
return (spa_special_class(spa));
else
return (spa_normal_class(spa));
}
/*
* Allow small file blocks in special class in some cases (like
* for the dRAID vdev feature). But always leave a reserve of
* zfs_special_class_metadata_reserve_pct exclusively for metadata.
*/
if (DMU_OT_IS_FILE(objtype) &&
has_special_class && size <= special_smallblk) {
metaslab_class_t *special = spa_special_class(spa);
uint64_t alloc = metaslab_class_get_alloc(special);
uint64_t space = metaslab_class_get_space(special);
uint64_t limit =
(space * (100 - zfs_special_class_metadata_reserve_pct))
/ 100;
if (alloc < limit)
return (special);
}
return (spa_normal_class(spa));
}
void
spa_evicting_os_register(spa_t *spa, objset_t *os)
{
mutex_enter(&spa->spa_evicting_os_lock);
list_insert_head(&spa->spa_evicting_os_list, os);
mutex_exit(&spa->spa_evicting_os_lock);
}
void
spa_evicting_os_deregister(spa_t *spa, objset_t *os)
{
mutex_enter(&spa->spa_evicting_os_lock);
list_remove(&spa->spa_evicting_os_list, os);
cv_broadcast(&spa->spa_evicting_os_cv);
mutex_exit(&spa->spa_evicting_os_lock);
}
void
spa_evicting_os_wait(spa_t *spa)
{
mutex_enter(&spa->spa_evicting_os_lock);
while (!list_is_empty(&spa->spa_evicting_os_list))
cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock);
mutex_exit(&spa->spa_evicting_os_lock);
dmu_buf_user_evict_wait();
}
int
spa_max_replication(spa_t *spa)
{
/*
* As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to
* handle BPs with more than one DVA allocated. Set our max
* replication level accordingly.
*/
if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS)
return (1);
return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override));
}
int
spa_prev_software_version(spa_t *spa)
{
return (spa->spa_prev_software_version);
}
uint64_t
spa_deadman_synctime(spa_t *spa)
{
return (spa->spa_deadman_synctime);
}
spa_autotrim_t
spa_get_autotrim(spa_t *spa)
{
return (spa->spa_autotrim);
}
uint64_t
spa_deadman_ziotime(spa_t *spa)
{
return (spa->spa_deadman_ziotime);
}
uint64_t
spa_get_deadman_failmode(spa_t *spa)
{
return (spa->spa_deadman_failmode);
}
void
spa_set_deadman_failmode(spa_t *spa, const char *failmode)
{
if (strcmp(failmode, "wait") == 0)
spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT;
else if (strcmp(failmode, "continue") == 0)
spa->spa_deadman_failmode = ZIO_FAILURE_MODE_CONTINUE;
else if (strcmp(failmode, "panic") == 0)
spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC;
else
spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT;
}
void
spa_set_deadman_ziotime(hrtime_t ns)
{
spa_t *spa = NULL;
if (spa_mode_global != SPA_MODE_UNINIT) {
mutex_enter(&spa_namespace_lock);
while ((spa = spa_next(spa)) != NULL)
spa->spa_deadman_ziotime = ns;
mutex_exit(&spa_namespace_lock);
}
}
void
spa_set_deadman_synctime(hrtime_t ns)
{
spa_t *spa = NULL;
if (spa_mode_global != SPA_MODE_UNINIT) {
mutex_enter(&spa_namespace_lock);
while ((spa = spa_next(spa)) != NULL)
spa->spa_deadman_synctime = ns;
mutex_exit(&spa_namespace_lock);
}
}
uint64_t
dva_get_dsize_sync(spa_t *spa, const dva_t *dva)
{
uint64_t asize = DVA_GET_ASIZE(dva);
uint64_t dsize = asize;
ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
if (asize != 0 && spa->spa_deflate) {
vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));
if (vd != NULL)
dsize = (asize >> SPA_MINBLOCKSHIFT) *
vd->vdev_deflate_ratio;
}
return (dsize);
}
uint64_t
bp_get_dsize_sync(spa_t *spa, const blkptr_t *bp)
{
uint64_t dsize = 0;
for (int d = 0; d < BP_GET_NDVAS(bp); d++)
dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);
return (dsize);
}
uint64_t
bp_get_dsize(spa_t *spa, const blkptr_t *bp)
{
uint64_t dsize = 0;
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
for (int d = 0; d < BP_GET_NDVAS(bp); d++)
dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);
spa_config_exit(spa, SCL_VDEV, FTAG);
return (dsize);
}
uint64_t
spa_dirty_data(spa_t *spa)
{
return (spa->spa_dsl_pool->dp_dirty_total);
}
/*
* ==========================================================================
* SPA Import Progress Routines
* ==========================================================================
*/
typedef struct spa_import_progress {
uint64_t pool_guid; /* unique id for updates */
char *pool_name;
spa_load_state_t spa_load_state;
uint64_t mmp_sec_remaining; /* MMP activity check */
uint64_t spa_load_max_txg; /* rewind txg */
procfs_list_node_t smh_node;
} spa_import_progress_t;
spa_history_list_t *spa_import_progress_list = NULL;
static int
spa_import_progress_show_header(struct seq_file *f)
{
seq_printf(f, "%-20s %-14s %-14s %-12s %s\n", "pool_guid",
"load_state", "multihost_secs", "max_txg",
"pool_name");
return (0);
}
static int
spa_import_progress_show(struct seq_file *f, void *data)
{
spa_import_progress_t *sip = (spa_import_progress_t *)data;
seq_printf(f, "%-20llu %-14llu %-14llu %-12llu %s\n",
(u_longlong_t)sip->pool_guid, (u_longlong_t)sip->spa_load_state,
(u_longlong_t)sip->mmp_sec_remaining,
(u_longlong_t)sip->spa_load_max_txg,
(sip->pool_name ? sip->pool_name : "-"));
return (0);
}
/* Remove oldest elements from list until there are no more than 'size' left */
static void
spa_import_progress_truncate(spa_history_list_t *shl, unsigned int size)
{
spa_import_progress_t *sip;
while (shl->size > size) {
sip = list_remove_head(&shl->procfs_list.pl_list);
if (sip->pool_name)
spa_strfree(sip->pool_name);
kmem_free(sip, sizeof (spa_import_progress_t));
shl->size--;
}
IMPLY(size == 0, list_is_empty(&shl->procfs_list.pl_list));
}
static void
spa_import_progress_init(void)
{
spa_import_progress_list = kmem_zalloc(sizeof (spa_history_list_t),
KM_SLEEP);
spa_import_progress_list->size = 0;
spa_import_progress_list->procfs_list.pl_private =
spa_import_progress_list;
procfs_list_install("zfs",
NULL,
"import_progress",
0644,
&spa_import_progress_list->procfs_list,
spa_import_progress_show,
spa_import_progress_show_header,
NULL,
offsetof(spa_import_progress_t, smh_node));
}
static void
spa_import_progress_destroy(void)
{
spa_history_list_t *shl = spa_import_progress_list;
procfs_list_uninstall(&shl->procfs_list);
spa_import_progress_truncate(shl, 0);
procfs_list_destroy(&shl->procfs_list);
kmem_free(shl, sizeof (spa_history_list_t));
}
int
spa_import_progress_set_state(uint64_t pool_guid,
spa_load_state_t load_state)
{
spa_history_list_t *shl = spa_import_progress_list;
spa_import_progress_t *sip;
int error = ENOENT;
if (shl->size == 0)
return (0);
mutex_enter(&shl->procfs_list.pl_lock);
for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
sip = list_prev(&shl->procfs_list.pl_list, sip)) {
if (sip->pool_guid == pool_guid) {
sip->spa_load_state = load_state;
error = 0;
break;
}
}
mutex_exit(&shl->procfs_list.pl_lock);
return (error);
}
int
spa_import_progress_set_max_txg(uint64_t pool_guid, uint64_t load_max_txg)
{
spa_history_list_t *shl = spa_import_progress_list;
spa_import_progress_t *sip;
int error = ENOENT;
if (shl->size == 0)
return (0);
mutex_enter(&shl->procfs_list.pl_lock);
for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
sip = list_prev(&shl->procfs_list.pl_list, sip)) {
if (sip->pool_guid == pool_guid) {
sip->spa_load_max_txg = load_max_txg;
error = 0;
break;
}
}
mutex_exit(&shl->procfs_list.pl_lock);
return (error);
}
int
spa_import_progress_set_mmp_check(uint64_t pool_guid,
uint64_t mmp_sec_remaining)
{
spa_history_list_t *shl = spa_import_progress_list;
spa_import_progress_t *sip;
int error = ENOENT;
if (shl->size == 0)
return (0);
mutex_enter(&shl->procfs_list.pl_lock);
for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
sip = list_prev(&shl->procfs_list.pl_list, sip)) {
if (sip->pool_guid == pool_guid) {
sip->mmp_sec_remaining = mmp_sec_remaining;
error = 0;
break;
}
}
mutex_exit(&shl->procfs_list.pl_lock);
return (error);
}
/*
* A new import is in progress, add an entry.
*/
void
spa_import_progress_add(spa_t *spa)
{
spa_history_list_t *shl = spa_import_progress_list;
spa_import_progress_t *sip;
char *poolname = NULL;
sip = kmem_zalloc(sizeof (spa_import_progress_t), KM_SLEEP);
sip->pool_guid = spa_guid(spa);
(void) nvlist_lookup_string(spa->spa_config, ZPOOL_CONFIG_POOL_NAME,
&poolname);
if (poolname == NULL)
poolname = spa_name(spa);
sip->pool_name = spa_strdup(poolname);
sip->spa_load_state = spa_load_state(spa);
mutex_enter(&shl->procfs_list.pl_lock);
procfs_list_add(&shl->procfs_list, sip);
shl->size++;
mutex_exit(&shl->procfs_list.pl_lock);
}
void
spa_import_progress_remove(uint64_t pool_guid)
{
spa_history_list_t *shl = spa_import_progress_list;
spa_import_progress_t *sip;
mutex_enter(&shl->procfs_list.pl_lock);
for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
sip = list_prev(&shl->procfs_list.pl_list, sip)) {
if (sip->pool_guid == pool_guid) {
if (sip->pool_name)
spa_strfree(sip->pool_name);
list_remove(&shl->procfs_list.pl_list, sip);
shl->size--;
kmem_free(sip, sizeof (spa_import_progress_t));
break;
}
}
mutex_exit(&shl->procfs_list.pl_lock);
}
/*
* ==========================================================================
* Initialization and Termination
* ==========================================================================
*/
static int
spa_name_compare(const void *a1, const void *a2)
{
const spa_t *s1 = a1;
const spa_t *s2 = a2;
int s;
s = strcmp(s1->spa_name, s2->spa_name);
return (TREE_ISIGN(s));
}
void
spa_boot_init(void)
{
spa_config_load();
}
void
spa_init(spa_mode_t mode)
{
mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL);
avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t),
offsetof(spa_t, spa_avl));
avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t),
offsetof(spa_aux_t, aux_avl));
avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t),
offsetof(spa_aux_t, aux_avl));
spa_mode_global = mode;
#ifndef _KERNEL
if (spa_mode_global != SPA_MODE_READ && dprintf_find_string("watch")) {
struct sigaction sa;
sa.sa_flags = SA_SIGINFO;
sigemptyset(&sa.sa_mask);
sa.sa_sigaction = arc_buf_sigsegv;
if (sigaction(SIGSEGV, &sa, NULL) == -1) {
perror("could not enable watchpoints: "
"sigaction(SIGSEGV, ...) = ");
} else {
arc_watch = B_TRUE;
}
}
#endif
fm_init();
zfs_refcount_init();
unique_init();
zfs_btree_init();
metaslab_stat_init();
ddt_init();
zio_init();
dmu_init();
zil_init();
vdev_cache_stat_init();
vdev_mirror_stat_init();
vdev_raidz_math_init();
vdev_file_init();
zfs_prop_init();
zpool_prop_init();
zpool_feature_init();
spa_config_load();
vdev_prop_init();
l2arc_start();
scan_init();
qat_init();
spa_import_progress_init();
}
void
spa_fini(void)
{
l2arc_stop();
spa_evict_all();
vdev_file_fini();
vdev_cache_stat_fini();
vdev_mirror_stat_fini();
vdev_raidz_math_fini();
zil_fini();
dmu_fini();
zio_fini();
ddt_fini();
metaslab_stat_fini();
zfs_btree_fini();
unique_fini();
zfs_refcount_fini();
fm_fini();
scan_fini();
qat_fini();
spa_import_progress_destroy();
avl_destroy(&spa_namespace_avl);
avl_destroy(&spa_spare_avl);
avl_destroy(&spa_l2cache_avl);
cv_destroy(&spa_namespace_cv);
mutex_destroy(&spa_namespace_lock);
mutex_destroy(&spa_spare_lock);
mutex_destroy(&spa_l2cache_lock);
}
/*
* Return whether this pool has a dedicated slog device. No locking needed.
* It's not a problem if the wrong answer is returned as it's only for
* performance and not correctness.
*/
boolean_t
spa_has_slogs(spa_t *spa)
{
return (spa->spa_log_class->mc_groups != 0);
}
spa_log_state_t
spa_get_log_state(spa_t *spa)
{
return (spa->spa_log_state);
}
void
spa_set_log_state(spa_t *spa, spa_log_state_t state)
{
spa->spa_log_state = state;
}
boolean_t
spa_is_root(spa_t *spa)
{
return (spa->spa_is_root);
}
boolean_t
spa_writeable(spa_t *spa)
{
return (!!(spa->spa_mode & SPA_MODE_WRITE) && spa->spa_trust_config);
}
/*
* Returns true if there is a pending sync task in any of the current
* syncing txg, the current quiescing txg, or the current open txg.
*/
boolean_t
spa_has_pending_synctask(spa_t *spa)
{
return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks) ||
!txg_all_lists_empty(&spa->spa_dsl_pool->dp_early_sync_tasks));
}
spa_mode_t
spa_mode(spa_t *spa)
{
return (spa->spa_mode);
}
uint64_t
spa_bootfs(spa_t *spa)
{
return (spa->spa_bootfs);
}
uint64_t
spa_delegation(spa_t *spa)
{
return (spa->spa_delegation);
}
objset_t *
spa_meta_objset(spa_t *spa)
{
return (spa->spa_meta_objset);
}
enum zio_checksum
spa_dedup_checksum(spa_t *spa)
{
return (spa->spa_dedup_checksum);
}
/*
* Reset pool scan stat per scan pass (or reboot).
*/
void
spa_scan_stat_init(spa_t *spa)
{
/* data not stored on disk */
spa->spa_scan_pass_start = gethrestime_sec();
if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan))
spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start;
else
spa->spa_scan_pass_scrub_pause = 0;
spa->spa_scan_pass_scrub_spent_paused = 0;
spa->spa_scan_pass_exam = 0;
spa->spa_scan_pass_issued = 0;
vdev_scan_stat_init(spa->spa_root_vdev);
}
/*
* Get scan stats for zpool status reports
*/
int
spa_scan_get_stats(spa_t *spa, pool_scan_stat_t *ps)
{
dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL;
if (scn == NULL || scn->scn_phys.scn_func == POOL_SCAN_NONE)
return (SET_ERROR(ENOENT));
bzero(ps, sizeof (pool_scan_stat_t));
/* data stored on disk */
ps->pss_func = scn->scn_phys.scn_func;
ps->pss_state = scn->scn_phys.scn_state;
ps->pss_start_time = scn->scn_phys.scn_start_time;
ps->pss_end_time = scn->scn_phys.scn_end_time;
ps->pss_to_examine = scn->scn_phys.scn_to_examine;
ps->pss_examined = scn->scn_phys.scn_examined;
ps->pss_to_process = scn->scn_phys.scn_to_process;
ps->pss_processed = scn->scn_phys.scn_processed;
ps->pss_errors = scn->scn_phys.scn_errors;
/* data not stored on disk */
ps->pss_pass_exam = spa->spa_scan_pass_exam;
ps->pss_pass_start = spa->spa_scan_pass_start;
ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause;
ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused;
ps->pss_pass_issued = spa->spa_scan_pass_issued;
ps->pss_issued =
scn->scn_issued_before_pass + spa->spa_scan_pass_issued;
return (0);
}
int
spa_maxblocksize(spa_t *spa)
{
if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS))
return (SPA_MAXBLOCKSIZE);
else
return (SPA_OLD_MAXBLOCKSIZE);
}
/*
* Returns the txg that the last device removal completed. No indirect mappings
* have been added since this txg.
*/
uint64_t
spa_get_last_removal_txg(spa_t *spa)
{
uint64_t vdevid;
uint64_t ret = -1ULL;
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
/*
* sr_prev_indirect_vdev is only modified while holding all the
* config locks, so it is sufficient to hold SCL_VDEV as reader when
* examining it.
*/
vdevid = spa->spa_removing_phys.sr_prev_indirect_vdev;
while (vdevid != -1ULL) {
vdev_t *vd = vdev_lookup_top(spa, vdevid);
vdev_indirect_births_t *vib = vd->vdev_indirect_births;
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
/*
* If the removal did not remap any data, we don't care.
*/
if (vdev_indirect_births_count(vib) != 0) {
ret = vdev_indirect_births_last_entry_txg(vib);
break;
}
vdevid = vd->vdev_indirect_config.vic_prev_indirect_vdev;
}
spa_config_exit(spa, SCL_VDEV, FTAG);
IMPLY(ret != -1ULL,
spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL));
return (ret);
}
int
spa_maxdnodesize(spa_t *spa)
{
if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE))
return (DNODE_MAX_SIZE);
else
return (DNODE_MIN_SIZE);
}
boolean_t
spa_multihost(spa_t *spa)
{
return (spa->spa_multihost ? B_TRUE : B_FALSE);
}
uint32_t
spa_get_hostid(spa_t *spa)
{
return (spa->spa_hostid);
}
boolean_t
spa_trust_config(spa_t *spa)
{
return (spa->spa_trust_config);
}
uint64_t
spa_missing_tvds_allowed(spa_t *spa)
{
return (spa->spa_missing_tvds_allowed);
}
space_map_t *
spa_syncing_log_sm(spa_t *spa)
{
return (spa->spa_syncing_log_sm);
}
void
spa_set_missing_tvds(spa_t *spa, uint64_t missing)
{
spa->spa_missing_tvds = missing;
}
/*
* Return the pool state string ("ONLINE", "DEGRADED", "SUSPENDED", etc).
*/
const char *
spa_state_to_name(spa_t *spa)
{
ASSERT3P(spa, !=, NULL);
/*
* it is possible for the spa to exist, without root vdev
* as the spa transitions during import/export
*/
vdev_t *rvd = spa->spa_root_vdev;
if (rvd == NULL) {
return ("TRANSITIONING");
}
vdev_state_t state = rvd->vdev_state;
vdev_aux_t aux = rvd->vdev_stat.vs_aux;
if (spa_suspended(spa) &&
(spa_get_failmode(spa) != ZIO_FAILURE_MODE_CONTINUE))
return ("SUSPENDED");
switch (state) {
case VDEV_STATE_CLOSED:
case VDEV_STATE_OFFLINE:
return ("OFFLINE");
case VDEV_STATE_REMOVED:
return ("REMOVED");
case VDEV_STATE_CANT_OPEN:
if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
return ("FAULTED");
else if (aux == VDEV_AUX_SPLIT_POOL)
return ("SPLIT");
else
return ("UNAVAIL");
case VDEV_STATE_FAULTED:
return ("FAULTED");
case VDEV_STATE_DEGRADED:
return ("DEGRADED");
case VDEV_STATE_HEALTHY:
return ("ONLINE");
default:
break;
}
return ("UNKNOWN");
}
boolean_t
spa_top_vdevs_spacemap_addressable(spa_t *spa)
{
vdev_t *rvd = spa->spa_root_vdev;
for (uint64_t c = 0; c < rvd->vdev_children; c++) {
if (!vdev_is_spacemap_addressable(rvd->vdev_child[c]))
return (B_FALSE);
}
return (B_TRUE);
}
boolean_t
spa_has_checkpoint(spa_t *spa)
{
return (spa->spa_checkpoint_txg != 0);
}
boolean_t
spa_importing_readonly_checkpoint(spa_t *spa)
{
return ((spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT) &&
spa->spa_mode == SPA_MODE_READ);
}
uint64_t
spa_min_claim_txg(spa_t *spa)
{
uint64_t checkpoint_txg = spa->spa_uberblock.ub_checkpoint_txg;
if (checkpoint_txg != 0)
return (checkpoint_txg + 1);
return (spa->spa_first_txg);
}
/*
* If there is a checkpoint, async destroys may consume more space from
* the pool instead of freeing it. In an attempt to save the pool from
* getting suspended when it is about to run out of space, we stop
* processing async destroys.
*/
boolean_t
spa_suspend_async_destroy(spa_t *spa)
{
dsl_pool_t *dp = spa_get_dsl(spa);
uint64_t unreserved = dsl_pool_unreserved_space(dp,
ZFS_SPACE_CHECK_EXTRA_RESERVED);
uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes;
uint64_t avail = (unreserved > used) ? (unreserved - used) : 0;
if (spa_has_checkpoint(spa) && avail == 0)
return (B_TRUE);
return (B_FALSE);
}
#if defined(_KERNEL)
int
param_set_deadman_failmode_common(const char *val)
{
spa_t *spa = NULL;
char *p;
if (val == NULL)
return (SET_ERROR(EINVAL));
if ((p = strchr(val, '\n')) != NULL)
*p = '\0';
if (strcmp(val, "wait") != 0 && strcmp(val, "continue") != 0 &&
strcmp(val, "panic"))
return (SET_ERROR(EINVAL));
if (spa_mode_global != SPA_MODE_UNINIT) {
mutex_enter(&spa_namespace_lock);
while ((spa = spa_next(spa)) != NULL)
spa_set_deadman_failmode(spa, val);
mutex_exit(&spa_namespace_lock);
}
return (0);
}
#endif
/* Namespace manipulation */
EXPORT_SYMBOL(spa_lookup);
EXPORT_SYMBOL(spa_add);
EXPORT_SYMBOL(spa_remove);
EXPORT_SYMBOL(spa_next);
/* Refcount functions */
EXPORT_SYMBOL(spa_open_ref);
EXPORT_SYMBOL(spa_close);
EXPORT_SYMBOL(spa_refcount_zero);
/* Pool configuration lock */
EXPORT_SYMBOL(spa_config_tryenter);
EXPORT_SYMBOL(spa_config_enter);
EXPORT_SYMBOL(spa_config_exit);
EXPORT_SYMBOL(spa_config_held);
/* Pool vdev add/remove lock */
EXPORT_SYMBOL(spa_vdev_enter);
EXPORT_SYMBOL(spa_vdev_exit);
/* Pool vdev state change lock */
EXPORT_SYMBOL(spa_vdev_state_enter);
EXPORT_SYMBOL(spa_vdev_state_exit);
/* Accessor functions */
EXPORT_SYMBOL(spa_shutting_down);
EXPORT_SYMBOL(spa_get_dsl);
EXPORT_SYMBOL(spa_get_rootblkptr);
EXPORT_SYMBOL(spa_set_rootblkptr);
EXPORT_SYMBOL(spa_altroot);
EXPORT_SYMBOL(spa_sync_pass);
EXPORT_SYMBOL(spa_name);
EXPORT_SYMBOL(spa_guid);
EXPORT_SYMBOL(spa_last_synced_txg);
EXPORT_SYMBOL(spa_first_txg);
EXPORT_SYMBOL(spa_syncing_txg);
EXPORT_SYMBOL(spa_version);
EXPORT_SYMBOL(spa_state);
EXPORT_SYMBOL(spa_load_state);
EXPORT_SYMBOL(spa_freeze_txg);
EXPORT_SYMBOL(spa_get_dspace);
EXPORT_SYMBOL(spa_update_dspace);
EXPORT_SYMBOL(spa_deflate);
EXPORT_SYMBOL(spa_normal_class);
EXPORT_SYMBOL(spa_log_class);
EXPORT_SYMBOL(spa_special_class);
EXPORT_SYMBOL(spa_preferred_class);
EXPORT_SYMBOL(spa_max_replication);
EXPORT_SYMBOL(spa_prev_software_version);
EXPORT_SYMBOL(spa_get_failmode);
EXPORT_SYMBOL(spa_suspended);
EXPORT_SYMBOL(spa_bootfs);
EXPORT_SYMBOL(spa_delegation);
EXPORT_SYMBOL(spa_meta_objset);
EXPORT_SYMBOL(spa_maxblocksize);
EXPORT_SYMBOL(spa_maxdnodesize);
/* Miscellaneous support routines */
EXPORT_SYMBOL(spa_guid_exists);
EXPORT_SYMBOL(spa_strdup);
EXPORT_SYMBOL(spa_strfree);
EXPORT_SYMBOL(spa_generate_guid);
EXPORT_SYMBOL(snprintf_blkptr);
EXPORT_SYMBOL(spa_freeze);
EXPORT_SYMBOL(spa_upgrade);
EXPORT_SYMBOL(spa_evict_all);
EXPORT_SYMBOL(spa_lookup_by_guid);
EXPORT_SYMBOL(spa_has_spare);
EXPORT_SYMBOL(dva_get_dsize_sync);
EXPORT_SYMBOL(bp_get_dsize_sync);
EXPORT_SYMBOL(bp_get_dsize);
EXPORT_SYMBOL(spa_has_slogs);
EXPORT_SYMBOL(spa_is_root);
EXPORT_SYMBOL(spa_writeable);
EXPORT_SYMBOL(spa_mode);
EXPORT_SYMBOL(spa_namespace_lock);
EXPORT_SYMBOL(spa_trust_config);
EXPORT_SYMBOL(spa_missing_tvds_allowed);
EXPORT_SYMBOL(spa_set_missing_tvds);
EXPORT_SYMBOL(spa_state_to_name);
EXPORT_SYMBOL(spa_importing_readonly_checkpoint);
EXPORT_SYMBOL(spa_min_claim_txg);
EXPORT_SYMBOL(spa_suspend_async_destroy);
EXPORT_SYMBOL(spa_has_checkpoint);
EXPORT_SYMBOL(spa_top_vdevs_spacemap_addressable);
ZFS_MODULE_PARAM(zfs, zfs_, flags, UINT, ZMOD_RW,
"Set additional debugging flags");
ZFS_MODULE_PARAM(zfs, zfs_, recover, INT, ZMOD_RW,
"Set to attempt to recover from fatal errors");
ZFS_MODULE_PARAM(zfs, zfs_, free_leak_on_eio, INT, ZMOD_RW,
"Set to ignore IO errors during free and permanently leak the space");
ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, checktime_ms, ULONG, ZMOD_RW,
"Dead I/O check interval in milliseconds");
ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, enabled, INT, ZMOD_RW,
"Enable deadman timer");
ZFS_MODULE_PARAM(zfs_spa, spa_, asize_inflation, INT, ZMOD_RW,
"SPA size estimate multiplication factor");
ZFS_MODULE_PARAM(zfs, zfs_, ddt_data_is_special, INT, ZMOD_RW,
"Place DDT data into the special class");
ZFS_MODULE_PARAM(zfs, zfs_, user_indirect_is_special, INT, ZMOD_RW,
"Place user data indirect blocks into the special class");
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, failmode,
param_set_deadman_failmode, param_get_charp, ZMOD_RW,
"Failmode for deadman timer");
ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, synctime_ms,
param_set_deadman_synctime, param_get_ulong, ZMOD_RW,
"Pool sync expiration time in milliseconds");
ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, ziotime_ms,
param_set_deadman_ziotime, param_get_ulong, ZMOD_RW,
"IO expiration time in milliseconds");
ZFS_MODULE_PARAM(zfs, zfs_, special_class_metadata_reserve_pct, INT, ZMOD_RW,
"Small file blocks in special vdevs depends on this much "
"free space available");
/* END CSTYLED */
ZFS_MODULE_PARAM_CALL(zfs_spa, spa_, slop_shift, param_set_slop_shift,
param_get_int, ZMOD_RW, "Reserved free space in pool");
diff --git a/module/zfs/spa_stats.c b/module/zfs/spa_stats.c
index 534ac72fee7b..d89f79480ce3 100644
--- a/module/zfs/spa_stats.c
+++ b/module/zfs/spa_stats.c
@@ -1,979 +1,979 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
#include <sys/zfs_context.h>
#include <sys/spa_impl.h>
#include <sys/vdev_impl.h>
#include <sys/spa.h>
#include <zfs_comutil.h>
/*
* Keeps stats on last N reads per spa_t, disabled by default.
*/
-int zfs_read_history = 0;
+static int zfs_read_history = B_FALSE;
/*
* Include cache hits in history, disabled by default.
*/
-int zfs_read_history_hits = 0;
+static int zfs_read_history_hits = B_FALSE;
/*
* Keeps stats on the last 100 txgs by default.
*/
-int zfs_txg_history = 100;
+static int zfs_txg_history = 100;
/*
* Keeps stats on the last N MMP updates, disabled by default.
*/
-int zfs_multihost_history = 0;
+int zfs_multihost_history = B_FALSE;
/*
* ==========================================================================
* SPA Read History Routines
* ==========================================================================
*/
/*
* Read statistics - Information exported regarding each arc_read call
*/
typedef struct spa_read_history {
hrtime_t start; /* time read completed */
uint64_t objset; /* read from this objset */
uint64_t object; /* read of this object number */
uint64_t level; /* block's indirection level */
uint64_t blkid; /* read of this block id */
char origin[24]; /* read originated from here */
uint32_t aflags; /* ARC flags (cached, prefetch, etc.) */
pid_t pid; /* PID of task doing read */
char comm[16]; /* process name of task doing read */
procfs_list_node_t srh_node;
} spa_read_history_t;
static int
spa_read_history_show_header(struct seq_file *f)
{
seq_printf(f, "%-8s %-16s %-8s %-8s %-8s %-8s %-8s "
"%-24s %-8s %-16s\n", "UID", "start", "objset", "object",
"level", "blkid", "aflags", "origin", "pid", "process");
return (0);
}
static int
spa_read_history_show(struct seq_file *f, void *data)
{
spa_read_history_t *srh = (spa_read_history_t *)data;
seq_printf(f, "%-8llu %-16llu 0x%-6llx "
"%-8lli %-8lli %-8lli 0x%-6x %-24s %-8i %-16s\n",
(u_longlong_t)srh->srh_node.pln_id, srh->start,
(longlong_t)srh->objset, (longlong_t)srh->object,
(longlong_t)srh->level, (longlong_t)srh->blkid,
srh->aflags, srh->origin, srh->pid, srh->comm);
return (0);
}
/* Remove oldest elements from list until there are no more than 'size' left */
static void
spa_read_history_truncate(spa_history_list_t *shl, unsigned int size)
{
spa_read_history_t *srh;
while (shl->size > size) {
srh = list_remove_head(&shl->procfs_list.pl_list);
ASSERT3P(srh, !=, NULL);
kmem_free(srh, sizeof (spa_read_history_t));
shl->size--;
}
if (size == 0)
ASSERT(list_is_empty(&shl->procfs_list.pl_list));
}
static int
spa_read_history_clear(procfs_list_t *procfs_list)
{
spa_history_list_t *shl = procfs_list->pl_private;
mutex_enter(&procfs_list->pl_lock);
spa_read_history_truncate(shl, 0);
mutex_exit(&procfs_list->pl_lock);
return (0);
}
static void
spa_read_history_init(spa_t *spa)
{
spa_history_list_t *shl = &spa->spa_stats.read_history;
shl->size = 0;
shl->procfs_list.pl_private = shl;
procfs_list_install("zfs",
spa_name(spa),
"reads",
0600,
&shl->procfs_list,
spa_read_history_show,
spa_read_history_show_header,
spa_read_history_clear,
offsetof(spa_read_history_t, srh_node));
}
static void
spa_read_history_destroy(spa_t *spa)
{
spa_history_list_t *shl = &spa->spa_stats.read_history;
procfs_list_uninstall(&shl->procfs_list);
spa_read_history_truncate(shl, 0);
procfs_list_destroy(&shl->procfs_list);
}
void
spa_read_history_add(spa_t *spa, const zbookmark_phys_t *zb, uint32_t aflags)
{
spa_history_list_t *shl = &spa->spa_stats.read_history;
spa_read_history_t *srh;
ASSERT3P(spa, !=, NULL);
ASSERT3P(zb, !=, NULL);
if (zfs_read_history == 0 && shl->size == 0)
return;
if (zfs_read_history_hits == 0 && (aflags & ARC_FLAG_CACHED))
return;
srh = kmem_zalloc(sizeof (spa_read_history_t), KM_SLEEP);
strlcpy(srh->comm, getcomm(), sizeof (srh->comm));
srh->start = gethrtime();
srh->objset = zb->zb_objset;
srh->object = zb->zb_object;
srh->level = zb->zb_level;
srh->blkid = zb->zb_blkid;
srh->aflags = aflags;
srh->pid = getpid();
mutex_enter(&shl->procfs_list.pl_lock);
procfs_list_add(&shl->procfs_list, srh);
shl->size++;
spa_read_history_truncate(shl, zfs_read_history);
mutex_exit(&shl->procfs_list.pl_lock);
}
/*
* ==========================================================================
* SPA TXG History Routines
* ==========================================================================
*/
/*
* Txg statistics - Information exported regarding each txg sync
*/
typedef struct spa_txg_history {
uint64_t txg; /* txg id */
txg_state_t state; /* active txg state */
uint64_t nread; /* number of bytes read */
uint64_t nwritten; /* number of bytes written */
uint64_t reads; /* number of read operations */
uint64_t writes; /* number of write operations */
uint64_t ndirty; /* number of dirty bytes */
hrtime_t times[TXG_STATE_COMMITTED]; /* completion times */
procfs_list_node_t sth_node;
} spa_txg_history_t;
static int
spa_txg_history_show_header(struct seq_file *f)
{
seq_printf(f, "%-8s %-16s %-5s %-12s %-12s %-12s "
"%-8s %-8s %-12s %-12s %-12s %-12s\n", "txg", "birth", "state",
"ndirty", "nread", "nwritten", "reads", "writes",
"otime", "qtime", "wtime", "stime");
return (0);
}
static int
spa_txg_history_show(struct seq_file *f, void *data)
{
spa_txg_history_t *sth = (spa_txg_history_t *)data;
uint64_t open = 0, quiesce = 0, wait = 0, sync = 0;
char state;
switch (sth->state) {
case TXG_STATE_BIRTH: state = 'B'; break;
case TXG_STATE_OPEN: state = 'O'; break;
case TXG_STATE_QUIESCED: state = 'Q'; break;
case TXG_STATE_WAIT_FOR_SYNC: state = 'W'; break;
case TXG_STATE_SYNCED: state = 'S'; break;
case TXG_STATE_COMMITTED: state = 'C'; break;
default: state = '?'; break;
}
if (sth->times[TXG_STATE_OPEN])
open = sth->times[TXG_STATE_OPEN] -
sth->times[TXG_STATE_BIRTH];
if (sth->times[TXG_STATE_QUIESCED])
quiesce = sth->times[TXG_STATE_QUIESCED] -
sth->times[TXG_STATE_OPEN];
if (sth->times[TXG_STATE_WAIT_FOR_SYNC])
wait = sth->times[TXG_STATE_WAIT_FOR_SYNC] -
sth->times[TXG_STATE_QUIESCED];
if (sth->times[TXG_STATE_SYNCED])
sync = sth->times[TXG_STATE_SYNCED] -
sth->times[TXG_STATE_WAIT_FOR_SYNC];
seq_printf(f, "%-8llu %-16llu %-5c %-12llu "
"%-12llu %-12llu %-8llu %-8llu %-12llu %-12llu %-12llu %-12llu\n",
(longlong_t)sth->txg, sth->times[TXG_STATE_BIRTH], state,
(u_longlong_t)sth->ndirty,
(u_longlong_t)sth->nread, (u_longlong_t)sth->nwritten,
(u_longlong_t)sth->reads, (u_longlong_t)sth->writes,
(u_longlong_t)open, (u_longlong_t)quiesce, (u_longlong_t)wait,
(u_longlong_t)sync);
return (0);
}
/* Remove oldest elements from list until there are no more than 'size' left */
static void
spa_txg_history_truncate(spa_history_list_t *shl, unsigned int size)
{
spa_txg_history_t *sth;
while (shl->size > size) {
sth = list_remove_head(&shl->procfs_list.pl_list);
ASSERT3P(sth, !=, NULL);
kmem_free(sth, sizeof (spa_txg_history_t));
shl->size--;
}
if (size == 0)
ASSERT(list_is_empty(&shl->procfs_list.pl_list));
}
static int
spa_txg_history_clear(procfs_list_t *procfs_list)
{
spa_history_list_t *shl = procfs_list->pl_private;
mutex_enter(&procfs_list->pl_lock);
spa_txg_history_truncate(shl, 0);
mutex_exit(&procfs_list->pl_lock);
return (0);
}
static void
spa_txg_history_init(spa_t *spa)
{
spa_history_list_t *shl = &spa->spa_stats.txg_history;
shl->size = 0;
shl->procfs_list.pl_private = shl;
procfs_list_install("zfs",
spa_name(spa),
"txgs",
0644,
&shl->procfs_list,
spa_txg_history_show,
spa_txg_history_show_header,
spa_txg_history_clear,
offsetof(spa_txg_history_t, sth_node));
}
static void
spa_txg_history_destroy(spa_t *spa)
{
spa_history_list_t *shl = &spa->spa_stats.txg_history;
procfs_list_uninstall(&shl->procfs_list);
spa_txg_history_truncate(shl, 0);
procfs_list_destroy(&shl->procfs_list);
}
/*
* Add a new txg to historical record.
*/
void
spa_txg_history_add(spa_t *spa, uint64_t txg, hrtime_t birth_time)
{
spa_history_list_t *shl = &spa->spa_stats.txg_history;
spa_txg_history_t *sth;
if (zfs_txg_history == 0 && shl->size == 0)
return;
sth = kmem_zalloc(sizeof (spa_txg_history_t), KM_SLEEP);
sth->txg = txg;
sth->state = TXG_STATE_OPEN;
sth->times[TXG_STATE_BIRTH] = birth_time;
mutex_enter(&shl->procfs_list.pl_lock);
procfs_list_add(&shl->procfs_list, sth);
shl->size++;
spa_txg_history_truncate(shl, zfs_txg_history);
mutex_exit(&shl->procfs_list.pl_lock);
}
/*
* Set txg state completion time and increment current state.
*/
int
spa_txg_history_set(spa_t *spa, uint64_t txg, txg_state_t completed_state,
hrtime_t completed_time)
{
spa_history_list_t *shl = &spa->spa_stats.txg_history;
spa_txg_history_t *sth;
int error = ENOENT;
if (zfs_txg_history == 0)
return (0);
mutex_enter(&shl->procfs_list.pl_lock);
for (sth = list_tail(&shl->procfs_list.pl_list); sth != NULL;
sth = list_prev(&shl->procfs_list.pl_list, sth)) {
if (sth->txg == txg) {
sth->times[completed_state] = completed_time;
sth->state++;
error = 0;
break;
}
}
mutex_exit(&shl->procfs_list.pl_lock);
return (error);
}
/*
* Set txg IO stats.
*/
static int
spa_txg_history_set_io(spa_t *spa, uint64_t txg, uint64_t nread,
uint64_t nwritten, uint64_t reads, uint64_t writes, uint64_t ndirty)
{
spa_history_list_t *shl = &spa->spa_stats.txg_history;
spa_txg_history_t *sth;
int error = ENOENT;
if (zfs_txg_history == 0)
return (0);
mutex_enter(&shl->procfs_list.pl_lock);
for (sth = list_tail(&shl->procfs_list.pl_list); sth != NULL;
sth = list_prev(&shl->procfs_list.pl_list, sth)) {
if (sth->txg == txg) {
sth->nread = nread;
sth->nwritten = nwritten;
sth->reads = reads;
sth->writes = writes;
sth->ndirty = ndirty;
error = 0;
break;
}
}
mutex_exit(&shl->procfs_list.pl_lock);
return (error);
}
txg_stat_t *
spa_txg_history_init_io(spa_t *spa, uint64_t txg, dsl_pool_t *dp)
{
txg_stat_t *ts;
if (zfs_txg_history == 0)
return (NULL);
ts = kmem_alloc(sizeof (txg_stat_t), KM_SLEEP);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vdev_get_stats(spa->spa_root_vdev, &ts->vs1);
spa_config_exit(spa, SCL_CONFIG, FTAG);
ts->txg = txg;
ts->ndirty = dp->dp_dirty_pertxg[txg & TXG_MASK];
spa_txg_history_set(spa, txg, TXG_STATE_WAIT_FOR_SYNC, gethrtime());
return (ts);
}
void
spa_txg_history_fini_io(spa_t *spa, txg_stat_t *ts)
{
if (ts == NULL)
return;
if (zfs_txg_history == 0) {
kmem_free(ts, sizeof (txg_stat_t));
return;
}
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vdev_get_stats(spa->spa_root_vdev, &ts->vs2);
spa_config_exit(spa, SCL_CONFIG, FTAG);
spa_txg_history_set(spa, ts->txg, TXG_STATE_SYNCED, gethrtime());
spa_txg_history_set_io(spa, ts->txg,
ts->vs2.vs_bytes[ZIO_TYPE_READ] - ts->vs1.vs_bytes[ZIO_TYPE_READ],
ts->vs2.vs_bytes[ZIO_TYPE_WRITE] - ts->vs1.vs_bytes[ZIO_TYPE_WRITE],
ts->vs2.vs_ops[ZIO_TYPE_READ] - ts->vs1.vs_ops[ZIO_TYPE_READ],
ts->vs2.vs_ops[ZIO_TYPE_WRITE] - ts->vs1.vs_ops[ZIO_TYPE_WRITE],
ts->ndirty);
kmem_free(ts, sizeof (txg_stat_t));
}
/*
* ==========================================================================
* SPA TX Assign Histogram Routines
* ==========================================================================
*/
/*
* Tx statistics - Information exported regarding dmu_tx_assign time.
*/
/*
* When the kstat is written zero all buckets. When the kstat is read
* count the number of trailing buckets set to zero and update ks_ndata
* such that they are not output.
*/
static int
spa_tx_assign_update(kstat_t *ksp, int rw)
{
spa_t *spa = ksp->ks_private;
spa_history_kstat_t *shk = &spa->spa_stats.tx_assign_histogram;
int i;
if (rw == KSTAT_WRITE) {
for (i = 0; i < shk->count; i++)
((kstat_named_t *)shk->priv)[i].value.ui64 = 0;
}
for (i = shk->count; i > 0; i--)
if (((kstat_named_t *)shk->priv)[i-1].value.ui64 != 0)
break;
ksp->ks_ndata = i;
ksp->ks_data_size = i * sizeof (kstat_named_t);
return (0);
}
static void
spa_tx_assign_init(spa_t *spa)
{
spa_history_kstat_t *shk = &spa->spa_stats.tx_assign_histogram;
char *name;
kstat_named_t *ks;
kstat_t *ksp;
int i;
mutex_init(&shk->lock, NULL, MUTEX_DEFAULT, NULL);
shk->count = 42; /* power of two buckets for 1ns to 2,199s */
shk->size = shk->count * sizeof (kstat_named_t);
shk->priv = kmem_alloc(shk->size, KM_SLEEP);
name = kmem_asprintf("zfs/%s", spa_name(spa));
for (i = 0; i < shk->count; i++) {
ks = &((kstat_named_t *)shk->priv)[i];
ks->data_type = KSTAT_DATA_UINT64;
ks->value.ui64 = 0;
(void) snprintf(ks->name, KSTAT_STRLEN, "%llu ns",
(u_longlong_t)1 << i);
}
ksp = kstat_create(name, 0, "dmu_tx_assign", "misc",
KSTAT_TYPE_NAMED, 0, KSTAT_FLAG_VIRTUAL);
shk->kstat = ksp;
if (ksp) {
ksp->ks_lock = &shk->lock;
ksp->ks_data = shk->priv;
ksp->ks_ndata = shk->count;
ksp->ks_data_size = shk->size;
ksp->ks_private = spa;
ksp->ks_update = spa_tx_assign_update;
kstat_install(ksp);
}
kmem_strfree(name);
}
static void
spa_tx_assign_destroy(spa_t *spa)
{
spa_history_kstat_t *shk = &spa->spa_stats.tx_assign_histogram;
kstat_t *ksp;
ksp = shk->kstat;
if (ksp)
kstat_delete(ksp);
kmem_free(shk->priv, shk->size);
mutex_destroy(&shk->lock);
}
void
spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs)
{
spa_history_kstat_t *shk = &spa->spa_stats.tx_assign_histogram;
uint64_t idx = 0;
while (((1ULL << idx) < nsecs) && (idx < shk->size - 1))
idx++;
atomic_inc_64(&((kstat_named_t *)shk->priv)[idx].value.ui64);
}
/*
* ==========================================================================
* SPA MMP History Routines
* ==========================================================================
*/
/*
* MMP statistics - Information exported regarding attempted MMP writes
* For MMP writes issued, fields used as per comments below.
* For MMP writes skipped, an entry represents a span of time when
* writes were skipped for same reason (error from mmp_random_leaf).
* Differences are:
* timestamp time first write skipped, if >1 skipped in a row
* mmp_delay delay value at timestamp
* vdev_guid number of writes skipped
* io_error one of enum mmp_error
* duration time span (ns) of skipped writes
*/
typedef struct spa_mmp_history {
uint64_t mmp_node_id; /* unique # for updates */
uint64_t txg; /* txg of last sync */
uint64_t timestamp; /* UTC time MMP write issued */
uint64_t mmp_delay; /* mmp_thread.mmp_delay at timestamp */
uint64_t vdev_guid; /* unique ID of leaf vdev */
char *vdev_path;
int vdev_label; /* vdev label */
int io_error; /* error status of MMP write */
hrtime_t error_start; /* hrtime of start of error period */
hrtime_t duration; /* time from submission to completion */
procfs_list_node_t smh_node;
} spa_mmp_history_t;
static int
spa_mmp_history_show_header(struct seq_file *f)
{
seq_printf(f, "%-10s %-10s %-10s %-6s %-10s %-12s %-24s "
"%-10s %s\n", "id", "txg", "timestamp", "error", "duration",
"mmp_delay", "vdev_guid", "vdev_label", "vdev_path");
return (0);
}
static int
spa_mmp_history_show(struct seq_file *f, void *data)
{
spa_mmp_history_t *smh = (spa_mmp_history_t *)data;
char skip_fmt[] = "%-10llu %-10llu %10llu %#6llx %10lld %12llu %-24llu "
"%-10lld %s\n";
char write_fmt[] = "%-10llu %-10llu %10llu %6lld %10lld %12llu %-24llu "
"%-10lld %s\n";
seq_printf(f, (smh->error_start ? skip_fmt : write_fmt),
(u_longlong_t)smh->mmp_node_id, (u_longlong_t)smh->txg,
(u_longlong_t)smh->timestamp, (longlong_t)smh->io_error,
(longlong_t)smh->duration, (u_longlong_t)smh->mmp_delay,
(u_longlong_t)smh->vdev_guid, (u_longlong_t)smh->vdev_label,
(smh->vdev_path ? smh->vdev_path : "-"));
return (0);
}
/* Remove oldest elements from list until there are no more than 'size' left */
static void
spa_mmp_history_truncate(spa_history_list_t *shl, unsigned int size)
{
spa_mmp_history_t *smh;
while (shl->size > size) {
smh = list_remove_head(&shl->procfs_list.pl_list);
if (smh->vdev_path)
kmem_strfree(smh->vdev_path);
kmem_free(smh, sizeof (spa_mmp_history_t));
shl->size--;
}
if (size == 0)
ASSERT(list_is_empty(&shl->procfs_list.pl_list));
}
static int
spa_mmp_history_clear(procfs_list_t *procfs_list)
{
spa_history_list_t *shl = procfs_list->pl_private;
mutex_enter(&procfs_list->pl_lock);
spa_mmp_history_truncate(shl, 0);
mutex_exit(&procfs_list->pl_lock);
return (0);
}
static void
spa_mmp_history_init(spa_t *spa)
{
spa_history_list_t *shl = &spa->spa_stats.mmp_history;
shl->size = 0;
shl->procfs_list.pl_private = shl;
procfs_list_install("zfs",
spa_name(spa),
"multihost",
0644,
&shl->procfs_list,
spa_mmp_history_show,
spa_mmp_history_show_header,
spa_mmp_history_clear,
offsetof(spa_mmp_history_t, smh_node));
}
static void
spa_mmp_history_destroy(spa_t *spa)
{
spa_history_list_t *shl = &spa->spa_stats.mmp_history;
procfs_list_uninstall(&shl->procfs_list);
spa_mmp_history_truncate(shl, 0);
procfs_list_destroy(&shl->procfs_list);
}
/*
* Set duration in existing "skip" record to how long we have waited for a leaf
* vdev to become available.
*
* Important that we start search at the tail of the list where new
* records are inserted, so this is normally an O(1) operation.
*/
int
spa_mmp_history_set_skip(spa_t *spa, uint64_t mmp_node_id)
{
spa_history_list_t *shl = &spa->spa_stats.mmp_history;
spa_mmp_history_t *smh;
int error = ENOENT;
if (zfs_multihost_history == 0 && shl->size == 0)
return (0);
mutex_enter(&shl->procfs_list.pl_lock);
for (smh = list_tail(&shl->procfs_list.pl_list); smh != NULL;
smh = list_prev(&shl->procfs_list.pl_list, smh)) {
if (smh->mmp_node_id == mmp_node_id) {
ASSERT3U(smh->io_error, !=, 0);
smh->duration = gethrtime() - smh->error_start;
smh->vdev_guid++;
error = 0;
break;
}
}
mutex_exit(&shl->procfs_list.pl_lock);
return (error);
}
/*
* Set MMP write duration and error status in existing record.
* See comment re: search order above spa_mmp_history_set_skip().
*/
int
spa_mmp_history_set(spa_t *spa, uint64_t mmp_node_id, int io_error,
hrtime_t duration)
{
spa_history_list_t *shl = &spa->spa_stats.mmp_history;
spa_mmp_history_t *smh;
int error = ENOENT;
if (zfs_multihost_history == 0 && shl->size == 0)
return (0);
mutex_enter(&shl->procfs_list.pl_lock);
for (smh = list_tail(&shl->procfs_list.pl_list); smh != NULL;
smh = list_prev(&shl->procfs_list.pl_list, smh)) {
if (smh->mmp_node_id == mmp_node_id) {
ASSERT(smh->io_error == 0);
smh->io_error = io_error;
smh->duration = duration;
error = 0;
break;
}
}
mutex_exit(&shl->procfs_list.pl_lock);
return (error);
}
/*
* Add a new MMP historical record.
* error == 0 : a write was issued.
* error != 0 : a write was not issued because no leaves were found.
*/
void
spa_mmp_history_add(spa_t *spa, uint64_t txg, uint64_t timestamp,
uint64_t mmp_delay, vdev_t *vd, int label, uint64_t mmp_node_id,
int error)
{
spa_history_list_t *shl = &spa->spa_stats.mmp_history;
spa_mmp_history_t *smh;
if (zfs_multihost_history == 0 && shl->size == 0)
return;
smh = kmem_zalloc(sizeof (spa_mmp_history_t), KM_SLEEP);
smh->txg = txg;
smh->timestamp = timestamp;
smh->mmp_delay = mmp_delay;
if (vd) {
smh->vdev_guid = vd->vdev_guid;
if (vd->vdev_path)
smh->vdev_path = kmem_strdup(vd->vdev_path);
}
smh->vdev_label = label;
smh->mmp_node_id = mmp_node_id;
if (error) {
smh->io_error = error;
smh->error_start = gethrtime();
smh->vdev_guid = 1;
}
mutex_enter(&shl->procfs_list.pl_lock);
procfs_list_add(&shl->procfs_list, smh);
shl->size++;
spa_mmp_history_truncate(shl, zfs_multihost_history);
mutex_exit(&shl->procfs_list.pl_lock);
}
static void *
spa_state_addr(kstat_t *ksp, loff_t n)
{
if (n == 0)
return (ksp->ks_private); /* return the spa_t */
return (NULL);
}
static int
spa_state_data(char *buf, size_t size, void *data)
{
spa_t *spa = (spa_t *)data;
(void) snprintf(buf, size, "%s\n", spa_state_to_name(spa));
return (0);
}
/*
* Return the state of the pool in /proc/spl/kstat/zfs/<pool>/state.
*
* This is a lock-less read of the pool's state (unlike using 'zpool', which
* can potentially block for seconds). Because it doesn't block, it can useful
* as a pool heartbeat value.
*/
static void
spa_state_init(spa_t *spa)
{
spa_history_kstat_t *shk = &spa->spa_stats.state;
char *name;
kstat_t *ksp;
mutex_init(&shk->lock, NULL, MUTEX_DEFAULT, NULL);
name = kmem_asprintf("zfs/%s", spa_name(spa));
ksp = kstat_create(name, 0, "state", "misc",
KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);
shk->kstat = ksp;
if (ksp) {
ksp->ks_lock = &shk->lock;
ksp->ks_data = NULL;
ksp->ks_private = spa;
ksp->ks_flags |= KSTAT_FLAG_NO_HEADERS;
kstat_set_raw_ops(ksp, NULL, spa_state_data, spa_state_addr);
kstat_install(ksp);
}
kmem_strfree(name);
}
static void
spa_health_destroy(spa_t *spa)
{
spa_history_kstat_t *shk = &spa->spa_stats.state;
kstat_t *ksp = shk->kstat;
if (ksp)
kstat_delete(ksp);
mutex_destroy(&shk->lock);
}
-static spa_iostats_t spa_iostats_template = {
+static const spa_iostats_t spa_iostats_template = {
{ "trim_extents_written", KSTAT_DATA_UINT64 },
{ "trim_bytes_written", KSTAT_DATA_UINT64 },
{ "trim_extents_skipped", KSTAT_DATA_UINT64 },
{ "trim_bytes_skipped", KSTAT_DATA_UINT64 },
{ "trim_extents_failed", KSTAT_DATA_UINT64 },
{ "trim_bytes_failed", KSTAT_DATA_UINT64 },
{ "autotrim_extents_written", KSTAT_DATA_UINT64 },
{ "autotrim_bytes_written", KSTAT_DATA_UINT64 },
{ "autotrim_extents_skipped", KSTAT_DATA_UINT64 },
{ "autotrim_bytes_skipped", KSTAT_DATA_UINT64 },
{ "autotrim_extents_failed", KSTAT_DATA_UINT64 },
{ "autotrim_bytes_failed", KSTAT_DATA_UINT64 },
{ "simple_trim_extents_written", KSTAT_DATA_UINT64 },
{ "simple_trim_bytes_written", KSTAT_DATA_UINT64 },
{ "simple_trim_extents_skipped", KSTAT_DATA_UINT64 },
{ "simple_trim_bytes_skipped", KSTAT_DATA_UINT64 },
{ "simple_trim_extents_failed", KSTAT_DATA_UINT64 },
{ "simple_trim_bytes_failed", KSTAT_DATA_UINT64 },
};
#define SPA_IOSTATS_ADD(stat, val) \
atomic_add_64(&iostats->stat.value.ui64, (val));
void
spa_iostats_trim_add(spa_t *spa, trim_type_t type,
uint64_t extents_written, uint64_t bytes_written,
uint64_t extents_skipped, uint64_t bytes_skipped,
uint64_t extents_failed, uint64_t bytes_failed)
{
spa_history_kstat_t *shk = &spa->spa_stats.iostats;
kstat_t *ksp = shk->kstat;
spa_iostats_t *iostats;
if (ksp == NULL)
return;
iostats = ksp->ks_data;
if (type == TRIM_TYPE_MANUAL) {
SPA_IOSTATS_ADD(trim_extents_written, extents_written);
SPA_IOSTATS_ADD(trim_bytes_written, bytes_written);
SPA_IOSTATS_ADD(trim_extents_skipped, extents_skipped);
SPA_IOSTATS_ADD(trim_bytes_skipped, bytes_skipped);
SPA_IOSTATS_ADD(trim_extents_failed, extents_failed);
SPA_IOSTATS_ADD(trim_bytes_failed, bytes_failed);
} else if (type == TRIM_TYPE_AUTO) {
SPA_IOSTATS_ADD(autotrim_extents_written, extents_written);
SPA_IOSTATS_ADD(autotrim_bytes_written, bytes_written);
SPA_IOSTATS_ADD(autotrim_extents_skipped, extents_skipped);
SPA_IOSTATS_ADD(autotrim_bytes_skipped, bytes_skipped);
SPA_IOSTATS_ADD(autotrim_extents_failed, extents_failed);
SPA_IOSTATS_ADD(autotrim_bytes_failed, bytes_failed);
} else {
SPA_IOSTATS_ADD(simple_trim_extents_written, extents_written);
SPA_IOSTATS_ADD(simple_trim_bytes_written, bytes_written);
SPA_IOSTATS_ADD(simple_trim_extents_skipped, extents_skipped);
SPA_IOSTATS_ADD(simple_trim_bytes_skipped, bytes_skipped);
SPA_IOSTATS_ADD(simple_trim_extents_failed, extents_failed);
SPA_IOSTATS_ADD(simple_trim_bytes_failed, bytes_failed);
}
}
static int
spa_iostats_update(kstat_t *ksp, int rw)
{
if (rw == KSTAT_WRITE) {
memcpy(ksp->ks_data, &spa_iostats_template,
sizeof (spa_iostats_t));
}
return (0);
}
static void
spa_iostats_init(spa_t *spa)
{
spa_history_kstat_t *shk = &spa->spa_stats.iostats;
mutex_init(&shk->lock, NULL, MUTEX_DEFAULT, NULL);
char *name = kmem_asprintf("zfs/%s", spa_name(spa));
kstat_t *ksp = kstat_create(name, 0, "iostats", "misc",
KSTAT_TYPE_NAMED, sizeof (spa_iostats_t) / sizeof (kstat_named_t),
KSTAT_FLAG_VIRTUAL);
shk->kstat = ksp;
if (ksp) {
int size = sizeof (spa_iostats_t);
ksp->ks_lock = &shk->lock;
ksp->ks_private = spa;
ksp->ks_update = spa_iostats_update;
ksp->ks_data = kmem_alloc(size, KM_SLEEP);
memcpy(ksp->ks_data, &spa_iostats_template, size);
kstat_install(ksp);
}
kmem_strfree(name);
}
static void
spa_iostats_destroy(spa_t *spa)
{
spa_history_kstat_t *shk = &spa->spa_stats.iostats;
kstat_t *ksp = shk->kstat;
if (ksp) {
kmem_free(ksp->ks_data, sizeof (spa_iostats_t));
kstat_delete(ksp);
}
mutex_destroy(&shk->lock);
}
void
spa_stats_init(spa_t *spa)
{
spa_read_history_init(spa);
spa_txg_history_init(spa);
spa_tx_assign_init(spa);
spa_mmp_history_init(spa);
spa_state_init(spa);
spa_iostats_init(spa);
}
void
spa_stats_destroy(spa_t *spa)
{
spa_iostats_destroy(spa);
spa_health_destroy(spa);
spa_tx_assign_destroy(spa);
spa_txg_history_destroy(spa);
spa_read_history_destroy(spa);
spa_mmp_history_destroy(spa);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, read_history, INT, ZMOD_RW,
"Historical statistics for the last N reads");
ZFS_MODULE_PARAM(zfs, zfs_, read_history_hits, INT, ZMOD_RW,
"Include cache hits in read history");
ZFS_MODULE_PARAM(zfs_txg, zfs_txg_, history, INT, ZMOD_RW,
"Historical statistics for the last N txgs");
ZFS_MODULE_PARAM(zfs_multihost, zfs_multihost_, history, INT, ZMOD_RW,
"Historical statistics for last N multihost writes");
/* END CSTYLED */
diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
index 451d1b9a0b6c..5481902f2a6c 100644
--- a/module/zfs/vdev.c
+++ b/module/zfs/vdev.c
@@ -1,6081 +1,6081 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2021 by Delphix. All rights reserved.
* Copyright 2017 Nexenta Systems, Inc.
* Copyright (c) 2014 Integros [integros.com]
* Copyright 2016 Toomas Soome <tsoome@me.com>
* Copyright 2017 Joyent, Inc.
* Copyright (c) 2017, Intel Corporation.
* Copyright (c) 2019, Datto Inc. All rights reserved.
* Copyright (c) 2021, Klara Inc.
* Copyright [2021] Hewlett Packard Enterprise Development LP
*/
#include <sys/zfs_context.h>
#include <sys/fm/fs/zfs.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/bpobj.h>
#include <sys/dmu.h>
#include <sys/dmu_tx.h>
#include <sys/dsl_dir.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_rebuild.h>
#include <sys/vdev_draid.h>
#include <sys/uberblock_impl.h>
#include <sys/metaslab.h>
#include <sys/metaslab_impl.h>
#include <sys/space_map.h>
#include <sys/space_reftree.h>
#include <sys/zio.h>
#include <sys/zap.h>
#include <sys/fs/zfs.h>
#include <sys/arc.h>
#include <sys/zil.h>
#include <sys/dsl_scan.h>
#include <sys/vdev_raidz.h>
#include <sys/abd.h>
#include <sys/vdev_initialize.h>
#include <sys/vdev_trim.h>
#include <sys/zvol.h>
#include <sys/zfs_ratelimit.h>
#include "zfs_prop.h"
/*
* One metaslab from each (normal-class) vdev is used by the ZIL. These are
* called "embedded slog metaslabs", are referenced by vdev_log_mg, and are
* part of the spa_embedded_log_class. The metaslab with the most free space
* in each vdev is selected for this purpose when the pool is opened (or a
* vdev is added). See vdev_metaslab_init().
*
* Log blocks can be allocated from the following locations. Each one is tried
* in order until the allocation succeeds:
* 1. dedicated log vdevs, aka "slog" (spa_log_class)
* 2. embedded slog metaslabs (spa_embedded_log_class)
* 3. other metaslabs in normal vdevs (spa_normal_class)
*
* zfs_embedded_slog_min_ms disables the embedded slog if there are fewer
* than this number of metaslabs in the vdev. This ensures that we don't set
* aside an unreasonable amount of space for the ZIL. If set to less than
* 1 << (spa_slop_shift + 1), on small pools the usable space may be reduced
* (by more than 1<<spa_slop_shift) due to the embedded slog metaslab.
*/
-int zfs_embedded_slog_min_ms = 64;
+static int zfs_embedded_slog_min_ms = 64;
/* default target for number of metaslabs per top-level vdev */
-int zfs_vdev_default_ms_count = 200;
+static int zfs_vdev_default_ms_count = 200;
/* minimum number of metaslabs per top-level vdev */
-int zfs_vdev_min_ms_count = 16;
+static int zfs_vdev_min_ms_count = 16;
/* practical upper limit of total metaslabs per top-level vdev */
-int zfs_vdev_ms_count_limit = 1ULL << 17;
+static int zfs_vdev_ms_count_limit = 1ULL << 17;
/* lower limit for metaslab size (512M) */
-int zfs_vdev_default_ms_shift = 29;
+static int zfs_vdev_default_ms_shift = 29;
/* upper limit for metaslab size (16G) */
-int zfs_vdev_max_ms_shift = 34;
+static const int zfs_vdev_max_ms_shift = 34;
int vdev_validate_skip = B_FALSE;
/*
* Since the DTL space map of a vdev is not expected to have a lot of
* entries, we default its block size to 4K.
*/
int zfs_vdev_dtl_sm_blksz = (1 << 12);
/*
* Rate limit slow IO (delay) events to this many per second.
*/
-unsigned int zfs_slow_io_events_per_second = 20;
+static unsigned int zfs_slow_io_events_per_second = 20;
/*
* Rate limit checksum events after this many checksum errors per second.
*/
-unsigned int zfs_checksum_events_per_second = 20;
+static unsigned int zfs_checksum_events_per_second = 20;
/*
* Ignore errors during scrub/resilver. Allows to work around resilver
* upon import when there are pool errors.
*/
-int zfs_scan_ignore_errors = 0;
+static int zfs_scan_ignore_errors = 0;
/*
* vdev-wide space maps that have lots of entries written to them at
* the end of each transaction can benefit from a higher I/O bandwidth
* (e.g. vdev_obsolete_sm), thus we default their block size to 128K.
*/
int zfs_vdev_standard_sm_blksz = (1 << 17);
/*
* Tunable parameter for debugging or performance analysis. Setting this
* will cause pool corruption on power loss if a volatile out-of-order
* write cache is enabled.
*/
int zfs_nocacheflush = 0;
uint64_t zfs_vdev_max_auto_ashift = ASHIFT_MAX;
uint64_t zfs_vdev_min_auto_ashift = ASHIFT_MIN;
void
vdev_dbgmsg(vdev_t *vd, const char *fmt, ...)
{
va_list adx;
char buf[256];
va_start(adx, fmt);
(void) vsnprintf(buf, sizeof (buf), fmt, adx);
va_end(adx);
if (vd->vdev_path != NULL) {
zfs_dbgmsg("%s vdev '%s': %s", vd->vdev_ops->vdev_op_type,
vd->vdev_path, buf);
} else {
zfs_dbgmsg("%s-%llu vdev (guid %llu): %s",
vd->vdev_ops->vdev_op_type,
(u_longlong_t)vd->vdev_id,
(u_longlong_t)vd->vdev_guid, buf);
}
}
void
vdev_dbgmsg_print_tree(vdev_t *vd, int indent)
{
char state[20];
if (vd->vdev_ishole || vd->vdev_ops == &vdev_missing_ops) {
zfs_dbgmsg("%*svdev %llu: %s", indent, "",
(u_longlong_t)vd->vdev_id,
vd->vdev_ops->vdev_op_type);
return;
}
switch (vd->vdev_state) {
case VDEV_STATE_UNKNOWN:
(void) snprintf(state, sizeof (state), "unknown");
break;
case VDEV_STATE_CLOSED:
(void) snprintf(state, sizeof (state), "closed");
break;
case VDEV_STATE_OFFLINE:
(void) snprintf(state, sizeof (state), "offline");
break;
case VDEV_STATE_REMOVED:
(void) snprintf(state, sizeof (state), "removed");
break;
case VDEV_STATE_CANT_OPEN:
(void) snprintf(state, sizeof (state), "can't open");
break;
case VDEV_STATE_FAULTED:
(void) snprintf(state, sizeof (state), "faulted");
break;
case VDEV_STATE_DEGRADED:
(void) snprintf(state, sizeof (state), "degraded");
break;
case VDEV_STATE_HEALTHY:
(void) snprintf(state, sizeof (state), "healthy");
break;
default:
(void) snprintf(state, sizeof (state), "<state %u>",
(uint_t)vd->vdev_state);
}
zfs_dbgmsg("%*svdev %u: %s%s, guid: %llu, path: %s, %s", indent,
"", (int)vd->vdev_id, vd->vdev_ops->vdev_op_type,
vd->vdev_islog ? " (log)" : "",
(u_longlong_t)vd->vdev_guid,
vd->vdev_path ? vd->vdev_path : "N/A", state);
for (uint64_t i = 0; i < vd->vdev_children; i++)
vdev_dbgmsg_print_tree(vd->vdev_child[i], indent + 2);
}
/*
* Virtual device management.
*/
-static vdev_ops_t *vdev_ops_table[] = {
+static const vdev_ops_t *const vdev_ops_table[] = {
&vdev_root_ops,
&vdev_raidz_ops,
&vdev_draid_ops,
&vdev_draid_spare_ops,
&vdev_mirror_ops,
&vdev_replacing_ops,
&vdev_spare_ops,
&vdev_disk_ops,
&vdev_file_ops,
&vdev_missing_ops,
&vdev_hole_ops,
&vdev_indirect_ops,
NULL
};
/*
* Given a vdev type, return the appropriate ops vector.
*/
static vdev_ops_t *
vdev_getops(const char *type)
{
- vdev_ops_t *ops, **opspp;
+ const vdev_ops_t *ops, *const *opspp;
for (opspp = vdev_ops_table; (ops = *opspp) != NULL; opspp++)
if (strcmp(ops->vdev_op_type, type) == 0)
break;
return (ops);
}
/*
* Given a vdev and a metaslab class, find which metaslab group we're
* interested in. All vdevs may belong to two different metaslab classes.
* Dedicated slog devices use only the primary metaslab group, rather than a
* separate log group. For embedded slogs, the vdev_log_mg will be non-NULL.
*/
metaslab_group_t *
vdev_get_mg(vdev_t *vd, metaslab_class_t *mc)
{
if (mc == spa_embedded_log_class(vd->vdev_spa) &&
vd->vdev_log_mg != NULL)
return (vd->vdev_log_mg);
else
return (vd->vdev_mg);
}
void
vdev_default_xlate(vdev_t *vd, const range_seg64_t *logical_rs,
range_seg64_t *physical_rs, range_seg64_t *remain_rs)
{
(void) vd, (void) remain_rs;
physical_rs->rs_start = logical_rs->rs_start;
physical_rs->rs_end = logical_rs->rs_end;
}
/*
* Derive the enumerated allocation bias from string input.
* String origin is either the per-vdev zap or zpool(8).
*/
static vdev_alloc_bias_t
vdev_derive_alloc_bias(const char *bias)
{
vdev_alloc_bias_t alloc_bias = VDEV_BIAS_NONE;
if (strcmp(bias, VDEV_ALLOC_BIAS_LOG) == 0)
alloc_bias = VDEV_BIAS_LOG;
else if (strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0)
alloc_bias = VDEV_BIAS_SPECIAL;
else if (strcmp(bias, VDEV_ALLOC_BIAS_DEDUP) == 0)
alloc_bias = VDEV_BIAS_DEDUP;
return (alloc_bias);
}
/*
* Default asize function: return the MAX of psize with the asize of
* all children. This is what's used by anything other than RAID-Z.
*/
uint64_t
vdev_default_asize(vdev_t *vd, uint64_t psize)
{
uint64_t asize = P2ROUNDUP(psize, 1ULL << vd->vdev_top->vdev_ashift);
uint64_t csize;
for (int c = 0; c < vd->vdev_children; c++) {
csize = vdev_psize_to_asize(vd->vdev_child[c], psize);
asize = MAX(asize, csize);
}
return (asize);
}
uint64_t
vdev_default_min_asize(vdev_t *vd)
{
return (vd->vdev_min_asize);
}
/*
* Get the minimum allocatable size. We define the allocatable size as
* the vdev's asize rounded to the nearest metaslab. This allows us to
* replace or attach devices which don't have the same physical size but
* can still satisfy the same number of allocations.
*/
uint64_t
vdev_get_min_asize(vdev_t *vd)
{
vdev_t *pvd = vd->vdev_parent;
/*
* If our parent is NULL (inactive spare or cache) or is the root,
* just return our own asize.
*/
if (pvd == NULL)
return (vd->vdev_asize);
/*
* The top-level vdev just returns the allocatable size rounded
* to the nearest metaslab.
*/
if (vd == vd->vdev_top)
return (P2ALIGN(vd->vdev_asize, 1ULL << vd->vdev_ms_shift));
return (pvd->vdev_ops->vdev_op_min_asize(pvd));
}
void
vdev_set_min_asize(vdev_t *vd)
{
vd->vdev_min_asize = vdev_get_min_asize(vd);
for (int c = 0; c < vd->vdev_children; c++)
vdev_set_min_asize(vd->vdev_child[c]);
}
/*
* Get the minimal allocation size for the top-level vdev.
*/
uint64_t
vdev_get_min_alloc(vdev_t *vd)
{
uint64_t min_alloc = 1ULL << vd->vdev_ashift;
if (vd->vdev_ops->vdev_op_min_alloc != NULL)
min_alloc = vd->vdev_ops->vdev_op_min_alloc(vd);
return (min_alloc);
}
/*
* Get the parity level for a top-level vdev.
*/
uint64_t
vdev_get_nparity(vdev_t *vd)
{
uint64_t nparity = 0;
if (vd->vdev_ops->vdev_op_nparity != NULL)
nparity = vd->vdev_ops->vdev_op_nparity(vd);
return (nparity);
}
/*
* Get the number of data disks for a top-level vdev.
*/
uint64_t
vdev_get_ndisks(vdev_t *vd)
{
uint64_t ndisks = 1;
if (vd->vdev_ops->vdev_op_ndisks != NULL)
ndisks = vd->vdev_ops->vdev_op_ndisks(vd);
return (ndisks);
}
vdev_t *
vdev_lookup_top(spa_t *spa, uint64_t vdev)
{
vdev_t *rvd = spa->spa_root_vdev;
ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
if (vdev < rvd->vdev_children) {
ASSERT(rvd->vdev_child[vdev] != NULL);
return (rvd->vdev_child[vdev]);
}
return (NULL);
}
vdev_t *
vdev_lookup_by_guid(vdev_t *vd, uint64_t guid)
{
vdev_t *mvd;
if (vd->vdev_guid == guid)
return (vd);
for (int c = 0; c < vd->vdev_children; c++)
if ((mvd = vdev_lookup_by_guid(vd->vdev_child[c], guid)) !=
NULL)
return (mvd);
return (NULL);
}
static int
vdev_count_leaves_impl(vdev_t *vd)
{
int n = 0;
if (vd->vdev_ops->vdev_op_leaf)
return (1);
for (int c = 0; c < vd->vdev_children; c++)
n += vdev_count_leaves_impl(vd->vdev_child[c]);
return (n);
}
int
vdev_count_leaves(spa_t *spa)
{
int rc;
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
rc = vdev_count_leaves_impl(spa->spa_root_vdev);
spa_config_exit(spa, SCL_VDEV, FTAG);
return (rc);
}
void
vdev_add_child(vdev_t *pvd, vdev_t *cvd)
{
size_t oldsize, newsize;
uint64_t id = cvd->vdev_id;
vdev_t **newchild;
ASSERT(spa_config_held(cvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
ASSERT(cvd->vdev_parent == NULL);
cvd->vdev_parent = pvd;
if (pvd == NULL)
return;
ASSERT(id >= pvd->vdev_children || pvd->vdev_child[id] == NULL);
oldsize = pvd->vdev_children * sizeof (vdev_t *);
pvd->vdev_children = MAX(pvd->vdev_children, id + 1);
newsize = pvd->vdev_children * sizeof (vdev_t *);
newchild = kmem_alloc(newsize, KM_SLEEP);
if (pvd->vdev_child != NULL) {
bcopy(pvd->vdev_child, newchild, oldsize);
kmem_free(pvd->vdev_child, oldsize);
}
pvd->vdev_child = newchild;
pvd->vdev_child[id] = cvd;
cvd->vdev_top = (pvd->vdev_top ? pvd->vdev_top: cvd);
ASSERT(cvd->vdev_top->vdev_parent->vdev_parent == NULL);
/*
* Walk up all ancestors to update guid sum.
*/
for (; pvd != NULL; pvd = pvd->vdev_parent)
pvd->vdev_guid_sum += cvd->vdev_guid_sum;
if (cvd->vdev_ops->vdev_op_leaf) {
list_insert_head(&cvd->vdev_spa->spa_leaf_list, cvd);
cvd->vdev_spa->spa_leaf_list_gen++;
}
}
void
vdev_remove_child(vdev_t *pvd, vdev_t *cvd)
{
int c;
uint_t id = cvd->vdev_id;
ASSERT(cvd->vdev_parent == pvd);
if (pvd == NULL)
return;
ASSERT(id < pvd->vdev_children);
ASSERT(pvd->vdev_child[id] == cvd);
pvd->vdev_child[id] = NULL;
cvd->vdev_parent = NULL;
for (c = 0; c < pvd->vdev_children; c++)
if (pvd->vdev_child[c])
break;
if (c == pvd->vdev_children) {
kmem_free(pvd->vdev_child, c * sizeof (vdev_t *));
pvd->vdev_child = NULL;
pvd->vdev_children = 0;
}
if (cvd->vdev_ops->vdev_op_leaf) {
spa_t *spa = cvd->vdev_spa;
list_remove(&spa->spa_leaf_list, cvd);
spa->spa_leaf_list_gen++;
}
/*
* Walk up all ancestors to update guid sum.
*/
for (; pvd != NULL; pvd = pvd->vdev_parent)
pvd->vdev_guid_sum -= cvd->vdev_guid_sum;
}
/*
* Remove any holes in the child array.
*/
void
vdev_compact_children(vdev_t *pvd)
{
vdev_t **newchild, *cvd;
int oldc = pvd->vdev_children;
int newc;
ASSERT(spa_config_held(pvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
if (oldc == 0)
return;
for (int c = newc = 0; c < oldc; c++)
if (pvd->vdev_child[c])
newc++;
if (newc > 0) {
newchild = kmem_zalloc(newc * sizeof (vdev_t *), KM_SLEEP);
for (int c = newc = 0; c < oldc; c++) {
if ((cvd = pvd->vdev_child[c]) != NULL) {
newchild[newc] = cvd;
cvd->vdev_id = newc++;
}
}
} else {
newchild = NULL;
}
kmem_free(pvd->vdev_child, oldc * sizeof (vdev_t *));
pvd->vdev_child = newchild;
pvd->vdev_children = newc;
}
/*
* Allocate and minimally initialize a vdev_t.
*/
vdev_t *
vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops)
{
vdev_t *vd;
vdev_indirect_config_t *vic;
vd = kmem_zalloc(sizeof (vdev_t), KM_SLEEP);
vic = &vd->vdev_indirect_config;
if (spa->spa_root_vdev == NULL) {
ASSERT(ops == &vdev_root_ops);
spa->spa_root_vdev = vd;
spa->spa_load_guid = spa_generate_guid(NULL);
}
if (guid == 0 && ops != &vdev_hole_ops) {
if (spa->spa_root_vdev == vd) {
/*
* The root vdev's guid will also be the pool guid,
* which must be unique among all pools.
*/
guid = spa_generate_guid(NULL);
} else {
/*
* Any other vdev's guid must be unique within the pool.
*/
guid = spa_generate_guid(spa);
}
ASSERT(!spa_guid_exists(spa_guid(spa), guid));
}
vd->vdev_spa = spa;
vd->vdev_id = id;
vd->vdev_guid = guid;
vd->vdev_guid_sum = guid;
vd->vdev_ops = ops;
vd->vdev_state = VDEV_STATE_CLOSED;
vd->vdev_ishole = (ops == &vdev_hole_ops);
vic->vic_prev_indirect_vdev = UINT64_MAX;
rw_init(&vd->vdev_indirect_rwlock, NULL, RW_DEFAULT, NULL);
mutex_init(&vd->vdev_obsolete_lock, NULL, MUTEX_DEFAULT, NULL);
vd->vdev_obsolete_segments = range_tree_create(NULL, RANGE_SEG64, NULL,
0, 0);
/*
* Initialize rate limit structs for events. We rate limit ZIO delay
* and checksum events so that we don't overwhelm ZED with thousands
* of events when a disk is acting up.
*/
zfs_ratelimit_init(&vd->vdev_delay_rl, &zfs_slow_io_events_per_second,
1);
zfs_ratelimit_init(&vd->vdev_deadman_rl, &zfs_slow_io_events_per_second,
1);
zfs_ratelimit_init(&vd->vdev_checksum_rl,
&zfs_checksum_events_per_second, 1);
list_link_init(&vd->vdev_config_dirty_node);
list_link_init(&vd->vdev_state_dirty_node);
list_link_init(&vd->vdev_initialize_node);
list_link_init(&vd->vdev_leaf_node);
list_link_init(&vd->vdev_trim_node);
mutex_init(&vd->vdev_dtl_lock, NULL, MUTEX_NOLOCKDEP, NULL);
mutex_init(&vd->vdev_stat_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&vd->vdev_probe_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&vd->vdev_scan_io_queue_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&vd->vdev_initialize_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&vd->vdev_initialize_io_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&vd->vdev_initialize_cv, NULL, CV_DEFAULT, NULL);
cv_init(&vd->vdev_initialize_io_cv, NULL, CV_DEFAULT, NULL);
mutex_init(&vd->vdev_trim_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&vd->vdev_autotrim_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&vd->vdev_trim_io_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&vd->vdev_trim_cv, NULL, CV_DEFAULT, NULL);
cv_init(&vd->vdev_autotrim_cv, NULL, CV_DEFAULT, NULL);
cv_init(&vd->vdev_trim_io_cv, NULL, CV_DEFAULT, NULL);
mutex_init(&vd->vdev_rebuild_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&vd->vdev_rebuild_cv, NULL, CV_DEFAULT, NULL);
for (int t = 0; t < DTL_TYPES; t++) {
vd->vdev_dtl[t] = range_tree_create(NULL, RANGE_SEG64, NULL, 0,
0);
}
txg_list_create(&vd->vdev_ms_list, spa,
offsetof(struct metaslab, ms_txg_node));
txg_list_create(&vd->vdev_dtl_list, spa,
offsetof(struct vdev, vdev_dtl_node));
vd->vdev_stat.vs_timestamp = gethrtime();
vdev_queue_init(vd);
vdev_cache_init(vd);
return (vd);
}
/*
* Allocate a new vdev. The 'alloctype' is used to control whether we are
* creating a new vdev or loading an existing one - the behavior is slightly
* different for each case.
*/
int
vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
int alloctype)
{
vdev_ops_t *ops;
char *type;
uint64_t guid = 0, islog;
vdev_t *vd;
vdev_indirect_config_t *vic;
char *tmp = NULL;
int rc;
vdev_alloc_bias_t alloc_bias = VDEV_BIAS_NONE;
boolean_t top_level = (parent && !parent->vdev_parent);
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
return (SET_ERROR(EINVAL));
if ((ops = vdev_getops(type)) == NULL)
return (SET_ERROR(EINVAL));
/*
* If this is a load, get the vdev guid from the nvlist.
* Otherwise, vdev_alloc_common() will generate one for us.
*/
if (alloctype == VDEV_ALLOC_LOAD) {
uint64_t label_id;
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, &label_id) ||
label_id != id)
return (SET_ERROR(EINVAL));
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
return (SET_ERROR(EINVAL));
} else if (alloctype == VDEV_ALLOC_SPARE) {
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
return (SET_ERROR(EINVAL));
} else if (alloctype == VDEV_ALLOC_L2CACHE) {
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
return (SET_ERROR(EINVAL));
} else if (alloctype == VDEV_ALLOC_ROOTPOOL) {
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
return (SET_ERROR(EINVAL));
}
/*
* The first allocated vdev must be of type 'root'.
*/
if (ops != &vdev_root_ops && spa->spa_root_vdev == NULL)
return (SET_ERROR(EINVAL));
/*
* Determine whether we're a log vdev.
*/
islog = 0;
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &islog);
if (islog && spa_version(spa) < SPA_VERSION_SLOGS)
return (SET_ERROR(ENOTSUP));
if (ops == &vdev_hole_ops && spa_version(spa) < SPA_VERSION_HOLES)
return (SET_ERROR(ENOTSUP));
if (top_level && alloctype == VDEV_ALLOC_ADD) {
char *bias;
/*
* If creating a top-level vdev, check for allocation
* classes input.
*/
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_ALLOCATION_BIAS,
&bias) == 0) {
alloc_bias = vdev_derive_alloc_bias(bias);
/* spa_vdev_add() expects feature to be enabled */
if (spa->spa_load_state != SPA_LOAD_CREATE &&
!spa_feature_is_enabled(spa,
SPA_FEATURE_ALLOCATION_CLASSES)) {
return (SET_ERROR(ENOTSUP));
}
}
/* spa_vdev_add() expects feature to be enabled */
if (ops == &vdev_draid_ops &&
spa->spa_load_state != SPA_LOAD_CREATE &&
!spa_feature_is_enabled(spa, SPA_FEATURE_DRAID)) {
return (SET_ERROR(ENOTSUP));
}
}
/*
* Initialize the vdev specific data. This is done before calling
* vdev_alloc_common() since it may fail and this simplifies the
* error reporting and cleanup code paths.
*/
void *tsd = NULL;
if (ops->vdev_op_init != NULL) {
rc = ops->vdev_op_init(spa, nv, &tsd);
if (rc != 0) {
return (rc);
}
}
vd = vdev_alloc_common(spa, id, guid, ops);
vd->vdev_tsd = tsd;
vd->vdev_islog = islog;
if (top_level && alloc_bias != VDEV_BIAS_NONE)
vd->vdev_alloc_bias = alloc_bias;
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &vd->vdev_path) == 0)
vd->vdev_path = spa_strdup(vd->vdev_path);
/*
* ZPOOL_CONFIG_AUX_STATE = "external" means we previously forced a
* fault on a vdev and want it to persist across imports (like with
* zpool offline -f).
*/
rc = nvlist_lookup_string(nv, ZPOOL_CONFIG_AUX_STATE, &tmp);
if (rc == 0 && tmp != NULL && strcmp(tmp, "external") == 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_EXTERNAL;
vd->vdev_faulted = 1;
vd->vdev_label_aux = VDEV_AUX_EXTERNAL;
}
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &vd->vdev_devid) == 0)
vd->vdev_devid = spa_strdup(vd->vdev_devid);
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH,
&vd->vdev_physpath) == 0)
vd->vdev_physpath = spa_strdup(vd->vdev_physpath);
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH,
&vd->vdev_enc_sysfs_path) == 0)
vd->vdev_enc_sysfs_path = spa_strdup(vd->vdev_enc_sysfs_path);
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_FRU, &vd->vdev_fru) == 0)
vd->vdev_fru = spa_strdup(vd->vdev_fru);
/*
* Set the whole_disk property. If it's not specified, leave the value
* as -1.
*/
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
&vd->vdev_wholedisk) != 0)
vd->vdev_wholedisk = -1ULL;
vic = &vd->vdev_indirect_config;
ASSERT0(vic->vic_mapping_object);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_INDIRECT_OBJECT,
&vic->vic_mapping_object);
ASSERT0(vic->vic_births_object);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_INDIRECT_BIRTHS,
&vic->vic_births_object);
ASSERT3U(vic->vic_prev_indirect_vdev, ==, UINT64_MAX);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_PREV_INDIRECT_VDEV,
&vic->vic_prev_indirect_vdev);
/*
* Look for the 'not present' flag. This will only be set if the device
* was not present at the time of import.
*/
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
&vd->vdev_not_present);
/*
* Get the alignment requirement.
*/
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASHIFT, &vd->vdev_ashift);
/*
* Retrieve the vdev creation time.
*/
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_CREATE_TXG,
&vd->vdev_crtxg);
/*
* If we're a top-level vdev, try to load the allocation parameters.
*/
if (top_level &&
(alloctype == VDEV_ALLOC_LOAD || alloctype == VDEV_ALLOC_SPLIT)) {
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY,
&vd->vdev_ms_array);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT,
&vd->vdev_ms_shift);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASIZE,
&vd->vdev_asize);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NONALLOCATING,
&vd->vdev_noalloc);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVING,
&vd->vdev_removing);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP,
&vd->vdev_top_zap);
} else {
ASSERT0(vd->vdev_top_zap);
}
if (top_level && alloctype != VDEV_ALLOC_ATTACH) {
ASSERT(alloctype == VDEV_ALLOC_LOAD ||
alloctype == VDEV_ALLOC_ADD ||
alloctype == VDEV_ALLOC_SPLIT ||
alloctype == VDEV_ALLOC_ROOTPOOL);
/* Note: metaslab_group_create() is now deferred */
}
if (vd->vdev_ops->vdev_op_leaf &&
(alloctype == VDEV_ALLOC_LOAD || alloctype == VDEV_ALLOC_SPLIT)) {
(void) nvlist_lookup_uint64(nv,
ZPOOL_CONFIG_VDEV_LEAF_ZAP, &vd->vdev_leaf_zap);
} else {
ASSERT0(vd->vdev_leaf_zap);
}
/*
* If we're a leaf vdev, try to load the DTL object and other state.
*/
if (vd->vdev_ops->vdev_op_leaf &&
(alloctype == VDEV_ALLOC_LOAD || alloctype == VDEV_ALLOC_L2CACHE ||
alloctype == VDEV_ALLOC_ROOTPOOL)) {
if (alloctype == VDEV_ALLOC_LOAD) {
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DTL,
&vd->vdev_dtl_object);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_UNSPARE,
&vd->vdev_unspare);
}
if (alloctype == VDEV_ALLOC_ROOTPOOL) {
uint64_t spare = 0;
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
&spare) == 0 && spare)
spa_spare_add(vd);
}
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE,
&vd->vdev_offline);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_RESILVER_TXG,
&vd->vdev_resilver_txg);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REBUILD_TXG,
&vd->vdev_rebuild_txg);
if (nvlist_exists(nv, ZPOOL_CONFIG_RESILVER_DEFER))
vdev_defer_resilver(vd);
/*
* In general, when importing a pool we want to ignore the
* persistent fault state, as the diagnosis made on another
* system may not be valid in the current context. The only
* exception is if we forced a vdev to a persistently faulted
* state with 'zpool offline -f'. The persistent fault will
* remain across imports until cleared.
*
* Local vdevs will remain in the faulted state.
*/
if (spa_load_state(spa) == SPA_LOAD_OPEN ||
spa_load_state(spa) == SPA_LOAD_IMPORT) {
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED,
&vd->vdev_faulted);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DEGRADED,
&vd->vdev_degraded);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED,
&vd->vdev_removed);
if (vd->vdev_faulted || vd->vdev_degraded) {
char *aux;
vd->vdev_label_aux =
VDEV_AUX_ERR_EXCEEDED;
if (nvlist_lookup_string(nv,
ZPOOL_CONFIG_AUX_STATE, &aux) == 0 &&
strcmp(aux, "external") == 0)
vd->vdev_label_aux = VDEV_AUX_EXTERNAL;
else
vd->vdev_faulted = 0ULL;
}
}
}
/*
* Add ourselves to the parent's list of children.
*/
vdev_add_child(parent, vd);
*vdp = vd;
return (0);
}
void
vdev_free(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
ASSERT3P(vd->vdev_trim_thread, ==, NULL);
ASSERT3P(vd->vdev_autotrim_thread, ==, NULL);
ASSERT3P(vd->vdev_rebuild_thread, ==, NULL);
/*
* Scan queues are normally destroyed at the end of a scan. If the
* queue exists here, that implies the vdev is being removed while
* the scan is still running.
*/
if (vd->vdev_scan_io_queue != NULL) {
mutex_enter(&vd->vdev_scan_io_queue_lock);
dsl_scan_io_queue_destroy(vd->vdev_scan_io_queue);
vd->vdev_scan_io_queue = NULL;
mutex_exit(&vd->vdev_scan_io_queue_lock);
}
/*
* vdev_free() implies closing the vdev first. This is simpler than
* trying to ensure complicated semantics for all callers.
*/
vdev_close(vd);
ASSERT(!list_link_active(&vd->vdev_config_dirty_node));
ASSERT(!list_link_active(&vd->vdev_state_dirty_node));
/*
* Free all children.
*/
for (int c = 0; c < vd->vdev_children; c++)
vdev_free(vd->vdev_child[c]);
ASSERT(vd->vdev_child == NULL);
ASSERT(vd->vdev_guid_sum == vd->vdev_guid);
if (vd->vdev_ops->vdev_op_fini != NULL)
vd->vdev_ops->vdev_op_fini(vd);
/*
* Discard allocation state.
*/
if (vd->vdev_mg != NULL) {
vdev_metaslab_fini(vd);
metaslab_group_destroy(vd->vdev_mg);
vd->vdev_mg = NULL;
}
if (vd->vdev_log_mg != NULL) {
ASSERT0(vd->vdev_ms_count);
metaslab_group_destroy(vd->vdev_log_mg);
vd->vdev_log_mg = NULL;
}
ASSERT0(vd->vdev_stat.vs_space);
ASSERT0(vd->vdev_stat.vs_dspace);
ASSERT0(vd->vdev_stat.vs_alloc);
/*
* Remove this vdev from its parent's child list.
*/
vdev_remove_child(vd->vdev_parent, vd);
ASSERT(vd->vdev_parent == NULL);
ASSERT(!list_link_active(&vd->vdev_leaf_node));
/*
* Clean up vdev structure.
*/
vdev_queue_fini(vd);
vdev_cache_fini(vd);
if (vd->vdev_path)
spa_strfree(vd->vdev_path);
if (vd->vdev_devid)
spa_strfree(vd->vdev_devid);
if (vd->vdev_physpath)
spa_strfree(vd->vdev_physpath);
if (vd->vdev_enc_sysfs_path)
spa_strfree(vd->vdev_enc_sysfs_path);
if (vd->vdev_fru)
spa_strfree(vd->vdev_fru);
if (vd->vdev_isspare)
spa_spare_remove(vd);
if (vd->vdev_isl2cache)
spa_l2cache_remove(vd);
txg_list_destroy(&vd->vdev_ms_list);
txg_list_destroy(&vd->vdev_dtl_list);
mutex_enter(&vd->vdev_dtl_lock);
space_map_close(vd->vdev_dtl_sm);
for (int t = 0; t < DTL_TYPES; t++) {
range_tree_vacate(vd->vdev_dtl[t], NULL, NULL);
range_tree_destroy(vd->vdev_dtl[t]);
}
mutex_exit(&vd->vdev_dtl_lock);
EQUIV(vd->vdev_indirect_births != NULL,
vd->vdev_indirect_mapping != NULL);
if (vd->vdev_indirect_births != NULL) {
vdev_indirect_mapping_close(vd->vdev_indirect_mapping);
vdev_indirect_births_close(vd->vdev_indirect_births);
}
if (vd->vdev_obsolete_sm != NULL) {
ASSERT(vd->vdev_removing ||
vd->vdev_ops == &vdev_indirect_ops);
space_map_close(vd->vdev_obsolete_sm);
vd->vdev_obsolete_sm = NULL;
}
range_tree_destroy(vd->vdev_obsolete_segments);
rw_destroy(&vd->vdev_indirect_rwlock);
mutex_destroy(&vd->vdev_obsolete_lock);
mutex_destroy(&vd->vdev_dtl_lock);
mutex_destroy(&vd->vdev_stat_lock);
mutex_destroy(&vd->vdev_probe_lock);
mutex_destroy(&vd->vdev_scan_io_queue_lock);
mutex_destroy(&vd->vdev_initialize_lock);
mutex_destroy(&vd->vdev_initialize_io_lock);
cv_destroy(&vd->vdev_initialize_io_cv);
cv_destroy(&vd->vdev_initialize_cv);
mutex_destroy(&vd->vdev_trim_lock);
mutex_destroy(&vd->vdev_autotrim_lock);
mutex_destroy(&vd->vdev_trim_io_lock);
cv_destroy(&vd->vdev_trim_cv);
cv_destroy(&vd->vdev_autotrim_cv);
cv_destroy(&vd->vdev_trim_io_cv);
mutex_destroy(&vd->vdev_rebuild_lock);
cv_destroy(&vd->vdev_rebuild_cv);
zfs_ratelimit_fini(&vd->vdev_delay_rl);
zfs_ratelimit_fini(&vd->vdev_deadman_rl);
zfs_ratelimit_fini(&vd->vdev_checksum_rl);
if (vd == spa->spa_root_vdev)
spa->spa_root_vdev = NULL;
kmem_free(vd, sizeof (vdev_t));
}
/*
* Transfer top-level vdev state from svd to tvd.
*/
static void
vdev_top_transfer(vdev_t *svd, vdev_t *tvd)
{
spa_t *spa = svd->vdev_spa;
metaslab_t *msp;
vdev_t *vd;
int t;
ASSERT(tvd == tvd->vdev_top);
tvd->vdev_pending_fastwrite = svd->vdev_pending_fastwrite;
tvd->vdev_ms_array = svd->vdev_ms_array;
tvd->vdev_ms_shift = svd->vdev_ms_shift;
tvd->vdev_ms_count = svd->vdev_ms_count;
tvd->vdev_top_zap = svd->vdev_top_zap;
svd->vdev_ms_array = 0;
svd->vdev_ms_shift = 0;
svd->vdev_ms_count = 0;
svd->vdev_top_zap = 0;
if (tvd->vdev_mg)
ASSERT3P(tvd->vdev_mg, ==, svd->vdev_mg);
if (tvd->vdev_log_mg)
ASSERT3P(tvd->vdev_log_mg, ==, svd->vdev_log_mg);
tvd->vdev_mg = svd->vdev_mg;
tvd->vdev_log_mg = svd->vdev_log_mg;
tvd->vdev_ms = svd->vdev_ms;
svd->vdev_mg = NULL;
svd->vdev_log_mg = NULL;
svd->vdev_ms = NULL;
if (tvd->vdev_mg != NULL)
tvd->vdev_mg->mg_vd = tvd;
if (tvd->vdev_log_mg != NULL)
tvd->vdev_log_mg->mg_vd = tvd;
tvd->vdev_checkpoint_sm = svd->vdev_checkpoint_sm;
svd->vdev_checkpoint_sm = NULL;
tvd->vdev_alloc_bias = svd->vdev_alloc_bias;
svd->vdev_alloc_bias = VDEV_BIAS_NONE;
tvd->vdev_stat.vs_alloc = svd->vdev_stat.vs_alloc;
tvd->vdev_stat.vs_space = svd->vdev_stat.vs_space;
tvd->vdev_stat.vs_dspace = svd->vdev_stat.vs_dspace;
svd->vdev_stat.vs_alloc = 0;
svd->vdev_stat.vs_space = 0;
svd->vdev_stat.vs_dspace = 0;
/*
* State which may be set on a top-level vdev that's in the
* process of being removed.
*/
ASSERT0(tvd->vdev_indirect_config.vic_births_object);
ASSERT0(tvd->vdev_indirect_config.vic_mapping_object);
ASSERT3U(tvd->vdev_indirect_config.vic_prev_indirect_vdev, ==, -1ULL);
ASSERT3P(tvd->vdev_indirect_mapping, ==, NULL);
ASSERT3P(tvd->vdev_indirect_births, ==, NULL);
ASSERT3P(tvd->vdev_obsolete_sm, ==, NULL);
ASSERT0(tvd->vdev_noalloc);
ASSERT0(tvd->vdev_removing);
ASSERT0(tvd->vdev_rebuilding);
tvd->vdev_noalloc = svd->vdev_noalloc;
tvd->vdev_removing = svd->vdev_removing;
tvd->vdev_rebuilding = svd->vdev_rebuilding;
tvd->vdev_rebuild_config = svd->vdev_rebuild_config;
tvd->vdev_indirect_config = svd->vdev_indirect_config;
tvd->vdev_indirect_mapping = svd->vdev_indirect_mapping;
tvd->vdev_indirect_births = svd->vdev_indirect_births;
range_tree_swap(&svd->vdev_obsolete_segments,
&tvd->vdev_obsolete_segments);
tvd->vdev_obsolete_sm = svd->vdev_obsolete_sm;
svd->vdev_indirect_config.vic_mapping_object = 0;
svd->vdev_indirect_config.vic_births_object = 0;
svd->vdev_indirect_config.vic_prev_indirect_vdev = -1ULL;
svd->vdev_indirect_mapping = NULL;
svd->vdev_indirect_births = NULL;
svd->vdev_obsolete_sm = NULL;
svd->vdev_noalloc = 0;
svd->vdev_removing = 0;
svd->vdev_rebuilding = 0;
for (t = 0; t < TXG_SIZE; t++) {
while ((msp = txg_list_remove(&svd->vdev_ms_list, t)) != NULL)
(void) txg_list_add(&tvd->vdev_ms_list, msp, t);
while ((vd = txg_list_remove(&svd->vdev_dtl_list, t)) != NULL)
(void) txg_list_add(&tvd->vdev_dtl_list, vd, t);
if (txg_list_remove_this(&spa->spa_vdev_txg_list, svd, t))
(void) txg_list_add(&spa->spa_vdev_txg_list, tvd, t);
}
if (list_link_active(&svd->vdev_config_dirty_node)) {
vdev_config_clean(svd);
vdev_config_dirty(tvd);
}
if (list_link_active(&svd->vdev_state_dirty_node)) {
vdev_state_clean(svd);
vdev_state_dirty(tvd);
}
tvd->vdev_deflate_ratio = svd->vdev_deflate_ratio;
svd->vdev_deflate_ratio = 0;
tvd->vdev_islog = svd->vdev_islog;
svd->vdev_islog = 0;
dsl_scan_io_queue_vdev_xfer(svd, tvd);
}
static void
vdev_top_update(vdev_t *tvd, vdev_t *vd)
{
if (vd == NULL)
return;
vd->vdev_top = tvd;
for (int c = 0; c < vd->vdev_children; c++)
vdev_top_update(tvd, vd->vdev_child[c]);
}
/*
* Add a mirror/replacing vdev above an existing vdev. There is no need to
* call .vdev_op_init() since mirror/replacing vdevs do not have private state.
*/
vdev_t *
vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops)
{
spa_t *spa = cvd->vdev_spa;
vdev_t *pvd = cvd->vdev_parent;
vdev_t *mvd;
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
mvd = vdev_alloc_common(spa, cvd->vdev_id, 0, ops);
mvd->vdev_asize = cvd->vdev_asize;
mvd->vdev_min_asize = cvd->vdev_min_asize;
mvd->vdev_max_asize = cvd->vdev_max_asize;
mvd->vdev_psize = cvd->vdev_psize;
mvd->vdev_ashift = cvd->vdev_ashift;
mvd->vdev_logical_ashift = cvd->vdev_logical_ashift;
mvd->vdev_physical_ashift = cvd->vdev_physical_ashift;
mvd->vdev_state = cvd->vdev_state;
mvd->vdev_crtxg = cvd->vdev_crtxg;
vdev_remove_child(pvd, cvd);
vdev_add_child(pvd, mvd);
cvd->vdev_id = mvd->vdev_children;
vdev_add_child(mvd, cvd);
vdev_top_update(cvd->vdev_top, cvd->vdev_top);
if (mvd == mvd->vdev_top)
vdev_top_transfer(cvd, mvd);
return (mvd);
}
/*
* Remove a 1-way mirror/replacing vdev from the tree.
*/
void
vdev_remove_parent(vdev_t *cvd)
{
vdev_t *mvd = cvd->vdev_parent;
vdev_t *pvd = mvd->vdev_parent;
ASSERT(spa_config_held(cvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
ASSERT(mvd->vdev_children == 1);
ASSERT(mvd->vdev_ops == &vdev_mirror_ops ||
mvd->vdev_ops == &vdev_replacing_ops ||
mvd->vdev_ops == &vdev_spare_ops);
cvd->vdev_ashift = mvd->vdev_ashift;
cvd->vdev_logical_ashift = mvd->vdev_logical_ashift;
cvd->vdev_physical_ashift = mvd->vdev_physical_ashift;
vdev_remove_child(mvd, cvd);
vdev_remove_child(pvd, mvd);
/*
* If cvd will replace mvd as a top-level vdev, preserve mvd's guid.
* Otherwise, we could have detached an offline device, and when we
* go to import the pool we'll think we have two top-level vdevs,
* instead of a different version of the same top-level vdev.
*/
if (mvd->vdev_top == mvd) {
uint64_t guid_delta = mvd->vdev_guid - cvd->vdev_guid;
cvd->vdev_orig_guid = cvd->vdev_guid;
cvd->vdev_guid += guid_delta;
cvd->vdev_guid_sum += guid_delta;
/*
* If pool not set for autoexpand, we need to also preserve
* mvd's asize to prevent automatic expansion of cvd.
* Otherwise if we are adjusting the mirror by attaching and
* detaching children of non-uniform sizes, the mirror could
* autoexpand, unexpectedly requiring larger devices to
* re-establish the mirror.
*/
if (!cvd->vdev_spa->spa_autoexpand)
cvd->vdev_asize = mvd->vdev_asize;
}
cvd->vdev_id = mvd->vdev_id;
vdev_add_child(pvd, cvd);
vdev_top_update(cvd->vdev_top, cvd->vdev_top);
if (cvd == cvd->vdev_top)
vdev_top_transfer(mvd, cvd);
ASSERT(mvd->vdev_children == 0);
vdev_free(mvd);
}
void
vdev_metaslab_group_create(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
/*
* metaslab_group_create was delayed until allocation bias was available
*/
if (vd->vdev_mg == NULL) {
metaslab_class_t *mc;
if (vd->vdev_islog && vd->vdev_alloc_bias == VDEV_BIAS_NONE)
vd->vdev_alloc_bias = VDEV_BIAS_LOG;
ASSERT3U(vd->vdev_islog, ==,
(vd->vdev_alloc_bias == VDEV_BIAS_LOG));
switch (vd->vdev_alloc_bias) {
case VDEV_BIAS_LOG:
mc = spa_log_class(spa);
break;
case VDEV_BIAS_SPECIAL:
mc = spa_special_class(spa);
break;
case VDEV_BIAS_DEDUP:
mc = spa_dedup_class(spa);
break;
default:
mc = spa_normal_class(spa);
}
vd->vdev_mg = metaslab_group_create(mc, vd,
spa->spa_alloc_count);
if (!vd->vdev_islog) {
vd->vdev_log_mg = metaslab_group_create(
spa_embedded_log_class(spa), vd, 1);
}
/*
* The spa ashift min/max only apply for the normal metaslab
* class. Class destination is late binding so ashift boundary
* setting had to wait until now.
*/
if (vd->vdev_top == vd && vd->vdev_ashift != 0 &&
mc == spa_normal_class(spa) && vd->vdev_aux == NULL) {
if (vd->vdev_ashift > spa->spa_max_ashift)
spa->spa_max_ashift = vd->vdev_ashift;
if (vd->vdev_ashift < spa->spa_min_ashift)
spa->spa_min_ashift = vd->vdev_ashift;
uint64_t min_alloc = vdev_get_min_alloc(vd);
if (min_alloc < spa->spa_min_alloc)
spa->spa_min_alloc = min_alloc;
}
}
}
int
vdev_metaslab_init(vdev_t *vd, uint64_t txg)
{
spa_t *spa = vd->vdev_spa;
uint64_t oldc = vd->vdev_ms_count;
uint64_t newc = vd->vdev_asize >> vd->vdev_ms_shift;
metaslab_t **mspp;
int error;
boolean_t expanding = (oldc != 0);
ASSERT(txg == 0 || spa_config_held(spa, SCL_ALLOC, RW_WRITER));
/*
* This vdev is not being allocated from yet or is a hole.
*/
if (vd->vdev_ms_shift == 0)
return (0);
ASSERT(!vd->vdev_ishole);
ASSERT(oldc <= newc);
mspp = vmem_zalloc(newc * sizeof (*mspp), KM_SLEEP);
if (expanding) {
bcopy(vd->vdev_ms, mspp, oldc * sizeof (*mspp));
vmem_free(vd->vdev_ms, oldc * sizeof (*mspp));
}
vd->vdev_ms = mspp;
vd->vdev_ms_count = newc;
for (uint64_t m = oldc; m < newc; m++) {
uint64_t object = 0;
/*
* vdev_ms_array may be 0 if we are creating the "fake"
* metaslabs for an indirect vdev for zdb's leak detection.
* See zdb_leak_init().
*/
if (txg == 0 && vd->vdev_ms_array != 0) {
error = dmu_read(spa->spa_meta_objset,
vd->vdev_ms_array,
m * sizeof (uint64_t), sizeof (uint64_t), &object,
DMU_READ_PREFETCH);
if (error != 0) {
vdev_dbgmsg(vd, "unable to read the metaslab "
"array [error=%d]", error);
return (error);
}
}
error = metaslab_init(vd->vdev_mg, m, object, txg,
&(vd->vdev_ms[m]));
if (error != 0) {
vdev_dbgmsg(vd, "metaslab_init failed [error=%d]",
error);
return (error);
}
}
/*
* Find the emptiest metaslab on the vdev and mark it for use for
* embedded slog by moving it from the regular to the log metaslab
* group.
*/
if (vd->vdev_mg->mg_class == spa_normal_class(spa) &&
vd->vdev_ms_count > zfs_embedded_slog_min_ms &&
avl_is_empty(&vd->vdev_log_mg->mg_metaslab_tree)) {
uint64_t slog_msid = 0;
uint64_t smallest = UINT64_MAX;
/*
* Note, we only search the new metaslabs, because the old
* (pre-existing) ones may be active (e.g. have non-empty
* range_tree's), and we don't move them to the new
* metaslab_t.
*/
for (uint64_t m = oldc; m < newc; m++) {
uint64_t alloc =
space_map_allocated(vd->vdev_ms[m]->ms_sm);
if (alloc < smallest) {
slog_msid = m;
smallest = alloc;
}
}
metaslab_t *slog_ms = vd->vdev_ms[slog_msid];
/*
* The metaslab was marked as dirty at the end of
* metaslab_init(). Remove it from the dirty list so that we
* can uninitialize and reinitialize it to the new class.
*/
if (txg != 0) {
(void) txg_list_remove_this(&vd->vdev_ms_list,
slog_ms, txg);
}
uint64_t sm_obj = space_map_object(slog_ms->ms_sm);
metaslab_fini(slog_ms);
VERIFY0(metaslab_init(vd->vdev_log_mg, slog_msid, sm_obj, txg,
&vd->vdev_ms[slog_msid]));
}
if (txg == 0)
spa_config_enter(spa, SCL_ALLOC, FTAG, RW_WRITER);
/*
* If the vdev is marked as non-allocating then don't
* activate the metaslabs since we want to ensure that
* no allocations are performed on this device.
*/
if (vd->vdev_noalloc) {
/* track non-allocating vdev space */
spa->spa_nonallocating_dspace += spa_deflate(spa) ?
vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;
} else if (!expanding) {
metaslab_group_activate(vd->vdev_mg);
if (vd->vdev_log_mg != NULL)
metaslab_group_activate(vd->vdev_log_mg);
}
if (txg == 0)
spa_config_exit(spa, SCL_ALLOC, FTAG);
/*
* Regardless whether this vdev was just added or it is being
* expanded, the metaslab count has changed. Recalculate the
* block limit.
*/
spa_log_sm_set_blocklimit(spa);
return (0);
}
void
vdev_metaslab_fini(vdev_t *vd)
{
if (vd->vdev_checkpoint_sm != NULL) {
ASSERT(spa_feature_is_active(vd->vdev_spa,
SPA_FEATURE_POOL_CHECKPOINT));
space_map_close(vd->vdev_checkpoint_sm);
/*
* Even though we close the space map, we need to set its
* pointer to NULL. The reason is that vdev_metaslab_fini()
* may be called multiple times for certain operations
* (i.e. when destroying a pool) so we need to ensure that
* this clause never executes twice. This logic is similar
* to the one used for the vdev_ms clause below.
*/
vd->vdev_checkpoint_sm = NULL;
}
if (vd->vdev_ms != NULL) {
metaslab_group_t *mg = vd->vdev_mg;
metaslab_group_passivate(mg);
if (vd->vdev_log_mg != NULL) {
ASSERT(!vd->vdev_islog);
metaslab_group_passivate(vd->vdev_log_mg);
}
uint64_t count = vd->vdev_ms_count;
for (uint64_t m = 0; m < count; m++) {
metaslab_t *msp = vd->vdev_ms[m];
if (msp != NULL)
metaslab_fini(msp);
}
vmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
vd->vdev_ms = NULL;
vd->vdev_ms_count = 0;
for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
ASSERT0(mg->mg_histogram[i]);
if (vd->vdev_log_mg != NULL)
ASSERT0(vd->vdev_log_mg->mg_histogram[i]);
}
}
ASSERT0(vd->vdev_ms_count);
ASSERT3U(vd->vdev_pending_fastwrite, ==, 0);
}
typedef struct vdev_probe_stats {
boolean_t vps_readable;
boolean_t vps_writeable;
int vps_flags;
} vdev_probe_stats_t;
static void
vdev_probe_done(zio_t *zio)
{
spa_t *spa = zio->io_spa;
vdev_t *vd = zio->io_vd;
vdev_probe_stats_t *vps = zio->io_private;
ASSERT(vd->vdev_probe_zio != NULL);
if (zio->io_type == ZIO_TYPE_READ) {
if (zio->io_error == 0)
vps->vps_readable = 1;
if (zio->io_error == 0 && spa_writeable(spa)) {
zio_nowait(zio_write_phys(vd->vdev_probe_zio, vd,
zio->io_offset, zio->io_size, zio->io_abd,
ZIO_CHECKSUM_OFF, vdev_probe_done, vps,
ZIO_PRIORITY_SYNC_WRITE, vps->vps_flags, B_TRUE));
} else {
abd_free(zio->io_abd);
}
} else if (zio->io_type == ZIO_TYPE_WRITE) {
if (zio->io_error == 0)
vps->vps_writeable = 1;
abd_free(zio->io_abd);
} else if (zio->io_type == ZIO_TYPE_NULL) {
zio_t *pio;
zio_link_t *zl;
vd->vdev_cant_read |= !vps->vps_readable;
vd->vdev_cant_write |= !vps->vps_writeable;
if (vdev_readable(vd) &&
(vdev_writeable(vd) || !spa_writeable(spa))) {
zio->io_error = 0;
} else {
ASSERT(zio->io_error != 0);
vdev_dbgmsg(vd, "failed probe");
(void) zfs_ereport_post(FM_EREPORT_ZFS_PROBE_FAILURE,
spa, vd, NULL, NULL, 0);
zio->io_error = SET_ERROR(ENXIO);
}
mutex_enter(&vd->vdev_probe_lock);
ASSERT(vd->vdev_probe_zio == zio);
vd->vdev_probe_zio = NULL;
mutex_exit(&vd->vdev_probe_lock);
zl = NULL;
while ((pio = zio_walk_parents(zio, &zl)) != NULL)
if (!vdev_accessible(vd, pio))
pio->io_error = SET_ERROR(ENXIO);
kmem_free(vps, sizeof (*vps));
}
}
/*
* Determine whether this device is accessible.
*
* Read and write to several known locations: the pad regions of each
* vdev label but the first, which we leave alone in case it contains
* a VTOC.
*/
zio_t *
vdev_probe(vdev_t *vd, zio_t *zio)
{
spa_t *spa = vd->vdev_spa;
vdev_probe_stats_t *vps = NULL;
zio_t *pio;
ASSERT(vd->vdev_ops->vdev_op_leaf);
/*
* Don't probe the probe.
*/
if (zio && (zio->io_flags & ZIO_FLAG_PROBE))
return (NULL);
/*
* To prevent 'probe storms' when a device fails, we create
* just one probe i/o at a time. All zios that want to probe
* this vdev will become parents of the probe io.
*/
mutex_enter(&vd->vdev_probe_lock);
if ((pio = vd->vdev_probe_zio) == NULL) {
vps = kmem_zalloc(sizeof (*vps), KM_SLEEP);
vps->vps_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_PROBE |
ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_AGGREGATE |
ZIO_FLAG_TRYHARD;
if (spa_config_held(spa, SCL_ZIO, RW_WRITER)) {
/*
* vdev_cant_read and vdev_cant_write can only
* transition from TRUE to FALSE when we have the
* SCL_ZIO lock as writer; otherwise they can only
* transition from FALSE to TRUE. This ensures that
* any zio looking at these values can assume that
* failures persist for the life of the I/O. That's
* important because when a device has intermittent
* connectivity problems, we want to ensure that
* they're ascribed to the device (ENXIO) and not
* the zio (EIO).
*
* Since we hold SCL_ZIO as writer here, clear both
* values so the probe can reevaluate from first
* principles.
*/
vps->vps_flags |= ZIO_FLAG_CONFIG_WRITER;
vd->vdev_cant_read = B_FALSE;
vd->vdev_cant_write = B_FALSE;
}
vd->vdev_probe_zio = pio = zio_null(NULL, spa, vd,
vdev_probe_done, vps,
vps->vps_flags | ZIO_FLAG_DONT_PROPAGATE);
/*
* We can't change the vdev state in this context, so we
* kick off an async task to do it on our behalf.
*/
if (zio != NULL) {
vd->vdev_probe_wanted = B_TRUE;
spa_async_request(spa, SPA_ASYNC_PROBE);
}
}
if (zio != NULL)
zio_add_child(zio, pio);
mutex_exit(&vd->vdev_probe_lock);
if (vps == NULL) {
ASSERT(zio != NULL);
return (NULL);
}
for (int l = 1; l < VDEV_LABELS; l++) {
zio_nowait(zio_read_phys(pio, vd,
vdev_label_offset(vd->vdev_psize, l,
offsetof(vdev_label_t, vl_be)), VDEV_PAD_SIZE,
abd_alloc_for_io(VDEV_PAD_SIZE, B_TRUE),
ZIO_CHECKSUM_OFF, vdev_probe_done, vps,
ZIO_PRIORITY_SYNC_READ, vps->vps_flags, B_TRUE));
}
if (zio == NULL)
return (pio);
zio_nowait(pio);
return (NULL);
}
static void
vdev_load_child(void *arg)
{
vdev_t *vd = arg;
vd->vdev_load_error = vdev_load(vd);
}
static void
vdev_open_child(void *arg)
{
vdev_t *vd = arg;
vd->vdev_open_thread = curthread;
vd->vdev_open_error = vdev_open(vd);
vd->vdev_open_thread = NULL;
}
static boolean_t
vdev_uses_zvols(vdev_t *vd)
{
#ifdef _KERNEL
if (zvol_is_zvol(vd->vdev_path))
return (B_TRUE);
#endif
for (int c = 0; c < vd->vdev_children; c++)
if (vdev_uses_zvols(vd->vdev_child[c]))
return (B_TRUE);
return (B_FALSE);
}
/*
* Returns B_TRUE if the passed child should be opened.
*/
static boolean_t
vdev_default_open_children_func(vdev_t *vd)
{
(void) vd;
return (B_TRUE);
}
/*
* Open the requested child vdevs. If any of the leaf vdevs are using
* a ZFS volume then do the opens in a single thread. This avoids a
* deadlock when the current thread is holding the spa_namespace_lock.
*/
static void
vdev_open_children_impl(vdev_t *vd, vdev_open_children_func_t *open_func)
{
int children = vd->vdev_children;
taskq_t *tq = taskq_create("vdev_open", children, minclsyspri,
children, children, TASKQ_PREPOPULATE);
vd->vdev_nonrot = B_TRUE;
for (int c = 0; c < children; c++) {
vdev_t *cvd = vd->vdev_child[c];
if (open_func(cvd) == B_FALSE)
continue;
if (tq == NULL || vdev_uses_zvols(vd)) {
cvd->vdev_open_error = vdev_open(cvd);
} else {
VERIFY(taskq_dispatch(tq, vdev_open_child,
cvd, TQ_SLEEP) != TASKQID_INVALID);
}
vd->vdev_nonrot &= cvd->vdev_nonrot;
}
if (tq != NULL) {
taskq_wait(tq);
taskq_destroy(tq);
}
}
/*
* Open all child vdevs.
*/
void
vdev_open_children(vdev_t *vd)
{
vdev_open_children_impl(vd, vdev_default_open_children_func);
}
/*
* Conditionally open a subset of child vdevs.
*/
void
vdev_open_children_subset(vdev_t *vd, vdev_open_children_func_t *open_func)
{
vdev_open_children_impl(vd, open_func);
}
/*
* Compute the raidz-deflation ratio. Note, we hard-code
* in 128k (1 << 17) because it is the "typical" blocksize.
* Even though SPA_MAXBLOCKSIZE changed, this algorithm can not change,
* otherwise it would inconsistently account for existing bp's.
*/
static void
vdev_set_deflate_ratio(vdev_t *vd)
{
if (vd == vd->vdev_top && !vd->vdev_ishole && vd->vdev_ashift != 0) {
vd->vdev_deflate_ratio = (1 << 17) /
(vdev_psize_to_asize(vd, 1 << 17) >> SPA_MINBLOCKSHIFT);
}
}
/*
* Maximize performance by inflating the configured ashift for top level
* vdevs to be as close to the physical ashift as possible while maintaining
* administrator defined limits and ensuring it doesn't go below the
* logical ashift.
*/
static void
vdev_ashift_optimize(vdev_t *vd)
{
ASSERT(vd == vd->vdev_top);
if (vd->vdev_ashift < vd->vdev_physical_ashift) {
vd->vdev_ashift = MIN(
MAX(zfs_vdev_max_auto_ashift, vd->vdev_ashift),
MAX(zfs_vdev_min_auto_ashift,
vd->vdev_physical_ashift));
} else {
/*
* If the logical and physical ashifts are the same, then
* we ensure that the top-level vdev's ashift is not smaller
* than our minimum ashift value. For the unusual case
* where logical ashift > physical ashift, we can't cap
* the calculated ashift based on max ashift as that
* would cause failures.
* We still check if we need to increase it to match
* the min ashift.
*/
vd->vdev_ashift = MAX(zfs_vdev_min_auto_ashift,
vd->vdev_ashift);
}
}
/*
* Prepare a virtual device for access.
*/
int
vdev_open(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
int error;
uint64_t osize = 0;
uint64_t max_osize = 0;
uint64_t asize, max_asize, psize;
uint64_t logical_ashift = 0;
uint64_t physical_ashift = 0;
ASSERT(vd->vdev_open_thread == curthread ||
spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
ASSERT(vd->vdev_state == VDEV_STATE_CLOSED ||
vd->vdev_state == VDEV_STATE_CANT_OPEN ||
vd->vdev_state == VDEV_STATE_OFFLINE);
vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
vd->vdev_cant_read = B_FALSE;
vd->vdev_cant_write = B_FALSE;
vd->vdev_min_asize = vdev_get_min_asize(vd);
/*
* If this vdev is not removed, check its fault status. If it's
* faulted, bail out of the open.
*/
if (!vd->vdev_removed && vd->vdev_faulted) {
ASSERT(vd->vdev_children == 0);
ASSERT(vd->vdev_label_aux == VDEV_AUX_ERR_EXCEEDED ||
vd->vdev_label_aux == VDEV_AUX_EXTERNAL);
vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
vd->vdev_label_aux);
return (SET_ERROR(ENXIO));
} else if (vd->vdev_offline) {
ASSERT(vd->vdev_children == 0);
vdev_set_state(vd, B_TRUE, VDEV_STATE_OFFLINE, VDEV_AUX_NONE);
return (SET_ERROR(ENXIO));
}
error = vd->vdev_ops->vdev_op_open(vd, &osize, &max_osize,
&logical_ashift, &physical_ashift);
/*
* Physical volume size should never be larger than its max size, unless
* the disk has shrunk while we were reading it or the device is buggy
* or damaged: either way it's not safe for use, bail out of the open.
*/
if (osize > max_osize) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_OPEN_FAILED);
return (SET_ERROR(ENXIO));
}
/*
* Reset the vdev_reopening flag so that we actually close
* the vdev on error.
*/
vd->vdev_reopening = B_FALSE;
if (zio_injection_enabled && error == 0)
error = zio_handle_device_injection(vd, NULL, SET_ERROR(ENXIO));
if (error) {
if (vd->vdev_removed &&
vd->vdev_stat.vs_aux != VDEV_AUX_OPEN_FAILED)
vd->vdev_removed = B_FALSE;
if (vd->vdev_stat.vs_aux == VDEV_AUX_CHILDREN_OFFLINE) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_OFFLINE,
vd->vdev_stat.vs_aux);
} else {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
vd->vdev_stat.vs_aux);
}
return (error);
}
vd->vdev_removed = B_FALSE;
/*
* Recheck the faulted flag now that we have confirmed that
* the vdev is accessible. If we're faulted, bail.
*/
if (vd->vdev_faulted) {
ASSERT(vd->vdev_children == 0);
ASSERT(vd->vdev_label_aux == VDEV_AUX_ERR_EXCEEDED ||
vd->vdev_label_aux == VDEV_AUX_EXTERNAL);
vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
vd->vdev_label_aux);
return (SET_ERROR(ENXIO));
}
if (vd->vdev_degraded) {
ASSERT(vd->vdev_children == 0);
vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED,
VDEV_AUX_ERR_EXCEEDED);
} else {
vdev_set_state(vd, B_TRUE, VDEV_STATE_HEALTHY, 0);
}
/*
* For hole or missing vdevs we just return success.
*/
if (vd->vdev_ishole || vd->vdev_ops == &vdev_missing_ops)
return (0);
for (int c = 0; c < vd->vdev_children; c++) {
if (vd->vdev_child[c]->vdev_state != VDEV_STATE_HEALTHY) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED,
VDEV_AUX_NONE);
break;
}
}
osize = P2ALIGN(osize, (uint64_t)sizeof (vdev_label_t));
max_osize = P2ALIGN(max_osize, (uint64_t)sizeof (vdev_label_t));
if (vd->vdev_children == 0) {
if (osize < SPA_MINDEVSIZE) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_TOO_SMALL);
return (SET_ERROR(EOVERFLOW));
}
psize = osize;
asize = osize - (VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE);
max_asize = max_osize - (VDEV_LABEL_START_SIZE +
VDEV_LABEL_END_SIZE);
} else {
if (vd->vdev_parent != NULL && osize < SPA_MINDEVSIZE -
(VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE)) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_TOO_SMALL);
return (SET_ERROR(EOVERFLOW));
}
psize = 0;
asize = osize;
max_asize = max_osize;
}
/*
* If the vdev was expanded, record this so that we can re-create the
* uberblock rings in labels {2,3}, during the next sync.
*/
if ((psize > vd->vdev_psize) && (vd->vdev_psize != 0))
vd->vdev_copy_uberblocks = B_TRUE;
vd->vdev_psize = psize;
/*
* Make sure the allocatable size hasn't shrunk too much.
*/
if (asize < vd->vdev_min_asize) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_BAD_LABEL);
return (SET_ERROR(EINVAL));
}
/*
* We can always set the logical/physical ashift members since
* their values are only used to calculate the vdev_ashift when
* the device is first added to the config. These values should
* not be used for anything else since they may change whenever
* the device is reopened and we don't store them in the label.
*/
vd->vdev_physical_ashift =
MAX(physical_ashift, vd->vdev_physical_ashift);
vd->vdev_logical_ashift = MAX(logical_ashift,
vd->vdev_logical_ashift);
if (vd->vdev_asize == 0) {
/*
* This is the first-ever open, so use the computed values.
* For compatibility, a different ashift can be requested.
*/
vd->vdev_asize = asize;
vd->vdev_max_asize = max_asize;
/*
* If the vdev_ashift was not overridden at creation time,
* then set it the logical ashift and optimize the ashift.
*/
if (vd->vdev_ashift == 0) {
vd->vdev_ashift = vd->vdev_logical_ashift;
if (vd->vdev_logical_ashift > ASHIFT_MAX) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_ASHIFT_TOO_BIG);
return (SET_ERROR(EDOM));
}
if (vd->vdev_top == vd) {
vdev_ashift_optimize(vd);
}
}
if (vd->vdev_ashift != 0 && (vd->vdev_ashift < ASHIFT_MIN ||
vd->vdev_ashift > ASHIFT_MAX)) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_BAD_ASHIFT);
return (SET_ERROR(EDOM));
}
} else {
/*
* Make sure the alignment required hasn't increased.
*/
if (vd->vdev_ashift > vd->vdev_top->vdev_ashift &&
vd->vdev_ops->vdev_op_leaf) {
(void) zfs_ereport_post(
FM_EREPORT_ZFS_DEVICE_BAD_ASHIFT,
spa, vd, NULL, NULL, 0);
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_BAD_LABEL);
return (SET_ERROR(EDOM));
}
vd->vdev_max_asize = max_asize;
}
/*
* If all children are healthy we update asize if either:
* The asize has increased, due to a device expansion caused by dynamic
* LUN growth or vdev replacement, and automatic expansion is enabled;
* making the additional space available.
*
* The asize has decreased, due to a device shrink usually caused by a
* vdev replace with a smaller device. This ensures that calculations
* based of max_asize and asize e.g. esize are always valid. It's safe
* to do this as we've already validated that asize is greater than
* vdev_min_asize.
*/
if (vd->vdev_state == VDEV_STATE_HEALTHY &&
((asize > vd->vdev_asize &&
(vd->vdev_expanding || spa->spa_autoexpand)) ||
(asize < vd->vdev_asize)))
vd->vdev_asize = asize;
vdev_set_min_asize(vd);
/*
* Ensure we can issue some IO before declaring the
* vdev open for business.
*/
if (vd->vdev_ops->vdev_op_leaf &&
(error = zio_wait(vdev_probe(vd, NULL))) != 0) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
VDEV_AUX_ERR_EXCEEDED);
return (error);
}
/*
* Track the minimum allocation size.
*/
if (vd->vdev_top == vd && vd->vdev_ashift != 0 &&
vd->vdev_islog == 0 && vd->vdev_aux == NULL) {
uint64_t min_alloc = vdev_get_min_alloc(vd);
if (min_alloc < spa->spa_min_alloc)
spa->spa_min_alloc = min_alloc;
}
/*
* If this is a leaf vdev, assess whether a resilver is needed.
* But don't do this if we are doing a reopen for a scrub, since
* this would just restart the scrub we are already doing.
*/
if (vd->vdev_ops->vdev_op_leaf && !spa->spa_scrub_reopen)
dsl_scan_assess_vdev(spa->spa_dsl_pool, vd);
return (0);
}
static void
vdev_validate_child(void *arg)
{
vdev_t *vd = arg;
vd->vdev_validate_thread = curthread;
vd->vdev_validate_error = vdev_validate(vd);
vd->vdev_validate_thread = NULL;
}
/*
* Called once the vdevs are all opened, this routine validates the label
* contents. This needs to be done before vdev_load() so that we don't
* inadvertently do repair I/Os to the wrong device.
*
* This function will only return failure if one of the vdevs indicates that it
* has since been destroyed or exported. This is only possible if
* /etc/zfs/zpool.cache was readonly at the time. Otherwise, the vdev state
* will be updated but the function will return 0.
*/
int
vdev_validate(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
taskq_t *tq = NULL;
nvlist_t *label;
uint64_t guid = 0, aux_guid = 0, top_guid;
uint64_t state;
nvlist_t *nvl;
uint64_t txg;
int children = vd->vdev_children;
if (vdev_validate_skip)
return (0);
if (children > 0) {
tq = taskq_create("vdev_validate", children, minclsyspri,
children, children, TASKQ_PREPOPULATE);
}
for (uint64_t c = 0; c < children; c++) {
vdev_t *cvd = vd->vdev_child[c];
if (tq == NULL || vdev_uses_zvols(cvd)) {
vdev_validate_child(cvd);
} else {
VERIFY(taskq_dispatch(tq, vdev_validate_child, cvd,
TQ_SLEEP) != TASKQID_INVALID);
}
}
if (tq != NULL) {
taskq_wait(tq);
taskq_destroy(tq);
}
for (int c = 0; c < children; c++) {
int error = vd->vdev_child[c]->vdev_validate_error;
if (error != 0)
return (SET_ERROR(EBADF));
}
/*
* If the device has already failed, or was marked offline, don't do
* any further validation. Otherwise, label I/O will fail and we will
* overwrite the previous state.
*/
if (!vd->vdev_ops->vdev_op_leaf || !vdev_readable(vd))
return (0);
/*
* If we are performing an extreme rewind, we allow for a label that
* was modified at a point after the current txg.
* If config lock is not held do not check for the txg. spa_sync could
* be updating the vdev's label before updating spa_last_synced_txg.
*/
if (spa->spa_extreme_rewind || spa_last_synced_txg(spa) == 0 ||
spa_config_held(spa, SCL_CONFIG, RW_WRITER) != SCL_CONFIG)
txg = UINT64_MAX;
else
txg = spa_last_synced_txg(spa);
if ((label = vdev_label_read_config(vd, txg)) == NULL) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_BAD_LABEL);
vdev_dbgmsg(vd, "vdev_validate: failed reading config for "
"txg %llu", (u_longlong_t)txg);
return (0);
}
/*
* Determine if this vdev has been split off into another
* pool. If so, then refuse to open it.
*/
if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_SPLIT_GUID,
&aux_guid) == 0 && aux_guid == spa_guid(spa)) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_SPLIT_POOL);
nvlist_free(label);
vdev_dbgmsg(vd, "vdev_validate: vdev split into other pool");
return (0);
}
if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, &guid) != 0) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
nvlist_free(label);
vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
ZPOOL_CONFIG_POOL_GUID);
return (0);
}
/*
* If config is not trusted then ignore the spa guid check. This is
* necessary because if the machine crashed during a re-guid the new
* guid might have been written to all of the vdev labels, but not the
* cached config. The check will be performed again once we have the
* trusted config from the MOS.
*/
if (spa->spa_trust_config && guid != spa_guid(spa)) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
nvlist_free(label);
vdev_dbgmsg(vd, "vdev_validate: vdev label pool_guid doesn't "
"match config (%llu != %llu)", (u_longlong_t)guid,
(u_longlong_t)spa_guid(spa));
return (0);
}
if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_VDEV_TREE, &nvl)
!= 0 || nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_ORIG_GUID,
&aux_guid) != 0)
aux_guid = 0;
if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) != 0) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
nvlist_free(label);
vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
ZPOOL_CONFIG_GUID);
return (0);
}
if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_TOP_GUID, &top_guid)
!= 0) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
nvlist_free(label);
vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
ZPOOL_CONFIG_TOP_GUID);
return (0);
}
/*
* If this vdev just became a top-level vdev because its sibling was
* detached, it will have adopted the parent's vdev guid -- but the
* label may or may not be on disk yet. Fortunately, either version
* of the label will have the same top guid, so if we're a top-level
* vdev, we can safely compare to that instead.
* However, if the config comes from a cachefile that failed to update
* after the detach, a top-level vdev will appear as a non top-level
* vdev in the config. Also relax the constraints if we perform an
* extreme rewind.
*
* If we split this vdev off instead, then we also check the
* original pool's guid. We don't want to consider the vdev
* corrupt if it is partway through a split operation.
*/
if (vd->vdev_guid != guid && vd->vdev_guid != aux_guid) {
boolean_t mismatch = B_FALSE;
if (spa->spa_trust_config && !spa->spa_extreme_rewind) {
if (vd != vd->vdev_top || vd->vdev_guid != top_guid)
mismatch = B_TRUE;
} else {
if (vd->vdev_guid != top_guid &&
vd->vdev_top->vdev_guid != guid)
mismatch = B_TRUE;
}
if (mismatch) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
nvlist_free(label);
vdev_dbgmsg(vd, "vdev_validate: config guid "
"doesn't match label guid");
vdev_dbgmsg(vd, "CONFIG: guid %llu, top_guid %llu",
(u_longlong_t)vd->vdev_guid,
(u_longlong_t)vd->vdev_top->vdev_guid);
vdev_dbgmsg(vd, "LABEL: guid %llu, top_guid %llu, "
"aux_guid %llu", (u_longlong_t)guid,
(u_longlong_t)top_guid, (u_longlong_t)aux_guid);
return (0);
}
}
if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE,
&state) != 0) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
nvlist_free(label);
vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
ZPOOL_CONFIG_POOL_STATE);
return (0);
}
nvlist_free(label);
/*
* If this is a verbatim import, no need to check the
* state of the pool.
*/
if (!(spa->spa_import_flags & ZFS_IMPORT_VERBATIM) &&
spa_load_state(spa) == SPA_LOAD_OPEN &&
state != POOL_STATE_ACTIVE) {
vdev_dbgmsg(vd, "vdev_validate: invalid pool state (%llu) "
"for spa %s", (u_longlong_t)state, spa->spa_name);
return (SET_ERROR(EBADF));
}
/*
* If we were able to open and validate a vdev that was
* previously marked permanently unavailable, clear that state
* now.
*/
if (vd->vdev_not_present)
vd->vdev_not_present = 0;
return (0);
}
static void
vdev_copy_path_impl(vdev_t *svd, vdev_t *dvd)
{
char *old, *new;
if (svd->vdev_path != NULL && dvd->vdev_path != NULL) {
if (strcmp(svd->vdev_path, dvd->vdev_path) != 0) {
zfs_dbgmsg("vdev_copy_path: vdev %llu: path changed "
"from '%s' to '%s'", (u_longlong_t)dvd->vdev_guid,
dvd->vdev_path, svd->vdev_path);
spa_strfree(dvd->vdev_path);
dvd->vdev_path = spa_strdup(svd->vdev_path);
}
} else if (svd->vdev_path != NULL) {
dvd->vdev_path = spa_strdup(svd->vdev_path);
zfs_dbgmsg("vdev_copy_path: vdev %llu: path set to '%s'",
(u_longlong_t)dvd->vdev_guid, dvd->vdev_path);
}
/*
* Our enclosure sysfs path may have changed between imports
*/
old = dvd->vdev_enc_sysfs_path;
new = svd->vdev_enc_sysfs_path;
if ((old != NULL && new == NULL) ||
(old == NULL && new != NULL) ||
((old != NULL && new != NULL) && strcmp(new, old) != 0)) {
zfs_dbgmsg("vdev_copy_path: vdev %llu: vdev_enc_sysfs_path "
"changed from '%s' to '%s'", (u_longlong_t)dvd->vdev_guid,
old, new);
if (dvd->vdev_enc_sysfs_path)
spa_strfree(dvd->vdev_enc_sysfs_path);
if (svd->vdev_enc_sysfs_path) {
dvd->vdev_enc_sysfs_path = spa_strdup(
svd->vdev_enc_sysfs_path);
} else {
dvd->vdev_enc_sysfs_path = NULL;
}
}
}
/*
* Recursively copy vdev paths from one vdev to another. Source and destination
* vdev trees must have same geometry otherwise return error. Intended to copy
* paths from userland config into MOS config.
*/
int
vdev_copy_path_strict(vdev_t *svd, vdev_t *dvd)
{
if ((svd->vdev_ops == &vdev_missing_ops) ||
(svd->vdev_ishole && dvd->vdev_ishole) ||
(dvd->vdev_ops == &vdev_indirect_ops))
return (0);
if (svd->vdev_ops != dvd->vdev_ops) {
vdev_dbgmsg(svd, "vdev_copy_path: vdev type mismatch: %s != %s",
svd->vdev_ops->vdev_op_type, dvd->vdev_ops->vdev_op_type);
return (SET_ERROR(EINVAL));
}
if (svd->vdev_guid != dvd->vdev_guid) {
vdev_dbgmsg(svd, "vdev_copy_path: guids mismatch (%llu != "
"%llu)", (u_longlong_t)svd->vdev_guid,
(u_longlong_t)dvd->vdev_guid);
return (SET_ERROR(EINVAL));
}
if (svd->vdev_children != dvd->vdev_children) {
vdev_dbgmsg(svd, "vdev_copy_path: children count mismatch: "
"%llu != %llu", (u_longlong_t)svd->vdev_children,
(u_longlong_t)dvd->vdev_children);
return (SET_ERROR(EINVAL));
}
for (uint64_t i = 0; i < svd->vdev_children; i++) {
int error = vdev_copy_path_strict(svd->vdev_child[i],
dvd->vdev_child[i]);
if (error != 0)
return (error);
}
if (svd->vdev_ops->vdev_op_leaf)
vdev_copy_path_impl(svd, dvd);
return (0);
}
static void
vdev_copy_path_search(vdev_t *stvd, vdev_t *dvd)
{
ASSERT(stvd->vdev_top == stvd);
ASSERT3U(stvd->vdev_id, ==, dvd->vdev_top->vdev_id);
for (uint64_t i = 0; i < dvd->vdev_children; i++) {
vdev_copy_path_search(stvd, dvd->vdev_child[i]);
}
if (!dvd->vdev_ops->vdev_op_leaf || !vdev_is_concrete(dvd))
return;
/*
* The idea here is that while a vdev can shift positions within
* a top vdev (when replacing, attaching mirror, etc.) it cannot
* step outside of it.
*/
vdev_t *vd = vdev_lookup_by_guid(stvd, dvd->vdev_guid);
if (vd == NULL || vd->vdev_ops != dvd->vdev_ops)
return;
ASSERT(vd->vdev_ops->vdev_op_leaf);
vdev_copy_path_impl(vd, dvd);
}
/*
* Recursively copy vdev paths from one root vdev to another. Source and
* destination vdev trees may differ in geometry. For each destination leaf
* vdev, search a vdev with the same guid and top vdev id in the source.
* Intended to copy paths from userland config into MOS config.
*/
void
vdev_copy_path_relaxed(vdev_t *srvd, vdev_t *drvd)
{
uint64_t children = MIN(srvd->vdev_children, drvd->vdev_children);
ASSERT(srvd->vdev_ops == &vdev_root_ops);
ASSERT(drvd->vdev_ops == &vdev_root_ops);
for (uint64_t i = 0; i < children; i++) {
vdev_copy_path_search(srvd->vdev_child[i],
drvd->vdev_child[i]);
}
}
/*
* Close a virtual device.
*/
void
vdev_close(vdev_t *vd)
{
vdev_t *pvd = vd->vdev_parent;
spa_t *spa __maybe_unused = vd->vdev_spa;
ASSERT(vd != NULL);
ASSERT(vd->vdev_open_thread == curthread ||
spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
/*
* If our parent is reopening, then we are as well, unless we are
* going offline.
*/
if (pvd != NULL && pvd->vdev_reopening)
vd->vdev_reopening = (pvd->vdev_reopening && !vd->vdev_offline);
vd->vdev_ops->vdev_op_close(vd);
vdev_cache_purge(vd);
/*
* We record the previous state before we close it, so that if we are
* doing a reopen(), we don't generate FMA ereports if we notice that
* it's still faulted.
*/
vd->vdev_prevstate = vd->vdev_state;
if (vd->vdev_offline)
vd->vdev_state = VDEV_STATE_OFFLINE;
else
vd->vdev_state = VDEV_STATE_CLOSED;
vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
}
void
vdev_hold(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
ASSERT(spa_is_root(spa));
if (spa->spa_state == POOL_STATE_UNINITIALIZED)
return;
for (int c = 0; c < vd->vdev_children; c++)
vdev_hold(vd->vdev_child[c]);
if (vd->vdev_ops->vdev_op_leaf && vd->vdev_ops->vdev_op_hold != NULL)
vd->vdev_ops->vdev_op_hold(vd);
}
void
vdev_rele(vdev_t *vd)
{
ASSERT(spa_is_root(vd->vdev_spa));
for (int c = 0; c < vd->vdev_children; c++)
vdev_rele(vd->vdev_child[c]);
if (vd->vdev_ops->vdev_op_leaf && vd->vdev_ops->vdev_op_rele != NULL)
vd->vdev_ops->vdev_op_rele(vd);
}
/*
* Reopen all interior vdevs and any unopened leaves. We don't actually
* reopen leaf vdevs which had previously been opened as they might deadlock
* on the spa_config_lock. Instead we only obtain the leaf's physical size.
* If the leaf has never been opened then open it, as usual.
*/
void
vdev_reopen(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
/* set the reopening flag unless we're taking the vdev offline */
vd->vdev_reopening = !vd->vdev_offline;
vdev_close(vd);
(void) vdev_open(vd);
/*
* Call vdev_validate() here to make sure we have the same device.
* Otherwise, a device with an invalid label could be successfully
* opened in response to vdev_reopen().
*/
if (vd->vdev_aux) {
(void) vdev_validate_aux(vd);
if (vdev_readable(vd) && vdev_writeable(vd) &&
vd->vdev_aux == &spa->spa_l2cache) {
/*
* In case the vdev is present we should evict all ARC
* buffers and pointers to log blocks and reclaim their
* space before restoring its contents to L2ARC.
*/
if (l2arc_vdev_present(vd)) {
l2arc_rebuild_vdev(vd, B_TRUE);
} else {
l2arc_add_vdev(spa, vd);
}
spa_async_request(spa, SPA_ASYNC_L2CACHE_REBUILD);
spa_async_request(spa, SPA_ASYNC_L2CACHE_TRIM);
}
} else {
(void) vdev_validate(vd);
}
/*
* Reassess parent vdev's health.
*/
vdev_propagate_state(vd);
}
int
vdev_create(vdev_t *vd, uint64_t txg, boolean_t isreplacing)
{
int error;
/*
* Normally, partial opens (e.g. of a mirror) are allowed.
* For a create, however, we want to fail the request if
* there are any components we can't open.
*/
error = vdev_open(vd);
if (error || vd->vdev_state != VDEV_STATE_HEALTHY) {
vdev_close(vd);
return (error ? error : SET_ERROR(ENXIO));
}
/*
* Recursively load DTLs and initialize all labels.
*/
if ((error = vdev_dtl_load(vd)) != 0 ||
(error = vdev_label_init(vd, txg, isreplacing ?
VDEV_LABEL_REPLACE : VDEV_LABEL_CREATE)) != 0) {
vdev_close(vd);
return (error);
}
return (0);
}
void
vdev_metaslab_set_size(vdev_t *vd)
{
uint64_t asize = vd->vdev_asize;
uint64_t ms_count = asize >> zfs_vdev_default_ms_shift;
uint64_t ms_shift;
/*
* There are two dimensions to the metaslab sizing calculation:
* the size of the metaslab and the count of metaslabs per vdev.
*
* The default values used below are a good balance between memory
* usage (larger metaslab size means more memory needed for loaded
* metaslabs; more metaslabs means more memory needed for the
* metaslab_t structs), metaslab load time (larger metaslabs take
* longer to load), and metaslab sync time (more metaslabs means
* more time spent syncing all of them).
*
* In general, we aim for zfs_vdev_default_ms_count (200) metaslabs.
* The range of the dimensions are as follows:
*
* 2^29 <= ms_size <= 2^34
* 16 <= ms_count <= 131,072
*
* On the lower end of vdev sizes, we aim for metaslabs sizes of
* at least 512MB (2^29) to minimize fragmentation effects when
* testing with smaller devices. However, the count constraint
* of at least 16 metaslabs will override this minimum size goal.
*
* On the upper end of vdev sizes, we aim for a maximum metaslab
* size of 16GB. However, we will cap the total count to 2^17
* metaslabs to keep our memory footprint in check and let the
* metaslab size grow from there if that limit is hit.
*
* The net effect of applying above constrains is summarized below.
*
* vdev size metaslab count
* --------------|-----------------
* < 8GB ~16
* 8GB - 100GB one per 512MB
* 100GB - 3TB ~200
* 3TB - 2PB one per 16GB
* > 2PB ~131,072
* --------------------------------
*
* Finally, note that all of the above calculate the initial
* number of metaslabs. Expanding a top-level vdev will result
* in additional metaslabs being allocated making it possible
* to exceed the zfs_vdev_ms_count_limit.
*/
if (ms_count < zfs_vdev_min_ms_count)
ms_shift = highbit64(asize / zfs_vdev_min_ms_count);
else if (ms_count > zfs_vdev_default_ms_count)
ms_shift = highbit64(asize / zfs_vdev_default_ms_count);
else
ms_shift = zfs_vdev_default_ms_shift;
if (ms_shift < SPA_MAXBLOCKSHIFT) {
ms_shift = SPA_MAXBLOCKSHIFT;
} else if (ms_shift > zfs_vdev_max_ms_shift) {
ms_shift = zfs_vdev_max_ms_shift;
/* cap the total count to constrain memory footprint */
if ((asize >> ms_shift) > zfs_vdev_ms_count_limit)
ms_shift = highbit64(asize / zfs_vdev_ms_count_limit);
}
vd->vdev_ms_shift = ms_shift;
ASSERT3U(vd->vdev_ms_shift, >=, SPA_MAXBLOCKSHIFT);
}
void
vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg)
{
ASSERT(vd == vd->vdev_top);
/* indirect vdevs don't have metaslabs or dtls */
ASSERT(vdev_is_concrete(vd) || flags == 0);
ASSERT(ISP2(flags));
ASSERT(spa_writeable(vd->vdev_spa));
if (flags & VDD_METASLAB)
(void) txg_list_add(&vd->vdev_ms_list, arg, txg);
if (flags & VDD_DTL)
(void) txg_list_add(&vd->vdev_dtl_list, arg, txg);
(void) txg_list_add(&vd->vdev_spa->spa_vdev_txg_list, vd, txg);
}
void
vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg)
{
for (int c = 0; c < vd->vdev_children; c++)
vdev_dirty_leaves(vd->vdev_child[c], flags, txg);
if (vd->vdev_ops->vdev_op_leaf)
vdev_dirty(vd->vdev_top, flags, vd, txg);
}
/*
* DTLs.
*
* A vdev's DTL (dirty time log) is the set of transaction groups for which
* the vdev has less than perfect replication. There are four kinds of DTL:
*
* DTL_MISSING: txgs for which the vdev has no valid copies of the data
*
* DTL_PARTIAL: txgs for which data is available, but not fully replicated
*
* DTL_SCRUB: the txgs that could not be repaired by the last scrub; upon
* scrub completion, DTL_SCRUB replaces DTL_MISSING in the range of
* txgs that was scrubbed.
*
* DTL_OUTAGE: txgs which cannot currently be read, whether due to
* persistent errors or just some device being offline.
* Unlike the other three, the DTL_OUTAGE map is not generally
* maintained; it's only computed when needed, typically to
* determine whether a device can be detached.
*
* For leaf vdevs, DTL_MISSING and DTL_PARTIAL are identical: the device
* either has the data or it doesn't.
*
* For interior vdevs such as mirror and RAID-Z the picture is more complex.
* A vdev's DTL_PARTIAL is the union of its children's DTL_PARTIALs, because
* if any child is less than fully replicated, then so is its parent.
* A vdev's DTL_MISSING is a modified union of its children's DTL_MISSINGs,
* comprising only those txgs which appear in 'maxfaults' or more children;
* those are the txgs we don't have enough replication to read. For example,
* double-parity RAID-Z can tolerate up to two missing devices (maxfaults == 2);
* thus, its DTL_MISSING consists of the set of txgs that appear in more than
* two child DTL_MISSING maps.
*
* It should be clear from the above that to compute the DTLs and outage maps
* for all vdevs, it suffices to know just the leaf vdevs' DTL_MISSING maps.
* Therefore, that is all we keep on disk. When loading the pool, or after
* a configuration change, we generate all other DTLs from first principles.
*/
void
vdev_dtl_dirty(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
{
range_tree_t *rt = vd->vdev_dtl[t];
ASSERT(t < DTL_TYPES);
ASSERT(vd != vd->vdev_spa->spa_root_vdev);
ASSERT(spa_writeable(vd->vdev_spa));
mutex_enter(&vd->vdev_dtl_lock);
if (!range_tree_contains(rt, txg, size))
range_tree_add(rt, txg, size);
mutex_exit(&vd->vdev_dtl_lock);
}
boolean_t
vdev_dtl_contains(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
{
range_tree_t *rt = vd->vdev_dtl[t];
boolean_t dirty = B_FALSE;
ASSERT(t < DTL_TYPES);
ASSERT(vd != vd->vdev_spa->spa_root_vdev);
/*
* While we are loading the pool, the DTLs have not been loaded yet.
* This isn't a problem but it can result in devices being tried
* which are known to not have the data. In which case, the import
* is relying on the checksum to ensure that we get the right data.
* Note that while importing we are only reading the MOS, which is
* always checksummed.
*/
mutex_enter(&vd->vdev_dtl_lock);
if (!range_tree_is_empty(rt))
dirty = range_tree_contains(rt, txg, size);
mutex_exit(&vd->vdev_dtl_lock);
return (dirty);
}
boolean_t
vdev_dtl_empty(vdev_t *vd, vdev_dtl_type_t t)
{
range_tree_t *rt = vd->vdev_dtl[t];
boolean_t empty;
mutex_enter(&vd->vdev_dtl_lock);
empty = range_tree_is_empty(rt);
mutex_exit(&vd->vdev_dtl_lock);
return (empty);
}
/*
* Check if the txg falls within the range which must be
* resilvered. DVAs outside this range can always be skipped.
*/
boolean_t
vdev_default_need_resilver(vdev_t *vd, const dva_t *dva, size_t psize,
uint64_t phys_birth)
{
(void) dva, (void) psize;
/* Set by sequential resilver. */
if (phys_birth == TXG_UNKNOWN)
return (B_TRUE);
return (vdev_dtl_contains(vd, DTL_PARTIAL, phys_birth, 1));
}
/*
* Returns B_TRUE if the vdev determines the DVA needs to be resilvered.
*/
boolean_t
vdev_dtl_need_resilver(vdev_t *vd, const dva_t *dva, size_t psize,
uint64_t phys_birth)
{
ASSERT(vd != vd->vdev_spa->spa_root_vdev);
if (vd->vdev_ops->vdev_op_need_resilver == NULL ||
vd->vdev_ops->vdev_op_leaf)
return (B_TRUE);
return (vd->vdev_ops->vdev_op_need_resilver(vd, dva, psize,
phys_birth));
}
/*
* Returns the lowest txg in the DTL range.
*/
static uint64_t
vdev_dtl_min(vdev_t *vd)
{
ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
ASSERT0(vd->vdev_children);
return (range_tree_min(vd->vdev_dtl[DTL_MISSING]) - 1);
}
/*
* Returns the highest txg in the DTL.
*/
static uint64_t
vdev_dtl_max(vdev_t *vd)
{
ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
ASSERT0(vd->vdev_children);
return (range_tree_max(vd->vdev_dtl[DTL_MISSING]));
}
/*
* Determine if a resilvering vdev should remove any DTL entries from
* its range. If the vdev was resilvering for the entire duration of the
* scan then it should excise that range from its DTLs. Otherwise, this
* vdev is considered partially resilvered and should leave its DTL
* entries intact. The comment in vdev_dtl_reassess() describes how we
* excise the DTLs.
*/
static boolean_t
vdev_dtl_should_excise(vdev_t *vd, boolean_t rebuild_done)
{
ASSERT0(vd->vdev_children);
if (vd->vdev_state < VDEV_STATE_DEGRADED)
return (B_FALSE);
if (vd->vdev_resilver_deferred)
return (B_FALSE);
if (range_tree_is_empty(vd->vdev_dtl[DTL_MISSING]))
return (B_TRUE);
if (rebuild_done) {
vdev_rebuild_t *vr = &vd->vdev_top->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
/* Rebuild not initiated by attach */
if (vd->vdev_rebuild_txg == 0)
return (B_TRUE);
/*
* When a rebuild completes without error then all missing data
* up to the rebuild max txg has been reconstructed and the DTL
* is eligible for excision.
*/
if (vrp->vrp_rebuild_state == VDEV_REBUILD_COMPLETE &&
vdev_dtl_max(vd) <= vrp->vrp_max_txg) {
ASSERT3U(vrp->vrp_min_txg, <=, vdev_dtl_min(vd));
ASSERT3U(vrp->vrp_min_txg, <, vd->vdev_rebuild_txg);
ASSERT3U(vd->vdev_rebuild_txg, <=, vrp->vrp_max_txg);
return (B_TRUE);
}
} else {
dsl_scan_t *scn = vd->vdev_spa->spa_dsl_pool->dp_scan;
dsl_scan_phys_t *scnp __maybe_unused = &scn->scn_phys;
/* Resilver not initiated by attach */
if (vd->vdev_resilver_txg == 0)
return (B_TRUE);
/*
* When a resilver is initiated the scan will assign the
* scn_max_txg value to the highest txg value that exists
* in all DTLs. If this device's max DTL is not part of this
* scan (i.e. it is not in the range (scn_min_txg, scn_max_txg]
* then it is not eligible for excision.
*/
if (vdev_dtl_max(vd) <= scn->scn_phys.scn_max_txg) {
ASSERT3U(scnp->scn_min_txg, <=, vdev_dtl_min(vd));
ASSERT3U(scnp->scn_min_txg, <, vd->vdev_resilver_txg);
ASSERT3U(vd->vdev_resilver_txg, <=, scnp->scn_max_txg);
return (B_TRUE);
}
}
return (B_FALSE);
}
/*
* Reassess DTLs after a config change or scrub completion. If txg == 0 no
* write operations will be issued to the pool.
*/
void
vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
boolean_t scrub_done, boolean_t rebuild_done)
{
spa_t *spa = vd->vdev_spa;
avl_tree_t reftree;
int minref;
ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
for (int c = 0; c < vd->vdev_children; c++)
vdev_dtl_reassess(vd->vdev_child[c], txg,
scrub_txg, scrub_done, rebuild_done);
if (vd == spa->spa_root_vdev || !vdev_is_concrete(vd) || vd->vdev_aux)
return;
if (vd->vdev_ops->vdev_op_leaf) {
dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;
vdev_rebuild_t *vr = &vd->vdev_top->vdev_rebuild_config;
boolean_t check_excise = B_FALSE;
boolean_t wasempty = B_TRUE;
mutex_enter(&vd->vdev_dtl_lock);
/*
* If requested, pretend the scan or rebuild completed cleanly.
*/
if (zfs_scan_ignore_errors) {
if (scn != NULL)
scn->scn_phys.scn_errors = 0;
if (vr != NULL)
vr->vr_rebuild_phys.vrp_errors = 0;
}
if (scrub_txg != 0 &&
!range_tree_is_empty(vd->vdev_dtl[DTL_MISSING])) {
wasempty = B_FALSE;
zfs_dbgmsg("guid:%llu txg:%llu scrub:%llu started:%d "
"dtl:%llu/%llu errors:%llu",
(u_longlong_t)vd->vdev_guid, (u_longlong_t)txg,
(u_longlong_t)scrub_txg, spa->spa_scrub_started,
(u_longlong_t)vdev_dtl_min(vd),
(u_longlong_t)vdev_dtl_max(vd),
(u_longlong_t)(scn ? scn->scn_phys.scn_errors : 0));
}
/*
* If we've completed a scrub/resilver or a rebuild cleanly
* then determine if this vdev should remove any DTLs. We
* only want to excise regions on vdevs that were available
* during the entire duration of this scan.
*/
if (rebuild_done &&
vr != NULL && vr->vr_rebuild_phys.vrp_errors == 0) {
check_excise = B_TRUE;
} else {
if (spa->spa_scrub_started ||
(scn != NULL && scn->scn_phys.scn_errors == 0)) {
check_excise = B_TRUE;
}
}
if (scrub_txg && check_excise &&
vdev_dtl_should_excise(vd, rebuild_done)) {
/*
* We completed a scrub, resilver or rebuild up to
* scrub_txg. If we did it without rebooting, then
* the scrub dtl will be valid, so excise the old
* region and fold in the scrub dtl. Otherwise,
* leave the dtl as-is if there was an error.
*
* There's little trick here: to excise the beginning
* of the DTL_MISSING map, we put it into a reference
* tree and then add a segment with refcnt -1 that
* covers the range [0, scrub_txg). This means
* that each txg in that range has refcnt -1 or 0.
* We then add DTL_SCRUB with a refcnt of 2, so that
* entries in the range [0, scrub_txg) will have a
* positive refcnt -- either 1 or 2. We then convert
* the reference tree into the new DTL_MISSING map.
*/
space_reftree_create(&reftree);
space_reftree_add_map(&reftree,
vd->vdev_dtl[DTL_MISSING], 1);
space_reftree_add_seg(&reftree, 0, scrub_txg, -1);
space_reftree_add_map(&reftree,
vd->vdev_dtl[DTL_SCRUB], 2);
space_reftree_generate_map(&reftree,
vd->vdev_dtl[DTL_MISSING], 1);
space_reftree_destroy(&reftree);
if (!range_tree_is_empty(vd->vdev_dtl[DTL_MISSING])) {
zfs_dbgmsg("update DTL_MISSING:%llu/%llu",
(u_longlong_t)vdev_dtl_min(vd),
(u_longlong_t)vdev_dtl_max(vd));
} else if (!wasempty) {
zfs_dbgmsg("DTL_MISSING is now empty");
}
}
range_tree_vacate(vd->vdev_dtl[DTL_PARTIAL], NULL, NULL);
range_tree_walk(vd->vdev_dtl[DTL_MISSING],
range_tree_add, vd->vdev_dtl[DTL_PARTIAL]);
if (scrub_done)
range_tree_vacate(vd->vdev_dtl[DTL_SCRUB], NULL, NULL);
range_tree_vacate(vd->vdev_dtl[DTL_OUTAGE], NULL, NULL);
if (!vdev_readable(vd))
range_tree_add(vd->vdev_dtl[DTL_OUTAGE], 0, -1ULL);
else
range_tree_walk(vd->vdev_dtl[DTL_MISSING],
range_tree_add, vd->vdev_dtl[DTL_OUTAGE]);
/*
* If the vdev was resilvering or rebuilding and no longer
* has any DTLs then reset the appropriate flag and dirty
* the top level so that we persist the change.
*/
if (txg != 0 &&
range_tree_is_empty(vd->vdev_dtl[DTL_MISSING]) &&
range_tree_is_empty(vd->vdev_dtl[DTL_OUTAGE])) {
if (vd->vdev_rebuild_txg != 0) {
vd->vdev_rebuild_txg = 0;
vdev_config_dirty(vd->vdev_top);
} else if (vd->vdev_resilver_txg != 0) {
vd->vdev_resilver_txg = 0;
vdev_config_dirty(vd->vdev_top);
}
}
mutex_exit(&vd->vdev_dtl_lock);
if (txg != 0)
vdev_dirty(vd->vdev_top, VDD_DTL, vd, txg);
return;
}
mutex_enter(&vd->vdev_dtl_lock);
for (int t = 0; t < DTL_TYPES; t++) {
/* account for child's outage in parent's missing map */
int s = (t == DTL_MISSING) ? DTL_OUTAGE: t;
if (t == DTL_SCRUB)
continue; /* leaf vdevs only */
if (t == DTL_PARTIAL)
minref = 1; /* i.e. non-zero */
else if (vdev_get_nparity(vd) != 0)
minref = vdev_get_nparity(vd) + 1; /* RAID-Z, dRAID */
else
minref = vd->vdev_children; /* any kind of mirror */
space_reftree_create(&reftree);
for (int c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
mutex_enter(&cvd->vdev_dtl_lock);
space_reftree_add_map(&reftree, cvd->vdev_dtl[s], 1);
mutex_exit(&cvd->vdev_dtl_lock);
}
space_reftree_generate_map(&reftree, vd->vdev_dtl[t], minref);
space_reftree_destroy(&reftree);
}
mutex_exit(&vd->vdev_dtl_lock);
}
int
vdev_dtl_load(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
objset_t *mos = spa->spa_meta_objset;
range_tree_t *rt;
int error = 0;
if (vd->vdev_ops->vdev_op_leaf && vd->vdev_dtl_object != 0) {
ASSERT(vdev_is_concrete(vd));
/*
* If the dtl cannot be sync'd there is no need to open it.
*/
if (spa->spa_mode == SPA_MODE_READ && !spa->spa_read_spacemaps)
return (0);
error = space_map_open(&vd->vdev_dtl_sm, mos,
vd->vdev_dtl_object, 0, -1ULL, 0);
if (error)
return (error);
ASSERT(vd->vdev_dtl_sm != NULL);
rt = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
error = space_map_load(vd->vdev_dtl_sm, rt, SM_ALLOC);
if (error == 0) {
mutex_enter(&vd->vdev_dtl_lock);
range_tree_walk(rt, range_tree_add,
vd->vdev_dtl[DTL_MISSING]);
mutex_exit(&vd->vdev_dtl_lock);
}
range_tree_vacate(rt, NULL, NULL);
range_tree_destroy(rt);
return (error);
}
for (int c = 0; c < vd->vdev_children; c++) {
error = vdev_dtl_load(vd->vdev_child[c]);
if (error != 0)
break;
}
return (error);
}
static void
vdev_zap_allocation_data(vdev_t *vd, dmu_tx_t *tx)
{
spa_t *spa = vd->vdev_spa;
objset_t *mos = spa->spa_meta_objset;
vdev_alloc_bias_t alloc_bias = vd->vdev_alloc_bias;
const char *string;
ASSERT(alloc_bias != VDEV_BIAS_NONE);
string =
(alloc_bias == VDEV_BIAS_LOG) ? VDEV_ALLOC_BIAS_LOG :
(alloc_bias == VDEV_BIAS_SPECIAL) ? VDEV_ALLOC_BIAS_SPECIAL :
(alloc_bias == VDEV_BIAS_DEDUP) ? VDEV_ALLOC_BIAS_DEDUP : NULL;
ASSERT(string != NULL);
VERIFY0(zap_add(mos, vd->vdev_top_zap, VDEV_TOP_ZAP_ALLOCATION_BIAS,
1, strlen(string) + 1, string, tx));
if (alloc_bias == VDEV_BIAS_SPECIAL || alloc_bias == VDEV_BIAS_DEDUP) {
spa_activate_allocation_classes(spa, tx);
}
}
void
vdev_destroy_unlink_zap(vdev_t *vd, uint64_t zapobj, dmu_tx_t *tx)
{
spa_t *spa = vd->vdev_spa;
VERIFY0(zap_destroy(spa->spa_meta_objset, zapobj, tx));
VERIFY0(zap_remove_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps,
zapobj, tx));
}
uint64_t
vdev_create_link_zap(vdev_t *vd, dmu_tx_t *tx)
{
spa_t *spa = vd->vdev_spa;
uint64_t zap = zap_create(spa->spa_meta_objset, DMU_OTN_ZAP_METADATA,
DMU_OT_NONE, 0, tx);
ASSERT(zap != 0);
VERIFY0(zap_add_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps,
zap, tx));
return (zap);
}
void
vdev_construct_zaps(vdev_t *vd, dmu_tx_t *tx)
{
if (vd->vdev_ops != &vdev_hole_ops &&
vd->vdev_ops != &vdev_missing_ops &&
vd->vdev_ops != &vdev_root_ops &&
!vd->vdev_top->vdev_removing) {
if (vd->vdev_ops->vdev_op_leaf && vd->vdev_leaf_zap == 0) {
vd->vdev_leaf_zap = vdev_create_link_zap(vd, tx);
}
if (vd == vd->vdev_top && vd->vdev_top_zap == 0) {
vd->vdev_top_zap = vdev_create_link_zap(vd, tx);
if (vd->vdev_alloc_bias != VDEV_BIAS_NONE)
vdev_zap_allocation_data(vd, tx);
}
}
for (uint64_t i = 0; i < vd->vdev_children; i++) {
vdev_construct_zaps(vd->vdev_child[i], tx);
}
}
static void
vdev_dtl_sync(vdev_t *vd, uint64_t txg)
{
spa_t *spa = vd->vdev_spa;
range_tree_t *rt = vd->vdev_dtl[DTL_MISSING];
objset_t *mos = spa->spa_meta_objset;
range_tree_t *rtsync;
dmu_tx_t *tx;
uint64_t object = space_map_object(vd->vdev_dtl_sm);
ASSERT(vdev_is_concrete(vd));
ASSERT(vd->vdev_ops->vdev_op_leaf);
tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
if (vd->vdev_detached || vd->vdev_top->vdev_removing) {
mutex_enter(&vd->vdev_dtl_lock);
space_map_free(vd->vdev_dtl_sm, tx);
space_map_close(vd->vdev_dtl_sm);
vd->vdev_dtl_sm = NULL;
mutex_exit(&vd->vdev_dtl_lock);
/*
* We only destroy the leaf ZAP for detached leaves or for
* removed log devices. Removed data devices handle leaf ZAP
* cleanup later, once cancellation is no longer possible.
*/
if (vd->vdev_leaf_zap != 0 && (vd->vdev_detached ||
vd->vdev_top->vdev_islog)) {
vdev_destroy_unlink_zap(vd, vd->vdev_leaf_zap, tx);
vd->vdev_leaf_zap = 0;
}
dmu_tx_commit(tx);
return;
}
if (vd->vdev_dtl_sm == NULL) {
uint64_t new_object;
new_object = space_map_alloc(mos, zfs_vdev_dtl_sm_blksz, tx);
VERIFY3U(new_object, !=, 0);
VERIFY0(space_map_open(&vd->vdev_dtl_sm, mos, new_object,
0, -1ULL, 0));
ASSERT(vd->vdev_dtl_sm != NULL);
}
rtsync = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
mutex_enter(&vd->vdev_dtl_lock);
range_tree_walk(rt, range_tree_add, rtsync);
mutex_exit(&vd->vdev_dtl_lock);
space_map_truncate(vd->vdev_dtl_sm, zfs_vdev_dtl_sm_blksz, tx);
space_map_write(vd->vdev_dtl_sm, rtsync, SM_ALLOC, SM_NO_VDEVID, tx);
range_tree_vacate(rtsync, NULL, NULL);
range_tree_destroy(rtsync);
/*
* If the object for the space map has changed then dirty
* the top level so that we update the config.
*/
if (object != space_map_object(vd->vdev_dtl_sm)) {
vdev_dbgmsg(vd, "txg %llu, spa %s, DTL old object %llu, "
"new object %llu", (u_longlong_t)txg, spa_name(spa),
(u_longlong_t)object,
(u_longlong_t)space_map_object(vd->vdev_dtl_sm));
vdev_config_dirty(vd->vdev_top);
}
dmu_tx_commit(tx);
}
/*
* Determine whether the specified vdev can be offlined/detached/removed
* without losing data.
*/
boolean_t
vdev_dtl_required(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
vdev_t *tvd = vd->vdev_top;
uint8_t cant_read = vd->vdev_cant_read;
boolean_t required;
ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
if (vd == spa->spa_root_vdev || vd == tvd)
return (B_TRUE);
/*
* Temporarily mark the device as unreadable, and then determine
* whether this results in any DTL outages in the top-level vdev.
* If not, we can safely offline/detach/remove the device.
*/
vd->vdev_cant_read = B_TRUE;
vdev_dtl_reassess(tvd, 0, 0, B_FALSE, B_FALSE);
required = !vdev_dtl_empty(tvd, DTL_OUTAGE);
vd->vdev_cant_read = cant_read;
vdev_dtl_reassess(tvd, 0, 0, B_FALSE, B_FALSE);
if (!required && zio_injection_enabled) {
required = !!zio_handle_device_injection(vd, NULL,
SET_ERROR(ECHILD));
}
return (required);
}
/*
* Determine if resilver is needed, and if so the txg range.
*/
boolean_t
vdev_resilver_needed(vdev_t *vd, uint64_t *minp, uint64_t *maxp)
{
boolean_t needed = B_FALSE;
uint64_t thismin = UINT64_MAX;
uint64_t thismax = 0;
if (vd->vdev_children == 0) {
mutex_enter(&vd->vdev_dtl_lock);
if (!range_tree_is_empty(vd->vdev_dtl[DTL_MISSING]) &&
vdev_writeable(vd)) {
thismin = vdev_dtl_min(vd);
thismax = vdev_dtl_max(vd);
needed = B_TRUE;
}
mutex_exit(&vd->vdev_dtl_lock);
} else {
for (int c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
uint64_t cmin, cmax;
if (vdev_resilver_needed(cvd, &cmin, &cmax)) {
thismin = MIN(thismin, cmin);
thismax = MAX(thismax, cmax);
needed = B_TRUE;
}
}
}
if (needed && minp) {
*minp = thismin;
*maxp = thismax;
}
return (needed);
}
/*
* Gets the checkpoint space map object from the vdev's ZAP. On success sm_obj
* will contain either the checkpoint spacemap object or zero if none exists.
* All other errors are returned to the caller.
*/
int
vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj)
{
ASSERT0(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));
if (vd->vdev_top_zap == 0) {
*sm_obj = 0;
return (0);
}
int error = zap_lookup(spa_meta_objset(vd->vdev_spa), vd->vdev_top_zap,
VDEV_TOP_ZAP_POOL_CHECKPOINT_SM, sizeof (uint64_t), 1, sm_obj);
if (error == ENOENT) {
*sm_obj = 0;
error = 0;
}
return (error);
}
int
vdev_load(vdev_t *vd)
{
int children = vd->vdev_children;
int error = 0;
taskq_t *tq = NULL;
/*
* It's only worthwhile to use the taskq for the root vdev, because the
* slow part is metaslab_init, and that only happens for top-level
* vdevs.
*/
if (vd->vdev_ops == &vdev_root_ops && vd->vdev_children > 0) {
tq = taskq_create("vdev_load", children, minclsyspri,
children, children, TASKQ_PREPOPULATE);
}
/*
* Recursively load all children.
*/
for (int c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
if (tq == NULL || vdev_uses_zvols(cvd)) {
cvd->vdev_load_error = vdev_load(cvd);
} else {
VERIFY(taskq_dispatch(tq, vdev_load_child,
cvd, TQ_SLEEP) != TASKQID_INVALID);
}
}
if (tq != NULL) {
taskq_wait(tq);
taskq_destroy(tq);
}
for (int c = 0; c < vd->vdev_children; c++) {
int error = vd->vdev_child[c]->vdev_load_error;
if (error != 0)
return (error);
}
vdev_set_deflate_ratio(vd);
/*
* On spa_load path, grab the allocation bias from our zap
*/
if (vd == vd->vdev_top && vd->vdev_top_zap != 0) {
spa_t *spa = vd->vdev_spa;
char bias_str[64];
error = zap_lookup(spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_ALLOCATION_BIAS, 1, sizeof (bias_str),
bias_str);
if (error == 0) {
ASSERT(vd->vdev_alloc_bias == VDEV_BIAS_NONE);
vd->vdev_alloc_bias = vdev_derive_alloc_bias(bias_str);
} else if (error != ENOENT) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
vdev_dbgmsg(vd, "vdev_load: zap_lookup(top_zap=%llu) "
"failed [error=%d]",
(u_longlong_t)vd->vdev_top_zap, error);
return (error);
}
}
/*
* Load any rebuild state from the top-level vdev zap.
*/
if (vd == vd->vdev_top && vd->vdev_top_zap != 0) {
error = vdev_rebuild_load(vd);
if (error && error != ENOTSUP) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
vdev_dbgmsg(vd, "vdev_load: vdev_rebuild_load "
"failed [error=%d]", error);
return (error);
}
}
/*
* If this is a top-level vdev, initialize its metaslabs.
*/
if (vd == vd->vdev_top && vdev_is_concrete(vd)) {
vdev_metaslab_group_create(vd);
if (vd->vdev_ashift == 0 || vd->vdev_asize == 0) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
vdev_dbgmsg(vd, "vdev_load: invalid size. ashift=%llu, "
"asize=%llu", (u_longlong_t)vd->vdev_ashift,
(u_longlong_t)vd->vdev_asize);
return (SET_ERROR(ENXIO));
}
error = vdev_metaslab_init(vd, 0);
if (error != 0) {
vdev_dbgmsg(vd, "vdev_load: metaslab_init failed "
"[error=%d]", error);
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
return (error);
}
uint64_t checkpoint_sm_obj;
error = vdev_checkpoint_sm_object(vd, &checkpoint_sm_obj);
if (error == 0 && checkpoint_sm_obj != 0) {
objset_t *mos = spa_meta_objset(vd->vdev_spa);
ASSERT(vd->vdev_asize != 0);
ASSERT3P(vd->vdev_checkpoint_sm, ==, NULL);
error = space_map_open(&vd->vdev_checkpoint_sm,
mos, checkpoint_sm_obj, 0, vd->vdev_asize,
vd->vdev_ashift);
if (error != 0) {
vdev_dbgmsg(vd, "vdev_load: space_map_open "
"failed for checkpoint spacemap (obj %llu) "
"[error=%d]",
(u_longlong_t)checkpoint_sm_obj, error);
return (error);
}
ASSERT3P(vd->vdev_checkpoint_sm, !=, NULL);
/*
* Since the checkpoint_sm contains free entries
* exclusively we can use space_map_allocated() to
* indicate the cumulative checkpointed space that
* has been freed.
*/
vd->vdev_stat.vs_checkpoint_space =
-space_map_allocated(vd->vdev_checkpoint_sm);
vd->vdev_spa->spa_checkpoint_info.sci_dspace +=
vd->vdev_stat.vs_checkpoint_space;
} else if (error != 0) {
vdev_dbgmsg(vd, "vdev_load: failed to retrieve "
"checkpoint space map object from vdev ZAP "
"[error=%d]", error);
return (error);
}
}
/*
* If this is a leaf vdev, load its DTL.
*/
if (vd->vdev_ops->vdev_op_leaf && (error = vdev_dtl_load(vd)) != 0) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
vdev_dbgmsg(vd, "vdev_load: vdev_dtl_load failed "
"[error=%d]", error);
return (error);
}
uint64_t obsolete_sm_object;
error = vdev_obsolete_sm_object(vd, &obsolete_sm_object);
if (error == 0 && obsolete_sm_object != 0) {
objset_t *mos = vd->vdev_spa->spa_meta_objset;
ASSERT(vd->vdev_asize != 0);
ASSERT3P(vd->vdev_obsolete_sm, ==, NULL);
if ((error = space_map_open(&vd->vdev_obsolete_sm, mos,
obsolete_sm_object, 0, vd->vdev_asize, 0))) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
vdev_dbgmsg(vd, "vdev_load: space_map_open failed for "
"obsolete spacemap (obj %llu) [error=%d]",
(u_longlong_t)obsolete_sm_object, error);
return (error);
}
} else if (error != 0) {
vdev_dbgmsg(vd, "vdev_load: failed to retrieve obsolete "
"space map object from vdev ZAP [error=%d]", error);
return (error);
}
return (0);
}
/*
* The special vdev case is used for hot spares and l2cache devices. Its
* sole purpose it to set the vdev state for the associated vdev. To do this,
* we make sure that we can open the underlying device, then try to read the
* label, and make sure that the label is sane and that it hasn't been
* repurposed to another pool.
*/
int
vdev_validate_aux(vdev_t *vd)
{
nvlist_t *label;
uint64_t guid, version;
uint64_t state;
if (!vdev_readable(vd))
return (0);
if ((label = vdev_label_read_config(vd, -1ULL)) == NULL) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
return (-1);
}
if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_VERSION, &version) != 0 ||
!SPA_VERSION_IS_SUPPORTED(version) ||
nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) != 0 ||
guid != vd->vdev_guid ||
nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, &state) != 0) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
nvlist_free(label);
return (-1);
}
/*
* We don't actually check the pool state here. If it's in fact in
* use by another pool, we update this fact on the fly when requested.
*/
nvlist_free(label);
return (0);
}
static void
vdev_destroy_ms_flush_data(vdev_t *vd, dmu_tx_t *tx)
{
objset_t *mos = spa_meta_objset(vd->vdev_spa);
if (vd->vdev_top_zap == 0)
return;
uint64_t object = 0;
int err = zap_lookup(mos, vd->vdev_top_zap,
VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, sizeof (uint64_t), 1, &object);
if (err == ENOENT)
return;
VERIFY0(err);
VERIFY0(dmu_object_free(mos, object, tx));
VERIFY0(zap_remove(mos, vd->vdev_top_zap,
VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, tx));
}
/*
* Free the objects used to store this vdev's spacemaps, and the array
* that points to them.
*/
void
vdev_destroy_spacemaps(vdev_t *vd, dmu_tx_t *tx)
{
if (vd->vdev_ms_array == 0)
return;
objset_t *mos = vd->vdev_spa->spa_meta_objset;
uint64_t array_count = vd->vdev_asize >> vd->vdev_ms_shift;
size_t array_bytes = array_count * sizeof (uint64_t);
uint64_t *smobj_array = kmem_alloc(array_bytes, KM_SLEEP);
VERIFY0(dmu_read(mos, vd->vdev_ms_array, 0,
array_bytes, smobj_array, 0));
for (uint64_t i = 0; i < array_count; i++) {
uint64_t smobj = smobj_array[i];
if (smobj == 0)
continue;
space_map_free_obj(mos, smobj, tx);
}
kmem_free(smobj_array, array_bytes);
VERIFY0(dmu_object_free(mos, vd->vdev_ms_array, tx));
vdev_destroy_ms_flush_data(vd, tx);
vd->vdev_ms_array = 0;
}
static void
vdev_remove_empty_log(vdev_t *vd, uint64_t txg)
{
spa_t *spa = vd->vdev_spa;
ASSERT(vd->vdev_islog);
ASSERT(vd == vd->vdev_top);
ASSERT3U(txg, ==, spa_syncing_txg(spa));
dmu_tx_t *tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);
vdev_destroy_spacemaps(vd, tx);
if (vd->vdev_top_zap != 0) {
vdev_destroy_unlink_zap(vd, vd->vdev_top_zap, tx);
vd->vdev_top_zap = 0;
}
dmu_tx_commit(tx);
}
void
vdev_sync_done(vdev_t *vd, uint64_t txg)
{
metaslab_t *msp;
boolean_t reassess = !txg_list_empty(&vd->vdev_ms_list, TXG_CLEAN(txg));
ASSERT(vdev_is_concrete(vd));
while ((msp = txg_list_remove(&vd->vdev_ms_list, TXG_CLEAN(txg)))
!= NULL)
metaslab_sync_done(msp, txg);
if (reassess) {
metaslab_sync_reassess(vd->vdev_mg);
if (vd->vdev_log_mg != NULL)
metaslab_sync_reassess(vd->vdev_log_mg);
}
}
void
vdev_sync(vdev_t *vd, uint64_t txg)
{
spa_t *spa = vd->vdev_spa;
vdev_t *lvd;
metaslab_t *msp;
ASSERT3U(txg, ==, spa->spa_syncing_txg);
dmu_tx_t *tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
if (range_tree_space(vd->vdev_obsolete_segments) > 0) {
ASSERT(vd->vdev_removing ||
vd->vdev_ops == &vdev_indirect_ops);
vdev_indirect_sync_obsolete(vd, tx);
/*
* If the vdev is indirect, it can't have dirty
* metaslabs or DTLs.
*/
if (vd->vdev_ops == &vdev_indirect_ops) {
ASSERT(txg_list_empty(&vd->vdev_ms_list, txg));
ASSERT(txg_list_empty(&vd->vdev_dtl_list, txg));
dmu_tx_commit(tx);
return;
}
}
ASSERT(vdev_is_concrete(vd));
if (vd->vdev_ms_array == 0 && vd->vdev_ms_shift != 0 &&
!vd->vdev_removing) {
ASSERT(vd == vd->vdev_top);
ASSERT0(vd->vdev_indirect_config.vic_mapping_object);
vd->vdev_ms_array = dmu_object_alloc(spa->spa_meta_objset,
DMU_OT_OBJECT_ARRAY, 0, DMU_OT_NONE, 0, tx);
ASSERT(vd->vdev_ms_array != 0);
vdev_config_dirty(vd);
}
while ((msp = txg_list_remove(&vd->vdev_ms_list, txg)) != NULL) {
metaslab_sync(msp, txg);
(void) txg_list_add(&vd->vdev_ms_list, msp, TXG_CLEAN(txg));
}
while ((lvd = txg_list_remove(&vd->vdev_dtl_list, txg)) != NULL)
vdev_dtl_sync(lvd, txg);
/*
* If this is an empty log device being removed, destroy the
* metadata associated with it.
*/
if (vd->vdev_islog && vd->vdev_stat.vs_alloc == 0 && vd->vdev_removing)
vdev_remove_empty_log(vd, txg);
(void) txg_list_add(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg));
dmu_tx_commit(tx);
}
uint64_t
vdev_psize_to_asize(vdev_t *vd, uint64_t psize)
{
return (vd->vdev_ops->vdev_op_asize(vd, psize));
}
/*
* Mark the given vdev faulted. A faulted vdev behaves as if the device could
* not be opened, and no I/O is attempted.
*/
int
vdev_fault(spa_t *spa, uint64_t guid, vdev_aux_t aux)
{
vdev_t *vd, *tvd;
spa_vdev_state_enter(spa, SCL_NONE);
if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));
if (!vd->vdev_ops->vdev_op_leaf)
return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));
tvd = vd->vdev_top;
/*
* If user did a 'zpool offline -f' then make the fault persist across
* reboots.
*/
if (aux == VDEV_AUX_EXTERNAL_PERSIST) {
/*
* There are two kinds of forced faults: temporary and
* persistent. Temporary faults go away at pool import, while
* persistent faults stay set. Both types of faults can be
* cleared with a zpool clear.
*
* We tell if a vdev is persistently faulted by looking at the
* ZPOOL_CONFIG_AUX_STATE nvpair. If it's set to "external" at
* import then it's a persistent fault. Otherwise, it's
* temporary. We get ZPOOL_CONFIG_AUX_STATE set to "external"
* by setting vd.vdev_stat.vs_aux to VDEV_AUX_EXTERNAL. This
* tells vdev_config_generate() (which gets run later) to set
* ZPOOL_CONFIG_AUX_STATE to "external" in the nvlist.
*/
vd->vdev_stat.vs_aux = VDEV_AUX_EXTERNAL;
vd->vdev_tmpoffline = B_FALSE;
aux = VDEV_AUX_EXTERNAL;
} else {
vd->vdev_tmpoffline = B_TRUE;
}
/*
* We don't directly use the aux state here, but if we do a
* vdev_reopen(), we need this value to be present to remember why we
* were faulted.
*/
vd->vdev_label_aux = aux;
/*
* Faulted state takes precedence over degraded.
*/
vd->vdev_delayed_close = B_FALSE;
vd->vdev_faulted = 1ULL;
vd->vdev_degraded = 0ULL;
vdev_set_state(vd, B_FALSE, VDEV_STATE_FAULTED, aux);
/*
* If this device has the only valid copy of the data, then
* back off and simply mark the vdev as degraded instead.
*/
if (!tvd->vdev_islog && vd->vdev_aux == NULL && vdev_dtl_required(vd)) {
vd->vdev_degraded = 1ULL;
vd->vdev_faulted = 0ULL;
/*
* If we reopen the device and it's not dead, only then do we
* mark it degraded.
*/
vdev_reopen(tvd);
if (vdev_readable(vd))
vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, aux);
}
return (spa_vdev_state_exit(spa, vd, 0));
}
/*
* Mark the given vdev degraded. A degraded vdev is purely an indication to the
* user that something is wrong. The vdev continues to operate as normal as far
* as I/O is concerned.
*/
int
vdev_degrade(spa_t *spa, uint64_t guid, vdev_aux_t aux)
{
vdev_t *vd;
spa_vdev_state_enter(spa, SCL_NONE);
if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));
if (!vd->vdev_ops->vdev_op_leaf)
return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));
/*
* If the vdev is already faulted, then don't do anything.
*/
if (vd->vdev_faulted || vd->vdev_degraded)
return (spa_vdev_state_exit(spa, NULL, 0));
vd->vdev_degraded = 1ULL;
if (!vdev_is_dead(vd))
vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED,
aux);
return (spa_vdev_state_exit(spa, vd, 0));
}
/*
* Online the given vdev.
*
* If 'ZFS_ONLINE_UNSPARE' is set, it implies two things. First, any attached
* spare device should be detached when the device finishes resilvering.
* Second, the online should be treated like a 'test' online case, so no FMA
* events are generated if the device fails to open.
*/
int
vdev_online(spa_t *spa, uint64_t guid, uint64_t flags, vdev_state_t *newstate)
{
vdev_t *vd, *tvd, *pvd, *rvd = spa->spa_root_vdev;
boolean_t wasoffline;
vdev_state_t oldstate;
spa_vdev_state_enter(spa, SCL_NONE);
if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));
if (!vd->vdev_ops->vdev_op_leaf)
return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));
wasoffline = (vd->vdev_offline || vd->vdev_tmpoffline);
oldstate = vd->vdev_state;
tvd = vd->vdev_top;
vd->vdev_offline = B_FALSE;
vd->vdev_tmpoffline = B_FALSE;
vd->vdev_checkremove = !!(flags & ZFS_ONLINE_CHECKREMOVE);
vd->vdev_forcefault = !!(flags & ZFS_ONLINE_FORCEFAULT);
/* XXX - L2ARC 1.0 does not support expansion */
if (!vd->vdev_aux) {
for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
pvd->vdev_expanding = !!((flags & ZFS_ONLINE_EXPAND) ||
spa->spa_autoexpand);
vd->vdev_expansion_time = gethrestime_sec();
}
vdev_reopen(tvd);
vd->vdev_checkremove = vd->vdev_forcefault = B_FALSE;
if (!vd->vdev_aux) {
for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
pvd->vdev_expanding = B_FALSE;
}
if (newstate)
*newstate = vd->vdev_state;
if ((flags & ZFS_ONLINE_UNSPARE) &&
!vdev_is_dead(vd) && vd->vdev_parent &&
vd->vdev_parent->vdev_ops == &vdev_spare_ops &&
vd->vdev_parent->vdev_child[0] == vd)
vd->vdev_unspare = B_TRUE;
if ((flags & ZFS_ONLINE_EXPAND) || spa->spa_autoexpand) {
/* XXX - L2ARC 1.0 does not support expansion */
if (vd->vdev_aux)
return (spa_vdev_state_exit(spa, vd, ENOTSUP));
spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
}
/* Restart initializing if necessary */
mutex_enter(&vd->vdev_initialize_lock);
if (vdev_writeable(vd) &&
vd->vdev_initialize_thread == NULL &&
vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE) {
(void) vdev_initialize(vd);
}
mutex_exit(&vd->vdev_initialize_lock);
/*
* Restart trimming if necessary. We do not restart trimming for cache
* devices here. This is triggered by l2arc_rebuild_vdev()
* asynchronously for the whole device or in l2arc_evict() as it evicts
* space for upcoming writes.
*/
mutex_enter(&vd->vdev_trim_lock);
if (vdev_writeable(vd) && !vd->vdev_isl2cache &&
vd->vdev_trim_thread == NULL &&
vd->vdev_trim_state == VDEV_TRIM_ACTIVE) {
(void) vdev_trim(vd, vd->vdev_trim_rate, vd->vdev_trim_partial,
vd->vdev_trim_secure);
}
mutex_exit(&vd->vdev_trim_lock);
if (wasoffline ||
(oldstate < VDEV_STATE_DEGRADED &&
vd->vdev_state >= VDEV_STATE_DEGRADED))
spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_ONLINE);
return (spa_vdev_state_exit(spa, vd, 0));
}
static int
vdev_offline_locked(spa_t *spa, uint64_t guid, uint64_t flags)
{
vdev_t *vd, *tvd;
int error = 0;
uint64_t generation;
metaslab_group_t *mg;
top:
spa_vdev_state_enter(spa, SCL_ALLOC);
if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));
if (!vd->vdev_ops->vdev_op_leaf)
return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));
if (vd->vdev_ops == &vdev_draid_spare_ops)
return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
tvd = vd->vdev_top;
mg = tvd->vdev_mg;
generation = spa->spa_config_generation + 1;
/*
* If the device isn't already offline, try to offline it.
*/
if (!vd->vdev_offline) {
/*
* If this device has the only valid copy of some data,
* don't allow it to be offlined. Log devices are always
* expendable.
*/
if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
vdev_dtl_required(vd))
return (spa_vdev_state_exit(spa, NULL,
SET_ERROR(EBUSY)));
/*
* If the top-level is a slog and it has had allocations
* then proceed. We check that the vdev's metaslab group
* is not NULL since it's possible that we may have just
* added this vdev but not yet initialized its metaslabs.
*/
if (tvd->vdev_islog && mg != NULL) {
/*
* Prevent any future allocations.
*/
ASSERT3P(tvd->vdev_log_mg, ==, NULL);
metaslab_group_passivate(mg);
(void) spa_vdev_state_exit(spa, vd, 0);
error = spa_reset_logs(spa);
/*
* If the log device was successfully reset but has
* checkpointed data, do not offline it.
*/
if (error == 0 &&
tvd->vdev_checkpoint_sm != NULL) {
ASSERT3U(space_map_allocated(
tvd->vdev_checkpoint_sm), !=, 0);
error = ZFS_ERR_CHECKPOINT_EXISTS;
}
spa_vdev_state_enter(spa, SCL_ALLOC);
/*
* Check to see if the config has changed.
*/
if (error || generation != spa->spa_config_generation) {
metaslab_group_activate(mg);
if (error)
return (spa_vdev_state_exit(spa,
vd, error));
(void) spa_vdev_state_exit(spa, vd, 0);
goto top;
}
ASSERT0(tvd->vdev_stat.vs_alloc);
}
/*
* Offline this device and reopen its top-level vdev.
* If the top-level vdev is a log device then just offline
* it. Otherwise, if this action results in the top-level
* vdev becoming unusable, undo it and fail the request.
*/
vd->vdev_offline = B_TRUE;
vdev_reopen(tvd);
if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
vdev_is_dead(tvd)) {
vd->vdev_offline = B_FALSE;
vdev_reopen(tvd);
return (spa_vdev_state_exit(spa, NULL,
SET_ERROR(EBUSY)));
}
/*
* Add the device back into the metaslab rotor so that
* once we online the device it's open for business.
*/
if (tvd->vdev_islog && mg != NULL)
metaslab_group_activate(mg);
}
vd->vdev_tmpoffline = !!(flags & ZFS_OFFLINE_TEMPORARY);
return (spa_vdev_state_exit(spa, vd, 0));
}
int
vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags)
{
int error;
mutex_enter(&spa->spa_vdev_top_lock);
error = vdev_offline_locked(spa, guid, flags);
mutex_exit(&spa->spa_vdev_top_lock);
return (error);
}
/*
* Clear the error counts associated with this vdev. Unlike vdev_online() and
* vdev_offline(), we assume the spa config is locked. We also clear all
* children. If 'vd' is NULL, then the user wants to clear all vdevs.
*/
void
vdev_clear(spa_t *spa, vdev_t *vd)
{
vdev_t *rvd = spa->spa_root_vdev;
ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
if (vd == NULL)
vd = rvd;
vd->vdev_stat.vs_read_errors = 0;
vd->vdev_stat.vs_write_errors = 0;
vd->vdev_stat.vs_checksum_errors = 0;
vd->vdev_stat.vs_slow_ios = 0;
for (int c = 0; c < vd->vdev_children; c++)
vdev_clear(spa, vd->vdev_child[c]);
/*
* It makes no sense to "clear" an indirect vdev.
*/
if (!vdev_is_concrete(vd))
return;
/*
* If we're in the FAULTED state or have experienced failed I/O, then
* clear the persistent state and attempt to reopen the device. We
* also mark the vdev config dirty, so that the new faulted state is
* written out to disk.
*/
if (vd->vdev_faulted || vd->vdev_degraded ||
!vdev_readable(vd) || !vdev_writeable(vd)) {
/*
* When reopening in response to a clear event, it may be due to
* a fmadm repair request. In this case, if the device is
* still broken, we want to still post the ereport again.
*/
vd->vdev_forcefault = B_TRUE;
vd->vdev_faulted = vd->vdev_degraded = 0ULL;
vd->vdev_cant_read = B_FALSE;
vd->vdev_cant_write = B_FALSE;
vd->vdev_stat.vs_aux = 0;
vdev_reopen(vd == rvd ? rvd : vd->vdev_top);
vd->vdev_forcefault = B_FALSE;
if (vd != rvd && vdev_writeable(vd->vdev_top))
vdev_state_dirty(vd->vdev_top);
/* If a resilver isn't required, check if vdevs can be culled */
if (vd->vdev_aux == NULL && !vdev_is_dead(vd) &&
!dsl_scan_resilvering(spa->spa_dsl_pool) &&
!dsl_scan_resilver_scheduled(spa->spa_dsl_pool))
spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);
spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_CLEAR);
}
/*
* When clearing a FMA-diagnosed fault, we always want to
* unspare the device, as we assume that the original spare was
* done in response to the FMA fault.
*/
if (!vdev_is_dead(vd) && vd->vdev_parent != NULL &&
vd->vdev_parent->vdev_ops == &vdev_spare_ops &&
vd->vdev_parent->vdev_child[0] == vd)
vd->vdev_unspare = B_TRUE;
/* Clear recent error events cache (i.e. duplicate events tracking) */
zfs_ereport_clear(spa, vd);
}
boolean_t
vdev_is_dead(vdev_t *vd)
{
/*
* Holes and missing devices are always considered "dead".
* This simplifies the code since we don't have to check for
* these types of devices in the various code paths.
* Instead we rely on the fact that we skip over dead devices
* before issuing I/O to them.
*/
return (vd->vdev_state < VDEV_STATE_DEGRADED ||
vd->vdev_ops == &vdev_hole_ops ||
vd->vdev_ops == &vdev_missing_ops);
}
boolean_t
vdev_readable(vdev_t *vd)
{
return (!vdev_is_dead(vd) && !vd->vdev_cant_read);
}
boolean_t
vdev_writeable(vdev_t *vd)
{
return (!vdev_is_dead(vd) && !vd->vdev_cant_write &&
vdev_is_concrete(vd));
}
boolean_t
vdev_allocatable(vdev_t *vd)
{
uint64_t state = vd->vdev_state;
/*
* We currently allow allocations from vdevs which may be in the
* process of reopening (i.e. VDEV_STATE_CLOSED). If the device
* fails to reopen then we'll catch it later when we're holding
* the proper locks. Note that we have to get the vdev state
* in a local variable because although it changes atomically,
* we're asking two separate questions about it.
*/
return (!(state < VDEV_STATE_DEGRADED && state != VDEV_STATE_CLOSED) &&
!vd->vdev_cant_write && vdev_is_concrete(vd) &&
vd->vdev_mg->mg_initialized);
}
boolean_t
vdev_accessible(vdev_t *vd, zio_t *zio)
{
ASSERT(zio->io_vd == vd);
if (vdev_is_dead(vd) || vd->vdev_remove_wanted)
return (B_FALSE);
if (zio->io_type == ZIO_TYPE_READ)
return (!vd->vdev_cant_read);
if (zio->io_type == ZIO_TYPE_WRITE)
return (!vd->vdev_cant_write);
return (B_TRUE);
}
static void
vdev_get_child_stat(vdev_t *cvd, vdev_stat_t *vs, vdev_stat_t *cvs)
{
/*
* Exclude the dRAID spare when aggregating to avoid double counting
* the ops and bytes. These IOs are counted by the physical leaves.
*/
if (cvd->vdev_ops == &vdev_draid_spare_ops)
return;
for (int t = 0; t < VS_ZIO_TYPES; t++) {
vs->vs_ops[t] += cvs->vs_ops[t];
vs->vs_bytes[t] += cvs->vs_bytes[t];
}
cvs->vs_scan_removing = cvd->vdev_removing;
}
/*
* Get extended stats
*/
static void
vdev_get_child_stat_ex(vdev_t *cvd, vdev_stat_ex_t *vsx, vdev_stat_ex_t *cvsx)
{
(void) cvd;
int t, b;
for (t = 0; t < ZIO_TYPES; t++) {
for (b = 0; b < ARRAY_SIZE(vsx->vsx_disk_histo[0]); b++)
vsx->vsx_disk_histo[t][b] += cvsx->vsx_disk_histo[t][b];
for (b = 0; b < ARRAY_SIZE(vsx->vsx_total_histo[0]); b++) {
vsx->vsx_total_histo[t][b] +=
cvsx->vsx_total_histo[t][b];
}
}
for (t = 0; t < ZIO_PRIORITY_NUM_QUEUEABLE; t++) {
for (b = 0; b < ARRAY_SIZE(vsx->vsx_queue_histo[0]); b++) {
vsx->vsx_queue_histo[t][b] +=
cvsx->vsx_queue_histo[t][b];
}
vsx->vsx_active_queue[t] += cvsx->vsx_active_queue[t];
vsx->vsx_pend_queue[t] += cvsx->vsx_pend_queue[t];
for (b = 0; b < ARRAY_SIZE(vsx->vsx_ind_histo[0]); b++)
vsx->vsx_ind_histo[t][b] += cvsx->vsx_ind_histo[t][b];
for (b = 0; b < ARRAY_SIZE(vsx->vsx_agg_histo[0]); b++)
vsx->vsx_agg_histo[t][b] += cvsx->vsx_agg_histo[t][b];
}
}
boolean_t
vdev_is_spacemap_addressable(vdev_t *vd)
{
if (spa_feature_is_active(vd->vdev_spa, SPA_FEATURE_SPACEMAP_V2))
return (B_TRUE);
/*
* If double-word space map entries are not enabled we assume
* 47 bits of the space map entry are dedicated to the entry's
* offset (see SM_OFFSET_BITS in space_map.h). We then use that
* to calculate the maximum address that can be described by a
* space map entry for the given device.
*/
uint64_t shift = vd->vdev_ashift + SM_OFFSET_BITS;
if (shift >= 63) /* detect potential overflow */
return (B_TRUE);
return (vd->vdev_asize < (1ULL << shift));
}
/*
* Get statistics for the given vdev.
*/
static void
vdev_get_stats_ex_impl(vdev_t *vd, vdev_stat_t *vs, vdev_stat_ex_t *vsx)
{
int t;
/*
* If we're getting stats on the root vdev, aggregate the I/O counts
* over all top-level vdevs (i.e. the direct children of the root).
*/
if (!vd->vdev_ops->vdev_op_leaf) {
if (vs) {
memset(vs->vs_ops, 0, sizeof (vs->vs_ops));
memset(vs->vs_bytes, 0, sizeof (vs->vs_bytes));
}
if (vsx)
memset(vsx, 0, sizeof (*vsx));
for (int c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
vdev_stat_t *cvs = &cvd->vdev_stat;
vdev_stat_ex_t *cvsx = &cvd->vdev_stat_ex;
vdev_get_stats_ex_impl(cvd, cvs, cvsx);
if (vs)
vdev_get_child_stat(cvd, vs, cvs);
if (vsx)
vdev_get_child_stat_ex(cvd, vsx, cvsx);
}
} else {
/*
* We're a leaf. Just copy our ZIO active queue stats in. The
* other leaf stats are updated in vdev_stat_update().
*/
if (!vsx)
return;
memcpy(vsx, &vd->vdev_stat_ex, sizeof (vd->vdev_stat_ex));
for (t = 0; t < ARRAY_SIZE(vd->vdev_queue.vq_class); t++) {
vsx->vsx_active_queue[t] =
vd->vdev_queue.vq_class[t].vqc_active;
vsx->vsx_pend_queue[t] = avl_numnodes(
&vd->vdev_queue.vq_class[t].vqc_queued_tree);
}
}
}
void
vdev_get_stats_ex(vdev_t *vd, vdev_stat_t *vs, vdev_stat_ex_t *vsx)
{
vdev_t *tvd = vd->vdev_top;
mutex_enter(&vd->vdev_stat_lock);
if (vs) {
bcopy(&vd->vdev_stat, vs, sizeof (*vs));
vs->vs_timestamp = gethrtime() - vs->vs_timestamp;
vs->vs_state = vd->vdev_state;
vs->vs_rsize = vdev_get_min_asize(vd);
if (vd->vdev_ops->vdev_op_leaf) {
vs->vs_rsize += VDEV_LABEL_START_SIZE +
VDEV_LABEL_END_SIZE;
/*
* Report initializing progress. Since we don't
* have the initializing locks held, this is only
* an estimate (although a fairly accurate one).
*/
vs->vs_initialize_bytes_done =
vd->vdev_initialize_bytes_done;
vs->vs_initialize_bytes_est =
vd->vdev_initialize_bytes_est;
vs->vs_initialize_state = vd->vdev_initialize_state;
vs->vs_initialize_action_time =
vd->vdev_initialize_action_time;
/*
* Report manual TRIM progress. Since we don't have
* the manual TRIM locks held, this is only an
* estimate (although fairly accurate one).
*/
vs->vs_trim_notsup = !vd->vdev_has_trim;
vs->vs_trim_bytes_done = vd->vdev_trim_bytes_done;
vs->vs_trim_bytes_est = vd->vdev_trim_bytes_est;
vs->vs_trim_state = vd->vdev_trim_state;
vs->vs_trim_action_time = vd->vdev_trim_action_time;
/* Set when there is a deferred resilver. */
vs->vs_resilver_deferred = vd->vdev_resilver_deferred;
}
/*
* Report expandable space on top-level, non-auxiliary devices
* only. The expandable space is reported in terms of metaslab
* sized units since that determines how much space the pool
* can expand.
*/
if (vd->vdev_aux == NULL && tvd != NULL) {
vs->vs_esize = P2ALIGN(
vd->vdev_max_asize - vd->vdev_asize,
1ULL << tvd->vdev_ms_shift);
}
vs->vs_configured_ashift = vd->vdev_top != NULL
? vd->vdev_top->vdev_ashift : vd->vdev_ashift;
vs->vs_logical_ashift = vd->vdev_logical_ashift;
vs->vs_physical_ashift = vd->vdev_physical_ashift;
/*
* Report fragmentation and rebuild progress for top-level,
* non-auxiliary, concrete devices.
*/
if (vd->vdev_aux == NULL && vd == vd->vdev_top &&
vdev_is_concrete(vd)) {
/*
* The vdev fragmentation rating doesn't take into
* account the embedded slog metaslab (vdev_log_mg).
* Since it's only one metaslab, it would have a tiny
* impact on the overall fragmentation.
*/
vs->vs_fragmentation = (vd->vdev_mg != NULL) ?
vd->vdev_mg->mg_fragmentation : 0;
}
vs->vs_noalloc = MAX(vd->vdev_noalloc,
tvd ? tvd->vdev_noalloc : 0);
}
vdev_get_stats_ex_impl(vd, vs, vsx);
mutex_exit(&vd->vdev_stat_lock);
}
void
vdev_get_stats(vdev_t *vd, vdev_stat_t *vs)
{
return (vdev_get_stats_ex(vd, vs, NULL));
}
void
vdev_clear_stats(vdev_t *vd)
{
mutex_enter(&vd->vdev_stat_lock);
vd->vdev_stat.vs_space = 0;
vd->vdev_stat.vs_dspace = 0;
vd->vdev_stat.vs_alloc = 0;
mutex_exit(&vd->vdev_stat_lock);
}
void
vdev_scan_stat_init(vdev_t *vd)
{
vdev_stat_t *vs = &vd->vdev_stat;
for (int c = 0; c < vd->vdev_children; c++)
vdev_scan_stat_init(vd->vdev_child[c]);
mutex_enter(&vd->vdev_stat_lock);
vs->vs_scan_processed = 0;
mutex_exit(&vd->vdev_stat_lock);
}
void
vdev_stat_update(zio_t *zio, uint64_t psize)
{
spa_t *spa = zio->io_spa;
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *vd = zio->io_vd ? zio->io_vd : rvd;
vdev_t *pvd;
uint64_t txg = zio->io_txg;
vdev_stat_t *vs = &vd->vdev_stat;
vdev_stat_ex_t *vsx = &vd->vdev_stat_ex;
zio_type_t type = zio->io_type;
int flags = zio->io_flags;
/*
* If this i/o is a gang leader, it didn't do any actual work.
*/
if (zio->io_gang_tree)
return;
if (zio->io_error == 0) {
/*
* If this is a root i/o, don't count it -- we've already
* counted the top-level vdevs, and vdev_get_stats() will
* aggregate them when asked. This reduces contention on
* the root vdev_stat_lock and implicitly handles blocks
* that compress away to holes, for which there is no i/o.
* (Holes never create vdev children, so all the counters
* remain zero, which is what we want.)
*
* Note: this only applies to successful i/o (io_error == 0)
* because unlike i/o counts, errors are not additive.
* When reading a ditto block, for example, failure of
* one top-level vdev does not imply a root-level error.
*/
if (vd == rvd)
return;
ASSERT(vd == zio->io_vd);
if (flags & ZIO_FLAG_IO_BYPASS)
return;
mutex_enter(&vd->vdev_stat_lock);
if (flags & ZIO_FLAG_IO_REPAIR) {
/*
* Repair is the result of a resilver issued by the
* scan thread (spa_sync).
*/
if (flags & ZIO_FLAG_SCAN_THREAD) {
dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;
dsl_scan_phys_t *scn_phys = &scn->scn_phys;
uint64_t *processed = &scn_phys->scn_processed;
if (vd->vdev_ops->vdev_op_leaf)
atomic_add_64(processed, psize);
vs->vs_scan_processed += psize;
}
/*
* Repair is the result of a rebuild issued by the
* rebuild thread (vdev_rebuild_thread). To avoid
* double counting repaired bytes the virtual dRAID
* spare vdev is excluded from the processed bytes.
*/
if (zio->io_priority == ZIO_PRIORITY_REBUILD) {
vdev_t *tvd = vd->vdev_top;
vdev_rebuild_t *vr = &tvd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
uint64_t *rebuilt = &vrp->vrp_bytes_rebuilt;
if (vd->vdev_ops->vdev_op_leaf &&
vd->vdev_ops != &vdev_draid_spare_ops) {
atomic_add_64(rebuilt, psize);
}
vs->vs_rebuild_processed += psize;
}
if (flags & ZIO_FLAG_SELF_HEAL)
vs->vs_self_healed += psize;
}
/*
* The bytes/ops/histograms are recorded at the leaf level and
* aggregated into the higher level vdevs in vdev_get_stats().
*/
if (vd->vdev_ops->vdev_op_leaf &&
(zio->io_priority < ZIO_PRIORITY_NUM_QUEUEABLE)) {
zio_type_t vs_type = type;
zio_priority_t priority = zio->io_priority;
/*
* TRIM ops and bytes are reported to user space as
* ZIO_TYPE_IOCTL. This is done to preserve the
* vdev_stat_t structure layout for user space.
*/
if (type == ZIO_TYPE_TRIM)
vs_type = ZIO_TYPE_IOCTL;
/*
* Solely for the purposes of 'zpool iostat -lqrw'
* reporting use the priority to categorize the IO.
* Only the following are reported to user space:
*
* ZIO_PRIORITY_SYNC_READ,
* ZIO_PRIORITY_SYNC_WRITE,
* ZIO_PRIORITY_ASYNC_READ,
* ZIO_PRIORITY_ASYNC_WRITE,
* ZIO_PRIORITY_SCRUB,
* ZIO_PRIORITY_TRIM,
* ZIO_PRIORITY_REBUILD.
*/
if (priority == ZIO_PRIORITY_INITIALIZING) {
ASSERT3U(type, ==, ZIO_TYPE_WRITE);
priority = ZIO_PRIORITY_ASYNC_WRITE;
} else if (priority == ZIO_PRIORITY_REMOVAL) {
priority = ((type == ZIO_TYPE_WRITE) ?
ZIO_PRIORITY_ASYNC_WRITE :
ZIO_PRIORITY_ASYNC_READ);
}
vs->vs_ops[vs_type]++;
vs->vs_bytes[vs_type] += psize;
if (flags & ZIO_FLAG_DELEGATED) {
vsx->vsx_agg_histo[priority]
[RQ_HISTO(zio->io_size)]++;
} else {
vsx->vsx_ind_histo[priority]
[RQ_HISTO(zio->io_size)]++;
}
if (zio->io_delta && zio->io_delay) {
vsx->vsx_queue_histo[priority]
[L_HISTO(zio->io_delta - zio->io_delay)]++;
vsx->vsx_disk_histo[type]
[L_HISTO(zio->io_delay)]++;
vsx->vsx_total_histo[type]
[L_HISTO(zio->io_delta)]++;
}
}
mutex_exit(&vd->vdev_stat_lock);
return;
}
if (flags & ZIO_FLAG_SPECULATIVE)
return;
/*
* If this is an I/O error that is going to be retried, then ignore the
* error. Otherwise, the user may interpret B_FAILFAST I/O errors as
* hard errors, when in reality they can happen for any number of
* innocuous reasons (bus resets, MPxIO link failure, etc).
*/
if (zio->io_error == EIO &&
!(zio->io_flags & ZIO_FLAG_IO_RETRY))
return;
/*
* Intent logs writes won't propagate their error to the root
* I/O so don't mark these types of failures as pool-level
* errors.
*/
if (zio->io_vd == NULL && (zio->io_flags & ZIO_FLAG_DONT_PROPAGATE))
return;
if (type == ZIO_TYPE_WRITE && txg != 0 &&
(!(flags & ZIO_FLAG_IO_REPAIR) ||
(flags & ZIO_FLAG_SCAN_THREAD) ||
spa->spa_claiming)) {
/*
* This is either a normal write (not a repair), or it's
* a repair induced by the scrub thread, or it's a repair
* made by zil_claim() during spa_load() in the first txg.
* In the normal case, we commit the DTL change in the same
* txg as the block was born. In the scrub-induced repair
* case, we know that scrubs run in first-pass syncing context,
* so we commit the DTL change in spa_syncing_txg(spa).
* In the zil_claim() case, we commit in spa_first_txg(spa).
*
* We currently do not make DTL entries for failed spontaneous
* self-healing writes triggered by normal (non-scrubbing)
* reads, because we have no transactional context in which to
* do so -- and it's not clear that it'd be desirable anyway.
*/
if (vd->vdev_ops->vdev_op_leaf) {
uint64_t commit_txg = txg;
if (flags & ZIO_FLAG_SCAN_THREAD) {
ASSERT(flags & ZIO_FLAG_IO_REPAIR);
ASSERT(spa_sync_pass(spa) == 1);
vdev_dtl_dirty(vd, DTL_SCRUB, txg, 1);
commit_txg = spa_syncing_txg(spa);
} else if (spa->spa_claiming) {
ASSERT(flags & ZIO_FLAG_IO_REPAIR);
commit_txg = spa_first_txg(spa);
}
ASSERT(commit_txg >= spa_syncing_txg(spa));
if (vdev_dtl_contains(vd, DTL_MISSING, txg, 1))
return;
for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
vdev_dtl_dirty(pvd, DTL_PARTIAL, txg, 1);
vdev_dirty(vd->vdev_top, VDD_DTL, vd, commit_txg);
}
if (vd != rvd)
vdev_dtl_dirty(vd, DTL_MISSING, txg, 1);
}
}
int64_t
vdev_deflated_space(vdev_t *vd, int64_t space)
{
ASSERT((space & (SPA_MINBLOCKSIZE-1)) == 0);
ASSERT(vd->vdev_deflate_ratio != 0 || vd->vdev_isl2cache);
return ((space >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio);
}
/*
* Update the in-core space usage stats for this vdev, its metaslab class,
* and the root vdev.
*/
void
vdev_space_update(vdev_t *vd, int64_t alloc_delta, int64_t defer_delta,
int64_t space_delta)
{
(void) defer_delta;
int64_t dspace_delta;
spa_t *spa = vd->vdev_spa;
vdev_t *rvd = spa->spa_root_vdev;
ASSERT(vd == vd->vdev_top);
/*
* Apply the inverse of the psize-to-asize (ie. RAID-Z) space-expansion
* factor. We must calculate this here and not at the root vdev
* because the root vdev's psize-to-asize is simply the max of its
* children's, thus not accurate enough for us.
*/
dspace_delta = vdev_deflated_space(vd, space_delta);
mutex_enter(&vd->vdev_stat_lock);
/* ensure we won't underflow */
if (alloc_delta < 0) {
ASSERT3U(vd->vdev_stat.vs_alloc, >=, -alloc_delta);
}
vd->vdev_stat.vs_alloc += alloc_delta;
vd->vdev_stat.vs_space += space_delta;
vd->vdev_stat.vs_dspace += dspace_delta;
mutex_exit(&vd->vdev_stat_lock);
/* every class but log contributes to root space stats */
if (vd->vdev_mg != NULL && !vd->vdev_islog) {
ASSERT(!vd->vdev_isl2cache);
mutex_enter(&rvd->vdev_stat_lock);
rvd->vdev_stat.vs_alloc += alloc_delta;
rvd->vdev_stat.vs_space += space_delta;
rvd->vdev_stat.vs_dspace += dspace_delta;
mutex_exit(&rvd->vdev_stat_lock);
}
/* Note: metaslab_class_space_update moved to metaslab_space_update */
}
/*
* Mark a top-level vdev's config as dirty, placing it on the dirty list
* so that it will be written out next time the vdev configuration is synced.
* If the root vdev is specified (vdev_top == NULL), dirty all top-level vdevs.
*/
void
vdev_config_dirty(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
vdev_t *rvd = spa->spa_root_vdev;
int c;
ASSERT(spa_writeable(spa));
/*
* If this is an aux vdev (as with l2cache and spare devices), then we
* update the vdev config manually and set the sync flag.
*/
if (vd->vdev_aux != NULL) {
spa_aux_vdev_t *sav = vd->vdev_aux;
nvlist_t **aux;
uint_t naux;
for (c = 0; c < sav->sav_count; c++) {
if (sav->sav_vdevs[c] == vd)
break;
}
if (c == sav->sav_count) {
/*
* We're being removed. There's nothing more to do.
*/
ASSERT(sav->sav_sync == B_TRUE);
return;
}
sav->sav_sync = B_TRUE;
if (nvlist_lookup_nvlist_array(sav->sav_config,
ZPOOL_CONFIG_L2CACHE, &aux, &naux) != 0) {
VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
ZPOOL_CONFIG_SPARES, &aux, &naux) == 0);
}
ASSERT(c < naux);
/*
* Setting the nvlist in the middle if the array is a little
* sketchy, but it will work.
*/
nvlist_free(aux[c]);
aux[c] = vdev_config_generate(spa, vd, B_TRUE, 0);
return;
}
/*
* The dirty list is protected by the SCL_CONFIG lock. The caller
* must either hold SCL_CONFIG as writer, or must be the sync thread
* (which holds SCL_CONFIG as reader). There's only one sync thread,
* so this is sufficient to ensure mutual exclusion.
*/
ASSERT(spa_config_held(spa, SCL_CONFIG, RW_WRITER) ||
(dsl_pool_sync_context(spa_get_dsl(spa)) &&
spa_config_held(spa, SCL_CONFIG, RW_READER)));
if (vd == rvd) {
for (c = 0; c < rvd->vdev_children; c++)
vdev_config_dirty(rvd->vdev_child[c]);
} else {
ASSERT(vd == vd->vdev_top);
if (!list_link_active(&vd->vdev_config_dirty_node) &&
vdev_is_concrete(vd)) {
list_insert_head(&spa->spa_config_dirty_list, vd);
}
}
}
void
vdev_config_clean(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
ASSERT(spa_config_held(spa, SCL_CONFIG, RW_WRITER) ||
(dsl_pool_sync_context(spa_get_dsl(spa)) &&
spa_config_held(spa, SCL_CONFIG, RW_READER)));
ASSERT(list_link_active(&vd->vdev_config_dirty_node));
list_remove(&spa->spa_config_dirty_list, vd);
}
/*
* Mark a top-level vdev's state as dirty, so that the next pass of
* spa_sync() can convert this into vdev_config_dirty(). We distinguish
* the state changes from larger config changes because they require
* much less locking, and are often needed for administrative actions.
*/
void
vdev_state_dirty(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
ASSERT(spa_writeable(spa));
ASSERT(vd == vd->vdev_top);
/*
* The state list is protected by the SCL_STATE lock. The caller
* must either hold SCL_STATE as writer, or must be the sync thread
* (which holds SCL_STATE as reader). There's only one sync thread,
* so this is sufficient to ensure mutual exclusion.
*/
ASSERT(spa_config_held(spa, SCL_STATE, RW_WRITER) ||
(dsl_pool_sync_context(spa_get_dsl(spa)) &&
spa_config_held(spa, SCL_STATE, RW_READER)));
if (!list_link_active(&vd->vdev_state_dirty_node) &&
vdev_is_concrete(vd))
list_insert_head(&spa->spa_state_dirty_list, vd);
}
void
vdev_state_clean(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
ASSERT(spa_config_held(spa, SCL_STATE, RW_WRITER) ||
(dsl_pool_sync_context(spa_get_dsl(spa)) &&
spa_config_held(spa, SCL_STATE, RW_READER)));
ASSERT(list_link_active(&vd->vdev_state_dirty_node));
list_remove(&spa->spa_state_dirty_list, vd);
}
/*
* Propagate vdev state up from children to parent.
*/
void
vdev_propagate_state(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
vdev_t *rvd = spa->spa_root_vdev;
int degraded = 0, faulted = 0;
int corrupted = 0;
vdev_t *child;
if (vd->vdev_children > 0) {
for (int c = 0; c < vd->vdev_children; c++) {
child = vd->vdev_child[c];
/*
* Don't factor holes or indirect vdevs into the
* decision.
*/
if (!vdev_is_concrete(child))
continue;
if (!vdev_readable(child) ||
(!vdev_writeable(child) && spa_writeable(spa))) {
/*
* Root special: if there is a top-level log
* device, treat the root vdev as if it were
* degraded.
*/
if (child->vdev_islog && vd == rvd)
degraded++;
else
faulted++;
} else if (child->vdev_state <= VDEV_STATE_DEGRADED) {
degraded++;
}
if (child->vdev_stat.vs_aux == VDEV_AUX_CORRUPT_DATA)
corrupted++;
}
vd->vdev_ops->vdev_op_state_change(vd, faulted, degraded);
/*
* Root special: if there is a top-level vdev that cannot be
* opened due to corrupted metadata, then propagate the root
* vdev's aux state as 'corrupt' rather than 'insufficient
* replicas'.
*/
if (corrupted && vd == rvd &&
rvd->vdev_state == VDEV_STATE_CANT_OPEN)
vdev_set_state(rvd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_CORRUPT_DATA);
}
if (vd->vdev_parent)
vdev_propagate_state(vd->vdev_parent);
}
/*
* Set a vdev's state. If this is during an open, we don't update the parent
* state, because we're in the process of opening children depth-first.
* Otherwise, we propagate the change to the parent.
*
* If this routine places a device in a faulted state, an appropriate ereport is
* generated.
*/
void
vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state, vdev_aux_t aux)
{
uint64_t save_state;
spa_t *spa = vd->vdev_spa;
if (state == vd->vdev_state) {
/*
* Since vdev_offline() code path is already in an offline
* state we can miss a statechange event to OFFLINE. Check
* the previous state to catch this condition.
*/
if (vd->vdev_ops->vdev_op_leaf &&
(state == VDEV_STATE_OFFLINE) &&
(vd->vdev_prevstate >= VDEV_STATE_FAULTED)) {
/* post an offline state change */
zfs_post_state_change(spa, vd, vd->vdev_prevstate);
}
vd->vdev_stat.vs_aux = aux;
return;
}
save_state = vd->vdev_state;
vd->vdev_state = state;
vd->vdev_stat.vs_aux = aux;
/*
* If we are setting the vdev state to anything but an open state, then
* always close the underlying device unless the device has requested
* a delayed close (i.e. we're about to remove or fault the device).
* Otherwise, we keep accessible but invalid devices open forever.
* We don't call vdev_close() itself, because that implies some extra
* checks (offline, etc) that we don't want here. This is limited to
* leaf devices, because otherwise closing the device will affect other
* children.
*/
if (!vd->vdev_delayed_close && vdev_is_dead(vd) &&
vd->vdev_ops->vdev_op_leaf)
vd->vdev_ops->vdev_op_close(vd);
if (vd->vdev_removed &&
state == VDEV_STATE_CANT_OPEN &&
(aux == VDEV_AUX_OPEN_FAILED || vd->vdev_checkremove)) {
/*
* If the previous state is set to VDEV_STATE_REMOVED, then this
* device was previously marked removed and someone attempted to
* reopen it. If this failed due to a nonexistent device, then
* keep the device in the REMOVED state. We also let this be if
* it is one of our special test online cases, which is only
* attempting to online the device and shouldn't generate an FMA
* fault.
*/
vd->vdev_state = VDEV_STATE_REMOVED;
vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
} else if (state == VDEV_STATE_REMOVED) {
vd->vdev_removed = B_TRUE;
} else if (state == VDEV_STATE_CANT_OPEN) {
/*
* If we fail to open a vdev during an import or recovery, we
* mark it as "not available", which signifies that it was
* never there to begin with. Failure to open such a device
* is not considered an error.
*/
if ((spa_load_state(spa) == SPA_LOAD_IMPORT ||
spa_load_state(spa) == SPA_LOAD_RECOVER) &&
vd->vdev_ops->vdev_op_leaf)
vd->vdev_not_present = 1;
/*
* Post the appropriate ereport. If the 'prevstate' field is
* set to something other than VDEV_STATE_UNKNOWN, it indicates
* that this is part of a vdev_reopen(). In this case, we don't
* want to post the ereport if the device was already in the
* CANT_OPEN state beforehand.
*
* If the 'checkremove' flag is set, then this is an attempt to
* online the device in response to an insertion event. If we
* hit this case, then we have detected an insertion event for a
* faulted or offline device that wasn't in the removed state.
* In this scenario, we don't post an ereport because we are
* about to replace the device, or attempt an online with
* vdev_forcefault, which will generate the fault for us.
*/
if ((vd->vdev_prevstate != state || vd->vdev_forcefault) &&
!vd->vdev_not_present && !vd->vdev_checkremove &&
vd != spa->spa_root_vdev) {
const char *class;
switch (aux) {
case VDEV_AUX_OPEN_FAILED:
class = FM_EREPORT_ZFS_DEVICE_OPEN_FAILED;
break;
case VDEV_AUX_CORRUPT_DATA:
class = FM_EREPORT_ZFS_DEVICE_CORRUPT_DATA;
break;
case VDEV_AUX_NO_REPLICAS:
class = FM_EREPORT_ZFS_DEVICE_NO_REPLICAS;
break;
case VDEV_AUX_BAD_GUID_SUM:
class = FM_EREPORT_ZFS_DEVICE_BAD_GUID_SUM;
break;
case VDEV_AUX_TOO_SMALL:
class = FM_EREPORT_ZFS_DEVICE_TOO_SMALL;
break;
case VDEV_AUX_BAD_LABEL:
class = FM_EREPORT_ZFS_DEVICE_BAD_LABEL;
break;
case VDEV_AUX_BAD_ASHIFT:
class = FM_EREPORT_ZFS_DEVICE_BAD_ASHIFT;
break;
default:
class = FM_EREPORT_ZFS_DEVICE_UNKNOWN;
}
(void) zfs_ereport_post(class, spa, vd, NULL, NULL,
save_state);
}
/* Erase any notion of persistent removed state */
vd->vdev_removed = B_FALSE;
} else {
vd->vdev_removed = B_FALSE;
}
/*
* Notify ZED of any significant state-change on a leaf vdev.
*
*/
if (vd->vdev_ops->vdev_op_leaf) {
/* preserve original state from a vdev_reopen() */
if ((vd->vdev_prevstate != VDEV_STATE_UNKNOWN) &&
(vd->vdev_prevstate != vd->vdev_state) &&
(save_state <= VDEV_STATE_CLOSED))
save_state = vd->vdev_prevstate;
/* filter out state change due to initial vdev_open */
if (save_state > VDEV_STATE_CLOSED)
zfs_post_state_change(spa, vd, save_state);
}
if (!isopen && vd->vdev_parent)
vdev_propagate_state(vd->vdev_parent);
}
boolean_t
vdev_children_are_offline(vdev_t *vd)
{
ASSERT(!vd->vdev_ops->vdev_op_leaf);
for (uint64_t i = 0; i < vd->vdev_children; i++) {
if (vd->vdev_child[i]->vdev_state != VDEV_STATE_OFFLINE)
return (B_FALSE);
}
return (B_TRUE);
}
/*
* Check the vdev configuration to ensure that it's capable of supporting
* a root pool. We do not support partial configuration.
*/
boolean_t
vdev_is_bootable(vdev_t *vd)
{
if (!vd->vdev_ops->vdev_op_leaf) {
const char *vdev_type = vd->vdev_ops->vdev_op_type;
if (strcmp(vdev_type, VDEV_TYPE_MISSING) == 0)
return (B_FALSE);
}
for (int c = 0; c < vd->vdev_children; c++) {
if (!vdev_is_bootable(vd->vdev_child[c]))
return (B_FALSE);
}
return (B_TRUE);
}
boolean_t
vdev_is_concrete(vdev_t *vd)
{
vdev_ops_t *ops = vd->vdev_ops;
if (ops == &vdev_indirect_ops || ops == &vdev_hole_ops ||
ops == &vdev_missing_ops || ops == &vdev_root_ops) {
return (B_FALSE);
} else {
return (B_TRUE);
}
}
/*
* Determine if a log device has valid content. If the vdev was
* removed or faulted in the MOS config then we know that
* the content on the log device has already been written to the pool.
*/
boolean_t
vdev_log_state_valid(vdev_t *vd)
{
if (vd->vdev_ops->vdev_op_leaf && !vd->vdev_faulted &&
!vd->vdev_removed)
return (B_TRUE);
for (int c = 0; c < vd->vdev_children; c++)
if (vdev_log_state_valid(vd->vdev_child[c]))
return (B_TRUE);
return (B_FALSE);
}
/*
* Expand a vdev if possible.
*/
void
vdev_expand(vdev_t *vd, uint64_t txg)
{
ASSERT(vd->vdev_top == vd);
ASSERT(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
ASSERT(vdev_is_concrete(vd));
vdev_set_deflate_ratio(vd);
if ((vd->vdev_asize >> vd->vdev_ms_shift) > vd->vdev_ms_count &&
vdev_is_concrete(vd)) {
vdev_metaslab_group_create(vd);
VERIFY(vdev_metaslab_init(vd, txg) == 0);
vdev_config_dirty(vd);
}
}
/*
* Split a vdev.
*/
void
vdev_split(vdev_t *vd)
{
vdev_t *cvd, *pvd = vd->vdev_parent;
vdev_remove_child(pvd, vd);
vdev_compact_children(pvd);
cvd = pvd->vdev_child[0];
if (pvd->vdev_children == 1) {
vdev_remove_parent(cvd);
cvd->vdev_splitting = B_TRUE;
}
vdev_propagate_state(cvd);
}
void
vdev_deadman(vdev_t *vd, char *tag)
{
for (int c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
vdev_deadman(cvd, tag);
}
if (vd->vdev_ops->vdev_op_leaf) {
vdev_queue_t *vq = &vd->vdev_queue;
mutex_enter(&vq->vq_lock);
if (avl_numnodes(&vq->vq_active_tree) > 0) {
spa_t *spa = vd->vdev_spa;
zio_t *fio;
uint64_t delta;
zfs_dbgmsg("slow vdev: %s has %lu active IOs",
vd->vdev_path, avl_numnodes(&vq->vq_active_tree));
/*
* Look at the head of all the pending queues,
* if any I/O has been outstanding for longer than
* the spa_deadman_synctime invoke the deadman logic.
*/
fio = avl_first(&vq->vq_active_tree);
delta = gethrtime() - fio->io_timestamp;
if (delta > spa_deadman_synctime(spa))
zio_deadman(fio, tag);
}
mutex_exit(&vq->vq_lock);
}
}
void
vdev_defer_resilver(vdev_t *vd)
{
ASSERT(vd->vdev_ops->vdev_op_leaf);
vd->vdev_resilver_deferred = B_TRUE;
vd->vdev_spa->spa_resilver_deferred = B_TRUE;
}
/*
* Clears the resilver deferred flag on all leaf devs under vd. Returns
* B_TRUE if we have devices that need to be resilvered and are available to
* accept resilver I/Os.
*/
boolean_t
vdev_clear_resilver_deferred(vdev_t *vd, dmu_tx_t *tx)
{
boolean_t resilver_needed = B_FALSE;
spa_t *spa = vd->vdev_spa;
for (int c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
resilver_needed |= vdev_clear_resilver_deferred(cvd, tx);
}
if (vd == spa->spa_root_vdev &&
spa_feature_is_active(spa, SPA_FEATURE_RESILVER_DEFER)) {
spa_feature_decr(spa, SPA_FEATURE_RESILVER_DEFER, tx);
vdev_config_dirty(vd);
spa->spa_resilver_deferred = B_FALSE;
return (resilver_needed);
}
if (!vdev_is_concrete(vd) || vd->vdev_aux ||
!vd->vdev_ops->vdev_op_leaf)
return (resilver_needed);
vd->vdev_resilver_deferred = B_FALSE;
return (!vdev_is_dead(vd) && !vd->vdev_offline &&
vdev_resilver_needed(vd, NULL, NULL));
}
boolean_t
vdev_xlate_is_empty(range_seg64_t *rs)
{
return (rs->rs_start == rs->rs_end);
}
/*
* Translate a logical range to the first contiguous physical range for the
* specified vdev_t. This function is initially called with a leaf vdev and
* will walk each parent vdev until it reaches a top-level vdev. Once the
* top-level is reached the physical range is initialized and the recursive
* function begins to unwind. As it unwinds it calls the parent's vdev
* specific translation function to do the real conversion.
*/
void
vdev_xlate(vdev_t *vd, const range_seg64_t *logical_rs,
range_seg64_t *physical_rs, range_seg64_t *remain_rs)
{
/*
* Walk up the vdev tree
*/
if (vd != vd->vdev_top) {
vdev_xlate(vd->vdev_parent, logical_rs, physical_rs,
remain_rs);
} else {
/*
* We've reached the top-level vdev, initialize the physical
* range to the logical range and set an empty remaining
* range then start to unwind.
*/
physical_rs->rs_start = logical_rs->rs_start;
physical_rs->rs_end = logical_rs->rs_end;
remain_rs->rs_start = logical_rs->rs_start;
remain_rs->rs_end = logical_rs->rs_start;
return;
}
vdev_t *pvd = vd->vdev_parent;
ASSERT3P(pvd, !=, NULL);
ASSERT3P(pvd->vdev_ops->vdev_op_xlate, !=, NULL);
/*
* As this recursive function unwinds, translate the logical
* range into its physical and any remaining components by calling
* the vdev specific translate function.
*/
range_seg64_t intermediate = { 0 };
pvd->vdev_ops->vdev_op_xlate(vd, physical_rs, &intermediate, remain_rs);
physical_rs->rs_start = intermediate.rs_start;
physical_rs->rs_end = intermediate.rs_end;
}
void
vdev_xlate_walk(vdev_t *vd, const range_seg64_t *logical_rs,
vdev_xlate_func_t *func, void *arg)
{
range_seg64_t iter_rs = *logical_rs;
range_seg64_t physical_rs;
range_seg64_t remain_rs;
while (!vdev_xlate_is_empty(&iter_rs)) {
vdev_xlate(vd, &iter_rs, &physical_rs, &remain_rs);
/*
* With raidz and dRAID, it's possible that the logical range
* does not live on this leaf vdev. Only when there is a non-
* zero physical size call the provided function.
*/
if (!vdev_xlate_is_empty(&physical_rs))
func(arg, &physical_rs);
iter_rs = remain_rs;
}
}
static char *
vdev_name(vdev_t *vd, char *buf, int buflen)
{
if (vd->vdev_path == NULL) {
if (strcmp(vd->vdev_ops->vdev_op_type, "root") == 0) {
strlcpy(buf, vd->vdev_spa->spa_name, buflen);
} else if (!vd->vdev_ops->vdev_op_leaf) {
snprintf(buf, buflen, "%s-%llu",
vd->vdev_ops->vdev_op_type,
(u_longlong_t)vd->vdev_id);
}
} else {
strlcpy(buf, vd->vdev_path, buflen);
}
return (buf);
}
/*
* Look at the vdev tree and determine whether any devices are currently being
* replaced.
*/
boolean_t
vdev_replace_in_progress(vdev_t *vdev)
{
ASSERT(spa_config_held(vdev->vdev_spa, SCL_ALL, RW_READER) != 0);
if (vdev->vdev_ops == &vdev_replacing_ops)
return (B_TRUE);
/*
* A 'spare' vdev indicates that we have a replace in progress, unless
* it has exactly two children, and the second, the hot spare, has
* finished being resilvered.
*/
if (vdev->vdev_ops == &vdev_spare_ops && (vdev->vdev_children > 2 ||
!vdev_dtl_empty(vdev->vdev_child[1], DTL_MISSING)))
return (B_TRUE);
for (int i = 0; i < vdev->vdev_children; i++) {
if (vdev_replace_in_progress(vdev->vdev_child[i]))
return (B_TRUE);
}
return (B_FALSE);
}
/*
* Add a (source=src, propname=propval) list to an nvlist.
*/
static void
vdev_prop_add_list(nvlist_t *nvl, const char *propname, char *strval,
uint64_t intval, zprop_source_t src)
{
nvlist_t *propval;
propval = fnvlist_alloc();
fnvlist_add_uint64(propval, ZPROP_SOURCE, src);
if (strval != NULL)
fnvlist_add_string(propval, ZPROP_VALUE, strval);
else
fnvlist_add_uint64(propval, ZPROP_VALUE, intval);
fnvlist_add_nvlist(nvl, propname, propval);
nvlist_free(propval);
}
static void
vdev_props_set_sync(void *arg, dmu_tx_t *tx)
{
vdev_t *vd;
nvlist_t *nvp = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
objset_t *mos = spa->spa_meta_objset;
nvpair_t *elem = NULL;
uint64_t vdev_guid;
nvlist_t *nvprops;
vdev_guid = fnvlist_lookup_uint64(nvp, ZPOOL_VDEV_PROPS_SET_VDEV);
nvprops = fnvlist_lookup_nvlist(nvp, ZPOOL_VDEV_PROPS_SET_PROPS);
vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE);
/* this vdev could get removed while waiting for this sync task */
if (vd == NULL)
return;
mutex_enter(&spa->spa_props_lock);
while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
uint64_t intval, objid = 0;
char *strval;
vdev_prop_t prop;
const char *propname = nvpair_name(elem);
zprop_type_t proptype;
/*
* Set vdev property values in the vdev props mos object.
*/
if (vd->vdev_top_zap != 0) {
objid = vd->vdev_top_zap;
} else if (vd->vdev_leaf_zap != 0) {
objid = vd->vdev_leaf_zap;
} else {
panic("vdev not top or leaf");
}
switch (prop = vdev_name_to_prop(propname)) {
case VDEV_PROP_USER:
if (vdev_prop_user(propname)) {
strval = fnvpair_value_string(elem);
if (strlen(strval) == 0) {
/* remove the property if value == "" */
(void) zap_remove(mos, objid, propname,
tx);
} else {
VERIFY0(zap_update(mos, objid, propname,
1, strlen(strval) + 1, strval, tx));
}
spa_history_log_internal(spa, "vdev set", tx,
"vdev_guid=%llu: %s=%s",
(u_longlong_t)vdev_guid, nvpair_name(elem),
strval);
}
break;
default:
/* normalize the property name */
propname = vdev_prop_to_name(prop);
proptype = vdev_prop_get_type(prop);
if (nvpair_type(elem) == DATA_TYPE_STRING) {
ASSERT(proptype == PROP_TYPE_STRING);
strval = fnvpair_value_string(elem);
VERIFY0(zap_update(mos, objid, propname,
1, strlen(strval) + 1, strval, tx));
spa_history_log_internal(spa, "vdev set", tx,
"vdev_guid=%llu: %s=%s",
(u_longlong_t)vdev_guid, nvpair_name(elem),
strval);
} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
intval = fnvpair_value_uint64(elem);
if (proptype == PROP_TYPE_INDEX) {
const char *unused;
VERIFY0(vdev_prop_index_to_string(
prop, intval, &unused));
}
VERIFY0(zap_update(mos, objid, propname,
sizeof (uint64_t), 1, &intval, tx));
spa_history_log_internal(spa, "vdev set", tx,
"vdev_guid=%llu: %s=%lld",
(u_longlong_t)vdev_guid,
nvpair_name(elem), (longlong_t)intval);
} else {
panic("invalid vdev property type %u",
nvpair_type(elem));
}
}
}
mutex_exit(&spa->spa_props_lock);
}
int
vdev_prop_set(vdev_t *vd, nvlist_t *innvl, nvlist_t *outnvl)
{
spa_t *spa = vd->vdev_spa;
nvpair_t *elem = NULL;
uint64_t vdev_guid;
nvlist_t *nvprops;
int error;
ASSERT(vd != NULL);
if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
&vdev_guid) != 0)
return (SET_ERROR(EINVAL));
if (nvlist_lookup_nvlist(innvl, ZPOOL_VDEV_PROPS_SET_PROPS,
&nvprops) != 0)
return (SET_ERROR(EINVAL));
if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL)
return (SET_ERROR(EINVAL));
while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
char *propname = nvpair_name(elem);
vdev_prop_t prop = vdev_name_to_prop(propname);
uint64_t intval = 0;
char *strval = NULL;
if (prop == VDEV_PROP_USER && !vdev_prop_user(propname)) {
error = EINVAL;
goto end;
}
if (vdev_prop_readonly(prop)) {
error = EROFS;
goto end;
}
/* Special Processing */
switch (prop) {
case VDEV_PROP_PATH:
if (vd->vdev_path == NULL) {
error = EROFS;
break;
}
if (nvpair_value_string(elem, &strval) != 0) {
error = EINVAL;
break;
}
/* New path must start with /dev/ */
if (strncmp(strval, "/dev/", 5)) {
error = EINVAL;
break;
}
error = spa_vdev_setpath(spa, vdev_guid, strval);
break;
case VDEV_PROP_ALLOCATING:
if (nvpair_value_uint64(elem, &intval) != 0) {
error = EINVAL;
break;
}
if (intval != vd->vdev_noalloc)
break;
if (intval == 0)
error = spa_vdev_noalloc(spa, vdev_guid);
else
error = spa_vdev_alloc(spa, vdev_guid);
break;
default:
/* Most processing is done in vdev_props_set_sync */
break;
}
end:
if (error != 0) {
intval = error;
vdev_prop_add_list(outnvl, propname, strval, intval, 0);
return (error);
}
}
return (dsl_sync_task(spa->spa_name, NULL, vdev_props_set_sync,
innvl, 6, ZFS_SPACE_CHECK_EXTRA_RESERVED));
}
int
vdev_prop_get(vdev_t *vd, nvlist_t *innvl, nvlist_t *outnvl)
{
spa_t *spa = vd->vdev_spa;
objset_t *mos = spa->spa_meta_objset;
int err = 0;
uint64_t objid;
uint64_t vdev_guid;
nvpair_t *elem = NULL;
nvlist_t *nvprops = NULL;
uint64_t intval = 0;
char *strval = NULL;
const char *propname = NULL;
vdev_prop_t prop;
ASSERT(vd != NULL);
ASSERT(mos != NULL);
if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_GET_VDEV,
&vdev_guid) != 0)
return (SET_ERROR(EINVAL));
nvlist_lookup_nvlist(innvl, ZPOOL_VDEV_PROPS_GET_PROPS, &nvprops);
if (vd->vdev_top_zap != 0) {
objid = vd->vdev_top_zap;
} else if (vd->vdev_leaf_zap != 0) {
objid = vd->vdev_leaf_zap;
} else {
return (SET_ERROR(EINVAL));
}
ASSERT(objid != 0);
mutex_enter(&spa->spa_props_lock);
if (nvprops != NULL) {
char namebuf[64] = { 0 };
while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
intval = 0;
strval = NULL;
propname = nvpair_name(elem);
prop = vdev_name_to_prop(propname);
zprop_source_t src = ZPROP_SRC_DEFAULT;
uint64_t integer_size, num_integers;
switch (prop) {
/* Special Read-only Properties */
case VDEV_PROP_NAME:
strval = vdev_name(vd, namebuf,
sizeof (namebuf));
if (strval == NULL)
continue;
vdev_prop_add_list(outnvl, propname, strval, 0,
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_CAPACITY:
/* percent used */
intval = (vd->vdev_stat.vs_dspace == 0) ? 0 :
(vd->vdev_stat.vs_alloc * 100 /
vd->vdev_stat.vs_dspace);
vdev_prop_add_list(outnvl, propname, NULL,
intval, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_STATE:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_state, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_GUID:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_guid, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_ASIZE:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_asize, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_PSIZE:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_psize, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_ASHIFT:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_ashift, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_SIZE:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_dspace, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_FREE:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_dspace -
vd->vdev_stat.vs_alloc, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_ALLOCATED:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_alloc, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_EXPANDSZ:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_esize, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_FRAGMENTATION:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_fragmentation,
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_PARITY:
vdev_prop_add_list(outnvl, propname, NULL,
vdev_get_nparity(vd), ZPROP_SRC_NONE);
continue;
case VDEV_PROP_PATH:
if (vd->vdev_path == NULL)
continue;
vdev_prop_add_list(outnvl, propname,
vd->vdev_path, 0, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_DEVID:
if (vd->vdev_devid == NULL)
continue;
vdev_prop_add_list(outnvl, propname,
vd->vdev_devid, 0, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_PHYS_PATH:
if (vd->vdev_physpath == NULL)
continue;
vdev_prop_add_list(outnvl, propname,
vd->vdev_physpath, 0, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_ENC_PATH:
if (vd->vdev_enc_sysfs_path == NULL)
continue;
vdev_prop_add_list(outnvl, propname,
vd->vdev_enc_sysfs_path, 0, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_FRU:
if (vd->vdev_fru == NULL)
continue;
vdev_prop_add_list(outnvl, propname,
vd->vdev_fru, 0, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_PARENT:
if (vd->vdev_parent != NULL) {
strval = vdev_name(vd->vdev_parent,
namebuf, sizeof (namebuf));
vdev_prop_add_list(outnvl, propname,
strval, 0, ZPROP_SRC_NONE);
}
continue;
case VDEV_PROP_CHILDREN:
if (vd->vdev_children > 0)
strval = kmem_zalloc(ZAP_MAXVALUELEN,
KM_SLEEP);
for (uint64_t i = 0; i < vd->vdev_children;
i++) {
char *vname;
vname = vdev_name(vd->vdev_child[i],
namebuf, sizeof (namebuf));
if (vname == NULL)
vname = "(unknown)";
if (strlen(strval) > 0)
strlcat(strval, ",",
ZAP_MAXVALUELEN);
strlcat(strval, vname, ZAP_MAXVALUELEN);
}
if (strval != NULL) {
vdev_prop_add_list(outnvl, propname,
strval, 0, ZPROP_SRC_NONE);
kmem_free(strval, ZAP_MAXVALUELEN);
}
continue;
case VDEV_PROP_NUMCHILDREN:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_children, ZPROP_SRC_NONE);
continue;
case VDEV_PROP_READ_ERRORS:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_read_errors,
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_WRITE_ERRORS:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_write_errors,
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_CHECKSUM_ERRORS:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_checksum_errors,
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_INITIALIZE_ERRORS:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_initialize_errors,
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_OPS_NULL:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_ops[ZIO_TYPE_NULL],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_OPS_READ:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_ops[ZIO_TYPE_READ],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_OPS_WRITE:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_ops[ZIO_TYPE_WRITE],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_OPS_FREE:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_ops[ZIO_TYPE_FREE],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_OPS_CLAIM:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_ops[ZIO_TYPE_CLAIM],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_OPS_TRIM:
/*
* TRIM ops and bytes are reported to user
* space as ZIO_TYPE_IOCTL. This is done to
* preserve the vdev_stat_t structure layout
* for user space.
*/
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_ops[ZIO_TYPE_IOCTL],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_BYTES_NULL:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_bytes[ZIO_TYPE_NULL],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_BYTES_READ:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_bytes[ZIO_TYPE_READ],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_BYTES_WRITE:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_bytes[ZIO_TYPE_WRITE],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_BYTES_FREE:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_bytes[ZIO_TYPE_FREE],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_BYTES_CLAIM:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_bytes[ZIO_TYPE_CLAIM],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_BYTES_TRIM:
/*
* TRIM ops and bytes are reported to user
* space as ZIO_TYPE_IOCTL. This is done to
* preserve the vdev_stat_t structure layout
* for user space.
*/
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_stat.vs_bytes[ZIO_TYPE_IOCTL],
ZPROP_SRC_NONE);
continue;
case VDEV_PROP_REMOVING:
vdev_prop_add_list(outnvl, propname, NULL,
vd->vdev_removing, ZPROP_SRC_NONE);
continue;
/* Numeric Properites */
case VDEV_PROP_ALLOCATING:
src = ZPROP_SRC_LOCAL;
strval = NULL;
err = zap_lookup(mos, objid, nvpair_name(elem),
sizeof (uint64_t), 1, &intval);
if (err == ENOENT) {
intval =
vdev_prop_default_numeric(prop);
err = 0;
} else if (err)
break;
if (intval == vdev_prop_default_numeric(prop))
src = ZPROP_SRC_DEFAULT;
/* Leaf vdevs cannot have this property */
if (vd->vdev_mg == NULL &&
vd->vdev_top != NULL) {
src = ZPROP_SRC_NONE;
intval = ZPROP_BOOLEAN_NA;
}
vdev_prop_add_list(outnvl, propname, strval,
intval, src);
break;
/* Text Properties */
case VDEV_PROP_COMMENT:
/* Exists in the ZAP below */
/* FALLTHRU */
case VDEV_PROP_USER:
/* User Properites */
src = ZPROP_SRC_LOCAL;
err = zap_length(mos, objid, nvpair_name(elem),
&integer_size, &num_integers);
if (err)
break;
switch (integer_size) {
case 8:
/* User properties cannot be integers */
err = EINVAL;
break;
case 1:
/* string property */
strval = kmem_alloc(num_integers,
KM_SLEEP);
err = zap_lookup(mos, objid,
nvpair_name(elem), 1,
num_integers, strval);
if (err) {
kmem_free(strval,
num_integers);
break;
}
vdev_prop_add_list(outnvl, propname,
strval, 0, src);
kmem_free(strval, num_integers);
break;
}
break;
default:
err = ENOENT;
break;
}
if (err)
break;
}
} else {
/*
* Get all properties from the MOS vdev property object.
*/
zap_cursor_t zc;
zap_attribute_t za;
for (zap_cursor_init(&zc, mos, objid);
(err = zap_cursor_retrieve(&zc, &za)) == 0;
zap_cursor_advance(&zc)) {
intval = 0;
strval = NULL;
zprop_source_t src = ZPROP_SRC_DEFAULT;
propname = za.za_name;
prop = vdev_name_to_prop(propname);
switch (za.za_integer_length) {
case 8:
/* We do not allow integer user properties */
/* This is likely an internal value */
break;
case 1:
/* string property */
strval = kmem_alloc(za.za_num_integers,
KM_SLEEP);
err = zap_lookup(mos, objid, za.za_name, 1,
za.za_num_integers, strval);
if (err) {
kmem_free(strval, za.za_num_integers);
break;
}
vdev_prop_add_list(outnvl, propname, strval, 0,
src);
kmem_free(strval, za.za_num_integers);
break;
default:
break;
}
}
zap_cursor_fini(&zc);
}
mutex_exit(&spa->spa_props_lock);
if (err && err != ENOENT) {
return (err);
}
return (0);
}
EXPORT_SYMBOL(vdev_fault);
EXPORT_SYMBOL(vdev_degrade);
EXPORT_SYMBOL(vdev_online);
EXPORT_SYMBOL(vdev_offline);
EXPORT_SYMBOL(vdev_clear);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, default_ms_count, INT, ZMOD_RW,
"Target number of metaslabs per top-level vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, default_ms_shift, INT, ZMOD_RW,
"Default limit for metaslab size");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, min_ms_count, INT, ZMOD_RW,
"Minimum number of metaslabs per top-level vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, ms_count_limit, INT, ZMOD_RW,
"Practical upper limit of total metaslabs per top-level vdev");
ZFS_MODULE_PARAM(zfs, zfs_, slow_io_events_per_second, UINT, ZMOD_RW,
"Rate limit slow IO (delay) events to this many per second");
ZFS_MODULE_PARAM(zfs, zfs_, checksum_events_per_second, UINT, ZMOD_RW,
"Rate limit checksum events to this many checksum errors per second "
"(do not set below zed threshold).");
ZFS_MODULE_PARAM(zfs, zfs_, scan_ignore_errors, INT, ZMOD_RW,
"Ignore errors during resilver/scrub");
ZFS_MODULE_PARAM(zfs_vdev, vdev_, validate_skip, INT, ZMOD_RW,
"Bypass vdev_validate()");
ZFS_MODULE_PARAM(zfs, zfs_, nocacheflush, INT, ZMOD_RW,
"Disable cache flushes");
ZFS_MODULE_PARAM(zfs, zfs_, embedded_slog_min_ms, INT, ZMOD_RW,
"Minimum number of metaslabs required to dedicate one for log blocks");
ZFS_MODULE_PARAM_CALL(zfs_vdev, zfs_vdev_, min_auto_ashift,
param_set_min_auto_ashift, param_get_ulong, ZMOD_RW,
"Minimum ashift used when creating new top-level vdevs");
ZFS_MODULE_PARAM_CALL(zfs_vdev, zfs_vdev_, max_auto_ashift,
param_set_max_auto_ashift, param_get_ulong, ZMOD_RW,
"Maximum ashift used when optimizing for logical -> physical sector "
"size on new top-level vdevs");
/* END CSTYLED */
diff --git a/module/zfs/vdev_cache.c b/module/zfs/vdev_cache.c
index 35ed1a3352fe..b6e680334d5c 100644
--- a/module/zfs/vdev_cache.c
+++ b/module/zfs/vdev_cache.c
@@ -1,438 +1,438 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright (c) 2013, 2016 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/vdev_impl.h>
#include <sys/zio.h>
#include <sys/kstat.h>
#include <sys/abd.h>
/*
* Virtual device read-ahead caching.
*
* This file implements a simple LRU read-ahead cache. When the DMU reads
* a given block, it will often want other, nearby blocks soon thereafter.
* We take advantage of this by reading a larger disk region and caching
* the result. In the best case, this can turn 128 back-to-back 512-byte
* reads into a single 64k read followed by 127 cache hits; this reduces
* latency dramatically. In the worst case, it can turn an isolated 512-byte
* read into a 64k read, which doesn't affect latency all that much but is
* terribly wasteful of bandwidth. A more intelligent version of the cache
* could keep track of access patterns and not do read-ahead unless it sees
* at least two temporally close I/Os to the same region. Currently, only
* metadata I/O is inflated. A further enhancement could take advantage of
* more semantic information about the I/O. And it could use something
* faster than an AVL tree; that was chosen solely for convenience.
*
* There are five cache operations: allocate, fill, read, write, evict.
*
* (1) Allocate. This reserves a cache entry for the specified region.
* We separate the allocate and fill operations so that multiple threads
* don't generate I/O for the same cache miss.
*
* (2) Fill. When the I/O for a cache miss completes, the fill routine
* places the data in the previously allocated cache entry.
*
* (3) Read. Read data from the cache.
*
* (4) Write. Update cache contents after write completion.
*
* (5) Evict. When allocating a new entry, we evict the oldest (LRU) entry
* if the total cache size exceeds zfs_vdev_cache_size.
*/
/*
* These tunables are for performance analysis.
*/
/*
* All i/os smaller than zfs_vdev_cache_max will be turned into
* 1<<zfs_vdev_cache_bshift byte reads by the vdev_cache (aka software
* track buffer). At most zfs_vdev_cache_size bytes will be kept in each
* vdev's vdev_cache.
*
* TODO: Note that with the current ZFS code, it turns out that the
* vdev cache is not helpful, and in some cases actually harmful. It
* is better if we disable this. Once some time has passed, we should
* actually remove this to simplify the code. For now we just disable
* it by setting the zfs_vdev_cache_size to zero. Note that Solaris 11
* has made these same changes.
*/
-int zfs_vdev_cache_max = 1<<14; /* 16KB */
-int zfs_vdev_cache_size = 0;
-int zfs_vdev_cache_bshift = 16;
+static int zfs_vdev_cache_max = 1<<14; /* 16KB */
+static int zfs_vdev_cache_size = 0;
+static int zfs_vdev_cache_bshift = 16;
#define VCBS (1 << zfs_vdev_cache_bshift) /* 64KB */
-kstat_t *vdc_ksp = NULL;
+static kstat_t *vdc_ksp = NULL;
typedef struct vdc_stats {
kstat_named_t vdc_stat_delegations;
kstat_named_t vdc_stat_hits;
kstat_named_t vdc_stat_misses;
} vdc_stats_t;
static vdc_stats_t vdc_stats = {
{ "delegations", KSTAT_DATA_UINT64 },
{ "hits", KSTAT_DATA_UINT64 },
{ "misses", KSTAT_DATA_UINT64 }
};
#define VDCSTAT_BUMP(stat) atomic_inc_64(&vdc_stats.stat.value.ui64);
static inline int
vdev_cache_offset_compare(const void *a1, const void *a2)
{
const vdev_cache_entry_t *ve1 = (const vdev_cache_entry_t *)a1;
const vdev_cache_entry_t *ve2 = (const vdev_cache_entry_t *)a2;
return (TREE_CMP(ve1->ve_offset, ve2->ve_offset));
}
static int
vdev_cache_lastused_compare(const void *a1, const void *a2)
{
const vdev_cache_entry_t *ve1 = (const vdev_cache_entry_t *)a1;
const vdev_cache_entry_t *ve2 = (const vdev_cache_entry_t *)a2;
int cmp = TREE_CMP(ve1->ve_lastused, ve2->ve_lastused);
if (likely(cmp))
return (cmp);
/*
* Among equally old entries, sort by offset to ensure uniqueness.
*/
return (vdev_cache_offset_compare(a1, a2));
}
/*
* Evict the specified entry from the cache.
*/
static void
vdev_cache_evict(vdev_cache_t *vc, vdev_cache_entry_t *ve)
{
ASSERT(MUTEX_HELD(&vc->vc_lock));
ASSERT3P(ve->ve_fill_io, ==, NULL);
ASSERT3P(ve->ve_abd, !=, NULL);
avl_remove(&vc->vc_lastused_tree, ve);
avl_remove(&vc->vc_offset_tree, ve);
abd_free(ve->ve_abd);
kmem_free(ve, sizeof (vdev_cache_entry_t));
}
/*
* Allocate an entry in the cache. At the point we don't have the data,
* we're just creating a placeholder so that multiple threads don't all
* go off and read the same blocks.
*/
static vdev_cache_entry_t *
vdev_cache_allocate(zio_t *zio)
{
vdev_cache_t *vc = &zio->io_vd->vdev_cache;
uint64_t offset = P2ALIGN(zio->io_offset, VCBS);
vdev_cache_entry_t *ve;
ASSERT(MUTEX_HELD(&vc->vc_lock));
if (zfs_vdev_cache_size == 0)
return (NULL);
/*
* If adding a new entry would exceed the cache size,
* evict the oldest entry (LRU).
*/
if ((avl_numnodes(&vc->vc_lastused_tree) << zfs_vdev_cache_bshift) >
zfs_vdev_cache_size) {
ve = avl_first(&vc->vc_lastused_tree);
if (ve->ve_fill_io != NULL)
return (NULL);
ASSERT3U(ve->ve_hits, !=, 0);
vdev_cache_evict(vc, ve);
}
ve = kmem_zalloc(sizeof (vdev_cache_entry_t), KM_SLEEP);
ve->ve_offset = offset;
ve->ve_lastused = ddi_get_lbolt();
ve->ve_abd = abd_alloc_for_io(VCBS, B_TRUE);
avl_add(&vc->vc_offset_tree, ve);
avl_add(&vc->vc_lastused_tree, ve);
return (ve);
}
static void
vdev_cache_hit(vdev_cache_t *vc, vdev_cache_entry_t *ve, zio_t *zio)
{
uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);
ASSERT(MUTEX_HELD(&vc->vc_lock));
ASSERT3P(ve->ve_fill_io, ==, NULL);
if (ve->ve_lastused != ddi_get_lbolt()) {
avl_remove(&vc->vc_lastused_tree, ve);
ve->ve_lastused = ddi_get_lbolt();
avl_add(&vc->vc_lastused_tree, ve);
}
ve->ve_hits++;
abd_copy_off(zio->io_abd, ve->ve_abd, 0, cache_phase, zio->io_size);
}
/*
* Fill a previously allocated cache entry with data.
*/
static void
vdev_cache_fill(zio_t *fio)
{
vdev_t *vd = fio->io_vd;
vdev_cache_t *vc = &vd->vdev_cache;
vdev_cache_entry_t *ve = fio->io_private;
zio_t *pio;
ASSERT3U(fio->io_size, ==, VCBS);
/*
* Add data to the cache.
*/
mutex_enter(&vc->vc_lock);
ASSERT3P(ve->ve_fill_io, ==, fio);
ASSERT3U(ve->ve_offset, ==, fio->io_offset);
ASSERT3P(ve->ve_abd, ==, fio->io_abd);
ve->ve_fill_io = NULL;
/*
* Even if this cache line was invalidated by a missed write update,
* any reads that were queued up before the missed update are still
* valid, so we can satisfy them from this line before we evict it.
*/
zio_link_t *zl = NULL;
while ((pio = zio_walk_parents(fio, &zl)) != NULL)
vdev_cache_hit(vc, ve, pio);
if (fio->io_error || ve->ve_missed_update)
vdev_cache_evict(vc, ve);
mutex_exit(&vc->vc_lock);
}
/*
* Read data from the cache. Returns B_TRUE cache hit, B_FALSE on miss.
*/
boolean_t
vdev_cache_read(zio_t *zio)
{
vdev_cache_t *vc = &zio->io_vd->vdev_cache;
vdev_cache_entry_t *ve, ve_search;
uint64_t cache_offset = P2ALIGN(zio->io_offset, VCBS);
zio_t *fio;
uint64_t cache_phase __maybe_unused = P2PHASE(zio->io_offset, VCBS);
ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ);
if (zfs_vdev_cache_size == 0)
return (B_FALSE);
if (zio->io_flags & ZIO_FLAG_DONT_CACHE)
return (B_FALSE);
if (zio->io_size > zfs_vdev_cache_max)
return (B_FALSE);
/*
* If the I/O straddles two or more cache blocks, don't cache it.
*/
if (P2BOUNDARY(zio->io_offset, zio->io_size, VCBS))
return (B_FALSE);
ASSERT3U(cache_phase + zio->io_size, <=, VCBS);
mutex_enter(&vc->vc_lock);
ve_search.ve_offset = cache_offset;
ve = avl_find(&vc->vc_offset_tree, &ve_search, NULL);
if (ve != NULL) {
if (ve->ve_missed_update) {
mutex_exit(&vc->vc_lock);
return (B_FALSE);
}
if ((fio = ve->ve_fill_io) != NULL) {
zio_vdev_io_bypass(zio);
zio_add_child(zio, fio);
mutex_exit(&vc->vc_lock);
VDCSTAT_BUMP(vdc_stat_delegations);
return (B_TRUE);
}
vdev_cache_hit(vc, ve, zio);
zio_vdev_io_bypass(zio);
mutex_exit(&vc->vc_lock);
VDCSTAT_BUMP(vdc_stat_hits);
return (B_TRUE);
}
ve = vdev_cache_allocate(zio);
if (ve == NULL) {
mutex_exit(&vc->vc_lock);
return (B_FALSE);
}
fio = zio_vdev_delegated_io(zio->io_vd, cache_offset,
ve->ve_abd, VCBS, ZIO_TYPE_READ, ZIO_PRIORITY_NOW,
ZIO_FLAG_DONT_CACHE, vdev_cache_fill, ve);
ve->ve_fill_io = fio;
zio_vdev_io_bypass(zio);
zio_add_child(zio, fio);
mutex_exit(&vc->vc_lock);
zio_nowait(fio);
VDCSTAT_BUMP(vdc_stat_misses);
return (B_TRUE);
}
/*
* Update cache contents upon write completion.
*/
void
vdev_cache_write(zio_t *zio)
{
vdev_cache_t *vc = &zio->io_vd->vdev_cache;
vdev_cache_entry_t *ve, ve_search;
uint64_t io_start = zio->io_offset;
uint64_t io_end = io_start + zio->io_size;
uint64_t min_offset = P2ALIGN(io_start, VCBS);
uint64_t max_offset = P2ROUNDUP(io_end, VCBS);
avl_index_t where;
ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);
mutex_enter(&vc->vc_lock);
ve_search.ve_offset = min_offset;
ve = avl_find(&vc->vc_offset_tree, &ve_search, &where);
if (ve == NULL)
ve = avl_nearest(&vc->vc_offset_tree, where, AVL_AFTER);
while (ve != NULL && ve->ve_offset < max_offset) {
uint64_t start = MAX(ve->ve_offset, io_start);
uint64_t end = MIN(ve->ve_offset + VCBS, io_end);
if (ve->ve_fill_io != NULL) {
ve->ve_missed_update = 1;
} else {
abd_copy_off(ve->ve_abd, zio->io_abd,
start - ve->ve_offset, start - io_start,
end - start);
}
ve = AVL_NEXT(&vc->vc_offset_tree, ve);
}
mutex_exit(&vc->vc_lock);
}
void
vdev_cache_purge(vdev_t *vd)
{
vdev_cache_t *vc = &vd->vdev_cache;
vdev_cache_entry_t *ve;
mutex_enter(&vc->vc_lock);
while ((ve = avl_first(&vc->vc_offset_tree)) != NULL)
vdev_cache_evict(vc, ve);
mutex_exit(&vc->vc_lock);
}
void
vdev_cache_init(vdev_t *vd)
{
vdev_cache_t *vc = &vd->vdev_cache;
mutex_init(&vc->vc_lock, NULL, MUTEX_DEFAULT, NULL);
avl_create(&vc->vc_offset_tree, vdev_cache_offset_compare,
sizeof (vdev_cache_entry_t),
offsetof(struct vdev_cache_entry, ve_offset_node));
avl_create(&vc->vc_lastused_tree, vdev_cache_lastused_compare,
sizeof (vdev_cache_entry_t),
offsetof(struct vdev_cache_entry, ve_lastused_node));
}
void
vdev_cache_fini(vdev_t *vd)
{
vdev_cache_t *vc = &vd->vdev_cache;
vdev_cache_purge(vd);
avl_destroy(&vc->vc_offset_tree);
avl_destroy(&vc->vc_lastused_tree);
mutex_destroy(&vc->vc_lock);
}
void
vdev_cache_stat_init(void)
{
vdc_ksp = kstat_create("zfs", 0, "vdev_cache_stats", "misc",
KSTAT_TYPE_NAMED, sizeof (vdc_stats) / sizeof (kstat_named_t),
KSTAT_FLAG_VIRTUAL);
if (vdc_ksp != NULL) {
vdc_ksp->ks_data = &vdc_stats;
kstat_install(vdc_ksp);
}
}
void
vdev_cache_stat_fini(void)
{
if (vdc_ksp != NULL) {
kstat_delete(vdc_ksp);
vdc_ksp = NULL;
}
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, cache_max, INT, ZMOD_RW,
"Inflate reads small than max");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, cache_size, INT, ZMOD_RD,
"Total size of the per-disk cache");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, cache_bshift, INT, ZMOD_RW,
"Shift size to inflate reads too");
/* END CSTYLED */
diff --git a/module/zfs/vdev_indirect.c b/module/zfs/vdev_indirect.c
index 8762855d46aa..aa4801e67442 100644
--- a/module/zfs/vdev_indirect.c
+++ b/module/zfs/vdev_indirect.c
@@ -1,1905 +1,1905 @@
/*
* CDDL HEADER START
*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2014, 2017 by Delphix. All rights reserved.
* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
* Copyright (c) 2014, 2020 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/vdev_impl.h>
#include <sys/fs/zfs.h>
#include <sys/zio.h>
#include <sys/zio_checksum.h>
#include <sys/metaslab.h>
#include <sys/dmu.h>
#include <sys/vdev_indirect_mapping.h>
#include <sys/dmu_tx.h>
#include <sys/dsl_synctask.h>
#include <sys/zap.h>
#include <sys/abd.h>
#include <sys/zthr.h>
/*
* An indirect vdev corresponds to a vdev that has been removed. Since
* we cannot rewrite block pointers of snapshots, etc., we keep a
* mapping from old location on the removed device to the new location
* on another device in the pool and use this mapping whenever we need
* to access the DVA. Unfortunately, this mapping did not respect
* logical block boundaries when it was first created, and so a DVA on
* this indirect vdev may be "split" into multiple sections that each
* map to a different location. As a consequence, not all DVAs can be
* translated to an equivalent new DVA. Instead we must provide a
* "vdev_remap" operation that executes a callback on each contiguous
* segment of the new location. This function is used in multiple ways:
*
* - i/os to this vdev use the callback to determine where the
* data is now located, and issue child i/os for each segment's new
* location.
*
* - frees and claims to this vdev use the callback to free or claim
* each mapped segment. (Note that we don't actually need to claim
* log blocks on indirect vdevs, because we don't allocate to
* removing vdevs. However, zdb uses zio_claim() for its leak
* detection.)
*/
/*
* "Big theory statement" for how we mark blocks obsolete.
*
* When a block on an indirect vdev is freed or remapped, a section of
* that vdev's mapping may no longer be referenced (aka "obsolete"). We
* keep track of how much of each mapping entry is obsolete. When
* an entry becomes completely obsolete, we can remove it, thus reducing
* the memory used by the mapping. The complete picture of obsolescence
* is given by the following data structures, described below:
* - the entry-specific obsolete count
* - the vdev-specific obsolete spacemap
* - the pool-specific obsolete bpobj
*
* == On disk data structures used ==
*
* We track the obsolete space for the pool using several objects. Each
* of these objects is created on demand and freed when no longer
* needed, and is assumed to be empty if it does not exist.
* SPA_FEATURE_OBSOLETE_COUNTS includes the count of these objects.
*
* - Each vic_mapping_object (associated with an indirect vdev) can
* have a vimp_counts_object. This is an array of uint32_t's
* with the same number of entries as the vic_mapping_object. When
* the mapping is condensed, entries from the vic_obsolete_sm_object
* (see below) are folded into the counts. Therefore, each
* obsolete_counts entry tells us the number of bytes in the
* corresponding mapping entry that were not referenced when the
* mapping was last condensed.
*
* - Each indirect or removing vdev can have a vic_obsolete_sm_object.
* This is a space map containing an alloc entry for every DVA that
* has been obsoleted since the last time this indirect vdev was
* condensed. We use this object in order to improve performance
* when marking a DVA as obsolete. Instead of modifying an arbitrary
* offset of the vimp_counts_object, we only need to append an entry
* to the end of this object. When a DVA becomes obsolete, it is
* added to the obsolete space map. This happens when the DVA is
* freed, remapped and not referenced by a snapshot, or the last
* snapshot referencing it is destroyed.
*
* - Each dataset can have a ds_remap_deadlist object. This is a
* deadlist object containing all blocks that were remapped in this
* dataset but referenced in a previous snapshot. Blocks can *only*
* appear on this list if they were remapped (dsl_dataset_block_remapped);
* blocks that were killed in a head dataset are put on the normal
* ds_deadlist and marked obsolete when they are freed.
*
* - The pool can have a dp_obsolete_bpobj. This is a list of blocks
* in the pool that need to be marked obsolete. When a snapshot is
* destroyed, we move some of the ds_remap_deadlist to the obsolete
* bpobj (see dsl_destroy_snapshot_handle_remaps()). We then
* asynchronously process the obsolete bpobj, moving its entries to
* the specific vdevs' obsolete space maps.
*
* == Summary of how we mark blocks as obsolete ==
*
* - When freeing a block: if any DVA is on an indirect vdev, append to
* vic_obsolete_sm_object.
* - When remapping a block, add dva to ds_remap_deadlist (if prev snap
* references; otherwise append to vic_obsolete_sm_object).
* - When freeing a snapshot: move parts of ds_remap_deadlist to
* dp_obsolete_bpobj (same algorithm as ds_deadlist).
* - When syncing the spa: process dp_obsolete_bpobj, moving ranges to
* individual vdev's vic_obsolete_sm_object.
*/
/*
* "Big theory statement" for how we condense indirect vdevs.
*
* Condensing an indirect vdev's mapping is the process of determining
* the precise counts of obsolete space for each mapping entry (by
* integrating the obsolete spacemap into the obsolete counts) and
* writing out a new mapping that contains only referenced entries.
*
* We condense a vdev when we expect the mapping to shrink (see
* vdev_indirect_should_condense()), but only perform one condense at a
* time to limit the memory usage. In addition, we use a separate
* open-context thread (spa_condense_indirect_thread) to incrementally
* create the new mapping object in a way that minimizes the impact on
* the rest of the system.
*
* == Generating a new mapping ==
*
* To generate a new mapping, we follow these steps:
*
* 1. Save the old obsolete space map and create a new mapping object
* (see spa_condense_indirect_start_sync()). This initializes the
* spa_condensing_indirect_phys with the "previous obsolete space map",
* which is now read only. Newly obsolete DVAs will be added to a
* new (initially empty) obsolete space map, and will not be
* considered as part of this condense operation.
*
* 2. Construct in memory the precise counts of obsolete space for each
* mapping entry, by incorporating the obsolete space map into the
* counts. (See vdev_indirect_mapping_load_obsolete_{counts,spacemap}().)
*
* 3. Iterate through each mapping entry, writing to the new mapping any
* entries that are not completely obsolete (i.e. which don't have
* obsolete count == mapping length). (See
* spa_condense_indirect_generate_new_mapping().)
*
* 4. Destroy the old mapping object and switch over to the new one
* (spa_condense_indirect_complete_sync).
*
* == Restarting from failure ==
*
* To restart the condense when we import/open the pool, we must start
* at the 2nd step above: reconstruct the precise counts in memory,
* based on the space map + counts. Then in the 3rd step, we start
* iterating where we left off: at vimp_max_offset of the new mapping
* object.
*/
-int zfs_condense_indirect_vdevs_enable = B_TRUE;
+static int zfs_condense_indirect_vdevs_enable = B_TRUE;
/*
* Condense if at least this percent of the bytes in the mapping is
* obsolete. With the default of 25%, the amount of space mapped
* will be reduced to 1% of its original size after at most 16
* condenses. Higher values will condense less often (causing less
* i/o); lower values will reduce the mapping size more quickly.
*/
-int zfs_condense_indirect_obsolete_pct = 25;
+static int zfs_condense_indirect_obsolete_pct = 25;
/*
* Condense if the obsolete space map takes up more than this amount of
* space on disk (logically). This limits the amount of disk space
* consumed by the obsolete space map; the default of 1GB is small enough
* that we typically don't mind "wasting" it.
*/
-unsigned long zfs_condense_max_obsolete_bytes = 1024 * 1024 * 1024;
+static unsigned long zfs_condense_max_obsolete_bytes = 1024 * 1024 * 1024;
/*
* Don't bother condensing if the mapping uses less than this amount of
* memory. The default of 128KB is considered a "trivial" amount of
* memory and not worth reducing.
*/
-unsigned long zfs_condense_min_mapping_bytes = 128 * 1024;
+static unsigned long zfs_condense_min_mapping_bytes = 128 * 1024;
/*
* This is used by the test suite so that it can ensure that certain
* actions happen while in the middle of a condense (which might otherwise
* complete too quickly). If used to reduce the performance impact of
* condensing in production, a maximum value of 1 should be sufficient.
*/
-int zfs_condense_indirect_commit_entry_delay_ms = 0;
+static int zfs_condense_indirect_commit_entry_delay_ms = 0;
/*
* If an indirect split block contains more than this many possible unique
* combinations when being reconstructed, consider it too computationally
* expensive to check them all. Instead, try at most 100 randomly-selected
* combinations each time the block is accessed. This allows all segment
* copies to participate fairly in the reconstruction when all combinations
* cannot be checked and prevents repeated use of one bad copy.
*/
int zfs_reconstruct_indirect_combinations_max = 4096;
/*
* Enable to simulate damaged segments and validate reconstruction. This
* is intentionally not exposed as a module parameter.
*/
unsigned long zfs_reconstruct_indirect_damage_fraction = 0;
/*
* The indirect_child_t represents the vdev that we will read from, when we
* need to read all copies of the data (e.g. for scrub or reconstruction).
* For plain (non-mirror) top-level vdevs (i.e. is_vdev is not a mirror),
* ic_vdev is the same as is_vdev. However, for mirror top-level vdevs,
* ic_vdev is a child of the mirror.
*/
typedef struct indirect_child {
abd_t *ic_data;
vdev_t *ic_vdev;
/*
* ic_duplicate is NULL when the ic_data contents are unique, when it
* is determined to be a duplicate it references the primary child.
*/
struct indirect_child *ic_duplicate;
list_node_t ic_node; /* node on is_unique_child */
int ic_error; /* set when a child does not contain the data */
} indirect_child_t;
/*
* The indirect_split_t represents one mapped segment of an i/o to the
* indirect vdev. For non-split (contiguously-mapped) blocks, there will be
* only one indirect_split_t, with is_split_offset==0 and is_size==io_size.
* For split blocks, there will be several of these.
*/
typedef struct indirect_split {
list_node_t is_node; /* link on iv_splits */
/*
* is_split_offset is the offset into the i/o.
* This is the sum of the previous splits' is_size's.
*/
uint64_t is_split_offset;
vdev_t *is_vdev; /* top-level vdev */
uint64_t is_target_offset; /* offset on is_vdev */
uint64_t is_size;
int is_children; /* number of entries in is_child[] */
int is_unique_children; /* number of entries in is_unique_child */
list_t is_unique_child;
/*
* is_good_child is the child that we are currently using to
* attempt reconstruction.
*/
indirect_child_t *is_good_child;
indirect_child_t is_child[1]; /* variable-length */
} indirect_split_t;
/*
* The indirect_vsd_t is associated with each i/o to the indirect vdev.
* It is the "Vdev-Specific Data" in the zio_t's io_vsd.
*/
typedef struct indirect_vsd {
boolean_t iv_split_block;
boolean_t iv_reconstruct;
uint64_t iv_unique_combinations;
uint64_t iv_attempts;
uint64_t iv_attempts_max;
list_t iv_splits; /* list of indirect_split_t's */
} indirect_vsd_t;
static void
vdev_indirect_map_free(zio_t *zio)
{
indirect_vsd_t *iv = zio->io_vsd;
indirect_split_t *is;
while ((is = list_head(&iv->iv_splits)) != NULL) {
for (int c = 0; c < is->is_children; c++) {
indirect_child_t *ic = &is->is_child[c];
if (ic->ic_data != NULL)
abd_free(ic->ic_data);
}
list_remove(&iv->iv_splits, is);
indirect_child_t *ic;
while ((ic = list_head(&is->is_unique_child)) != NULL)
list_remove(&is->is_unique_child, ic);
list_destroy(&is->is_unique_child);
kmem_free(is,
offsetof(indirect_split_t, is_child[is->is_children]));
}
kmem_free(iv, sizeof (*iv));
}
static const zio_vsd_ops_t vdev_indirect_vsd_ops = {
.vsd_free = vdev_indirect_map_free,
};
/*
* Mark the given offset and size as being obsolete.
*/
void
vdev_indirect_mark_obsolete(vdev_t *vd, uint64_t offset, uint64_t size)
{
spa_t *spa = vd->vdev_spa;
ASSERT3U(vd->vdev_indirect_config.vic_mapping_object, !=, 0);
ASSERT(vd->vdev_removing || vd->vdev_ops == &vdev_indirect_ops);
ASSERT(size > 0);
VERIFY(vdev_indirect_mapping_entry_for_offset(
vd->vdev_indirect_mapping, offset) != NULL);
if (spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS)) {
mutex_enter(&vd->vdev_obsolete_lock);
range_tree_add(vd->vdev_obsolete_segments, offset, size);
mutex_exit(&vd->vdev_obsolete_lock);
vdev_dirty(vd, 0, NULL, spa_syncing_txg(spa));
}
}
/*
* Mark the DVA vdev_id:offset:size as being obsolete in the given tx. This
* wrapper is provided because the DMU does not know about vdev_t's and
* cannot directly call vdev_indirect_mark_obsolete.
*/
void
spa_vdev_indirect_mark_obsolete(spa_t *spa, uint64_t vdev_id, uint64_t offset,
uint64_t size, dmu_tx_t *tx)
{
vdev_t *vd = vdev_lookup_top(spa, vdev_id);
ASSERT(dmu_tx_is_syncing(tx));
/* The DMU can only remap indirect vdevs. */
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
vdev_indirect_mark_obsolete(vd, offset, size);
}
static spa_condensing_indirect_t *
spa_condensing_indirect_create(spa_t *spa)
{
spa_condensing_indirect_phys_t *scip =
&spa->spa_condensing_indirect_phys;
spa_condensing_indirect_t *sci = kmem_zalloc(sizeof (*sci), KM_SLEEP);
objset_t *mos = spa->spa_meta_objset;
for (int i = 0; i < TXG_SIZE; i++) {
list_create(&sci->sci_new_mapping_entries[i],
sizeof (vdev_indirect_mapping_entry_t),
offsetof(vdev_indirect_mapping_entry_t, vime_node));
}
sci->sci_new_mapping =
vdev_indirect_mapping_open(mos, scip->scip_next_mapping_object);
return (sci);
}
static void
spa_condensing_indirect_destroy(spa_condensing_indirect_t *sci)
{
for (int i = 0; i < TXG_SIZE; i++)
list_destroy(&sci->sci_new_mapping_entries[i]);
if (sci->sci_new_mapping != NULL)
vdev_indirect_mapping_close(sci->sci_new_mapping);
kmem_free(sci, sizeof (*sci));
}
boolean_t
vdev_indirect_should_condense(vdev_t *vd)
{
vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
spa_t *spa = vd->vdev_spa;
ASSERT(dsl_pool_sync_context(spa->spa_dsl_pool));
if (!zfs_condense_indirect_vdevs_enable)
return (B_FALSE);
/*
* We can only condense one indirect vdev at a time.
*/
if (spa->spa_condensing_indirect != NULL)
return (B_FALSE);
if (spa_shutting_down(spa))
return (B_FALSE);
/*
* The mapping object size must not change while we are
* condensing, so we can only condense indirect vdevs
* (not vdevs that are still in the middle of being removed).
*/
if (vd->vdev_ops != &vdev_indirect_ops)
return (B_FALSE);
/*
* If nothing new has been marked obsolete, there is no
* point in condensing.
*/
uint64_t obsolete_sm_obj __maybe_unused;
ASSERT0(vdev_obsolete_sm_object(vd, &obsolete_sm_obj));
if (vd->vdev_obsolete_sm == NULL) {
ASSERT0(obsolete_sm_obj);
return (B_FALSE);
}
ASSERT(vd->vdev_obsolete_sm != NULL);
ASSERT3U(obsolete_sm_obj, ==, space_map_object(vd->vdev_obsolete_sm));
uint64_t bytes_mapped = vdev_indirect_mapping_bytes_mapped(vim);
uint64_t bytes_obsolete = space_map_allocated(vd->vdev_obsolete_sm);
uint64_t mapping_size = vdev_indirect_mapping_size(vim);
uint64_t obsolete_sm_size = space_map_length(vd->vdev_obsolete_sm);
ASSERT3U(bytes_obsolete, <=, bytes_mapped);
/*
* If a high percentage of the bytes that are mapped have become
* obsolete, condense (unless the mapping is already small enough).
* This has a good chance of reducing the amount of memory used
* by the mapping.
*/
if (bytes_obsolete * 100 / bytes_mapped >=
zfs_condense_indirect_obsolete_pct &&
mapping_size > zfs_condense_min_mapping_bytes) {
zfs_dbgmsg("should condense vdev %llu because obsolete "
"spacemap covers %d%% of %lluMB mapping",
(u_longlong_t)vd->vdev_id,
(int)(bytes_obsolete * 100 / bytes_mapped),
(u_longlong_t)bytes_mapped / 1024 / 1024);
return (B_TRUE);
}
/*
* If the obsolete space map takes up too much space on disk,
* condense in order to free up this disk space.
*/
if (obsolete_sm_size >= zfs_condense_max_obsolete_bytes) {
zfs_dbgmsg("should condense vdev %llu because obsolete sm "
"length %lluMB >= max size %lluMB",
(u_longlong_t)vd->vdev_id,
(u_longlong_t)obsolete_sm_size / 1024 / 1024,
(u_longlong_t)zfs_condense_max_obsolete_bytes /
1024 / 1024);
return (B_TRUE);
}
return (B_FALSE);
}
/*
* This sync task completes (finishes) a condense, deleting the old
* mapping and replacing it with the new one.
*/
static void
spa_condense_indirect_complete_sync(void *arg, dmu_tx_t *tx)
{
spa_condensing_indirect_t *sci = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
spa_condensing_indirect_phys_t *scip =
&spa->spa_condensing_indirect_phys;
vdev_t *vd = vdev_lookup_top(spa, scip->scip_vdev);
vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
objset_t *mos = spa->spa_meta_objset;
vdev_indirect_mapping_t *old_mapping = vd->vdev_indirect_mapping;
uint64_t old_count = vdev_indirect_mapping_num_entries(old_mapping);
uint64_t new_count =
vdev_indirect_mapping_num_entries(sci->sci_new_mapping);
ASSERT(dmu_tx_is_syncing(tx));
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
ASSERT3P(sci, ==, spa->spa_condensing_indirect);
for (int i = 0; i < TXG_SIZE; i++) {
ASSERT(list_is_empty(&sci->sci_new_mapping_entries[i]));
}
ASSERT(vic->vic_mapping_object != 0);
ASSERT3U(vd->vdev_id, ==, scip->scip_vdev);
ASSERT(scip->scip_next_mapping_object != 0);
ASSERT(scip->scip_prev_obsolete_sm_object != 0);
/*
* Reset vdev_indirect_mapping to refer to the new object.
*/
rw_enter(&vd->vdev_indirect_rwlock, RW_WRITER);
vdev_indirect_mapping_close(vd->vdev_indirect_mapping);
vd->vdev_indirect_mapping = sci->sci_new_mapping;
rw_exit(&vd->vdev_indirect_rwlock);
sci->sci_new_mapping = NULL;
vdev_indirect_mapping_free(mos, vic->vic_mapping_object, tx);
vic->vic_mapping_object = scip->scip_next_mapping_object;
scip->scip_next_mapping_object = 0;
space_map_free_obj(mos, scip->scip_prev_obsolete_sm_object, tx);
spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
scip->scip_prev_obsolete_sm_object = 0;
scip->scip_vdev = 0;
VERIFY0(zap_remove(mos, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_CONDENSING_INDIRECT, tx));
spa_condensing_indirect_destroy(spa->spa_condensing_indirect);
spa->spa_condensing_indirect = NULL;
zfs_dbgmsg("finished condense of vdev %llu in txg %llu: "
"new mapping object %llu has %llu entries "
"(was %llu entries)",
(u_longlong_t)vd->vdev_id, (u_longlong_t)dmu_tx_get_txg(tx),
(u_longlong_t)vic->vic_mapping_object,
(u_longlong_t)new_count, (u_longlong_t)old_count);
vdev_config_dirty(spa->spa_root_vdev);
}
/*
* This sync task appends entries to the new mapping object.
*/
static void
spa_condense_indirect_commit_sync(void *arg, dmu_tx_t *tx)
{
spa_condensing_indirect_t *sci = arg;
uint64_t txg = dmu_tx_get_txg(tx);
spa_t *spa __maybe_unused = dmu_tx_pool(tx)->dp_spa;
ASSERT(dmu_tx_is_syncing(tx));
ASSERT3P(sci, ==, spa->spa_condensing_indirect);
vdev_indirect_mapping_add_entries(sci->sci_new_mapping,
&sci->sci_new_mapping_entries[txg & TXG_MASK], tx);
ASSERT(list_is_empty(&sci->sci_new_mapping_entries[txg & TXG_MASK]));
}
/*
* Open-context function to add one entry to the new mapping. The new
* entry will be remembered and written from syncing context.
*/
static void
spa_condense_indirect_commit_entry(spa_t *spa,
vdev_indirect_mapping_entry_phys_t *vimep, uint32_t count)
{
spa_condensing_indirect_t *sci = spa->spa_condensing_indirect;
ASSERT3U(count, <, DVA_GET_ASIZE(&vimep->vimep_dst));
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
dmu_tx_hold_space(tx, sizeof (*vimep) + sizeof (count));
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;
/*
* If we are the first entry committed this txg, kick off the sync
* task to write to the MOS on our behalf.
*/
if (list_is_empty(&sci->sci_new_mapping_entries[txgoff])) {
dsl_sync_task_nowait(dmu_tx_pool(tx),
spa_condense_indirect_commit_sync, sci, tx);
}
vdev_indirect_mapping_entry_t *vime =
kmem_alloc(sizeof (*vime), KM_SLEEP);
vime->vime_mapping = *vimep;
vime->vime_obsolete_count = count;
list_insert_tail(&sci->sci_new_mapping_entries[txgoff], vime);
dmu_tx_commit(tx);
}
static void
spa_condense_indirect_generate_new_mapping(vdev_t *vd,
uint32_t *obsolete_counts, uint64_t start_index, zthr_t *zthr)
{
spa_t *spa = vd->vdev_spa;
uint64_t mapi = start_index;
vdev_indirect_mapping_t *old_mapping = vd->vdev_indirect_mapping;
uint64_t old_num_entries =
vdev_indirect_mapping_num_entries(old_mapping);
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
ASSERT3U(vd->vdev_id, ==, spa->spa_condensing_indirect_phys.scip_vdev);
zfs_dbgmsg("starting condense of vdev %llu from index %llu",
(u_longlong_t)vd->vdev_id,
(u_longlong_t)mapi);
while (mapi < old_num_entries) {
if (zthr_iscancelled(zthr)) {
zfs_dbgmsg("pausing condense of vdev %llu "
"at index %llu", (u_longlong_t)vd->vdev_id,
(u_longlong_t)mapi);
break;
}
vdev_indirect_mapping_entry_phys_t *entry =
&old_mapping->vim_entries[mapi];
uint64_t entry_size = DVA_GET_ASIZE(&entry->vimep_dst);
ASSERT3U(obsolete_counts[mapi], <=, entry_size);
if (obsolete_counts[mapi] < entry_size) {
spa_condense_indirect_commit_entry(spa, entry,
obsolete_counts[mapi]);
/*
* This delay may be requested for testing, debugging,
* or performance reasons.
*/
hrtime_t now = gethrtime();
hrtime_t sleep_until = now + MSEC2NSEC(
zfs_condense_indirect_commit_entry_delay_ms);
zfs_sleep_until(sleep_until);
}
mapi++;
}
}
static boolean_t
spa_condense_indirect_thread_check(void *arg, zthr_t *zthr)
{
(void) zthr;
spa_t *spa = arg;
return (spa->spa_condensing_indirect != NULL);
}
static void
spa_condense_indirect_thread(void *arg, zthr_t *zthr)
{
spa_t *spa = arg;
vdev_t *vd;
ASSERT3P(spa->spa_condensing_indirect, !=, NULL);
spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
vd = vdev_lookup_top(spa, spa->spa_condensing_indirect_phys.scip_vdev);
ASSERT3P(vd, !=, NULL);
spa_config_exit(spa, SCL_VDEV, FTAG);
spa_condensing_indirect_t *sci = spa->spa_condensing_indirect;
spa_condensing_indirect_phys_t *scip =
&spa->spa_condensing_indirect_phys;
uint32_t *counts;
uint64_t start_index;
vdev_indirect_mapping_t *old_mapping = vd->vdev_indirect_mapping;
space_map_t *prev_obsolete_sm = NULL;
ASSERT3U(vd->vdev_id, ==, scip->scip_vdev);
ASSERT(scip->scip_next_mapping_object != 0);
ASSERT(scip->scip_prev_obsolete_sm_object != 0);
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
for (int i = 0; i < TXG_SIZE; i++) {
/*
* The list must start out empty in order for the
* _commit_sync() sync task to be properly registered
* on the first call to _commit_entry(); so it's wise
* to double check and ensure we actually are starting
* with empty lists.
*/
ASSERT(list_is_empty(&sci->sci_new_mapping_entries[i]));
}
VERIFY0(space_map_open(&prev_obsolete_sm, spa->spa_meta_objset,
scip->scip_prev_obsolete_sm_object, 0, vd->vdev_asize, 0));
counts = vdev_indirect_mapping_load_obsolete_counts(old_mapping);
if (prev_obsolete_sm != NULL) {
vdev_indirect_mapping_load_obsolete_spacemap(old_mapping,
counts, prev_obsolete_sm);
}
space_map_close(prev_obsolete_sm);
/*
* Generate new mapping. Determine what index to continue from
* based on the max offset that we've already written in the
* new mapping.
*/
uint64_t max_offset =
vdev_indirect_mapping_max_offset(sci->sci_new_mapping);
if (max_offset == 0) {
/* We haven't written anything to the new mapping yet. */
start_index = 0;
} else {
/*
* Pick up from where we left off. _entry_for_offset()
* returns a pointer into the vim_entries array. If
* max_offset is greater than any of the mappings
* contained in the table NULL will be returned and
* that indicates we've exhausted our iteration of the
* old_mapping.
*/
vdev_indirect_mapping_entry_phys_t *entry =
vdev_indirect_mapping_entry_for_offset_or_next(old_mapping,
max_offset);
if (entry == NULL) {
/*
* We've already written the whole new mapping.
* This special value will cause us to skip the
* generate_new_mapping step and just do the sync
* task to complete the condense.
*/
start_index = UINT64_MAX;
} else {
start_index = entry - old_mapping->vim_entries;
ASSERT3U(start_index, <,
vdev_indirect_mapping_num_entries(old_mapping));
}
}
spa_condense_indirect_generate_new_mapping(vd, counts,
start_index, zthr);
vdev_indirect_mapping_free_obsolete_counts(old_mapping, counts);
/*
* If the zthr has received a cancellation signal while running
* in generate_new_mapping() or at any point after that, then bail
* early. We don't want to complete the condense if the spa is
* shutting down.
*/
if (zthr_iscancelled(zthr))
return;
VERIFY0(dsl_sync_task(spa_name(spa), NULL,
spa_condense_indirect_complete_sync, sci, 0,
ZFS_SPACE_CHECK_EXTRA_RESERVED));
}
/*
* Sync task to begin the condensing process.
*/
void
spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx)
{
spa_t *spa = vd->vdev_spa;
spa_condensing_indirect_phys_t *scip =
&spa->spa_condensing_indirect_phys;
ASSERT0(scip->scip_next_mapping_object);
ASSERT0(scip->scip_prev_obsolete_sm_object);
ASSERT0(scip->scip_vdev);
ASSERT(dmu_tx_is_syncing(tx));
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_OBSOLETE_COUNTS));
ASSERT(vdev_indirect_mapping_num_entries(vd->vdev_indirect_mapping));
uint64_t obsolete_sm_obj;
VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_obj));
ASSERT3U(obsolete_sm_obj, !=, 0);
scip->scip_vdev = vd->vdev_id;
scip->scip_next_mapping_object =
vdev_indirect_mapping_alloc(spa->spa_meta_objset, tx);
scip->scip_prev_obsolete_sm_object = obsolete_sm_obj;
/*
* We don't need to allocate a new space map object, since
* vdev_indirect_sync_obsolete will allocate one when needed.
*/
space_map_close(vd->vdev_obsolete_sm);
vd->vdev_obsolete_sm = NULL;
VERIFY0(zap_remove(spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM, tx));
VERIFY0(zap_add(spa->spa_dsl_pool->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_CONDENSING_INDIRECT, sizeof (uint64_t),
sizeof (*scip) / sizeof (uint64_t), scip, tx));
ASSERT3P(spa->spa_condensing_indirect, ==, NULL);
spa->spa_condensing_indirect = spa_condensing_indirect_create(spa);
zfs_dbgmsg("starting condense of vdev %llu in txg %llu: "
"posm=%llu nm=%llu",
(u_longlong_t)vd->vdev_id, (u_longlong_t)dmu_tx_get_txg(tx),
(u_longlong_t)scip->scip_prev_obsolete_sm_object,
(u_longlong_t)scip->scip_next_mapping_object);
zthr_wakeup(spa->spa_condense_zthr);
}
/*
* Sync to the given vdev's obsolete space map any segments that are no longer
* referenced as of the given txg.
*
* If the obsolete space map doesn't exist yet, create and open it.
*/
void
vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx)
{
spa_t *spa = vd->vdev_spa;
vdev_indirect_config_t *vic __maybe_unused = &vd->vdev_indirect_config;
ASSERT3U(vic->vic_mapping_object, !=, 0);
ASSERT(range_tree_space(vd->vdev_obsolete_segments) > 0);
ASSERT(vd->vdev_removing || vd->vdev_ops == &vdev_indirect_ops);
ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS));
uint64_t obsolete_sm_object;
VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
if (obsolete_sm_object == 0) {
obsolete_sm_object = space_map_alloc(spa->spa_meta_objset,
zfs_vdev_standard_sm_blksz, tx);
ASSERT(vd->vdev_top_zap != 0);
VERIFY0(zap_add(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM,
sizeof (obsolete_sm_object), 1, &obsolete_sm_object, tx));
ASSERT0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
ASSERT3U(obsolete_sm_object, !=, 0);
spa_feature_incr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
VERIFY0(space_map_open(&vd->vdev_obsolete_sm,
spa->spa_meta_objset, obsolete_sm_object,
0, vd->vdev_asize, 0));
}
ASSERT(vd->vdev_obsolete_sm != NULL);
ASSERT3U(obsolete_sm_object, ==,
space_map_object(vd->vdev_obsolete_sm));
space_map_write(vd->vdev_obsolete_sm,
vd->vdev_obsolete_segments, SM_ALLOC, SM_NO_VDEVID, tx);
range_tree_vacate(vd->vdev_obsolete_segments, NULL, NULL);
}
int
spa_condense_init(spa_t *spa)
{
int error = zap_lookup(spa->spa_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_CONDENSING_INDIRECT, sizeof (uint64_t),
sizeof (spa->spa_condensing_indirect_phys) / sizeof (uint64_t),
&spa->spa_condensing_indirect_phys);
if (error == 0) {
if (spa_writeable(spa)) {
spa->spa_condensing_indirect =
spa_condensing_indirect_create(spa);
}
return (0);
} else if (error == ENOENT) {
return (0);
} else {
return (error);
}
}
void
spa_condense_fini(spa_t *spa)
{
if (spa->spa_condensing_indirect != NULL) {
spa_condensing_indirect_destroy(spa->spa_condensing_indirect);
spa->spa_condensing_indirect = NULL;
}
}
void
spa_start_indirect_condensing_thread(spa_t *spa)
{
ASSERT3P(spa->spa_condense_zthr, ==, NULL);
spa->spa_condense_zthr = zthr_create("z_indirect_condense",
spa_condense_indirect_thread_check,
spa_condense_indirect_thread, spa, minclsyspri);
}
/*
* Gets the obsolete spacemap object from the vdev's ZAP. On success sm_obj
* will contain either the obsolete spacemap object or zero if none exists.
* All other errors are returned to the caller.
*/
int
vdev_obsolete_sm_object(vdev_t *vd, uint64_t *sm_obj)
{
ASSERT0(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));
if (vd->vdev_top_zap == 0) {
*sm_obj = 0;
return (0);
}
int error = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM, sizeof (uint64_t), 1, sm_obj);
if (error == ENOENT) {
*sm_obj = 0;
error = 0;
}
return (error);
}
/*
* Gets the obsolete count are precise spacemap object from the vdev's ZAP.
* On success are_precise will be set to reflect if the counts are precise.
* All other errors are returned to the caller.
*/
int
vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise)
{
ASSERT0(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));
if (vd->vdev_top_zap == 0) {
*are_precise = B_FALSE;
return (0);
}
uint64_t val = 0;
int error = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_OBSOLETE_COUNTS_ARE_PRECISE, sizeof (val), 1, &val);
if (error == 0) {
*are_precise = (val != 0);
} else if (error == ENOENT) {
*are_precise = B_FALSE;
error = 0;
}
return (error);
}
static void
vdev_indirect_close(vdev_t *vd)
{
(void) vd;
}
static int
vdev_indirect_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
uint64_t *logical_ashift, uint64_t *physical_ashift)
{
*psize = *max_psize = vd->vdev_asize +
VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE;
*logical_ashift = vd->vdev_ashift;
*physical_ashift = vd->vdev_physical_ashift;
return (0);
}
typedef struct remap_segment {
vdev_t *rs_vd;
uint64_t rs_offset;
uint64_t rs_asize;
uint64_t rs_split_offset;
list_node_t rs_node;
} remap_segment_t;
static remap_segment_t *
rs_alloc(vdev_t *vd, uint64_t offset, uint64_t asize, uint64_t split_offset)
{
remap_segment_t *rs = kmem_alloc(sizeof (remap_segment_t), KM_SLEEP);
rs->rs_vd = vd;
rs->rs_offset = offset;
rs->rs_asize = asize;
rs->rs_split_offset = split_offset;
return (rs);
}
/*
* Given an indirect vdev and an extent on that vdev, it duplicates the
* physical entries of the indirect mapping that correspond to the extent
* to a new array and returns a pointer to it. In addition, copied_entries
* is populated with the number of mapping entries that were duplicated.
*
* Note that the function assumes that the caller holds vdev_indirect_rwlock.
* This ensures that the mapping won't change due to condensing as we
* copy over its contents.
*
* Finally, since we are doing an allocation, it is up to the caller to
* free the array allocated in this function.
*/
static vdev_indirect_mapping_entry_phys_t *
vdev_indirect_mapping_duplicate_adjacent_entries(vdev_t *vd, uint64_t offset,
uint64_t asize, uint64_t *copied_entries)
{
vdev_indirect_mapping_entry_phys_t *duplicate_mappings = NULL;
vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
uint64_t entries = 0;
ASSERT(RW_READ_HELD(&vd->vdev_indirect_rwlock));
vdev_indirect_mapping_entry_phys_t *first_mapping =
vdev_indirect_mapping_entry_for_offset(vim, offset);
ASSERT3P(first_mapping, !=, NULL);
vdev_indirect_mapping_entry_phys_t *m = first_mapping;
while (asize > 0) {
uint64_t size = DVA_GET_ASIZE(&m->vimep_dst);
ASSERT3U(offset, >=, DVA_MAPPING_GET_SRC_OFFSET(m));
ASSERT3U(offset, <, DVA_MAPPING_GET_SRC_OFFSET(m) + size);
uint64_t inner_offset = offset - DVA_MAPPING_GET_SRC_OFFSET(m);
uint64_t inner_size = MIN(asize, size - inner_offset);
offset += inner_size;
asize -= inner_size;
entries++;
m++;
}
size_t copy_length = entries * sizeof (*first_mapping);
duplicate_mappings = kmem_alloc(copy_length, KM_SLEEP);
bcopy(first_mapping, duplicate_mappings, copy_length);
*copied_entries = entries;
return (duplicate_mappings);
}
/*
* Goes through the relevant indirect mappings until it hits a concrete vdev
* and issues the callback. On the way to the concrete vdev, if any other
* indirect vdevs are encountered, then the callback will also be called on
* each of those indirect vdevs. For example, if the segment is mapped to
* segment A on indirect vdev 1, and then segment A on indirect vdev 1 is
* mapped to segment B on concrete vdev 2, then the callback will be called on
* both vdev 1 and vdev 2.
*
* While the callback passed to vdev_indirect_remap() is called on every vdev
* the function encounters, certain callbacks only care about concrete vdevs.
* These types of callbacks should return immediately and explicitly when they
* are called on an indirect vdev.
*
* Because there is a possibility that a DVA section in the indirect device
* has been split into multiple sections in our mapping, we keep track
* of the relevant contiguous segments of the new location (remap_segment_t)
* in a stack. This way we can call the callback for each of the new sections
* created by a single section of the indirect device. Note though, that in
* this scenario the callbacks in each split block won't occur in-order in
* terms of offset, so callers should not make any assumptions about that.
*
* For callbacks that don't handle split blocks and immediately return when
* they encounter them (as is the case for remap_blkptr_cb), the caller can
* assume that its callback will be applied from the first indirect vdev
* encountered to the last one and then the concrete vdev, in that order.
*/
static void
vdev_indirect_remap(vdev_t *vd, uint64_t offset, uint64_t asize,
void (*func)(uint64_t, vdev_t *, uint64_t, uint64_t, void *), void *arg)
{
list_t stack;
spa_t *spa = vd->vdev_spa;
list_create(&stack, sizeof (remap_segment_t),
offsetof(remap_segment_t, rs_node));
for (remap_segment_t *rs = rs_alloc(vd, offset, asize, 0);
rs != NULL; rs = list_remove_head(&stack)) {
vdev_t *v = rs->rs_vd;
uint64_t num_entries = 0;
ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
ASSERT(rs->rs_asize > 0);
/*
* Note: As this function can be called from open context
* (e.g. zio_read()), we need the following rwlock to
* prevent the mapping from being changed by condensing.
*
* So we grab the lock and we make a copy of the entries
* that are relevant to the extent that we are working on.
* Once that is done, we drop the lock and iterate over
* our copy of the mapping. Once we are done with the with
* the remap segment and we free it, we also free our copy
* of the indirect mapping entries that are relevant to it.
*
* This way we don't need to wait until the function is
* finished with a segment, to condense it. In addition, we
* don't need a recursive rwlock for the case that a call to
* vdev_indirect_remap() needs to call itself (through the
* codepath of its callback) for the same vdev in the middle
* of its execution.
*/
rw_enter(&v->vdev_indirect_rwlock, RW_READER);
ASSERT3P(v->vdev_indirect_mapping, !=, NULL);
vdev_indirect_mapping_entry_phys_t *mapping =
vdev_indirect_mapping_duplicate_adjacent_entries(v,
rs->rs_offset, rs->rs_asize, &num_entries);
ASSERT3P(mapping, !=, NULL);
ASSERT3U(num_entries, >, 0);
rw_exit(&v->vdev_indirect_rwlock);
for (uint64_t i = 0; i < num_entries; i++) {
/*
* Note: the vdev_indirect_mapping can not change
* while we are running. It only changes while the
* removal is in progress, and then only from syncing
* context. While a removal is in progress, this
* function is only called for frees, which also only
* happen from syncing context.
*/
vdev_indirect_mapping_entry_phys_t *m = &mapping[i];
ASSERT3P(m, !=, NULL);
ASSERT3U(rs->rs_asize, >, 0);
uint64_t size = DVA_GET_ASIZE(&m->vimep_dst);
uint64_t dst_offset = DVA_GET_OFFSET(&m->vimep_dst);
uint64_t dst_vdev = DVA_GET_VDEV(&m->vimep_dst);
ASSERT3U(rs->rs_offset, >=,
DVA_MAPPING_GET_SRC_OFFSET(m));
ASSERT3U(rs->rs_offset, <,
DVA_MAPPING_GET_SRC_OFFSET(m) + size);
ASSERT3U(dst_vdev, !=, v->vdev_id);
uint64_t inner_offset = rs->rs_offset -
DVA_MAPPING_GET_SRC_OFFSET(m);
uint64_t inner_size =
MIN(rs->rs_asize, size - inner_offset);
vdev_t *dst_v = vdev_lookup_top(spa, dst_vdev);
ASSERT3P(dst_v, !=, NULL);
if (dst_v->vdev_ops == &vdev_indirect_ops) {
list_insert_head(&stack,
rs_alloc(dst_v, dst_offset + inner_offset,
inner_size, rs->rs_split_offset));
}
if ((zfs_flags & ZFS_DEBUG_INDIRECT_REMAP) &&
IS_P2ALIGNED(inner_size, 2 * SPA_MINBLOCKSIZE)) {
/*
* Note: This clause exists only solely for
* testing purposes. We use it to ensure that
* split blocks work and that the callbacks
* using them yield the same result if issued
* in reverse order.
*/
uint64_t inner_half = inner_size / 2;
func(rs->rs_split_offset + inner_half, dst_v,
dst_offset + inner_offset + inner_half,
inner_half, arg);
func(rs->rs_split_offset, dst_v,
dst_offset + inner_offset,
inner_half, arg);
} else {
func(rs->rs_split_offset, dst_v,
dst_offset + inner_offset,
inner_size, arg);
}
rs->rs_offset += inner_size;
rs->rs_asize -= inner_size;
rs->rs_split_offset += inner_size;
}
VERIFY0(rs->rs_asize);
kmem_free(mapping, num_entries * sizeof (*mapping));
kmem_free(rs, sizeof (remap_segment_t));
}
list_destroy(&stack);
}
static void
vdev_indirect_child_io_done(zio_t *zio)
{
zio_t *pio = zio->io_private;
mutex_enter(&pio->io_lock);
pio->io_error = zio_worst_error(pio->io_error, zio->io_error);
mutex_exit(&pio->io_lock);
abd_free(zio->io_abd);
}
/*
* This is a callback for vdev_indirect_remap() which allocates an
* indirect_split_t for each split segment and adds it to iv_splits.
*/
static void
vdev_indirect_gather_splits(uint64_t split_offset, vdev_t *vd, uint64_t offset,
uint64_t size, void *arg)
{
zio_t *zio = arg;
indirect_vsd_t *iv = zio->io_vsd;
ASSERT3P(vd, !=, NULL);
if (vd->vdev_ops == &vdev_indirect_ops)
return;
int n = 1;
if (vd->vdev_ops == &vdev_mirror_ops)
n = vd->vdev_children;
indirect_split_t *is =
kmem_zalloc(offsetof(indirect_split_t, is_child[n]), KM_SLEEP);
is->is_children = n;
is->is_size = size;
is->is_split_offset = split_offset;
is->is_target_offset = offset;
is->is_vdev = vd;
list_create(&is->is_unique_child, sizeof (indirect_child_t),
offsetof(indirect_child_t, ic_node));
/*
* Note that we only consider multiple copies of the data for
* *mirror* vdevs. We don't for "replacing" or "spare" vdevs, even
* though they use the same ops as mirror, because there's only one
* "good" copy under the replacing/spare.
*/
if (vd->vdev_ops == &vdev_mirror_ops) {
for (int i = 0; i < n; i++) {
is->is_child[i].ic_vdev = vd->vdev_child[i];
list_link_init(&is->is_child[i].ic_node);
}
} else {
is->is_child[0].ic_vdev = vd;
}
list_insert_tail(&iv->iv_splits, is);
}
static void
vdev_indirect_read_split_done(zio_t *zio)
{
indirect_child_t *ic = zio->io_private;
if (zio->io_error != 0) {
/*
* Clear ic_data to indicate that we do not have data for this
* child.
*/
abd_free(ic->ic_data);
ic->ic_data = NULL;
}
}
/*
* Issue reads for all copies (mirror children) of all splits.
*/
static void
vdev_indirect_read_all(zio_t *zio)
{
indirect_vsd_t *iv = zio->io_vsd;
ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ);
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
for (int i = 0; i < is->is_children; i++) {
indirect_child_t *ic = &is->is_child[i];
if (!vdev_readable(ic->ic_vdev))
continue;
/*
* If a child is missing the data, set ic_error. Used
* in vdev_indirect_repair(). We perform the read
* nevertheless which provides the opportunity to
* reconstruct the split block if at all possible.
*/
if (vdev_dtl_contains(ic->ic_vdev, DTL_MISSING,
zio->io_txg, 1))
ic->ic_error = SET_ERROR(ESTALE);
ic->ic_data = abd_alloc_sametype(zio->io_abd,
is->is_size);
ic->ic_duplicate = NULL;
zio_nowait(zio_vdev_child_io(zio, NULL,
ic->ic_vdev, is->is_target_offset, ic->ic_data,
is->is_size, zio->io_type, zio->io_priority, 0,
vdev_indirect_read_split_done, ic));
}
}
iv->iv_reconstruct = B_TRUE;
}
static void
vdev_indirect_io_start(zio_t *zio)
{
spa_t *spa __maybe_unused = zio->io_spa;
indirect_vsd_t *iv = kmem_zalloc(sizeof (*iv), KM_SLEEP);
list_create(&iv->iv_splits,
sizeof (indirect_split_t), offsetof(indirect_split_t, is_node));
zio->io_vsd = iv;
zio->io_vsd_ops = &vdev_indirect_vsd_ops;
ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
if (zio->io_type != ZIO_TYPE_READ) {
ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);
/*
* Note: this code can handle other kinds of writes,
* but we don't expect them.
*/
ASSERT((zio->io_flags & (ZIO_FLAG_SELF_HEAL |
ZIO_FLAG_RESILVER | ZIO_FLAG_INDUCE_DAMAGE)) != 0);
}
vdev_indirect_remap(zio->io_vd, zio->io_offset, zio->io_size,
vdev_indirect_gather_splits, zio);
indirect_split_t *first = list_head(&iv->iv_splits);
if (first->is_size == zio->io_size) {
/*
* This is not a split block; we are pointing to the entire
* data, which will checksum the same as the original data.
* Pass the BP down so that the child i/o can verify the
* checksum, and try a different location if available
* (e.g. on a mirror).
*
* While this special case could be handled the same as the
* general (split block) case, doing it this way ensures
* that the vast majority of blocks on indirect vdevs
* (which are not split) are handled identically to blocks
* on non-indirect vdevs. This allows us to be less strict
* about performance in the general (but rare) case.
*/
ASSERT0(first->is_split_offset);
ASSERT3P(list_next(&iv->iv_splits, first), ==, NULL);
zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
first->is_vdev, first->is_target_offset,
abd_get_offset(zio->io_abd, 0),
zio->io_size, zio->io_type, zio->io_priority, 0,
vdev_indirect_child_io_done, zio));
} else {
iv->iv_split_block = B_TRUE;
if (zio->io_type == ZIO_TYPE_READ &&
zio->io_flags & (ZIO_FLAG_SCRUB | ZIO_FLAG_RESILVER)) {
/*
* Read all copies. Note that for simplicity,
* we don't bother consulting the DTL in the
* resilver case.
*/
vdev_indirect_read_all(zio);
} else {
/*
* If this is a read zio, we read one copy of each
* split segment, from the top-level vdev. Since
* we don't know the checksum of each split
* individually, the child zio can't ensure that
* we get the right data. E.g. if it's a mirror,
* it will just read from a random (healthy) leaf
* vdev. We have to verify the checksum in
* vdev_indirect_io_done().
*
* For write zios, the vdev code will ensure we write
* to all children.
*/
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
zio_nowait(zio_vdev_child_io(zio, NULL,
is->is_vdev, is->is_target_offset,
abd_get_offset(zio->io_abd,
is->is_split_offset), is->is_size,
zio->io_type, zio->io_priority, 0,
vdev_indirect_child_io_done, zio));
}
}
}
zio_execute(zio);
}
/*
* Report a checksum error for a child.
*/
static void
vdev_indirect_checksum_error(zio_t *zio,
indirect_split_t *is, indirect_child_t *ic)
{
vdev_t *vd = ic->ic_vdev;
if (zio->io_flags & ZIO_FLAG_SPECULATIVE)
return;
mutex_enter(&vd->vdev_stat_lock);
vd->vdev_stat.vs_checksum_errors++;
mutex_exit(&vd->vdev_stat_lock);
zio_bad_cksum_t zbc = {{{ 0 }}};
abd_t *bad_abd = ic->ic_data;
abd_t *good_abd = is->is_good_child->ic_data;
(void) zfs_ereport_post_checksum(zio->io_spa, vd, NULL, zio,
is->is_target_offset, is->is_size, good_abd, bad_abd, &zbc);
}
/*
* Issue repair i/os for any incorrect copies. We do this by comparing
* each split segment's correct data (is_good_child's ic_data) with each
* other copy of the data. If they differ, then we overwrite the bad data
* with the good copy. The DTL is checked in vdev_indirect_read_all() and
* if a vdev is missing a copy of the data we set ic_error and the read is
* performed. This provides the opportunity to reconstruct the split block
* if at all possible. ic_error is checked here and if set it suppresses
* incrementing the checksum counter. Aside from this DTLs are not checked,
* which simplifies this code and also issues the optimal number of writes
* (based on which copies actually read bad data, as opposed to which we
* think might be wrong). For the same reason, we always use
* ZIO_FLAG_SELF_HEAL, to bypass the DTL check in zio_vdev_io_start().
*/
static void
vdev_indirect_repair(zio_t *zio)
{
indirect_vsd_t *iv = zio->io_vsd;
if (!spa_writeable(zio->io_spa))
return;
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
for (int c = 0; c < is->is_children; c++) {
indirect_child_t *ic = &is->is_child[c];
if (ic == is->is_good_child)
continue;
if (ic->ic_data == NULL)
continue;
if (ic->ic_duplicate == is->is_good_child)
continue;
zio_nowait(zio_vdev_child_io(zio, NULL,
ic->ic_vdev, is->is_target_offset,
is->is_good_child->ic_data, is->is_size,
ZIO_TYPE_WRITE, ZIO_PRIORITY_ASYNC_WRITE,
ZIO_FLAG_IO_REPAIR | ZIO_FLAG_SELF_HEAL,
NULL, NULL));
/*
* If ic_error is set the current child does not have
* a copy of the data, so suppress incrementing the
* checksum counter.
*/
if (ic->ic_error == ESTALE)
continue;
vdev_indirect_checksum_error(zio, is, ic);
}
}
}
/*
* Report checksum errors on all children that we read from.
*/
static void
vdev_indirect_all_checksum_errors(zio_t *zio)
{
indirect_vsd_t *iv = zio->io_vsd;
if (zio->io_flags & ZIO_FLAG_SPECULATIVE)
return;
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
for (int c = 0; c < is->is_children; c++) {
indirect_child_t *ic = &is->is_child[c];
if (ic->ic_data == NULL)
continue;
vdev_t *vd = ic->ic_vdev;
(void) zfs_ereport_post_checksum(zio->io_spa, vd,
NULL, zio, is->is_target_offset, is->is_size,
NULL, NULL, NULL);
mutex_enter(&vd->vdev_stat_lock);
vd->vdev_stat.vs_checksum_errors++;
mutex_exit(&vd->vdev_stat_lock);
}
}
}
/*
* Copy data from all the splits to a main zio then validate the checksum.
* If then checksum is successfully validated return success.
*/
static int
vdev_indirect_splits_checksum_validate(indirect_vsd_t *iv, zio_t *zio)
{
zio_bad_cksum_t zbc;
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
ASSERT3P(is->is_good_child->ic_data, !=, NULL);
ASSERT3P(is->is_good_child->ic_duplicate, ==, NULL);
abd_copy_off(zio->io_abd, is->is_good_child->ic_data,
is->is_split_offset, 0, is->is_size);
}
return (zio_checksum_error(zio, &zbc));
}
/*
* There are relatively few possible combinations making it feasible to
* deterministically check them all. We do this by setting the good_child
* to the next unique split version. If we reach the end of the list then
* "carry over" to the next unique split version (like counting in base
* is_unique_children, but each digit can have a different base).
*/
static int
vdev_indirect_splits_enumerate_all(indirect_vsd_t *iv, zio_t *zio)
{
boolean_t more = B_TRUE;
iv->iv_attempts = 0;
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is))
is->is_good_child = list_head(&is->is_unique_child);
while (more == B_TRUE) {
iv->iv_attempts++;
more = B_FALSE;
if (vdev_indirect_splits_checksum_validate(iv, zio) == 0)
return (0);
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
is->is_good_child = list_next(&is->is_unique_child,
is->is_good_child);
if (is->is_good_child != NULL) {
more = B_TRUE;
break;
}
is->is_good_child = list_head(&is->is_unique_child);
}
}
ASSERT3S(iv->iv_attempts, <=, iv->iv_unique_combinations);
return (SET_ERROR(ECKSUM));
}
/*
* There are too many combinations to try all of them in a reasonable amount
* of time. So try a fixed number of random combinations from the unique
* split versions, after which we'll consider the block unrecoverable.
*/
static int
vdev_indirect_splits_enumerate_randomly(indirect_vsd_t *iv, zio_t *zio)
{
iv->iv_attempts = 0;
while (iv->iv_attempts < iv->iv_attempts_max) {
iv->iv_attempts++;
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
indirect_child_t *ic = list_head(&is->is_unique_child);
int children = is->is_unique_children;
for (int i = random_in_range(children); i > 0; i--)
ic = list_next(&is->is_unique_child, ic);
ASSERT3P(ic, !=, NULL);
is->is_good_child = ic;
}
if (vdev_indirect_splits_checksum_validate(iv, zio) == 0)
return (0);
}
return (SET_ERROR(ECKSUM));
}
/*
* This is a validation function for reconstruction. It randomly selects
* a good combination, if one can be found, and then it intentionally
* damages all other segment copes by zeroing them. This forces the
* reconstruction algorithm to locate the one remaining known good copy.
*/
static int
vdev_indirect_splits_damage(indirect_vsd_t *iv, zio_t *zio)
{
int error;
/* Presume all the copies are unique for initial selection. */
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
is->is_unique_children = 0;
for (int i = 0; i < is->is_children; i++) {
indirect_child_t *ic = &is->is_child[i];
if (ic->ic_data != NULL) {
is->is_unique_children++;
list_insert_tail(&is->is_unique_child, ic);
}
}
if (list_is_empty(&is->is_unique_child)) {
error = SET_ERROR(EIO);
goto out;
}
}
/*
* Set each is_good_child to a randomly-selected child which
* is known to contain validated data.
*/
error = vdev_indirect_splits_enumerate_randomly(iv, zio);
if (error)
goto out;
/*
* Damage all but the known good copy by zeroing it. This will
* result in two or less unique copies per indirect_child_t.
* Both may need to be checked in order to reconstruct the block.
* Set iv->iv_attempts_max such that all unique combinations will
* enumerated, but limit the damage to at most 12 indirect splits.
*/
iv->iv_attempts_max = 1;
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
for (int c = 0; c < is->is_children; c++) {
indirect_child_t *ic = &is->is_child[c];
if (ic == is->is_good_child)
continue;
if (ic->ic_data == NULL)
continue;
abd_zero(ic->ic_data, abd_get_size(ic->ic_data));
}
iv->iv_attempts_max *= 2;
if (iv->iv_attempts_max >= (1ULL << 12)) {
iv->iv_attempts_max = UINT64_MAX;
break;
}
}
out:
/* Empty the unique children lists so they can be reconstructed. */
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
indirect_child_t *ic;
while ((ic = list_head(&is->is_unique_child)) != NULL)
list_remove(&is->is_unique_child, ic);
is->is_unique_children = 0;
}
return (error);
}
/*
* This function is called when we have read all copies of the data and need
* to try to find a combination of copies that gives us the right checksum.
*
* If we pointed to any mirror vdevs, this effectively does the job of the
* mirror. The mirror vdev code can't do its own job because we don't know
* the checksum of each split segment individually.
*
* We have to try every unique combination of copies of split segments, until
* we find one that checksums correctly. Duplicate segment copies are first
* identified and latter skipped during reconstruction. This optimization
* reduces the search space and ensures that of the remaining combinations
* at most one is correct.
*
* When the total number of combinations is small they can all be checked.
* For example, if we have 3 segments in the split, and each points to a
* 2-way mirror with unique copies, we will have the following pieces of data:
*
* | mirror child
* split | [0] [1]
* ======|=====================
* A | data_A_0 data_A_1
* B | data_B_0 data_B_1
* C | data_C_0 data_C_1
*
* We will try the following (mirror children)^(number of splits) (2^3=8)
* combinations, which is similar to bitwise-little-endian counting in
* binary. In general each "digit" corresponds to a split segment, and the
* base of each digit is is_children, which can be different for each
* digit.
*
* "low bit" "high bit"
* v v
* data_A_0 data_B_0 data_C_0
* data_A_1 data_B_0 data_C_0
* data_A_0 data_B_1 data_C_0
* data_A_1 data_B_1 data_C_0
* data_A_0 data_B_0 data_C_1
* data_A_1 data_B_0 data_C_1
* data_A_0 data_B_1 data_C_1
* data_A_1 data_B_1 data_C_1
*
* Note that the split segments may be on the same or different top-level
* vdevs. In either case, we may need to try lots of combinations (see
* zfs_reconstruct_indirect_combinations_max). This ensures that if a mirror
* has small silent errors on all of its children, we can still reconstruct
* the correct data, as long as those errors are at sufficiently-separated
* offsets (specifically, separated by the largest block size - default of
* 128KB, but up to 16MB).
*/
static void
vdev_indirect_reconstruct_io_done(zio_t *zio)
{
indirect_vsd_t *iv = zio->io_vsd;
boolean_t known_good = B_FALSE;
int error;
iv->iv_unique_combinations = 1;
iv->iv_attempts_max = UINT64_MAX;
if (zfs_reconstruct_indirect_combinations_max > 0)
iv->iv_attempts_max = zfs_reconstruct_indirect_combinations_max;
/*
* If nonzero, every 1/x blocks will be damaged, in order to validate
* reconstruction when there are split segments with damaged copies.
* Known_good will be TRUE when reconstruction is known to be possible.
*/
if (zfs_reconstruct_indirect_damage_fraction != 0 &&
random_in_range(zfs_reconstruct_indirect_damage_fraction) == 0)
known_good = (vdev_indirect_splits_damage(iv, zio) == 0);
/*
* Determine the unique children for a split segment and add them
* to the is_unique_child list. By restricting reconstruction
* to these children, only unique combinations will be considered.
* This can vastly reduce the search space when there are a large
* number of indirect splits.
*/
for (indirect_split_t *is = list_head(&iv->iv_splits);
is != NULL; is = list_next(&iv->iv_splits, is)) {
is->is_unique_children = 0;
for (int i = 0; i < is->is_children; i++) {
indirect_child_t *ic_i = &is->is_child[i];
if (ic_i->ic_data == NULL ||
ic_i->ic_duplicate != NULL)
continue;
for (int j = i + 1; j < is->is_children; j++) {
indirect_child_t *ic_j = &is->is_child[j];
if (ic_j->ic_data == NULL ||
ic_j->ic_duplicate != NULL)
continue;
if (abd_cmp(ic_i->ic_data, ic_j->ic_data) == 0)
ic_j->ic_duplicate = ic_i;
}
is->is_unique_children++;
list_insert_tail(&is->is_unique_child, ic_i);
}
/* Reconstruction is impossible, no valid children */
EQUIV(list_is_empty(&is->is_unique_child),
is->is_unique_children == 0);
if (list_is_empty(&is->is_unique_child)) {
zio->io_error = EIO;
vdev_indirect_all_checksum_errors(zio);
zio_checksum_verified(zio);
return;
}
iv->iv_unique_combinations *= is->is_unique_children;
}
if (iv->iv_unique_combinations <= iv->iv_attempts_max)
error = vdev_indirect_splits_enumerate_all(iv, zio);
else
error = vdev_indirect_splits_enumerate_randomly(iv, zio);
if (error != 0) {
/* All attempted combinations failed. */
ASSERT3B(known_good, ==, B_FALSE);
zio->io_error = error;
vdev_indirect_all_checksum_errors(zio);
} else {
/*
* The checksum has been successfully validated. Issue
* repair I/Os to any copies of splits which don't match
* the validated version.
*/
ASSERT0(vdev_indirect_splits_checksum_validate(iv, zio));
vdev_indirect_repair(zio);
zio_checksum_verified(zio);
}
}
static void
vdev_indirect_io_done(zio_t *zio)
{
indirect_vsd_t *iv = zio->io_vsd;
if (iv->iv_reconstruct) {
/*
* We have read all copies of the data (e.g. from mirrors),
* either because this was a scrub/resilver, or because the
* one-copy read didn't checksum correctly.
*/
vdev_indirect_reconstruct_io_done(zio);
return;
}
if (!iv->iv_split_block) {
/*
* This was not a split block, so we passed the BP down,
* and the checksum was handled by the (one) child zio.
*/
return;
}
zio_bad_cksum_t zbc;
int ret = zio_checksum_error(zio, &zbc);
if (ret == 0) {
zio_checksum_verified(zio);
return;
}
/*
* The checksum didn't match. Read all copies of all splits, and
* then we will try to reconstruct. The next time
* vdev_indirect_io_done() is called, iv_reconstruct will be set.
*/
vdev_indirect_read_all(zio);
zio_vdev_io_redone(zio);
}
vdev_ops_t vdev_indirect_ops = {
.vdev_op_init = NULL,
.vdev_op_fini = NULL,
.vdev_op_open = vdev_indirect_open,
.vdev_op_close = vdev_indirect_close,
.vdev_op_asize = vdev_default_asize,
.vdev_op_min_asize = vdev_default_min_asize,
.vdev_op_min_alloc = NULL,
.vdev_op_io_start = vdev_indirect_io_start,
.vdev_op_io_done = vdev_indirect_io_done,
.vdev_op_state_change = NULL,
.vdev_op_need_resilver = NULL,
.vdev_op_hold = NULL,
.vdev_op_rele = NULL,
.vdev_op_remap = vdev_indirect_remap,
.vdev_op_xlate = NULL,
.vdev_op_rebuild_asize = NULL,
.vdev_op_metaslab_init = NULL,
.vdev_op_config_generate = NULL,
.vdev_op_nparity = NULL,
.vdev_op_ndisks = NULL,
.vdev_op_type = VDEV_TYPE_INDIRECT, /* name of this vdev type */
.vdev_op_leaf = B_FALSE /* leaf vdev */
};
EXPORT_SYMBOL(spa_condense_fini);
EXPORT_SYMBOL(spa_start_indirect_condensing_thread);
EXPORT_SYMBOL(spa_condense_indirect_start_sync);
EXPORT_SYMBOL(spa_condense_init);
EXPORT_SYMBOL(spa_vdev_indirect_mark_obsolete);
EXPORT_SYMBOL(vdev_indirect_mark_obsolete);
EXPORT_SYMBOL(vdev_indirect_should_condense);
EXPORT_SYMBOL(vdev_indirect_sync_obsolete);
EXPORT_SYMBOL(vdev_obsolete_counts_are_precise);
EXPORT_SYMBOL(vdev_obsolete_sm_object);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, indirect_vdevs_enable, INT, ZMOD_RW,
"Whether to attempt condensing indirect vdev mappings");
ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, indirect_obsolete_pct, INT, ZMOD_RW,
"Minimum obsolete percent of bytes in the mapping to attempt condensing");
ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, min_mapping_bytes, ULONG, ZMOD_RW,
"Don't bother condensing if the mapping uses less than this amount of "
"memory");
ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, max_obsolete_bytes, ULONG, ZMOD_RW,
"Minimum size obsolete spacemap to attempt condensing");
ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, indirect_commit_entry_delay_ms, INT, ZMOD_RW,
"Used by tests to ensure certain actions happen in the middle of a "
"condense. A maximum value of 1 should be sufficient.");
ZFS_MODULE_PARAM(zfs_reconstruct, zfs_reconstruct_, indirect_combinations_max, INT, ZMOD_RW,
"Maximum number of combinations when reconstructing split segments");
/* END CSTYLED */
diff --git a/module/zfs/vdev_initialize.c b/module/zfs/vdev_initialize.c
index 6ffd0d618fdd..eda71faeacdc 100644
--- a/module/zfs/vdev_initialize.c
+++ b/module/zfs/vdev_initialize.c
@@ -1,774 +1,774 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2016, 2019 by Delphix. All rights reserved.
*/
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/txg.h>
#include <sys/vdev_impl.h>
#include <sys/metaslab_impl.h>
#include <sys/dsl_synctask.h>
#include <sys/zap.h>
#include <sys/dmu_tx.h>
#include <sys/vdev_initialize.h>
/*
* Value that is written to disk during initialization.
*/
#ifdef _ILP32
-unsigned long zfs_initialize_value = 0xdeadbeefUL;
+static unsigned long zfs_initialize_value = 0xdeadbeefUL;
#else
-unsigned long zfs_initialize_value = 0xdeadbeefdeadbeeeULL;
+static unsigned long zfs_initialize_value = 0xdeadbeefdeadbeeeULL;
#endif
/* maximum number of I/Os outstanding per leaf vdev */
-int zfs_initialize_limit = 1;
+static const int zfs_initialize_limit = 1;
/* size of initializing writes; default 1MiB, see zfs_remove_max_segment */
-unsigned long zfs_initialize_chunk_size = 1024 * 1024;
+static unsigned long zfs_initialize_chunk_size = 1024 * 1024;
static boolean_t
vdev_initialize_should_stop(vdev_t *vd)
{
return (vd->vdev_initialize_exit_wanted || !vdev_writeable(vd) ||
vd->vdev_detached || vd->vdev_top->vdev_removing);
}
static void
vdev_initialize_zap_update_sync(void *arg, dmu_tx_t *tx)
{
/*
* We pass in the guid instead of the vdev_t since the vdev may
* have been freed prior to the sync task being processed. This
* happens when a vdev is detached as we call spa_config_vdev_exit(),
* stop the initializing thread, schedule the sync task, and free
* the vdev. Later when the scheduled sync task is invoked, it would
* find that the vdev has been freed.
*/
uint64_t guid = *(uint64_t *)arg;
uint64_t txg = dmu_tx_get_txg(tx);
kmem_free(arg, sizeof (uint64_t));
vdev_t *vd = spa_lookup_by_guid(tx->tx_pool->dp_spa, guid, B_FALSE);
if (vd == NULL || vd->vdev_top->vdev_removing || !vdev_is_concrete(vd))
return;
uint64_t last_offset = vd->vdev_initialize_offset[txg & TXG_MASK];
vd->vdev_initialize_offset[txg & TXG_MASK] = 0;
VERIFY(vd->vdev_leaf_zap != 0);
objset_t *mos = vd->vdev_spa->spa_meta_objset;
if (last_offset > 0) {
vd->vdev_initialize_last_offset = last_offset;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET,
sizeof (last_offset), 1, &last_offset, tx));
}
if (vd->vdev_initialize_action_time > 0) {
uint64_t val = (uint64_t)vd->vdev_initialize_action_time;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME, sizeof (val),
1, &val, tx));
}
uint64_t initialize_state = vd->vdev_initialize_state;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
VDEV_LEAF_ZAP_INITIALIZE_STATE, sizeof (initialize_state), 1,
&initialize_state, tx));
}
static void
vdev_initialize_change_state(vdev_t *vd, vdev_initializing_state_t new_state)
{
ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock));
spa_t *spa = vd->vdev_spa;
if (new_state == vd->vdev_initialize_state)
return;
/*
* Copy the vd's guid, this will be freed by the sync task.
*/
uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP);
*guid = vd->vdev_guid;
/*
* If we're suspending, then preserving the original start time.
*/
if (vd->vdev_initialize_state != VDEV_INITIALIZE_SUSPENDED) {
vd->vdev_initialize_action_time = gethrestime_sec();
}
vdev_initializing_state_t old_state = vd->vdev_initialize_state;
vd->vdev_initialize_state = new_state;
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
dsl_sync_task_nowait(spa_get_dsl(spa), vdev_initialize_zap_update_sync,
guid, tx);
switch (new_state) {
case VDEV_INITIALIZE_ACTIVE:
spa_history_log_internal(spa, "initialize", tx,
"vdev=%s activated", vd->vdev_path);
break;
case VDEV_INITIALIZE_SUSPENDED:
spa_history_log_internal(spa, "initialize", tx,
"vdev=%s suspended", vd->vdev_path);
break;
case VDEV_INITIALIZE_CANCELED:
if (old_state == VDEV_INITIALIZE_ACTIVE ||
old_state == VDEV_INITIALIZE_SUSPENDED)
spa_history_log_internal(spa, "initialize", tx,
"vdev=%s canceled", vd->vdev_path);
break;
case VDEV_INITIALIZE_COMPLETE:
spa_history_log_internal(spa, "initialize", tx,
"vdev=%s complete", vd->vdev_path);
break;
default:
panic("invalid state %llu", (unsigned long long)new_state);
}
dmu_tx_commit(tx);
if (new_state != VDEV_INITIALIZE_ACTIVE)
spa_notify_waiters(spa);
}
static void
vdev_initialize_cb(zio_t *zio)
{
vdev_t *vd = zio->io_vd;
mutex_enter(&vd->vdev_initialize_io_lock);
if (zio->io_error == ENXIO && !vdev_writeable(vd)) {
/*
* The I/O failed because the vdev was unavailable; roll the
* last offset back. (This works because spa_sync waits on
* spa_txg_zio before it runs sync tasks.)
*/
uint64_t *off =
&vd->vdev_initialize_offset[zio->io_txg & TXG_MASK];
*off = MIN(*off, zio->io_offset);
} else {
/*
* Since initializing is best-effort, we ignore I/O errors and
* rely on vdev_probe to determine if the errors are more
* critical.
*/
if (zio->io_error != 0)
vd->vdev_stat.vs_initialize_errors++;
vd->vdev_initialize_bytes_done += zio->io_orig_size;
}
ASSERT3U(vd->vdev_initialize_inflight, >, 0);
vd->vdev_initialize_inflight--;
cv_broadcast(&vd->vdev_initialize_io_cv);
mutex_exit(&vd->vdev_initialize_io_lock);
spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
}
/* Takes care of physical writing and limiting # of concurrent ZIOs. */
static int
vdev_initialize_write(vdev_t *vd, uint64_t start, uint64_t size, abd_t *data)
{
spa_t *spa = vd->vdev_spa;
/* Limit inflight initializing I/Os */
mutex_enter(&vd->vdev_initialize_io_lock);
while (vd->vdev_initialize_inflight >= zfs_initialize_limit) {
cv_wait(&vd->vdev_initialize_io_cv,
&vd->vdev_initialize_io_lock);
}
vd->vdev_initialize_inflight++;
mutex_exit(&vd->vdev_initialize_io_lock);
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
uint64_t txg = dmu_tx_get_txg(tx);
spa_config_enter(spa, SCL_STATE_ALL, vd, RW_READER);
mutex_enter(&vd->vdev_initialize_lock);
if (vd->vdev_initialize_offset[txg & TXG_MASK] == 0) {
uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP);
*guid = vd->vdev_guid;
/* This is the first write of this txg. */
dsl_sync_task_nowait(spa_get_dsl(spa),
vdev_initialize_zap_update_sync, guid, tx);
}
/*
* We know the vdev struct will still be around since all
* consumers of vdev_free must stop the initialization first.
*/
if (vdev_initialize_should_stop(vd)) {
mutex_enter(&vd->vdev_initialize_io_lock);
ASSERT3U(vd->vdev_initialize_inflight, >, 0);
vd->vdev_initialize_inflight--;
mutex_exit(&vd->vdev_initialize_io_lock);
spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
mutex_exit(&vd->vdev_initialize_lock);
dmu_tx_commit(tx);
return (SET_ERROR(EINTR));
}
mutex_exit(&vd->vdev_initialize_lock);
vd->vdev_initialize_offset[txg & TXG_MASK] = start + size;
zio_nowait(zio_write_phys(spa->spa_txg_zio[txg & TXG_MASK], vd, start,
size, data, ZIO_CHECKSUM_OFF, vdev_initialize_cb, NULL,
ZIO_PRIORITY_INITIALIZING, ZIO_FLAG_CANFAIL, B_FALSE));
/* vdev_initialize_cb releases SCL_STATE_ALL */
dmu_tx_commit(tx);
return (0);
}
/*
* Callback to fill each ABD chunk with zfs_initialize_value. len must be
* divisible by sizeof (uint64_t), and buf must be 8-byte aligned. The ABD
* allocation will guarantee these for us.
*/
static int
vdev_initialize_block_fill(void *buf, size_t len, void *unused)
{
(void) unused;
ASSERT0(len % sizeof (uint64_t));
#ifdef _ILP32
for (uint64_t i = 0; i < len; i += sizeof (uint32_t)) {
*(uint32_t *)((char *)(buf) + i) = zfs_initialize_value;
}
#else
for (uint64_t i = 0; i < len; i += sizeof (uint64_t)) {
*(uint64_t *)((char *)(buf) + i) = zfs_initialize_value;
}
#endif
return (0);
}
static abd_t *
vdev_initialize_block_alloc(void)
{
/* Allocate ABD for filler data */
abd_t *data = abd_alloc_for_io(zfs_initialize_chunk_size, B_FALSE);
ASSERT0(zfs_initialize_chunk_size % sizeof (uint64_t));
(void) abd_iterate_func(data, 0, zfs_initialize_chunk_size,
vdev_initialize_block_fill, NULL);
return (data);
}
static void
vdev_initialize_block_free(abd_t *data)
{
abd_free(data);
}
static int
vdev_initialize_ranges(vdev_t *vd, abd_t *data)
{
range_tree_t *rt = vd->vdev_initialize_tree;
zfs_btree_t *bt = &rt->rt_root;
zfs_btree_index_t where;
for (range_seg_t *rs = zfs_btree_first(bt, &where); rs != NULL;
rs = zfs_btree_next(bt, &where, &where)) {
uint64_t size = rs_get_end(rs, rt) - rs_get_start(rs, rt);
/* Split range into legally-sized physical chunks */
uint64_t writes_required =
((size - 1) / zfs_initialize_chunk_size) + 1;
for (uint64_t w = 0; w < writes_required; w++) {
int error;
error = vdev_initialize_write(vd,
VDEV_LABEL_START_SIZE + rs_get_start(rs, rt) +
(w * zfs_initialize_chunk_size),
MIN(size - (w * zfs_initialize_chunk_size),
zfs_initialize_chunk_size), data);
if (error != 0)
return (error);
}
}
return (0);
}
static void
vdev_initialize_xlate_last_rs_end(void *arg, range_seg64_t *physical_rs)
{
uint64_t *last_rs_end = (uint64_t *)arg;
if (physical_rs->rs_end > *last_rs_end)
*last_rs_end = physical_rs->rs_end;
}
static void
vdev_initialize_xlate_progress(void *arg, range_seg64_t *physical_rs)
{
vdev_t *vd = (vdev_t *)arg;
uint64_t size = physical_rs->rs_end - physical_rs->rs_start;
vd->vdev_initialize_bytes_est += size;
if (vd->vdev_initialize_last_offset > physical_rs->rs_end) {
vd->vdev_initialize_bytes_done += size;
} else if (vd->vdev_initialize_last_offset > physical_rs->rs_start &&
vd->vdev_initialize_last_offset < physical_rs->rs_end) {
vd->vdev_initialize_bytes_done +=
vd->vdev_initialize_last_offset - physical_rs->rs_start;
}
}
static void
vdev_initialize_calculate_progress(vdev_t *vd)
{
ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) ||
spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER));
ASSERT(vd->vdev_leaf_zap != 0);
vd->vdev_initialize_bytes_est = 0;
vd->vdev_initialize_bytes_done = 0;
for (uint64_t i = 0; i < vd->vdev_top->vdev_ms_count; i++) {
metaslab_t *msp = vd->vdev_top->vdev_ms[i];
mutex_enter(&msp->ms_lock);
uint64_t ms_free = (msp->ms_size -
metaslab_allocated_space(msp)) /
vdev_get_ndisks(vd->vdev_top);
/*
* Convert the metaslab range to a physical range
* on our vdev. We use this to determine if we are
* in the middle of this metaslab range.
*/
range_seg64_t logical_rs, physical_rs, remain_rs;
logical_rs.rs_start = msp->ms_start;
logical_rs.rs_end = msp->ms_start + msp->ms_size;
/* Metaslab space after this offset has not been initialized */
vdev_xlate(vd, &logical_rs, &physical_rs, &remain_rs);
if (vd->vdev_initialize_last_offset <= physical_rs.rs_start) {
vd->vdev_initialize_bytes_est += ms_free;
mutex_exit(&msp->ms_lock);
continue;
}
/* Metaslab space before this offset has been initialized */
uint64_t last_rs_end = physical_rs.rs_end;
if (!vdev_xlate_is_empty(&remain_rs)) {
vdev_xlate_walk(vd, &remain_rs,
vdev_initialize_xlate_last_rs_end, &last_rs_end);
}
if (vd->vdev_initialize_last_offset > last_rs_end) {
vd->vdev_initialize_bytes_done += ms_free;
vd->vdev_initialize_bytes_est += ms_free;
mutex_exit(&msp->ms_lock);
continue;
}
/*
* If we get here, we're in the middle of initializing this
* metaslab. Load it and walk the free tree for more accurate
* progress estimation.
*/
VERIFY0(metaslab_load(msp));
zfs_btree_index_t where;
range_tree_t *rt = msp->ms_allocatable;
for (range_seg_t *rs =
zfs_btree_first(&rt->rt_root, &where); rs;
rs = zfs_btree_next(&rt->rt_root, &where,
&where)) {
logical_rs.rs_start = rs_get_start(rs, rt);
logical_rs.rs_end = rs_get_end(rs, rt);
vdev_xlate_walk(vd, &logical_rs,
vdev_initialize_xlate_progress, vd);
}
mutex_exit(&msp->ms_lock);
}
}
static int
vdev_initialize_load(vdev_t *vd)
{
int err = 0;
ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) ||
spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER));
ASSERT(vd->vdev_leaf_zap != 0);
if (vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE ||
vd->vdev_initialize_state == VDEV_INITIALIZE_SUSPENDED) {
err = zap_lookup(vd->vdev_spa->spa_meta_objset,
vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET,
sizeof (vd->vdev_initialize_last_offset), 1,
&vd->vdev_initialize_last_offset);
if (err == ENOENT) {
vd->vdev_initialize_last_offset = 0;
err = 0;
}
}
vdev_initialize_calculate_progress(vd);
return (err);
}
static void
vdev_initialize_xlate_range_add(void *arg, range_seg64_t *physical_rs)
{
vdev_t *vd = arg;
/* Only add segments that we have not visited yet */
if (physical_rs->rs_end <= vd->vdev_initialize_last_offset)
return;
/* Pick up where we left off mid-range. */
if (vd->vdev_initialize_last_offset > physical_rs->rs_start) {
zfs_dbgmsg("range write: vd %s changed (%llu, %llu) to "
"(%llu, %llu)", vd->vdev_path,
(u_longlong_t)physical_rs->rs_start,
(u_longlong_t)physical_rs->rs_end,
(u_longlong_t)vd->vdev_initialize_last_offset,
(u_longlong_t)physical_rs->rs_end);
ASSERT3U(physical_rs->rs_end, >,
vd->vdev_initialize_last_offset);
physical_rs->rs_start = vd->vdev_initialize_last_offset;
}
ASSERT3U(physical_rs->rs_end, >, physical_rs->rs_start);
range_tree_add(vd->vdev_initialize_tree, physical_rs->rs_start,
physical_rs->rs_end - physical_rs->rs_start);
}
/*
* Convert the logical range into a physical range and add it to our
* avl tree.
*/
static void
vdev_initialize_range_add(void *arg, uint64_t start, uint64_t size)
{
vdev_t *vd = arg;
range_seg64_t logical_rs;
logical_rs.rs_start = start;
logical_rs.rs_end = start + size;
ASSERT(vd->vdev_ops->vdev_op_leaf);
vdev_xlate_walk(vd, &logical_rs, vdev_initialize_xlate_range_add, arg);
}
static void
vdev_initialize_thread(void *arg)
{
vdev_t *vd = arg;
spa_t *spa = vd->vdev_spa;
int error = 0;
uint64_t ms_count = 0;
ASSERT(vdev_is_concrete(vd));
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vd->vdev_initialize_last_offset = 0;
VERIFY0(vdev_initialize_load(vd));
abd_t *deadbeef = vdev_initialize_block_alloc();
vd->vdev_initialize_tree = range_tree_create(NULL, RANGE_SEG64, NULL,
0, 0);
for (uint64_t i = 0; !vd->vdev_detached &&
i < vd->vdev_top->vdev_ms_count; i++) {
metaslab_t *msp = vd->vdev_top->vdev_ms[i];
boolean_t unload_when_done = B_FALSE;
/*
* If we've expanded the top-level vdev or it's our
* first pass, calculate our progress.
*/
if (vd->vdev_top->vdev_ms_count != ms_count) {
vdev_initialize_calculate_progress(vd);
ms_count = vd->vdev_top->vdev_ms_count;
}
spa_config_exit(spa, SCL_CONFIG, FTAG);
metaslab_disable(msp);
mutex_enter(&msp->ms_lock);
if (!msp->ms_loaded && !msp->ms_loading)
unload_when_done = B_TRUE;
VERIFY0(metaslab_load(msp));
range_tree_walk(msp->ms_allocatable, vdev_initialize_range_add,
vd);
mutex_exit(&msp->ms_lock);
error = vdev_initialize_ranges(vd, deadbeef);
metaslab_enable(msp, B_TRUE, unload_when_done);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
range_tree_vacate(vd->vdev_initialize_tree, NULL, NULL);
if (error != 0)
break;
}
spa_config_exit(spa, SCL_CONFIG, FTAG);
mutex_enter(&vd->vdev_initialize_io_lock);
while (vd->vdev_initialize_inflight > 0) {
cv_wait(&vd->vdev_initialize_io_cv,
&vd->vdev_initialize_io_lock);
}
mutex_exit(&vd->vdev_initialize_io_lock);
range_tree_destroy(vd->vdev_initialize_tree);
vdev_initialize_block_free(deadbeef);
vd->vdev_initialize_tree = NULL;
mutex_enter(&vd->vdev_initialize_lock);
if (!vd->vdev_initialize_exit_wanted) {
if (vdev_writeable(vd)) {
vdev_initialize_change_state(vd,
VDEV_INITIALIZE_COMPLETE);
} else if (vd->vdev_faulted) {
vdev_initialize_change_state(vd,
VDEV_INITIALIZE_CANCELED);
}
}
ASSERT(vd->vdev_initialize_thread != NULL ||
vd->vdev_initialize_inflight == 0);
/*
* Drop the vdev_initialize_lock while we sync out the
* txg since it's possible that a device might be trying to
* come online and must check to see if it needs to restart an
* initialization. That thread will be holding the spa_config_lock
* which would prevent the txg_wait_synced from completing.
*/
mutex_exit(&vd->vdev_initialize_lock);
txg_wait_synced(spa_get_dsl(spa), 0);
mutex_enter(&vd->vdev_initialize_lock);
vd->vdev_initialize_thread = NULL;
cv_broadcast(&vd->vdev_initialize_cv);
mutex_exit(&vd->vdev_initialize_lock);
thread_exit();
}
/*
* Initiates a device. Caller must hold vdev_initialize_lock.
* Device must be a leaf and not already be initializing.
*/
void
vdev_initialize(vdev_t *vd)
{
ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock));
ASSERT(vd->vdev_ops->vdev_op_leaf);
ASSERT(vdev_is_concrete(vd));
ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
ASSERT(!vd->vdev_detached);
ASSERT(!vd->vdev_initialize_exit_wanted);
ASSERT(!vd->vdev_top->vdev_removing);
vdev_initialize_change_state(vd, VDEV_INITIALIZE_ACTIVE);
vd->vdev_initialize_thread = thread_create(NULL, 0,
vdev_initialize_thread, vd, 0, &p0, TS_RUN, maxclsyspri);
}
/*
* Wait for the initialize thread to be terminated (cancelled or stopped).
*/
static void
vdev_initialize_stop_wait_impl(vdev_t *vd)
{
ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock));
while (vd->vdev_initialize_thread != NULL)
cv_wait(&vd->vdev_initialize_cv, &vd->vdev_initialize_lock);
ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
vd->vdev_initialize_exit_wanted = B_FALSE;
}
/*
* Wait for vdev initialize threads which were either to cleanly exit.
*/
void
vdev_initialize_stop_wait(spa_t *spa, list_t *vd_list)
{
(void) spa;
vdev_t *vd;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
while ((vd = list_remove_head(vd_list)) != NULL) {
mutex_enter(&vd->vdev_initialize_lock);
vdev_initialize_stop_wait_impl(vd);
mutex_exit(&vd->vdev_initialize_lock);
}
}
/*
* Stop initializing a device, with the resultant initializing state being
* tgt_state. For blocking behavior pass NULL for vd_list. Otherwise, when
* a list_t is provided the stopping vdev is inserted in to the list. Callers
* are then required to call vdev_initialize_stop_wait() to block for all the
* initialization threads to exit. The caller must hold vdev_initialize_lock
* and must not be writing to the spa config, as the initializing thread may
* try to enter the config as a reader before exiting.
*/
void
vdev_initialize_stop(vdev_t *vd, vdev_initializing_state_t tgt_state,
list_t *vd_list)
{
ASSERT(!spa_config_held(vd->vdev_spa, SCL_CONFIG|SCL_STATE, RW_WRITER));
ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock));
ASSERT(vd->vdev_ops->vdev_op_leaf);
ASSERT(vdev_is_concrete(vd));
/*
* Allow cancel requests to proceed even if the initialize thread
* has stopped.
*/
if (vd->vdev_initialize_thread == NULL &&
tgt_state != VDEV_INITIALIZE_CANCELED) {
return;
}
vdev_initialize_change_state(vd, tgt_state);
vd->vdev_initialize_exit_wanted = B_TRUE;
if (vd_list == NULL) {
vdev_initialize_stop_wait_impl(vd);
} else {
ASSERT(MUTEX_HELD(&spa_namespace_lock));
list_insert_tail(vd_list, vd);
}
}
static void
vdev_initialize_stop_all_impl(vdev_t *vd, vdev_initializing_state_t tgt_state,
list_t *vd_list)
{
if (vd->vdev_ops->vdev_op_leaf && vdev_is_concrete(vd)) {
mutex_enter(&vd->vdev_initialize_lock);
vdev_initialize_stop(vd, tgt_state, vd_list);
mutex_exit(&vd->vdev_initialize_lock);
return;
}
for (uint64_t i = 0; i < vd->vdev_children; i++) {
vdev_initialize_stop_all_impl(vd->vdev_child[i], tgt_state,
vd_list);
}
}
/*
* Convenience function to stop initializing of a vdev tree and set all
* initialize thread pointers to NULL.
*/
void
vdev_initialize_stop_all(vdev_t *vd, vdev_initializing_state_t tgt_state)
{
spa_t *spa = vd->vdev_spa;
list_t vd_list;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
list_create(&vd_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_initialize_node));
vdev_initialize_stop_all_impl(vd, tgt_state, &vd_list);
vdev_initialize_stop_wait(spa, &vd_list);
if (vd->vdev_spa->spa_sync_on) {
/* Make sure that our state has been synced to disk */
txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0);
}
list_destroy(&vd_list);
}
void
vdev_initialize_restart(vdev_t *vd)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
ASSERT(!spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));
if (vd->vdev_leaf_zap != 0) {
mutex_enter(&vd->vdev_initialize_lock);
uint64_t initialize_state = VDEV_INITIALIZE_NONE;
int err = zap_lookup(vd->vdev_spa->spa_meta_objset,
vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_STATE,
sizeof (initialize_state), 1, &initialize_state);
ASSERT(err == 0 || err == ENOENT);
vd->vdev_initialize_state = initialize_state;
uint64_t timestamp = 0;
err = zap_lookup(vd->vdev_spa->spa_meta_objset,
vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME,
sizeof (timestamp), 1, &timestamp);
ASSERT(err == 0 || err == ENOENT);
vd->vdev_initialize_action_time = timestamp;
if (vd->vdev_initialize_state == VDEV_INITIALIZE_SUSPENDED ||
vd->vdev_offline) {
/* load progress for reporting, but don't resume */
VERIFY0(vdev_initialize_load(vd));
} else if (vd->vdev_initialize_state ==
VDEV_INITIALIZE_ACTIVE && vdev_writeable(vd) &&
!vd->vdev_top->vdev_removing &&
vd->vdev_initialize_thread == NULL) {
vdev_initialize(vd);
}
mutex_exit(&vd->vdev_initialize_lock);
}
for (uint64_t i = 0; i < vd->vdev_children; i++) {
vdev_initialize_restart(vd->vdev_child[i]);
}
}
EXPORT_SYMBOL(vdev_initialize);
EXPORT_SYMBOL(vdev_initialize_stop);
EXPORT_SYMBOL(vdev_initialize_stop_all);
EXPORT_SYMBOL(vdev_initialize_stop_wait);
EXPORT_SYMBOL(vdev_initialize_restart);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, initialize_value, ULONG, ZMOD_RW,
"Value written during zpool initialize");
ZFS_MODULE_PARAM(zfs, zfs_, initialize_chunk_size, ULONG, ZMOD_RW,
"Size in bytes of writes by zpool initialize");
/* END CSTYLED */
diff --git a/module/zfs/vdev_mirror.c b/module/zfs/vdev_mirror.c
index 45b744b2ec89..30d0e7de54b0 100644
--- a/module/zfs/vdev_mirror.c
+++ b/module/zfs/vdev_mirror.c
@@ -1,982 +1,982 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2010 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright (c) 2012, 2015 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_scan.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_draid.h>
#include <sys/zio.h>
#include <sys/abd.h>
#include <sys/fs/zfs.h>
/*
* Vdev mirror kstats
*/
static kstat_t *mirror_ksp = NULL;
typedef struct mirror_stats {
kstat_named_t vdev_mirror_stat_rotating_linear;
kstat_named_t vdev_mirror_stat_rotating_offset;
kstat_named_t vdev_mirror_stat_rotating_seek;
kstat_named_t vdev_mirror_stat_non_rotating_linear;
kstat_named_t vdev_mirror_stat_non_rotating_seek;
kstat_named_t vdev_mirror_stat_preferred_found;
kstat_named_t vdev_mirror_stat_preferred_not_found;
} mirror_stats_t;
static mirror_stats_t mirror_stats = {
/* New I/O follows directly the last I/O */
{ "rotating_linear", KSTAT_DATA_UINT64 },
/* New I/O is within zfs_vdev_mirror_rotating_seek_offset of the last */
{ "rotating_offset", KSTAT_DATA_UINT64 },
/* New I/O requires random seek */
{ "rotating_seek", KSTAT_DATA_UINT64 },
/* New I/O follows directly the last I/O (nonrot) */
{ "non_rotating_linear", KSTAT_DATA_UINT64 },
/* New I/O requires random seek (nonrot) */
{ "non_rotating_seek", KSTAT_DATA_UINT64 },
/* Preferred child vdev found */
{ "preferred_found", KSTAT_DATA_UINT64 },
/* Preferred child vdev not found or equal load */
{ "preferred_not_found", KSTAT_DATA_UINT64 },
};
#define MIRROR_STAT(stat) (mirror_stats.stat.value.ui64)
#define MIRROR_INCR(stat, val) atomic_add_64(&MIRROR_STAT(stat), val)
#define MIRROR_BUMP(stat) MIRROR_INCR(stat, 1)
void
vdev_mirror_stat_init(void)
{
mirror_ksp = kstat_create("zfs", 0, "vdev_mirror_stats",
"misc", KSTAT_TYPE_NAMED,
sizeof (mirror_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
if (mirror_ksp != NULL) {
mirror_ksp->ks_data = &mirror_stats;
kstat_install(mirror_ksp);
}
}
void
vdev_mirror_stat_fini(void)
{
if (mirror_ksp != NULL) {
kstat_delete(mirror_ksp);
mirror_ksp = NULL;
}
}
/*
* Virtual device vector for mirroring.
*/
typedef struct mirror_child {
vdev_t *mc_vd;
uint64_t mc_offset;
int mc_error;
int mc_load;
uint8_t mc_tried;
uint8_t mc_skipped;
uint8_t mc_speculative;
uint8_t mc_rebuilding;
} mirror_child_t;
typedef struct mirror_map {
int *mm_preferred;
int mm_preferred_cnt;
int mm_children;
boolean_t mm_resilvering;
boolean_t mm_rebuilding;
boolean_t mm_root;
mirror_child_t mm_child[];
} mirror_map_t;
-static int vdev_mirror_shift = 21;
+static const int vdev_mirror_shift = 21;
/*
* The load configuration settings below are tuned by default for
* the case where all devices are of the same rotational type.
*
* If there is a mixture of rotating and non-rotating media, setting
* zfs_vdev_mirror_non_rotating_seek_inc to 0 may well provide better results
* as it will direct more reads to the non-rotating vdevs which are more likely
* to have a higher performance.
*/
/* Rotating media load calculation configuration. */
static int zfs_vdev_mirror_rotating_inc = 0;
static int zfs_vdev_mirror_rotating_seek_inc = 5;
static int zfs_vdev_mirror_rotating_seek_offset = 1 * 1024 * 1024;
/* Non-rotating media load calculation configuration. */
static int zfs_vdev_mirror_non_rotating_inc = 0;
static int zfs_vdev_mirror_non_rotating_seek_inc = 1;
static inline size_t
vdev_mirror_map_size(int children)
{
return (offsetof(mirror_map_t, mm_child[children]) +
sizeof (int) * children);
}
static inline mirror_map_t *
vdev_mirror_map_alloc(int children, boolean_t resilvering, boolean_t root)
{
mirror_map_t *mm;
mm = kmem_zalloc(vdev_mirror_map_size(children), KM_SLEEP);
mm->mm_children = children;
mm->mm_resilvering = resilvering;
mm->mm_root = root;
mm->mm_preferred = (int *)((uintptr_t)mm +
offsetof(mirror_map_t, mm_child[children]));
return (mm);
}
static void
vdev_mirror_map_free(zio_t *zio)
{
mirror_map_t *mm = zio->io_vsd;
kmem_free(mm, vdev_mirror_map_size(mm->mm_children));
}
static const zio_vsd_ops_t vdev_mirror_vsd_ops = {
.vsd_free = vdev_mirror_map_free,
};
static int
vdev_mirror_load(mirror_map_t *mm, vdev_t *vd, uint64_t zio_offset)
{
uint64_t last_offset;
int64_t offset_diff;
int load;
/* All DVAs have equal weight at the root. */
if (mm->mm_root)
return (INT_MAX);
/*
* We don't return INT_MAX if the device is resilvering i.e.
* vdev_resilver_txg != 0 as when tested performance was slightly
* worse overall when resilvering with compared to without.
*/
/* Fix zio_offset for leaf vdevs */
if (vd->vdev_ops->vdev_op_leaf)
zio_offset += VDEV_LABEL_START_SIZE;
/* Standard load based on pending queue length. */
load = vdev_queue_length(vd);
last_offset = vdev_queue_last_offset(vd);
if (vd->vdev_nonrot) {
/* Non-rotating media. */
if (last_offset == zio_offset) {
MIRROR_BUMP(vdev_mirror_stat_non_rotating_linear);
return (load + zfs_vdev_mirror_non_rotating_inc);
}
/*
* Apply a seek penalty even for non-rotating devices as
* sequential I/O's can be aggregated into fewer operations on
* the device, thus avoiding unnecessary per-command overhead
* and boosting performance.
*/
MIRROR_BUMP(vdev_mirror_stat_non_rotating_seek);
return (load + zfs_vdev_mirror_non_rotating_seek_inc);
}
/* Rotating media I/O's which directly follow the last I/O. */
if (last_offset == zio_offset) {
MIRROR_BUMP(vdev_mirror_stat_rotating_linear);
return (load + zfs_vdev_mirror_rotating_inc);
}
/*
* Apply half the seek increment to I/O's within seek offset
* of the last I/O issued to this vdev as they should incur less
* of a seek increment.
*/
offset_diff = (int64_t)(last_offset - zio_offset);
if (ABS(offset_diff) < zfs_vdev_mirror_rotating_seek_offset) {
MIRROR_BUMP(vdev_mirror_stat_rotating_offset);
return (load + (zfs_vdev_mirror_rotating_seek_inc / 2));
}
/* Apply the full seek increment to all other I/O's. */
MIRROR_BUMP(vdev_mirror_stat_rotating_seek);
return (load + zfs_vdev_mirror_rotating_seek_inc);
}
static boolean_t
vdev_mirror_rebuilding(vdev_t *vd)
{
if (vd->vdev_ops->vdev_op_leaf && vd->vdev_rebuild_txg)
return (B_TRUE);
for (int i = 0; i < vd->vdev_children; i++) {
if (vdev_mirror_rebuilding(vd->vdev_child[i])) {
return (B_TRUE);
}
}
return (B_FALSE);
}
/*
* Avoid inlining the function to keep vdev_mirror_io_start(), which
* is this functions only caller, as small as possible on the stack.
*/
noinline static mirror_map_t *
vdev_mirror_map_init(zio_t *zio)
{
mirror_map_t *mm = NULL;
mirror_child_t *mc;
vdev_t *vd = zio->io_vd;
int c;
if (vd == NULL) {
dva_t *dva = zio->io_bp->blk_dva;
spa_t *spa = zio->io_spa;
dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;
dva_t dva_copy[SPA_DVAS_PER_BP];
/*
* The sequential scrub code sorts and issues all DVAs
* of a bp separately. Each of these IOs includes all
* original DVA copies so that repairs can be performed
* in the event of an error, but we only actually want
* to check the first DVA since the others will be
* checked by their respective sorted IOs. Only if we
* hit an error will we try all DVAs upon retrying.
*
* Note: This check is safe even if the user switches
* from a legacy scrub to a sequential one in the middle
* of processing, since scn_is_sorted isn't updated until
* all outstanding IOs from the previous scrub pass
* complete.
*/
if ((zio->io_flags & ZIO_FLAG_SCRUB) &&
!(zio->io_flags & ZIO_FLAG_IO_RETRY) &&
dsl_scan_scrubbing(spa->spa_dsl_pool) &&
scn->scn_is_sorted) {
c = 1;
} else {
c = BP_GET_NDVAS(zio->io_bp);
}
/*
* If the pool cannot be written to, then infer that some
* DVAs might be invalid or point to vdevs that do not exist.
* We skip them.
*/
if (!spa_writeable(spa)) {
ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ);
int j = 0;
for (int i = 0; i < c; i++) {
if (zfs_dva_valid(spa, &dva[i], zio->io_bp))
dva_copy[j++] = dva[i];
}
if (j == 0) {
zio->io_vsd = NULL;
zio->io_error = ENXIO;
return (NULL);
}
if (j < c) {
dva = dva_copy;
c = j;
}
}
mm = vdev_mirror_map_alloc(c, B_FALSE, B_TRUE);
for (c = 0; c < mm->mm_children; c++) {
mc = &mm->mm_child[c];
mc->mc_vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[c]));
mc->mc_offset = DVA_GET_OFFSET(&dva[c]);
if (mc->mc_vd == NULL) {
kmem_free(mm, vdev_mirror_map_size(
mm->mm_children));
zio->io_vsd = NULL;
zio->io_error = ENXIO;
return (NULL);
}
}
} else {
/*
* If we are resilvering, then we should handle scrub reads
* differently; we shouldn't issue them to the resilvering
* device because it might not have those blocks.
*
* We are resilvering iff:
* 1) We are a replacing vdev (ie our name is "replacing-1" or
* "spare-1" or something like that), and
* 2) The pool is currently being resilvered.
*
* We cannot simply check vd->vdev_resilver_txg, because it's
* not set in this path.
*
* Nor can we just check our vdev_ops; there are cases (such as
* when a user types "zpool replace pool odev spare_dev" and
* spare_dev is in the spare list, or when a spare device is
* automatically used to replace a DEGRADED device) when
* resilvering is complete but both the original vdev and the
* spare vdev remain in the pool. That behavior is intentional.
* It helps implement the policy that a spare should be
* automatically removed from the pool after the user replaces
* the device that originally failed.
*
* If a spa load is in progress, then spa_dsl_pool may be
* uninitialized. But we shouldn't be resilvering during a spa
* load anyway.
*/
boolean_t replacing = (vd->vdev_ops == &vdev_replacing_ops ||
vd->vdev_ops == &vdev_spare_ops) &&
spa_load_state(vd->vdev_spa) == SPA_LOAD_NONE &&
dsl_scan_resilvering(vd->vdev_spa->spa_dsl_pool);
mm = vdev_mirror_map_alloc(vd->vdev_children, replacing,
B_FALSE);
for (c = 0; c < mm->mm_children; c++) {
mc = &mm->mm_child[c];
mc->mc_vd = vd->vdev_child[c];
mc->mc_offset = zio->io_offset;
if (vdev_mirror_rebuilding(mc->mc_vd))
mm->mm_rebuilding = mc->mc_rebuilding = B_TRUE;
}
}
return (mm);
}
static int
vdev_mirror_open(vdev_t *vd, uint64_t *asize, uint64_t *max_asize,
uint64_t *logical_ashift, uint64_t *physical_ashift)
{
int numerrors = 0;
int lasterror = 0;
if (vd->vdev_children == 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
vdev_open_children(vd);
for (int c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
if (cvd->vdev_open_error) {
lasterror = cvd->vdev_open_error;
numerrors++;
continue;
}
*asize = MIN(*asize - 1, cvd->vdev_asize - 1) + 1;
*max_asize = MIN(*max_asize - 1, cvd->vdev_max_asize - 1) + 1;
*logical_ashift = MAX(*logical_ashift, cvd->vdev_ashift);
*physical_ashift = MAX(*physical_ashift,
cvd->vdev_physical_ashift);
}
if (numerrors == vd->vdev_children) {
if (vdev_children_are_offline(vd))
vd->vdev_stat.vs_aux = VDEV_AUX_CHILDREN_OFFLINE;
else
vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
return (lasterror);
}
return (0);
}
static void
vdev_mirror_close(vdev_t *vd)
{
for (int c = 0; c < vd->vdev_children; c++)
vdev_close(vd->vdev_child[c]);
}
static void
vdev_mirror_child_done(zio_t *zio)
{
mirror_child_t *mc = zio->io_private;
mc->mc_error = zio->io_error;
mc->mc_tried = 1;
mc->mc_skipped = 0;
}
static void
vdev_mirror_scrub_done(zio_t *zio)
{
mirror_child_t *mc = zio->io_private;
if (zio->io_error == 0) {
zio_t *pio;
zio_link_t *zl = NULL;
mutex_enter(&zio->io_lock);
while ((pio = zio_walk_parents(zio, &zl)) != NULL) {
mutex_enter(&pio->io_lock);
ASSERT3U(zio->io_size, >=, pio->io_size);
abd_copy(pio->io_abd, zio->io_abd, pio->io_size);
mutex_exit(&pio->io_lock);
}
mutex_exit(&zio->io_lock);
}
abd_free(zio->io_abd);
mc->mc_error = zio->io_error;
mc->mc_tried = 1;
mc->mc_skipped = 0;
}
/*
* Check the other, lower-index DVAs to see if they're on the same
* vdev as the child we picked. If they are, use them since they
* are likely to have been allocated from the primary metaslab in
* use at the time, and hence are more likely to have locality with
* single-copy data.
*/
static int
vdev_mirror_dva_select(zio_t *zio, int p)
{
dva_t *dva = zio->io_bp->blk_dva;
mirror_map_t *mm = zio->io_vsd;
int preferred;
int c;
preferred = mm->mm_preferred[p];
for (p--; p >= 0; p--) {
c = mm->mm_preferred[p];
if (DVA_GET_VDEV(&dva[c]) == DVA_GET_VDEV(&dva[preferred]))
preferred = c;
}
return (preferred);
}
static int
vdev_mirror_preferred_child_randomize(zio_t *zio)
{
mirror_map_t *mm = zio->io_vsd;
int p;
if (mm->mm_root) {
p = random_in_range(mm->mm_preferred_cnt);
return (vdev_mirror_dva_select(zio, p));
}
/*
* To ensure we don't always favour the first matching vdev,
* which could lead to wear leveling issues on SSD's, we
* use the I/O offset as a pseudo random seed into the vdevs
* which have the lowest load.
*/
p = (zio->io_offset >> vdev_mirror_shift) % mm->mm_preferred_cnt;
return (mm->mm_preferred[p]);
}
static boolean_t
vdev_mirror_child_readable(mirror_child_t *mc)
{
vdev_t *vd = mc->mc_vd;
if (vd->vdev_top != NULL && vd->vdev_top->vdev_ops == &vdev_draid_ops)
return (vdev_draid_readable(vd, mc->mc_offset));
else
return (vdev_readable(vd));
}
static boolean_t
vdev_mirror_child_missing(mirror_child_t *mc, uint64_t txg, uint64_t size)
{
vdev_t *vd = mc->mc_vd;
if (vd->vdev_top != NULL && vd->vdev_top->vdev_ops == &vdev_draid_ops)
return (vdev_draid_missing(vd, mc->mc_offset, txg, size));
else
return (vdev_dtl_contains(vd, DTL_MISSING, txg, size));
}
/*
* Try to find a vdev whose DTL doesn't contain the block we want to read
* preferring vdevs based on determined load. If we can't, try the read on
* any vdev we haven't already tried.
*
* Distributed spares are an exception to the above load rule. They are
* always preferred in order to detect gaps in the distributed spare which
* are created when another disk in the dRAID fails. In order to restore
* redundancy those gaps must be read to trigger the required repair IO.
*/
static int
vdev_mirror_child_select(zio_t *zio)
{
mirror_map_t *mm = zio->io_vsd;
uint64_t txg = zio->io_txg;
int c, lowest_load;
ASSERT(zio->io_bp == NULL || BP_PHYSICAL_BIRTH(zio->io_bp) == txg);
lowest_load = INT_MAX;
mm->mm_preferred_cnt = 0;
for (c = 0; c < mm->mm_children; c++) {
mirror_child_t *mc;
mc = &mm->mm_child[c];
if (mc->mc_tried || mc->mc_skipped)
continue;
if (mc->mc_vd == NULL ||
!vdev_mirror_child_readable(mc)) {
mc->mc_error = SET_ERROR(ENXIO);
mc->mc_tried = 1; /* don't even try */
mc->mc_skipped = 1;
continue;
}
if (vdev_mirror_child_missing(mc, txg, 1)) {
mc->mc_error = SET_ERROR(ESTALE);
mc->mc_skipped = 1;
mc->mc_speculative = 1;
continue;
}
if (mc->mc_vd->vdev_ops == &vdev_draid_spare_ops) {
mm->mm_preferred[0] = c;
mm->mm_preferred_cnt = 1;
break;
}
mc->mc_load = vdev_mirror_load(mm, mc->mc_vd, mc->mc_offset);
if (mc->mc_load > lowest_load)
continue;
if (mc->mc_load < lowest_load) {
lowest_load = mc->mc_load;
mm->mm_preferred_cnt = 0;
}
mm->mm_preferred[mm->mm_preferred_cnt] = c;
mm->mm_preferred_cnt++;
}
if (mm->mm_preferred_cnt == 1) {
MIRROR_BUMP(vdev_mirror_stat_preferred_found);
return (mm->mm_preferred[0]);
}
if (mm->mm_preferred_cnt > 1) {
MIRROR_BUMP(vdev_mirror_stat_preferred_not_found);
return (vdev_mirror_preferred_child_randomize(zio));
}
/*
* Every device is either missing or has this txg in its DTL.
* Look for any child we haven't already tried before giving up.
*/
for (c = 0; c < mm->mm_children; c++) {
if (!mm->mm_child[c].mc_tried)
return (c);
}
/*
* Every child failed. There's no place left to look.
*/
return (-1);
}
static void
vdev_mirror_io_start(zio_t *zio)
{
mirror_map_t *mm;
mirror_child_t *mc;
int c, children;
mm = vdev_mirror_map_init(zio);
zio->io_vsd = mm;
zio->io_vsd_ops = &vdev_mirror_vsd_ops;
if (mm == NULL) {
ASSERT(!spa_trust_config(zio->io_spa));
ASSERT(zio->io_type == ZIO_TYPE_READ);
zio_execute(zio);
return;
}
if (zio->io_type == ZIO_TYPE_READ) {
if (zio->io_bp != NULL &&
(zio->io_flags & ZIO_FLAG_SCRUB) && !mm->mm_resilvering) {
/*
* For scrubbing reads (if we can verify the
* checksum here, as indicated by io_bp being
* non-NULL) we need to allocate a read buffer for
* each child and issue reads to all children. If
* any child succeeds, it will copy its data into
* zio->io_data in vdev_mirror_scrub_done.
*/
for (c = 0; c < mm->mm_children; c++) {
mc = &mm->mm_child[c];
/* Don't issue ZIOs to offline children */
if (!vdev_mirror_child_readable(mc)) {
mc->mc_error = SET_ERROR(ENXIO);
mc->mc_tried = 1;
mc->mc_skipped = 1;
continue;
}
zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
mc->mc_vd, mc->mc_offset,
abd_alloc_sametype(zio->io_abd,
zio->io_size), zio->io_size,
zio->io_type, zio->io_priority, 0,
vdev_mirror_scrub_done, mc));
}
zio_execute(zio);
return;
}
/*
* For normal reads just pick one child.
*/
c = vdev_mirror_child_select(zio);
children = (c >= 0);
} else {
ASSERT(zio->io_type == ZIO_TYPE_WRITE);
/*
* Writes go to all children.
*/
c = 0;
children = mm->mm_children;
}
while (children--) {
mc = &mm->mm_child[c];
c++;
/*
* When sequentially resilvering only issue write repair
* IOs to the vdev which is being rebuilt since performance
* is limited by the slowest child. This is an issue for
* faster replacement devices such as distributed spares.
*/
if ((zio->io_priority == ZIO_PRIORITY_REBUILD) &&
(zio->io_flags & ZIO_FLAG_IO_REPAIR) &&
!(zio->io_flags & ZIO_FLAG_SCRUB) &&
mm->mm_rebuilding && !mc->mc_rebuilding) {
continue;
}
zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
mc->mc_vd, mc->mc_offset, zio->io_abd, zio->io_size,
zio->io_type, zio->io_priority, 0,
vdev_mirror_child_done, mc));
}
zio_execute(zio);
}
static int
vdev_mirror_worst_error(mirror_map_t *mm)
{
int error[2] = { 0, 0 };
for (int c = 0; c < mm->mm_children; c++) {
mirror_child_t *mc = &mm->mm_child[c];
int s = mc->mc_speculative;
error[s] = zio_worst_error(error[s], mc->mc_error);
}
return (error[0] ? error[0] : error[1]);
}
static void
vdev_mirror_io_done(zio_t *zio)
{
mirror_map_t *mm = zio->io_vsd;
mirror_child_t *mc;
int c;
int good_copies = 0;
int unexpected_errors = 0;
if (mm == NULL)
return;
for (c = 0; c < mm->mm_children; c++) {
mc = &mm->mm_child[c];
if (mc->mc_error) {
if (!mc->mc_skipped)
unexpected_errors++;
} else if (mc->mc_tried) {
good_copies++;
}
}
if (zio->io_type == ZIO_TYPE_WRITE) {
/*
* XXX -- for now, treat partial writes as success.
*
* Now that we support write reallocation, it would be better
* to treat partial failure as real failure unless there are
* no non-degraded top-level vdevs left, and not update DTLs
* if we intend to reallocate.
*/
/* XXPOLICY */
if (good_copies != mm->mm_children) {
/*
* Always require at least one good copy.
*
* For ditto blocks (io_vd == NULL), require
* all copies to be good.
*
* XXX -- for replacing vdevs, there's no great answer.
* If the old device is really dead, we may not even
* be able to access it -- so we only want to
* require good writes to the new device. But if
* the new device turns out to be flaky, we want
* to be able to detach it -- which requires all
* writes to the old device to have succeeded.
*/
if (good_copies == 0 || zio->io_vd == NULL)
zio->io_error = vdev_mirror_worst_error(mm);
}
return;
}
ASSERT(zio->io_type == ZIO_TYPE_READ);
/*
* If we don't have a good copy yet, keep trying other children.
*/
/* XXPOLICY */
if (good_copies == 0 && (c = vdev_mirror_child_select(zio)) != -1) {
ASSERT(c >= 0 && c < mm->mm_children);
mc = &mm->mm_child[c];
zio_vdev_io_redone(zio);
zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
mc->mc_vd, mc->mc_offset, zio->io_abd, zio->io_size,
ZIO_TYPE_READ, zio->io_priority, 0,
vdev_mirror_child_done, mc));
return;
}
/* XXPOLICY */
if (good_copies == 0) {
zio->io_error = vdev_mirror_worst_error(mm);
ASSERT(zio->io_error != 0);
}
if (good_copies && spa_writeable(zio->io_spa) &&
(unexpected_errors ||
(zio->io_flags & ZIO_FLAG_RESILVER) ||
((zio->io_flags & ZIO_FLAG_SCRUB) && mm->mm_resilvering))) {
/*
* Use the good data we have in hand to repair damaged children.
*/
for (c = 0; c < mm->mm_children; c++) {
/*
* Don't rewrite known good children.
* Not only is it unnecessary, it could
* actually be harmful: if the system lost
* power while rewriting the only good copy,
* there would be no good copies left!
*/
mc = &mm->mm_child[c];
if (mc->mc_error == 0) {
vdev_ops_t *ops = mc->mc_vd->vdev_ops;
if (mc->mc_tried)
continue;
/*
* We didn't try this child. We need to
* repair it if:
* 1. it's a scrub (in which case we have
* tried everything that was healthy)
* - or -
* 2. it's an indirect or distributed spare
* vdev (in which case it could point to any
* other vdev, which might have a bad DTL)
* - or -
* 3. the DTL indicates that this data is
* missing from this vdev
*/
if (!(zio->io_flags & ZIO_FLAG_SCRUB) &&
ops != &vdev_indirect_ops &&
ops != &vdev_draid_spare_ops &&
!vdev_dtl_contains(mc->mc_vd, DTL_PARTIAL,
zio->io_txg, 1))
continue;
mc->mc_error = SET_ERROR(ESTALE);
}
zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
mc->mc_vd, mc->mc_offset,
zio->io_abd, zio->io_size, ZIO_TYPE_WRITE,
zio->io_priority == ZIO_PRIORITY_REBUILD ?
ZIO_PRIORITY_REBUILD : ZIO_PRIORITY_ASYNC_WRITE,
ZIO_FLAG_IO_REPAIR | (unexpected_errors ?
ZIO_FLAG_SELF_HEAL : 0), NULL, NULL));
}
}
}
static void
vdev_mirror_state_change(vdev_t *vd, int faulted, int degraded)
{
if (faulted == vd->vdev_children) {
if (vdev_children_are_offline(vd)) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_OFFLINE,
VDEV_AUX_CHILDREN_OFFLINE);
} else {
vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_NO_REPLICAS);
}
} else if (degraded + faulted != 0) {
vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE);
} else {
vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE);
}
}
/*
* Return the maximum asize for a rebuild zio in the provided range.
*/
static uint64_t
vdev_mirror_rebuild_asize(vdev_t *vd, uint64_t start, uint64_t asize,
uint64_t max_segment)
{
(void) start;
uint64_t psize = MIN(P2ROUNDUP(max_segment, 1 << vd->vdev_ashift),
SPA_MAXBLOCKSIZE);
return (MIN(asize, vdev_psize_to_asize(vd, psize)));
}
vdev_ops_t vdev_mirror_ops = {
.vdev_op_init = NULL,
.vdev_op_fini = NULL,
.vdev_op_open = vdev_mirror_open,
.vdev_op_close = vdev_mirror_close,
.vdev_op_asize = vdev_default_asize,
.vdev_op_min_asize = vdev_default_min_asize,
.vdev_op_min_alloc = NULL,
.vdev_op_io_start = vdev_mirror_io_start,
.vdev_op_io_done = vdev_mirror_io_done,
.vdev_op_state_change = vdev_mirror_state_change,
.vdev_op_need_resilver = vdev_default_need_resilver,
.vdev_op_hold = NULL,
.vdev_op_rele = NULL,
.vdev_op_remap = NULL,
.vdev_op_xlate = vdev_default_xlate,
.vdev_op_rebuild_asize = vdev_mirror_rebuild_asize,
.vdev_op_metaslab_init = NULL,
.vdev_op_config_generate = NULL,
.vdev_op_nparity = NULL,
.vdev_op_ndisks = NULL,
.vdev_op_type = VDEV_TYPE_MIRROR, /* name of this vdev type */
.vdev_op_leaf = B_FALSE /* not a leaf vdev */
};
vdev_ops_t vdev_replacing_ops = {
.vdev_op_init = NULL,
.vdev_op_fini = NULL,
.vdev_op_open = vdev_mirror_open,
.vdev_op_close = vdev_mirror_close,
.vdev_op_asize = vdev_default_asize,
.vdev_op_min_asize = vdev_default_min_asize,
.vdev_op_min_alloc = NULL,
.vdev_op_io_start = vdev_mirror_io_start,
.vdev_op_io_done = vdev_mirror_io_done,
.vdev_op_state_change = vdev_mirror_state_change,
.vdev_op_need_resilver = vdev_default_need_resilver,
.vdev_op_hold = NULL,
.vdev_op_rele = NULL,
.vdev_op_remap = NULL,
.vdev_op_xlate = vdev_default_xlate,
.vdev_op_rebuild_asize = vdev_mirror_rebuild_asize,
.vdev_op_metaslab_init = NULL,
.vdev_op_config_generate = NULL,
.vdev_op_nparity = NULL,
.vdev_op_ndisks = NULL,
.vdev_op_type = VDEV_TYPE_REPLACING, /* name of this vdev type */
.vdev_op_leaf = B_FALSE /* not a leaf vdev */
};
vdev_ops_t vdev_spare_ops = {
.vdev_op_init = NULL,
.vdev_op_fini = NULL,
.vdev_op_open = vdev_mirror_open,
.vdev_op_close = vdev_mirror_close,
.vdev_op_asize = vdev_default_asize,
.vdev_op_min_asize = vdev_default_min_asize,
.vdev_op_min_alloc = NULL,
.vdev_op_io_start = vdev_mirror_io_start,
.vdev_op_io_done = vdev_mirror_io_done,
.vdev_op_state_change = vdev_mirror_state_change,
.vdev_op_need_resilver = vdev_default_need_resilver,
.vdev_op_hold = NULL,
.vdev_op_rele = NULL,
.vdev_op_remap = NULL,
.vdev_op_xlate = vdev_default_xlate,
.vdev_op_rebuild_asize = vdev_mirror_rebuild_asize,
.vdev_op_metaslab_init = NULL,
.vdev_op_config_generate = NULL,
.vdev_op_nparity = NULL,
.vdev_op_ndisks = NULL,
.vdev_op_type = VDEV_TYPE_SPARE, /* name of this vdev type */
.vdev_op_leaf = B_FALSE /* not a leaf vdev */
};
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_vdev_mirror, zfs_vdev_mirror_, rotating_inc, INT, ZMOD_RW,
"Rotating media load increment for non-seeking I/O's");
ZFS_MODULE_PARAM(zfs_vdev_mirror, zfs_vdev_mirror_, rotating_seek_inc, INT, ZMOD_RW,
"Rotating media load increment for seeking I/O's");
ZFS_MODULE_PARAM(zfs_vdev_mirror, zfs_vdev_mirror_, rotating_seek_offset, INT, ZMOD_RW,
"Offset in bytes from the last I/O which triggers "
"a reduced rotating media seek increment");
ZFS_MODULE_PARAM(zfs_vdev_mirror, zfs_vdev_mirror_, non_rotating_inc, INT, ZMOD_RW,
"Non-rotating media load increment for non-seeking I/O's");
ZFS_MODULE_PARAM(zfs_vdev_mirror, zfs_vdev_mirror_, non_rotating_seek_inc, INT, ZMOD_RW,
"Non-rotating media load increment for seeking I/O's");
/* END CSTYLED */
diff --git a/module/zfs/vdev_queue.c b/module/zfs/vdev_queue.c
index af612ba9c9c6..ae0a322f86c7 100644
--- a/module/zfs/vdev_queue.c
+++ b/module/zfs/vdev_queue.c
@@ -1,1121 +1,1121 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/vdev_impl.h>
#include <sys/spa_impl.h>
#include <sys/zio.h>
#include <sys/avl.h>
#include <sys/dsl_pool.h>
#include <sys/metaslab_impl.h>
#include <sys/spa.h>
#include <sys/abd.h>
/*
* ZFS I/O Scheduler
* ---------------
*
* ZFS issues I/O operations to leaf vdevs to satisfy and complete zios. The
* I/O scheduler determines when and in what order those operations are
* issued. The I/O scheduler divides operations into five I/O classes
* prioritized in the following order: sync read, sync write, async read,
* async write, and scrub/resilver. Each queue defines the minimum and
* maximum number of concurrent operations that may be issued to the device.
* In addition, the device has an aggregate maximum. Note that the sum of the
* per-queue minimums must not exceed the aggregate maximum. If the
* sum of the per-queue maximums exceeds the aggregate maximum, then the
* number of active i/os may reach zfs_vdev_max_active, in which case no
* further i/os will be issued regardless of whether all per-queue
* minimums have been met.
*
* For many physical devices, throughput increases with the number of
* concurrent operations, but latency typically suffers. Further, physical
* devices typically have a limit at which more concurrent operations have no
* effect on throughput or can actually cause it to decrease.
*
* The scheduler selects the next operation to issue by first looking for an
* I/O class whose minimum has not been satisfied. Once all are satisfied and
* the aggregate maximum has not been hit, the scheduler looks for classes
* whose maximum has not been satisfied. Iteration through the I/O classes is
* done in the order specified above. No further operations are issued if the
* aggregate maximum number of concurrent operations has been hit or if there
* are no operations queued for an I/O class that has not hit its maximum.
* Every time an i/o is queued or an operation completes, the I/O scheduler
* looks for new operations to issue.
*
* All I/O classes have a fixed maximum number of outstanding operations
* except for the async write class. Asynchronous writes represent the data
* that is committed to stable storage during the syncing stage for
* transaction groups (see txg.c). Transaction groups enter the syncing state
* periodically so the number of queued async writes will quickly burst up and
* then bleed down to zero. Rather than servicing them as quickly as possible,
* the I/O scheduler changes the maximum number of active async write i/os
* according to the amount of dirty data in the pool (see dsl_pool.c). Since
* both throughput and latency typically increase with the number of
* concurrent operations issued to physical devices, reducing the burstiness
* in the number of concurrent operations also stabilizes the response time of
* operations from other -- and in particular synchronous -- queues. In broad
* strokes, the I/O scheduler will issue more concurrent operations from the
* async write queue as there's more dirty data in the pool.
*
* Async Writes
*
* The number of concurrent operations issued for the async write I/O class
* follows a piece-wise linear function defined by a few adjustable points.
*
* | o---------| <-- zfs_vdev_async_write_max_active
* ^ | /^ |
* | | / | |
* active | / | |
* I/O | / | |
* count | / | |
* | / | |
* |------------o | | <-- zfs_vdev_async_write_min_active
* 0|____________^______|_________|
* 0% | | 100% of zfs_dirty_data_max
* | |
* | `-- zfs_vdev_async_write_active_max_dirty_percent
* `--------- zfs_vdev_async_write_active_min_dirty_percent
*
* Until the amount of dirty data exceeds a minimum percentage of the dirty
* data allowed in the pool, the I/O scheduler will limit the number of
* concurrent operations to the minimum. As that threshold is crossed, the
* number of concurrent operations issued increases linearly to the maximum at
* the specified maximum percentage of the dirty data allowed in the pool.
*
* Ideally, the amount of dirty data on a busy pool will stay in the sloped
* part of the function between zfs_vdev_async_write_active_min_dirty_percent
* and zfs_vdev_async_write_active_max_dirty_percent. If it exceeds the
* maximum percentage, this indicates that the rate of incoming data is
* greater than the rate that the backend storage can handle. In this case, we
* must further throttle incoming writes (see dmu_tx_delay() for details).
*/
/*
* The maximum number of i/os active to each device. Ideally, this will be >=
* the sum of each queue's max_active.
*/
uint32_t zfs_vdev_max_active = 1000;
/*
* Per-queue limits on the number of i/os active to each device. If the
* number of active i/os is < zfs_vdev_max_active, then the min_active comes
* into play. We will send min_active from each queue round-robin, and then
* send from queues in the order defined by zio_priority_t up to max_active.
* Some queues have additional mechanisms to limit number of active I/Os in
* addition to min_active and max_active, see below.
*
* In general, smaller max_active's will lead to lower latency of synchronous
* operations. Larger max_active's may lead to higher overall throughput,
* depending on underlying storage.
*
* The ratio of the queues' max_actives determines the balance of performance
* between reads, writes, and scrubs. E.g., increasing
* zfs_vdev_scrub_max_active will cause the scrub or resilver to complete
* more quickly, but reads and writes to have higher latency and lower
* throughput.
*/
-uint32_t zfs_vdev_sync_read_min_active = 10;
-uint32_t zfs_vdev_sync_read_max_active = 10;
-uint32_t zfs_vdev_sync_write_min_active = 10;
-uint32_t zfs_vdev_sync_write_max_active = 10;
-uint32_t zfs_vdev_async_read_min_active = 1;
-uint32_t zfs_vdev_async_read_max_active = 3;
-uint32_t zfs_vdev_async_write_min_active = 2;
-uint32_t zfs_vdev_async_write_max_active = 10;
-uint32_t zfs_vdev_scrub_min_active = 1;
-uint32_t zfs_vdev_scrub_max_active = 3;
-uint32_t zfs_vdev_removal_min_active = 1;
-uint32_t zfs_vdev_removal_max_active = 2;
-uint32_t zfs_vdev_initializing_min_active = 1;
-uint32_t zfs_vdev_initializing_max_active = 1;
-uint32_t zfs_vdev_trim_min_active = 1;
-uint32_t zfs_vdev_trim_max_active = 2;
-uint32_t zfs_vdev_rebuild_min_active = 1;
-uint32_t zfs_vdev_rebuild_max_active = 3;
+static uint32_t zfs_vdev_sync_read_min_active = 10;
+static uint32_t zfs_vdev_sync_read_max_active = 10;
+static uint32_t zfs_vdev_sync_write_min_active = 10;
+static uint32_t zfs_vdev_sync_write_max_active = 10;
+static uint32_t zfs_vdev_async_read_min_active = 1;
+/* */ uint32_t zfs_vdev_async_read_max_active = 3;
+static uint32_t zfs_vdev_async_write_min_active = 2;
+/* */ uint32_t zfs_vdev_async_write_max_active = 10;
+static uint32_t zfs_vdev_scrub_min_active = 1;
+static uint32_t zfs_vdev_scrub_max_active = 3;
+static uint32_t zfs_vdev_removal_min_active = 1;
+static uint32_t zfs_vdev_removal_max_active = 2;
+static uint32_t zfs_vdev_initializing_min_active = 1;
+static uint32_t zfs_vdev_initializing_max_active = 1;
+static uint32_t zfs_vdev_trim_min_active = 1;
+static uint32_t zfs_vdev_trim_max_active = 2;
+static uint32_t zfs_vdev_rebuild_min_active = 1;
+static uint32_t zfs_vdev_rebuild_max_active = 3;
/*
* When the pool has less than zfs_vdev_async_write_active_min_dirty_percent
* dirty data, use zfs_vdev_async_write_min_active. When it has more than
* zfs_vdev_async_write_active_max_dirty_percent, use
* zfs_vdev_async_write_max_active. The value is linearly interpolated
* between min and max.
*/
int zfs_vdev_async_write_active_min_dirty_percent = 30;
int zfs_vdev_async_write_active_max_dirty_percent = 60;
/*
* For non-interactive I/O (scrub, resilver, removal, initialize and rebuild),
* the number of concurrently-active I/O's is limited to *_min_active, unless
* the vdev is "idle". When there are no interactive I/Os active (sync or
* async), and zfs_vdev_nia_delay I/Os have completed since the last
* interactive I/O, then the vdev is considered to be "idle", and the number
* of concurrently-active non-interactive I/O's is increased to *_max_active.
*/
-uint_t zfs_vdev_nia_delay = 5;
+static uint_t zfs_vdev_nia_delay = 5;
/*
* Some HDDs tend to prioritize sequential I/O so high that concurrent
* random I/O latency reaches several seconds. On some HDDs it happens
* even if sequential I/Os are submitted one at a time, and so setting
* *_max_active to 1 does not help. To prevent non-interactive I/Os, like
* scrub, from monopolizing the device no more than zfs_vdev_nia_credit
* I/Os can be sent while there are outstanding incomplete interactive
* I/Os. This enforced wait ensures the HDD services the interactive I/O
* within a reasonable amount of time.
*/
-uint_t zfs_vdev_nia_credit = 5;
+static uint_t zfs_vdev_nia_credit = 5;
/*
* To reduce IOPs, we aggregate small adjacent I/Os into one large I/O.
* For read I/Os, we also aggregate across small adjacency gaps; for writes
* we include spans of optional I/Os to aid aggregation at the disk even when
* they aren't able to help us aggregate at this level.
*/
-int zfs_vdev_aggregation_limit = 1 << 20;
-int zfs_vdev_aggregation_limit_non_rotating = SPA_OLD_MAXBLOCKSIZE;
-int zfs_vdev_read_gap_limit = 32 << 10;
-int zfs_vdev_write_gap_limit = 4 << 10;
+static int zfs_vdev_aggregation_limit = 1 << 20;
+static int zfs_vdev_aggregation_limit_non_rotating = SPA_OLD_MAXBLOCKSIZE;
+static int zfs_vdev_read_gap_limit = 32 << 10;
+static int zfs_vdev_write_gap_limit = 4 << 10;
/*
* Define the queue depth percentage for each top-level. This percentage is
* used in conjunction with zfs_vdev_async_max_active to determine how many
* allocations a specific top-level vdev should handle. Once the queue depth
* reaches zfs_vdev_queue_depth_pct * zfs_vdev_async_write_max_active / 100
* then allocator will stop allocating blocks on that top-level device.
* The default kernel setting is 1000% which will yield 100 allocations per
* device. For userland testing, the default setting is 300% which equates
* to 30 allocations per device.
*/
#ifdef _KERNEL
int zfs_vdev_queue_depth_pct = 1000;
#else
int zfs_vdev_queue_depth_pct = 300;
#endif
/*
* When performing allocations for a given metaslab, we want to make sure that
* there are enough IOs to aggregate together to improve throughput. We want to
* ensure that there are at least 128k worth of IOs that can be aggregated, and
* we assume that the average allocation size is 4k, so we need the queue depth
* to be 32 per allocator to get good aggregation of sequential writes.
*/
int zfs_vdev_def_queue_depth = 32;
/*
* Allow TRIM I/Os to be aggregated. This should normally not be needed since
* TRIM I/O for extents up to zfs_trim_extent_bytes_max (128M) can be submitted
* by the TRIM code in zfs_trim.c.
*/
-int zfs_vdev_aggregate_trim = 0;
+static int zfs_vdev_aggregate_trim = 0;
static int
vdev_queue_offset_compare(const void *x1, const void *x2)
{
const zio_t *z1 = (const zio_t *)x1;
const zio_t *z2 = (const zio_t *)x2;
int cmp = TREE_CMP(z1->io_offset, z2->io_offset);
if (likely(cmp))
return (cmp);
return (TREE_PCMP(z1, z2));
}
static inline avl_tree_t *
vdev_queue_class_tree(vdev_queue_t *vq, zio_priority_t p)
{
return (&vq->vq_class[p].vqc_queued_tree);
}
static inline avl_tree_t *
vdev_queue_type_tree(vdev_queue_t *vq, zio_type_t t)
{
ASSERT(t == ZIO_TYPE_READ || t == ZIO_TYPE_WRITE || t == ZIO_TYPE_TRIM);
if (t == ZIO_TYPE_READ)
return (&vq->vq_read_offset_tree);
else if (t == ZIO_TYPE_WRITE)
return (&vq->vq_write_offset_tree);
else
return (&vq->vq_trim_offset_tree);
}
static int
vdev_queue_timestamp_compare(const void *x1, const void *x2)
{
const zio_t *z1 = (const zio_t *)x1;
const zio_t *z2 = (const zio_t *)x2;
int cmp = TREE_CMP(z1->io_timestamp, z2->io_timestamp);
if (likely(cmp))
return (cmp);
return (TREE_PCMP(z1, z2));
}
static int
vdev_queue_class_min_active(vdev_queue_t *vq, zio_priority_t p)
{
switch (p) {
case ZIO_PRIORITY_SYNC_READ:
return (zfs_vdev_sync_read_min_active);
case ZIO_PRIORITY_SYNC_WRITE:
return (zfs_vdev_sync_write_min_active);
case ZIO_PRIORITY_ASYNC_READ:
return (zfs_vdev_async_read_min_active);
case ZIO_PRIORITY_ASYNC_WRITE:
return (zfs_vdev_async_write_min_active);
case ZIO_PRIORITY_SCRUB:
return (vq->vq_ia_active == 0 ? zfs_vdev_scrub_min_active :
MIN(vq->vq_nia_credit, zfs_vdev_scrub_min_active));
case ZIO_PRIORITY_REMOVAL:
return (vq->vq_ia_active == 0 ? zfs_vdev_removal_min_active :
MIN(vq->vq_nia_credit, zfs_vdev_removal_min_active));
case ZIO_PRIORITY_INITIALIZING:
return (vq->vq_ia_active == 0 ?zfs_vdev_initializing_min_active:
MIN(vq->vq_nia_credit, zfs_vdev_initializing_min_active));
case ZIO_PRIORITY_TRIM:
return (zfs_vdev_trim_min_active);
case ZIO_PRIORITY_REBUILD:
return (vq->vq_ia_active == 0 ? zfs_vdev_rebuild_min_active :
MIN(vq->vq_nia_credit, zfs_vdev_rebuild_min_active));
default:
panic("invalid priority %u", p);
return (0);
}
}
static int
vdev_queue_max_async_writes(spa_t *spa)
{
int writes;
uint64_t dirty = 0;
dsl_pool_t *dp = spa_get_dsl(spa);
uint64_t min_bytes = zfs_dirty_data_max *
zfs_vdev_async_write_active_min_dirty_percent / 100;
uint64_t max_bytes = zfs_dirty_data_max *
zfs_vdev_async_write_active_max_dirty_percent / 100;
/*
* Async writes may occur before the assignment of the spa's
* dsl_pool_t if a self-healing zio is issued prior to the
* completion of dmu_objset_open_impl().
*/
if (dp == NULL)
return (zfs_vdev_async_write_max_active);
/*
* Sync tasks correspond to interactive user actions. To reduce the
* execution time of those actions we push data out as fast as possible.
*/
dirty = dp->dp_dirty_total;
if (dirty > max_bytes || spa_has_pending_synctask(spa))
return (zfs_vdev_async_write_max_active);
if (dirty < min_bytes)
return (zfs_vdev_async_write_min_active);
/*
* linear interpolation:
* slope = (max_writes - min_writes) / (max_bytes - min_bytes)
* move right by min_bytes
* move up by min_writes
*/
writes = (dirty - min_bytes) *
(zfs_vdev_async_write_max_active -
zfs_vdev_async_write_min_active) /
(max_bytes - min_bytes) +
zfs_vdev_async_write_min_active;
ASSERT3U(writes, >=, zfs_vdev_async_write_min_active);
ASSERT3U(writes, <=, zfs_vdev_async_write_max_active);
return (writes);
}
static int
vdev_queue_class_max_active(spa_t *spa, vdev_queue_t *vq, zio_priority_t p)
{
switch (p) {
case ZIO_PRIORITY_SYNC_READ:
return (zfs_vdev_sync_read_max_active);
case ZIO_PRIORITY_SYNC_WRITE:
return (zfs_vdev_sync_write_max_active);
case ZIO_PRIORITY_ASYNC_READ:
return (zfs_vdev_async_read_max_active);
case ZIO_PRIORITY_ASYNC_WRITE:
return (vdev_queue_max_async_writes(spa));
case ZIO_PRIORITY_SCRUB:
if (vq->vq_ia_active > 0) {
return (MIN(vq->vq_nia_credit,
zfs_vdev_scrub_min_active));
} else if (vq->vq_nia_credit < zfs_vdev_nia_delay)
return (MAX(1, zfs_vdev_scrub_min_active));
return (zfs_vdev_scrub_max_active);
case ZIO_PRIORITY_REMOVAL:
if (vq->vq_ia_active > 0) {
return (MIN(vq->vq_nia_credit,
zfs_vdev_removal_min_active));
} else if (vq->vq_nia_credit < zfs_vdev_nia_delay)
return (MAX(1, zfs_vdev_removal_min_active));
return (zfs_vdev_removal_max_active);
case ZIO_PRIORITY_INITIALIZING:
if (vq->vq_ia_active > 0) {
return (MIN(vq->vq_nia_credit,
zfs_vdev_initializing_min_active));
} else if (vq->vq_nia_credit < zfs_vdev_nia_delay)
return (MAX(1, zfs_vdev_initializing_min_active));
return (zfs_vdev_initializing_max_active);
case ZIO_PRIORITY_TRIM:
return (zfs_vdev_trim_max_active);
case ZIO_PRIORITY_REBUILD:
if (vq->vq_ia_active > 0) {
return (MIN(vq->vq_nia_credit,
zfs_vdev_rebuild_min_active));
} else if (vq->vq_nia_credit < zfs_vdev_nia_delay)
return (MAX(1, zfs_vdev_rebuild_min_active));
return (zfs_vdev_rebuild_max_active);
default:
panic("invalid priority %u", p);
return (0);
}
}
/*
* Return the i/o class to issue from, or ZIO_PRIORITY_NUM_QUEUEABLE if
* there is no eligible class.
*/
static zio_priority_t
vdev_queue_class_to_issue(vdev_queue_t *vq)
{
spa_t *spa = vq->vq_vdev->vdev_spa;
zio_priority_t p, n;
if (avl_numnodes(&vq->vq_active_tree) >= zfs_vdev_max_active)
return (ZIO_PRIORITY_NUM_QUEUEABLE);
/*
* Find a queue that has not reached its minimum # outstanding i/os.
* Do round-robin to reduce starvation due to zfs_vdev_max_active
* and vq_nia_credit limits.
*/
for (n = 0; n < ZIO_PRIORITY_NUM_QUEUEABLE; n++) {
p = (vq->vq_last_prio + n + 1) % ZIO_PRIORITY_NUM_QUEUEABLE;
if (avl_numnodes(vdev_queue_class_tree(vq, p)) > 0 &&
vq->vq_class[p].vqc_active <
vdev_queue_class_min_active(vq, p)) {
vq->vq_last_prio = p;
return (p);
}
}
/*
* If we haven't found a queue, look for one that hasn't reached its
* maximum # outstanding i/os.
*/
for (p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) {
if (avl_numnodes(vdev_queue_class_tree(vq, p)) > 0 &&
vq->vq_class[p].vqc_active <
vdev_queue_class_max_active(spa, vq, p)) {
vq->vq_last_prio = p;
return (p);
}
}
/* No eligible queued i/os */
return (ZIO_PRIORITY_NUM_QUEUEABLE);
}
void
vdev_queue_init(vdev_t *vd)
{
vdev_queue_t *vq = &vd->vdev_queue;
zio_priority_t p;
mutex_init(&vq->vq_lock, NULL, MUTEX_DEFAULT, NULL);
vq->vq_vdev = vd;
taskq_init_ent(&vd->vdev_queue.vq_io_search.io_tqent);
avl_create(&vq->vq_active_tree, vdev_queue_offset_compare,
sizeof (zio_t), offsetof(struct zio, io_queue_node));
avl_create(vdev_queue_type_tree(vq, ZIO_TYPE_READ),
vdev_queue_offset_compare, sizeof (zio_t),
offsetof(struct zio, io_offset_node));
avl_create(vdev_queue_type_tree(vq, ZIO_TYPE_WRITE),
vdev_queue_offset_compare, sizeof (zio_t),
offsetof(struct zio, io_offset_node));
avl_create(vdev_queue_type_tree(vq, ZIO_TYPE_TRIM),
vdev_queue_offset_compare, sizeof (zio_t),
offsetof(struct zio, io_offset_node));
for (p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) {
int (*compfn) (const void *, const void *);
/*
* The synchronous/trim i/o queues are dispatched in FIFO rather
* than LBA order. This provides more consistent latency for
* these i/os.
*/
if (p == ZIO_PRIORITY_SYNC_READ ||
p == ZIO_PRIORITY_SYNC_WRITE ||
p == ZIO_PRIORITY_TRIM) {
compfn = vdev_queue_timestamp_compare;
} else {
compfn = vdev_queue_offset_compare;
}
avl_create(vdev_queue_class_tree(vq, p), compfn,
sizeof (zio_t), offsetof(struct zio, io_queue_node));
}
vq->vq_last_offset = 0;
}
void
vdev_queue_fini(vdev_t *vd)
{
vdev_queue_t *vq = &vd->vdev_queue;
for (zio_priority_t p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++)
avl_destroy(vdev_queue_class_tree(vq, p));
avl_destroy(&vq->vq_active_tree);
avl_destroy(vdev_queue_type_tree(vq, ZIO_TYPE_READ));
avl_destroy(vdev_queue_type_tree(vq, ZIO_TYPE_WRITE));
avl_destroy(vdev_queue_type_tree(vq, ZIO_TYPE_TRIM));
mutex_destroy(&vq->vq_lock);
}
static void
vdev_queue_io_add(vdev_queue_t *vq, zio_t *zio)
{
ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
avl_add(vdev_queue_class_tree(vq, zio->io_priority), zio);
avl_add(vdev_queue_type_tree(vq, zio->io_type), zio);
}
static void
vdev_queue_io_remove(vdev_queue_t *vq, zio_t *zio)
{
ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
avl_remove(vdev_queue_class_tree(vq, zio->io_priority), zio);
avl_remove(vdev_queue_type_tree(vq, zio->io_type), zio);
}
static boolean_t
vdev_queue_is_interactive(zio_priority_t p)
{
switch (p) {
case ZIO_PRIORITY_SCRUB:
case ZIO_PRIORITY_REMOVAL:
case ZIO_PRIORITY_INITIALIZING:
case ZIO_PRIORITY_REBUILD:
return (B_FALSE);
default:
return (B_TRUE);
}
}
static void
vdev_queue_pending_add(vdev_queue_t *vq, zio_t *zio)
{
ASSERT(MUTEX_HELD(&vq->vq_lock));
ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
vq->vq_class[zio->io_priority].vqc_active++;
if (vdev_queue_is_interactive(zio->io_priority)) {
if (++vq->vq_ia_active == 1)
vq->vq_nia_credit = 1;
} else if (vq->vq_ia_active > 0) {
vq->vq_nia_credit--;
}
avl_add(&vq->vq_active_tree, zio);
}
static void
vdev_queue_pending_remove(vdev_queue_t *vq, zio_t *zio)
{
ASSERT(MUTEX_HELD(&vq->vq_lock));
ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
vq->vq_class[zio->io_priority].vqc_active--;
if (vdev_queue_is_interactive(zio->io_priority)) {
if (--vq->vq_ia_active == 0)
vq->vq_nia_credit = 0;
else
vq->vq_nia_credit = zfs_vdev_nia_credit;
} else if (vq->vq_ia_active == 0)
vq->vq_nia_credit++;
avl_remove(&vq->vq_active_tree, zio);
}
static void
vdev_queue_agg_io_done(zio_t *aio)
{
abd_free(aio->io_abd);
}
/*
* Compute the range spanned by two i/os, which is the endpoint of the last
* (lio->io_offset + lio->io_size) minus start of the first (fio->io_offset).
* Conveniently, the gap between fio and lio is given by -IO_SPAN(lio, fio);
* thus fio and lio are adjacent if and only if IO_SPAN(lio, fio) == 0.
*/
#define IO_SPAN(fio, lio) ((lio)->io_offset + (lio)->io_size - (fio)->io_offset)
#define IO_GAP(fio, lio) (-IO_SPAN(lio, fio))
/*
* Sufficiently adjacent io_offset's in ZIOs will be aggregated. We do this
* by creating a gang ABD from the adjacent ZIOs io_abd's. By using
* a gang ABD we avoid doing memory copies to and from the parent,
* child ZIOs. The gang ABD also accounts for gaps between adjacent
* io_offsets by simply getting the zero ABD for writes or allocating
* a new ABD for reads and placing them in the gang ABD as well.
*/
static zio_t *
vdev_queue_aggregate(vdev_queue_t *vq, zio_t *zio)
{
zio_t *first, *last, *aio, *dio, *mandatory, *nio;
uint64_t maxgap = 0;
uint64_t size;
uint64_t limit;
int maxblocksize;
boolean_t stretch = B_FALSE;
avl_tree_t *t = vdev_queue_type_tree(vq, zio->io_type);
enum zio_flag flags = zio->io_flags & ZIO_FLAG_AGG_INHERIT;
uint64_t next_offset;
abd_t *abd;
maxblocksize = spa_maxblocksize(vq->vq_vdev->vdev_spa);
if (vq->vq_vdev->vdev_nonrot)
limit = zfs_vdev_aggregation_limit_non_rotating;
else
limit = zfs_vdev_aggregation_limit;
limit = MAX(MIN(limit, maxblocksize), 0);
if (zio->io_flags & ZIO_FLAG_DONT_AGGREGATE || limit == 0)
return (NULL);
/*
* While TRIM commands could be aggregated based on offset this
* behavior is disabled until it's determined to be beneficial.
*/
if (zio->io_type == ZIO_TYPE_TRIM && !zfs_vdev_aggregate_trim)
return (NULL);
/*
* I/Os to distributed spares are directly dispatched to the dRAID
* leaf vdevs for aggregation. See the comment at the end of the
* zio_vdev_io_start() function.
*/
ASSERT(vq->vq_vdev->vdev_ops != &vdev_draid_spare_ops);
first = last = zio;
if (zio->io_type == ZIO_TYPE_READ)
maxgap = zfs_vdev_read_gap_limit;
/*
* We can aggregate I/Os that are sufficiently adjacent and of
* the same flavor, as expressed by the AGG_INHERIT flags.
* The latter requirement is necessary so that certain
* attributes of the I/O, such as whether it's a normal I/O
* or a scrub/resilver, can be preserved in the aggregate.
* We can include optional I/Os, but don't allow them
* to begin a range as they add no benefit in that situation.
*/
/*
* We keep track of the last non-optional I/O.
*/
mandatory = (first->io_flags & ZIO_FLAG_OPTIONAL) ? NULL : first;
/*
* Walk backwards through sufficiently contiguous I/Os
* recording the last non-optional I/O.
*/
while ((dio = AVL_PREV(t, first)) != NULL &&
(dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags &&
IO_SPAN(dio, last) <= limit &&
IO_GAP(dio, first) <= maxgap &&
dio->io_type == zio->io_type) {
first = dio;
if (mandatory == NULL && !(first->io_flags & ZIO_FLAG_OPTIONAL))
mandatory = first;
}
/*
* Skip any initial optional I/Os.
*/
while ((first->io_flags & ZIO_FLAG_OPTIONAL) && first != last) {
first = AVL_NEXT(t, first);
ASSERT(first != NULL);
}
/*
* Walk forward through sufficiently contiguous I/Os.
* The aggregation limit does not apply to optional i/os, so that
* we can issue contiguous writes even if they are larger than the
* aggregation limit.
*/
while ((dio = AVL_NEXT(t, last)) != NULL &&
(dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags &&
(IO_SPAN(first, dio) <= limit ||
(dio->io_flags & ZIO_FLAG_OPTIONAL)) &&
IO_SPAN(first, dio) <= maxblocksize &&
IO_GAP(last, dio) <= maxgap &&
dio->io_type == zio->io_type) {
last = dio;
if (!(last->io_flags & ZIO_FLAG_OPTIONAL))
mandatory = last;
}
/*
* Now that we've established the range of the I/O aggregation
* we must decide what to do with trailing optional I/Os.
* For reads, there's nothing to do. While we are unable to
* aggregate further, it's possible that a trailing optional
* I/O would allow the underlying device to aggregate with
* subsequent I/Os. We must therefore determine if the next
* non-optional I/O is close enough to make aggregation
* worthwhile.
*/
if (zio->io_type == ZIO_TYPE_WRITE && mandatory != NULL) {
zio_t *nio = last;
while ((dio = AVL_NEXT(t, nio)) != NULL &&
IO_GAP(nio, dio) == 0 &&
IO_GAP(mandatory, dio) <= zfs_vdev_write_gap_limit) {
nio = dio;
if (!(nio->io_flags & ZIO_FLAG_OPTIONAL)) {
stretch = B_TRUE;
break;
}
}
}
if (stretch) {
/*
* We are going to include an optional io in our aggregated
* span, thus closing the write gap. Only mandatory i/os can
* start aggregated spans, so make sure that the next i/o
* after our span is mandatory.
*/
dio = AVL_NEXT(t, last);
dio->io_flags &= ~ZIO_FLAG_OPTIONAL;
} else {
/* do not include the optional i/o */
while (last != mandatory && last != first) {
ASSERT(last->io_flags & ZIO_FLAG_OPTIONAL);
last = AVL_PREV(t, last);
ASSERT(last != NULL);
}
}
if (first == last)
return (NULL);
size = IO_SPAN(first, last);
ASSERT3U(size, <=, maxblocksize);
abd = abd_alloc_gang();
if (abd == NULL)
return (NULL);
aio = zio_vdev_delegated_io(first->io_vd, first->io_offset,
abd, size, first->io_type, zio->io_priority,
flags | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE,
vdev_queue_agg_io_done, NULL);
aio->io_timestamp = first->io_timestamp;
nio = first;
next_offset = first->io_offset;
do {
dio = nio;
nio = AVL_NEXT(t, dio);
zio_add_child(dio, aio);
vdev_queue_io_remove(vq, dio);
if (dio->io_offset != next_offset) {
/* allocate a buffer for a read gap */
ASSERT3U(dio->io_type, ==, ZIO_TYPE_READ);
ASSERT3U(dio->io_offset, >, next_offset);
abd = abd_alloc_for_io(
dio->io_offset - next_offset, B_TRUE);
abd_gang_add(aio->io_abd, abd, B_TRUE);
}
if (dio->io_abd &&
(dio->io_size != abd_get_size(dio->io_abd))) {
/* abd size not the same as IO size */
ASSERT3U(abd_get_size(dio->io_abd), >, dio->io_size);
abd = abd_get_offset_size(dio->io_abd, 0, dio->io_size);
abd_gang_add(aio->io_abd, abd, B_TRUE);
} else {
if (dio->io_flags & ZIO_FLAG_NODATA) {
/* allocate a buffer for a write gap */
ASSERT3U(dio->io_type, ==, ZIO_TYPE_WRITE);
ASSERT3P(dio->io_abd, ==, NULL);
abd_gang_add(aio->io_abd,
abd_get_zeros(dio->io_size), B_TRUE);
} else {
/*
* We pass B_FALSE to abd_gang_add()
* because we did not allocate a new
* ABD, so it is assumed the caller
* will free this ABD.
*/
abd_gang_add(aio->io_abd, dio->io_abd,
B_FALSE);
}
}
next_offset = dio->io_offset + dio->io_size;
} while (dio != last);
ASSERT3U(abd_get_size(aio->io_abd), ==, aio->io_size);
/*
* Callers must call zio_vdev_io_bypass() and zio_execute() for
* aggregated (parent) I/Os so that we could avoid dropping the
* queue's lock here to avoid a deadlock that we could encounter
* due to lock order reversal between vq_lock and io_lock in
* zio_change_priority().
*/
return (aio);
}
static zio_t *
vdev_queue_io_to_issue(vdev_queue_t *vq)
{
zio_t *zio, *aio;
zio_priority_t p;
avl_index_t idx;
avl_tree_t *tree;
again:
ASSERT(MUTEX_HELD(&vq->vq_lock));
p = vdev_queue_class_to_issue(vq);
if (p == ZIO_PRIORITY_NUM_QUEUEABLE) {
/* No eligible queued i/os */
return (NULL);
}
/*
* For LBA-ordered queues (async / scrub / initializing), issue the
* i/o which follows the most recently issued i/o in LBA (offset) order.
*
* For FIFO queues (sync/trim), issue the i/o with the lowest timestamp.
*/
tree = vdev_queue_class_tree(vq, p);
vq->vq_io_search.io_timestamp = 0;
vq->vq_io_search.io_offset = vq->vq_last_offset - 1;
VERIFY3P(avl_find(tree, &vq->vq_io_search, &idx), ==, NULL);
zio = avl_nearest(tree, idx, AVL_AFTER);
if (zio == NULL)
zio = avl_first(tree);
ASSERT3U(zio->io_priority, ==, p);
aio = vdev_queue_aggregate(vq, zio);
if (aio != NULL) {
zio = aio;
} else {
vdev_queue_io_remove(vq, zio);
/*
* If the I/O is or was optional and therefore has no data, we
* need to simply discard it. We need to drop the vdev queue's
* lock to avoid a deadlock that we could encounter since this
* I/O will complete immediately.
*/
if (zio->io_flags & ZIO_FLAG_NODATA) {
mutex_exit(&vq->vq_lock);
zio_vdev_io_bypass(zio);
zio_execute(zio);
mutex_enter(&vq->vq_lock);
goto again;
}
}
vdev_queue_pending_add(vq, zio);
vq->vq_last_offset = zio->io_offset + zio->io_size;
return (zio);
}
zio_t *
vdev_queue_io(zio_t *zio)
{
vdev_queue_t *vq = &zio->io_vd->vdev_queue;
zio_t *dio, *nio;
zio_link_t *zl = NULL;
if (zio->io_flags & ZIO_FLAG_DONT_QUEUE)
return (zio);
/*
* Children i/os inherent their parent's priority, which might
* not match the child's i/o type. Fix it up here.
*/
if (zio->io_type == ZIO_TYPE_READ) {
ASSERT(zio->io_priority != ZIO_PRIORITY_TRIM);
if (zio->io_priority != ZIO_PRIORITY_SYNC_READ &&
zio->io_priority != ZIO_PRIORITY_ASYNC_READ &&
zio->io_priority != ZIO_PRIORITY_SCRUB &&
zio->io_priority != ZIO_PRIORITY_REMOVAL &&
zio->io_priority != ZIO_PRIORITY_INITIALIZING &&
zio->io_priority != ZIO_PRIORITY_REBUILD) {
zio->io_priority = ZIO_PRIORITY_ASYNC_READ;
}
} else if (zio->io_type == ZIO_TYPE_WRITE) {
ASSERT(zio->io_priority != ZIO_PRIORITY_TRIM);
if (zio->io_priority != ZIO_PRIORITY_SYNC_WRITE &&
zio->io_priority != ZIO_PRIORITY_ASYNC_WRITE &&
zio->io_priority != ZIO_PRIORITY_REMOVAL &&
zio->io_priority != ZIO_PRIORITY_INITIALIZING &&
zio->io_priority != ZIO_PRIORITY_REBUILD) {
zio->io_priority = ZIO_PRIORITY_ASYNC_WRITE;
}
} else {
ASSERT(zio->io_type == ZIO_TYPE_TRIM);
ASSERT(zio->io_priority == ZIO_PRIORITY_TRIM);
}
zio->io_flags |= ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE;
zio->io_timestamp = gethrtime();
mutex_enter(&vq->vq_lock);
vdev_queue_io_add(vq, zio);
nio = vdev_queue_io_to_issue(vq);
mutex_exit(&vq->vq_lock);
if (nio == NULL)
return (NULL);
if (nio->io_done == vdev_queue_agg_io_done) {
while ((dio = zio_walk_parents(nio, &zl)) != NULL) {
ASSERT3U(dio->io_type, ==, nio->io_type);
zio_vdev_io_bypass(dio);
zio_execute(dio);
}
zio_nowait(nio);
return (NULL);
}
return (nio);
}
void
vdev_queue_io_done(zio_t *zio)
{
vdev_queue_t *vq = &zio->io_vd->vdev_queue;
zio_t *dio, *nio;
zio_link_t *zl = NULL;
hrtime_t now = gethrtime();
vq->vq_io_complete_ts = now;
vq->vq_io_delta_ts = zio->io_delta = now - zio->io_timestamp;
mutex_enter(&vq->vq_lock);
vdev_queue_pending_remove(vq, zio);
while ((nio = vdev_queue_io_to_issue(vq)) != NULL) {
mutex_exit(&vq->vq_lock);
if (nio->io_done == vdev_queue_agg_io_done) {
while ((dio = zio_walk_parents(nio, &zl)) != NULL) {
ASSERT3U(dio->io_type, ==, nio->io_type);
zio_vdev_io_bypass(dio);
zio_execute(dio);
}
zio_nowait(nio);
} else {
zio_vdev_io_reissue(nio);
zio_execute(nio);
}
mutex_enter(&vq->vq_lock);
}
mutex_exit(&vq->vq_lock);
}
void
vdev_queue_change_io_priority(zio_t *zio, zio_priority_t priority)
{
vdev_queue_t *vq = &zio->io_vd->vdev_queue;
avl_tree_t *tree;
/*
* ZIO_PRIORITY_NOW is used by the vdev cache code and the aggregate zio
* code to issue IOs without adding them to the vdev queue. In this
* case, the zio is already going to be issued as quickly as possible
* and so it doesn't need any reprioritization to help.
*/
if (zio->io_priority == ZIO_PRIORITY_NOW)
return;
ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
ASSERT3U(priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
if (zio->io_type == ZIO_TYPE_READ) {
if (priority != ZIO_PRIORITY_SYNC_READ &&
priority != ZIO_PRIORITY_ASYNC_READ &&
priority != ZIO_PRIORITY_SCRUB)
priority = ZIO_PRIORITY_ASYNC_READ;
} else {
ASSERT(zio->io_type == ZIO_TYPE_WRITE);
if (priority != ZIO_PRIORITY_SYNC_WRITE &&
priority != ZIO_PRIORITY_ASYNC_WRITE)
priority = ZIO_PRIORITY_ASYNC_WRITE;
}
mutex_enter(&vq->vq_lock);
/*
* If the zio is in none of the queues we can simply change
* the priority. If the zio is waiting to be submitted we must
* remove it from the queue and re-insert it with the new priority.
* Otherwise, the zio is currently active and we cannot change its
* priority.
*/
tree = vdev_queue_class_tree(vq, zio->io_priority);
if (avl_find(tree, zio, NULL) == zio) {
avl_remove(vdev_queue_class_tree(vq, zio->io_priority), zio);
zio->io_priority = priority;
avl_add(vdev_queue_class_tree(vq, zio->io_priority), zio);
} else if (avl_find(&vq->vq_active_tree, zio, NULL) != zio) {
zio->io_priority = priority;
}
mutex_exit(&vq->vq_lock);
}
/*
* As these two methods are only used for load calculations we're not
* concerned if we get an incorrect value on 32bit platforms due to lack of
* vq_lock mutex use here, instead we prefer to keep it lock free for
* performance.
*/
int
vdev_queue_length(vdev_t *vd)
{
return (avl_numnodes(&vd->vdev_queue.vq_active_tree));
}
uint64_t
vdev_queue_last_offset(vdev_t *vd)
{
return (vd->vdev_queue.vq_last_offset);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, aggregation_limit, INT, ZMOD_RW,
"Max vdev I/O aggregation size");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, aggregation_limit_non_rotating, INT, ZMOD_RW,
"Max vdev I/O aggregation size for non-rotating media");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, aggregate_trim, INT, ZMOD_RW,
"Allow TRIM I/O to be aggregated");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, read_gap_limit, INT, ZMOD_RW,
"Aggregate read I/O over gap");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, write_gap_limit, INT, ZMOD_RW,
"Aggregate write I/O over gap");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, max_active, INT, ZMOD_RW,
"Maximum number of active I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_active_max_dirty_percent, INT, ZMOD_RW,
"Async write concurrency max threshold");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_active_min_dirty_percent, INT, ZMOD_RW,
"Async write concurrency min threshold");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_read_max_active, INT, ZMOD_RW,
"Max active async read I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_read_min_active, INT, ZMOD_RW,
"Min active async read I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_max_active, INT, ZMOD_RW,
"Max active async write I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_min_active, INT, ZMOD_RW,
"Min active async write I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, initializing_max_active, INT, ZMOD_RW,
"Max active initializing I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, initializing_min_active, INT, ZMOD_RW,
"Min active initializing I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, removal_max_active, INT, ZMOD_RW,
"Max active removal I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, removal_min_active, INT, ZMOD_RW,
"Min active removal I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, scrub_max_active, INT, ZMOD_RW,
"Max active scrub I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, scrub_min_active, INT, ZMOD_RW,
"Min active scrub I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_read_max_active, INT, ZMOD_RW,
"Max active sync read I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_read_min_active, INT, ZMOD_RW,
"Min active sync read I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_write_max_active, INT, ZMOD_RW,
"Max active sync write I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_write_min_active, INT, ZMOD_RW,
"Min active sync write I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, trim_max_active, INT, ZMOD_RW,
"Max active trim/discard I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, trim_min_active, INT, ZMOD_RW,
"Min active trim/discard I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, rebuild_max_active, INT, ZMOD_RW,
"Max active rebuild I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, rebuild_min_active, INT, ZMOD_RW,
"Min active rebuild I/Os per vdev");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, nia_credit, INT, ZMOD_RW,
"Number of non-interactive I/Os to allow in sequence");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, nia_delay, INT, ZMOD_RW,
"Number of non-interactive I/Os before _max_active");
ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, queue_depth_pct, INT, ZMOD_RW,
"Queue depth percentage for each top-level vdev");
/* END CSTYLED */
diff --git a/module/zfs/vdev_raidz_math.c b/module/zfs/vdev_raidz_math.c
index 03df2df5adaf..50b8dab74848 100644
--- a/module/zfs/vdev_raidz_math.c
+++ b/module/zfs/vdev_raidz_math.c
@@ -1,673 +1,670 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (C) 2016 Gvozden Nešković. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/types.h>
#include <sys/zio.h>
#include <sys/debug.h>
#include <sys/zfs_debug.h>
#include <sys/vdev_raidz.h>
#include <sys/vdev_raidz_impl.h>
#include <sys/simd.h>
/* Opaque implementation with NULL methods to represent original methods */
static const raidz_impl_ops_t vdev_raidz_original_impl = {
.name = "original",
.is_supported = raidz_will_scalar_work,
};
/* RAIDZ parity op that contain the fastest methods */
static raidz_impl_ops_t vdev_raidz_fastest_impl = {
.name = "fastest"
};
/* All compiled in implementations */
-const raidz_impl_ops_t *raidz_all_maths[] = {
+static const raidz_impl_ops_t *const raidz_all_maths[] = {
&vdev_raidz_original_impl,
&vdev_raidz_scalar_impl,
#if defined(__x86_64) && defined(HAVE_SSE2) /* only x86_64 for now */
&vdev_raidz_sse2_impl,
#endif
#if defined(__x86_64) && defined(HAVE_SSSE3) /* only x86_64 for now */
&vdev_raidz_ssse3_impl,
#endif
#if defined(__x86_64) && defined(HAVE_AVX2) /* only x86_64 for now */
&vdev_raidz_avx2_impl,
#endif
#if defined(__x86_64) && defined(HAVE_AVX512F) /* only x86_64 for now */
&vdev_raidz_avx512f_impl,
#endif
#if defined(__x86_64) && defined(HAVE_AVX512BW) /* only x86_64 for now */
&vdev_raidz_avx512bw_impl,
#endif
#if defined(__aarch64__) && !defined(__FreeBSD__)
&vdev_raidz_aarch64_neon_impl,
&vdev_raidz_aarch64_neonx2_impl,
#endif
#if defined(__powerpc__) && defined(__altivec__)
&vdev_raidz_powerpc_altivec_impl,
#endif
};
/* Indicate that benchmark has been completed */
static boolean_t raidz_math_initialized = B_FALSE;
/* Select raidz implementation */
#define IMPL_FASTEST (UINT32_MAX)
#define IMPL_CYCLE (UINT32_MAX - 1)
#define IMPL_ORIGINAL (0)
#define IMPL_SCALAR (1)
#define RAIDZ_IMPL_READ(i) (*(volatile uint32_t *) &(i))
static uint32_t zfs_vdev_raidz_impl = IMPL_SCALAR;
static uint32_t user_sel_impl = IMPL_FASTEST;
/* Hold all supported implementations */
static size_t raidz_supp_impl_cnt = 0;
static raidz_impl_ops_t *raidz_supp_impl[ARRAY_SIZE(raidz_all_maths)];
#if defined(_KERNEL)
/*
* kstats values for supported implementations
* Values represent per disk throughput of 8 disk+parity raidz vdev [B/s]
*/
static raidz_impl_kstat_t raidz_impl_kstats[ARRAY_SIZE(raidz_all_maths) + 1];
/* kstat for benchmarked implementations */
static kstat_t *raidz_math_kstat = NULL;
#endif
/*
* Returns the RAIDZ operations for raidz_map() parity calculations. When
* a SIMD implementation is not allowed in the current context, then fallback
* to the fastest generic implementation.
*/
const raidz_impl_ops_t *
vdev_raidz_math_get_ops(void)
{
if (!kfpu_allowed())
return (&vdev_raidz_scalar_impl);
raidz_impl_ops_t *ops = NULL;
const uint32_t impl = RAIDZ_IMPL_READ(zfs_vdev_raidz_impl);
switch (impl) {
case IMPL_FASTEST:
ASSERT(raidz_math_initialized);
ops = &vdev_raidz_fastest_impl;
break;
case IMPL_CYCLE:
/* Cycle through all supported implementations */
ASSERT(raidz_math_initialized);
ASSERT3U(raidz_supp_impl_cnt, >, 0);
static size_t cycle_impl_idx = 0;
size_t idx = (++cycle_impl_idx) % raidz_supp_impl_cnt;
ops = raidz_supp_impl[idx];
break;
case IMPL_ORIGINAL:
ops = (raidz_impl_ops_t *)&vdev_raidz_original_impl;
break;
case IMPL_SCALAR:
ops = (raidz_impl_ops_t *)&vdev_raidz_scalar_impl;
break;
default:
ASSERT3U(impl, <, raidz_supp_impl_cnt);
ASSERT3U(raidz_supp_impl_cnt, >, 0);
if (impl < ARRAY_SIZE(raidz_all_maths))
ops = raidz_supp_impl[impl];
break;
}
ASSERT3P(ops, !=, NULL);
return (ops);
}
/*
* Select parity generation method for raidz_map
*/
int
vdev_raidz_math_generate(raidz_map_t *rm, raidz_row_t *rr)
{
raidz_gen_f gen_parity = NULL;
switch (raidz_parity(rm)) {
case 1:
gen_parity = rm->rm_ops->gen[RAIDZ_GEN_P];
break;
case 2:
gen_parity = rm->rm_ops->gen[RAIDZ_GEN_PQ];
break;
case 3:
gen_parity = rm->rm_ops->gen[RAIDZ_GEN_PQR];
break;
default:
gen_parity = NULL;
cmn_err(CE_PANIC, "invalid RAID-Z configuration %llu",
(u_longlong_t)raidz_parity(rm));
break;
}
/* if method is NULL execute the original implementation */
if (gen_parity == NULL)
return (RAIDZ_ORIGINAL_IMPL);
gen_parity(rr);
return (0);
}
static raidz_rec_f
reconstruct_fun_p_sel(raidz_map_t *rm, const int *parity_valid,
const int nbaddata)
{
if (nbaddata == 1 && parity_valid[CODE_P]) {
return (rm->rm_ops->rec[RAIDZ_REC_P]);
}
return ((raidz_rec_f) NULL);
}
static raidz_rec_f
reconstruct_fun_pq_sel(raidz_map_t *rm, const int *parity_valid,
const int nbaddata)
{
if (nbaddata == 1) {
if (parity_valid[CODE_P]) {
return (rm->rm_ops->rec[RAIDZ_REC_P]);
} else if (parity_valid[CODE_Q]) {
return (rm->rm_ops->rec[RAIDZ_REC_Q]);
}
} else if (nbaddata == 2 &&
parity_valid[CODE_P] && parity_valid[CODE_Q]) {
return (rm->rm_ops->rec[RAIDZ_REC_PQ]);
}
return ((raidz_rec_f) NULL);
}
static raidz_rec_f
reconstruct_fun_pqr_sel(raidz_map_t *rm, const int *parity_valid,
const int nbaddata)
{
if (nbaddata == 1) {
if (parity_valid[CODE_P]) {
return (rm->rm_ops->rec[RAIDZ_REC_P]);
} else if (parity_valid[CODE_Q]) {
return (rm->rm_ops->rec[RAIDZ_REC_Q]);
} else if (parity_valid[CODE_R]) {
return (rm->rm_ops->rec[RAIDZ_REC_R]);
}
} else if (nbaddata == 2) {
if (parity_valid[CODE_P] && parity_valid[CODE_Q]) {
return (rm->rm_ops->rec[RAIDZ_REC_PQ]);
} else if (parity_valid[CODE_P] && parity_valid[CODE_R]) {
return (rm->rm_ops->rec[RAIDZ_REC_PR]);
} else if (parity_valid[CODE_Q] && parity_valid[CODE_R]) {
return (rm->rm_ops->rec[RAIDZ_REC_QR]);
}
} else if (nbaddata == 3 &&
parity_valid[CODE_P] && parity_valid[CODE_Q] &&
parity_valid[CODE_R]) {
return (rm->rm_ops->rec[RAIDZ_REC_PQR]);
}
return ((raidz_rec_f) NULL);
}
/*
* Select data reconstruction method for raidz_map
* @parity_valid - Parity validity flag
* @dt - Failed data index array
* @nbaddata - Number of failed data columns
*/
int
vdev_raidz_math_reconstruct(raidz_map_t *rm, raidz_row_t *rr,
const int *parity_valid, const int *dt, const int nbaddata)
{
raidz_rec_f rec_fn = NULL;
switch (raidz_parity(rm)) {
case PARITY_P:
rec_fn = reconstruct_fun_p_sel(rm, parity_valid, nbaddata);
break;
case PARITY_PQ:
rec_fn = reconstruct_fun_pq_sel(rm, parity_valid, nbaddata);
break;
case PARITY_PQR:
rec_fn = reconstruct_fun_pqr_sel(rm, parity_valid, nbaddata);
break;
default:
cmn_err(CE_PANIC, "invalid RAID-Z configuration %llu",
(u_longlong_t)raidz_parity(rm));
break;
}
if (rec_fn == NULL)
return (RAIDZ_ORIGINAL_IMPL);
else
return (rec_fn(rr, dt));
}
-const char *raidz_gen_name[] = {
+const char *const raidz_gen_name[] = {
"gen_p", "gen_pq", "gen_pqr"
};
-const char *raidz_rec_name[] = {
+const char *const raidz_rec_name[] = {
"rec_p", "rec_q", "rec_r",
"rec_pq", "rec_pr", "rec_qr", "rec_pqr"
};
#if defined(_KERNEL)
#define RAIDZ_KSTAT_LINE_LEN (17 + 10*12 + 1)
static int
raidz_math_kstat_headers(char *buf, size_t size)
{
- int i;
- ssize_t off;
-
ASSERT3U(size, >=, RAIDZ_KSTAT_LINE_LEN);
- off = snprintf(buf, size, "%-17s", "implementation");
+ ssize_t off = snprintf(buf, size, "%-17s", "implementation");
- for (i = 0; i < ARRAY_SIZE(raidz_gen_name); i++)
+ for (int i = 0; i < ARRAY_SIZE(raidz_gen_name); i++)
off += snprintf(buf + off, size - off, "%-16s",
raidz_gen_name[i]);
- for (i = 0; i < ARRAY_SIZE(raidz_rec_name); i++)
+ for (int i = 0; i < ARRAY_SIZE(raidz_rec_name); i++)
off += snprintf(buf + off, size - off, "%-16s",
raidz_rec_name[i]);
(void) snprintf(buf + off, size - off, "\n");
return (0);
}
static int
raidz_math_kstat_data(char *buf, size_t size, void *data)
{
raidz_impl_kstat_t *fstat = &raidz_impl_kstats[raidz_supp_impl_cnt];
raidz_impl_kstat_t *cstat = (raidz_impl_kstat_t *)data;
ssize_t off = 0;
int i;
ASSERT3U(size, >=, RAIDZ_KSTAT_LINE_LEN);
if (cstat == fstat) {
off += snprintf(buf + off, size - off, "%-17s", "fastest");
for (i = 0; i < ARRAY_SIZE(raidz_gen_name); i++) {
int id = fstat->gen[i];
off += snprintf(buf + off, size - off, "%-16s",
raidz_supp_impl[id]->name);
}
for (i = 0; i < ARRAY_SIZE(raidz_rec_name); i++) {
int id = fstat->rec[i];
off += snprintf(buf + off, size - off, "%-16s",
raidz_supp_impl[id]->name);
}
} else {
ptrdiff_t id = cstat - raidz_impl_kstats;
off += snprintf(buf + off, size - off, "%-17s",
raidz_supp_impl[id]->name);
for (i = 0; i < ARRAY_SIZE(raidz_gen_name); i++)
off += snprintf(buf + off, size - off, "%-16llu",
(u_longlong_t)cstat->gen[i]);
for (i = 0; i < ARRAY_SIZE(raidz_rec_name); i++)
off += snprintf(buf + off, size - off, "%-16llu",
(u_longlong_t)cstat->rec[i]);
}
(void) snprintf(buf + off, size - off, "\n");
return (0);
}
static void *
raidz_math_kstat_addr(kstat_t *ksp, loff_t n)
{
if (n <= raidz_supp_impl_cnt)
ksp->ks_private = (void *) (raidz_impl_kstats + n);
else
ksp->ks_private = NULL;
return (ksp->ks_private);
}
#define BENCH_D_COLS (8ULL)
#define BENCH_COLS (BENCH_D_COLS + PARITY_PQR)
#define BENCH_ZIO_SIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT) /* 128 kiB */
#define BENCH_NS MSEC2NSEC(1) /* 1ms */
typedef void (*benchmark_fn)(raidz_map_t *rm, const int fn);
static void
benchmark_gen_impl(raidz_map_t *rm, const int fn)
{
(void) fn;
vdev_raidz_generate_parity(rm);
}
static void
benchmark_rec_impl(raidz_map_t *rm, const int fn)
{
static const int rec_tgt[7][3] = {
{1, 2, 3}, /* rec_p: bad QR & D[0] */
{0, 2, 3}, /* rec_q: bad PR & D[0] */
{0, 1, 3}, /* rec_r: bad PQ & D[0] */
{2, 3, 4}, /* rec_pq: bad R & D[0][1] */
{1, 3, 4}, /* rec_pr: bad Q & D[0][1] */
{0, 3, 4}, /* rec_qr: bad P & D[0][1] */
{3, 4, 5} /* rec_pqr: bad & D[0][1][2] */
};
vdev_raidz_reconstruct(rm, rec_tgt[fn], 3);
}
/*
* Benchmarking of all supported implementations (raidz_supp_impl_cnt)
* is performed by setting the rm_ops pointer and calling the top level
* generate/reconstruct methods of bench_rm.
*/
static void
benchmark_raidz_impl(raidz_map_t *bench_rm, const int fn, benchmark_fn bench_fn)
{
uint64_t run_cnt, speed, best_speed = 0;
hrtime_t t_start, t_diff;
raidz_impl_ops_t *curr_impl;
raidz_impl_kstat_t *fstat = &raidz_impl_kstats[raidz_supp_impl_cnt];
int impl, i;
for (impl = 0; impl < raidz_supp_impl_cnt; impl++) {
/* set an implementation to benchmark */
curr_impl = raidz_supp_impl[impl];
bench_rm->rm_ops = curr_impl;
run_cnt = 0;
t_start = gethrtime();
do {
for (i = 0; i < 5; i++, run_cnt++)
bench_fn(bench_rm, fn);
t_diff = gethrtime() - t_start;
} while (t_diff < BENCH_NS);
speed = run_cnt * BENCH_ZIO_SIZE * NANOSEC;
speed /= (t_diff * BENCH_COLS);
if (bench_fn == benchmark_gen_impl)
raidz_impl_kstats[impl].gen[fn] = speed;
else
raidz_impl_kstats[impl].rec[fn] = speed;
/* Update fastest implementation method */
if (speed > best_speed) {
best_speed = speed;
if (bench_fn == benchmark_gen_impl) {
fstat->gen[fn] = impl;
vdev_raidz_fastest_impl.gen[fn] =
curr_impl->gen[fn];
} else {
fstat->rec[fn] = impl;
vdev_raidz_fastest_impl.rec[fn] =
curr_impl->rec[fn];
}
}
}
}
#endif
/*
* Initialize and benchmark all supported implementations.
*/
static void
benchmark_raidz(void)
{
raidz_impl_ops_t *curr_impl;
int i, c;
/* Move supported impl into raidz_supp_impl */
for (i = 0, c = 0; i < ARRAY_SIZE(raidz_all_maths); i++) {
curr_impl = (raidz_impl_ops_t *)raidz_all_maths[i];
if (curr_impl->init)
curr_impl->init();
if (curr_impl->is_supported())
raidz_supp_impl[c++] = (raidz_impl_ops_t *)curr_impl;
}
membar_producer(); /* complete raidz_supp_impl[] init */
raidz_supp_impl_cnt = c; /* number of supported impl */
#if defined(_KERNEL)
abd_t *pabd;
zio_t *bench_zio = NULL;
raidz_map_t *bench_rm = NULL;
uint64_t bench_parity;
/* Fake a zio and run the benchmark on a warmed up buffer */
bench_zio = kmem_zalloc(sizeof (zio_t), KM_SLEEP);
bench_zio->io_offset = 0;
bench_zio->io_size = BENCH_ZIO_SIZE; /* only data columns */
bench_zio->io_abd = abd_alloc_linear(BENCH_ZIO_SIZE, B_TRUE);
memset(abd_to_buf(bench_zio->io_abd), 0xAA, BENCH_ZIO_SIZE);
/* Benchmark parity generation methods */
for (int fn = 0; fn < RAIDZ_GEN_NUM; fn++) {
bench_parity = fn + 1;
/* New raidz_map is needed for each generate_p/q/r */
bench_rm = vdev_raidz_map_alloc(bench_zio, SPA_MINBLOCKSHIFT,
BENCH_D_COLS + bench_parity, bench_parity);
benchmark_raidz_impl(bench_rm, fn, benchmark_gen_impl);
vdev_raidz_map_free(bench_rm);
}
/* Benchmark data reconstruction methods */
bench_rm = vdev_raidz_map_alloc(bench_zio, SPA_MINBLOCKSHIFT,
BENCH_COLS, PARITY_PQR);
/* Ensure that fake parity blocks are initialized */
for (c = 0; c < bench_rm->rm_row[0]->rr_firstdatacol; c++) {
pabd = bench_rm->rm_row[0]->rr_col[c].rc_abd;
memset(abd_to_buf(pabd), 0xAA, abd_get_size(pabd));
}
for (int fn = 0; fn < RAIDZ_REC_NUM; fn++)
benchmark_raidz_impl(bench_rm, fn, benchmark_rec_impl);
vdev_raidz_map_free(bench_rm);
/* cleanup the bench zio */
abd_free(bench_zio->io_abd);
kmem_free(bench_zio, sizeof (zio_t));
#else
/*
* Skip the benchmark in user space to avoid impacting libzpool
* consumers (zdb, zhack, zinject, ztest). The last implementation
* is assumed to be the fastest and used by default.
*/
memcpy(&vdev_raidz_fastest_impl,
raidz_supp_impl[raidz_supp_impl_cnt - 1],
sizeof (vdev_raidz_fastest_impl));
strcpy(vdev_raidz_fastest_impl.name, "fastest");
#endif /* _KERNEL */
}
void
vdev_raidz_math_init(void)
{
/* Determine the fastest available implementation. */
benchmark_raidz();
#if defined(_KERNEL)
/* Install kstats for all implementations */
raidz_math_kstat = kstat_create("zfs", 0, "vdev_raidz_bench", "misc",
KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);
if (raidz_math_kstat != NULL) {
raidz_math_kstat->ks_data = NULL;
raidz_math_kstat->ks_ndata = UINT32_MAX;
kstat_set_raw_ops(raidz_math_kstat,
raidz_math_kstat_headers,
raidz_math_kstat_data,
raidz_math_kstat_addr);
kstat_install(raidz_math_kstat);
}
#endif
/* Finish initialization */
atomic_swap_32(&zfs_vdev_raidz_impl, user_sel_impl);
raidz_math_initialized = B_TRUE;
}
void
vdev_raidz_math_fini(void)
{
raidz_impl_ops_t const *curr_impl;
#if defined(_KERNEL)
if (raidz_math_kstat != NULL) {
kstat_delete(raidz_math_kstat);
raidz_math_kstat = NULL;
}
#endif
for (int i = 0; i < ARRAY_SIZE(raidz_all_maths); i++) {
curr_impl = raidz_all_maths[i];
if (curr_impl->fini)
curr_impl->fini();
}
}
static const struct {
char *name;
uint32_t sel;
} math_impl_opts[] = {
{ "cycle", IMPL_CYCLE },
{ "fastest", IMPL_FASTEST },
{ "original", IMPL_ORIGINAL },
{ "scalar", IMPL_SCALAR }
};
/*
* Function sets desired raidz implementation.
*
* If we are called before init(), user preference will be saved in
* user_sel_impl, and applied in later init() call. This occurs when module
* parameter is specified on module load. Otherwise, directly update
* zfs_vdev_raidz_impl.
*
* @val Name of raidz implementation to use
* @param Unused.
*/
int
vdev_raidz_impl_set(const char *val)
{
int err = -EINVAL;
char req_name[RAIDZ_IMPL_NAME_MAX];
uint32_t impl = RAIDZ_IMPL_READ(user_sel_impl);
size_t i;
/* sanitize input */
i = strnlen(val, RAIDZ_IMPL_NAME_MAX);
if (i == 0 || i == RAIDZ_IMPL_NAME_MAX)
return (err);
strlcpy(req_name, val, RAIDZ_IMPL_NAME_MAX);
while (i > 0 && !!isspace(req_name[i-1]))
i--;
req_name[i] = '\0';
/* Check mandatory options */
for (i = 0; i < ARRAY_SIZE(math_impl_opts); i++) {
if (strcmp(req_name, math_impl_opts[i].name) == 0) {
impl = math_impl_opts[i].sel;
err = 0;
break;
}
}
/* check all supported impl if init() was already called */
if (err != 0 && raidz_math_initialized) {
/* check all supported implementations */
for (i = 0; i < raidz_supp_impl_cnt; i++) {
if (strcmp(req_name, raidz_supp_impl[i]->name) == 0) {
impl = i;
err = 0;
break;
}
}
}
if (err == 0) {
if (raidz_math_initialized)
atomic_swap_32(&zfs_vdev_raidz_impl, impl);
else
atomic_swap_32(&user_sel_impl, impl);
}
return (err);
}
#if defined(_KERNEL) && defined(__linux__)
static int
zfs_vdev_raidz_impl_set(const char *val, zfs_kernel_param_t *kp)
{
return (vdev_raidz_impl_set(val));
}
static int
zfs_vdev_raidz_impl_get(char *buffer, zfs_kernel_param_t *kp)
{
int i, cnt = 0;
char *fmt;
const uint32_t impl = RAIDZ_IMPL_READ(zfs_vdev_raidz_impl);
ASSERT(raidz_math_initialized);
/* list mandatory options */
for (i = 0; i < ARRAY_SIZE(math_impl_opts) - 2; i++) {
fmt = (impl == math_impl_opts[i].sel) ? "[%s] " : "%s ";
cnt += sprintf(buffer + cnt, fmt, math_impl_opts[i].name);
}
/* list all supported implementations */
for (i = 0; i < raidz_supp_impl_cnt; i++) {
fmt = (i == impl) ? "[%s] " : "%s ";
cnt += sprintf(buffer + cnt, fmt, raidz_supp_impl[i]->name);
}
return (cnt);
}
module_param_call(zfs_vdev_raidz_impl, zfs_vdev_raidz_impl_set,
zfs_vdev_raidz_impl_get, NULL, 0644);
MODULE_PARM_DESC(zfs_vdev_raidz_impl, "Select raidz implementation.");
#endif
diff --git a/module/zfs/vdev_rebuild.c b/module/zfs/vdev_rebuild.c
index 4d7de0c6c44c..fd2490c0aed6 100644
--- a/module/zfs/vdev_rebuild.c
+++ b/module/zfs/vdev_rebuild.c
@@ -1,1150 +1,1150 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
*
* Copyright (c) 2018, Intel Corporation.
* Copyright (c) 2020 by Lawrence Livermore National Security, LLC.
*/
#include <sys/vdev_impl.h>
#include <sys/vdev_draid.h>
#include <sys/dsl_scan.h>
#include <sys/spa_impl.h>
#include <sys/metaslab_impl.h>
#include <sys/vdev_rebuild.h>
#include <sys/zio.h>
#include <sys/dmu_tx.h>
#include <sys/arc.h>
#include <sys/zap.h>
/*
* This file contains the sequential reconstruction implementation for
* resilvering. This form of resilvering is internally referred to as device
* rebuild to avoid conflating it with the traditional healing reconstruction
* performed by the dsl scan code.
*
* When replacing a device, or scrubbing the pool, ZFS has historically used
* a process called resilvering which is a form of healing reconstruction.
* This approach has the advantage that as blocks are read from disk their
* checksums can be immediately verified and the data repaired. Unfortunately,
* it also results in a random IO pattern to the disk even when extra care
* is taken to sequentialize the IO as much as possible. This substantially
* increases the time required to resilver the pool and restore redundancy.
*
* For mirrored devices it's possible to implement an alternate sequential
* reconstruction strategy when resilvering. Sequential reconstruction
* behaves like a traditional RAID rebuild and reconstructs a device in LBA
* order without verifying the checksum. After this phase completes a second
* scrub phase is started to verify all of the checksums. This two phase
* process will take longer than the healing reconstruction described above.
* However, it has that advantage that after the reconstruction first phase
* completes redundancy has been restored. At this point the pool can incur
* another device failure without risking data loss.
*
* There are a few noteworthy limitations and other advantages of resilvering
* using sequential reconstruction vs healing reconstruction.
*
* Limitations:
*
* - Sequential reconstruction is not possible on RAIDZ due to its
* variable stripe width. Note dRAID uses a fixed stripe width which
* avoids this issue, but comes at the expense of some usable capacity.
*
* - Block checksums are not verified during sequential reconstruction.
* Similar to traditional RAID the parity/mirror data is reconstructed
* but cannot be immediately double checked. For this reason when the
* last active resilver completes the pool is automatically scrubbed
* by default.
*
* - Deferred resilvers using sequential reconstruction are not currently
* supported. When adding another vdev to an active top-level resilver
* it must be restarted.
*
* Advantages:
*
* - Sequential reconstruction is performed in LBA order which may be faster
* than healing reconstruction particularly when using HDDs (or
* especially with SMR devices). Only allocated capacity is resilvered.
*
* - Sequential reconstruction is not constrained by ZFS block boundaries.
* This allows it to issue larger IOs to disk which span multiple blocks
* allowing all of these logical blocks to be repaired with a single IO.
*
* - Unlike a healing resilver or scrub which are pool wide operations,
* sequential reconstruction is handled by the top-level vdevs. This
* allows for it to be started or canceled on a top-level vdev without
* impacting any other top-level vdevs in the pool.
*
* - Data only referenced by a pool checkpoint will be repaired because
* that space is reflected in the space maps. This differs for a
* healing resilver or scrub which will not repair that data.
*/
/*
* Size of rebuild reads; defaults to 1MiB per data disk and is capped at
* SPA_MAXBLOCKSIZE.
*/
-unsigned long zfs_rebuild_max_segment = 1024 * 1024;
+static unsigned long zfs_rebuild_max_segment = 1024 * 1024;
/*
* Maximum number of parallelly executed bytes per leaf vdev caused by a
* sequential resilver. We attempt to strike a balance here between keeping
* the vdev queues full of I/Os at all times and not overflowing the queues
* to cause long latency, which would cause long txg sync times.
*
* A large default value can be safely used here because the default target
* segment size is also large (zfs_rebuild_max_segment=1M). This helps keep
* the queue depth short.
*
* 32MB was selected as the default value to achieve good performance with
* a large 90-drive dRAID HDD configuration (draid2:8d:90c:2s). A sequential
* rebuild was unable to saturate all of the drives using smaller values.
* With a value of 32MB the sequential resilver write rate was measured at
* 800MB/s sustained while rebuilding to a distributed spare.
*/
-unsigned long zfs_rebuild_vdev_limit = 32 << 20;
+static unsigned long zfs_rebuild_vdev_limit = 32 << 20;
/*
* Automatically start a pool scrub when the last active sequential resilver
* completes in order to verify the checksums of all blocks which have been
* resilvered. This option is enabled by default and is strongly recommended.
*/
-int zfs_rebuild_scrub_enabled = 1;
+static int zfs_rebuild_scrub_enabled = 1;
/*
* For vdev_rebuild_initiate_sync() and vdev_rebuild_reset_sync().
*/
static void vdev_rebuild_thread(void *arg);
/*
* Clear the per-vdev rebuild bytes value for a vdev tree.
*/
static void
clear_rebuild_bytes(vdev_t *vd)
{
vdev_stat_t *vs = &vd->vdev_stat;
for (uint64_t i = 0; i < vd->vdev_children; i++)
clear_rebuild_bytes(vd->vdev_child[i]);
mutex_enter(&vd->vdev_stat_lock);
vs->vs_rebuild_processed = 0;
mutex_exit(&vd->vdev_stat_lock);
}
/*
* Determines whether a vdev_rebuild_thread() should be stopped.
*/
static boolean_t
vdev_rebuild_should_stop(vdev_t *vd)
{
return (!vdev_writeable(vd) || vd->vdev_removing ||
vd->vdev_rebuild_exit_wanted ||
vd->vdev_rebuild_cancel_wanted ||
vd->vdev_rebuild_reset_wanted);
}
/*
* Determine if the rebuild should be canceled. This may happen when all
* vdevs with MISSING DTLs are detached.
*/
static boolean_t
vdev_rebuild_should_cancel(vdev_t *vd)
{
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
if (!vdev_resilver_needed(vd, &vrp->vrp_min_txg, &vrp->vrp_max_txg))
return (B_TRUE);
return (B_FALSE);
}
/*
* The sync task for updating the on-disk state of a rebuild. This is
* scheduled by vdev_rebuild_range().
*/
static void
vdev_rebuild_update_sync(void *arg, dmu_tx_t *tx)
{
int vdev_id = (uintptr_t)arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *vd = vdev_lookup_top(spa, vdev_id);
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
uint64_t txg = dmu_tx_get_txg(tx);
mutex_enter(&vd->vdev_rebuild_lock);
if (vr->vr_scan_offset[txg & TXG_MASK] > 0) {
vrp->vrp_last_offset = vr->vr_scan_offset[txg & TXG_MASK];
vr->vr_scan_offset[txg & TXG_MASK] = 0;
}
vrp->vrp_scan_time_ms = vr->vr_prev_scan_time_ms +
NSEC2MSEC(gethrtime() - vr->vr_pass_start_time);
VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
REBUILD_PHYS_ENTRIES, vrp, tx));
mutex_exit(&vd->vdev_rebuild_lock);
}
/*
* Initialize the on-disk state for a new rebuild, start the rebuild thread.
*/
static void
vdev_rebuild_initiate_sync(void *arg, dmu_tx_t *tx)
{
int vdev_id = (uintptr_t)arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *vd = vdev_lookup_top(spa, vdev_id);
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
ASSERT(vd->vdev_rebuilding);
spa_feature_incr(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD, tx);
mutex_enter(&vd->vdev_rebuild_lock);
bzero(vrp, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES);
vrp->vrp_rebuild_state = VDEV_REBUILD_ACTIVE;
vrp->vrp_min_txg = 0;
vrp->vrp_max_txg = dmu_tx_get_txg(tx);
vrp->vrp_start_time = gethrestime_sec();
vrp->vrp_scan_time_ms = 0;
vr->vr_prev_scan_time_ms = 0;
/*
* Rebuilds are currently only used when replacing a device, in which
* case there must be DTL_MISSING entries. In the future, we could
* allow rebuilds to be used in a way similar to a scrub. This would
* be useful because it would allow us to rebuild the space used by
* pool checkpoints.
*/
VERIFY(vdev_resilver_needed(vd, &vrp->vrp_min_txg, &vrp->vrp_max_txg));
VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
REBUILD_PHYS_ENTRIES, vrp, tx));
spa_history_log_internal(spa, "rebuild", tx,
"vdev_id=%llu vdev_guid=%llu started",
(u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid);
ASSERT3P(vd->vdev_rebuild_thread, ==, NULL);
vd->vdev_rebuild_thread = thread_create(NULL, 0,
vdev_rebuild_thread, vd, 0, &p0, TS_RUN, maxclsyspri);
mutex_exit(&vd->vdev_rebuild_lock);
}
static void
vdev_rebuild_log_notify(spa_t *spa, vdev_t *vd, char *name)
{
nvlist_t *aux = fnvlist_alloc();
fnvlist_add_string(aux, ZFS_EV_RESILVER_TYPE, "sequential");
spa_event_notify(spa, vd, aux, name);
nvlist_free(aux);
}
/*
* Called to request that a new rebuild be started. The feature will remain
* active for the duration of the rebuild, then revert to the enabled state.
*/
static void
vdev_rebuild_initiate(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
ASSERT(vd->vdev_top == vd);
ASSERT(MUTEX_HELD(&vd->vdev_rebuild_lock));
ASSERT(!vd->vdev_rebuilding);
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
vd->vdev_rebuilding = B_TRUE;
dsl_sync_task_nowait(spa_get_dsl(spa), vdev_rebuild_initiate_sync,
(void *)(uintptr_t)vd->vdev_id, tx);
dmu_tx_commit(tx);
vdev_rebuild_log_notify(spa, vd, ESC_ZFS_RESILVER_START);
}
/*
* Update the on-disk state to completed when a rebuild finishes.
*/
static void
vdev_rebuild_complete_sync(void *arg, dmu_tx_t *tx)
{
int vdev_id = (uintptr_t)arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *vd = vdev_lookup_top(spa, vdev_id);
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
mutex_enter(&vd->vdev_rebuild_lock);
vrp->vrp_rebuild_state = VDEV_REBUILD_COMPLETE;
vrp->vrp_end_time = gethrestime_sec();
VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
REBUILD_PHYS_ENTRIES, vrp, tx));
vdev_dtl_reassess(vd, tx->tx_txg, vrp->vrp_max_txg, B_TRUE, B_TRUE);
spa_feature_decr(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD, tx);
spa_history_log_internal(spa, "rebuild", tx,
"vdev_id=%llu vdev_guid=%llu complete",
(u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid);
vdev_rebuild_log_notify(spa, vd, ESC_ZFS_RESILVER_FINISH);
/* Handles detaching of spares */
spa_async_request(spa, SPA_ASYNC_REBUILD_DONE);
vd->vdev_rebuilding = B_FALSE;
mutex_exit(&vd->vdev_rebuild_lock);
/*
* While we're in syncing context take the opportunity to
* setup the scrub when there are no more active rebuilds.
*/
pool_scan_func_t func = POOL_SCAN_SCRUB;
if (dsl_scan_setup_check(&func, tx) == 0 &&
zfs_rebuild_scrub_enabled) {
dsl_scan_setup_sync(&func, tx);
}
cv_broadcast(&vd->vdev_rebuild_cv);
/* Clear recent error events (i.e. duplicate events tracking) */
zfs_ereport_clear(spa, NULL);
}
/*
* Update the on-disk state to canceled when a rebuild finishes.
*/
static void
vdev_rebuild_cancel_sync(void *arg, dmu_tx_t *tx)
{
int vdev_id = (uintptr_t)arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *vd = vdev_lookup_top(spa, vdev_id);
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
mutex_enter(&vd->vdev_rebuild_lock);
vrp->vrp_rebuild_state = VDEV_REBUILD_CANCELED;
vrp->vrp_end_time = gethrestime_sec();
VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
REBUILD_PHYS_ENTRIES, vrp, tx));
spa_feature_decr(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD, tx);
spa_history_log_internal(spa, "rebuild", tx,
"vdev_id=%llu vdev_guid=%llu canceled",
(u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid);
vdev_rebuild_log_notify(spa, vd, ESC_ZFS_RESILVER_FINISH);
vd->vdev_rebuild_cancel_wanted = B_FALSE;
vd->vdev_rebuilding = B_FALSE;
mutex_exit(&vd->vdev_rebuild_lock);
spa_notify_waiters(spa);
cv_broadcast(&vd->vdev_rebuild_cv);
}
/*
* Resets the progress of a running rebuild. This will occur when a new
* vdev is added to rebuild.
*/
static void
vdev_rebuild_reset_sync(void *arg, dmu_tx_t *tx)
{
int vdev_id = (uintptr_t)arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *vd = vdev_lookup_top(spa, vdev_id);
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
mutex_enter(&vd->vdev_rebuild_lock);
ASSERT(vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE);
ASSERT3P(vd->vdev_rebuild_thread, ==, NULL);
vrp->vrp_last_offset = 0;
vrp->vrp_min_txg = 0;
vrp->vrp_max_txg = dmu_tx_get_txg(tx);
vrp->vrp_bytes_scanned = 0;
vrp->vrp_bytes_issued = 0;
vrp->vrp_bytes_rebuilt = 0;
vrp->vrp_bytes_est = 0;
vrp->vrp_scan_time_ms = 0;
vr->vr_prev_scan_time_ms = 0;
/* See vdev_rebuild_initiate_sync comment */
VERIFY(vdev_resilver_needed(vd, &vrp->vrp_min_txg, &vrp->vrp_max_txg));
VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
REBUILD_PHYS_ENTRIES, vrp, tx));
spa_history_log_internal(spa, "rebuild", tx,
"vdev_id=%llu vdev_guid=%llu reset",
(u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid);
vd->vdev_rebuild_reset_wanted = B_FALSE;
ASSERT(vd->vdev_rebuilding);
vd->vdev_rebuild_thread = thread_create(NULL, 0,
vdev_rebuild_thread, vd, 0, &p0, TS_RUN, maxclsyspri);
mutex_exit(&vd->vdev_rebuild_lock);
}
/*
* Clear the last rebuild status.
*/
void
vdev_rebuild_clear_sync(void *arg, dmu_tx_t *tx)
{
int vdev_id = (uintptr_t)arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *vd = vdev_lookup_top(spa, vdev_id);
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
objset_t *mos = spa_meta_objset(spa);
mutex_enter(&vd->vdev_rebuild_lock);
if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD) ||
vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE) {
mutex_exit(&vd->vdev_rebuild_lock);
return;
}
clear_rebuild_bytes(vd);
bzero(vrp, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES);
if (vd->vdev_top_zap != 0 && zap_contains(mos, vd->vdev_top_zap,
VDEV_TOP_ZAP_VDEV_REBUILD_PHYS) == 0) {
VERIFY0(zap_update(mos, vd->vdev_top_zap,
VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
REBUILD_PHYS_ENTRIES, vrp, tx));
}
mutex_exit(&vd->vdev_rebuild_lock);
}
/*
* The zio_done_func_t callback for each rebuild I/O issued. It's responsible
* for updating the rebuild stats and limiting the number of in flight I/Os.
*/
static void
vdev_rebuild_cb(zio_t *zio)
{
vdev_rebuild_t *vr = zio->io_private;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
vdev_t *vd = vr->vr_top_vdev;
mutex_enter(&vr->vr_io_lock);
if (zio->io_error == ENXIO && !vdev_writeable(vd)) {
/*
* The I/O failed because the top-level vdev was unavailable.
* Attempt to roll back to the last completed offset, in order
* resume from the correct location if the pool is resumed.
* (This works because spa_sync waits on spa_txg_zio before
* it runs sync tasks.)
*/
uint64_t *off = &vr->vr_scan_offset[zio->io_txg & TXG_MASK];
*off = MIN(*off, zio->io_offset);
} else if (zio->io_error) {
vrp->vrp_errors++;
}
abd_free(zio->io_abd);
ASSERT3U(vr->vr_bytes_inflight, >, 0);
vr->vr_bytes_inflight -= zio->io_size;
cv_broadcast(&vr->vr_io_cv);
mutex_exit(&vr->vr_io_lock);
spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
}
/*
* Initialize a block pointer that can be used to read the given segment
* for sequential rebuild.
*/
static void
vdev_rebuild_blkptr_init(blkptr_t *bp, vdev_t *vd, uint64_t start,
uint64_t asize)
{
ASSERT(vd->vdev_ops == &vdev_draid_ops ||
vd->vdev_ops == &vdev_mirror_ops ||
vd->vdev_ops == &vdev_replacing_ops ||
vd->vdev_ops == &vdev_spare_ops);
uint64_t psize = vd->vdev_ops == &vdev_draid_ops ?
vdev_draid_asize_to_psize(vd, asize) : asize;
BP_ZERO(bp);
DVA_SET_VDEV(&bp->blk_dva[0], vd->vdev_id);
DVA_SET_OFFSET(&bp->blk_dva[0], start);
DVA_SET_GANG(&bp->blk_dva[0], 0);
DVA_SET_ASIZE(&bp->blk_dva[0], asize);
BP_SET_BIRTH(bp, TXG_INITIAL, TXG_INITIAL);
BP_SET_LSIZE(bp, psize);
BP_SET_PSIZE(bp, psize);
BP_SET_COMPRESS(bp, ZIO_COMPRESS_OFF);
BP_SET_CHECKSUM(bp, ZIO_CHECKSUM_OFF);
BP_SET_TYPE(bp, DMU_OT_NONE);
BP_SET_LEVEL(bp, 0);
BP_SET_DEDUP(bp, 0);
BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);
}
/*
* Issues a rebuild I/O and takes care of rate limiting the number of queued
* rebuild I/Os. The provided start and size must be properly aligned for the
* top-level vdev type being rebuilt.
*/
static int
vdev_rebuild_range(vdev_rebuild_t *vr, uint64_t start, uint64_t size)
{
uint64_t ms_id __maybe_unused = vr->vr_scan_msp->ms_id;
vdev_t *vd = vr->vr_top_vdev;
spa_t *spa = vd->vdev_spa;
blkptr_t blk;
ASSERT3U(ms_id, ==, start >> vd->vdev_ms_shift);
ASSERT3U(ms_id, ==, (start + size - 1) >> vd->vdev_ms_shift);
vr->vr_pass_bytes_scanned += size;
vr->vr_rebuild_phys.vrp_bytes_scanned += size;
/*
* Rebuild the data in this range by constructing a special block
* pointer. It has no relation to any existing blocks in the pool.
* However, by disabling checksum verification and issuing a scrub IO
* we can reconstruct and repair any children with missing data.
*/
vdev_rebuild_blkptr_init(&blk, vd, start, size);
uint64_t psize = BP_GET_PSIZE(&blk);
if (!vdev_dtl_need_resilver(vd, &blk.blk_dva[0], psize, TXG_UNKNOWN))
return (0);
mutex_enter(&vr->vr_io_lock);
/* Limit in flight rebuild I/Os */
while (vr->vr_bytes_inflight >= vr->vr_bytes_inflight_max)
cv_wait(&vr->vr_io_cv, &vr->vr_io_lock);
vr->vr_bytes_inflight += psize;
mutex_exit(&vr->vr_io_lock);
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
uint64_t txg = dmu_tx_get_txg(tx);
spa_config_enter(spa, SCL_STATE_ALL, vd, RW_READER);
mutex_enter(&vd->vdev_rebuild_lock);
/* This is the first I/O for this txg. */
if (vr->vr_scan_offset[txg & TXG_MASK] == 0) {
vr->vr_scan_offset[txg & TXG_MASK] = start;
dsl_sync_task_nowait(spa_get_dsl(spa),
vdev_rebuild_update_sync,
(void *)(uintptr_t)vd->vdev_id, tx);
}
/* When exiting write out our progress. */
if (vdev_rebuild_should_stop(vd)) {
mutex_enter(&vr->vr_io_lock);
vr->vr_bytes_inflight -= psize;
mutex_exit(&vr->vr_io_lock);
spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
mutex_exit(&vd->vdev_rebuild_lock);
dmu_tx_commit(tx);
return (SET_ERROR(EINTR));
}
mutex_exit(&vd->vdev_rebuild_lock);
dmu_tx_commit(tx);
vr->vr_scan_offset[txg & TXG_MASK] = start + size;
vr->vr_pass_bytes_issued += size;
vr->vr_rebuild_phys.vrp_bytes_issued += size;
zio_nowait(zio_read(spa->spa_txg_zio[txg & TXG_MASK], spa, &blk,
abd_alloc(psize, B_FALSE), psize, vdev_rebuild_cb, vr,
ZIO_PRIORITY_REBUILD, ZIO_FLAG_RAW | ZIO_FLAG_CANFAIL |
ZIO_FLAG_RESILVER, NULL));
return (0);
}
/*
* Issues rebuild I/Os for all ranges in the provided vr->vr_tree range tree.
*/
static int
vdev_rebuild_ranges(vdev_rebuild_t *vr)
{
vdev_t *vd = vr->vr_top_vdev;
zfs_btree_t *t = &vr->vr_scan_tree->rt_root;
zfs_btree_index_t idx;
int error;
for (range_seg_t *rs = zfs_btree_first(t, &idx); rs != NULL;
rs = zfs_btree_next(t, &idx, &idx)) {
uint64_t start = rs_get_start(rs, vr->vr_scan_tree);
uint64_t size = rs_get_end(rs, vr->vr_scan_tree) - start;
/*
* zfs_scan_suspend_progress can be set to disable rebuild
* progress for testing. See comment in dsl_scan_sync().
*/
while (zfs_scan_suspend_progress &&
!vdev_rebuild_should_stop(vd)) {
delay(hz);
}
while (size > 0) {
uint64_t chunk_size;
/*
* Split range into legally-sized logical chunks
* given the constraints of the top-level vdev
* being rebuilt (dRAID or mirror).
*/
ASSERT3P(vd->vdev_ops, !=, NULL);
chunk_size = vd->vdev_ops->vdev_op_rebuild_asize(vd,
start, size, zfs_rebuild_max_segment);
error = vdev_rebuild_range(vr, start, chunk_size);
if (error != 0)
return (error);
size -= chunk_size;
start += chunk_size;
}
}
return (0);
}
/*
* Calculates the estimated capacity which remains to be scanned. Since
* we traverse the pool in metaslab order only allocated capacity beyond
* the vrp_last_offset need be considered. All lower offsets must have
* already been rebuilt and are thus already included in vrp_bytes_scanned.
*/
static void
vdev_rebuild_update_bytes_est(vdev_t *vd, uint64_t ms_id)
{
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
uint64_t bytes_est = vrp->vrp_bytes_scanned;
if (vrp->vrp_last_offset < vd->vdev_ms[ms_id]->ms_start)
return;
for (uint64_t i = ms_id; i < vd->vdev_ms_count; i++) {
metaslab_t *msp = vd->vdev_ms[i];
mutex_enter(&msp->ms_lock);
bytes_est += metaslab_allocated_space(msp);
mutex_exit(&msp->ms_lock);
}
vrp->vrp_bytes_est = bytes_est;
}
/*
* Load from disk the top-level vdev's rebuild information.
*/
int
vdev_rebuild_load(vdev_t *vd)
{
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
spa_t *spa = vd->vdev_spa;
int err = 0;
mutex_enter(&vd->vdev_rebuild_lock);
vd->vdev_rebuilding = B_FALSE;
if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD)) {
bzero(vrp, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES);
mutex_exit(&vd->vdev_rebuild_lock);
return (SET_ERROR(ENOTSUP));
}
ASSERT(vd->vdev_top == vd);
err = zap_lookup(spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
REBUILD_PHYS_ENTRIES, vrp);
/*
* A missing or damaged VDEV_TOP_ZAP_VDEV_REBUILD_PHYS should
* not prevent a pool from being imported. Clear the rebuild
* status allowing a new resilver/rebuild to be started.
*/
if (err == ENOENT || err == EOVERFLOW || err == ECKSUM) {
bzero(vrp, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES);
} else if (err) {
mutex_exit(&vd->vdev_rebuild_lock);
return (err);
}
vr->vr_prev_scan_time_ms = vrp->vrp_scan_time_ms;
vr->vr_top_vdev = vd;
mutex_exit(&vd->vdev_rebuild_lock);
return (0);
}
/*
* Each scan thread is responsible for rebuilding a top-level vdev. The
* rebuild progress in tracked on-disk in VDEV_TOP_ZAP_VDEV_REBUILD_PHYS.
*/
static void
vdev_rebuild_thread(void *arg)
{
vdev_t *vd = arg;
spa_t *spa = vd->vdev_spa;
int error = 0;
/*
* If there's a scrub in process request that it be stopped. This
* is not required for a correct rebuild, but we do want rebuilds to
* emulate the resilver behavior as much as possible.
*/
dsl_pool_t *dsl = spa_get_dsl(spa);
if (dsl_scan_scrubbing(dsl))
dsl_scan_cancel(dsl);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
mutex_enter(&vd->vdev_rebuild_lock);
ASSERT3P(vd->vdev_top, ==, vd);
ASSERT3P(vd->vdev_rebuild_thread, !=, NULL);
ASSERT(vd->vdev_rebuilding);
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REBUILD));
ASSERT3B(vd->vdev_rebuild_cancel_wanted, ==, B_FALSE);
ASSERT3B(vd->vdev_rebuild_reset_wanted, ==, B_FALSE);
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
vr->vr_top_vdev = vd;
vr->vr_scan_msp = NULL;
vr->vr_scan_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
mutex_init(&vr->vr_io_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&vr->vr_io_cv, NULL, CV_DEFAULT, NULL);
vr->vr_pass_start_time = gethrtime();
vr->vr_pass_bytes_scanned = 0;
vr->vr_pass_bytes_issued = 0;
vr->vr_bytes_inflight_max = MAX(1ULL << 20,
zfs_rebuild_vdev_limit * vd->vdev_children);
uint64_t update_est_time = gethrtime();
vdev_rebuild_update_bytes_est(vd, 0);
clear_rebuild_bytes(vr->vr_top_vdev);
mutex_exit(&vd->vdev_rebuild_lock);
/*
* Systematically walk the metaslabs and issue rebuild I/Os for
* all ranges in the allocated space map.
*/
for (uint64_t i = 0; i < vd->vdev_ms_count; i++) {
metaslab_t *msp = vd->vdev_ms[i];
vr->vr_scan_msp = msp;
/*
* Removal of vdevs from the vdev tree may eliminate the need
* for the rebuild, in which case it should be canceled. The
* vdev_rebuild_cancel_wanted flag is set until the sync task
* completes. This may be after the rebuild thread exits.
*/
if (vdev_rebuild_should_cancel(vd)) {
vd->vdev_rebuild_cancel_wanted = B_TRUE;
error = EINTR;
break;
}
ASSERT0(range_tree_space(vr->vr_scan_tree));
/* Disable any new allocations to this metaslab */
spa_config_exit(spa, SCL_CONFIG, FTAG);
metaslab_disable(msp);
mutex_enter(&msp->ms_sync_lock);
mutex_enter(&msp->ms_lock);
/*
* If there are outstanding allocations wait for them to be
* synced. This is needed to ensure all allocated ranges are
* on disk and therefore will be rebuilt.
*/
for (int j = 0; j < TXG_SIZE; j++) {
if (range_tree_space(msp->ms_allocating[j])) {
mutex_exit(&msp->ms_lock);
mutex_exit(&msp->ms_sync_lock);
txg_wait_synced(dsl, 0);
mutex_enter(&msp->ms_sync_lock);
mutex_enter(&msp->ms_lock);
break;
}
}
/*
* When a metaslab has been allocated from read its allocated
* ranges from the space map object into the vr_scan_tree.
* Then add inflight / unflushed ranges and remove inflight /
* unflushed frees. This is the minimum range to be rebuilt.
*/
if (msp->ms_sm != NULL) {
VERIFY0(space_map_load(msp->ms_sm,
vr->vr_scan_tree, SM_ALLOC));
for (int i = 0; i < TXG_SIZE; i++) {
ASSERT0(range_tree_space(
msp->ms_allocating[i]));
}
range_tree_walk(msp->ms_unflushed_allocs,
range_tree_add, vr->vr_scan_tree);
range_tree_walk(msp->ms_unflushed_frees,
range_tree_remove, vr->vr_scan_tree);
/*
* Remove ranges which have already been rebuilt based
* on the last offset. This can happen when restarting
* a scan after exporting and re-importing the pool.
*/
range_tree_clear(vr->vr_scan_tree, 0,
vrp->vrp_last_offset);
}
mutex_exit(&msp->ms_lock);
mutex_exit(&msp->ms_sync_lock);
/*
* To provide an accurate estimate re-calculate the estimated
* size every 5 minutes to account for recent allocations and
* frees made to space maps which have not yet been rebuilt.
*/
if (gethrtime() > update_est_time + SEC2NSEC(300)) {
update_est_time = gethrtime();
vdev_rebuild_update_bytes_est(vd, i);
}
/*
* Walk the allocated space map and issue the rebuild I/O.
*/
error = vdev_rebuild_ranges(vr);
range_tree_vacate(vr->vr_scan_tree, NULL, NULL);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
metaslab_enable(msp, B_FALSE, B_FALSE);
if (error != 0)
break;
}
range_tree_destroy(vr->vr_scan_tree);
spa_config_exit(spa, SCL_CONFIG, FTAG);
/* Wait for any remaining rebuild I/O to complete */
mutex_enter(&vr->vr_io_lock);
while (vr->vr_bytes_inflight > 0)
cv_wait(&vr->vr_io_cv, &vr->vr_io_lock);
mutex_exit(&vr->vr_io_lock);
mutex_destroy(&vr->vr_io_lock);
cv_destroy(&vr->vr_io_cv);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
dsl_pool_t *dp = spa_get_dsl(spa);
dmu_tx_t *tx = dmu_tx_create_dd(dp->dp_mos_dir);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
mutex_enter(&vd->vdev_rebuild_lock);
if (error == 0) {
/*
* After a successful rebuild clear the DTLs of all ranges
* which were missing when the rebuild was started. These
* ranges must have been rebuilt as a consequence of rebuilding
* all allocated space. Note that unlike a scrub or resilver
* the rebuild operation will reconstruct data only referenced
* by a pool checkpoint. See the dsl_scan_done() comments.
*/
dsl_sync_task_nowait(dp, vdev_rebuild_complete_sync,
(void *)(uintptr_t)vd->vdev_id, tx);
} else if (vd->vdev_rebuild_cancel_wanted) {
/*
* The rebuild operation was canceled. This will occur when
* a device participating in the rebuild is detached.
*/
dsl_sync_task_nowait(dp, vdev_rebuild_cancel_sync,
(void *)(uintptr_t)vd->vdev_id, tx);
} else if (vd->vdev_rebuild_reset_wanted) {
/*
* Reset the running rebuild without canceling and restarting
* it. This will occur when a new device is attached and must
* participate in the rebuild.
*/
dsl_sync_task_nowait(dp, vdev_rebuild_reset_sync,
(void *)(uintptr_t)vd->vdev_id, tx);
} else {
/*
* The rebuild operation should be suspended. This may occur
* when detaching a child vdev or when exporting the pool. The
* rebuild is left in the active state so it will be resumed.
*/
ASSERT(vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE);
vd->vdev_rebuilding = B_FALSE;
}
dmu_tx_commit(tx);
vd->vdev_rebuild_thread = NULL;
mutex_exit(&vd->vdev_rebuild_lock);
spa_config_exit(spa, SCL_CONFIG, FTAG);
cv_broadcast(&vd->vdev_rebuild_cv);
thread_exit();
}
/*
* Returns B_TRUE if any top-level vdev are rebuilding.
*/
boolean_t
vdev_rebuild_active(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
boolean_t ret = B_FALSE;
if (vd == spa->spa_root_vdev) {
for (uint64_t i = 0; i < vd->vdev_children; i++) {
ret = vdev_rebuild_active(vd->vdev_child[i]);
if (ret)
return (ret);
}
} else if (vd->vdev_top_zap != 0) {
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
mutex_enter(&vd->vdev_rebuild_lock);
ret = (vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE);
mutex_exit(&vd->vdev_rebuild_lock);
}
return (ret);
}
/*
* Start a rebuild operation. The rebuild may be restarted when the
* top-level vdev is currently actively rebuilding.
*/
void
vdev_rebuild(vdev_t *vd)
{
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp __maybe_unused = &vr->vr_rebuild_phys;
ASSERT(vd->vdev_top == vd);
ASSERT(vdev_is_concrete(vd));
ASSERT(!vd->vdev_removing);
ASSERT(spa_feature_is_enabled(vd->vdev_spa,
SPA_FEATURE_DEVICE_REBUILD));
mutex_enter(&vd->vdev_rebuild_lock);
if (vd->vdev_rebuilding) {
ASSERT3U(vrp->vrp_rebuild_state, ==, VDEV_REBUILD_ACTIVE);
/*
* Signal a running rebuild operation that it should restart
* from the beginning because a new device was attached. The
* vdev_rebuild_reset_wanted flag is set until the sync task
* completes. This may be after the rebuild thread exits.
*/
if (!vd->vdev_rebuild_reset_wanted)
vd->vdev_rebuild_reset_wanted = B_TRUE;
} else {
vdev_rebuild_initiate(vd);
}
mutex_exit(&vd->vdev_rebuild_lock);
}
static void
vdev_rebuild_restart_impl(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
if (vd == spa->spa_root_vdev) {
for (uint64_t i = 0; i < vd->vdev_children; i++)
vdev_rebuild_restart_impl(vd->vdev_child[i]);
} else if (vd->vdev_top_zap != 0) {
vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
mutex_enter(&vd->vdev_rebuild_lock);
if (vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE &&
vdev_writeable(vd) && !vd->vdev_rebuilding) {
ASSERT(spa_feature_is_active(spa,
SPA_FEATURE_DEVICE_REBUILD));
vd->vdev_rebuilding = B_TRUE;
vd->vdev_rebuild_thread = thread_create(NULL, 0,
vdev_rebuild_thread, vd, 0, &p0, TS_RUN,
maxclsyspri);
}
mutex_exit(&vd->vdev_rebuild_lock);
}
}
/*
* Conditionally restart all of the vdev_rebuild_thread's for a pool. The
* feature flag must be active and the rebuild in the active state. This
* cannot be used to start a new rebuild.
*/
void
vdev_rebuild_restart(spa_t *spa)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
vdev_rebuild_restart_impl(spa->spa_root_vdev);
}
/*
* Stop and wait for all of the vdev_rebuild_thread's associated with the
* vdev tree provide to be terminated (canceled or stopped).
*/
void
vdev_rebuild_stop_wait(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (vd == spa->spa_root_vdev) {
for (uint64_t i = 0; i < vd->vdev_children; i++)
vdev_rebuild_stop_wait(vd->vdev_child[i]);
} else if (vd->vdev_top_zap != 0) {
ASSERT(vd == vd->vdev_top);
mutex_enter(&vd->vdev_rebuild_lock);
if (vd->vdev_rebuild_thread != NULL) {
vd->vdev_rebuild_exit_wanted = B_TRUE;
while (vd->vdev_rebuilding) {
cv_wait(&vd->vdev_rebuild_cv,
&vd->vdev_rebuild_lock);
}
vd->vdev_rebuild_exit_wanted = B_FALSE;
}
mutex_exit(&vd->vdev_rebuild_lock);
}
}
/*
* Stop all rebuild operations but leave them in the active state so they
* will be resumed when importing the pool.
*/
void
vdev_rebuild_stop_all(spa_t *spa)
{
vdev_rebuild_stop_wait(spa->spa_root_vdev);
}
/*
* Rebuild statistics reported per top-level vdev.
*/
int
vdev_rebuild_get_stats(vdev_t *tvd, vdev_rebuild_stat_t *vrs)
{
spa_t *spa = tvd->vdev_spa;
if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD))
return (SET_ERROR(ENOTSUP));
if (tvd != tvd->vdev_top || tvd->vdev_top_zap == 0)
return (SET_ERROR(EINVAL));
int error = zap_contains(spa_meta_objset(spa),
tvd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS);
if (error == ENOENT) {
bzero(vrs, sizeof (vdev_rebuild_stat_t));
vrs->vrs_state = VDEV_REBUILD_NONE;
error = 0;
} else if (error == 0) {
vdev_rebuild_t *vr = &tvd->vdev_rebuild_config;
vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
mutex_enter(&tvd->vdev_rebuild_lock);
vrs->vrs_state = vrp->vrp_rebuild_state;
vrs->vrs_start_time = vrp->vrp_start_time;
vrs->vrs_end_time = vrp->vrp_end_time;
vrs->vrs_scan_time_ms = vrp->vrp_scan_time_ms;
vrs->vrs_bytes_scanned = vrp->vrp_bytes_scanned;
vrs->vrs_bytes_issued = vrp->vrp_bytes_issued;
vrs->vrs_bytes_rebuilt = vrp->vrp_bytes_rebuilt;
vrs->vrs_bytes_est = vrp->vrp_bytes_est;
vrs->vrs_errors = vrp->vrp_errors;
vrs->vrs_pass_time_ms = NSEC2MSEC(gethrtime() -
vr->vr_pass_start_time);
vrs->vrs_pass_bytes_scanned = vr->vr_pass_bytes_scanned;
vrs->vrs_pass_bytes_issued = vr->vr_pass_bytes_issued;
mutex_exit(&tvd->vdev_rebuild_lock);
}
return (error);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, rebuild_max_segment, ULONG, ZMOD_RW,
"Max segment size in bytes of rebuild reads");
ZFS_MODULE_PARAM(zfs, zfs_, rebuild_vdev_limit, ULONG, ZMOD_RW,
"Max bytes in flight per leaf vdev for sequential resilvers");
ZFS_MODULE_PARAM(zfs, zfs_, rebuild_scrub_enabled, INT, ZMOD_RW,
"Automatically scrub after sequential resilver completes");
/* END CSTYLED */
diff --git a/module/zfs/vdev_removal.c b/module/zfs/vdev_removal.c
index 2a3ee90fe020..149de633dca8 100644
--- a/module/zfs/vdev_removal.c
+++ b/module/zfs/vdev_removal.c
@@ -1,2570 +1,2570 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa_impl.h>
#include <sys/dmu.h>
#include <sys/dmu_tx.h>
#include <sys/zap.h>
#include <sys/vdev_impl.h>
#include <sys/metaslab.h>
#include <sys/metaslab_impl.h>
#include <sys/uberblock_impl.h>
#include <sys/txg.h>
#include <sys/avl.h>
#include <sys/bpobj.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_synctask.h>
#include <sys/dsl_dir.h>
#include <sys/arc.h>
#include <sys/zfeature.h>
#include <sys/vdev_indirect_births.h>
#include <sys/vdev_indirect_mapping.h>
#include <sys/abd.h>
#include <sys/vdev_initialize.h>
#include <sys/vdev_trim.h>
#include <sys/trace_zfs.h>
/*
* This file contains the necessary logic to remove vdevs from a
* storage pool. Currently, the only devices that can be removed
* are log, cache, and spare devices; and top level vdevs from a pool
* w/o raidz or mirrors. (Note that members of a mirror can be removed
* by the detach operation.)
*
* Log vdevs are removed by evacuating them and then turning the vdev
* into a hole vdev while holding spa config locks.
*
* Top level vdevs are removed and converted into an indirect vdev via
* a multi-step process:
*
* - Disable allocations from this device (spa_vdev_remove_top).
*
* - From a new thread (spa_vdev_remove_thread), copy data from
* the removing vdev to a different vdev. The copy happens in open
* context (spa_vdev_copy_impl) and issues a sync task
* (vdev_mapping_sync) so the sync thread can update the partial
* indirect mappings in core and on disk.
*
* - If a free happens during a removal, it is freed from the
* removing vdev, and if it has already been copied, from the new
* location as well (free_from_removing_vdev).
*
* - After the removal is completed, the copy thread converts the vdev
* into an indirect vdev (vdev_remove_complete) before instructing
* the sync thread to destroy the space maps and finish the removal
* (spa_finish_removal).
*/
typedef struct vdev_copy_arg {
metaslab_t *vca_msp;
uint64_t vca_outstanding_bytes;
uint64_t vca_read_error_bytes;
uint64_t vca_write_error_bytes;
kcondvar_t vca_cv;
kmutex_t vca_lock;
} vdev_copy_arg_t;
/*
* The maximum amount of memory we can use for outstanding i/o while
* doing a device removal. This determines how much i/o we can have
* in flight concurrently.
*/
-int zfs_remove_max_copy_bytes = 64 * 1024 * 1024;
+static const int zfs_remove_max_copy_bytes = 64 * 1024 * 1024;
/*
* The largest contiguous segment that we will attempt to allocate when
* removing a device. This can be no larger than SPA_MAXBLOCKSIZE. If
* there is a performance problem with attempting to allocate large blocks,
* consider decreasing this.
*
* See also the accessor function spa_remove_max_segment().
*/
int zfs_remove_max_segment = SPA_MAXBLOCKSIZE;
/*
* Ignore hard IO errors during device removal. When set if a device
* encounters hard IO error during the removal process the removal will
* not be cancelled. This can result in a normally recoverable block
* becoming permanently damaged and is not recommended.
*/
-int zfs_removal_ignore_errors = 0;
+static int zfs_removal_ignore_errors = 0;
/*
* Allow a remap segment to span free chunks of at most this size. The main
* impact of a larger span is that we will read and write larger, more
* contiguous chunks, with more "unnecessary" data -- trading off bandwidth
* for iops. The value here was chosen to align with
* zfs_vdev_read_gap_limit, which is a similar concept when doing regular
* reads (but there's no reason it has to be the same).
*
* Additionally, a higher span will have the following relatively minor
* effects:
* - the mapping will be smaller, since one entry can cover more allocated
* segments
* - more of the fragmentation in the removing device will be preserved
* - we'll do larger allocations, which may fail and fall back on smaller
* allocations
*/
int vdev_removal_max_span = 32 * 1024;
/*
* This is used by the test suite so that it can ensure that certain
* actions happen while in the middle of a removal.
*/
int zfs_removal_suspend_progress = 0;
#define VDEV_REMOVAL_ZAP_OBJS "lzap"
static void spa_vdev_remove_thread(void *arg);
static int spa_vdev_remove_cancel_impl(spa_t *spa);
static void
spa_sync_removing_state(spa_t *spa, dmu_tx_t *tx)
{
VERIFY0(zap_update(spa->spa_dsl_pool->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_REMOVING, sizeof (uint64_t),
sizeof (spa->spa_removing_phys) / sizeof (uint64_t),
&spa->spa_removing_phys, tx));
}
static nvlist_t *
spa_nvlist_lookup_by_guid(nvlist_t **nvpp, int count, uint64_t target_guid)
{
for (int i = 0; i < count; i++) {
uint64_t guid =
fnvlist_lookup_uint64(nvpp[i], ZPOOL_CONFIG_GUID);
if (guid == target_guid)
return (nvpp[i]);
}
return (NULL);
}
static void
vdev_activate(vdev_t *vd)
{
metaslab_group_t *mg = vd->vdev_mg;
spa_t *spa = vd->vdev_spa;
uint64_t vdev_space = spa_deflate(spa) ?
vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;
ASSERT(!vd->vdev_islog);
ASSERT(vd->vdev_noalloc);
metaslab_group_activate(mg);
metaslab_group_activate(vd->vdev_log_mg);
ASSERT3U(spa->spa_nonallocating_dspace, >=, vdev_space);
spa->spa_nonallocating_dspace -= vdev_space;
vd->vdev_noalloc = B_FALSE;
}
static int
vdev_passivate(vdev_t *vd, uint64_t *txg)
{
spa_t *spa = vd->vdev_spa;
int error;
ASSERT(!vd->vdev_noalloc);
vdev_t *rvd = spa->spa_root_vdev;
metaslab_group_t *mg = vd->vdev_mg;
metaslab_class_t *normal = spa_normal_class(spa);
if (mg->mg_class == normal) {
/*
* We must check that this is not the only allocating device in
* the pool before passivating, otherwise we will not be able
* to make progress because we can't allocate from any vdevs.
*/
boolean_t last = B_TRUE;
for (uint64_t id = 0; id < rvd->vdev_children; id++) {
vdev_t *cvd = rvd->vdev_child[id];
if (cvd == vd ||
cvd->vdev_ops == &vdev_indirect_ops)
continue;
metaslab_class_t *mc = cvd->vdev_mg->mg_class;
if (mc != normal)
continue;
if (!cvd->vdev_noalloc) {
last = B_FALSE;
break;
}
}
if (last)
return (SET_ERROR(EINVAL));
}
metaslab_group_passivate(mg);
ASSERT(!vd->vdev_islog);
metaslab_group_passivate(vd->vdev_log_mg);
/*
* Wait for the youngest allocations and frees to sync,
* and then wait for the deferral of those frees to finish.
*/
spa_vdev_config_exit(spa, NULL,
*txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG);
/*
* We must ensure that no "stubby" log blocks are allocated
* on the device to be removed. These blocks could be
* written at any time, including while we are in the middle
* of copying them.
*/
error = spa_reset_logs(spa);
*txg = spa_vdev_config_enter(spa);
if (error != 0) {
metaslab_group_activate(mg);
ASSERT(!vd->vdev_islog);
if (vd->vdev_log_mg != NULL)
metaslab_group_activate(vd->vdev_log_mg);
return (error);
}
spa->spa_nonallocating_dspace += spa_deflate(spa) ?
vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;
vd->vdev_noalloc = B_TRUE;
return (0);
}
/*
* Turn off allocations for a top-level device from the pool.
*
* Turning off allocations for a top-level device can take a significant
* amount of time. As a result we use the spa_vdev_config_[enter/exit]
* functions which allow us to grab and release the spa_config_lock while
* still holding the namespace lock. During each step the configuration
* is synced out.
*/
int
spa_vdev_noalloc(spa_t *spa, uint64_t guid)
{
vdev_t *vd;
uint64_t txg;
int error = 0;
ASSERT(!MUTEX_HELD(&spa_namespace_lock));
ASSERT(spa_writeable(spa));
txg = spa_vdev_enter(spa);
ASSERT(MUTEX_HELD(&spa_namespace_lock));
vd = spa_lookup_by_guid(spa, guid, B_FALSE);
if (vd == NULL)
error = SET_ERROR(ENOENT);
else if (vd->vdev_mg == NULL)
error = SET_ERROR(ZFS_ERR_VDEV_NOTSUP);
else if (!vd->vdev_noalloc)
error = vdev_passivate(vd, &txg);
if (error == 0) {
vdev_dirty_leaves(vd, VDD_DTL, txg);
vdev_config_dirty(vd);
}
error = spa_vdev_exit(spa, NULL, txg, error);
return (error);
}
int
spa_vdev_alloc(spa_t *spa, uint64_t guid)
{
vdev_t *vd;
uint64_t txg;
int error = 0;
ASSERT(!MUTEX_HELD(&spa_namespace_lock));
ASSERT(spa_writeable(spa));
txg = spa_vdev_enter(spa);
ASSERT(MUTEX_HELD(&spa_namespace_lock));
vd = spa_lookup_by_guid(spa, guid, B_FALSE);
if (vd == NULL)
error = SET_ERROR(ENOENT);
else if (vd->vdev_mg == NULL)
error = SET_ERROR(ZFS_ERR_VDEV_NOTSUP);
else if (!vd->vdev_removing)
vdev_activate(vd);
if (error == 0) {
vdev_dirty_leaves(vd, VDD_DTL, txg);
vdev_config_dirty(vd);
}
(void) spa_vdev_exit(spa, NULL, txg, error);
return (error);
}
static void
spa_vdev_remove_aux(nvlist_t *config, char *name, nvlist_t **dev, int count,
nvlist_t *dev_to_remove)
{
nvlist_t **newdev = NULL;
if (count > 1)
newdev = kmem_alloc((count - 1) * sizeof (void *), KM_SLEEP);
for (int i = 0, j = 0; i < count; i++) {
if (dev[i] == dev_to_remove)
continue;
VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_SLEEP) == 0);
}
VERIFY(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY) == 0);
fnvlist_add_nvlist_array(config, name, (const nvlist_t * const *)newdev,
count - 1);
for (int i = 0; i < count - 1; i++)
nvlist_free(newdev[i]);
if (count > 1)
kmem_free(newdev, (count - 1) * sizeof (void *));
}
static spa_vdev_removal_t *
spa_vdev_removal_create(vdev_t *vd)
{
spa_vdev_removal_t *svr = kmem_zalloc(sizeof (*svr), KM_SLEEP);
mutex_init(&svr->svr_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&svr->svr_cv, NULL, CV_DEFAULT, NULL);
svr->svr_allocd_segs = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
svr->svr_vdev_id = vd->vdev_id;
for (int i = 0; i < TXG_SIZE; i++) {
svr->svr_frees[i] = range_tree_create(NULL, RANGE_SEG64, NULL,
0, 0);
list_create(&svr->svr_new_segments[i],
sizeof (vdev_indirect_mapping_entry_t),
offsetof(vdev_indirect_mapping_entry_t, vime_node));
}
return (svr);
}
void
spa_vdev_removal_destroy(spa_vdev_removal_t *svr)
{
for (int i = 0; i < TXG_SIZE; i++) {
ASSERT0(svr->svr_bytes_done[i]);
ASSERT0(svr->svr_max_offset_to_sync[i]);
range_tree_destroy(svr->svr_frees[i]);
list_destroy(&svr->svr_new_segments[i]);
}
range_tree_destroy(svr->svr_allocd_segs);
mutex_destroy(&svr->svr_lock);
cv_destroy(&svr->svr_cv);
kmem_free(svr, sizeof (*svr));
}
/*
* This is called as a synctask in the txg in which we will mark this vdev
* as removing (in the config stored in the MOS).
*
* It begins the evacuation of a toplevel vdev by:
* - initializing the spa_removing_phys which tracks this removal
* - computing the amount of space to remove for accounting purposes
* - dirtying all dbufs in the spa_config_object
* - creating the spa_vdev_removal
* - starting the spa_vdev_remove_thread
*/
static void
vdev_remove_initiate_sync(void *arg, dmu_tx_t *tx)
{
int vdev_id = (uintptr_t)arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *vd = vdev_lookup_top(spa, vdev_id);
vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
objset_t *mos = spa->spa_dsl_pool->dp_meta_objset;
spa_vdev_removal_t *svr = NULL;
uint64_t txg __maybe_unused = dmu_tx_get_txg(tx);
ASSERT0(vdev_get_nparity(vd));
svr = spa_vdev_removal_create(vd);
ASSERT(vd->vdev_removing);
ASSERT3P(vd->vdev_indirect_mapping, ==, NULL);
spa_feature_incr(spa, SPA_FEATURE_DEVICE_REMOVAL, tx);
if (spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS)) {
/*
* By activating the OBSOLETE_COUNTS feature, we prevent
* the pool from being downgraded and ensure that the
* refcounts are precise.
*/
spa_feature_incr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
uint64_t one = 1;
VERIFY0(zap_add(spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_OBSOLETE_COUNTS_ARE_PRECISE, sizeof (one), 1,
&one, tx));
boolean_t are_precise __maybe_unused;
ASSERT0(vdev_obsolete_counts_are_precise(vd, &are_precise));
ASSERT3B(are_precise, ==, B_TRUE);
}
vic->vic_mapping_object = vdev_indirect_mapping_alloc(mos, tx);
vd->vdev_indirect_mapping =
vdev_indirect_mapping_open(mos, vic->vic_mapping_object);
vic->vic_births_object = vdev_indirect_births_alloc(mos, tx);
vd->vdev_indirect_births =
vdev_indirect_births_open(mos, vic->vic_births_object);
spa->spa_removing_phys.sr_removing_vdev = vd->vdev_id;
spa->spa_removing_phys.sr_start_time = gethrestime_sec();
spa->spa_removing_phys.sr_end_time = 0;
spa->spa_removing_phys.sr_state = DSS_SCANNING;
spa->spa_removing_phys.sr_to_copy = 0;
spa->spa_removing_phys.sr_copied = 0;
/*
* Note: We can't use vdev_stat's vs_alloc for sr_to_copy, because
* there may be space in the defer tree, which is free, but still
* counted in vs_alloc.
*/
for (uint64_t i = 0; i < vd->vdev_ms_count; i++) {
metaslab_t *ms = vd->vdev_ms[i];
if (ms->ms_sm == NULL)
continue;
spa->spa_removing_phys.sr_to_copy +=
metaslab_allocated_space(ms);
/*
* Space which we are freeing this txg does not need to
* be copied.
*/
spa->spa_removing_phys.sr_to_copy -=
range_tree_space(ms->ms_freeing);
ASSERT0(range_tree_space(ms->ms_freed));
for (int t = 0; t < TXG_SIZE; t++)
ASSERT0(range_tree_space(ms->ms_allocating[t]));
}
/*
* Sync tasks are called before metaslab_sync(), so there should
* be no already-synced metaslabs in the TXG_CLEAN list.
*/
ASSERT3P(txg_list_head(&vd->vdev_ms_list, TXG_CLEAN(txg)), ==, NULL);
spa_sync_removing_state(spa, tx);
/*
* All blocks that we need to read the most recent mapping must be
* stored on concrete vdevs. Therefore, we must dirty anything that
* is read before spa_remove_init(). Specifically, the
* spa_config_object. (Note that although we already modified the
* spa_config_object in spa_sync_removing_state, that may not have
* modified all blocks of the object.)
*/
dmu_object_info_t doi;
VERIFY0(dmu_object_info(mos, DMU_POOL_DIRECTORY_OBJECT, &doi));
for (uint64_t offset = 0; offset < doi.doi_max_offset; ) {
dmu_buf_t *dbuf;
VERIFY0(dmu_buf_hold(mos, DMU_POOL_DIRECTORY_OBJECT,
offset, FTAG, &dbuf, 0));
dmu_buf_will_dirty(dbuf, tx);
offset += dbuf->db_size;
dmu_buf_rele(dbuf, FTAG);
}
/*
* Now that we've allocated the im_object, dirty the vdev to ensure
* that the object gets written to the config on disk.
*/
vdev_config_dirty(vd);
zfs_dbgmsg("starting removal thread for vdev %llu (%px) in txg %llu "
"im_obj=%llu", (u_longlong_t)vd->vdev_id, vd,
(u_longlong_t)dmu_tx_get_txg(tx),
(u_longlong_t)vic->vic_mapping_object);
spa_history_log_internal(spa, "vdev remove started", tx,
"%s vdev %llu %s", spa_name(spa), (u_longlong_t)vd->vdev_id,
(vd->vdev_path != NULL) ? vd->vdev_path : "-");
/*
* Setting spa_vdev_removal causes subsequent frees to call
* free_from_removing_vdev(). Note that we don't need any locking
* because we are the sync thread, and metaslab_free_impl() is only
* called from syncing context (potentially from a zio taskq thread,
* but in any case only when there are outstanding free i/os, which
* there are not).
*/
ASSERT3P(spa->spa_vdev_removal, ==, NULL);
spa->spa_vdev_removal = svr;
svr->svr_thread = thread_create(NULL, 0,
spa_vdev_remove_thread, spa, 0, &p0, TS_RUN, minclsyspri);
}
/*
* When we are opening a pool, we must read the mapping for each
* indirect vdev in order from most recently removed to least
* recently removed. We do this because the blocks for the mapping
* of older indirect vdevs may be stored on more recently removed vdevs.
* In order to read each indirect mapping object, we must have
* initialized all more recently removed vdevs.
*/
int
spa_remove_init(spa_t *spa)
{
int error;
error = zap_lookup(spa->spa_dsl_pool->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_REMOVING, sizeof (uint64_t),
sizeof (spa->spa_removing_phys) / sizeof (uint64_t),
&spa->spa_removing_phys);
if (error == ENOENT) {
spa->spa_removing_phys.sr_state = DSS_NONE;
spa->spa_removing_phys.sr_removing_vdev = -1;
spa->spa_removing_phys.sr_prev_indirect_vdev = -1;
spa->spa_indirect_vdevs_loaded = B_TRUE;
return (0);
} else if (error != 0) {
return (error);
}
if (spa->spa_removing_phys.sr_state == DSS_SCANNING) {
/*
* We are currently removing a vdev. Create and
* initialize a spa_vdev_removal_t from the bonus
* buffer of the removing vdevs vdev_im_object, and
* initialize its partial mapping.
*/
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
vdev_t *vd = vdev_lookup_top(spa,
spa->spa_removing_phys.sr_removing_vdev);
if (vd == NULL) {
spa_config_exit(spa, SCL_STATE, FTAG);
return (EINVAL);
}
vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
ASSERT(vdev_is_concrete(vd));
spa_vdev_removal_t *svr = spa_vdev_removal_create(vd);
ASSERT3U(svr->svr_vdev_id, ==, vd->vdev_id);
ASSERT(vd->vdev_removing);
vd->vdev_indirect_mapping = vdev_indirect_mapping_open(
spa->spa_meta_objset, vic->vic_mapping_object);
vd->vdev_indirect_births = vdev_indirect_births_open(
spa->spa_meta_objset, vic->vic_births_object);
spa_config_exit(spa, SCL_STATE, FTAG);
spa->spa_vdev_removal = svr;
}
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
uint64_t indirect_vdev_id =
spa->spa_removing_phys.sr_prev_indirect_vdev;
while (indirect_vdev_id != UINT64_MAX) {
vdev_t *vd = vdev_lookup_top(spa, indirect_vdev_id);
vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
vd->vdev_indirect_mapping = vdev_indirect_mapping_open(
spa->spa_meta_objset, vic->vic_mapping_object);
vd->vdev_indirect_births = vdev_indirect_births_open(
spa->spa_meta_objset, vic->vic_births_object);
indirect_vdev_id = vic->vic_prev_indirect_vdev;
}
spa_config_exit(spa, SCL_STATE, FTAG);
/*
* Now that we've loaded all the indirect mappings, we can allow
* reads from other blocks (e.g. via predictive prefetch).
*/
spa->spa_indirect_vdevs_loaded = B_TRUE;
return (0);
}
void
spa_restart_removal(spa_t *spa)
{
spa_vdev_removal_t *svr = spa->spa_vdev_removal;
if (svr == NULL)
return;
/*
* In general when this function is called there is no
* removal thread running. The only scenario where this
* is not true is during spa_import() where this function
* is called twice [once from spa_import_impl() and
* spa_async_resume()]. Thus, in the scenario where we
* import a pool that has an ongoing removal we don't
* want to spawn a second thread.
*/
if (svr->svr_thread != NULL)
return;
if (!spa_writeable(spa))
return;
zfs_dbgmsg("restarting removal of %llu",
(u_longlong_t)svr->svr_vdev_id);
svr->svr_thread = thread_create(NULL, 0, spa_vdev_remove_thread, spa,
0, &p0, TS_RUN, minclsyspri);
}
/*
* Process freeing from a device which is in the middle of being removed.
* We must handle this carefully so that we attempt to copy freed data,
* and we correctly free already-copied data.
*/
void
free_from_removing_vdev(vdev_t *vd, uint64_t offset, uint64_t size)
{
spa_t *spa = vd->vdev_spa;
spa_vdev_removal_t *svr = spa->spa_vdev_removal;
vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
uint64_t txg = spa_syncing_txg(spa);
uint64_t max_offset_yet = 0;
ASSERT(vd->vdev_indirect_config.vic_mapping_object != 0);
ASSERT3U(vd->vdev_indirect_config.vic_mapping_object, ==,
vdev_indirect_mapping_object(vim));
ASSERT3U(vd->vdev_id, ==, svr->svr_vdev_id);
mutex_enter(&svr->svr_lock);
/*
* Remove the segment from the removing vdev's spacemap. This
* ensures that we will not attempt to copy this space (if the
* removal thread has not yet visited it), and also ensures
* that we know what is actually allocated on the new vdevs
* (needed if we cancel the removal).
*
* Note: we must do the metaslab_free_concrete() with the svr_lock
* held, so that the remove_thread can not load this metaslab and then
* visit this offset between the time that we metaslab_free_concrete()
* and when we check to see if it has been visited.
*
* Note: The checkpoint flag is set to false as having/taking
* a checkpoint and removing a device can't happen at the same
* time.
*/
ASSERT(!spa_has_checkpoint(spa));
metaslab_free_concrete(vd, offset, size, B_FALSE);
uint64_t synced_size = 0;
uint64_t synced_offset = 0;
uint64_t max_offset_synced = vdev_indirect_mapping_max_offset(vim);
if (offset < max_offset_synced) {
/*
* The mapping for this offset is already on disk.
* Free from the new location.
*
* Note that we use svr_max_synced_offset because it is
* updated atomically with respect to the in-core mapping.
* By contrast, vim_max_offset is not.
*
* This block may be split between a synced entry and an
* in-flight or unvisited entry. Only process the synced
* portion of it here.
*/
synced_size = MIN(size, max_offset_synced - offset);
synced_offset = offset;
ASSERT3U(max_offset_yet, <=, max_offset_synced);
max_offset_yet = max_offset_synced;
DTRACE_PROBE3(remove__free__synced,
spa_t *, spa,
uint64_t, offset,
uint64_t, synced_size);
size -= synced_size;
offset += synced_size;
}
/*
* Look at all in-flight txgs starting from the currently syncing one
* and see if a section of this free is being copied. By starting from
* this txg and iterating forward, we might find that this region
* was copied in two different txgs and handle it appropriately.
*/
for (int i = 0; i < TXG_CONCURRENT_STATES; i++) {
int txgoff = (txg + i) & TXG_MASK;
if (size > 0 && offset < svr->svr_max_offset_to_sync[txgoff]) {
/*
* The mapping for this offset is in flight, and
* will be synced in txg+i.
*/
uint64_t inflight_size = MIN(size,
svr->svr_max_offset_to_sync[txgoff] - offset);
DTRACE_PROBE4(remove__free__inflight,
spa_t *, spa,
uint64_t, offset,
uint64_t, inflight_size,
uint64_t, txg + i);
/*
* We copy data in order of increasing offset.
* Therefore the max_offset_to_sync[] must increase
* (or be zero, indicating that nothing is being
* copied in that txg).
*/
if (svr->svr_max_offset_to_sync[txgoff] != 0) {
ASSERT3U(svr->svr_max_offset_to_sync[txgoff],
>=, max_offset_yet);
max_offset_yet =
svr->svr_max_offset_to_sync[txgoff];
}
/*
* We've already committed to copying this segment:
* we have allocated space elsewhere in the pool for
* it and have an IO outstanding to copy the data. We
* cannot free the space before the copy has
* completed, or else the copy IO might overwrite any
* new data. To free that space, we record the
* segment in the appropriate svr_frees tree and free
* the mapped space later, in the txg where we have
* completed the copy and synced the mapping (see
* vdev_mapping_sync).
*/
range_tree_add(svr->svr_frees[txgoff],
offset, inflight_size);
size -= inflight_size;
offset += inflight_size;
/*
* This space is already accounted for as being
* done, because it is being copied in txg+i.
* However, if i!=0, then it is being copied in
* a future txg. If we crash after this txg
* syncs but before txg+i syncs, then the space
* will be free. Therefore we must account
* for the space being done in *this* txg
* (when it is freed) rather than the future txg
* (when it will be copied).
*/
ASSERT3U(svr->svr_bytes_done[txgoff], >=,
inflight_size);
svr->svr_bytes_done[txgoff] -= inflight_size;
svr->svr_bytes_done[txg & TXG_MASK] += inflight_size;
}
}
ASSERT0(svr->svr_max_offset_to_sync[TXG_CLEAN(txg) & TXG_MASK]);
if (size > 0) {
/*
* The copy thread has not yet visited this offset. Ensure
* that it doesn't.
*/
DTRACE_PROBE3(remove__free__unvisited,
spa_t *, spa,
uint64_t, offset,
uint64_t, size);
if (svr->svr_allocd_segs != NULL)
range_tree_clear(svr->svr_allocd_segs, offset, size);
/*
* Since we now do not need to copy this data, for
* accounting purposes we have done our job and can count
* it as completed.
*/
svr->svr_bytes_done[txg & TXG_MASK] += size;
}
mutex_exit(&svr->svr_lock);
/*
* Now that we have dropped svr_lock, process the synced portion
* of this free.
*/
if (synced_size > 0) {
vdev_indirect_mark_obsolete(vd, synced_offset, synced_size);
/*
* Note: this can only be called from syncing context,
* and the vdev_indirect_mapping is only changed from the
* sync thread, so we don't need svr_lock while doing
* metaslab_free_impl_cb.
*/
boolean_t checkpoint = B_FALSE;
vdev_indirect_ops.vdev_op_remap(vd, synced_offset, synced_size,
metaslab_free_impl_cb, &checkpoint);
}
}
/*
* Stop an active removal and update the spa_removing phys.
*/
static void
spa_finish_removal(spa_t *spa, dsl_scan_state_t state, dmu_tx_t *tx)
{
spa_vdev_removal_t *svr = spa->spa_vdev_removal;
ASSERT3U(dmu_tx_get_txg(tx), ==, spa_syncing_txg(spa));
/* Ensure the removal thread has completed before we free the svr. */
spa_vdev_remove_suspend(spa);
ASSERT(state == DSS_FINISHED || state == DSS_CANCELED);
if (state == DSS_FINISHED) {
spa_removing_phys_t *srp = &spa->spa_removing_phys;
vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
if (srp->sr_prev_indirect_vdev != -1) {
vdev_t *pvd;
pvd = vdev_lookup_top(spa,
srp->sr_prev_indirect_vdev);
ASSERT3P(pvd->vdev_ops, ==, &vdev_indirect_ops);
}
vic->vic_prev_indirect_vdev = srp->sr_prev_indirect_vdev;
srp->sr_prev_indirect_vdev = vd->vdev_id;
}
spa->spa_removing_phys.sr_state = state;
spa->spa_removing_phys.sr_end_time = gethrestime_sec();
spa->spa_vdev_removal = NULL;
spa_vdev_removal_destroy(svr);
spa_sync_removing_state(spa, tx);
spa_notify_waiters(spa);
vdev_config_dirty(spa->spa_root_vdev);
}
static void
free_mapped_segment_cb(void *arg, uint64_t offset, uint64_t size)
{
vdev_t *vd = arg;
vdev_indirect_mark_obsolete(vd, offset, size);
boolean_t checkpoint = B_FALSE;
vdev_indirect_ops.vdev_op_remap(vd, offset, size,
metaslab_free_impl_cb, &checkpoint);
}
/*
* On behalf of the removal thread, syncs an incremental bit more of
* the indirect mapping to disk and updates the in-memory mapping.
* Called as a sync task in every txg that the removal thread makes progress.
*/
static void
vdev_mapping_sync(void *arg, dmu_tx_t *tx)
{
spa_vdev_removal_t *svr = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
vdev_indirect_config_t *vic __maybe_unused = &vd->vdev_indirect_config;
uint64_t txg = dmu_tx_get_txg(tx);
vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
ASSERT(vic->vic_mapping_object != 0);
ASSERT3U(txg, ==, spa_syncing_txg(spa));
vdev_indirect_mapping_add_entries(vim,
&svr->svr_new_segments[txg & TXG_MASK], tx);
vdev_indirect_births_add_entry(vd->vdev_indirect_births,
vdev_indirect_mapping_max_offset(vim), dmu_tx_get_txg(tx), tx);
/*
* Free the copied data for anything that was freed while the
* mapping entries were in flight.
*/
mutex_enter(&svr->svr_lock);
range_tree_vacate(svr->svr_frees[txg & TXG_MASK],
free_mapped_segment_cb, vd);
ASSERT3U(svr->svr_max_offset_to_sync[txg & TXG_MASK], >=,
vdev_indirect_mapping_max_offset(vim));
svr->svr_max_offset_to_sync[txg & TXG_MASK] = 0;
mutex_exit(&svr->svr_lock);
spa_sync_removing_state(spa, tx);
}
typedef struct vdev_copy_segment_arg {
spa_t *vcsa_spa;
dva_t *vcsa_dest_dva;
uint64_t vcsa_txg;
range_tree_t *vcsa_obsolete_segs;
} vdev_copy_segment_arg_t;
static void
unalloc_seg(void *arg, uint64_t start, uint64_t size)
{
vdev_copy_segment_arg_t *vcsa = arg;
spa_t *spa = vcsa->vcsa_spa;
blkptr_t bp = { { { {0} } } };
BP_SET_BIRTH(&bp, TXG_INITIAL, TXG_INITIAL);
BP_SET_LSIZE(&bp, size);
BP_SET_PSIZE(&bp, size);
BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_OFF);
BP_SET_TYPE(&bp, DMU_OT_NONE);
BP_SET_LEVEL(&bp, 0);
BP_SET_DEDUP(&bp, 0);
BP_SET_BYTEORDER(&bp, ZFS_HOST_BYTEORDER);
DVA_SET_VDEV(&bp.blk_dva[0], DVA_GET_VDEV(vcsa->vcsa_dest_dva));
DVA_SET_OFFSET(&bp.blk_dva[0],
DVA_GET_OFFSET(vcsa->vcsa_dest_dva) + start);
DVA_SET_ASIZE(&bp.blk_dva[0], size);
zio_free(spa, vcsa->vcsa_txg, &bp);
}
/*
* All reads and writes associated with a call to spa_vdev_copy_segment()
* are done.
*/
static void
spa_vdev_copy_segment_done(zio_t *zio)
{
vdev_copy_segment_arg_t *vcsa = zio->io_private;
range_tree_vacate(vcsa->vcsa_obsolete_segs,
unalloc_seg, vcsa);
range_tree_destroy(vcsa->vcsa_obsolete_segs);
kmem_free(vcsa, sizeof (*vcsa));
spa_config_exit(zio->io_spa, SCL_STATE, zio->io_spa);
}
/*
* The write of the new location is done.
*/
static void
spa_vdev_copy_segment_write_done(zio_t *zio)
{
vdev_copy_arg_t *vca = zio->io_private;
abd_free(zio->io_abd);
mutex_enter(&vca->vca_lock);
vca->vca_outstanding_bytes -= zio->io_size;
if (zio->io_error != 0)
vca->vca_write_error_bytes += zio->io_size;
cv_signal(&vca->vca_cv);
mutex_exit(&vca->vca_lock);
}
/*
* The read of the old location is done. The parent zio is the write to
* the new location. Allow it to start.
*/
static void
spa_vdev_copy_segment_read_done(zio_t *zio)
{
vdev_copy_arg_t *vca = zio->io_private;
if (zio->io_error != 0) {
mutex_enter(&vca->vca_lock);
vca->vca_read_error_bytes += zio->io_size;
mutex_exit(&vca->vca_lock);
}
zio_nowait(zio_unique_parent(zio));
}
/*
* If the old and new vdevs are mirrors, we will read both sides of the old
* mirror, and write each copy to the corresponding side of the new mirror.
* If the old and new vdevs have a different number of children, we will do
* this as best as possible. Since we aren't verifying checksums, this
* ensures that as long as there's a good copy of the data, we'll have a
* good copy after the removal, even if there's silent damage to one side
* of the mirror. If we're removing a mirror that has some silent damage,
* we'll have exactly the same damage in the new location (assuming that
* the new location is also a mirror).
*
* We accomplish this by creating a tree of zio_t's, with as many writes as
* there are "children" of the new vdev (a non-redundant vdev counts as one
* child, a 2-way mirror has 2 children, etc). Each write has an associated
* read from a child of the old vdev. Typically there will be the same
* number of children of the old and new vdevs. However, if there are more
* children of the new vdev, some child(ren) of the old vdev will be issued
* multiple reads. If there are more children of the old vdev, some copies
* will be dropped.
*
* For example, the tree of zio_t's for a 2-way mirror is:
*
* null
* / \
* write(new vdev, child 0) write(new vdev, child 1)
* | |
* read(old vdev, child 0) read(old vdev, child 1)
*
* Child zio's complete before their parents complete. However, zio's
* created with zio_vdev_child_io() may be issued before their children
* complete. In this case we need to make sure that the children (reads)
* complete before the parents (writes) are *issued*. We do this by not
* calling zio_nowait() on each write until its corresponding read has
* completed.
*
* The spa_config_lock must be held while zio's created by
* zio_vdev_child_io() are in progress, to ensure that the vdev tree does
* not change (e.g. due to a concurrent "zpool attach/detach"). The "null"
* zio is needed to release the spa_config_lock after all the reads and
* writes complete. (Note that we can't grab the config lock for each read,
* because it is not reentrant - we could deadlock with a thread waiting
* for a write lock.)
*/
static void
spa_vdev_copy_one_child(vdev_copy_arg_t *vca, zio_t *nzio,
vdev_t *source_vd, uint64_t source_offset,
vdev_t *dest_child_vd, uint64_t dest_offset, int dest_id, uint64_t size)
{
ASSERT3U(spa_config_held(nzio->io_spa, SCL_ALL, RW_READER), !=, 0);
/*
* If the destination child in unwritable then there is no point
* in issuing the source reads which cannot be written.
*/
if (!vdev_writeable(dest_child_vd))
return;
mutex_enter(&vca->vca_lock);
vca->vca_outstanding_bytes += size;
mutex_exit(&vca->vca_lock);
abd_t *abd = abd_alloc_for_io(size, B_FALSE);
vdev_t *source_child_vd = NULL;
if (source_vd->vdev_ops == &vdev_mirror_ops && dest_id != -1) {
/*
* Source and dest are both mirrors. Copy from the same
* child id as we are copying to (wrapping around if there
* are more dest children than source children). If the
* preferred source child is unreadable select another.
*/
for (int i = 0; i < source_vd->vdev_children; i++) {
source_child_vd = source_vd->vdev_child[
(dest_id + i) % source_vd->vdev_children];
if (vdev_readable(source_child_vd))
break;
}
} else {
source_child_vd = source_vd;
}
/*
* There should always be at least one readable source child or
* the pool would be in a suspended state. Somehow selecting an
* unreadable child would result in IO errors, the removal process
* being cancelled, and the pool reverting to its pre-removal state.
*/
ASSERT3P(source_child_vd, !=, NULL);
zio_t *write_zio = zio_vdev_child_io(nzio, NULL,
dest_child_vd, dest_offset, abd, size,
ZIO_TYPE_WRITE, ZIO_PRIORITY_REMOVAL,
ZIO_FLAG_CANFAIL,
spa_vdev_copy_segment_write_done, vca);
zio_nowait(zio_vdev_child_io(write_zio, NULL,
source_child_vd, source_offset, abd, size,
ZIO_TYPE_READ, ZIO_PRIORITY_REMOVAL,
ZIO_FLAG_CANFAIL,
spa_vdev_copy_segment_read_done, vca));
}
/*
* Allocate a new location for this segment, and create the zio_t's to
* read from the old location and write to the new location.
*/
static int
spa_vdev_copy_segment(vdev_t *vd, range_tree_t *segs,
uint64_t maxalloc, uint64_t txg,
vdev_copy_arg_t *vca, zio_alloc_list_t *zal)
{
metaslab_group_t *mg = vd->vdev_mg;
spa_t *spa = vd->vdev_spa;
spa_vdev_removal_t *svr = spa->spa_vdev_removal;
vdev_indirect_mapping_entry_t *entry;
dva_t dst = {{ 0 }};
uint64_t start = range_tree_min(segs);
ASSERT0(P2PHASE(start, 1 << spa->spa_min_ashift));
ASSERT3U(maxalloc, <=, SPA_MAXBLOCKSIZE);
ASSERT0(P2PHASE(maxalloc, 1 << spa->spa_min_ashift));
uint64_t size = range_tree_span(segs);
if (range_tree_span(segs) > maxalloc) {
/*
* We can't allocate all the segments. Prefer to end
* the allocation at the end of a segment, thus avoiding
* additional split blocks.
*/
range_seg_max_t search;
zfs_btree_index_t where;
rs_set_start(&search, segs, start + maxalloc);
rs_set_end(&search, segs, start + maxalloc);
(void) zfs_btree_find(&segs->rt_root, &search, &where);
range_seg_t *rs = zfs_btree_prev(&segs->rt_root, &where,
&where);
if (rs != NULL) {
size = rs_get_end(rs, segs) - start;
} else {
/*
* There are no segments that end before maxalloc.
* I.e. the first segment is larger than maxalloc,
* so we must split it.
*/
size = maxalloc;
}
}
ASSERT3U(size, <=, maxalloc);
ASSERT0(P2PHASE(size, 1 << spa->spa_min_ashift));
/*
* An allocation class might not have any remaining vdevs or space
*/
metaslab_class_t *mc = mg->mg_class;
if (mc->mc_groups == 0)
mc = spa_normal_class(spa);
int error = metaslab_alloc_dva(spa, mc, size, &dst, 0, NULL, txg, 0,
zal, 0);
if (error == ENOSPC && mc != spa_normal_class(spa)) {
error = metaslab_alloc_dva(spa, spa_normal_class(spa), size,
&dst, 0, NULL, txg, 0, zal, 0);
}
if (error != 0)
return (error);
/*
* Determine the ranges that are not actually needed. Offsets are
* relative to the start of the range to be copied (i.e. relative to the
* local variable "start").
*/
range_tree_t *obsolete_segs = range_tree_create(NULL, RANGE_SEG64, NULL,
0, 0);
zfs_btree_index_t where;
range_seg_t *rs = zfs_btree_first(&segs->rt_root, &where);
ASSERT3U(rs_get_start(rs, segs), ==, start);
uint64_t prev_seg_end = rs_get_end(rs, segs);
while ((rs = zfs_btree_next(&segs->rt_root, &where, &where)) != NULL) {
if (rs_get_start(rs, segs) >= start + size) {
break;
} else {
range_tree_add(obsolete_segs,
prev_seg_end - start,
rs_get_start(rs, segs) - prev_seg_end);
}
prev_seg_end = rs_get_end(rs, segs);
}
/* We don't end in the middle of an obsolete range */
ASSERT3U(start + size, <=, prev_seg_end);
range_tree_clear(segs, start, size);
/*
* We can't have any padding of the allocated size, otherwise we will
* misunderstand what's allocated, and the size of the mapping. We
* prevent padding by ensuring that all devices in the pool have the
* same ashift, and the allocation size is a multiple of the ashift.
*/
VERIFY3U(DVA_GET_ASIZE(&dst), ==, size);
entry = kmem_zalloc(sizeof (vdev_indirect_mapping_entry_t), KM_SLEEP);
DVA_MAPPING_SET_SRC_OFFSET(&entry->vime_mapping, start);
entry->vime_mapping.vimep_dst = dst;
if (spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS)) {
entry->vime_obsolete_count = range_tree_space(obsolete_segs);
}
vdev_copy_segment_arg_t *vcsa = kmem_zalloc(sizeof (*vcsa), KM_SLEEP);
vcsa->vcsa_dest_dva = &entry->vime_mapping.vimep_dst;
vcsa->vcsa_obsolete_segs = obsolete_segs;
vcsa->vcsa_spa = spa;
vcsa->vcsa_txg = txg;
/*
* See comment before spa_vdev_copy_one_child().
*/
spa_config_enter(spa, SCL_STATE, spa, RW_READER);
zio_t *nzio = zio_null(spa->spa_txg_zio[txg & TXG_MASK], spa, NULL,
spa_vdev_copy_segment_done, vcsa, 0);
vdev_t *dest_vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dst));
if (dest_vd->vdev_ops == &vdev_mirror_ops) {
for (int i = 0; i < dest_vd->vdev_children; i++) {
vdev_t *child = dest_vd->vdev_child[i];
spa_vdev_copy_one_child(vca, nzio, vd, start,
child, DVA_GET_OFFSET(&dst), i, size);
}
} else {
spa_vdev_copy_one_child(vca, nzio, vd, start,
dest_vd, DVA_GET_OFFSET(&dst), -1, size);
}
zio_nowait(nzio);
list_insert_tail(&svr->svr_new_segments[txg & TXG_MASK], entry);
ASSERT3U(start + size, <=, vd->vdev_ms_count << vd->vdev_ms_shift);
vdev_dirty(vd, 0, NULL, txg);
return (0);
}
/*
* Complete the removal of a toplevel vdev. This is called as a
* synctask in the same txg that we will sync out the new config (to the
* MOS object) which indicates that this vdev is indirect.
*/
static void
vdev_remove_complete_sync(void *arg, dmu_tx_t *tx)
{
spa_vdev_removal_t *svr = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
for (int i = 0; i < TXG_SIZE; i++) {
ASSERT0(svr->svr_bytes_done[i]);
}
ASSERT3U(spa->spa_removing_phys.sr_copied, ==,
spa->spa_removing_phys.sr_to_copy);
vdev_destroy_spacemaps(vd, tx);
/* destroy leaf zaps, if any */
ASSERT3P(svr->svr_zaplist, !=, NULL);
for (nvpair_t *pair = nvlist_next_nvpair(svr->svr_zaplist, NULL);
pair != NULL;
pair = nvlist_next_nvpair(svr->svr_zaplist, pair)) {
vdev_destroy_unlink_zap(vd, fnvpair_value_uint64(pair), tx);
}
fnvlist_free(svr->svr_zaplist);
spa_finish_removal(dmu_tx_pool(tx)->dp_spa, DSS_FINISHED, tx);
/* vd->vdev_path is not available here */
spa_history_log_internal(spa, "vdev remove completed", tx,
"%s vdev %llu", spa_name(spa), (u_longlong_t)vd->vdev_id);
}
static void
vdev_remove_enlist_zaps(vdev_t *vd, nvlist_t *zlist)
{
ASSERT3P(zlist, !=, NULL);
ASSERT0(vdev_get_nparity(vd));
if (vd->vdev_leaf_zap != 0) {
char zkey[32];
(void) snprintf(zkey, sizeof (zkey), "%s-%llu",
VDEV_REMOVAL_ZAP_OBJS, (u_longlong_t)vd->vdev_leaf_zap);
fnvlist_add_uint64(zlist, zkey, vd->vdev_leaf_zap);
}
for (uint64_t id = 0; id < vd->vdev_children; id++) {
vdev_remove_enlist_zaps(vd->vdev_child[id], zlist);
}
}
static void
vdev_remove_replace_with_indirect(vdev_t *vd, uint64_t txg)
{
vdev_t *ivd;
dmu_tx_t *tx;
spa_t *spa = vd->vdev_spa;
spa_vdev_removal_t *svr = spa->spa_vdev_removal;
/*
* First, build a list of leaf zaps to be destroyed.
* This is passed to the sync context thread,
* which does the actual unlinking.
*/
svr->svr_zaplist = fnvlist_alloc();
vdev_remove_enlist_zaps(vd, svr->svr_zaplist);
ivd = vdev_add_parent(vd, &vdev_indirect_ops);
ivd->vdev_removing = 0;
vd->vdev_leaf_zap = 0;
vdev_remove_child(ivd, vd);
vdev_compact_children(ivd);
ASSERT(!list_link_active(&vd->vdev_state_dirty_node));
mutex_enter(&svr->svr_lock);
svr->svr_thread = NULL;
cv_broadcast(&svr->svr_cv);
mutex_exit(&svr->svr_lock);
/* After this, we can not use svr. */
tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
dsl_sync_task_nowait(spa->spa_dsl_pool,
vdev_remove_complete_sync, svr, tx);
dmu_tx_commit(tx);
}
/*
* Complete the removal of a toplevel vdev. This is called in open
* context by the removal thread after we have copied all vdev's data.
*/
static void
vdev_remove_complete(spa_t *spa)
{
uint64_t txg;
/*
* Wait for any deferred frees to be synced before we call
* vdev_metaslab_fini()
*/
txg_wait_synced(spa->spa_dsl_pool, 0);
txg = spa_vdev_enter(spa);
vdev_t *vd = vdev_lookup_top(spa, spa->spa_vdev_removal->svr_vdev_id);
ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
ASSERT3P(vd->vdev_trim_thread, ==, NULL);
ASSERT3P(vd->vdev_autotrim_thread, ==, NULL);
uint64_t vdev_space = spa_deflate(spa) ?
vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;
sysevent_t *ev = spa_event_create(spa, vd, NULL,
ESC_ZFS_VDEV_REMOVE_DEV);
zfs_dbgmsg("finishing device removal for vdev %llu in txg %llu",
(u_longlong_t)vd->vdev_id, (u_longlong_t)txg);
ASSERT3U(0, !=, vdev_space);
ASSERT3U(spa->spa_nonallocating_dspace, >=, vdev_space);
/* the vdev is no longer part of the dspace */
spa->spa_nonallocating_dspace -= vdev_space;
/*
* Discard allocation state.
*/
if (vd->vdev_mg != NULL) {
vdev_metaslab_fini(vd);
metaslab_group_destroy(vd->vdev_mg);
vd->vdev_mg = NULL;
spa_log_sm_set_blocklimit(spa);
}
if (vd->vdev_log_mg != NULL) {
ASSERT0(vd->vdev_ms_count);
metaslab_group_destroy(vd->vdev_log_mg);
vd->vdev_log_mg = NULL;
}
ASSERT0(vd->vdev_stat.vs_space);
ASSERT0(vd->vdev_stat.vs_dspace);
vdev_remove_replace_with_indirect(vd, txg);
/*
* We now release the locks, allowing spa_sync to run and finish the
* removal via vdev_remove_complete_sync in syncing context.
*
* Note that we hold on to the vdev_t that has been replaced. Since
* it isn't part of the vdev tree any longer, it can't be concurrently
* manipulated, even while we don't have the config lock.
*/
(void) spa_vdev_exit(spa, NULL, txg, 0);
/*
* Top ZAP should have been transferred to the indirect vdev in
* vdev_remove_replace_with_indirect.
*/
ASSERT0(vd->vdev_top_zap);
/*
* Leaf ZAP should have been moved in vdev_remove_replace_with_indirect.
*/
ASSERT0(vd->vdev_leaf_zap);
txg = spa_vdev_enter(spa);
(void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE);
/*
* Request to update the config and the config cachefile.
*/
vdev_config_dirty(spa->spa_root_vdev);
(void) spa_vdev_exit(spa, vd, txg, 0);
if (ev != NULL)
spa_event_post(ev);
}
/*
* Evacuates a segment of size at most max_alloc from the vdev
* via repeated calls to spa_vdev_copy_segment. If an allocation
* fails, the pool is probably too fragmented to handle such a
* large size, so decrease max_alloc so that the caller will not try
* this size again this txg.
*/
static void
spa_vdev_copy_impl(vdev_t *vd, spa_vdev_removal_t *svr, vdev_copy_arg_t *vca,
uint64_t *max_alloc, dmu_tx_t *tx)
{
uint64_t txg = dmu_tx_get_txg(tx);
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
mutex_enter(&svr->svr_lock);
/*
* Determine how big of a chunk to copy. We can allocate up
* to max_alloc bytes, and we can span up to vdev_removal_max_span
* bytes of unallocated space at a time. "segs" will track the
* allocated segments that we are copying. We may also be copying
* free segments (of up to vdev_removal_max_span bytes).
*/
range_tree_t *segs = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
for (;;) {
range_tree_t *rt = svr->svr_allocd_segs;
range_seg_t *rs = range_tree_first(rt);
if (rs == NULL)
break;
uint64_t seg_length;
if (range_tree_is_empty(segs)) {
/* need to truncate the first seg based on max_alloc */
seg_length = MIN(rs_get_end(rs, rt) - rs_get_start(rs,
rt), *max_alloc);
} else {
if (rs_get_start(rs, rt) - range_tree_max(segs) >
vdev_removal_max_span) {
/*
* Including this segment would cause us to
* copy a larger unneeded chunk than is allowed.
*/
break;
} else if (rs_get_end(rs, rt) - range_tree_min(segs) >
*max_alloc) {
/*
* This additional segment would extend past
* max_alloc. Rather than splitting this
* segment, leave it for the next mapping.
*/
break;
} else {
seg_length = rs_get_end(rs, rt) -
rs_get_start(rs, rt);
}
}
range_tree_add(segs, rs_get_start(rs, rt), seg_length);
range_tree_remove(svr->svr_allocd_segs,
rs_get_start(rs, rt), seg_length);
}
if (range_tree_is_empty(segs)) {
mutex_exit(&svr->svr_lock);
range_tree_destroy(segs);
return;
}
if (svr->svr_max_offset_to_sync[txg & TXG_MASK] == 0) {
dsl_sync_task_nowait(dmu_tx_pool(tx), vdev_mapping_sync,
svr, tx);
}
svr->svr_max_offset_to_sync[txg & TXG_MASK] = range_tree_max(segs);
/*
* Note: this is the amount of *allocated* space
* that we are taking care of each txg.
*/
svr->svr_bytes_done[txg & TXG_MASK] += range_tree_space(segs);
mutex_exit(&svr->svr_lock);
zio_alloc_list_t zal;
metaslab_trace_init(&zal);
uint64_t thismax = SPA_MAXBLOCKSIZE;
while (!range_tree_is_empty(segs)) {
int error = spa_vdev_copy_segment(vd,
segs, thismax, txg, vca, &zal);
if (error == ENOSPC) {
/*
* Cut our segment in half, and don't try this
* segment size again this txg. Note that the
* allocation size must be aligned to the highest
* ashift in the pool, so that the allocation will
* not be padded out to a multiple of the ashift,
* which could cause us to think that this mapping
* is larger than we intended.
*/
ASSERT3U(spa->spa_max_ashift, >=, SPA_MINBLOCKSHIFT);
ASSERT3U(spa->spa_max_ashift, ==, spa->spa_min_ashift);
uint64_t attempted =
MIN(range_tree_span(segs), thismax);
thismax = P2ROUNDUP(attempted / 2,
1 << spa->spa_max_ashift);
/*
* The minimum-size allocation can not fail.
*/
ASSERT3U(attempted, >, 1 << spa->spa_max_ashift);
*max_alloc = attempted - (1 << spa->spa_max_ashift);
} else {
ASSERT0(error);
/*
* We've performed an allocation, so reset the
* alloc trace list.
*/
metaslab_trace_fini(&zal);
metaslab_trace_init(&zal);
}
}
metaslab_trace_fini(&zal);
range_tree_destroy(segs);
}
/*
* The size of each removal mapping is limited by the tunable
* zfs_remove_max_segment, but we must adjust this to be a multiple of the
* pool's ashift, so that we don't try to split individual sectors regardless
* of the tunable value. (Note that device removal requires that all devices
* have the same ashift, so there's no difference between spa_min_ashift and
* spa_max_ashift.) The raw tunable should not be used elsewhere.
*/
uint64_t
spa_remove_max_segment(spa_t *spa)
{
return (P2ROUNDUP(zfs_remove_max_segment, 1 << spa->spa_max_ashift));
}
/*
* The removal thread operates in open context. It iterates over all
* allocated space in the vdev, by loading each metaslab's spacemap.
* For each contiguous segment of allocated space (capping the segment
* size at SPA_MAXBLOCKSIZE), we:
* - Allocate space for it on another vdev.
* - Create a new mapping from the old location to the new location
* (as a record in svr_new_segments).
* - Initiate a physical read zio to get the data off the removing disk.
* - In the read zio's done callback, initiate a physical write zio to
* write it to the new vdev.
* Note that all of this will take effect when a particular TXG syncs.
* The sync thread ensures that all the phys reads and writes for the syncing
* TXG have completed (see spa_txg_zio) and writes the new mappings to disk
* (see vdev_mapping_sync()).
*/
static void
spa_vdev_remove_thread(void *arg)
{
spa_t *spa = arg;
spa_vdev_removal_t *svr = spa->spa_vdev_removal;
vdev_copy_arg_t vca;
uint64_t max_alloc = spa_remove_max_segment(spa);
uint64_t last_txg = 0;
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
uint64_t start_offset = vdev_indirect_mapping_max_offset(vim);
ASSERT3P(vd->vdev_ops, !=, &vdev_indirect_ops);
ASSERT(vdev_is_concrete(vd));
ASSERT(vd->vdev_removing);
ASSERT(vd->vdev_indirect_config.vic_mapping_object != 0);
ASSERT(vim != NULL);
mutex_init(&vca.vca_lock, NULL, MUTEX_DEFAULT, NULL);
cv_init(&vca.vca_cv, NULL, CV_DEFAULT, NULL);
vca.vca_outstanding_bytes = 0;
vca.vca_read_error_bytes = 0;
vca.vca_write_error_bytes = 0;
mutex_enter(&svr->svr_lock);
/*
* Start from vim_max_offset so we pick up where we left off
* if we are restarting the removal after opening the pool.
*/
uint64_t msi;
for (msi = start_offset >> vd->vdev_ms_shift;
msi < vd->vdev_ms_count && !svr->svr_thread_exit; msi++) {
metaslab_t *msp = vd->vdev_ms[msi];
ASSERT3U(msi, <=, vd->vdev_ms_count);
ASSERT0(range_tree_space(svr->svr_allocd_segs));
mutex_enter(&msp->ms_sync_lock);
mutex_enter(&msp->ms_lock);
/*
* Assert nothing in flight -- ms_*tree is empty.
*/
for (int i = 0; i < TXG_SIZE; i++) {
ASSERT0(range_tree_space(msp->ms_allocating[i]));
}
/*
* If the metaslab has ever been allocated from (ms_sm!=NULL),
* read the allocated segments from the space map object
* into svr_allocd_segs. Since we do this while holding
* svr_lock and ms_sync_lock, concurrent frees (which
* would have modified the space map) will wait for us
* to finish loading the spacemap, and then take the
* appropriate action (see free_from_removing_vdev()).
*/
if (msp->ms_sm != NULL) {
VERIFY0(space_map_load(msp->ms_sm,
svr->svr_allocd_segs, SM_ALLOC));
range_tree_walk(msp->ms_unflushed_allocs,
range_tree_add, svr->svr_allocd_segs);
range_tree_walk(msp->ms_unflushed_frees,
range_tree_remove, svr->svr_allocd_segs);
range_tree_walk(msp->ms_freeing,
range_tree_remove, svr->svr_allocd_segs);
/*
* When we are resuming from a paused removal (i.e.
* when importing a pool with a removal in progress),
* discard any state that we have already processed.
*/
range_tree_clear(svr->svr_allocd_segs, 0, start_offset);
}
mutex_exit(&msp->ms_lock);
mutex_exit(&msp->ms_sync_lock);
vca.vca_msp = msp;
zfs_dbgmsg("copying %llu segments for metaslab %llu",
(u_longlong_t)zfs_btree_numnodes(
&svr->svr_allocd_segs->rt_root),
(u_longlong_t)msp->ms_id);
while (!svr->svr_thread_exit &&
!range_tree_is_empty(svr->svr_allocd_segs)) {
mutex_exit(&svr->svr_lock);
/*
* We need to periodically drop the config lock so that
* writers can get in. Additionally, we can't wait
* for a txg to sync while holding a config lock
* (since a waiting writer could cause a 3-way deadlock
* with the sync thread, which also gets a config
* lock for reader). So we can't hold the config lock
* while calling dmu_tx_assign().
*/
spa_config_exit(spa, SCL_CONFIG, FTAG);
/*
* This delay will pause the removal around the point
* specified by zfs_removal_suspend_progress. We do this
* solely from the test suite or during debugging.
*/
while (zfs_removal_suspend_progress &&
!svr->svr_thread_exit)
delay(hz);
mutex_enter(&vca.vca_lock);
while (vca.vca_outstanding_bytes >
zfs_remove_max_copy_bytes) {
cv_wait(&vca.vca_cv, &vca.vca_lock);
}
mutex_exit(&vca.vca_lock);
dmu_tx_t *tx =
dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
uint64_t txg = dmu_tx_get_txg(tx);
/*
* Reacquire the vdev_config lock. The vdev_t
* that we're removing may have changed, e.g. due
* to a vdev_attach or vdev_detach.
*/
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vd = vdev_lookup_top(spa, svr->svr_vdev_id);
if (txg != last_txg)
max_alloc = spa_remove_max_segment(spa);
last_txg = txg;
spa_vdev_copy_impl(vd, svr, &vca, &max_alloc, tx);
dmu_tx_commit(tx);
mutex_enter(&svr->svr_lock);
}
mutex_enter(&vca.vca_lock);
if (zfs_removal_ignore_errors == 0 &&
(vca.vca_read_error_bytes > 0 ||
vca.vca_write_error_bytes > 0)) {
svr->svr_thread_exit = B_TRUE;
}
mutex_exit(&vca.vca_lock);
}
mutex_exit(&svr->svr_lock);
spa_config_exit(spa, SCL_CONFIG, FTAG);
/*
* Wait for all copies to finish before cleaning up the vca.
*/
txg_wait_synced(spa->spa_dsl_pool, 0);
ASSERT0(vca.vca_outstanding_bytes);
mutex_destroy(&vca.vca_lock);
cv_destroy(&vca.vca_cv);
if (svr->svr_thread_exit) {
mutex_enter(&svr->svr_lock);
range_tree_vacate(svr->svr_allocd_segs, NULL, NULL);
svr->svr_thread = NULL;
cv_broadcast(&svr->svr_cv);
mutex_exit(&svr->svr_lock);
/*
* During the removal process an unrecoverable read or write
* error was encountered. The removal process must be
* cancelled or this damage may become permanent.
*/
if (zfs_removal_ignore_errors == 0 &&
(vca.vca_read_error_bytes > 0 ||
vca.vca_write_error_bytes > 0)) {
zfs_dbgmsg("canceling removal due to IO errors: "
"[read_error_bytes=%llu] [write_error_bytes=%llu]",
(u_longlong_t)vca.vca_read_error_bytes,
(u_longlong_t)vca.vca_write_error_bytes);
spa_vdev_remove_cancel_impl(spa);
}
} else {
ASSERT0(range_tree_space(svr->svr_allocd_segs));
vdev_remove_complete(spa);
}
thread_exit();
}
void
spa_vdev_remove_suspend(spa_t *spa)
{
spa_vdev_removal_t *svr = spa->spa_vdev_removal;
if (svr == NULL)
return;
mutex_enter(&svr->svr_lock);
svr->svr_thread_exit = B_TRUE;
while (svr->svr_thread != NULL)
cv_wait(&svr->svr_cv, &svr->svr_lock);
svr->svr_thread_exit = B_FALSE;
mutex_exit(&svr->svr_lock);
}
/*
* Return true if the "allocating" property has been set to "off"
*/
static boolean_t
vdev_prop_allocating_off(vdev_t *vd)
{
uint64_t objid = vd->vdev_top_zap;
uint64_t allocating = 1;
/* no vdev property object => no props */
if (objid != 0) {
spa_t *spa = vd->vdev_spa;
objset_t *mos = spa->spa_meta_objset;
mutex_enter(&spa->spa_props_lock);
(void) zap_lookup(mos, objid, "allocating", sizeof (uint64_t),
1, &allocating);
mutex_exit(&spa->spa_props_lock);
}
return (allocating == 0);
}
static int
spa_vdev_remove_cancel_check(void *arg, dmu_tx_t *tx)
{
(void) arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
if (spa->spa_vdev_removal == NULL)
return (ENOTACTIVE);
return (0);
}
/*
* Cancel a removal by freeing all entries from the partial mapping
* and marking the vdev as no longer being removing.
*/
static void
spa_vdev_remove_cancel_sync(void *arg, dmu_tx_t *tx)
{
(void) arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
spa_vdev_removal_t *svr = spa->spa_vdev_removal;
vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
objset_t *mos = spa->spa_meta_objset;
ASSERT3P(svr->svr_thread, ==, NULL);
spa_feature_decr(spa, SPA_FEATURE_DEVICE_REMOVAL, tx);
boolean_t are_precise;
VERIFY0(vdev_obsolete_counts_are_precise(vd, &are_precise));
if (are_precise) {
spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
VERIFY0(zap_remove(spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_OBSOLETE_COUNTS_ARE_PRECISE, tx));
}
uint64_t obsolete_sm_object;
VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
if (obsolete_sm_object != 0) {
ASSERT(vd->vdev_obsolete_sm != NULL);
ASSERT3U(obsolete_sm_object, ==,
space_map_object(vd->vdev_obsolete_sm));
space_map_free(vd->vdev_obsolete_sm, tx);
VERIFY0(zap_remove(spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM, tx));
space_map_close(vd->vdev_obsolete_sm);
vd->vdev_obsolete_sm = NULL;
spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
}
for (int i = 0; i < TXG_SIZE; i++) {
ASSERT(list_is_empty(&svr->svr_new_segments[i]));
ASSERT3U(svr->svr_max_offset_to_sync[i], <=,
vdev_indirect_mapping_max_offset(vim));
}
for (uint64_t msi = 0; msi < vd->vdev_ms_count; msi++) {
metaslab_t *msp = vd->vdev_ms[msi];
if (msp->ms_start >= vdev_indirect_mapping_max_offset(vim))
break;
ASSERT0(range_tree_space(svr->svr_allocd_segs));
mutex_enter(&msp->ms_lock);
/*
* Assert nothing in flight -- ms_*tree is empty.
*/
for (int i = 0; i < TXG_SIZE; i++)
ASSERT0(range_tree_space(msp->ms_allocating[i]));
for (int i = 0; i < TXG_DEFER_SIZE; i++)
ASSERT0(range_tree_space(msp->ms_defer[i]));
ASSERT0(range_tree_space(msp->ms_freed));
if (msp->ms_sm != NULL) {
mutex_enter(&svr->svr_lock);
VERIFY0(space_map_load(msp->ms_sm,
svr->svr_allocd_segs, SM_ALLOC));
range_tree_walk(msp->ms_unflushed_allocs,
range_tree_add, svr->svr_allocd_segs);
range_tree_walk(msp->ms_unflushed_frees,
range_tree_remove, svr->svr_allocd_segs);
range_tree_walk(msp->ms_freeing,
range_tree_remove, svr->svr_allocd_segs);
/*
* Clear everything past what has been synced,
* because we have not allocated mappings for it yet.
*/
uint64_t syncd = vdev_indirect_mapping_max_offset(vim);
uint64_t sm_end = msp->ms_sm->sm_start +
msp->ms_sm->sm_size;
if (sm_end > syncd)
range_tree_clear(svr->svr_allocd_segs,
syncd, sm_end - syncd);
mutex_exit(&svr->svr_lock);
}
mutex_exit(&msp->ms_lock);
mutex_enter(&svr->svr_lock);
range_tree_vacate(svr->svr_allocd_segs,
free_mapped_segment_cb, vd);
mutex_exit(&svr->svr_lock);
}
/*
* Note: this must happen after we invoke free_mapped_segment_cb,
* because it adds to the obsolete_segments.
*/
range_tree_vacate(vd->vdev_obsolete_segments, NULL, NULL);
ASSERT3U(vic->vic_mapping_object, ==,
vdev_indirect_mapping_object(vd->vdev_indirect_mapping));
vdev_indirect_mapping_close(vd->vdev_indirect_mapping);
vd->vdev_indirect_mapping = NULL;
vdev_indirect_mapping_free(mos, vic->vic_mapping_object, tx);
vic->vic_mapping_object = 0;
ASSERT3U(vic->vic_births_object, ==,
vdev_indirect_births_object(vd->vdev_indirect_births));
vdev_indirect_births_close(vd->vdev_indirect_births);
vd->vdev_indirect_births = NULL;
vdev_indirect_births_free(mos, vic->vic_births_object, tx);
vic->vic_births_object = 0;
/*
* We may have processed some frees from the removing vdev in this
* txg, thus increasing svr_bytes_done; discard that here to
* satisfy the assertions in spa_vdev_removal_destroy().
* Note that future txg's can not have any bytes_done, because
* future TXG's are only modified from open context, and we have
* already shut down the copying thread.
*/
svr->svr_bytes_done[dmu_tx_get_txg(tx) & TXG_MASK] = 0;
spa_finish_removal(spa, DSS_CANCELED, tx);
vd->vdev_removing = B_FALSE;
if (!vdev_prop_allocating_off(vd)) {
spa_config_enter(spa, SCL_ALLOC | SCL_VDEV, FTAG, RW_WRITER);
vdev_activate(vd);
spa_config_exit(spa, SCL_ALLOC | SCL_VDEV, FTAG);
}
vdev_config_dirty(vd);
zfs_dbgmsg("canceled device removal for vdev %llu in %llu",
(u_longlong_t)vd->vdev_id, (u_longlong_t)dmu_tx_get_txg(tx));
spa_history_log_internal(spa, "vdev remove canceled", tx,
"%s vdev %llu %s", spa_name(spa),
(u_longlong_t)vd->vdev_id,
(vd->vdev_path != NULL) ? vd->vdev_path : "-");
}
static int
spa_vdev_remove_cancel_impl(spa_t *spa)
{
int error = dsl_sync_task(spa->spa_name, spa_vdev_remove_cancel_check,
spa_vdev_remove_cancel_sync, NULL, 0,
ZFS_SPACE_CHECK_EXTRA_RESERVED);
return (error);
}
int
spa_vdev_remove_cancel(spa_t *spa)
{
spa_vdev_remove_suspend(spa);
if (spa->spa_vdev_removal == NULL)
return (ENOTACTIVE);
return (spa_vdev_remove_cancel_impl(spa));
}
void
svr_sync(spa_t *spa, dmu_tx_t *tx)
{
spa_vdev_removal_t *svr = spa->spa_vdev_removal;
int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;
if (svr == NULL)
return;
/*
* This check is necessary so that we do not dirty the
* DIRECTORY_OBJECT via spa_sync_removing_state() when there
* is nothing to do. Dirtying it every time would prevent us
* from syncing-to-convergence.
*/
if (svr->svr_bytes_done[txgoff] == 0)
return;
/*
* Update progress accounting.
*/
spa->spa_removing_phys.sr_copied += svr->svr_bytes_done[txgoff];
svr->svr_bytes_done[txgoff] = 0;
spa_sync_removing_state(spa, tx);
}
static void
vdev_remove_make_hole_and_free(vdev_t *vd)
{
uint64_t id = vd->vdev_id;
spa_t *spa = vd->vdev_spa;
vdev_t *rvd = spa->spa_root_vdev;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
vdev_free(vd);
vd = vdev_alloc_common(spa, id, 0, &vdev_hole_ops);
vdev_add_child(rvd, vd);
vdev_config_dirty(rvd);
/*
* Reassess the health of our root vdev.
*/
vdev_reopen(rvd);
}
/*
* Remove a log device. The config lock is held for the specified TXG.
*/
static int
spa_vdev_remove_log(vdev_t *vd, uint64_t *txg)
{
metaslab_group_t *mg = vd->vdev_mg;
spa_t *spa = vd->vdev_spa;
int error = 0;
ASSERT(vd->vdev_islog);
ASSERT(vd == vd->vdev_top);
ASSERT3P(vd->vdev_log_mg, ==, NULL);
ASSERT(MUTEX_HELD(&spa_namespace_lock));
/*
* Stop allocating from this vdev.
*/
metaslab_group_passivate(mg);
/*
* Wait for the youngest allocations and frees to sync,
* and then wait for the deferral of those frees to finish.
*/
spa_vdev_config_exit(spa, NULL,
*txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG);
/*
* Cancel any initialize or TRIM which was in progress.
*/
vdev_initialize_stop_all(vd, VDEV_INITIALIZE_CANCELED);
vdev_trim_stop_all(vd, VDEV_TRIM_CANCELED);
vdev_autotrim_stop_wait(vd);
/*
* Evacuate the device. We don't hold the config lock as
* writer since we need to do I/O but we do keep the
* spa_namespace_lock held. Once this completes the device
* should no longer have any blocks allocated on it.
*/
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (vd->vdev_stat.vs_alloc != 0)
error = spa_reset_logs(spa);
*txg = spa_vdev_config_enter(spa);
if (error != 0) {
metaslab_group_activate(mg);
ASSERT3P(vd->vdev_log_mg, ==, NULL);
return (error);
}
ASSERT0(vd->vdev_stat.vs_alloc);
/*
* The evacuation succeeded. Remove any remaining MOS metadata
* associated with this vdev, and wait for these changes to sync.
*/
vd->vdev_removing = B_TRUE;
vdev_dirty_leaves(vd, VDD_DTL, *txg);
vdev_config_dirty(vd);
/*
* When the log space map feature is enabled we look at
* the vdev's top_zap to find the on-disk flush data of
* the metaslab we just flushed. Thus, while removing a
* log vdev we make sure to call vdev_metaslab_fini()
* first, which removes all metaslabs of this vdev from
* spa_metaslabs_by_flushed before vdev_remove_empty()
* destroys the top_zap of this log vdev.
*
* This avoids the scenario where we flush a metaslab
* from the log vdev being removed that doesn't have a
* top_zap and end up failing to lookup its on-disk flush
* data.
*
* We don't call metaslab_group_destroy() right away
* though (it will be called in vdev_free() later) as
* during metaslab_sync() of metaslabs from other vdevs
* we may touch the metaslab group of this vdev through
* metaslab_class_histogram_verify()
*/
vdev_metaslab_fini(vd);
spa_log_sm_set_blocklimit(spa);
spa_vdev_config_exit(spa, NULL, *txg, 0, FTAG);
*txg = spa_vdev_config_enter(spa);
sysevent_t *ev = spa_event_create(spa, vd, NULL,
ESC_ZFS_VDEV_REMOVE_DEV);
ASSERT(MUTEX_HELD(&spa_namespace_lock));
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
/* The top ZAP should have been destroyed by vdev_remove_empty. */
ASSERT0(vd->vdev_top_zap);
/* The leaf ZAP should have been destroyed by vdev_dtl_sync. */
ASSERT0(vd->vdev_leaf_zap);
(void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE);
if (list_link_active(&vd->vdev_state_dirty_node))
vdev_state_clean(vd);
if (list_link_active(&vd->vdev_config_dirty_node))
vdev_config_clean(vd);
ASSERT0(vd->vdev_stat.vs_alloc);
/*
* Clean up the vdev namespace.
*/
vdev_remove_make_hole_and_free(vd);
if (ev != NULL)
spa_event_post(ev);
return (0);
}
static int
spa_vdev_remove_top_check(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
if (vd != vd->vdev_top)
return (SET_ERROR(ENOTSUP));
if (!vdev_is_concrete(vd))
return (SET_ERROR(ENOTSUP));
if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REMOVAL))
return (SET_ERROR(ENOTSUP));
/*
* This device is already being removed
*/
if (vd->vdev_removing)
return (SET_ERROR(EALREADY));
metaslab_class_t *mc = vd->vdev_mg->mg_class;
metaslab_class_t *normal = spa_normal_class(spa);
if (mc != normal) {
/*
* Space allocated from the special (or dedup) class is
* included in the DMU's space usage, but it's not included
* in spa_dspace (or dsl_pool_adjustedsize()). Therefore
* there is always at least as much free space in the normal
* class, as is allocated from the special (and dedup) class.
* As a backup check, we will return ENOSPC if this is
* violated. See also spa_update_dspace().
*/
uint64_t available = metaslab_class_get_space(normal) -
metaslab_class_get_alloc(normal);
ASSERT3U(available, >=, vd->vdev_stat.vs_alloc);
if (available < vd->vdev_stat.vs_alloc)
return (SET_ERROR(ENOSPC));
} else if (!vd->vdev_noalloc) {
/* available space in the pool's normal class */
uint64_t available = dsl_dir_space_available(
spa->spa_dsl_pool->dp_root_dir, NULL, 0, B_TRUE);
if (available < vd->vdev_stat.vs_dspace)
return (SET_ERROR(ENOSPC));
}
/*
* There can not be a removal in progress.
*/
if (spa->spa_removing_phys.sr_state == DSS_SCANNING)
return (SET_ERROR(EBUSY));
/*
* The device must have all its data.
*/
if (!vdev_dtl_empty(vd, DTL_MISSING) ||
!vdev_dtl_empty(vd, DTL_OUTAGE))
return (SET_ERROR(EBUSY));
/*
* The device must be healthy.
*/
if (!vdev_readable(vd))
return (SET_ERROR(EIO));
/*
* All vdevs in normal class must have the same ashift.
*/
if (spa->spa_max_ashift != spa->spa_min_ashift) {
return (SET_ERROR(EINVAL));
}
/*
* A removed special/dedup vdev must have same ashift as normal class.
*/
ASSERT(!vd->vdev_islog);
if (vd->vdev_alloc_bias != VDEV_BIAS_NONE &&
vd->vdev_ashift != spa->spa_max_ashift) {
return (SET_ERROR(EINVAL));
}
/*
* All vdevs in normal class must have the same ashift
* and not be raidz or draid.
*/
vdev_t *rvd = spa->spa_root_vdev;
int num_indirect = 0;
for (uint64_t id = 0; id < rvd->vdev_children; id++) {
vdev_t *cvd = rvd->vdev_child[id];
/*
* A removed special/dedup vdev must have the same ashift
* across all vdevs in its class.
*/
if (vd->vdev_alloc_bias != VDEV_BIAS_NONE &&
cvd->vdev_alloc_bias == vd->vdev_alloc_bias &&
cvd->vdev_ashift != vd->vdev_ashift) {
return (SET_ERROR(EINVAL));
}
if (cvd->vdev_ashift != 0 &&
cvd->vdev_alloc_bias == VDEV_BIAS_NONE)
ASSERT3U(cvd->vdev_ashift, ==, spa->spa_max_ashift);
if (cvd->vdev_ops == &vdev_indirect_ops)
num_indirect++;
if (!vdev_is_concrete(cvd))
continue;
if (vdev_get_nparity(cvd) != 0)
return (SET_ERROR(EINVAL));
/*
* Need the mirror to be mirror of leaf vdevs only
*/
if (cvd->vdev_ops == &vdev_mirror_ops) {
for (uint64_t cid = 0;
cid < cvd->vdev_children; cid++) {
if (!cvd->vdev_child[cid]->vdev_ops->
vdev_op_leaf)
return (SET_ERROR(EINVAL));
}
}
}
return (0);
}
/*
* Initiate removal of a top-level vdev, reducing the total space in the pool.
* The config lock is held for the specified TXG. Once initiated,
* evacuation of all allocated space (copying it to other vdevs) happens
* in the background (see spa_vdev_remove_thread()), and can be canceled
* (see spa_vdev_remove_cancel()). If successful, the vdev will
* be transformed to an indirect vdev (see spa_vdev_remove_complete()).
*/
static int
spa_vdev_remove_top(vdev_t *vd, uint64_t *txg)
{
spa_t *spa = vd->vdev_spa;
boolean_t set_noalloc = B_FALSE;
int error;
/*
* Check for errors up-front, so that we don't waste time
* passivating the metaslab group and clearing the ZIL if there
* are errors.
*/
error = spa_vdev_remove_top_check(vd);
/*
* Stop allocating from this vdev. Note that we must check
* that this is not the only device in the pool before
* passivating, otherwise we will not be able to make
* progress because we can't allocate from any vdevs.
* The above check for sufficient free space serves this
* purpose.
*/
if (error == 0 && !vd->vdev_noalloc) {
set_noalloc = B_TRUE;
error = vdev_passivate(vd, txg);
}
if (error != 0)
return (error);
/*
* We stop any initializing and TRIM that is currently in progress
* but leave the state as "active". This will allow the process to
* resume if the removal is canceled sometime later.
*/
spa_vdev_config_exit(spa, NULL, *txg, 0, FTAG);
vdev_initialize_stop_all(vd, VDEV_INITIALIZE_ACTIVE);
vdev_trim_stop_all(vd, VDEV_TRIM_ACTIVE);
vdev_autotrim_stop_wait(vd);
*txg = spa_vdev_config_enter(spa);
/*
* Things might have changed while the config lock was dropped
* (e.g. space usage). Check for errors again.
*/
error = spa_vdev_remove_top_check(vd);
if (error != 0) {
if (set_noalloc)
vdev_activate(vd);
spa_async_request(spa, SPA_ASYNC_INITIALIZE_RESTART);
spa_async_request(spa, SPA_ASYNC_TRIM_RESTART);
spa_async_request(spa, SPA_ASYNC_AUTOTRIM_RESTART);
return (error);
}
vd->vdev_removing = B_TRUE;
vdev_dirty_leaves(vd, VDD_DTL, *txg);
vdev_config_dirty(vd);
dmu_tx_t *tx = dmu_tx_create_assigned(spa->spa_dsl_pool, *txg);
dsl_sync_task_nowait(spa->spa_dsl_pool,
vdev_remove_initiate_sync, (void *)(uintptr_t)vd->vdev_id, tx);
dmu_tx_commit(tx);
return (0);
}
/*
* Remove a device from the pool.
*
* Removing a device from the vdev namespace requires several steps
* and can take a significant amount of time. As a result we use
* the spa_vdev_config_[enter/exit] functions which allow us to
* grab and release the spa_config_lock while still holding the namespace
* lock. During each step the configuration is synced out.
*/
int
spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare)
{
vdev_t *vd;
nvlist_t **spares, **l2cache, *nv;
uint64_t txg = 0;
uint_t nspares, nl2cache;
int error = 0, error_log;
boolean_t locked = MUTEX_HELD(&spa_namespace_lock);
sysevent_t *ev = NULL;
char *vd_type = NULL, *vd_path = NULL;
ASSERT(spa_writeable(spa));
if (!locked)
txg = spa_vdev_enter(spa);
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
error = (spa_has_checkpoint(spa)) ?
ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
if (!locked)
return (spa_vdev_exit(spa, NULL, txg, error));
return (error);
}
vd = spa_lookup_by_guid(spa, guid, B_FALSE);
if (spa->spa_spares.sav_vdevs != NULL &&
nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0 &&
(nv = spa_nvlist_lookup_by_guid(spares, nspares, guid)) != NULL) {
/*
* Only remove the hot spare if it's not currently in use
* in this pool.
*/
if (vd == NULL || unspare) {
char *type;
boolean_t draid_spare = B_FALSE;
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type)
== 0 && strcmp(type, VDEV_TYPE_DRAID_SPARE) == 0)
draid_spare = B_TRUE;
if (vd == NULL && draid_spare) {
error = SET_ERROR(ENOTSUP);
} else {
if (vd == NULL)
vd = spa_lookup_by_guid(spa,
guid, B_TRUE);
ev = spa_event_create(spa, vd, NULL,
ESC_ZFS_VDEV_REMOVE_AUX);
vd_type = VDEV_TYPE_SPARE;
vd_path = spa_strdup(fnvlist_lookup_string(
nv, ZPOOL_CONFIG_PATH));
spa_vdev_remove_aux(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, spares, nspares, nv);
spa_load_spares(spa);
spa->spa_spares.sav_sync = B_TRUE;
}
} else {
error = SET_ERROR(EBUSY);
}
} else if (spa->spa_l2cache.sav_vdevs != NULL &&
nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0 &&
(nv = spa_nvlist_lookup_by_guid(l2cache, nl2cache, guid)) != NULL) {
vd_type = VDEV_TYPE_L2CACHE;
vd_path = spa_strdup(fnvlist_lookup_string(
nv, ZPOOL_CONFIG_PATH));
/*
* Cache devices can always be removed.
*/
vd = spa_lookup_by_guid(spa, guid, B_TRUE);
/*
* Stop trimming the cache device. We need to release the
* config lock to allow the syncing of TRIM transactions
* without releasing the spa_namespace_lock. The same
* strategy is employed in spa_vdev_remove_top().
*/
spa_vdev_config_exit(spa, NULL,
txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG);
mutex_enter(&vd->vdev_trim_lock);
vdev_trim_stop(vd, VDEV_TRIM_CANCELED, NULL);
mutex_exit(&vd->vdev_trim_lock);
txg = spa_vdev_config_enter(spa);
ev = spa_event_create(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE_AUX);
spa_vdev_remove_aux(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv);
spa_load_l2cache(spa);
spa->spa_l2cache.sav_sync = B_TRUE;
} else if (vd != NULL && vd->vdev_islog) {
ASSERT(!locked);
vd_type = VDEV_TYPE_LOG;
vd_path = spa_strdup((vd->vdev_path != NULL) ?
vd->vdev_path : "-");
error = spa_vdev_remove_log(vd, &txg);
} else if (vd != NULL) {
ASSERT(!locked);
error = spa_vdev_remove_top(vd, &txg);
} else {
/*
* There is no vdev of any kind with the specified guid.
*/
error = SET_ERROR(ENOENT);
}
error_log = error;
if (!locked)
error = spa_vdev_exit(spa, NULL, txg, error);
/*
* Logging must be done outside the spa config lock. Otherwise,
* this code path could end up holding the spa config lock while
* waiting for a txg_sync so it can write to the internal log.
* Doing that would prevent the txg sync from actually happening,
* causing a deadlock.
*/
if (error_log == 0 && vd_type != NULL && vd_path != NULL) {
spa_history_log_internal(spa, "vdev remove", NULL,
"%s vdev (%s) %s", spa_name(spa), vd_type, vd_path);
}
if (vd_path != NULL)
spa_strfree(vd_path);
if (ev != NULL)
spa_event_post(ev);
return (error);
}
int
spa_removal_get_stats(spa_t *spa, pool_removal_stat_t *prs)
{
prs->prs_state = spa->spa_removing_phys.sr_state;
if (prs->prs_state == DSS_NONE)
return (SET_ERROR(ENOENT));
prs->prs_removing_vdev = spa->spa_removing_phys.sr_removing_vdev;
prs->prs_start_time = spa->spa_removing_phys.sr_start_time;
prs->prs_end_time = spa->spa_removing_phys.sr_end_time;
prs->prs_to_copy = spa->spa_removing_phys.sr_to_copy;
prs->prs_copied = spa->spa_removing_phys.sr_copied;
prs->prs_mapping_memory = 0;
uint64_t indirect_vdev_id =
spa->spa_removing_phys.sr_prev_indirect_vdev;
while (indirect_vdev_id != -1) {
vdev_t *vd = spa->spa_root_vdev->vdev_child[indirect_vdev_id];
vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
prs->prs_mapping_memory += vdev_indirect_mapping_size(vim);
indirect_vdev_id = vic->vic_prev_indirect_vdev;
}
return (0);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_vdev, zfs_, removal_ignore_errors, INT, ZMOD_RW,
"Ignore hard IO errors when removing device");
ZFS_MODULE_PARAM(zfs_vdev, zfs_, remove_max_segment, INT, ZMOD_RW,
"Largest contiguous segment to allocate when removing device");
ZFS_MODULE_PARAM(zfs_vdev, vdev_, removal_max_span, INT, ZMOD_RW,
"Largest span of free chunks a remap segment can span");
ZFS_MODULE_PARAM(zfs_vdev, zfs_, removal_suspend_progress, INT, ZMOD_RW,
"Pause device removal after this many bytes are copied "
"(debug use only - causes removal to hang)");
/* END CSTYLED */
EXPORT_SYMBOL(free_from_removing_vdev);
EXPORT_SYMBOL(spa_removal_get_stats);
EXPORT_SYMBOL(spa_remove_init);
EXPORT_SYMBOL(spa_restart_removal);
EXPORT_SYMBOL(spa_vdev_removal_destroy);
EXPORT_SYMBOL(spa_vdev_remove);
EXPORT_SYMBOL(spa_vdev_remove_cancel);
EXPORT_SYMBOL(spa_vdev_remove_suspend);
EXPORT_SYMBOL(svr_sync);
diff --git a/module/zfs/vdev_trim.c b/module/zfs/vdev_trim.c
index 2bae33b2b532..18aa2b3bfcb0 100644
--- a/module/zfs/vdev_trim.c
+++ b/module/zfs/vdev_trim.c
@@ -1,1727 +1,1727 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2016 by Delphix. All rights reserved.
* Copyright (c) 2019 by Lawrence Livermore National Security, LLC.
* Copyright (c) 2021 Hewlett Packard Enterprise Development LP
*/
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/txg.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_trim.h>
#include <sys/metaslab_impl.h>
#include <sys/dsl_synctask.h>
#include <sys/zap.h>
#include <sys/dmu_tx.h>
#include <sys/arc_impl.h>
/*
* TRIM is a feature which is used to notify a SSD that some previously
* written space is no longer allocated by the pool. This is useful because
* writes to a SSD must be performed to blocks which have first been erased.
* Ensuring the SSD always has a supply of erased blocks for new writes
* helps prevent the performance from deteriorating.
*
* There are two supported TRIM methods; manual and automatic.
*
* Manual TRIM:
*
* A manual TRIM is initiated by running the 'zpool trim' command. A single
* 'vdev_trim' thread is created for each leaf vdev, and it is responsible for
* managing that vdev TRIM process. This involves iterating over all the
* metaslabs, calculating the unallocated space ranges, and then issuing the
* required TRIM I/Os.
*
* While a metaslab is being actively trimmed it is not eligible to perform
* new allocations. After traversing all of the metaslabs the thread is
* terminated. Finally, both the requested options and current progress of
* the TRIM are regularly written to the pool. This allows the TRIM to be
* suspended and resumed as needed.
*
* Automatic TRIM:
*
* An automatic TRIM is enabled by setting the 'autotrim' pool property
* to 'on'. When enabled, a `vdev_autotrim' thread is created for each
* top-level (not leaf) vdev in the pool. These threads perform the same
* core TRIM process as a manual TRIM, but with a few key differences.
*
* 1) Automatic TRIM happens continuously in the background and operates
* solely on recently freed blocks (ms_trim not ms_allocatable).
*
* 2) Each thread is associated with a top-level (not leaf) vdev. This has
* the benefit of simplifying the threading model, it makes it easier
* to coordinate administrative commands, and it ensures only a single
* metaslab is disabled at a time. Unlike manual TRIM, this means each
* 'vdev_autotrim' thread is responsible for issuing TRIM I/Os for its
* children.
*
* 3) There is no automatic TRIM progress information stored on disk, nor
* is it reported by 'zpool status'.
*
* While the automatic TRIM process is highly effective it is more likely
* than a manual TRIM to encounter tiny ranges. Ranges less than or equal to
* 'zfs_trim_extent_bytes_min' (32k) are considered too small to efficiently
* TRIM and are skipped. This means small amounts of freed space may not
* be automatically trimmed.
*
* Furthermore, devices with attached hot spares and devices being actively
* replaced are skipped. This is done to avoid adding additional stress to
* a potentially unhealthy device and to minimize the required rebuild time.
*
* For this reason it may be beneficial to occasionally manually TRIM a pool
* even when automatic TRIM is enabled.
*/
/*
* Maximum size of TRIM I/O, ranges will be chunked in to 128MiB lengths.
*/
-unsigned int zfs_trim_extent_bytes_max = 128 * 1024 * 1024;
+static unsigned int zfs_trim_extent_bytes_max = 128 * 1024 * 1024;
/*
* Minimum size of TRIM I/O, extents smaller than 32Kib will be skipped.
*/
-unsigned int zfs_trim_extent_bytes_min = 32 * 1024;
+static unsigned int zfs_trim_extent_bytes_min = 32 * 1024;
/*
* Skip uninitialized metaslabs during the TRIM process. This option is
* useful for pools constructed from large thinly-provisioned devices where
* TRIM operations are slow. As a pool ages an increasing fraction of
* the pools metaslabs will be initialized progressively degrading the
* usefulness of this option. This setting is stored when starting a
* manual TRIM and will persist for the duration of the requested TRIM.
*/
unsigned int zfs_trim_metaslab_skip = 0;
/*
* Maximum number of queued TRIM I/Os per leaf vdev. The number of
* concurrent TRIM I/Os issued to the device is controlled by the
* zfs_vdev_trim_min_active and zfs_vdev_trim_max_active module options.
*/
-unsigned int zfs_trim_queue_limit = 10;
+static unsigned int zfs_trim_queue_limit = 10;
/*
* The minimum number of transaction groups between automatic trims of a
* metaslab. This setting represents a trade-off between issuing more
* efficient TRIM operations, by allowing them to be aggregated longer,
* and issuing them promptly so the trimmed space is available. Note
* that this value is a minimum; metaslabs can be trimmed less frequently
* when there are a large number of ranges which need to be trimmed.
*
* Increasing this value will allow frees to be aggregated for a longer
* time. This can result is larger TRIM operations, and increased memory
* usage in order to track the ranges to be trimmed. Decreasing this value
* has the opposite effect. The default value of 32 was determined though
* testing to be a reasonable compromise.
*/
-unsigned int zfs_trim_txg_batch = 32;
+static unsigned int zfs_trim_txg_batch = 32;
/*
* The trim_args are a control structure which describe how a leaf vdev
* should be trimmed. The core elements are the vdev, the metaslab being
* trimmed and a range tree containing the extents to TRIM. All provided
* ranges must be within the metaslab.
*/
typedef struct trim_args {
/*
* These fields are set by the caller of vdev_trim_ranges().
*/
vdev_t *trim_vdev; /* Leaf vdev to TRIM */
metaslab_t *trim_msp; /* Disabled metaslab */
range_tree_t *trim_tree; /* TRIM ranges (in metaslab) */
trim_type_t trim_type; /* Manual or auto TRIM */
uint64_t trim_extent_bytes_max; /* Maximum TRIM I/O size */
uint64_t trim_extent_bytes_min; /* Minimum TRIM I/O size */
enum trim_flag trim_flags; /* TRIM flags (secure) */
/*
* These fields are updated by vdev_trim_ranges().
*/
hrtime_t trim_start_time; /* Start time */
uint64_t trim_bytes_done; /* Bytes trimmed */
} trim_args_t;
/*
* Determines whether a vdev_trim_thread() should be stopped.
*/
static boolean_t
vdev_trim_should_stop(vdev_t *vd)
{
return (vd->vdev_trim_exit_wanted || !vdev_writeable(vd) ||
vd->vdev_detached || vd->vdev_top->vdev_removing);
}
/*
* Determines whether a vdev_autotrim_thread() should be stopped.
*/
static boolean_t
vdev_autotrim_should_stop(vdev_t *tvd)
{
return (tvd->vdev_autotrim_exit_wanted ||
!vdev_writeable(tvd) || tvd->vdev_removing ||
spa_get_autotrim(tvd->vdev_spa) == SPA_AUTOTRIM_OFF);
}
/*
* The sync task for updating the on-disk state of a manual TRIM. This
* is scheduled by vdev_trim_change_state().
*/
static void
vdev_trim_zap_update_sync(void *arg, dmu_tx_t *tx)
{
/*
* We pass in the guid instead of the vdev_t since the vdev may
* have been freed prior to the sync task being processed. This
* happens when a vdev is detached as we call spa_config_vdev_exit(),
* stop the trimming thread, schedule the sync task, and free
* the vdev. Later when the scheduled sync task is invoked, it would
* find that the vdev has been freed.
*/
uint64_t guid = *(uint64_t *)arg;
uint64_t txg = dmu_tx_get_txg(tx);
kmem_free(arg, sizeof (uint64_t));
vdev_t *vd = spa_lookup_by_guid(tx->tx_pool->dp_spa, guid, B_FALSE);
if (vd == NULL || vd->vdev_top->vdev_removing || !vdev_is_concrete(vd))
return;
uint64_t last_offset = vd->vdev_trim_offset[txg & TXG_MASK];
vd->vdev_trim_offset[txg & TXG_MASK] = 0;
VERIFY3U(vd->vdev_leaf_zap, !=, 0);
objset_t *mos = vd->vdev_spa->spa_meta_objset;
if (last_offset > 0 || vd->vdev_trim_last_offset == UINT64_MAX) {
if (vd->vdev_trim_last_offset == UINT64_MAX)
last_offset = 0;
vd->vdev_trim_last_offset = last_offset;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
VDEV_LEAF_ZAP_TRIM_LAST_OFFSET,
sizeof (last_offset), 1, &last_offset, tx));
}
if (vd->vdev_trim_action_time > 0) {
uint64_t val = (uint64_t)vd->vdev_trim_action_time;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
VDEV_LEAF_ZAP_TRIM_ACTION_TIME, sizeof (val),
1, &val, tx));
}
if (vd->vdev_trim_rate > 0) {
uint64_t rate = (uint64_t)vd->vdev_trim_rate;
if (rate == UINT64_MAX)
rate = 0;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
VDEV_LEAF_ZAP_TRIM_RATE, sizeof (rate), 1, &rate, tx));
}
uint64_t partial = vd->vdev_trim_partial;
if (partial == UINT64_MAX)
partial = 0;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_PARTIAL,
sizeof (partial), 1, &partial, tx));
uint64_t secure = vd->vdev_trim_secure;
if (secure == UINT64_MAX)
secure = 0;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_SECURE,
sizeof (secure), 1, &secure, tx));
uint64_t trim_state = vd->vdev_trim_state;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_STATE,
sizeof (trim_state), 1, &trim_state, tx));
}
/*
* Update the on-disk state of a manual TRIM. This is called to request
* that a TRIM be started/suspended/canceled, or to change one of the
* TRIM options (partial, secure, rate).
*/
static void
vdev_trim_change_state(vdev_t *vd, vdev_trim_state_t new_state,
uint64_t rate, boolean_t partial, boolean_t secure)
{
ASSERT(MUTEX_HELD(&vd->vdev_trim_lock));
spa_t *spa = vd->vdev_spa;
if (new_state == vd->vdev_trim_state)
return;
/*
* Copy the vd's guid, this will be freed by the sync task.
*/
uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP);
*guid = vd->vdev_guid;
/*
* If we're suspending, then preserve the original start time.
*/
if (vd->vdev_trim_state != VDEV_TRIM_SUSPENDED) {
vd->vdev_trim_action_time = gethrestime_sec();
}
/*
* If we're activating, then preserve the requested rate and trim
* method. Setting the last offset and rate to UINT64_MAX is used
* as a sentinel to indicate they should be reset to default values.
*/
if (new_state == VDEV_TRIM_ACTIVE) {
if (vd->vdev_trim_state == VDEV_TRIM_COMPLETE ||
vd->vdev_trim_state == VDEV_TRIM_CANCELED) {
vd->vdev_trim_last_offset = UINT64_MAX;
vd->vdev_trim_rate = UINT64_MAX;
vd->vdev_trim_partial = UINT64_MAX;
vd->vdev_trim_secure = UINT64_MAX;
}
if (rate != 0)
vd->vdev_trim_rate = rate;
if (partial != 0)
vd->vdev_trim_partial = partial;
if (secure != 0)
vd->vdev_trim_secure = secure;
}
vdev_trim_state_t old_state = vd->vdev_trim_state;
boolean_t resumed = (old_state == VDEV_TRIM_SUSPENDED);
vd->vdev_trim_state = new_state;
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
dsl_sync_task_nowait(spa_get_dsl(spa), vdev_trim_zap_update_sync,
guid, tx);
switch (new_state) {
case VDEV_TRIM_ACTIVE:
spa_event_notify(spa, vd, NULL,
resumed ? ESC_ZFS_TRIM_RESUME : ESC_ZFS_TRIM_START);
spa_history_log_internal(spa, "trim", tx,
"vdev=%s activated", vd->vdev_path);
break;
case VDEV_TRIM_SUSPENDED:
spa_event_notify(spa, vd, NULL, ESC_ZFS_TRIM_SUSPEND);
spa_history_log_internal(spa, "trim", tx,
"vdev=%s suspended", vd->vdev_path);
break;
case VDEV_TRIM_CANCELED:
if (old_state == VDEV_TRIM_ACTIVE ||
old_state == VDEV_TRIM_SUSPENDED) {
spa_event_notify(spa, vd, NULL, ESC_ZFS_TRIM_CANCEL);
spa_history_log_internal(spa, "trim", tx,
"vdev=%s canceled", vd->vdev_path);
}
break;
case VDEV_TRIM_COMPLETE:
spa_event_notify(spa, vd, NULL, ESC_ZFS_TRIM_FINISH);
spa_history_log_internal(spa, "trim", tx,
"vdev=%s complete", vd->vdev_path);
break;
default:
panic("invalid state %llu", (unsigned long long)new_state);
}
dmu_tx_commit(tx);
if (new_state != VDEV_TRIM_ACTIVE)
spa_notify_waiters(spa);
}
/*
* The zio_done_func_t done callback for each manual TRIM issued. It is
* responsible for updating the TRIM stats, reissuing failed TRIM I/Os,
* and limiting the number of in flight TRIM I/Os.
*/
static void
vdev_trim_cb(zio_t *zio)
{
vdev_t *vd = zio->io_vd;
mutex_enter(&vd->vdev_trim_io_lock);
if (zio->io_error == ENXIO && !vdev_writeable(vd)) {
/*
* The I/O failed because the vdev was unavailable; roll the
* last offset back. (This works because spa_sync waits on
* spa_txg_zio before it runs sync tasks.)
*/
uint64_t *offset =
&vd->vdev_trim_offset[zio->io_txg & TXG_MASK];
*offset = MIN(*offset, zio->io_offset);
} else {
if (zio->io_error != 0) {
vd->vdev_stat.vs_trim_errors++;
spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_MANUAL,
0, 0, 0, 0, 1, zio->io_orig_size);
} else {
spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_MANUAL,
1, zio->io_orig_size, 0, 0, 0, 0);
}
vd->vdev_trim_bytes_done += zio->io_orig_size;
}
ASSERT3U(vd->vdev_trim_inflight[TRIM_TYPE_MANUAL], >, 0);
vd->vdev_trim_inflight[TRIM_TYPE_MANUAL]--;
cv_broadcast(&vd->vdev_trim_io_cv);
mutex_exit(&vd->vdev_trim_io_lock);
spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
}
/*
* The zio_done_func_t done callback for each automatic TRIM issued. It
* is responsible for updating the TRIM stats and limiting the number of
* in flight TRIM I/Os. Automatic TRIM I/Os are best effort and are
* never reissued on failure.
*/
static void
vdev_autotrim_cb(zio_t *zio)
{
vdev_t *vd = zio->io_vd;
mutex_enter(&vd->vdev_trim_io_lock);
if (zio->io_error != 0) {
vd->vdev_stat.vs_trim_errors++;
spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_AUTO,
0, 0, 0, 0, 1, zio->io_orig_size);
} else {
spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_AUTO,
1, zio->io_orig_size, 0, 0, 0, 0);
}
ASSERT3U(vd->vdev_trim_inflight[TRIM_TYPE_AUTO], >, 0);
vd->vdev_trim_inflight[TRIM_TYPE_AUTO]--;
cv_broadcast(&vd->vdev_trim_io_cv);
mutex_exit(&vd->vdev_trim_io_lock);
spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
}
/*
* The zio_done_func_t done callback for each TRIM issued via
* vdev_trim_simple(). It is responsible for updating the TRIM stats and
* limiting the number of in flight TRIM I/Os. Simple TRIM I/Os are best
* effort and are never reissued on failure.
*/
static void
vdev_trim_simple_cb(zio_t *zio)
{
vdev_t *vd = zio->io_vd;
mutex_enter(&vd->vdev_trim_io_lock);
if (zio->io_error != 0) {
vd->vdev_stat.vs_trim_errors++;
spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_SIMPLE,
0, 0, 0, 0, 1, zio->io_orig_size);
} else {
spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_SIMPLE,
1, zio->io_orig_size, 0, 0, 0, 0);
}
ASSERT3U(vd->vdev_trim_inflight[TRIM_TYPE_SIMPLE], >, 0);
vd->vdev_trim_inflight[TRIM_TYPE_SIMPLE]--;
cv_broadcast(&vd->vdev_trim_io_cv);
mutex_exit(&vd->vdev_trim_io_lock);
spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
}
/*
* Returns the average trim rate in bytes/sec for the ta->trim_vdev.
*/
static uint64_t
vdev_trim_calculate_rate(trim_args_t *ta)
{
return (ta->trim_bytes_done * 1000 /
(NSEC2MSEC(gethrtime() - ta->trim_start_time) + 1));
}
/*
* Issues a physical TRIM and takes care of rate limiting (bytes/sec)
* and number of concurrent TRIM I/Os.
*/
static int
vdev_trim_range(trim_args_t *ta, uint64_t start, uint64_t size)
{
vdev_t *vd = ta->trim_vdev;
spa_t *spa = vd->vdev_spa;
void *cb;
mutex_enter(&vd->vdev_trim_io_lock);
/*
* Limit manual TRIM I/Os to the requested rate. This does not
* apply to automatic TRIM since no per vdev rate can be specified.
*/
if (ta->trim_type == TRIM_TYPE_MANUAL) {
while (vd->vdev_trim_rate != 0 && !vdev_trim_should_stop(vd) &&
vdev_trim_calculate_rate(ta) > vd->vdev_trim_rate) {
cv_timedwait_idle(&vd->vdev_trim_io_cv,
&vd->vdev_trim_io_lock, ddi_get_lbolt() +
MSEC_TO_TICK(10));
}
}
ta->trim_bytes_done += size;
/* Limit in flight trimming I/Os */
while (vd->vdev_trim_inflight[0] + vd->vdev_trim_inflight[1] +
vd->vdev_trim_inflight[2] >= zfs_trim_queue_limit) {
cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock);
}
vd->vdev_trim_inflight[ta->trim_type]++;
mutex_exit(&vd->vdev_trim_io_lock);
dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
uint64_t txg = dmu_tx_get_txg(tx);
spa_config_enter(spa, SCL_STATE_ALL, vd, RW_READER);
mutex_enter(&vd->vdev_trim_lock);
if (ta->trim_type == TRIM_TYPE_MANUAL &&
vd->vdev_trim_offset[txg & TXG_MASK] == 0) {
uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP);
*guid = vd->vdev_guid;
/* This is the first write of this txg. */
dsl_sync_task_nowait(spa_get_dsl(spa),
vdev_trim_zap_update_sync, guid, tx);
}
/*
* We know the vdev_t will still be around since all consumers of
* vdev_free must stop the trimming first.
*/
if ((ta->trim_type == TRIM_TYPE_MANUAL &&
vdev_trim_should_stop(vd)) ||
(ta->trim_type == TRIM_TYPE_AUTO &&
vdev_autotrim_should_stop(vd->vdev_top))) {
mutex_enter(&vd->vdev_trim_io_lock);
vd->vdev_trim_inflight[ta->trim_type]--;
mutex_exit(&vd->vdev_trim_io_lock);
spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
mutex_exit(&vd->vdev_trim_lock);
dmu_tx_commit(tx);
return (SET_ERROR(EINTR));
}
mutex_exit(&vd->vdev_trim_lock);
if (ta->trim_type == TRIM_TYPE_MANUAL)
vd->vdev_trim_offset[txg & TXG_MASK] = start + size;
if (ta->trim_type == TRIM_TYPE_MANUAL) {
cb = vdev_trim_cb;
} else if (ta->trim_type == TRIM_TYPE_AUTO) {
cb = vdev_autotrim_cb;
} else {
cb = vdev_trim_simple_cb;
}
zio_nowait(zio_trim(spa->spa_txg_zio[txg & TXG_MASK], vd,
start, size, cb, NULL, ZIO_PRIORITY_TRIM, ZIO_FLAG_CANFAIL,
ta->trim_flags));
/* vdev_trim_cb and vdev_autotrim_cb release SCL_STATE_ALL */
dmu_tx_commit(tx);
return (0);
}
/*
* Issues TRIM I/Os for all ranges in the provided ta->trim_tree range tree.
* Additional parameters describing how the TRIM should be performed must
* be set in the trim_args structure. See the trim_args definition for
* additional information.
*/
static int
vdev_trim_ranges(trim_args_t *ta)
{
vdev_t *vd = ta->trim_vdev;
zfs_btree_t *t = &ta->trim_tree->rt_root;
zfs_btree_index_t idx;
uint64_t extent_bytes_max = ta->trim_extent_bytes_max;
uint64_t extent_bytes_min = ta->trim_extent_bytes_min;
spa_t *spa = vd->vdev_spa;
ta->trim_start_time = gethrtime();
ta->trim_bytes_done = 0;
for (range_seg_t *rs = zfs_btree_first(t, &idx); rs != NULL;
rs = zfs_btree_next(t, &idx, &idx)) {
uint64_t size = rs_get_end(rs, ta->trim_tree) - rs_get_start(rs,
ta->trim_tree);
if (extent_bytes_min && size < extent_bytes_min) {
spa_iostats_trim_add(spa, ta->trim_type,
0, 0, 1, size, 0, 0);
continue;
}
/* Split range into legally-sized physical chunks */
uint64_t writes_required = ((size - 1) / extent_bytes_max) + 1;
for (uint64_t w = 0; w < writes_required; w++) {
int error;
error = vdev_trim_range(ta, VDEV_LABEL_START_SIZE +
rs_get_start(rs, ta->trim_tree) +
(w *extent_bytes_max), MIN(size -
(w * extent_bytes_max), extent_bytes_max));
if (error != 0) {
return (error);
}
}
}
return (0);
}
static void
vdev_trim_xlate_last_rs_end(void *arg, range_seg64_t *physical_rs)
{
uint64_t *last_rs_end = (uint64_t *)arg;
if (physical_rs->rs_end > *last_rs_end)
*last_rs_end = physical_rs->rs_end;
}
static void
vdev_trim_xlate_progress(void *arg, range_seg64_t *physical_rs)
{
vdev_t *vd = (vdev_t *)arg;
uint64_t size = physical_rs->rs_end - physical_rs->rs_start;
vd->vdev_trim_bytes_est += size;
if (vd->vdev_trim_last_offset >= physical_rs->rs_end) {
vd->vdev_trim_bytes_done += size;
} else if (vd->vdev_trim_last_offset > physical_rs->rs_start &&
vd->vdev_trim_last_offset <= physical_rs->rs_end) {
vd->vdev_trim_bytes_done +=
vd->vdev_trim_last_offset - physical_rs->rs_start;
}
}
/*
* Calculates the completion percentage of a manual TRIM.
*/
static void
vdev_trim_calculate_progress(vdev_t *vd)
{
ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) ||
spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER));
ASSERT(vd->vdev_leaf_zap != 0);
vd->vdev_trim_bytes_est = 0;
vd->vdev_trim_bytes_done = 0;
for (uint64_t i = 0; i < vd->vdev_top->vdev_ms_count; i++) {
metaslab_t *msp = vd->vdev_top->vdev_ms[i];
mutex_enter(&msp->ms_lock);
uint64_t ms_free = (msp->ms_size -
metaslab_allocated_space(msp)) /
vdev_get_ndisks(vd->vdev_top);
/*
* Convert the metaslab range to a physical range
* on our vdev. We use this to determine if we are
* in the middle of this metaslab range.
*/
range_seg64_t logical_rs, physical_rs, remain_rs;
logical_rs.rs_start = msp->ms_start;
logical_rs.rs_end = msp->ms_start + msp->ms_size;
/* Metaslab space after this offset has not been trimmed. */
vdev_xlate(vd, &logical_rs, &physical_rs, &remain_rs);
if (vd->vdev_trim_last_offset <= physical_rs.rs_start) {
vd->vdev_trim_bytes_est += ms_free;
mutex_exit(&msp->ms_lock);
continue;
}
/* Metaslab space before this offset has been trimmed */
uint64_t last_rs_end = physical_rs.rs_end;
if (!vdev_xlate_is_empty(&remain_rs)) {
vdev_xlate_walk(vd, &remain_rs,
vdev_trim_xlate_last_rs_end, &last_rs_end);
}
if (vd->vdev_trim_last_offset > last_rs_end) {
vd->vdev_trim_bytes_done += ms_free;
vd->vdev_trim_bytes_est += ms_free;
mutex_exit(&msp->ms_lock);
continue;
}
/*
* If we get here, we're in the middle of trimming this
* metaslab. Load it and walk the free tree for more
* accurate progress estimation.
*/
VERIFY0(metaslab_load(msp));
range_tree_t *rt = msp->ms_allocatable;
zfs_btree_t *bt = &rt->rt_root;
zfs_btree_index_t idx;
for (range_seg_t *rs = zfs_btree_first(bt, &idx);
rs != NULL; rs = zfs_btree_next(bt, &idx, &idx)) {
logical_rs.rs_start = rs_get_start(rs, rt);
logical_rs.rs_end = rs_get_end(rs, rt);
vdev_xlate_walk(vd, &logical_rs,
vdev_trim_xlate_progress, vd);
}
mutex_exit(&msp->ms_lock);
}
}
/*
* Load from disk the vdev's manual TRIM information. This includes the
* state, progress, and options provided when initiating the manual TRIM.
*/
static int
vdev_trim_load(vdev_t *vd)
{
int err = 0;
ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) ||
spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER));
ASSERT(vd->vdev_leaf_zap != 0);
if (vd->vdev_trim_state == VDEV_TRIM_ACTIVE ||
vd->vdev_trim_state == VDEV_TRIM_SUSPENDED) {
err = zap_lookup(vd->vdev_spa->spa_meta_objset,
vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_LAST_OFFSET,
sizeof (vd->vdev_trim_last_offset), 1,
&vd->vdev_trim_last_offset);
if (err == ENOENT) {
vd->vdev_trim_last_offset = 0;
err = 0;
}
if (err == 0) {
err = zap_lookup(vd->vdev_spa->spa_meta_objset,
vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_RATE,
sizeof (vd->vdev_trim_rate), 1,
&vd->vdev_trim_rate);
if (err == ENOENT) {
vd->vdev_trim_rate = 0;
err = 0;
}
}
if (err == 0) {
err = zap_lookup(vd->vdev_spa->spa_meta_objset,
vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_PARTIAL,
sizeof (vd->vdev_trim_partial), 1,
&vd->vdev_trim_partial);
if (err == ENOENT) {
vd->vdev_trim_partial = 0;
err = 0;
}
}
if (err == 0) {
err = zap_lookup(vd->vdev_spa->spa_meta_objset,
vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_SECURE,
sizeof (vd->vdev_trim_secure), 1,
&vd->vdev_trim_secure);
if (err == ENOENT) {
vd->vdev_trim_secure = 0;
err = 0;
}
}
}
vdev_trim_calculate_progress(vd);
return (err);
}
static void
vdev_trim_xlate_range_add(void *arg, range_seg64_t *physical_rs)
{
trim_args_t *ta = arg;
vdev_t *vd = ta->trim_vdev;
/*
* Only a manual trim will be traversing the vdev sequentially.
* For an auto trim all valid ranges should be added.
*/
if (ta->trim_type == TRIM_TYPE_MANUAL) {
/* Only add segments that we have not visited yet */
if (physical_rs->rs_end <= vd->vdev_trim_last_offset)
return;
/* Pick up where we left off mid-range. */
if (vd->vdev_trim_last_offset > physical_rs->rs_start) {
ASSERT3U(physical_rs->rs_end, >,
vd->vdev_trim_last_offset);
physical_rs->rs_start = vd->vdev_trim_last_offset;
}
}
ASSERT3U(physical_rs->rs_end, >, physical_rs->rs_start);
range_tree_add(ta->trim_tree, physical_rs->rs_start,
physical_rs->rs_end - physical_rs->rs_start);
}
/*
* Convert the logical range into physical ranges and add them to the
* range tree passed in the trim_args_t.
*/
static void
vdev_trim_range_add(void *arg, uint64_t start, uint64_t size)
{
trim_args_t *ta = arg;
vdev_t *vd = ta->trim_vdev;
range_seg64_t logical_rs;
logical_rs.rs_start = start;
logical_rs.rs_end = start + size;
/*
* Every range to be trimmed must be part of ms_allocatable.
* When ZFS_DEBUG_TRIM is set load the metaslab to verify this
* is always the case.
*/
if (zfs_flags & ZFS_DEBUG_TRIM) {
metaslab_t *msp = ta->trim_msp;
VERIFY0(metaslab_load(msp));
VERIFY3B(msp->ms_loaded, ==, B_TRUE);
VERIFY(range_tree_contains(msp->ms_allocatable, start, size));
}
ASSERT(vd->vdev_ops->vdev_op_leaf);
vdev_xlate_walk(vd, &logical_rs, vdev_trim_xlate_range_add, arg);
}
/*
* Each manual TRIM thread is responsible for trimming the unallocated
* space for each leaf vdev. This is accomplished by sequentially iterating
* over its top-level metaslabs and issuing TRIM I/O for the space described
* by its ms_allocatable. While a metaslab is undergoing trimming it is
* not eligible for new allocations.
*/
static void
vdev_trim_thread(void *arg)
{
vdev_t *vd = arg;
spa_t *spa = vd->vdev_spa;
trim_args_t ta;
int error = 0;
/*
* The VDEV_LEAF_ZAP_TRIM_* entries may have been updated by
* vdev_trim(). Wait for the updated values to be reflected
* in the zap in order to start with the requested settings.
*/
txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0);
ASSERT(vdev_is_concrete(vd));
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vd->vdev_trim_last_offset = 0;
vd->vdev_trim_rate = 0;
vd->vdev_trim_partial = 0;
vd->vdev_trim_secure = 0;
VERIFY0(vdev_trim_load(vd));
ta.trim_vdev = vd;
ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max;
ta.trim_extent_bytes_min = zfs_trim_extent_bytes_min;
ta.trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
ta.trim_type = TRIM_TYPE_MANUAL;
ta.trim_flags = 0;
/*
* When a secure TRIM has been requested infer that the intent
* is that everything must be trimmed. Override the default
* minimum TRIM size to prevent ranges from being skipped.
*/
if (vd->vdev_trim_secure) {
ta.trim_flags |= ZIO_TRIM_SECURE;
ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE;
}
uint64_t ms_count = 0;
for (uint64_t i = 0; !vd->vdev_detached &&
i < vd->vdev_top->vdev_ms_count; i++) {
metaslab_t *msp = vd->vdev_top->vdev_ms[i];
/*
* If we've expanded the top-level vdev or it's our
* first pass, calculate our progress.
*/
if (vd->vdev_top->vdev_ms_count != ms_count) {
vdev_trim_calculate_progress(vd);
ms_count = vd->vdev_top->vdev_ms_count;
}
spa_config_exit(spa, SCL_CONFIG, FTAG);
metaslab_disable(msp);
mutex_enter(&msp->ms_lock);
VERIFY0(metaslab_load(msp));
/*
* If a partial TRIM was requested skip metaslabs which have
* never been initialized and thus have never been written.
*/
if (msp->ms_sm == NULL && vd->vdev_trim_partial) {
mutex_exit(&msp->ms_lock);
metaslab_enable(msp, B_FALSE, B_FALSE);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vdev_trim_calculate_progress(vd);
continue;
}
ta.trim_msp = msp;
range_tree_walk(msp->ms_allocatable, vdev_trim_range_add, &ta);
range_tree_vacate(msp->ms_trim, NULL, NULL);
mutex_exit(&msp->ms_lock);
error = vdev_trim_ranges(&ta);
metaslab_enable(msp, B_TRUE, B_FALSE);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
range_tree_vacate(ta.trim_tree, NULL, NULL);
if (error != 0)
break;
}
spa_config_exit(spa, SCL_CONFIG, FTAG);
mutex_enter(&vd->vdev_trim_io_lock);
while (vd->vdev_trim_inflight[0] > 0) {
cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock);
}
mutex_exit(&vd->vdev_trim_io_lock);
range_tree_destroy(ta.trim_tree);
mutex_enter(&vd->vdev_trim_lock);
if (!vd->vdev_trim_exit_wanted) {
if (vdev_writeable(vd)) {
vdev_trim_change_state(vd, VDEV_TRIM_COMPLETE,
vd->vdev_trim_rate, vd->vdev_trim_partial,
vd->vdev_trim_secure);
} else if (vd->vdev_faulted) {
vdev_trim_change_state(vd, VDEV_TRIM_CANCELED,
vd->vdev_trim_rate, vd->vdev_trim_partial,
vd->vdev_trim_secure);
}
}
ASSERT(vd->vdev_trim_thread != NULL || vd->vdev_trim_inflight[0] == 0);
/*
* Drop the vdev_trim_lock while we sync out the txg since it's
* possible that a device might be trying to come online and must
* check to see if it needs to restart a trim. That thread will be
* holding the spa_config_lock which would prevent the txg_wait_synced
* from completing.
*/
mutex_exit(&vd->vdev_trim_lock);
txg_wait_synced(spa_get_dsl(spa), 0);
mutex_enter(&vd->vdev_trim_lock);
vd->vdev_trim_thread = NULL;
cv_broadcast(&vd->vdev_trim_cv);
mutex_exit(&vd->vdev_trim_lock);
thread_exit();
}
/*
* Initiates a manual TRIM for the vdev_t. Callers must hold vdev_trim_lock,
* the vdev_t must be a leaf and cannot already be manually trimming.
*/
void
vdev_trim(vdev_t *vd, uint64_t rate, boolean_t partial, boolean_t secure)
{
ASSERT(MUTEX_HELD(&vd->vdev_trim_lock));
ASSERT(vd->vdev_ops->vdev_op_leaf);
ASSERT(vdev_is_concrete(vd));
ASSERT3P(vd->vdev_trim_thread, ==, NULL);
ASSERT(!vd->vdev_detached);
ASSERT(!vd->vdev_trim_exit_wanted);
ASSERT(!vd->vdev_top->vdev_removing);
vdev_trim_change_state(vd, VDEV_TRIM_ACTIVE, rate, partial, secure);
vd->vdev_trim_thread = thread_create(NULL, 0,
vdev_trim_thread, vd, 0, &p0, TS_RUN, maxclsyspri);
}
/*
* Wait for the trimming thread to be terminated (canceled or stopped).
*/
static void
vdev_trim_stop_wait_impl(vdev_t *vd)
{
ASSERT(MUTEX_HELD(&vd->vdev_trim_lock));
while (vd->vdev_trim_thread != NULL)
cv_wait(&vd->vdev_trim_cv, &vd->vdev_trim_lock);
ASSERT3P(vd->vdev_trim_thread, ==, NULL);
vd->vdev_trim_exit_wanted = B_FALSE;
}
/*
* Wait for vdev trim threads which were listed to cleanly exit.
*/
void
vdev_trim_stop_wait(spa_t *spa, list_t *vd_list)
{
(void) spa;
vdev_t *vd;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
while ((vd = list_remove_head(vd_list)) != NULL) {
mutex_enter(&vd->vdev_trim_lock);
vdev_trim_stop_wait_impl(vd);
mutex_exit(&vd->vdev_trim_lock);
}
}
/*
* Stop trimming a device, with the resultant trimming state being tgt_state.
* For blocking behavior pass NULL for vd_list. Otherwise, when a list_t is
* provided the stopping vdev is inserted in to the list. Callers are then
* required to call vdev_trim_stop_wait() to block for all the trim threads
* to exit. The caller must hold vdev_trim_lock and must not be writing to
* the spa config, as the trimming thread may try to enter the config as a
* reader before exiting.
*/
void
vdev_trim_stop(vdev_t *vd, vdev_trim_state_t tgt_state, list_t *vd_list)
{
ASSERT(!spa_config_held(vd->vdev_spa, SCL_CONFIG|SCL_STATE, RW_WRITER));
ASSERT(MUTEX_HELD(&vd->vdev_trim_lock));
ASSERT(vd->vdev_ops->vdev_op_leaf);
ASSERT(vdev_is_concrete(vd));
/*
* Allow cancel requests to proceed even if the trim thread has
* stopped.
*/
if (vd->vdev_trim_thread == NULL && tgt_state != VDEV_TRIM_CANCELED)
return;
vdev_trim_change_state(vd, tgt_state, 0, 0, 0);
vd->vdev_trim_exit_wanted = B_TRUE;
if (vd_list == NULL) {
vdev_trim_stop_wait_impl(vd);
} else {
ASSERT(MUTEX_HELD(&spa_namespace_lock));
list_insert_tail(vd_list, vd);
}
}
/*
* Requests that all listed vdevs stop trimming.
*/
static void
vdev_trim_stop_all_impl(vdev_t *vd, vdev_trim_state_t tgt_state,
list_t *vd_list)
{
if (vd->vdev_ops->vdev_op_leaf && vdev_is_concrete(vd)) {
mutex_enter(&vd->vdev_trim_lock);
vdev_trim_stop(vd, tgt_state, vd_list);
mutex_exit(&vd->vdev_trim_lock);
return;
}
for (uint64_t i = 0; i < vd->vdev_children; i++) {
vdev_trim_stop_all_impl(vd->vdev_child[i], tgt_state,
vd_list);
}
}
/*
* Convenience function to stop trimming of a vdev tree and set all trim
* thread pointers to NULL.
*/
void
vdev_trim_stop_all(vdev_t *vd, vdev_trim_state_t tgt_state)
{
spa_t *spa = vd->vdev_spa;
list_t vd_list;
vdev_t *vd_l2cache;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
list_create(&vd_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_trim_node));
vdev_trim_stop_all_impl(vd, tgt_state, &vd_list);
/*
* Iterate over cache devices and request stop trimming the
* whole device in case we export the pool or remove the cache
* device prematurely.
*/
for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
vd_l2cache = spa->spa_l2cache.sav_vdevs[i];
vdev_trim_stop_all_impl(vd_l2cache, tgt_state, &vd_list);
}
vdev_trim_stop_wait(spa, &vd_list);
if (vd->vdev_spa->spa_sync_on) {
/* Make sure that our state has been synced to disk */
txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0);
}
list_destroy(&vd_list);
}
/*
* Conditionally restarts a manual TRIM given its on-disk state.
*/
void
vdev_trim_restart(vdev_t *vd)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
ASSERT(!spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));
if (vd->vdev_leaf_zap != 0) {
mutex_enter(&vd->vdev_trim_lock);
uint64_t trim_state = VDEV_TRIM_NONE;
int err = zap_lookup(vd->vdev_spa->spa_meta_objset,
vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_STATE,
sizeof (trim_state), 1, &trim_state);
ASSERT(err == 0 || err == ENOENT);
vd->vdev_trim_state = trim_state;
uint64_t timestamp = 0;
err = zap_lookup(vd->vdev_spa->spa_meta_objset,
vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_ACTION_TIME,
sizeof (timestamp), 1, &timestamp);
ASSERT(err == 0 || err == ENOENT);
vd->vdev_trim_action_time = timestamp;
if (vd->vdev_trim_state == VDEV_TRIM_SUSPENDED ||
vd->vdev_offline) {
/* load progress for reporting, but don't resume */
VERIFY0(vdev_trim_load(vd));
} else if (vd->vdev_trim_state == VDEV_TRIM_ACTIVE &&
vdev_writeable(vd) && !vd->vdev_top->vdev_removing &&
vd->vdev_trim_thread == NULL) {
VERIFY0(vdev_trim_load(vd));
vdev_trim(vd, vd->vdev_trim_rate,
vd->vdev_trim_partial, vd->vdev_trim_secure);
}
mutex_exit(&vd->vdev_trim_lock);
}
for (uint64_t i = 0; i < vd->vdev_children; i++) {
vdev_trim_restart(vd->vdev_child[i]);
}
}
/*
* Used by the automatic TRIM when ZFS_DEBUG_TRIM is set to verify that
* every TRIM range is contained within ms_allocatable.
*/
static void
vdev_trim_range_verify(void *arg, uint64_t start, uint64_t size)
{
trim_args_t *ta = arg;
metaslab_t *msp = ta->trim_msp;
VERIFY3B(msp->ms_loaded, ==, B_TRUE);
VERIFY3U(msp->ms_disabled, >, 0);
VERIFY(range_tree_contains(msp->ms_allocatable, start, size));
}
/*
* Each automatic TRIM thread is responsible for managing the trimming of a
* top-level vdev in the pool. No automatic TRIM state is maintained on-disk.
*
* N.B. This behavior is different from a manual TRIM where a thread
* is created for each leaf vdev, instead of each top-level vdev.
*/
static void
vdev_autotrim_thread(void *arg)
{
vdev_t *vd = arg;
spa_t *spa = vd->vdev_spa;
int shift = 0;
mutex_enter(&vd->vdev_autotrim_lock);
ASSERT3P(vd->vdev_top, ==, vd);
ASSERT3P(vd->vdev_autotrim_thread, !=, NULL);
mutex_exit(&vd->vdev_autotrim_lock);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
uint64_t extent_bytes_max = zfs_trim_extent_bytes_max;
uint64_t extent_bytes_min = zfs_trim_extent_bytes_min;
while (!vdev_autotrim_should_stop(vd)) {
int txgs_per_trim = MAX(zfs_trim_txg_batch, 1);
boolean_t issued_trim = B_FALSE;
/*
* All of the metaslabs are divided in to groups of size
* num_metaslabs / zfs_trim_txg_batch. Each of these groups
* is composed of metaslabs which are spread evenly over the
* device.
*
* For example, when zfs_trim_txg_batch = 32 (default) then
* group 0 will contain metaslabs 0, 32, 64, ...;
* group 1 will contain metaslabs 1, 33, 65, ...;
* group 2 will contain metaslabs 2, 34, 66, ...; and so on.
*
* On each pass through the while() loop one of these groups
* is selected. This is accomplished by using a shift value
* to select the starting metaslab, then striding over the
* metaslabs using the zfs_trim_txg_batch size. This is
* done to accomplish two things.
*
* 1) By dividing the metaslabs in to groups, and making sure
* that each group takes a minimum of one txg to process.
* Then zfs_trim_txg_batch controls the minimum number of
* txgs which must occur before a metaslab is revisited.
*
* 2) Selecting non-consecutive metaslabs distributes the
* TRIM commands for a group evenly over the entire device.
* This can be advantageous for certain types of devices.
*/
for (uint64_t i = shift % txgs_per_trim; i < vd->vdev_ms_count;
i += txgs_per_trim) {
metaslab_t *msp = vd->vdev_ms[i];
range_tree_t *trim_tree;
spa_config_exit(spa, SCL_CONFIG, FTAG);
metaslab_disable(msp);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
mutex_enter(&msp->ms_lock);
/*
* Skip the metaslab when it has never been allocated
* or when there are no recent frees to trim.
*/
if (msp->ms_sm == NULL ||
range_tree_is_empty(msp->ms_trim)) {
mutex_exit(&msp->ms_lock);
metaslab_enable(msp, B_FALSE, B_FALSE);
continue;
}
/*
* Skip the metaslab when it has already been disabled.
* This may happen when a manual TRIM or initialize
* operation is running concurrently. In the case
* of a manual TRIM, the ms_trim tree will have been
* vacated. Only ranges added after the manual TRIM
* disabled the metaslab will be included in the tree.
* These will be processed when the automatic TRIM
* next revisits this metaslab.
*/
if (msp->ms_disabled > 1) {
mutex_exit(&msp->ms_lock);
metaslab_enable(msp, B_FALSE, B_FALSE);
continue;
}
/*
* Allocate an empty range tree which is swapped in
* for the existing ms_trim tree while it is processed.
*/
trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL,
0, 0);
range_tree_swap(&msp->ms_trim, &trim_tree);
ASSERT(range_tree_is_empty(msp->ms_trim));
/*
* There are two cases when constructing the per-vdev
* trim trees for a metaslab. If the top-level vdev
* has no children then it is also a leaf and should
* be trimmed. Otherwise our children are the leaves
* and a trim tree should be constructed for each.
*/
trim_args_t *tap;
uint64_t children = vd->vdev_children;
if (children == 0) {
children = 1;
tap = kmem_zalloc(sizeof (trim_args_t) *
children, KM_SLEEP);
tap[0].trim_vdev = vd;
} else {
tap = kmem_zalloc(sizeof (trim_args_t) *
children, KM_SLEEP);
for (uint64_t c = 0; c < children; c++) {
tap[c].trim_vdev = vd->vdev_child[c];
}
}
for (uint64_t c = 0; c < children; c++) {
trim_args_t *ta = &tap[c];
vdev_t *cvd = ta->trim_vdev;
ta->trim_msp = msp;
ta->trim_extent_bytes_max = extent_bytes_max;
ta->trim_extent_bytes_min = extent_bytes_min;
ta->trim_type = TRIM_TYPE_AUTO;
ta->trim_flags = 0;
if (cvd->vdev_detached ||
!vdev_writeable(cvd) ||
!cvd->vdev_has_trim ||
cvd->vdev_trim_thread != NULL) {
continue;
}
/*
* When a device has an attached hot spare, or
* is being replaced it will not be trimmed.
* This is done to avoid adding additional
* stress to a potentially unhealthy device,
* and to minimize the required rebuild time.
*/
if (!cvd->vdev_ops->vdev_op_leaf)
continue;
ta->trim_tree = range_tree_create(NULL,
RANGE_SEG64, NULL, 0, 0);
range_tree_walk(trim_tree,
vdev_trim_range_add, ta);
}
mutex_exit(&msp->ms_lock);
spa_config_exit(spa, SCL_CONFIG, FTAG);
/*
* Issue the TRIM I/Os for all ranges covered by the
* TRIM trees. These ranges are safe to TRIM because
* no new allocations will be performed until the call
* to metaslab_enabled() below.
*/
for (uint64_t c = 0; c < children; c++) {
trim_args_t *ta = &tap[c];
/*
* Always yield to a manual TRIM if one has
* been started for the child vdev.
*/
if (ta->trim_tree == NULL ||
ta->trim_vdev->vdev_trim_thread != NULL) {
continue;
}
/*
* After this point metaslab_enable() must be
* called with the sync flag set. This is done
* here because vdev_trim_ranges() is allowed
* to be interrupted (EINTR) before issuing all
* of the required TRIM I/Os.
*/
issued_trim = B_TRUE;
int error = vdev_trim_ranges(ta);
if (error)
break;
}
/*
* Verify every range which was trimmed is still
* contained within the ms_allocatable tree.
*/
if (zfs_flags & ZFS_DEBUG_TRIM) {
mutex_enter(&msp->ms_lock);
VERIFY0(metaslab_load(msp));
VERIFY3P(tap[0].trim_msp, ==, msp);
range_tree_walk(trim_tree,
vdev_trim_range_verify, &tap[0]);
mutex_exit(&msp->ms_lock);
}
range_tree_vacate(trim_tree, NULL, NULL);
range_tree_destroy(trim_tree);
metaslab_enable(msp, issued_trim, B_FALSE);
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
for (uint64_t c = 0; c < children; c++) {
trim_args_t *ta = &tap[c];
if (ta->trim_tree == NULL)
continue;
range_tree_vacate(ta->trim_tree, NULL, NULL);
range_tree_destroy(ta->trim_tree);
}
kmem_free(tap, sizeof (trim_args_t) * children);
}
spa_config_exit(spa, SCL_CONFIG, FTAG);
/*
* After completing the group of metaslabs wait for the next
* open txg. This is done to make sure that a minimum of
* zfs_trim_txg_batch txgs will occur before these metaslabs
* are trimmed again.
*/
txg_wait_open(spa_get_dsl(spa), 0, issued_trim);
shift++;
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
}
for (uint64_t c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
mutex_enter(&cvd->vdev_trim_io_lock);
while (cvd->vdev_trim_inflight[1] > 0) {
cv_wait(&cvd->vdev_trim_io_cv,
&cvd->vdev_trim_io_lock);
}
mutex_exit(&cvd->vdev_trim_io_lock);
}
spa_config_exit(spa, SCL_CONFIG, FTAG);
/*
* When exiting because the autotrim property was set to off, then
* abandon any unprocessed ms_trim ranges to reclaim the memory.
*/
if (spa_get_autotrim(spa) == SPA_AUTOTRIM_OFF) {
for (uint64_t i = 0; i < vd->vdev_ms_count; i++) {
metaslab_t *msp = vd->vdev_ms[i];
mutex_enter(&msp->ms_lock);
range_tree_vacate(msp->ms_trim, NULL, NULL);
mutex_exit(&msp->ms_lock);
}
}
mutex_enter(&vd->vdev_autotrim_lock);
ASSERT(vd->vdev_autotrim_thread != NULL);
vd->vdev_autotrim_thread = NULL;
cv_broadcast(&vd->vdev_autotrim_cv);
mutex_exit(&vd->vdev_autotrim_lock);
thread_exit();
}
/*
* Starts an autotrim thread, if needed, for each top-level vdev which can be
* trimmed. A top-level vdev which has been evacuated will never be trimmed.
*/
void
vdev_autotrim(spa_t *spa)
{
vdev_t *root_vd = spa->spa_root_vdev;
for (uint64_t i = 0; i < root_vd->vdev_children; i++) {
vdev_t *tvd = root_vd->vdev_child[i];
mutex_enter(&tvd->vdev_autotrim_lock);
if (vdev_writeable(tvd) && !tvd->vdev_removing &&
tvd->vdev_autotrim_thread == NULL) {
ASSERT3P(tvd->vdev_top, ==, tvd);
tvd->vdev_autotrim_thread = thread_create(NULL, 0,
vdev_autotrim_thread, tvd, 0, &p0, TS_RUN,
maxclsyspri);
ASSERT(tvd->vdev_autotrim_thread != NULL);
}
mutex_exit(&tvd->vdev_autotrim_lock);
}
}
/*
* Wait for the vdev_autotrim_thread associated with the passed top-level
* vdev to be terminated (canceled or stopped).
*/
void
vdev_autotrim_stop_wait(vdev_t *tvd)
{
mutex_enter(&tvd->vdev_autotrim_lock);
if (tvd->vdev_autotrim_thread != NULL) {
tvd->vdev_autotrim_exit_wanted = B_TRUE;
while (tvd->vdev_autotrim_thread != NULL) {
cv_wait(&tvd->vdev_autotrim_cv,
&tvd->vdev_autotrim_lock);
}
ASSERT3P(tvd->vdev_autotrim_thread, ==, NULL);
tvd->vdev_autotrim_exit_wanted = B_FALSE;
}
mutex_exit(&tvd->vdev_autotrim_lock);
}
/*
* Wait for all of the vdev_autotrim_thread associated with the pool to
* be terminated (canceled or stopped).
*/
void
vdev_autotrim_stop_all(spa_t *spa)
{
vdev_t *root_vd = spa->spa_root_vdev;
for (uint64_t i = 0; i < root_vd->vdev_children; i++)
vdev_autotrim_stop_wait(root_vd->vdev_child[i]);
}
/*
* Conditionally restart all of the vdev_autotrim_thread's for the pool.
*/
void
vdev_autotrim_restart(spa_t *spa)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (spa->spa_autotrim)
vdev_autotrim(spa);
}
static void
vdev_trim_l2arc_thread(void *arg)
{
vdev_t *vd = arg;
spa_t *spa = vd->vdev_spa;
l2arc_dev_t *dev = l2arc_vdev_get(vd);
trim_args_t ta;
range_seg64_t physical_rs;
ASSERT(vdev_is_concrete(vd));
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
vd->vdev_trim_last_offset = 0;
vd->vdev_trim_rate = 0;
vd->vdev_trim_partial = 0;
vd->vdev_trim_secure = 0;
bzero(&ta, sizeof (ta));
ta.trim_vdev = vd;
ta.trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
ta.trim_type = TRIM_TYPE_MANUAL;
ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max;
ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE;
ta.trim_flags = 0;
physical_rs.rs_start = vd->vdev_trim_bytes_done = 0;
physical_rs.rs_end = vd->vdev_trim_bytes_est =
vdev_get_min_asize(vd);
range_tree_add(ta.trim_tree, physical_rs.rs_start,
physical_rs.rs_end - physical_rs.rs_start);
mutex_enter(&vd->vdev_trim_lock);
vdev_trim_change_state(vd, VDEV_TRIM_ACTIVE, 0, 0, 0);
mutex_exit(&vd->vdev_trim_lock);
(void) vdev_trim_ranges(&ta);
spa_config_exit(spa, SCL_CONFIG, FTAG);
mutex_enter(&vd->vdev_trim_io_lock);
while (vd->vdev_trim_inflight[TRIM_TYPE_MANUAL] > 0) {
cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock);
}
mutex_exit(&vd->vdev_trim_io_lock);
range_tree_vacate(ta.trim_tree, NULL, NULL);
range_tree_destroy(ta.trim_tree);
mutex_enter(&vd->vdev_trim_lock);
if (!vd->vdev_trim_exit_wanted && vdev_writeable(vd)) {
vdev_trim_change_state(vd, VDEV_TRIM_COMPLETE,
vd->vdev_trim_rate, vd->vdev_trim_partial,
vd->vdev_trim_secure);
}
ASSERT(vd->vdev_trim_thread != NULL ||
vd->vdev_trim_inflight[TRIM_TYPE_MANUAL] == 0);
/*
* Drop the vdev_trim_lock while we sync out the txg since it's
* possible that a device might be trying to come online and
* must check to see if it needs to restart a trim. That thread
* will be holding the spa_config_lock which would prevent the
* txg_wait_synced from completing. Same strategy as in
* vdev_trim_thread().
*/
mutex_exit(&vd->vdev_trim_lock);
txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0);
mutex_enter(&vd->vdev_trim_lock);
/*
* Update the header of the cache device here, before
* broadcasting vdev_trim_cv which may lead to the removal
* of the device. The same applies for setting l2ad_trim_all to
* false.
*/
spa_config_enter(vd->vdev_spa, SCL_L2ARC, vd,
RW_READER);
bzero(dev->l2ad_dev_hdr, dev->l2ad_dev_hdr_asize);
l2arc_dev_hdr_update(dev);
spa_config_exit(vd->vdev_spa, SCL_L2ARC, vd);
vd->vdev_trim_thread = NULL;
if (vd->vdev_trim_state == VDEV_TRIM_COMPLETE)
dev->l2ad_trim_all = B_FALSE;
cv_broadcast(&vd->vdev_trim_cv);
mutex_exit(&vd->vdev_trim_lock);
thread_exit();
}
/*
* Punches out TRIM threads for the L2ARC devices in a spa and assigns them
* to vd->vdev_trim_thread variable. This facilitates the management of
* trimming the whole cache device using TRIM_TYPE_MANUAL upon addition
* to a pool or pool creation or when the header of the device is invalid.
*/
void
vdev_trim_l2arc(spa_t *spa)
{
ASSERT(MUTEX_HELD(&spa_namespace_lock));
/*
* Locate the spa's l2arc devices and kick off TRIM threads.
*/
for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
vdev_t *vd = spa->spa_l2cache.sav_vdevs[i];
l2arc_dev_t *dev = l2arc_vdev_get(vd);
if (dev == NULL || !dev->l2ad_trim_all) {
/*
* Don't attempt TRIM if the vdev is UNAVAIL or if the
* cache device was not marked for whole device TRIM
* (ie l2arc_trim_ahead = 0, or the L2ARC device header
* is valid with trim_state = VDEV_TRIM_COMPLETE and
* l2ad_log_entries > 0).
*/
continue;
}
mutex_enter(&vd->vdev_trim_lock);
ASSERT(vd->vdev_ops->vdev_op_leaf);
ASSERT(vdev_is_concrete(vd));
ASSERT3P(vd->vdev_trim_thread, ==, NULL);
ASSERT(!vd->vdev_detached);
ASSERT(!vd->vdev_trim_exit_wanted);
ASSERT(!vd->vdev_top->vdev_removing);
vdev_trim_change_state(vd, VDEV_TRIM_ACTIVE, 0, 0, 0);
vd->vdev_trim_thread = thread_create(NULL, 0,
vdev_trim_l2arc_thread, vd, 0, &p0, TS_RUN, maxclsyspri);
mutex_exit(&vd->vdev_trim_lock);
}
}
/*
* A wrapper which calls vdev_trim_ranges(). It is intended to be called
* on leaf vdevs.
*/
int
vdev_trim_simple(vdev_t *vd, uint64_t start, uint64_t size)
{
trim_args_t ta;
range_seg64_t physical_rs;
int error;
physical_rs.rs_start = start;
physical_rs.rs_end = start + size;
ASSERT(vdev_is_concrete(vd));
ASSERT(vd->vdev_ops->vdev_op_leaf);
ASSERT(!vd->vdev_detached);
ASSERT(!vd->vdev_top->vdev_removing);
bzero(&ta, sizeof (ta));
ta.trim_vdev = vd;
ta.trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
ta.trim_type = TRIM_TYPE_SIMPLE;
ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max;
ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE;
ta.trim_flags = 0;
ASSERT3U(physical_rs.rs_end, >=, physical_rs.rs_start);
if (physical_rs.rs_end > physical_rs.rs_start) {
range_tree_add(ta.trim_tree, physical_rs.rs_start,
physical_rs.rs_end - physical_rs.rs_start);
} else {
ASSERT3U(physical_rs.rs_end, ==, physical_rs.rs_start);
}
error = vdev_trim_ranges(&ta);
mutex_enter(&vd->vdev_trim_io_lock);
while (vd->vdev_trim_inflight[TRIM_TYPE_SIMPLE] > 0) {
cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock);
}
mutex_exit(&vd->vdev_trim_io_lock);
range_tree_vacate(ta.trim_tree, NULL, NULL);
range_tree_destroy(ta.trim_tree);
return (error);
}
EXPORT_SYMBOL(vdev_trim);
EXPORT_SYMBOL(vdev_trim_stop);
EXPORT_SYMBOL(vdev_trim_stop_all);
EXPORT_SYMBOL(vdev_trim_stop_wait);
EXPORT_SYMBOL(vdev_trim_restart);
EXPORT_SYMBOL(vdev_autotrim);
EXPORT_SYMBOL(vdev_autotrim_stop_all);
EXPORT_SYMBOL(vdev_autotrim_stop_wait);
EXPORT_SYMBOL(vdev_autotrim_restart);
EXPORT_SYMBOL(vdev_trim_l2arc);
EXPORT_SYMBOL(vdev_trim_simple);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, extent_bytes_max, UINT, ZMOD_RW,
"Max size of TRIM commands, larger will be split");
ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, extent_bytes_min, UINT, ZMOD_RW,
"Min size of TRIM commands, smaller will be skipped");
ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, metaslab_skip, UINT, ZMOD_RW,
"Skip metaslabs which have never been initialized");
ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, txg_batch, UINT, ZMOD_RW,
"Min number of txgs to aggregate frees before issuing TRIM");
ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, queue_limit, UINT, ZMOD_RW,
"Max queued TRIMs outstanding per leaf vdev");
/* END CSTYLED */
diff --git a/module/zfs/zap.c b/module/zfs/zap.c
index 6f03beef3bdb..98ed284c992c 100644
--- a/module/zfs/zap.c
+++ b/module/zfs/zap.c
@@ -1,1386 +1,1386 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
*/
/*
* This file contains the top half of the zfs directory structure
* implementation. The bottom half is in zap_leaf.c.
*
* The zdir is an extendable hash data structure. There is a table of
* pointers to buckets (zap_t->zd_data->zd_leafs). The buckets are
* each a constant size and hold a variable number of directory entries.
* The buckets (aka "leaf nodes") are implemented in zap_leaf.c.
*
* The pointer table holds a power of 2 number of pointers.
* (1<<zap_t->zd_data->zd_phys->zd_prefix_len). The bucket pointed to
* by the pointer at index i in the table holds entries whose hash value
* has a zd_prefix_len - bit prefix
*/
#include <sys/spa.h>
#include <sys/dmu.h>
#include <sys/zfs_context.h>
#include <sys/zfs_znode.h>
#include <sys/fs/zfs.h>
#include <sys/zap.h>
#include <sys/zap_impl.h>
#include <sys/zap_leaf.h>
/*
* If zap_iterate_prefetch is set, we will prefetch the entire ZAP object
* (all leaf blocks) when we start iterating over it.
*
* For zap_cursor_init(), the callers all intend to iterate through all the
* entries. There are a few cases where an error (typically i/o error) could
* cause it to bail out early.
*
* For zap_cursor_init_serialized(), there are callers that do the iteration
* outside of ZFS. Typically they would iterate over everything, but we
* don't have control of that. E.g. zfs_ioc_snapshot_list_next(),
* zcp_snapshots_iter(), and other iterators over things in the MOS - these
* are called by /sbin/zfs and channel programs. The other example is
* zfs_readdir() which iterates over directory entries for the getdents()
* syscall. /sbin/ls iterates to the end (unless it receives a signal), but
* userland doesn't have to.
*
* Given that the ZAP entries aren't returned in a specific order, the only
* legitimate use cases for partial iteration would be:
*
* 1. Pagination: e.g. you only want to display 100 entries at a time, so you
* get the first 100 and then wait for the user to hit "next page", which
* they may never do).
*
* 2. You want to know if there are more than X entries, without relying on
* the zfs-specific implementation of the directory's st_size (which is
* the number of entries).
*/
-int zap_iterate_prefetch = B_TRUE;
+static int zap_iterate_prefetch = B_TRUE;
int fzap_default_block_shift = 14; /* 16k blocksize */
extern inline zap_phys_t *zap_f_phys(zap_t *zap);
static uint64_t zap_allocate_blocks(zap_t *zap, int nblocks);
void
fzap_byteswap(void *vbuf, size_t size)
{
uint64_t block_type = *(uint64_t *)vbuf;
if (block_type == ZBT_LEAF || block_type == BSWAP_64(ZBT_LEAF))
zap_leaf_byteswap(vbuf, size);
else {
/* it's a ptrtbl block */
byteswap_uint64_array(vbuf, size);
}
}
void
fzap_upgrade(zap_t *zap, dmu_tx_t *tx, zap_flags_t flags)
{
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
zap->zap_ismicro = FALSE;
zap->zap_dbu.dbu_evict_func_sync = zap_evict_sync;
zap->zap_dbu.dbu_evict_func_async = NULL;
mutex_init(&zap->zap_f.zap_num_entries_mtx, 0, MUTEX_DEFAULT, 0);
zap->zap_f.zap_block_shift = highbit64(zap->zap_dbuf->db_size) - 1;
zap_phys_t *zp = zap_f_phys(zap);
/*
* explicitly zero it since it might be coming from an
* initialized microzap
*/
bzero(zap->zap_dbuf->db_data, zap->zap_dbuf->db_size);
zp->zap_block_type = ZBT_HEADER;
zp->zap_magic = ZAP_MAGIC;
zp->zap_ptrtbl.zt_shift = ZAP_EMBEDDED_PTRTBL_SHIFT(zap);
zp->zap_freeblk = 2; /* block 1 will be the first leaf */
zp->zap_num_leafs = 1;
zp->zap_num_entries = 0;
zp->zap_salt = zap->zap_salt;
zp->zap_normflags = zap->zap_normflags;
zp->zap_flags = flags;
/* block 1 will be the first leaf */
for (int i = 0; i < (1<<zp->zap_ptrtbl.zt_shift); i++)
ZAP_EMBEDDED_PTRTBL_ENT(zap, i) = 1;
/*
* set up block 1 - the first leaf
*/
dmu_buf_t *db;
VERIFY0(dmu_buf_hold(zap->zap_objset, zap->zap_object,
1<<FZAP_BLOCK_SHIFT(zap), FTAG, &db, DMU_READ_NO_PREFETCH));
dmu_buf_will_dirty(db, tx);
zap_leaf_t *l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
l->l_dbuf = db;
zap_leaf_init(l, zp->zap_normflags != 0);
kmem_free(l, sizeof (zap_leaf_t));
dmu_buf_rele(db, FTAG);
}
static int
zap_tryupgradedir(zap_t *zap, dmu_tx_t *tx)
{
if (RW_WRITE_HELD(&zap->zap_rwlock))
return (1);
if (rw_tryupgrade(&zap->zap_rwlock)) {
dmu_buf_will_dirty(zap->zap_dbuf, tx);
return (1);
}
return (0);
}
/*
* Generic routines for dealing with the pointer & cookie tables.
*/
static int
zap_table_grow(zap_t *zap, zap_table_phys_t *tbl,
void (*transfer_func)(const uint64_t *src, uint64_t *dst, int n),
dmu_tx_t *tx)
{
uint64_t newblk;
int bs = FZAP_BLOCK_SHIFT(zap);
int hepb = 1<<(bs-4);
/* hepb = half the number of entries in a block */
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
ASSERT(tbl->zt_blk != 0);
ASSERT(tbl->zt_numblks > 0);
if (tbl->zt_nextblk != 0) {
newblk = tbl->zt_nextblk;
} else {
newblk = zap_allocate_blocks(zap, tbl->zt_numblks * 2);
tbl->zt_nextblk = newblk;
ASSERT0(tbl->zt_blks_copied);
dmu_prefetch(zap->zap_objset, zap->zap_object, 0,
tbl->zt_blk << bs, tbl->zt_numblks << bs,
ZIO_PRIORITY_SYNC_READ);
}
/*
* Copy the ptrtbl from the old to new location.
*/
uint64_t b = tbl->zt_blks_copied;
dmu_buf_t *db_old;
int err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
(tbl->zt_blk + b) << bs, FTAG, &db_old, DMU_READ_NO_PREFETCH);
if (err != 0)
return (err);
/* first half of entries in old[b] go to new[2*b+0] */
dmu_buf_t *db_new;
VERIFY0(dmu_buf_hold(zap->zap_objset, zap->zap_object,
(newblk + 2*b+0) << bs, FTAG, &db_new, DMU_READ_NO_PREFETCH));
dmu_buf_will_dirty(db_new, tx);
transfer_func(db_old->db_data, db_new->db_data, hepb);
dmu_buf_rele(db_new, FTAG);
/* second half of entries in old[b] go to new[2*b+1] */
VERIFY0(dmu_buf_hold(zap->zap_objset, zap->zap_object,
(newblk + 2*b+1) << bs, FTAG, &db_new, DMU_READ_NO_PREFETCH));
dmu_buf_will_dirty(db_new, tx);
transfer_func((uint64_t *)db_old->db_data + hepb,
db_new->db_data, hepb);
dmu_buf_rele(db_new, FTAG);
dmu_buf_rele(db_old, FTAG);
tbl->zt_blks_copied++;
dprintf("copied block %llu of %llu\n",
(u_longlong_t)tbl->zt_blks_copied,
(u_longlong_t)tbl->zt_numblks);
if (tbl->zt_blks_copied == tbl->zt_numblks) {
(void) dmu_free_range(zap->zap_objset, zap->zap_object,
tbl->zt_blk << bs, tbl->zt_numblks << bs, tx);
tbl->zt_blk = newblk;
tbl->zt_numblks *= 2;
tbl->zt_shift++;
tbl->zt_nextblk = 0;
tbl->zt_blks_copied = 0;
dprintf("finished; numblocks now %llu (%uk entries)\n",
(u_longlong_t)tbl->zt_numblks, 1<<(tbl->zt_shift-10));
}
return (0);
}
static int
zap_table_store(zap_t *zap, zap_table_phys_t *tbl, uint64_t idx, uint64_t val,
dmu_tx_t *tx)
{
int bs = FZAP_BLOCK_SHIFT(zap);
ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
ASSERT(tbl->zt_blk != 0);
dprintf("storing %llx at index %llx\n", (u_longlong_t)val,
(u_longlong_t)idx);
uint64_t blk = idx >> (bs-3);
uint64_t off = idx & ((1<<(bs-3))-1);
dmu_buf_t *db;
int err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
(tbl->zt_blk + blk) << bs, FTAG, &db, DMU_READ_NO_PREFETCH);
if (err != 0)
return (err);
dmu_buf_will_dirty(db, tx);
if (tbl->zt_nextblk != 0) {
uint64_t idx2 = idx * 2;
uint64_t blk2 = idx2 >> (bs-3);
uint64_t off2 = idx2 & ((1<<(bs-3))-1);
dmu_buf_t *db2;
err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
(tbl->zt_nextblk + blk2) << bs, FTAG, &db2,
DMU_READ_NO_PREFETCH);
if (err != 0) {
dmu_buf_rele(db, FTAG);
return (err);
}
dmu_buf_will_dirty(db2, tx);
((uint64_t *)db2->db_data)[off2] = val;
((uint64_t *)db2->db_data)[off2+1] = val;
dmu_buf_rele(db2, FTAG);
}
((uint64_t *)db->db_data)[off] = val;
dmu_buf_rele(db, FTAG);
return (0);
}
static int
zap_table_load(zap_t *zap, zap_table_phys_t *tbl, uint64_t idx, uint64_t *valp)
{
int bs = FZAP_BLOCK_SHIFT(zap);
ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
uint64_t blk = idx >> (bs-3);
uint64_t off = idx & ((1<<(bs-3))-1);
/*
* Note: this is equivalent to dmu_buf_hold(), but we use
* _dnode_enter / _by_dnode because it's faster because we don't
* have to hold the dnode.
*/
dnode_t *dn = dmu_buf_dnode_enter(zap->zap_dbuf);
dmu_buf_t *db;
int err = dmu_buf_hold_by_dnode(dn,
(tbl->zt_blk + blk) << bs, FTAG, &db, DMU_READ_NO_PREFETCH);
dmu_buf_dnode_exit(zap->zap_dbuf);
if (err != 0)
return (err);
*valp = ((uint64_t *)db->db_data)[off];
dmu_buf_rele(db, FTAG);
if (tbl->zt_nextblk != 0) {
/*
* read the nextblk for the sake of i/o error checking,
* so that zap_table_load() will catch errors for
* zap_table_store.
*/
blk = (idx*2) >> (bs-3);
dn = dmu_buf_dnode_enter(zap->zap_dbuf);
err = dmu_buf_hold_by_dnode(dn,
(tbl->zt_nextblk + blk) << bs, FTAG, &db,
DMU_READ_NO_PREFETCH);
dmu_buf_dnode_exit(zap->zap_dbuf);
if (err == 0)
dmu_buf_rele(db, FTAG);
}
return (err);
}
/*
* Routines for growing the ptrtbl.
*/
static void
zap_ptrtbl_transfer(const uint64_t *src, uint64_t *dst, int n)
{
for (int i = 0; i < n; i++) {
uint64_t lb = src[i];
dst[2 * i + 0] = lb;
dst[2 * i + 1] = lb;
}
}
static int
zap_grow_ptrtbl(zap_t *zap, dmu_tx_t *tx)
{
/*
* The pointer table should never use more hash bits than we
* have (otherwise we'd be using useless zero bits to index it).
* If we are within 2 bits of running out, stop growing, since
* this is already an aberrant condition.
*/
if (zap_f_phys(zap)->zap_ptrtbl.zt_shift >= zap_hashbits(zap) - 2)
return (SET_ERROR(ENOSPC));
if (zap_f_phys(zap)->zap_ptrtbl.zt_numblks == 0) {
/*
* We are outgrowing the "embedded" ptrtbl (the one
* stored in the header block). Give it its own entire
* block, which will double the size of the ptrtbl.
*/
ASSERT3U(zap_f_phys(zap)->zap_ptrtbl.zt_shift, ==,
ZAP_EMBEDDED_PTRTBL_SHIFT(zap));
ASSERT0(zap_f_phys(zap)->zap_ptrtbl.zt_blk);
uint64_t newblk = zap_allocate_blocks(zap, 1);
dmu_buf_t *db_new;
int err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
newblk << FZAP_BLOCK_SHIFT(zap), FTAG, &db_new,
DMU_READ_NO_PREFETCH);
if (err != 0)
return (err);
dmu_buf_will_dirty(db_new, tx);
zap_ptrtbl_transfer(&ZAP_EMBEDDED_PTRTBL_ENT(zap, 0),
db_new->db_data, 1 << ZAP_EMBEDDED_PTRTBL_SHIFT(zap));
dmu_buf_rele(db_new, FTAG);
zap_f_phys(zap)->zap_ptrtbl.zt_blk = newblk;
zap_f_phys(zap)->zap_ptrtbl.zt_numblks = 1;
zap_f_phys(zap)->zap_ptrtbl.zt_shift++;
ASSERT3U(1ULL << zap_f_phys(zap)->zap_ptrtbl.zt_shift, ==,
zap_f_phys(zap)->zap_ptrtbl.zt_numblks <<
(FZAP_BLOCK_SHIFT(zap)-3));
return (0);
} else {
return (zap_table_grow(zap, &zap_f_phys(zap)->zap_ptrtbl,
zap_ptrtbl_transfer, tx));
}
}
static void
zap_increment_num_entries(zap_t *zap, int delta, dmu_tx_t *tx)
{
dmu_buf_will_dirty(zap->zap_dbuf, tx);
mutex_enter(&zap->zap_f.zap_num_entries_mtx);
ASSERT(delta > 0 || zap_f_phys(zap)->zap_num_entries >= -delta);
zap_f_phys(zap)->zap_num_entries += delta;
mutex_exit(&zap->zap_f.zap_num_entries_mtx);
}
static uint64_t
zap_allocate_blocks(zap_t *zap, int nblocks)
{
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
uint64_t newblk = zap_f_phys(zap)->zap_freeblk;
zap_f_phys(zap)->zap_freeblk += nblocks;
return (newblk);
}
static void
zap_leaf_evict_sync(void *dbu)
{
zap_leaf_t *l = dbu;
rw_destroy(&l->l_rwlock);
kmem_free(l, sizeof (zap_leaf_t));
}
static zap_leaf_t *
zap_create_leaf(zap_t *zap, dmu_tx_t *tx)
{
zap_leaf_t *l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
rw_init(&l->l_rwlock, NULL, RW_NOLOCKDEP, NULL);
rw_enter(&l->l_rwlock, RW_WRITER);
l->l_blkid = zap_allocate_blocks(zap, 1);
l->l_dbuf = NULL;
VERIFY0(dmu_buf_hold(zap->zap_objset, zap->zap_object,
l->l_blkid << FZAP_BLOCK_SHIFT(zap), NULL, &l->l_dbuf,
DMU_READ_NO_PREFETCH));
dmu_buf_init_user(&l->l_dbu, zap_leaf_evict_sync, NULL, &l->l_dbuf);
VERIFY3P(NULL, ==, dmu_buf_set_user(l->l_dbuf, &l->l_dbu));
dmu_buf_will_dirty(l->l_dbuf, tx);
zap_leaf_init(l, zap->zap_normflags != 0);
zap_f_phys(zap)->zap_num_leafs++;
return (l);
}
int
fzap_count(zap_t *zap, uint64_t *count)
{
ASSERT(!zap->zap_ismicro);
mutex_enter(&zap->zap_f.zap_num_entries_mtx); /* unnecessary */
*count = zap_f_phys(zap)->zap_num_entries;
mutex_exit(&zap->zap_f.zap_num_entries_mtx);
return (0);
}
/*
* Routines for obtaining zap_leaf_t's
*/
void
zap_put_leaf(zap_leaf_t *l)
{
rw_exit(&l->l_rwlock);
dmu_buf_rele(l->l_dbuf, NULL);
}
static zap_leaf_t *
zap_open_leaf(uint64_t blkid, dmu_buf_t *db)
{
ASSERT(blkid != 0);
zap_leaf_t *l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
rw_init(&l->l_rwlock, NULL, RW_DEFAULT, NULL);
rw_enter(&l->l_rwlock, RW_WRITER);
l->l_blkid = blkid;
l->l_bs = highbit64(db->db_size) - 1;
l->l_dbuf = db;
dmu_buf_init_user(&l->l_dbu, zap_leaf_evict_sync, NULL, &l->l_dbuf);
zap_leaf_t *winner = dmu_buf_set_user(db, &l->l_dbu);
rw_exit(&l->l_rwlock);
if (winner != NULL) {
/* someone else set it first */
zap_leaf_evict_sync(&l->l_dbu);
l = winner;
}
/*
* lhr_pad was previously used for the next leaf in the leaf
* chain. There should be no chained leafs (as we have removed
* support for them).
*/
ASSERT0(zap_leaf_phys(l)->l_hdr.lh_pad1);
/*
* There should be more hash entries than there can be
* chunks to put in the hash table
*/
ASSERT3U(ZAP_LEAF_HASH_NUMENTRIES(l), >, ZAP_LEAF_NUMCHUNKS(l) / 3);
/* The chunks should begin at the end of the hash table */
ASSERT3P(&ZAP_LEAF_CHUNK(l, 0), ==, (zap_leaf_chunk_t *)
&zap_leaf_phys(l)->l_hash[ZAP_LEAF_HASH_NUMENTRIES(l)]);
/* The chunks should end at the end of the block */
ASSERT3U((uintptr_t)&ZAP_LEAF_CHUNK(l, ZAP_LEAF_NUMCHUNKS(l)) -
(uintptr_t)zap_leaf_phys(l), ==, l->l_dbuf->db_size);
return (l);
}
static int
zap_get_leaf_byblk(zap_t *zap, uint64_t blkid, dmu_tx_t *tx, krw_t lt,
zap_leaf_t **lp)
{
dmu_buf_t *db;
ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
/*
* If system crashed just after dmu_free_long_range in zfs_rmnode, we
* would be left with an empty xattr dir in delete queue. blkid=0
* would be passed in when doing zfs_purgedir. If that's the case we
* should just return immediately. The underlying objects should
* already be freed, so this should be perfectly fine.
*/
if (blkid == 0)
return (SET_ERROR(ENOENT));
int bs = FZAP_BLOCK_SHIFT(zap);
dnode_t *dn = dmu_buf_dnode_enter(zap->zap_dbuf);
int err = dmu_buf_hold_by_dnode(dn,
blkid << bs, NULL, &db, DMU_READ_NO_PREFETCH);
dmu_buf_dnode_exit(zap->zap_dbuf);
if (err != 0)
return (err);
ASSERT3U(db->db_object, ==, zap->zap_object);
ASSERT3U(db->db_offset, ==, blkid << bs);
ASSERT3U(db->db_size, ==, 1 << bs);
ASSERT(blkid != 0);
zap_leaf_t *l = dmu_buf_get_user(db);
if (l == NULL)
l = zap_open_leaf(blkid, db);
rw_enter(&l->l_rwlock, lt);
/*
* Must lock before dirtying, otherwise zap_leaf_phys(l) could change,
* causing ASSERT below to fail.
*/
if (lt == RW_WRITER)
dmu_buf_will_dirty(db, tx);
ASSERT3U(l->l_blkid, ==, blkid);
ASSERT3P(l->l_dbuf, ==, db);
ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_block_type, ==, ZBT_LEAF);
ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_magic, ==, ZAP_LEAF_MAGIC);
*lp = l;
return (0);
}
static int
zap_idx_to_blk(zap_t *zap, uint64_t idx, uint64_t *valp)
{
ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
if (zap_f_phys(zap)->zap_ptrtbl.zt_numblks == 0) {
ASSERT3U(idx, <,
(1ULL << zap_f_phys(zap)->zap_ptrtbl.zt_shift));
*valp = ZAP_EMBEDDED_PTRTBL_ENT(zap, idx);
return (0);
} else {
return (zap_table_load(zap, &zap_f_phys(zap)->zap_ptrtbl,
idx, valp));
}
}
static int
zap_set_idx_to_blk(zap_t *zap, uint64_t idx, uint64_t blk, dmu_tx_t *tx)
{
ASSERT(tx != NULL);
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
if (zap_f_phys(zap)->zap_ptrtbl.zt_blk == 0) {
ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) = blk;
return (0);
} else {
return (zap_table_store(zap, &zap_f_phys(zap)->zap_ptrtbl,
idx, blk, tx));
}
}
static int
zap_deref_leaf(zap_t *zap, uint64_t h, dmu_tx_t *tx, krw_t lt, zap_leaf_t **lp)
{
uint64_t blk;
ASSERT(zap->zap_dbuf == NULL ||
zap_f_phys(zap) == zap->zap_dbuf->db_data);
/* Reality check for corrupt zap objects (leaf or header). */
if ((zap_f_phys(zap)->zap_block_type != ZBT_LEAF &&
zap_f_phys(zap)->zap_block_type != ZBT_HEADER) ||
zap_f_phys(zap)->zap_magic != ZAP_MAGIC) {
return (SET_ERROR(EIO));
}
uint64_t idx = ZAP_HASH_IDX(h, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
int err = zap_idx_to_blk(zap, idx, &blk);
if (err != 0)
return (err);
err = zap_get_leaf_byblk(zap, blk, tx, lt, lp);
ASSERT(err ||
ZAP_HASH_IDX(h, zap_leaf_phys(*lp)->l_hdr.lh_prefix_len) ==
zap_leaf_phys(*lp)->l_hdr.lh_prefix);
return (err);
}
static int
zap_expand_leaf(zap_name_t *zn, zap_leaf_t *l,
void *tag, dmu_tx_t *tx, zap_leaf_t **lp)
{
zap_t *zap = zn->zn_zap;
uint64_t hash = zn->zn_hash;
int err;
int old_prefix_len = zap_leaf_phys(l)->l_hdr.lh_prefix_len;
ASSERT3U(old_prefix_len, <=, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
ASSERT3U(ZAP_HASH_IDX(hash, old_prefix_len), ==,
zap_leaf_phys(l)->l_hdr.lh_prefix);
if (zap_tryupgradedir(zap, tx) == 0 ||
old_prefix_len == zap_f_phys(zap)->zap_ptrtbl.zt_shift) {
/* We failed to upgrade, or need to grow the pointer table */
objset_t *os = zap->zap_objset;
uint64_t object = zap->zap_object;
zap_put_leaf(l);
zap_unlockdir(zap, tag);
err = zap_lockdir(os, object, tx, RW_WRITER,
FALSE, FALSE, tag, &zn->zn_zap);
zap = zn->zn_zap;
if (err != 0)
return (err);
ASSERT(!zap->zap_ismicro);
while (old_prefix_len ==
zap_f_phys(zap)->zap_ptrtbl.zt_shift) {
err = zap_grow_ptrtbl(zap, tx);
if (err != 0)
return (err);
}
err = zap_deref_leaf(zap, hash, tx, RW_WRITER, &l);
if (err != 0)
return (err);
if (zap_leaf_phys(l)->l_hdr.lh_prefix_len != old_prefix_len) {
/* it split while our locks were down */
*lp = l;
return (0);
}
}
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
ASSERT3U(old_prefix_len, <, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
ASSERT3U(ZAP_HASH_IDX(hash, old_prefix_len), ==,
zap_leaf_phys(l)->l_hdr.lh_prefix);
int prefix_diff = zap_f_phys(zap)->zap_ptrtbl.zt_shift -
(old_prefix_len + 1);
uint64_t sibling =
(ZAP_HASH_IDX(hash, old_prefix_len + 1) | 1) << prefix_diff;
/* check for i/o errors before doing zap_leaf_split */
for (int i = 0; i < (1ULL << prefix_diff); i++) {
uint64_t blk;
err = zap_idx_to_blk(zap, sibling + i, &blk);
if (err != 0)
return (err);
ASSERT3U(blk, ==, l->l_blkid);
}
zap_leaf_t *nl = zap_create_leaf(zap, tx);
zap_leaf_split(l, nl, zap->zap_normflags != 0);
/* set sibling pointers */
for (int i = 0; i < (1ULL << prefix_diff); i++) {
err = zap_set_idx_to_blk(zap, sibling + i, nl->l_blkid, tx);
ASSERT0(err); /* we checked for i/o errors above */
}
ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_prefix_len, >, 0);
if (hash & (1ULL << (64 - zap_leaf_phys(l)->l_hdr.lh_prefix_len))) {
/* we want the sibling */
zap_put_leaf(l);
*lp = nl;
} else {
zap_put_leaf(nl);
*lp = l;
}
return (0);
}
static void
zap_put_leaf_maybe_grow_ptrtbl(zap_name_t *zn, zap_leaf_t *l,
void *tag, dmu_tx_t *tx)
{
zap_t *zap = zn->zn_zap;
int shift = zap_f_phys(zap)->zap_ptrtbl.zt_shift;
int leaffull = (zap_leaf_phys(l)->l_hdr.lh_prefix_len == shift &&
zap_leaf_phys(l)->l_hdr.lh_nfree < ZAP_LEAF_LOW_WATER);
zap_put_leaf(l);
if (leaffull || zap_f_phys(zap)->zap_ptrtbl.zt_nextblk) {
/*
* We are in the middle of growing the pointer table, or
* this leaf will soon make us grow it.
*/
if (zap_tryupgradedir(zap, tx) == 0) {
objset_t *os = zap->zap_objset;
uint64_t zapobj = zap->zap_object;
zap_unlockdir(zap, tag);
int err = zap_lockdir(os, zapobj, tx,
RW_WRITER, FALSE, FALSE, tag, &zn->zn_zap);
zap = zn->zn_zap;
if (err != 0)
return;
}
/* could have finished growing while our locks were down */
if (zap_f_phys(zap)->zap_ptrtbl.zt_shift == shift)
(void) zap_grow_ptrtbl(zap, tx);
}
}
static int
fzap_checkname(zap_name_t *zn)
{
if (zn->zn_key_orig_numints * zn->zn_key_intlen > ZAP_MAXNAMELEN)
return (SET_ERROR(ENAMETOOLONG));
return (0);
}
static int
fzap_checksize(uint64_t integer_size, uint64_t num_integers)
{
/* Only integer sizes supported by C */
switch (integer_size) {
case 1:
case 2:
case 4:
case 8:
break;
default:
return (SET_ERROR(EINVAL));
}
if (integer_size * num_integers > ZAP_MAXVALUELEN)
return (SET_ERROR(E2BIG));
return (0);
}
static int
fzap_check(zap_name_t *zn, uint64_t integer_size, uint64_t num_integers)
{
int err = fzap_checkname(zn);
if (err != 0)
return (err);
return (fzap_checksize(integer_size, num_integers));
}
/*
* Routines for manipulating attributes.
*/
int
fzap_lookup(zap_name_t *zn,
uint64_t integer_size, uint64_t num_integers, void *buf,
char *realname, int rn_len, boolean_t *ncp)
{
zap_leaf_t *l;
zap_entry_handle_t zeh;
int err = fzap_checkname(zn);
if (err != 0)
return (err);
err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, NULL, RW_READER, &l);
if (err != 0)
return (err);
err = zap_leaf_lookup(l, zn, &zeh);
if (err == 0) {
if ((err = fzap_checksize(integer_size, num_integers)) != 0) {
zap_put_leaf(l);
return (err);
}
err = zap_entry_read(&zeh, integer_size, num_integers, buf);
(void) zap_entry_read_name(zn->zn_zap, &zeh, rn_len, realname);
if (ncp) {
*ncp = zap_entry_normalization_conflict(&zeh,
zn, NULL, zn->zn_zap);
}
}
zap_put_leaf(l);
return (err);
}
int
fzap_add_cd(zap_name_t *zn,
uint64_t integer_size, uint64_t num_integers,
const void *val, uint32_t cd, void *tag, dmu_tx_t *tx)
{
zap_leaf_t *l;
int err;
zap_entry_handle_t zeh;
zap_t *zap = zn->zn_zap;
ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
ASSERT(!zap->zap_ismicro);
ASSERT(fzap_check(zn, integer_size, num_integers) == 0);
err = zap_deref_leaf(zap, zn->zn_hash, tx, RW_WRITER, &l);
if (err != 0)
return (err);
retry:
err = zap_leaf_lookup(l, zn, &zeh);
if (err == 0) {
err = SET_ERROR(EEXIST);
goto out;
}
if (err != ENOENT)
goto out;
err = zap_entry_create(l, zn, cd,
integer_size, num_integers, val, &zeh);
if (err == 0) {
zap_increment_num_entries(zap, 1, tx);
} else if (err == EAGAIN) {
err = zap_expand_leaf(zn, l, tag, tx, &l);
zap = zn->zn_zap; /* zap_expand_leaf() may change zap */
if (err == 0) {
goto retry;
} else if (err == ENOSPC) {
/*
* If we failed to expand the leaf, then bailout
* as there is no point trying
* zap_put_leaf_maybe_grow_ptrtbl().
*/
return (err);
}
}
out:
if (zap != NULL)
zap_put_leaf_maybe_grow_ptrtbl(zn, l, tag, tx);
return (err);
}
int
fzap_add(zap_name_t *zn,
uint64_t integer_size, uint64_t num_integers,
const void *val, void *tag, dmu_tx_t *tx)
{
int err = fzap_check(zn, integer_size, num_integers);
if (err != 0)
return (err);
return (fzap_add_cd(zn, integer_size, num_integers,
val, ZAP_NEED_CD, tag, tx));
}
int
fzap_update(zap_name_t *zn,
int integer_size, uint64_t num_integers, const void *val,
void *tag, dmu_tx_t *tx)
{
zap_leaf_t *l;
int err;
boolean_t create;
zap_entry_handle_t zeh;
zap_t *zap = zn->zn_zap;
ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
err = fzap_check(zn, integer_size, num_integers);
if (err != 0)
return (err);
err = zap_deref_leaf(zap, zn->zn_hash, tx, RW_WRITER, &l);
if (err != 0)
return (err);
retry:
err = zap_leaf_lookup(l, zn, &zeh);
create = (err == ENOENT);
ASSERT(err == 0 || err == ENOENT);
if (create) {
err = zap_entry_create(l, zn, ZAP_NEED_CD,
integer_size, num_integers, val, &zeh);
if (err == 0)
zap_increment_num_entries(zap, 1, tx);
} else {
err = zap_entry_update(&zeh, integer_size, num_integers, val);
}
if (err == EAGAIN) {
err = zap_expand_leaf(zn, l, tag, tx, &l);
zap = zn->zn_zap; /* zap_expand_leaf() may change zap */
if (err == 0)
goto retry;
}
if (zap != NULL)
zap_put_leaf_maybe_grow_ptrtbl(zn, l, tag, tx);
return (err);
}
int
fzap_length(zap_name_t *zn,
uint64_t *integer_size, uint64_t *num_integers)
{
zap_leaf_t *l;
int err;
zap_entry_handle_t zeh;
err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, NULL, RW_READER, &l);
if (err != 0)
return (err);
err = zap_leaf_lookup(l, zn, &zeh);
if (err != 0)
goto out;
if (integer_size != 0)
*integer_size = zeh.zeh_integer_size;
if (num_integers != 0)
*num_integers = zeh.zeh_num_integers;
out:
zap_put_leaf(l);
return (err);
}
int
fzap_remove(zap_name_t *zn, dmu_tx_t *tx)
{
zap_leaf_t *l;
int err;
zap_entry_handle_t zeh;
err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, tx, RW_WRITER, &l);
if (err != 0)
return (err);
err = zap_leaf_lookup(l, zn, &zeh);
if (err == 0) {
zap_entry_remove(&zeh);
zap_increment_num_entries(zn->zn_zap, -1, tx);
}
zap_put_leaf(l);
return (err);
}
void
fzap_prefetch(zap_name_t *zn)
{
uint64_t blk;
zap_t *zap = zn->zn_zap;
uint64_t idx = ZAP_HASH_IDX(zn->zn_hash,
zap_f_phys(zap)->zap_ptrtbl.zt_shift);
if (zap_idx_to_blk(zap, idx, &blk) != 0)
return;
int bs = FZAP_BLOCK_SHIFT(zap);
dmu_prefetch(zap->zap_objset, zap->zap_object, 0, blk << bs, 1 << bs,
ZIO_PRIORITY_SYNC_READ);
}
/*
* Helper functions for consumers.
*/
uint64_t
zap_create_link(objset_t *os, dmu_object_type_t ot, uint64_t parent_obj,
const char *name, dmu_tx_t *tx)
{
return (zap_create_link_dnsize(os, ot, parent_obj, name, 0, tx));
}
uint64_t
zap_create_link_dnsize(objset_t *os, dmu_object_type_t ot, uint64_t parent_obj,
const char *name, int dnodesize, dmu_tx_t *tx)
{
uint64_t new_obj;
new_obj = zap_create_dnsize(os, ot, DMU_OT_NONE, 0, dnodesize, tx);
VERIFY(new_obj != 0);
VERIFY0(zap_add(os, parent_obj, name, sizeof (uint64_t), 1, &new_obj,
tx));
return (new_obj);
}
int
zap_value_search(objset_t *os, uint64_t zapobj, uint64_t value, uint64_t mask,
char *name)
{
zap_cursor_t zc;
int err;
if (mask == 0)
mask = -1ULL;
zap_attribute_t *za = kmem_alloc(sizeof (*za), KM_SLEEP);
for (zap_cursor_init(&zc, os, zapobj);
(err = zap_cursor_retrieve(&zc, za)) == 0;
zap_cursor_advance(&zc)) {
if ((za->za_first_integer & mask) == (value & mask)) {
(void) strlcpy(name, za->za_name, MAXNAMELEN);
break;
}
}
zap_cursor_fini(&zc);
kmem_free(za, sizeof (*za));
return (err);
}
int
zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx)
{
zap_cursor_t zc;
int err = 0;
zap_attribute_t *za = kmem_alloc(sizeof (*za), KM_SLEEP);
for (zap_cursor_init(&zc, os, fromobj);
zap_cursor_retrieve(&zc, za) == 0;
(void) zap_cursor_advance(&zc)) {
if (za->za_integer_length != 8 || za->za_num_integers != 1) {
err = SET_ERROR(EINVAL);
break;
}
err = zap_add(os, intoobj, za->za_name,
8, 1, &za->za_first_integer, tx);
if (err != 0)
break;
}
zap_cursor_fini(&zc);
kmem_free(za, sizeof (*za));
return (err);
}
int
zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj,
uint64_t value, dmu_tx_t *tx)
{
zap_cursor_t zc;
int err = 0;
zap_attribute_t *za = kmem_alloc(sizeof (*za), KM_SLEEP);
for (zap_cursor_init(&zc, os, fromobj);
zap_cursor_retrieve(&zc, za) == 0;
(void) zap_cursor_advance(&zc)) {
if (za->za_integer_length != 8 || za->za_num_integers != 1) {
err = SET_ERROR(EINVAL);
break;
}
err = zap_add(os, intoobj, za->za_name,
8, 1, &value, tx);
if (err != 0)
break;
}
zap_cursor_fini(&zc);
kmem_free(za, sizeof (*za));
return (err);
}
int
zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj,
dmu_tx_t *tx)
{
zap_cursor_t zc;
int err = 0;
zap_attribute_t *za = kmem_alloc(sizeof (*za), KM_SLEEP);
for (zap_cursor_init(&zc, os, fromobj);
zap_cursor_retrieve(&zc, za) == 0;
(void) zap_cursor_advance(&zc)) {
uint64_t delta = 0;
if (za->za_integer_length != 8 || za->za_num_integers != 1) {
err = SET_ERROR(EINVAL);
break;
}
err = zap_lookup(os, intoobj, za->za_name, 8, 1, &delta);
if (err != 0 && err != ENOENT)
break;
delta += za->za_first_integer;
err = zap_update(os, intoobj, za->za_name, 8, 1, &delta, tx);
if (err != 0)
break;
}
zap_cursor_fini(&zc);
kmem_free(za, sizeof (*za));
return (err);
}
int
zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx)
{
char name[20];
(void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
return (zap_add(os, obj, name, 8, 1, &value, tx));
}
int
zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx)
{
char name[20];
(void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
return (zap_remove(os, obj, name, tx));
}
int
zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value)
{
char name[20];
(void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
return (zap_lookup(os, obj, name, 8, 1, &value));
}
int
zap_add_int_key(objset_t *os, uint64_t obj,
uint64_t key, uint64_t value, dmu_tx_t *tx)
{
char name[20];
(void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
return (zap_add(os, obj, name, 8, 1, &value, tx));
}
int
zap_update_int_key(objset_t *os, uint64_t obj,
uint64_t key, uint64_t value, dmu_tx_t *tx)
{
char name[20];
(void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
return (zap_update(os, obj, name, 8, 1, &value, tx));
}
int
zap_lookup_int_key(objset_t *os, uint64_t obj, uint64_t key, uint64_t *valuep)
{
char name[20];
(void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
return (zap_lookup(os, obj, name, 8, 1, valuep));
}
int
zap_increment(objset_t *os, uint64_t obj, const char *name, int64_t delta,
dmu_tx_t *tx)
{
uint64_t value = 0;
if (delta == 0)
return (0);
int err = zap_lookup(os, obj, name, 8, 1, &value);
if (err != 0 && err != ENOENT)
return (err);
value += delta;
if (value == 0)
err = zap_remove(os, obj, name, tx);
else
err = zap_update(os, obj, name, 8, 1, &value, tx);
return (err);
}
int
zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta,
dmu_tx_t *tx)
{
char name[20];
(void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
return (zap_increment(os, obj, name, delta, tx));
}
/*
* Routines for iterating over the attributes.
*/
int
fzap_cursor_retrieve(zap_t *zap, zap_cursor_t *zc, zap_attribute_t *za)
{
int err = ENOENT;
zap_entry_handle_t zeh;
zap_leaf_t *l;
/* retrieve the next entry at or after zc_hash/zc_cd */
/* if no entry, return ENOENT */
/*
* If we are reading from the beginning, we're almost certain to
* iterate over the entire ZAP object. If there are multiple leaf
* blocks (freeblk > 2), prefetch the whole object (up to
* dmu_prefetch_max bytes), so that we read the leaf blocks
* concurrently. (Unless noprefetch was requested via
* zap_cursor_init_noprefetch()).
*/
if (zc->zc_hash == 0 && zap_iterate_prefetch &&
zc->zc_prefetch && zap_f_phys(zap)->zap_freeblk > 2) {
dmu_prefetch(zc->zc_objset, zc->zc_zapobj, 0, 0,
zap_f_phys(zap)->zap_freeblk << FZAP_BLOCK_SHIFT(zap),
ZIO_PRIORITY_ASYNC_READ);
}
if (zc->zc_leaf &&
(ZAP_HASH_IDX(zc->zc_hash,
zap_leaf_phys(zc->zc_leaf)->l_hdr.lh_prefix_len) !=
zap_leaf_phys(zc->zc_leaf)->l_hdr.lh_prefix)) {
rw_enter(&zc->zc_leaf->l_rwlock, RW_READER);
zap_put_leaf(zc->zc_leaf);
zc->zc_leaf = NULL;
}
again:
if (zc->zc_leaf == NULL) {
err = zap_deref_leaf(zap, zc->zc_hash, NULL, RW_READER,
&zc->zc_leaf);
if (err != 0)
return (err);
} else {
rw_enter(&zc->zc_leaf->l_rwlock, RW_READER);
}
l = zc->zc_leaf;
err = zap_leaf_lookup_closest(l, zc->zc_hash, zc->zc_cd, &zeh);
if (err == ENOENT) {
if (zap_leaf_phys(l)->l_hdr.lh_prefix_len == 0) {
zc->zc_hash = -1ULL;
zc->zc_cd = 0;
} else {
uint64_t nocare = (1ULL <<
(64 - zap_leaf_phys(l)->l_hdr.lh_prefix_len)) - 1;
zc->zc_hash = (zc->zc_hash & ~nocare) + nocare + 1;
zc->zc_cd = 0;
if (zc->zc_hash == 0) {
zc->zc_hash = -1ULL;
} else {
zap_put_leaf(zc->zc_leaf);
zc->zc_leaf = NULL;
goto again;
}
}
}
if (err == 0) {
zc->zc_hash = zeh.zeh_hash;
zc->zc_cd = zeh.zeh_cd;
za->za_integer_length = zeh.zeh_integer_size;
za->za_num_integers = zeh.zeh_num_integers;
if (zeh.zeh_num_integers == 0) {
za->za_first_integer = 0;
} else {
err = zap_entry_read(&zeh, 8, 1, &za->za_first_integer);
ASSERT(err == 0 || err == EOVERFLOW);
}
err = zap_entry_read_name(zap, &zeh,
sizeof (za->za_name), za->za_name);
ASSERT(err == 0);
za->za_normalization_conflict =
zap_entry_normalization_conflict(&zeh,
NULL, za->za_name, zap);
}
rw_exit(&zc->zc_leaf->l_rwlock);
return (err);
}
static void
zap_stats_ptrtbl(zap_t *zap, uint64_t *tbl, int len, zap_stats_t *zs)
{
uint64_t lastblk = 0;
/*
* NB: if a leaf has more pointers than an entire ptrtbl block
* can hold, then it'll be accounted for more than once, since
* we won't have lastblk.
*/
for (int i = 0; i < len; i++) {
zap_leaf_t *l;
if (tbl[i] == lastblk)
continue;
lastblk = tbl[i];
int err = zap_get_leaf_byblk(zap, tbl[i], NULL, RW_READER, &l);
if (err == 0) {
zap_leaf_stats(zap, l, zs);
zap_put_leaf(l);
}
}
}
void
fzap_get_stats(zap_t *zap, zap_stats_t *zs)
{
int bs = FZAP_BLOCK_SHIFT(zap);
zs->zs_blocksize = 1ULL << bs;
/*
* Set zap_phys_t fields
*/
zs->zs_num_leafs = zap_f_phys(zap)->zap_num_leafs;
zs->zs_num_entries = zap_f_phys(zap)->zap_num_entries;
zs->zs_num_blocks = zap_f_phys(zap)->zap_freeblk;
zs->zs_block_type = zap_f_phys(zap)->zap_block_type;
zs->zs_magic = zap_f_phys(zap)->zap_magic;
zs->zs_salt = zap_f_phys(zap)->zap_salt;
/*
* Set zap_ptrtbl fields
*/
zs->zs_ptrtbl_len = 1ULL << zap_f_phys(zap)->zap_ptrtbl.zt_shift;
zs->zs_ptrtbl_nextblk = zap_f_phys(zap)->zap_ptrtbl.zt_nextblk;
zs->zs_ptrtbl_blks_copied =
zap_f_phys(zap)->zap_ptrtbl.zt_blks_copied;
zs->zs_ptrtbl_zt_blk = zap_f_phys(zap)->zap_ptrtbl.zt_blk;
zs->zs_ptrtbl_zt_numblks = zap_f_phys(zap)->zap_ptrtbl.zt_numblks;
zs->zs_ptrtbl_zt_shift = zap_f_phys(zap)->zap_ptrtbl.zt_shift;
if (zap_f_phys(zap)->zap_ptrtbl.zt_numblks == 0) {
/* the ptrtbl is entirely in the header block. */
zap_stats_ptrtbl(zap, &ZAP_EMBEDDED_PTRTBL_ENT(zap, 0),
1 << ZAP_EMBEDDED_PTRTBL_SHIFT(zap), zs);
} else {
dmu_prefetch(zap->zap_objset, zap->zap_object, 0,
zap_f_phys(zap)->zap_ptrtbl.zt_blk << bs,
zap_f_phys(zap)->zap_ptrtbl.zt_numblks << bs,
ZIO_PRIORITY_SYNC_READ);
for (int b = 0; b < zap_f_phys(zap)->zap_ptrtbl.zt_numblks;
b++) {
dmu_buf_t *db;
int err;
err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
(zap_f_phys(zap)->zap_ptrtbl.zt_blk + b) << bs,
FTAG, &db, DMU_READ_NO_PREFETCH);
if (err == 0) {
zap_stats_ptrtbl(zap, db->db_data,
1<<(bs-3), zs);
dmu_buf_rele(db, FTAG);
}
}
}
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, , zap_iterate_prefetch, INT, ZMOD_RW,
"When iterating ZAP object, prefetch it");
/* END CSTYLED */
diff --git a/module/zfs/zcp.c b/module/zfs/zcp.c
index f200b928bc6d..0a79068384a5 100644
--- a/module/zfs/zcp.c
+++ b/module/zfs/zcp.c
@@ -1,1452 +1,1452 @@
/*
* CDDL HEADER START
*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2016, 2018 by Delphix. All rights reserved.
*/
/*
* ZFS Channel Programs (ZCP)
*
* The ZCP interface allows various ZFS commands and operations ZFS
* administrative operations (e.g. creating and destroying snapshots, typically
* performed via an ioctl to /dev/zfs by the zfs(8) command and
* libzfs/libzfs_core) to be run * programmatically as a Lua script. A ZCP
* script is run as a dsl_sync_task and fully executed during one transaction
* group sync. This ensures that no other changes can be written concurrently
* with a running Lua script. Combining multiple calls to the exposed ZFS
* functions into one script gives a number of benefits:
*
* 1. Atomicity. For some compound or iterative operations, it's useful to be
* able to guarantee that the state of a pool has not changed between calls to
* ZFS.
*
* 2. Performance. If a large number of changes need to be made (e.g. deleting
* many filesystems), there can be a significant performance penalty as a
* result of the need to wait for a transaction group sync to pass for every
* single operation. When expressed as a single ZCP script, all these changes
* can be performed at once in one txg sync.
*
* A modified version of the Lua 5.2 interpreter is used to run channel program
* scripts. The Lua 5.2 manual can be found at:
*
* http://www.lua.org/manual/5.2/
*
* If being run by a user (via an ioctl syscall), executing a ZCP script
* requires root privileges in the global zone.
*
* Scripts are passed to zcp_eval() as a string, then run in a synctask by
* zcp_eval_sync(). Arguments can be passed into the Lua script as an nvlist,
* which will be converted to a Lua table. Similarly, values returned from
* a ZCP script will be converted to an nvlist. See zcp_lua_to_nvlist_impl()
* for details on exact allowed types and conversion.
*
* ZFS functionality is exposed to a ZCP script as a library of function calls.
* These calls are sorted into submodules, such as zfs.list and zfs.sync, for
* iterators and synctasks, respectively. Each of these submodules resides in
* its own source file, with a zcp_*_info structure describing each library
* call in the submodule.
*
* Error handling in ZCP scripts is handled by a number of different methods
* based on severity:
*
* 1. Memory and time limits are in place to prevent a channel program from
* consuming excessive system or running forever. If one of these limits is
* hit, the channel program will be stopped immediately and return from
* zcp_eval() with an error code. No attempt will be made to roll back or undo
* any changes made by the channel program before the error occurred.
* Consumers invoking zcp_eval() from elsewhere in the kernel may pass a time
* limit of 0, disabling the time limit.
*
* 2. Internal Lua errors can occur as a result of a syntax error, calling a
* library function with incorrect arguments, invoking the error() function,
* failing an assert(), or other runtime errors. In these cases the channel
* program will stop executing and return from zcp_eval() with an error code.
* In place of a return value, an error message will also be returned in the
* 'result' nvlist containing information about the error. No attempt will be
* made to roll back or undo any changes made by the channel program before the
* error occurred.
*
* 3. If an error occurs inside a ZFS library call which returns an error code,
* the error is returned to the Lua script to be handled as desired.
*
* In the first two cases, Lua's error-throwing mechanism is used, which
* longjumps out of the script execution with luaL_error() and returns with the
* error.
*
* See zfs-program(8) for more information on high level usage.
*/
#include <sys/lua/lua.h>
#include <sys/lua/lualib.h>
#include <sys/lua/lauxlib.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_synctask.h>
#include <sys/dsl_dataset.h>
#include <sys/zcp.h>
#include <sys/zcp_iter.h>
#include <sys/zcp_prop.h>
#include <sys/zcp_global.h>
#include <sys/zvol.h>
#ifndef KM_NORMALPRI
#define KM_NORMALPRI 0
#endif
#define ZCP_NVLIST_MAX_DEPTH 20
-uint64_t zfs_lua_check_instrlimit_interval = 100;
+static const uint64_t zfs_lua_check_instrlimit_interval = 100;
unsigned long zfs_lua_max_instrlimit = ZCP_MAX_INSTRLIMIT;
unsigned long zfs_lua_max_memlimit = ZCP_MAX_MEMLIMIT;
/*
* Forward declarations for mutually recursive functions
*/
static int zcp_nvpair_value_to_lua(lua_State *, nvpair_t *, char *, int);
static int zcp_lua_to_nvlist_impl(lua_State *, int, nvlist_t *, const char *,
int);
/*
* The outer-most error callback handler for use with lua_pcall(). On
* error Lua will call this callback with a single argument that
* represents the error value. In most cases this will be a string
* containing an error message, but channel programs can use Lua's
* error() function to return arbitrary objects as errors. This callback
* returns (on the Lua stack) the original error object along with a traceback.
*
* Fatal Lua errors can occur while resources are held, so we also call any
* registered cleanup function here.
*/
static int
zcp_error_handler(lua_State *state)
{
const char *msg;
zcp_cleanup(state);
VERIFY3U(1, ==, lua_gettop(state));
msg = lua_tostring(state, 1);
luaL_traceback(state, state, msg, 1);
return (1);
}
int
zcp_argerror(lua_State *state, int narg, const char *msg, ...)
{
va_list alist;
va_start(alist, msg);
const char *buf = lua_pushvfstring(state, msg, alist);
va_end(alist);
return (luaL_argerror(state, narg, buf));
}
/*
* Install a new cleanup function, which will be invoked with the given
* opaque argument if a fatal error causes the Lua interpreter to longjump out
* of a function call.
*
* If an error occurs, the cleanup function will be invoked exactly once and
* then unregistered.
*
* Returns the registered cleanup handler so the caller can deregister it
* if no error occurs.
*/
zcp_cleanup_handler_t *
zcp_register_cleanup(lua_State *state, zcp_cleanup_t cleanfunc, void *cleanarg)
{
zcp_run_info_t *ri = zcp_run_info(state);
zcp_cleanup_handler_t *zch = kmem_alloc(sizeof (*zch), KM_SLEEP);
zch->zch_cleanup_func = cleanfunc;
zch->zch_cleanup_arg = cleanarg;
list_insert_head(&ri->zri_cleanup_handlers, zch);
return (zch);
}
void
zcp_deregister_cleanup(lua_State *state, zcp_cleanup_handler_t *zch)
{
zcp_run_info_t *ri = zcp_run_info(state);
list_remove(&ri->zri_cleanup_handlers, zch);
kmem_free(zch, sizeof (*zch));
}
/*
* Execute the currently registered cleanup handlers then free them and
* destroy the handler list.
*/
void
zcp_cleanup(lua_State *state)
{
zcp_run_info_t *ri = zcp_run_info(state);
for (zcp_cleanup_handler_t *zch =
list_remove_head(&ri->zri_cleanup_handlers); zch != NULL;
zch = list_remove_head(&ri->zri_cleanup_handlers)) {
zch->zch_cleanup_func(zch->zch_cleanup_arg);
kmem_free(zch, sizeof (*zch));
}
}
/*
* Convert the lua table at the given index on the Lua stack to an nvlist
* and return it.
*
* If the table can not be converted for any reason, NULL is returned and
* an error message is pushed onto the Lua stack.
*/
static nvlist_t *
zcp_table_to_nvlist(lua_State *state, int index, int depth)
{
nvlist_t *nvl;
/*
* Converting a Lua table to an nvlist with key uniqueness checking is
* O(n^2) in the number of keys in the nvlist, which can take a long
* time when we return a large table from a channel program.
* Furthermore, Lua's table interface *almost* guarantees unique keys
* on its own (details below). Therefore, we don't use fnvlist_alloc()
* here to avoid the built-in uniqueness checking.
*
* The *almost* is because it's possible to have key collisions between
* e.g. the string "1" and the number 1, or the string "true" and the
* boolean true, so we explicitly check that when we're looking at a
* key which is an integer / boolean or a string that can be parsed as
* one of those types. In the worst case this could still devolve into
* O(n^2), so we only start doing these checks on boolean/integer keys
* once we've seen a string key which fits this weird usage pattern.
*
* Ultimately, we still want callers to know that the keys in this
* nvlist are unique, so before we return this we set the nvlist's
* flags to reflect that.
*/
VERIFY0(nvlist_alloc(&nvl, 0, KM_SLEEP));
/*
* Push an empty stack slot where lua_next() will store each
* table key.
*/
lua_pushnil(state);
boolean_t saw_str_could_collide = B_FALSE;
while (lua_next(state, index) != 0) {
/*
* The next key-value pair from the table at index is
* now on the stack, with the key at stack slot -2 and
* the value at slot -1.
*/
int err = 0;
char buf[32];
const char *key = NULL;
boolean_t key_could_collide = B_FALSE;
switch (lua_type(state, -2)) {
case LUA_TSTRING:
key = lua_tostring(state, -2);
/* check if this could collide with a number or bool */
long long tmp;
int parselen;
if ((sscanf(key, "%lld%n", &tmp, &parselen) > 0 &&
parselen == strlen(key)) ||
strcmp(key, "true") == 0 ||
strcmp(key, "false") == 0) {
key_could_collide = B_TRUE;
saw_str_could_collide = B_TRUE;
}
break;
case LUA_TBOOLEAN:
key = (lua_toboolean(state, -2) == B_TRUE ?
"true" : "false");
if (saw_str_could_collide) {
key_could_collide = B_TRUE;
}
break;
case LUA_TNUMBER:
VERIFY3U(sizeof (buf), >,
snprintf(buf, sizeof (buf), "%lld",
(longlong_t)lua_tonumber(state, -2)));
key = buf;
if (saw_str_could_collide) {
key_could_collide = B_TRUE;
}
break;
default:
fnvlist_free(nvl);
(void) lua_pushfstring(state, "Invalid key "
"type '%s' in table",
lua_typename(state, lua_type(state, -2)));
return (NULL);
}
/*
* Check for type-mismatched key collisions, and throw an error.
*/
if (key_could_collide && nvlist_exists(nvl, key)) {
fnvlist_free(nvl);
(void) lua_pushfstring(state, "Collision of "
"key '%s' in table", key);
return (NULL);
}
/*
* Recursively convert the table value and insert into
* the new nvlist with the parsed key. To prevent
* stack overflow on circular or heavily nested tables,
* we track the current nvlist depth.
*/
if (depth >= ZCP_NVLIST_MAX_DEPTH) {
fnvlist_free(nvl);
(void) lua_pushfstring(state, "Maximum table "
"depth (%d) exceeded for table",
ZCP_NVLIST_MAX_DEPTH);
return (NULL);
}
err = zcp_lua_to_nvlist_impl(state, -1, nvl, key,
depth + 1);
if (err != 0) {
fnvlist_free(nvl);
/*
* Error message has been pushed to the lua
* stack by the recursive call.
*/
return (NULL);
}
/*
* Pop the value pushed by lua_next().
*/
lua_pop(state, 1);
}
/*
* Mark the nvlist as having unique keys. This is a little ugly, but we
* ensured above that there are no duplicate keys in the nvlist.
*/
nvl->nvl_nvflag |= NV_UNIQUE_NAME;
return (nvl);
}
/*
* Convert a value from the given index into the lua stack to an nvpair, adding
* it to an nvlist with the given key.
*
* Values are converted as follows:
*
* string -> string
* number -> int64
* boolean -> boolean
* nil -> boolean (no value)
*
* Lua tables are converted to nvlists and then inserted. The table's keys
* are converted to strings then used as keys in the nvlist to store each table
* element. Keys are converted as follows:
*
* string -> no change
* number -> "%lld"
* boolean -> "true" | "false"
* nil -> error
*
* In the case of a key collision, an error is thrown.
*
* If an error is encountered, a nonzero error code is returned, and an error
* string will be pushed onto the Lua stack.
*/
static int
zcp_lua_to_nvlist_impl(lua_State *state, int index, nvlist_t *nvl,
const char *key, int depth)
{
/*
* Verify that we have enough remaining space in the lua stack to parse
* a key-value pair and push an error.
*/
if (!lua_checkstack(state, 3)) {
(void) lua_pushstring(state, "Lua stack overflow");
return (1);
}
index = lua_absindex(state, index);
switch (lua_type(state, index)) {
case LUA_TNIL:
fnvlist_add_boolean(nvl, key);
break;
case LUA_TBOOLEAN:
fnvlist_add_boolean_value(nvl, key,
lua_toboolean(state, index));
break;
case LUA_TNUMBER:
fnvlist_add_int64(nvl, key, lua_tonumber(state, index));
break;
case LUA_TSTRING:
fnvlist_add_string(nvl, key, lua_tostring(state, index));
break;
case LUA_TTABLE: {
nvlist_t *value_nvl = zcp_table_to_nvlist(state, index, depth);
if (value_nvl == NULL)
return (SET_ERROR(EINVAL));
fnvlist_add_nvlist(nvl, key, value_nvl);
fnvlist_free(value_nvl);
break;
}
default:
(void) lua_pushfstring(state,
"Invalid value type '%s' for key '%s'",
lua_typename(state, lua_type(state, index)), key);
return (SET_ERROR(EINVAL));
}
return (0);
}
/*
* Convert a lua value to an nvpair, adding it to an nvlist with the given key.
*/
static void
zcp_lua_to_nvlist(lua_State *state, int index, nvlist_t *nvl, const char *key)
{
/*
* On error, zcp_lua_to_nvlist_impl pushes an error string onto the Lua
* stack before returning with a nonzero error code. If an error is
* returned, throw a fatal lua error with the given string.
*/
if (zcp_lua_to_nvlist_impl(state, index, nvl, key, 0) != 0)
(void) lua_error(state);
}
static int
zcp_lua_to_nvlist_helper(lua_State *state)
{
nvlist_t *nv = (nvlist_t *)lua_touserdata(state, 2);
const char *key = (const char *)lua_touserdata(state, 1);
zcp_lua_to_nvlist(state, 3, nv, key);
return (0);
}
static void
zcp_convert_return_values(lua_State *state, nvlist_t *nvl,
const char *key, int *result)
{
int err;
VERIFY3U(1, ==, lua_gettop(state));
lua_pushcfunction(state, zcp_lua_to_nvlist_helper);
lua_pushlightuserdata(state, (char *)key);
lua_pushlightuserdata(state, nvl);
lua_pushvalue(state, 1);
lua_remove(state, 1);
err = lua_pcall(state, 3, 0, 0); /* zcp_lua_to_nvlist_helper */
if (err != 0) {
zcp_lua_to_nvlist(state, 1, nvl, ZCP_RET_ERROR);
*result = SET_ERROR(ECHRNG);
}
}
/*
* Push a Lua table representing nvl onto the stack. If it can't be
* converted, return EINVAL, fill in errbuf, and push nothing. errbuf may
* be specified as NULL, in which case no error string will be output.
*
* Most nvlists are converted as simple key->value Lua tables, but we make
* an exception for the case where all nvlist entries are BOOLEANs (a string
* key without a value). In Lua, a table key pointing to a value of Nil
* (no value) is equivalent to the key not existing, so a BOOLEAN nvlist
* entry can't be directly converted to a Lua table entry. Nvlists of entirely
* BOOLEAN entries are frequently used to pass around lists of datasets, so for
* convenience we check for this case, and convert it to a simple Lua array of
* strings.
*/
int
zcp_nvlist_to_lua(lua_State *state, nvlist_t *nvl,
char *errbuf, int errbuf_len)
{
nvpair_t *pair;
lua_newtable(state);
boolean_t has_values = B_FALSE;
/*
* If the list doesn't have any values, just convert it to a string
* array.
*/
for (pair = nvlist_next_nvpair(nvl, NULL);
pair != NULL; pair = nvlist_next_nvpair(nvl, pair)) {
if (nvpair_type(pair) != DATA_TYPE_BOOLEAN) {
has_values = B_TRUE;
break;
}
}
if (!has_values) {
int i = 1;
for (pair = nvlist_next_nvpair(nvl, NULL);
pair != NULL; pair = nvlist_next_nvpair(nvl, pair)) {
(void) lua_pushinteger(state, i);
(void) lua_pushstring(state, nvpair_name(pair));
(void) lua_settable(state, -3);
i++;
}
} else {
for (pair = nvlist_next_nvpair(nvl, NULL);
pair != NULL; pair = nvlist_next_nvpair(nvl, pair)) {
int err = zcp_nvpair_value_to_lua(state, pair,
errbuf, errbuf_len);
if (err != 0) {
lua_pop(state, 1);
return (err);
}
(void) lua_setfield(state, -2, nvpair_name(pair));
}
}
return (0);
}
/*
* Push a Lua object representing the value of "pair" onto the stack.
*
* Only understands boolean_value, string, int64, nvlist,
* string_array, and int64_array type values. For other
* types, returns EINVAL, fills in errbuf, and pushes nothing.
*/
static int
zcp_nvpair_value_to_lua(lua_State *state, nvpair_t *pair,
char *errbuf, int errbuf_len)
{
int err = 0;
if (pair == NULL) {
lua_pushnil(state);
return (0);
}
switch (nvpair_type(pair)) {
case DATA_TYPE_BOOLEAN_VALUE:
(void) lua_pushboolean(state,
fnvpair_value_boolean_value(pair));
break;
case DATA_TYPE_STRING:
(void) lua_pushstring(state, fnvpair_value_string(pair));
break;
case DATA_TYPE_INT64:
(void) lua_pushinteger(state, fnvpair_value_int64(pair));
break;
case DATA_TYPE_NVLIST:
err = zcp_nvlist_to_lua(state,
fnvpair_value_nvlist(pair), errbuf, errbuf_len);
break;
case DATA_TYPE_STRING_ARRAY: {
char **strarr;
uint_t nelem;
(void) nvpair_value_string_array(pair, &strarr, &nelem);
lua_newtable(state);
for (int i = 0; i < nelem; i++) {
(void) lua_pushinteger(state, i + 1);
(void) lua_pushstring(state, strarr[i]);
(void) lua_settable(state, -3);
}
break;
}
case DATA_TYPE_UINT64_ARRAY: {
uint64_t *intarr;
uint_t nelem;
(void) nvpair_value_uint64_array(pair, &intarr, &nelem);
lua_newtable(state);
for (int i = 0; i < nelem; i++) {
(void) lua_pushinteger(state, i + 1);
(void) lua_pushinteger(state, intarr[i]);
(void) lua_settable(state, -3);
}
break;
}
case DATA_TYPE_INT64_ARRAY: {
int64_t *intarr;
uint_t nelem;
(void) nvpair_value_int64_array(pair, &intarr, &nelem);
lua_newtable(state);
for (int i = 0; i < nelem; i++) {
(void) lua_pushinteger(state, i + 1);
(void) lua_pushinteger(state, intarr[i]);
(void) lua_settable(state, -3);
}
break;
}
default: {
if (errbuf != NULL) {
(void) snprintf(errbuf, errbuf_len,
"Unhandled nvpair type %d for key '%s'",
nvpair_type(pair), nvpair_name(pair));
}
return (SET_ERROR(EINVAL));
}
}
return (err);
}
int
zcp_dataset_hold_error(lua_State *state, dsl_pool_t *dp, const char *dsname,
int error)
{
if (error == ENOENT) {
(void) zcp_argerror(state, 1, "no such dataset '%s'", dsname);
return (0); /* not reached; zcp_argerror will longjmp */
} else if (error == EXDEV) {
(void) zcp_argerror(state, 1,
"dataset '%s' is not in the target pool '%s'",
dsname, spa_name(dp->dp_spa));
return (0); /* not reached; zcp_argerror will longjmp */
} else if (error == EIO) {
(void) luaL_error(state,
"I/O error while accessing dataset '%s'", dsname);
return (0); /* not reached; luaL_error will longjmp */
} else if (error != 0) {
(void) luaL_error(state,
"unexpected error %d while accessing dataset '%s'",
error, dsname);
return (0); /* not reached; luaL_error will longjmp */
}
return (0);
}
/*
* Note: will longjmp (via lua_error()) on error.
* Assumes that the dsname is argument #1 (for error reporting purposes).
*/
dsl_dataset_t *
zcp_dataset_hold(lua_State *state, dsl_pool_t *dp, const char *dsname,
void *tag)
{
dsl_dataset_t *ds;
int error = dsl_dataset_hold(dp, dsname, tag, &ds);
(void) zcp_dataset_hold_error(state, dp, dsname, error);
return (ds);
}
static int zcp_debug(lua_State *);
-static zcp_lib_info_t zcp_debug_info = {
+static const zcp_lib_info_t zcp_debug_info = {
.name = "debug",
.func = zcp_debug,
.pargs = {
- { .za_name = "debug string", .za_lua_type = LUA_TSTRING},
+ { .za_name = "debug string", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
static int
zcp_debug(lua_State *state)
{
const char *dbgstring;
zcp_run_info_t *ri = zcp_run_info(state);
- zcp_lib_info_t *libinfo = &zcp_debug_info;
+ const zcp_lib_info_t *libinfo = &zcp_debug_info;
zcp_parse_args(state, libinfo->name, libinfo->pargs, libinfo->kwargs);
dbgstring = lua_tostring(state, 1);
zfs_dbgmsg("txg %lld ZCP: %s", (longlong_t)ri->zri_tx->tx_txg,
dbgstring);
return (0);
}
static int zcp_exists(lua_State *);
-static zcp_lib_info_t zcp_exists_info = {
+static const zcp_lib_info_t zcp_exists_info = {
.name = "exists",
.func = zcp_exists,
.pargs = {
- { .za_name = "dataset", .za_lua_type = LUA_TSTRING},
+ { .za_name = "dataset", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
static int
zcp_exists(lua_State *state)
{
zcp_run_info_t *ri = zcp_run_info(state);
dsl_pool_t *dp = ri->zri_pool;
- zcp_lib_info_t *libinfo = &zcp_exists_info;
+ const zcp_lib_info_t *libinfo = &zcp_exists_info;
zcp_parse_args(state, libinfo->name, libinfo->pargs, libinfo->kwargs);
const char *dsname = lua_tostring(state, 1);
dsl_dataset_t *ds;
int error = dsl_dataset_hold(dp, dsname, FTAG, &ds);
if (error == 0) {
dsl_dataset_rele(ds, FTAG);
lua_pushboolean(state, B_TRUE);
} else if (error == ENOENT) {
lua_pushboolean(state, B_FALSE);
} else if (error == EXDEV) {
return (luaL_error(state, "dataset '%s' is not in the "
"target pool", dsname));
} else if (error == EIO) {
return (luaL_error(state, "I/O error opening dataset '%s'",
dsname));
} else if (error != 0) {
return (luaL_error(state, "unexpected error %d", error));
}
return (1);
}
/*
* Allocate/realloc/free a buffer for the lua interpreter.
*
* When nsize is 0, behaves as free() and returns NULL.
*
* If ptr is NULL, behaves as malloc() and returns an allocated buffer of size
* at least nsize.
*
* Otherwise, behaves as realloc(), changing the allocation from osize to nsize.
* Shrinking the buffer size never fails.
*
* The original allocated buffer size is stored as a uint64 at the beginning of
* the buffer to avoid actually reallocating when shrinking a buffer, since lua
* requires that this operation never fail.
*/
static void *
zcp_lua_alloc(void *ud, void *ptr, size_t osize, size_t nsize)
{
zcp_alloc_arg_t *allocargs = ud;
if (nsize == 0) {
if (ptr != NULL) {
int64_t *allocbuf = (int64_t *)ptr - 1;
int64_t allocsize = *allocbuf;
ASSERT3S(allocsize, >, 0);
ASSERT3S(allocargs->aa_alloc_remaining + allocsize, <=,
allocargs->aa_alloc_limit);
allocargs->aa_alloc_remaining += allocsize;
vmem_free(allocbuf, allocsize);
}
return (NULL);
} else if (ptr == NULL) {
int64_t *allocbuf;
int64_t allocsize = nsize + sizeof (int64_t);
if (!allocargs->aa_must_succeed &&
(allocsize <= 0 ||
allocsize > allocargs->aa_alloc_remaining)) {
return (NULL);
}
allocbuf = vmem_alloc(allocsize, KM_SLEEP);
allocargs->aa_alloc_remaining -= allocsize;
*allocbuf = allocsize;
return (allocbuf + 1);
} else if (nsize <= osize) {
/*
* If shrinking the buffer, lua requires that the reallocation
* never fail.
*/
return (ptr);
} else {
ASSERT3U(nsize, >, osize);
uint64_t *luabuf = zcp_lua_alloc(ud, NULL, 0, nsize);
if (luabuf == NULL) {
return (NULL);
}
(void) memcpy(luabuf, ptr, osize);
VERIFY3P(zcp_lua_alloc(ud, ptr, osize, 0), ==, NULL);
return (luabuf);
}
}
static void
zcp_lua_counthook(lua_State *state, lua_Debug *ar)
{
(void) ar;
lua_getfield(state, LUA_REGISTRYINDEX, ZCP_RUN_INFO_KEY);
zcp_run_info_t *ri = lua_touserdata(state, -1);
/*
* Check if we were canceled while waiting for the
* txg to sync or from our open context thread
*/
if (ri->zri_canceled ||
(!ri->zri_sync && issig(JUSTLOOKING) && issig(FORREAL))) {
ri->zri_canceled = B_TRUE;
(void) lua_pushstring(state, "Channel program was canceled.");
(void) lua_error(state);
/* Unreachable */
}
/*
* Check how many instructions the channel program has
* executed so far, and compare against the limit.
*/
ri->zri_curinstrs += zfs_lua_check_instrlimit_interval;
if (ri->zri_maxinstrs != 0 && ri->zri_curinstrs > ri->zri_maxinstrs) {
ri->zri_timed_out = B_TRUE;
(void) lua_pushstring(state,
"Channel program timed out.");
(void) lua_error(state);
/* Unreachable */
}
}
static int
zcp_panic_cb(lua_State *state)
{
panic("unprotected error in call to Lua API (%s)\n",
lua_tostring(state, -1));
return (0);
}
static void
zcp_eval_impl(dmu_tx_t *tx, zcp_run_info_t *ri)
{
int err;
lua_State *state = ri->zri_state;
VERIFY3U(3, ==, lua_gettop(state));
/* finish initializing our runtime state */
ri->zri_pool = dmu_tx_pool(tx);
ri->zri_tx = tx;
list_create(&ri->zri_cleanup_handlers, sizeof (zcp_cleanup_handler_t),
offsetof(zcp_cleanup_handler_t, zch_node));
/*
* Store the zcp_run_info_t struct for this run in the Lua registry.
* Registry entries are not directly accessible by the Lua scripts but
* can be accessed by our callbacks.
*/
lua_pushlightuserdata(state, ri);
lua_setfield(state, LUA_REGISTRYINDEX, ZCP_RUN_INFO_KEY);
VERIFY3U(3, ==, lua_gettop(state));
/*
* Tell the Lua interpreter to call our handler every count
* instructions. Channel programs that execute too many instructions
* should die with ETIME.
*/
(void) lua_sethook(state, zcp_lua_counthook, LUA_MASKCOUNT,
zfs_lua_check_instrlimit_interval);
/*
* Tell the Lua memory allocator to stop using KM_SLEEP before handing
* off control to the channel program. Channel programs that use too
* much memory should die with ENOSPC.
*/
ri->zri_allocargs->aa_must_succeed = B_FALSE;
/*
* Call the Lua function that open-context passed us. This pops the
* function and its input from the stack and pushes any return
* or error values.
*/
err = lua_pcall(state, 1, LUA_MULTRET, 1);
/*
* Let Lua use KM_SLEEP while we interpret the return values.
*/
ri->zri_allocargs->aa_must_succeed = B_TRUE;
/*
* Remove the error handler callback from the stack. At this point,
* there shouldn't be any cleanup handler registered in the handler
* list (zri_cleanup_handlers), regardless of whether it ran or not.
*/
list_destroy(&ri->zri_cleanup_handlers);
lua_remove(state, 1);
switch (err) {
case LUA_OK: {
/*
* Lua supports returning multiple values in a single return
* statement. Return values will have been pushed onto the
* stack:
* 1: Return value 1
* 2: Return value 2
* 3: etc...
* To simplify the process of retrieving a return value from a
* channel program, we disallow returning more than one value
* to ZFS from the Lua script, yielding a singleton return
* nvlist of the form { "return": Return value 1 }.
*/
int return_count = lua_gettop(state);
if (return_count == 1) {
ri->zri_result = 0;
zcp_convert_return_values(state, ri->zri_outnvl,
ZCP_RET_RETURN, &ri->zri_result);
} else if (return_count > 1) {
ri->zri_result = SET_ERROR(ECHRNG);
lua_settop(state, 0);
(void) lua_pushfstring(state, "Multiple return "
"values not supported");
zcp_convert_return_values(state, ri->zri_outnvl,
ZCP_RET_ERROR, &ri->zri_result);
}
break;
}
case LUA_ERRRUN:
case LUA_ERRGCMM: {
/*
* The channel program encountered a fatal error within the
* script, such as failing an assertion, or calling a function
* with incompatible arguments. The error value and the
* traceback generated by zcp_error_handler() should be on the
* stack.
*/
VERIFY3U(1, ==, lua_gettop(state));
if (ri->zri_timed_out) {
ri->zri_result = SET_ERROR(ETIME);
} else if (ri->zri_canceled) {
ri->zri_result = SET_ERROR(EINTR);
} else {
ri->zri_result = SET_ERROR(ECHRNG);
}
zcp_convert_return_values(state, ri->zri_outnvl,
ZCP_RET_ERROR, &ri->zri_result);
if (ri->zri_result == ETIME && ri->zri_outnvl != NULL) {
(void) nvlist_add_uint64(ri->zri_outnvl,
ZCP_ARG_INSTRLIMIT, ri->zri_curinstrs);
}
break;
}
case LUA_ERRERR: {
/*
* The channel program encountered a fatal error within the
* script, and we encountered another error while trying to
* compute the traceback in zcp_error_handler(). We can only
* return the error message.
*/
VERIFY3U(1, ==, lua_gettop(state));
if (ri->zri_timed_out) {
ri->zri_result = SET_ERROR(ETIME);
} else if (ri->zri_canceled) {
ri->zri_result = SET_ERROR(EINTR);
} else {
ri->zri_result = SET_ERROR(ECHRNG);
}
zcp_convert_return_values(state, ri->zri_outnvl,
ZCP_RET_ERROR, &ri->zri_result);
break;
}
case LUA_ERRMEM:
/*
* Lua ran out of memory while running the channel program.
* There's not much we can do.
*/
ri->zri_result = SET_ERROR(ENOSPC);
break;
default:
VERIFY0(err);
}
}
static void
zcp_pool_error(zcp_run_info_t *ri, const char *poolname)
{
ri->zri_result = SET_ERROR(ECHRNG);
lua_settop(ri->zri_state, 0);
(void) lua_pushfstring(ri->zri_state, "Could not open pool: %s",
poolname);
zcp_convert_return_values(ri->zri_state, ri->zri_outnvl,
ZCP_RET_ERROR, &ri->zri_result);
}
/*
* This callback is called when txg_wait_synced_sig encountered a signal.
* The txg_wait_synced_sig will continue to wait for the txg to complete
* after calling this callback.
*/
static void
zcp_eval_sig(void *arg, dmu_tx_t *tx)
{
(void) tx;
zcp_run_info_t *ri = arg;
ri->zri_canceled = B_TRUE;
}
static void
zcp_eval_sync(void *arg, dmu_tx_t *tx)
{
zcp_run_info_t *ri = arg;
/*
* Open context should have setup the stack to contain:
* 1: Error handler callback
* 2: Script to run (converted to a Lua function)
* 3: nvlist input to function (converted to Lua table or nil)
*/
VERIFY3U(3, ==, lua_gettop(ri->zri_state));
zcp_eval_impl(tx, ri);
}
static void
zcp_eval_open(zcp_run_info_t *ri, const char *poolname)
{
int error;
dsl_pool_t *dp;
dmu_tx_t *tx;
/*
* See comment from the same assertion in zcp_eval_sync().
*/
VERIFY3U(3, ==, lua_gettop(ri->zri_state));
error = dsl_pool_hold(poolname, FTAG, &dp);
if (error != 0) {
zcp_pool_error(ri, poolname);
return;
}
/*
* As we are running in open-context, we have no transaction associated
* with the channel program. At the same time, functions from the
* zfs.check submodule need to be associated with a transaction as
* they are basically dry-runs of their counterparts in the zfs.sync
* submodule. These functions should be able to run in open-context.
* Therefore we create a new transaction that we later abort once
* the channel program has been evaluated.
*/
tx = dmu_tx_create_dd(dp->dp_mos_dir);
zcp_eval_impl(tx, ri);
dmu_tx_abort(tx);
dsl_pool_rele(dp, FTAG);
}
int
zcp_eval(const char *poolname, const char *program, boolean_t sync,
uint64_t instrlimit, uint64_t memlimit, nvpair_t *nvarg, nvlist_t *outnvl)
{
int err;
lua_State *state;
zcp_run_info_t runinfo;
if (instrlimit > zfs_lua_max_instrlimit)
return (SET_ERROR(EINVAL));
if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
return (SET_ERROR(EINVAL));
zcp_alloc_arg_t allocargs = {
.aa_must_succeed = B_TRUE,
.aa_alloc_remaining = (int64_t)memlimit,
.aa_alloc_limit = (int64_t)memlimit,
};
/*
* Creates a Lua state with a memory allocator that uses KM_SLEEP.
* This should never fail.
*/
state = lua_newstate(zcp_lua_alloc, &allocargs);
VERIFY(state != NULL);
(void) lua_atpanic(state, zcp_panic_cb);
/*
* Load core Lua libraries we want access to.
*/
VERIFY3U(1, ==, luaopen_base(state));
lua_pop(state, 1);
VERIFY3U(1, ==, luaopen_coroutine(state));
lua_setglobal(state, LUA_COLIBNAME);
VERIFY0(lua_gettop(state));
VERIFY3U(1, ==, luaopen_string(state));
lua_setglobal(state, LUA_STRLIBNAME);
VERIFY0(lua_gettop(state));
VERIFY3U(1, ==, luaopen_table(state));
lua_setglobal(state, LUA_TABLIBNAME);
VERIFY0(lua_gettop(state));
/*
* Load globally visible variables such as errno aliases.
*/
zcp_load_globals(state);
VERIFY0(lua_gettop(state));
/*
* Load ZFS-specific modules.
*/
lua_newtable(state);
VERIFY3U(1, ==, zcp_load_list_lib(state));
lua_setfield(state, -2, "list");
VERIFY3U(1, ==, zcp_load_synctask_lib(state, B_FALSE));
lua_setfield(state, -2, "check");
VERIFY3U(1, ==, zcp_load_synctask_lib(state, B_TRUE));
lua_setfield(state, -2, "sync");
VERIFY3U(1, ==, zcp_load_get_lib(state));
lua_pushcclosure(state, zcp_debug_info.func, 0);
lua_setfield(state, -2, zcp_debug_info.name);
lua_pushcclosure(state, zcp_exists_info.func, 0);
lua_setfield(state, -2, zcp_exists_info.name);
lua_setglobal(state, "zfs");
VERIFY0(lua_gettop(state));
/*
* Push the error-callback that calculates Lua stack traces on
* unexpected failures.
*/
lua_pushcfunction(state, zcp_error_handler);
VERIFY3U(1, ==, lua_gettop(state));
/*
* Load the actual script as a function onto the stack as text ("t").
* The only valid error condition is a syntax error in the script.
* ERRMEM should not be possible because our allocator is using
* KM_SLEEP. ERRGCMM should not be possible because we have not added
* any objects with __gc metamethods to the interpreter that could
* fail.
*/
err = luaL_loadbufferx(state, program, strlen(program),
"channel program", "t");
if (err == LUA_ERRSYNTAX) {
fnvlist_add_string(outnvl, ZCP_RET_ERROR,
lua_tostring(state, -1));
lua_close(state);
return (SET_ERROR(EINVAL));
}
VERIFY0(err);
VERIFY3U(2, ==, lua_gettop(state));
/*
* Convert the input nvlist to a Lua object and put it on top of the
* stack.
*/
char errmsg[128];
err = zcp_nvpair_value_to_lua(state, nvarg,
errmsg, sizeof (errmsg));
if (err != 0) {
fnvlist_add_string(outnvl, ZCP_RET_ERROR, errmsg);
lua_close(state);
return (SET_ERROR(EINVAL));
}
VERIFY3U(3, ==, lua_gettop(state));
runinfo.zri_state = state;
runinfo.zri_allocargs = &allocargs;
runinfo.zri_outnvl = outnvl;
runinfo.zri_result = 0;
runinfo.zri_cred = CRED();
runinfo.zri_proc = curproc;
runinfo.zri_timed_out = B_FALSE;
runinfo.zri_canceled = B_FALSE;
runinfo.zri_sync = sync;
runinfo.zri_space_used = 0;
runinfo.zri_curinstrs = 0;
runinfo.zri_maxinstrs = instrlimit;
runinfo.zri_new_zvols = fnvlist_alloc();
if (sync) {
err = dsl_sync_task_sig(poolname, NULL, zcp_eval_sync,
zcp_eval_sig, &runinfo, 0, ZFS_SPACE_CHECK_ZCP_EVAL);
if (err != 0)
zcp_pool_error(&runinfo, poolname);
} else {
zcp_eval_open(&runinfo, poolname);
}
lua_close(state);
/*
* Create device minor nodes for any new zvols.
*/
for (nvpair_t *pair = nvlist_next_nvpair(runinfo.zri_new_zvols, NULL);
pair != NULL;
pair = nvlist_next_nvpair(runinfo.zri_new_zvols, pair)) {
zvol_create_minor(nvpair_name(pair));
}
fnvlist_free(runinfo.zri_new_zvols);
return (runinfo.zri_result);
}
/*
* Retrieve metadata about the currently running channel program.
*/
zcp_run_info_t *
zcp_run_info(lua_State *state)
{
zcp_run_info_t *ri;
lua_getfield(state, LUA_REGISTRYINDEX, ZCP_RUN_INFO_KEY);
ri = lua_touserdata(state, -1);
lua_pop(state, 1);
return (ri);
}
/*
* Argument Parsing
* ================
*
* The Lua language allows methods to be called with any number
* of arguments of any type. When calling back into ZFS we need to sanitize
* arguments from channel programs to make sure unexpected arguments or
* arguments of the wrong type result in clear error messages. To do this
* in a uniform way all callbacks from channel programs should use the
* zcp_parse_args() function to interpret inputs.
*
* Positional vs Keyword Arguments
* ===============================
*
* Every callback function takes a fixed set of required positional arguments
* and optional keyword arguments. For example, the destroy function takes
* a single positional string argument (the name of the dataset to destroy)
* and an optional "defer" keyword boolean argument. When calling lua functions
* with parentheses, only positional arguments can be used:
*
* zfs.sync.snapshot("rpool@snap")
*
* To use keyword arguments functions should be called with a single argument
* that is a lua table containing mappings of integer -> positional arguments
* and string -> keyword arguments:
*
* zfs.sync.snapshot({1="rpool@snap", defer=true})
*
* The lua language allows curly braces to be used in place of parenthesis as
* syntactic sugar for this calling convention:
*
* zfs.sync.snapshot{"rpool@snap", defer=true}
*/
/*
* Throw an error and print the given arguments. If there are too many
* arguments to fit in the output buffer, only the error format string is
* output.
*/
static void
zcp_args_error(lua_State *state, const char *fname, const zcp_arg_t *pargs,
const zcp_arg_t *kwargs, const char *fmt, ...)
{
int i;
char errmsg[512];
size_t len = sizeof (errmsg);
size_t msglen = 0;
va_list argp;
va_start(argp, fmt);
VERIFY3U(len, >, vsnprintf(errmsg, len, fmt, argp));
va_end(argp);
/*
* Calculate the total length of the final string, including extra
* formatting characters. If the argument dump would be too large,
* only print the error string.
*/
msglen = strlen(errmsg);
msglen += strlen(fname) + 4; /* : + {} + null terminator */
for (i = 0; pargs[i].za_name != NULL; i++) {
msglen += strlen(pargs[i].za_name);
msglen += strlen(lua_typename(state, pargs[i].za_lua_type));
if (pargs[i + 1].za_name != NULL || kwargs[0].za_name != NULL)
msglen += 5; /* < + ( + )> + , */
else
msglen += 4; /* < + ( + )> */
}
for (i = 0; kwargs[i].za_name != NULL; i++) {
msglen += strlen(kwargs[i].za_name);
msglen += strlen(lua_typename(state, kwargs[i].za_lua_type));
if (kwargs[i + 1].za_name != NULL)
msglen += 4; /* =( + ) + , */
else
msglen += 3; /* =( + ) */
}
if (msglen >= len)
(void) luaL_error(state, errmsg);
VERIFY3U(len, >, strlcat(errmsg, ": ", len));
VERIFY3U(len, >, strlcat(errmsg, fname, len));
VERIFY3U(len, >, strlcat(errmsg, "{", len));
for (i = 0; pargs[i].za_name != NULL; i++) {
VERIFY3U(len, >, strlcat(errmsg, "<", len));
VERIFY3U(len, >, strlcat(errmsg, pargs[i].za_name, len));
VERIFY3U(len, >, strlcat(errmsg, "(", len));
VERIFY3U(len, >, strlcat(errmsg,
lua_typename(state, pargs[i].za_lua_type), len));
VERIFY3U(len, >, strlcat(errmsg, ")>", len));
if (pargs[i + 1].za_name != NULL || kwargs[0].za_name != NULL) {
VERIFY3U(len, >, strlcat(errmsg, ", ", len));
}
}
for (i = 0; kwargs[i].za_name != NULL; i++) {
VERIFY3U(len, >, strlcat(errmsg, kwargs[i].za_name, len));
VERIFY3U(len, >, strlcat(errmsg, "=(", len));
VERIFY3U(len, >, strlcat(errmsg,
lua_typename(state, kwargs[i].za_lua_type), len));
VERIFY3U(len, >, strlcat(errmsg, ")", len));
if (kwargs[i + 1].za_name != NULL) {
VERIFY3U(len, >, strlcat(errmsg, ", ", len));
}
}
VERIFY3U(len, >, strlcat(errmsg, "}", len));
(void) luaL_error(state, errmsg);
panic("unreachable code");
}
static void
zcp_parse_table_args(lua_State *state, const char *fname,
const zcp_arg_t *pargs, const zcp_arg_t *kwargs)
{
int i;
int type;
for (i = 0; pargs[i].za_name != NULL; i++) {
/*
* Check the table for this positional argument, leaving it
* on the top of the stack once we finish validating it.
*/
lua_pushinteger(state, i + 1);
lua_gettable(state, 1);
type = lua_type(state, -1);
if (type == LUA_TNIL) {
zcp_args_error(state, fname, pargs, kwargs,
"too few arguments");
panic("unreachable code");
} else if (type != pargs[i].za_lua_type) {
zcp_args_error(state, fname, pargs, kwargs,
"arg %d wrong type (is '%s', expected '%s')",
i + 1, lua_typename(state, type),
lua_typename(state, pargs[i].za_lua_type));
panic("unreachable code");
}
/*
* Remove the positional argument from the table.
*/
lua_pushinteger(state, i + 1);
lua_pushnil(state);
lua_settable(state, 1);
}
for (i = 0; kwargs[i].za_name != NULL; i++) {
/*
* Check the table for this keyword argument, which may be
* nil if it was omitted. Leave the value on the top of
* the stack after validating it.
*/
lua_getfield(state, 1, kwargs[i].za_name);
type = lua_type(state, -1);
if (type != LUA_TNIL && type != kwargs[i].za_lua_type) {
zcp_args_error(state, fname, pargs, kwargs,
"kwarg '%s' wrong type (is '%s', expected '%s')",
kwargs[i].za_name, lua_typename(state, type),
lua_typename(state, kwargs[i].za_lua_type));
panic("unreachable code");
}
/*
* Remove the keyword argument from the table.
*/
lua_pushnil(state);
lua_setfield(state, 1, kwargs[i].za_name);
}
/*
* Any entries remaining in the table are invalid inputs, print
* an error message based on what the entry is.
*/
lua_pushnil(state);
if (lua_next(state, 1)) {
if (lua_isnumber(state, -2) && lua_tointeger(state, -2) > 0) {
zcp_args_error(state, fname, pargs, kwargs,
"too many positional arguments");
} else if (lua_isstring(state, -2)) {
zcp_args_error(state, fname, pargs, kwargs,
"invalid kwarg '%s'", lua_tostring(state, -2));
} else {
zcp_args_error(state, fname, pargs, kwargs,
"kwarg keys must be strings");
}
panic("unreachable code");
}
lua_remove(state, 1);
}
static void
zcp_parse_pos_args(lua_State *state, const char *fname, const zcp_arg_t *pargs,
const zcp_arg_t *kwargs)
{
int i;
int type;
for (i = 0; pargs[i].za_name != NULL; i++) {
type = lua_type(state, i + 1);
if (type == LUA_TNONE) {
zcp_args_error(state, fname, pargs, kwargs,
"too few arguments");
panic("unreachable code");
} else if (type != pargs[i].za_lua_type) {
zcp_args_error(state, fname, pargs, kwargs,
"arg %d wrong type (is '%s', expected '%s')",
i + 1, lua_typename(state, type),
lua_typename(state, pargs[i].za_lua_type));
panic("unreachable code");
}
}
if (lua_gettop(state) != i) {
zcp_args_error(state, fname, pargs, kwargs,
"too many positional arguments");
panic("unreachable code");
}
for (i = 0; kwargs[i].za_name != NULL; i++) {
lua_pushnil(state);
}
}
/*
* Checks the current Lua stack against an expected set of positional and
* keyword arguments. If the stack does not match the expected arguments
* aborts the current channel program with a useful error message, otherwise
* it re-arranges the stack so that it contains the positional arguments
* followed by the keyword argument values in declaration order. Any missing
* keyword argument will be represented by a nil value on the stack.
*
* If the stack contains exactly one argument of type LUA_TTABLE the curly
* braces calling convention is assumed, otherwise the stack is parsed for
* positional arguments only.
*
* This function should be used by every function callback. It should be called
* before the callback manipulates the Lua stack as it assumes the stack
* represents the function arguments.
*/
void
zcp_parse_args(lua_State *state, const char *fname, const zcp_arg_t *pargs,
const zcp_arg_t *kwargs)
{
if (lua_gettop(state) == 1 && lua_istable(state, 1)) {
zcp_parse_table_args(state, fname, pargs, kwargs);
} else {
zcp_parse_pos_args(state, fname, pargs, kwargs);
}
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_lua, zfs_lua_, max_instrlimit, ULONG, ZMOD_RW,
"Max instruction limit that can be specified for a channel program");
ZFS_MODULE_PARAM(zfs_lua, zfs_lua_, max_memlimit, ULONG, ZMOD_RW,
"Max memory limit that can be specified for a channel program");
/* END CSTYLED */
diff --git a/module/zfs/zcp_get.c b/module/zfs/zcp_get.c
index 7256e4de1915..fe712afd7ade 100644
--- a/module/zfs/zcp_get.c
+++ b/module/zfs/zcp_get.c
@@ -1,813 +1,813 @@
/*
* CDDL HEADER START
*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2016 by Delphix. All rights reserved.
*/
#include <sys/lua/lua.h>
#include <sys/lua/lualib.h>
#include <sys/lua/lauxlib.h>
#include <zfs_prop.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_synctask.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dmu_objset.h>
#include <sys/mntent.h>
#include <sys/sunddi.h>
#include <sys/zap.h>
#include <sys/zcp.h>
#include <sys/zcp_iter.h>
#include <sys/zcp_global.h>
#include <sys/zcp_prop.h>
#include <sys/zfs_ioctl.h>
#include <sys/zfs_znode.h>
#include <sys/zvol.h>
#ifdef _KERNEL
#include <sys/zfs_quota.h>
#include <sys/zfs_vfsops.h>
#endif
static int
get_objset_type(dsl_dataset_t *ds, zfs_type_t *type)
{
int error;
objset_t *os;
error = dmu_objset_from_ds(ds, &os);
if (error != 0)
return (error);
if (ds->ds_is_snapshot) {
*type = ZFS_TYPE_SNAPSHOT;
} else {
switch (os->os_phys->os_type) {
case DMU_OST_ZFS:
*type = ZFS_TYPE_FILESYSTEM;
break;
case DMU_OST_ZVOL:
*type = ZFS_TYPE_VOLUME;
break;
default:
return (EINVAL);
}
}
return (0);
}
/*
* Returns the string name of ds's type in str (a buffer which should be
* at least 12 bytes long).
*/
static int
get_objset_type_name(dsl_dataset_t *ds, char *str)
{
int error;
zfs_type_t type;
error = get_objset_type(ds, &type);
if (error != 0)
return (error);
switch (type) {
case ZFS_TYPE_SNAPSHOT:
(void) strlcpy(str, "snapshot", ZAP_MAXVALUELEN);
break;
case ZFS_TYPE_FILESYSTEM:
(void) strlcpy(str, "filesystem", ZAP_MAXVALUELEN);
break;
case ZFS_TYPE_VOLUME:
(void) strlcpy(str, "volume", ZAP_MAXVALUELEN);
break;
default:
return (EINVAL);
}
return (0);
}
/*
* Determines the source of a property given its setpoint and
* property type. It pushes the source to the lua stack.
*/
static void
get_prop_src(lua_State *state, const char *setpoint, zfs_prop_t prop)
{
if (zfs_prop_readonly(prop) || (prop == ZFS_PROP_VERSION)) {
lua_pushnil(state);
} else {
const char *src;
if (strcmp("", setpoint) == 0) {
src = "default";
} else {
src = setpoint;
}
(void) lua_pushstring(state, src);
}
}
/*
* Given an error encountered while getting properties, either longjmp's for
* a fatal error or pushes nothing to the stack for a non fatal one.
*/
static int
zcp_handle_error(lua_State *state, const char *dataset_name,
const char *property_name, int error)
{
ASSERT3S(error, !=, 0);
if (error == ENOENT) {
return (0);
} else if (error == EINVAL) {
return (luaL_error(state,
"property '%s' is not a valid property on dataset '%s'",
property_name, dataset_name));
} else if (error == EIO) {
return (luaL_error(state,
"I/O error while retrieving property '%s' on dataset '%s'",
property_name, dataset_name));
} else {
return (luaL_error(state, "unexpected error %d while "
"retrieving property '%s' on dataset '%s'",
error, property_name, dataset_name));
}
}
/*
* Look up a user defined property in the zap object. If it exists, push it
* and the setpoint onto the stack, otherwise don't push anything.
*/
static int
zcp_get_user_prop(lua_State *state, dsl_pool_t *dp, const char *dataset_name,
const char *property_name)
{
int error;
char *buf;
char setpoint[ZFS_MAX_DATASET_NAME_LEN];
/*
* zcp_dataset_hold will either successfully return the requested
* dataset or throw a lua error and longjmp out of the zfs.get_prop call
* without returning.
*/
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dataset_name, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
buf = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP);
error = dsl_prop_get_ds(ds, property_name, 1, ZAP_MAXVALUELEN,
buf, setpoint);
dsl_dataset_rele(ds, FTAG);
if (error != 0) {
kmem_free(buf, ZAP_MAXVALUELEN);
return (zcp_handle_error(state, dataset_name, property_name,
error));
}
(void) lua_pushstring(state, buf);
(void) lua_pushstring(state, setpoint);
kmem_free(buf, ZAP_MAXVALUELEN);
return (2);
}
/*
* Check if the property we're looking for is stored in the ds_dir. If so,
* return it in the 'val' argument. Return 0 on success and ENOENT and if
* the property is not present.
*/
static int
get_dsl_dir_prop(dsl_dataset_t *ds, zfs_prop_t zfs_prop,
uint64_t *val)
{
dsl_dir_t *dd = ds->ds_dir;
mutex_enter(&dd->dd_lock);
switch (zfs_prop) {
case ZFS_PROP_USEDSNAP:
*val = dsl_dir_get_usedsnap(dd);
break;
case ZFS_PROP_USEDCHILD:
*val = dsl_dir_get_usedchild(dd);
break;
case ZFS_PROP_USEDDS:
*val = dsl_dir_get_usedds(dd);
break;
case ZFS_PROP_USEDREFRESERV:
*val = dsl_dir_get_usedrefreserv(dd);
break;
case ZFS_PROP_LOGICALUSED:
*val = dsl_dir_get_logicalused(dd);
break;
default:
mutex_exit(&dd->dd_lock);
return (SET_ERROR(ENOENT));
}
mutex_exit(&dd->dd_lock);
return (0);
}
/*
* Check if the property we're looking for is stored at the dsl_dataset or
* dsl_dir level. If so, push the property value and source onto the lua stack
* and return 0. If it is not present or a failure occurs in lookup, return a
* non-zero error value.
*/
static int
get_special_prop(lua_State *state, dsl_dataset_t *ds, const char *dsname,
zfs_prop_t zfs_prop)
{
int error = 0;
objset_t *os;
uint64_t numval = 0;
char *strval = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP);
char setpoint[ZFS_MAX_DATASET_NAME_LEN] =
"Internal error - setpoint not determined";
zfs_type_t ds_type;
zprop_type_t prop_type = zfs_prop_get_type(zfs_prop);
(void) get_objset_type(ds, &ds_type);
switch (zfs_prop) {
case ZFS_PROP_REFRATIO:
numval = dsl_get_refratio(ds);
break;
case ZFS_PROP_USED:
numval = dsl_get_used(ds);
break;
case ZFS_PROP_CLONES: {
nvlist_t *clones = fnvlist_alloc();
error = get_clones_stat_impl(ds, clones);
if (error == 0) {
/* push list to lua stack */
VERIFY0(zcp_nvlist_to_lua(state, clones, NULL, 0ULL));
/* source */
(void) lua_pushnil(state);
}
nvlist_free(clones);
kmem_free(strval, ZAP_MAXVALUELEN);
return (error);
}
case ZFS_PROP_COMPRESSRATIO:
numval = dsl_get_compressratio(ds);
break;
case ZFS_PROP_CREATION:
numval = dsl_get_creation(ds);
break;
case ZFS_PROP_REFERENCED:
numval = dsl_get_referenced(ds);
break;
case ZFS_PROP_AVAILABLE:
numval = dsl_get_available(ds);
break;
case ZFS_PROP_LOGICALREFERENCED:
numval = dsl_get_logicalreferenced(ds);
break;
case ZFS_PROP_CREATETXG:
numval = dsl_get_creationtxg(ds);
break;
case ZFS_PROP_GUID:
numval = dsl_get_guid(ds);
break;
case ZFS_PROP_UNIQUE:
numval = dsl_get_unique(ds);
break;
case ZFS_PROP_OBJSETID:
numval = dsl_get_objsetid(ds);
break;
case ZFS_PROP_ORIGIN:
dsl_dir_get_origin(ds->ds_dir, strval);
break;
case ZFS_PROP_USERACCOUNTING:
error = dmu_objset_from_ds(ds, &os);
if (error == 0)
numval = dmu_objset_userspace_present(os);
break;
case ZFS_PROP_WRITTEN:
error = dsl_get_written(ds, &numval);
break;
case ZFS_PROP_TYPE:
error = get_objset_type_name(ds, strval);
break;
case ZFS_PROP_PREV_SNAP:
error = dsl_get_prev_snap(ds, strval);
break;
case ZFS_PROP_NAME:
dsl_dataset_name(ds, strval);
break;
case ZFS_PROP_MOUNTPOINT:
error = dsl_get_mountpoint(ds, dsname, strval, setpoint);
break;
case ZFS_PROP_VERSION:
/* should be a snapshot or filesystem */
ASSERT(ds_type != ZFS_TYPE_VOLUME);
error = dmu_objset_from_ds(ds, &os);
/* look in the master node for the version */
if (error == 0) {
error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
sizeof (numval), 1, &numval);
}
break;
case ZFS_PROP_DEFER_DESTROY:
numval = dsl_get_defer_destroy(ds);
break;
case ZFS_PROP_USERREFS:
numval = dsl_get_userrefs(ds);
break;
case ZFS_PROP_FILESYSTEM_COUNT:
error = dsl_dir_get_filesystem_count(ds->ds_dir, &numval);
(void) strlcpy(setpoint, "", ZFS_MAX_DATASET_NAME_LEN);
break;
case ZFS_PROP_SNAPSHOT_COUNT:
error = dsl_dir_get_snapshot_count(ds->ds_dir, &numval);
(void) strlcpy(setpoint, "", ZFS_MAX_DATASET_NAME_LEN);
break;
case ZFS_PROP_NUMCLONES:
numval = dsl_get_numclones(ds);
break;
case ZFS_PROP_INCONSISTENT:
numval = dsl_get_inconsistent(ds);
break;
case ZFS_PROP_IVSET_GUID:
if (dsl_dataset_is_zapified(ds)) {
error = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset,
ds->ds_object, DS_FIELD_IVSET_GUID,
sizeof (numval), 1, &numval);
} else {
error = ENOENT;
}
break;
case ZFS_PROP_RECEIVE_RESUME_TOKEN: {
char *token = get_receive_resume_stats_impl(ds);
(void) strlcpy(strval, token, ZAP_MAXVALUELEN);
if (strcmp(strval, "") == 0) {
char *childval = get_child_receive_stats(ds);
(void) strlcpy(strval, childval, ZAP_MAXVALUELEN);
if (strcmp(strval, "") == 0)
error = ENOENT;
kmem_strfree(childval);
}
kmem_strfree(token);
break;
}
case ZFS_PROP_VOLSIZE:
ASSERT(ds_type == ZFS_TYPE_VOLUME ||
ds_type == ZFS_TYPE_SNAPSHOT);
error = dmu_objset_from_ds(ds, &os);
if (error == 0) {
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size",
sizeof (numval), 1, &numval);
}
if (error == 0)
(void) strlcpy(setpoint, dsname,
ZFS_MAX_DATASET_NAME_LEN);
break;
case ZFS_PROP_VOLBLOCKSIZE: {
ASSERT(ds_type == ZFS_TYPE_VOLUME);
dmu_object_info_t doi;
error = dmu_objset_from_ds(ds, &os);
if (error == 0) {
error = dmu_object_info(os, ZVOL_OBJ, &doi);
if (error == 0)
numval = doi.doi_data_block_size;
}
break;
}
case ZFS_PROP_KEYSTATUS:
case ZFS_PROP_KEYFORMAT: {
/* provide defaults in case no crypto obj exists */
setpoint[0] = '\0';
if (zfs_prop == ZFS_PROP_KEYSTATUS)
numval = ZFS_KEYSTATUS_NONE;
else
numval = ZFS_KEYFORMAT_NONE;
nvlist_t *nvl, *propval;
nvl = fnvlist_alloc();
dsl_dataset_crypt_stats(ds, nvl);
if (nvlist_lookup_nvlist(nvl, zfs_prop_to_name(zfs_prop),
&propval) == 0) {
char *source;
(void) nvlist_lookup_uint64(propval, ZPROP_VALUE,
&numval);
if (nvlist_lookup_string(propval, ZPROP_SOURCE,
&source) == 0)
strlcpy(setpoint, source, sizeof (setpoint));
}
nvlist_free(nvl);
break;
}
default:
/* Did not match these props, check in the dsl_dir */
error = get_dsl_dir_prop(ds, zfs_prop, &numval);
}
if (error != 0) {
kmem_free(strval, ZAP_MAXVALUELEN);
return (error);
}
switch (prop_type) {
case PROP_TYPE_NUMBER: {
(void) lua_pushnumber(state, numval);
break;
}
case PROP_TYPE_STRING: {
(void) lua_pushstring(state, strval);
break;
}
case PROP_TYPE_INDEX: {
const char *propval;
error = zfs_prop_index_to_string(zfs_prop, numval, &propval);
if (error != 0) {
kmem_free(strval, ZAP_MAXVALUELEN);
return (error);
}
(void) lua_pushstring(state, propval);
break;
}
}
kmem_free(strval, ZAP_MAXVALUELEN);
/* Push the source to the stack */
get_prop_src(state, setpoint, zfs_prop);
return (0);
}
/*
* Look up a property and its source in the zap object. If the value is
* present and successfully retrieved, push the value and source on the
* lua stack and return 0. On failure, return a non-zero error value.
*/
static int
get_zap_prop(lua_State *state, dsl_dataset_t *ds, zfs_prop_t zfs_prop)
{
int error = 0;
char setpoint[ZFS_MAX_DATASET_NAME_LEN];
char *strval = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP);
uint64_t numval;
const char *prop_name = zfs_prop_to_name(zfs_prop);
zprop_type_t prop_type = zfs_prop_get_type(zfs_prop);
if (prop_type == PROP_TYPE_STRING) {
/* Push value to lua stack */
error = dsl_prop_get_ds(ds, prop_name, 1,
ZAP_MAXVALUELEN, strval, setpoint);
if (error == 0)
(void) lua_pushstring(state, strval);
} else {
error = dsl_prop_get_ds(ds, prop_name, sizeof (numval),
1, &numval, setpoint);
#ifdef _KERNEL
/* Fill in temporary value for prop, if applicable */
(void) zfs_get_temporary_prop(ds, zfs_prop, &numval, setpoint);
#else
return (luaL_error(state,
"temporary properties only supported in kernel mode",
prop_name));
#endif
/* Push value to lua stack */
if (prop_type == PROP_TYPE_INDEX) {
const char *propval;
error = zfs_prop_index_to_string(zfs_prop, numval,
&propval);
if (error == 0)
(void) lua_pushstring(state, propval);
} else {
if (error == 0)
(void) lua_pushnumber(state, numval);
}
}
kmem_free(strval, ZAP_MAXVALUELEN);
if (error == 0)
get_prop_src(state, setpoint, zfs_prop);
return (error);
}
/*
* Determine whether property is valid for a given dataset
*/
boolean_t
prop_valid_for_ds(dsl_dataset_t *ds, zfs_prop_t zfs_prop)
{
int error;
zfs_type_t zfs_type;
/* properties not supported */
if ((zfs_prop == ZFS_PROP_ISCSIOPTIONS) ||
(zfs_prop == ZFS_PROP_MOUNTED))
return (B_FALSE);
/* if we want the origin prop, ds must be a clone */
if ((zfs_prop == ZFS_PROP_ORIGIN) && (!dsl_dir_is_clone(ds->ds_dir)))
return (B_FALSE);
error = get_objset_type(ds, &zfs_type);
if (error != 0)
return (B_FALSE);
return (zfs_prop_valid_for_type(zfs_prop, zfs_type, B_FALSE));
}
/*
* Look up a given dataset property. On success return 2, the number of
* values pushed to the lua stack (property value and source). On a fatal
* error, longjmp. On a non fatal error push nothing.
*/
static int
zcp_get_system_prop(lua_State *state, dsl_pool_t *dp, const char *dataset_name,
zfs_prop_t zfs_prop)
{
int error;
/*
* zcp_dataset_hold will either successfully return the requested
* dataset or throw a lua error and longjmp out of the zfs.get_prop call
* without returning.
*/
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dataset_name, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
/* Check that the property is valid for the given dataset */
const char *prop_name = zfs_prop_to_name(zfs_prop);
if (!prop_valid_for_ds(ds, zfs_prop)) {
dsl_dataset_rele(ds, FTAG);
return (0);
}
/* Check if the property can be accessed directly */
error = get_special_prop(state, ds, dataset_name, zfs_prop);
if (error == 0) {
dsl_dataset_rele(ds, FTAG);
/* The value and source have been pushed by get_special_prop */
return (2);
}
if (error != ENOENT) {
dsl_dataset_rele(ds, FTAG);
return (zcp_handle_error(state, dataset_name,
prop_name, error));
}
/* If we were unable to find it, look in the zap object */
error = get_zap_prop(state, ds, zfs_prop);
dsl_dataset_rele(ds, FTAG);
if (error != 0) {
return (zcp_handle_error(state, dataset_name,
prop_name, error));
}
/* The value and source have been pushed by get_zap_prop */
return (2);
}
#ifdef _KERNEL
static zfs_userquota_prop_t
get_userquota_prop(const char *prop_name)
{
zfs_userquota_prop_t type;
/* Figure out the property type ({user|group}{quota|used}) */
for (type = 0; type < ZFS_NUM_USERQUOTA_PROPS; type++) {
if (strncmp(prop_name, zfs_userquota_prop_prefixes[type],
strlen(zfs_userquota_prop_prefixes[type])) == 0)
break;
}
return (type);
}
/*
* Given the name of a zfs_userquota_prop, this function determines the
* prop type as well as the numeric group/user ids based on the string
* following the '@' in the property name. On success, returns 0. On failure,
* returns a non-zero error.
* 'domain' must be free'd by caller using kmem_strfree()
*/
static int
parse_userquota_prop(const char *prop_name, zfs_userquota_prop_t *type,
char **domain, uint64_t *rid)
{
char *cp, *end, *domain_val;
*type = get_userquota_prop(prop_name);
if (*type >= ZFS_NUM_USERQUOTA_PROPS)
return (EINVAL);
*rid = 0;
cp = strchr(prop_name, '@') + 1;
if (strncmp(cp, "S-1-", 4) == 0) {
/*
* It's a numeric SID (eg "S-1-234-567-89") and we want to
* separate the domain id and the rid
*/
int domain_len = strrchr(cp, '-') - cp;
domain_val = kmem_alloc(domain_len + 1, KM_SLEEP);
(void) strncpy(domain_val, cp, domain_len);
domain_val[domain_len] = '\0';
cp += domain_len + 1;
(void) ddi_strtoll(cp, &end, 10, (longlong_t *)rid);
if (*end != '\0') {
kmem_strfree(domain_val);
return (EINVAL);
}
} else {
/* It's only a user/group ID (eg "12345"), just get the rid */
domain_val = NULL;
(void) ddi_strtoll(cp, &end, 10, (longlong_t *)rid);
if (*end != '\0')
return (EINVAL);
}
*domain = domain_val;
return (0);
}
/*
* Look up {user|group}{quota|used} property for given dataset. On success
* push the value (quota or used amount) and the setpoint. On failure, push
* a lua error.
*/
static int
zcp_get_userquota_prop(lua_State *state, dsl_pool_t *dp,
const char *dataset_name, const char *prop_name)
{
zfsvfs_t *zfvp;
zfsvfs_t *zfsvfs;
int error;
zfs_userquota_prop_t type;
char *domain;
uint64_t rid, value = 0;
objset_t *os;
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dataset_name, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
error = parse_userquota_prop(prop_name, &type, &domain, &rid);
if (error == 0) {
error = dmu_objset_from_ds(ds, &os);
if (error == 0) {
zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
error = zfsvfs_create_impl(&zfvp, zfsvfs, os);
if (error == 0) {
error = zfs_userspace_one(zfvp, type, domain,
rid, &value);
zfsvfs_free(zfvp);
}
}
if (domain != NULL)
kmem_strfree(domain);
}
dsl_dataset_rele(ds, FTAG);
if ((value == 0) && ((type == ZFS_PROP_USERQUOTA) ||
(type == ZFS_PROP_GROUPQUOTA)))
error = SET_ERROR(ENOENT);
if (error != 0) {
return (zcp_handle_error(state, dataset_name,
prop_name, error));
}
(void) lua_pushnumber(state, value);
(void) lua_pushstring(state, dataset_name);
return (2);
}
#endif
/*
* Determines the name of the snapshot referenced in the written property
* name. Returns snapshot name in snap_name, a buffer that must be at least
* as large as ZFS_MAX_DATASET_NAME_LEN
*/
static void
parse_written_prop(const char *dataset_name, const char *prop_name,
char *snap_name)
{
ASSERT(zfs_prop_written(prop_name));
const char *name = prop_name + ZFS_WRITTEN_PROP_PREFIX_LEN;
if (strchr(name, '@') == NULL) {
(void) snprintf(snap_name, ZFS_MAX_DATASET_NAME_LEN, "%s@%s",
dataset_name, name);
} else {
(void) strlcpy(snap_name, name, ZFS_MAX_DATASET_NAME_LEN);
}
}
/*
* Look up written@ property for given dataset. On success
* push the value and the setpoint. If error is fatal, we will
* longjmp, otherwise push nothing.
*/
static int
zcp_get_written_prop(lua_State *state, dsl_pool_t *dp,
const char *dataset_name, const char *prop_name)
{
char snap_name[ZFS_MAX_DATASET_NAME_LEN];
uint64_t used, comp, uncomp;
dsl_dataset_t *old;
int error = 0;
parse_written_prop(dataset_name, prop_name, snap_name);
dsl_dataset_t *new = zcp_dataset_hold(state, dp, dataset_name, FTAG);
if (new == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
error = dsl_dataset_hold(dp, snap_name, FTAG, &old);
if (error != 0) {
dsl_dataset_rele(new, FTAG);
return (zcp_dataset_hold_error(state, dp, snap_name,
error));
}
error = dsl_dataset_space_written(old, new,
&used, &comp, &uncomp);
dsl_dataset_rele(old, FTAG);
dsl_dataset_rele(new, FTAG);
if (error != 0) {
return (zcp_handle_error(state, dataset_name,
snap_name, error));
}
(void) lua_pushnumber(state, used);
(void) lua_pushstring(state, dataset_name);
return (2);
}
static int zcp_get_prop(lua_State *state);
-static zcp_lib_info_t zcp_get_prop_info = {
+static const zcp_lib_info_t zcp_get_prop_info = {
.name = "get_prop",
.func = zcp_get_prop,
.pargs = {
- { .za_name = "dataset", .za_lua_type = LUA_TSTRING},
- { .za_name = "property", .za_lua_type = LUA_TSTRING},
+ { .za_name = "dataset", .za_lua_type = LUA_TSTRING },
+ { .za_name = "property", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
static int
zcp_get_prop(lua_State *state)
{
const char *dataset_name;
const char *property_name;
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
- zcp_lib_info_t *libinfo = &zcp_get_prop_info;
+ const zcp_lib_info_t *libinfo = &zcp_get_prop_info;
zcp_parse_args(state, libinfo->name, libinfo->pargs, libinfo->kwargs);
dataset_name = lua_tostring(state, 1);
property_name = lua_tostring(state, 2);
/* User defined property */
if (zfs_prop_user(property_name)) {
return (zcp_get_user_prop(state, dp,
dataset_name, property_name));
}
/* userspace property */
if (zfs_prop_userquota(property_name)) {
#ifdef _KERNEL
return (zcp_get_userquota_prop(state, dp,
dataset_name, property_name));
#else
return (luaL_error(state,
"user quota properties only supported in kernel mode",
property_name));
#endif
}
/* written@ property */
if (zfs_prop_written(property_name)) {
return (zcp_get_written_prop(state, dp,
dataset_name, property_name));
}
zfs_prop_t zfs_prop = zfs_name_to_prop(property_name);
/* Valid system property */
if (zfs_prop != ZPROP_INVAL) {
return (zcp_get_system_prop(state, dp, dataset_name,
zfs_prop));
}
/* Invalid property name */
return (luaL_error(state,
"'%s' is not a valid property", property_name));
}
int
zcp_load_get_lib(lua_State *state)
{
lua_pushcclosure(state, zcp_get_prop_info.func, 0);
lua_setfield(state, -2, zcp_get_prop_info.name);
return (1);
}
diff --git a/module/zfs/zcp_iter.c b/module/zfs/zcp_iter.c
index f727c56f212d..ed575738fc8b 100644
--- a/module/zfs/zcp_iter.c
+++ b/module/zfs/zcp_iter.c
@@ -1,751 +1,749 @@
/*
* CDDL HEADER START
*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2016, 2018 by Delphix. All rights reserved.
*/
#include <sys/lua/lua.h>
#include <sys/lua/lauxlib.h>
#include <sys/dmu.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_synctask.h>
#include <sys/dsl_bookmark.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_pool.h>
#include <sys/dmu_tx.h>
#include <sys/dmu_objset.h>
#include <sys/zap.h>
#include <sys/dsl_dir.h>
#include <sys/zcp_prop.h>
#include <sys/zcp.h>
#include "zfs_comutil.h"
typedef int (zcp_list_func_t)(lua_State *);
typedef struct zcp_list_info {
const char *name;
zcp_list_func_t *func;
zcp_list_func_t *gc;
const zcp_arg_t pargs[4];
const zcp_arg_t kwargs[2];
} zcp_list_info_t;
static int
zcp_clones_iter(lua_State *state)
{
int err;
char clonename[ZFS_MAX_DATASET_NAME_LEN];
uint64_t dsobj = lua_tonumber(state, lua_upvalueindex(1));
uint64_t cursor = lua_tonumber(state, lua_upvalueindex(2));
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
dsl_dataset_t *ds, *clone;
zap_attribute_t za;
zap_cursor_t zc;
err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
if (err == ENOENT) {
return (0);
} else if (err != 0) {
return (luaL_error(state,
"unexpected error %d from dsl_dataset_hold_obj(dsobj)",
err));
}
if (dsl_dataset_phys(ds)->ds_next_clones_obj == 0) {
dsl_dataset_rele(ds, FTAG);
return (0);
}
zap_cursor_init_serialized(&zc, dp->dp_meta_objset,
dsl_dataset_phys(ds)->ds_next_clones_obj, cursor);
dsl_dataset_rele(ds, FTAG);
err = zap_cursor_retrieve(&zc, &za);
if (err != 0) {
zap_cursor_fini(&zc);
if (err != ENOENT) {
return (luaL_error(state,
"unexpected error %d from zap_cursor_retrieve()",
err));
}
return (0);
}
zap_cursor_advance(&zc);
cursor = zap_cursor_serialize(&zc);
zap_cursor_fini(&zc);
err = dsl_dataset_hold_obj(dp, za.za_first_integer, FTAG, &clone);
if (err != 0) {
return (luaL_error(state,
"unexpected error %d from "
"dsl_dataset_hold_obj(za_first_integer)", err));
}
dsl_dir_name(clone->ds_dir, clonename);
dsl_dataset_rele(clone, FTAG);
lua_pushnumber(state, cursor);
lua_replace(state, lua_upvalueindex(2));
(void) lua_pushstring(state, clonename);
return (1);
}
static int zcp_clones_list(lua_State *);
-static zcp_list_info_t zcp_clones_list_info = {
+static const zcp_list_info_t zcp_clones_list_info = {
.name = "clones",
.func = zcp_clones_list,
.gc = NULL,
.pargs = {
- { .za_name = "snapshot", .za_lua_type = LUA_TSTRING},
+ { .za_name = "snapshot", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
static int
zcp_clones_list(lua_State *state)
{
const char *snapname = lua_tostring(state, 1);
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
/*
* zcp_dataset_hold will either successfully return the requested
* dataset or throw a lua error and longjmp out of the zfs.list.clones
* call without returning.
*/
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, snapname, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
boolean_t issnap = ds->ds_is_snapshot;
uint64_t cursor = 0;
uint64_t dsobj = ds->ds_object;
dsl_dataset_rele(ds, FTAG);
if (!issnap) {
return (zcp_argerror(state, 1, "%s is not a snapshot",
snapname));
}
lua_pushnumber(state, dsobj);
lua_pushnumber(state, cursor);
lua_pushcclosure(state, &zcp_clones_iter, 2);
return (1);
}
static int
zcp_snapshots_iter(lua_State *state)
{
int err;
char snapname[ZFS_MAX_DATASET_NAME_LEN];
uint64_t dsobj = lua_tonumber(state, lua_upvalueindex(1));
uint64_t cursor = lua_tonumber(state, lua_upvalueindex(2));
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
dsl_dataset_t *ds;
objset_t *os;
char *p;
err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
if (err != 0) {
return (luaL_error(state,
"unexpected error %d from dsl_dataset_hold_obj(dsobj)",
err));
}
dsl_dataset_name(ds, snapname);
VERIFY3U(sizeof (snapname), >,
strlcat(snapname, "@", sizeof (snapname)));
p = strchr(snapname, '\0');
VERIFY0(dmu_objset_from_ds(ds, &os));
err = dmu_snapshot_list_next(os,
sizeof (snapname) - (p - snapname), p, NULL, &cursor, NULL);
dsl_dataset_rele(ds, FTAG);
if (err == ENOENT) {
return (0);
} else if (err != 0) {
return (luaL_error(state,
"unexpected error %d from dmu_snapshot_list_next()", err));
}
lua_pushnumber(state, cursor);
lua_replace(state, lua_upvalueindex(2));
(void) lua_pushstring(state, snapname);
return (1);
}
static int zcp_snapshots_list(lua_State *);
-static zcp_list_info_t zcp_snapshots_list_info = {
+static const zcp_list_info_t zcp_snapshots_list_info = {
.name = "snapshots",
.func = zcp_snapshots_list,
.gc = NULL,
.pargs = {
- { .za_name = "filesystem | volume", .za_lua_type = LUA_TSTRING},
+ { .za_name = "filesystem | volume", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
static int
zcp_snapshots_list(lua_State *state)
{
const char *fsname = lua_tostring(state, 1);
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
boolean_t issnap;
uint64_t dsobj;
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, fsname, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
issnap = ds->ds_is_snapshot;
dsobj = ds->ds_object;
dsl_dataset_rele(ds, FTAG);
if (issnap) {
return (zcp_argerror(state, 1,
"argument %s cannot be a snapshot", fsname));
}
lua_pushnumber(state, dsobj);
lua_pushnumber(state, 0);
lua_pushcclosure(state, &zcp_snapshots_iter, 2);
return (1);
}
static int
zcp_children_iter(lua_State *state)
{
int err;
char childname[ZFS_MAX_DATASET_NAME_LEN];
uint64_t dsobj = lua_tonumber(state, lua_upvalueindex(1));
uint64_t cursor = lua_tonumber(state, lua_upvalueindex(2));
zcp_run_info_t *ri = zcp_run_info(state);
dsl_pool_t *dp = ri->zri_pool;
dsl_dataset_t *ds;
objset_t *os;
char *p;
err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
if (err != 0) {
return (luaL_error(state,
"unexpected error %d from dsl_dataset_hold_obj(dsobj)",
err));
}
dsl_dataset_name(ds, childname);
VERIFY3U(sizeof (childname), >,
strlcat(childname, "/", sizeof (childname)));
p = strchr(childname, '\0');
VERIFY0(dmu_objset_from_ds(ds, &os));
do {
err = dmu_dir_list_next(os,
sizeof (childname) - (p - childname), p, NULL, &cursor);
} while (err == 0 && zfs_dataset_name_hidden(childname));
dsl_dataset_rele(ds, FTAG);
if (err == ENOENT) {
return (0);
} else if (err != 0) {
return (luaL_error(state,
"unexpected error %d from dmu_dir_list_next()",
err));
}
lua_pushnumber(state, cursor);
lua_replace(state, lua_upvalueindex(2));
(void) lua_pushstring(state, childname);
return (1);
}
static int zcp_children_list(lua_State *);
-static zcp_list_info_t zcp_children_list_info = {
+static const zcp_list_info_t zcp_children_list_info = {
.name = "children",
.func = zcp_children_list,
.gc = NULL,
.pargs = {
- { .za_name = "filesystem | volume", .za_lua_type = LUA_TSTRING},
+ { .za_name = "filesystem | volume", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
static int
zcp_children_list(lua_State *state)
{
const char *fsname = lua_tostring(state, 1);
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
boolean_t issnap;
uint64_t dsobj;
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, fsname, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
issnap = ds->ds_is_snapshot;
dsobj = ds->ds_object;
dsl_dataset_rele(ds, FTAG);
if (issnap) {
return (zcp_argerror(state, 1,
"argument %s cannot be a snapshot", fsname));
}
lua_pushnumber(state, dsobj);
lua_pushnumber(state, 0);
lua_pushcclosure(state, &zcp_children_iter, 2);
return (1);
}
static int
zcp_user_props_list_gc(lua_State *state)
{
nvlist_t **props = lua_touserdata(state, 1);
if (*props != NULL)
fnvlist_free(*props);
return (0);
}
static int
zcp_user_props_iter(lua_State *state)
{
char *source, *val;
nvlist_t *nvprop;
nvlist_t **props = lua_touserdata(state, lua_upvalueindex(1));
nvpair_t *pair = lua_touserdata(state, lua_upvalueindex(2));
do {
pair = nvlist_next_nvpair(*props, pair);
if (pair == NULL) {
fnvlist_free(*props);
*props = NULL;
return (0);
}
} while (!zfs_prop_user(nvpair_name(pair)));
lua_pushlightuserdata(state, pair);
lua_replace(state, lua_upvalueindex(2));
nvprop = fnvpair_value_nvlist(pair);
val = fnvlist_lookup_string(nvprop, ZPROP_VALUE);
source = fnvlist_lookup_string(nvprop, ZPROP_SOURCE);
(void) lua_pushstring(state, nvpair_name(pair));
(void) lua_pushstring(state, val);
(void) lua_pushstring(state, source);
return (3);
}
static int zcp_user_props_list(lua_State *);
-static zcp_list_info_t zcp_user_props_list_info = {
+static const zcp_list_info_t zcp_user_props_list_info = {
.name = "user_properties",
.func = zcp_user_props_list,
.gc = zcp_user_props_list_gc,
.pargs = {
{ .za_name = "filesystem | snapshot | volume",
- .za_lua_type = LUA_TSTRING},
+ .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
/*
* 'properties' was the initial name for 'user_properties' seen
* above. 'user_properties' is a better name as it distinguishes
* these properties from 'system_properties' which are different.
* In order to avoid breaking compatibility between different
* versions of ZFS, we declare 'properties' as an alias for
* 'user_properties'.
*/
-static zcp_list_info_t zcp_props_list_info = {
+static const zcp_list_info_t zcp_props_list_info = {
.name = "properties",
.func = zcp_user_props_list,
.gc = zcp_user_props_list_gc,
.pargs = {
{ .za_name = "filesystem | snapshot | volume",
- .za_lua_type = LUA_TSTRING},
+ .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
static int
zcp_user_props_list(lua_State *state)
{
const char *dsname = lua_tostring(state, 1);
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
objset_t *os;
nvlist_t **props = lua_newuserdata(state, sizeof (nvlist_t *));
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dsname, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
VERIFY0(dmu_objset_from_ds(ds, &os));
VERIFY0(dsl_prop_get_all(os, props));
dsl_dataset_rele(ds, FTAG);
/*
* Set the metatable for the properties list to free it on
* completion.
*/
luaL_getmetatable(state, zcp_user_props_list_info.name);
(void) lua_setmetatable(state, -2);
lua_pushlightuserdata(state, NULL);
lua_pushcclosure(state, &zcp_user_props_iter, 2);
return (1);
}
/*
* Populate nv with all valid system properties and their values for the given
* dataset.
*/
static void
zcp_dataset_system_props(dsl_dataset_t *ds, nvlist_t *nv)
{
for (int prop = ZFS_PROP_TYPE; prop < ZFS_NUM_PROPS; prop++) {
/* Do not display hidden props */
if (!zfs_prop_visible(prop))
continue;
/* Do not display props not valid for this dataset */
if (!prop_valid_for_ds(ds, prop))
continue;
fnvlist_add_boolean(nv, zfs_prop_to_name(prop));
}
}
static int zcp_system_props_list(lua_State *);
-static zcp_list_info_t zcp_system_props_list_info = {
+static const zcp_list_info_t zcp_system_props_list_info = {
.name = "system_properties",
.func = zcp_system_props_list,
.pargs = {
- { .za_name = "dataset", .za_lua_type = LUA_TSTRING},
+ { .za_name = "dataset", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
/*
* Get a list of all visible system properties and their values for a given
* dataset. Returned on the stack as a Lua table.
*/
static int
zcp_system_props_list(lua_State *state)
{
int error;
char errbuf[128];
const char *dataset_name;
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
- zcp_list_info_t *libinfo = &zcp_system_props_list_info;
+ const zcp_list_info_t *libinfo = &zcp_system_props_list_info;
zcp_parse_args(state, libinfo->name, libinfo->pargs, libinfo->kwargs);
dataset_name = lua_tostring(state, 1);
nvlist_t *nv = fnvlist_alloc();
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dataset_name, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
/* Get the names of all valid system properties for this dataset */
zcp_dataset_system_props(ds, nv);
dsl_dataset_rele(ds, FTAG);
/* push list as lua table */
error = zcp_nvlist_to_lua(state, nv, errbuf, sizeof (errbuf));
nvlist_free(nv);
if (error != 0) {
return (luaL_error(state,
"Error returning nvlist: %s", errbuf));
}
return (1);
}
static int
zcp_bookmarks_iter(lua_State *state)
{
char ds_name[ZFS_MAX_DATASET_NAME_LEN];
char bookmark_name[ZFS_MAX_DATASET_NAME_LEN];
uint64_t dsobj = lua_tonumber(state, lua_upvalueindex(1));
uint64_t cursor = lua_tonumber(state, lua_upvalueindex(2));
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
dsl_dataset_t *ds;
zap_attribute_t za;
zap_cursor_t zc;
int err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
if (err == ENOENT) {
return (0);
} else if (err != 0) {
return (luaL_error(state,
"unexpected error %d from dsl_dataset_hold_obj(dsobj)",
err));
}
if (!dsl_dataset_is_zapified(ds)) {
dsl_dataset_rele(ds, FTAG);
return (0);
}
err = zap_lookup(dp->dp_meta_objset, ds->ds_object,
DS_FIELD_BOOKMARK_NAMES, sizeof (ds->ds_bookmarks_obj), 1,
&ds->ds_bookmarks_obj);
if (err != 0 && err != ENOENT) {
dsl_dataset_rele(ds, FTAG);
return (luaL_error(state,
"unexpected error %d from zap_lookup()", err));
}
if (ds->ds_bookmarks_obj == 0) {
dsl_dataset_rele(ds, FTAG);
return (0);
}
/* Store the dataset's name so we can append the bookmark's name */
dsl_dataset_name(ds, ds_name);
zap_cursor_init_serialized(&zc, ds->ds_dir->dd_pool->dp_meta_objset,
ds->ds_bookmarks_obj, cursor);
dsl_dataset_rele(ds, FTAG);
err = zap_cursor_retrieve(&zc, &za);
if (err != 0) {
zap_cursor_fini(&zc);
if (err != ENOENT) {
return (luaL_error(state,
"unexpected error %d from zap_cursor_retrieve()",
err));
}
return (0);
}
zap_cursor_advance(&zc);
cursor = zap_cursor_serialize(&zc);
zap_cursor_fini(&zc);
/* Create the full "pool/fs#bookmark" string to return */
int n = snprintf(bookmark_name, ZFS_MAX_DATASET_NAME_LEN, "%s#%s",
ds_name, za.za_name);
if (n >= ZFS_MAX_DATASET_NAME_LEN) {
return (luaL_error(state,
"unexpected error %d from snprintf()", ENAMETOOLONG));
}
lua_pushnumber(state, cursor);
lua_replace(state, lua_upvalueindex(2));
(void) lua_pushstring(state, bookmark_name);
return (1);
}
static int zcp_bookmarks_list(lua_State *);
-static zcp_list_info_t zcp_bookmarks_list_info = {
+static const zcp_list_info_t zcp_bookmarks_list_info = {
.name = "bookmarks",
.func = zcp_bookmarks_list,
.pargs = {
- { .za_name = "dataset", .za_lua_type = LUA_TSTRING},
+ { .za_name = "dataset", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
static int
zcp_bookmarks_list(lua_State *state)
{
const char *dsname = lua_tostring(state, 1);
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dsname, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
boolean_t issnap = ds->ds_is_snapshot;
uint64_t dsobj = ds->ds_object;
uint64_t cursor = 0;
dsl_dataset_rele(ds, FTAG);
if (issnap) {
return (zcp_argerror(state, 1, "%s is a snapshot", dsname));
}
lua_pushnumber(state, dsobj);
lua_pushnumber(state, cursor);
lua_pushcclosure(state, &zcp_bookmarks_iter, 2);
return (1);
}
static int
zcp_holds_iter(lua_State *state)
{
uint64_t dsobj = lua_tonumber(state, lua_upvalueindex(1));
uint64_t cursor = lua_tonumber(state, lua_upvalueindex(2));
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
dsl_dataset_t *ds;
zap_attribute_t za;
zap_cursor_t zc;
int err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
if (err == ENOENT) {
return (0);
} else if (err != 0) {
return (luaL_error(state,
"unexpected error %d from dsl_dataset_hold_obj(dsobj)",
err));
}
if (dsl_dataset_phys(ds)->ds_userrefs_obj == 0) {
dsl_dataset_rele(ds, FTAG);
return (0);
}
zap_cursor_init_serialized(&zc, ds->ds_dir->dd_pool->dp_meta_objset,
dsl_dataset_phys(ds)->ds_userrefs_obj, cursor);
dsl_dataset_rele(ds, FTAG);
err = zap_cursor_retrieve(&zc, &za);
if (err != 0) {
zap_cursor_fini(&zc);
if (err != ENOENT) {
return (luaL_error(state,
"unexpected error %d from zap_cursor_retrieve()",
err));
}
return (0);
}
zap_cursor_advance(&zc);
cursor = zap_cursor_serialize(&zc);
zap_cursor_fini(&zc);
lua_pushnumber(state, cursor);
lua_replace(state, lua_upvalueindex(2));
(void) lua_pushstring(state, za.za_name);
(void) lua_pushnumber(state, za.za_first_integer);
return (2);
}
static int zcp_holds_list(lua_State *);
-static zcp_list_info_t zcp_holds_list_info = {
+static const zcp_list_info_t zcp_holds_list_info = {
.name = "holds",
.func = zcp_holds_list,
.gc = NULL,
.pargs = {
- { .za_name = "snapshot", .za_lua_type = LUA_TSTRING},
+ { .za_name = "snapshot", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
}
};
/*
* Iterate over all the holds for a given dataset. Each iteration returns
* a hold's tag and its timestamp as an integer.
*/
static int
zcp_holds_list(lua_State *state)
{
const char *snapname = lua_tostring(state, 1);
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, snapname, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
boolean_t issnap = ds->ds_is_snapshot;
uint64_t dsobj = ds->ds_object;
uint64_t cursor = 0;
dsl_dataset_rele(ds, FTAG);
if (!issnap) {
return (zcp_argerror(state, 1, "%s is not a snapshot",
snapname));
}
lua_pushnumber(state, dsobj);
lua_pushnumber(state, cursor);
lua_pushcclosure(state, &zcp_holds_iter, 2);
return (1);
}
static int
zcp_list_func(lua_State *state)
{
zcp_list_info_t *info = lua_touserdata(state, lua_upvalueindex(1));
zcp_parse_args(state, info->name, info->pargs, info->kwargs);
return (info->func(state));
}
int
zcp_load_list_lib(lua_State *state)
{
- int i;
- zcp_list_info_t *zcp_list_funcs[] = {
+ const zcp_list_info_t *zcp_list_funcs[] = {
&zcp_children_list_info,
&zcp_snapshots_list_info,
&zcp_user_props_list_info,
&zcp_props_list_info,
&zcp_clones_list_info,
&zcp_system_props_list_info,
&zcp_bookmarks_list_info,
&zcp_holds_list_info,
NULL
};
lua_newtable(state);
- for (i = 0; zcp_list_funcs[i] != NULL; i++) {
- zcp_list_info_t *info = zcp_list_funcs[i];
+ for (int i = 0; zcp_list_funcs[i] != NULL; i++) {
+ const zcp_list_info_t *info = zcp_list_funcs[i];
if (info->gc != NULL) {
/*
* If the function requires garbage collection, create
* a metatable with its name and register the __gc
* function.
*/
(void) luaL_newmetatable(state, info->name);
(void) lua_pushstring(state, "__gc");
lua_pushcfunction(state, info->gc);
lua_settable(state, -3);
lua_pop(state, 1);
}
- lua_pushlightuserdata(state, info);
+ lua_pushlightuserdata(state, (void *)(uintptr_t)info);
lua_pushcclosure(state, &zcp_list_func, 1);
lua_setfield(state, -2, info->name);
- info++;
}
return (1);
}
diff --git a/module/zfs/zcp_synctask.c b/module/zfs/zcp_synctask.c
index bfcdbcf9c27f..403856ae3571 100644
--- a/module/zfs/zcp_synctask.c
+++ b/module/zfs/zcp_synctask.c
@@ -1,551 +1,549 @@
/*
* CDDL HEADER START
*
* This file and its contents are supplied under the terms of the
* Common Development and Distribution License ("CDDL"), version 1.0.
* You may only use this file in accordance with the terms of version
* 1.0 of the CDDL.
*
* A full copy of the text of the CDDL should have accompanied this
* source. A copy of the CDDL is also available via the Internet at
* http://www.illumos.org/license/CDDL.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2016, 2017 by Delphix. All rights reserved.
* Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved.
* Copyright 2020 Joyent, Inc.
*/
#include <sys/lua/lua.h>
#include <sys/lua/lauxlib.h>
#include <sys/zcp.h>
#include <sys/zcp_set.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_synctask.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_bookmark.h>
#include <sys/dsl_destroy.h>
#include <sys/dmu_objset.h>
#include <sys/zfs_znode.h>
#include <sys/zfeature.h>
#include <sys/metaslab.h>
#define DST_AVG_BLKSHIFT 14
typedef struct zcp_inherit_prop_arg {
lua_State *zipa_state;
const char *zipa_prop;
dsl_props_set_arg_t zipa_dpsa;
} zcp_inherit_prop_arg_t;
typedef int (zcp_synctask_func_t)(lua_State *, boolean_t, nvlist_t *);
typedef struct zcp_synctask_info {
const char *name;
zcp_synctask_func_t *func;
const zcp_arg_t pargs[4];
const zcp_arg_t kwargs[2];
zfs_space_check_t space_check;
int blocks_modified;
} zcp_synctask_info_t;
static void
zcp_synctask_cleanup(void *arg)
{
fnvlist_free(arg);
}
/*
* Generic synctask interface for channel program syncfuncs.
*
* To perform some action in syncing context, we'd generally call
* dsl_sync_task(), but since the Lua script is already running inside a
* synctask we need to leave out some actions (such as acquiring the config
* rwlock and performing space checks).
*
* If 'sync' is false, executes a dry run and returns the error code.
*
* If we are not running in syncing context and we are not doing a dry run
* (meaning we are running a zfs.sync function in open-context) then we
* return a Lua error.
*
* This function also handles common fatal error cases for channel program
* library functions. If a fatal error occurs, err_dsname will be the dataset
* name reported in error messages, if supplied.
*/
static int
zcp_sync_task(lua_State *state, dsl_checkfunc_t *checkfunc,
dsl_syncfunc_t *syncfunc, void *arg, boolean_t sync, const char *err_dsname)
{
int err;
zcp_run_info_t *ri = zcp_run_info(state);
err = checkfunc(arg, ri->zri_tx);
if (!sync)
return (err);
if (!ri->zri_sync) {
return (luaL_error(state, "running functions from the zfs.sync "
"submodule requires passing sync=TRUE to "
"lzc_channel_program() (i.e. do not specify the \"-n\" "
"command line argument)"));
}
if (err == 0) {
syncfunc(arg, ri->zri_tx);
} else if (err == EIO) {
if (err_dsname != NULL) {
return (luaL_error(state,
"I/O error while accessing dataset '%s'",
err_dsname));
} else {
return (luaL_error(state,
"I/O error while accessing dataset."));
}
}
return (err);
}
static int zcp_synctask_destroy(lua_State *, boolean_t, nvlist_t *);
-static zcp_synctask_info_t zcp_synctask_destroy_info = {
+static const zcp_synctask_info_t zcp_synctask_destroy_info = {
.name = "destroy",
.func = zcp_synctask_destroy,
.pargs = {
- {.za_name = "filesystem | snapshot", .za_lua_type = LUA_TSTRING},
+ {.za_name = "filesystem | snapshot", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
- {.za_name = "defer", .za_lua_type = LUA_TBOOLEAN},
+ {.za_name = "defer", .za_lua_type = LUA_TBOOLEAN },
{NULL, 0}
},
.space_check = ZFS_SPACE_CHECK_DESTROY,
.blocks_modified = 0
};
static int
zcp_synctask_destroy(lua_State *state, boolean_t sync, nvlist_t *err_details)
{
(void) err_details;
int err;
const char *dsname = lua_tostring(state, 1);
boolean_t issnap = (strchr(dsname, '@') != NULL);
if (!issnap && !lua_isnil(state, 2)) {
return (luaL_error(state,
"'deferred' kwarg only supported for snapshots: %s",
dsname));
}
if (issnap) {
dsl_destroy_snapshot_arg_t ddsa = { 0 };
ddsa.ddsa_name = dsname;
if (!lua_isnil(state, 2)) {
ddsa.ddsa_defer = lua_toboolean(state, 2);
} else {
ddsa.ddsa_defer = B_FALSE;
}
err = zcp_sync_task(state, dsl_destroy_snapshot_check,
dsl_destroy_snapshot_sync, &ddsa, sync, dsname);
} else {
dsl_destroy_head_arg_t ddha = { 0 };
ddha.ddha_name = dsname;
err = zcp_sync_task(state, dsl_destroy_head_check,
dsl_destroy_head_sync, &ddha, sync, dsname);
}
return (err);
}
static int zcp_synctask_promote(lua_State *, boolean_t, nvlist_t *);
-static zcp_synctask_info_t zcp_synctask_promote_info = {
+static const zcp_synctask_info_t zcp_synctask_promote_info = {
.name = "promote",
.func = zcp_synctask_promote,
.pargs = {
- {.za_name = "clone", .za_lua_type = LUA_TSTRING},
+ {.za_name = "clone", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
},
.space_check = ZFS_SPACE_CHECK_RESERVED,
.blocks_modified = 3
};
static int
zcp_synctask_promote(lua_State *state, boolean_t sync, nvlist_t *err_details)
{
int err;
dsl_dataset_promote_arg_t ddpa = { 0 };
const char *dsname = lua_tostring(state, 1);
zcp_run_info_t *ri = zcp_run_info(state);
ddpa.ddpa_clonename = dsname;
ddpa.err_ds = err_details;
ddpa.cr = ri->zri_cred;
ddpa.proc = ri->zri_proc;
/*
* If there was a snapshot name conflict, then err_ds will be filled
* with a list of conflicting snapshot names.
*/
err = zcp_sync_task(state, dsl_dataset_promote_check,
dsl_dataset_promote_sync, &ddpa, sync, dsname);
return (err);
}
static int zcp_synctask_rollback(lua_State *, boolean_t, nvlist_t *err_details);
-static zcp_synctask_info_t zcp_synctask_rollback_info = {
+static const zcp_synctask_info_t zcp_synctask_rollback_info = {
.name = "rollback",
.func = zcp_synctask_rollback,
.space_check = ZFS_SPACE_CHECK_RESERVED,
.blocks_modified = 1,
.pargs = {
- {.za_name = "filesystem", .za_lua_type = LUA_TSTRING},
+ {.za_name = "filesystem", .za_lua_type = LUA_TSTRING },
{0, 0}
},
.kwargs = {
{0, 0}
}
};
static int
zcp_synctask_rollback(lua_State *state, boolean_t sync, nvlist_t *err_details)
{
int err;
const char *dsname = lua_tostring(state, 1);
dsl_dataset_rollback_arg_t ddra = { 0 };
ddra.ddra_fsname = dsname;
ddra.ddra_result = err_details;
err = zcp_sync_task(state, dsl_dataset_rollback_check,
dsl_dataset_rollback_sync, &ddra, sync, dsname);
return (err);
}
static int zcp_synctask_snapshot(lua_State *, boolean_t, nvlist_t *);
-static zcp_synctask_info_t zcp_synctask_snapshot_info = {
+static const zcp_synctask_info_t zcp_synctask_snapshot_info = {
.name = "snapshot",
.func = zcp_synctask_snapshot,
.pargs = {
{.za_name = "filesystem@snapname | volume@snapname",
- .za_lua_type = LUA_TSTRING},
+ .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
},
.space_check = ZFS_SPACE_CHECK_NORMAL,
.blocks_modified = 3
};
static int
zcp_synctask_snapshot(lua_State *state, boolean_t sync, nvlist_t *err_details)
{
(void) err_details;
int err;
dsl_dataset_snapshot_arg_t ddsa = { 0 };
const char *dsname = lua_tostring(state, 1);
zcp_run_info_t *ri = zcp_run_info(state);
/*
* On old pools, the ZIL must not be active when a snapshot is created,
* but we can't suspend the ZIL because we're already in syncing
* context.
*/
if (spa_version(ri->zri_pool->dp_spa) < SPA_VERSION_FAST_SNAP) {
return (SET_ERROR(ENOTSUP));
}
/*
* We only allow for a single snapshot rather than a list, so the
* error list output is unnecessary.
*/
ddsa.ddsa_errors = NULL;
ddsa.ddsa_props = NULL;
ddsa.ddsa_cr = ri->zri_cred;
ddsa.ddsa_proc = ri->zri_proc;
ddsa.ddsa_snaps = fnvlist_alloc();
fnvlist_add_boolean(ddsa.ddsa_snaps, dsname);
zcp_cleanup_handler_t *zch = zcp_register_cleanup(state,
zcp_synctask_cleanup, ddsa.ddsa_snaps);
err = zcp_sync_task(state, dsl_dataset_snapshot_check,
dsl_dataset_snapshot_sync, &ddsa, sync, dsname);
if (err == 0) {
/*
* We may need to create a new device minor node for this
* dataset (if it is a zvol and the "snapdev" property is set).
* Save it in the nvlist so that it can be processed in open
* context.
*/
fnvlist_add_boolean(ri->zri_new_zvols, dsname);
}
zcp_deregister_cleanup(state, zch);
fnvlist_free(ddsa.ddsa_snaps);
return (err);
}
static int zcp_synctask_inherit_prop(lua_State *, boolean_t,
nvlist_t *err_details);
-static zcp_synctask_info_t zcp_synctask_inherit_prop_info = {
+static const zcp_synctask_info_t zcp_synctask_inherit_prop_info = {
.name = "inherit",
.func = zcp_synctask_inherit_prop,
.space_check = ZFS_SPACE_CHECK_RESERVED,
.blocks_modified = 2, /* 2 * numprops */
.pargs = {
{ .za_name = "dataset", .za_lua_type = LUA_TSTRING },
{ .za_name = "property", .za_lua_type = LUA_TSTRING },
{ NULL, 0 }
},
.kwargs = {
{ NULL, 0 }
},
};
static int
zcp_synctask_inherit_prop_check(void *arg, dmu_tx_t *tx)
{
zcp_inherit_prop_arg_t *args = arg;
zfs_prop_t prop = zfs_name_to_prop(args->zipa_prop);
if (prop == ZPROP_INVAL) {
if (zfs_prop_user(args->zipa_prop))
return (0);
return (EINVAL);
}
if (zfs_prop_readonly(prop))
return (EINVAL);
if (!zfs_prop_inheritable(prop))
return (EINVAL);
return (dsl_props_set_check(&args->zipa_dpsa, tx));
}
static void
zcp_synctask_inherit_prop_sync(void *arg, dmu_tx_t *tx)
{
zcp_inherit_prop_arg_t *args = arg;
dsl_props_set_arg_t *dpsa = &args->zipa_dpsa;
dsl_props_set_sync(dpsa, tx);
}
static int
zcp_synctask_inherit_prop(lua_State *state, boolean_t sync,
nvlist_t *err_details)
{
(void) err_details;
int err;
zcp_inherit_prop_arg_t zipa = { 0 };
dsl_props_set_arg_t *dpsa = &zipa.zipa_dpsa;
const char *dsname = lua_tostring(state, 1);
const char *prop = lua_tostring(state, 2);
zipa.zipa_state = state;
zipa.zipa_prop = prop;
dpsa->dpsa_dsname = dsname;
dpsa->dpsa_source = ZPROP_SRC_INHERITED;
dpsa->dpsa_props = fnvlist_alloc();
fnvlist_add_boolean(dpsa->dpsa_props, prop);
zcp_cleanup_handler_t *zch = zcp_register_cleanup(state,
zcp_synctask_cleanup, dpsa->dpsa_props);
err = zcp_sync_task(state, zcp_synctask_inherit_prop_check,
zcp_synctask_inherit_prop_sync, &zipa, sync, dsname);
zcp_deregister_cleanup(state, zch);
fnvlist_free(dpsa->dpsa_props);
return (err);
}
static int zcp_synctask_bookmark(lua_State *, boolean_t, nvlist_t *);
-static zcp_synctask_info_t zcp_synctask_bookmark_info = {
+static const zcp_synctask_info_t zcp_synctask_bookmark_info = {
.name = "bookmark",
.func = zcp_synctask_bookmark,
.pargs = {
- {.za_name = "snapshot | bookmark", .za_lua_type = LUA_TSTRING},
- {.za_name = "bookmark", .za_lua_type = LUA_TSTRING},
+ {.za_name = "snapshot | bookmark", .za_lua_type = LUA_TSTRING },
+ {.za_name = "bookmark", .za_lua_type = LUA_TSTRING },
{NULL, 0}
},
.kwargs = {
{NULL, 0}
},
.space_check = ZFS_SPACE_CHECK_NORMAL,
.blocks_modified = 1,
};
static int
zcp_synctask_bookmark(lua_State *state, boolean_t sync, nvlist_t *err_details)
{
(void) err_details;
int err;
const char *source = lua_tostring(state, 1);
const char *new = lua_tostring(state, 2);
nvlist_t *bmarks = fnvlist_alloc();
fnvlist_add_string(bmarks, new, source);
zcp_cleanup_handler_t *zch = zcp_register_cleanup(state,
zcp_synctask_cleanup, bmarks);
dsl_bookmark_create_arg_t dbca = {
.dbca_bmarks = bmarks,
.dbca_errors = NULL,
};
err = zcp_sync_task(state, dsl_bookmark_create_check,
dsl_bookmark_create_sync, &dbca, sync, source);
zcp_deregister_cleanup(state, zch);
fnvlist_free(bmarks);
return (err);
}
static int zcp_synctask_set_prop(lua_State *, boolean_t, nvlist_t *err_details);
-static zcp_synctask_info_t zcp_synctask_set_prop_info = {
+static const zcp_synctask_info_t zcp_synctask_set_prop_info = {
.name = "set_prop",
.func = zcp_synctask_set_prop,
.space_check = ZFS_SPACE_CHECK_RESERVED,
.blocks_modified = 2,
.pargs = {
- { .za_name = "dataset", .za_lua_type = LUA_TSTRING},
- { .za_name = "property", .za_lua_type = LUA_TSTRING},
- { .za_name = "value", .za_lua_type = LUA_TSTRING},
+ { .za_name = "dataset", .za_lua_type = LUA_TSTRING },
+ { .za_name = "property", .za_lua_type = LUA_TSTRING },
+ { .za_name = "value", .za_lua_type = LUA_TSTRING },
{ NULL, 0 }
},
.kwargs = {
{ NULL, 0 }
}
};
static int
zcp_synctask_set_prop(lua_State *state, boolean_t sync, nvlist_t *err_details)
{
(void) err_details;
int err;
zcp_set_prop_arg_t args = { 0 };
const char *dsname = lua_tostring(state, 1);
const char *prop = lua_tostring(state, 2);
const char *val = lua_tostring(state, 3);
args.state = state;
args.dsname = dsname;
args.prop = prop;
args.val = val;
err = zcp_sync_task(state, zcp_set_prop_check, zcp_set_prop_sync,
&args, sync, dsname);
return (err);
}
static int
zcp_synctask_wrapper(lua_State *state)
{
int err;
zcp_cleanup_handler_t *zch;
int num_ret = 1;
nvlist_t *err_details = fnvlist_alloc();
/*
* Make sure err_details is properly freed, even if a fatal error is
* thrown during the synctask.
*/
zch = zcp_register_cleanup(state, zcp_synctask_cleanup, err_details);
zcp_synctask_info_t *info = lua_touserdata(state, lua_upvalueindex(1));
boolean_t sync = lua_toboolean(state, lua_upvalueindex(2));
zcp_run_info_t *ri = zcp_run_info(state);
dsl_pool_t *dp = ri->zri_pool;
/* MOS space is triple-dittoed, so we multiply by 3. */
uint64_t funcspace =
((uint64_t)info->blocks_modified << DST_AVG_BLKSHIFT) * 3;
zcp_parse_args(state, info->name, info->pargs, info->kwargs);
err = 0;
if (info->space_check != ZFS_SPACE_CHECK_NONE) {
uint64_t quota = dsl_pool_unreserved_space(dp,
info->space_check);
uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes +
ri->zri_space_used;
if (used + funcspace > quota) {
err = SET_ERROR(ENOSPC);
}
}
if (err == 0) {
err = info->func(state, sync, err_details);
}
if (err == 0) {
ri->zri_space_used += funcspace;
}
lua_pushnumber(state, (lua_Number)err);
if (fnvlist_num_pairs(err_details) > 0) {
(void) zcp_nvlist_to_lua(state, err_details, NULL, 0);
num_ret++;
}
zcp_deregister_cleanup(state, zch);
fnvlist_free(err_details);
return (num_ret);
}
int
zcp_load_synctask_lib(lua_State *state, boolean_t sync)
{
- int i;
- zcp_synctask_info_t *zcp_synctask_funcs[] = {
+ const zcp_synctask_info_t *zcp_synctask_funcs[] = {
&zcp_synctask_destroy_info,
&zcp_synctask_promote_info,
&zcp_synctask_rollback_info,
&zcp_synctask_snapshot_info,
&zcp_synctask_inherit_prop_info,
&zcp_synctask_bookmark_info,
&zcp_synctask_set_prop_info,
NULL
};
lua_newtable(state);
- for (i = 0; zcp_synctask_funcs[i] != NULL; i++) {
- zcp_synctask_info_t *info = zcp_synctask_funcs[i];
- lua_pushlightuserdata(state, info);
+ for (int i = 0; zcp_synctask_funcs[i] != NULL; i++) {
+ const zcp_synctask_info_t *info = zcp_synctask_funcs[i];
+ lua_pushlightuserdata(state, (void *)(uintptr_t)info);
lua_pushboolean(state, sync);
lua_pushcclosure(state, &zcp_synctask_wrapper, 2);
lua_setfield(state, -2, info->name);
- info++;
}
return (1);
}
diff --git a/module/zfs/zfs_fm.c b/module/zfs/zfs_fm.c
index a05ff7330458..828385b430b4 100644
--- a/module/zfs/zfs_fm.c
+++ b/module/zfs/zfs_fm.c
@@ -1,1521 +1,1521 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright (c) 2012,2021 by Delphix. All rights reserved.
*/
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/vdev.h>
#include <sys/vdev_impl.h>
#include <sys/zio.h>
#include <sys/zio_checksum.h>
#include <sys/fm/fs/zfs.h>
#include <sys/fm/protocol.h>
#include <sys/fm/util.h>
#include <sys/sysevent.h>
/*
* This general routine is responsible for generating all the different ZFS
* ereports. The payload is dependent on the class, and which arguments are
* supplied to the function:
*
* EREPORT POOL VDEV IO
* block X X X
* data X X
* device X X
* pool X
*
* If we are in a loading state, all errors are chained together by the same
* SPA-wide ENA (Error Numeric Association).
*
* For isolated I/O requests, we get the ENA from the zio_t. The propagation
* gets very complicated due to RAID-Z, gang blocks, and vdev caching. We want
* to chain together all ereports associated with a logical piece of data. For
* read I/Os, there are basically three 'types' of I/O, which form a roughly
* layered diagram:
*
* +---------------+
* | Aggregate I/O | No associated logical data or device
* +---------------+
* |
* V
* +---------------+ Reads associated with a piece of logical data.
* | Read I/O | This includes reads on behalf of RAID-Z,
* +---------------+ mirrors, gang blocks, retries, etc.
* |
* V
* +---------------+ Reads associated with a particular device, but
* | Physical I/O | no logical data. Issued as part of vdev caching
* +---------------+ and I/O aggregation.
*
* Note that 'physical I/O' here is not the same terminology as used in the rest
* of ZIO. Typically, 'physical I/O' simply means that there is no attached
* blockpointer. But I/O with no associated block pointer can still be related
* to a logical piece of data (i.e. RAID-Z requests).
*
* Purely physical I/O always have unique ENAs. They are not related to a
* particular piece of logical data, and therefore cannot be chained together.
* We still generate an ereport, but the DE doesn't correlate it with any
* logical piece of data. When such an I/O fails, the delegated I/O requests
* will issue a retry, which will trigger the 'real' ereport with the correct
* ENA.
*
* We keep track of the ENA for a ZIO chain through the 'io_logical' member.
* When a new logical I/O is issued, we set this to point to itself. Child I/Os
* then inherit this pointer, so that when it is first set subsequent failures
* will use the same ENA. For vdev cache fill and queue aggregation I/O,
* this pointer is set to NULL, and no ereport will be generated (since it
* doesn't actually correspond to any particular device or piece of data,
* and the caller will always retry without caching or queueing anyway).
*
* For checksum errors, we want to include more information about the actual
* error which occurs. Accordingly, we build an ereport when the error is
* noticed, but instead of sending it in immediately, we hang it off of the
* io_cksum_report field of the logical IO. When the logical IO completes
* (successfully or not), zfs_ereport_finish_checksum() is called with the
* good and bad versions of the buffer (if available), and we annotate the
* ereport with information about the differences.
*/
#ifdef _KERNEL
/*
* Duplicate ereport Detection
*
* Some ereports are retained momentarily for detecting duplicates. These
* are kept in a recent_events_node_t in both a time-ordered list and an AVL
* tree of recent unique ereports.
*
* The lifespan of these recent ereports is bounded (15 mins) and a cleaner
* task is used to purge stale entries.
*/
static list_t recent_events_list;
static avl_tree_t recent_events_tree;
static kmutex_t recent_events_lock;
static taskqid_t recent_events_cleaner_tqid;
/*
* Each node is about 128 bytes so 2,000 would consume 1/4 MiB.
*
* This setting can be changed dynamically and setting it to zero
* disables duplicate detection.
*/
-unsigned int zfs_zevent_retain_max = 2000;
+static unsigned int zfs_zevent_retain_max = 2000;
/*
* The lifespan for a recent ereport entry. The default of 15 minutes is
* intended to outlive the zfs diagnosis engine's threshold of 10 errors
* over a period of 10 minutes.
*/
-unsigned int zfs_zevent_retain_expire_secs = 900;
+static unsigned int zfs_zevent_retain_expire_secs = 900;
typedef enum zfs_subclass {
ZSC_IO,
ZSC_DATA,
ZSC_CHECKSUM
} zfs_subclass_t;
typedef struct {
/* common criteria */
uint64_t re_pool_guid;
uint64_t re_vdev_guid;
int re_io_error;
uint64_t re_io_size;
uint64_t re_io_offset;
zfs_subclass_t re_subclass;
zio_priority_t re_io_priority;
/* logical zio criteria (optional) */
zbookmark_phys_t re_io_bookmark;
/* internal state */
avl_node_t re_tree_link;
list_node_t re_list_link;
uint64_t re_timestamp;
} recent_events_node_t;
static int
recent_events_compare(const void *a, const void *b)
{
const recent_events_node_t *node1 = a;
const recent_events_node_t *node2 = b;
int cmp;
/*
* The comparison order here is somewhat arbitrary.
* What's important is that if every criteria matches, then it
* is a duplicate (i.e. compare returns 0)
*/
if ((cmp = TREE_CMP(node1->re_subclass, node2->re_subclass)) != 0)
return (cmp);
if ((cmp = TREE_CMP(node1->re_pool_guid, node2->re_pool_guid)) != 0)
return (cmp);
if ((cmp = TREE_CMP(node1->re_vdev_guid, node2->re_vdev_guid)) != 0)
return (cmp);
if ((cmp = TREE_CMP(node1->re_io_error, node2->re_io_error)) != 0)
return (cmp);
if ((cmp = TREE_CMP(node1->re_io_priority, node2->re_io_priority)) != 0)
return (cmp);
if ((cmp = TREE_CMP(node1->re_io_size, node2->re_io_size)) != 0)
return (cmp);
if ((cmp = TREE_CMP(node1->re_io_offset, node2->re_io_offset)) != 0)
return (cmp);
const zbookmark_phys_t *zb1 = &node1->re_io_bookmark;
const zbookmark_phys_t *zb2 = &node2->re_io_bookmark;
if ((cmp = TREE_CMP(zb1->zb_objset, zb2->zb_objset)) != 0)
return (cmp);
if ((cmp = TREE_CMP(zb1->zb_object, zb2->zb_object)) != 0)
return (cmp);
if ((cmp = TREE_CMP(zb1->zb_level, zb2->zb_level)) != 0)
return (cmp);
if ((cmp = TREE_CMP(zb1->zb_blkid, zb2->zb_blkid)) != 0)
return (cmp);
return (0);
}
static void zfs_ereport_schedule_cleaner(void);
/*
* background task to clean stale recent event nodes.
*/
static void
zfs_ereport_cleaner(void *arg)
{
recent_events_node_t *entry;
uint64_t now = gethrtime();
/*
* purge expired entries
*/
mutex_enter(&recent_events_lock);
while ((entry = list_tail(&recent_events_list)) != NULL) {
uint64_t age = NSEC2SEC(now - entry->re_timestamp);
if (age <= zfs_zevent_retain_expire_secs)
break;
/* remove expired node */
avl_remove(&recent_events_tree, entry);
list_remove(&recent_events_list, entry);
kmem_free(entry, sizeof (*entry));
}
/* Restart the cleaner if more entries remain */
recent_events_cleaner_tqid = 0;
if (!list_is_empty(&recent_events_list))
zfs_ereport_schedule_cleaner();
mutex_exit(&recent_events_lock);
}
static void
zfs_ereport_schedule_cleaner(void)
{
ASSERT(MUTEX_HELD(&recent_events_lock));
uint64_t timeout = SEC2NSEC(zfs_zevent_retain_expire_secs + 1);
recent_events_cleaner_tqid = taskq_dispatch_delay(
system_delay_taskq, zfs_ereport_cleaner, NULL, TQ_SLEEP,
ddi_get_lbolt() + NSEC_TO_TICK(timeout));
}
/*
* Clear entries for a given vdev or all vdevs in a pool when vdev == NULL
*/
void
zfs_ereport_clear(spa_t *spa, vdev_t *vd)
{
uint64_t vdev_guid, pool_guid;
int cnt = 0;
ASSERT(vd != NULL || spa != NULL);
if (vd == NULL) {
vdev_guid = 0;
pool_guid = spa_guid(spa);
} else {
vdev_guid = vd->vdev_guid;
pool_guid = 0;
}
mutex_enter(&recent_events_lock);
recent_events_node_t *next = list_head(&recent_events_list);
while (next != NULL) {
recent_events_node_t *entry = next;
next = list_next(&recent_events_list, next);
if (entry->re_vdev_guid == vdev_guid ||
entry->re_pool_guid == pool_guid) {
avl_remove(&recent_events_tree, entry);
list_remove(&recent_events_list, entry);
kmem_free(entry, sizeof (*entry));
cnt++;
}
}
mutex_exit(&recent_events_lock);
}
/*
* Check if an ereport would be a duplicate of one recently posted.
*
* An ereport is considered a duplicate if the set of criteria in
* recent_events_node_t all match.
*
* Only FM_EREPORT_ZFS_IO, FM_EREPORT_ZFS_DATA, and FM_EREPORT_ZFS_CHECKSUM
* are candidates for duplicate checking.
*/
static boolean_t
zfs_ereport_is_duplicate(const char *subclass, spa_t *spa, vdev_t *vd,
const zbookmark_phys_t *zb, zio_t *zio, uint64_t offset, uint64_t size)
{
recent_events_node_t search = {0}, *entry;
if (vd == NULL || zio == NULL)
return (B_FALSE);
if (zfs_zevent_retain_max == 0)
return (B_FALSE);
if (strcmp(subclass, FM_EREPORT_ZFS_IO) == 0)
search.re_subclass = ZSC_IO;
else if (strcmp(subclass, FM_EREPORT_ZFS_DATA) == 0)
search.re_subclass = ZSC_DATA;
else if (strcmp(subclass, FM_EREPORT_ZFS_CHECKSUM) == 0)
search.re_subclass = ZSC_CHECKSUM;
else
return (B_FALSE);
search.re_pool_guid = spa_guid(spa);
search.re_vdev_guid = vd->vdev_guid;
search.re_io_error = zio->io_error;
search.re_io_priority = zio->io_priority;
/* if size is supplied use it over what's in zio */
if (size) {
search.re_io_size = size;
search.re_io_offset = offset;
} else {
search.re_io_size = zio->io_size;
search.re_io_offset = zio->io_offset;
}
/* grab optional logical zio criteria */
if (zb != NULL) {
search.re_io_bookmark.zb_objset = zb->zb_objset;
search.re_io_bookmark.zb_object = zb->zb_object;
search.re_io_bookmark.zb_level = zb->zb_level;
search.re_io_bookmark.zb_blkid = zb->zb_blkid;
}
uint64_t now = gethrtime();
mutex_enter(&recent_events_lock);
/* check if we have seen this one recently */
entry = avl_find(&recent_events_tree, &search, NULL);
if (entry != NULL) {
uint64_t age = NSEC2SEC(now - entry->re_timestamp);
/*
* There is still an active cleaner (since we're here).
* Reset the last seen time for this duplicate entry
* so that its lifespand gets extended.
*/
list_remove(&recent_events_list, entry);
list_insert_head(&recent_events_list, entry);
entry->re_timestamp = now;
zfs_zevent_track_duplicate();
mutex_exit(&recent_events_lock);
return (age <= zfs_zevent_retain_expire_secs);
}
if (avl_numnodes(&recent_events_tree) >= zfs_zevent_retain_max) {
/* recycle oldest node */
entry = list_tail(&recent_events_list);
ASSERT(entry != NULL);
list_remove(&recent_events_list, entry);
avl_remove(&recent_events_tree, entry);
} else {
entry = kmem_alloc(sizeof (recent_events_node_t), KM_SLEEP);
}
/* record this as a recent ereport */
*entry = search;
avl_add(&recent_events_tree, entry);
list_insert_head(&recent_events_list, entry);
entry->re_timestamp = now;
/* Start a cleaner if not already scheduled */
if (recent_events_cleaner_tqid == 0)
zfs_ereport_schedule_cleaner();
mutex_exit(&recent_events_lock);
return (B_FALSE);
}
void
zfs_zevent_post_cb(nvlist_t *nvl, nvlist_t *detector)
{
if (nvl)
fm_nvlist_destroy(nvl, FM_NVA_FREE);
if (detector)
fm_nvlist_destroy(detector, FM_NVA_FREE);
}
/*
* We want to rate limit ZIO delay, deadman, and checksum events so as to not
* flood zevent consumers when a disk is acting up.
*
* Returns 1 if we're ratelimiting, 0 if not.
*/
static int
zfs_is_ratelimiting_event(const char *subclass, vdev_t *vd)
{
int rc = 0;
/*
* zfs_ratelimit() returns 1 if we're *not* ratelimiting and 0 if we
* are. Invert it to get our return value.
*/
if (strcmp(subclass, FM_EREPORT_ZFS_DELAY) == 0) {
rc = !zfs_ratelimit(&vd->vdev_delay_rl);
} else if (strcmp(subclass, FM_EREPORT_ZFS_DEADMAN) == 0) {
rc = !zfs_ratelimit(&vd->vdev_deadman_rl);
} else if (strcmp(subclass, FM_EREPORT_ZFS_CHECKSUM) == 0) {
rc = !zfs_ratelimit(&vd->vdev_checksum_rl);
}
if (rc) {
/* We're rate limiting */
fm_erpt_dropped_increment();
}
return (rc);
}
/*
* Return B_TRUE if the event actually posted, B_FALSE if not.
*/
static boolean_t
zfs_ereport_start(nvlist_t **ereport_out, nvlist_t **detector_out,
const char *subclass, spa_t *spa, vdev_t *vd, const zbookmark_phys_t *zb,
zio_t *zio, uint64_t stateoroffset, uint64_t size)
{
nvlist_t *ereport, *detector;
uint64_t ena;
char class[64];
if ((ereport = fm_nvlist_create(NULL)) == NULL)
return (B_FALSE);
if ((detector = fm_nvlist_create(NULL)) == NULL) {
fm_nvlist_destroy(ereport, FM_NVA_FREE);
return (B_FALSE);
}
/*
* Serialize ereport generation
*/
mutex_enter(&spa->spa_errlist_lock);
/*
* Determine the ENA to use for this event. If we are in a loading
* state, use a SPA-wide ENA. Otherwise, if we are in an I/O state, use
* a root zio-wide ENA. Otherwise, simply use a unique ENA.
*/
if (spa_load_state(spa) != SPA_LOAD_NONE) {
if (spa->spa_ena == 0)
spa->spa_ena = fm_ena_generate(0, FM_ENA_FMT1);
ena = spa->spa_ena;
} else if (zio != NULL && zio->io_logical != NULL) {
if (zio->io_logical->io_ena == 0)
zio->io_logical->io_ena =
fm_ena_generate(0, FM_ENA_FMT1);
ena = zio->io_logical->io_ena;
} else {
ena = fm_ena_generate(0, FM_ENA_FMT1);
}
/*
* Construct the full class, detector, and other standard FMA fields.
*/
(void) snprintf(class, sizeof (class), "%s.%s",
ZFS_ERROR_CLASS, subclass);
fm_fmri_zfs_set(detector, FM_ZFS_SCHEME_VERSION, spa_guid(spa),
vd != NULL ? vd->vdev_guid : 0);
fm_ereport_set(ereport, FM_EREPORT_VERSION, class, ena, detector, NULL);
/*
* Construct the per-ereport payload, depending on which parameters are
* passed in.
*/
/*
* Generic payload members common to all ereports.
*/
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_POOL, DATA_TYPE_STRING, spa_name(spa),
FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, DATA_TYPE_UINT64, spa_guid(spa),
FM_EREPORT_PAYLOAD_ZFS_POOL_STATE, DATA_TYPE_UINT64,
(uint64_t)spa_state(spa),
FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, DATA_TYPE_INT32,
(int32_t)spa_load_state(spa), NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE,
DATA_TYPE_STRING,
spa_get_failmode(spa) == ZIO_FAILURE_MODE_WAIT ?
FM_EREPORT_FAILMODE_WAIT :
spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE ?
FM_EREPORT_FAILMODE_CONTINUE : FM_EREPORT_FAILMODE_PANIC,
NULL);
if (vd != NULL) {
vdev_t *pvd = vd->vdev_parent;
vdev_queue_t *vq = &vd->vdev_queue;
vdev_stat_t *vs = &vd->vdev_stat;
vdev_t *spare_vd;
uint64_t *spare_guids;
char **spare_paths;
int i, spare_count;
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID,
DATA_TYPE_UINT64, vd->vdev_guid,
FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
DATA_TYPE_STRING, vd->vdev_ops->vdev_op_type, NULL);
if (vd->vdev_path != NULL)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH,
DATA_TYPE_STRING, vd->vdev_path, NULL);
if (vd->vdev_devid != NULL)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID,
DATA_TYPE_STRING, vd->vdev_devid, NULL);
if (vd->vdev_fru != NULL)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU,
DATA_TYPE_STRING, vd->vdev_fru, NULL);
if (vd->vdev_enc_sysfs_path != NULL)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH,
DATA_TYPE_STRING, vd->vdev_enc_sysfs_path, NULL);
if (vd->vdev_ashift)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_ASHIFT,
DATA_TYPE_UINT64, vd->vdev_ashift, NULL);
if (vq != NULL) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_COMP_TS,
DATA_TYPE_UINT64, vq->vq_io_complete_ts, NULL);
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_DELTA_TS,
DATA_TYPE_UINT64, vq->vq_io_delta_ts, NULL);
}
if (vs != NULL) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_READ_ERRORS,
DATA_TYPE_UINT64, vs->vs_read_errors,
FM_EREPORT_PAYLOAD_ZFS_VDEV_WRITE_ERRORS,
DATA_TYPE_UINT64, vs->vs_write_errors,
FM_EREPORT_PAYLOAD_ZFS_VDEV_CKSUM_ERRORS,
DATA_TYPE_UINT64, vs->vs_checksum_errors,
FM_EREPORT_PAYLOAD_ZFS_VDEV_DELAYS,
DATA_TYPE_UINT64, vs->vs_slow_ios,
NULL);
}
if (pvd != NULL) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_PARENT_GUID,
DATA_TYPE_UINT64, pvd->vdev_guid,
FM_EREPORT_PAYLOAD_ZFS_PARENT_TYPE,
DATA_TYPE_STRING, pvd->vdev_ops->vdev_op_type,
NULL);
if (pvd->vdev_path)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_PARENT_PATH,
DATA_TYPE_STRING, pvd->vdev_path, NULL);
if (pvd->vdev_devid)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_PARENT_DEVID,
DATA_TYPE_STRING, pvd->vdev_devid, NULL);
}
spare_count = spa->spa_spares.sav_count;
spare_paths = kmem_zalloc(sizeof (char *) * spare_count,
KM_SLEEP);
spare_guids = kmem_zalloc(sizeof (uint64_t) * spare_count,
KM_SLEEP);
for (i = 0; i < spare_count; i++) {
spare_vd = spa->spa_spares.sav_vdevs[i];
if (spare_vd) {
spare_paths[i] = spare_vd->vdev_path;
spare_guids[i] = spare_vd->vdev_guid;
}
}
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_SPARE_PATHS,
DATA_TYPE_STRING_ARRAY, spare_count, spare_paths,
FM_EREPORT_PAYLOAD_ZFS_VDEV_SPARE_GUIDS,
DATA_TYPE_UINT64_ARRAY, spare_count, spare_guids, NULL);
kmem_free(spare_guids, sizeof (uint64_t) * spare_count);
kmem_free(spare_paths, sizeof (char *) * spare_count);
}
if (zio != NULL) {
/*
* Payload common to all I/Os.
*/
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_ERR,
DATA_TYPE_INT32, zio->io_error, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_FLAGS,
DATA_TYPE_INT32, zio->io_flags, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_STAGE,
DATA_TYPE_UINT32, zio->io_stage, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_PIPELINE,
DATA_TYPE_UINT32, zio->io_pipeline, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELAY,
DATA_TYPE_UINT64, zio->io_delay, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_TIMESTAMP,
DATA_TYPE_UINT64, zio->io_timestamp, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELTA,
DATA_TYPE_UINT64, zio->io_delta, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_PRIORITY,
DATA_TYPE_UINT32, zio->io_priority, NULL);
/*
* If the 'size' parameter is non-zero, it indicates this is a
* RAID-Z or other I/O where the physical offset and length are
* provided for us, instead of within the zio_t.
*/
if (vd != NULL) {
if (size)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
DATA_TYPE_UINT64, stateoroffset,
FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
DATA_TYPE_UINT64, size, NULL);
else
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
DATA_TYPE_UINT64, zio->io_offset,
FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
DATA_TYPE_UINT64, zio->io_size, NULL);
}
} else if (vd != NULL) {
/*
* If we have a vdev but no zio, this is a device fault, and the
* 'stateoroffset' parameter indicates the previous state of the
* vdev.
*/
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_PREV_STATE,
DATA_TYPE_UINT64, stateoroffset, NULL);
}
/*
* Payload for I/Os with corresponding logical information.
*/
if (zb != NULL && (zio == NULL || zio->io_logical != NULL)) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJSET,
DATA_TYPE_UINT64, zb->zb_objset,
FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJECT,
DATA_TYPE_UINT64, zb->zb_object,
FM_EREPORT_PAYLOAD_ZFS_ZIO_LEVEL,
DATA_TYPE_INT64, zb->zb_level,
FM_EREPORT_PAYLOAD_ZFS_ZIO_BLKID,
DATA_TYPE_UINT64, zb->zb_blkid, NULL);
}
mutex_exit(&spa->spa_errlist_lock);
*ereport_out = ereport;
*detector_out = detector;
return (B_TRUE);
}
/* if it's <= 128 bytes, save the corruption directly */
#define ZFM_MAX_INLINE (128 / sizeof (uint64_t))
#define MAX_RANGES 16
typedef struct zfs_ecksum_info {
/* histograms of set and cleared bits by bit number in a 64-bit word */
uint32_t zei_histogram_set[sizeof (uint64_t) * NBBY];
uint32_t zei_histogram_cleared[sizeof (uint64_t) * NBBY];
/* inline arrays of bits set and cleared. */
uint64_t zei_bits_set[ZFM_MAX_INLINE];
uint64_t zei_bits_cleared[ZFM_MAX_INLINE];
/*
* for each range, the number of bits set and cleared. The Hamming
* distance between the good and bad buffers is the sum of them all.
*/
uint32_t zei_range_sets[MAX_RANGES];
uint32_t zei_range_clears[MAX_RANGES];
struct zei_ranges {
uint32_t zr_start;
uint32_t zr_end;
} zei_ranges[MAX_RANGES];
size_t zei_range_count;
uint32_t zei_mingap;
uint32_t zei_allowed_mingap;
} zfs_ecksum_info_t;
static void
update_histogram(uint64_t value_arg, uint32_t *hist, uint32_t *count)
{
size_t i;
size_t bits = 0;
uint64_t value = BE_64(value_arg);
/* We store the bits in big-endian (largest-first) order */
for (i = 0; i < 64; i++) {
if (value & (1ull << i)) {
hist[63 - i]++;
++bits;
}
}
/* update the count of bits changed */
*count += bits;
}
/*
* We've now filled up the range array, and need to increase "mingap" and
* shrink the range list accordingly. zei_mingap is always the smallest
* distance between array entries, so we set the new_allowed_gap to be
* one greater than that. We then go through the list, joining together
* any ranges which are closer than the new_allowed_gap.
*
* By construction, there will be at least one. We also update zei_mingap
* to the new smallest gap, to prepare for our next invocation.
*/
static void
zei_shrink_ranges(zfs_ecksum_info_t *eip)
{
uint32_t mingap = UINT32_MAX;
uint32_t new_allowed_gap = eip->zei_mingap + 1;
size_t idx, output;
size_t max = eip->zei_range_count;
struct zei_ranges *r = eip->zei_ranges;
ASSERT3U(eip->zei_range_count, >, 0);
ASSERT3U(eip->zei_range_count, <=, MAX_RANGES);
output = idx = 0;
while (idx < max - 1) {
uint32_t start = r[idx].zr_start;
uint32_t end = r[idx].zr_end;
while (idx < max - 1) {
idx++;
uint32_t nstart = r[idx].zr_start;
uint32_t nend = r[idx].zr_end;
uint32_t gap = nstart - end;
if (gap < new_allowed_gap) {
end = nend;
continue;
}
if (gap < mingap)
mingap = gap;
break;
}
r[output].zr_start = start;
r[output].zr_end = end;
output++;
}
ASSERT3U(output, <, eip->zei_range_count);
eip->zei_range_count = output;
eip->zei_mingap = mingap;
eip->zei_allowed_mingap = new_allowed_gap;
}
static void
zei_add_range(zfs_ecksum_info_t *eip, int start, int end)
{
struct zei_ranges *r = eip->zei_ranges;
size_t count = eip->zei_range_count;
if (count >= MAX_RANGES) {
zei_shrink_ranges(eip);
count = eip->zei_range_count;
}
if (count == 0) {
eip->zei_mingap = UINT32_MAX;
eip->zei_allowed_mingap = 1;
} else {
int gap = start - r[count - 1].zr_end;
if (gap < eip->zei_allowed_mingap) {
r[count - 1].zr_end = end;
return;
}
if (gap < eip->zei_mingap)
eip->zei_mingap = gap;
}
r[count].zr_start = start;
r[count].zr_end = end;
eip->zei_range_count++;
}
static size_t
zei_range_total_size(zfs_ecksum_info_t *eip)
{
struct zei_ranges *r = eip->zei_ranges;
size_t count = eip->zei_range_count;
size_t result = 0;
size_t idx;
for (idx = 0; idx < count; idx++)
result += (r[idx].zr_end - r[idx].zr_start);
return (result);
}
static zfs_ecksum_info_t *
annotate_ecksum(nvlist_t *ereport, zio_bad_cksum_t *info,
const abd_t *goodabd, const abd_t *badabd, size_t size,
boolean_t drop_if_identical)
{
const uint64_t *good;
const uint64_t *bad;
size_t nui64s = size / sizeof (uint64_t);
size_t inline_size;
int no_inline = 0;
size_t idx;
size_t range;
size_t offset = 0;
ssize_t start = -1;
zfs_ecksum_info_t *eip = kmem_zalloc(sizeof (*eip), KM_SLEEP);
/* don't do any annotation for injected checksum errors */
if (info != NULL && info->zbc_injected)
return (eip);
if (info != NULL && info->zbc_has_cksum) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_CKSUM_EXPECTED,
DATA_TYPE_UINT64_ARRAY,
sizeof (info->zbc_expected) / sizeof (uint64_t),
(uint64_t *)&info->zbc_expected,
FM_EREPORT_PAYLOAD_ZFS_CKSUM_ACTUAL,
DATA_TYPE_UINT64_ARRAY,
sizeof (info->zbc_actual) / sizeof (uint64_t),
(uint64_t *)&info->zbc_actual,
FM_EREPORT_PAYLOAD_ZFS_CKSUM_ALGO,
DATA_TYPE_STRING,
info->zbc_checksum_name,
NULL);
if (info->zbc_byteswapped) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_CKSUM_BYTESWAP,
DATA_TYPE_BOOLEAN, 1,
NULL);
}
}
if (badabd == NULL || goodabd == NULL)
return (eip);
ASSERT3U(nui64s, <=, UINT32_MAX);
ASSERT3U(size, ==, nui64s * sizeof (uint64_t));
ASSERT3U(size, <=, SPA_MAXBLOCKSIZE);
ASSERT3U(size, <=, UINT32_MAX);
good = (const uint64_t *) abd_borrow_buf_copy((abd_t *)goodabd, size);
bad = (const uint64_t *) abd_borrow_buf_copy((abd_t *)badabd, size);
/* build up the range list by comparing the two buffers. */
for (idx = 0; idx < nui64s; idx++) {
if (good[idx] == bad[idx]) {
if (start == -1)
continue;
zei_add_range(eip, start, idx);
start = -1;
} else {
if (start != -1)
continue;
start = idx;
}
}
if (start != -1)
zei_add_range(eip, start, idx);
/* See if it will fit in our inline buffers */
inline_size = zei_range_total_size(eip);
if (inline_size > ZFM_MAX_INLINE)
no_inline = 1;
/*
* If there is no change and we want to drop if the buffers are
* identical, do so.
*/
if (inline_size == 0 && drop_if_identical) {
kmem_free(eip, sizeof (*eip));
abd_return_buf((abd_t *)goodabd, (void *)good, size);
abd_return_buf((abd_t *)badabd, (void *)bad, size);
return (NULL);
}
/*
* Now walk through the ranges, filling in the details of the
* differences. Also convert our uint64_t-array offsets to byte
* offsets.
*/
for (range = 0; range < eip->zei_range_count; range++) {
size_t start = eip->zei_ranges[range].zr_start;
size_t end = eip->zei_ranges[range].zr_end;
for (idx = start; idx < end; idx++) {
uint64_t set, cleared;
// bits set in bad, but not in good
set = ((~good[idx]) & bad[idx]);
// bits set in good, but not in bad
cleared = (good[idx] & (~bad[idx]));
if (!no_inline) {
ASSERT3U(offset, <, inline_size);
eip->zei_bits_set[offset] = set;
eip->zei_bits_cleared[offset] = cleared;
offset++;
}
update_histogram(set, eip->zei_histogram_set,
&eip->zei_range_sets[range]);
update_histogram(cleared, eip->zei_histogram_cleared,
&eip->zei_range_clears[range]);
}
/* convert to byte offsets */
eip->zei_ranges[range].zr_start *= sizeof (uint64_t);
eip->zei_ranges[range].zr_end *= sizeof (uint64_t);
}
abd_return_buf((abd_t *)goodabd, (void *)good, size);
abd_return_buf((abd_t *)badabd, (void *)bad, size);
eip->zei_allowed_mingap *= sizeof (uint64_t);
inline_size *= sizeof (uint64_t);
/* fill in ereport */
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_BAD_OFFSET_RANGES,
DATA_TYPE_UINT32_ARRAY, 2 * eip->zei_range_count,
(uint32_t *)eip->zei_ranges,
FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_MIN_GAP,
DATA_TYPE_UINT32, eip->zei_allowed_mingap,
FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_SETS,
DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_sets,
FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_CLEARS,
DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_clears,
NULL);
if (!no_inline) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_BAD_SET_BITS,
DATA_TYPE_UINT8_ARRAY,
inline_size, (uint8_t *)eip->zei_bits_set,
FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_BITS,
DATA_TYPE_UINT8_ARRAY,
inline_size, (uint8_t *)eip->zei_bits_cleared,
NULL);
} else {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_BAD_SET_HISTOGRAM,
DATA_TYPE_UINT32_ARRAY,
NBBY * sizeof (uint64_t), eip->zei_histogram_set,
FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_HISTOGRAM,
DATA_TYPE_UINT32_ARRAY,
NBBY * sizeof (uint64_t), eip->zei_histogram_cleared,
NULL);
}
return (eip);
}
#else
void
zfs_ereport_clear(spa_t *spa, vdev_t *vd)
{
(void) spa, (void) vd;
}
#endif
/*
* Make sure our event is still valid for the given zio/vdev/pool. For example,
* we don't want to keep logging events for a faulted or missing vdev.
*/
boolean_t
zfs_ereport_is_valid(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio)
{
#ifdef _KERNEL
/*
* If we are doing a spa_tryimport() or in recovery mode,
* ignore errors.
*/
if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT ||
spa_load_state(spa) == SPA_LOAD_RECOVER)
return (B_FALSE);
/*
* If we are in the middle of opening a pool, and the previous attempt
* failed, don't bother logging any new ereports - we're just going to
* get the same diagnosis anyway.
*/
if (spa_load_state(spa) != SPA_LOAD_NONE &&
spa->spa_last_open_failed)
return (B_FALSE);
if (zio != NULL) {
/*
* If this is not a read or write zio, ignore the error. This
* can occur if the DKIOCFLUSHWRITECACHE ioctl fails.
*/
if (zio->io_type != ZIO_TYPE_READ &&
zio->io_type != ZIO_TYPE_WRITE)
return (B_FALSE);
if (vd != NULL) {
/*
* If the vdev has already been marked as failing due
* to a failed probe, then ignore any subsequent I/O
* errors, as the DE will automatically fault the vdev
* on the first such failure. This also catches cases
* where vdev_remove_wanted is set and the device has
* not yet been asynchronously placed into the REMOVED
* state.
*/
if (zio->io_vd == vd && !vdev_accessible(vd, zio))
return (B_FALSE);
/*
* Ignore checksum errors for reads from DTL regions of
* leaf vdevs.
*/
if (zio->io_type == ZIO_TYPE_READ &&
zio->io_error == ECKSUM &&
vd->vdev_ops->vdev_op_leaf &&
vdev_dtl_contains(vd, DTL_MISSING, zio->io_txg, 1))
return (B_FALSE);
}
}
/*
* For probe failure, we want to avoid posting ereports if we've
* already removed the device in the meantime.
*/
if (vd != NULL &&
strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) == 0 &&
(vd->vdev_remove_wanted || vd->vdev_state == VDEV_STATE_REMOVED))
return (B_FALSE);
/* Ignore bogus delay events (like from ioctls or unqueued IOs) */
if ((strcmp(subclass, FM_EREPORT_ZFS_DELAY) == 0) &&
(zio != NULL) && (!zio->io_timestamp)) {
return (B_FALSE);
}
#else
(void) subclass, (void) spa, (void) vd, (void) zio;
#endif
return (B_TRUE);
}
/*
* Post an ereport for the given subclass
*
* Returns
* - 0 if an event was posted
* - EINVAL if there was a problem posting event
* - EBUSY if the event was rate limited
* - EALREADY if the event was already posted (duplicate)
*/
int
zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd,
const zbookmark_phys_t *zb, zio_t *zio, uint64_t state)
{
int rc = 0;
#ifdef _KERNEL
nvlist_t *ereport = NULL;
nvlist_t *detector = NULL;
if (!zfs_ereport_is_valid(subclass, spa, vd, zio))
return (EINVAL);
if (zfs_ereport_is_duplicate(subclass, spa, vd, zb, zio, 0, 0))
return (SET_ERROR(EALREADY));
if (zfs_is_ratelimiting_event(subclass, vd))
return (SET_ERROR(EBUSY));
if (!zfs_ereport_start(&ereport, &detector, subclass, spa, vd,
zb, zio, state, 0))
return (SET_ERROR(EINVAL)); /* couldn't post event */
if (ereport == NULL)
return (SET_ERROR(EINVAL));
/* Cleanup is handled by the callback function */
rc = zfs_zevent_post(ereport, detector, zfs_zevent_post_cb);
#else
(void) subclass, (void) spa, (void) vd, (void) zb, (void) zio,
(void) state;
#endif
return (rc);
}
/*
* Prepare a checksum ereport
*
* Returns
* - 0 if an event was posted
* - EINVAL if there was a problem posting event
* - EBUSY if the event was rate limited
* - EALREADY if the event was already posted (duplicate)
*/
int
zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, const zbookmark_phys_t *zb,
struct zio *zio, uint64_t offset, uint64_t length, zio_bad_cksum_t *info)
{
zio_cksum_report_t *report;
#ifdef _KERNEL
if (!zfs_ereport_is_valid(FM_EREPORT_ZFS_CHECKSUM, spa, vd, zio))
return (SET_ERROR(EINVAL));
if (zfs_ereport_is_duplicate(FM_EREPORT_ZFS_CHECKSUM, spa, vd, zb, zio,
offset, length))
return (SET_ERROR(EALREADY));
if (zfs_is_ratelimiting_event(FM_EREPORT_ZFS_CHECKSUM, vd))
return (SET_ERROR(EBUSY));
#else
(void) zb, (void) offset;
#endif
report = kmem_zalloc(sizeof (*report), KM_SLEEP);
zio_vsd_default_cksum_report(zio, report);
/* copy the checksum failure information if it was provided */
if (info != NULL) {
report->zcr_ckinfo = kmem_zalloc(sizeof (*info), KM_SLEEP);
bcopy(info, report->zcr_ckinfo, sizeof (*info));
}
report->zcr_sector = 1ULL << vd->vdev_top->vdev_ashift;
report->zcr_align =
vdev_psize_to_asize(vd->vdev_top, report->zcr_sector);
report->zcr_length = length;
#ifdef _KERNEL
(void) zfs_ereport_start(&report->zcr_ereport, &report->zcr_detector,
FM_EREPORT_ZFS_CHECKSUM, spa, vd, zb, zio, offset, length);
if (report->zcr_ereport == NULL) {
zfs_ereport_free_checksum(report);
return (0);
}
#endif
mutex_enter(&spa->spa_errlist_lock);
report->zcr_next = zio->io_logical->io_cksum_report;
zio->io_logical->io_cksum_report = report;
mutex_exit(&spa->spa_errlist_lock);
return (0);
}
void
zfs_ereport_finish_checksum(zio_cksum_report_t *report, const abd_t *good_data,
const abd_t *bad_data, boolean_t drop_if_identical)
{
#ifdef _KERNEL
zfs_ecksum_info_t *info;
info = annotate_ecksum(report->zcr_ereport, report->zcr_ckinfo,
good_data, bad_data, report->zcr_length, drop_if_identical);
if (info != NULL)
zfs_zevent_post(report->zcr_ereport,
report->zcr_detector, zfs_zevent_post_cb);
else
zfs_zevent_post_cb(report->zcr_ereport, report->zcr_detector);
report->zcr_ereport = report->zcr_detector = NULL;
if (info != NULL)
kmem_free(info, sizeof (*info));
#else
(void) report, (void) good_data, (void) bad_data,
(void) drop_if_identical;
#endif
}
void
zfs_ereport_free_checksum(zio_cksum_report_t *rpt)
{
#ifdef _KERNEL
if (rpt->zcr_ereport != NULL) {
fm_nvlist_destroy(rpt->zcr_ereport,
FM_NVA_FREE);
fm_nvlist_destroy(rpt->zcr_detector,
FM_NVA_FREE);
}
#endif
rpt->zcr_free(rpt->zcr_cbdata, rpt->zcr_cbinfo);
if (rpt->zcr_ckinfo != NULL)
kmem_free(rpt->zcr_ckinfo, sizeof (*rpt->zcr_ckinfo));
kmem_free(rpt, sizeof (*rpt));
}
/*
* Post a checksum ereport
*
* Returns
* - 0 if an event was posted
* - EINVAL if there was a problem posting event
* - EBUSY if the event was rate limited
* - EALREADY if the event was already posted (duplicate)
*/
int
zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd, const zbookmark_phys_t *zb,
struct zio *zio, uint64_t offset, uint64_t length,
const abd_t *good_data, const abd_t *bad_data, zio_bad_cksum_t *zbc)
{
int rc = 0;
#ifdef _KERNEL
nvlist_t *ereport = NULL;
nvlist_t *detector = NULL;
zfs_ecksum_info_t *info;
if (!zfs_ereport_is_valid(FM_EREPORT_ZFS_CHECKSUM, spa, vd, zio))
return (SET_ERROR(EINVAL));
if (zfs_ereport_is_duplicate(FM_EREPORT_ZFS_CHECKSUM, spa, vd, zb, zio,
offset, length))
return (SET_ERROR(EALREADY));
if (zfs_is_ratelimiting_event(FM_EREPORT_ZFS_CHECKSUM, vd))
return (SET_ERROR(EBUSY));
if (!zfs_ereport_start(&ereport, &detector, FM_EREPORT_ZFS_CHECKSUM,
spa, vd, zb, zio, offset, length) || (ereport == NULL)) {
return (SET_ERROR(EINVAL));
}
info = annotate_ecksum(ereport, zbc, good_data, bad_data, length,
B_FALSE);
if (info != NULL) {
rc = zfs_zevent_post(ereport, detector, zfs_zevent_post_cb);
kmem_free(info, sizeof (*info));
}
#else
(void) spa, (void) vd, (void) zb, (void) zio, (void) offset,
(void) length, (void) good_data, (void) bad_data, (void) zbc;
#endif
return (rc);
}
/*
* The 'sysevent.fs.zfs.*' events are signals posted to notify user space of
* change in the pool. All sysevents are listed in sys/sysevent/eventdefs.h
* and are designed to be consumed by the ZFS Event Daemon (ZED). For
* additional details refer to the zed(8) man page.
*/
nvlist_t *
zfs_event_create(spa_t *spa, vdev_t *vd, const char *type, const char *name,
nvlist_t *aux)
{
nvlist_t *resource = NULL;
#ifdef _KERNEL
char class[64];
if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT)
return (NULL);
if ((resource = fm_nvlist_create(NULL)) == NULL)
return (NULL);
(void) snprintf(class, sizeof (class), "%s.%s.%s", type,
ZFS_ERROR_CLASS, name);
VERIFY0(nvlist_add_uint8(resource, FM_VERSION, FM_RSRC_VERSION));
VERIFY0(nvlist_add_string(resource, FM_CLASS, class));
VERIFY0(nvlist_add_string(resource,
FM_EREPORT_PAYLOAD_ZFS_POOL, spa_name(spa)));
VERIFY0(nvlist_add_uint64(resource,
FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, spa_guid(spa)));
VERIFY0(nvlist_add_uint64(resource,
FM_EREPORT_PAYLOAD_ZFS_POOL_STATE, spa_state(spa)));
VERIFY0(nvlist_add_int32(resource,
FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, spa_load_state(spa)));
if (vd) {
VERIFY0(nvlist_add_uint64(resource,
FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, vd->vdev_guid));
VERIFY0(nvlist_add_uint64(resource,
FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE, vd->vdev_state));
if (vd->vdev_path != NULL)
VERIFY0(nvlist_add_string(resource,
FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH, vd->vdev_path));
if (vd->vdev_devid != NULL)
VERIFY0(nvlist_add_string(resource,
FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID, vd->vdev_devid));
if (vd->vdev_fru != NULL)
VERIFY0(nvlist_add_string(resource,
FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU, vd->vdev_fru));
if (vd->vdev_enc_sysfs_path != NULL)
VERIFY0(nvlist_add_string(resource,
FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH,
vd->vdev_enc_sysfs_path));
}
/* also copy any optional payload data */
if (aux) {
nvpair_t *elem = NULL;
while ((elem = nvlist_next_nvpair(aux, elem)) != NULL)
(void) nvlist_add_nvpair(resource, elem);
}
#else
(void) spa, (void) vd, (void) type, (void) name, (void) aux;
#endif
return (resource);
}
static void
zfs_post_common(spa_t *spa, vdev_t *vd, const char *type, const char *name,
nvlist_t *aux)
{
#ifdef _KERNEL
nvlist_t *resource;
resource = zfs_event_create(spa, vd, type, name, aux);
if (resource)
zfs_zevent_post(resource, NULL, zfs_zevent_post_cb);
#else
(void) spa, (void) vd, (void) type, (void) name, (void) aux;
#endif
}
/*
* The 'resource.fs.zfs.removed' event is an internal signal that the given vdev
* has been removed from the system. This will cause the DE to ignore any
* recent I/O errors, inferring that they are due to the asynchronous device
* removal.
*/
void
zfs_post_remove(spa_t *spa, vdev_t *vd)
{
zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_REMOVED, NULL);
}
/*
* The 'resource.fs.zfs.autoreplace' event is an internal signal that the pool
* has the 'autoreplace' property set, and therefore any broken vdevs will be
* handled by higher level logic, and no vdev fault should be generated.
*/
void
zfs_post_autoreplace(spa_t *spa, vdev_t *vd)
{
zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_AUTOREPLACE, NULL);
}
/*
* The 'resource.fs.zfs.statechange' event is an internal signal that the
* given vdev has transitioned its state to DEGRADED or HEALTHY. This will
* cause the retire agent to repair any outstanding fault management cases
* open because the device was not found (fault.fs.zfs.device).
*/
void
zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate)
{
#ifdef _KERNEL
nvlist_t *aux;
/*
* Add optional supplemental keys to payload
*/
aux = fm_nvlist_create(NULL);
if (vd && aux) {
if (vd->vdev_physpath) {
(void) nvlist_add_string(aux,
FM_EREPORT_PAYLOAD_ZFS_VDEV_PHYSPATH,
vd->vdev_physpath);
}
if (vd->vdev_enc_sysfs_path) {
(void) nvlist_add_string(aux,
FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH,
vd->vdev_enc_sysfs_path);
}
(void) nvlist_add_uint64(aux,
FM_EREPORT_PAYLOAD_ZFS_VDEV_LASTSTATE, laststate);
}
zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_STATECHANGE,
aux);
if (aux)
fm_nvlist_destroy(aux, FM_NVA_FREE);
#else
(void) spa, (void) vd, (void) laststate;
#endif
}
#ifdef _KERNEL
void
zfs_ereport_init(void)
{
mutex_init(&recent_events_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&recent_events_list, sizeof (recent_events_node_t),
offsetof(recent_events_node_t, re_list_link));
avl_create(&recent_events_tree, recent_events_compare,
sizeof (recent_events_node_t), offsetof(recent_events_node_t,
re_tree_link));
}
/*
* This 'early' fini needs to run before zfs_fini() which on Linux waits
* for the system_delay_taskq to drain.
*/
void
zfs_ereport_taskq_fini(void)
{
mutex_enter(&recent_events_lock);
if (recent_events_cleaner_tqid != 0) {
taskq_cancel_id(system_delay_taskq, recent_events_cleaner_tqid);
recent_events_cleaner_tqid = 0;
}
mutex_exit(&recent_events_lock);
}
void
zfs_ereport_fini(void)
{
recent_events_node_t *entry;
while ((entry = list_head(&recent_events_list)) != NULL) {
avl_remove(&recent_events_tree, entry);
list_remove(&recent_events_list, entry);
kmem_free(entry, sizeof (*entry));
}
avl_destroy(&recent_events_tree);
list_destroy(&recent_events_list);
mutex_destroy(&recent_events_lock);
}
void
zfs_ereport_snapshot_post(const char *subclass, spa_t *spa, const char *name)
{
nvlist_t *aux;
aux = fm_nvlist_create(NULL);
nvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_SNAPSHOT_NAME, name);
zfs_post_common(spa, NULL, FM_RSRC_CLASS, subclass, aux);
fm_nvlist_destroy(aux, FM_NVA_FREE);
}
/*
* Post when a event when a zvol is created or removed
*
* This is currently only used by macOS, since it uses the event to create
* symlinks between the volume name (mypool/myvol) and the actual /dev
* device (/dev/disk3). For example:
*
* /var/run/zfs/dsk/mypool/myvol -> /dev/disk3
*
* name: The full name of the zvol ("mypool/myvol")
* dev_name: The full /dev name for the zvol ("/dev/disk3")
* raw_name: The raw /dev name for the zvol ("/dev/rdisk3")
*/
void
zfs_ereport_zvol_post(const char *subclass, const char *name,
const char *dev_name, const char *raw_name)
{
nvlist_t *aux;
char *r;
boolean_t locked = mutex_owned(&spa_namespace_lock);
if (!locked) mutex_enter(&spa_namespace_lock);
spa_t *spa = spa_lookup(name);
if (!locked) mutex_exit(&spa_namespace_lock);
if (spa == NULL)
return;
aux = fm_nvlist_create(NULL);
nvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_DEVICE_NAME, dev_name);
nvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_RAW_DEVICE_NAME,
raw_name);
r = strchr(name, '/');
if (r && r[1])
nvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_VOLUME, &r[1]);
zfs_post_common(spa, NULL, FM_RSRC_CLASS, subclass, aux);
fm_nvlist_destroy(aux, FM_NVA_FREE);
}
EXPORT_SYMBOL(zfs_ereport_post);
EXPORT_SYMBOL(zfs_ereport_is_valid);
EXPORT_SYMBOL(zfs_ereport_post_checksum);
EXPORT_SYMBOL(zfs_post_remove);
EXPORT_SYMBOL(zfs_post_autoreplace);
EXPORT_SYMBOL(zfs_post_state_change);
ZFS_MODULE_PARAM(zfs_zevent, zfs_zevent_, retain_max, UINT, ZMOD_RW,
"Maximum recent zevents records to retain for duplicate checking");
ZFS_MODULE_PARAM(zfs_zevent, zfs_zevent_, retain_expire_secs, UINT, ZMOD_RW,
"Expiration time for recent zevents records");
#endif /* _KERNEL */
diff --git a/module/zfs/zfs_ioctl.c b/module/zfs/zfs_ioctl.c
index ca2da561220b..e592734eea09 100644
--- a/module/zfs/zfs_ioctl.c
+++ b/module/zfs/zfs_ioctl.c
@@ -1,7875 +1,7871 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Portions Copyright 2011 Martin Matuska
* Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
* Portions Copyright 2012 Pawel Jakub Dawidek <pawel@dawidek.net>
* Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2014, Joyent, Inc. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright (c) 2013 Steven Hartland. All rights reserved.
* Copyright (c) 2014 Integros [integros.com]
* Copyright 2016 Toomas Soome <tsoome@me.com>
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
* Copyright 2017 RackTop Systems.
* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
* Copyright (c) 2019 Datto Inc.
* Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved.
* Copyright (c) 2019, 2021, Klara Inc.
* Copyright (c) 2019, Allan Jude
*/
/*
* ZFS ioctls.
*
* This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
* pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
*
* There are two ways that we handle ioctls: the legacy way where almost
* all of the logic is in the ioctl callback, and the new way where most
* of the marshalling is handled in the common entry point, zfsdev_ioctl().
*
* Non-legacy ioctls should be registered by calling
* zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
* from userland by lzc_ioctl().
*
* The registration arguments are as follows:
*
* const char *name
* The name of the ioctl. This is used for history logging. If the
* ioctl returns successfully (the callback returns 0), and allow_log
* is true, then a history log entry will be recorded with the input &
* output nvlists. The log entry can be printed with "zpool history -i".
*
* zfs_ioc_t ioc
* The ioctl request number, which userland will pass to ioctl(2).
* We want newer versions of libzfs and libzfs_core to run against
* existing zfs kernel modules (i.e. a deferred reboot after an update).
* Therefore the ioctl numbers cannot change from release to release.
*
* zfs_secpolicy_func_t *secpolicy
* This function will be called before the zfs_ioc_func_t, to
* determine if this operation is permitted. It should return EPERM
* on failure, and 0 on success. Checks include determining if the
* dataset is visible in this zone, and if the user has either all
* zfs privileges in the zone (SYS_MOUNT), or has been granted permission
* to do this operation on this dataset with "zfs allow".
*
* zfs_ioc_namecheck_t namecheck
* This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
* name, a dataset name, or nothing. If the name is not well-formed,
* the ioctl will fail and the callback will not be called.
* Therefore, the callback can assume that the name is well-formed
* (e.g. is null-terminated, doesn't have more than one '@' character,
* doesn't have invalid characters).
*
* zfs_ioc_poolcheck_t pool_check
* This specifies requirements on the pool state. If the pool does
* not meet them (is suspended or is readonly), the ioctl will fail
* and the callback will not be called. If any checks are specified
* (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
* Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
* POOL_CHECK_READONLY).
*
* zfs_ioc_key_t *nvl_keys
* The list of expected/allowable innvl input keys. This list is used
* to validate the nvlist input to the ioctl.
*
* boolean_t smush_outnvlist
* If smush_outnvlist is true, then the output is presumed to be a
* list of errors, and it will be "smushed" down to fit into the
* caller's buffer, by removing some entries and replacing them with a
* single "N_MORE_ERRORS" entry indicating how many were removed. See
* nvlist_smush() for details. If smush_outnvlist is false, and the
* outnvlist does not fit into the userland-provided buffer, then the
* ioctl will fail with ENOMEM.
*
* zfs_ioc_func_t *func
* The callback function that will perform the operation.
*
* The callback should return 0 on success, or an error number on
* failure. If the function fails, the userland ioctl will return -1,
* and errno will be set to the callback's return value. The callback
* will be called with the following arguments:
*
* const char *name
* The name of the pool or dataset to operate on, from
* zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
* expected type (pool, dataset, or none).
*
* nvlist_t *innvl
* The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
* NULL if no input nvlist was provided. Changes to this nvlist are
* ignored. If the input nvlist could not be deserialized, the
* ioctl will fail and the callback will not be called.
*
* nvlist_t *outnvl
* The output nvlist, initially empty. The callback can fill it in,
* and it will be returned to userland by serializing it into
* zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
* fails (e.g. because the caller didn't supply a large enough
* buffer), then the overall ioctl will fail. See the
* 'smush_nvlist' argument above for additional behaviors.
*
* There are two typical uses of the output nvlist:
* - To return state, e.g. property values. In this case,
* smush_outnvlist should be false. If the buffer was not large
* enough, the caller will reallocate a larger buffer and try
* the ioctl again.
*
* - To return multiple errors from an ioctl which makes on-disk
* changes. In this case, smush_outnvlist should be true.
* Ioctls which make on-disk modifications should generally not
* use the outnvl if they succeed, because the caller can not
* distinguish between the operation failing, and
* deserialization failing.
*
* IOCTL Interface Errors
*
* The following ioctl input errors can be returned:
* ZFS_ERR_IOC_CMD_UNAVAIL the ioctl number is not supported by kernel
* ZFS_ERR_IOC_ARG_UNAVAIL an input argument is not supported by kernel
* ZFS_ERR_IOC_ARG_REQUIRED a required input argument is missing
* ZFS_ERR_IOC_ARG_BADTYPE an input argument has an invalid type
*/
#include <sys/types.h>
#include <sys/param.h>
#include <sys/errno.h>
#include <sys/uio_impl.h>
#include <sys/file.h>
#include <sys/kmem.h>
#include <sys/cmn_err.h>
#include <sys/stat.h>
#include <sys/zfs_ioctl.h>
#include <sys/zfs_quota.h>
#include <sys/zfs_vfsops.h>
#include <sys/zfs_znode.h>
#include <sys/zap.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/vdev.h>
#include <sys/vdev_impl.h>
#include <sys/dmu.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_deleg.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_impl.h>
#include <sys/dmu_redact.h>
#include <sys/dmu_tx.h>
#include <sys/sunddi.h>
#include <sys/policy.h>
#include <sys/zone.h>
#include <sys/nvpair.h>
#include <sys/pathname.h>
#include <sys/fs/zfs.h>
#include <sys/zfs_ctldir.h>
#include <sys/zfs_dir.h>
#include <sys/zfs_onexit.h>
#include <sys/zvol.h>
#include <sys/dsl_scan.h>
#include <sys/fm/util.h>
#include <sys/dsl_crypt.h>
#include <sys/rrwlock.h>
#include <sys/zfs_file.h>
#include <sys/dmu_recv.h>
#include <sys/dmu_send.h>
#include <sys/dmu_recv.h>
#include <sys/dsl_destroy.h>
#include <sys/dsl_bookmark.h>
#include <sys/dsl_userhold.h>
#include <sys/zfeature.h>
#include <sys/zcp.h>
#include <sys/zio_checksum.h>
#include <sys/vdev_removal.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_initialize.h>
#include <sys/vdev_trim.h>
#include "zfs_namecheck.h"
#include "zfs_prop.h"
#include "zfs_deleg.h"
#include "zfs_comutil.h"
#include <sys/lua/lua.h>
#include <sys/lua/lauxlib.h>
#include <sys/zfs_ioctl_impl.h>
kmutex_t zfsdev_state_lock;
-zfsdev_state_t *zfsdev_state_list;
+static zfsdev_state_t *zfsdev_state_list;
/*
* Limit maximum nvlist size. We don't want users passing in insane values
* for zc->zc_nvlist_src_size, since we will need to allocate that much memory.
* Defaults to 0=auto which is handled by platform code.
*/
unsigned long zfs_max_nvlist_src_size = 0;
/*
* When logging the output nvlist of an ioctl in the on-disk history, limit
* the logged size to this many bytes. This must be less than DMU_MAX_ACCESS.
* This applies primarily to zfs_ioc_channel_program().
*/
-unsigned long zfs_history_output_max = 1024 * 1024;
+static unsigned long zfs_history_output_max = 1024 * 1024;
uint_t zfs_fsyncer_key;
uint_t zfs_allow_log_key;
/* DATA_TYPE_ANY is used when zkey_type can vary. */
#define DATA_TYPE_ANY DATA_TYPE_UNKNOWN
typedef struct zfs_ioc_vec {
zfs_ioc_legacy_func_t *zvec_legacy_func;
zfs_ioc_func_t *zvec_func;
zfs_secpolicy_func_t *zvec_secpolicy;
zfs_ioc_namecheck_t zvec_namecheck;
boolean_t zvec_allow_log;
zfs_ioc_poolcheck_t zvec_pool_check;
boolean_t zvec_smush_outnvlist;
const char *zvec_name;
const zfs_ioc_key_t *zvec_nvl_keys;
size_t zvec_nvl_key_count;
} zfs_ioc_vec_t;
/* This array is indexed by zfs_userquota_prop_t */
static const char *userquota_perms[] = {
ZFS_DELEG_PERM_USERUSED,
ZFS_DELEG_PERM_USERQUOTA,
ZFS_DELEG_PERM_GROUPUSED,
ZFS_DELEG_PERM_GROUPQUOTA,
ZFS_DELEG_PERM_USEROBJUSED,
ZFS_DELEG_PERM_USEROBJQUOTA,
ZFS_DELEG_PERM_GROUPOBJUSED,
ZFS_DELEG_PERM_GROUPOBJQUOTA,
ZFS_DELEG_PERM_PROJECTUSED,
ZFS_DELEG_PERM_PROJECTQUOTA,
ZFS_DELEG_PERM_PROJECTOBJUSED,
ZFS_DELEG_PERM_PROJECTOBJQUOTA,
};
static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
static int zfs_check_settable(const char *name, nvpair_t *property,
cred_t *cr);
static int zfs_check_clearable(const char *dataset, nvlist_t *props,
nvlist_t **errors);
static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
boolean_t *);
int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
static void
history_str_free(char *buf)
{
kmem_free(buf, HIS_MAX_RECORD_LEN);
}
static char *
history_str_get(zfs_cmd_t *zc)
{
char *buf;
if (zc->zc_history == 0)
return (NULL);
buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
if (copyinstr((void *)(uintptr_t)zc->zc_history,
buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
history_str_free(buf);
return (NULL);
}
buf[HIS_MAX_RECORD_LEN -1] = '\0';
return (buf);
}
/*
* Return non-zero if the spa version is less than requested version.
*/
static int
zfs_earlier_version(const char *name, int version)
{
spa_t *spa;
if (spa_open(name, &spa, FTAG) == 0) {
if (spa_version(spa) < version) {
spa_close(spa, FTAG);
return (1);
}
spa_close(spa, FTAG);
}
return (0);
}
/*
* Return TRUE if the ZPL version is less than requested version.
*/
static boolean_t
zpl_earlier_version(const char *name, int version)
{
objset_t *os;
boolean_t rc = B_TRUE;
if (dmu_objset_hold(name, FTAG, &os) == 0) {
uint64_t zplversion;
if (dmu_objset_type(os) != DMU_OST_ZFS) {
dmu_objset_rele(os, FTAG);
return (B_TRUE);
}
/* XXX reading from non-owned objset */
if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
rc = zplversion < version;
dmu_objset_rele(os, FTAG);
}
return (rc);
}
static void
zfs_log_history(zfs_cmd_t *zc)
{
spa_t *spa;
char *buf;
if ((buf = history_str_get(zc)) == NULL)
return;
if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
(void) spa_history_log(spa, buf);
spa_close(spa, FTAG);
}
history_str_free(buf);
}
/*
* Policy for top-level read operations (list pools). Requires no privileges,
* and can be used in the local zone, as there is no associated dataset.
*/
/* ARGSUSED */
static int
zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (0);
}
/*
* Policy for dataset read operations (list children, get statistics). Requires
* no privileges, but must be visible in the local zone.
*/
/* ARGSUSED */
static int
zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
if (INGLOBALZONE(curproc) ||
zone_dataset_visible(zc->zc_name, NULL))
return (0);
return (SET_ERROR(ENOENT));
}
static int
zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
{
int writable = 1;
/*
* The dataset must be visible by this zone -- check this first
* so they don't see EPERM on something they shouldn't know about.
*/
if (!INGLOBALZONE(curproc) &&
!zone_dataset_visible(dataset, &writable))
return (SET_ERROR(ENOENT));
if (INGLOBALZONE(curproc)) {
/*
* If the fs is zoned, only root can access it from the
* global zone.
*/
if (secpolicy_zfs(cr) && zoned)
return (SET_ERROR(EPERM));
} else {
/*
* If we are in a local zone, the 'zoned' property must be set.
*/
if (!zoned)
return (SET_ERROR(EPERM));
/* must be writable by this zone */
if (!writable)
return (SET_ERROR(EPERM));
}
return (0);
}
static int
zfs_dozonecheck(const char *dataset, cred_t *cr)
{
uint64_t zoned;
if (dsl_prop_get_integer(dataset, zfs_prop_to_name(ZFS_PROP_ZONED),
&zoned, NULL))
return (SET_ERROR(ENOENT));
return (zfs_dozonecheck_impl(dataset, zoned, cr));
}
static int
zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
{
uint64_t zoned;
if (dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_ZONED), &zoned))
return (SET_ERROR(ENOENT));
return (zfs_dozonecheck_impl(dataset, zoned, cr));
}
static int
zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
const char *perm, cred_t *cr)
{
int error;
error = zfs_dozonecheck_ds(name, ds, cr);
if (error == 0) {
error = secpolicy_zfs(cr);
if (error != 0)
error = dsl_deleg_access_impl(ds, perm, cr);
}
return (error);
}
static int
zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
{
int error;
dsl_dataset_t *ds;
dsl_pool_t *dp;
/*
* First do a quick check for root in the global zone, which
* is allowed to do all write_perms. This ensures that zfs_ioc_*
* will get to handle nonexistent datasets.
*/
if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
return (0);
error = dsl_pool_hold(name, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold(dp, name, FTAG, &ds);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
/*
* Policy for setting the security label property.
*
* Returns 0 for success, non-zero for access and other errors.
*/
static int
zfs_set_slabel_policy(const char *name, const char *strval, cred_t *cr)
{
#ifdef HAVE_MLSLABEL
char ds_hexsl[MAXNAMELEN];
bslabel_t ds_sl, new_sl;
boolean_t new_default = FALSE;
uint64_t zoned;
int needed_priv = -1;
int error;
/* First get the existing dataset label. */
error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
1, sizeof (ds_hexsl), &ds_hexsl, NULL);
if (error != 0)
return (SET_ERROR(EPERM));
if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
new_default = TRUE;
/* The label must be translatable */
if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
return (SET_ERROR(EINVAL));
/*
* In a non-global zone, disallow attempts to set a label that
* doesn't match that of the zone; otherwise no other checks
* are needed.
*/
if (!INGLOBALZONE(curproc)) {
if (new_default || !blequal(&new_sl, CR_SL(CRED())))
return (SET_ERROR(EPERM));
return (0);
}
/*
* For global-zone datasets (i.e., those whose zoned property is
* "off", verify that the specified new label is valid for the
* global zone.
*/
if (dsl_prop_get_integer(name,
zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
return (SET_ERROR(EPERM));
if (!zoned) {
if (zfs_check_global_label(name, strval) != 0)
return (SET_ERROR(EPERM));
}
/*
* If the existing dataset label is nondefault, check if the
* dataset is mounted (label cannot be changed while mounted).
* Get the zfsvfs_t; if there isn't one, then the dataset isn't
* mounted (or isn't a dataset, doesn't exist, ...).
*/
if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
objset_t *os;
static const char *setsl_tag = "setsl_tag";
/*
* Try to own the dataset; abort if there is any error,
* (e.g., already mounted, in use, or other error).
*/
error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
setsl_tag, &os);
if (error != 0)
return (SET_ERROR(EPERM));
dmu_objset_disown(os, B_TRUE, setsl_tag);
if (new_default) {
needed_priv = PRIV_FILE_DOWNGRADE_SL;
goto out_check;
}
if (hexstr_to_label(strval, &new_sl) != 0)
return (SET_ERROR(EPERM));
if (blstrictdom(&ds_sl, &new_sl))
needed_priv = PRIV_FILE_DOWNGRADE_SL;
else if (blstrictdom(&new_sl, &ds_sl))
needed_priv = PRIV_FILE_UPGRADE_SL;
} else {
/* dataset currently has a default label */
if (!new_default)
needed_priv = PRIV_FILE_UPGRADE_SL;
}
out_check:
if (needed_priv != -1)
return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
return (0);
#else
return (SET_ERROR(ENOTSUP));
#endif /* HAVE_MLSLABEL */
}
static int
zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
cred_t *cr)
{
char *strval;
/*
* Check permissions for special properties.
*/
switch (prop) {
default:
break;
case ZFS_PROP_ZONED:
/*
* Disallow setting of 'zoned' from within a local zone.
*/
if (!INGLOBALZONE(curproc))
return (SET_ERROR(EPERM));
break;
case ZFS_PROP_QUOTA:
case ZFS_PROP_FILESYSTEM_LIMIT:
case ZFS_PROP_SNAPSHOT_LIMIT:
if (!INGLOBALZONE(curproc)) {
uint64_t zoned;
char setpoint[ZFS_MAX_DATASET_NAME_LEN];
/*
* Unprivileged users are allowed to modify the
* limit on things *under* (ie. contained by)
* the thing they own.
*/
if (dsl_prop_get_integer(dsname,
zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, setpoint))
return (SET_ERROR(EPERM));
if (!zoned || strlen(dsname) <= strlen(setpoint))
return (SET_ERROR(EPERM));
}
break;
case ZFS_PROP_MLSLABEL:
if (!is_system_labeled())
return (SET_ERROR(EPERM));
if (nvpair_value_string(propval, &strval) == 0) {
int err;
err = zfs_set_slabel_policy(dsname, strval, CRED());
if (err != 0)
return (err);
}
break;
}
return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
}
/* ARGSUSED */
static int
zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
int error;
error = zfs_dozonecheck(zc->zc_name, cr);
if (error != 0)
return (error);
/*
* permission to set permissions will be evaluated later in
* dsl_deleg_can_allow()
*/
return (0);
}
/* ARGSUSED */
static int
zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_ROLLBACK, cr));
}
/* ARGSUSED */
static int
zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
const char *cp;
int error;
/*
* Generate the current snapshot name from the given objsetid, then
* use that name for the secpolicy/zone checks.
*/
cp = strchr(zc->zc_name, '@');
if (cp == NULL)
return (SET_ERROR(EINVAL));
error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
dsl_dataset_name(ds, zc->zc_name);
error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
ZFS_DELEG_PERM_SEND, cr);
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
/* ARGSUSED */
static int
zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_SEND, cr));
}
static int
zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (SET_ERROR(ENOTSUP));
}
static int
zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (SET_ERROR(ENOTSUP));
}
static int
zfs_get_parent(const char *datasetname, char *parent, int parentsize)
{
char *cp;
/*
* Remove the @bla or /bla from the end of the name to get the parent.
*/
(void) strncpy(parent, datasetname, parentsize);
cp = strrchr(parent, '@');
if (cp != NULL) {
cp[0] = '\0';
} else {
cp = strrchr(parent, '/');
if (cp == NULL)
return (SET_ERROR(ENOENT));
cp[0] = '\0';
}
return (0);
}
int
zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
{
int error;
if ((error = zfs_secpolicy_write_perms(name,
ZFS_DELEG_PERM_MOUNT, cr)) != 0)
return (error);
return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
}
/* ARGSUSED */
static int
zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
}
/*
* Destroying snapshots with delegated permissions requires
* descendant mount and destroy permissions.
*/
/* ARGSUSED */
static int
zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
nvlist_t *snaps;
nvpair_t *pair, *nextpair;
int error = 0;
snaps = fnvlist_lookup_nvlist(innvl, "snaps");
for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
pair = nextpair) {
nextpair = nvlist_next_nvpair(snaps, pair);
error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
if (error == ENOENT) {
/*
* Ignore any snapshots that don't exist (we consider
* them "already destroyed"). Remove the name from the
* nvl here in case the snapshot is created between
* now and when we try to destroy it (in which case
* we don't want to destroy it since we haven't
* checked for permission).
*/
fnvlist_remove_nvpair(snaps, pair);
error = 0;
}
if (error != 0)
break;
}
return (error);
}
int
zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
{
char parentname[ZFS_MAX_DATASET_NAME_LEN];
int error;
if ((error = zfs_secpolicy_write_perms(from,
ZFS_DELEG_PERM_RENAME, cr)) != 0)
return (error);
if ((error = zfs_secpolicy_write_perms(from,
ZFS_DELEG_PERM_MOUNT, cr)) != 0)
return (error);
if ((error = zfs_get_parent(to, parentname,
sizeof (parentname))) != 0)
return (error);
if ((error = zfs_secpolicy_write_perms(parentname,
ZFS_DELEG_PERM_CREATE, cr)) != 0)
return (error);
if ((error = zfs_secpolicy_write_perms(parentname,
ZFS_DELEG_PERM_MOUNT, cr)) != 0)
return (error);
return (error);
}
/* ARGSUSED */
static int
zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
}
/* ARGSUSED */
static int
zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
dsl_pool_t *dp;
dsl_dataset_t *clone;
int error;
error = zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_PROMOTE, cr);
if (error != 0)
return (error);
error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
if (error == 0) {
char parentname[ZFS_MAX_DATASET_NAME_LEN];
dsl_dataset_t *origin = NULL;
dsl_dir_t *dd;
dd = clone->ds_dir;
error = dsl_dataset_hold_obj(dd->dd_pool,
dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
if (error != 0) {
dsl_dataset_rele(clone, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
ZFS_DELEG_PERM_MOUNT, cr);
dsl_dataset_name(origin, parentname);
if (error == 0) {
error = zfs_secpolicy_write_perms_ds(parentname, origin,
ZFS_DELEG_PERM_PROMOTE, cr);
}
dsl_dataset_rele(clone, FTAG);
dsl_dataset_rele(origin, FTAG);
}
dsl_pool_rele(dp, FTAG);
return (error);
}
/* ARGSUSED */
static int
zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
int error;
if ((error = zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
return (error);
if ((error = zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_MOUNT, cr)) != 0)
return (error);
return (zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_CREATE, cr));
}
/* ARGSUSED */
static int
zfs_secpolicy_recv_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (zfs_secpolicy_recv(zc, innvl, cr));
}
int
zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
{
return (zfs_secpolicy_write_perms(name,
ZFS_DELEG_PERM_SNAPSHOT, cr));
}
/*
* Check for permission to create each snapshot in the nvlist.
*/
/* ARGSUSED */
static int
zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
nvlist_t *snaps;
int error = 0;
nvpair_t *pair;
snaps = fnvlist_lookup_nvlist(innvl, "snaps");
for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
pair = nvlist_next_nvpair(snaps, pair)) {
char *name = nvpair_name(pair);
char *atp = strchr(name, '@');
if (atp == NULL) {
error = SET_ERROR(EINVAL);
break;
}
*atp = '\0';
error = zfs_secpolicy_snapshot_perms(name, cr);
*atp = '@';
if (error != 0)
break;
}
return (error);
}
/*
* Check for permission to create each bookmark in the nvlist.
*/
/* ARGSUSED */
static int
zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
int error = 0;
for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
char *name = nvpair_name(pair);
char *hashp = strchr(name, '#');
if (hashp == NULL) {
error = SET_ERROR(EINVAL);
break;
}
*hashp = '\0';
error = zfs_secpolicy_write_perms(name,
ZFS_DELEG_PERM_BOOKMARK, cr);
*hashp = '#';
if (error != 0)
break;
}
return (error);
}
/* ARGSUSED */
static int
zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
nvpair_t *pair, *nextpair;
int error = 0;
for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
pair = nextpair) {
char *name = nvpair_name(pair);
char *hashp = strchr(name, '#');
nextpair = nvlist_next_nvpair(innvl, pair);
if (hashp == NULL) {
error = SET_ERROR(EINVAL);
break;
}
*hashp = '\0';
error = zfs_secpolicy_write_perms(name,
ZFS_DELEG_PERM_DESTROY, cr);
*hashp = '#';
if (error == ENOENT) {
/*
* Ignore any filesystems that don't exist (we consider
* their bookmarks "already destroyed"). Remove
* the name from the nvl here in case the filesystem
* is created between now and when we try to destroy
* the bookmark (in which case we don't want to
* destroy it since we haven't checked for permission).
*/
fnvlist_remove_nvpair(innvl, pair);
error = 0;
}
if (error != 0)
break;
}
return (error);
}
/* ARGSUSED */
static int
zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
/*
* Even root must have a proper TSD so that we know what pool
* to log to.
*/
if (tsd_get(zfs_allow_log_key) == NULL)
return (SET_ERROR(EPERM));
return (0);
}
static int
zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
char parentname[ZFS_MAX_DATASET_NAME_LEN];
int error;
char *origin;
if ((error = zfs_get_parent(zc->zc_name, parentname,
sizeof (parentname))) != 0)
return (error);
if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
(error = zfs_secpolicy_write_perms(origin,
ZFS_DELEG_PERM_CLONE, cr)) != 0)
return (error);
if ((error = zfs_secpolicy_write_perms(parentname,
ZFS_DELEG_PERM_CREATE, cr)) != 0)
return (error);
return (zfs_secpolicy_write_perms(parentname,
ZFS_DELEG_PERM_MOUNT, cr));
}
/*
* Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
* SYS_CONFIG privilege, which is not available in a local zone.
*/
/* ARGSUSED */
int
zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
if (secpolicy_sys_config(cr, B_FALSE) != 0)
return (SET_ERROR(EPERM));
return (0);
}
/*
* Policy for object to name lookups.
*/
/* ARGSUSED */
static int
zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
int error;
if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
return (0);
error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
return (error);
}
/*
* Policy for fault injection. Requires all privileges.
*/
/* ARGSUSED */
static int
zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (secpolicy_zinject(cr));
}
/* ARGSUSED */
static int
zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
if (prop == ZPROP_INVAL) {
if (!zfs_prop_user(zc->zc_value))
return (SET_ERROR(EINVAL));
return (zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_USERPROP, cr));
} else {
return (zfs_secpolicy_setprop(zc->zc_name, prop,
NULL, cr));
}
}
static int
zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
int err = zfs_secpolicy_read(zc, innvl, cr);
if (err)
return (err);
if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
return (SET_ERROR(EINVAL));
if (zc->zc_value[0] == 0) {
/*
* They are asking about a posix uid/gid. If it's
* themself, allow it.
*/
if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
zc->zc_objset_type == ZFS_PROP_USERQUOTA ||
zc->zc_objset_type == ZFS_PROP_USEROBJUSED ||
zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
if (zc->zc_guid == crgetuid(cr))
return (0);
} else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED ||
zc->zc_objset_type == ZFS_PROP_GROUPQUOTA ||
zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED ||
zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) {
if (groupmember(zc->zc_guid, cr))
return (0);
}
/* else is for project quota/used */
}
return (zfs_secpolicy_write_perms(zc->zc_name,
userquota_perms[zc->zc_objset_type], cr));
}
static int
zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
int err = zfs_secpolicy_read(zc, innvl, cr);
if (err)
return (err);
if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
return (SET_ERROR(EINVAL));
return (zfs_secpolicy_write_perms(zc->zc_name,
userquota_perms[zc->zc_objset_type], cr));
}
/* ARGSUSED */
static int
zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
NULL, cr));
}
/* ARGSUSED */
static int
zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
nvpair_t *pair;
nvlist_t *holds;
int error;
holds = fnvlist_lookup_nvlist(innvl, "holds");
for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
pair = nvlist_next_nvpair(holds, pair)) {
char fsname[ZFS_MAX_DATASET_NAME_LEN];
error = dmu_fsname(nvpair_name(pair), fsname);
if (error != 0)
return (error);
error = zfs_secpolicy_write_perms(fsname,
ZFS_DELEG_PERM_HOLD, cr);
if (error != 0)
return (error);
}
return (0);
}
/* ARGSUSED */
static int
zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
nvpair_t *pair;
int error;
for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
pair = nvlist_next_nvpair(innvl, pair)) {
char fsname[ZFS_MAX_DATASET_NAME_LEN];
error = dmu_fsname(nvpair_name(pair), fsname);
if (error != 0)
return (error);
error = zfs_secpolicy_write_perms(fsname,
ZFS_DELEG_PERM_RELEASE, cr);
if (error != 0)
return (error);
}
return (0);
}
/*
* Policy for allowing temporary snapshots to be taken or released
*/
static int
zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
/*
* A temporary snapshot is the same as a snapshot,
* hold, destroy and release all rolled into one.
* Delegated diff alone is sufficient that we allow this.
*/
int error;
if ((error = zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_DIFF, cr)) == 0)
return (0);
error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
if (innvl != NULL) {
if (error == 0)
error = zfs_secpolicy_hold(zc, innvl, cr);
if (error == 0)
error = zfs_secpolicy_release(zc, innvl, cr);
if (error == 0)
error = zfs_secpolicy_destroy(zc, innvl, cr);
}
return (error);
}
static int
zfs_secpolicy_load_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_LOAD_KEY, cr));
}
static int
zfs_secpolicy_change_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
{
return (zfs_secpolicy_write_perms(zc->zc_name,
ZFS_DELEG_PERM_CHANGE_KEY, cr));
}
/*
* Returns the nvlist as specified by the user in the zfs_cmd_t.
*/
static int
get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
{
char *packed;
int error;
nvlist_t *list = NULL;
/*
* Read in and unpack the user-supplied nvlist.
*/
if (size == 0)
return (SET_ERROR(EINVAL));
packed = vmem_alloc(size, KM_SLEEP);
if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
iflag)) != 0) {
vmem_free(packed, size);
return (SET_ERROR(EFAULT));
}
if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
vmem_free(packed, size);
return (error);
}
vmem_free(packed, size);
*nvp = list;
return (0);
}
/*
* Reduce the size of this nvlist until it can be serialized in 'max' bytes.
* Entries will be removed from the end of the nvlist, and one int32 entry
* named "N_MORE_ERRORS" will be added indicating how many entries were
* removed.
*/
static int
nvlist_smush(nvlist_t *errors, size_t max)
{
size_t size;
size = fnvlist_size(errors);
if (size > max) {
nvpair_t *more_errors;
int n = 0;
if (max < 1024)
return (SET_ERROR(ENOMEM));
fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
more_errors = nvlist_prev_nvpair(errors, NULL);
do {
nvpair_t *pair = nvlist_prev_nvpair(errors,
more_errors);
fnvlist_remove_nvpair(errors, pair);
n++;
size = fnvlist_size(errors);
} while (size > max);
fnvlist_remove_nvpair(errors, more_errors);
fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
ASSERT3U(fnvlist_size(errors), <=, max);
}
return (0);
}
static int
put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
{
char *packed = NULL;
int error = 0;
size_t size;
size = fnvlist_size(nvl);
if (size > zc->zc_nvlist_dst_size) {
error = SET_ERROR(ENOMEM);
} else {
packed = fnvlist_pack(nvl, &size);
if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
size, zc->zc_iflags) != 0)
error = SET_ERROR(EFAULT);
fnvlist_pack_free(packed, size);
}
zc->zc_nvlist_dst_size = size;
zc->zc_nvlist_dst_filled = B_TRUE;
return (error);
}
int
getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
{
int error = 0;
if (dmu_objset_type(os) != DMU_OST_ZFS) {
return (SET_ERROR(EINVAL));
}
mutex_enter(&os->os_user_ptr_lock);
*zfvp = dmu_objset_get_user(os);
/* bump s_active only when non-zero to prevent umount race */
error = zfs_vfs_ref(zfvp);
mutex_exit(&os->os_user_ptr_lock);
return (error);
}
int
getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
{
objset_t *os;
int error;
error = dmu_objset_hold(dsname, FTAG, &os);
if (error != 0)
return (error);
error = getzfsvfs_impl(os, zfvp);
dmu_objset_rele(os, FTAG);
return (error);
}
/*
* Find a zfsvfs_t for a mounted filesystem, or create our own, in which
* case its z_sb will be NULL, and it will be opened as the owner.
* If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
* which prevents all inode ops from running.
*/
static int
zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
{
int error = 0;
if (getzfsvfs(name, zfvp) != 0)
error = zfsvfs_create(name, B_FALSE, zfvp);
if (error == 0) {
if (writer)
ZFS_TEARDOWN_ENTER_WRITE(*zfvp, tag);
else
ZFS_TEARDOWN_ENTER_READ(*zfvp, tag);
if ((*zfvp)->z_unmounted) {
/*
* XXX we could probably try again, since the unmounting
* thread should be just about to disassociate the
* objset from the zfsvfs.
*/
ZFS_TEARDOWN_EXIT(*zfvp, tag);
return (SET_ERROR(EBUSY));
}
}
return (error);
}
static void
zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
{
ZFS_TEARDOWN_EXIT(zfsvfs, tag);
if (zfs_vfs_held(zfsvfs)) {
zfs_vfs_rele(zfsvfs);
} else {
dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
zfsvfs_free(zfsvfs);
}
}
static int
zfs_ioc_pool_create(zfs_cmd_t *zc)
{
int error;
nvlist_t *config, *props = NULL;
nvlist_t *rootprops = NULL;
nvlist_t *zplprops = NULL;
dsl_crypto_params_t *dcp = NULL;
const char *spa_name = zc->zc_name;
boolean_t unload_wkey = B_TRUE;
if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
zc->zc_iflags, &config)))
return (error);
if (zc->zc_nvlist_src_size != 0 && (error =
get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
zc->zc_iflags, &props))) {
nvlist_free(config);
return (error);
}
if (props) {
nvlist_t *nvl = NULL;
nvlist_t *hidden_args = NULL;
uint64_t version = SPA_VERSION;
char *tname;
(void) nvlist_lookup_uint64(props,
zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
if (!SPA_VERSION_IS_SUPPORTED(version)) {
error = SET_ERROR(EINVAL);
goto pool_props_bad;
}
(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
if (nvl) {
error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
if (error != 0)
goto pool_props_bad;
(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
}
(void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
&hidden_args);
error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
rootprops, hidden_args, &dcp);
if (error != 0)
goto pool_props_bad;
(void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);
VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
error = zfs_fill_zplprops_root(version, rootprops,
zplprops, NULL);
if (error != 0)
goto pool_props_bad;
if (nvlist_lookup_string(props,
zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
spa_name = tname;
}
error = spa_create(zc->zc_name, config, props, zplprops, dcp);
/*
* Set the remaining root properties
*/
if (!error && (error = zfs_set_prop_nvlist(spa_name,
ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) {
(void) spa_destroy(spa_name);
unload_wkey = B_FALSE; /* spa_destroy() unloads wrapping keys */
}
pool_props_bad:
nvlist_free(rootprops);
nvlist_free(zplprops);
nvlist_free(config);
nvlist_free(props);
dsl_crypto_params_free(dcp, unload_wkey && !!error);
return (error);
}
static int
zfs_ioc_pool_destroy(zfs_cmd_t *zc)
{
int error;
zfs_log_history(zc);
error = spa_destroy(zc->zc_name);
return (error);
}
static int
zfs_ioc_pool_import(zfs_cmd_t *zc)
{
nvlist_t *config, *props = NULL;
uint64_t guid;
int error;
if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
zc->zc_iflags, &config)) != 0)
return (error);
if (zc->zc_nvlist_src_size != 0 && (error =
get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
zc->zc_iflags, &props))) {
nvlist_free(config);
return (error);
}
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
guid != zc->zc_guid)
error = SET_ERROR(EINVAL);
else
error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
if (zc->zc_nvlist_dst != 0) {
int err;
if ((err = put_nvlist(zc, config)) != 0)
error = err;
}
nvlist_free(config);
nvlist_free(props);
return (error);
}
static int
zfs_ioc_pool_export(zfs_cmd_t *zc)
{
int error;
boolean_t force = (boolean_t)zc->zc_cookie;
boolean_t hardforce = (boolean_t)zc->zc_guid;
zfs_log_history(zc);
error = spa_export(zc->zc_name, NULL, force, hardforce);
return (error);
}
static int
zfs_ioc_pool_configs(zfs_cmd_t *zc)
{
nvlist_t *configs;
int error;
if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
return (SET_ERROR(EEXIST));
error = put_nvlist(zc, configs);
nvlist_free(configs);
return (error);
}
/*
* inputs:
* zc_name name of the pool
*
* outputs:
* zc_cookie real errno
* zc_nvlist_dst config nvlist
* zc_nvlist_dst_size size of config nvlist
*/
static int
zfs_ioc_pool_stats(zfs_cmd_t *zc)
{
nvlist_t *config;
int error;
int ret = 0;
error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
sizeof (zc->zc_value));
if (config != NULL) {
ret = put_nvlist(zc, config);
nvlist_free(config);
/*
* The config may be present even if 'error' is non-zero.
* In this case we return success, and preserve the real errno
* in 'zc_cookie'.
*/
zc->zc_cookie = error;
} else {
ret = error;
}
return (ret);
}
/*
* Try to import the given pool, returning pool stats as appropriate so that
* user land knows which devices are available and overall pool health.
*/
static int
zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
{
nvlist_t *tryconfig, *config = NULL;
int error;
if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
zc->zc_iflags, &tryconfig)) != 0)
return (error);
config = spa_tryimport(tryconfig);
nvlist_free(tryconfig);
if (config == NULL)
return (SET_ERROR(EINVAL));
error = put_nvlist(zc, config);
nvlist_free(config);
return (error);
}
/*
* inputs:
* zc_name name of the pool
* zc_cookie scan func (pool_scan_func_t)
* zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
*/
static int
zfs_ioc_pool_scan(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
return (SET_ERROR(EINVAL));
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
return (error);
if (zc->zc_flags == POOL_SCRUB_PAUSE)
error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
else if (zc->zc_cookie == POOL_SCAN_NONE)
error = spa_scan_stop(spa);
else
error = spa_scan(spa, zc->zc_cookie);
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_pool_freeze(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
error = spa_open(zc->zc_name, &spa, FTAG);
if (error == 0) {
spa_freeze(spa);
spa_close(spa, FTAG);
}
return (error);
}
static int
zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
return (error);
if (zc->zc_cookie < spa_version(spa) ||
!SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
spa_close(spa, FTAG);
return (SET_ERROR(EINVAL));
}
spa_upgrade(spa, zc->zc_cookie);
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_pool_get_history(zfs_cmd_t *zc)
{
spa_t *spa;
char *hist_buf;
uint64_t size;
int error;
if ((size = zc->zc_history_len) == 0)
return (SET_ERROR(EINVAL));
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
return (error);
if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOTSUP));
}
hist_buf = vmem_alloc(size, KM_SLEEP);
if ((error = spa_history_get(spa, &zc->zc_history_offset,
&zc->zc_history_len, hist_buf)) == 0) {
error = ddi_copyout(hist_buf,
(void *)(uintptr_t)zc->zc_history,
zc->zc_history_len, zc->zc_iflags);
}
spa_close(spa, FTAG);
vmem_free(hist_buf, size);
return (error);
}
static int
zfs_ioc_pool_reguid(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
error = spa_open(zc->zc_name, &spa, FTAG);
if (error == 0) {
error = spa_change_guid(spa);
spa_close(spa, FTAG);
}
return (error);
}
static int
zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
{
return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
}
/*
* inputs:
* zc_name name of filesystem
* zc_obj object to find
*
* outputs:
* zc_value name of object
*/
static int
zfs_ioc_obj_to_path(zfs_cmd_t *zc)
{
objset_t *os;
int error;
/* XXX reading from objset not owned */
if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
FTAG, &os)) != 0)
return (error);
if (dmu_objset_type(os) != DMU_OST_ZFS) {
dmu_objset_rele_flags(os, B_TRUE, FTAG);
return (SET_ERROR(EINVAL));
}
error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
sizeof (zc->zc_value));
dmu_objset_rele_flags(os, B_TRUE, FTAG);
return (error);
}
/*
* inputs:
* zc_name name of filesystem
* zc_obj object to find
*
* outputs:
* zc_stat stats on object
* zc_value path to object
*/
static int
zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
{
objset_t *os;
int error;
/* XXX reading from objset not owned */
if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
FTAG, &os)) != 0)
return (error);
if (dmu_objset_type(os) != DMU_OST_ZFS) {
dmu_objset_rele_flags(os, B_TRUE, FTAG);
return (SET_ERROR(EINVAL));
}
error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
sizeof (zc->zc_value));
dmu_objset_rele_flags(os, B_TRUE, FTAG);
return (error);
}
static int
zfs_ioc_vdev_add(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
nvlist_t *config;
error = spa_open(zc->zc_name, &spa, FTAG);
if (error != 0)
return (error);
error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
zc->zc_iflags, &config);
if (error == 0) {
error = spa_vdev_add(spa, config);
nvlist_free(config);
}
spa_close(spa, FTAG);
return (error);
}
/*
* inputs:
* zc_name name of the pool
* zc_guid guid of vdev to remove
* zc_cookie cancel removal
*/
static int
zfs_ioc_vdev_remove(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
error = spa_open(zc->zc_name, &spa, FTAG);
if (error != 0)
return (error);
if (zc->zc_cookie != 0) {
error = spa_vdev_remove_cancel(spa);
} else {
error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
}
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
vdev_state_t newstate = VDEV_STATE_UNKNOWN;
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
return (error);
switch (zc->zc_cookie) {
case VDEV_STATE_ONLINE:
error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
break;
case VDEV_STATE_OFFLINE:
error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
break;
case VDEV_STATE_FAULTED:
if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
zc->zc_obj != VDEV_AUX_EXTERNAL &&
zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
break;
case VDEV_STATE_DEGRADED:
if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
zc->zc_obj != VDEV_AUX_EXTERNAL)
zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
break;
default:
error = SET_ERROR(EINVAL);
}
zc->zc_cookie = newstate;
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_vdev_attach(zfs_cmd_t *zc)
{
spa_t *spa;
nvlist_t *config;
int replacing = zc->zc_cookie;
int rebuild = zc->zc_simple;
int error;
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
return (error);
if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
zc->zc_iflags, &config)) == 0) {
error = spa_vdev_attach(spa, zc->zc_guid, config, replacing,
rebuild);
nvlist_free(config);
}
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_vdev_detach(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
return (error);
error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_vdev_split(zfs_cmd_t *zc)
{
spa_t *spa;
nvlist_t *config, *props = NULL;
int error;
boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
return (error);
if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
zc->zc_iflags, &config))) {
spa_close(spa, FTAG);
return (error);
}
if (zc->zc_nvlist_src_size != 0 && (error =
get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
zc->zc_iflags, &props))) {
spa_close(spa, FTAG);
nvlist_free(config);
return (error);
}
error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
spa_close(spa, FTAG);
nvlist_free(config);
nvlist_free(props);
return (error);
}
static int
zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
{
spa_t *spa;
const char *path = zc->zc_value;
uint64_t guid = zc->zc_guid;
int error;
error = spa_open(zc->zc_name, &spa, FTAG);
if (error != 0)
return (error);
error = spa_vdev_setpath(spa, guid, path);
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
{
spa_t *spa;
const char *fru = zc->zc_value;
uint64_t guid = zc->zc_guid;
int error;
error = spa_open(zc->zc_name, &spa, FTAG);
if (error != 0)
return (error);
error = spa_vdev_setfru(spa, guid, fru);
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
{
int error = 0;
nvlist_t *nv;
dmu_objset_fast_stat(os, &zc->zc_objset_stats);
if (zc->zc_nvlist_dst != 0 &&
(error = dsl_prop_get_all(os, &nv)) == 0) {
dmu_objset_stats(os, nv);
/*
* NB: zvol_get_stats() will read the objset contents,
* which we aren't supposed to do with a
* DS_MODE_USER hold, because it could be
* inconsistent. So this is a bit of a workaround...
* XXX reading without owning
*/
if (!zc->zc_objset_stats.dds_inconsistent &&
dmu_objset_type(os) == DMU_OST_ZVOL) {
error = zvol_get_stats(os, nv);
if (error == EIO) {
nvlist_free(nv);
return (error);
}
VERIFY0(error);
}
if (error == 0)
error = put_nvlist(zc, nv);
nvlist_free(nv);
}
return (error);
}
/*
* inputs:
* zc_name name of filesystem
* zc_nvlist_dst_size size of buffer for property nvlist
*
* outputs:
* zc_objset_stats stats
* zc_nvlist_dst property nvlist
* zc_nvlist_dst_size size of property nvlist
*/
static int
zfs_ioc_objset_stats(zfs_cmd_t *zc)
{
objset_t *os;
int error;
error = dmu_objset_hold(zc->zc_name, FTAG, &os);
if (error == 0) {
error = zfs_ioc_objset_stats_impl(zc, os);
dmu_objset_rele(os, FTAG);
}
return (error);
}
/*
* inputs:
* zc_name name of filesystem
* zc_nvlist_dst_size size of buffer for property nvlist
*
* outputs:
* zc_nvlist_dst received property nvlist
* zc_nvlist_dst_size size of received property nvlist
*
* Gets received properties (distinct from local properties on or after
* SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
* local property values.
*/
static int
zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
{
int error = 0;
nvlist_t *nv;
/*
* Without this check, we would return local property values if the
* caller has not already received properties on or after
* SPA_VERSION_RECVD_PROPS.
*/
if (!dsl_prop_get_hasrecvd(zc->zc_name))
return (SET_ERROR(ENOTSUP));
if (zc->zc_nvlist_dst != 0 &&
(error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
error = put_nvlist(zc, nv);
nvlist_free(nv);
}
return (error);
}
static int
nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
{
uint64_t value;
int error;
/*
* zfs_get_zplprop() will either find a value or give us
* the default value (if there is one).
*/
if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
return (error);
VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
return (0);
}
/*
* inputs:
* zc_name name of filesystem
* zc_nvlist_dst_size size of buffer for zpl property nvlist
*
* outputs:
* zc_nvlist_dst zpl property nvlist
* zc_nvlist_dst_size size of zpl property nvlist
*/
static int
zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
{
objset_t *os;
int err;
/* XXX reading without owning */
if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
return (err);
dmu_objset_fast_stat(os, &zc->zc_objset_stats);
/*
* NB: nvl_add_zplprop() will read the objset contents,
* which we aren't supposed to do with a DS_MODE_USER
* hold, because it could be inconsistent.
*/
if (zc->zc_nvlist_dst != 0 &&
!zc->zc_objset_stats.dds_inconsistent &&
dmu_objset_type(os) == DMU_OST_ZFS) {
nvlist_t *nv;
VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
(err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
(err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
(err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
err = put_nvlist(zc, nv);
nvlist_free(nv);
} else {
err = SET_ERROR(ENOENT);
}
dmu_objset_rele(os, FTAG);
return (err);
}
/*
* inputs:
* zc_name name of filesystem
* zc_cookie zap cursor
* zc_nvlist_dst_size size of buffer for property nvlist
*
* outputs:
* zc_name name of next filesystem
* zc_cookie zap cursor
* zc_objset_stats stats
* zc_nvlist_dst property nvlist
* zc_nvlist_dst_size size of property nvlist
*/
static int
zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
{
objset_t *os;
int error;
char *p;
size_t orig_len = strlen(zc->zc_name);
top:
if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
if (error == ENOENT)
error = SET_ERROR(ESRCH);
return (error);
}
p = strrchr(zc->zc_name, '/');
if (p == NULL || p[1] != '\0')
(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
p = zc->zc_name + strlen(zc->zc_name);
do {
error = dmu_dir_list_next(os,
sizeof (zc->zc_name) - (p - zc->zc_name), p,
NULL, &zc->zc_cookie);
if (error == ENOENT)
error = SET_ERROR(ESRCH);
} while (error == 0 && zfs_dataset_name_hidden(zc->zc_name));
dmu_objset_rele(os, FTAG);
/*
* If it's an internal dataset (ie. with a '$' in its name),
* don't try to get stats for it, otherwise we'll return ENOENT.
*/
if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
error = zfs_ioc_objset_stats(zc); /* fill in the stats */
if (error == ENOENT) {
/* We lost a race with destroy, get the next one. */
zc->zc_name[orig_len] = '\0';
goto top;
}
}
return (error);
}
/*
* inputs:
* zc_name name of filesystem
* zc_cookie zap cursor
* zc_nvlist_src iteration range nvlist
* zc_nvlist_src_size size of iteration range nvlist
*
* outputs:
* zc_name name of next snapshot
* zc_objset_stats stats
* zc_nvlist_dst property nvlist
* zc_nvlist_dst_size size of property nvlist
*/
static int
zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
{
int error;
objset_t *os, *ossnap;
dsl_dataset_t *ds;
uint64_t min_txg = 0, max_txg = 0;
if (zc->zc_nvlist_src_size != 0) {
nvlist_t *props = NULL;
error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
zc->zc_iflags, &props);
if (error != 0)
return (error);
(void) nvlist_lookup_uint64(props, SNAP_ITER_MIN_TXG,
&min_txg);
(void) nvlist_lookup_uint64(props, SNAP_ITER_MAX_TXG,
&max_txg);
nvlist_free(props);
}
error = dmu_objset_hold(zc->zc_name, FTAG, &os);
if (error != 0) {
return (error == ENOENT ? SET_ERROR(ESRCH) : error);
}
/*
* A dataset name of maximum length cannot have any snapshots,
* so exit immediately.
*/
if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
ZFS_MAX_DATASET_NAME_LEN) {
dmu_objset_rele(os, FTAG);
return (SET_ERROR(ESRCH));
}
while (error == 0) {
if (issig(JUSTLOOKING) && issig(FORREAL)) {
error = SET_ERROR(EINTR);
break;
}
error = dmu_snapshot_list_next(os,
sizeof (zc->zc_name) - strlen(zc->zc_name),
zc->zc_name + strlen(zc->zc_name), &zc->zc_obj,
&zc->zc_cookie, NULL);
if (error == ENOENT) {
error = SET_ERROR(ESRCH);
break;
} else if (error != 0) {
break;
}
error = dsl_dataset_hold_obj(dmu_objset_pool(os), zc->zc_obj,
FTAG, &ds);
if (error != 0)
break;
if ((min_txg != 0 && dsl_get_creationtxg(ds) < min_txg) ||
(max_txg != 0 && dsl_get_creationtxg(ds) > max_txg)) {
dsl_dataset_rele(ds, FTAG);
/* undo snapshot name append */
*(strchr(zc->zc_name, '@') + 1) = '\0';
/* skip snapshot */
continue;
}
if (zc->zc_simple) {
dsl_dataset_rele(ds, FTAG);
break;
}
if ((error = dmu_objset_from_ds(ds, &ossnap)) != 0) {
dsl_dataset_rele(ds, FTAG);
break;
}
if ((error = zfs_ioc_objset_stats_impl(zc, ossnap)) != 0) {
dsl_dataset_rele(ds, FTAG);
break;
}
dsl_dataset_rele(ds, FTAG);
break;
}
dmu_objset_rele(os, FTAG);
/* if we failed, undo the @ that we tacked on to zc_name */
if (error != 0)
*strchr(zc->zc_name, '@') = '\0';
return (error);
}
static int
zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
{
const char *propname = nvpair_name(pair);
uint64_t *valary;
unsigned int vallen;
const char *dash, *domain;
zfs_userquota_prop_t type;
uint64_t rid;
uint64_t quota;
zfsvfs_t *zfsvfs;
int err;
if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
nvlist_t *attrs;
VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
&pair) != 0)
return (SET_ERROR(EINVAL));
}
/*
* A correctly constructed propname is encoded as
* userquota@<rid>-<domain>.
*/
if ((dash = strchr(propname, '-')) == NULL ||
nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
vallen != 3)
return (SET_ERROR(EINVAL));
domain = dash + 1;
type = valary[0];
rid = valary[1];
quota = valary[2];
err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
if (err == 0) {
err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
zfsvfs_rele(zfsvfs, FTAG);
}
return (err);
}
/*
* If the named property is one that has a special function to set its value,
* return 0 on success and a positive error code on failure; otherwise if it is
* not one of the special properties handled by this function, return -1.
*
* XXX: It would be better for callers of the property interface if we handled
* these special cases in dsl_prop.c (in the dsl layer).
*/
static int
zfs_prop_set_special(const char *dsname, zprop_source_t source,
nvpair_t *pair)
{
const char *propname = nvpair_name(pair);
zfs_prop_t prop = zfs_name_to_prop(propname);
uint64_t intval = 0;
const char *strval = NULL;
int err = -1;
if (prop == ZPROP_INVAL) {
if (zfs_prop_userquota(propname))
return (zfs_prop_set_userquota(dsname, pair));
return (-1);
}
if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
nvlist_t *attrs;
VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
&pair) == 0);
}
/* all special properties are numeric except for keylocation */
if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
strval = fnvpair_value_string(pair);
} else {
intval = fnvpair_value_uint64(pair);
}
switch (prop) {
case ZFS_PROP_QUOTA:
err = dsl_dir_set_quota(dsname, source, intval);
break;
case ZFS_PROP_REFQUOTA:
err = dsl_dataset_set_refquota(dsname, source, intval);
break;
case ZFS_PROP_FILESYSTEM_LIMIT:
case ZFS_PROP_SNAPSHOT_LIMIT:
if (intval == UINT64_MAX) {
/* clearing the limit, just do it */
err = 0;
} else {
err = dsl_dir_activate_fs_ss_limit(dsname);
}
/*
* Set err to -1 to force the zfs_set_prop_nvlist code down the
* default path to set the value in the nvlist.
*/
if (err == 0)
err = -1;
break;
case ZFS_PROP_KEYLOCATION:
err = dsl_crypto_can_set_keylocation(dsname, strval);
/*
* Set err to -1 to force the zfs_set_prop_nvlist code down the
* default path to set the value in the nvlist.
*/
if (err == 0)
err = -1;
break;
case ZFS_PROP_RESERVATION:
err = dsl_dir_set_reservation(dsname, source, intval);
break;
case ZFS_PROP_REFRESERVATION:
err = dsl_dataset_set_refreservation(dsname, source, intval);
break;
case ZFS_PROP_COMPRESSION:
err = dsl_dataset_set_compression(dsname, source, intval);
/*
* Set err to -1 to force the zfs_set_prop_nvlist code down the
* default path to set the value in the nvlist.
*/
if (err == 0)
err = -1;
break;
case ZFS_PROP_VOLSIZE:
err = zvol_set_volsize(dsname, intval);
break;
case ZFS_PROP_SNAPDEV:
err = zvol_set_snapdev(dsname, source, intval);
break;
case ZFS_PROP_VOLMODE:
err = zvol_set_volmode(dsname, source, intval);
break;
case ZFS_PROP_VERSION:
{
zfsvfs_t *zfsvfs;
if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
break;
err = zfs_set_version(zfsvfs, intval);
zfsvfs_rele(zfsvfs, FTAG);
if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
zfs_cmd_t *zc;
zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
(void) strlcpy(zc->zc_name, dsname,
sizeof (zc->zc_name));
(void) zfs_ioc_userspace_upgrade(zc);
(void) zfs_ioc_id_quota_upgrade(zc);
kmem_free(zc, sizeof (zfs_cmd_t));
}
break;
}
default:
err = -1;
}
return (err);
}
static boolean_t
zfs_is_namespace_prop(zfs_prop_t prop)
{
switch (prop) {
case ZFS_PROP_ATIME:
case ZFS_PROP_RELATIME:
case ZFS_PROP_DEVICES:
case ZFS_PROP_EXEC:
case ZFS_PROP_SETUID:
case ZFS_PROP_READONLY:
case ZFS_PROP_XATTR:
case ZFS_PROP_NBMAND:
return (B_TRUE);
default:
return (B_FALSE);
}
}
/*
* This function is best effort. If it fails to set any of the given properties,
* it continues to set as many as it can and returns the last error
* encountered. If the caller provides a non-NULL errlist, it will be filled in
* with the list of names of all the properties that failed along with the
* corresponding error numbers.
*
* If every property is set successfully, zero is returned and errlist is not
* modified.
*/
int
zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
nvlist_t *errlist)
{
nvpair_t *pair;
nvpair_t *propval;
int rv = 0;
uint64_t intval;
const char *strval;
boolean_t should_update_mount_cache = B_FALSE;
nvlist_t *genericnvl = fnvlist_alloc();
nvlist_t *retrynvl = fnvlist_alloc();
retry:
pair = NULL;
while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
const char *propname = nvpair_name(pair);
zfs_prop_t prop = zfs_name_to_prop(propname);
int err = 0;
/* decode the property value */
propval = pair;
if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
nvlist_t *attrs;
attrs = fnvpair_value_nvlist(pair);
if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
&propval) != 0)
err = SET_ERROR(EINVAL);
}
/* Validate value type */
if (err == 0 && source == ZPROP_SRC_INHERITED) {
/* inherited properties are expected to be booleans */
if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
err = SET_ERROR(EINVAL);
} else if (err == 0 && prop == ZPROP_INVAL) {
if (zfs_prop_user(propname)) {
if (nvpair_type(propval) != DATA_TYPE_STRING)
err = SET_ERROR(EINVAL);
} else if (zfs_prop_userquota(propname)) {
if (nvpair_type(propval) !=
DATA_TYPE_UINT64_ARRAY)
err = SET_ERROR(EINVAL);
} else {
err = SET_ERROR(EINVAL);
}
} else if (err == 0) {
if (nvpair_type(propval) == DATA_TYPE_STRING) {
if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
err = SET_ERROR(EINVAL);
} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
const char *unused;
intval = fnvpair_value_uint64(propval);
switch (zfs_prop_get_type(prop)) {
case PROP_TYPE_NUMBER:
break;
case PROP_TYPE_STRING:
err = SET_ERROR(EINVAL);
break;
case PROP_TYPE_INDEX:
if (zfs_prop_index_to_string(prop,
intval, &unused) != 0)
err =
SET_ERROR(ZFS_ERR_BADPROP);
break;
default:
cmn_err(CE_PANIC,
"unknown property type");
}
} else {
err = SET_ERROR(EINVAL);
}
}
/* Validate permissions */
if (err == 0)
err = zfs_check_settable(dsname, pair, CRED());
if (err == 0) {
if (source == ZPROP_SRC_INHERITED)
err = -1; /* does not need special handling */
else
err = zfs_prop_set_special(dsname, source,
pair);
if (err == -1) {
/*
* For better performance we build up a list of
* properties to set in a single transaction.
*/
err = nvlist_add_nvpair(genericnvl, pair);
} else if (err != 0 && nvl != retrynvl) {
/*
* This may be a spurious error caused by
* receiving quota and reservation out of order.
* Try again in a second pass.
*/
err = nvlist_add_nvpair(retrynvl, pair);
}
}
if (err != 0) {
if (errlist != NULL)
fnvlist_add_int32(errlist, propname, err);
rv = err;
}
if (zfs_is_namespace_prop(prop))
should_update_mount_cache = B_TRUE;
}
if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
nvl = retrynvl;
goto retry;
}
if (!nvlist_empty(genericnvl) &&
dsl_props_set(dsname, source, genericnvl) != 0) {
/*
* If this fails, we still want to set as many properties as we
* can, so try setting them individually.
*/
pair = NULL;
while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
const char *propname = nvpair_name(pair);
int err = 0;
propval = pair;
if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
nvlist_t *attrs;
attrs = fnvpair_value_nvlist(pair);
propval = fnvlist_lookup_nvpair(attrs,
ZPROP_VALUE);
}
if (nvpair_type(propval) == DATA_TYPE_STRING) {
strval = fnvpair_value_string(propval);
err = dsl_prop_set_string(dsname, propname,
source, strval);
} else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
err = dsl_prop_inherit(dsname, propname,
source);
} else {
intval = fnvpair_value_uint64(propval);
err = dsl_prop_set_int(dsname, propname, source,
intval);
}
if (err != 0) {
if (errlist != NULL) {
fnvlist_add_int32(errlist, propname,
err);
}
rv = err;
}
}
}
if (should_update_mount_cache)
zfs_ioctl_update_mount_cache(dsname);
nvlist_free(genericnvl);
nvlist_free(retrynvl);
return (rv);
}
/*
* Check that all the properties are valid user properties.
*/
static int
zfs_check_userprops(nvlist_t *nvl)
{
nvpair_t *pair = NULL;
while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
const char *propname = nvpair_name(pair);
if (!zfs_prop_user(propname) ||
nvpair_type(pair) != DATA_TYPE_STRING)
return (SET_ERROR(EINVAL));
if (strlen(propname) >= ZAP_MAXNAMELEN)
return (SET_ERROR(ENAMETOOLONG));
if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
return (SET_ERROR(E2BIG));
}
return (0);
}
static void
props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
{
nvpair_t *pair;
VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
pair = NULL;
while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
if (nvlist_exists(skipped, nvpair_name(pair)))
continue;
VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
}
}
static int
clear_received_props(const char *dsname, nvlist_t *props,
nvlist_t *skipped)
{
int err = 0;
nvlist_t *cleared_props = NULL;
props_skip(props, skipped, &cleared_props);
if (!nvlist_empty(cleared_props)) {
/*
* Acts on local properties until the dataset has received
* properties at least once on or after SPA_VERSION_RECVD_PROPS.
*/
zprop_source_t flags = (ZPROP_SRC_NONE |
(dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
}
nvlist_free(cleared_props);
return (err);
}
/*
* inputs:
* zc_name name of filesystem
* zc_value name of property to set
* zc_nvlist_src{_size} nvlist of properties to apply
* zc_cookie received properties flag
*
* outputs:
* zc_nvlist_dst{_size} error for each unapplied received property
*/
static int
zfs_ioc_set_prop(zfs_cmd_t *zc)
{
nvlist_t *nvl;
boolean_t received = zc->zc_cookie;
zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
ZPROP_SRC_LOCAL);
nvlist_t *errors;
int error;
if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
zc->zc_iflags, &nvl)) != 0)
return (error);
if (received) {
nvlist_t *origprops;
if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
(void) clear_received_props(zc->zc_name,
origprops, nvl);
nvlist_free(origprops);
}
error = dsl_prop_set_hasrecvd(zc->zc_name);
}
errors = fnvlist_alloc();
if (error == 0)
error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
if (zc->zc_nvlist_dst != 0 && errors != NULL) {
(void) put_nvlist(zc, errors);
}
nvlist_free(errors);
nvlist_free(nvl);
return (error);
}
/*
* inputs:
* zc_name name of filesystem
* zc_value name of property to inherit
* zc_cookie revert to received value if TRUE
*
* outputs: none
*/
static int
zfs_ioc_inherit_prop(zfs_cmd_t *zc)
{
const char *propname = zc->zc_value;
zfs_prop_t prop = zfs_name_to_prop(propname);
boolean_t received = zc->zc_cookie;
zprop_source_t source = (received
? ZPROP_SRC_NONE /* revert to received value, if any */
: ZPROP_SRC_INHERITED); /* explicitly inherit */
nvlist_t *dummy;
nvpair_t *pair;
zprop_type_t type;
int err;
if (!received) {
/*
* Only check this in the non-received case. We want to allow
* 'inherit -S' to revert non-inheritable properties like quota
* and reservation to the received or default values even though
* they are not considered inheritable.
*/
if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
return (SET_ERROR(EINVAL));
}
if (prop == ZPROP_INVAL) {
if (!zfs_prop_user(propname))
return (SET_ERROR(EINVAL));
type = PROP_TYPE_STRING;
} else if (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION) {
return (SET_ERROR(EINVAL));
} else {
type = zfs_prop_get_type(prop);
}
/*
* zfs_prop_set_special() expects properties in the form of an
* nvpair with type info.
*/
dummy = fnvlist_alloc();
switch (type) {
case PROP_TYPE_STRING:
VERIFY(0 == nvlist_add_string(dummy, propname, ""));
break;
case PROP_TYPE_NUMBER:
case PROP_TYPE_INDEX:
VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
break;
default:
err = SET_ERROR(EINVAL);
goto errout;
}
pair = nvlist_next_nvpair(dummy, NULL);
if (pair == NULL) {
err = SET_ERROR(EINVAL);
} else {
err = zfs_prop_set_special(zc->zc_name, source, pair);
if (err == -1) /* property is not "special", needs handling */
err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
source);
}
errout:
nvlist_free(dummy);
return (err);
}
static int
zfs_ioc_pool_set_props(zfs_cmd_t *zc)
{
nvlist_t *props;
spa_t *spa;
int error;
nvpair_t *pair;
if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
zc->zc_iflags, &props)))
return (error);
/*
* If the only property is the configfile, then just do a spa_lookup()
* to handle the faulted case.
*/
pair = nvlist_next_nvpair(props, NULL);
if (pair != NULL && strcmp(nvpair_name(pair),
zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
nvlist_next_nvpair(props, pair) == NULL) {
mutex_enter(&spa_namespace_lock);
if ((spa = spa_lookup(zc->zc_name)) != NULL) {
spa_configfile_set(spa, props, B_FALSE);
spa_write_cachefile(spa, B_FALSE, B_TRUE);
}
mutex_exit(&spa_namespace_lock);
if (spa != NULL) {
nvlist_free(props);
return (0);
}
}
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
nvlist_free(props);
return (error);
}
error = spa_prop_set(spa, props);
nvlist_free(props);
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_pool_get_props(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
nvlist_t *nvp = NULL;
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
/*
* If the pool is faulted, there may be properties we can still
* get (such as altroot and cachefile), so attempt to get them
* anyway.
*/
mutex_enter(&spa_namespace_lock);
if ((spa = spa_lookup(zc->zc_name)) != NULL)
error = spa_prop_get(spa, &nvp);
mutex_exit(&spa_namespace_lock);
} else {
error = spa_prop_get(spa, &nvp);
spa_close(spa, FTAG);
}
if (error == 0 && zc->zc_nvlist_dst != 0)
error = put_nvlist(zc, nvp);
else
error = SET_ERROR(EFAULT);
nvlist_free(nvp);
return (error);
}
/*
* innvl: {
* "vdevprops_set_vdev" -> guid
* "vdevprops_set_props" -> { prop -> value }
* }
*
* outnvl: propname -> error code (int32)
*/
static const zfs_ioc_key_t zfs_keys_vdev_set_props[] = {
{ZPOOL_VDEV_PROPS_SET_VDEV, DATA_TYPE_UINT64, 0},
{ZPOOL_VDEV_PROPS_SET_PROPS, DATA_TYPE_NVLIST, 0}
};
static int
zfs_ioc_vdev_set_props(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
{
spa_t *spa;
int error;
vdev_t *vd;
uint64_t vdev_guid;
/* Early validation */
if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
&vdev_guid) != 0)
return (SET_ERROR(EINVAL));
if (outnvl == NULL)
return (SET_ERROR(EINVAL));
if ((error = spa_open(poolname, &spa, FTAG)) != 0)
return (error);
ASSERT(spa_writeable(spa));
if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOENT));
}
error = vdev_prop_set(vd, innvl, outnvl);
spa_close(spa, FTAG);
return (error);
}
/*
* innvl: {
* "vdevprops_get_vdev" -> guid
* (optional) "vdevprops_get_props" -> { propname -> propid }
* }
*
* outnvl: propname -> value
*/
static const zfs_ioc_key_t zfs_keys_vdev_get_props[] = {
{ZPOOL_VDEV_PROPS_GET_VDEV, DATA_TYPE_UINT64, 0},
{ZPOOL_VDEV_PROPS_GET_PROPS, DATA_TYPE_NVLIST, ZK_OPTIONAL}
};
static int
zfs_ioc_vdev_get_props(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
{
spa_t *spa;
int error;
vdev_t *vd;
uint64_t vdev_guid;
/* Early validation */
if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_GET_VDEV,
&vdev_guid) != 0)
return (SET_ERROR(EINVAL));
if (outnvl == NULL)
return (SET_ERROR(EINVAL));
if ((error = spa_open(poolname, &spa, FTAG)) != 0)
return (error);
if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOENT));
}
error = vdev_prop_get(vd, innvl, outnvl);
spa_close(spa, FTAG);
return (error);
}
/*
* inputs:
* zc_name name of filesystem
* zc_nvlist_src{_size} nvlist of delegated permissions
* zc_perm_action allow/unallow flag
*
* outputs: none
*/
static int
zfs_ioc_set_fsacl(zfs_cmd_t *zc)
{
int error;
nvlist_t *fsaclnv = NULL;
if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
zc->zc_iflags, &fsaclnv)) != 0)
return (error);
/*
* Verify nvlist is constructed correctly
*/
if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
nvlist_free(fsaclnv);
return (SET_ERROR(EINVAL));
}
/*
* If we don't have PRIV_SYS_MOUNT, then validate
* that user is allowed to hand out each permission in
* the nvlist(s)
*/
error = secpolicy_zfs(CRED());
if (error != 0) {
if (zc->zc_perm_action == B_FALSE) {
error = dsl_deleg_can_allow(zc->zc_name,
fsaclnv, CRED());
} else {
error = dsl_deleg_can_unallow(zc->zc_name,
fsaclnv, CRED());
}
}
if (error == 0)
error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
nvlist_free(fsaclnv);
return (error);
}
/*
* inputs:
* zc_name name of filesystem
*
* outputs:
* zc_nvlist_src{_size} nvlist of delegated permissions
*/
static int
zfs_ioc_get_fsacl(zfs_cmd_t *zc)
{
nvlist_t *nvp;
int error;
if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
error = put_nvlist(zc, nvp);
nvlist_free(nvp);
}
return (error);
}
/* ARGSUSED */
static void
zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
{
zfs_creat_t *zct = arg;
zfs_create_fs(os, cr, zct->zct_zplprops, tx);
}
#define ZFS_PROP_UNDEFINED ((uint64_t)-1)
/*
* inputs:
* os parent objset pointer (NULL if root fs)
* fuids_ok fuids allowed in this version of the spa?
* sa_ok SAs allowed in this version of the spa?
* createprops list of properties requested by creator
*
* outputs:
* zplprops values for the zplprops we attach to the master node object
* is_ci true if requested file system will be purely case-insensitive
*
* Determine the settings for utf8only, normalization and
* casesensitivity. Specific values may have been requested by the
* creator and/or we can inherit values from the parent dataset. If
* the file system is of too early a vintage, a creator can not
* request settings for these properties, even if the requested
* setting is the default value. We don't actually want to create dsl
* properties for these, so remove them from the source nvlist after
* processing.
*/
static int
zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
nvlist_t *zplprops, boolean_t *is_ci)
{
uint64_t sense = ZFS_PROP_UNDEFINED;
uint64_t norm = ZFS_PROP_UNDEFINED;
uint64_t u8 = ZFS_PROP_UNDEFINED;
int error;
ASSERT(zplprops != NULL);
/* parent dataset must be a filesystem */
if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
/*
* Pull out creator prop choices, if any.
*/
if (createprops) {
(void) nvlist_lookup_uint64(createprops,
zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
(void) nvlist_lookup_uint64(createprops,
zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
(void) nvlist_remove_all(createprops,
zfs_prop_to_name(ZFS_PROP_NORMALIZE));
(void) nvlist_lookup_uint64(createprops,
zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
(void) nvlist_remove_all(createprops,
zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
(void) nvlist_lookup_uint64(createprops,
zfs_prop_to_name(ZFS_PROP_CASE), &sense);
(void) nvlist_remove_all(createprops,
zfs_prop_to_name(ZFS_PROP_CASE));
}
/*
* If the zpl version requested is whacky or the file system
* or pool is version is too "young" to support normalization
* and the creator tried to set a value for one of the props,
* error out.
*/
if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
(zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
(zplver >= ZPL_VERSION_SA && !sa_ok) ||
(zplver < ZPL_VERSION_NORMALIZATION &&
(norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
sense != ZFS_PROP_UNDEFINED)))
return (SET_ERROR(ENOTSUP));
/*
* Put the version in the zplprops
*/
VERIFY(nvlist_add_uint64(zplprops,
zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
if (norm == ZFS_PROP_UNDEFINED &&
(error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
return (error);
VERIFY(nvlist_add_uint64(zplprops,
zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
/*
* If we're normalizing, names must always be valid UTF-8 strings.
*/
if (norm)
u8 = 1;
if (u8 == ZFS_PROP_UNDEFINED &&
(error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
return (error);
VERIFY(nvlist_add_uint64(zplprops,
zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
if (sense == ZFS_PROP_UNDEFINED &&
(error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
return (error);
VERIFY(nvlist_add_uint64(zplprops,
zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
if (is_ci)
*is_ci = (sense == ZFS_CASE_INSENSITIVE);
return (0);
}
static int
zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
nvlist_t *zplprops, boolean_t *is_ci)
{
boolean_t fuids_ok, sa_ok;
uint64_t zplver = ZPL_VERSION;
objset_t *os = NULL;
char parentname[ZFS_MAX_DATASET_NAME_LEN];
spa_t *spa;
uint64_t spa_vers;
int error;
zfs_get_parent(dataset, parentname, sizeof (parentname));
if ((error = spa_open(dataset, &spa, FTAG)) != 0)
return (error);
spa_vers = spa_version(spa);
spa_close(spa, FTAG);
zplver = zfs_zpl_version_map(spa_vers);
fuids_ok = (zplver >= ZPL_VERSION_FUID);
sa_ok = (zplver >= ZPL_VERSION_SA);
/*
* Open parent object set so we can inherit zplprop values.
*/
if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
return (error);
error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
zplprops, is_ci);
dmu_objset_rele(os, FTAG);
return (error);
}
static int
zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
nvlist_t *zplprops, boolean_t *is_ci)
{
boolean_t fuids_ok;
boolean_t sa_ok;
uint64_t zplver = ZPL_VERSION;
int error;
zplver = zfs_zpl_version_map(spa_vers);
fuids_ok = (zplver >= ZPL_VERSION_FUID);
sa_ok = (zplver >= ZPL_VERSION_SA);
error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
createprops, zplprops, is_ci);
return (error);
}
/*
* innvl: {
* "type" -> dmu_objset_type_t (int32)
* (optional) "props" -> { prop -> value }
* (optional) "hidden_args" -> { "wkeydata" -> value }
* raw uint8_t array of encryption wrapping key data (32 bytes)
* }
*
* outnvl: propname -> error code (int32)
*/
static const zfs_ioc_key_t zfs_keys_create[] = {
{"type", DATA_TYPE_INT32, 0},
{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
};
static int
zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
{
int error = 0;
zfs_creat_t zct = { 0 };
nvlist_t *nvprops = NULL;
nvlist_t *hidden_args = NULL;
void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
dmu_objset_type_t type;
boolean_t is_insensitive = B_FALSE;
dsl_crypto_params_t *dcp = NULL;
type = (dmu_objset_type_t)fnvlist_lookup_int32(innvl, "type");
(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
switch (type) {
case DMU_OST_ZFS:
cbfunc = zfs_create_cb;
break;
case DMU_OST_ZVOL:
cbfunc = zvol_create_cb;
break;
default:
cbfunc = NULL;
break;
}
if (strchr(fsname, '@') ||
strchr(fsname, '%'))
return (SET_ERROR(EINVAL));
zct.zct_props = nvprops;
if (cbfunc == NULL)
return (SET_ERROR(EINVAL));
if (type == DMU_OST_ZVOL) {
uint64_t volsize, volblocksize;
if (nvprops == NULL)
return (SET_ERROR(EINVAL));
if (nvlist_lookup_uint64(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
return (SET_ERROR(EINVAL));
if ((error = nvlist_lookup_uint64(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
&volblocksize)) != 0 && error != ENOENT)
return (SET_ERROR(EINVAL));
if (error != 0)
volblocksize = zfs_prop_default_numeric(
ZFS_PROP_VOLBLOCKSIZE);
if ((error = zvol_check_volblocksize(fsname,
volblocksize)) != 0 ||
(error = zvol_check_volsize(volsize,
volblocksize)) != 0)
return (error);
} else if (type == DMU_OST_ZFS) {
int error;
/*
* We have to have normalization and
* case-folding flags correct when we do the
* file system creation, so go figure them out
* now.
*/
VERIFY(nvlist_alloc(&zct.zct_zplprops,
NV_UNIQUE_NAME, KM_SLEEP) == 0);
error = zfs_fill_zplprops(fsname, nvprops,
zct.zct_zplprops, &is_insensitive);
if (error != 0) {
nvlist_free(zct.zct_zplprops);
return (error);
}
}
error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
hidden_args, &dcp);
if (error != 0) {
nvlist_free(zct.zct_zplprops);
return (error);
}
error = dmu_objset_create(fsname, type,
is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);
nvlist_free(zct.zct_zplprops);
dsl_crypto_params_free(dcp, !!error);
/*
* It would be nice to do this atomically.
*/
if (error == 0) {
error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
nvprops, outnvl);
if (error != 0) {
spa_t *spa;
int error2;
/*
* Volumes will return EBUSY and cannot be destroyed
* until all asynchronous minor handling (e.g. from
* setting the volmode property) has completed. Wait for
* the spa_zvol_taskq to drain then retry.
*/
error2 = dsl_destroy_head(fsname);
while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
error2 = spa_open(fsname, &spa, FTAG);
if (error2 == 0) {
taskq_wait(spa->spa_zvol_taskq);
spa_close(spa, FTAG);
}
error2 = dsl_destroy_head(fsname);
}
}
}
return (error);
}
/*
* innvl: {
* "origin" -> name of origin snapshot
* (optional) "props" -> { prop -> value }
* (optional) "hidden_args" -> { "wkeydata" -> value }
* raw uint8_t array of encryption wrapping key data (32 bytes)
* }
*
* outputs:
* outnvl: propname -> error code (int32)
*/
static const zfs_ioc_key_t zfs_keys_clone[] = {
{"origin", DATA_TYPE_STRING, 0},
{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
};
static int
zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
{
int error = 0;
nvlist_t *nvprops = NULL;
const char *origin_name;
origin_name = fnvlist_lookup_string(innvl, "origin");
(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
if (strchr(fsname, '@') ||
strchr(fsname, '%'))
return (SET_ERROR(EINVAL));
if (dataset_namecheck(origin_name, NULL, NULL) != 0)
return (SET_ERROR(EINVAL));
error = dmu_objset_clone(fsname, origin_name);
/*
* It would be nice to do this atomically.
*/
if (error == 0) {
error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
nvprops, outnvl);
if (error != 0)
(void) dsl_destroy_head(fsname);
}
return (error);
}
static const zfs_ioc_key_t zfs_keys_remap[] = {
/* no nvl keys */
};
/* ARGSUSED */
static int
zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
{
/* This IOCTL is no longer supported. */
return (0);
}
/*
* innvl: {
* "snaps" -> { snapshot1, snapshot2 }
* (optional) "props" -> { prop -> value (string) }
* }
*
* outnvl: snapshot -> error code (int32)
*/
static const zfs_ioc_key_t zfs_keys_snapshot[] = {
{"snaps", DATA_TYPE_NVLIST, 0},
{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
};
static int
zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
{
nvlist_t *snaps;
nvlist_t *props = NULL;
int error, poollen;
nvpair_t *pair;
(void) nvlist_lookup_nvlist(innvl, "props", &props);
if (!nvlist_empty(props) &&
zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
return (SET_ERROR(ENOTSUP));
if ((error = zfs_check_userprops(props)) != 0)
return (error);
snaps = fnvlist_lookup_nvlist(innvl, "snaps");
poollen = strlen(poolname);
for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
pair = nvlist_next_nvpair(snaps, pair)) {
const char *name = nvpair_name(pair);
char *cp = strchr(name, '@');
/*
* The snap name must contain an @, and the part after it must
* contain only valid characters.
*/
if (cp == NULL ||
zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
return (SET_ERROR(EINVAL));
/*
* The snap must be in the specified pool.
*/
if (strncmp(name, poolname, poollen) != 0 ||
(name[poollen] != '/' && name[poollen] != '@'))
return (SET_ERROR(EXDEV));
/*
* Check for permission to set the properties on the fs.
*/
if (!nvlist_empty(props)) {
*cp = '\0';
error = zfs_secpolicy_write_perms(name,
ZFS_DELEG_PERM_USERPROP, CRED());
*cp = '@';
if (error != 0)
return (error);
}
/* This must be the only snap of this fs. */
for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
if (strncmp(name, nvpair_name(pair2), cp - name + 1)
== 0) {
return (SET_ERROR(EXDEV));
}
}
}
error = dsl_dataset_snapshot(snaps, props, outnvl);
return (error);
}
/*
* innvl: "message" -> string
*/
static const zfs_ioc_key_t zfs_keys_log_history[] = {
{"message", DATA_TYPE_STRING, 0},
};
/* ARGSUSED */
static int
zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
{
const char *message;
char *poolname;
spa_t *spa;
int error;
/*
* The poolname in the ioctl is not set, we get it from the TSD,
* which was set at the end of the last successful ioctl that allows
* logging. The secpolicy func already checked that it is set.
* Only one log ioctl is allowed after each successful ioctl, so
* we clear the TSD here.
*/
poolname = tsd_get(zfs_allow_log_key);
if (poolname == NULL)
return (SET_ERROR(EINVAL));
(void) tsd_set(zfs_allow_log_key, NULL);
error = spa_open(poolname, &spa, FTAG);
kmem_strfree(poolname);
if (error != 0)
return (error);
message = fnvlist_lookup_string(innvl, "message");
if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOTSUP));
}
error = spa_history_log(spa, message);
spa_close(spa, FTAG);
return (error);
}
/*
* This ioctl is used to set the bootenv configuration on the current
* pool. This configuration is stored in the second padding area of the label,
* and it is used by the bootloader(s) to store the bootloader and/or system
* specific data.
* The data is stored as nvlist data stream, and is protected by
* an embedded checksum.
* The version can have two possible values:
* VB_RAW: nvlist should have key GRUB_ENVMAP, value DATA_TYPE_STRING.
* VB_NVLIST: nvlist with arbitrary <key, value> pairs.
*/
static const zfs_ioc_key_t zfs_keys_set_bootenv[] = {
{"version", DATA_TYPE_UINT64, 0},
{"<keys>", DATA_TYPE_ANY, ZK_OPTIONAL | ZK_WILDCARDLIST},
};
static int
zfs_ioc_set_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
{
int error;
spa_t *spa;
if ((error = spa_open(name, &spa, FTAG)) != 0)
return (error);
spa_vdev_state_enter(spa, SCL_ALL);
error = vdev_label_write_bootenv(spa->spa_root_vdev, innvl);
(void) spa_vdev_state_exit(spa, NULL, 0);
spa_close(spa, FTAG);
return (error);
}
static const zfs_ioc_key_t zfs_keys_get_bootenv[] = {
/* no nvl keys */
};
static int
zfs_ioc_get_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
{
spa_t *spa;
int error;
if ((error = spa_open(name, &spa, FTAG)) != 0)
return (error);
spa_vdev_state_enter(spa, SCL_ALL);
error = vdev_label_read_bootenv(spa->spa_root_vdev, outnvl);
(void) spa_vdev_state_exit(spa, NULL, 0);
spa_close(spa, FTAG);
return (error);
}
/*
* The dp_config_rwlock must not be held when calling this, because the
* unmount may need to write out data.
*
* This function is best-effort. Callers must deal gracefully if it
* remains mounted (or is remounted after this call).
*
* Returns 0 if the argument is not a snapshot, or it is not currently a
* filesystem, or we were able to unmount it. Returns error code otherwise.
*/
void
zfs_unmount_snap(const char *snapname)
{
if (strchr(snapname, '@') == NULL)
return;
(void) zfsctl_snapshot_unmount(snapname, MNT_FORCE);
}
/* ARGSUSED */
static int
zfs_unmount_snap_cb(const char *snapname, void *arg)
{
zfs_unmount_snap(snapname);
return (0);
}
/*
* When a clone is destroyed, its origin may also need to be destroyed,
* in which case it must be unmounted. This routine will do that unmount
* if necessary.
*/
void
zfs_destroy_unmount_origin(const char *fsname)
{
int error;
objset_t *os;
dsl_dataset_t *ds;
error = dmu_objset_hold(fsname, FTAG, &os);
if (error != 0)
return;
ds = dmu_objset_ds(os);
if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
char originname[ZFS_MAX_DATASET_NAME_LEN];
dsl_dataset_name(ds->ds_prev, originname);
dmu_objset_rele(os, FTAG);
zfs_unmount_snap(originname);
} else {
dmu_objset_rele(os, FTAG);
}
}
/*
* innvl: {
* "snaps" -> { snapshot1, snapshot2 }
* (optional boolean) "defer"
* }
*
* outnvl: snapshot -> error code (int32)
*/
static const zfs_ioc_key_t zfs_keys_destroy_snaps[] = {
{"snaps", DATA_TYPE_NVLIST, 0},
{"defer", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
};
/* ARGSUSED */
static int
zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
{
int poollen;
nvlist_t *snaps;
nvpair_t *pair;
boolean_t defer;
spa_t *spa;
snaps = fnvlist_lookup_nvlist(innvl, "snaps");
defer = nvlist_exists(innvl, "defer");
poollen = strlen(poolname);
for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
pair = nvlist_next_nvpair(snaps, pair)) {
const char *name = nvpair_name(pair);
/*
* The snap must be in the specified pool to prevent the
* invalid removal of zvol minors below.
*/
if (strncmp(name, poolname, poollen) != 0 ||
(name[poollen] != '/' && name[poollen] != '@'))
return (SET_ERROR(EXDEV));
zfs_unmount_snap(nvpair_name(pair));
if (spa_open(name, &spa, FTAG) == 0) {
zvol_remove_minors(spa, name, B_TRUE);
spa_close(spa, FTAG);
}
}
return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
}
/*
* Create bookmarks. The bookmark names are of the form <fs>#<bmark>.
* All bookmarks and snapshots must be in the same pool.
* dsl_bookmark_create_nvl_validate describes the nvlist schema in more detail.
*
* innvl: {
* new_bookmark1 -> existing_snapshot,
* new_bookmark2 -> existing_bookmark,
* }
*
* outnvl: bookmark -> error code (int32)
*
*/
static const zfs_ioc_key_t zfs_keys_bookmark[] = {
{"<bookmark>...", DATA_TYPE_STRING, ZK_WILDCARDLIST},
};
/* ARGSUSED */
static int
zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
{
return (dsl_bookmark_create(innvl, outnvl));
}
/*
* innvl: {
* property 1, property 2, ...
* }
*
* outnvl: {
* bookmark name 1 -> { property 1, property 2, ... },
* bookmark name 2 -> { property 1, property 2, ... }
* }
*
*/
static const zfs_ioc_key_t zfs_keys_get_bookmarks[] = {
{"<property>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST | ZK_OPTIONAL},
};
static int
zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
{
return (dsl_get_bookmarks(fsname, innvl, outnvl));
}
/*
* innvl is not used.
*
* outnvl: {
* property 1, property 2, ...
* }
*
*/
static const zfs_ioc_key_t zfs_keys_get_bookmark_props[] = {
/* no nvl keys */
};
/* ARGSUSED */
static int
zfs_ioc_get_bookmark_props(const char *bookmark, nvlist_t *innvl,
nvlist_t *outnvl)
{
char fsname[ZFS_MAX_DATASET_NAME_LEN];
char *bmname;
bmname = strchr(bookmark, '#');
if (bmname == NULL)
return (SET_ERROR(EINVAL));
bmname++;
(void) strlcpy(fsname, bookmark, sizeof (fsname));
*(strchr(fsname, '#')) = '\0';
return (dsl_get_bookmark_props(fsname, bmname, outnvl));
}
/*
* innvl: {
* bookmark name 1, bookmark name 2
* }
*
* outnvl: bookmark -> error code (int32)
*
*/
static const zfs_ioc_key_t zfs_keys_destroy_bookmarks[] = {
{"<bookmark>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST},
};
static int
zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
nvlist_t *outnvl)
{
int error, poollen;
poollen = strlen(poolname);
for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
const char *name = nvpair_name(pair);
const char *cp = strchr(name, '#');
/*
* The bookmark name must contain an #, and the part after it
* must contain only valid characters.
*/
if (cp == NULL ||
zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
return (SET_ERROR(EINVAL));
/*
* The bookmark must be in the specified pool.
*/
if (strncmp(name, poolname, poollen) != 0 ||
(name[poollen] != '/' && name[poollen] != '#'))
return (SET_ERROR(EXDEV));
}
error = dsl_bookmark_destroy(innvl, outnvl);
return (error);
}
static const zfs_ioc_key_t zfs_keys_channel_program[] = {
{"program", DATA_TYPE_STRING, 0},
{"arg", DATA_TYPE_ANY, 0},
{"sync", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
{"instrlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
{"memlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
};
static int
zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
nvlist_t *outnvl)
{
char *program;
uint64_t instrlimit, memlimit;
boolean_t sync_flag;
nvpair_t *nvarg = NULL;
program = fnvlist_lookup_string(innvl, ZCP_ARG_PROGRAM);
if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
sync_flag = B_TRUE;
}
if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
instrlimit = ZCP_DEFAULT_INSTRLIMIT;
}
if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
memlimit = ZCP_DEFAULT_MEMLIMIT;
}
nvarg = fnvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST);
if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
return (SET_ERROR(EINVAL));
if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
return (SET_ERROR(EINVAL));
return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
nvarg, outnvl));
}
/*
* innvl: unused
* outnvl: empty
*/
static const zfs_ioc_key_t zfs_keys_pool_checkpoint[] = {
/* no nvl keys */
};
/* ARGSUSED */
static int
zfs_ioc_pool_checkpoint(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
{
return (spa_checkpoint(poolname));
}
/*
* innvl: unused
* outnvl: empty
*/
static const zfs_ioc_key_t zfs_keys_pool_discard_checkpoint[] = {
/* no nvl keys */
};
/* ARGSUSED */
static int
zfs_ioc_pool_discard_checkpoint(const char *poolname, nvlist_t *innvl,
nvlist_t *outnvl)
{
return (spa_checkpoint_discard(poolname));
}
/*
* inputs:
* zc_name name of dataset to destroy
* zc_defer_destroy mark for deferred destroy
*
* outputs: none
*/
static int
zfs_ioc_destroy(zfs_cmd_t *zc)
{
objset_t *os;
dmu_objset_type_t ost;
int err;
err = dmu_objset_hold(zc->zc_name, FTAG, &os);
if (err != 0)
return (err);
ost = dmu_objset_type(os);
dmu_objset_rele(os, FTAG);
if (ost == DMU_OST_ZFS)
zfs_unmount_snap(zc->zc_name);
if (strchr(zc->zc_name, '@')) {
err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
} else {
err = dsl_destroy_head(zc->zc_name);
if (err == EEXIST) {
/*
* It is possible that the given DS may have
* hidden child (%recv) datasets - "leftovers"
* resulting from the previously interrupted
* 'zfs receive'.
*
* 6 extra bytes for /%recv
*/
char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];
if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
zc->zc_name, recv_clone_name) >=
sizeof (namebuf))
return (SET_ERROR(EINVAL));
/*
* Try to remove the hidden child (%recv) and after
* that try to remove the target dataset.
* If the hidden child (%recv) does not exist
* the original error (EEXIST) will be returned
*/
err = dsl_destroy_head(namebuf);
if (err == 0)
err = dsl_destroy_head(zc->zc_name);
else if (err == ENOENT)
err = SET_ERROR(EEXIST);
}
}
return (err);
}
/*
* innvl: {
* "initialize_command" -> POOL_INITIALIZE_{CANCEL|START|SUSPEND} (uint64)
* "initialize_vdevs": { -> guids to initialize (nvlist)
* "vdev_path_1": vdev_guid_1, (uint64),
* "vdev_path_2": vdev_guid_2, (uint64),
* ...
* },
* }
*
* outnvl: {
* "initialize_vdevs": { -> initialization errors (nvlist)
* "vdev_path_1": errno, see function body for possible errnos (uint64)
* "vdev_path_2": errno, ... (uint64)
* ...
* }
* }
*
* EINVAL is returned for an unknown commands or if any of the provided vdev
* guids have be specified with a type other than uint64.
*/
static const zfs_ioc_key_t zfs_keys_pool_initialize[] = {
{ZPOOL_INITIALIZE_COMMAND, DATA_TYPE_UINT64, 0},
{ZPOOL_INITIALIZE_VDEVS, DATA_TYPE_NVLIST, 0}
};
static int
zfs_ioc_pool_initialize(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
{
uint64_t cmd_type;
if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
&cmd_type) != 0) {
return (SET_ERROR(EINVAL));
}
if (!(cmd_type == POOL_INITIALIZE_CANCEL ||
cmd_type == POOL_INITIALIZE_START ||
cmd_type == POOL_INITIALIZE_SUSPEND)) {
return (SET_ERROR(EINVAL));
}
nvlist_t *vdev_guids;
if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
&vdev_guids) != 0) {
return (SET_ERROR(EINVAL));
}
for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
uint64_t vdev_guid;
if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
return (SET_ERROR(EINVAL));
}
}
spa_t *spa;
int error = spa_open(poolname, &spa, FTAG);
if (error != 0)
return (error);
nvlist_t *vdev_errlist = fnvlist_alloc();
int total_errors = spa_vdev_initialize(spa, vdev_guids, cmd_type,
vdev_errlist);
if (fnvlist_size(vdev_errlist) > 0) {
fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
vdev_errlist);
}
fnvlist_free(vdev_errlist);
spa_close(spa, FTAG);
return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
}
/*
* innvl: {
* "trim_command" -> POOL_TRIM_{CANCEL|START|SUSPEND} (uint64)
* "trim_vdevs": { -> guids to TRIM (nvlist)
* "vdev_path_1": vdev_guid_1, (uint64),
* "vdev_path_2": vdev_guid_2, (uint64),
* ...
* },
* "trim_rate" -> Target TRIM rate in bytes/sec.
* "trim_secure" -> Set to request a secure TRIM.
* }
*
* outnvl: {
* "trim_vdevs": { -> TRIM errors (nvlist)
* "vdev_path_1": errno, see function body for possible errnos (uint64)
* "vdev_path_2": errno, ... (uint64)
* ...
* }
* }
*
* EINVAL is returned for an unknown commands or if any of the provided vdev
* guids have be specified with a type other than uint64.
*/
static const zfs_ioc_key_t zfs_keys_pool_trim[] = {
{ZPOOL_TRIM_COMMAND, DATA_TYPE_UINT64, 0},
{ZPOOL_TRIM_VDEVS, DATA_TYPE_NVLIST, 0},
{ZPOOL_TRIM_RATE, DATA_TYPE_UINT64, ZK_OPTIONAL},
{ZPOOL_TRIM_SECURE, DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
};
static int
zfs_ioc_pool_trim(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
{
uint64_t cmd_type;
if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_COMMAND, &cmd_type) != 0)
return (SET_ERROR(EINVAL));
if (!(cmd_type == POOL_TRIM_CANCEL ||
cmd_type == POOL_TRIM_START ||
cmd_type == POOL_TRIM_SUSPEND)) {
return (SET_ERROR(EINVAL));
}
nvlist_t *vdev_guids;
if (nvlist_lookup_nvlist(innvl, ZPOOL_TRIM_VDEVS, &vdev_guids) != 0)
return (SET_ERROR(EINVAL));
for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
uint64_t vdev_guid;
if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
return (SET_ERROR(EINVAL));
}
}
/* Optional, defaults to maximum rate when not provided */
uint64_t rate;
if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_RATE, &rate) != 0)
rate = 0;
/* Optional, defaults to standard TRIM when not provided */
boolean_t secure;
if (nvlist_lookup_boolean_value(innvl, ZPOOL_TRIM_SECURE,
&secure) != 0) {
secure = B_FALSE;
}
spa_t *spa;
int error = spa_open(poolname, &spa, FTAG);
if (error != 0)
return (error);
nvlist_t *vdev_errlist = fnvlist_alloc();
int total_errors = spa_vdev_trim(spa, vdev_guids, cmd_type,
rate, !!zfs_trim_metaslab_skip, secure, vdev_errlist);
if (fnvlist_size(vdev_errlist) > 0)
fnvlist_add_nvlist(outnvl, ZPOOL_TRIM_VDEVS, vdev_errlist);
fnvlist_free(vdev_errlist);
spa_close(spa, FTAG);
return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
}
/*
* This ioctl waits for activity of a particular type to complete. If there is
* no activity of that type in progress, it returns immediately, and the
* returned value "waited" is false. If there is activity in progress, and no
* tag is passed in, the ioctl blocks until all activity of that type is
* complete, and then returns with "waited" set to true.
*
* If a tag is provided, it identifies a particular instance of an activity to
* wait for. Currently, this is only valid for use with 'initialize', because
* that is the only activity for which there can be multiple instances running
* concurrently. In the case of 'initialize', the tag corresponds to the guid of
* the vdev on which to wait.
*
* If a thread waiting in the ioctl receives a signal, the call will return
* immediately, and the return value will be EINTR.
*
* innvl: {
* "wait_activity" -> int32_t
* (optional) "wait_tag" -> uint64_t
* }
*
* outnvl: "waited" -> boolean_t
*/
static const zfs_ioc_key_t zfs_keys_pool_wait[] = {
{ZPOOL_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
{ZPOOL_WAIT_TAG, DATA_TYPE_UINT64, ZK_OPTIONAL},
};
static int
zfs_ioc_wait(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
{
int32_t activity;
uint64_t tag;
boolean_t waited;
int error;
if (nvlist_lookup_int32(innvl, ZPOOL_WAIT_ACTIVITY, &activity) != 0)
return (EINVAL);
if (nvlist_lookup_uint64(innvl, ZPOOL_WAIT_TAG, &tag) == 0)
error = spa_wait_tag(name, activity, tag, &waited);
else
error = spa_wait(name, activity, &waited);
if (error == 0)
fnvlist_add_boolean_value(outnvl, ZPOOL_WAIT_WAITED, waited);
return (error);
}
/*
* This ioctl waits for activity of a particular type to complete. If there is
* no activity of that type in progress, it returns immediately, and the
* returned value "waited" is false. If there is activity in progress, and no
* tag is passed in, the ioctl blocks until all activity of that type is
* complete, and then returns with "waited" set to true.
*
* If a thread waiting in the ioctl receives a signal, the call will return
* immediately, and the return value will be EINTR.
*
* innvl: {
* "wait_activity" -> int32_t
* }
*
* outnvl: "waited" -> boolean_t
*/
static const zfs_ioc_key_t zfs_keys_fs_wait[] = {
{ZFS_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
};
static int
zfs_ioc_wait_fs(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
{
int32_t activity;
boolean_t waited = B_FALSE;
int error;
dsl_pool_t *dp;
dsl_dir_t *dd;
dsl_dataset_t *ds;
if (nvlist_lookup_int32(innvl, ZFS_WAIT_ACTIVITY, &activity) != 0)
return (SET_ERROR(EINVAL));
if (activity >= ZFS_WAIT_NUM_ACTIVITIES || activity < 0)
return (SET_ERROR(EINVAL));
if ((error = dsl_pool_hold(name, FTAG, &dp)) != 0)
return (error);
if ((error = dsl_dataset_hold(dp, name, FTAG, &ds)) != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
dd = ds->ds_dir;
mutex_enter(&dd->dd_activity_lock);
dd->dd_activity_waiters++;
/*
* We get a long-hold here so that the dsl_dataset_t and dsl_dir_t
* aren't evicted while we're waiting. Normally this is prevented by
* holding the pool, but we can't do that while we're waiting since
* that would prevent TXGs from syncing out. Some of the functionality
* of long-holds (e.g. preventing deletion) is unnecessary for this
* case, since we would cancel the waiters before proceeding with a
* deletion. An alternative mechanism for keeping the dataset around
* could be developed but this is simpler.
*/
dsl_dataset_long_hold(ds, FTAG);
dsl_pool_rele(dp, FTAG);
error = dsl_dir_wait(dd, ds, activity, &waited);
dsl_dataset_long_rele(ds, FTAG);
dd->dd_activity_waiters--;
if (dd->dd_activity_waiters == 0)
cv_signal(&dd->dd_activity_cv);
mutex_exit(&dd->dd_activity_lock);
dsl_dataset_rele(ds, FTAG);
if (error == 0)
fnvlist_add_boolean_value(outnvl, ZFS_WAIT_WAITED, waited);
return (error);
}
/*
* fsname is name of dataset to rollback (to most recent snapshot)
*
* innvl may contain name of expected target snapshot
*
* outnvl: "target" -> name of most recent snapshot
* }
*/
static const zfs_ioc_key_t zfs_keys_rollback[] = {
{"target", DATA_TYPE_STRING, ZK_OPTIONAL},
};
/* ARGSUSED */
static int
zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
{
zfsvfs_t *zfsvfs;
zvol_state_handle_t *zv;
char *target = NULL;
int error;
(void) nvlist_lookup_string(innvl, "target", &target);
if (target != NULL) {
const char *cp = strchr(target, '@');
/*
* The snap name must contain an @, and the part after it must
* contain only valid characters.
*/
if (cp == NULL ||
zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
return (SET_ERROR(EINVAL));
}
if (getzfsvfs(fsname, &zfsvfs) == 0) {
dsl_dataset_t *ds;
ds = dmu_objset_ds(zfsvfs->z_os);
error = zfs_suspend_fs(zfsvfs);
if (error == 0) {
int resume_err;
error = dsl_dataset_rollback(fsname, target, zfsvfs,
outnvl);
resume_err = zfs_resume_fs(zfsvfs, ds);
error = error ? error : resume_err;
}
zfs_vfs_rele(zfsvfs);
} else if ((zv = zvol_suspend(fsname)) != NULL) {
error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
outnvl);
zvol_resume(zv);
} else {
error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
}
return (error);
}
static int
recursive_unmount(const char *fsname, void *arg)
{
const char *snapname = arg;
char *fullname;
fullname = kmem_asprintf("%s@%s", fsname, snapname);
zfs_unmount_snap(fullname);
kmem_strfree(fullname);
return (0);
}
/*
*
* snapname is the snapshot to redact.
* innvl: {
* "bookname" -> (string)
* shortname of the redaction bookmark to generate
* "snapnv" -> (nvlist, values ignored)
* snapshots to redact snapname with respect to
* }
*
* outnvl is unused
*/
/* ARGSUSED */
static const zfs_ioc_key_t zfs_keys_redact[] = {
{"bookname", DATA_TYPE_STRING, 0},
{"snapnv", DATA_TYPE_NVLIST, 0},
};
static int
zfs_ioc_redact(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
{
nvlist_t *redactnvl = NULL;
char *redactbook = NULL;
if (nvlist_lookup_nvlist(innvl, "snapnv", &redactnvl) != 0)
return (SET_ERROR(EINVAL));
if (fnvlist_num_pairs(redactnvl) == 0)
return (SET_ERROR(ENXIO));
if (nvlist_lookup_string(innvl, "bookname", &redactbook) != 0)
return (SET_ERROR(EINVAL));
return (dmu_redact_snap(snapname, redactnvl, redactbook));
}
/*
* inputs:
* zc_name old name of dataset
* zc_value new name of dataset
* zc_cookie recursive flag (only valid for snapshots)
*
* outputs: none
*/
static int
zfs_ioc_rename(zfs_cmd_t *zc)
{
objset_t *os;
dmu_objset_type_t ost;
boolean_t recursive = zc->zc_cookie & 1;
boolean_t nounmount = !!(zc->zc_cookie & 2);
char *at;
int err;
/* "zfs rename" from and to ...%recv datasets should both fail */
zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
return (SET_ERROR(EINVAL));
err = dmu_objset_hold(zc->zc_name, FTAG, &os);
if (err != 0)
return (err);
ost = dmu_objset_type(os);
dmu_objset_rele(os, FTAG);
at = strchr(zc->zc_name, '@');
if (at != NULL) {
/* snaps must be in same fs */
int error;
if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
return (SET_ERROR(EXDEV));
*at = '\0';
if (ost == DMU_OST_ZFS && !nounmount) {
error = dmu_objset_find(zc->zc_name,
recursive_unmount, at + 1,
recursive ? DS_FIND_CHILDREN : 0);
if (error != 0) {
*at = '@';
return (error);
}
}
error = dsl_dataset_rename_snapshot(zc->zc_name,
at + 1, strchr(zc->zc_value, '@') + 1, recursive);
*at = '@';
return (error);
} else {
return (dsl_dir_rename(zc->zc_name, zc->zc_value));
}
}
static int
zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
{
const char *propname = nvpair_name(pair);
boolean_t issnap = (strchr(dsname, '@') != NULL);
zfs_prop_t prop = zfs_name_to_prop(propname);
uint64_t intval, compval;
int err;
if (prop == ZPROP_INVAL) {
if (zfs_prop_user(propname)) {
if ((err = zfs_secpolicy_write_perms(dsname,
ZFS_DELEG_PERM_USERPROP, cr)))
return (err);
return (0);
}
if (!issnap && zfs_prop_userquota(propname)) {
const char *perm = NULL;
const char *uq_prefix =
zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
const char *gq_prefix =
zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
const char *uiq_prefix =
zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
const char *giq_prefix =
zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
const char *pq_prefix =
zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
const char *piq_prefix = zfs_userquota_prop_prefixes[\
ZFS_PROP_PROJECTOBJQUOTA];
if (strncmp(propname, uq_prefix,
strlen(uq_prefix)) == 0) {
perm = ZFS_DELEG_PERM_USERQUOTA;
} else if (strncmp(propname, uiq_prefix,
strlen(uiq_prefix)) == 0) {
perm = ZFS_DELEG_PERM_USEROBJQUOTA;
} else if (strncmp(propname, gq_prefix,
strlen(gq_prefix)) == 0) {
perm = ZFS_DELEG_PERM_GROUPQUOTA;
} else if (strncmp(propname, giq_prefix,
strlen(giq_prefix)) == 0) {
perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
} else if (strncmp(propname, pq_prefix,
strlen(pq_prefix)) == 0) {
perm = ZFS_DELEG_PERM_PROJECTQUOTA;
} else if (strncmp(propname, piq_prefix,
strlen(piq_prefix)) == 0) {
perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
} else {
/* {USER|GROUP|PROJECT}USED are read-only */
return (SET_ERROR(EINVAL));
}
if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
return (err);
return (0);
}
return (SET_ERROR(EINVAL));
}
if (issnap)
return (SET_ERROR(EINVAL));
if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
/*
* dsl_prop_get_all_impl() returns properties in this
* format.
*/
nvlist_t *attrs;
VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
&pair) == 0);
}
/*
* Check that this value is valid for this pool version
*/
switch (prop) {
case ZFS_PROP_COMPRESSION:
/*
* If the user specified gzip compression, make sure
* the SPA supports it. We ignore any errors here since
* we'll catch them later.
*/
if (nvpair_value_uint64(pair, &intval) == 0) {
compval = ZIO_COMPRESS_ALGO(intval);
if (compval >= ZIO_COMPRESS_GZIP_1 &&
compval <= ZIO_COMPRESS_GZIP_9 &&
zfs_earlier_version(dsname,
SPA_VERSION_GZIP_COMPRESSION)) {
return (SET_ERROR(ENOTSUP));
}
if (compval == ZIO_COMPRESS_ZLE &&
zfs_earlier_version(dsname,
SPA_VERSION_ZLE_COMPRESSION))
return (SET_ERROR(ENOTSUP));
if (compval == ZIO_COMPRESS_LZ4) {
spa_t *spa;
if ((err = spa_open(dsname, &spa, FTAG)) != 0)
return (err);
if (!spa_feature_is_enabled(spa,
SPA_FEATURE_LZ4_COMPRESS)) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOTSUP));
}
spa_close(spa, FTAG);
}
if (compval == ZIO_COMPRESS_ZSTD) {
spa_t *spa;
if ((err = spa_open(dsname, &spa, FTAG)) != 0)
return (err);
if (!spa_feature_is_enabled(spa,
SPA_FEATURE_ZSTD_COMPRESS)) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOTSUP));
}
spa_close(spa, FTAG);
}
}
break;
case ZFS_PROP_COPIES:
if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
return (SET_ERROR(ENOTSUP));
break;
case ZFS_PROP_VOLBLOCKSIZE:
case ZFS_PROP_RECORDSIZE:
/* Record sizes above 128k need the feature to be enabled */
if (nvpair_value_uint64(pair, &intval) == 0 &&
intval > SPA_OLD_MAXBLOCKSIZE) {
spa_t *spa;
/*
* We don't allow setting the property above 1MB,
* unless the tunable has been changed.
*/
if (intval > zfs_max_recordsize ||
intval > SPA_MAXBLOCKSIZE)
return (SET_ERROR(ERANGE));
if ((err = spa_open(dsname, &spa, FTAG)) != 0)
return (err);
if (!spa_feature_is_enabled(spa,
SPA_FEATURE_LARGE_BLOCKS)) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOTSUP));
}
spa_close(spa, FTAG);
}
break;
case ZFS_PROP_DNODESIZE:
/* Dnode sizes above 512 need the feature to be enabled */
if (nvpair_value_uint64(pair, &intval) == 0 &&
intval != ZFS_DNSIZE_LEGACY) {
spa_t *spa;
if ((err = spa_open(dsname, &spa, FTAG)) != 0)
return (err);
if (!spa_feature_is_enabled(spa,
SPA_FEATURE_LARGE_DNODE)) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOTSUP));
}
spa_close(spa, FTAG);
}
break;
case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
/*
* This property could require the allocation classes
* feature to be active for setting, however we allow
* it so that tests of settable properties succeed.
* The CLI will issue a warning in this case.
*/
break;
case ZFS_PROP_SHARESMB:
if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
return (SET_ERROR(ENOTSUP));
break;
case ZFS_PROP_ACLINHERIT:
if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
nvpair_value_uint64(pair, &intval) == 0) {
if (intval == ZFS_ACL_PASSTHROUGH_X &&
zfs_earlier_version(dsname,
SPA_VERSION_PASSTHROUGH_X))
return (SET_ERROR(ENOTSUP));
}
break;
case ZFS_PROP_CHECKSUM:
case ZFS_PROP_DEDUP:
{
spa_feature_t feature;
spa_t *spa;
int err;
/* dedup feature version checks */
if (prop == ZFS_PROP_DEDUP &&
zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
return (SET_ERROR(ENOTSUP));
if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
nvpair_value_uint64(pair, &intval) == 0) {
/* check prop value is enabled in features */
feature = zio_checksum_to_feature(
intval & ZIO_CHECKSUM_MASK);
if (feature == SPA_FEATURE_NONE)
break;
if ((err = spa_open(dsname, &spa, FTAG)) != 0)
return (err);
if (!spa_feature_is_enabled(spa, feature)) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOTSUP));
}
spa_close(spa, FTAG);
}
break;
}
default:
break;
}
return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
}
/*
* Removes properties from the given props list that fail permission checks
* needed to clear them and to restore them in case of a receive error. For each
* property, make sure we have both set and inherit permissions.
*
* Returns the first error encountered if any permission checks fail. If the
* caller provides a non-NULL errlist, it also gives the complete list of names
* of all the properties that failed a permission check along with the
* corresponding error numbers. The caller is responsible for freeing the
* returned errlist.
*
* If every property checks out successfully, zero is returned and the list
* pointed at by errlist is NULL.
*/
static int
zfs_check_clearable(const char *dataset, nvlist_t *props, nvlist_t **errlist)
{
zfs_cmd_t *zc;
nvpair_t *pair, *next_pair;
nvlist_t *errors;
int err, rv = 0;
if (props == NULL)
return (0);
VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
(void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
pair = nvlist_next_nvpair(props, NULL);
while (pair != NULL) {
next_pair = nvlist_next_nvpair(props, pair);
(void) strlcpy(zc->zc_value, nvpair_name(pair),
sizeof (zc->zc_value));
if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
(err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
VERIFY(nvlist_remove_nvpair(props, pair) == 0);
VERIFY(nvlist_add_int32(errors,
zc->zc_value, err) == 0);
}
pair = next_pair;
}
kmem_free(zc, sizeof (zfs_cmd_t));
if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
nvlist_free(errors);
errors = NULL;
} else {
VERIFY(nvpair_value_int32(pair, &rv) == 0);
}
if (errlist == NULL)
nvlist_free(errors);
else
*errlist = errors;
return (rv);
}
static boolean_t
propval_equals(nvpair_t *p1, nvpair_t *p2)
{
if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
/* dsl_prop_get_all_impl() format */
nvlist_t *attrs;
VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
&p1) == 0);
}
if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
nvlist_t *attrs;
VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
&p2) == 0);
}
if (nvpair_type(p1) != nvpair_type(p2))
return (B_FALSE);
if (nvpair_type(p1) == DATA_TYPE_STRING) {
char *valstr1, *valstr2;
VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
return (strcmp(valstr1, valstr2) == 0);
} else {
uint64_t intval1, intval2;
VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
return (intval1 == intval2);
}
}
/*
* Remove properties from props if they are not going to change (as determined
* by comparison with origprops). Remove them from origprops as well, since we
* do not need to clear or restore properties that won't change.
*/
static void
props_reduce(nvlist_t *props, nvlist_t *origprops)
{
nvpair_t *pair, *next_pair;
if (origprops == NULL)
return; /* all props need to be received */
pair = nvlist_next_nvpair(props, NULL);
while (pair != NULL) {
const char *propname = nvpair_name(pair);
nvpair_t *match;
next_pair = nvlist_next_nvpair(props, pair);
if ((nvlist_lookup_nvpair(origprops, propname,
&match) != 0) || !propval_equals(pair, match))
goto next; /* need to set received value */
/* don't clear the existing received value */
(void) nvlist_remove_nvpair(origprops, match);
/* don't bother receiving the property */
(void) nvlist_remove_nvpair(props, pair);
next:
pair = next_pair;
}
}
/*
* Extract properties that cannot be set PRIOR to the receipt of a dataset.
* For example, refquota cannot be set until after the receipt of a dataset,
* because in replication streams, an older/earlier snapshot may exceed the
* refquota. We want to receive the older/earlier snapshot, but setting
* refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
* the older/earlier snapshot from being received (with EDQUOT).
*
* The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
*
* libzfs will need to be judicious handling errors encountered by props
* extracted by this function.
*/
static nvlist_t *
extract_delay_props(nvlist_t *props)
{
nvlist_t *delayprops;
nvpair_t *nvp, *tmp;
static const zfs_prop_t delayable[] = {
ZFS_PROP_REFQUOTA,
ZFS_PROP_KEYLOCATION,
0
};
int i;
VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
nvp = nvlist_next_nvpair(props, nvp)) {
/*
* strcmp() is safe because zfs_prop_to_name() always returns
* a bounded string.
*/
for (i = 0; delayable[i] != 0; i++) {
if (strcmp(zfs_prop_to_name(delayable[i]),
nvpair_name(nvp)) == 0) {
break;
}
}
if (delayable[i] != 0) {
tmp = nvlist_prev_nvpair(props, nvp);
VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
nvp = tmp;
}
}
if (nvlist_empty(delayprops)) {
nvlist_free(delayprops);
delayprops = NULL;
}
return (delayprops);
}
static void
zfs_allow_log_destroy(void *arg)
{
char *poolname = arg;
if (poolname != NULL)
kmem_strfree(poolname);
}
#ifdef ZFS_DEBUG
static boolean_t zfs_ioc_recv_inject_err;
#endif
/*
* nvlist 'errors' is always allocated. It will contain descriptions of
* encountered errors, if any. It's the callers responsibility to free.
*/
static int
zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
nvlist_t *localprops, nvlist_t *hidden_args, boolean_t force,
boolean_t resumable, int input_fd,
dmu_replay_record_t *begin_record, uint64_t *read_bytes,
uint64_t *errflags, nvlist_t **errors)
{
dmu_recv_cookie_t drc;
int error = 0;
int props_error = 0;
offset_t off, noff;
nvlist_t *local_delayprops = NULL;
nvlist_t *recv_delayprops = NULL;
nvlist_t *origprops = NULL; /* existing properties */
nvlist_t *origrecvd = NULL; /* existing received properties */
boolean_t first_recvd_props = B_FALSE;
boolean_t tofs_was_redacted;
zfs_file_t *input_fp;
*read_bytes = 0;
*errflags = 0;
*errors = fnvlist_alloc();
off = 0;
if ((input_fp = zfs_file_get(input_fd)) == NULL)
return (SET_ERROR(EBADF));
noff = off = zfs_file_off(input_fp);
error = dmu_recv_begin(tofs, tosnap, begin_record, force,
resumable, localprops, hidden_args, origin, &drc, input_fp,
&off);
if (error != 0)
goto out;
tofs_was_redacted = dsl_get_redacted(drc.drc_ds);
/*
* Set properties before we receive the stream so that they are applied
* to the new data. Note that we must call dmu_recv_stream() if
* dmu_recv_begin() succeeds.
*/
if (recvprops != NULL && !drc.drc_newfs) {
if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
SPA_VERSION_RECVD_PROPS &&
!dsl_prop_get_hasrecvd(tofs))
first_recvd_props = B_TRUE;
/*
* If new received properties are supplied, they are to
* completely replace the existing received properties,
* so stash away the existing ones.
*/
if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
nvlist_t *errlist = NULL;
/*
* Don't bother writing a property if its value won't
* change (and avoid the unnecessary security checks).
*
* The first receive after SPA_VERSION_RECVD_PROPS is a
* special case where we blow away all local properties
* regardless.
*/
if (!first_recvd_props)
props_reduce(recvprops, origrecvd);
if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
(void) nvlist_merge(*errors, errlist, 0);
nvlist_free(errlist);
if (clear_received_props(tofs, origrecvd,
first_recvd_props ? NULL : recvprops) != 0)
*errflags |= ZPROP_ERR_NOCLEAR;
} else {
*errflags |= ZPROP_ERR_NOCLEAR;
}
}
/*
* Stash away existing properties so we can restore them on error unless
* we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
* case "origrecvd" will take care of that.
*/
if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
objset_t *os;
if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
if (dsl_prop_get_all(os, &origprops) != 0) {
*errflags |= ZPROP_ERR_NOCLEAR;
}
dmu_objset_rele(os, FTAG);
} else {
*errflags |= ZPROP_ERR_NOCLEAR;
}
}
if (recvprops != NULL) {
props_error = dsl_prop_set_hasrecvd(tofs);
if (props_error == 0) {
recv_delayprops = extract_delay_props(recvprops);
(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
recvprops, *errors);
}
}
if (localprops != NULL) {
nvlist_t *oprops = fnvlist_alloc();
nvlist_t *xprops = fnvlist_alloc();
nvpair_t *nvp = NULL;
while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
/* -x property */
const char *name = nvpair_name(nvp);
zfs_prop_t prop = zfs_name_to_prop(name);
if (prop != ZPROP_INVAL) {
if (!zfs_prop_inheritable(prop))
continue;
} else if (!zfs_prop_user(name))
continue;
fnvlist_add_boolean(xprops, name);
} else {
/* -o property=value */
fnvlist_add_nvpair(oprops, nvp);
}
}
local_delayprops = extract_delay_props(oprops);
(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
oprops, *errors);
(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
xprops, *errors);
nvlist_free(oprops);
nvlist_free(xprops);
}
error = dmu_recv_stream(&drc, &off);
if (error == 0) {
zfsvfs_t *zfsvfs = NULL;
zvol_state_handle_t *zv = NULL;
if (getzfsvfs(tofs, &zfsvfs) == 0) {
/* online recv */
dsl_dataset_t *ds;
int end_err;
boolean_t stream_is_redacted = DMU_GET_FEATUREFLAGS(
begin_record->drr_u.drr_begin.
drr_versioninfo) & DMU_BACKUP_FEATURE_REDACTED;
ds = dmu_objset_ds(zfsvfs->z_os);
error = zfs_suspend_fs(zfsvfs);
/*
* If the suspend fails, then the recv_end will
* likely also fail, and clean up after itself.
*/
end_err = dmu_recv_end(&drc, zfsvfs);
/*
* If the dataset was not redacted, but we received a
* redacted stream onto it, we need to unmount the
* dataset. Otherwise, resume the filesystem.
*/
if (error == 0 && !drc.drc_newfs &&
stream_is_redacted && !tofs_was_redacted) {
error = zfs_end_fs(zfsvfs, ds);
} else if (error == 0) {
error = zfs_resume_fs(zfsvfs, ds);
}
error = error ? error : end_err;
zfs_vfs_rele(zfsvfs);
} else if ((zv = zvol_suspend(tofs)) != NULL) {
error = dmu_recv_end(&drc, zvol_tag(zv));
zvol_resume(zv);
} else {
error = dmu_recv_end(&drc, NULL);
}
/* Set delayed properties now, after we're done receiving. */
if (recv_delayprops != NULL && error == 0) {
(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
recv_delayprops, *errors);
}
if (local_delayprops != NULL && error == 0) {
(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
local_delayprops, *errors);
}
}
/*
* Merge delayed props back in with initial props, in case
* we're DEBUG and zfs_ioc_recv_inject_err is set (which means
* we have to make sure clear_received_props() includes
* the delayed properties).
*
* Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
* using ASSERT() will be just like a VERIFY.
*/
if (recv_delayprops != NULL) {
ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0);
nvlist_free(recv_delayprops);
}
if (local_delayprops != NULL) {
ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0);
nvlist_free(local_delayprops);
}
*read_bytes = off - noff;
#ifdef ZFS_DEBUG
if (zfs_ioc_recv_inject_err) {
zfs_ioc_recv_inject_err = B_FALSE;
error = 1;
}
#endif
/*
* On error, restore the original props.
*/
if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
if (clear_received_props(tofs, recvprops, NULL) != 0) {
/*
* We failed to clear the received properties.
* Since we may have left a $recvd value on the
* system, we can't clear the $hasrecvd flag.
*/
*errflags |= ZPROP_ERR_NORESTORE;
} else if (first_recvd_props) {
dsl_prop_unset_hasrecvd(tofs);
}
if (origrecvd == NULL && !drc.drc_newfs) {
/* We failed to stash the original properties. */
*errflags |= ZPROP_ERR_NORESTORE;
}
/*
* dsl_props_set() will not convert RECEIVED to LOCAL on or
* after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
* explicitly if we're restoring local properties cleared in the
* first new-style receive.
*/
if (origrecvd != NULL &&
zfs_set_prop_nvlist(tofs, (first_recvd_props ?
ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
origrecvd, NULL) != 0) {
/*
* We stashed the original properties but failed to
* restore them.
*/
*errflags |= ZPROP_ERR_NORESTORE;
}
}
if (error != 0 && localprops != NULL && !drc.drc_newfs &&
!first_recvd_props) {
nvlist_t *setprops;
nvlist_t *inheritprops;
nvpair_t *nvp;
if (origprops == NULL) {
/* We failed to stash the original properties. */
*errflags |= ZPROP_ERR_NORESTORE;
goto out;
}
/* Restore original props */
setprops = fnvlist_alloc();
inheritprops = fnvlist_alloc();
nvp = NULL;
while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
const char *name = nvpair_name(nvp);
const char *source;
nvlist_t *attrs;
if (!nvlist_exists(origprops, name)) {
/*
* Property was not present or was explicitly
* inherited before the receive, restore this.
*/
fnvlist_add_boolean(inheritprops, name);
continue;
}
attrs = fnvlist_lookup_nvlist(origprops, name);
source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);
/* Skip received properties */
if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
continue;
if (strcmp(source, tofs) == 0) {
/* Property was locally set */
fnvlist_add_nvlist(setprops, name, attrs);
} else {
/* Property was implicitly inherited */
fnvlist_add_boolean(inheritprops, name);
}
}
if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
NULL) != 0)
*errflags |= ZPROP_ERR_NORESTORE;
if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
NULL) != 0)
*errflags |= ZPROP_ERR_NORESTORE;
nvlist_free(setprops);
nvlist_free(inheritprops);
}
out:
zfs_file_put(input_fp);
nvlist_free(origrecvd);
nvlist_free(origprops);
if (error == 0)
error = props_error;
return (error);
}
/*
* inputs:
* zc_name name of containing filesystem (unused)
* zc_nvlist_src{_size} nvlist of properties to apply
* zc_nvlist_conf{_size} nvlist of properties to exclude
* (DATA_TYPE_BOOLEAN) and override (everything else)
* zc_value name of snapshot to create
* zc_string name of clone origin (if DRR_FLAG_CLONE)
* zc_cookie file descriptor to recv from
* zc_begin_record the BEGIN record of the stream (not byteswapped)
* zc_guid force flag
*
* outputs:
* zc_cookie number of bytes read
* zc_obj zprop_errflags_t
* zc_nvlist_dst{_size} error for each unapplied received property
*/
static int
zfs_ioc_recv(zfs_cmd_t *zc)
{
dmu_replay_record_t begin_record;
nvlist_t *errors = NULL;
nvlist_t *recvdprops = NULL;
nvlist_t *localprops = NULL;
char *origin = NULL;
char *tosnap;
char tofs[ZFS_MAX_DATASET_NAME_LEN];
int error = 0;
if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
strchr(zc->zc_value, '@') == NULL ||
strchr(zc->zc_value, '%'))
return (SET_ERROR(EINVAL));
(void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
tosnap = strchr(tofs, '@');
*tosnap++ = '\0';
if (zc->zc_nvlist_src != 0 &&
(error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
zc->zc_iflags, &recvdprops)) != 0)
return (error);
if (zc->zc_nvlist_conf != 0 &&
(error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
zc->zc_iflags, &localprops)) != 0)
return (error);
if (zc->zc_string[0])
origin = zc->zc_string;
begin_record.drr_type = DRR_BEGIN;
begin_record.drr_payloadlen = 0;
begin_record.drr_u.drr_begin = zc->zc_begin_record;
error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
NULL, zc->zc_guid, B_FALSE, zc->zc_cookie, &begin_record,
&zc->zc_cookie, &zc->zc_obj, &errors);
nvlist_free(recvdprops);
nvlist_free(localprops);
/*
* Now that all props, initial and delayed, are set, report the prop
* errors to the caller.
*/
if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
(nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
put_nvlist(zc, errors) != 0)) {
/*
* Caller made zc->zc_nvlist_dst less than the minimum expected
* size or supplied an invalid address.
*/
error = SET_ERROR(EINVAL);
}
nvlist_free(errors);
return (error);
}
/*
* innvl: {
* "snapname" -> full name of the snapshot to create
* (optional) "props" -> received properties to set (nvlist)
* (optional) "localprops" -> override and exclude properties (nvlist)
* (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
* "begin_record" -> non-byteswapped dmu_replay_record_t
* "input_fd" -> file descriptor to read stream from (int32)
* (optional) "force" -> force flag (value ignored)
* (optional) "resumable" -> resumable flag (value ignored)
* (optional) "cleanup_fd" -> unused
* (optional) "action_handle" -> unused
* (optional) "hidden_args" -> { "wkeydata" -> value }
* }
*
* outnvl: {
* "read_bytes" -> number of bytes read
* "error_flags" -> zprop_errflags_t
* "errors" -> error for each unapplied received property (nvlist)
* }
*/
static const zfs_ioc_key_t zfs_keys_recv_new[] = {
{"snapname", DATA_TYPE_STRING, 0},
{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
{"localprops", DATA_TYPE_NVLIST, ZK_OPTIONAL},
{"origin", DATA_TYPE_STRING, ZK_OPTIONAL},
{"begin_record", DATA_TYPE_BYTE_ARRAY, 0},
{"input_fd", DATA_TYPE_INT32, 0},
{"force", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"resumable", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
{"action_handle", DATA_TYPE_UINT64, ZK_OPTIONAL},
{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
};
static int
zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
{
dmu_replay_record_t *begin_record;
uint_t begin_record_size;
nvlist_t *errors = NULL;
nvlist_t *recvprops = NULL;
nvlist_t *localprops = NULL;
nvlist_t *hidden_args = NULL;
char *snapname;
char *origin = NULL;
char *tosnap;
char tofs[ZFS_MAX_DATASET_NAME_LEN];
boolean_t force;
boolean_t resumable;
uint64_t read_bytes = 0;
uint64_t errflags = 0;
int input_fd = -1;
int error;
snapname = fnvlist_lookup_string(innvl, "snapname");
if (dataset_namecheck(snapname, NULL, NULL) != 0 ||
strchr(snapname, '@') == NULL ||
strchr(snapname, '%'))
return (SET_ERROR(EINVAL));
(void) strlcpy(tofs, snapname, sizeof (tofs));
tosnap = strchr(tofs, '@');
*tosnap++ = '\0';
error = nvlist_lookup_string(innvl, "origin", &origin);
if (error && error != ENOENT)
return (error);
error = nvlist_lookup_byte_array(innvl, "begin_record",
(uchar_t **)&begin_record, &begin_record_size);
if (error != 0 || begin_record_size != sizeof (*begin_record))
return (SET_ERROR(EINVAL));
input_fd = fnvlist_lookup_int32(innvl, "input_fd");
force = nvlist_exists(innvl, "force");
resumable = nvlist_exists(innvl, "resumable");
/* we still use "props" here for backwards compatibility */
error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
if (error && error != ENOENT)
return (error);
error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
if (error && error != ENOENT)
return (error);
error = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
if (error && error != ENOENT)
return (error);
error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
hidden_args, force, resumable, input_fd, begin_record,
&read_bytes, &errflags, &errors);
fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
fnvlist_add_uint64(outnvl, "error_flags", errflags);
fnvlist_add_nvlist(outnvl, "errors", errors);
nvlist_free(errors);
nvlist_free(recvprops);
nvlist_free(localprops);
return (error);
}
typedef struct dump_bytes_io {
zfs_file_t *dbi_fp;
caddr_t dbi_buf;
int dbi_len;
int dbi_err;
} dump_bytes_io_t;
static void
dump_bytes_cb(void *arg)
{
dump_bytes_io_t *dbi = (dump_bytes_io_t *)arg;
zfs_file_t *fp;
caddr_t buf;
fp = dbi->dbi_fp;
buf = dbi->dbi_buf;
dbi->dbi_err = zfs_file_write(fp, buf, dbi->dbi_len, NULL);
}
static int
dump_bytes(objset_t *os, void *buf, int len, void *arg)
{
dump_bytes_io_t dbi;
dbi.dbi_fp = arg;
dbi.dbi_buf = buf;
dbi.dbi_len = len;
#if defined(HAVE_LARGE_STACKS)
dump_bytes_cb(&dbi);
#else
/*
* The vn_rdwr() call is performed in a taskq to ensure that there is
* always enough stack space to write safely to the target filesystem.
* The ZIO_TYPE_FREE threads are used because there can be a lot of
* them and they are used in vdev_file.c for a similar purpose.
*/
spa_taskq_dispatch_sync(dmu_objset_spa(os), ZIO_TYPE_FREE,
ZIO_TASKQ_ISSUE, dump_bytes_cb, &dbi, TQ_SLEEP);
#endif /* HAVE_LARGE_STACKS */
return (dbi.dbi_err);
}
/*
* inputs:
* zc_name name of snapshot to send
* zc_cookie file descriptor to send stream to
* zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
* zc_sendobj objsetid of snapshot to send
* zc_fromobj objsetid of incremental fromsnap (may be zero)
* zc_guid if set, estimate size of stream only. zc_cookie is ignored.
* output size in zc_objset_type.
* zc_flags lzc_send_flags
*
* outputs:
* zc_objset_type estimated size, if zc_guid is set
*
* NOTE: This is no longer the preferred interface, any new functionality
* should be added to zfs_ioc_send_new() instead.
*/
static int
zfs_ioc_send(zfs_cmd_t *zc)
{
int error;
offset_t off;
boolean_t estimate = (zc->zc_guid != 0);
boolean_t embedok = (zc->zc_flags & 0x1);
boolean_t large_block_ok = (zc->zc_flags & 0x2);
boolean_t compressok = (zc->zc_flags & 0x4);
boolean_t rawok = (zc->zc_flags & 0x8);
boolean_t savedok = (zc->zc_flags & 0x10);
if (zc->zc_obj != 0) {
dsl_pool_t *dp;
dsl_dataset_t *tosnap;
error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
if (dsl_dir_is_clone(tosnap->ds_dir))
zc->zc_fromobj =
dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
dsl_dataset_rele(tosnap, FTAG);
dsl_pool_rele(dp, FTAG);
}
if (estimate) {
dsl_pool_t *dp;
dsl_dataset_t *tosnap;
dsl_dataset_t *fromsnap = NULL;
error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
FTAG, &tosnap);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
if (zc->zc_fromobj != 0) {
error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
FTAG, &fromsnap);
if (error != 0) {
dsl_dataset_rele(tosnap, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
}
error = dmu_send_estimate_fast(tosnap, fromsnap, NULL,
compressok || rawok, savedok, &zc->zc_objset_type);
if (fromsnap != NULL)
dsl_dataset_rele(fromsnap, FTAG);
dsl_dataset_rele(tosnap, FTAG);
dsl_pool_rele(dp, FTAG);
} else {
zfs_file_t *fp;
dmu_send_outparams_t out = {0};
if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
return (SET_ERROR(EBADF));
off = zfs_file_off(fp);
out.dso_outfunc = dump_bytes;
out.dso_arg = fp;
out.dso_dryrun = B_FALSE;
error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
zc->zc_fromobj, embedok, large_block_ok, compressok,
rawok, savedok, zc->zc_cookie, &off, &out);
zfs_file_put(fp);
}
return (error);
}
/*
* inputs:
* zc_name name of snapshot on which to report progress
* zc_cookie file descriptor of send stream
*
* outputs:
* zc_cookie number of bytes written in send stream thus far
* zc_objset_type logical size of data traversed by send thus far
*/
static int
zfs_ioc_send_progress(zfs_cmd_t *zc)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
dmu_sendstatus_t *dsp = NULL;
int error;
error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
mutex_enter(&ds->ds_sendstream_lock);
/*
* Iterate over all the send streams currently active on this dataset.
* If there's one which matches the specified file descriptor _and_ the
* stream was started by the current process, return the progress of
* that stream.
*/
for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
dsp = list_next(&ds->ds_sendstreams, dsp)) {
if (dsp->dss_outfd == zc->zc_cookie &&
zfs_proc_is_caller(dsp->dss_proc))
break;
}
if (dsp != NULL) {
zc->zc_cookie = atomic_cas_64((volatile uint64_t *)dsp->dss_off,
0, 0);
/* This is the closest thing we have to atomic_read_64. */
zc->zc_objset_type = atomic_cas_64(&dsp->dss_blocks, 0, 0);
} else {
error = SET_ERROR(ENOENT);
}
mutex_exit(&ds->ds_sendstream_lock);
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
static int
zfs_ioc_inject_fault(zfs_cmd_t *zc)
{
int id, error;
error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
&zc->zc_inject_record);
if (error == 0)
zc->zc_guid = (uint64_t)id;
return (error);
}
static int
zfs_ioc_clear_fault(zfs_cmd_t *zc)
{
return (zio_clear_fault((int)zc->zc_guid));
}
static int
zfs_ioc_inject_list_next(zfs_cmd_t *zc)
{
int id = (int)zc->zc_guid;
int error;
error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
&zc->zc_inject_record);
zc->zc_guid = id;
return (error);
}
static int
zfs_ioc_error_log(zfs_cmd_t *zc)
{
spa_t *spa;
int error;
size_t count = (size_t)zc->zc_nvlist_dst_size;
if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
return (error);
error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
&count);
if (error == 0)
zc->zc_nvlist_dst_size = count;
else
zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
spa_close(spa, FTAG);
return (error);
}
static int
zfs_ioc_clear(zfs_cmd_t *zc)
{
spa_t *spa;
vdev_t *vd;
int error;
/*
* On zpool clear we also fix up missing slogs
*/
mutex_enter(&spa_namespace_lock);
spa = spa_lookup(zc->zc_name);
if (spa == NULL) {
mutex_exit(&spa_namespace_lock);
return (SET_ERROR(EIO));
}
if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
/* we need to let spa_open/spa_load clear the chains */
spa_set_log_state(spa, SPA_LOG_CLEAR);
}
spa->spa_last_open_failed = 0;
mutex_exit(&spa_namespace_lock);
if (zc->zc_cookie & ZPOOL_NO_REWIND) {
error = spa_open(zc->zc_name, &spa, FTAG);
} else {
nvlist_t *policy;
nvlist_t *config = NULL;
if (zc->zc_nvlist_src == 0)
return (SET_ERROR(EINVAL));
if ((error = get_nvlist(zc->zc_nvlist_src,
zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
error = spa_open_rewind(zc->zc_name, &spa, FTAG,
policy, &config);
if (config != NULL) {
int err;
if ((err = put_nvlist(zc, config)) != 0)
error = err;
nvlist_free(config);
}
nvlist_free(policy);
}
}
if (error != 0)
return (error);
/*
* If multihost is enabled, resuming I/O is unsafe as another
* host may have imported the pool.
*/
if (spa_multihost(spa) && spa_suspended(spa))
return (SET_ERROR(EINVAL));
spa_vdev_state_enter(spa, SCL_NONE);
if (zc->zc_guid == 0) {
vd = NULL;
} else {
vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
if (vd == NULL) {
error = SET_ERROR(ENODEV);
(void) spa_vdev_state_exit(spa, NULL, error);
spa_close(spa, FTAG);
return (error);
}
}
vdev_clear(spa, vd);
(void) spa_vdev_state_exit(spa, spa_suspended(spa) ?
NULL : spa->spa_root_vdev, 0);
/*
* Resume any suspended I/Os.
*/
if (zio_resume(spa) != 0)
error = SET_ERROR(EIO);
spa_close(spa, FTAG);
return (error);
}
/*
* Reopen all the vdevs associated with the pool.
*
* innvl: {
* "scrub_restart" -> when true and scrub is running, allow to restart
* scrub as the side effect of the reopen (boolean).
* }
*
* outnvl is unused
*/
static const zfs_ioc_key_t zfs_keys_pool_reopen[] = {
{"scrub_restart", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
};
/* ARGSUSED */
static int
zfs_ioc_pool_reopen(const char *pool, nvlist_t *innvl, nvlist_t *outnvl)
{
spa_t *spa;
int error;
boolean_t rc, scrub_restart = B_TRUE;
if (innvl) {
error = nvlist_lookup_boolean_value(innvl,
"scrub_restart", &rc);
if (error == 0)
scrub_restart = rc;
}
error = spa_open(pool, &spa, FTAG);
if (error != 0)
return (error);
spa_vdev_state_enter(spa, SCL_NONE);
/*
* If the scrub_restart flag is B_FALSE and a scrub is already
* in progress then set spa_scrub_reopen flag to B_TRUE so that
* we don't restart the scrub as a side effect of the reopen.
* Otherwise, let vdev_open() decided if a resilver is required.
*/
spa->spa_scrub_reopen = (!scrub_restart &&
dsl_scan_scrubbing(spa->spa_dsl_pool));
vdev_reopen(spa->spa_root_vdev);
spa->spa_scrub_reopen = B_FALSE;
(void) spa_vdev_state_exit(spa, NULL, 0);
spa_close(spa, FTAG);
return (0);
}
/*
* inputs:
* zc_name name of filesystem
*
* outputs:
* zc_string name of conflicting snapshot, if there is one
*/
static int
zfs_ioc_promote(zfs_cmd_t *zc)
{
dsl_pool_t *dp;
dsl_dataset_t *ds, *ods;
char origin[ZFS_MAX_DATASET_NAME_LEN];
char *cp;
int error;
zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
strchr(zc->zc_name, '%'))
return (SET_ERROR(EINVAL));
error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
if (!dsl_dir_is_clone(ds->ds_dir)) {
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (SET_ERROR(EINVAL));
}
error = dsl_dataset_hold_obj(dp,
dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
if (error != 0) {
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
dsl_dataset_name(ods, origin);
dsl_dataset_rele(ods, FTAG);
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
/*
* We don't need to unmount *all* the origin fs's snapshots, but
* it's easier.
*/
cp = strchr(origin, '@');
if (cp)
*cp = '\0';
(void) dmu_objset_find(origin,
zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
}
/*
* Retrieve a single {user|group|project}{used|quota}@... property.
*
* inputs:
* zc_name name of filesystem
* zc_objset_type zfs_userquota_prop_t
* zc_value domain name (eg. "S-1-234-567-89")
* zc_guid RID/UID/GID
*
* outputs:
* zc_cookie property value
*/
static int
zfs_ioc_userspace_one(zfs_cmd_t *zc)
{
zfsvfs_t *zfsvfs;
int error;
if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
return (SET_ERROR(EINVAL));
error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
if (error != 0)
return (error);
error = zfs_userspace_one(zfsvfs,
zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
zfsvfs_rele(zfsvfs, FTAG);
return (error);
}
/*
* inputs:
* zc_name name of filesystem
* zc_cookie zap cursor
* zc_objset_type zfs_userquota_prop_t
* zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
*
* outputs:
* zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
* zc_cookie zap cursor
*/
static int
zfs_ioc_userspace_many(zfs_cmd_t *zc)
{
zfsvfs_t *zfsvfs;
int bufsize = zc->zc_nvlist_dst_size;
if (bufsize <= 0)
return (SET_ERROR(ENOMEM));
int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
if (error != 0)
return (error);
void *buf = vmem_alloc(bufsize, KM_SLEEP);
error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
buf, &zc->zc_nvlist_dst_size);
if (error == 0) {
error = xcopyout(buf,
(void *)(uintptr_t)zc->zc_nvlist_dst,
zc->zc_nvlist_dst_size);
}
vmem_free(buf, bufsize);
zfsvfs_rele(zfsvfs, FTAG);
return (error);
}
/*
* inputs:
* zc_name name of filesystem
*
* outputs:
* none
*/
static int
zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
{
int error = 0;
zfsvfs_t *zfsvfs;
if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
/*
* If userused is not enabled, it may be because the
* objset needs to be closed & reopened (to grow the
* objset_phys_t). Suspend/resume the fs will do that.
*/
dsl_dataset_t *ds, *newds;
ds = dmu_objset_ds(zfsvfs->z_os);
error = zfs_suspend_fs(zfsvfs);
if (error == 0) {
dmu_objset_refresh_ownership(ds, &newds,
B_TRUE, zfsvfs);
error = zfs_resume_fs(zfsvfs, newds);
}
}
if (error == 0) {
mutex_enter(&zfsvfs->z_os->os_upgrade_lock);
if (zfsvfs->z_os->os_upgrade_id == 0) {
/* clear potential error code and retry */
zfsvfs->z_os->os_upgrade_status = 0;
mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
dsl_pool_config_enter(
dmu_objset_pool(zfsvfs->z_os), FTAG);
dmu_objset_userspace_upgrade(zfsvfs->z_os);
dsl_pool_config_exit(
dmu_objset_pool(zfsvfs->z_os), FTAG);
} else {
mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
}
taskq_wait_id(zfsvfs->z_os->os_spa->spa_upgrade_taskq,
zfsvfs->z_os->os_upgrade_id);
error = zfsvfs->z_os->os_upgrade_status;
}
zfs_vfs_rele(zfsvfs);
} else {
objset_t *os;
/* XXX kind of reading contents without owning */
error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
if (error != 0)
return (error);
mutex_enter(&os->os_upgrade_lock);
if (os->os_upgrade_id == 0) {
/* clear potential error code and retry */
os->os_upgrade_status = 0;
mutex_exit(&os->os_upgrade_lock);
dmu_objset_userspace_upgrade(os);
} else {
mutex_exit(&os->os_upgrade_lock);
}
dsl_pool_rele(dmu_objset_pool(os), FTAG);
taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
error = os->os_upgrade_status;
dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT,
FTAG);
}
return (error);
}
/*
* inputs:
* zc_name name of filesystem
*
* outputs:
* none
*/
static int
zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
{
objset_t *os;
int error;
error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
if (error != 0)
return (error);
if (dmu_objset_userobjspace_upgradable(os) ||
dmu_objset_projectquota_upgradable(os)) {
mutex_enter(&os->os_upgrade_lock);
if (os->os_upgrade_id == 0) {
/* clear potential error code and retry */
os->os_upgrade_status = 0;
mutex_exit(&os->os_upgrade_lock);
dmu_objset_id_quota_upgrade(os);
} else {
mutex_exit(&os->os_upgrade_lock);
}
dsl_pool_rele(dmu_objset_pool(os), FTAG);
taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
error = os->os_upgrade_status;
} else {
dsl_pool_rele(dmu_objset_pool(os), FTAG);
}
dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT, FTAG);
return (error);
}
static int
zfs_ioc_share(zfs_cmd_t *zc)
{
return (SET_ERROR(ENOSYS));
}
-ace_t full_access[] = {
- {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
-};
-
/*
* inputs:
* zc_name name of containing filesystem
* zc_obj object # beyond which we want next in-use object #
*
* outputs:
* zc_obj next in-use object #
*/
static int
zfs_ioc_next_obj(zfs_cmd_t *zc)
{
objset_t *os = NULL;
int error;
error = dmu_objset_hold(zc->zc_name, FTAG, &os);
if (error != 0)
return (error);
error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);
dmu_objset_rele(os, FTAG);
return (error);
}
/*
* inputs:
* zc_name name of filesystem
* zc_value prefix name for snapshot
* zc_cleanup_fd cleanup-on-exit file descriptor for calling process
*
* outputs:
* zc_value short name of new snapshot
*/
static int
zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
{
char *snap_name;
char *hold_name;
minor_t minor;
zfs_file_t *fp = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
if (fp == NULL)
return (SET_ERROR(EBADF));
snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
(u_longlong_t)ddi_get_lbolt64());
hold_name = kmem_asprintf("%%%s", zc->zc_value);
int error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
hold_name);
if (error == 0)
(void) strlcpy(zc->zc_value, snap_name,
sizeof (zc->zc_value));
kmem_strfree(snap_name);
kmem_strfree(hold_name);
zfs_onexit_fd_rele(fp);
return (error);
}
/*
* inputs:
* zc_name name of "to" snapshot
* zc_value name of "from" snapshot
* zc_cookie file descriptor to write diff data on
*
* outputs:
* dmu_diff_record_t's to the file descriptor
*/
static int
zfs_ioc_diff(zfs_cmd_t *zc)
{
zfs_file_t *fp;
offset_t off;
int error;
if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
return (SET_ERROR(EBADF));
off = zfs_file_off(fp);
error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
zfs_file_put(fp);
return (error);
}
static int
zfs_ioc_smb_acl(zfs_cmd_t *zc)
{
return (SET_ERROR(ENOTSUP));
}
/*
* innvl: {
* "holds" -> { snapname -> holdname (string), ... }
* (optional) "cleanup_fd" -> fd (int32)
* }
*
* outnvl: {
* snapname -> error value (int32)
* ...
* }
*/
static const zfs_ioc_key_t zfs_keys_hold[] = {
{"holds", DATA_TYPE_NVLIST, 0},
{"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
};
/* ARGSUSED */
static int
zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
{
nvpair_t *pair;
nvlist_t *holds;
int cleanup_fd = -1;
int error;
minor_t minor = 0;
zfs_file_t *fp = NULL;
holds = fnvlist_lookup_nvlist(args, "holds");
/* make sure the user didn't pass us any invalid (empty) tags */
for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
pair = nvlist_next_nvpair(holds, pair)) {
char *htag;
error = nvpair_value_string(pair, &htag);
if (error != 0)
return (SET_ERROR(error));
if (strlen(htag) == 0)
return (SET_ERROR(EINVAL));
}
if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
fp = zfs_onexit_fd_hold(cleanup_fd, &minor);
if (fp == NULL)
return (SET_ERROR(EBADF));
}
error = dsl_dataset_user_hold(holds, minor, errlist);
if (fp != NULL) {
ASSERT3U(minor, !=, 0);
zfs_onexit_fd_rele(fp);
}
return (SET_ERROR(error));
}
/*
* innvl is not used.
*
* outnvl: {
* holdname -> time added (uint64 seconds since epoch)
* ...
* }
*/
static const zfs_ioc_key_t zfs_keys_get_holds[] = {
/* no nvl keys */
};
/* ARGSUSED */
static int
zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
{
return (dsl_dataset_get_holds(snapname, outnvl));
}
/*
* innvl: {
* snapname -> { holdname, ... }
* ...
* }
*
* outnvl: {
* snapname -> error value (int32)
* ...
* }
*/
static const zfs_ioc_key_t zfs_keys_release[] = {
{"<snapname>...", DATA_TYPE_NVLIST, ZK_WILDCARDLIST},
};
/* ARGSUSED */
static int
zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
{
return (dsl_dataset_user_release(holds, errlist));
}
/*
* inputs:
* zc_guid flags (ZEVENT_NONBLOCK)
* zc_cleanup_fd zevent file descriptor
*
* outputs:
* zc_nvlist_dst next nvlist event
* zc_cookie dropped events since last get
*/
static int
zfs_ioc_events_next(zfs_cmd_t *zc)
{
zfs_zevent_t *ze;
nvlist_t *event = NULL;
minor_t minor;
uint64_t dropped = 0;
int error;
zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
if (fp == NULL)
return (SET_ERROR(EBADF));
do {
error = zfs_zevent_next(ze, &event,
&zc->zc_nvlist_dst_size, &dropped);
if (event != NULL) {
zc->zc_cookie = dropped;
error = put_nvlist(zc, event);
nvlist_free(event);
}
if (zc->zc_guid & ZEVENT_NONBLOCK)
break;
if ((error == 0) || (error != ENOENT))
break;
error = zfs_zevent_wait(ze);
if (error != 0)
break;
} while (1);
zfs_zevent_fd_rele(fp);
return (error);
}
/*
* outputs:
* zc_cookie cleared events count
*/
static int
zfs_ioc_events_clear(zfs_cmd_t *zc)
{
int count;
zfs_zevent_drain_all(&count);
zc->zc_cookie = count;
return (0);
}
/*
* inputs:
* zc_guid eid | ZEVENT_SEEK_START | ZEVENT_SEEK_END
* zc_cleanup zevent file descriptor
*/
static int
zfs_ioc_events_seek(zfs_cmd_t *zc)
{
zfs_zevent_t *ze;
minor_t minor;
int error;
zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
if (fp == NULL)
return (SET_ERROR(EBADF));
error = zfs_zevent_seek(ze, zc->zc_guid);
zfs_zevent_fd_rele(fp);
return (error);
}
/*
* inputs:
* zc_name name of later filesystem or snapshot
* zc_value full name of old snapshot or bookmark
*
* outputs:
* zc_cookie space in bytes
* zc_objset_type compressed space in bytes
* zc_perm_action uncompressed space in bytes
*/
static int
zfs_ioc_space_written(zfs_cmd_t *zc)
{
int error;
dsl_pool_t *dp;
dsl_dataset_t *new;
error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
if (strchr(zc->zc_value, '#') != NULL) {
zfs_bookmark_phys_t bmp;
error = dsl_bookmark_lookup(dp, zc->zc_value,
new, &bmp);
if (error == 0) {
error = dsl_dataset_space_written_bookmark(&bmp, new,
&zc->zc_cookie,
&zc->zc_objset_type, &zc->zc_perm_action);
}
} else {
dsl_dataset_t *old;
error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
if (error == 0) {
error = dsl_dataset_space_written(old, new,
&zc->zc_cookie,
&zc->zc_objset_type, &zc->zc_perm_action);
dsl_dataset_rele(old, FTAG);
}
}
dsl_dataset_rele(new, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
/*
* innvl: {
* "firstsnap" -> snapshot name
* }
*
* outnvl: {
* "used" -> space in bytes
* "compressed" -> compressed space in bytes
* "uncompressed" -> uncompressed space in bytes
* }
*/
static const zfs_ioc_key_t zfs_keys_space_snaps[] = {
{"firstsnap", DATA_TYPE_STRING, 0},
};
static int
zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
{
int error;
dsl_pool_t *dp;
dsl_dataset_t *new, *old;
char *firstsnap;
uint64_t used, comp, uncomp;
firstsnap = fnvlist_lookup_string(innvl, "firstsnap");
error = dsl_pool_hold(lastsnap, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
if (error == 0 && !new->ds_is_snapshot) {
dsl_dataset_rele(new, FTAG);
error = SET_ERROR(EINVAL);
}
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
if (error == 0 && !old->ds_is_snapshot) {
dsl_dataset_rele(old, FTAG);
error = SET_ERROR(EINVAL);
}
if (error != 0) {
dsl_dataset_rele(new, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
dsl_dataset_rele(old, FTAG);
dsl_dataset_rele(new, FTAG);
dsl_pool_rele(dp, FTAG);
fnvlist_add_uint64(outnvl, "used", used);
fnvlist_add_uint64(outnvl, "compressed", comp);
fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
return (error);
}
/*
* innvl: {
* "fd" -> file descriptor to write stream to (int32)
* (optional) "fromsnap" -> full snap name to send an incremental from
* (optional) "largeblockok" -> (value ignored)
* indicates that blocks > 128KB are permitted
* (optional) "embedok" -> (value ignored)
* presence indicates DRR_WRITE_EMBEDDED records are permitted
* (optional) "compressok" -> (value ignored)
* presence indicates compressed DRR_WRITE records are permitted
* (optional) "rawok" -> (value ignored)
* presence indicates raw encrypted records should be used.
* (optional) "savedok" -> (value ignored)
* presence indicates we should send a partially received snapshot
* (optional) "resume_object" and "resume_offset" -> (uint64)
* if present, resume send stream from specified object and offset.
* (optional) "redactbook" -> (string)
* if present, use this bookmark's redaction list to generate a redacted
* send stream
* }
*
* outnvl is unused
*/
static const zfs_ioc_key_t zfs_keys_send_new[] = {
{"fd", DATA_TYPE_INT32, 0},
{"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
{"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"savedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL},
{"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL},
{"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
};
/* ARGSUSED */
static int
zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
{
int error;
offset_t off;
char *fromname = NULL;
int fd;
zfs_file_t *fp;
boolean_t largeblockok;
boolean_t embedok;
boolean_t compressok;
boolean_t rawok;
boolean_t savedok;
uint64_t resumeobj = 0;
uint64_t resumeoff = 0;
char *redactbook = NULL;
fd = fnvlist_lookup_int32(innvl, "fd");
(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
largeblockok = nvlist_exists(innvl, "largeblockok");
embedok = nvlist_exists(innvl, "embedok");
compressok = nvlist_exists(innvl, "compressok");
rawok = nvlist_exists(innvl, "rawok");
savedok = nvlist_exists(innvl, "savedok");
(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
(void) nvlist_lookup_string(innvl, "redactbook", &redactbook);
if ((fp = zfs_file_get(fd)) == NULL)
return (SET_ERROR(EBADF));
off = zfs_file_off(fp);
dmu_send_outparams_t out = {0};
out.dso_outfunc = dump_bytes;
out.dso_arg = fp;
out.dso_dryrun = B_FALSE;
error = dmu_send(snapname, fromname, embedok, largeblockok,
compressok, rawok, savedok, resumeobj, resumeoff,
redactbook, fd, &off, &out);
zfs_file_put(fp);
return (error);
}
/* ARGSUSED */
static int
send_space_sum(objset_t *os, void *buf, int len, void *arg)
{
uint64_t *size = arg;
*size += len;
return (0);
}
/*
* Determine approximately how large a zfs send stream will be -- the number
* of bytes that will be written to the fd supplied to zfs_ioc_send_new().
*
* innvl: {
* (optional) "from" -> full snap or bookmark name to send an incremental
* from
* (optional) "largeblockok" -> (value ignored)
* indicates that blocks > 128KB are permitted
* (optional) "embedok" -> (value ignored)
* presence indicates DRR_WRITE_EMBEDDED records are permitted
* (optional) "compressok" -> (value ignored)
* presence indicates compressed DRR_WRITE records are permitted
* (optional) "rawok" -> (value ignored)
* presence indicates raw encrypted records should be used.
* (optional) "resume_object" and "resume_offset" -> (uint64)
* if present, resume send stream from specified object and offset.
* (optional) "fd" -> file descriptor to use as a cookie for progress
* tracking (int32)
* }
*
* outnvl: {
* "space" -> bytes of space (uint64)
* }
*/
static const zfs_ioc_key_t zfs_keys_send_space[] = {
{"from", DATA_TYPE_STRING, ZK_OPTIONAL},
{"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
{"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
{"fd", DATA_TYPE_INT32, ZK_OPTIONAL},
{"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
{"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL},
{"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL},
{"bytes", DATA_TYPE_UINT64, ZK_OPTIONAL},
};
static int
zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
{
dsl_pool_t *dp;
dsl_dataset_t *tosnap;
dsl_dataset_t *fromsnap = NULL;
int error;
char *fromname = NULL;
char *redactlist_book = NULL;
boolean_t largeblockok;
boolean_t embedok;
boolean_t compressok;
boolean_t rawok;
boolean_t savedok;
uint64_t space = 0;
boolean_t full_estimate = B_FALSE;
uint64_t resumeobj = 0;
uint64_t resumeoff = 0;
uint64_t resume_bytes = 0;
int32_t fd = -1;
zfs_bookmark_phys_t zbm = {0};
error = dsl_pool_hold(snapname, FTAG, &dp);
if (error != 0)
return (error);
error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
if (error != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
(void) nvlist_lookup_int32(innvl, "fd", &fd);
largeblockok = nvlist_exists(innvl, "largeblockok");
embedok = nvlist_exists(innvl, "embedok");
compressok = nvlist_exists(innvl, "compressok");
rawok = nvlist_exists(innvl, "rawok");
savedok = nvlist_exists(innvl, "savedok");
boolean_t from = (nvlist_lookup_string(innvl, "from", &fromname) == 0);
boolean_t altbook = (nvlist_lookup_string(innvl, "redactbook",
&redactlist_book) == 0);
(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
(void) nvlist_lookup_uint64(innvl, "bytes", &resume_bytes);
if (altbook) {
full_estimate = B_TRUE;
} else if (from) {
if (strchr(fromname, '#')) {
error = dsl_bookmark_lookup(dp, fromname, tosnap, &zbm);
/*
* dsl_bookmark_lookup() will fail with EXDEV if
* the from-bookmark and tosnap are at the same txg.
* However, it's valid to do a send (and therefore,
* a send estimate) from and to the same time point,
* if the bookmark is redacted (the incremental send
* can change what's redacted on the target). In
* this case, dsl_bookmark_lookup() fills in zbm
* but returns EXDEV. Ignore this error.
*/
if (error == EXDEV && zbm.zbm_redaction_obj != 0 &&
zbm.zbm_guid ==
dsl_dataset_phys(tosnap)->ds_guid)
error = 0;
if (error != 0) {
dsl_dataset_rele(tosnap, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
if (zbm.zbm_redaction_obj != 0 || !(zbm.zbm_flags &
ZBM_FLAG_HAS_FBN)) {
full_estimate = B_TRUE;
}
} else if (strchr(fromname, '@')) {
error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
if (error != 0) {
dsl_dataset_rele(tosnap, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
if (!dsl_dataset_is_before(tosnap, fromsnap, 0)) {
full_estimate = B_TRUE;
dsl_dataset_rele(fromsnap, FTAG);
}
} else {
/*
* from is not properly formatted as a snapshot or
* bookmark
*/
dsl_dataset_rele(tosnap, FTAG);
dsl_pool_rele(dp, FTAG);
return (SET_ERROR(EINVAL));
}
}
if (full_estimate) {
dmu_send_outparams_t out = {0};
offset_t off = 0;
out.dso_outfunc = send_space_sum;
out.dso_arg = &space;
out.dso_dryrun = B_TRUE;
/*
* We have to release these holds so dmu_send can take them. It
* will do all the error checking we need.
*/
dsl_dataset_rele(tosnap, FTAG);
dsl_pool_rele(dp, FTAG);
error = dmu_send(snapname, fromname, embedok, largeblockok,
compressok, rawok, savedok, resumeobj, resumeoff,
redactlist_book, fd, &off, &out);
} else {
error = dmu_send_estimate_fast(tosnap, fromsnap,
(from && strchr(fromname, '#') != NULL ? &zbm : NULL),
compressok || rawok, savedok, &space);
space -= resume_bytes;
if (fromsnap != NULL)
dsl_dataset_rele(fromsnap, FTAG);
dsl_dataset_rele(tosnap, FTAG);
dsl_pool_rele(dp, FTAG);
}
fnvlist_add_uint64(outnvl, "space", space);
return (error);
}
/*
* Sync the currently open TXG to disk for the specified pool.
* This is somewhat similar to 'zfs_sync()'.
* For cases that do not result in error this ioctl will wait for
* the currently open TXG to commit before returning back to the caller.
*
* innvl: {
* "force" -> when true, force uberblock update even if there is no dirty data.
* In addition this will cause the vdev configuration to be written
* out including updating the zpool cache file. (boolean_t)
* }
*
* onvl is unused
*/
static const zfs_ioc_key_t zfs_keys_pool_sync[] = {
{"force", DATA_TYPE_BOOLEAN_VALUE, 0},
};
/* ARGSUSED */
static int
zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
{
int err;
boolean_t rc, force = B_FALSE;
spa_t *spa;
if ((err = spa_open(pool, &spa, FTAG)) != 0)
return (err);
if (innvl) {
err = nvlist_lookup_boolean_value(innvl, "force", &rc);
if (err == 0)
force = rc;
}
if (force) {
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
vdev_config_dirty(spa->spa_root_vdev);
spa_config_exit(spa, SCL_CONFIG, FTAG);
}
txg_wait_synced(spa_get_dsl(spa), 0);
spa_close(spa, FTAG);
return (0);
}
/*
* Load a user's wrapping key into the kernel.
* innvl: {
* "hidden_args" -> { "wkeydata" -> value }
* raw uint8_t array of encryption wrapping key data (32 bytes)
* (optional) "noop" -> (value ignored)
* presence indicated key should only be verified, not loaded
* }
*/
static const zfs_ioc_key_t zfs_keys_load_key[] = {
{"hidden_args", DATA_TYPE_NVLIST, 0},
{"noop", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
};
/* ARGSUSED */
static int
zfs_ioc_load_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
{
int ret;
dsl_crypto_params_t *dcp = NULL;
nvlist_t *hidden_args;
boolean_t noop = nvlist_exists(innvl, "noop");
if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
ret = SET_ERROR(EINVAL);
goto error;
}
hidden_args = fnvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS);
ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
hidden_args, &dcp);
if (ret != 0)
goto error;
ret = spa_keystore_load_wkey(dsname, dcp, noop);
if (ret != 0)
goto error;
dsl_crypto_params_free(dcp, noop);
return (0);
error:
dsl_crypto_params_free(dcp, B_TRUE);
return (ret);
}
/*
* Unload a user's wrapping key from the kernel.
* Both innvl and outnvl are unused.
*/
static const zfs_ioc_key_t zfs_keys_unload_key[] = {
/* no nvl keys */
};
/* ARGSUSED */
static int
zfs_ioc_unload_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
{
int ret = 0;
if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
ret = (SET_ERROR(EINVAL));
goto out;
}
ret = spa_keystore_unload_wkey(dsname);
if (ret != 0)
goto out;
out:
return (ret);
}
/*
* Changes a user's wrapping key used to decrypt a dataset. The keyformat,
* keylocation, pbkdf2salt, and pbkdf2iters properties can also be specified
* here to change how the key is derived in userspace.
*
* innvl: {
* "hidden_args" (optional) -> { "wkeydata" -> value }
* raw uint8_t array of new encryption wrapping key data (32 bytes)
* "props" (optional) -> { prop -> value }
* }
*
* outnvl is unused
*/
static const zfs_ioc_key_t zfs_keys_change_key[] = {
{"crypt_cmd", DATA_TYPE_UINT64, ZK_OPTIONAL},
{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
};
/* ARGSUSED */
static int
zfs_ioc_change_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
{
int ret;
uint64_t cmd = DCP_CMD_NONE;
dsl_crypto_params_t *dcp = NULL;
nvlist_t *args = NULL, *hidden_args = NULL;
if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
ret = (SET_ERROR(EINVAL));
goto error;
}
(void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
(void) nvlist_lookup_nvlist(innvl, "props", &args);
(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
if (ret != 0)
goto error;
ret = spa_keystore_change_key(dsname, dcp);
if (ret != 0)
goto error;
dsl_crypto_params_free(dcp, B_FALSE);
return (0);
error:
dsl_crypto_params_free(dcp, B_TRUE);
return (ret);
}
static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
static void
zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
{
zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
ASSERT3U(ioc, <, ZFS_IOC_LAST);
ASSERT3P(vec->zvec_legacy_func, ==, NULL);
ASSERT3P(vec->zvec_func, ==, NULL);
vec->zvec_legacy_func = func;
vec->zvec_secpolicy = secpolicy;
vec->zvec_namecheck = namecheck;
vec->zvec_allow_log = log_history;
vec->zvec_pool_check = pool_check;
}
/*
* See the block comment at the beginning of this file for details on
* each argument to this function.
*/
void
zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
boolean_t allow_log, const zfs_ioc_key_t *nvl_keys, size_t num_keys)
{
zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
ASSERT3U(ioc, <, ZFS_IOC_LAST);
ASSERT3P(vec->zvec_legacy_func, ==, NULL);
ASSERT3P(vec->zvec_func, ==, NULL);
/* if we are logging, the name must be valid */
ASSERT(!allow_log || namecheck != NO_NAME);
vec->zvec_name = name;
vec->zvec_func = func;
vec->zvec_secpolicy = secpolicy;
vec->zvec_namecheck = namecheck;
vec->zvec_pool_check = pool_check;
vec->zvec_smush_outnvlist = smush_outnvlist;
vec->zvec_allow_log = allow_log;
vec->zvec_nvl_keys = nvl_keys;
vec->zvec_nvl_key_count = num_keys;
}
static void
zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
zfs_ioc_poolcheck_t pool_check)
{
zfs_ioctl_register_legacy(ioc, func, secpolicy,
POOL_NAME, log_history, pool_check);
}
void
zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
{
zfs_ioctl_register_legacy(ioc, func, secpolicy,
DATASET_NAME, B_FALSE, pool_check);
}
static void
zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
{
zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
}
static void
zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
zfs_secpolicy_func_t *secpolicy)
{
zfs_ioctl_register_legacy(ioc, func, secpolicy,
NO_NAME, B_FALSE, POOL_CHECK_NONE);
}
static void
zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
{
zfs_ioctl_register_legacy(ioc, func, secpolicy,
DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
}
static void
zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
{
zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
zfs_secpolicy_read);
}
static void
zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
zfs_secpolicy_func_t *secpolicy)
{
zfs_ioctl_register_legacy(ioc, func, secpolicy,
DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
}
static void
zfs_ioctl_init(void)
{
zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_snapshot, ARRAY_SIZE(zfs_keys_snapshot));
zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
zfs_keys_log_history, ARRAY_SIZE(zfs_keys_log_history));
zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
zfs_keys_space_snaps, ARRAY_SIZE(zfs_keys_space_snaps));
zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
zfs_keys_send_new, ARRAY_SIZE(zfs_keys_send_new));
zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
zfs_keys_send_space, ARRAY_SIZE(zfs_keys_send_space));
zfs_ioctl_register("create", ZFS_IOC_CREATE,
zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_create, ARRAY_SIZE(zfs_keys_create));
zfs_ioctl_register("clone", ZFS_IOC_CLONE,
zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_clone, ARRAY_SIZE(zfs_keys_clone));
zfs_ioctl_register("remap", ZFS_IOC_REMAP,
zfs_ioc_remap, zfs_secpolicy_none, DATASET_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
zfs_keys_remap, ARRAY_SIZE(zfs_keys_remap));
zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_destroy_snaps, ARRAY_SIZE(zfs_keys_destroy_snaps));
zfs_ioctl_register("hold", ZFS_IOC_HOLD,
zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_hold, ARRAY_SIZE(zfs_keys_hold));
zfs_ioctl_register("release", ZFS_IOC_RELEASE,
zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_release, ARRAY_SIZE(zfs_keys_release));
zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
zfs_keys_get_holds, ARRAY_SIZE(zfs_keys_get_holds));
zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
zfs_keys_rollback, ARRAY_SIZE(zfs_keys_rollback));
zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_bookmark, ARRAY_SIZE(zfs_keys_bookmark));
zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
zfs_keys_get_bookmarks, ARRAY_SIZE(zfs_keys_get_bookmarks));
zfs_ioctl_register("get_bookmark_props", ZFS_IOC_GET_BOOKMARK_PROPS,
zfs_ioc_get_bookmark_props, zfs_secpolicy_read, ENTITY_NAME,
POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, zfs_keys_get_bookmark_props,
ARRAY_SIZE(zfs_keys_get_bookmark_props));
zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_destroy_bookmarks,
ARRAY_SIZE(zfs_keys_destroy_bookmarks));
zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
zfs_ioc_recv_new, zfs_secpolicy_recv_new, DATASET_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_recv_new, ARRAY_SIZE(zfs_keys_recv_new));
zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
zfs_ioc_load_key, zfs_secpolicy_load_key,
DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
zfs_keys_load_key, ARRAY_SIZE(zfs_keys_load_key));
zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
zfs_ioc_unload_key, zfs_secpolicy_load_key,
DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
zfs_keys_unload_key, ARRAY_SIZE(zfs_keys_unload_key));
zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
zfs_ioc_change_key, zfs_secpolicy_change_key,
DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY,
B_TRUE, B_TRUE, zfs_keys_change_key,
ARRAY_SIZE(zfs_keys_change_key));
zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
zfs_keys_pool_sync, ARRAY_SIZE(zfs_keys_pool_sync));
zfs_ioctl_register("reopen", ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
zfs_secpolicy_config, POOL_NAME, POOL_CHECK_SUSPENDED, B_TRUE,
B_TRUE, zfs_keys_pool_reopen, ARRAY_SIZE(zfs_keys_pool_reopen));
zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
zfs_ioc_channel_program, zfs_secpolicy_config,
POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
B_TRUE, zfs_keys_channel_program,
ARRAY_SIZE(zfs_keys_channel_program));
zfs_ioctl_register("redact", ZFS_IOC_REDACT,
zfs_ioc_redact, zfs_secpolicy_config, DATASET_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_redact, ARRAY_SIZE(zfs_keys_redact));
zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_pool_checkpoint, ARRAY_SIZE(zfs_keys_pool_checkpoint));
zfs_ioctl_register("zpool_discard_checkpoint",
ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
zfs_secpolicy_config, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_pool_discard_checkpoint,
ARRAY_SIZE(zfs_keys_pool_discard_checkpoint));
zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_pool_initialize, ARRAY_SIZE(zfs_keys_pool_initialize));
zfs_ioctl_register("trim", ZFS_IOC_POOL_TRIM,
zfs_ioc_pool_trim, zfs_secpolicy_config, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
zfs_keys_pool_trim, ARRAY_SIZE(zfs_keys_pool_trim));
zfs_ioctl_register("wait", ZFS_IOC_WAIT,
zfs_ioc_wait, zfs_secpolicy_none, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
zfs_keys_pool_wait, ARRAY_SIZE(zfs_keys_pool_wait));
zfs_ioctl_register("wait_fs", ZFS_IOC_WAIT_FS,
zfs_ioc_wait_fs, zfs_secpolicy_none, DATASET_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
zfs_keys_fs_wait, ARRAY_SIZE(zfs_keys_fs_wait));
zfs_ioctl_register("set_bootenv", ZFS_IOC_SET_BOOTENV,
zfs_ioc_set_bootenv, zfs_secpolicy_config, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
zfs_keys_set_bootenv, ARRAY_SIZE(zfs_keys_set_bootenv));
zfs_ioctl_register("get_bootenv", ZFS_IOC_GET_BOOTENV,
zfs_ioc_get_bootenv, zfs_secpolicy_none, POOL_NAME,
POOL_CHECK_SUSPENDED, B_FALSE, B_TRUE,
zfs_keys_get_bootenv, ARRAY_SIZE(zfs_keys_get_bootenv));
zfs_ioctl_register("zpool_vdev_get_props", ZFS_IOC_VDEV_GET_PROPS,
zfs_ioc_vdev_get_props, zfs_secpolicy_read, POOL_NAME,
POOL_CHECK_NONE, B_FALSE, B_FALSE, zfs_keys_vdev_get_props,
ARRAY_SIZE(zfs_keys_vdev_get_props));
zfs_ioctl_register("zpool_vdev_set_props", ZFS_IOC_VDEV_SET_PROPS,
zfs_ioc_vdev_set_props, zfs_secpolicy_config, POOL_NAME,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
zfs_keys_vdev_set_props, ARRAY_SIZE(zfs_keys_vdev_set_props));
/* IOCTLS that use the legacy function signature */
zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
zfs_ioc_pool_scan);
zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
zfs_ioc_pool_upgrade);
zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
zfs_ioc_vdev_add);
zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
zfs_ioc_vdev_remove);
zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
zfs_ioc_vdev_set_state);
zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
zfs_ioc_vdev_attach);
zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
zfs_ioc_vdev_detach);
zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
zfs_ioc_vdev_setpath);
zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
zfs_ioc_vdev_setfru);
zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
zfs_ioc_pool_set_props);
zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
zfs_ioc_vdev_split);
zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
zfs_ioc_pool_reguid);
zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
zfs_ioc_pool_configs, zfs_secpolicy_none);
zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
zfs_ioc_pool_tryimport, zfs_secpolicy_config);
zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
zfs_ioc_inject_fault, zfs_secpolicy_inject);
zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
zfs_ioc_clear_fault, zfs_secpolicy_inject);
zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
zfs_ioc_inject_list_next, zfs_secpolicy_inject);
/*
* pool destroy, and export don't log the history as part of
* zfsdev_ioctl, but rather zfs_ioc_pool_export
* does the logging of those commands.
*/
zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
zfs_ioc_dsobj_to_dsname,
zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
zfs_ioc_pool_get_history,
zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
zfs_ioc_space_written);
zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
zfs_ioc_objset_recvd_props);
zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
zfs_ioc_next_obj);
zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
zfs_ioc_get_fsacl);
zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
zfs_ioc_objset_stats);
zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
zfs_ioc_objset_zplprops);
zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
zfs_ioc_dataset_list_next);
zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
zfs_ioc_snapshot_list_next);
zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
zfs_ioc_send_progress);
zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
zfs_ioc_diff, zfs_secpolicy_diff);
zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
zfs_ioc_obj_to_path, zfs_secpolicy_diff);
zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
zfs_ioc_send, zfs_secpolicy_send);
zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
zfs_secpolicy_none);
zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
zfs_secpolicy_destroy);
zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
zfs_secpolicy_rename);
zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
zfs_secpolicy_recv);
zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
zfs_secpolicy_promote);
zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
zfs_secpolicy_set_fsacl);
zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
zfs_secpolicy_share, POOL_CHECK_NONE);
zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
zfs_ioctl_init_os();
}
/*
* Verify that for non-legacy ioctls the input nvlist
* pairs match against the expected input.
*
* Possible errors are:
* ZFS_ERR_IOC_ARG_UNAVAIL An unrecognized nvpair was encountered
* ZFS_ERR_IOC_ARG_REQUIRED A required nvpair is missing
* ZFS_ERR_IOC_ARG_BADTYPE Invalid type for nvpair
*/
static int
zfs_check_input_nvpairs(nvlist_t *innvl, const zfs_ioc_vec_t *vec)
{
const zfs_ioc_key_t *nvl_keys = vec->zvec_nvl_keys;
boolean_t required_keys_found = B_FALSE;
/*
* examine each input pair
*/
for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
char *name = nvpair_name(pair);
data_type_t type = nvpair_type(pair);
boolean_t identified = B_FALSE;
/*
* check pair against the documented names and type
*/
for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
/* if not a wild card name, check for an exact match */
if ((nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) == 0 &&
strcmp(nvl_keys[k].zkey_name, name) != 0)
continue;
identified = B_TRUE;
if (nvl_keys[k].zkey_type != DATA_TYPE_ANY &&
nvl_keys[k].zkey_type != type) {
return (SET_ERROR(ZFS_ERR_IOC_ARG_BADTYPE));
}
if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
continue;
required_keys_found = B_TRUE;
break;
}
/* allow an 'optional' key, everything else is invalid */
if (!identified &&
(strcmp(name, "optional") != 0 ||
type != DATA_TYPE_NVLIST)) {
return (SET_ERROR(ZFS_ERR_IOC_ARG_UNAVAIL));
}
}
/* verify that all required keys were found */
for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
continue;
if (nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) {
/* at least one non-optional key is expected here */
if (!required_keys_found)
return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
continue;
}
if (!nvlist_exists(innvl, nvl_keys[k].zkey_name))
return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
}
return (0);
}
static int
pool_status_check(const char *name, zfs_ioc_namecheck_t type,
zfs_ioc_poolcheck_t check)
{
spa_t *spa;
int error;
ASSERT(type == POOL_NAME || type == DATASET_NAME ||
type == ENTITY_NAME);
if (check & POOL_CHECK_NONE)
return (0);
error = spa_open(name, &spa, FTAG);
if (error == 0) {
if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
error = SET_ERROR(EAGAIN);
else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
error = SET_ERROR(EROFS);
spa_close(spa, FTAG);
}
return (error);
}
int
zfsdev_getminor(zfs_file_t *fp, minor_t *minorp)
{
zfsdev_state_t *zs, *fpd;
ASSERT(!MUTEX_HELD(&zfsdev_state_lock));
fpd = zfs_file_private(fp);
if (fpd == NULL)
return (SET_ERROR(EBADF));
mutex_enter(&zfsdev_state_lock);
for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
if (zs->zs_minor == -1)
continue;
if (fpd == zs) {
*minorp = fpd->zs_minor;
mutex_exit(&zfsdev_state_lock);
return (0);
}
}
mutex_exit(&zfsdev_state_lock);
return (SET_ERROR(EBADF));
}
void *
zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
{
zfsdev_state_t *zs;
for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
if (zs->zs_minor == minor) {
smp_rmb();
switch (which) {
case ZST_ONEXIT:
return (zs->zs_onexit);
case ZST_ZEVENT:
return (zs->zs_zevent);
case ZST_ALL:
return (zs);
}
}
}
return (NULL);
}
/*
* Find a free minor number. The zfsdev_state_list is expected to
* be short since it is only a list of currently open file handles.
*/
static minor_t
zfsdev_minor_alloc(void)
{
static minor_t last_minor = 0;
minor_t m;
ASSERT(MUTEX_HELD(&zfsdev_state_lock));
for (m = last_minor + 1; m != last_minor; m++) {
if (m > ZFSDEV_MAX_MINOR)
m = 1;
if (zfsdev_get_state(m, ZST_ALL) == NULL) {
last_minor = m;
return (m);
}
}
return (0);
}
int
zfsdev_state_init(void *priv)
{
zfsdev_state_t *zs, *zsprev = NULL;
minor_t minor;
boolean_t newzs = B_FALSE;
ASSERT(MUTEX_HELD(&zfsdev_state_lock));
minor = zfsdev_minor_alloc();
if (minor == 0)
return (SET_ERROR(ENXIO));
for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
if (zs->zs_minor == -1)
break;
zsprev = zs;
}
if (!zs) {
zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
newzs = B_TRUE;
}
zfsdev_private_set_state(priv, zs);
zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
/*
* In order to provide for lock-free concurrent read access
* to the minor list in zfsdev_get_state(), new entries
* must be completely written before linking them into the
* list whereas existing entries are already linked; the last
* operation must be updating zs_minor (from -1 to the new
* value).
*/
if (newzs) {
zs->zs_minor = minor;
membar_producer();
zsprev->zs_next = zs;
} else {
membar_producer();
zs->zs_minor = minor;
}
return (0);
}
void
zfsdev_state_destroy(void *priv)
{
zfsdev_state_t *zs = zfsdev_private_get_state(priv);
ASSERT(zs != NULL);
ASSERT3S(zs->zs_minor, >, 0);
/*
* The last reference to this zfsdev file descriptor is being dropped.
* We don't have to worry about lookup grabbing this state object, and
* zfsdev_state_init() will not try to reuse this object until it is
* invalidated by setting zs_minor to -1. Invalidation must be done
* last, with a memory barrier to ensure ordering. This lets us avoid
* taking the global zfsdev state lock around destruction.
*/
zfs_onexit_destroy(zs->zs_onexit);
zfs_zevent_destroy(zs->zs_zevent);
zs->zs_onexit = NULL;
zs->zs_zevent = NULL;
membar_producer();
zs->zs_minor = -1;
}
long
zfsdev_ioctl_common(uint_t vecnum, zfs_cmd_t *zc, int flag)
{
int error, cmd;
const zfs_ioc_vec_t *vec;
char *saved_poolname = NULL;
uint64_t max_nvlist_src_size;
size_t saved_poolname_len = 0;
nvlist_t *innvl = NULL;
fstrans_cookie_t cookie;
hrtime_t start_time = gethrtime();
cmd = vecnum;
error = 0;
if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
vec = &zfs_ioc_vec[vecnum];
/*
* The registered ioctl list may be sparse, verify that either
* a normal or legacy handler are registered.
*/
if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
zc->zc_iflags = flag & FKIOCTL;
max_nvlist_src_size = zfs_max_nvlist_src_size_os();
if (zc->zc_nvlist_src_size > max_nvlist_src_size) {
/*
* Make sure the user doesn't pass in an insane value for
* zc_nvlist_src_size. We have to check, since we will end
* up allocating that much memory inside of get_nvlist(). This
* prevents a nefarious user from allocating tons of kernel
* memory.
*
* Also, we return EINVAL instead of ENOMEM here. The reason
* being that returning ENOMEM from an ioctl() has a special
* connotation; that the user's size value is too small and
* needs to be expanded to hold the nvlist. See
* zcmd_expand_dst_nvlist() for details.
*/
error = SET_ERROR(EINVAL); /* User's size too big */
} else if (zc->zc_nvlist_src_size != 0) {
error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
zc->zc_iflags, &innvl);
if (error != 0)
goto out;
}
/*
* Ensure that all pool/dataset names are valid before we pass down to
* the lower layers.
*/
zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
switch (vec->zvec_namecheck) {
case POOL_NAME:
if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
error = SET_ERROR(EINVAL);
else
error = pool_status_check(zc->zc_name,
vec->zvec_namecheck, vec->zvec_pool_check);
break;
case DATASET_NAME:
if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
error = SET_ERROR(EINVAL);
else
error = pool_status_check(zc->zc_name,
vec->zvec_namecheck, vec->zvec_pool_check);
break;
case ENTITY_NAME:
if (entity_namecheck(zc->zc_name, NULL, NULL) != 0) {
error = SET_ERROR(EINVAL);
} else {
error = pool_status_check(zc->zc_name,
vec->zvec_namecheck, vec->zvec_pool_check);
}
break;
case NO_NAME:
break;
}
/*
* Ensure that all input pairs are valid before we pass them down
* to the lower layers.
*
* The vectored functions can use fnvlist_lookup_{type} for any
* required pairs since zfs_check_input_nvpairs() confirmed that
* they exist and are of the correct type.
*/
if (error == 0 && vec->zvec_func != NULL) {
error = zfs_check_input_nvpairs(innvl, vec);
if (error != 0)
goto out;
}
if (error == 0) {
cookie = spl_fstrans_mark();
error = vec->zvec_secpolicy(zc, innvl, CRED());
spl_fstrans_unmark(cookie);
}
if (error != 0)
goto out;
/* legacy ioctls can modify zc_name */
/*
* Can't use kmem_strdup() as we might truncate the string and
* kmem_strfree() would then free with incorrect size.
*/
saved_poolname_len = strlen(zc->zc_name) + 1;
saved_poolname = kmem_alloc(saved_poolname_len, KM_SLEEP);
strlcpy(saved_poolname, zc->zc_name, saved_poolname_len);
saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';
if (vec->zvec_func != NULL) {
nvlist_t *outnvl;
int puterror = 0;
spa_t *spa;
nvlist_t *lognv = NULL;
ASSERT(vec->zvec_legacy_func == NULL);
/*
* Add the innvl to the lognv before calling the func,
* in case the func changes the innvl.
*/
if (vec->zvec_allow_log) {
lognv = fnvlist_alloc();
fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
vec->zvec_name);
if (!nvlist_empty(innvl)) {
fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
innvl);
}
}
outnvl = fnvlist_alloc();
cookie = spl_fstrans_mark();
error = vec->zvec_func(zc->zc_name, innvl, outnvl);
spl_fstrans_unmark(cookie);
/*
* Some commands can partially execute, modify state, and still
* return an error. In these cases, attempt to record what
* was modified.
*/
if ((error == 0 ||
(cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
vec->zvec_allow_log &&
spa_open(zc->zc_name, &spa, FTAG) == 0) {
if (!nvlist_empty(outnvl)) {
size_t out_size = fnvlist_size(outnvl);
if (out_size > zfs_history_output_max) {
fnvlist_add_int64(lognv,
ZPOOL_HIST_OUTPUT_SIZE, out_size);
} else {
fnvlist_add_nvlist(lognv,
ZPOOL_HIST_OUTPUT_NVL, outnvl);
}
}
if (error != 0) {
fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
error);
}
fnvlist_add_int64(lognv, ZPOOL_HIST_ELAPSED_NS,
gethrtime() - start_time);
(void) spa_history_log_nvl(spa, lognv);
spa_close(spa, FTAG);
}
fnvlist_free(lognv);
if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
int smusherror = 0;
if (vec->zvec_smush_outnvlist) {
smusherror = nvlist_smush(outnvl,
zc->zc_nvlist_dst_size);
}
if (smusherror == 0)
puterror = put_nvlist(zc, outnvl);
}
if (puterror != 0)
error = puterror;
nvlist_free(outnvl);
} else {
cookie = spl_fstrans_mark();
error = vec->zvec_legacy_func(zc);
spl_fstrans_unmark(cookie);
}
out:
nvlist_free(innvl);
if (error == 0 && vec->zvec_allow_log) {
char *s = tsd_get(zfs_allow_log_key);
if (s != NULL)
kmem_strfree(s);
(void) tsd_set(zfs_allow_log_key, kmem_strdup(saved_poolname));
}
if (saved_poolname != NULL)
kmem_free(saved_poolname, saved_poolname_len);
return (error);
}
int
zfs_kmod_init(void)
{
int error;
if ((error = zvol_init()) != 0)
return (error);
spa_init(SPA_MODE_READ | SPA_MODE_WRITE);
zfs_init();
zfs_ioctl_init();
mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
zfsdev_state_list = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
zfsdev_state_list->zs_minor = -1;
if ((error = zfsdev_attach()) != 0)
goto out;
tsd_create(&zfs_fsyncer_key, NULL);
tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
return (0);
out:
zfs_fini();
spa_fini();
zvol_fini();
return (error);
}
void
zfs_kmod_fini(void)
{
zfsdev_state_t *zs, *zsnext = NULL;
zfsdev_detach();
mutex_destroy(&zfsdev_state_lock);
for (zs = zfsdev_state_list; zs != NULL; zs = zsnext) {
zsnext = zs->zs_next;
if (zs->zs_onexit)
zfs_onexit_destroy(zs->zs_onexit);
if (zs->zs_zevent)
zfs_zevent_destroy(zs->zs_zevent);
kmem_free(zs, sizeof (zfsdev_state_t));
}
zfs_ereport_taskq_fini(); /* run before zfs_fini() on Linux */
zfs_fini();
spa_fini();
zvol_fini();
tsd_destroy(&zfs_fsyncer_key);
tsd_destroy(&rrw_tsd_key);
tsd_destroy(&zfs_allow_log_key);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, max_nvlist_src_size, ULONG, ZMOD_RW,
"Maximum size in bytes allowed for src nvlist passed with ZFS ioctls");
ZFS_MODULE_PARAM(zfs, zfs_, history_output_max, ULONG, ZMOD_RW,
"Maximum size in bytes of ZFS ioctl output that will be logged");
/* END CSTYLED */
diff --git a/module/zfs/zfs_log.c b/module/zfs/zfs_log.c
index e248dc3cc4e8..2f3eab67993e 100644
--- a/module/zfs/zfs_log.c
+++ b/module/zfs/zfs_log.c
@@ -1,792 +1,792 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2015, 2018 by Delphix. All rights reserved.
*/
#include <sys/types.h>
#include <sys/param.h>
#include <sys/sysmacros.h>
#include <sys/cmn_err.h>
#include <sys/kmem.h>
#include <sys/thread.h>
#include <sys/file.h>
#include <sys/vfs.h>
#include <sys/zfs_znode.h>
#include <sys/zfs_dir.h>
#include <sys/zil.h>
#include <sys/zil_impl.h>
#include <sys/byteorder.h>
#include <sys/policy.h>
#include <sys/stat.h>
#include <sys/acl.h>
#include <sys/dmu.h>
#include <sys/dbuf.h>
#include <sys/spa.h>
#include <sys/zfs_fuid.h>
#include <sys/dsl_dataset.h>
/*
* These zfs_log_* functions must be called within a dmu tx, in one
* of 2 contexts depending on zilog->z_replay:
*
* Non replay mode
* ---------------
* We need to record the transaction so that if it is committed to
* the Intent Log then it can be replayed. An intent log transaction
* structure (itx_t) is allocated and all the information necessary to
* possibly replay the transaction is saved in it. The itx is then assigned
* a sequence number and inserted in the in-memory list anchored in the zilog.
*
* Replay mode
* -----------
* We need to mark the intent log record as replayed in the log header.
* This is done in the same transaction as the replay so that they
* commit atomically.
*/
int
zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
{
int isxvattr = (vap->va_mask & ATTR_XVATTR);
switch (type) {
case Z_FILE:
if (vsecp == NULL && !isxvattr)
return (TX_CREATE);
if (vsecp && isxvattr)
return (TX_CREATE_ACL_ATTR);
if (vsecp)
return (TX_CREATE_ACL);
else
return (TX_CREATE_ATTR);
case Z_DIR:
if (vsecp == NULL && !isxvattr)
return (TX_MKDIR);
if (vsecp && isxvattr)
return (TX_MKDIR_ACL_ATTR);
if (vsecp)
return (TX_MKDIR_ACL);
else
return (TX_MKDIR_ATTR);
case Z_XATTRDIR:
return (TX_MKXATTR);
}
ASSERT(0);
return (TX_MAX_TYPE);
}
/*
* build up the log data necessary for logging xvattr_t
* First lr_attr_t is initialized. following the lr_attr_t
* is the mapsize and attribute bitmap copied from the xvattr_t.
* Following the bitmap and bitmapsize two 64 bit words are reserved
* for the create time which may be set. Following the create time
* records a single 64 bit integer which has the bits to set on
* replay for the xvattr.
*/
static void
zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
{
uint32_t *bitmap;
uint64_t *attrs;
uint64_t *crtime;
xoptattr_t *xoap;
void *scanstamp;
int i;
xoap = xva_getxoptattr(xvap);
ASSERT(xoap);
lrattr->lr_attr_masksize = xvap->xva_mapsize;
bitmap = &lrattr->lr_attr_bitmap;
for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) {
*bitmap = xvap->xva_reqattrmap[i];
}
/* Now pack the attributes up in a single uint64_t */
attrs = (uint64_t *)bitmap;
*attrs = 0;
crtime = attrs + 1;
bzero(crtime, 2 * sizeof (uint64_t));
scanstamp = (caddr_t)(crtime + 2);
bzero(scanstamp, AV_SCANSTAMP_SZ);
if (XVA_ISSET_REQ(xvap, XAT_READONLY))
*attrs |= (xoap->xoa_readonly == 0) ? 0 :
XAT0_READONLY;
if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
*attrs |= (xoap->xoa_hidden == 0) ? 0 :
XAT0_HIDDEN;
if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
*attrs |= (xoap->xoa_system == 0) ? 0 :
XAT0_SYSTEM;
if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
*attrs |= (xoap->xoa_archive == 0) ? 0 :
XAT0_ARCHIVE;
if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
*attrs |= (xoap->xoa_immutable == 0) ? 0 :
XAT0_IMMUTABLE;
if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
*attrs |= (xoap->xoa_nounlink == 0) ? 0 :
XAT0_NOUNLINK;
if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
*attrs |= (xoap->xoa_appendonly == 0) ? 0 :
XAT0_APPENDONLY;
if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
*attrs |= (xoap->xoa_opaque == 0) ? 0 :
XAT0_APPENDONLY;
if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
*attrs |= (xoap->xoa_nodump == 0) ? 0 :
XAT0_NODUMP;
if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
*attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
XAT0_AV_QUARANTINED;
if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
*attrs |= (xoap->xoa_av_modified == 0) ? 0 :
XAT0_AV_MODIFIED;
if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime);
if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
ASSERT(!XVA_ISSET_REQ(xvap, XAT_PROJID));
bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ);
} else if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
/*
* XAT_PROJID and XAT_AV_SCANSTAMP will never be valid
* at the same time, so we can share the same space.
*/
bcopy(&xoap->xoa_projid, scanstamp, sizeof (uint64_t));
}
if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
*attrs |= (xoap->xoa_reparse == 0) ? 0 :
XAT0_REPARSE;
if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
*attrs |= (xoap->xoa_offline == 0) ? 0 :
XAT0_OFFLINE;
if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
*attrs |= (xoap->xoa_sparse == 0) ? 0 :
XAT0_SPARSE;
if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT))
*attrs |= (xoap->xoa_projinherit == 0) ? 0 :
XAT0_PROJINHERIT;
}
static void *
zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
{
zfs_fuid_t *zfuid;
uint64_t *fuidloc = start;
/* First copy in the ACE FUIDs */
for (zfuid = list_head(&fuidp->z_fuids); zfuid;
zfuid = list_next(&fuidp->z_fuids, zfuid)) {
*fuidloc++ = zfuid->z_logfuid;
}
return (fuidloc);
}
static void *
zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
{
zfs_fuid_domain_t *zdomain;
/* now copy in the domain info, if any */
if (fuidp->z_domain_str_sz != 0) {
for (zdomain = list_head(&fuidp->z_domains); zdomain;
zdomain = list_next(&fuidp->z_domains, zdomain)) {
bcopy((void *)zdomain->z_domain, start,
strlen(zdomain->z_domain) + 1);
start = (caddr_t)start +
strlen(zdomain->z_domain) + 1;
}
}
return (start);
}
/*
* If zp is an xattr node, check whether the xattr owner is unlinked.
* We don't want to log anything if the owner is unlinked.
*/
static int
zfs_xattr_owner_unlinked(znode_t *zp)
{
int unlinked = 0;
znode_t *dzp;
#ifdef __FreeBSD__
znode_t *tzp = zp;
/*
* zrele drops the vnode lock which violates the VOP locking contract
* on FreeBSD. See comment at the top of zfs_replay.c for more detail.
*/
/*
* if zp is XATTR node, keep walking up via z_xattr_parent until we
* get the owner
*/
while (tzp->z_pflags & ZFS_XATTR) {
ASSERT3U(zp->z_xattr_parent, !=, 0);
if (zfs_zget(ZTOZSB(tzp), tzp->z_xattr_parent, &dzp) != 0) {
unlinked = 1;
break;
}
if (tzp != zp)
zrele(tzp);
tzp = dzp;
unlinked = tzp->z_unlinked;
}
if (tzp != zp)
zrele(tzp);
#else
zhold(zp);
/*
* if zp is XATTR node, keep walking up via z_xattr_parent until we
* get the owner
*/
while (zp->z_pflags & ZFS_XATTR) {
ASSERT3U(zp->z_xattr_parent, !=, 0);
if (zfs_zget(ZTOZSB(zp), zp->z_xattr_parent, &dzp) != 0) {
unlinked = 1;
break;
}
zrele(zp);
zp = dzp;
unlinked = zp->z_unlinked;
}
zrele(zp);
#endif
return (unlinked);
}
/*
* Handles TX_CREATE, TX_CREATE_ATTR, TX_MKDIR, TX_MKDIR_ATTR and
* TK_MKXATTR transactions.
*
* TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
* domain information appended prior to the name. In this case the
* uid/gid in the log record will be a log centric FUID.
*
* TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
* may contain attributes, ACL and optional fuid information.
*
* TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
* and ACL and normal users/groups in the ACEs.
*
* There may be an optional xvattr attribute information similar
* to zfs_log_setattr.
*
* Also, after the file name "domain" strings may be appended.
*/
void
zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
znode_t *dzp, znode_t *zp, const char *name, vsecattr_t *vsecp,
zfs_fuid_info_t *fuidp, vattr_t *vap)
{
itx_t *itx;
lr_create_t *lr;
lr_acl_create_t *lracl;
size_t aclsize = 0;
size_t xvatsize = 0;
size_t txsize;
xvattr_t *xvap = (xvattr_t *)vap;
void *end;
size_t lrsize;
size_t namesize = strlen(name) + 1;
size_t fuidsz = 0;
if (zil_replaying(zilog, tx) || zfs_xattr_owner_unlinked(dzp))
return;
/*
* If we have FUIDs present then add in space for
* domains and ACE fuid's if any.
*/
if (fuidp) {
fuidsz += fuidp->z_domain_str_sz;
fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
}
if (vap->va_mask & ATTR_XVATTR)
xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
(int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
(int)txtype == TX_MKXATTR) {
txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
lrsize = sizeof (*lr);
} else {
txsize =
sizeof (lr_acl_create_t) + namesize + fuidsz +
ZIL_ACE_LENGTH(aclsize) + xvatsize;
lrsize = sizeof (lr_acl_create_t);
}
itx = zil_itx_create(txtype, txsize);
lr = (lr_create_t *)&itx->itx_lr;
lr->lr_doid = dzp->z_id;
lr->lr_foid = zp->z_id;
/* Store dnode slot count in 8 bits above object id. */
LR_FOID_SET_SLOTS(lr->lr_foid, zp->z_dnodesize >> DNODE_SHIFT);
lr->lr_mode = zp->z_mode;
if (!IS_EPHEMERAL(KUID_TO_SUID(ZTOUID(zp)))) {
lr->lr_uid = (uint64_t)KUID_TO_SUID(ZTOUID(zp));
} else {
lr->lr_uid = fuidp->z_fuid_owner;
}
if (!IS_EPHEMERAL(KGID_TO_SGID(ZTOGID(zp)))) {
lr->lr_gid = (uint64_t)KGID_TO_SGID(ZTOGID(zp));
} else {
lr->lr_gid = fuidp->z_fuid_group;
}
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
sizeof (uint64_t));
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
lr->lr_crtime, sizeof (uint64_t) * 2);
if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(ZTOZSB(zp)), &lr->lr_rdev,
sizeof (lr->lr_rdev)) != 0)
lr->lr_rdev = 0;
/*
* Fill in xvattr info if any
*/
if (vap->va_mask & ATTR_XVATTR) {
zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
end = (caddr_t)lr + lrsize + xvatsize;
} else {
end = (caddr_t)lr + lrsize;
}
/* Now fill in any ACL info */
if (vsecp) {
lracl = (lr_acl_create_t *)&itx->itx_lr;
lracl->lr_aclcnt = vsecp->vsa_aclcnt;
lracl->lr_acl_bytes = aclsize;
lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
lracl->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
else
lracl->lr_acl_flags = 0;
bcopy(vsecp->vsa_aclentp, end, aclsize);
end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
}
/* drop in FUID info */
if (fuidp) {
end = zfs_log_fuid_ids(fuidp, end);
end = zfs_log_fuid_domains(fuidp, end);
}
/*
* Now place file name in log record
*/
bcopy(name, end, namesize);
zil_itx_assign(zilog, itx, tx);
}
/*
* Handles both TX_REMOVE and TX_RMDIR transactions.
*/
void
zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
znode_t *dzp, const char *name, uint64_t foid, boolean_t unlinked)
{
itx_t *itx;
lr_remove_t *lr;
size_t namesize = strlen(name) + 1;
if (zil_replaying(zilog, tx) || zfs_xattr_owner_unlinked(dzp))
return;
itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
lr = (lr_remove_t *)&itx->itx_lr;
lr->lr_doid = dzp->z_id;
bcopy(name, (char *)(lr + 1), namesize);
itx->itx_oid = foid;
/*
* Object ids can be re-instantiated in the next txg so
* remove any async transactions to avoid future leaks.
* This can happen if a fsync occurs on the re-instantiated
* object for a WR_INDIRECT or WR_NEED_COPY write, which gets
* the new file data and flushes a write record for the old object.
*/
if (unlinked) {
ASSERT((txtype & ~TX_CI) == TX_REMOVE);
zil_remove_async(zilog, foid);
}
zil_itx_assign(zilog, itx, tx);
}
/*
* Handles TX_LINK transactions.
*/
void
zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
znode_t *dzp, znode_t *zp, const char *name)
{
itx_t *itx;
lr_link_t *lr;
size_t namesize = strlen(name) + 1;
if (zil_replaying(zilog, tx))
return;
itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
lr = (lr_link_t *)&itx->itx_lr;
lr->lr_doid = dzp->z_id;
lr->lr_link_obj = zp->z_id;
bcopy(name, (char *)(lr + 1), namesize);
zil_itx_assign(zilog, itx, tx);
}
/*
* Handles TX_SYMLINK transactions.
*/
void
zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
znode_t *dzp, znode_t *zp, const char *name, const char *link)
{
itx_t *itx;
lr_create_t *lr;
size_t namesize = strlen(name) + 1;
size_t linksize = strlen(link) + 1;
if (zil_replaying(zilog, tx))
return;
itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
lr = (lr_create_t *)&itx->itx_lr;
lr->lr_doid = dzp->z_id;
lr->lr_foid = zp->z_id;
lr->lr_uid = KUID_TO_SUID(ZTOUID(zp));
lr->lr_gid = KGID_TO_SGID(ZTOGID(zp));
lr->lr_mode = zp->z_mode;
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
sizeof (uint64_t));
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
lr->lr_crtime, sizeof (uint64_t) * 2);
bcopy(name, (char *)(lr + 1), namesize);
bcopy(link, (char *)(lr + 1) + namesize, linksize);
zil_itx_assign(zilog, itx, tx);
}
/*
* Handles TX_RENAME transactions.
*/
void
zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, znode_t *sdzp,
const char *sname, znode_t *tdzp, const char *dname, znode_t *szp)
{
itx_t *itx;
lr_rename_t *lr;
size_t snamesize = strlen(sname) + 1;
size_t dnamesize = strlen(dname) + 1;
if (zil_replaying(zilog, tx))
return;
itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
lr = (lr_rename_t *)&itx->itx_lr;
lr->lr_sdoid = sdzp->z_id;
lr->lr_tdoid = tdzp->z_id;
bcopy(sname, (char *)(lr + 1), snamesize);
bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
itx->itx_oid = szp->z_id;
zil_itx_assign(zilog, itx, tx);
}
/*
* zfs_log_write() handles TX_WRITE transactions. The specified callback is
* called as soon as the write is on stable storage (be it via a DMU sync or a
* ZIL commit).
*/
-long zfs_immediate_write_sz = 32768;
+static long zfs_immediate_write_sz = 32768;
void
zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
znode_t *zp, offset_t off, ssize_t resid, int ioflag,
zil_callback_t callback, void *callback_data)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)sa_get_db(zp->z_sa_hdl);
uint32_t blocksize = zp->z_blksz;
itx_wr_state_t write_state;
uintptr_t fsync_cnt;
uint64_t gen = 0;
ssize_t size = resid;
if (zil_replaying(zilog, tx) || zp->z_unlinked ||
zfs_xattr_owner_unlinked(zp)) {
if (callback != NULL)
callback(callback_data);
return;
}
if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
write_state = WR_INDIRECT;
else if (!spa_has_slogs(zilog->zl_spa) &&
resid >= zfs_immediate_write_sz)
write_state = WR_INDIRECT;
else if (ioflag & (O_SYNC | O_DSYNC))
write_state = WR_COPIED;
else
write_state = WR_NEED_COPY;
if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
(void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
}
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &gen,
sizeof (gen));
while (resid) {
itx_t *itx;
lr_write_t *lr;
itx_wr_state_t wr_state = write_state;
ssize_t len = resid;
/*
* A WR_COPIED record must fit entirely in one log block.
* Large writes can use WR_NEED_COPY, which the ZIL will
* split into multiple records across several log blocks
* if necessary.
*/
if (wr_state == WR_COPIED &&
resid > zil_max_copied_data(zilog))
wr_state = WR_NEED_COPY;
else if (wr_state == WR_INDIRECT)
len = MIN(blocksize - P2PHASE(off, blocksize), resid);
itx = zil_itx_create(txtype, sizeof (*lr) +
(wr_state == WR_COPIED ? len : 0));
lr = (lr_write_t *)&itx->itx_lr;
/*
* For WR_COPIED records, copy the data into the lr_write_t.
*/
if (wr_state == WR_COPIED) {
int err;
DB_DNODE_ENTER(db);
err = dmu_read_by_dnode(DB_DNODE(db), off, len, lr + 1,
DMU_READ_NO_PREFETCH);
if (err != 0) {
zil_itx_destroy(itx);
itx = zil_itx_create(txtype, sizeof (*lr));
lr = (lr_write_t *)&itx->itx_lr;
wr_state = WR_NEED_COPY;
}
DB_DNODE_EXIT(db);
}
itx->itx_wr_state = wr_state;
lr->lr_foid = zp->z_id;
lr->lr_offset = off;
lr->lr_length = len;
lr->lr_blkoff = 0;
BP_ZERO(&lr->lr_blkptr);
itx->itx_private = ZTOZSB(zp);
itx->itx_gen = gen;
if (!(ioflag & (O_SYNC | O_DSYNC)) && (zp->z_sync_cnt == 0) &&
(fsync_cnt == 0))
itx->itx_sync = B_FALSE;
itx->itx_callback = callback;
itx->itx_callback_data = callback_data;
zil_itx_assign(zilog, itx, tx);
off += len;
resid -= len;
}
if (write_state == WR_COPIED || write_state == WR_NEED_COPY) {
dsl_pool_wrlog_count(zilog->zl_dmu_pool, size, tx->tx_txg);
}
}
/*
* Handles TX_TRUNCATE transactions.
*/
void
zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
znode_t *zp, uint64_t off, uint64_t len)
{
itx_t *itx;
lr_truncate_t *lr;
if (zil_replaying(zilog, tx) || zp->z_unlinked ||
zfs_xattr_owner_unlinked(zp))
return;
itx = zil_itx_create(txtype, sizeof (*lr));
lr = (lr_truncate_t *)&itx->itx_lr;
lr->lr_foid = zp->z_id;
lr->lr_offset = off;
lr->lr_length = len;
itx->itx_sync = (zp->z_sync_cnt != 0);
zil_itx_assign(zilog, itx, tx);
}
/*
* Handles TX_SETATTR transactions.
*/
void
zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
{
itx_t *itx;
lr_setattr_t *lr;
xvattr_t *xvap = (xvattr_t *)vap;
size_t recsize = sizeof (lr_setattr_t);
void *start;
if (zil_replaying(zilog, tx) || zp->z_unlinked)
return;
/*
* If XVATTR set, then log record size needs to allow
* for lr_attr_t + xvattr mask, mapsize and create time
* plus actual attribute values
*/
if (vap->va_mask & ATTR_XVATTR)
recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
if (fuidp)
recsize += fuidp->z_domain_str_sz;
itx = zil_itx_create(txtype, recsize);
lr = (lr_setattr_t *)&itx->itx_lr;
lr->lr_foid = zp->z_id;
lr->lr_mask = (uint64_t)mask_applied;
lr->lr_mode = (uint64_t)vap->va_mode;
if ((mask_applied & ATTR_UID) && IS_EPHEMERAL(vap->va_uid))
lr->lr_uid = fuidp->z_fuid_owner;
else
lr->lr_uid = (uint64_t)vap->va_uid;
if ((mask_applied & ATTR_GID) && IS_EPHEMERAL(vap->va_gid))
lr->lr_gid = fuidp->z_fuid_group;
else
lr->lr_gid = (uint64_t)vap->va_gid;
lr->lr_size = (uint64_t)vap->va_size;
ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
start = (lr_setattr_t *)(lr + 1);
if (vap->va_mask & ATTR_XVATTR) {
zfs_log_xvattr((lr_attr_t *)start, xvap);
start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
}
/*
* Now stick on domain information if any on end
*/
if (fuidp)
(void) zfs_log_fuid_domains(fuidp, start);
itx->itx_sync = (zp->z_sync_cnt != 0);
zil_itx_assign(zilog, itx, tx);
}
/*
* Handles TX_ACL transactions.
*/
void
zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
{
itx_t *itx;
lr_acl_v0_t *lrv0;
lr_acl_t *lr;
int txtype;
int lrsize;
size_t txsize;
size_t aclbytes = vsecp->vsa_aclentsz;
if (zil_replaying(zilog, tx) || zp->z_unlinked)
return;
txtype = (ZTOZSB(zp)->z_version < ZPL_VERSION_FUID) ?
TX_ACL_V0 : TX_ACL;
if (txtype == TX_ACL)
lrsize = sizeof (*lr);
else
lrsize = sizeof (*lrv0);
txsize = lrsize +
((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
(fuidp ? fuidp->z_domain_str_sz : 0) +
sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
itx = zil_itx_create(txtype, txsize);
lr = (lr_acl_t *)&itx->itx_lr;
lr->lr_foid = zp->z_id;
if (txtype == TX_ACL) {
lr->lr_acl_bytes = aclbytes;
lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
else
lr->lr_acl_flags = 0;
}
lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
if (txtype == TX_ACL_V0) {
lrv0 = (lr_acl_v0_t *)lr;
bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes);
} else {
void *start = (ace_t *)(lr + 1);
bcopy(vsecp->vsa_aclentp, start, aclbytes);
start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
if (fuidp) {
start = zfs_log_fuid_ids(fuidp, start);
(void) zfs_log_fuid_domains(fuidp, start);
}
}
itx->itx_sync = (zp->z_sync_cnt != 0);
zil_itx_assign(zilog, itx, tx);
}
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, immediate_write_sz, LONG, ZMOD_RW,
"Largest data block to write to zil");
/* END CSTYLED */
diff --git a/module/zfs/zfs_replay.c b/module/zfs/zfs_replay.c
index e6ed3e738e40..f76725f0b716 100644
--- a/module/zfs/zfs_replay.c
+++ b/module/zfs/zfs_replay.c
@@ -1,992 +1,992 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012 Cyril Plisko. All rights reserved.
* Copyright (c) 2013, 2017 by Delphix. All rights reserved.
*/
#include <sys/types.h>
#include <sys/param.h>
#include <sys/sysmacros.h>
#include <sys/cmn_err.h>
#include <sys/kmem.h>
#include <sys/thread.h>
#include <sys/file.h>
#include <sys/fcntl.h>
#include <sys/vfs.h>
#include <sys/fs/zfs.h>
#include <sys/zfs_znode.h>
#include <sys/zfs_dir.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_fuid.h>
#include <sys/zfs_vnops.h>
#include <sys/spa.h>
#include <sys/zil.h>
#include <sys/byteorder.h>
#include <sys/stat.h>
#include <sys/acl.h>
#include <sys/atomic.h>
#include <sys/cred.h>
#include <sys/zpl.h>
/*
* NB: FreeBSD expects to be able to do vnode locking in lookup and
* hold the locks across all subsequent VOPs until vput is called.
* This means that its zfs vnops routines can't do any internal locking.
* In order to have the same contract as the Linux vnops there would
* needed to be duplicate locked vnops. If the vnops were used more widely
* in common code this would likely be preferable. However, currently
* this is the only file where this is the case.
*/
/*
* Functions to replay ZFS intent log (ZIL) records
* The functions are called through a function vector (zfs_replay_vector)
* which is indexed by the transaction type.
*/
static void
zfs_init_vattr(vattr_t *vap, uint64_t mask, uint64_t mode,
uint64_t uid, uint64_t gid, uint64_t rdev, uint64_t nodeid)
{
bzero(vap, sizeof (*vap));
vap->va_mask = (uint_t)mask;
vap->va_mode = mode;
#if defined(__FreeBSD__) || defined(__APPLE__)
vap->va_type = IFTOVT(mode);
#endif
vap->va_uid = (uid_t)(IS_EPHEMERAL(uid)) ? -1 : uid;
vap->va_gid = (gid_t)(IS_EPHEMERAL(gid)) ? -1 : gid;
vap->va_rdev = zfs_cmpldev(rdev);
vap->va_nodeid = nodeid;
}
/* ARGSUSED */
static int
zfs_replay_error(void *arg1, void *arg2, boolean_t byteswap)
{
return (SET_ERROR(ENOTSUP));
}
static void
zfs_replay_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
{
xoptattr_t *xoap = NULL;
uint64_t *attrs;
uint64_t *crtime;
uint32_t *bitmap;
void *scanstamp;
int i;
xvap->xva_vattr.va_mask |= ATTR_XVATTR;
if ((xoap = xva_getxoptattr(xvap)) == NULL) {
xvap->xva_vattr.va_mask &= ~ATTR_XVATTR; /* shouldn't happen */
return;
}
ASSERT(lrattr->lr_attr_masksize == xvap->xva_mapsize);
bitmap = &lrattr->lr_attr_bitmap;
for (i = 0; i != lrattr->lr_attr_masksize; i++, bitmap++)
xvap->xva_reqattrmap[i] = *bitmap;
attrs = (uint64_t *)(lrattr + lrattr->lr_attr_masksize - 1);
crtime = attrs + 1;
scanstamp = (caddr_t)(crtime + 2);
if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
xoap->xoa_hidden = ((*attrs & XAT0_HIDDEN) != 0);
if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
xoap->xoa_system = ((*attrs & XAT0_SYSTEM) != 0);
if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
xoap->xoa_archive = ((*attrs & XAT0_ARCHIVE) != 0);
if (XVA_ISSET_REQ(xvap, XAT_READONLY))
xoap->xoa_readonly = ((*attrs & XAT0_READONLY) != 0);
if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
xoap->xoa_immutable = ((*attrs & XAT0_IMMUTABLE) != 0);
if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
xoap->xoa_nounlink = ((*attrs & XAT0_NOUNLINK) != 0);
if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
xoap->xoa_appendonly = ((*attrs & XAT0_APPENDONLY) != 0);
if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
xoap->xoa_nodump = ((*attrs & XAT0_NODUMP) != 0);
if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
xoap->xoa_opaque = ((*attrs & XAT0_OPAQUE) != 0);
if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
xoap->xoa_av_modified = ((*attrs & XAT0_AV_MODIFIED) != 0);
if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
xoap->xoa_av_quarantined =
((*attrs & XAT0_AV_QUARANTINED) != 0);
if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
ZFS_TIME_DECODE(&xoap->xoa_createtime, crtime);
if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
ASSERT(!XVA_ISSET_REQ(xvap, XAT_PROJID));
bcopy(scanstamp, xoap->xoa_av_scanstamp, AV_SCANSTAMP_SZ);
} else if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
/*
* XAT_PROJID and XAT_AV_SCANSTAMP will never be valid
* at the same time, so we can share the same space.
*/
bcopy(scanstamp, &xoap->xoa_projid, sizeof (uint64_t));
}
if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
xoap->xoa_reparse = ((*attrs & XAT0_REPARSE) != 0);
if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
xoap->xoa_offline = ((*attrs & XAT0_OFFLINE) != 0);
if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
xoap->xoa_sparse = ((*attrs & XAT0_SPARSE) != 0);
if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT))
xoap->xoa_projinherit = ((*attrs & XAT0_PROJINHERIT) != 0);
}
static int
zfs_replay_domain_cnt(uint64_t uid, uint64_t gid)
{
uint64_t uid_idx;
uint64_t gid_idx;
int domcnt = 0;
uid_idx = FUID_INDEX(uid);
gid_idx = FUID_INDEX(gid);
if (uid_idx)
domcnt++;
if (gid_idx > 0 && gid_idx != uid_idx)
domcnt++;
return (domcnt);
}
static void *
zfs_replay_fuid_domain_common(zfs_fuid_info_t *fuid_infop, void *start,
int domcnt)
{
int i;
for (i = 0; i != domcnt; i++) {
fuid_infop->z_domain_table[i] = start;
start = (caddr_t)start + strlen(start) + 1;
}
return (start);
}
/*
* Set the uid/gid in the fuid_info structure.
*/
static void
zfs_replay_fuid_ugid(zfs_fuid_info_t *fuid_infop, uint64_t uid, uint64_t gid)
{
/*
* If owner or group are log specific FUIDs then slurp up
* domain information and build zfs_fuid_info_t
*/
if (IS_EPHEMERAL(uid))
fuid_infop->z_fuid_owner = uid;
if (IS_EPHEMERAL(gid))
fuid_infop->z_fuid_group = gid;
}
/*
* Load fuid domains into fuid_info_t
*/
static zfs_fuid_info_t *
zfs_replay_fuid_domain(void *buf, void **end, uint64_t uid, uint64_t gid)
{
int domcnt;
zfs_fuid_info_t *fuid_infop;
fuid_infop = zfs_fuid_info_alloc();
domcnt = zfs_replay_domain_cnt(uid, gid);
if (domcnt == 0)
return (fuid_infop);
fuid_infop->z_domain_table =
kmem_zalloc(domcnt * sizeof (char *), KM_SLEEP);
zfs_replay_fuid_ugid(fuid_infop, uid, gid);
fuid_infop->z_domain_cnt = domcnt;
*end = zfs_replay_fuid_domain_common(fuid_infop, buf, domcnt);
return (fuid_infop);
}
/*
* load zfs_fuid_t's and fuid_domains into fuid_info_t
*/
static zfs_fuid_info_t *
zfs_replay_fuids(void *start, void **end, int idcnt, int domcnt, uint64_t uid,
uint64_t gid)
{
uint64_t *log_fuid = (uint64_t *)start;
zfs_fuid_info_t *fuid_infop;
int i;
fuid_infop = zfs_fuid_info_alloc();
fuid_infop->z_domain_cnt = domcnt;
fuid_infop->z_domain_table =
kmem_zalloc(domcnt * sizeof (char *), KM_SLEEP);
for (i = 0; i != idcnt; i++) {
zfs_fuid_t *zfuid;
zfuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
zfuid->z_logfuid = *log_fuid;
zfuid->z_id = -1;
zfuid->z_domidx = 0;
list_insert_tail(&fuid_infop->z_fuids, zfuid);
log_fuid++;
}
zfs_replay_fuid_ugid(fuid_infop, uid, gid);
*end = zfs_replay_fuid_domain_common(fuid_infop, log_fuid, domcnt);
return (fuid_infop);
}
static void
zfs_replay_swap_attrs(lr_attr_t *lrattr)
{
/* swap the lr_attr structure */
byteswap_uint32_array(lrattr, sizeof (*lrattr));
/* swap the bitmap */
byteswap_uint32_array(lrattr + 1, (lrattr->lr_attr_masksize - 1) *
sizeof (uint32_t));
/* swap the attributes, create time + 64 bit word for attributes */
byteswap_uint64_array((caddr_t)(lrattr + 1) + (sizeof (uint32_t) *
(lrattr->lr_attr_masksize - 1)), 3 * sizeof (uint64_t));
}
/*
* Replay file create with optional ACL, xvattr information as well
* as option FUID information.
*/
static int
zfs_replay_create_acl(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_acl_create_t *lracl = arg2;
char *name = NULL; /* location determined later */
lr_create_t *lr = (lr_create_t *)lracl;
znode_t *dzp;
znode_t *zp;
xvattr_t xva;
int vflg = 0;
vsecattr_t vsec = { 0 };
lr_attr_t *lrattr;
void *aclstart;
void *fuidstart;
size_t xvatlen = 0;
uint64_t txtype;
uint64_t objid;
uint64_t dnodesize;
int error;
txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
if (byteswap) {
byteswap_uint64_array(lracl, sizeof (*lracl));
if (txtype == TX_CREATE_ACL_ATTR ||
txtype == TX_MKDIR_ACL_ATTR) {
lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
zfs_replay_swap_attrs(lrattr);
xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
}
aclstart = (caddr_t)(lracl + 1) + xvatlen;
zfs_ace_byteswap(aclstart, lracl->lr_acl_bytes, B_FALSE);
/* swap fuids */
if (lracl->lr_fuidcnt) {
byteswap_uint64_array((caddr_t)aclstart +
ZIL_ACE_LENGTH(lracl->lr_acl_bytes),
lracl->lr_fuidcnt * sizeof (uint64_t));
}
}
if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
return (error);
objid = LR_FOID_GET_OBJ(lr->lr_foid);
dnodesize = LR_FOID_GET_SLOTS(lr->lr_foid) << DNODE_SHIFT;
xva_init(&xva);
zfs_init_vattr(&xva.xva_vattr, ATTR_MODE | ATTR_UID | ATTR_GID,
lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, objid);
/*
* All forms of zfs create (create, mkdir, mkxattrdir, symlink)
* eventually end up in zfs_mknode(), which assigns the object's
* creation time, generation number, and dnode size. The generic
* zfs_create() has no concept of these attributes, so we smuggle
* the values inside the vattr's otherwise unused va_ctime,
* va_nblocks, and va_fsid fields.
*/
ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
xva.xva_vattr.va_nblocks = lr->lr_gen;
xva.xva_vattr.va_fsid = dnodesize;
error = dnode_try_claim(zfsvfs->z_os, objid, dnodesize >> DNODE_SHIFT);
if (error)
goto bail;
if (lr->lr_common.lrc_txtype & TX_CI)
vflg |= FIGNORECASE;
switch (txtype) {
case TX_CREATE_ACL:
aclstart = (caddr_t)(lracl + 1);
fuidstart = (caddr_t)aclstart +
ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
(void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
lr->lr_uid, lr->lr_gid);
fallthrough;
case TX_CREATE_ACL_ATTR:
if (name == NULL) {
lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
xva.xva_vattr.va_mask |= ATTR_XVATTR;
zfs_replay_xvattr(lrattr, &xva);
}
vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
vsec.vsa_aclcnt = lracl->lr_aclcnt;
vsec.vsa_aclentsz = lracl->lr_acl_bytes;
vsec.vsa_aclflags = lracl->lr_acl_flags;
if (zfsvfs->z_fuid_replay == NULL) {
fuidstart = (caddr_t)(lracl + 1) + xvatlen +
ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
zfsvfs->z_fuid_replay =
zfs_replay_fuids(fuidstart,
(void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
lr->lr_uid, lr->lr_gid);
}
error = zfs_create(dzp, name, &xva.xva_vattr,
0, 0, &zp, kcred, vflg, &vsec);
break;
case TX_MKDIR_ACL:
aclstart = (caddr_t)(lracl + 1);
fuidstart = (caddr_t)aclstart +
ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
(void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
lr->lr_uid, lr->lr_gid);
fallthrough;
case TX_MKDIR_ACL_ATTR:
if (name == NULL) {
lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
zfs_replay_xvattr(lrattr, &xva);
}
vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
vsec.vsa_aclcnt = lracl->lr_aclcnt;
vsec.vsa_aclentsz = lracl->lr_acl_bytes;
vsec.vsa_aclflags = lracl->lr_acl_flags;
if (zfsvfs->z_fuid_replay == NULL) {
fuidstart = (caddr_t)(lracl + 1) + xvatlen +
ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
zfsvfs->z_fuid_replay =
zfs_replay_fuids(fuidstart,
(void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
lr->lr_uid, lr->lr_gid);
}
error = zfs_mkdir(dzp, name, &xva.xva_vattr,
&zp, kcred, vflg, &vsec);
break;
default:
error = SET_ERROR(ENOTSUP);
}
bail:
if (error == 0 && zp != NULL) {
#ifdef __FreeBSD__
VOP_UNLOCK1(ZTOV(zp));
#endif
zrele(zp);
}
zrele(dzp);
if (zfsvfs->z_fuid_replay)
zfs_fuid_info_free(zfsvfs->z_fuid_replay);
zfsvfs->z_fuid_replay = NULL;
return (error);
}
static int
zfs_replay_create(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_create_t *lr = arg2;
char *name = NULL; /* location determined later */
char *link; /* symlink content follows name */
znode_t *dzp;
znode_t *zp = NULL;
xvattr_t xva;
int vflg = 0;
size_t lrsize = sizeof (lr_create_t);
lr_attr_t *lrattr;
void *start;
size_t xvatlen;
uint64_t txtype;
uint64_t objid;
uint64_t dnodesize;
int error;
txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
if (byteswap) {
byteswap_uint64_array(lr, sizeof (*lr));
if (txtype == TX_CREATE_ATTR || txtype == TX_MKDIR_ATTR)
zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
}
if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
return (error);
objid = LR_FOID_GET_OBJ(lr->lr_foid);
dnodesize = LR_FOID_GET_SLOTS(lr->lr_foid) << DNODE_SHIFT;
xva_init(&xva);
zfs_init_vattr(&xva.xva_vattr, ATTR_MODE | ATTR_UID | ATTR_GID,
lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, objid);
/*
* All forms of zfs create (create, mkdir, mkxattrdir, symlink)
* eventually end up in zfs_mknode(), which assigns the object's
* creation time, generation number, and dnode slot count. The
* generic zfs_create() has no concept of these attributes, so
* we smuggle the values inside the vattr's otherwise unused
* va_ctime, va_nblocks, and va_fsid fields.
*/
ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
xva.xva_vattr.va_nblocks = lr->lr_gen;
xva.xva_vattr.va_fsid = dnodesize;
error = dnode_try_claim(zfsvfs->z_os, objid, dnodesize >> DNODE_SHIFT);
if (error)
goto out;
if (lr->lr_common.lrc_txtype & TX_CI)
vflg |= FIGNORECASE;
/*
* Symlinks don't have fuid info, and CIFS never creates
* symlinks.
*
* The _ATTR versions will grab the fuid info in their subcases.
*/
if ((int)lr->lr_common.lrc_txtype != TX_SYMLINK &&
(int)lr->lr_common.lrc_txtype != TX_MKDIR_ATTR &&
(int)lr->lr_common.lrc_txtype != TX_CREATE_ATTR) {
start = (lr + 1);
zfsvfs->z_fuid_replay =
zfs_replay_fuid_domain(start, &start,
lr->lr_uid, lr->lr_gid);
}
switch (txtype) {
case TX_CREATE_ATTR:
lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
start = (caddr_t)(lr + 1) + xvatlen;
zfsvfs->z_fuid_replay =
zfs_replay_fuid_domain(start, &start,
lr->lr_uid, lr->lr_gid);
name = (char *)start;
fallthrough;
case TX_CREATE:
if (name == NULL)
name = (char *)start;
error = zfs_create(dzp, name, &xva.xva_vattr,
0, 0, &zp, kcred, vflg, NULL);
break;
case TX_MKDIR_ATTR:
lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
start = (caddr_t)(lr + 1) + xvatlen;
zfsvfs->z_fuid_replay =
zfs_replay_fuid_domain(start, &start,
lr->lr_uid, lr->lr_gid);
name = (char *)start;
fallthrough;
case TX_MKDIR:
if (name == NULL)
name = (char *)(lr + 1);
error = zfs_mkdir(dzp, name, &xva.xva_vattr,
&zp, kcred, vflg, NULL);
break;
case TX_MKXATTR:
error = zfs_make_xattrdir(dzp, &xva.xva_vattr, &zp, kcred);
break;
case TX_SYMLINK:
name = (char *)(lr + 1);
link = name + strlen(name) + 1;
error = zfs_symlink(dzp, name, &xva.xva_vattr,
link, &zp, kcred, vflg);
break;
default:
error = SET_ERROR(ENOTSUP);
}
out:
if (error == 0 && zp != NULL) {
#ifdef __FreeBSD__
VOP_UNLOCK1(ZTOV(zp));
#endif
zrele(zp);
}
zrele(dzp);
if (zfsvfs->z_fuid_replay)
zfs_fuid_info_free(zfsvfs->z_fuid_replay);
zfsvfs->z_fuid_replay = NULL;
return (error);
}
static int
zfs_replay_remove(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_remove_t *lr = arg2;
char *name = (char *)(lr + 1); /* name follows lr_remove_t */
znode_t *dzp;
int error;
int vflg = 0;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
return (error);
if (lr->lr_common.lrc_txtype & TX_CI)
vflg |= FIGNORECASE;
switch ((int)lr->lr_common.lrc_txtype) {
case TX_REMOVE:
error = zfs_remove(dzp, name, kcred, vflg);
break;
case TX_RMDIR:
error = zfs_rmdir(dzp, name, NULL, kcred, vflg);
break;
default:
error = SET_ERROR(ENOTSUP);
}
zrele(dzp);
return (error);
}
static int
zfs_replay_link(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_link_t *lr = arg2;
char *name = (char *)(lr + 1); /* name follows lr_link_t */
znode_t *dzp, *zp;
int error;
int vflg = 0;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
return (error);
if ((error = zfs_zget(zfsvfs, lr->lr_link_obj, &zp)) != 0) {
zrele(dzp);
return (error);
}
if (lr->lr_common.lrc_txtype & TX_CI)
vflg |= FIGNORECASE;
error = zfs_link(dzp, zp, name, kcred, vflg);
zrele(zp);
zrele(dzp);
return (error);
}
static int
zfs_replay_rename(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_rename_t *lr = arg2;
char *sname = (char *)(lr + 1); /* sname and tname follow lr_rename_t */
char *tname = sname + strlen(sname) + 1;
znode_t *sdzp, *tdzp;
int error;
int vflg = 0;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0)
return (error);
if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) {
zrele(sdzp);
return (error);
}
if (lr->lr_common.lrc_txtype & TX_CI)
vflg |= FIGNORECASE;
error = zfs_rename(sdzp, sname, tdzp, tname, kcred, vflg);
zrele(tdzp);
zrele(sdzp);
return (error);
}
static int
zfs_replay_write(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_write_t *lr = arg2;
char *data = (char *)(lr + 1); /* data follows lr_write_t */
znode_t *zp;
int error;
uint64_t eod, offset, length;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
/*
* As we can log writes out of order, it's possible the
* file has been removed. In this case just drop the write
* and return success.
*/
if (error == ENOENT)
error = 0;
return (error);
}
offset = lr->lr_offset;
length = lr->lr_length;
eod = offset + length; /* end of data for this write */
/*
* This may be a write from a dmu_sync() for a whole block,
* and may extend beyond the current end of the file.
* We can't just replay what was written for this TX_WRITE as
* a future TX_WRITE2 may extend the eof and the data for that
* write needs to be there. So we write the whole block and
* reduce the eof. This needs to be done within the single dmu
* transaction created within vn_rdwr -> zfs_write. So a possible
* new end of file is passed through in zfsvfs->z_replay_eof
*/
zfsvfs->z_replay_eof = 0; /* 0 means don't change end of file */
/* If it's a dmu_sync() block, write the whole block */
if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
if (length < blocksize) {
offset -= offset % blocksize;
length = blocksize;
}
if (zp->z_size < eod)
zfsvfs->z_replay_eof = eod;
}
error = zfs_write_simple(zp, data, length, offset, NULL);
zrele(zp);
zfsvfs->z_replay_eof = 0; /* safety */
return (error);
}
/*
* TX_WRITE2 are only generated when dmu_sync() returns EALREADY
* meaning the pool block is already being synced. So now that we always write
* out full blocks, all we have to do is expand the eof if
* the file is grown.
*/
static int
zfs_replay_write2(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_write_t *lr = arg2;
znode_t *zp;
int error;
uint64_t end;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
return (error);
top:
end = lr->lr_offset + lr->lr_length;
if (end > zp->z_size) {
dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
zp->z_size = end;
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
zrele(zp);
if (error == ERESTART) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
return (error);
}
(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
(void *)&zp->z_size, sizeof (uint64_t), tx);
/* Ensure the replayed seq is updated */
(void) zil_replaying(zfsvfs->z_log, tx);
dmu_tx_commit(tx);
}
zrele(zp);
return (error);
}
static int
zfs_replay_truncate(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_truncate_t *lr = arg2;
znode_t *zp;
flock64_t fl;
int error;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
return (error);
bzero(&fl, sizeof (fl));
fl.l_type = F_WRLCK;
fl.l_whence = SEEK_SET;
fl.l_start = lr->lr_offset;
fl.l_len = lr->lr_length;
error = zfs_space(zp, F_FREESP, &fl, O_RDWR | O_LARGEFILE,
lr->lr_offset, kcred);
zrele(zp);
return (error);
}
static int
zfs_replay_setattr(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_setattr_t *lr = arg2;
znode_t *zp;
xvattr_t xva;
vattr_t *vap = &xva.xva_vattr;
int error;
void *start;
xva_init(&xva);
if (byteswap) {
byteswap_uint64_array(lr, sizeof (*lr));
if ((lr->lr_mask & ATTR_XVATTR) &&
zfsvfs->z_version >= ZPL_VERSION_INITIAL)
zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
}
if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
return (error);
zfs_init_vattr(vap, lr->lr_mask, lr->lr_mode,
lr->lr_uid, lr->lr_gid, 0, lr->lr_foid);
vap->va_size = lr->lr_size;
ZFS_TIME_DECODE(&vap->va_atime, lr->lr_atime);
ZFS_TIME_DECODE(&vap->va_mtime, lr->lr_mtime);
gethrestime(&vap->va_ctime);
vap->va_mask |= ATTR_CTIME;
/*
* Fill in xvattr_t portions if necessary.
*/
start = (lr_setattr_t *)(lr + 1);
if (vap->va_mask & ATTR_XVATTR) {
zfs_replay_xvattr((lr_attr_t *)start, &xva);
start = (caddr_t)start +
ZIL_XVAT_SIZE(((lr_attr_t *)start)->lr_attr_masksize);
} else
xva.xva_vattr.va_mask &= ~ATTR_XVATTR;
zfsvfs->z_fuid_replay = zfs_replay_fuid_domain(start, &start,
lr->lr_uid, lr->lr_gid);
error = zfs_setattr(zp, vap, 0, kcred);
zfs_fuid_info_free(zfsvfs->z_fuid_replay);
zfsvfs->z_fuid_replay = NULL;
zrele(zp);
return (error);
}
static int
zfs_replay_acl_v0(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_acl_v0_t *lr = arg2;
ace_t *ace = (ace_t *)(lr + 1); /* ace array follows lr_acl_t */
vsecattr_t vsa;
znode_t *zp;
int error;
if (byteswap) {
byteswap_uint64_array(lr, sizeof (*lr));
zfs_oldace_byteswap(ace, lr->lr_aclcnt);
}
if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
return (error);
bzero(&vsa, sizeof (vsa));
vsa.vsa_mask = VSA_ACE | VSA_ACECNT;
vsa.vsa_aclcnt = lr->lr_aclcnt;
vsa.vsa_aclentsz = sizeof (ace_t) * vsa.vsa_aclcnt;
vsa.vsa_aclflags = 0;
vsa.vsa_aclentp = ace;
error = zfs_setsecattr(zp, &vsa, 0, kcred);
zrele(zp);
return (error);
}
/*
* Replaying ACLs is complicated by FUID support.
* The log record may contain some optional data
* to be used for replaying FUID's. These pieces
* are the actual FUIDs that were created initially.
* The FUID table index may no longer be valid and
* during zfs_create() a new index may be assigned.
* Because of this the log will contain the original
* domain+rid in order to create a new FUID.
*
* The individual ACEs may contain an ephemeral uid/gid which is no
* longer valid and will need to be replaced with an actual FUID.
*
*/
static int
zfs_replay_acl(void *arg1, void *arg2, boolean_t byteswap)
{
zfsvfs_t *zfsvfs = arg1;
lr_acl_t *lr = arg2;
ace_t *ace = (ace_t *)(lr + 1);
vsecattr_t vsa;
znode_t *zp;
int error;
if (byteswap) {
byteswap_uint64_array(lr, sizeof (*lr));
zfs_ace_byteswap(ace, lr->lr_acl_bytes, B_FALSE);
if (lr->lr_fuidcnt) {
byteswap_uint64_array((caddr_t)ace +
ZIL_ACE_LENGTH(lr->lr_acl_bytes),
lr->lr_fuidcnt * sizeof (uint64_t));
}
}
if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
return (error);
bzero(&vsa, sizeof (vsa));
vsa.vsa_mask = VSA_ACE | VSA_ACECNT | VSA_ACE_ACLFLAGS;
vsa.vsa_aclcnt = lr->lr_aclcnt;
vsa.vsa_aclentp = ace;
vsa.vsa_aclentsz = lr->lr_acl_bytes;
vsa.vsa_aclflags = lr->lr_acl_flags;
if (lr->lr_fuidcnt) {
void *fuidstart = (caddr_t)ace +
ZIL_ACE_LENGTH(lr->lr_acl_bytes);
zfsvfs->z_fuid_replay =
zfs_replay_fuids(fuidstart, &fuidstart,
lr->lr_fuidcnt, lr->lr_domcnt, 0, 0);
}
error = zfs_setsecattr(zp, &vsa, 0, kcred);
if (zfsvfs->z_fuid_replay)
zfs_fuid_info_free(zfsvfs->z_fuid_replay);
zfsvfs->z_fuid_replay = NULL;
zrele(zp);
return (error);
}
/*
* Callback vectors for replaying records
*/
-zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE] = {
+zil_replay_func_t *const zfs_replay_vector[TX_MAX_TYPE] = {
zfs_replay_error, /* no such type */
zfs_replay_create, /* TX_CREATE */
zfs_replay_create, /* TX_MKDIR */
zfs_replay_create, /* TX_MKXATTR */
zfs_replay_create, /* TX_SYMLINK */
zfs_replay_remove, /* TX_REMOVE */
zfs_replay_remove, /* TX_RMDIR */
zfs_replay_link, /* TX_LINK */
zfs_replay_rename, /* TX_RENAME */
zfs_replay_write, /* TX_WRITE */
zfs_replay_truncate, /* TX_TRUNCATE */
zfs_replay_setattr, /* TX_SETATTR */
zfs_replay_acl_v0, /* TX_ACL_V0 */
zfs_replay_acl, /* TX_ACL */
zfs_replay_create_acl, /* TX_CREATE_ACL */
zfs_replay_create, /* TX_CREATE_ATTR */
zfs_replay_create_acl, /* TX_CREATE_ACL_ATTR */
zfs_replay_create_acl, /* TX_MKDIR_ACL */
zfs_replay_create, /* TX_MKDIR_ATTR */
zfs_replay_create_acl, /* TX_MKDIR_ACL_ATTR */
zfs_replay_write2, /* TX_WRITE2 */
};
diff --git a/module/zfs/zfs_sa.c b/module/zfs/zfs_sa.c
index 67be131da63b..817f63048789 100644
--- a/module/zfs/zfs_sa.c
+++ b/module/zfs/zfs_sa.c
@@ -1,446 +1,446 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/vnode.h>
#include <sys/sa.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_sa.h>
#include <sys/dmu_objset.h>
#include <sys/sa_impl.h>
/*
* ZPL attribute registration table.
* Order of attributes doesn't matter
* a unique value will be assigned for each
* attribute that is file system specific
*
* This is just the set of ZPL attributes that this
* version of ZFS deals with natively. The file system
* could have other attributes stored in files, but they will be
* ignored. The SA framework will preserve them, just that
* this version of ZFS won't change or delete them.
*/
-sa_attr_reg_t zfs_attr_table[ZPL_END+1] = {
+const sa_attr_reg_t zfs_attr_table[ZPL_END+1] = {
{"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
{"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
{"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
{"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
{"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
{"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
{"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
{"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
{"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
{"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
{"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
{"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
{"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
{"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
{"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
{"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
{"ZPL_DACL_COUNT", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
{"ZPL_SYMLINK", 0, SA_UINT8_ARRAY, 0},
{"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0},
{"ZPL_DACL_ACES", 0, SA_ACL, 0},
{"ZPL_DXATTR", 0, SA_UINT8_ARRAY, 0},
{"ZPL_PROJID", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
{NULL, 0, 0, 0}
};
#ifdef _KERNEL
int
zfs_sa_readlink(znode_t *zp, zfs_uio_t *uio)
{
dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
size_t bufsz;
int error;
bufsz = zp->z_size;
if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) {
error = zfs_uiomove((caddr_t)db->db_data +
ZFS_OLD_ZNODE_PHYS_SIZE,
MIN((size_t)bufsz, zfs_uio_resid(uio)), UIO_READ, uio);
} else {
dmu_buf_t *dbp;
if ((error = dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id,
0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) {
error = zfs_uiomove(dbp->db_data,
MIN((size_t)bufsz, zfs_uio_resid(uio)), UIO_READ,
uio);
dmu_buf_rele(dbp, FTAG);
}
}
return (error);
}
void
zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx)
{
dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
VERIFY0(dmu_set_bonus(db, len + ZFS_OLD_ZNODE_PHYS_SIZE, tx));
if (len) {
bcopy(link, (caddr_t)db->db_data +
ZFS_OLD_ZNODE_PHYS_SIZE, len);
}
} else {
dmu_buf_t *dbp;
zfs_grow_blocksize(zp, len, tx);
VERIFY0(dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id, 0, FTAG, &dbp,
DMU_READ_NO_PREFETCH));
dmu_buf_will_dirty(dbp, tx);
ASSERT3U(len, <=, dbp->db_size);
bcopy(link, dbp->db_data, len);
dmu_buf_rele(dbp, FTAG);
}
}
void
zfs_sa_get_scanstamp(znode_t *zp, xvattr_t *xvap)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
xoptattr_t *xoap;
ASSERT(MUTEX_HELD(&zp->z_lock));
VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
if (zp->z_is_sa) {
if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
&xoap->xoa_av_scanstamp,
sizeof (xoap->xoa_av_scanstamp)) != 0)
return;
} else {
dmu_object_info_t doi;
dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
int len;
if (!(zp->z_pflags & ZFS_BONUS_SCANSTAMP))
return;
sa_object_info(zp->z_sa_hdl, &doi);
len = sizeof (xoap->xoa_av_scanstamp) +
ZFS_OLD_ZNODE_PHYS_SIZE;
if (len <= doi.doi_bonus_size) {
(void) memcpy(xoap->xoa_av_scanstamp,
(caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
sizeof (xoap->xoa_av_scanstamp));
}
}
XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
}
void
zfs_sa_set_scanstamp(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
xoptattr_t *xoap;
ASSERT(MUTEX_HELD(&zp->z_lock));
VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
if (zp->z_is_sa)
VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
&xoap->xoa_av_scanstamp,
sizeof (xoap->xoa_av_scanstamp), tx));
else {
dmu_object_info_t doi;
dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
int len;
sa_object_info(zp->z_sa_hdl, &doi);
len = sizeof (xoap->xoa_av_scanstamp) +
ZFS_OLD_ZNODE_PHYS_SIZE;
if (len > doi.doi_bonus_size)
VERIFY(dmu_set_bonus(db, len, tx) == 0);
(void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp));
zp->z_pflags |= ZFS_BONUS_SCANSTAMP;
VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
&zp->z_pflags, sizeof (uint64_t), tx));
}
}
int
zfs_sa_get_xattr(znode_t *zp)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
char *obj;
int size;
int error;
ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
ASSERT(!zp->z_xattr_cached);
ASSERT(zp->z_is_sa);
error = sa_size(zp->z_sa_hdl, SA_ZPL_DXATTR(zfsvfs), &size);
if (error) {
if (error == ENOENT)
return nvlist_alloc(&zp->z_xattr_cached,
NV_UNIQUE_NAME, KM_SLEEP);
else
return (error);
}
obj = vmem_alloc(size, KM_SLEEP);
error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DXATTR(zfsvfs), obj, size);
if (error == 0)
error = nvlist_unpack(obj, size, &zp->z_xattr_cached, KM_SLEEP);
vmem_free(obj, size);
return (error);
}
int
zfs_sa_set_xattr(znode_t *zp)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
dmu_tx_t *tx;
char *obj;
size_t size;
int error;
ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
ASSERT(zp->z_xattr_cached);
ASSERT(zp->z_is_sa);
error = nvlist_size(zp->z_xattr_cached, &size, NV_ENCODE_XDR);
if ((error == 0) && (size > SA_ATTR_MAX_LEN))
error = SET_ERROR(EFBIG);
if (error)
goto out;
obj = vmem_alloc(size, KM_SLEEP);
error = nvlist_pack(zp->z_xattr_cached, &obj, &size,
NV_ENCODE_XDR, KM_SLEEP);
if (error)
goto out_free;
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_sa_create(tx, size);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
} else {
int count = 0;
sa_bulk_attr_t bulk[2];
uint64_t ctime[2];
zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DXATTR(zfsvfs),
NULL, obj, size);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
NULL, &ctime, 16);
VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
dmu_tx_commit(tx);
}
out_free:
vmem_free(obj, size);
out:
return (error);
}
/*
* I'm not convinced we should do any of this upgrade.
* since the SA code can read both old/new znode formats
* with probably little to no performance difference.
*
* All new files will be created with the new format.
*/
void
zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
{
dmu_buf_t *db = sa_get_db(hdl);
znode_t *zp = sa_get_userdata(hdl);
zfsvfs_t *zfsvfs = ZTOZSB(zp);
int count = 0;
sa_bulk_attr_t *bulk, *sa_attrs;
zfs_acl_locator_cb_t locate = { 0 };
uint64_t uid, gid, mode, rdev, xattr, parent, tmp_gen;
uint64_t crtime[2], mtime[2], ctime[2], atime[2];
uint64_t links;
zfs_acl_phys_t znode_acl;
char scanstamp[AV_SCANSTAMP_SZ];
boolean_t drop_lock = B_FALSE;
/*
* No upgrade if ACL isn't cached
* since we won't know which locks are held
* and ready the ACL would require special "locked"
* interfaces that would be messy
*/
if (zp->z_acl_cached == NULL || Z_ISLNK(ZTOTYPE(zp)))
return;
/*
* If the z_lock is held and we aren't the owner
* the just return since we don't want to deadlock
* trying to update the status of z_is_sa. This
* file can then be upgraded at a later time.
*
* Otherwise, we know we are doing the
* sa_update() that caused us to enter this function.
*/
if (MUTEX_NOT_HELD(&zp->z_lock)) {
if (mutex_tryenter(&zp->z_lock) == 0)
return;
else
drop_lock = B_TRUE;
}
/* First do a bulk query of the attributes that aren't cached */
bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, &xattr, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8);
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
&znode_acl, 88);
if (sa_bulk_lookup_locked(hdl, bulk, count) != 0)
goto done;
if (dmu_objset_projectquota_enabled(hdl->sa_os) &&
!(zp->z_pflags & ZFS_PROJID)) {
zp->z_pflags |= ZFS_PROJID;
zp->z_projid = ZFS_DEFAULT_PROJID;
}
/*
* While the order here doesn't matter its best to try and organize
* it is such a way to pick up an already existing layout number
*/
count = 0;
sa_attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
&zp->z_size, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zfsvfs),
NULL, &tmp_gen, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zfsvfs),
NULL, &parent, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
&zp->z_pflags, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zfsvfs), NULL,
&atime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zfsvfs), NULL,
&mtime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zfsvfs), NULL,
&ctime, 16);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
&crtime, 16);
links = ZTONLNK(zp);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL,
&links, 8);
if (dmu_objset_projectquota_enabled(hdl->sa_os))
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PROJID(zfsvfs), NULL,
&zp->z_projid, 8);
if (Z_ISBLK(ZTOTYPE(zp)) || Z_ISCHR(ZTOTYPE(zp)))
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
&rdev, 8);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
&zp->z_acl_cached->z_acl_count, 8);
if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
zfs_acl_xform(zp, zp->z_acl_cached, CRED());
locate.cb_aclp = zp->z_acl_cached;
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
if (xattr)
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zfsvfs),
NULL, &xattr, 8);
/* if scanstamp then add scanstamp */
if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
scanstamp, AV_SCANSTAMP_SZ);
SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zfsvfs),
NULL, scanstamp, AV_SCANSTAMP_SZ);
zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
}
VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
count, tx) == 0);
if (znode_acl.z_acl_extern_obj)
VERIFY(0 == dmu_object_free(zfsvfs->z_os,
znode_acl.z_acl_extern_obj, tx));
zp->z_is_sa = B_TRUE;
kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
done:
kmem_free(bulk, sizeof (sa_bulk_attr_t) * ZPL_END);
if (drop_lock)
mutex_exit(&zp->z_lock);
}
void
zfs_sa_upgrade_txholds(dmu_tx_t *tx, znode_t *zp)
{
if (!ZTOZSB(zp)->z_use_sa || zp->z_is_sa)
return;
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
if (zfs_external_acl(zp)) {
dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
DMU_OBJECT_END);
}
}
EXPORT_SYMBOL(zfs_attr_table);
EXPORT_SYMBOL(zfs_sa_readlink);
EXPORT_SYMBOL(zfs_sa_symlink);
EXPORT_SYMBOL(zfs_sa_get_scanstamp);
EXPORT_SYMBOL(zfs_sa_set_scanstamp);
EXPORT_SYMBOL(zfs_sa_get_xattr);
EXPORT_SYMBOL(zfs_sa_set_xattr);
EXPORT_SYMBOL(zfs_sa_upgrade);
EXPORT_SYMBOL(zfs_sa_upgrade_txholds);
#endif
diff --git a/module/zfs/zil.c b/module/zfs/zil.c
index b9f177daee53..85a17f10b38f 100644
--- a/module/zfs/zil.c
+++ b/module/zfs/zil.c
@@ -1,3734 +1,3735 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
* Copyright (c) 2014 Integros [integros.com]
* Copyright (c) 2018 Datto Inc.
*/
/* Portions Copyright 2010 Robert Milkowski */
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/dmu.h>
#include <sys/zap.h>
#include <sys/arc.h>
#include <sys/stat.h>
#include <sys/zil.h>
#include <sys/zil_impl.h>
#include <sys/dsl_dataset.h>
#include <sys/vdev_impl.h>
#include <sys/dmu_tx.h>
#include <sys/dsl_pool.h>
#include <sys/metaslab.h>
#include <sys/trace_zfs.h>
#include <sys/abd.h>
/*
* The ZFS Intent Log (ZIL) saves "transaction records" (itxs) of system
* calls that change the file system. Each itx has enough information to
* be able to replay them after a system crash, power loss, or
* equivalent failure mode. These are stored in memory until either:
*
* 1. they are committed to the pool by the DMU transaction group
* (txg), at which point they can be discarded; or
* 2. they are committed to the on-disk ZIL for the dataset being
* modified (e.g. due to an fsync, O_DSYNC, or other synchronous
* requirement).
*
* In the event of a crash or power loss, the itxs contained by each
* dataset's on-disk ZIL will be replayed when that dataset is first
* instantiated (e.g. if the dataset is a normal filesystem, when it is
* first mounted).
*
* As hinted at above, there is one ZIL per dataset (both the in-memory
* representation, and the on-disk representation). The on-disk format
* consists of 3 parts:
*
* - a single, per-dataset, ZIL header; which points to a chain of
* - zero or more ZIL blocks; each of which contains
* - zero or more ZIL records
*
* A ZIL record holds the information necessary to replay a single
* system call transaction. A ZIL block can hold many ZIL records, and
* the blocks are chained together, similarly to a singly linked list.
*
* Each ZIL block contains a block pointer (blkptr_t) to the next ZIL
* block in the chain, and the ZIL header points to the first block in
* the chain.
*
* Note, there is not a fixed place in the pool to hold these ZIL
* blocks; they are dynamically allocated and freed as needed from the
* blocks available on the pool, though they can be preferentially
* allocated from a dedicated "log" vdev.
*/
/*
* This controls the amount of time that a ZIL block (lwb) will remain
* "open" when it isn't "full", and it has a thread waiting for it to be
* committed to stable storage. Please refer to the zil_commit_waiter()
* function (and the comments within it) for more details.
*/
-int zfs_commit_timeout_pct = 5;
+static int zfs_commit_timeout_pct = 5;
/*
* See zil.h for more information about these fields.
*/
-zil_stats_t zil_stats = {
+static zil_stats_t zil_stats = {
{ "zil_commit_count", KSTAT_DATA_UINT64 },
{ "zil_commit_writer_count", KSTAT_DATA_UINT64 },
{ "zil_itx_count", KSTAT_DATA_UINT64 },
{ "zil_itx_indirect_count", KSTAT_DATA_UINT64 },
{ "zil_itx_indirect_bytes", KSTAT_DATA_UINT64 },
{ "zil_itx_copied_count", KSTAT_DATA_UINT64 },
{ "zil_itx_copied_bytes", KSTAT_DATA_UINT64 },
{ "zil_itx_needcopy_count", KSTAT_DATA_UINT64 },
{ "zil_itx_needcopy_bytes", KSTAT_DATA_UINT64 },
{ "zil_itx_metaslab_normal_count", KSTAT_DATA_UINT64 },
{ "zil_itx_metaslab_normal_bytes", KSTAT_DATA_UINT64 },
{ "zil_itx_metaslab_slog_count", KSTAT_DATA_UINT64 },
{ "zil_itx_metaslab_slog_bytes", KSTAT_DATA_UINT64 },
};
static kstat_t *zil_ksp;
/*
* Disable intent logging replay. This global ZIL switch affects all pools.
*/
int zil_replay_disable = 0;
/*
* Disable the DKIOCFLUSHWRITECACHE commands that are normally sent to
* the disk(s) by the ZIL after an LWB write has completed. Setting this
* will cause ZIL corruption on power loss if a volatile out-of-order
* write cache is enabled.
*/
-int zil_nocacheflush = 0;
+static int zil_nocacheflush = 0;
/*
* Limit SLOG write size per commit executed with synchronous priority.
* Any writes above that will be executed with lower (asynchronous) priority
* to limit potential SLOG device abuse by single active ZIL writer.
*/
-unsigned long zil_slog_bulk = 768 * 1024;
+static unsigned long zil_slog_bulk = 768 * 1024;
static kmem_cache_t *zil_lwb_cache;
static kmem_cache_t *zil_zcw_cache;
#define LWB_EMPTY(lwb) ((BP_GET_LSIZE(&lwb->lwb_blk) - \
sizeof (zil_chain_t)) == (lwb->lwb_sz - lwb->lwb_nused))
static int
zil_bp_compare(const void *x1, const void *x2)
{
const dva_t *dva1 = &((zil_bp_node_t *)x1)->zn_dva;
const dva_t *dva2 = &((zil_bp_node_t *)x2)->zn_dva;
int cmp = TREE_CMP(DVA_GET_VDEV(dva1), DVA_GET_VDEV(dva2));
if (likely(cmp))
return (cmp);
return (TREE_CMP(DVA_GET_OFFSET(dva1), DVA_GET_OFFSET(dva2)));
}
static void
zil_bp_tree_init(zilog_t *zilog)
{
avl_create(&zilog->zl_bp_tree, zil_bp_compare,
sizeof (zil_bp_node_t), offsetof(zil_bp_node_t, zn_node));
}
static void
zil_bp_tree_fini(zilog_t *zilog)
{
avl_tree_t *t = &zilog->zl_bp_tree;
zil_bp_node_t *zn;
void *cookie = NULL;
while ((zn = avl_destroy_nodes(t, &cookie)) != NULL)
kmem_free(zn, sizeof (zil_bp_node_t));
avl_destroy(t);
}
int
zil_bp_tree_add(zilog_t *zilog, const blkptr_t *bp)
{
avl_tree_t *t = &zilog->zl_bp_tree;
const dva_t *dva;
zil_bp_node_t *zn;
avl_index_t where;
if (BP_IS_EMBEDDED(bp))
return (0);
dva = BP_IDENTITY(bp);
if (avl_find(t, dva, &where) != NULL)
return (SET_ERROR(EEXIST));
zn = kmem_alloc(sizeof (zil_bp_node_t), KM_SLEEP);
zn->zn_dva = *dva;
avl_insert(t, zn, where);
return (0);
}
static zil_header_t *
zil_header_in_syncing_context(zilog_t *zilog)
{
return ((zil_header_t *)zilog->zl_header);
}
static void
zil_init_log_chain(zilog_t *zilog, blkptr_t *bp)
{
zio_cksum_t *zc = &bp->blk_cksum;
(void) random_get_pseudo_bytes((void *)&zc->zc_word[ZIL_ZC_GUID_0],
sizeof (zc->zc_word[ZIL_ZC_GUID_0]));
(void) random_get_pseudo_bytes((void *)&zc->zc_word[ZIL_ZC_GUID_1],
sizeof (zc->zc_word[ZIL_ZC_GUID_1]));
zc->zc_word[ZIL_ZC_OBJSET] = dmu_objset_id(zilog->zl_os);
zc->zc_word[ZIL_ZC_SEQ] = 1ULL;
}
/*
* Read a log block and make sure it's valid.
*/
static int
zil_read_log_block(zilog_t *zilog, boolean_t decrypt, const blkptr_t *bp,
blkptr_t *nbp, void *dst, char **end)
{
enum zio_flag zio_flags = ZIO_FLAG_CANFAIL;
arc_flags_t aflags = ARC_FLAG_WAIT;
arc_buf_t *abuf = NULL;
zbookmark_phys_t zb;
int error;
if (zilog->zl_header->zh_claim_txg == 0)
zio_flags |= ZIO_FLAG_SPECULATIVE | ZIO_FLAG_SCRUB;
if (!(zilog->zl_header->zh_flags & ZIL_CLAIM_LR_SEQ_VALID))
zio_flags |= ZIO_FLAG_SPECULATIVE;
if (!decrypt)
zio_flags |= ZIO_FLAG_RAW;
SET_BOOKMARK(&zb, bp->blk_cksum.zc_word[ZIL_ZC_OBJSET],
ZB_ZIL_OBJECT, ZB_ZIL_LEVEL, bp->blk_cksum.zc_word[ZIL_ZC_SEQ]);
error = arc_read(NULL, zilog->zl_spa, bp, arc_getbuf_func,
&abuf, ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb);
if (error == 0) {
zio_cksum_t cksum = bp->blk_cksum;
/*
* Validate the checksummed log block.
*
* Sequence numbers should be... sequential. The checksum
* verifier for the next block should be bp's checksum plus 1.
*
* Also check the log chain linkage and size used.
*/
cksum.zc_word[ZIL_ZC_SEQ]++;
if (BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_ZILOG2) {
zil_chain_t *zilc = abuf->b_data;
char *lr = (char *)(zilc + 1);
uint64_t len = zilc->zc_nused - sizeof (zil_chain_t);
if (bcmp(&cksum, &zilc->zc_next_blk.blk_cksum,
sizeof (cksum)) || BP_IS_HOLE(&zilc->zc_next_blk)) {
error = SET_ERROR(ECKSUM);
} else {
ASSERT3U(len, <=, SPA_OLD_MAXBLOCKSIZE);
bcopy(lr, dst, len);
*end = (char *)dst + len;
*nbp = zilc->zc_next_blk;
}
} else {
char *lr = abuf->b_data;
uint64_t size = BP_GET_LSIZE(bp);
zil_chain_t *zilc = (zil_chain_t *)(lr + size) - 1;
if (bcmp(&cksum, &zilc->zc_next_blk.blk_cksum,
sizeof (cksum)) || BP_IS_HOLE(&zilc->zc_next_blk) ||
(zilc->zc_nused > (size - sizeof (*zilc)))) {
error = SET_ERROR(ECKSUM);
} else {
ASSERT3U(zilc->zc_nused, <=,
SPA_OLD_MAXBLOCKSIZE);
bcopy(lr, dst, zilc->zc_nused);
*end = (char *)dst + zilc->zc_nused;
*nbp = zilc->zc_next_blk;
}
}
arc_buf_destroy(abuf, &abuf);
}
return (error);
}
/*
* Read a TX_WRITE log data block.
*/
static int
zil_read_log_data(zilog_t *zilog, const lr_write_t *lr, void *wbuf)
{
enum zio_flag zio_flags = ZIO_FLAG_CANFAIL;
const blkptr_t *bp = &lr->lr_blkptr;
arc_flags_t aflags = ARC_FLAG_WAIT;
arc_buf_t *abuf = NULL;
zbookmark_phys_t zb;
int error;
if (BP_IS_HOLE(bp)) {
if (wbuf != NULL)
bzero(wbuf, MAX(BP_GET_LSIZE(bp), lr->lr_length));
return (0);
}
if (zilog->zl_header->zh_claim_txg == 0)
zio_flags |= ZIO_FLAG_SPECULATIVE | ZIO_FLAG_SCRUB;
/*
* If we are not using the resulting data, we are just checking that
* it hasn't been corrupted so we don't need to waste CPU time
* decompressing and decrypting it.
*/
if (wbuf == NULL)
zio_flags |= ZIO_FLAG_RAW;
SET_BOOKMARK(&zb, dmu_objset_id(zilog->zl_os), lr->lr_foid,
ZB_ZIL_LEVEL, lr->lr_offset / BP_GET_LSIZE(bp));
error = arc_read(NULL, zilog->zl_spa, bp, arc_getbuf_func, &abuf,
ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb);
if (error == 0) {
if (wbuf != NULL)
bcopy(abuf->b_data, wbuf, arc_buf_size(abuf));
arc_buf_destroy(abuf, &abuf);
}
return (error);
}
/*
* Parse the intent log, and call parse_func for each valid record within.
*/
int
zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func,
zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg,
boolean_t decrypt)
{
const zil_header_t *zh = zilog->zl_header;
boolean_t claimed = !!zh->zh_claim_txg;
uint64_t claim_blk_seq = claimed ? zh->zh_claim_blk_seq : UINT64_MAX;
uint64_t claim_lr_seq = claimed ? zh->zh_claim_lr_seq : UINT64_MAX;
uint64_t max_blk_seq = 0;
uint64_t max_lr_seq = 0;
uint64_t blk_count = 0;
uint64_t lr_count = 0;
blkptr_t blk, next_blk;
char *lrbuf, *lrp;
int error = 0;
bzero(&next_blk, sizeof (blkptr_t));
/*
* Old logs didn't record the maximum zh_claim_lr_seq.
*/
if (!(zh->zh_flags & ZIL_CLAIM_LR_SEQ_VALID))
claim_lr_seq = UINT64_MAX;
/*
* Starting at the block pointed to by zh_log we read the log chain.
* For each block in the chain we strongly check that block to
* ensure its validity. We stop when an invalid block is found.
* For each block pointer in the chain we call parse_blk_func().
* For each record in each valid block we call parse_lr_func().
* If the log has been claimed, stop if we encounter a sequence
* number greater than the highest claimed sequence number.
*/
lrbuf = zio_buf_alloc(SPA_OLD_MAXBLOCKSIZE);
zil_bp_tree_init(zilog);
for (blk = zh->zh_log; !BP_IS_HOLE(&blk); blk = next_blk) {
uint64_t blk_seq = blk.blk_cksum.zc_word[ZIL_ZC_SEQ];
int reclen;
char *end = NULL;
if (blk_seq > claim_blk_seq)
break;
error = parse_blk_func(zilog, &blk, arg, txg);
if (error != 0)
break;
ASSERT3U(max_blk_seq, <, blk_seq);
max_blk_seq = blk_seq;
blk_count++;
if (max_lr_seq == claim_lr_seq && max_blk_seq == claim_blk_seq)
break;
error = zil_read_log_block(zilog, decrypt, &blk, &next_blk,
lrbuf, &end);
if (error != 0)
break;
for (lrp = lrbuf; lrp < end; lrp += reclen) {
lr_t *lr = (lr_t *)lrp;
reclen = lr->lrc_reclen;
ASSERT3U(reclen, >=, sizeof (lr_t));
if (lr->lrc_seq > claim_lr_seq)
goto done;
error = parse_lr_func(zilog, lr, arg, txg);
if (error != 0)
goto done;
ASSERT3U(max_lr_seq, <, lr->lrc_seq);
max_lr_seq = lr->lrc_seq;
lr_count++;
}
}
done:
zilog->zl_parse_error = error;
zilog->zl_parse_blk_seq = max_blk_seq;
zilog->zl_parse_lr_seq = max_lr_seq;
zilog->zl_parse_blk_count = blk_count;
zilog->zl_parse_lr_count = lr_count;
ASSERT(!claimed || !(zh->zh_flags & ZIL_CLAIM_LR_SEQ_VALID) ||
(max_blk_seq == claim_blk_seq && max_lr_seq == claim_lr_seq) ||
(decrypt && error == EIO));
zil_bp_tree_fini(zilog);
zio_buf_free(lrbuf, SPA_OLD_MAXBLOCKSIZE);
return (error);
}
static int
zil_clear_log_block(zilog_t *zilog, const blkptr_t *bp, void *tx,
uint64_t first_txg)
{
(void) tx;
ASSERT(!BP_IS_HOLE(bp));
/*
* As we call this function from the context of a rewind to a
* checkpoint, each ZIL block whose txg is later than the txg
* that we rewind to is invalid. Thus, we return -1 so
* zil_parse() doesn't attempt to read it.
*/
if (bp->blk_birth >= first_txg)
return (-1);
if (zil_bp_tree_add(zilog, bp) != 0)
return (0);
zio_free(zilog->zl_spa, first_txg, bp);
return (0);
}
static int
zil_noop_log_record(zilog_t *zilog, const lr_t *lrc, void *tx,
uint64_t first_txg)
{
(void) zilog, (void) lrc, (void) tx, (void) first_txg;
return (0);
}
static int
zil_claim_log_block(zilog_t *zilog, const blkptr_t *bp, void *tx,
uint64_t first_txg)
{
/*
* Claim log block if not already committed and not already claimed.
* If tx == NULL, just verify that the block is claimable.
*/
if (BP_IS_HOLE(bp) || bp->blk_birth < first_txg ||
zil_bp_tree_add(zilog, bp) != 0)
return (0);
return (zio_wait(zio_claim(NULL, zilog->zl_spa,
tx == NULL ? 0 : first_txg, bp, spa_claim_notify, NULL,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_SCRUB)));
}
static int
zil_claim_log_record(zilog_t *zilog, const lr_t *lrc, void *tx,
uint64_t first_txg)
{
lr_write_t *lr = (lr_write_t *)lrc;
int error;
if (lrc->lrc_txtype != TX_WRITE)
return (0);
/*
* If the block is not readable, don't claim it. This can happen
* in normal operation when a log block is written to disk before
* some of the dmu_sync() blocks it points to. In this case, the
* transaction cannot have been committed to anyone (we would have
* waited for all writes to be stable first), so it is semantically
* correct to declare this the end of the log.
*/
if (lr->lr_blkptr.blk_birth >= first_txg) {
error = zil_read_log_data(zilog, lr, NULL);
if (error != 0)
return (error);
}
return (zil_claim_log_block(zilog, &lr->lr_blkptr, tx, first_txg));
}
static int
zil_free_log_block(zilog_t *zilog, const blkptr_t *bp, void *tx,
uint64_t claim_txg)
{
(void) claim_txg;
zio_free(zilog->zl_spa, dmu_tx_get_txg(tx), bp);
return (0);
}
static int
zil_free_log_record(zilog_t *zilog, const lr_t *lrc, void *tx,
uint64_t claim_txg)
{
lr_write_t *lr = (lr_write_t *)lrc;
blkptr_t *bp = &lr->lr_blkptr;
/*
* If we previously claimed it, we need to free it.
*/
if (claim_txg != 0 && lrc->lrc_txtype == TX_WRITE &&
bp->blk_birth >= claim_txg && zil_bp_tree_add(zilog, bp) == 0 &&
!BP_IS_HOLE(bp))
zio_free(zilog->zl_spa, dmu_tx_get_txg(tx), bp);
return (0);
}
static int
zil_lwb_vdev_compare(const void *x1, const void *x2)
{
const uint64_t v1 = ((zil_vdev_node_t *)x1)->zv_vdev;
const uint64_t v2 = ((zil_vdev_node_t *)x2)->zv_vdev;
return (TREE_CMP(v1, v2));
}
static lwb_t *
zil_alloc_lwb(zilog_t *zilog, blkptr_t *bp, boolean_t slog, uint64_t txg,
boolean_t fastwrite)
{
lwb_t *lwb;
lwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP);
lwb->lwb_zilog = zilog;
lwb->lwb_blk = *bp;
lwb->lwb_fastwrite = fastwrite;
lwb->lwb_slog = slog;
lwb->lwb_state = LWB_STATE_CLOSED;
lwb->lwb_buf = zio_buf_alloc(BP_GET_LSIZE(bp));
lwb->lwb_max_txg = txg;
lwb->lwb_write_zio = NULL;
lwb->lwb_root_zio = NULL;
lwb->lwb_tx = NULL;
lwb->lwb_issued_timestamp = 0;
if (BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_ZILOG2) {
lwb->lwb_nused = sizeof (zil_chain_t);
lwb->lwb_sz = BP_GET_LSIZE(bp);
} else {
lwb->lwb_nused = 0;
lwb->lwb_sz = BP_GET_LSIZE(bp) - sizeof (zil_chain_t);
}
mutex_enter(&zilog->zl_lock);
list_insert_tail(&zilog->zl_lwb_list, lwb);
mutex_exit(&zilog->zl_lock);
ASSERT(!MUTEX_HELD(&lwb->lwb_vdev_lock));
ASSERT(avl_is_empty(&lwb->lwb_vdev_tree));
VERIFY(list_is_empty(&lwb->lwb_waiters));
VERIFY(list_is_empty(&lwb->lwb_itxs));
return (lwb);
}
static void
zil_free_lwb(zilog_t *zilog, lwb_t *lwb)
{
ASSERT(MUTEX_HELD(&zilog->zl_lock));
ASSERT(!MUTEX_HELD(&lwb->lwb_vdev_lock));
VERIFY(list_is_empty(&lwb->lwb_waiters));
VERIFY(list_is_empty(&lwb->lwb_itxs));
ASSERT(avl_is_empty(&lwb->lwb_vdev_tree));
ASSERT3P(lwb->lwb_write_zio, ==, NULL);
ASSERT3P(lwb->lwb_root_zio, ==, NULL);
ASSERT3U(lwb->lwb_max_txg, <=, spa_syncing_txg(zilog->zl_spa));
ASSERT(lwb->lwb_state == LWB_STATE_CLOSED ||
lwb->lwb_state == LWB_STATE_FLUSH_DONE);
/*
* Clear the zilog's field to indicate this lwb is no longer
* valid, and prevent use-after-free errors.
*/
if (zilog->zl_last_lwb_opened == lwb)
zilog->zl_last_lwb_opened = NULL;
kmem_cache_free(zil_lwb_cache, lwb);
}
/*
* Called when we create in-memory log transactions so that we know
* to cleanup the itxs at the end of spa_sync().
*/
static void
zilog_dirty(zilog_t *zilog, uint64_t txg)
{
dsl_pool_t *dp = zilog->zl_dmu_pool;
dsl_dataset_t *ds = dmu_objset_ds(zilog->zl_os);
ASSERT(spa_writeable(zilog->zl_spa));
if (ds->ds_is_snapshot)
panic("dirtying snapshot!");
if (txg_list_add(&dp->dp_dirty_zilogs, zilog, txg)) {
/* up the hold count until we can be written out */
dmu_buf_add_ref(ds->ds_dbuf, zilog);
zilog->zl_dirty_max_txg = MAX(txg, zilog->zl_dirty_max_txg);
}
}
/*
* Determine if the zil is dirty in the specified txg. Callers wanting to
* ensure that the dirty state does not change must hold the itxg_lock for
* the specified txg. Holding the lock will ensure that the zil cannot be
* dirtied (zil_itx_assign) or cleaned (zil_clean) while we check its current
* state.
*/
static boolean_t __maybe_unused
zilog_is_dirty_in_txg(zilog_t *zilog, uint64_t txg)
{
dsl_pool_t *dp = zilog->zl_dmu_pool;
if (txg_list_member(&dp->dp_dirty_zilogs, zilog, txg & TXG_MASK))
return (B_TRUE);
return (B_FALSE);
}
/*
* Determine if the zil is dirty. The zil is considered dirty if it has
* any pending itx records that have not been cleaned by zil_clean().
*/
static boolean_t
zilog_is_dirty(zilog_t *zilog)
{
dsl_pool_t *dp = zilog->zl_dmu_pool;
for (int t = 0; t < TXG_SIZE; t++) {
if (txg_list_member(&dp->dp_dirty_zilogs, zilog, t))
return (B_TRUE);
}
return (B_FALSE);
}
/*
* Create an on-disk intent log.
*/
static lwb_t *
zil_create(zilog_t *zilog)
{
const zil_header_t *zh = zilog->zl_header;
lwb_t *lwb = NULL;
uint64_t txg = 0;
dmu_tx_t *tx = NULL;
blkptr_t blk;
int error = 0;
boolean_t fastwrite = FALSE;
boolean_t slog = FALSE;
/*
* Wait for any previous destroy to complete.
*/
txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);
ASSERT(zh->zh_claim_txg == 0);
ASSERT(zh->zh_replay_seq == 0);
blk = zh->zh_log;
/*
* Allocate an initial log block if:
* - there isn't one already
* - the existing block is the wrong endianness
*/
if (BP_IS_HOLE(&blk) || BP_SHOULD_BYTESWAP(&blk)) {
tx = dmu_tx_create(zilog->zl_os);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
txg = dmu_tx_get_txg(tx);
if (!BP_IS_HOLE(&blk)) {
zio_free(zilog->zl_spa, txg, &blk);
BP_ZERO(&blk);
}
error = zio_alloc_zil(zilog->zl_spa, zilog->zl_os, txg, &blk,
ZIL_MIN_BLKSZ, &slog);
fastwrite = TRUE;
if (error == 0)
zil_init_log_chain(zilog, &blk);
}
/*
* Allocate a log write block (lwb) for the first log block.
*/
if (error == 0)
lwb = zil_alloc_lwb(zilog, &blk, slog, txg, fastwrite);
/*
* If we just allocated the first log block, commit our transaction
* and wait for zil_sync() to stuff the block pointer into zh_log.
* (zh is part of the MOS, so we cannot modify it in open context.)
*/
if (tx != NULL) {
dmu_tx_commit(tx);
txg_wait_synced(zilog->zl_dmu_pool, txg);
}
ASSERT(error != 0 || bcmp(&blk, &zh->zh_log, sizeof (blk)) == 0);
IMPLY(error == 0, lwb != NULL);
return (lwb);
}
/*
* In one tx, free all log blocks and clear the log header. If keep_first
* is set, then we're replaying a log with no content. We want to keep the
* first block, however, so that the first synchronous transaction doesn't
* require a txg_wait_synced() in zil_create(). We don't need to
* txg_wait_synced() here either when keep_first is set, because both
* zil_create() and zil_destroy() will wait for any in-progress destroys
* to complete.
*/
void
zil_destroy(zilog_t *zilog, boolean_t keep_first)
{
const zil_header_t *zh = zilog->zl_header;
lwb_t *lwb;
dmu_tx_t *tx;
uint64_t txg;
/*
* Wait for any previous destroy to complete.
*/
txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);
zilog->zl_old_header = *zh; /* debugging aid */
if (BP_IS_HOLE(&zh->zh_log))
return;
tx = dmu_tx_create(zilog->zl_os);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
txg = dmu_tx_get_txg(tx);
mutex_enter(&zilog->zl_lock);
ASSERT3U(zilog->zl_destroy_txg, <, txg);
zilog->zl_destroy_txg = txg;
zilog->zl_keep_first = keep_first;
if (!list_is_empty(&zilog->zl_lwb_list)) {
ASSERT(zh->zh_claim_txg == 0);
VERIFY(!keep_first);
while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) {
if (lwb->lwb_fastwrite)
metaslab_fastwrite_unmark(zilog->zl_spa,
&lwb->lwb_blk);
list_remove(&zilog->zl_lwb_list, lwb);
if (lwb->lwb_buf != NULL)
zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
zio_free(zilog->zl_spa, txg, &lwb->lwb_blk);
zil_free_lwb(zilog, lwb);
}
} else if (!keep_first) {
zil_destroy_sync(zilog, tx);
}
mutex_exit(&zilog->zl_lock);
dmu_tx_commit(tx);
}
void
zil_destroy_sync(zilog_t *zilog, dmu_tx_t *tx)
{
ASSERT(list_is_empty(&zilog->zl_lwb_list));
(void) zil_parse(zilog, zil_free_log_block,
zil_free_log_record, tx, zilog->zl_header->zh_claim_txg, B_FALSE);
}
int
zil_claim(dsl_pool_t *dp, dsl_dataset_t *ds, void *txarg)
{
dmu_tx_t *tx = txarg;
zilog_t *zilog;
uint64_t first_txg;
zil_header_t *zh;
objset_t *os;
int error;
error = dmu_objset_own_obj(dp, ds->ds_object,
DMU_OST_ANY, B_FALSE, B_FALSE, FTAG, &os);
if (error != 0) {
/*
* EBUSY indicates that the objset is inconsistent, in which
* case it can not have a ZIL.
*/
if (error != EBUSY) {
cmn_err(CE_WARN, "can't open objset for %llu, error %u",
(unsigned long long)ds->ds_object, error);
}
return (0);
}
zilog = dmu_objset_zil(os);
zh = zil_header_in_syncing_context(zilog);
ASSERT3U(tx->tx_txg, ==, spa_first_txg(zilog->zl_spa));
first_txg = spa_min_claim_txg(zilog->zl_spa);
/*
* If the spa_log_state is not set to be cleared, check whether
* the current uberblock is a checkpoint one and if the current
* header has been claimed before moving on.
*
* If the current uberblock is a checkpointed uberblock then
* one of the following scenarios took place:
*
* 1] We are currently rewinding to the checkpoint of the pool.
* 2] We crashed in the middle of a checkpoint rewind but we
* did manage to write the checkpointed uberblock to the
* vdev labels, so when we tried to import the pool again
* the checkpointed uberblock was selected from the import
* procedure.
*
* In both cases we want to zero out all the ZIL blocks, except
* the ones that have been claimed at the time of the checkpoint
* (their zh_claim_txg != 0). The reason is that these blocks
* may be corrupted since we may have reused their locations on
* disk after we took the checkpoint.
*
* We could try to set spa_log_state to SPA_LOG_CLEAR earlier
* when we first figure out whether the current uberblock is
* checkpointed or not. Unfortunately, that would discard all
* the logs, including the ones that are claimed, and we would
* leak space.
*/
if (spa_get_log_state(zilog->zl_spa) == SPA_LOG_CLEAR ||
(zilog->zl_spa->spa_uberblock.ub_checkpoint_txg != 0 &&
zh->zh_claim_txg == 0)) {
if (!BP_IS_HOLE(&zh->zh_log)) {
(void) zil_parse(zilog, zil_clear_log_block,
zil_noop_log_record, tx, first_txg, B_FALSE);
}
BP_ZERO(&zh->zh_log);
if (os->os_encrypted)
os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_TRUE;
dsl_dataset_dirty(dmu_objset_ds(os), tx);
dmu_objset_disown(os, B_FALSE, FTAG);
return (0);
}
/*
* If we are not rewinding and opening the pool normally, then
* the min_claim_txg should be equal to the first txg of the pool.
*/
ASSERT3U(first_txg, ==, spa_first_txg(zilog->zl_spa));
/*
* Claim all log blocks if we haven't already done so, and remember
* the highest claimed sequence number. This ensures that if we can
* read only part of the log now (e.g. due to a missing device),
* but we can read the entire log later, we will not try to replay
* or destroy beyond the last block we successfully claimed.
*/
ASSERT3U(zh->zh_claim_txg, <=, first_txg);
if (zh->zh_claim_txg == 0 && !BP_IS_HOLE(&zh->zh_log)) {
(void) zil_parse(zilog, zil_claim_log_block,
zil_claim_log_record, tx, first_txg, B_FALSE);
zh->zh_claim_txg = first_txg;
zh->zh_claim_blk_seq = zilog->zl_parse_blk_seq;
zh->zh_claim_lr_seq = zilog->zl_parse_lr_seq;
if (zilog->zl_parse_lr_count || zilog->zl_parse_blk_count > 1)
zh->zh_flags |= ZIL_REPLAY_NEEDED;
zh->zh_flags |= ZIL_CLAIM_LR_SEQ_VALID;
if (os->os_encrypted)
os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_TRUE;
dsl_dataset_dirty(dmu_objset_ds(os), tx);
}
ASSERT3U(first_txg, ==, (spa_last_synced_txg(zilog->zl_spa) + 1));
dmu_objset_disown(os, B_FALSE, FTAG);
return (0);
}
/*
* Check the log by walking the log chain.
* Checksum errors are ok as they indicate the end of the chain.
* Any other error (no device or read failure) returns an error.
*/
int
zil_check_log_chain(dsl_pool_t *dp, dsl_dataset_t *ds, void *tx)
{
(void) dp;
zilog_t *zilog;
objset_t *os;
blkptr_t *bp;
int error;
ASSERT(tx == NULL);
error = dmu_objset_from_ds(ds, &os);
if (error != 0) {
cmn_err(CE_WARN, "can't open objset %llu, error %d",
(unsigned long long)ds->ds_object, error);
return (0);
}
zilog = dmu_objset_zil(os);
bp = (blkptr_t *)&zilog->zl_header->zh_log;
if (!BP_IS_HOLE(bp)) {
vdev_t *vd;
boolean_t valid = B_TRUE;
/*
* Check the first block and determine if it's on a log device
* which may have been removed or faulted prior to loading this
* pool. If so, there's no point in checking the rest of the
* log as its content should have already been synced to the
* pool.
*/
spa_config_enter(os->os_spa, SCL_STATE, FTAG, RW_READER);
vd = vdev_lookup_top(os->os_spa, DVA_GET_VDEV(&bp->blk_dva[0]));
if (vd->vdev_islog && vdev_is_dead(vd))
valid = vdev_log_state_valid(vd);
spa_config_exit(os->os_spa, SCL_STATE, FTAG);
if (!valid)
return (0);
/*
* Check whether the current uberblock is checkpointed (e.g.
* we are rewinding) and whether the current header has been
* claimed or not. If it hasn't then skip verifying it. We
* do this because its ZIL blocks may be part of the pool's
* state before the rewind, which is no longer valid.
*/
zil_header_t *zh = zil_header_in_syncing_context(zilog);
if (zilog->zl_spa->spa_uberblock.ub_checkpoint_txg != 0 &&
zh->zh_claim_txg == 0)
return (0);
}
/*
* Because tx == NULL, zil_claim_log_block() will not actually claim
* any blocks, but just determine whether it is possible to do so.
* In addition to checking the log chain, zil_claim_log_block()
* will invoke zio_claim() with a done func of spa_claim_notify(),
* which will update spa_max_claim_txg. See spa_load() for details.
*/
error = zil_parse(zilog, zil_claim_log_block, zil_claim_log_record, tx,
zilog->zl_header->zh_claim_txg ? -1ULL :
spa_min_claim_txg(os->os_spa), B_FALSE);
return ((error == ECKSUM || error == ENOENT) ? 0 : error);
}
/*
* When an itx is "skipped", this function is used to properly mark the
* waiter as "done, and signal any thread(s) waiting on it. An itx can
* be skipped (and not committed to an lwb) for a variety of reasons,
* one of them being that the itx was committed via spa_sync(), prior to
* it being committed to an lwb; this can happen if a thread calling
* zil_commit() is racing with spa_sync().
*/
static void
zil_commit_waiter_skip(zil_commit_waiter_t *zcw)
{
mutex_enter(&zcw->zcw_lock);
ASSERT3B(zcw->zcw_done, ==, B_FALSE);
zcw->zcw_done = B_TRUE;
cv_broadcast(&zcw->zcw_cv);
mutex_exit(&zcw->zcw_lock);
}
/*
* This function is used when the given waiter is to be linked into an
* lwb's "lwb_waiter" list; i.e. when the itx is committed to the lwb.
* At this point, the waiter will no longer be referenced by the itx,
* and instead, will be referenced by the lwb.
*/
static void
zil_commit_waiter_link_lwb(zil_commit_waiter_t *zcw, lwb_t *lwb)
{
/*
* The lwb_waiters field of the lwb is protected by the zilog's
* zl_lock, thus it must be held when calling this function.
*/
ASSERT(MUTEX_HELD(&lwb->lwb_zilog->zl_lock));
mutex_enter(&zcw->zcw_lock);
ASSERT(!list_link_active(&zcw->zcw_node));
ASSERT3P(zcw->zcw_lwb, ==, NULL);
ASSERT3P(lwb, !=, NULL);
ASSERT(lwb->lwb_state == LWB_STATE_OPENED ||
lwb->lwb_state == LWB_STATE_ISSUED ||
lwb->lwb_state == LWB_STATE_WRITE_DONE);
list_insert_tail(&lwb->lwb_waiters, zcw);
zcw->zcw_lwb = lwb;
mutex_exit(&zcw->zcw_lock);
}
/*
* This function is used when zio_alloc_zil() fails to allocate a ZIL
* block, and the given waiter must be linked to the "nolwb waiters"
* list inside of zil_process_commit_list().
*/
static void
zil_commit_waiter_link_nolwb(zil_commit_waiter_t *zcw, list_t *nolwb)
{
mutex_enter(&zcw->zcw_lock);
ASSERT(!list_link_active(&zcw->zcw_node));
ASSERT3P(zcw->zcw_lwb, ==, NULL);
list_insert_tail(nolwb, zcw);
mutex_exit(&zcw->zcw_lock);
}
void
zil_lwb_add_block(lwb_t *lwb, const blkptr_t *bp)
{
avl_tree_t *t = &lwb->lwb_vdev_tree;
avl_index_t where;
zil_vdev_node_t *zv, zvsearch;
int ndvas = BP_GET_NDVAS(bp);
int i;
if (zil_nocacheflush)
return;
mutex_enter(&lwb->lwb_vdev_lock);
for (i = 0; i < ndvas; i++) {
zvsearch.zv_vdev = DVA_GET_VDEV(&bp->blk_dva[i]);
if (avl_find(t, &zvsearch, &where) == NULL) {
zv = kmem_alloc(sizeof (*zv), KM_SLEEP);
zv->zv_vdev = zvsearch.zv_vdev;
avl_insert(t, zv, where);
}
}
mutex_exit(&lwb->lwb_vdev_lock);
}
static void
zil_lwb_flush_defer(lwb_t *lwb, lwb_t *nlwb)
{
avl_tree_t *src = &lwb->lwb_vdev_tree;
avl_tree_t *dst = &nlwb->lwb_vdev_tree;
void *cookie = NULL;
zil_vdev_node_t *zv;
ASSERT3S(lwb->lwb_state, ==, LWB_STATE_WRITE_DONE);
ASSERT3S(nlwb->lwb_state, !=, LWB_STATE_WRITE_DONE);
ASSERT3S(nlwb->lwb_state, !=, LWB_STATE_FLUSH_DONE);
/*
* While 'lwb' is at a point in its lifetime where lwb_vdev_tree does
* not need the protection of lwb_vdev_lock (it will only be modified
* while holding zilog->zl_lock) as its writes and those of its
* children have all completed. The younger 'nlwb' may be waiting on
* future writes to additional vdevs.
*/
mutex_enter(&nlwb->lwb_vdev_lock);
/*
* Tear down the 'lwb' vdev tree, ensuring that entries which do not
* exist in 'nlwb' are moved to it, freeing any would-be duplicates.
*/
while ((zv = avl_destroy_nodes(src, &cookie)) != NULL) {
avl_index_t where;
if (avl_find(dst, zv, &where) == NULL) {
avl_insert(dst, zv, where);
} else {
kmem_free(zv, sizeof (*zv));
}
}
mutex_exit(&nlwb->lwb_vdev_lock);
}
void
zil_lwb_add_txg(lwb_t *lwb, uint64_t txg)
{
lwb->lwb_max_txg = MAX(lwb->lwb_max_txg, txg);
}
/*
* This function is a called after all vdevs associated with a given lwb
* write have completed their DKIOCFLUSHWRITECACHE command; or as soon
* as the lwb write completes, if "zil_nocacheflush" is set. Further,
* all "previous" lwb's will have completed before this function is
* called; i.e. this function is called for all previous lwbs before
* it's called for "this" lwb (enforced via zio the dependencies
* configured in zil_lwb_set_zio_dependency()).
*
* The intention is for this function to be called as soon as the
* contents of an lwb are considered "stable" on disk, and will survive
* any sudden loss of power. At this point, any threads waiting for the
* lwb to reach this state are signalled, and the "waiter" structures
* are marked "done".
*/
static void
zil_lwb_flush_vdevs_done(zio_t *zio)
{
lwb_t *lwb = zio->io_private;
zilog_t *zilog = lwb->lwb_zilog;
dmu_tx_t *tx = lwb->lwb_tx;
zil_commit_waiter_t *zcw;
itx_t *itx;
spa_config_exit(zilog->zl_spa, SCL_STATE, lwb);
zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
mutex_enter(&zilog->zl_lock);
/*
* Ensure the lwb buffer pointer is cleared before releasing the
* txg. If we have had an allocation failure and the txg is
* waiting to sync then we want zil_sync() to remove the lwb so
* that it's not picked up as the next new one in
* zil_process_commit_list(). zil_sync() will only remove the
* lwb if lwb_buf is null.
*/
lwb->lwb_buf = NULL;
lwb->lwb_tx = NULL;
ASSERT3U(lwb->lwb_issued_timestamp, >, 0);
zilog->zl_last_lwb_latency = gethrtime() - lwb->lwb_issued_timestamp;
lwb->lwb_root_zio = NULL;
ASSERT3S(lwb->lwb_state, ==, LWB_STATE_WRITE_DONE);
lwb->lwb_state = LWB_STATE_FLUSH_DONE;
if (zilog->zl_last_lwb_opened == lwb) {
/*
* Remember the highest committed log sequence number
* for ztest. We only update this value when all the log
* writes succeeded, because ztest wants to ASSERT that
* it got the whole log chain.
*/
zilog->zl_commit_lr_seq = zilog->zl_lr_seq;
}
while ((itx = list_head(&lwb->lwb_itxs)) != NULL) {
list_remove(&lwb->lwb_itxs, itx);
zil_itx_destroy(itx);
}
while ((zcw = list_head(&lwb->lwb_waiters)) != NULL) {
mutex_enter(&zcw->zcw_lock);
ASSERT(list_link_active(&zcw->zcw_node));
list_remove(&lwb->lwb_waiters, zcw);
ASSERT3P(zcw->zcw_lwb, ==, lwb);
zcw->zcw_lwb = NULL;
/*
* We expect any ZIO errors from child ZIOs to have been
* propagated "up" to this specific LWB's root ZIO, in
* order for this error handling to work correctly. This
* includes ZIO errors from either this LWB's write or
* flush, as well as any errors from other dependent LWBs
* (e.g. a root LWB ZIO that might be a child of this LWB).
*
* With that said, it's important to note that LWB flush
* errors are not propagated up to the LWB root ZIO.
* This is incorrect behavior, and results in VDEV flush
* errors not being handled correctly here. See the
* comment above the call to "zio_flush" for details.
*/
zcw->zcw_zio_error = zio->io_error;
ASSERT3B(zcw->zcw_done, ==, B_FALSE);
zcw->zcw_done = B_TRUE;
cv_broadcast(&zcw->zcw_cv);
mutex_exit(&zcw->zcw_lock);
}
mutex_exit(&zilog->zl_lock);
/*
* Now that we've written this log block, we have a stable pointer
* to the next block in the chain, so it's OK to let the txg in
* which we allocated the next block sync.
*/
dmu_tx_commit(tx);
}
/*
* This is called when an lwb's write zio completes. The callback's
* purpose is to issue the DKIOCFLUSHWRITECACHE commands for the vdevs
* in the lwb's lwb_vdev_tree. The tree will contain the vdevs involved
* in writing out this specific lwb's data, and in the case that cache
* flushes have been deferred, vdevs involved in writing the data for
* previous lwbs. The writes corresponding to all the vdevs in the
* lwb_vdev_tree will have completed by the time this is called, due to
* the zio dependencies configured in zil_lwb_set_zio_dependency(),
* which takes deferred flushes into account. The lwb will be "done"
* once zil_lwb_flush_vdevs_done() is called, which occurs in the zio
* completion callback for the lwb's root zio.
*/
static void
zil_lwb_write_done(zio_t *zio)
{
lwb_t *lwb = zio->io_private;
spa_t *spa = zio->io_spa;
zilog_t *zilog = lwb->lwb_zilog;
avl_tree_t *t = &lwb->lwb_vdev_tree;
void *cookie = NULL;
zil_vdev_node_t *zv;
lwb_t *nlwb;
ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), !=, 0);
ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF);
ASSERT(BP_GET_TYPE(zio->io_bp) == DMU_OT_INTENT_LOG);
ASSERT(BP_GET_LEVEL(zio->io_bp) == 0);
ASSERT(BP_GET_BYTEORDER(zio->io_bp) == ZFS_HOST_BYTEORDER);
ASSERT(!BP_IS_GANG(zio->io_bp));
ASSERT(!BP_IS_HOLE(zio->io_bp));
ASSERT(BP_GET_FILL(zio->io_bp) == 0);
abd_free(zio->io_abd);
mutex_enter(&zilog->zl_lock);
ASSERT3S(lwb->lwb_state, ==, LWB_STATE_ISSUED);
lwb->lwb_state = LWB_STATE_WRITE_DONE;
lwb->lwb_write_zio = NULL;
lwb->lwb_fastwrite = FALSE;
nlwb = list_next(&zilog->zl_lwb_list, lwb);
mutex_exit(&zilog->zl_lock);
if (avl_numnodes(t) == 0)
return;
/*
* If there was an IO error, we're not going to call zio_flush()
* on these vdevs, so we simply empty the tree and free the
* nodes. We avoid calling zio_flush() since there isn't any
* good reason for doing so, after the lwb block failed to be
* written out.
*
* Additionally, we don't perform any further error handling at
* this point (e.g. setting "zcw_zio_error" appropriately), as
* we expect that to occur in "zil_lwb_flush_vdevs_done" (thus,
* we expect any error seen here, to have been propagated to
* that function).
*/
if (zio->io_error != 0) {
while ((zv = avl_destroy_nodes(t, &cookie)) != NULL)
kmem_free(zv, sizeof (*zv));
return;
}
/*
* If this lwb does not have any threads waiting for it to
* complete, we want to defer issuing the DKIOCFLUSHWRITECACHE
* command to the vdevs written to by "this" lwb, and instead
* rely on the "next" lwb to handle the DKIOCFLUSHWRITECACHE
* command for those vdevs. Thus, we merge the vdev tree of
* "this" lwb with the vdev tree of the "next" lwb in the list,
* and assume the "next" lwb will handle flushing the vdevs (or
* deferring the flush(s) again).
*
* This is a useful performance optimization, especially for
* workloads with lots of async write activity and few sync
* write and/or fsync activity, as it has the potential to
* coalesce multiple flush commands to a vdev into one.
*/
if (list_head(&lwb->lwb_waiters) == NULL && nlwb != NULL) {
zil_lwb_flush_defer(lwb, nlwb);
ASSERT(avl_is_empty(&lwb->lwb_vdev_tree));
return;
}
while ((zv = avl_destroy_nodes(t, &cookie)) != NULL) {
vdev_t *vd = vdev_lookup_top(spa, zv->zv_vdev);
if (vd != NULL) {
/*
* The "ZIO_FLAG_DONT_PROPAGATE" is currently
* always used within "zio_flush". This means,
* any errors when flushing the vdev(s), will
* (unfortunately) not be handled correctly,
* since these "zio_flush" errors will not be
* propagated up to "zil_lwb_flush_vdevs_done".
*/
zio_flush(lwb->lwb_root_zio, vd);
}
kmem_free(zv, sizeof (*zv));
}
}
static void
zil_lwb_set_zio_dependency(zilog_t *zilog, lwb_t *lwb)
{
lwb_t *last_lwb_opened = zilog->zl_last_lwb_opened;
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
ASSERT(MUTEX_HELD(&zilog->zl_lock));
/*
* The zilog's "zl_last_lwb_opened" field is used to build the
* lwb/zio dependency chain, which is used to preserve the
* ordering of lwb completions that is required by the semantics
* of the ZIL. Each new lwb zio becomes a parent of the
* "previous" lwb zio, such that the new lwb's zio cannot
* complete until the "previous" lwb's zio completes.
*
* This is required by the semantics of zil_commit(); the commit
* waiters attached to the lwbs will be woken in the lwb zio's
* completion callback, so this zio dependency graph ensures the
* waiters are woken in the correct order (the same order the
* lwbs were created).
*/
if (last_lwb_opened != NULL &&
last_lwb_opened->lwb_state != LWB_STATE_FLUSH_DONE) {
ASSERT(last_lwb_opened->lwb_state == LWB_STATE_OPENED ||
last_lwb_opened->lwb_state == LWB_STATE_ISSUED ||
last_lwb_opened->lwb_state == LWB_STATE_WRITE_DONE);
ASSERT3P(last_lwb_opened->lwb_root_zio, !=, NULL);
zio_add_child(lwb->lwb_root_zio,
last_lwb_opened->lwb_root_zio);
/*
* If the previous lwb's write hasn't already completed,
* we also want to order the completion of the lwb write
* zios (above, we only order the completion of the lwb
* root zios). This is required because of how we can
* defer the DKIOCFLUSHWRITECACHE commands for each lwb.
*
* When the DKIOCFLUSHWRITECACHE commands are deferred,
* the previous lwb will rely on this lwb to flush the
* vdevs written to by that previous lwb. Thus, we need
* to ensure this lwb doesn't issue the flush until
* after the previous lwb's write completes. We ensure
* this ordering by setting the zio parent/child
* relationship here.
*
* Without this relationship on the lwb's write zio,
* it's possible for this lwb's write to complete prior
* to the previous lwb's write completing; and thus, the
* vdevs for the previous lwb would be flushed prior to
* that lwb's data being written to those vdevs (the
* vdevs are flushed in the lwb write zio's completion
* handler, zil_lwb_write_done()).
*/
if (last_lwb_opened->lwb_state != LWB_STATE_WRITE_DONE) {
ASSERT(last_lwb_opened->lwb_state == LWB_STATE_OPENED ||
last_lwb_opened->lwb_state == LWB_STATE_ISSUED);
ASSERT3P(last_lwb_opened->lwb_write_zio, !=, NULL);
zio_add_child(lwb->lwb_write_zio,
last_lwb_opened->lwb_write_zio);
}
}
}
/*
* This function's purpose is to "open" an lwb such that it is ready to
* accept new itxs being committed to it. To do this, the lwb's zio
* structures are created, and linked to the lwb. This function is
* idempotent; if the passed in lwb has already been opened, this
* function is essentially a no-op.
*/
static void
zil_lwb_write_open(zilog_t *zilog, lwb_t *lwb)
{
zbookmark_phys_t zb;
zio_priority_t prio;
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
ASSERT3P(lwb, !=, NULL);
EQUIV(lwb->lwb_root_zio == NULL, lwb->lwb_state == LWB_STATE_CLOSED);
EQUIV(lwb->lwb_root_zio != NULL, lwb->lwb_state == LWB_STATE_OPENED);
SET_BOOKMARK(&zb, lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_OBJSET],
ZB_ZIL_OBJECT, ZB_ZIL_LEVEL,
lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_SEQ]);
/* Lock so zil_sync() doesn't fastwrite_unmark after zio is created */
mutex_enter(&zilog->zl_lock);
if (lwb->lwb_root_zio == NULL) {
abd_t *lwb_abd = abd_get_from_buf(lwb->lwb_buf,
BP_GET_LSIZE(&lwb->lwb_blk));
if (!lwb->lwb_fastwrite) {
metaslab_fastwrite_mark(zilog->zl_spa, &lwb->lwb_blk);
lwb->lwb_fastwrite = 1;
}
if (!lwb->lwb_slog || zilog->zl_cur_used <= zil_slog_bulk)
prio = ZIO_PRIORITY_SYNC_WRITE;
else
prio = ZIO_PRIORITY_ASYNC_WRITE;
lwb->lwb_root_zio = zio_root(zilog->zl_spa,
zil_lwb_flush_vdevs_done, lwb, ZIO_FLAG_CANFAIL);
ASSERT3P(lwb->lwb_root_zio, !=, NULL);
lwb->lwb_write_zio = zio_rewrite(lwb->lwb_root_zio,
zilog->zl_spa, 0, &lwb->lwb_blk, lwb_abd,
BP_GET_LSIZE(&lwb->lwb_blk), zil_lwb_write_done, lwb,
prio, ZIO_FLAG_CANFAIL | ZIO_FLAG_FASTWRITE, &zb);
ASSERT3P(lwb->lwb_write_zio, !=, NULL);
lwb->lwb_state = LWB_STATE_OPENED;
zil_lwb_set_zio_dependency(zilog, lwb);
zilog->zl_last_lwb_opened = lwb;
}
mutex_exit(&zilog->zl_lock);
ASSERT3P(lwb->lwb_root_zio, !=, NULL);
ASSERT3P(lwb->lwb_write_zio, !=, NULL);
ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED);
}
/*
* Define a limited set of intent log block sizes.
*
* These must be a multiple of 4KB. Note only the amount used (again
* aligned to 4KB) actually gets written. However, we can't always just
* allocate SPA_OLD_MAXBLOCKSIZE as the slog space could be exhausted.
*/
-struct {
+static const struct {
uint64_t limit;
uint64_t blksz;
} zil_block_buckets[] = {
{ 4096, 4096 }, /* non TX_WRITE */
{ 8192 + 4096, 8192 + 4096 }, /* database */
{ 32768 + 4096, 32768 + 4096 }, /* NFS writes */
{ 65536 + 4096, 65536 + 4096 }, /* 64KB writes */
{ 131072, 131072 }, /* < 128KB writes */
{ 131072 +4096, 65536 + 4096 }, /* 128KB writes */
{ UINT64_MAX, SPA_OLD_MAXBLOCKSIZE}, /* > 128KB writes */
};
/*
* Maximum block size used by the ZIL. This is picked up when the ZIL is
* initialized. Otherwise this should not be used directly; see
* zl_max_block_size instead.
*/
-int zil_maxblocksize = SPA_OLD_MAXBLOCKSIZE;
+static int zil_maxblocksize = SPA_OLD_MAXBLOCKSIZE;
/*
* Start a log block write and advance to the next log block.
* Calls are serialized.
*/
static lwb_t *
zil_lwb_write_issue(zilog_t *zilog, lwb_t *lwb)
{
lwb_t *nlwb = NULL;
zil_chain_t *zilc;
spa_t *spa = zilog->zl_spa;
blkptr_t *bp;
dmu_tx_t *tx;
uint64_t txg;
uint64_t zil_blksz, wsz;
int i, error;
boolean_t slog;
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
ASSERT3P(lwb->lwb_root_zio, !=, NULL);
ASSERT3P(lwb->lwb_write_zio, !=, NULL);
ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED);
if (BP_GET_CHECKSUM(&lwb->lwb_blk) == ZIO_CHECKSUM_ZILOG2) {
zilc = (zil_chain_t *)lwb->lwb_buf;
bp = &zilc->zc_next_blk;
} else {
zilc = (zil_chain_t *)(lwb->lwb_buf + lwb->lwb_sz);
bp = &zilc->zc_next_blk;
}
ASSERT(lwb->lwb_nused <= lwb->lwb_sz);
/*
* Allocate the next block and save its address in this block
* before writing it in order to establish the log chain.
* Note that if the allocation of nlwb synced before we wrote
* the block that points at it (lwb), we'd leak it if we crashed.
* Therefore, we don't do dmu_tx_commit() until zil_lwb_write_done().
* We dirty the dataset to ensure that zil_sync() will be called
* to clean up in the event of allocation failure or I/O failure.
*/
tx = dmu_tx_create(zilog->zl_os);
/*
* Since we are not going to create any new dirty data, and we
* can even help with clearing the existing dirty data, we
* should not be subject to the dirty data based delays. We
* use TXG_NOTHROTTLE to bypass the delay mechanism.
*/
VERIFY0(dmu_tx_assign(tx, TXG_WAIT | TXG_NOTHROTTLE));
dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
txg = dmu_tx_get_txg(tx);
lwb->lwb_tx = tx;
/*
* Log blocks are pre-allocated. Here we select the size of the next
* block, based on size used in the last block.
* - first find the smallest bucket that will fit the block from a
* limited set of block sizes. This is because it's faster to write
* blocks allocated from the same metaslab as they are adjacent or
* close.
* - next find the maximum from the new suggested size and an array of
* previous sizes. This lessens a picket fence effect of wrongly
* guessing the size if we have a stream of say 2k, 64k, 2k, 64k
* requests.
*
* Note we only write what is used, but we can't just allocate
* the maximum block size because we can exhaust the available
* pool log space.
*/
zil_blksz = zilog->zl_cur_used + sizeof (zil_chain_t);
for (i = 0; zil_blksz > zil_block_buckets[i].limit; i++)
continue;
zil_blksz = MIN(zil_block_buckets[i].blksz, zilog->zl_max_block_size);
zilog->zl_prev_blks[zilog->zl_prev_rotor] = zil_blksz;
for (i = 0; i < ZIL_PREV_BLKS; i++)
zil_blksz = MAX(zil_blksz, zilog->zl_prev_blks[i]);
zilog->zl_prev_rotor = (zilog->zl_prev_rotor + 1) & (ZIL_PREV_BLKS - 1);
BP_ZERO(bp);
error = zio_alloc_zil(spa, zilog->zl_os, txg, bp, zil_blksz, &slog);
if (slog) {
ZIL_STAT_BUMP(zil_itx_metaslab_slog_count);
ZIL_STAT_INCR(zil_itx_metaslab_slog_bytes, lwb->lwb_nused);
} else {
ZIL_STAT_BUMP(zil_itx_metaslab_normal_count);
ZIL_STAT_INCR(zil_itx_metaslab_normal_bytes, lwb->lwb_nused);
}
if (error == 0) {
ASSERT3U(bp->blk_birth, ==, txg);
bp->blk_cksum = lwb->lwb_blk.blk_cksum;
bp->blk_cksum.zc_word[ZIL_ZC_SEQ]++;
/*
* Allocate a new log write block (lwb).
*/
nlwb = zil_alloc_lwb(zilog, bp, slog, txg, TRUE);
}
if (BP_GET_CHECKSUM(&lwb->lwb_blk) == ZIO_CHECKSUM_ZILOG2) {
/* For Slim ZIL only write what is used. */
wsz = P2ROUNDUP_TYPED(lwb->lwb_nused, ZIL_MIN_BLKSZ, uint64_t);
ASSERT3U(wsz, <=, lwb->lwb_sz);
zio_shrink(lwb->lwb_write_zio, wsz);
} else {
wsz = lwb->lwb_sz;
}
zilc->zc_pad = 0;
zilc->zc_nused = lwb->lwb_nused;
zilc->zc_eck.zec_cksum = lwb->lwb_blk.blk_cksum;
/*
* clear unused data for security
*/
bzero(lwb->lwb_buf + lwb->lwb_nused, wsz - lwb->lwb_nused);
spa_config_enter(zilog->zl_spa, SCL_STATE, lwb, RW_READER);
zil_lwb_add_block(lwb, &lwb->lwb_blk);
lwb->lwb_issued_timestamp = gethrtime();
lwb->lwb_state = LWB_STATE_ISSUED;
zio_nowait(lwb->lwb_root_zio);
zio_nowait(lwb->lwb_write_zio);
/*
* If there was an allocation failure then nlwb will be null which
* forces a txg_wait_synced().
*/
return (nlwb);
}
/*
* Maximum amount of write data that can be put into single log block.
*/
uint64_t
zil_max_log_data(zilog_t *zilog)
{
return (zilog->zl_max_block_size -
sizeof (zil_chain_t) - sizeof (lr_write_t));
}
/*
* Maximum amount of log space we agree to waste to reduce number of
* WR_NEED_COPY chunks to reduce zl_get_data() overhead (~12%).
*/
static inline uint64_t
zil_max_waste_space(zilog_t *zilog)
{
return (zil_max_log_data(zilog) / 8);
}
/*
* Maximum amount of write data for WR_COPIED. For correctness, consumers
* must fall back to WR_NEED_COPY if we can't fit the entire record into one
* maximum sized log block, because each WR_COPIED record must fit in a
* single log block. For space efficiency, we want to fit two records into a
* max-sized log block.
*/
uint64_t
zil_max_copied_data(zilog_t *zilog)
{
return ((zilog->zl_max_block_size - sizeof (zil_chain_t)) / 2 -
sizeof (lr_write_t));
}
static lwb_t *
zil_lwb_commit(zilog_t *zilog, itx_t *itx, lwb_t *lwb)
{
lr_t *lrcb, *lrc;
lr_write_t *lrwb, *lrw;
char *lr_buf;
uint64_t dlen, dnow, dpad, lwb_sp, reclen, txg, max_log_data;
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
ASSERT3P(lwb, !=, NULL);
ASSERT3P(lwb->lwb_buf, !=, NULL);
zil_lwb_write_open(zilog, lwb);
lrc = &itx->itx_lr;
lrw = (lr_write_t *)lrc;
/*
* A commit itx doesn't represent any on-disk state; instead
* it's simply used as a place holder on the commit list, and
* provides a mechanism for attaching a "commit waiter" onto the
* correct lwb (such that the waiter can be signalled upon
* completion of that lwb). Thus, we don't process this itx's
* log record if it's a commit itx (these itx's don't have log
* records), and instead link the itx's waiter onto the lwb's
* list of waiters.
*
* For more details, see the comment above zil_commit().
*/
if (lrc->lrc_txtype == TX_COMMIT) {
mutex_enter(&zilog->zl_lock);
zil_commit_waiter_link_lwb(itx->itx_private, lwb);
itx->itx_private = NULL;
mutex_exit(&zilog->zl_lock);
return (lwb);
}
if (lrc->lrc_txtype == TX_WRITE && itx->itx_wr_state == WR_NEED_COPY) {
dlen = P2ROUNDUP_TYPED(
lrw->lr_length, sizeof (uint64_t), uint64_t);
dpad = dlen - lrw->lr_length;
} else {
dlen = dpad = 0;
}
reclen = lrc->lrc_reclen;
zilog->zl_cur_used += (reclen + dlen);
txg = lrc->lrc_txg;
ASSERT3U(zilog->zl_cur_used, <, UINT64_MAX - (reclen + dlen));
cont:
/*
* If this record won't fit in the current log block, start a new one.
* For WR_NEED_COPY optimize layout for minimal number of chunks.
*/
lwb_sp = lwb->lwb_sz - lwb->lwb_nused;
max_log_data = zil_max_log_data(zilog);
if (reclen > lwb_sp || (reclen + dlen > lwb_sp &&
lwb_sp < zil_max_waste_space(zilog) &&
(dlen % max_log_data == 0 ||
lwb_sp < reclen + dlen % max_log_data))) {
lwb = zil_lwb_write_issue(zilog, lwb);
if (lwb == NULL)
return (NULL);
zil_lwb_write_open(zilog, lwb);
ASSERT(LWB_EMPTY(lwb));
lwb_sp = lwb->lwb_sz - lwb->lwb_nused;
/*
* There must be enough space in the new, empty log block to
* hold reclen. For WR_COPIED, we need to fit the whole
* record in one block, and reclen is the header size + the
* data size. For WR_NEED_COPY, we can create multiple
* records, splitting the data into multiple blocks, so we
* only need to fit one word of data per block; in this case
* reclen is just the header size (no data).
*/
ASSERT3U(reclen + MIN(dlen, sizeof (uint64_t)), <=, lwb_sp);
}
dnow = MIN(dlen, lwb_sp - reclen);
lr_buf = lwb->lwb_buf + lwb->lwb_nused;
bcopy(lrc, lr_buf, reclen);
lrcb = (lr_t *)lr_buf; /* Like lrc, but inside lwb. */
lrwb = (lr_write_t *)lrcb; /* Like lrw, but inside lwb. */
ZIL_STAT_BUMP(zil_itx_count);
/*
* If it's a write, fetch the data or get its blkptr as appropriate.
*/
if (lrc->lrc_txtype == TX_WRITE) {
if (txg > spa_freeze_txg(zilog->zl_spa))
txg_wait_synced(zilog->zl_dmu_pool, txg);
if (itx->itx_wr_state == WR_COPIED) {
ZIL_STAT_BUMP(zil_itx_copied_count);
ZIL_STAT_INCR(zil_itx_copied_bytes, lrw->lr_length);
} else {
char *dbuf;
int error;
if (itx->itx_wr_state == WR_NEED_COPY) {
dbuf = lr_buf + reclen;
lrcb->lrc_reclen += dnow;
if (lrwb->lr_length > dnow)
lrwb->lr_length = dnow;
lrw->lr_offset += dnow;
lrw->lr_length -= dnow;
ZIL_STAT_BUMP(zil_itx_needcopy_count);
ZIL_STAT_INCR(zil_itx_needcopy_bytes, dnow);
} else {
ASSERT3S(itx->itx_wr_state, ==, WR_INDIRECT);
dbuf = NULL;
ZIL_STAT_BUMP(zil_itx_indirect_count);
ZIL_STAT_INCR(zil_itx_indirect_bytes,
lrw->lr_length);
}
/*
* We pass in the "lwb_write_zio" rather than
* "lwb_root_zio" so that the "lwb_write_zio"
* becomes the parent of any zio's created by
* the "zl_get_data" callback. The vdevs are
* flushed after the "lwb_write_zio" completes,
* so we want to make sure that completion
* callback waits for these additional zio's,
* such that the vdevs used by those zio's will
* be included in the lwb's vdev tree, and those
* vdevs will be properly flushed. If we passed
* in "lwb_root_zio" here, then these additional
* vdevs may not be flushed; e.g. if these zio's
* completed after "lwb_write_zio" completed.
*/
error = zilog->zl_get_data(itx->itx_private,
itx->itx_gen, lrwb, dbuf, lwb,
lwb->lwb_write_zio);
if (dbuf != NULL && error == 0 && dnow == dlen)
/* Zero any padding bytes in the last block. */
bzero((char *)dbuf + lrwb->lr_length, dpad);
if (error == EIO) {
txg_wait_synced(zilog->zl_dmu_pool, txg);
return (lwb);
}
if (error != 0) {
ASSERT(error == ENOENT || error == EEXIST ||
error == EALREADY);
return (lwb);
}
}
}
/*
* We're actually making an entry, so update lrc_seq to be the
* log record sequence number. Note that this is generally not
* equal to the itx sequence number because not all transactions
* are synchronous, and sometimes spa_sync() gets there first.
*/
lrcb->lrc_seq = ++zilog->zl_lr_seq;
lwb->lwb_nused += reclen + dnow;
zil_lwb_add_txg(lwb, txg);
ASSERT3U(lwb->lwb_nused, <=, lwb->lwb_sz);
ASSERT0(P2PHASE(lwb->lwb_nused, sizeof (uint64_t)));
dlen -= dnow;
if (dlen > 0) {
zilog->zl_cur_used += reclen;
goto cont;
}
return (lwb);
}
itx_t *
zil_itx_create(uint64_t txtype, size_t olrsize)
{
size_t itxsize, lrsize;
itx_t *itx;
lrsize = P2ROUNDUP_TYPED(olrsize, sizeof (uint64_t), size_t);
itxsize = offsetof(itx_t, itx_lr) + lrsize;
itx = zio_data_buf_alloc(itxsize);
itx->itx_lr.lrc_txtype = txtype;
itx->itx_lr.lrc_reclen = lrsize;
itx->itx_lr.lrc_seq = 0; /* defensive */
bzero((char *)&itx->itx_lr + olrsize, lrsize - olrsize);
itx->itx_sync = B_TRUE; /* default is synchronous */
itx->itx_callback = NULL;
itx->itx_callback_data = NULL;
itx->itx_size = itxsize;
return (itx);
}
void
zil_itx_destroy(itx_t *itx)
{
IMPLY(itx->itx_lr.lrc_txtype == TX_COMMIT, itx->itx_callback == NULL);
IMPLY(itx->itx_callback != NULL, itx->itx_lr.lrc_txtype != TX_COMMIT);
if (itx->itx_callback != NULL)
itx->itx_callback(itx->itx_callback_data);
zio_data_buf_free(itx, itx->itx_size);
}
/*
* Free up the sync and async itxs. The itxs_t has already been detached
* so no locks are needed.
*/
static void
zil_itxg_clean(void *arg)
{
itx_t *itx;
list_t *list;
avl_tree_t *t;
void *cookie;
itxs_t *itxs = arg;
itx_async_node_t *ian;
list = &itxs->i_sync_list;
while ((itx = list_head(list)) != NULL) {
/*
* In the general case, commit itxs will not be found
* here, as they'll be committed to an lwb via
* zil_lwb_commit(), and free'd in that function. Having
* said that, it is still possible for commit itxs to be
* found here, due to the following race:
*
* - a thread calls zil_commit() which assigns the
* commit itx to a per-txg i_sync_list
* - zil_itxg_clean() is called (e.g. via spa_sync())
* while the waiter is still on the i_sync_list
*
* There's nothing to prevent syncing the txg while the
* waiter is on the i_sync_list. This normally doesn't
* happen because spa_sync() is slower than zil_commit(),
* but if zil_commit() calls txg_wait_synced() (e.g.
* because zil_create() or zil_commit_writer_stall() is
* called) we will hit this case.
*/
if (itx->itx_lr.lrc_txtype == TX_COMMIT)
zil_commit_waiter_skip(itx->itx_private);
list_remove(list, itx);
zil_itx_destroy(itx);
}
cookie = NULL;
t = &itxs->i_async_tree;
while ((ian = avl_destroy_nodes(t, &cookie)) != NULL) {
list = &ian->ia_list;
while ((itx = list_head(list)) != NULL) {
list_remove(list, itx);
/* commit itxs should never be on the async lists. */
ASSERT3U(itx->itx_lr.lrc_txtype, !=, TX_COMMIT);
zil_itx_destroy(itx);
}
list_destroy(list);
kmem_free(ian, sizeof (itx_async_node_t));
}
avl_destroy(t);
kmem_free(itxs, sizeof (itxs_t));
}
static int
zil_aitx_compare(const void *x1, const void *x2)
{
const uint64_t o1 = ((itx_async_node_t *)x1)->ia_foid;
const uint64_t o2 = ((itx_async_node_t *)x2)->ia_foid;
return (TREE_CMP(o1, o2));
}
/*
* Remove all async itx with the given oid.
*/
void
zil_remove_async(zilog_t *zilog, uint64_t oid)
{
uint64_t otxg, txg;
itx_async_node_t *ian;
avl_tree_t *t;
avl_index_t where;
list_t clean_list;
itx_t *itx;
ASSERT(oid != 0);
list_create(&clean_list, sizeof (itx_t), offsetof(itx_t, itx_node));
if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */
otxg = ZILTEST_TXG;
else
otxg = spa_last_synced_txg(zilog->zl_spa) + 1;
for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) {
itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK];
mutex_enter(&itxg->itxg_lock);
if (itxg->itxg_txg != txg) {
mutex_exit(&itxg->itxg_lock);
continue;
}
/*
* Locate the object node and append its list.
*/
t = &itxg->itxg_itxs->i_async_tree;
ian = avl_find(t, &oid, &where);
if (ian != NULL)
list_move_tail(&clean_list, &ian->ia_list);
mutex_exit(&itxg->itxg_lock);
}
while ((itx = list_head(&clean_list)) != NULL) {
list_remove(&clean_list, itx);
/* commit itxs should never be on the async lists. */
ASSERT3U(itx->itx_lr.lrc_txtype, !=, TX_COMMIT);
zil_itx_destroy(itx);
}
list_destroy(&clean_list);
}
void
zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx)
{
uint64_t txg;
itxg_t *itxg;
itxs_t *itxs, *clean = NULL;
/*
* Ensure the data of a renamed file is committed before the rename.
*/
if ((itx->itx_lr.lrc_txtype & ~TX_CI) == TX_RENAME)
zil_async_to_sync(zilog, itx->itx_oid);
if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX)
txg = ZILTEST_TXG;
else
txg = dmu_tx_get_txg(tx);
itxg = &zilog->zl_itxg[txg & TXG_MASK];
mutex_enter(&itxg->itxg_lock);
itxs = itxg->itxg_itxs;
if (itxg->itxg_txg != txg) {
if (itxs != NULL) {
/*
* The zil_clean callback hasn't got around to cleaning
* this itxg. Save the itxs for release below.
* This should be rare.
*/
zfs_dbgmsg("zil_itx_assign: missed itx cleanup for "
"txg %llu", (u_longlong_t)itxg->itxg_txg);
clean = itxg->itxg_itxs;
}
itxg->itxg_txg = txg;
itxs = itxg->itxg_itxs = kmem_zalloc(sizeof (itxs_t),
KM_SLEEP);
list_create(&itxs->i_sync_list, sizeof (itx_t),
offsetof(itx_t, itx_node));
avl_create(&itxs->i_async_tree, zil_aitx_compare,
sizeof (itx_async_node_t),
offsetof(itx_async_node_t, ia_node));
}
if (itx->itx_sync) {
list_insert_tail(&itxs->i_sync_list, itx);
} else {
avl_tree_t *t = &itxs->i_async_tree;
uint64_t foid =
LR_FOID_GET_OBJ(((lr_ooo_t *)&itx->itx_lr)->lr_foid);
itx_async_node_t *ian;
avl_index_t where;
ian = avl_find(t, &foid, &where);
if (ian == NULL) {
ian = kmem_alloc(sizeof (itx_async_node_t),
KM_SLEEP);
list_create(&ian->ia_list, sizeof (itx_t),
offsetof(itx_t, itx_node));
ian->ia_foid = foid;
avl_insert(t, ian, where);
}
list_insert_tail(&ian->ia_list, itx);
}
itx->itx_lr.lrc_txg = dmu_tx_get_txg(tx);
/*
* We don't want to dirty the ZIL using ZILTEST_TXG, because
* zil_clean() will never be called using ZILTEST_TXG. Thus, we
* need to be careful to always dirty the ZIL using the "real"
* TXG (not itxg_txg) even when the SPA is frozen.
*/
zilog_dirty(zilog, dmu_tx_get_txg(tx));
mutex_exit(&itxg->itxg_lock);
/* Release the old itxs now we've dropped the lock */
if (clean != NULL)
zil_itxg_clean(clean);
}
/*
* If there are any in-memory intent log transactions which have now been
* synced then start up a taskq to free them. We should only do this after we
* have written out the uberblocks (i.e. txg has been committed) so that
* don't inadvertently clean out in-memory log records that would be required
* by zil_commit().
*/
void
zil_clean(zilog_t *zilog, uint64_t synced_txg)
{
itxg_t *itxg = &zilog->zl_itxg[synced_txg & TXG_MASK];
itxs_t *clean_me;
ASSERT3U(synced_txg, <, ZILTEST_TXG);
mutex_enter(&itxg->itxg_lock);
if (itxg->itxg_itxs == NULL || itxg->itxg_txg == ZILTEST_TXG) {
mutex_exit(&itxg->itxg_lock);
return;
}
ASSERT3U(itxg->itxg_txg, <=, synced_txg);
ASSERT3U(itxg->itxg_txg, !=, 0);
clean_me = itxg->itxg_itxs;
itxg->itxg_itxs = NULL;
itxg->itxg_txg = 0;
mutex_exit(&itxg->itxg_lock);
/*
* Preferably start a task queue to free up the old itxs but
* if taskq_dispatch can't allocate resources to do that then
* free it in-line. This should be rare. Note, using TQ_SLEEP
* created a bad performance problem.
*/
ASSERT3P(zilog->zl_dmu_pool, !=, NULL);
ASSERT3P(zilog->zl_dmu_pool->dp_zil_clean_taskq, !=, NULL);
taskqid_t id = taskq_dispatch(zilog->zl_dmu_pool->dp_zil_clean_taskq,
zil_itxg_clean, clean_me, TQ_NOSLEEP);
if (id == TASKQID_INVALID)
zil_itxg_clean(clean_me);
}
/*
* This function will traverse the queue of itxs that need to be
* committed, and move them onto the ZIL's zl_itx_commit_list.
*/
static void
zil_get_commit_list(zilog_t *zilog)
{
uint64_t otxg, txg;
list_t *commit_list = &zilog->zl_itx_commit_list;
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */
otxg = ZILTEST_TXG;
else
otxg = spa_last_synced_txg(zilog->zl_spa) + 1;
/*
* This is inherently racy, since there is nothing to prevent
* the last synced txg from changing. That's okay since we'll
* only commit things in the future.
*/
for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) {
itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK];
mutex_enter(&itxg->itxg_lock);
if (itxg->itxg_txg != txg) {
mutex_exit(&itxg->itxg_lock);
continue;
}
/*
* If we're adding itx records to the zl_itx_commit_list,
* then the zil better be dirty in this "txg". We can assert
* that here since we're holding the itxg_lock which will
* prevent spa_sync from cleaning it. Once we add the itxs
* to the zl_itx_commit_list we must commit it to disk even
* if it's unnecessary (i.e. the txg was synced).
*/
ASSERT(zilog_is_dirty_in_txg(zilog, txg) ||
spa_freeze_txg(zilog->zl_spa) != UINT64_MAX);
list_move_tail(commit_list, &itxg->itxg_itxs->i_sync_list);
mutex_exit(&itxg->itxg_lock);
}
}
/*
* Move the async itxs for a specified object to commit into sync lists.
*/
void
zil_async_to_sync(zilog_t *zilog, uint64_t foid)
{
uint64_t otxg, txg;
itx_async_node_t *ian;
avl_tree_t *t;
avl_index_t where;
if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */
otxg = ZILTEST_TXG;
else
otxg = spa_last_synced_txg(zilog->zl_spa) + 1;
/*
* This is inherently racy, since there is nothing to prevent
* the last synced txg from changing.
*/
for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) {
itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK];
mutex_enter(&itxg->itxg_lock);
if (itxg->itxg_txg != txg) {
mutex_exit(&itxg->itxg_lock);
continue;
}
/*
* If a foid is specified then find that node and append its
* list. Otherwise walk the tree appending all the lists
* to the sync list. We add to the end rather than the
* beginning to ensure the create has happened.
*/
t = &itxg->itxg_itxs->i_async_tree;
if (foid != 0) {
ian = avl_find(t, &foid, &where);
if (ian != NULL) {
list_move_tail(&itxg->itxg_itxs->i_sync_list,
&ian->ia_list);
}
} else {
void *cookie = NULL;
while ((ian = avl_destroy_nodes(t, &cookie)) != NULL) {
list_move_tail(&itxg->itxg_itxs->i_sync_list,
&ian->ia_list);
list_destroy(&ian->ia_list);
kmem_free(ian, sizeof (itx_async_node_t));
}
}
mutex_exit(&itxg->itxg_lock);
}
}
/*
* This function will prune commit itxs that are at the head of the
* commit list (it won't prune past the first non-commit itx), and
* either: a) attach them to the last lwb that's still pending
* completion, or b) skip them altogether.
*
* This is used as a performance optimization to prevent commit itxs
* from generating new lwbs when it's unnecessary to do so.
*/
static void
zil_prune_commit_list(zilog_t *zilog)
{
itx_t *itx;
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
while ((itx = list_head(&zilog->zl_itx_commit_list)) != NULL) {
lr_t *lrc = &itx->itx_lr;
if (lrc->lrc_txtype != TX_COMMIT)
break;
mutex_enter(&zilog->zl_lock);
lwb_t *last_lwb = zilog->zl_last_lwb_opened;
if (last_lwb == NULL ||
last_lwb->lwb_state == LWB_STATE_FLUSH_DONE) {
/*
* All of the itxs this waiter was waiting on
* must have already completed (or there were
* never any itx's for it to wait on), so it's
* safe to skip this waiter and mark it done.
*/
zil_commit_waiter_skip(itx->itx_private);
} else {
zil_commit_waiter_link_lwb(itx->itx_private, last_lwb);
itx->itx_private = NULL;
}
mutex_exit(&zilog->zl_lock);
list_remove(&zilog->zl_itx_commit_list, itx);
zil_itx_destroy(itx);
}
IMPLY(itx != NULL, itx->itx_lr.lrc_txtype != TX_COMMIT);
}
static void
zil_commit_writer_stall(zilog_t *zilog)
{
/*
* When zio_alloc_zil() fails to allocate the next lwb block on
* disk, we must call txg_wait_synced() to ensure all of the
* lwbs in the zilog's zl_lwb_list are synced and then freed (in
* zil_sync()), such that any subsequent ZIL writer (i.e. a call
* to zil_process_commit_list()) will have to call zil_create(),
* and start a new ZIL chain.
*
* Since zil_alloc_zil() failed, the lwb that was previously
* issued does not have a pointer to the "next" lwb on disk.
* Thus, if another ZIL writer thread was to allocate the "next"
* on-disk lwb, that block could be leaked in the event of a
* crash (because the previous lwb on-disk would not point to
* it).
*
* We must hold the zilog's zl_issuer_lock while we do this, to
* ensure no new threads enter zil_process_commit_list() until
* all lwb's in the zl_lwb_list have been synced and freed
* (which is achieved via the txg_wait_synced() call).
*/
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
txg_wait_synced(zilog->zl_dmu_pool, 0);
ASSERT3P(list_tail(&zilog->zl_lwb_list), ==, NULL);
}
/*
* This function will traverse the commit list, creating new lwbs as
* needed, and committing the itxs from the commit list to these newly
* created lwbs. Additionally, as a new lwb is created, the previous
* lwb will be issued to the zio layer to be written to disk.
*/
static void
zil_process_commit_list(zilog_t *zilog)
{
spa_t *spa = zilog->zl_spa;
list_t nolwb_itxs;
list_t nolwb_waiters;
lwb_t *lwb;
itx_t *itx;
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
/*
* Return if there's nothing to commit before we dirty the fs by
* calling zil_create().
*/
if (list_head(&zilog->zl_itx_commit_list) == NULL)
return;
list_create(&nolwb_itxs, sizeof (itx_t), offsetof(itx_t, itx_node));
list_create(&nolwb_waiters, sizeof (zil_commit_waiter_t),
offsetof(zil_commit_waiter_t, zcw_node));
lwb = list_tail(&zilog->zl_lwb_list);
if (lwb == NULL) {
lwb = zil_create(zilog);
} else {
ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED);
ASSERT3S(lwb->lwb_state, !=, LWB_STATE_WRITE_DONE);
ASSERT3S(lwb->lwb_state, !=, LWB_STATE_FLUSH_DONE);
}
while ((itx = list_head(&zilog->zl_itx_commit_list)) != NULL) {
lr_t *lrc = &itx->itx_lr;
uint64_t txg = lrc->lrc_txg;
ASSERT3U(txg, !=, 0);
if (lrc->lrc_txtype == TX_COMMIT) {
DTRACE_PROBE2(zil__process__commit__itx,
zilog_t *, zilog, itx_t *, itx);
} else {
DTRACE_PROBE2(zil__process__normal__itx,
zilog_t *, zilog, itx_t *, itx);
}
list_remove(&zilog->zl_itx_commit_list, itx);
boolean_t synced = txg <= spa_last_synced_txg(spa);
boolean_t frozen = txg > spa_freeze_txg(spa);
/*
* If the txg of this itx has already been synced out, then
* we don't need to commit this itx to an lwb. This is
* because the data of this itx will have already been
* written to the main pool. This is inherently racy, and
* it's still ok to commit an itx whose txg has already
* been synced; this will result in a write that's
* unnecessary, but will do no harm.
*
* With that said, we always want to commit TX_COMMIT itxs
* to an lwb, regardless of whether or not that itx's txg
* has been synced out. We do this to ensure any OPENED lwb
* will always have at least one zil_commit_waiter_t linked
* to the lwb.
*
* As a counter-example, if we skipped TX_COMMIT itx's
* whose txg had already been synced, the following
* situation could occur if we happened to be racing with
* spa_sync:
*
* 1. We commit a non-TX_COMMIT itx to an lwb, where the
* itx's txg is 10 and the last synced txg is 9.
* 2. spa_sync finishes syncing out txg 10.
* 3. We move to the next itx in the list, it's a TX_COMMIT
* whose txg is 10, so we skip it rather than committing
* it to the lwb used in (1).
*
* If the itx that is skipped in (3) is the last TX_COMMIT
* itx in the commit list, than it's possible for the lwb
* used in (1) to remain in the OPENED state indefinitely.
*
* To prevent the above scenario from occurring, ensuring
* that once an lwb is OPENED it will transition to ISSUED
* and eventually DONE, we always commit TX_COMMIT itx's to
* an lwb here, even if that itx's txg has already been
* synced.
*
* Finally, if the pool is frozen, we _always_ commit the
* itx. The point of freezing the pool is to prevent data
* from being written to the main pool via spa_sync, and
* instead rely solely on the ZIL to persistently store the
* data; i.e. when the pool is frozen, the last synced txg
* value can't be trusted.
*/
if (frozen || !synced || lrc->lrc_txtype == TX_COMMIT) {
if (lwb != NULL) {
lwb = zil_lwb_commit(zilog, itx, lwb);
if (lwb == NULL)
list_insert_tail(&nolwb_itxs, itx);
else
list_insert_tail(&lwb->lwb_itxs, itx);
} else {
if (lrc->lrc_txtype == TX_COMMIT) {
zil_commit_waiter_link_nolwb(
itx->itx_private, &nolwb_waiters);
}
list_insert_tail(&nolwb_itxs, itx);
}
} else {
ASSERT3S(lrc->lrc_txtype, !=, TX_COMMIT);
zil_itx_destroy(itx);
}
}
if (lwb == NULL) {
/*
* This indicates zio_alloc_zil() failed to allocate the
* "next" lwb on-disk. When this happens, we must stall
* the ZIL write pipeline; see the comment within
* zil_commit_writer_stall() for more details.
*/
zil_commit_writer_stall(zilog);
/*
* Additionally, we have to signal and mark the "nolwb"
* waiters as "done" here, since without an lwb, we
* can't do this via zil_lwb_flush_vdevs_done() like
* normal.
*/
zil_commit_waiter_t *zcw;
while ((zcw = list_head(&nolwb_waiters)) != NULL) {
zil_commit_waiter_skip(zcw);
list_remove(&nolwb_waiters, zcw);
}
/*
* And finally, we have to destroy the itx's that
* couldn't be committed to an lwb; this will also call
* the itx's callback if one exists for the itx.
*/
while ((itx = list_head(&nolwb_itxs)) != NULL) {
list_remove(&nolwb_itxs, itx);
zil_itx_destroy(itx);
}
} else {
ASSERT(list_is_empty(&nolwb_waiters));
ASSERT3P(lwb, !=, NULL);
ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED);
ASSERT3S(lwb->lwb_state, !=, LWB_STATE_WRITE_DONE);
ASSERT3S(lwb->lwb_state, !=, LWB_STATE_FLUSH_DONE);
/*
* At this point, the ZIL block pointed at by the "lwb"
* variable is in one of the following states: "closed"
* or "open".
*
* If it's "closed", then no itxs have been committed to
* it, so there's no point in issuing its zio (i.e. it's
* "empty").
*
* If it's "open", then it contains one or more itxs that
* eventually need to be committed to stable storage. In
* this case we intentionally do not issue the lwb's zio
* to disk yet, and instead rely on one of the following
* two mechanisms for issuing the zio:
*
* 1. Ideally, there will be more ZIL activity occurring
* on the system, such that this function will be
* immediately called again (not necessarily by the same
* thread) and this lwb's zio will be issued via
* zil_lwb_commit(). This way, the lwb is guaranteed to
* be "full" when it is issued to disk, and we'll make
* use of the lwb's size the best we can.
*
* 2. If there isn't sufficient ZIL activity occurring on
* the system, such that this lwb's zio isn't issued via
* zil_lwb_commit(), zil_commit_waiter() will issue the
* lwb's zio. If this occurs, the lwb is not guaranteed
* to be "full" by the time its zio is issued, and means
* the size of the lwb was "too large" given the amount
* of ZIL activity occurring on the system at that time.
*
* We do this for a couple of reasons:
*
* 1. To try and reduce the number of IOPs needed to
* write the same number of itxs. If an lwb has space
* available in its buffer for more itxs, and more itxs
* will be committed relatively soon (relative to the
* latency of performing a write), then it's beneficial
* to wait for these "next" itxs. This way, more itxs
* can be committed to stable storage with fewer writes.
*
* 2. To try and use the largest lwb block size that the
* incoming rate of itxs can support. Again, this is to
* try and pack as many itxs into as few lwbs as
* possible, without significantly impacting the latency
* of each individual itx.
*/
}
}
/*
* This function is responsible for ensuring the passed in commit waiter
* (and associated commit itx) is committed to an lwb. If the waiter is
* not already committed to an lwb, all itxs in the zilog's queue of
* itxs will be processed. The assumption is the passed in waiter's
* commit itx will found in the queue just like the other non-commit
* itxs, such that when the entire queue is processed, the waiter will
* have been committed to an lwb.
*
* The lwb associated with the passed in waiter is not guaranteed to
* have been issued by the time this function completes. If the lwb is
* not issued, we rely on future calls to zil_commit_writer() to issue
* the lwb, or the timeout mechanism found in zil_commit_waiter().
*/
static void
zil_commit_writer(zilog_t *zilog, zil_commit_waiter_t *zcw)
{
ASSERT(!MUTEX_HELD(&zilog->zl_lock));
ASSERT(spa_writeable(zilog->zl_spa));
mutex_enter(&zilog->zl_issuer_lock);
if (zcw->zcw_lwb != NULL || zcw->zcw_done) {
/*
* It's possible that, while we were waiting to acquire
* the "zl_issuer_lock", another thread committed this
* waiter to an lwb. If that occurs, we bail out early,
* without processing any of the zilog's queue of itxs.
*
* On certain workloads and system configurations, the
* "zl_issuer_lock" can become highly contended. In an
* attempt to reduce this contention, we immediately drop
* the lock if the waiter has already been processed.
*
* We've measured this optimization to reduce CPU spent
* contending on this lock by up to 5%, using a system
* with 32 CPUs, low latency storage (~50 usec writes),
* and 1024 threads performing sync writes.
*/
goto out;
}
ZIL_STAT_BUMP(zil_commit_writer_count);
zil_get_commit_list(zilog);
zil_prune_commit_list(zilog);
zil_process_commit_list(zilog);
out:
mutex_exit(&zilog->zl_issuer_lock);
}
static void
zil_commit_waiter_timeout(zilog_t *zilog, zil_commit_waiter_t *zcw)
{
ASSERT(!MUTEX_HELD(&zilog->zl_issuer_lock));
ASSERT(MUTEX_HELD(&zcw->zcw_lock));
ASSERT3B(zcw->zcw_done, ==, B_FALSE);
lwb_t *lwb = zcw->zcw_lwb;
ASSERT3P(lwb, !=, NULL);
ASSERT3S(lwb->lwb_state, !=, LWB_STATE_CLOSED);
/*
* If the lwb has already been issued by another thread, we can
* immediately return since there's no work to be done (the
* point of this function is to issue the lwb). Additionally, we
* do this prior to acquiring the zl_issuer_lock, to avoid
* acquiring it when it's not necessary to do so.
*/
if (lwb->lwb_state == LWB_STATE_ISSUED ||
lwb->lwb_state == LWB_STATE_WRITE_DONE ||
lwb->lwb_state == LWB_STATE_FLUSH_DONE)
return;
/*
* In order to call zil_lwb_write_issue() we must hold the
* zilog's "zl_issuer_lock". We can't simply acquire that lock,
* since we're already holding the commit waiter's "zcw_lock",
* and those two locks are acquired in the opposite order
* elsewhere.
*/
mutex_exit(&zcw->zcw_lock);
mutex_enter(&zilog->zl_issuer_lock);
mutex_enter(&zcw->zcw_lock);
/*
* Since we just dropped and re-acquired the commit waiter's
* lock, we have to re-check to see if the waiter was marked
* "done" during that process. If the waiter was marked "done",
* the "lwb" pointer is no longer valid (it can be free'd after
* the waiter is marked "done"), so without this check we could
* wind up with a use-after-free error below.
*/
if (zcw->zcw_done)
goto out;
ASSERT3P(lwb, ==, zcw->zcw_lwb);
/*
* We've already checked this above, but since we hadn't acquired
* the zilog's zl_issuer_lock, we have to perform this check a
* second time while holding the lock.
*
* We don't need to hold the zl_lock since the lwb cannot transition
* from OPENED to ISSUED while we hold the zl_issuer_lock. The lwb
* _can_ transition from ISSUED to DONE, but it's OK to race with
* that transition since we treat the lwb the same, whether it's in
* the ISSUED or DONE states.
*
* The important thing, is we treat the lwb differently depending on
* if it's ISSUED or OPENED, and block any other threads that might
* attempt to issue this lwb. For that reason we hold the
* zl_issuer_lock when checking the lwb_state; we must not call
* zil_lwb_write_issue() if the lwb had already been issued.
*
* See the comment above the lwb_state_t structure definition for
* more details on the lwb states, and locking requirements.
*/
if (lwb->lwb_state == LWB_STATE_ISSUED ||
lwb->lwb_state == LWB_STATE_WRITE_DONE ||
lwb->lwb_state == LWB_STATE_FLUSH_DONE)
goto out;
ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED);
/*
* As described in the comments above zil_commit_waiter() and
* zil_process_commit_list(), we need to issue this lwb's zio
* since we've reached the commit waiter's timeout and it still
* hasn't been issued.
*/
lwb_t *nlwb = zil_lwb_write_issue(zilog, lwb);
IMPLY(nlwb != NULL, lwb->lwb_state != LWB_STATE_OPENED);
/*
* Since the lwb's zio hadn't been issued by the time this thread
* reached its timeout, we reset the zilog's "zl_cur_used" field
* to influence the zil block size selection algorithm.
*
* By having to issue the lwb's zio here, it means the size of the
* lwb was too large, given the incoming throughput of itxs. By
* setting "zl_cur_used" to zero, we communicate this fact to the
* block size selection algorithm, so it can take this information
* into account, and potentially select a smaller size for the
* next lwb block that is allocated.
*/
zilog->zl_cur_used = 0;
if (nlwb == NULL) {
/*
* When zil_lwb_write_issue() returns NULL, this
* indicates zio_alloc_zil() failed to allocate the
* "next" lwb on-disk. When this occurs, the ZIL write
* pipeline must be stalled; see the comment within the
* zil_commit_writer_stall() function for more details.
*
* We must drop the commit waiter's lock prior to
* calling zil_commit_writer_stall() or else we can wind
* up with the following deadlock:
*
* - This thread is waiting for the txg to sync while
* holding the waiter's lock; txg_wait_synced() is
* used within txg_commit_writer_stall().
*
* - The txg can't sync because it is waiting for this
* lwb's zio callback to call dmu_tx_commit().
*
* - The lwb's zio callback can't call dmu_tx_commit()
* because it's blocked trying to acquire the waiter's
* lock, which occurs prior to calling dmu_tx_commit()
*/
mutex_exit(&zcw->zcw_lock);
zil_commit_writer_stall(zilog);
mutex_enter(&zcw->zcw_lock);
}
out:
mutex_exit(&zilog->zl_issuer_lock);
ASSERT(MUTEX_HELD(&zcw->zcw_lock));
}
/*
* This function is responsible for performing the following two tasks:
*
* 1. its primary responsibility is to block until the given "commit
* waiter" is considered "done".
*
* 2. its secondary responsibility is to issue the zio for the lwb that
* the given "commit waiter" is waiting on, if this function has
* waited "long enough" and the lwb is still in the "open" state.
*
* Given a sufficient amount of itxs being generated and written using
* the ZIL, the lwb's zio will be issued via the zil_lwb_commit()
* function. If this does not occur, this secondary responsibility will
* ensure the lwb is issued even if there is not other synchronous
* activity on the system.
*
* For more details, see zil_process_commit_list(); more specifically,
* the comment at the bottom of that function.
*/
static void
zil_commit_waiter(zilog_t *zilog, zil_commit_waiter_t *zcw)
{
ASSERT(!MUTEX_HELD(&zilog->zl_lock));
ASSERT(!MUTEX_HELD(&zilog->zl_issuer_lock));
ASSERT(spa_writeable(zilog->zl_spa));
mutex_enter(&zcw->zcw_lock);
/*
* The timeout is scaled based on the lwb latency to avoid
* significantly impacting the latency of each individual itx.
* For more details, see the comment at the bottom of the
* zil_process_commit_list() function.
*/
int pct = MAX(zfs_commit_timeout_pct, 1);
hrtime_t sleep = (zilog->zl_last_lwb_latency * pct) / 100;
hrtime_t wakeup = gethrtime() + sleep;
boolean_t timedout = B_FALSE;
while (!zcw->zcw_done) {
ASSERT(MUTEX_HELD(&zcw->zcw_lock));
lwb_t *lwb = zcw->zcw_lwb;
/*
* Usually, the waiter will have a non-NULL lwb field here,
* but it's possible for it to be NULL as a result of
* zil_commit() racing with spa_sync().
*
* When zil_clean() is called, it's possible for the itxg
* list (which may be cleaned via a taskq) to contain
* commit itxs. When this occurs, the commit waiters linked
* off of these commit itxs will not be committed to an
* lwb. Additionally, these commit waiters will not be
* marked done until zil_commit_waiter_skip() is called via
* zil_itxg_clean().
*
* Thus, it's possible for this commit waiter (i.e. the
* "zcw" variable) to be found in this "in between" state;
* where it's "zcw_lwb" field is NULL, and it hasn't yet
* been skipped, so it's "zcw_done" field is still B_FALSE.
*/
IMPLY(lwb != NULL, lwb->lwb_state != LWB_STATE_CLOSED);
if (lwb != NULL && lwb->lwb_state == LWB_STATE_OPENED) {
ASSERT3B(timedout, ==, B_FALSE);
/*
* If the lwb hasn't been issued yet, then we
* need to wait with a timeout, in case this
* function needs to issue the lwb after the
* timeout is reached; responsibility (2) from
* the comment above this function.
*/
int rc = cv_timedwait_hires(&zcw->zcw_cv,
&zcw->zcw_lock, wakeup, USEC2NSEC(1),
CALLOUT_FLAG_ABSOLUTE);
if (rc != -1 || zcw->zcw_done)
continue;
timedout = B_TRUE;
zil_commit_waiter_timeout(zilog, zcw);
if (!zcw->zcw_done) {
/*
* If the commit waiter has already been
* marked "done", it's possible for the
* waiter's lwb structure to have already
* been freed. Thus, we can only reliably
* make these assertions if the waiter
* isn't done.
*/
ASSERT3P(lwb, ==, zcw->zcw_lwb);
ASSERT3S(lwb->lwb_state, !=, LWB_STATE_OPENED);
}
} else {
/*
* If the lwb isn't open, then it must have already
* been issued. In that case, there's no need to
* use a timeout when waiting for the lwb to
* complete.
*
* Additionally, if the lwb is NULL, the waiter
* will soon be signaled and marked done via
* zil_clean() and zil_itxg_clean(), so no timeout
* is required.
*/
IMPLY(lwb != NULL,
lwb->lwb_state == LWB_STATE_ISSUED ||
lwb->lwb_state == LWB_STATE_WRITE_DONE ||
lwb->lwb_state == LWB_STATE_FLUSH_DONE);
cv_wait(&zcw->zcw_cv, &zcw->zcw_lock);
}
}
mutex_exit(&zcw->zcw_lock);
}
static zil_commit_waiter_t *
zil_alloc_commit_waiter(void)
{
zil_commit_waiter_t *zcw = kmem_cache_alloc(zil_zcw_cache, KM_SLEEP);
cv_init(&zcw->zcw_cv, NULL, CV_DEFAULT, NULL);
mutex_init(&zcw->zcw_lock, NULL, MUTEX_DEFAULT, NULL);
list_link_init(&zcw->zcw_node);
zcw->zcw_lwb = NULL;
zcw->zcw_done = B_FALSE;
zcw->zcw_zio_error = 0;
return (zcw);
}
static void
zil_free_commit_waiter(zil_commit_waiter_t *zcw)
{
ASSERT(!list_link_active(&zcw->zcw_node));
ASSERT3P(zcw->zcw_lwb, ==, NULL);
ASSERT3B(zcw->zcw_done, ==, B_TRUE);
mutex_destroy(&zcw->zcw_lock);
cv_destroy(&zcw->zcw_cv);
kmem_cache_free(zil_zcw_cache, zcw);
}
/*
* This function is used to create a TX_COMMIT itx and assign it. This
* way, it will be linked into the ZIL's list of synchronous itxs, and
* then later committed to an lwb (or skipped) when
* zil_process_commit_list() is called.
*/
static void
zil_commit_itx_assign(zilog_t *zilog, zil_commit_waiter_t *zcw)
{
dmu_tx_t *tx = dmu_tx_create(zilog->zl_os);
VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
itx_t *itx = zil_itx_create(TX_COMMIT, sizeof (lr_t));
itx->itx_sync = B_TRUE;
itx->itx_private = zcw;
zil_itx_assign(zilog, itx, tx);
dmu_tx_commit(tx);
}
/*
* Commit ZFS Intent Log transactions (itxs) to stable storage.
*
* When writing ZIL transactions to the on-disk representation of the
* ZIL, the itxs are committed to a Log Write Block (lwb). Multiple
* itxs can be committed to a single lwb. Once a lwb is written and
* committed to stable storage (i.e. the lwb is written, and vdevs have
* been flushed), each itx that was committed to that lwb is also
* considered to be committed to stable storage.
*
* When an itx is committed to an lwb, the log record (lr_t) contained
* by the itx is copied into the lwb's zio buffer, and once this buffer
* is written to disk, it becomes an on-disk ZIL block.
*
* As itxs are generated, they're inserted into the ZIL's queue of
* uncommitted itxs. The semantics of zil_commit() are such that it will
* block until all itxs that were in the queue when it was called, are
* committed to stable storage.
*
* If "foid" is zero, this means all "synchronous" and "asynchronous"
* itxs, for all objects in the dataset, will be committed to stable
* storage prior to zil_commit() returning. If "foid" is non-zero, all
* "synchronous" itxs for all objects, but only "asynchronous" itxs
* that correspond to the foid passed in, will be committed to stable
* storage prior to zil_commit() returning.
*
* Generally speaking, when zil_commit() is called, the consumer doesn't
* actually care about _all_ of the uncommitted itxs. Instead, they're
* simply trying to waiting for a specific itx to be committed to disk,
* but the interface(s) for interacting with the ZIL don't allow such
* fine-grained communication. A better interface would allow a consumer
* to create and assign an itx, and then pass a reference to this itx to
* zil_commit(); such that zil_commit() would return as soon as that
* specific itx was committed to disk (instead of waiting for _all_
* itxs to be committed).
*
* When a thread calls zil_commit() a special "commit itx" will be
* generated, along with a corresponding "waiter" for this commit itx.
* zil_commit() will wait on this waiter's CV, such that when the waiter
* is marked done, and signaled, zil_commit() will return.
*
* This commit itx is inserted into the queue of uncommitted itxs. This
* provides an easy mechanism for determining which itxs were in the
* queue prior to zil_commit() having been called, and which itxs were
* added after zil_commit() was called.
*
* The commit it is special; it doesn't have any on-disk representation.
* When a commit itx is "committed" to an lwb, the waiter associated
* with it is linked onto the lwb's list of waiters. Then, when that lwb
* completes, each waiter on the lwb's list is marked done and signaled
* -- allowing the thread waiting on the waiter to return from zil_commit().
*
* It's important to point out a few critical factors that allow us
* to make use of the commit itxs, commit waiters, per-lwb lists of
* commit waiters, and zio completion callbacks like we're doing:
*
* 1. The list of waiters for each lwb is traversed, and each commit
* waiter is marked "done" and signaled, in the zio completion
* callback of the lwb's zio[*].
*
* * Actually, the waiters are signaled in the zio completion
* callback of the root zio for the DKIOCFLUSHWRITECACHE commands
* that are sent to the vdevs upon completion of the lwb zio.
*
* 2. When the itxs are inserted into the ZIL's queue of uncommitted
* itxs, the order in which they are inserted is preserved[*]; as
* itxs are added to the queue, they are added to the tail of
* in-memory linked lists.
*
* When committing the itxs to lwbs (to be written to disk), they
* are committed in the same order in which the itxs were added to
* the uncommitted queue's linked list(s); i.e. the linked list of
* itxs to commit is traversed from head to tail, and each itx is
* committed to an lwb in that order.
*
* * To clarify:
*
* - the order of "sync" itxs is preserved w.r.t. other
* "sync" itxs, regardless of the corresponding objects.
* - the order of "async" itxs is preserved w.r.t. other
* "async" itxs corresponding to the same object.
* - the order of "async" itxs is *not* preserved w.r.t. other
* "async" itxs corresponding to different objects.
* - the order of "sync" itxs w.r.t. "async" itxs (or vice
* versa) is *not* preserved, even for itxs that correspond
* to the same object.
*
* For more details, see: zil_itx_assign(), zil_async_to_sync(),
* zil_get_commit_list(), and zil_process_commit_list().
*
* 3. The lwbs represent a linked list of blocks on disk. Thus, any
* lwb cannot be considered committed to stable storage, until its
* "previous" lwb is also committed to stable storage. This fact,
* coupled with the fact described above, means that itxs are
* committed in (roughly) the order in which they were generated.
* This is essential because itxs are dependent on prior itxs.
* Thus, we *must not* deem an itx as being committed to stable
* storage, until *all* prior itxs have also been committed to
* stable storage.
*
* To enforce this ordering of lwb zio's, while still leveraging as
* much of the underlying storage performance as possible, we rely
* on two fundamental concepts:
*
* 1. The creation and issuance of lwb zio's is protected by
* the zilog's "zl_issuer_lock", which ensures only a single
* thread is creating and/or issuing lwb's at a time
* 2. The "previous" lwb is a child of the "current" lwb
* (leveraging the zio parent-child dependency graph)
*
* By relying on this parent-child zio relationship, we can have
* many lwb zio's concurrently issued to the underlying storage,
* but the order in which they complete will be the same order in
* which they were created.
*/
void
zil_commit(zilog_t *zilog, uint64_t foid)
{
/*
* We should never attempt to call zil_commit on a snapshot for
* a couple of reasons:
*
* 1. A snapshot may never be modified, thus it cannot have any
* in-flight itxs that would have modified the dataset.
*
* 2. By design, when zil_commit() is called, a commit itx will
* be assigned to this zilog; as a result, the zilog will be
* dirtied. We must not dirty the zilog of a snapshot; there's
* checks in the code that enforce this invariant, and will
* cause a panic if it's not upheld.
*/
ASSERT3B(dmu_objset_is_snapshot(zilog->zl_os), ==, B_FALSE);
if (zilog->zl_sync == ZFS_SYNC_DISABLED)
return;
if (!spa_writeable(zilog->zl_spa)) {
/*
* If the SPA is not writable, there should never be any
* pending itxs waiting to be committed to disk. If that
* weren't true, we'd skip writing those itxs out, and
* would break the semantics of zil_commit(); thus, we're
* verifying that truth before we return to the caller.
*/
ASSERT(list_is_empty(&zilog->zl_lwb_list));
ASSERT3P(zilog->zl_last_lwb_opened, ==, NULL);
for (int i = 0; i < TXG_SIZE; i++)
ASSERT3P(zilog->zl_itxg[i].itxg_itxs, ==, NULL);
return;
}
/*
* If the ZIL is suspended, we don't want to dirty it by calling
* zil_commit_itx_assign() below, nor can we write out
* lwbs like would be done in zil_commit_write(). Thus, we
* simply rely on txg_wait_synced() to maintain the necessary
* semantics, and avoid calling those functions altogether.
*/
if (zilog->zl_suspend > 0) {
txg_wait_synced(zilog->zl_dmu_pool, 0);
return;
}
zil_commit_impl(zilog, foid);
}
void
zil_commit_impl(zilog_t *zilog, uint64_t foid)
{
ZIL_STAT_BUMP(zil_commit_count);
/*
* Move the "async" itxs for the specified foid to the "sync"
* queues, such that they will be later committed (or skipped)
* to an lwb when zil_process_commit_list() is called.
*
* Since these "async" itxs must be committed prior to this
* call to zil_commit returning, we must perform this operation
* before we call zil_commit_itx_assign().
*/
zil_async_to_sync(zilog, foid);
/*
* We allocate a new "waiter" structure which will initially be
* linked to the commit itx using the itx's "itx_private" field.
* Since the commit itx doesn't represent any on-disk state,
* when it's committed to an lwb, rather than copying the its
* lr_t into the lwb's buffer, the commit itx's "waiter" will be
* added to the lwb's list of waiters. Then, when the lwb is
* committed to stable storage, each waiter in the lwb's list of
* waiters will be marked "done", and signalled.
*
* We must create the waiter and assign the commit itx prior to
* calling zil_commit_writer(), or else our specific commit itx
* is not guaranteed to be committed to an lwb prior to calling
* zil_commit_waiter().
*/
zil_commit_waiter_t *zcw = zil_alloc_commit_waiter();
zil_commit_itx_assign(zilog, zcw);
zil_commit_writer(zilog, zcw);
zil_commit_waiter(zilog, zcw);
if (zcw->zcw_zio_error != 0) {
/*
* If there was an error writing out the ZIL blocks that
* this thread is waiting on, then we fallback to
* relying on spa_sync() to write out the data this
* thread is waiting on. Obviously this has performance
* implications, but the expectation is for this to be
* an exceptional case, and shouldn't occur often.
*/
DTRACE_PROBE2(zil__commit__io__error,
zilog_t *, zilog, zil_commit_waiter_t *, zcw);
txg_wait_synced(zilog->zl_dmu_pool, 0);
}
zil_free_commit_waiter(zcw);
}
/*
* Called in syncing context to free committed log blocks and update log header.
*/
void
zil_sync(zilog_t *zilog, dmu_tx_t *tx)
{
zil_header_t *zh = zil_header_in_syncing_context(zilog);
uint64_t txg = dmu_tx_get_txg(tx);
spa_t *spa = zilog->zl_spa;
uint64_t *replayed_seq = &zilog->zl_replayed_seq[txg & TXG_MASK];
lwb_t *lwb;
/*
* We don't zero out zl_destroy_txg, so make sure we don't try
* to destroy it twice.
*/
if (spa_sync_pass(spa) != 1)
return;
mutex_enter(&zilog->zl_lock);
ASSERT(zilog->zl_stop_sync == 0);
if (*replayed_seq != 0) {
ASSERT(zh->zh_replay_seq < *replayed_seq);
zh->zh_replay_seq = *replayed_seq;
*replayed_seq = 0;
}
if (zilog->zl_destroy_txg == txg) {
blkptr_t blk = zh->zh_log;
ASSERT(list_head(&zilog->zl_lwb_list) == NULL);
bzero(zh, sizeof (zil_header_t));
bzero(zilog->zl_replayed_seq, sizeof (zilog->zl_replayed_seq));
if (zilog->zl_keep_first) {
/*
* If this block was part of log chain that couldn't
* be claimed because a device was missing during
* zil_claim(), but that device later returns,
* then this block could erroneously appear valid.
* To guard against this, assign a new GUID to the new
* log chain so it doesn't matter what blk points to.
*/
zil_init_log_chain(zilog, &blk);
zh->zh_log = blk;
}
}
while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) {
zh->zh_log = lwb->lwb_blk;
if (lwb->lwb_buf != NULL || lwb->lwb_max_txg > txg)
break;
list_remove(&zilog->zl_lwb_list, lwb);
zio_free(spa, txg, &lwb->lwb_blk);
zil_free_lwb(zilog, lwb);
/*
* If we don't have anything left in the lwb list then
* we've had an allocation failure and we need to zero
* out the zil_header blkptr so that we don't end
* up freeing the same block twice.
*/
if (list_head(&zilog->zl_lwb_list) == NULL)
BP_ZERO(&zh->zh_log);
}
/*
* Remove fastwrite on any blocks that have been pre-allocated for
* the next commit. This prevents fastwrite counter pollution by
* unused, long-lived LWBs.
*/
for (; lwb != NULL; lwb = list_next(&zilog->zl_lwb_list, lwb)) {
if (lwb->lwb_fastwrite && !lwb->lwb_write_zio) {
metaslab_fastwrite_unmark(zilog->zl_spa, &lwb->lwb_blk);
lwb->lwb_fastwrite = 0;
}
}
mutex_exit(&zilog->zl_lock);
}
static int
zil_lwb_cons(void *vbuf, void *unused, int kmflag)
{
(void) unused, (void) kmflag;
lwb_t *lwb = vbuf;
list_create(&lwb->lwb_itxs, sizeof (itx_t), offsetof(itx_t, itx_node));
list_create(&lwb->lwb_waiters, sizeof (zil_commit_waiter_t),
offsetof(zil_commit_waiter_t, zcw_node));
avl_create(&lwb->lwb_vdev_tree, zil_lwb_vdev_compare,
sizeof (zil_vdev_node_t), offsetof(zil_vdev_node_t, zv_node));
mutex_init(&lwb->lwb_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
return (0);
}
static void
zil_lwb_dest(void *vbuf, void *unused)
{
(void) unused;
lwb_t *lwb = vbuf;
mutex_destroy(&lwb->lwb_vdev_lock);
avl_destroy(&lwb->lwb_vdev_tree);
list_destroy(&lwb->lwb_waiters);
list_destroy(&lwb->lwb_itxs);
}
void
zil_init(void)
{
zil_lwb_cache = kmem_cache_create("zil_lwb_cache",
sizeof (lwb_t), 0, zil_lwb_cons, zil_lwb_dest, NULL, NULL, NULL, 0);
zil_zcw_cache = kmem_cache_create("zil_zcw_cache",
sizeof (zil_commit_waiter_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
zil_ksp = kstat_create("zfs", 0, "zil", "misc",
KSTAT_TYPE_NAMED, sizeof (zil_stats) / sizeof (kstat_named_t),
KSTAT_FLAG_VIRTUAL);
if (zil_ksp != NULL) {
zil_ksp->ks_data = &zil_stats;
kstat_install(zil_ksp);
}
}
void
zil_fini(void)
{
kmem_cache_destroy(zil_zcw_cache);
kmem_cache_destroy(zil_lwb_cache);
if (zil_ksp != NULL) {
kstat_delete(zil_ksp);
zil_ksp = NULL;
}
}
void
zil_set_sync(zilog_t *zilog, uint64_t sync)
{
zilog->zl_sync = sync;
}
void
zil_set_logbias(zilog_t *zilog, uint64_t logbias)
{
zilog->zl_logbias = logbias;
}
zilog_t *
zil_alloc(objset_t *os, zil_header_t *zh_phys)
{
zilog_t *zilog;
zilog = kmem_zalloc(sizeof (zilog_t), KM_SLEEP);
zilog->zl_header = zh_phys;
zilog->zl_os = os;
zilog->zl_spa = dmu_objset_spa(os);
zilog->zl_dmu_pool = dmu_objset_pool(os);
zilog->zl_destroy_txg = TXG_INITIAL - 1;
zilog->zl_logbias = dmu_objset_logbias(os);
zilog->zl_sync = dmu_objset_syncprop(os);
zilog->zl_dirty_max_txg = 0;
zilog->zl_last_lwb_opened = NULL;
zilog->zl_last_lwb_latency = 0;
zilog->zl_max_block_size = zil_maxblocksize;
mutex_init(&zilog->zl_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&zilog->zl_issuer_lock, NULL, MUTEX_DEFAULT, NULL);
for (int i = 0; i < TXG_SIZE; i++) {
mutex_init(&zilog->zl_itxg[i].itxg_lock, NULL,
MUTEX_DEFAULT, NULL);
}
list_create(&zilog->zl_lwb_list, sizeof (lwb_t),
offsetof(lwb_t, lwb_node));
list_create(&zilog->zl_itx_commit_list, sizeof (itx_t),
offsetof(itx_t, itx_node));
cv_init(&zilog->zl_cv_suspend, NULL, CV_DEFAULT, NULL);
return (zilog);
}
void
zil_free(zilog_t *zilog)
{
int i;
zilog->zl_stop_sync = 1;
ASSERT0(zilog->zl_suspend);
ASSERT0(zilog->zl_suspending);
ASSERT(list_is_empty(&zilog->zl_lwb_list));
list_destroy(&zilog->zl_lwb_list);
ASSERT(list_is_empty(&zilog->zl_itx_commit_list));
list_destroy(&zilog->zl_itx_commit_list);
for (i = 0; i < TXG_SIZE; i++) {
/*
* It's possible for an itx to be generated that doesn't dirty
* a txg (e.g. ztest TX_TRUNCATE). So there's no zil_clean()
* callback to remove the entry. We remove those here.
*
* Also free up the ziltest itxs.
*/
if (zilog->zl_itxg[i].itxg_itxs)
zil_itxg_clean(zilog->zl_itxg[i].itxg_itxs);
mutex_destroy(&zilog->zl_itxg[i].itxg_lock);
}
mutex_destroy(&zilog->zl_issuer_lock);
mutex_destroy(&zilog->zl_lock);
cv_destroy(&zilog->zl_cv_suspend);
kmem_free(zilog, sizeof (zilog_t));
}
/*
* Open an intent log.
*/
zilog_t *
zil_open(objset_t *os, zil_get_data_t *get_data)
{
zilog_t *zilog = dmu_objset_zil(os);
ASSERT3P(zilog->zl_get_data, ==, NULL);
ASSERT3P(zilog->zl_last_lwb_opened, ==, NULL);
ASSERT(list_is_empty(&zilog->zl_lwb_list));
zilog->zl_get_data = get_data;
return (zilog);
}
/*
* Close an intent log.
*/
void
zil_close(zilog_t *zilog)
{
lwb_t *lwb;
uint64_t txg;
if (!dmu_objset_is_snapshot(zilog->zl_os)) {
zil_commit(zilog, 0);
} else {
ASSERT3P(list_tail(&zilog->zl_lwb_list), ==, NULL);
ASSERT0(zilog->zl_dirty_max_txg);
ASSERT3B(zilog_is_dirty(zilog), ==, B_FALSE);
}
mutex_enter(&zilog->zl_lock);
lwb = list_tail(&zilog->zl_lwb_list);
if (lwb == NULL)
txg = zilog->zl_dirty_max_txg;
else
txg = MAX(zilog->zl_dirty_max_txg, lwb->lwb_max_txg);
mutex_exit(&zilog->zl_lock);
/*
* We need to use txg_wait_synced() to wait long enough for the
* ZIL to be clean, and to wait for all pending lwbs to be
* written out.
*/
if (txg != 0)
txg_wait_synced(zilog->zl_dmu_pool, txg);
if (zilog_is_dirty(zilog))
zfs_dbgmsg("zil (%px) is dirty, txg %llu", zilog,
(u_longlong_t)txg);
if (txg < spa_freeze_txg(zilog->zl_spa))
VERIFY(!zilog_is_dirty(zilog));
zilog->zl_get_data = NULL;
/*
* We should have only one lwb left on the list; remove it now.
*/
mutex_enter(&zilog->zl_lock);
lwb = list_head(&zilog->zl_lwb_list);
if (lwb != NULL) {
ASSERT3P(lwb, ==, list_tail(&zilog->zl_lwb_list));
ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED);
if (lwb->lwb_fastwrite)
metaslab_fastwrite_unmark(zilog->zl_spa, &lwb->lwb_blk);
list_remove(&zilog->zl_lwb_list, lwb);
zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
zil_free_lwb(zilog, lwb);
}
mutex_exit(&zilog->zl_lock);
}
static char *suspend_tag = "zil suspending";
/*
* Suspend an intent log. While in suspended mode, we still honor
* synchronous semantics, but we rely on txg_wait_synced() to do it.
* On old version pools, we suspend the log briefly when taking a
* snapshot so that it will have an empty intent log.
*
* Long holds are not really intended to be used the way we do here --
* held for such a short time. A concurrent caller of dsl_dataset_long_held()
* could fail. Therefore we take pains to only put a long hold if it is
* actually necessary. Fortunately, it will only be necessary if the
* objset is currently mounted (or the ZVOL equivalent). In that case it
* will already have a long hold, so we are not really making things any worse.
*
* Ideally, we would locate the existing long-holder (i.e. the zfsvfs_t or
* zvol_state_t), and use their mechanism to prevent their hold from being
* dropped (e.g. VFS_HOLD()). However, that would be even more pain for
* very little gain.
*
* if cookiep == NULL, this does both the suspend & resume.
* Otherwise, it returns with the dataset "long held", and the cookie
* should be passed into zil_resume().
*/
int
zil_suspend(const char *osname, void **cookiep)
{
objset_t *os;
zilog_t *zilog;
const zil_header_t *zh;
int error;
error = dmu_objset_hold(osname, suspend_tag, &os);
if (error != 0)
return (error);
zilog = dmu_objset_zil(os);
mutex_enter(&zilog->zl_lock);
zh = zilog->zl_header;
if (zh->zh_flags & ZIL_REPLAY_NEEDED) { /* unplayed log */
mutex_exit(&zilog->zl_lock);
dmu_objset_rele(os, suspend_tag);
return (SET_ERROR(EBUSY));
}
/*
* Don't put a long hold in the cases where we can avoid it. This
* is when there is no cookie so we are doing a suspend & resume
* (i.e. called from zil_vdev_offline()), and there's nothing to do
* for the suspend because it's already suspended, or there's no ZIL.
*/
if (cookiep == NULL && !zilog->zl_suspending &&
(zilog->zl_suspend > 0 || BP_IS_HOLE(&zh->zh_log))) {
mutex_exit(&zilog->zl_lock);
dmu_objset_rele(os, suspend_tag);
return (0);
}
dsl_dataset_long_hold(dmu_objset_ds(os), suspend_tag);
dsl_pool_rele(dmu_objset_pool(os), suspend_tag);
zilog->zl_suspend++;
if (zilog->zl_suspend > 1) {
/*
* Someone else is already suspending it.
* Just wait for them to finish.
*/
while (zilog->zl_suspending)
cv_wait(&zilog->zl_cv_suspend, &zilog->zl_lock);
mutex_exit(&zilog->zl_lock);
if (cookiep == NULL)
zil_resume(os);
else
*cookiep = os;
return (0);
}
/*
* If there is no pointer to an on-disk block, this ZIL must not
* be active (e.g. filesystem not mounted), so there's nothing
* to clean up.
*/
if (BP_IS_HOLE(&zh->zh_log)) {
ASSERT(cookiep != NULL); /* fast path already handled */
*cookiep = os;
mutex_exit(&zilog->zl_lock);
return (0);
}
/*
* The ZIL has work to do. Ensure that the associated encryption
* key will remain mapped while we are committing the log by
* grabbing a reference to it. If the key isn't loaded we have no
* choice but to return an error until the wrapping key is loaded.
*/
if (os->os_encrypted &&
dsl_dataset_create_key_mapping(dmu_objset_ds(os)) != 0) {
zilog->zl_suspend--;
mutex_exit(&zilog->zl_lock);
dsl_dataset_long_rele(dmu_objset_ds(os), suspend_tag);
dsl_dataset_rele(dmu_objset_ds(os), suspend_tag);
return (SET_ERROR(EACCES));
}
zilog->zl_suspending = B_TRUE;
mutex_exit(&zilog->zl_lock);
/*
* We need to use zil_commit_impl to ensure we wait for all
* LWB_STATE_OPENED and LWB_STATE_ISSUED lwbs to be committed
* to disk before proceeding. If we used zil_commit instead, it
* would just call txg_wait_synced(), because zl_suspend is set.
* txg_wait_synced() doesn't wait for these lwb's to be
* LWB_STATE_FLUSH_DONE before returning.
*/
zil_commit_impl(zilog, 0);
/*
* Now that we've ensured all lwb's are LWB_STATE_FLUSH_DONE, we
* use txg_wait_synced() to ensure the data from the zilog has
* migrated to the main pool before calling zil_destroy().
*/
txg_wait_synced(zilog->zl_dmu_pool, 0);
zil_destroy(zilog, B_FALSE);
mutex_enter(&zilog->zl_lock);
zilog->zl_suspending = B_FALSE;
cv_broadcast(&zilog->zl_cv_suspend);
mutex_exit(&zilog->zl_lock);
if (os->os_encrypted)
dsl_dataset_remove_key_mapping(dmu_objset_ds(os));
if (cookiep == NULL)
zil_resume(os);
else
*cookiep = os;
return (0);
}
void
zil_resume(void *cookie)
{
objset_t *os = cookie;
zilog_t *zilog = dmu_objset_zil(os);
mutex_enter(&zilog->zl_lock);
ASSERT(zilog->zl_suspend != 0);
zilog->zl_suspend--;
mutex_exit(&zilog->zl_lock);
dsl_dataset_long_rele(dmu_objset_ds(os), suspend_tag);
dsl_dataset_rele(dmu_objset_ds(os), suspend_tag);
}
typedef struct zil_replay_arg {
- zil_replay_func_t **zr_replay;
+ zil_replay_func_t *const *zr_replay;
void *zr_arg;
boolean_t zr_byteswap;
char *zr_lr;
} zil_replay_arg_t;
static int
zil_replay_error(zilog_t *zilog, const lr_t *lr, int error)
{
char name[ZFS_MAX_DATASET_NAME_LEN];
zilog->zl_replaying_seq--; /* didn't actually replay this one */
dmu_objset_name(zilog->zl_os, name);
cmn_err(CE_WARN, "ZFS replay transaction error %d, "
"dataset %s, seq 0x%llx, txtype %llu %s\n", error, name,
(u_longlong_t)lr->lrc_seq,
(u_longlong_t)(lr->lrc_txtype & ~TX_CI),
(lr->lrc_txtype & TX_CI) ? "CI" : "");
return (error);
}
static int
zil_replay_log_record(zilog_t *zilog, const lr_t *lr, void *zra,
uint64_t claim_txg)
{
zil_replay_arg_t *zr = zra;
const zil_header_t *zh = zilog->zl_header;
uint64_t reclen = lr->lrc_reclen;
uint64_t txtype = lr->lrc_txtype;
int error = 0;
zilog->zl_replaying_seq = lr->lrc_seq;
if (lr->lrc_seq <= zh->zh_replay_seq) /* already replayed */
return (0);
if (lr->lrc_txg < claim_txg) /* already committed */
return (0);
/* Strip case-insensitive bit, still present in log record */
txtype &= ~TX_CI;
if (txtype == 0 || txtype >= TX_MAX_TYPE)
return (zil_replay_error(zilog, lr, EINVAL));
/*
* If this record type can be logged out of order, the object
* (lr_foid) may no longer exist. That's legitimate, not an error.
*/
if (TX_OOO(txtype)) {
error = dmu_object_info(zilog->zl_os,
LR_FOID_GET_OBJ(((lr_ooo_t *)lr)->lr_foid), NULL);
if (error == ENOENT || error == EEXIST)
return (0);
}
/*
* Make a copy of the data so we can revise and extend it.
*/
bcopy(lr, zr->zr_lr, reclen);
/*
* If this is a TX_WRITE with a blkptr, suck in the data.
*/
if (txtype == TX_WRITE && reclen == sizeof (lr_write_t)) {
error = zil_read_log_data(zilog, (lr_write_t *)lr,
zr->zr_lr + reclen);
if (error != 0)
return (zil_replay_error(zilog, lr, error));
}
/*
* The log block containing this lr may have been byteswapped
* so that we can easily examine common fields like lrc_txtype.
* However, the log is a mix of different record types, and only the
* replay vectors know how to byteswap their records. Therefore, if
* the lr was byteswapped, undo it before invoking the replay vector.
*/
if (zr->zr_byteswap)
byteswap_uint64_array(zr->zr_lr, reclen);
/*
* We must now do two things atomically: replay this log record,
* and update the log header sequence number to reflect the fact that
* we did so. At the end of each replay function the sequence number
* is updated if we are in replay mode.
*/
error = zr->zr_replay[txtype](zr->zr_arg, zr->zr_lr, zr->zr_byteswap);
if (error != 0) {
/*
* The DMU's dnode layer doesn't see removes until the txg
* commits, so a subsequent claim can spuriously fail with
* EEXIST. So if we receive any error we try syncing out
* any removes then retry the transaction. Note that we
* specify B_FALSE for byteswap now, so we don't do it twice.
*/
txg_wait_synced(spa_get_dsl(zilog->zl_spa), 0);
error = zr->zr_replay[txtype](zr->zr_arg, zr->zr_lr, B_FALSE);
if (error != 0)
return (zil_replay_error(zilog, lr, error));
}
return (0);
}
static int
zil_incr_blks(zilog_t *zilog, const blkptr_t *bp, void *arg, uint64_t claim_txg)
{
(void) bp, (void) arg, (void) claim_txg;
zilog->zl_replay_blks++;
return (0);
}
/*
* If this dataset has a non-empty intent log, replay it and destroy it.
*/
void
-zil_replay(objset_t *os, void *arg, zil_replay_func_t *replay_func[TX_MAX_TYPE])
+zil_replay(objset_t *os, void *arg,
+ zil_replay_func_t *const replay_func[TX_MAX_TYPE])
{
zilog_t *zilog = dmu_objset_zil(os);
const zil_header_t *zh = zilog->zl_header;
zil_replay_arg_t zr;
if ((zh->zh_flags & ZIL_REPLAY_NEEDED) == 0) {
zil_destroy(zilog, B_TRUE);
return;
}
zr.zr_replay = replay_func;
zr.zr_arg = arg;
zr.zr_byteswap = BP_SHOULD_BYTESWAP(&zh->zh_log);
zr.zr_lr = vmem_alloc(2 * SPA_MAXBLOCKSIZE, KM_SLEEP);
/*
* Wait for in-progress removes to sync before starting replay.
*/
txg_wait_synced(zilog->zl_dmu_pool, 0);
zilog->zl_replay = B_TRUE;
zilog->zl_replay_time = ddi_get_lbolt();
ASSERT(zilog->zl_replay_blks == 0);
(void) zil_parse(zilog, zil_incr_blks, zil_replay_log_record, &zr,
zh->zh_claim_txg, B_TRUE);
vmem_free(zr.zr_lr, 2 * SPA_MAXBLOCKSIZE);
zil_destroy(zilog, B_FALSE);
txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);
zilog->zl_replay = B_FALSE;
}
boolean_t
zil_replaying(zilog_t *zilog, dmu_tx_t *tx)
{
if (zilog->zl_sync == ZFS_SYNC_DISABLED)
return (B_TRUE);
if (zilog->zl_replay) {
dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
zilog->zl_replayed_seq[dmu_tx_get_txg(tx) & TXG_MASK] =
zilog->zl_replaying_seq;
return (B_TRUE);
}
return (B_FALSE);
}
int
zil_reset(const char *osname, void *arg)
{
(void) arg;
int error = zil_suspend(osname, NULL);
/* EACCES means crypto key not loaded */
if ((error == EACCES) || (error == EBUSY))
return (SET_ERROR(error));
if (error != 0)
return (SET_ERROR(EEXIST));
return (0);
}
EXPORT_SYMBOL(zil_alloc);
EXPORT_SYMBOL(zil_free);
EXPORT_SYMBOL(zil_open);
EXPORT_SYMBOL(zil_close);
EXPORT_SYMBOL(zil_replay);
EXPORT_SYMBOL(zil_replaying);
EXPORT_SYMBOL(zil_destroy);
EXPORT_SYMBOL(zil_destroy_sync);
EXPORT_SYMBOL(zil_itx_create);
EXPORT_SYMBOL(zil_itx_destroy);
EXPORT_SYMBOL(zil_itx_assign);
EXPORT_SYMBOL(zil_commit);
EXPORT_SYMBOL(zil_claim);
EXPORT_SYMBOL(zil_check_log_chain);
EXPORT_SYMBOL(zil_sync);
EXPORT_SYMBOL(zil_clean);
EXPORT_SYMBOL(zil_suspend);
EXPORT_SYMBOL(zil_resume);
EXPORT_SYMBOL(zil_lwb_add_block);
EXPORT_SYMBOL(zil_bp_tree_add);
EXPORT_SYMBOL(zil_set_sync);
EXPORT_SYMBOL(zil_set_logbias);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, zfs_, commit_timeout_pct, INT, ZMOD_RW,
"ZIL block open timeout percentage");
ZFS_MODULE_PARAM(zfs_zil, zil_, replay_disable, INT, ZMOD_RW,
"Disable intent logging replay");
ZFS_MODULE_PARAM(zfs_zil, zil_, nocacheflush, INT, ZMOD_RW,
"Disable ZIL cache flushes");
ZFS_MODULE_PARAM(zfs_zil, zil_, slog_bulk, ULONG, ZMOD_RW,
"Limit in bytes slog sync writes per commit");
ZFS_MODULE_PARAM(zfs_zil, zil_, maxblocksize, INT, ZMOD_RW,
"Limit in bytes of ZIL log block size");
/* END CSTYLED */
diff --git a/module/zfs/zio.c b/module/zfs/zio.c
index e04be06bf315..92115488c544 100644
--- a/module/zfs/zio.c
+++ b/module/zfs/zio.c
@@ -1,5058 +1,5058 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
* Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2017, Intel Corporation.
* Copyright (c) 2019, Klara Inc.
* Copyright (c) 2019, Allan Jude
* Copyright (c) 2021, Datto, Inc.
*/
#include <sys/sysmacros.h>
#include <sys/zfs_context.h>
#include <sys/fm/fs/zfs.h>
#include <sys/spa.h>
#include <sys/txg.h>
#include <sys/spa_impl.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_trim.h>
#include <sys/zio_impl.h>
#include <sys/zio_compress.h>
#include <sys/zio_checksum.h>
#include <sys/dmu_objset.h>
#include <sys/arc.h>
#include <sys/ddt.h>
#include <sys/blkptr.h>
#include <sys/zfeature.h>
#include <sys/dsl_scan.h>
#include <sys/metaslab_impl.h>
#include <sys/time.h>
#include <sys/trace_zfs.h>
#include <sys/abd.h>
#include <sys/dsl_crypt.h>
#include <cityhash.h>
/*
* ==========================================================================
* I/O type descriptions
* ==========================================================================
*/
-const char *zio_type_name[ZIO_TYPES] = {
+const char *const zio_type_name[ZIO_TYPES] = {
/*
* Note: Linux kernel thread name length is limited
* so these names will differ from upstream open zfs.
*/
"z_null", "z_rd", "z_wr", "z_fr", "z_cl", "z_ioctl", "z_trim"
};
int zio_dva_throttle_enabled = B_TRUE;
-int zio_deadman_log_all = B_FALSE;
+static int zio_deadman_log_all = B_FALSE;
/*
* ==========================================================================
* I/O kmem caches
* ==========================================================================
*/
-kmem_cache_t *zio_cache;
-kmem_cache_t *zio_link_cache;
+static kmem_cache_t *zio_cache;
+static kmem_cache_t *zio_link_cache;
kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
#if defined(ZFS_DEBUG) && !defined(_KERNEL)
-uint64_t zio_buf_cache_allocs[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
-uint64_t zio_buf_cache_frees[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
+static uint64_t zio_buf_cache_allocs[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
+static uint64_t zio_buf_cache_frees[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
#endif
/* Mark IOs as "slow" if they take longer than 30 seconds */
-int zio_slow_io_ms = (30 * MILLISEC);
+static int zio_slow_io_ms = (30 * MILLISEC);
#define BP_SPANB(indblkshift, level) \
(((uint64_t)1) << ((level) * ((indblkshift) - SPA_BLKPTRSHIFT)))
#define COMPARE_META_LEVEL 0x80000000ul
/*
* The following actions directly effect the spa's sync-to-convergence logic.
* The values below define the sync pass when we start performing the action.
* Care should be taken when changing these values as they directly impact
* spa_sync() performance. Tuning these values may introduce subtle performance
* pathologies and should only be done in the context of performance analysis.
* These tunables will eventually be removed and replaced with #defines once
* enough analysis has been done to determine optimal values.
*
* The 'zfs_sync_pass_deferred_free' pass must be greater than 1 to ensure that
* regular blocks are not deferred.
*
* Starting in sync pass 8 (zfs_sync_pass_dont_compress), we disable
* compression (including of metadata). In practice, we don't have this
* many sync passes, so this has no effect.
*
* The original intent was that disabling compression would help the sync
* passes to converge. However, in practice disabling compression increases
* the average number of sync passes, because when we turn compression off, a
* lot of block's size will change and thus we have to re-allocate (not
* overwrite) them. It also increases the number of 128KB allocations (e.g.
* for indirect blocks and spacemaps) because these will not be compressed.
* The 128K allocations are especially detrimental to performance on highly
* fragmented systems, which may have very few free segments of this size,
* and may need to load new metaslabs to satisfy 128K allocations.
*/
int zfs_sync_pass_deferred_free = 2; /* defer frees starting in this pass */
-int zfs_sync_pass_dont_compress = 8; /* don't compress starting in this pass */
-int zfs_sync_pass_rewrite = 2; /* rewrite new bps starting in this pass */
+static int zfs_sync_pass_dont_compress = 8; /* don't compress s. i. t. p. */
+static int zfs_sync_pass_rewrite = 2; /* rewrite new bps s. i. t. p. */
/*
* An allocating zio is one that either currently has the DVA allocate
* stage set or will have it later in its lifetime.
*/
#define IO_IS_ALLOCATING(zio) ((zio)->io_orig_pipeline & ZIO_STAGE_DVA_ALLOCATE)
/*
* Enable smaller cores by excluding metadata
* allocations as well.
*/
int zio_exclude_metadata = 0;
-int zio_requeue_io_start_cut_in_line = 1;
+static int zio_requeue_io_start_cut_in_line = 1;
#ifdef ZFS_DEBUG
-int zio_buf_debug_limit = 16384;
+static const int zio_buf_debug_limit = 16384;
#else
-int zio_buf_debug_limit = 0;
+static const int zio_buf_debug_limit = 0;
#endif
static inline void __zio_execute(zio_t *zio);
static void zio_taskq_dispatch(zio_t *, zio_taskq_type_t, boolean_t);
void
zio_init(void)
{
size_t c;
zio_cache = kmem_cache_create("zio_cache",
sizeof (zio_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
zio_link_cache = kmem_cache_create("zio_link_cache",
sizeof (zio_link_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
/*
* For small buffers, we want a cache for each multiple of
* SPA_MINBLOCKSIZE. For larger buffers, we want a cache
* for each quarter-power of 2.
*/
for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) {
size_t size = (c + 1) << SPA_MINBLOCKSHIFT;
size_t p2 = size;
size_t align = 0;
size_t data_cflags, cflags;
data_cflags = KMC_NODEBUG;
cflags = (zio_exclude_metadata || size > zio_buf_debug_limit) ?
KMC_NODEBUG : 0;
#if defined(_ILP32) && defined(_KERNEL)
/*
* Cache size limited to 1M on 32-bit platforms until ARC
* buffers no longer require virtual address space.
*/
if (size > zfs_max_recordsize)
break;
#endif
while (!ISP2(p2))
p2 &= p2 - 1;
#ifndef _KERNEL
/*
* If we are using watchpoints, put each buffer on its own page,
* to eliminate the performance overhead of trapping to the
* kernel when modifying a non-watched buffer that shares the
* page with a watched buffer.
*/
if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE))
continue;
/*
* Here's the problem - on 4K native devices in userland on
* Linux using O_DIRECT, buffers must be 4K aligned or I/O
* will fail with EINVAL, causing zdb (and others) to coredump.
* Since userland probably doesn't need optimized buffer caches,
* we just force 4K alignment on everything.
*/
align = 8 * SPA_MINBLOCKSIZE;
#else
if (size < PAGESIZE) {
align = SPA_MINBLOCKSIZE;
} else if (IS_P2ALIGNED(size, p2 >> 2)) {
align = PAGESIZE;
}
#endif
if (align != 0) {
char name[36];
if (cflags == data_cflags) {
/*
* Resulting kmem caches would be identical.
* Save memory by creating only one.
*/
(void) snprintf(name, sizeof (name),
"zio_buf_comb_%lu", (ulong_t)size);
zio_buf_cache[c] = kmem_cache_create(name,
size, align, NULL, NULL, NULL, NULL, NULL,
cflags);
zio_data_buf_cache[c] = zio_buf_cache[c];
continue;
}
(void) snprintf(name, sizeof (name), "zio_buf_%lu",
(ulong_t)size);
zio_buf_cache[c] = kmem_cache_create(name, size,
align, NULL, NULL, NULL, NULL, NULL, cflags);
(void) snprintf(name, sizeof (name), "zio_data_buf_%lu",
(ulong_t)size);
zio_data_buf_cache[c] = kmem_cache_create(name, size,
align, NULL, NULL, NULL, NULL, NULL, data_cflags);
}
}
while (--c != 0) {
ASSERT(zio_buf_cache[c] != NULL);
if (zio_buf_cache[c - 1] == NULL)
zio_buf_cache[c - 1] = zio_buf_cache[c];
ASSERT(zio_data_buf_cache[c] != NULL);
if (zio_data_buf_cache[c - 1] == NULL)
zio_data_buf_cache[c - 1] = zio_data_buf_cache[c];
}
zio_inject_init();
lz4_init();
}
void
zio_fini(void)
{
size_t n = SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT;
#if defined(ZFS_DEBUG) && !defined(_KERNEL)
for (size_t i = 0; i < n; i++) {
if (zio_buf_cache_allocs[i] != zio_buf_cache_frees[i])
(void) printf("zio_fini: [%d] %llu != %llu\n",
(int)((i + 1) << SPA_MINBLOCKSHIFT),
(long long unsigned)zio_buf_cache_allocs[i],
(long long unsigned)zio_buf_cache_frees[i]);
}
#endif
/*
* The same kmem cache can show up multiple times in both zio_buf_cache
* and zio_data_buf_cache. Do a wasteful but trivially correct scan to
* sort it out.
*/
for (size_t i = 0; i < n; i++) {
kmem_cache_t *cache = zio_buf_cache[i];
if (cache == NULL)
continue;
for (size_t j = i; j < n; j++) {
if (cache == zio_buf_cache[j])
zio_buf_cache[j] = NULL;
if (cache == zio_data_buf_cache[j])
zio_data_buf_cache[j] = NULL;
}
kmem_cache_destroy(cache);
}
for (size_t i = 0; i < n; i++) {
kmem_cache_t *cache = zio_data_buf_cache[i];
if (cache == NULL)
continue;
for (size_t j = i; j < n; j++) {
if (cache == zio_data_buf_cache[j])
zio_data_buf_cache[j] = NULL;
}
kmem_cache_destroy(cache);
}
for (size_t i = 0; i < n; i++) {
VERIFY3P(zio_buf_cache[i], ==, NULL);
VERIFY3P(zio_data_buf_cache[i], ==, NULL);
}
kmem_cache_destroy(zio_link_cache);
kmem_cache_destroy(zio_cache);
zio_inject_fini();
lz4_fini();
}
/*
* ==========================================================================
* Allocate and free I/O buffers
* ==========================================================================
*/
/*
* Use zio_buf_alloc to allocate ZFS metadata. This data will appear in a
* crashdump if the kernel panics, so use it judiciously. Obviously, it's
* useful to inspect ZFS metadata, but if possible, we should avoid keeping
* excess / transient data in-core during a crashdump.
*/
void *
zio_buf_alloc(size_t size)
{
size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
#if defined(ZFS_DEBUG) && !defined(_KERNEL)
atomic_add_64(&zio_buf_cache_allocs[c], 1);
#endif
return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE));
}
/*
* Use zio_data_buf_alloc to allocate data. The data will not appear in a
* crashdump if the kernel panics. This exists so that we will limit the amount
* of ZFS data that shows up in a kernel crashdump. (Thus reducing the amount
* of kernel heap dumped to disk when the kernel panics)
*/
void *
zio_data_buf_alloc(size_t size)
{
size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE));
}
void
zio_buf_free(void *buf, size_t size)
{
size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
#if defined(ZFS_DEBUG) && !defined(_KERNEL)
atomic_add_64(&zio_buf_cache_frees[c], 1);
#endif
kmem_cache_free(zio_buf_cache[c], buf);
}
void
zio_data_buf_free(void *buf, size_t size)
{
size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
kmem_cache_free(zio_data_buf_cache[c], buf);
}
static void
zio_abd_free(void *abd, size_t size)
{
(void) size;
abd_free((abd_t *)abd);
}
/*
* ==========================================================================
* Push and pop I/O transform buffers
* ==========================================================================
*/
void
zio_push_transform(zio_t *zio, abd_t *data, uint64_t size, uint64_t bufsize,
zio_transform_func_t *transform)
{
zio_transform_t *zt = kmem_alloc(sizeof (zio_transform_t), KM_SLEEP);
zt->zt_orig_abd = zio->io_abd;
zt->zt_orig_size = zio->io_size;
zt->zt_bufsize = bufsize;
zt->zt_transform = transform;
zt->zt_next = zio->io_transform_stack;
zio->io_transform_stack = zt;
zio->io_abd = data;
zio->io_size = size;
}
void
zio_pop_transforms(zio_t *zio)
{
zio_transform_t *zt;
while ((zt = zio->io_transform_stack) != NULL) {
if (zt->zt_transform != NULL)
zt->zt_transform(zio,
zt->zt_orig_abd, zt->zt_orig_size);
if (zt->zt_bufsize != 0)
abd_free(zio->io_abd);
zio->io_abd = zt->zt_orig_abd;
zio->io_size = zt->zt_orig_size;
zio->io_transform_stack = zt->zt_next;
kmem_free(zt, sizeof (zio_transform_t));
}
}
/*
* ==========================================================================
* I/O transform callbacks for subblocks, decompression, and decryption
* ==========================================================================
*/
static void
zio_subblock(zio_t *zio, abd_t *data, uint64_t size)
{
ASSERT(zio->io_size > size);
if (zio->io_type == ZIO_TYPE_READ)
abd_copy(data, zio->io_abd, size);
}
static void
zio_decompress(zio_t *zio, abd_t *data, uint64_t size)
{
if (zio->io_error == 0) {
void *tmp = abd_borrow_buf(data, size);
int ret = zio_decompress_data(BP_GET_COMPRESS(zio->io_bp),
zio->io_abd, tmp, zio->io_size, size,
&zio->io_prop.zp_complevel);
abd_return_buf_copy(data, tmp, size);
if (zio_injection_enabled && ret == 0)
ret = zio_handle_fault_injection(zio, EINVAL);
if (ret != 0)
zio->io_error = SET_ERROR(EIO);
}
}
static void
zio_decrypt(zio_t *zio, abd_t *data, uint64_t size)
{
int ret;
void *tmp;
blkptr_t *bp = zio->io_bp;
spa_t *spa = zio->io_spa;
uint64_t dsobj = zio->io_bookmark.zb_objset;
uint64_t lsize = BP_GET_LSIZE(bp);
dmu_object_type_t ot = BP_GET_TYPE(bp);
uint8_t salt[ZIO_DATA_SALT_LEN];
uint8_t iv[ZIO_DATA_IV_LEN];
uint8_t mac[ZIO_DATA_MAC_LEN];
boolean_t no_crypt = B_FALSE;
ASSERT(BP_USES_CRYPT(bp));
ASSERT3U(size, !=, 0);
if (zio->io_error != 0)
return;
/*
* Verify the cksum of MACs stored in an indirect bp. It will always
* be possible to verify this since it does not require an encryption
* key.
*/
if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) {
zio_crypt_decode_mac_bp(bp, mac);
if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) {
/*
* We haven't decompressed the data yet, but
* zio_crypt_do_indirect_mac_checksum() requires
* decompressed data to be able to parse out the MACs
* from the indirect block. We decompress it now and
* throw away the result after we are finished.
*/
tmp = zio_buf_alloc(lsize);
ret = zio_decompress_data(BP_GET_COMPRESS(bp),
zio->io_abd, tmp, zio->io_size, lsize,
&zio->io_prop.zp_complevel);
if (ret != 0) {
ret = SET_ERROR(EIO);
goto error;
}
ret = zio_crypt_do_indirect_mac_checksum(B_FALSE,
tmp, lsize, BP_SHOULD_BYTESWAP(bp), mac);
zio_buf_free(tmp, lsize);
} else {
ret = zio_crypt_do_indirect_mac_checksum_abd(B_FALSE,
zio->io_abd, size, BP_SHOULD_BYTESWAP(bp), mac);
}
abd_copy(data, zio->io_abd, size);
if (zio_injection_enabled && ot != DMU_OT_DNODE && ret == 0) {
ret = zio_handle_decrypt_injection(spa,
&zio->io_bookmark, ot, ECKSUM);
}
if (ret != 0)
goto error;
return;
}
/*
* If this is an authenticated block, just check the MAC. It would be
* nice to separate this out into its own flag, but for the moment
* enum zio_flag is out of bits.
*/
if (BP_IS_AUTHENTICATED(bp)) {
if (ot == DMU_OT_OBJSET) {
ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa,
dsobj, zio->io_abd, size, BP_SHOULD_BYTESWAP(bp));
} else {
zio_crypt_decode_mac_bp(bp, mac);
ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj,
zio->io_abd, size, mac);
if (zio_injection_enabled && ret == 0) {
ret = zio_handle_decrypt_injection(spa,
&zio->io_bookmark, ot, ECKSUM);
}
}
abd_copy(data, zio->io_abd, size);
if (ret != 0)
goto error;
return;
}
zio_crypt_decode_params_bp(bp, salt, iv);
if (ot == DMU_OT_INTENT_LOG) {
tmp = abd_borrow_buf_copy(zio->io_abd, sizeof (zil_chain_t));
zio_crypt_decode_mac_zil(tmp, mac);
abd_return_buf(zio->io_abd, tmp, sizeof (zil_chain_t));
} else {
zio_crypt_decode_mac_bp(bp, mac);
}
ret = spa_do_crypt_abd(B_FALSE, spa, &zio->io_bookmark, BP_GET_TYPE(bp),
BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, size, data,
zio->io_abd, &no_crypt);
if (no_crypt)
abd_copy(data, zio->io_abd, size);
if (ret != 0)
goto error;
return;
error:
/* assert that the key was found unless this was speculative */
ASSERT(ret != EACCES || (zio->io_flags & ZIO_FLAG_SPECULATIVE));
/*
* If there was a decryption / authentication error return EIO as
* the io_error. If this was not a speculative zio, create an ereport.
*/
if (ret == ECKSUM) {
zio->io_error = SET_ERROR(EIO);
if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) {
spa_log_error(spa, &zio->io_bookmark);
(void) zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION,
spa, NULL, &zio->io_bookmark, zio, 0);
}
} else {
zio->io_error = ret;
}
}
/*
* ==========================================================================
* I/O parent/child relationships and pipeline interlocks
* ==========================================================================
*/
zio_t *
zio_walk_parents(zio_t *cio, zio_link_t **zl)
{
list_t *pl = &cio->io_parent_list;
*zl = (*zl == NULL) ? list_head(pl) : list_next(pl, *zl);
if (*zl == NULL)
return (NULL);
ASSERT((*zl)->zl_child == cio);
return ((*zl)->zl_parent);
}
zio_t *
zio_walk_children(zio_t *pio, zio_link_t **zl)
{
list_t *cl = &pio->io_child_list;
ASSERT(MUTEX_HELD(&pio->io_lock));
*zl = (*zl == NULL) ? list_head(cl) : list_next(cl, *zl);
if (*zl == NULL)
return (NULL);
ASSERT((*zl)->zl_parent == pio);
return ((*zl)->zl_child);
}
zio_t *
zio_unique_parent(zio_t *cio)
{
zio_link_t *zl = NULL;
zio_t *pio = zio_walk_parents(cio, &zl);
VERIFY3P(zio_walk_parents(cio, &zl), ==, NULL);
return (pio);
}
void
zio_add_child(zio_t *pio, zio_t *cio)
{
zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP);
/*
* Logical I/Os can have logical, gang, or vdev children.
* Gang I/Os can have gang or vdev children.
* Vdev I/Os can only have vdev children.
* The following ASSERT captures all of these constraints.
*/
ASSERT3S(cio->io_child_type, <=, pio->io_child_type);
zl->zl_parent = pio;
zl->zl_child = cio;
mutex_enter(&pio->io_lock);
mutex_enter(&cio->io_lock);
ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0);
for (int w = 0; w < ZIO_WAIT_TYPES; w++)
pio->io_children[cio->io_child_type][w] += !cio->io_state[w];
list_insert_head(&pio->io_child_list, zl);
list_insert_head(&cio->io_parent_list, zl);
pio->io_child_count++;
cio->io_parent_count++;
mutex_exit(&cio->io_lock);
mutex_exit(&pio->io_lock);
}
static void
zio_remove_child(zio_t *pio, zio_t *cio, zio_link_t *zl)
{
ASSERT(zl->zl_parent == pio);
ASSERT(zl->zl_child == cio);
mutex_enter(&pio->io_lock);
mutex_enter(&cio->io_lock);
list_remove(&pio->io_child_list, zl);
list_remove(&cio->io_parent_list, zl);
pio->io_child_count--;
cio->io_parent_count--;
mutex_exit(&cio->io_lock);
mutex_exit(&pio->io_lock);
kmem_cache_free(zio_link_cache, zl);
}
static boolean_t
zio_wait_for_children(zio_t *zio, uint8_t childbits, enum zio_wait_type wait)
{
boolean_t waiting = B_FALSE;
mutex_enter(&zio->io_lock);
ASSERT(zio->io_stall == NULL);
for (int c = 0; c < ZIO_CHILD_TYPES; c++) {
if (!(ZIO_CHILD_BIT_IS_SET(childbits, c)))
continue;
uint64_t *countp = &zio->io_children[c][wait];
if (*countp != 0) {
zio->io_stage >>= 1;
ASSERT3U(zio->io_stage, !=, ZIO_STAGE_OPEN);
zio->io_stall = countp;
waiting = B_TRUE;
break;
}
}
mutex_exit(&zio->io_lock);
return (waiting);
}
__attribute__((always_inline))
static inline void
zio_notify_parent(zio_t *pio, zio_t *zio, enum zio_wait_type wait,
zio_t **next_to_executep)
{
uint64_t *countp = &pio->io_children[zio->io_child_type][wait];
int *errorp = &pio->io_child_error[zio->io_child_type];
mutex_enter(&pio->io_lock);
if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE))
*errorp = zio_worst_error(*errorp, zio->io_error);
pio->io_reexecute |= zio->io_reexecute;
ASSERT3U(*countp, >, 0);
(*countp)--;
if (*countp == 0 && pio->io_stall == countp) {
zio_taskq_type_t type =
pio->io_stage < ZIO_STAGE_VDEV_IO_START ? ZIO_TASKQ_ISSUE :
ZIO_TASKQ_INTERRUPT;
pio->io_stall = NULL;
mutex_exit(&pio->io_lock);
/*
* If we can tell the caller to execute this parent next, do
* so. Otherwise dispatch the parent zio as its own task.
*
* Having the caller execute the parent when possible reduces
* locking on the zio taskq's, reduces context switch
* overhead, and has no recursion penalty. Note that one
* read from disk typically causes at least 3 zio's: a
* zio_null(), the logical zio_read(), and then a physical
* zio. When the physical ZIO completes, we are able to call
* zio_done() on all 3 of these zio's from one invocation of
* zio_execute() by returning the parent back to
* zio_execute(). Since the parent isn't executed until this
* thread returns back to zio_execute(), the caller should do
* so promptly.
*
* In other cases, dispatching the parent prevents
* overflowing the stack when we have deeply nested
* parent-child relationships, as we do with the "mega zio"
* of writes for spa_sync(), and the chain of ZIL blocks.
*/
if (next_to_executep != NULL && *next_to_executep == NULL) {
*next_to_executep = pio;
} else {
zio_taskq_dispatch(pio, type, B_FALSE);
}
} else {
mutex_exit(&pio->io_lock);
}
}
static void
zio_inherit_child_errors(zio_t *zio, enum zio_child c)
{
if (zio->io_child_error[c] != 0 && zio->io_error == 0)
zio->io_error = zio->io_child_error[c];
}
int
zio_bookmark_compare(const void *x1, const void *x2)
{
const zio_t *z1 = x1;
const zio_t *z2 = x2;
if (z1->io_bookmark.zb_objset < z2->io_bookmark.zb_objset)
return (-1);
if (z1->io_bookmark.zb_objset > z2->io_bookmark.zb_objset)
return (1);
if (z1->io_bookmark.zb_object < z2->io_bookmark.zb_object)
return (-1);
if (z1->io_bookmark.zb_object > z2->io_bookmark.zb_object)
return (1);
if (z1->io_bookmark.zb_level < z2->io_bookmark.zb_level)
return (-1);
if (z1->io_bookmark.zb_level > z2->io_bookmark.zb_level)
return (1);
if (z1->io_bookmark.zb_blkid < z2->io_bookmark.zb_blkid)
return (-1);
if (z1->io_bookmark.zb_blkid > z2->io_bookmark.zb_blkid)
return (1);
if (z1 < z2)
return (-1);
if (z1 > z2)
return (1);
return (0);
}
/*
* ==========================================================================
* Create the various types of I/O (read, write, free, etc)
* ==========================================================================
*/
static zio_t *
zio_create(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp,
abd_t *data, uint64_t lsize, uint64_t psize, zio_done_func_t *done,
void *private, zio_type_t type, zio_priority_t priority,
enum zio_flag flags, vdev_t *vd, uint64_t offset,
const zbookmark_phys_t *zb, enum zio_stage stage,
enum zio_stage pipeline)
{
zio_t *zio;
IMPLY(type != ZIO_TYPE_TRIM, psize <= SPA_MAXBLOCKSIZE);
ASSERT(P2PHASE(psize, SPA_MINBLOCKSIZE) == 0);
ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0);
ASSERT(!vd || spa_config_held(spa, SCL_STATE_ALL, RW_READER));
ASSERT(!bp || !(flags & ZIO_FLAG_CONFIG_WRITER));
ASSERT(vd || stage == ZIO_STAGE_OPEN);
IMPLY(lsize != psize, (flags & ZIO_FLAG_RAW_COMPRESS) != 0);
zio = kmem_cache_alloc(zio_cache, KM_SLEEP);
bzero(zio, sizeof (zio_t));
mutex_init(&zio->io_lock, NULL, MUTEX_NOLOCKDEP, NULL);
cv_init(&zio->io_cv, NULL, CV_DEFAULT, NULL);
list_create(&zio->io_parent_list, sizeof (zio_link_t),
offsetof(zio_link_t, zl_parent_node));
list_create(&zio->io_child_list, sizeof (zio_link_t),
offsetof(zio_link_t, zl_child_node));
metaslab_trace_init(&zio->io_alloc_list);
if (vd != NULL)
zio->io_child_type = ZIO_CHILD_VDEV;
else if (flags & ZIO_FLAG_GANG_CHILD)
zio->io_child_type = ZIO_CHILD_GANG;
else if (flags & ZIO_FLAG_DDT_CHILD)
zio->io_child_type = ZIO_CHILD_DDT;
else
zio->io_child_type = ZIO_CHILD_LOGICAL;
if (bp != NULL) {
zio->io_bp = (blkptr_t *)bp;
zio->io_bp_copy = *bp;
zio->io_bp_orig = *bp;
if (type != ZIO_TYPE_WRITE ||
zio->io_child_type == ZIO_CHILD_DDT)
zio->io_bp = &zio->io_bp_copy; /* so caller can free */
if (zio->io_child_type == ZIO_CHILD_LOGICAL)
zio->io_logical = zio;
if (zio->io_child_type > ZIO_CHILD_GANG && BP_IS_GANG(bp))
pipeline |= ZIO_GANG_STAGES;
}
zio->io_spa = spa;
zio->io_txg = txg;
zio->io_done = done;
zio->io_private = private;
zio->io_type = type;
zio->io_priority = priority;
zio->io_vd = vd;
zio->io_offset = offset;
zio->io_orig_abd = zio->io_abd = data;
zio->io_orig_size = zio->io_size = psize;
zio->io_lsize = lsize;
zio->io_orig_flags = zio->io_flags = flags;
zio->io_orig_stage = zio->io_stage = stage;
zio->io_orig_pipeline = zio->io_pipeline = pipeline;
zio->io_pipeline_trace = ZIO_STAGE_OPEN;
zio->io_state[ZIO_WAIT_READY] = (stage >= ZIO_STAGE_READY);
zio->io_state[ZIO_WAIT_DONE] = (stage >= ZIO_STAGE_DONE);
if (zb != NULL)
zio->io_bookmark = *zb;
if (pio != NULL) {
zio->io_metaslab_class = pio->io_metaslab_class;
if (zio->io_logical == NULL)
zio->io_logical = pio->io_logical;
if (zio->io_child_type == ZIO_CHILD_GANG)
zio->io_gang_leader = pio->io_gang_leader;
zio_add_child(pio, zio);
}
taskq_init_ent(&zio->io_tqent);
return (zio);
}
static void
zio_destroy(zio_t *zio)
{
metaslab_trace_fini(&zio->io_alloc_list);
list_destroy(&zio->io_parent_list);
list_destroy(&zio->io_child_list);
mutex_destroy(&zio->io_lock);
cv_destroy(&zio->io_cv);
kmem_cache_free(zio_cache, zio);
}
zio_t *
zio_null(zio_t *pio, spa_t *spa, vdev_t *vd, zio_done_func_t *done,
void *private, enum zio_flag flags)
{
zio_t *zio;
zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private,
ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL,
ZIO_STAGE_OPEN, ZIO_INTERLOCK_PIPELINE);
return (zio);
}
zio_t *
zio_root(spa_t *spa, zio_done_func_t *done, void *private, enum zio_flag flags)
{
return (zio_null(NULL, spa, NULL, done, private, flags));
}
static int
zfs_blkptr_verify_log(spa_t *spa, const blkptr_t *bp,
enum blk_verify_flag blk_verify, const char *fmt, ...)
{
va_list adx;
char buf[256];
va_start(adx, fmt);
(void) vsnprintf(buf, sizeof (buf), fmt, adx);
va_end(adx);
switch (blk_verify) {
case BLK_VERIFY_HALT:
dprintf_bp(bp, "blkptr at %p dprintf_bp():", bp);
zfs_panic_recover("%s: %s", spa_name(spa), buf);
break;
case BLK_VERIFY_LOG:
zfs_dbgmsg("%s: %s", spa_name(spa), buf);
break;
case BLK_VERIFY_ONLY:
break;
}
return (1);
}
/*
* Verify the block pointer fields contain reasonable values. This means
* it only contains known object types, checksum/compression identifiers,
* block sizes within the maximum allowed limits, valid DVAs, etc.
*
* If everything checks out B_TRUE is returned. The zfs_blkptr_verify
* argument controls the behavior when an invalid field is detected.
*
* Modes for zfs_blkptr_verify:
* 1) BLK_VERIFY_ONLY (evaluate the block)
* 2) BLK_VERIFY_LOG (evaluate the block and log problems)
* 3) BLK_VERIFY_HALT (call zfs_panic_recover on error)
*/
boolean_t
zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp, boolean_t config_held,
enum blk_verify_flag blk_verify)
{
int errors = 0;
if (!DMU_OT_IS_VALID(BP_GET_TYPE(bp))) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p has invalid TYPE %llu",
bp, (longlong_t)BP_GET_TYPE(bp));
}
if (BP_GET_CHECKSUM(bp) >= ZIO_CHECKSUM_FUNCTIONS ||
BP_GET_CHECKSUM(bp) <= ZIO_CHECKSUM_ON) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p has invalid CHECKSUM %llu",
bp, (longlong_t)BP_GET_CHECKSUM(bp));
}
if (BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_FUNCTIONS ||
BP_GET_COMPRESS(bp) <= ZIO_COMPRESS_ON) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p has invalid COMPRESS %llu",
bp, (longlong_t)BP_GET_COMPRESS(bp));
}
if (BP_GET_LSIZE(bp) > SPA_MAXBLOCKSIZE) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p has invalid LSIZE %llu",
bp, (longlong_t)BP_GET_LSIZE(bp));
}
if (BP_GET_PSIZE(bp) > SPA_MAXBLOCKSIZE) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p has invalid PSIZE %llu",
bp, (longlong_t)BP_GET_PSIZE(bp));
}
if (BP_IS_EMBEDDED(bp)) {
if (BPE_GET_ETYPE(bp) >= NUM_BP_EMBEDDED_TYPES) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p has invalid ETYPE %llu",
bp, (longlong_t)BPE_GET_ETYPE(bp));
}
}
/*
* Do not verify individual DVAs if the config is not trusted. This
* will be done once the zio is executed in vdev_mirror_map_alloc.
*/
if (!spa->spa_trust_config)
return (errors == 0);
if (!config_held)
spa_config_enter(spa, SCL_VDEV, bp, RW_READER);
else
ASSERT(spa_config_held(spa, SCL_VDEV, RW_WRITER));
/*
* Pool-specific checks.
*
* Note: it would be nice to verify that the blk_birth and
* BP_PHYSICAL_BIRTH() are not too large. However, spa_freeze()
* allows the birth time of log blocks (and dmu_sync()-ed blocks
* that are in the log) to be arbitrarily large.
*/
for (int i = 0; i < BP_GET_NDVAS(bp); i++) {
const dva_t *dva = &bp->blk_dva[i];
uint64_t vdevid = DVA_GET_VDEV(dva);
if (vdevid >= spa->spa_root_vdev->vdev_children) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p DVA %u has invalid VDEV %llu",
bp, i, (longlong_t)vdevid);
continue;
}
vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid];
if (vd == NULL) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p DVA %u has invalid VDEV %llu",
bp, i, (longlong_t)vdevid);
continue;
}
if (vd->vdev_ops == &vdev_hole_ops) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p DVA %u has hole VDEV %llu",
bp, i, (longlong_t)vdevid);
continue;
}
if (vd->vdev_ops == &vdev_missing_ops) {
/*
* "missing" vdevs are valid during import, but we
* don't have their detailed info (e.g. asize), so
* we can't perform any more checks on them.
*/
continue;
}
uint64_t offset = DVA_GET_OFFSET(dva);
uint64_t asize = DVA_GET_ASIZE(dva);
if (DVA_GET_GANG(dva))
asize = vdev_gang_header_asize(vd);
if (offset + asize > vd->vdev_asize) {
errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
"blkptr at %p DVA %u has invalid OFFSET %llu",
bp, i, (longlong_t)offset);
}
}
if (errors > 0)
dprintf_bp(bp, "blkptr at %p dprintf_bp():", bp);
if (!config_held)
spa_config_exit(spa, SCL_VDEV, bp);
return (errors == 0);
}
boolean_t
zfs_dva_valid(spa_t *spa, const dva_t *dva, const blkptr_t *bp)
{
(void) bp;
uint64_t vdevid = DVA_GET_VDEV(dva);
if (vdevid >= spa->spa_root_vdev->vdev_children)
return (B_FALSE);
vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid];
if (vd == NULL)
return (B_FALSE);
if (vd->vdev_ops == &vdev_hole_ops)
return (B_FALSE);
if (vd->vdev_ops == &vdev_missing_ops) {
return (B_FALSE);
}
uint64_t offset = DVA_GET_OFFSET(dva);
uint64_t asize = DVA_GET_ASIZE(dva);
if (DVA_GET_GANG(dva))
asize = vdev_gang_header_asize(vd);
if (offset + asize > vd->vdev_asize)
return (B_FALSE);
return (B_TRUE);
}
zio_t *
zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
abd_t *data, uint64_t size, zio_done_func_t *done, void *private,
zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb)
{
zio_t *zio;
zio = zio_create(pio, spa, BP_PHYSICAL_BIRTH(bp), bp,
data, size, size, done, private,
ZIO_TYPE_READ, priority, flags, NULL, 0, zb,
ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ?
ZIO_DDT_CHILD_READ_PIPELINE : ZIO_READ_PIPELINE);
return (zio);
}
zio_t *
zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
abd_t *data, uint64_t lsize, uint64_t psize, const zio_prop_t *zp,
zio_done_func_t *ready, zio_done_func_t *children_ready,
zio_done_func_t *physdone, zio_done_func_t *done,
void *private, zio_priority_t priority, enum zio_flag flags,
const zbookmark_phys_t *zb)
{
zio_t *zio;
ASSERT(zp->zp_checksum >= ZIO_CHECKSUM_OFF &&
zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS &&
zp->zp_compress >= ZIO_COMPRESS_OFF &&
zp->zp_compress < ZIO_COMPRESS_FUNCTIONS &&
DMU_OT_IS_VALID(zp->zp_type) &&
zp->zp_level < 32 &&
zp->zp_copies > 0 &&
zp->zp_copies <= spa_max_replication(spa));
zio = zio_create(pio, spa, txg, bp, data, lsize, psize, done, private,
ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb,
ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ?
ZIO_DDT_CHILD_WRITE_PIPELINE : ZIO_WRITE_PIPELINE);
zio->io_ready = ready;
zio->io_children_ready = children_ready;
zio->io_physdone = physdone;
zio->io_prop = *zp;
/*
* Data can be NULL if we are going to call zio_write_override() to
* provide the already-allocated BP. But we may need the data to
* verify a dedup hit (if requested). In this case, don't try to
* dedup (just take the already-allocated BP verbatim). Encrypted
* dedup blocks need data as well so we also disable dedup in this
* case.
*/
if (data == NULL &&
(zio->io_prop.zp_dedup_verify || zio->io_prop.zp_encrypt)) {
zio->io_prop.zp_dedup = zio->io_prop.zp_dedup_verify = B_FALSE;
}
return (zio);
}
zio_t *
zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, abd_t *data,
uint64_t size, zio_done_func_t *done, void *private,
zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb)
{
zio_t *zio;
zio = zio_create(pio, spa, txg, bp, data, size, size, done, private,
ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_IO_REWRITE, NULL, 0, zb,
ZIO_STAGE_OPEN, ZIO_REWRITE_PIPELINE);
return (zio);
}
void
zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite)
{
ASSERT(zio->io_type == ZIO_TYPE_WRITE);
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
ASSERT(zio->io_stage == ZIO_STAGE_OPEN);
ASSERT(zio->io_txg == spa_syncing_txg(zio->io_spa));
/*
* We must reset the io_prop to match the values that existed
* when the bp was first written by dmu_sync() keeping in mind
* that nopwrite and dedup are mutually exclusive.
*/
zio->io_prop.zp_dedup = nopwrite ? B_FALSE : zio->io_prop.zp_dedup;
zio->io_prop.zp_nopwrite = nopwrite;
zio->io_prop.zp_copies = copies;
zio->io_bp_override = bp;
}
void
zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp)
{
(void) zfs_blkptr_verify(spa, bp, B_FALSE, BLK_VERIFY_HALT);
/*
* The check for EMBEDDED is a performance optimization. We
* process the free here (by ignoring it) rather than
* putting it on the list and then processing it in zio_free_sync().
*/
if (BP_IS_EMBEDDED(bp))
return;
metaslab_check_free(spa, bp);
/*
* Frees that are for the currently-syncing txg, are not going to be
* deferred, and which will not need to do a read (i.e. not GANG or
* DEDUP), can be processed immediately. Otherwise, put them on the
* in-memory list for later processing.
*
* Note that we only defer frees after zfs_sync_pass_deferred_free
* when the log space map feature is disabled. [see relevant comment
* in spa_sync_iterate_to_convergence()]
*/
if (BP_IS_GANG(bp) ||
BP_GET_DEDUP(bp) ||
txg != spa->spa_syncing_txg ||
(spa_sync_pass(spa) >= zfs_sync_pass_deferred_free &&
!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))) {
bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp);
} else {
VERIFY3P(zio_free_sync(NULL, spa, txg, bp, 0), ==, NULL);
}
}
/*
* To improve performance, this function may return NULL if we were able
* to do the free immediately. This avoids the cost of creating a zio
* (and linking it to the parent, etc).
*/
zio_t *
zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp,
enum zio_flag flags)
{
ASSERT(!BP_IS_HOLE(bp));
ASSERT(spa_syncing_txg(spa) == txg);
if (BP_IS_EMBEDDED(bp))
return (NULL);
metaslab_check_free(spa, bp);
arc_freed(spa, bp);
dsl_scan_freed(spa, bp);
if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp)) {
/*
* GANG and DEDUP blocks can induce a read (for the gang block
* header, or the DDT), so issue them asynchronously so that
* this thread is not tied up.
*/
enum zio_stage stage =
ZIO_FREE_PIPELINE | ZIO_STAGE_ISSUE_ASYNC;
return (zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
BP_GET_PSIZE(bp), NULL, NULL,
ZIO_TYPE_FREE, ZIO_PRIORITY_NOW,
flags, NULL, 0, NULL, ZIO_STAGE_OPEN, stage));
} else {
metaslab_free(spa, bp, txg, B_FALSE);
return (NULL);
}
}
zio_t *
zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp,
zio_done_func_t *done, void *private, enum zio_flag flags)
{
zio_t *zio;
(void) zfs_blkptr_verify(spa, bp, flags & ZIO_FLAG_CONFIG_WRITER,
BLK_VERIFY_HALT);
if (BP_IS_EMBEDDED(bp))
return (zio_null(pio, spa, NULL, NULL, NULL, 0));
/*
* A claim is an allocation of a specific block. Claims are needed
* to support immediate writes in the intent log. The issue is that
* immediate writes contain committed data, but in a txg that was
* *not* committed. Upon opening the pool after an unclean shutdown,
* the intent log claims all blocks that contain immediate write data
* so that the SPA knows they're in use.
*
* All claims *must* be resolved in the first txg -- before the SPA
* starts allocating blocks -- so that nothing is allocated twice.
* If txg == 0 we just verify that the block is claimable.
*/
ASSERT3U(spa->spa_uberblock.ub_rootbp.blk_birth, <,
spa_min_claim_txg(spa));
ASSERT(txg == spa_min_claim_txg(spa) || txg == 0);
ASSERT(!BP_GET_DEDUP(bp) || !spa_writeable(spa)); /* zdb(8) */
zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
BP_GET_PSIZE(bp), done, private, ZIO_TYPE_CLAIM, ZIO_PRIORITY_NOW,
flags, NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_CLAIM_PIPELINE);
ASSERT0(zio->io_queued_timestamp);
return (zio);
}
zio_t *
zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
zio_done_func_t *done, void *private, enum zio_flag flags)
{
zio_t *zio;
int c;
if (vd->vdev_children == 0) {
zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private,
ZIO_TYPE_IOCTL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL,
ZIO_STAGE_OPEN, ZIO_IOCTL_PIPELINE);
zio->io_cmd = cmd;
} else {
zio = zio_null(pio, spa, NULL, NULL, NULL, flags);
for (c = 0; c < vd->vdev_children; c++)
zio_nowait(zio_ioctl(zio, spa, vd->vdev_child[c], cmd,
done, private, flags));
}
return (zio);
}
zio_t *
zio_trim(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
zio_done_func_t *done, void *private, zio_priority_t priority,
enum zio_flag flags, enum trim_flag trim_flags)
{
zio_t *zio;
ASSERT0(vd->vdev_children);
ASSERT0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
ASSERT0(P2PHASE(size, 1ULL << vd->vdev_ashift));
ASSERT3U(size, !=, 0);
zio = zio_create(pio, vd->vdev_spa, 0, NULL, NULL, size, size, done,
private, ZIO_TYPE_TRIM, priority, flags | ZIO_FLAG_PHYSICAL,
vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_TRIM_PIPELINE);
zio->io_trim_flags = trim_flags;
return (zio);
}
zio_t *
zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
abd_t *data, int checksum, zio_done_func_t *done, void *private,
zio_priority_t priority, enum zio_flag flags, boolean_t labels)
{
zio_t *zio;
ASSERT(vd->vdev_children == 0);
ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE ||
offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
ASSERT3U(offset + size, <=, vd->vdev_psize);
zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done,
private, ZIO_TYPE_READ, priority, flags | ZIO_FLAG_PHYSICAL, vd,
offset, NULL, ZIO_STAGE_OPEN, ZIO_READ_PHYS_PIPELINE);
zio->io_prop.zp_checksum = checksum;
return (zio);
}
zio_t *
zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
abd_t *data, int checksum, zio_done_func_t *done, void *private,
zio_priority_t priority, enum zio_flag flags, boolean_t labels)
{
zio_t *zio;
ASSERT(vd->vdev_children == 0);
ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE ||
offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
ASSERT3U(offset + size, <=, vd->vdev_psize);
zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done,
private, ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_PHYSICAL, vd,
offset, NULL, ZIO_STAGE_OPEN, ZIO_WRITE_PHYS_PIPELINE);
zio->io_prop.zp_checksum = checksum;
if (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_EMBEDDED) {
/*
* zec checksums are necessarily destructive -- they modify
* the end of the write buffer to hold the verifier/checksum.
* Therefore, we must make a local copy in case the data is
* being written to multiple places in parallel.
*/
abd_t *wbuf = abd_alloc_sametype(data, size);
abd_copy(wbuf, data, size);
zio_push_transform(zio, wbuf, size, size, NULL);
}
return (zio);
}
/*
* Create a child I/O to do some work for us.
*/
zio_t *
zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset,
abd_t *data, uint64_t size, int type, zio_priority_t priority,
enum zio_flag flags, zio_done_func_t *done, void *private)
{
enum zio_stage pipeline = ZIO_VDEV_CHILD_PIPELINE;
zio_t *zio;
/*
* vdev child I/Os do not propagate their error to the parent.
* Therefore, for correct operation the caller *must* check for
* and handle the error in the child i/o's done callback.
* The only exceptions are i/os that we don't care about
* (OPTIONAL or REPAIR).
*/
ASSERT((flags & ZIO_FLAG_OPTIONAL) || (flags & ZIO_FLAG_IO_REPAIR) ||
done != NULL);
if (type == ZIO_TYPE_READ && bp != NULL) {
/*
* If we have the bp, then the child should perform the
* checksum and the parent need not. This pushes error
* detection as close to the leaves as possible and
* eliminates redundant checksums in the interior nodes.
*/
pipeline |= ZIO_STAGE_CHECKSUM_VERIFY;
pio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
}
if (vd->vdev_ops->vdev_op_leaf) {
ASSERT0(vd->vdev_children);
offset += VDEV_LABEL_START_SIZE;
}
flags |= ZIO_VDEV_CHILD_FLAGS(pio);
/*
* If we've decided to do a repair, the write is not speculative --
* even if the original read was.
*/
if (flags & ZIO_FLAG_IO_REPAIR)
flags &= ~ZIO_FLAG_SPECULATIVE;
/*
* If we're creating a child I/O that is not associated with a
* top-level vdev, then the child zio is not an allocating I/O.
* If this is a retried I/O then we ignore it since we will
* have already processed the original allocating I/O.
*/
if (flags & ZIO_FLAG_IO_ALLOCATING &&
(vd != vd->vdev_top || (flags & ZIO_FLAG_IO_RETRY))) {
ASSERT(pio->io_metaslab_class != NULL);
ASSERT(pio->io_metaslab_class->mc_alloc_throttle_enabled);
ASSERT(type == ZIO_TYPE_WRITE);
ASSERT(priority == ZIO_PRIORITY_ASYNC_WRITE);
ASSERT(!(flags & ZIO_FLAG_IO_REPAIR));
ASSERT(!(pio->io_flags & ZIO_FLAG_IO_REWRITE) ||
pio->io_child_type == ZIO_CHILD_GANG);
flags &= ~ZIO_FLAG_IO_ALLOCATING;
}
zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size, size,
done, private, type, priority, flags, vd, offset, &pio->io_bookmark,
ZIO_STAGE_VDEV_IO_START >> 1, pipeline);
ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV);
zio->io_physdone = pio->io_physdone;
if (vd->vdev_ops->vdev_op_leaf && zio->io_logical != NULL)
zio->io_logical->io_phys_children++;
return (zio);
}
zio_t *
zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, abd_t *data, uint64_t size,
zio_type_t type, zio_priority_t priority, enum zio_flag flags,
zio_done_func_t *done, void *private)
{
zio_t *zio;
ASSERT(vd->vdev_ops->vdev_op_leaf);
zio = zio_create(NULL, vd->vdev_spa, 0, NULL,
data, size, size, done, private, type, priority,
flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_RETRY | ZIO_FLAG_DELEGATED,
vd, offset, NULL,
ZIO_STAGE_VDEV_IO_START >> 1, ZIO_VDEV_CHILD_PIPELINE);
return (zio);
}
void
zio_flush(zio_t *zio, vdev_t *vd)
{
zio_nowait(zio_ioctl(zio, zio->io_spa, vd, DKIOCFLUSHWRITECACHE,
NULL, NULL,
ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY));
}
void
zio_shrink(zio_t *zio, uint64_t size)
{
ASSERT3P(zio->io_executor, ==, NULL);
ASSERT3U(zio->io_orig_size, ==, zio->io_size);
ASSERT3U(size, <=, zio->io_size);
/*
* We don't shrink for raidz because of problems with the
* reconstruction when reading back less than the block size.
* Note, BP_IS_RAIDZ() assumes no compression.
*/
ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF);
if (!BP_IS_RAIDZ(zio->io_bp)) {
/* we are not doing a raw write */
ASSERT3U(zio->io_size, ==, zio->io_lsize);
zio->io_orig_size = zio->io_size = zio->io_lsize = size;
}
}
/*
* ==========================================================================
* Prepare to read and write logical blocks
* ==========================================================================
*/
static zio_t *
zio_read_bp_init(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
uint64_t psize =
BP_IS_EMBEDDED(bp) ? BPE_GET_PSIZE(bp) : BP_GET_PSIZE(bp);
ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);
if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF &&
zio->io_child_type == ZIO_CHILD_LOGICAL &&
!(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) {
zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize),
psize, psize, zio_decompress);
}
if (((BP_IS_PROTECTED(bp) && !(zio->io_flags & ZIO_FLAG_RAW_ENCRYPT)) ||
BP_HAS_INDIRECT_MAC_CKSUM(bp)) &&
zio->io_child_type == ZIO_CHILD_LOGICAL) {
zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize),
psize, psize, zio_decrypt);
}
if (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA) {
int psize = BPE_GET_PSIZE(bp);
void *data = abd_borrow_buf(zio->io_abd, psize);
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
decode_embedded_bp_compressed(bp, data);
abd_return_buf_copy(zio->io_abd, data, psize);
} else {
ASSERT(!BP_IS_EMBEDDED(bp));
ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);
}
if (!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) && BP_GET_LEVEL(bp) == 0)
zio->io_flags |= ZIO_FLAG_DONT_CACHE;
if (BP_GET_TYPE(bp) == DMU_OT_DDT_ZAP)
zio->io_flags |= ZIO_FLAG_DONT_CACHE;
if (BP_GET_DEDUP(bp) && zio->io_child_type == ZIO_CHILD_LOGICAL)
zio->io_pipeline = ZIO_DDT_READ_PIPELINE;
return (zio);
}
static zio_t *
zio_write_bp_init(zio_t *zio)
{
if (!IO_IS_ALLOCATING(zio))
return (zio);
ASSERT(zio->io_child_type != ZIO_CHILD_DDT);
if (zio->io_bp_override) {
blkptr_t *bp = zio->io_bp;
zio_prop_t *zp = &zio->io_prop;
ASSERT(bp->blk_birth != zio->io_txg);
ASSERT(BP_GET_DEDUP(zio->io_bp_override) == 0);
*bp = *zio->io_bp_override;
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
if (BP_IS_EMBEDDED(bp))
return (zio);
/*
* If we've been overridden and nopwrite is set then
* set the flag accordingly to indicate that a nopwrite
* has already occurred.
*/
if (!BP_IS_HOLE(bp) && zp->zp_nopwrite) {
ASSERT(!zp->zp_dedup);
ASSERT3U(BP_GET_CHECKSUM(bp), ==, zp->zp_checksum);
zio->io_flags |= ZIO_FLAG_NOPWRITE;
return (zio);
}
ASSERT(!zp->zp_nopwrite);
if (BP_IS_HOLE(bp) || !zp->zp_dedup)
return (zio);
ASSERT((zio_checksum_table[zp->zp_checksum].ci_flags &
ZCHECKSUM_FLAG_DEDUP) || zp->zp_dedup_verify);
if (BP_GET_CHECKSUM(bp) == zp->zp_checksum &&
!zp->zp_encrypt) {
BP_SET_DEDUP(bp, 1);
zio->io_pipeline |= ZIO_STAGE_DDT_WRITE;
return (zio);
}
/*
* We were unable to handle this as an override bp, treat
* it as a regular write I/O.
*/
zio->io_bp_override = NULL;
*bp = zio->io_bp_orig;
zio->io_pipeline = zio->io_orig_pipeline;
}
return (zio);
}
static zio_t *
zio_write_compress(zio_t *zio)
{
spa_t *spa = zio->io_spa;
zio_prop_t *zp = &zio->io_prop;
enum zio_compress compress = zp->zp_compress;
blkptr_t *bp = zio->io_bp;
uint64_t lsize = zio->io_lsize;
uint64_t psize = zio->io_size;
int pass = 1;
/*
* If our children haven't all reached the ready stage,
* wait for them and then repeat this pipeline stage.
*/
if (zio_wait_for_children(zio, ZIO_CHILD_LOGICAL_BIT |
ZIO_CHILD_GANG_BIT, ZIO_WAIT_READY)) {
return (NULL);
}
if (!IO_IS_ALLOCATING(zio))
return (zio);
if (zio->io_children_ready != NULL) {
/*
* Now that all our children are ready, run the callback
* associated with this zio in case it wants to modify the
* data to be written.
*/
ASSERT3U(zp->zp_level, >, 0);
zio->io_children_ready(zio);
}
ASSERT(zio->io_child_type != ZIO_CHILD_DDT);
ASSERT(zio->io_bp_override == NULL);
if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg) {
/*
* We're rewriting an existing block, which means we're
* working on behalf of spa_sync(). For spa_sync() to
* converge, it must eventually be the case that we don't
* have to allocate new blocks. But compression changes
* the blocksize, which forces a reallocate, and makes
* convergence take longer. Therefore, after the first
* few passes, stop compressing to ensure convergence.
*/
pass = spa_sync_pass(spa);
ASSERT(zio->io_txg == spa_syncing_txg(spa));
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
ASSERT(!BP_GET_DEDUP(bp));
if (pass >= zfs_sync_pass_dont_compress)
compress = ZIO_COMPRESS_OFF;
/* Make sure someone doesn't change their mind on overwrites */
ASSERT(BP_IS_EMBEDDED(bp) || MIN(zp->zp_copies + BP_IS_GANG(bp),
spa_max_replication(spa)) == BP_GET_NDVAS(bp));
}
/* If it's a compressed write that is not raw, compress the buffer. */
if (compress != ZIO_COMPRESS_OFF &&
!(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) {
void *cbuf = zio_buf_alloc(lsize);
psize = zio_compress_data(compress, zio->io_abd, cbuf, lsize,
zp->zp_complevel);
if (psize == 0 || psize >= lsize) {
compress = ZIO_COMPRESS_OFF;
zio_buf_free(cbuf, lsize);
} else if (!zp->zp_dedup && !zp->zp_encrypt &&
psize <= BPE_PAYLOAD_SIZE &&
zp->zp_level == 0 && !DMU_OT_HAS_FILL(zp->zp_type) &&
spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA)) {
encode_embedded_bp_compressed(bp,
cbuf, compress, lsize, psize);
BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_DATA);
BP_SET_TYPE(bp, zio->io_prop.zp_type);
BP_SET_LEVEL(bp, zio->io_prop.zp_level);
zio_buf_free(cbuf, lsize);
bp->blk_birth = zio->io_txg;
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
ASSERT(spa_feature_is_active(spa,
SPA_FEATURE_EMBEDDED_DATA));
return (zio);
} else {
/*
* Round compressed size up to the minimum allocation
* size of the smallest-ashift device, and zero the
* tail. This ensures that the compressed size of the
* BP (and thus compressratio property) are correct,
* in that we charge for the padding used to fill out
* the last sector.
*/
ASSERT3U(spa->spa_min_alloc, >=, SPA_MINBLOCKSHIFT);
size_t rounded = (size_t)roundup(psize,
spa->spa_min_alloc);
if (rounded >= lsize) {
compress = ZIO_COMPRESS_OFF;
zio_buf_free(cbuf, lsize);
psize = lsize;
} else {
abd_t *cdata = abd_get_from_buf(cbuf, lsize);
abd_take_ownership_of_buf(cdata, B_TRUE);
abd_zero_off(cdata, psize, rounded - psize);
psize = rounded;
zio_push_transform(zio, cdata,
psize, lsize, NULL);
}
}
/*
* We were unable to handle this as an override bp, treat
* it as a regular write I/O.
*/
zio->io_bp_override = NULL;
*bp = zio->io_bp_orig;
zio->io_pipeline = zio->io_orig_pipeline;
} else if ((zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) != 0 &&
zp->zp_type == DMU_OT_DNODE) {
/*
* The DMU actually relies on the zio layer's compression
* to free metadnode blocks that have had all contained
* dnodes freed. As a result, even when doing a raw
* receive, we must check whether the block can be compressed
* to a hole.
*/
psize = zio_compress_data(ZIO_COMPRESS_EMPTY,
zio->io_abd, NULL, lsize, zp->zp_complevel);
if (psize == 0 || psize >= lsize)
compress = ZIO_COMPRESS_OFF;
} else if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS) {
size_t rounded = MIN((size_t)roundup(psize,
spa->spa_min_alloc), lsize);
if (rounded != psize) {
abd_t *cdata = abd_alloc_linear(rounded, B_TRUE);
abd_zero_off(cdata, psize, rounded - psize);
abd_copy_off(cdata, zio->io_abd, 0, 0, psize);
psize = rounded;
zio_push_transform(zio, cdata,
psize, rounded, NULL);
}
} else {
ASSERT3U(psize, !=, 0);
}
/*
* The final pass of spa_sync() must be all rewrites, but the first
* few passes offer a trade-off: allocating blocks defers convergence,
* but newly allocated blocks are sequential, so they can be written
* to disk faster. Therefore, we allow the first few passes of
* spa_sync() to allocate new blocks, but force rewrites after that.
* There should only be a handful of blocks after pass 1 in any case.
*/
if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg &&
BP_GET_PSIZE(bp) == psize &&
pass >= zfs_sync_pass_rewrite) {
VERIFY3U(psize, !=, 0);
enum zio_stage gang_stages = zio->io_pipeline & ZIO_GANG_STAGES;
zio->io_pipeline = ZIO_REWRITE_PIPELINE | gang_stages;
zio->io_flags |= ZIO_FLAG_IO_REWRITE;
} else {
BP_ZERO(bp);
zio->io_pipeline = ZIO_WRITE_PIPELINE;
}
if (psize == 0) {
if (zio->io_bp_orig.blk_birth != 0 &&
spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) {
BP_SET_LSIZE(bp, lsize);
BP_SET_TYPE(bp, zp->zp_type);
BP_SET_LEVEL(bp, zp->zp_level);
BP_SET_BIRTH(bp, zio->io_txg, 0);
}
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
} else {
ASSERT(zp->zp_checksum != ZIO_CHECKSUM_GANG_HEADER);
BP_SET_LSIZE(bp, lsize);
BP_SET_TYPE(bp, zp->zp_type);
BP_SET_LEVEL(bp, zp->zp_level);
BP_SET_PSIZE(bp, psize);
BP_SET_COMPRESS(bp, compress);
BP_SET_CHECKSUM(bp, zp->zp_checksum);
BP_SET_DEDUP(bp, zp->zp_dedup);
BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);
if (zp->zp_dedup) {
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
ASSERT(!zp->zp_encrypt ||
DMU_OT_IS_ENCRYPTED(zp->zp_type));
zio->io_pipeline = ZIO_DDT_WRITE_PIPELINE;
}
if (zp->zp_nopwrite) {
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
zio->io_pipeline |= ZIO_STAGE_NOP_WRITE;
}
}
return (zio);
}
static zio_t *
zio_free_bp_init(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
if (zio->io_child_type == ZIO_CHILD_LOGICAL) {
if (BP_GET_DEDUP(bp))
zio->io_pipeline = ZIO_DDT_FREE_PIPELINE;
}
ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);
return (zio);
}
/*
* ==========================================================================
* Execute the I/O pipeline
* ==========================================================================
*/
static void
zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q, boolean_t cutinline)
{
spa_t *spa = zio->io_spa;
zio_type_t t = zio->io_type;
int flags = (cutinline ? TQ_FRONT : 0);
/*
* If we're a config writer or a probe, the normal issue and
* interrupt threads may all be blocked waiting for the config lock.
* In this case, select the otherwise-unused taskq for ZIO_TYPE_NULL.
*/
if (zio->io_flags & (ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_PROBE))
t = ZIO_TYPE_NULL;
/*
* A similar issue exists for the L2ARC write thread until L2ARC 2.0.
*/
if (t == ZIO_TYPE_WRITE && zio->io_vd && zio->io_vd->vdev_aux)
t = ZIO_TYPE_NULL;
/*
* If this is a high priority I/O, then use the high priority taskq if
* available.
*/
if ((zio->io_priority == ZIO_PRIORITY_NOW ||
zio->io_priority == ZIO_PRIORITY_SYNC_WRITE) &&
spa->spa_zio_taskq[t][q + 1].stqs_count != 0)
q++;
ASSERT3U(q, <, ZIO_TASKQ_TYPES);
/*
* NB: We are assuming that the zio can only be dispatched
* to a single taskq at a time. It would be a grievous error
* to dispatch the zio to another taskq at the same time.
*/
ASSERT(taskq_empty_ent(&zio->io_tqent));
spa_taskq_dispatch_ent(spa, t, q, zio_execute, zio, flags,
&zio->io_tqent);
}
static boolean_t
zio_taskq_member(zio_t *zio, zio_taskq_type_t q)
{
spa_t *spa = zio->io_spa;
taskq_t *tq = taskq_of_curthread();
for (zio_type_t t = 0; t < ZIO_TYPES; t++) {
spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
uint_t i;
for (i = 0; i < tqs->stqs_count; i++) {
if (tqs->stqs_taskq[i] == tq)
return (B_TRUE);
}
}
return (B_FALSE);
}
static zio_t *
zio_issue_async(zio_t *zio)
{
zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE);
return (NULL);
}
void
zio_interrupt(void *zio)
{
zio_taskq_dispatch(zio, ZIO_TASKQ_INTERRUPT, B_FALSE);
}
void
zio_delay_interrupt(zio_t *zio)
{
/*
* The timeout_generic() function isn't defined in userspace, so
* rather than trying to implement the function, the zio delay
* functionality has been disabled for userspace builds.
*/
#ifdef _KERNEL
/*
* If io_target_timestamp is zero, then no delay has been registered
* for this IO, thus jump to the end of this function and "skip" the
* delay; issuing it directly to the zio layer.
*/
if (zio->io_target_timestamp != 0) {
hrtime_t now = gethrtime();
if (now >= zio->io_target_timestamp) {
/*
* This IO has already taken longer than the target
* delay to complete, so we don't want to delay it
* any longer; we "miss" the delay and issue it
* directly to the zio layer. This is likely due to
* the target latency being set to a value less than
* the underlying hardware can satisfy (e.g. delay
* set to 1ms, but the disks take 10ms to complete an
* IO request).
*/
DTRACE_PROBE2(zio__delay__miss, zio_t *, zio,
hrtime_t, now);
zio_interrupt(zio);
} else {
taskqid_t tid;
hrtime_t diff = zio->io_target_timestamp - now;
clock_t expire_at_tick = ddi_get_lbolt() +
NSEC_TO_TICK(diff);
DTRACE_PROBE3(zio__delay__hit, zio_t *, zio,
hrtime_t, now, hrtime_t, diff);
if (NSEC_TO_TICK(diff) == 0) {
/* Our delay is less than a jiffy - just spin */
zfs_sleep_until(zio->io_target_timestamp);
zio_interrupt(zio);
} else {
/*
* Use taskq_dispatch_delay() in the place of
* OpenZFS's timeout_generic().
*/
tid = taskq_dispatch_delay(system_taskq,
zio_interrupt, zio, TQ_NOSLEEP,
expire_at_tick);
if (tid == TASKQID_INVALID) {
/*
* Couldn't allocate a task. Just
* finish the zio without a delay.
*/
zio_interrupt(zio);
}
}
}
return;
}
#endif
DTRACE_PROBE1(zio__delay__skip, zio_t *, zio);
zio_interrupt(zio);
}
static void
zio_deadman_impl(zio_t *pio, int ziodepth)
{
zio_t *cio, *cio_next;
zio_link_t *zl = NULL;
vdev_t *vd = pio->io_vd;
if (zio_deadman_log_all || (vd != NULL && vd->vdev_ops->vdev_op_leaf)) {
vdev_queue_t *vq = vd ? &vd->vdev_queue : NULL;
zbookmark_phys_t *zb = &pio->io_bookmark;
uint64_t delta = gethrtime() - pio->io_timestamp;
uint64_t failmode = spa_get_deadman_failmode(pio->io_spa);
zfs_dbgmsg("slow zio[%d]: zio=%px timestamp=%llu "
"delta=%llu queued=%llu io=%llu "
"path=%s "
"last=%llu type=%d "
"priority=%d flags=0x%x stage=0x%x "
"pipeline=0x%x pipeline-trace=0x%x "
"objset=%llu object=%llu "
"level=%llu blkid=%llu "
"offset=%llu size=%llu "
"error=%d",
ziodepth, pio, pio->io_timestamp,
(u_longlong_t)delta, pio->io_delta, pio->io_delay,
vd ? vd->vdev_path : "NULL",
vq ? vq->vq_io_complete_ts : 0, pio->io_type,
pio->io_priority, pio->io_flags, pio->io_stage,
pio->io_pipeline, pio->io_pipeline_trace,
(u_longlong_t)zb->zb_objset, (u_longlong_t)zb->zb_object,
(u_longlong_t)zb->zb_level, (u_longlong_t)zb->zb_blkid,
(u_longlong_t)pio->io_offset, (u_longlong_t)pio->io_size,
pio->io_error);
(void) zfs_ereport_post(FM_EREPORT_ZFS_DEADMAN,
pio->io_spa, vd, zb, pio, 0);
if (failmode == ZIO_FAILURE_MODE_CONTINUE &&
taskq_empty_ent(&pio->io_tqent)) {
zio_interrupt(pio);
}
}
mutex_enter(&pio->io_lock);
for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
cio_next = zio_walk_children(pio, &zl);
zio_deadman_impl(cio, ziodepth + 1);
}
mutex_exit(&pio->io_lock);
}
/*
* Log the critical information describing this zio and all of its children
* using the zfs_dbgmsg() interface then post deadman event for the ZED.
*/
void
zio_deadman(zio_t *pio, char *tag)
{
spa_t *spa = pio->io_spa;
char *name = spa_name(spa);
if (!zfs_deadman_enabled || spa_suspended(spa))
return;
zio_deadman_impl(pio, 0);
switch (spa_get_deadman_failmode(spa)) {
case ZIO_FAILURE_MODE_WAIT:
zfs_dbgmsg("%s waiting for hung I/O to pool '%s'", tag, name);
break;
case ZIO_FAILURE_MODE_CONTINUE:
zfs_dbgmsg("%s restarting hung I/O for pool '%s'", tag, name);
break;
case ZIO_FAILURE_MODE_PANIC:
fm_panic("%s determined I/O to pool '%s' is hung.", tag, name);
break;
}
}
/*
* Execute the I/O pipeline until one of the following occurs:
* (1) the I/O completes; (2) the pipeline stalls waiting for
* dependent child I/Os; (3) the I/O issues, so we're waiting
* for an I/O completion interrupt; (4) the I/O is delegated by
* vdev-level caching or aggregation; (5) the I/O is deferred
* due to vdev-level queueing; (6) the I/O is handed off to
* another thread. In all cases, the pipeline stops whenever
* there's no CPU work; it never burns a thread in cv_wait_io().
*
* There's no locking on io_stage because there's no legitimate way
* for multiple threads to be attempting to process the same I/O.
*/
static zio_pipe_stage_t *zio_pipeline[];
/*
* zio_execute() is a wrapper around the static function
* __zio_execute() so that we can force __zio_execute() to be
* inlined. This reduces stack overhead which is important
* because __zio_execute() is called recursively in several zio
* code paths. zio_execute() itself cannot be inlined because
* it is externally visible.
*/
void
zio_execute(void *zio)
{
fstrans_cookie_t cookie;
cookie = spl_fstrans_mark();
__zio_execute(zio);
spl_fstrans_unmark(cookie);
}
/*
* Used to determine if in the current context the stack is sized large
* enough to allow zio_execute() to be called recursively. A minimum
* stack size of 16K is required to avoid needing to re-dispatch the zio.
*/
static boolean_t
zio_execute_stack_check(zio_t *zio)
{
#if !defined(HAVE_LARGE_STACKS)
dsl_pool_t *dp = spa_get_dsl(zio->io_spa);
/* Executing in txg_sync_thread() context. */
if (dp && curthread == dp->dp_tx.tx_sync_thread)
return (B_TRUE);
/* Pool initialization outside of zio_taskq context. */
if (dp && spa_is_initializing(dp->dp_spa) &&
!zio_taskq_member(zio, ZIO_TASKQ_ISSUE) &&
!zio_taskq_member(zio, ZIO_TASKQ_ISSUE_HIGH))
return (B_TRUE);
#else
(void) zio;
#endif /* HAVE_LARGE_STACKS */
return (B_FALSE);
}
__attribute__((always_inline))
static inline void
__zio_execute(zio_t *zio)
{
ASSERT3U(zio->io_queued_timestamp, >, 0);
while (zio->io_stage < ZIO_STAGE_DONE) {
enum zio_stage pipeline = zio->io_pipeline;
enum zio_stage stage = zio->io_stage;
zio->io_executor = curthread;
ASSERT(!MUTEX_HELD(&zio->io_lock));
ASSERT(ISP2(stage));
ASSERT(zio->io_stall == NULL);
do {
stage <<= 1;
} while ((stage & pipeline) == 0);
ASSERT(stage <= ZIO_STAGE_DONE);
/*
* If we are in interrupt context and this pipeline stage
* will grab a config lock that is held across I/O,
* or may wait for an I/O that needs an interrupt thread
* to complete, issue async to avoid deadlock.
*
* For VDEV_IO_START, we cut in line so that the io will
* be sent to disk promptly.
*/
if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL &&
zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) {
boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
zio_requeue_io_start_cut_in_line : B_FALSE;
zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
return;
}
/*
* If the current context doesn't have large enough stacks
* the zio must be issued asynchronously to prevent overflow.
*/
if (zio_execute_stack_check(zio)) {
boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
zio_requeue_io_start_cut_in_line : B_FALSE;
zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
return;
}
zio->io_stage = stage;
zio->io_pipeline_trace |= zio->io_stage;
/*
* The zio pipeline stage returns the next zio to execute
* (typically the same as this one), or NULL if we should
* stop.
*/
zio = zio_pipeline[highbit64(stage) - 1](zio);
if (zio == NULL)
return;
}
}
/*
* ==========================================================================
* Initiate I/O, either sync or async
* ==========================================================================
*/
int
zio_wait(zio_t *zio)
{
/*
* Some routines, like zio_free_sync(), may return a NULL zio
* to avoid the performance overhead of creating and then destroying
* an unneeded zio. For the callers' simplicity, we accept a NULL
* zio and ignore it.
*/
if (zio == NULL)
return (0);
long timeout = MSEC_TO_TICK(zfs_deadman_ziotime_ms);
int error;
ASSERT3S(zio->io_stage, ==, ZIO_STAGE_OPEN);
ASSERT3P(zio->io_executor, ==, NULL);
zio->io_waiter = curthread;
ASSERT0(zio->io_queued_timestamp);
zio->io_queued_timestamp = gethrtime();
__zio_execute(zio);
mutex_enter(&zio->io_lock);
while (zio->io_executor != NULL) {
error = cv_timedwait_io(&zio->io_cv, &zio->io_lock,
ddi_get_lbolt() + timeout);
if (zfs_deadman_enabled && error == -1 &&
gethrtime() - zio->io_queued_timestamp >
spa_deadman_ziotime(zio->io_spa)) {
mutex_exit(&zio->io_lock);
timeout = MSEC_TO_TICK(zfs_deadman_checktime_ms);
zio_deadman(zio, FTAG);
mutex_enter(&zio->io_lock);
}
}
mutex_exit(&zio->io_lock);
error = zio->io_error;
zio_destroy(zio);
return (error);
}
void
zio_nowait(zio_t *zio)
{
/*
* See comment in zio_wait().
*/
if (zio == NULL)
return;
ASSERT3P(zio->io_executor, ==, NULL);
if (zio->io_child_type == ZIO_CHILD_LOGICAL &&
zio_unique_parent(zio) == NULL) {
zio_t *pio;
/*
* This is a logical async I/O with no parent to wait for it.
* We add it to the spa_async_root_zio "Godfather" I/O which
* will ensure they complete prior to unloading the pool.
*/
spa_t *spa = zio->io_spa;
pio = spa->spa_async_zio_root[CPU_SEQID_UNSTABLE];
zio_add_child(pio, zio);
}
ASSERT0(zio->io_queued_timestamp);
zio->io_queued_timestamp = gethrtime();
__zio_execute(zio);
}
/*
* ==========================================================================
* Reexecute, cancel, or suspend/resume failed I/O
* ==========================================================================
*/
static void
zio_reexecute(void *arg)
{
zio_t *pio = arg;
zio_t *cio, *cio_next;
ASSERT(pio->io_child_type == ZIO_CHILD_LOGICAL);
ASSERT(pio->io_orig_stage == ZIO_STAGE_OPEN);
ASSERT(pio->io_gang_leader == NULL);
ASSERT(pio->io_gang_tree == NULL);
pio->io_flags = pio->io_orig_flags;
pio->io_stage = pio->io_orig_stage;
pio->io_pipeline = pio->io_orig_pipeline;
pio->io_reexecute = 0;
pio->io_flags |= ZIO_FLAG_REEXECUTED;
pio->io_pipeline_trace = 0;
pio->io_error = 0;
for (int w = 0; w < ZIO_WAIT_TYPES; w++)
pio->io_state[w] = 0;
for (int c = 0; c < ZIO_CHILD_TYPES; c++)
pio->io_child_error[c] = 0;
if (IO_IS_ALLOCATING(pio))
BP_ZERO(pio->io_bp);
/*
* As we reexecute pio's children, new children could be created.
* New children go to the head of pio's io_child_list, however,
* so we will (correctly) not reexecute them. The key is that
* the remainder of pio's io_child_list, from 'cio_next' onward,
* cannot be affected by any side effects of reexecuting 'cio'.
*/
zio_link_t *zl = NULL;
mutex_enter(&pio->io_lock);
for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
cio_next = zio_walk_children(pio, &zl);
for (int w = 0; w < ZIO_WAIT_TYPES; w++)
pio->io_children[cio->io_child_type][w]++;
mutex_exit(&pio->io_lock);
zio_reexecute(cio);
mutex_enter(&pio->io_lock);
}
mutex_exit(&pio->io_lock);
/*
* Now that all children have been reexecuted, execute the parent.
* We don't reexecute "The Godfather" I/O here as it's the
* responsibility of the caller to wait on it.
*/
if (!(pio->io_flags & ZIO_FLAG_GODFATHER)) {
pio->io_queued_timestamp = gethrtime();
__zio_execute(pio);
}
}
void
zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t reason)
{
if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_PANIC)
fm_panic("Pool '%s' has encountered an uncorrectable I/O "
"failure and the failure mode property for this pool "
"is set to panic.", spa_name(spa));
cmn_err(CE_WARN, "Pool '%s' has encountered an uncorrectable I/O "
"failure and has been suspended.\n", spa_name(spa));
(void) zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL,
NULL, NULL, 0);
mutex_enter(&spa->spa_suspend_lock);
if (spa->spa_suspend_zio_root == NULL)
spa->spa_suspend_zio_root = zio_root(spa, NULL, NULL,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE |
ZIO_FLAG_GODFATHER);
spa->spa_suspended = reason;
if (zio != NULL) {
ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER));
ASSERT(zio != spa->spa_suspend_zio_root);
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
ASSERT(zio_unique_parent(zio) == NULL);
ASSERT(zio->io_stage == ZIO_STAGE_DONE);
zio_add_child(spa->spa_suspend_zio_root, zio);
}
mutex_exit(&spa->spa_suspend_lock);
}
int
zio_resume(spa_t *spa)
{
zio_t *pio;
/*
* Reexecute all previously suspended i/o.
*/
mutex_enter(&spa->spa_suspend_lock);
spa->spa_suspended = ZIO_SUSPEND_NONE;
cv_broadcast(&spa->spa_suspend_cv);
pio = spa->spa_suspend_zio_root;
spa->spa_suspend_zio_root = NULL;
mutex_exit(&spa->spa_suspend_lock);
if (pio == NULL)
return (0);
zio_reexecute(pio);
return (zio_wait(pio));
}
void
zio_resume_wait(spa_t *spa)
{
mutex_enter(&spa->spa_suspend_lock);
while (spa_suspended(spa))
cv_wait(&spa->spa_suspend_cv, &spa->spa_suspend_lock);
mutex_exit(&spa->spa_suspend_lock);
}
/*
* ==========================================================================
* Gang blocks.
*
* A gang block is a collection of small blocks that looks to the DMU
* like one large block. When zio_dva_allocate() cannot find a block
* of the requested size, due to either severe fragmentation or the pool
* being nearly full, it calls zio_write_gang_block() to construct the
* block from smaller fragments.
*
* A gang block consists of a gang header (zio_gbh_phys_t) and up to
* three (SPA_GBH_NBLKPTRS) gang members. The gang header is just like
* an indirect block: it's an array of block pointers. It consumes
* only one sector and hence is allocatable regardless of fragmentation.
* The gang header's bps point to its gang members, which hold the data.
*
* Gang blocks are self-checksumming, using the bp's <vdev, offset, txg>
* as the verifier to ensure uniqueness of the SHA256 checksum.
* Critically, the gang block bp's blk_cksum is the checksum of the data,
* not the gang header. This ensures that data block signatures (needed for
* deduplication) are independent of how the block is physically stored.
*
* Gang blocks can be nested: a gang member may itself be a gang block.
* Thus every gang block is a tree in which root and all interior nodes are
* gang headers, and the leaves are normal blocks that contain user data.
* The root of the gang tree is called the gang leader.
*
* To perform any operation (read, rewrite, free, claim) on a gang block,
* zio_gang_assemble() first assembles the gang tree (minus data leaves)
* in the io_gang_tree field of the original logical i/o by recursively
* reading the gang leader and all gang headers below it. This yields
* an in-core tree containing the contents of every gang header and the
* bps for every constituent of the gang block.
*
* With the gang tree now assembled, zio_gang_issue() just walks the gang tree
* and invokes a callback on each bp. To free a gang block, zio_gang_issue()
* calls zio_free_gang() -- a trivial wrapper around zio_free() -- for each bp.
* zio_claim_gang() provides a similarly trivial wrapper for zio_claim().
* zio_read_gang() is a wrapper around zio_read() that omits reading gang
* headers, since we already have those in io_gang_tree. zio_rewrite_gang()
* performs a zio_rewrite() of the data or, for gang headers, a zio_rewrite()
* of the gang header plus zio_checksum_compute() of the data to update the
* gang header's blk_cksum as described above.
*
* The two-phase assemble/issue model solves the problem of partial failure --
* what if you'd freed part of a gang block but then couldn't read the
* gang header for another part? Assembling the entire gang tree first
* ensures that all the necessary gang header I/O has succeeded before
* starting the actual work of free, claim, or write. Once the gang tree
* is assembled, free and claim are in-memory operations that cannot fail.
*
* In the event that a gang write fails, zio_dva_unallocate() walks the
* gang tree to immediately free (i.e. insert back into the space map)
* everything we've allocated. This ensures that we don't get ENOSPC
* errors during repeated suspend/resume cycles due to a flaky device.
*
* Gang rewrites only happen during sync-to-convergence. If we can't assemble
* the gang tree, we won't modify the block, so we can safely defer the free
* (knowing that the block is still intact). If we *can* assemble the gang
* tree, then even if some of the rewrites fail, zio_dva_unallocate() will free
* each constituent bp and we can allocate a new block on the next sync pass.
*
* In all cases, the gang tree allows complete recovery from partial failure.
* ==========================================================================
*/
static void
zio_gang_issue_func_done(zio_t *zio)
{
abd_free(zio->io_abd);
}
static zio_t *
zio_read_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data,
uint64_t offset)
{
if (gn != NULL)
return (pio);
return (zio_read(pio, pio->io_spa, bp, abd_get_offset(data, offset),
BP_GET_PSIZE(bp), zio_gang_issue_func_done,
NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
&pio->io_bookmark));
}
static zio_t *
zio_rewrite_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data,
uint64_t offset)
{
zio_t *zio;
if (gn != NULL) {
abd_t *gbh_abd =
abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE);
zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_issue_func_done, NULL,
pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
&pio->io_bookmark);
/*
* As we rewrite each gang header, the pipeline will compute
* a new gang block header checksum for it; but no one will
* compute a new data checksum, so we do that here. The one
* exception is the gang leader: the pipeline already computed
* its data checksum because that stage precedes gang assembly.
* (Presently, nothing actually uses interior data checksums;
* this is just good hygiene.)
*/
if (gn != pio->io_gang_leader->io_gang_tree) {
abd_t *buf = abd_get_offset(data, offset);
zio_checksum_compute(zio, BP_GET_CHECKSUM(bp),
buf, BP_GET_PSIZE(bp));
abd_free(buf);
}
/*
* If we are here to damage data for testing purposes,
* leave the GBH alone so that we can detect the damage.
*/
if (pio->io_gang_leader->io_flags & ZIO_FLAG_INDUCE_DAMAGE)
zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
} else {
zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
abd_get_offset(data, offset), BP_GET_PSIZE(bp),
zio_gang_issue_func_done, NULL, pio->io_priority,
ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);
}
return (zio);
}
static zio_t *
zio_free_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data,
uint64_t offset)
{
(void) gn, (void) data, (void) offset;
zio_t *zio = zio_free_sync(pio, pio->io_spa, pio->io_txg, bp,
ZIO_GANG_CHILD_FLAGS(pio));
if (zio == NULL) {
zio = zio_null(pio, pio->io_spa,
NULL, NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio));
}
return (zio);
}
static zio_t *
zio_claim_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data,
uint64_t offset)
{
(void) gn, (void) data, (void) offset;
return (zio_claim(pio, pio->io_spa, pio->io_txg, bp,
NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio)));
}
static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = {
NULL,
zio_read_gang,
zio_rewrite_gang,
zio_free_gang,
zio_claim_gang,
NULL
};
static void zio_gang_tree_assemble_done(zio_t *zio);
static zio_gang_node_t *
zio_gang_node_alloc(zio_gang_node_t **gnpp)
{
zio_gang_node_t *gn;
ASSERT(*gnpp == NULL);
gn = kmem_zalloc(sizeof (*gn), KM_SLEEP);
gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE);
*gnpp = gn;
return (gn);
}
static void
zio_gang_node_free(zio_gang_node_t **gnpp)
{
zio_gang_node_t *gn = *gnpp;
for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
ASSERT(gn->gn_child[g] == NULL);
zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE);
kmem_free(gn, sizeof (*gn));
*gnpp = NULL;
}
static void
zio_gang_tree_free(zio_gang_node_t **gnpp)
{
zio_gang_node_t *gn = *gnpp;
if (gn == NULL)
return;
for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
zio_gang_tree_free(&gn->gn_child[g]);
zio_gang_node_free(gnpp);
}
static void
zio_gang_tree_assemble(zio_t *gio, blkptr_t *bp, zio_gang_node_t **gnpp)
{
zio_gang_node_t *gn = zio_gang_node_alloc(gnpp);
abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE);
ASSERT(gio->io_gang_leader == gio);
ASSERT(BP_IS_GANG(bp));
zio_nowait(zio_read(gio, gio->io_spa, bp, gbh_abd, SPA_GANGBLOCKSIZE,
zio_gang_tree_assemble_done, gn, gio->io_priority,
ZIO_GANG_CHILD_FLAGS(gio), &gio->io_bookmark));
}
static void
zio_gang_tree_assemble_done(zio_t *zio)
{
zio_t *gio = zio->io_gang_leader;
zio_gang_node_t *gn = zio->io_private;
blkptr_t *bp = zio->io_bp;
ASSERT(gio == zio_unique_parent(zio));
ASSERT(zio->io_child_count == 0);
if (zio->io_error)
return;
/* this ABD was created from a linear buf in zio_gang_tree_assemble */
if (BP_SHOULD_BYTESWAP(bp))
byteswap_uint64_array(abd_to_buf(zio->io_abd), zio->io_size);
ASSERT3P(abd_to_buf(zio->io_abd), ==, gn->gn_gbh);
ASSERT(zio->io_size == SPA_GANGBLOCKSIZE);
ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC);
abd_free(zio->io_abd);
for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
if (!BP_IS_GANG(gbp))
continue;
zio_gang_tree_assemble(gio, gbp, &gn->gn_child[g]);
}
}
static void
zio_gang_tree_issue(zio_t *pio, zio_gang_node_t *gn, blkptr_t *bp, abd_t *data,
uint64_t offset)
{
zio_t *gio = pio->io_gang_leader;
zio_t *zio;
ASSERT(BP_IS_GANG(bp) == !!gn);
ASSERT(BP_GET_CHECKSUM(bp) == BP_GET_CHECKSUM(gio->io_bp));
ASSERT(BP_GET_LSIZE(bp) == BP_GET_PSIZE(bp) || gn == gio->io_gang_tree);
/*
* If you're a gang header, your data is in gn->gn_gbh.
* If you're a gang member, your data is in 'data' and gn == NULL.
*/
zio = zio_gang_issue_func[gio->io_type](pio, bp, gn, data, offset);
if (gn != NULL) {
ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC);
for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
if (BP_IS_HOLE(gbp))
continue;
zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data,
offset);
offset += BP_GET_PSIZE(gbp);
}
}
if (gn == gio->io_gang_tree)
ASSERT3U(gio->io_size, ==, offset);
if (zio != pio)
zio_nowait(zio);
}
static zio_t *
zio_gang_assemble(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == NULL);
ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
zio->io_gang_leader = zio;
zio_gang_tree_assemble(zio, bp, &zio->io_gang_tree);
return (zio);
}
static zio_t *
zio_gang_issue(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT, ZIO_WAIT_DONE)) {
return (NULL);
}
ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == zio);
ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
if (zio->io_child_error[ZIO_CHILD_GANG] == 0)
zio_gang_tree_issue(zio, zio->io_gang_tree, bp, zio->io_abd,
0);
else
zio_gang_tree_free(&zio->io_gang_tree);
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
return (zio);
}
static void
zio_write_gang_member_ready(zio_t *zio)
{
zio_t *pio = zio_unique_parent(zio);
dva_t *cdva = zio->io_bp->blk_dva;
dva_t *pdva = pio->io_bp->blk_dva;
uint64_t asize;
zio_t *gio __maybe_unused = zio->io_gang_leader;
if (BP_IS_HOLE(zio->io_bp))
return;
ASSERT(BP_IS_HOLE(&zio->io_bp_orig));
ASSERT(zio->io_child_type == ZIO_CHILD_GANG);
ASSERT3U(zio->io_prop.zp_copies, ==, gio->io_prop.zp_copies);
ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp));
ASSERT3U(pio->io_prop.zp_copies, <=, BP_GET_NDVAS(pio->io_bp));
ASSERT3U(BP_GET_NDVAS(zio->io_bp), <=, BP_GET_NDVAS(pio->io_bp));
mutex_enter(&pio->io_lock);
for (int d = 0; d < BP_GET_NDVAS(zio->io_bp); d++) {
ASSERT(DVA_GET_GANG(&pdva[d]));
asize = DVA_GET_ASIZE(&pdva[d]);
asize += DVA_GET_ASIZE(&cdva[d]);
DVA_SET_ASIZE(&pdva[d], asize);
}
mutex_exit(&pio->io_lock);
}
static void
zio_write_gang_done(zio_t *zio)
{
/*
* The io_abd field will be NULL for a zio with no data. The io_flags
* will initially have the ZIO_FLAG_NODATA bit flag set, but we can't
* check for it here as it is cleared in zio_ready.
*/
if (zio->io_abd != NULL)
abd_free(zio->io_abd);
}
static zio_t *
zio_write_gang_block(zio_t *pio, metaslab_class_t *mc)
{
spa_t *spa = pio->io_spa;
blkptr_t *bp = pio->io_bp;
zio_t *gio = pio->io_gang_leader;
zio_t *zio;
zio_gang_node_t *gn, **gnpp;
zio_gbh_phys_t *gbh;
abd_t *gbh_abd;
uint64_t txg = pio->io_txg;
uint64_t resid = pio->io_size;
uint64_t lsize;
int copies = gio->io_prop.zp_copies;
int gbh_copies;
zio_prop_t zp;
int error;
boolean_t has_data = !(pio->io_flags & ZIO_FLAG_NODATA);
/*
* encrypted blocks need DVA[2] free so encrypted gang headers can't
* have a third copy.
*/
gbh_copies = MIN(copies + 1, spa_max_replication(spa));
if (gio->io_prop.zp_encrypt && gbh_copies >= SPA_DVAS_PER_BP)
gbh_copies = SPA_DVAS_PER_BP - 1;
int flags = METASLAB_HINTBP_FAVOR | METASLAB_GANG_HEADER;
if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
ASSERT(has_data);
flags |= METASLAB_ASYNC_ALLOC;
VERIFY(zfs_refcount_held(&mc->mc_allocator[pio->io_allocator].
mca_alloc_slots, pio));
/*
* The logical zio has already placed a reservation for
* 'copies' allocation slots but gang blocks may require
* additional copies. These additional copies
* (i.e. gbh_copies - copies) are guaranteed to succeed
* since metaslab_class_throttle_reserve() always allows
* additional reservations for gang blocks.
*/
VERIFY(metaslab_class_throttle_reserve(mc, gbh_copies - copies,
pio->io_allocator, pio, flags));
}
error = metaslab_alloc(spa, mc, SPA_GANGBLOCKSIZE,
bp, gbh_copies, txg, pio == gio ? NULL : gio->io_bp, flags,
&pio->io_alloc_list, pio, pio->io_allocator);
if (error) {
if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
ASSERT(has_data);
/*
* If we failed to allocate the gang block header then
* we remove any additional allocation reservations that
* we placed here. The original reservation will
* be removed when the logical I/O goes to the ready
* stage.
*/
metaslab_class_throttle_unreserve(mc,
gbh_copies - copies, pio->io_allocator, pio);
}
pio->io_error = error;
return (pio);
}
if (pio == gio) {
gnpp = &gio->io_gang_tree;
} else {
gnpp = pio->io_private;
ASSERT(pio->io_ready == zio_write_gang_member_ready);
}
gn = zio_gang_node_alloc(gnpp);
gbh = gn->gn_gbh;
bzero(gbh, SPA_GANGBLOCKSIZE);
gbh_abd = abd_get_from_buf(gbh, SPA_GANGBLOCKSIZE);
/*
* Create the gang header.
*/
zio = zio_rewrite(pio, spa, txg, bp, gbh_abd, SPA_GANGBLOCKSIZE,
zio_write_gang_done, NULL, pio->io_priority,
ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);
/*
* Create and nowait the gang children.
*/
for (int g = 0; resid != 0; resid -= lsize, g++) {
lsize = P2ROUNDUP(resid / (SPA_GBH_NBLKPTRS - g),
SPA_MINBLOCKSIZE);
ASSERT(lsize >= SPA_MINBLOCKSIZE && lsize <= resid);
zp.zp_checksum = gio->io_prop.zp_checksum;
zp.zp_compress = ZIO_COMPRESS_OFF;
zp.zp_complevel = gio->io_prop.zp_complevel;
zp.zp_type = DMU_OT_NONE;
zp.zp_level = 0;
zp.zp_copies = gio->io_prop.zp_copies;
zp.zp_dedup = B_FALSE;
zp.zp_dedup_verify = B_FALSE;
zp.zp_nopwrite = B_FALSE;
zp.zp_encrypt = gio->io_prop.zp_encrypt;
zp.zp_byteorder = gio->io_prop.zp_byteorder;
bzero(zp.zp_salt, ZIO_DATA_SALT_LEN);
bzero(zp.zp_iv, ZIO_DATA_IV_LEN);
bzero(zp.zp_mac, ZIO_DATA_MAC_LEN);
zio_t *cio = zio_write(zio, spa, txg, &gbh->zg_blkptr[g],
has_data ? abd_get_offset(pio->io_abd, pio->io_size -
resid) : NULL, lsize, lsize, &zp,
zio_write_gang_member_ready, NULL, NULL,
zio_write_gang_done, &gn->gn_child[g], pio->io_priority,
ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);
if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
ASSERT(has_data);
/*
* Gang children won't throttle but we should
* account for their work, so reserve an allocation
* slot for them here.
*/
VERIFY(metaslab_class_throttle_reserve(mc,
zp.zp_copies, cio->io_allocator, cio, flags));
}
zio_nowait(cio);
}
/*
* Set pio's pipeline to just wait for zio to finish.
*/
pio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
/*
* We didn't allocate this bp, so make sure it doesn't get unmarked.
*/
pio->io_flags &= ~ZIO_FLAG_FASTWRITE;
zio_nowait(zio);
return (pio);
}
/*
* The zio_nop_write stage in the pipeline determines if allocating a
* new bp is necessary. The nopwrite feature can handle writes in
* either syncing or open context (i.e. zil writes) and as a result is
* mutually exclusive with dedup.
*
* By leveraging a cryptographically secure checksum, such as SHA256, we
* can compare the checksums of the new data and the old to determine if
* allocating a new block is required. Note that our requirements for
* cryptographic strength are fairly weak: there can't be any accidental
* hash collisions, but we don't need to be secure against intentional
* (malicious) collisions. To trigger a nopwrite, you have to be able
* to write the file to begin with, and triggering an incorrect (hash
* collision) nopwrite is no worse than simply writing to the file.
* That said, there are no known attacks against the checksum algorithms
* used for nopwrite, assuming that the salt and the checksums
* themselves remain secret.
*/
static zio_t *
zio_nop_write(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
blkptr_t *bp_orig = &zio->io_bp_orig;
zio_prop_t *zp = &zio->io_prop;
ASSERT(BP_GET_LEVEL(bp) == 0);
ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
ASSERT(zp->zp_nopwrite);
ASSERT(!zp->zp_dedup);
ASSERT(zio->io_bp_override == NULL);
ASSERT(IO_IS_ALLOCATING(zio));
/*
* Check to see if the original bp and the new bp have matching
* characteristics (i.e. same checksum, compression algorithms, etc).
* If they don't then just continue with the pipeline which will
* allocate a new bp.
*/
if (BP_IS_HOLE(bp_orig) ||
!(zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_flags &
ZCHECKSUM_FLAG_NOPWRITE) ||
BP_IS_ENCRYPTED(bp) || BP_IS_ENCRYPTED(bp_orig) ||
BP_GET_CHECKSUM(bp) != BP_GET_CHECKSUM(bp_orig) ||
BP_GET_COMPRESS(bp) != BP_GET_COMPRESS(bp_orig) ||
BP_GET_DEDUP(bp) != BP_GET_DEDUP(bp_orig) ||
zp->zp_copies != BP_GET_NDVAS(bp_orig))
return (zio);
/*
* If the checksums match then reset the pipeline so that we
* avoid allocating a new bp and issuing any I/O.
*/
if (ZIO_CHECKSUM_EQUAL(bp->blk_cksum, bp_orig->blk_cksum)) {
ASSERT(zio_checksum_table[zp->zp_checksum].ci_flags &
ZCHECKSUM_FLAG_NOPWRITE);
ASSERT3U(BP_GET_PSIZE(bp), ==, BP_GET_PSIZE(bp_orig));
ASSERT3U(BP_GET_LSIZE(bp), ==, BP_GET_LSIZE(bp_orig));
ASSERT(zp->zp_compress != ZIO_COMPRESS_OFF);
ASSERT(bcmp(&bp->blk_prop, &bp_orig->blk_prop,
sizeof (uint64_t)) == 0);
/*
* If we're overwriting a block that is currently on an
* indirect vdev, then ignore the nopwrite request and
* allow a new block to be allocated on a concrete vdev.
*/
spa_config_enter(zio->io_spa, SCL_VDEV, FTAG, RW_READER);
vdev_t *tvd = vdev_lookup_top(zio->io_spa,
DVA_GET_VDEV(&bp->blk_dva[0]));
if (tvd->vdev_ops == &vdev_indirect_ops) {
spa_config_exit(zio->io_spa, SCL_VDEV, FTAG);
return (zio);
}
spa_config_exit(zio->io_spa, SCL_VDEV, FTAG);
*bp = *bp_orig;
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
zio->io_flags |= ZIO_FLAG_NOPWRITE;
}
return (zio);
}
/*
* ==========================================================================
* Dedup
* ==========================================================================
*/
static void
zio_ddt_child_read_done(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
ddt_entry_t *dde = zio->io_private;
ddt_phys_t *ddp;
zio_t *pio = zio_unique_parent(zio);
mutex_enter(&pio->io_lock);
ddp = ddt_phys_select(dde, bp);
if (zio->io_error == 0)
ddt_phys_clear(ddp); /* this ddp doesn't need repair */
if (zio->io_error == 0 && dde->dde_repair_abd == NULL)
dde->dde_repair_abd = zio->io_abd;
else
abd_free(zio->io_abd);
mutex_exit(&pio->io_lock);
}
static zio_t *
zio_ddt_read_start(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
ASSERT(BP_GET_DEDUP(bp));
ASSERT(BP_GET_PSIZE(bp) == zio->io_size);
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
if (zio->io_child_error[ZIO_CHILD_DDT]) {
ddt_t *ddt = ddt_select(zio->io_spa, bp);
ddt_entry_t *dde = ddt_repair_start(ddt, bp);
ddt_phys_t *ddp = dde->dde_phys;
ddt_phys_t *ddp_self = ddt_phys_select(dde, bp);
blkptr_t blk;
ASSERT(zio->io_vsd == NULL);
zio->io_vsd = dde;
if (ddp_self == NULL)
return (zio);
for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
if (ddp->ddp_phys_birth == 0 || ddp == ddp_self)
continue;
ddt_bp_create(ddt->ddt_checksum, &dde->dde_key, ddp,
&blk);
zio_nowait(zio_read(zio, zio->io_spa, &blk,
abd_alloc_for_io(zio->io_size, B_TRUE),
zio->io_size, zio_ddt_child_read_done, dde,
zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio) |
ZIO_FLAG_DONT_PROPAGATE, &zio->io_bookmark));
}
return (zio);
}
zio_nowait(zio_read(zio, zio->io_spa, bp,
zio->io_abd, zio->io_size, NULL, NULL, zio->io_priority,
ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark));
return (zio);
}
static zio_t *
zio_ddt_read_done(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
if (zio_wait_for_children(zio, ZIO_CHILD_DDT_BIT, ZIO_WAIT_DONE)) {
return (NULL);
}
ASSERT(BP_GET_DEDUP(bp));
ASSERT(BP_GET_PSIZE(bp) == zio->io_size);
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
if (zio->io_child_error[ZIO_CHILD_DDT]) {
ddt_t *ddt = ddt_select(zio->io_spa, bp);
ddt_entry_t *dde = zio->io_vsd;
if (ddt == NULL) {
ASSERT(spa_load_state(zio->io_spa) != SPA_LOAD_NONE);
return (zio);
}
if (dde == NULL) {
zio->io_stage = ZIO_STAGE_DDT_READ_START >> 1;
zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE);
return (NULL);
}
if (dde->dde_repair_abd != NULL) {
abd_copy(zio->io_abd, dde->dde_repair_abd,
zio->io_size);
zio->io_child_error[ZIO_CHILD_DDT] = 0;
}
ddt_repair_done(ddt, dde);
zio->io_vsd = NULL;
}
ASSERT(zio->io_vsd == NULL);
return (zio);
}
static boolean_t
zio_ddt_collision(zio_t *zio, ddt_t *ddt, ddt_entry_t *dde)
{
spa_t *spa = zio->io_spa;
boolean_t do_raw = !!(zio->io_flags & ZIO_FLAG_RAW);
ASSERT(!(zio->io_bp_override && do_raw));
/*
* Note: we compare the original data, not the transformed data,
* because when zio->io_bp is an override bp, we will not have
* pushed the I/O transforms. That's an important optimization
* because otherwise we'd compress/encrypt all dmu_sync() data twice.
* However, we should never get a raw, override zio so in these
* cases we can compare the io_abd directly. This is useful because
* it allows us to do dedup verification even if we don't have access
* to the original data (for instance, if the encryption keys aren't
* loaded).
*/
for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) {
zio_t *lio = dde->dde_lead_zio[p];
if (lio != NULL && do_raw) {
return (lio->io_size != zio->io_size ||
abd_cmp(zio->io_abd, lio->io_abd) != 0);
} else if (lio != NULL) {
return (lio->io_orig_size != zio->io_orig_size ||
abd_cmp(zio->io_orig_abd, lio->io_orig_abd) != 0);
}
}
for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) {
ddt_phys_t *ddp = &dde->dde_phys[p];
if (ddp->ddp_phys_birth != 0 && do_raw) {
blkptr_t blk = *zio->io_bp;
uint64_t psize;
abd_t *tmpabd;
int error;
ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth);
psize = BP_GET_PSIZE(&blk);
if (psize != zio->io_size)
return (B_TRUE);
ddt_exit(ddt);
tmpabd = abd_alloc_for_io(psize, B_TRUE);
error = zio_wait(zio_read(NULL, spa, &blk, tmpabd,
psize, NULL, NULL, ZIO_PRIORITY_SYNC_READ,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE |
ZIO_FLAG_RAW, &zio->io_bookmark));
if (error == 0) {
if (abd_cmp(tmpabd, zio->io_abd) != 0)
error = SET_ERROR(ENOENT);
}
abd_free(tmpabd);
ddt_enter(ddt);
return (error != 0);
} else if (ddp->ddp_phys_birth != 0) {
arc_buf_t *abuf = NULL;
arc_flags_t aflags = ARC_FLAG_WAIT;
blkptr_t blk = *zio->io_bp;
int error;
ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth);
if (BP_GET_LSIZE(&blk) != zio->io_orig_size)
return (B_TRUE);
ddt_exit(ddt);
error = arc_read(NULL, spa, &blk,
arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
&aflags, &zio->io_bookmark);
if (error == 0) {
if (abd_cmp_buf(zio->io_orig_abd, abuf->b_data,
zio->io_orig_size) != 0)
error = SET_ERROR(ENOENT);
arc_buf_destroy(abuf, &abuf);
}
ddt_enter(ddt);
return (error != 0);
}
}
return (B_FALSE);
}
static void
zio_ddt_child_write_ready(zio_t *zio)
{
int p = zio->io_prop.zp_copies;
ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp);
ddt_entry_t *dde = zio->io_private;
ddt_phys_t *ddp = &dde->dde_phys[p];
zio_t *pio;
if (zio->io_error)
return;
ddt_enter(ddt);
ASSERT(dde->dde_lead_zio[p] == zio);
ddt_phys_fill(ddp, zio->io_bp);
zio_link_t *zl = NULL;
while ((pio = zio_walk_parents(zio, &zl)) != NULL)
ddt_bp_fill(ddp, pio->io_bp, zio->io_txg);
ddt_exit(ddt);
}
static void
zio_ddt_child_write_done(zio_t *zio)
{
int p = zio->io_prop.zp_copies;
ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp);
ddt_entry_t *dde = zio->io_private;
ddt_phys_t *ddp = &dde->dde_phys[p];
ddt_enter(ddt);
ASSERT(ddp->ddp_refcnt == 0);
ASSERT(dde->dde_lead_zio[p] == zio);
dde->dde_lead_zio[p] = NULL;
if (zio->io_error == 0) {
zio_link_t *zl = NULL;
while (zio_walk_parents(zio, &zl) != NULL)
ddt_phys_addref(ddp);
} else {
ddt_phys_clear(ddp);
}
ddt_exit(ddt);
}
static zio_t *
zio_ddt_write(zio_t *zio)
{
spa_t *spa = zio->io_spa;
blkptr_t *bp = zio->io_bp;
uint64_t txg = zio->io_txg;
zio_prop_t *zp = &zio->io_prop;
int p = zp->zp_copies;
zio_t *cio = NULL;
ddt_t *ddt = ddt_select(spa, bp);
ddt_entry_t *dde;
ddt_phys_t *ddp;
ASSERT(BP_GET_DEDUP(bp));
ASSERT(BP_GET_CHECKSUM(bp) == zp->zp_checksum);
ASSERT(BP_IS_HOLE(bp) || zio->io_bp_override);
ASSERT(!(zio->io_bp_override && (zio->io_flags & ZIO_FLAG_RAW)));
ddt_enter(ddt);
dde = ddt_lookup(ddt, bp, B_TRUE);
ddp = &dde->dde_phys[p];
if (zp->zp_dedup_verify && zio_ddt_collision(zio, ddt, dde)) {
/*
* If we're using a weak checksum, upgrade to a strong checksum
* and try again. If we're already using a strong checksum,
* we can't resolve it, so just convert to an ordinary write.
* (And automatically e-mail a paper to Nature?)
*/
if (!(zio_checksum_table[zp->zp_checksum].ci_flags &
ZCHECKSUM_FLAG_DEDUP)) {
zp->zp_checksum = spa_dedup_checksum(spa);
zio_pop_transforms(zio);
zio->io_stage = ZIO_STAGE_OPEN;
BP_ZERO(bp);
} else {
zp->zp_dedup = B_FALSE;
BP_SET_DEDUP(bp, B_FALSE);
}
ASSERT(!BP_GET_DEDUP(bp));
zio->io_pipeline = ZIO_WRITE_PIPELINE;
ddt_exit(ddt);
return (zio);
}
if (ddp->ddp_phys_birth != 0 || dde->dde_lead_zio[p] != NULL) {
if (ddp->ddp_phys_birth != 0)
ddt_bp_fill(ddp, bp, txg);
if (dde->dde_lead_zio[p] != NULL)
zio_add_child(zio, dde->dde_lead_zio[p]);
else
ddt_phys_addref(ddp);
} else if (zio->io_bp_override) {
ASSERT(bp->blk_birth == txg);
ASSERT(BP_EQUAL(bp, zio->io_bp_override));
ddt_phys_fill(ddp, bp);
ddt_phys_addref(ddp);
} else {
cio = zio_write(zio, spa, txg, bp, zio->io_orig_abd,
zio->io_orig_size, zio->io_orig_size, zp,
zio_ddt_child_write_ready, NULL, NULL,
zio_ddt_child_write_done, dde, zio->io_priority,
ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark);
zio_push_transform(cio, zio->io_abd, zio->io_size, 0, NULL);
dde->dde_lead_zio[p] = cio;
}
ddt_exit(ddt);
zio_nowait(cio);
return (zio);
}
ddt_entry_t *freedde; /* for debugging */
static zio_t *
zio_ddt_free(zio_t *zio)
{
spa_t *spa = zio->io_spa;
blkptr_t *bp = zio->io_bp;
ddt_t *ddt = ddt_select(spa, bp);
ddt_entry_t *dde;
ddt_phys_t *ddp;
ASSERT(BP_GET_DEDUP(bp));
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
ddt_enter(ddt);
freedde = dde = ddt_lookup(ddt, bp, B_TRUE);
if (dde) {
ddp = ddt_phys_select(dde, bp);
if (ddp)
ddt_phys_decref(ddp);
}
ddt_exit(ddt);
return (zio);
}
/*
* ==========================================================================
* Allocate and free blocks
* ==========================================================================
*/
static zio_t *
zio_io_to_allocate(spa_t *spa, int allocator)
{
zio_t *zio;
ASSERT(MUTEX_HELD(&spa->spa_allocs[allocator].spaa_lock));
zio = avl_first(&spa->spa_allocs[allocator].spaa_tree);
if (zio == NULL)
return (NULL);
ASSERT(IO_IS_ALLOCATING(zio));
/*
* Try to place a reservation for this zio. If we're unable to
* reserve then we throttle.
*/
ASSERT3U(zio->io_allocator, ==, allocator);
if (!metaslab_class_throttle_reserve(zio->io_metaslab_class,
zio->io_prop.zp_copies, allocator, zio, 0)) {
return (NULL);
}
avl_remove(&spa->spa_allocs[allocator].spaa_tree, zio);
ASSERT3U(zio->io_stage, <, ZIO_STAGE_DVA_ALLOCATE);
return (zio);
}
static zio_t *
zio_dva_throttle(zio_t *zio)
{
spa_t *spa = zio->io_spa;
zio_t *nio;
metaslab_class_t *mc;
/* locate an appropriate allocation class */
mc = spa_preferred_class(spa, zio->io_size, zio->io_prop.zp_type,
zio->io_prop.zp_level, zio->io_prop.zp_zpl_smallblk);
if (zio->io_priority == ZIO_PRIORITY_SYNC_WRITE ||
!mc->mc_alloc_throttle_enabled ||
zio->io_child_type == ZIO_CHILD_GANG ||
zio->io_flags & ZIO_FLAG_NODATA) {
return (zio);
}
ASSERT(zio->io_type == ZIO_TYPE_WRITE);
ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
ASSERT3U(zio->io_queued_timestamp, >, 0);
ASSERT(zio->io_stage == ZIO_STAGE_DVA_THROTTLE);
zbookmark_phys_t *bm = &zio->io_bookmark;
/*
* We want to try to use as many allocators as possible to help improve
* performance, but we also want logically adjacent IOs to be physically
* adjacent to improve sequential read performance. We chunk each object
* into 2^20 block regions, and then hash based on the objset, object,
* level, and region to accomplish both of these goals.
*/
int allocator = (uint_t)cityhash4(bm->zb_objset, bm->zb_object,
bm->zb_level, bm->zb_blkid >> 20) % spa->spa_alloc_count;
zio->io_allocator = allocator;
zio->io_metaslab_class = mc;
mutex_enter(&spa->spa_allocs[allocator].spaa_lock);
avl_add(&spa->spa_allocs[allocator].spaa_tree, zio);
nio = zio_io_to_allocate(spa, allocator);
mutex_exit(&spa->spa_allocs[allocator].spaa_lock);
return (nio);
}
static void
zio_allocate_dispatch(spa_t *spa, int allocator)
{
zio_t *zio;
mutex_enter(&spa->spa_allocs[allocator].spaa_lock);
zio = zio_io_to_allocate(spa, allocator);
mutex_exit(&spa->spa_allocs[allocator].spaa_lock);
if (zio == NULL)
return;
ASSERT3U(zio->io_stage, ==, ZIO_STAGE_DVA_THROTTLE);
ASSERT0(zio->io_error);
zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_TRUE);
}
static zio_t *
zio_dva_allocate(zio_t *zio)
{
spa_t *spa = zio->io_spa;
metaslab_class_t *mc;
blkptr_t *bp = zio->io_bp;
int error;
int flags = 0;
if (zio->io_gang_leader == NULL) {
ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
zio->io_gang_leader = zio;
}
ASSERT(BP_IS_HOLE(bp));
ASSERT0(BP_GET_NDVAS(bp));
ASSERT3U(zio->io_prop.zp_copies, >, 0);
ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa));
ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp));
flags |= (zio->io_flags & ZIO_FLAG_FASTWRITE) ? METASLAB_FASTWRITE : 0;
if (zio->io_flags & ZIO_FLAG_NODATA)
flags |= METASLAB_DONT_THROTTLE;
if (zio->io_flags & ZIO_FLAG_GANG_CHILD)
flags |= METASLAB_GANG_CHILD;
if (zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE)
flags |= METASLAB_ASYNC_ALLOC;
/*
* if not already chosen, locate an appropriate allocation class
*/
mc = zio->io_metaslab_class;
if (mc == NULL) {
mc = spa_preferred_class(spa, zio->io_size,
zio->io_prop.zp_type, zio->io_prop.zp_level,
zio->io_prop.zp_zpl_smallblk);
zio->io_metaslab_class = mc;
}
/*
* Try allocating the block in the usual metaslab class.
* If that's full, allocate it in the normal class.
* If that's full, allocate as a gang block,
* and if all are full, the allocation fails (which shouldn't happen).
*
* Note that we do not fall back on embedded slog (ZIL) space, to
* preserve unfragmented slog space, which is critical for decent
* sync write performance. If a log allocation fails, we will fall
* back to spa_sync() which is abysmal for performance.
*/
error = metaslab_alloc(spa, mc, zio->io_size, bp,
zio->io_prop.zp_copies, zio->io_txg, NULL, flags,
&zio->io_alloc_list, zio, zio->io_allocator);
/*
* Fallback to normal class when an alloc class is full
*/
if (error == ENOSPC && mc != spa_normal_class(spa)) {
/*
* If throttling, transfer reservation over to normal class.
* The io_allocator slot can remain the same even though we
* are switching classes.
*/
if (mc->mc_alloc_throttle_enabled &&
(zio->io_flags & ZIO_FLAG_IO_ALLOCATING)) {
metaslab_class_throttle_unreserve(mc,
zio->io_prop.zp_copies, zio->io_allocator, zio);
zio->io_flags &= ~ZIO_FLAG_IO_ALLOCATING;
VERIFY(metaslab_class_throttle_reserve(
spa_normal_class(spa),
zio->io_prop.zp_copies, zio->io_allocator, zio,
flags | METASLAB_MUST_RESERVE));
}
zio->io_metaslab_class = mc = spa_normal_class(spa);
if (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC) {
zfs_dbgmsg("%s: metaslab allocation failure, "
"trying normal class: zio %px, size %llu, error %d",
spa_name(spa), zio, (u_longlong_t)zio->io_size,
error);
}
error = metaslab_alloc(spa, mc, zio->io_size, bp,
zio->io_prop.zp_copies, zio->io_txg, NULL, flags,
&zio->io_alloc_list, zio, zio->io_allocator);
}
if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE) {
if (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC) {
zfs_dbgmsg("%s: metaslab allocation failure, "
"trying ganging: zio %px, size %llu, error %d",
spa_name(spa), zio, (u_longlong_t)zio->io_size,
error);
}
return (zio_write_gang_block(zio, mc));
}
if (error != 0) {
if (error != ENOSPC ||
(zfs_flags & ZFS_DEBUG_METASLAB_ALLOC)) {
zfs_dbgmsg("%s: metaslab allocation failure: zio %px, "
"size %llu, error %d",
spa_name(spa), zio, (u_longlong_t)zio->io_size,
error);
}
zio->io_error = error;
}
return (zio);
}
static zio_t *
zio_dva_free(zio_t *zio)
{
metaslab_free(zio->io_spa, zio->io_bp, zio->io_txg, B_FALSE);
return (zio);
}
static zio_t *
zio_dva_claim(zio_t *zio)
{
int error;
error = metaslab_claim(zio->io_spa, zio->io_bp, zio->io_txg);
if (error)
zio->io_error = error;
return (zio);
}
/*
* Undo an allocation. This is used by zio_done() when an I/O fails
* and we want to give back the block we just allocated.
* This handles both normal blocks and gang blocks.
*/
static void
zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp)
{
ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp));
ASSERT(zio->io_bp_override == NULL);
if (!BP_IS_HOLE(bp))
metaslab_free(zio->io_spa, bp, bp->blk_birth, B_TRUE);
if (gn != NULL) {
for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
zio_dva_unallocate(zio, gn->gn_child[g],
&gn->gn_gbh->zg_blkptr[g]);
}
}
}
/*
* Try to allocate an intent log block. Return 0 on success, errno on failure.
*/
int
zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp,
uint64_t size, boolean_t *slog)
{
int error = 1;
zio_alloc_list_t io_alloc_list;
ASSERT(txg > spa_syncing_txg(spa));
metaslab_trace_init(&io_alloc_list);
/*
* Block pointer fields are useful to metaslabs for stats and debugging.
* Fill in the obvious ones before calling into metaslab_alloc().
*/
BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
BP_SET_PSIZE(new_bp, size);
BP_SET_LEVEL(new_bp, 0);
/*
* When allocating a zil block, we don't have information about
* the final destination of the block except the objset it's part
* of, so we just hash the objset ID to pick the allocator to get
* some parallelism.
*/
int flags = METASLAB_FASTWRITE | METASLAB_ZIL;
int allocator = (uint_t)cityhash4(0, 0, 0,
os->os_dsl_dataset->ds_object) % spa->spa_alloc_count;
error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1,
txg, NULL, flags, &io_alloc_list, NULL, allocator);
*slog = (error == 0);
if (error != 0) {
error = metaslab_alloc(spa, spa_embedded_log_class(spa), size,
new_bp, 1, txg, NULL, flags,
&io_alloc_list, NULL, allocator);
}
if (error != 0) {
error = metaslab_alloc(spa, spa_normal_class(spa), size,
new_bp, 1, txg, NULL, flags,
&io_alloc_list, NULL, allocator);
}
metaslab_trace_fini(&io_alloc_list);
if (error == 0) {
BP_SET_LSIZE(new_bp, size);
BP_SET_PSIZE(new_bp, size);
BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF);
BP_SET_CHECKSUM(new_bp,
spa_version(spa) >= SPA_VERSION_SLIM_ZIL
? ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG);
BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
BP_SET_LEVEL(new_bp, 0);
BP_SET_DEDUP(new_bp, 0);
BP_SET_BYTEORDER(new_bp, ZFS_HOST_BYTEORDER);
/*
* encrypted blocks will require an IV and salt. We generate
* these now since we will not be rewriting the bp at
* rewrite time.
*/
if (os->os_encrypted) {
uint8_t iv[ZIO_DATA_IV_LEN];
uint8_t salt[ZIO_DATA_SALT_LEN];
BP_SET_CRYPT(new_bp, B_TRUE);
VERIFY0(spa_crypt_get_salt(spa,
dmu_objset_id(os), salt));
VERIFY0(zio_crypt_generate_iv(iv));
zio_crypt_encode_params_bp(new_bp, salt, iv);
}
} else {
zfs_dbgmsg("%s: zil block allocation failure: "
"size %llu, error %d", spa_name(spa), (u_longlong_t)size,
error);
}
return (error);
}
/*
* ==========================================================================
* Read and write to physical devices
* ==========================================================================
*/
/*
* Issue an I/O to the underlying vdev. Typically the issue pipeline
* stops after this stage and will resume upon I/O completion.
* However, there are instances where the vdev layer may need to
* continue the pipeline when an I/O was not issued. Since the I/O
* that was sent to the vdev layer might be different than the one
* currently active in the pipeline (see vdev_queue_io()), we explicitly
* force the underlying vdev layers to call either zio_execute() or
* zio_interrupt() to ensure that the pipeline continues with the correct I/O.
*/
static zio_t *
zio_vdev_io_start(zio_t *zio)
{
vdev_t *vd = zio->io_vd;
uint64_t align;
spa_t *spa = zio->io_spa;
zio->io_delay = 0;
ASSERT(zio->io_error == 0);
ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0);
if (vd == NULL) {
if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
spa_config_enter(spa, SCL_ZIO, zio, RW_READER);
/*
* The mirror_ops handle multiple DVAs in a single BP.
*/
vdev_mirror_ops.vdev_op_io_start(zio);
return (NULL);
}
ASSERT3P(zio->io_logical, !=, zio);
if (zio->io_type == ZIO_TYPE_WRITE) {
ASSERT(spa->spa_trust_config);
/*
* Note: the code can handle other kinds of writes,
* but we don't expect them.
*/
if (zio->io_vd->vdev_noalloc) {
ASSERT(zio->io_flags &
(ZIO_FLAG_PHYSICAL | ZIO_FLAG_SELF_HEAL |
ZIO_FLAG_RESILVER | ZIO_FLAG_INDUCE_DAMAGE));
}
}
align = 1ULL << vd->vdev_top->vdev_ashift;
if (!(zio->io_flags & ZIO_FLAG_PHYSICAL) &&
P2PHASE(zio->io_size, align) != 0) {
/* Transform logical writes to be a full physical block size. */
uint64_t asize = P2ROUNDUP(zio->io_size, align);
abd_t *abuf = abd_alloc_sametype(zio->io_abd, asize);
ASSERT(vd == vd->vdev_top);
if (zio->io_type == ZIO_TYPE_WRITE) {
abd_copy(abuf, zio->io_abd, zio->io_size);
abd_zero_off(abuf, zio->io_size, asize - zio->io_size);
}
zio_push_transform(zio, abuf, asize, asize, zio_subblock);
}
/*
* If this is not a physical io, make sure that it is properly aligned
* before proceeding.
*/
if (!(zio->io_flags & ZIO_FLAG_PHYSICAL)) {
ASSERT0(P2PHASE(zio->io_offset, align));
ASSERT0(P2PHASE(zio->io_size, align));
} else {
/*
* For physical writes, we allow 512b aligned writes and assume
* the device will perform a read-modify-write as necessary.
*/
ASSERT0(P2PHASE(zio->io_offset, SPA_MINBLOCKSIZE));
ASSERT0(P2PHASE(zio->io_size, SPA_MINBLOCKSIZE));
}
VERIFY(zio->io_type != ZIO_TYPE_WRITE || spa_writeable(spa));
/*
* If this is a repair I/O, and there's no self-healing involved --
* that is, we're just resilvering what we expect to resilver --
* then don't do the I/O unless zio's txg is actually in vd's DTL.
* This prevents spurious resilvering.
*
* There are a few ways that we can end up creating these spurious
* resilver i/os:
*
* 1. A resilver i/o will be issued if any DVA in the BP has a
* dirty DTL. The mirror code will issue resilver writes to
* each DVA, including the one(s) that are not on vdevs with dirty
* DTLs.
*
* 2. With nested replication, which happens when we have a
* "replacing" or "spare" vdev that's a child of a mirror or raidz.
* For example, given mirror(replacing(A+B), C), it's likely that
* only A is out of date (it's the new device). In this case, we'll
* read from C, then use the data to resilver A+B -- but we don't
* actually want to resilver B, just A. The top-level mirror has no
* way to know this, so instead we just discard unnecessary repairs
* as we work our way down the vdev tree.
*
* 3. ZTEST also creates mirrors of mirrors, mirrors of raidz, etc.
* The same logic applies to any form of nested replication: ditto
* + mirror, RAID-Z + replacing, etc.
*
* However, indirect vdevs point off to other vdevs which may have
* DTL's, so we never bypass them. The child i/os on concrete vdevs
* will be properly bypassed instead.
*
* Leaf DTL_PARTIAL can be empty when a legitimate write comes from
* a dRAID spare vdev. For example, when a dRAID spare is first
* used, its spare blocks need to be written to but the leaf vdev's
* of such blocks can have empty DTL_PARTIAL.
*
* There seemed no clean way to allow such writes while bypassing
* spurious ones. At this point, just avoid all bypassing for dRAID
* for correctness.
*/
if ((zio->io_flags & ZIO_FLAG_IO_REPAIR) &&
!(zio->io_flags & ZIO_FLAG_SELF_HEAL) &&
zio->io_txg != 0 && /* not a delegated i/o */
vd->vdev_ops != &vdev_indirect_ops &&
vd->vdev_top->vdev_ops != &vdev_draid_ops &&
!vdev_dtl_contains(vd, DTL_PARTIAL, zio->io_txg, 1)) {
ASSERT(zio->io_type == ZIO_TYPE_WRITE);
zio_vdev_io_bypass(zio);
return (zio);
}
/*
* Select the next best leaf I/O to process. Distributed spares are
* excluded since they dispatch the I/O directly to a leaf vdev after
* applying the dRAID mapping.
*/
if (vd->vdev_ops->vdev_op_leaf &&
vd->vdev_ops != &vdev_draid_spare_ops &&
(zio->io_type == ZIO_TYPE_READ ||
zio->io_type == ZIO_TYPE_WRITE ||
zio->io_type == ZIO_TYPE_TRIM)) {
if (zio->io_type == ZIO_TYPE_READ && vdev_cache_read(zio))
return (zio);
if ((zio = vdev_queue_io(zio)) == NULL)
return (NULL);
if (!vdev_accessible(vd, zio)) {
zio->io_error = SET_ERROR(ENXIO);
zio_interrupt(zio);
return (NULL);
}
zio->io_delay = gethrtime();
}
vd->vdev_ops->vdev_op_io_start(zio);
return (NULL);
}
static zio_t *
zio_vdev_io_done(zio_t *zio)
{
vdev_t *vd = zio->io_vd;
vdev_ops_t *ops = vd ? vd->vdev_ops : &vdev_mirror_ops;
boolean_t unexpected_error = B_FALSE;
if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) {
return (NULL);
}
ASSERT(zio->io_type == ZIO_TYPE_READ ||
zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_TRIM);
if (zio->io_delay)
zio->io_delay = gethrtime() - zio->io_delay;
if (vd != NULL && vd->vdev_ops->vdev_op_leaf &&
vd->vdev_ops != &vdev_draid_spare_ops) {
vdev_queue_io_done(zio);
if (zio->io_type == ZIO_TYPE_WRITE)
vdev_cache_write(zio);
if (zio_injection_enabled && zio->io_error == 0)
zio->io_error = zio_handle_device_injections(vd, zio,
EIO, EILSEQ);
if (zio_injection_enabled && zio->io_error == 0)
zio->io_error = zio_handle_label_injection(zio, EIO);
if (zio->io_error && zio->io_type != ZIO_TYPE_TRIM) {
if (!vdev_accessible(vd, zio)) {
zio->io_error = SET_ERROR(ENXIO);
} else {
unexpected_error = B_TRUE;
}
}
}
ops->vdev_op_io_done(zio);
if (unexpected_error)
VERIFY(vdev_probe(vd, zio) == NULL);
return (zio);
}
/*
* This function is used to change the priority of an existing zio that is
* currently in-flight. This is used by the arc to upgrade priority in the
* event that a demand read is made for a block that is currently queued
* as a scrub or async read IO. Otherwise, the high priority read request
* would end up having to wait for the lower priority IO.
*/
void
zio_change_priority(zio_t *pio, zio_priority_t priority)
{
zio_t *cio, *cio_next;
zio_link_t *zl = NULL;
ASSERT3U(priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
if (pio->io_vd != NULL && pio->io_vd->vdev_ops->vdev_op_leaf) {
vdev_queue_change_io_priority(pio, priority);
} else {
pio->io_priority = priority;
}
mutex_enter(&pio->io_lock);
for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
cio_next = zio_walk_children(pio, &zl);
zio_change_priority(cio, priority);
}
mutex_exit(&pio->io_lock);
}
/*
* For non-raidz ZIOs, we can just copy aside the bad data read from the
* disk, and use that to finish the checksum ereport later.
*/
static void
zio_vsd_default_cksum_finish(zio_cksum_report_t *zcr,
const abd_t *good_buf)
{
/* no processing needed */
zfs_ereport_finish_checksum(zcr, good_buf, zcr->zcr_cbdata, B_FALSE);
}
void
zio_vsd_default_cksum_report(zio_t *zio, zio_cksum_report_t *zcr)
{
void *abd = abd_alloc_sametype(zio->io_abd, zio->io_size);
abd_copy(abd, zio->io_abd, zio->io_size);
zcr->zcr_cbinfo = zio->io_size;
zcr->zcr_cbdata = abd;
zcr->zcr_finish = zio_vsd_default_cksum_finish;
zcr->zcr_free = zio_abd_free;
}
static zio_t *
zio_vdev_io_assess(zio_t *zio)
{
vdev_t *vd = zio->io_vd;
if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) {
return (NULL);
}
if (vd == NULL && !(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
spa_config_exit(zio->io_spa, SCL_ZIO, zio);
if (zio->io_vsd != NULL) {
zio->io_vsd_ops->vsd_free(zio);
zio->io_vsd = NULL;
}
if (zio_injection_enabled && zio->io_error == 0)
zio->io_error = zio_handle_fault_injection(zio, EIO);
/*
* If the I/O failed, determine whether we should attempt to retry it.
*
* On retry, we cut in line in the issue queue, since we don't want
* compression/checksumming/etc. work to prevent our (cheap) IO reissue.
*/
if (zio->io_error && vd == NULL &&
!(zio->io_flags & (ZIO_FLAG_DONT_RETRY | ZIO_FLAG_IO_RETRY))) {
ASSERT(!(zio->io_flags & ZIO_FLAG_DONT_QUEUE)); /* not a leaf */
ASSERT(!(zio->io_flags & ZIO_FLAG_IO_BYPASS)); /* not a leaf */
zio->io_error = 0;
zio->io_flags |= ZIO_FLAG_IO_RETRY |
ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_AGGREGATE;
zio->io_stage = ZIO_STAGE_VDEV_IO_START >> 1;
zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE,
zio_requeue_io_start_cut_in_line);
return (NULL);
}
/*
* If we got an error on a leaf device, convert it to ENXIO
* if the device is not accessible at all.
*/
if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf &&
!vdev_accessible(vd, zio))
zio->io_error = SET_ERROR(ENXIO);
/*
* If we can't write to an interior vdev (mirror or RAID-Z),
* set vdev_cant_write so that we stop trying to allocate from it.
*/
if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE &&
vd != NULL && !vd->vdev_ops->vdev_op_leaf) {
vdev_dbgmsg(vd, "zio_vdev_io_assess(zio=%px) setting "
"cant_write=TRUE due to write failure with ENXIO",
zio);
vd->vdev_cant_write = B_TRUE;
}
/*
* If a cache flush returns ENOTSUP or ENOTTY, we know that no future
* attempts will ever succeed. In this case we set a persistent
* boolean flag so that we don't bother with it in the future.
*/
if ((zio->io_error == ENOTSUP || zio->io_error == ENOTTY) &&
zio->io_type == ZIO_TYPE_IOCTL &&
zio->io_cmd == DKIOCFLUSHWRITECACHE && vd != NULL)
vd->vdev_nowritecache = B_TRUE;
if (zio->io_error)
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
if (vd != NULL && vd->vdev_ops->vdev_op_leaf &&
zio->io_physdone != NULL) {
ASSERT(!(zio->io_flags & ZIO_FLAG_DELEGATED));
ASSERT(zio->io_child_type == ZIO_CHILD_VDEV);
zio->io_physdone(zio->io_logical);
}
return (zio);
}
void
zio_vdev_io_reissue(zio_t *zio)
{
ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
ASSERT(zio->io_error == 0);
zio->io_stage >>= 1;
}
void
zio_vdev_io_redone(zio_t *zio)
{
ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_DONE);
zio->io_stage >>= 1;
}
void
zio_vdev_io_bypass(zio_t *zio)
{
ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
ASSERT(zio->io_error == 0);
zio->io_flags |= ZIO_FLAG_IO_BYPASS;
zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS >> 1;
}
/*
* ==========================================================================
* Encrypt and store encryption parameters
* ==========================================================================
*/
/*
* This function is used for ZIO_STAGE_ENCRYPT. It is responsible for
* managing the storage of encryption parameters and passing them to the
* lower-level encryption functions.
*/
static zio_t *
zio_encrypt(zio_t *zio)
{
zio_prop_t *zp = &zio->io_prop;
spa_t *spa = zio->io_spa;
blkptr_t *bp = zio->io_bp;
uint64_t psize = BP_GET_PSIZE(bp);
uint64_t dsobj = zio->io_bookmark.zb_objset;
dmu_object_type_t ot = BP_GET_TYPE(bp);
void *enc_buf = NULL;
abd_t *eabd = NULL;
uint8_t salt[ZIO_DATA_SALT_LEN];
uint8_t iv[ZIO_DATA_IV_LEN];
uint8_t mac[ZIO_DATA_MAC_LEN];
boolean_t no_crypt = B_FALSE;
/* the root zio already encrypted the data */
if (zio->io_child_type == ZIO_CHILD_GANG)
return (zio);
/* only ZIL blocks are re-encrypted on rewrite */
if (!IO_IS_ALLOCATING(zio) && ot != DMU_OT_INTENT_LOG)
return (zio);
if (!(zp->zp_encrypt || BP_IS_ENCRYPTED(bp))) {
BP_SET_CRYPT(bp, B_FALSE);
return (zio);
}
/* if we are doing raw encryption set the provided encryption params */
if (zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) {
ASSERT0(BP_GET_LEVEL(bp));
BP_SET_CRYPT(bp, B_TRUE);
BP_SET_BYTEORDER(bp, zp->zp_byteorder);
if (ot != DMU_OT_OBJSET)
zio_crypt_encode_mac_bp(bp, zp->zp_mac);
/* dnode blocks must be written out in the provided byteorder */
if (zp->zp_byteorder != ZFS_HOST_BYTEORDER &&
ot == DMU_OT_DNODE) {
void *bswap_buf = zio_buf_alloc(psize);
abd_t *babd = abd_get_from_buf(bswap_buf, psize);
ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
abd_copy_to_buf(bswap_buf, zio->io_abd, psize);
dmu_ot_byteswap[DMU_OT_BYTESWAP(ot)].ob_func(bswap_buf,
psize);
abd_take_ownership_of_buf(babd, B_TRUE);
zio_push_transform(zio, babd, psize, psize, NULL);
}
if (DMU_OT_IS_ENCRYPTED(ot))
zio_crypt_encode_params_bp(bp, zp->zp_salt, zp->zp_iv);
return (zio);
}
/* indirect blocks only maintain a cksum of the lower level MACs */
if (BP_GET_LEVEL(bp) > 0) {
BP_SET_CRYPT(bp, B_TRUE);
VERIFY0(zio_crypt_do_indirect_mac_checksum_abd(B_TRUE,
zio->io_orig_abd, BP_GET_LSIZE(bp), BP_SHOULD_BYTESWAP(bp),
mac));
zio_crypt_encode_mac_bp(bp, mac);
return (zio);
}
/*
* Objset blocks are a special case since they have 2 256-bit MACs
* embedded within them.
*/
if (ot == DMU_OT_OBJSET) {
ASSERT0(DMU_OT_IS_ENCRYPTED(ot));
ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
BP_SET_CRYPT(bp, B_TRUE);
VERIFY0(spa_do_crypt_objset_mac_abd(B_TRUE, spa, dsobj,
zio->io_abd, psize, BP_SHOULD_BYTESWAP(bp)));
return (zio);
}
/* unencrypted object types are only authenticated with a MAC */
if (!DMU_OT_IS_ENCRYPTED(ot)) {
BP_SET_CRYPT(bp, B_TRUE);
VERIFY0(spa_do_crypt_mac_abd(B_TRUE, spa, dsobj,
zio->io_abd, psize, mac));
zio_crypt_encode_mac_bp(bp, mac);
return (zio);
}
/*
* Later passes of sync-to-convergence may decide to rewrite data
* in place to avoid more disk reallocations. This presents a problem
* for encryption because this constitutes rewriting the new data with
* the same encryption key and IV. However, this only applies to blocks
* in the MOS (particularly the spacemaps) and we do not encrypt the
* MOS. We assert that the zio is allocating or an intent log write
* to enforce this.
*/
ASSERT(IO_IS_ALLOCATING(zio) || ot == DMU_OT_INTENT_LOG);
ASSERT(BP_GET_LEVEL(bp) == 0 || ot == DMU_OT_INTENT_LOG);
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_ENCRYPTION));
ASSERT3U(psize, !=, 0);
enc_buf = zio_buf_alloc(psize);
eabd = abd_get_from_buf(enc_buf, psize);
abd_take_ownership_of_buf(eabd, B_TRUE);
/*
* For an explanation of what encryption parameters are stored
* where, see the block comment in zio_crypt.c.
*/
if (ot == DMU_OT_INTENT_LOG) {
zio_crypt_decode_params_bp(bp, salt, iv);
} else {
BP_SET_CRYPT(bp, B_TRUE);
}
/* Perform the encryption. This should not fail */
VERIFY0(spa_do_crypt_abd(B_TRUE, spa, &zio->io_bookmark,
BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp),
salt, iv, mac, psize, zio->io_abd, eabd, &no_crypt));
/* encode encryption metadata into the bp */
if (ot == DMU_OT_INTENT_LOG) {
/*
* ZIL blocks store the MAC in the embedded checksum, so the
* transform must always be applied.
*/
zio_crypt_encode_mac_zil(enc_buf, mac);
zio_push_transform(zio, eabd, psize, psize, NULL);
} else {
BP_SET_CRYPT(bp, B_TRUE);
zio_crypt_encode_params_bp(bp, salt, iv);
zio_crypt_encode_mac_bp(bp, mac);
if (no_crypt) {
ASSERT3U(ot, ==, DMU_OT_DNODE);
abd_free(eabd);
} else {
zio_push_transform(zio, eabd, psize, psize, NULL);
}
}
return (zio);
}
/*
* ==========================================================================
* Generate and verify checksums
* ==========================================================================
*/
static zio_t *
zio_checksum_generate(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
enum zio_checksum checksum;
if (bp == NULL) {
/*
* This is zio_write_phys().
* We're either generating a label checksum, or none at all.
*/
checksum = zio->io_prop.zp_checksum;
if (checksum == ZIO_CHECKSUM_OFF)
return (zio);
ASSERT(checksum == ZIO_CHECKSUM_LABEL);
} else {
if (BP_IS_GANG(bp) && zio->io_child_type == ZIO_CHILD_GANG) {
ASSERT(!IO_IS_ALLOCATING(zio));
checksum = ZIO_CHECKSUM_GANG_HEADER;
} else {
checksum = BP_GET_CHECKSUM(bp);
}
}
zio_checksum_compute(zio, checksum, zio->io_abd, zio->io_size);
return (zio);
}
static zio_t *
zio_checksum_verify(zio_t *zio)
{
zio_bad_cksum_t info;
blkptr_t *bp = zio->io_bp;
int error;
ASSERT(zio->io_vd != NULL);
if (bp == NULL) {
/*
* This is zio_read_phys().
* We're either verifying a label checksum, or nothing at all.
*/
if (zio->io_prop.zp_checksum == ZIO_CHECKSUM_OFF)
return (zio);
ASSERT3U(zio->io_prop.zp_checksum, ==, ZIO_CHECKSUM_LABEL);
}
if ((error = zio_checksum_error(zio, &info)) != 0) {
zio->io_error = error;
if (error == ECKSUM &&
!(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
(void) zfs_ereport_start_checksum(zio->io_spa,
zio->io_vd, &zio->io_bookmark, zio,
zio->io_offset, zio->io_size, &info);
mutex_enter(&zio->io_vd->vdev_stat_lock);
zio->io_vd->vdev_stat.vs_checksum_errors++;
mutex_exit(&zio->io_vd->vdev_stat_lock);
}
}
return (zio);
}
/*
* Called by RAID-Z to ensure we don't compute the checksum twice.
*/
void
zio_checksum_verified(zio_t *zio)
{
zio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
}
/*
* ==========================================================================
* Error rank. Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other.
* An error of 0 indicates success. ENXIO indicates whole-device failure,
* which may be transient (e.g. unplugged) or permanent. ECKSUM and EIO
* indicate errors that are specific to one I/O, and most likely permanent.
* Any other error is presumed to be worse because we weren't expecting it.
* ==========================================================================
*/
int
zio_worst_error(int e1, int e2)
{
static int zio_error_rank[] = { 0, ENXIO, ECKSUM, EIO };
int r1, r2;
for (r1 = 0; r1 < sizeof (zio_error_rank) / sizeof (int); r1++)
if (e1 == zio_error_rank[r1])
break;
for (r2 = 0; r2 < sizeof (zio_error_rank) / sizeof (int); r2++)
if (e2 == zio_error_rank[r2])
break;
return (r1 > r2 ? e1 : e2);
}
/*
* ==========================================================================
* I/O completion
* ==========================================================================
*/
static zio_t *
zio_ready(zio_t *zio)
{
blkptr_t *bp = zio->io_bp;
zio_t *pio, *pio_next;
zio_link_t *zl = NULL;
if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT | ZIO_CHILD_DDT_BIT,
ZIO_WAIT_READY)) {
return (NULL);
}
if (zio->io_ready) {
ASSERT(IO_IS_ALLOCATING(zio));
ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp) ||
(zio->io_flags & ZIO_FLAG_NOPWRITE));
ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0);
zio->io_ready(zio);
}
if (bp != NULL && bp != &zio->io_bp_copy)
zio->io_bp_copy = *bp;
if (zio->io_error != 0) {
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
ASSERT(IO_IS_ALLOCATING(zio));
ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
ASSERT(zio->io_metaslab_class != NULL);
/*
* We were unable to allocate anything, unreserve and
* issue the next I/O to allocate.
*/
metaslab_class_throttle_unreserve(
zio->io_metaslab_class, zio->io_prop.zp_copies,
zio->io_allocator, zio);
zio_allocate_dispatch(zio->io_spa, zio->io_allocator);
}
}
mutex_enter(&zio->io_lock);
zio->io_state[ZIO_WAIT_READY] = 1;
pio = zio_walk_parents(zio, &zl);
mutex_exit(&zio->io_lock);
/*
* As we notify zio's parents, new parents could be added.
* New parents go to the head of zio's io_parent_list, however,
* so we will (correctly) not notify them. The remainder of zio's
* io_parent_list, from 'pio_next' onward, cannot change because
* all parents must wait for us to be done before they can be done.
*/
for (; pio != NULL; pio = pio_next) {
pio_next = zio_walk_parents(zio, &zl);
zio_notify_parent(pio, zio, ZIO_WAIT_READY, NULL);
}
if (zio->io_flags & ZIO_FLAG_NODATA) {
if (BP_IS_GANG(bp)) {
zio->io_flags &= ~ZIO_FLAG_NODATA;
} else {
ASSERT((uintptr_t)zio->io_abd < SPA_MAXBLOCKSIZE);
zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
}
}
if (zio_injection_enabled &&
zio->io_spa->spa_syncing_txg == zio->io_txg)
zio_handle_ignored_writes(zio);
return (zio);
}
/*
* Update the allocation throttle accounting.
*/
static void
zio_dva_throttle_done(zio_t *zio)
{
zio_t *lio __maybe_unused = zio->io_logical;
zio_t *pio = zio_unique_parent(zio);
vdev_t *vd = zio->io_vd;
int flags = METASLAB_ASYNC_ALLOC;
ASSERT3P(zio->io_bp, !=, NULL);
ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);
ASSERT3U(zio->io_priority, ==, ZIO_PRIORITY_ASYNC_WRITE);
ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV);
ASSERT(vd != NULL);
ASSERT3P(vd, ==, vd->vdev_top);
ASSERT(zio_injection_enabled || !(zio->io_flags & ZIO_FLAG_IO_RETRY));
ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR));
ASSERT(zio->io_flags & ZIO_FLAG_IO_ALLOCATING);
ASSERT(!(lio->io_flags & ZIO_FLAG_IO_REWRITE));
ASSERT(!(lio->io_orig_flags & ZIO_FLAG_NODATA));
/*
* Parents of gang children can have two flavors -- ones that
* allocated the gang header (will have ZIO_FLAG_IO_REWRITE set)
* and ones that allocated the constituent blocks. The allocation
* throttle needs to know the allocating parent zio so we must find
* it here.
*/
if (pio->io_child_type == ZIO_CHILD_GANG) {
/*
* If our parent is a rewrite gang child then our grandparent
* would have been the one that performed the allocation.
*/
if (pio->io_flags & ZIO_FLAG_IO_REWRITE)
pio = zio_unique_parent(pio);
flags |= METASLAB_GANG_CHILD;
}
ASSERT(IO_IS_ALLOCATING(pio));
ASSERT3P(zio, !=, zio->io_logical);
ASSERT(zio->io_logical != NULL);
ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR));
ASSERT0(zio->io_flags & ZIO_FLAG_NOPWRITE);
ASSERT(zio->io_metaslab_class != NULL);
mutex_enter(&pio->io_lock);
metaslab_group_alloc_decrement(zio->io_spa, vd->vdev_id, pio, flags,
pio->io_allocator, B_TRUE);
mutex_exit(&pio->io_lock);
metaslab_class_throttle_unreserve(zio->io_metaslab_class, 1,
pio->io_allocator, pio);
/*
* Call into the pipeline to see if there is more work that
* needs to be done. If there is work to be done it will be
* dispatched to another taskq thread.
*/
zio_allocate_dispatch(zio->io_spa, pio->io_allocator);
}
static zio_t *
zio_done(zio_t *zio)
{
/*
* Always attempt to keep stack usage minimal here since
* we can be called recursively up to 19 levels deep.
*/
const uint64_t psize = zio->io_size;
zio_t *pio, *pio_next;
zio_link_t *zl = NULL;
/*
* If our children haven't all completed,
* wait for them and then repeat this pipeline stage.
*/
if (zio_wait_for_children(zio, ZIO_CHILD_ALL_BITS, ZIO_WAIT_DONE)) {
return (NULL);
}
/*
* If the allocation throttle is enabled, then update the accounting.
* We only track child I/Os that are part of an allocating async
* write. We must do this since the allocation is performed
* by the logical I/O but the actual write is done by child I/Os.
*/
if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING &&
zio->io_child_type == ZIO_CHILD_VDEV) {
ASSERT(zio->io_metaslab_class != NULL);
ASSERT(zio->io_metaslab_class->mc_alloc_throttle_enabled);
zio_dva_throttle_done(zio);
}
/*
* If the allocation throttle is enabled, verify that
* we have decremented the refcounts for every I/O that was throttled.
*/
if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
ASSERT(zio->io_type == ZIO_TYPE_WRITE);
ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
ASSERT(zio->io_bp != NULL);
metaslab_group_alloc_verify(zio->io_spa, zio->io_bp, zio,
zio->io_allocator);
VERIFY(zfs_refcount_not_held(&zio->io_metaslab_class->
mc_allocator[zio->io_allocator].mca_alloc_slots, zio));
}
for (int c = 0; c < ZIO_CHILD_TYPES; c++)
for (int w = 0; w < ZIO_WAIT_TYPES; w++)
ASSERT(zio->io_children[c][w] == 0);
if (zio->io_bp != NULL && !BP_IS_EMBEDDED(zio->io_bp)) {
ASSERT(zio->io_bp->blk_pad[0] == 0);
ASSERT(zio->io_bp->blk_pad[1] == 0);
ASSERT(bcmp(zio->io_bp, &zio->io_bp_copy,
sizeof (blkptr_t)) == 0 ||
(zio->io_bp == zio_unique_parent(zio)->io_bp));
if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(zio->io_bp) &&
zio->io_bp_override == NULL &&
!(zio->io_flags & ZIO_FLAG_IO_REPAIR)) {
ASSERT3U(zio->io_prop.zp_copies, <=,
BP_GET_NDVAS(zio->io_bp));
ASSERT(BP_COUNT_GANG(zio->io_bp) == 0 ||
(BP_COUNT_GANG(zio->io_bp) ==
BP_GET_NDVAS(zio->io_bp)));
}
if (zio->io_flags & ZIO_FLAG_NOPWRITE)
VERIFY(BP_EQUAL(zio->io_bp, &zio->io_bp_orig));
}
/*
* If there were child vdev/gang/ddt errors, they apply to us now.
*/
zio_inherit_child_errors(zio, ZIO_CHILD_VDEV);
zio_inherit_child_errors(zio, ZIO_CHILD_GANG);
zio_inherit_child_errors(zio, ZIO_CHILD_DDT);
/*
* If the I/O on the transformed data was successful, generate any
* checksum reports now while we still have the transformed data.
*/
if (zio->io_error == 0) {
while (zio->io_cksum_report != NULL) {
zio_cksum_report_t *zcr = zio->io_cksum_report;
uint64_t align = zcr->zcr_align;
uint64_t asize = P2ROUNDUP(psize, align);
abd_t *adata = zio->io_abd;
if (adata != NULL && asize != psize) {
adata = abd_alloc(asize, B_TRUE);
abd_copy(adata, zio->io_abd, psize);
abd_zero_off(adata, psize, asize - psize);
}
zio->io_cksum_report = zcr->zcr_next;
zcr->zcr_next = NULL;
zcr->zcr_finish(zcr, adata);
zfs_ereport_free_checksum(zcr);
if (adata != NULL && asize != psize)
abd_free(adata);
}
}
zio_pop_transforms(zio); /* note: may set zio->io_error */
vdev_stat_update(zio, psize);
/*
* If this I/O is attached to a particular vdev is slow, exceeding
* 30 seconds to complete, post an error described the I/O delay.
* We ignore these errors if the device is currently unavailable.
*/
if (zio->io_delay >= MSEC2NSEC(zio_slow_io_ms)) {
if (zio->io_vd != NULL && !vdev_is_dead(zio->io_vd)) {
/*
* We want to only increment our slow IO counters if
* the IO is valid (i.e. not if the drive is removed).
*
* zfs_ereport_post() will also do these checks, but
* it can also ratelimit and have other failures, so we
* need to increment the slow_io counters independent
* of it.
*/
if (zfs_ereport_is_valid(FM_EREPORT_ZFS_DELAY,
zio->io_spa, zio->io_vd, zio)) {
mutex_enter(&zio->io_vd->vdev_stat_lock);
zio->io_vd->vdev_stat.vs_slow_ios++;
mutex_exit(&zio->io_vd->vdev_stat_lock);
(void) zfs_ereport_post(FM_EREPORT_ZFS_DELAY,
zio->io_spa, zio->io_vd, &zio->io_bookmark,
zio, 0);
}
}
}
if (zio->io_error) {
/*
* If this I/O is attached to a particular vdev,
* generate an error message describing the I/O failure
* at the block level. We ignore these errors if the
* device is currently unavailable.
*/
if (zio->io_error != ECKSUM && zio->io_vd != NULL &&
!vdev_is_dead(zio->io_vd)) {
int ret = zfs_ereport_post(FM_EREPORT_ZFS_IO,
zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, 0);
if (ret != EALREADY) {
mutex_enter(&zio->io_vd->vdev_stat_lock);
if (zio->io_type == ZIO_TYPE_READ)
zio->io_vd->vdev_stat.vs_read_errors++;
else if (zio->io_type == ZIO_TYPE_WRITE)
zio->io_vd->vdev_stat.vs_write_errors++;
mutex_exit(&zio->io_vd->vdev_stat_lock);
}
}
if ((zio->io_error == EIO || !(zio->io_flags &
(ZIO_FLAG_SPECULATIVE | ZIO_FLAG_DONT_PROPAGATE))) &&
zio == zio->io_logical) {
/*
* For logical I/O requests, tell the SPA to log the
* error and generate a logical data ereport.
*/
spa_log_error(zio->io_spa, &zio->io_bookmark);
(void) zfs_ereport_post(FM_EREPORT_ZFS_DATA,
zio->io_spa, NULL, &zio->io_bookmark, zio, 0);
}
}
if (zio->io_error && zio == zio->io_logical) {
/*
* Determine whether zio should be reexecuted. This will
* propagate all the way to the root via zio_notify_parent().
*/
ASSERT(zio->io_vd == NULL && zio->io_bp != NULL);
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
if (IO_IS_ALLOCATING(zio) &&
!(zio->io_flags & ZIO_FLAG_CANFAIL)) {
if (zio->io_error != ENOSPC)
zio->io_reexecute |= ZIO_REEXECUTE_NOW;
else
zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND;
}
if ((zio->io_type == ZIO_TYPE_READ ||
zio->io_type == ZIO_TYPE_FREE) &&
!(zio->io_flags & ZIO_FLAG_SCAN_THREAD) &&
zio->io_error == ENXIO &&
spa_load_state(zio->io_spa) == SPA_LOAD_NONE &&
spa_get_failmode(zio->io_spa) != ZIO_FAILURE_MODE_CONTINUE)
zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND;
if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute)
zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND;
/*
* Here is a possibly good place to attempt to do
* either combinatorial reconstruction or error correction
* based on checksums. It also might be a good place
* to send out preliminary ereports before we suspend
* processing.
*/
}
/*
* If there were logical child errors, they apply to us now.
* We defer this until now to avoid conflating logical child
* errors with errors that happened to the zio itself when
* updating vdev stats and reporting FMA events above.
*/
zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL);
if ((zio->io_error || zio->io_reexecute) &&
IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio &&
!(zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE)))
zio_dva_unallocate(zio, zio->io_gang_tree, zio->io_bp);
zio_gang_tree_free(&zio->io_gang_tree);
/*
* Godfather I/Os should never suspend.
*/
if ((zio->io_flags & ZIO_FLAG_GODFATHER) &&
(zio->io_reexecute & ZIO_REEXECUTE_SUSPEND))
zio->io_reexecute &= ~ZIO_REEXECUTE_SUSPEND;
if (zio->io_reexecute) {
/*
* This is a logical I/O that wants to reexecute.
*
* Reexecute is top-down. When an i/o fails, if it's not
* the root, it simply notifies its parent and sticks around.
* The parent, seeing that it still has children in zio_done(),
* does the same. This percolates all the way up to the root.
* The root i/o will reexecute or suspend the entire tree.
*
* This approach ensures that zio_reexecute() honors
* all the original i/o dependency relationships, e.g.
* parents not executing until children are ready.
*/
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
zio->io_gang_leader = NULL;
mutex_enter(&zio->io_lock);
zio->io_state[ZIO_WAIT_DONE] = 1;
mutex_exit(&zio->io_lock);
/*
* "The Godfather" I/O monitors its children but is
* not a true parent to them. It will track them through
* the pipeline but severs its ties whenever they get into
* trouble (e.g. suspended). This allows "The Godfather"
* I/O to return status without blocking.
*/
zl = NULL;
for (pio = zio_walk_parents(zio, &zl); pio != NULL;
pio = pio_next) {
zio_link_t *remove_zl = zl;
pio_next = zio_walk_parents(zio, &zl);
if ((pio->io_flags & ZIO_FLAG_GODFATHER) &&
(zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) {
zio_remove_child(pio, zio, remove_zl);
/*
* This is a rare code path, so we don't
* bother with "next_to_execute".
*/
zio_notify_parent(pio, zio, ZIO_WAIT_DONE,
NULL);
}
}
if ((pio = zio_unique_parent(zio)) != NULL) {
/*
* We're not a root i/o, so there's nothing to do
* but notify our parent. Don't propagate errors
* upward since we haven't permanently failed yet.
*/
ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER));
zio->io_flags |= ZIO_FLAG_DONT_PROPAGATE;
/*
* This is a rare code path, so we don't bother with
* "next_to_execute".
*/
zio_notify_parent(pio, zio, ZIO_WAIT_DONE, NULL);
} else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) {
/*
* We'd fail again if we reexecuted now, so suspend
* until conditions improve (e.g. device comes online).
*/
zio_suspend(zio->io_spa, zio, ZIO_SUSPEND_IOERR);
} else {
/*
* Reexecution is potentially a huge amount of work.
* Hand it off to the otherwise-unused claim taskq.
*/
ASSERT(taskq_empty_ent(&zio->io_tqent));
spa_taskq_dispatch_ent(zio->io_spa,
ZIO_TYPE_CLAIM, ZIO_TASKQ_ISSUE,
zio_reexecute, zio, 0, &zio->io_tqent);
}
return (NULL);
}
ASSERT(zio->io_child_count == 0);
ASSERT(zio->io_reexecute == 0);
ASSERT(zio->io_error == 0 || (zio->io_flags & ZIO_FLAG_CANFAIL));
/*
* Report any checksum errors, since the I/O is complete.
*/
while (zio->io_cksum_report != NULL) {
zio_cksum_report_t *zcr = zio->io_cksum_report;
zio->io_cksum_report = zcr->zcr_next;
zcr->zcr_next = NULL;
zcr->zcr_finish(zcr, NULL);
zfs_ereport_free_checksum(zcr);
}
if (zio->io_flags & ZIO_FLAG_FASTWRITE && zio->io_bp &&
!BP_IS_HOLE(zio->io_bp) && !BP_IS_EMBEDDED(zio->io_bp) &&
!(zio->io_flags & ZIO_FLAG_NOPWRITE)) {
metaslab_fastwrite_unmark(zio->io_spa, zio->io_bp);
}
/*
* It is the responsibility of the done callback to ensure that this
* particular zio is no longer discoverable for adoption, and as
* such, cannot acquire any new parents.
*/
if (zio->io_done)
zio->io_done(zio);
mutex_enter(&zio->io_lock);
zio->io_state[ZIO_WAIT_DONE] = 1;
mutex_exit(&zio->io_lock);
/*
* We are done executing this zio. We may want to execute a parent
* next. See the comment in zio_notify_parent().
*/
zio_t *next_to_execute = NULL;
zl = NULL;
for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) {
zio_link_t *remove_zl = zl;
pio_next = zio_walk_parents(zio, &zl);
zio_remove_child(pio, zio, remove_zl);
zio_notify_parent(pio, zio, ZIO_WAIT_DONE, &next_to_execute);
}
if (zio->io_waiter != NULL) {
mutex_enter(&zio->io_lock);
zio->io_executor = NULL;
cv_broadcast(&zio->io_cv);
mutex_exit(&zio->io_lock);
} else {
zio_destroy(zio);
}
return (next_to_execute);
}
/*
* ==========================================================================
* I/O pipeline definition
* ==========================================================================
*/
static zio_pipe_stage_t *zio_pipeline[] = {
NULL,
zio_read_bp_init,
zio_write_bp_init,
zio_free_bp_init,
zio_issue_async,
zio_write_compress,
zio_encrypt,
zio_checksum_generate,
zio_nop_write,
zio_ddt_read_start,
zio_ddt_read_done,
zio_ddt_write,
zio_ddt_free,
zio_gang_assemble,
zio_gang_issue,
zio_dva_throttle,
zio_dva_allocate,
zio_dva_free,
zio_dva_claim,
zio_ready,
zio_vdev_io_start,
zio_vdev_io_done,
zio_vdev_io_assess,
zio_checksum_verify,
zio_done
};
/*
* Compare two zbookmark_phys_t's to see which we would reach first in a
* pre-order traversal of the object tree.
*
* This is simple in every case aside from the meta-dnode object. For all other
* objects, we traverse them in order (object 1 before object 2, and so on).
* However, all of these objects are traversed while traversing object 0, since
* the data it points to is the list of objects. Thus, we need to convert to a
* canonical representation so we can compare meta-dnode bookmarks to
* non-meta-dnode bookmarks.
*
* We do this by calculating "equivalents" for each field of the zbookmark.
* zbookmarks outside of the meta-dnode use their own object and level, and
* calculate the level 0 equivalent (the first L0 blkid that is contained in the
* blocks this bookmark refers to) by multiplying their blkid by their span
* (the number of L0 blocks contained within one block at their level).
* zbookmarks inside the meta-dnode calculate their object equivalent
* (which is L0equiv * dnodes per data block), use 0 for their L0equiv, and use
* level + 1<<31 (any value larger than a level could ever be) for their level.
* This causes them to always compare before a bookmark in their object
* equivalent, compare appropriately to bookmarks in other objects, and to
* compare appropriately to other bookmarks in the meta-dnode.
*/
int
zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2,
const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2)
{
/*
* These variables represent the "equivalent" values for the zbookmark,
* after converting zbookmarks inside the meta dnode to their
* normal-object equivalents.
*/
uint64_t zb1obj, zb2obj;
uint64_t zb1L0, zb2L0;
uint64_t zb1level, zb2level;
if (zb1->zb_object == zb2->zb_object &&
zb1->zb_level == zb2->zb_level &&
zb1->zb_blkid == zb2->zb_blkid)
return (0);
IMPLY(zb1->zb_level > 0, ibs1 >= SPA_MINBLOCKSHIFT);
IMPLY(zb2->zb_level > 0, ibs2 >= SPA_MINBLOCKSHIFT);
/*
* BP_SPANB calculates the span in blocks.
*/
zb1L0 = (zb1->zb_blkid) * BP_SPANB(ibs1, zb1->zb_level);
zb2L0 = (zb2->zb_blkid) * BP_SPANB(ibs2, zb2->zb_level);
if (zb1->zb_object == DMU_META_DNODE_OBJECT) {
zb1obj = zb1L0 * (dbss1 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT));
zb1L0 = 0;
zb1level = zb1->zb_level + COMPARE_META_LEVEL;
} else {
zb1obj = zb1->zb_object;
zb1level = zb1->zb_level;
}
if (zb2->zb_object == DMU_META_DNODE_OBJECT) {
zb2obj = zb2L0 * (dbss2 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT));
zb2L0 = 0;
zb2level = zb2->zb_level + COMPARE_META_LEVEL;
} else {
zb2obj = zb2->zb_object;
zb2level = zb2->zb_level;
}
/* Now that we have a canonical representation, do the comparison. */
if (zb1obj != zb2obj)
return (zb1obj < zb2obj ? -1 : 1);
else if (zb1L0 != zb2L0)
return (zb1L0 < zb2L0 ? -1 : 1);
else if (zb1level != zb2level)
return (zb1level > zb2level ? -1 : 1);
/*
* This can (theoretically) happen if the bookmarks have the same object
* and level, but different blkids, if the block sizes are not the same.
* There is presently no way to change the indirect block sizes
*/
return (0);
}
/*
* This function checks the following: given that last_block is the place that
* our traversal stopped last time, does that guarantee that we've visited
* every node under subtree_root? Therefore, we can't just use the raw output
* of zbookmark_compare. We have to pass in a modified version of
* subtree_root; by incrementing the block id, and then checking whether
* last_block is before or equal to that, we can tell whether or not having
* visited last_block implies that all of subtree_root's children have been
* visited.
*/
boolean_t
zbookmark_subtree_completed(const dnode_phys_t *dnp,
const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block)
{
zbookmark_phys_t mod_zb = *subtree_root;
mod_zb.zb_blkid++;
ASSERT(last_block->zb_level == 0);
/* The objset_phys_t isn't before anything. */
if (dnp == NULL)
return (B_FALSE);
/*
* We pass in 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT) for the
* data block size in sectors, because that variable is only used if
* the bookmark refers to a block in the meta-dnode. Since we don't
* know without examining it what object it refers to, and there's no
* harm in passing in this value in other cases, we always pass it in.
*
* We pass in 0 for the indirect block size shift because zb2 must be
* level 0. The indirect block size is only used to calculate the span
* of the bookmark, but since the bookmark must be level 0, the span is
* always 1, so the math works out.
*
* If you make changes to how the zbookmark_compare code works, be sure
* to make sure that this code still works afterwards.
*/
return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift,
1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, &mod_zb,
last_block) <= 0);
}
EXPORT_SYMBOL(zio_type_name);
EXPORT_SYMBOL(zio_buf_alloc);
EXPORT_SYMBOL(zio_data_buf_alloc);
EXPORT_SYMBOL(zio_buf_free);
EXPORT_SYMBOL(zio_data_buf_free);
/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs_zio, zio_, slow_io_ms, INT, ZMOD_RW,
"Max I/O completion time (milliseconds) before marking it as slow");
ZFS_MODULE_PARAM(zfs_zio, zio_, requeue_io_start_cut_in_line, INT, ZMOD_RW,
"Prioritize requeued I/O");
ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_deferred_free, INT, ZMOD_RW,
"Defer frees starting in this pass");
ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_dont_compress, INT, ZMOD_RW,
"Don't compress starting in this pass");
ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_rewrite, INT, ZMOD_RW,
"Rewrite new bps starting in this pass");
ZFS_MODULE_PARAM(zfs_zio, zio_, dva_throttle_enabled, INT, ZMOD_RW,
"Throttle block allocations in the ZIO pipeline");
ZFS_MODULE_PARAM(zfs_zio, zio_, deadman_log_all, INT, ZMOD_RW,
"Log all slow ZIOs, not just those with vdevs");
/* END CSTYLED */
diff --git a/module/zfs/zvol.c b/module/zfs/zvol.c
index 31b3b91d6526..e7010e77a83a 100644
--- a/module/zfs/zvol.c
+++ b/module/zfs/zvol.c
@@ -1,1743 +1,1744 @@
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (C) 2008-2010 Lawrence Livermore National Security, LLC.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Rewritten for Linux by Brian Behlendorf <behlendorf1@llnl.gov>.
* LLNL-CODE-403049.
*
* ZFS volume emulation driver.
*
* Makes a DMU object look like a volume of arbitrary size, up to 2^64 bytes.
* Volumes are accessed through the symbolic links named:
*
* /dev/<pool_name>/<dataset_name>
*
* Volumes are persistent through reboot and module load. No user command
* needs to be run before opening and using a device.
*
* Copyright 2014 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
* Copyright (c) 2012, 2019 by Delphix. All rights reserved.
*/
/*
* Note on locking of zvol state structures.
*
* These structures are used to maintain internal state used to emulate block
* devices on top of zvols. In particular, management of device minor number
* operations - create, remove, rename, and set_snapdev - involves access to
* these structures. The zvol_state_lock is primarily used to protect the
* zvol_state_list. The zv->zv_state_lock is used to protect the contents
* of the zvol_state_t structures, as well as to make sure that when the
* time comes to remove the structure from the list, it is not in use, and
* therefore, it can be taken off zvol_state_list and freed.
*
* The zv_suspend_lock was introduced to allow for suspending I/O to a zvol,
* e.g. for the duration of receive and rollback operations. This lock can be
* held for significant periods of time. Given that it is undesirable to hold
* mutexes for long periods of time, the following lock ordering applies:
* - take zvol_state_lock if necessary, to protect zvol_state_list
* - take zv_suspend_lock if necessary, by the code path in question
* - take zv_state_lock to protect zvol_state_t
*
* The minor operations are issued to spa->spa_zvol_taskq queues, that are
* single-threaded (to preserve order of minor operations), and are executed
* through the zvol_task_cb that dispatches the specific operations. Therefore,
* these operations are serialized per pool. Consequently, we can be certain
* that for a given zvol, there is only one operation at a time in progress.
* That is why one can be sure that first, zvol_state_t for a given zvol is
* allocated and placed on zvol_state_list, and then other minor operations
* for this zvol are going to proceed in the order of issue.
*
*/
#include <sys/dataset_kstats.h>
#include <sys/dbuf.h>
#include <sys/dmu_traverse.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_dir.h>
#include <sys/zap.h>
#include <sys/zfeature.h>
#include <sys/zil_impl.h>
#include <sys/dmu_tx.h>
#include <sys/zio.h>
#include <sys/zfs_rlock.h>
#include <sys/spa_impl.h>
#include <sys/zvol.h>
#include <sys/zvol_impl.h>
unsigned int zvol_inhibit_dev = 0;
unsigned int zvol_volmode = ZFS_VOLMODE_GEOM;
struct hlist_head *zvol_htable;
-list_t zvol_state_list;
+static list_t zvol_state_list;
krwlock_t zvol_state_lock;
-const zvol_platform_ops_t *ops;
+static const zvol_platform_ops_t *ops;
typedef enum {
ZVOL_ASYNC_REMOVE_MINORS,
ZVOL_ASYNC_RENAME_MINORS,
ZVOL_ASYNC_SET_SNAPDEV,
ZVOL_ASYNC_SET_VOLMODE,
ZVOL_ASYNC_MAX
} zvol_async_op_t;
typedef struct {
zvol_async_op_t op;
char name1[MAXNAMELEN];
char name2[MAXNAMELEN];
uint64_t value;
} zvol_task_t;
uint64_t
zvol_name_hash(const char *name)
{
int i;
uint64_t crc = -1ULL;
const uint8_t *p = (const uint8_t *)name;
ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY);
for (i = 0; i < MAXNAMELEN - 1 && *p; i++, p++) {
crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (*p)) & 0xFF];
}
return (crc);
}
/*
* Find a zvol_state_t given the name and hash generated by zvol_name_hash.
* If found, return with zv_suspend_lock and zv_state_lock taken, otherwise,
* return (NULL) without the taking locks. The zv_suspend_lock is always taken
* before zv_state_lock. The mode argument indicates the mode (including none)
* for zv_suspend_lock to be taken.
*/
zvol_state_t *
zvol_find_by_name_hash(const char *name, uint64_t hash, int mode)
{
zvol_state_t *zv;
struct hlist_node *p = NULL;
rw_enter(&zvol_state_lock, RW_READER);
hlist_for_each(p, ZVOL_HT_HEAD(hash)) {
zv = hlist_entry(p, zvol_state_t, zv_hlink);
mutex_enter(&zv->zv_state_lock);
if (zv->zv_hash == hash &&
strncmp(zv->zv_name, name, MAXNAMELEN) == 0) {
/*
* this is the right zvol, take the locks in the
* right order
*/
if (mode != RW_NONE &&
!rw_tryenter(&zv->zv_suspend_lock, mode)) {
mutex_exit(&zv->zv_state_lock);
rw_enter(&zv->zv_suspend_lock, mode);
mutex_enter(&zv->zv_state_lock);
/*
* zvol cannot be renamed as we continue
* to hold zvol_state_lock
*/
ASSERT(zv->zv_hash == hash &&
strncmp(zv->zv_name, name, MAXNAMELEN)
== 0);
}
rw_exit(&zvol_state_lock);
return (zv);
}
mutex_exit(&zv->zv_state_lock);
}
rw_exit(&zvol_state_lock);
return (NULL);
}
/*
* Find a zvol_state_t given the name.
* If found, return with zv_suspend_lock and zv_state_lock taken, otherwise,
* return (NULL) without the taking locks. The zv_suspend_lock is always taken
* before zv_state_lock. The mode argument indicates the mode (including none)
* for zv_suspend_lock to be taken.
*/
static zvol_state_t *
zvol_find_by_name(const char *name, int mode)
{
return (zvol_find_by_name_hash(name, zvol_name_hash(name), mode));
}
/*
* ZFS_IOC_CREATE callback handles dmu zvol and zap object creation.
*/
void
zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
{
zfs_creat_t *zct = arg;
nvlist_t *nvprops = zct->zct_props;
int error;
uint64_t volblocksize, volsize;
VERIFY(nvlist_lookup_uint64(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0);
if (nvlist_lookup_uint64(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0)
volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE);
/*
* These properties must be removed from the list so the generic
* property setting step won't apply to them.
*/
VERIFY(nvlist_remove_all(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0);
(void) nvlist_remove_all(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE));
error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx);
ASSERT(error == 0);
}
/*
* ZFS_IOC_OBJSET_STATS entry point.
*/
int
zvol_get_stats(objset_t *os, nvlist_t *nv)
{
int error;
dmu_object_info_t *doi;
uint64_t val;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &val);
if (error)
return (SET_ERROR(error));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLSIZE, val);
doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);
error = dmu_object_info(os, ZVOL_OBJ, doi);
if (error == 0) {
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLBLOCKSIZE,
doi->doi_data_block_size);
}
kmem_free(doi, sizeof (dmu_object_info_t));
return (SET_ERROR(error));
}
/*
* Sanity check volume size.
*/
int
zvol_check_volsize(uint64_t volsize, uint64_t blocksize)
{
if (volsize == 0)
return (SET_ERROR(EINVAL));
if (volsize % blocksize != 0)
return (SET_ERROR(EINVAL));
#ifdef _ILP32
if (volsize - 1 > SPEC_MAXOFFSET_T)
return (SET_ERROR(EOVERFLOW));
#endif
return (0);
}
/*
* Ensure the zap is flushed then inform the VFS of the capacity change.
*/
static int
zvol_update_volsize(uint64_t volsize, objset_t *os)
{
dmu_tx_t *tx;
int error;
uint64_t txg;
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (SET_ERROR(error));
}
txg = dmu_tx_get_txg(tx);
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1,
&volsize, tx);
dmu_tx_commit(tx);
txg_wait_synced(dmu_objset_pool(os), txg);
if (error == 0)
error = dmu_free_long_range(os,
ZVOL_OBJ, volsize, DMU_OBJECT_END);
return (error);
}
/*
* Set ZFS_PROP_VOLSIZE set entry point. Note that modifying the volume
* size will result in a udev "change" event being generated.
*/
int
zvol_set_volsize(const char *name, uint64_t volsize)
{
objset_t *os = NULL;
uint64_t readonly;
int error;
boolean_t owned = B_FALSE;
error = dsl_prop_get_integer(name,
zfs_prop_to_name(ZFS_PROP_READONLY), &readonly, NULL);
if (error != 0)
return (SET_ERROR(error));
if (readonly)
return (SET_ERROR(EROFS));
zvol_state_t *zv = zvol_find_by_name(name, RW_READER);
ASSERT(zv == NULL || (MUTEX_HELD(&zv->zv_state_lock) &&
RW_READ_HELD(&zv->zv_suspend_lock)));
if (zv == NULL || zv->zv_objset == NULL) {
if (zv != NULL)
rw_exit(&zv->zv_suspend_lock);
if ((error = dmu_objset_own(name, DMU_OST_ZVOL, B_FALSE, B_TRUE,
FTAG, &os)) != 0) {
if (zv != NULL)
mutex_exit(&zv->zv_state_lock);
return (SET_ERROR(error));
}
owned = B_TRUE;
if (zv != NULL)
zv->zv_objset = os;
} else {
os = zv->zv_objset;
}
dmu_object_info_t *doi = kmem_alloc(sizeof (*doi), KM_SLEEP);
if ((error = dmu_object_info(os, ZVOL_OBJ, doi)) ||
(error = zvol_check_volsize(volsize, doi->doi_data_block_size)))
goto out;
error = zvol_update_volsize(volsize, os);
if (error == 0 && zv != NULL) {
zv->zv_volsize = volsize;
zv->zv_changed = 1;
}
out:
kmem_free(doi, sizeof (dmu_object_info_t));
if (owned) {
dmu_objset_disown(os, B_TRUE, FTAG);
if (zv != NULL)
zv->zv_objset = NULL;
} else {
rw_exit(&zv->zv_suspend_lock);
}
if (zv != NULL)
mutex_exit(&zv->zv_state_lock);
if (error == 0 && zv != NULL)
ops->zv_update_volsize(zv, volsize);
return (SET_ERROR(error));
}
/*
* Sanity check volume block size.
*/
int
zvol_check_volblocksize(const char *name, uint64_t volblocksize)
{
/* Record sizes above 128k need the feature to be enabled */
if (volblocksize > SPA_OLD_MAXBLOCKSIZE) {
spa_t *spa;
int error;
if ((error = spa_open(name, &spa, FTAG)) != 0)
return (error);
if (!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOTSUP));
}
/*
* We don't allow setting the property above 1MB,
* unless the tunable has been changed.
*/
if (volblocksize > zfs_max_recordsize)
return (SET_ERROR(EDOM));
spa_close(spa, FTAG);
}
if (volblocksize < SPA_MINBLOCKSIZE ||
volblocksize > SPA_MAXBLOCKSIZE ||
!ISP2(volblocksize))
return (SET_ERROR(EDOM));
return (0);
}
/*
* Replay a TX_TRUNCATE ZIL transaction if asked. TX_TRUNCATE is how we
* implement DKIOCFREE/free-long-range.
*/
static int
zvol_replay_truncate(void *arg1, void *arg2, boolean_t byteswap)
{
zvol_state_t *zv = arg1;
lr_truncate_t *lr = arg2;
uint64_t offset, length;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
offset = lr->lr_offset;
length = lr->lr_length;
dmu_tx_t *tx = dmu_tx_create(zv->zv_objset);
dmu_tx_mark_netfree(tx);
int error = dmu_tx_assign(tx, TXG_WAIT);
if (error != 0) {
dmu_tx_abort(tx);
} else {
zil_replaying(zv->zv_zilog, tx);
dmu_tx_commit(tx);
error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, offset,
length);
}
return (error);
}
/*
* Replay a TX_WRITE ZIL transaction that didn't get committed
* after a system failure
*/
static int
zvol_replay_write(void *arg1, void *arg2, boolean_t byteswap)
{
zvol_state_t *zv = arg1;
lr_write_t *lr = arg2;
objset_t *os = zv->zv_objset;
char *data = (char *)(lr + 1); /* data follows lr_write_t */
uint64_t offset, length;
dmu_tx_t *tx;
int error;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
offset = lr->lr_offset;
length = lr->lr_length;
/* If it's a dmu_sync() block, write the whole block */
if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
if (length < blocksize) {
offset -= offset % blocksize;
length = blocksize;
}
}
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, ZVOL_OBJ, offset, length);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
} else {
dmu_write(os, ZVOL_OBJ, offset, length, data, tx);
zil_replaying(zv->zv_zilog, tx);
dmu_tx_commit(tx);
}
return (error);
}
static int
zvol_replay_err(void *arg1, void *arg2, boolean_t byteswap)
{
+ (void) arg1, (void) arg2, (void) byteswap;
return (SET_ERROR(ENOTSUP));
}
/*
* Callback vectors for replaying records.
* Only TX_WRITE and TX_TRUNCATE are needed for zvol.
*/
-zil_replay_func_t *zvol_replay_vector[TX_MAX_TYPE] = {
+zil_replay_func_t *const zvol_replay_vector[TX_MAX_TYPE] = {
zvol_replay_err, /* no such transaction type */
zvol_replay_err, /* TX_CREATE */
zvol_replay_err, /* TX_MKDIR */
zvol_replay_err, /* TX_MKXATTR */
zvol_replay_err, /* TX_SYMLINK */
zvol_replay_err, /* TX_REMOVE */
zvol_replay_err, /* TX_RMDIR */
zvol_replay_err, /* TX_LINK */
zvol_replay_err, /* TX_RENAME */
zvol_replay_write, /* TX_WRITE */
zvol_replay_truncate, /* TX_TRUNCATE */
zvol_replay_err, /* TX_SETATTR */
zvol_replay_err, /* TX_ACL */
zvol_replay_err, /* TX_CREATE_ATTR */
zvol_replay_err, /* TX_CREATE_ACL_ATTR */
zvol_replay_err, /* TX_MKDIR_ACL */
zvol_replay_err, /* TX_MKDIR_ATTR */
zvol_replay_err, /* TX_MKDIR_ACL_ATTR */
zvol_replay_err, /* TX_WRITE2 */
};
/*
* zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions.
*
* We store data in the log buffers if it's small enough.
* Otherwise we will later flush the data out via dmu_sync().
*/
-ssize_t zvol_immediate_write_sz = 32768;
+static const ssize_t zvol_immediate_write_sz = 32768;
void
zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, uint64_t offset,
uint64_t size, int sync)
{
uint32_t blocksize = zv->zv_volblocksize;
zilog_t *zilog = zv->zv_zilog;
itx_wr_state_t write_state;
uint64_t sz = size;
if (zil_replaying(zilog, tx))
return;
if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
write_state = WR_INDIRECT;
else if (!spa_has_slogs(zilog->zl_spa) &&
size >= blocksize && blocksize > zvol_immediate_write_sz)
write_state = WR_INDIRECT;
else if (sync)
write_state = WR_COPIED;
else
write_state = WR_NEED_COPY;
while (size) {
itx_t *itx;
lr_write_t *lr;
itx_wr_state_t wr_state = write_state;
ssize_t len = size;
if (wr_state == WR_COPIED && size > zil_max_copied_data(zilog))
wr_state = WR_NEED_COPY;
else if (wr_state == WR_INDIRECT)
len = MIN(blocksize - P2PHASE(offset, blocksize), size);
itx = zil_itx_create(TX_WRITE, sizeof (*lr) +
(wr_state == WR_COPIED ? len : 0));
lr = (lr_write_t *)&itx->itx_lr;
if (wr_state == WR_COPIED && dmu_read_by_dnode(zv->zv_dn,
offset, len, lr+1, DMU_READ_NO_PREFETCH) != 0) {
zil_itx_destroy(itx);
itx = zil_itx_create(TX_WRITE, sizeof (*lr));
lr = (lr_write_t *)&itx->itx_lr;
wr_state = WR_NEED_COPY;
}
itx->itx_wr_state = wr_state;
lr->lr_foid = ZVOL_OBJ;
lr->lr_offset = offset;
lr->lr_length = len;
lr->lr_blkoff = 0;
BP_ZERO(&lr->lr_blkptr);
itx->itx_private = zv;
itx->itx_sync = sync;
(void) zil_itx_assign(zilog, itx, tx);
offset += len;
size -= len;
}
if (write_state == WR_COPIED || write_state == WR_NEED_COPY) {
dsl_pool_wrlog_count(zilog->zl_dmu_pool, sz, tx->tx_txg);
}
}
/*
* Log a DKIOCFREE/free-long-range to the ZIL with TX_TRUNCATE.
*/
void
zvol_log_truncate(zvol_state_t *zv, dmu_tx_t *tx, uint64_t off, uint64_t len,
boolean_t sync)
{
itx_t *itx;
lr_truncate_t *lr;
zilog_t *zilog = zv->zv_zilog;
if (zil_replaying(zilog, tx))
return;
itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr));
lr = (lr_truncate_t *)&itx->itx_lr;
lr->lr_foid = ZVOL_OBJ;
lr->lr_offset = off;
lr->lr_length = len;
itx->itx_sync = sync;
zil_itx_assign(zilog, itx, tx);
}
/* ARGSUSED */
static void
zvol_get_done(zgd_t *zgd, int error)
{
if (zgd->zgd_db)
dmu_buf_rele(zgd->zgd_db, zgd);
zfs_rangelock_exit(zgd->zgd_lr);
kmem_free(zgd, sizeof (zgd_t));
}
/*
* Get data to generate a TX_WRITE intent log record.
*/
int
zvol_get_data(void *arg, uint64_t arg2, lr_write_t *lr, char *buf,
struct lwb *lwb, zio_t *zio)
{
zvol_state_t *zv = arg;
uint64_t offset = lr->lr_offset;
uint64_t size = lr->lr_length;
dmu_buf_t *db;
zgd_t *zgd;
int error;
ASSERT3P(lwb, !=, NULL);
ASSERT3P(zio, !=, NULL);
ASSERT3U(size, !=, 0);
zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP);
zgd->zgd_lwb = lwb;
/*
* Write records come in two flavors: immediate and indirect.
* For small writes it's cheaper to store the data with the
* log record (immediate); for large writes it's cheaper to
* sync the data and get a pointer to it (indirect) so that
* we don't have to write the data twice.
*/
if (buf != NULL) { /* immediate write */
zgd->zgd_lr = zfs_rangelock_enter(&zv->zv_rangelock, offset,
size, RL_READER);
error = dmu_read_by_dnode(zv->zv_dn, offset, size, buf,
DMU_READ_NO_PREFETCH);
} else { /* indirect write */
/*
* Have to lock the whole block to ensure when it's written out
* and its checksum is being calculated that no one can change
* the data. Contrarily to zfs_get_data we need not re-check
* blocksize after we get the lock because it cannot be changed.
*/
size = zv->zv_volblocksize;
offset = P2ALIGN_TYPED(offset, size, uint64_t);
zgd->zgd_lr = zfs_rangelock_enter(&zv->zv_rangelock, offset,
size, RL_READER);
error = dmu_buf_hold_by_dnode(zv->zv_dn, offset, zgd, &db,
DMU_READ_NO_PREFETCH);
if (error == 0) {
blkptr_t *bp = &lr->lr_blkptr;
zgd->zgd_db = db;
zgd->zgd_bp = bp;
ASSERT(db != NULL);
ASSERT(db->db_offset == offset);
ASSERT(db->db_size == size);
error = dmu_sync(zio, lr->lr_common.lrc_txg,
zvol_get_done, zgd);
if (error == 0)
return (0);
}
}
zvol_get_done(zgd, error);
return (SET_ERROR(error));
}
/*
* The zvol_state_t's are inserted into zvol_state_list and zvol_htable.
*/
void
zvol_insert(zvol_state_t *zv)
{
ASSERT(RW_WRITE_HELD(&zvol_state_lock));
list_insert_head(&zvol_state_list, zv);
hlist_add_head(&zv->zv_hlink, ZVOL_HT_HEAD(zv->zv_hash));
}
/*
* Simply remove the zvol from to list of zvols.
*/
static void
zvol_remove(zvol_state_t *zv)
{
ASSERT(RW_WRITE_HELD(&zvol_state_lock));
list_remove(&zvol_state_list, zv);
hlist_del(&zv->zv_hlink);
}
/*
* Setup zv after we just own the zv->objset
*/
static int
zvol_setup_zv(zvol_state_t *zv)
{
uint64_t volsize;
int error;
uint64_t ro;
objset_t *os = zv->zv_objset;
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
ASSERT(RW_LOCK_HELD(&zv->zv_suspend_lock));
zv->zv_zilog = NULL;
zv->zv_flags &= ~ZVOL_WRITTEN_TO;
error = dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL);
if (error)
return (SET_ERROR(error));
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
return (SET_ERROR(error));
error = dnode_hold(os, ZVOL_OBJ, zv, &zv->zv_dn);
if (error)
return (SET_ERROR(error));
ops->zv_set_capacity(zv, volsize >> 9);
zv->zv_volsize = volsize;
if (ro || dmu_objset_is_snapshot(os) ||
!spa_writeable(dmu_objset_spa(os))) {
ops->zv_set_disk_ro(zv, 1);
zv->zv_flags |= ZVOL_RDONLY;
} else {
ops->zv_set_disk_ro(zv, 0);
zv->zv_flags &= ~ZVOL_RDONLY;
}
return (0);
}
/*
* Shutdown every zv_objset related stuff except zv_objset itself.
* The is the reverse of zvol_setup_zv.
*/
static void
zvol_shutdown_zv(zvol_state_t *zv)
{
ASSERT(MUTEX_HELD(&zv->zv_state_lock) &&
RW_LOCK_HELD(&zv->zv_suspend_lock));
if (zv->zv_flags & ZVOL_WRITTEN_TO) {
ASSERT(zv->zv_zilog != NULL);
zil_close(zv->zv_zilog);
}
zv->zv_zilog = NULL;
dnode_rele(zv->zv_dn, zv);
zv->zv_dn = NULL;
/*
* Evict cached data. We must write out any dirty data before
* disowning the dataset.
*/
if (zv->zv_flags & ZVOL_WRITTEN_TO)
txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);
(void) dmu_objset_evict_dbufs(zv->zv_objset);
}
/*
* return the proper tag for rollback and recv
*/
void *
zvol_tag(zvol_state_t *zv)
{
ASSERT(RW_WRITE_HELD(&zv->zv_suspend_lock));
return (zv->zv_open_count > 0 ? zv : NULL);
}
/*
* Suspend the zvol for recv and rollback.
*/
zvol_state_t *
zvol_suspend(const char *name)
{
zvol_state_t *zv;
zv = zvol_find_by_name(name, RW_WRITER);
if (zv == NULL)
return (NULL);
/* block all I/O, release in zvol_resume. */
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
ASSERT(RW_WRITE_HELD(&zv->zv_suspend_lock));
atomic_inc(&zv->zv_suspend_ref);
if (zv->zv_open_count > 0)
zvol_shutdown_zv(zv);
/*
* do not hold zv_state_lock across suspend/resume to
* avoid locking up zvol lookups
*/
mutex_exit(&zv->zv_state_lock);
/* zv_suspend_lock is released in zvol_resume() */
return (zv);
}
int
zvol_resume(zvol_state_t *zv)
{
int error = 0;
ASSERT(RW_WRITE_HELD(&zv->zv_suspend_lock));
mutex_enter(&zv->zv_state_lock);
if (zv->zv_open_count > 0) {
VERIFY0(dmu_objset_hold(zv->zv_name, zv, &zv->zv_objset));
VERIFY3P(zv->zv_objset->os_dsl_dataset->ds_owner, ==, zv);
VERIFY(dsl_dataset_long_held(zv->zv_objset->os_dsl_dataset));
dmu_objset_rele(zv->zv_objset, zv);
error = zvol_setup_zv(zv);
}
mutex_exit(&zv->zv_state_lock);
rw_exit(&zv->zv_suspend_lock);
/*
* We need this because we don't hold zvol_state_lock while releasing
* zv_suspend_lock. zvol_remove_minors_impl thus cannot check
* zv_suspend_lock to determine it is safe to free because rwlock is
* not inherent atomic.
*/
atomic_dec(&zv->zv_suspend_ref);
return (SET_ERROR(error));
}
int
zvol_first_open(zvol_state_t *zv, boolean_t readonly)
{
objset_t *os;
int error;
ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
ASSERT(mutex_owned(&spa_namespace_lock));
boolean_t ro = (readonly || (strchr(zv->zv_name, '@') != NULL));
error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, ro, B_TRUE, zv, &os);
if (error)
return (SET_ERROR(error));
zv->zv_objset = os;
error = zvol_setup_zv(zv);
if (error) {
dmu_objset_disown(os, 1, zv);
zv->zv_objset = NULL;
}
return (error);
}
void
zvol_last_close(zvol_state_t *zv)
{
ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
zvol_shutdown_zv(zv);
dmu_objset_disown(zv->zv_objset, 1, zv);
zv->zv_objset = NULL;
}
typedef struct minors_job {
list_t *list;
list_node_t link;
/* input */
char *name;
/* output */
int error;
} minors_job_t;
/*
* Prefetch zvol dnodes for the minors_job
*/
static void
zvol_prefetch_minors_impl(void *arg)
{
minors_job_t *job = arg;
char *dsname = job->name;
objset_t *os = NULL;
job->error = dmu_objset_own(dsname, DMU_OST_ZVOL, B_TRUE, B_TRUE,
FTAG, &os);
if (job->error == 0) {
dmu_prefetch(os, ZVOL_OBJ, 0, 0, 0, ZIO_PRIORITY_SYNC_READ);
dmu_objset_disown(os, B_TRUE, FTAG);
}
}
/*
* Mask errors to continue dmu_objset_find() traversal
*/
static int
zvol_create_snap_minor_cb(const char *dsname, void *arg)
{
minors_job_t *j = arg;
list_t *minors_list = j->list;
const char *name = j->name;
ASSERT0(MUTEX_HELD(&spa_namespace_lock));
/* skip the designated dataset */
if (name && strcmp(dsname, name) == 0)
return (0);
/* at this point, the dsname should name a snapshot */
if (strchr(dsname, '@') == 0) {
dprintf("zvol_create_snap_minor_cb(): "
"%s is not a snapshot name\n", dsname);
} else {
minors_job_t *job;
char *n = kmem_strdup(dsname);
if (n == NULL)
return (0);
job = kmem_alloc(sizeof (minors_job_t), KM_SLEEP);
job->name = n;
job->list = minors_list;
job->error = 0;
list_insert_tail(minors_list, job);
/* don't care if dispatch fails, because job->error is 0 */
taskq_dispatch(system_taskq, zvol_prefetch_minors_impl, job,
TQ_SLEEP);
}
return (0);
}
/*
* If spa_keystore_load_wkey() is called for an encrypted zvol,
* we need to look for any clones also using the key. This function
* is "best effort" - so we just skip over it if there are failures.
*/
static void
zvol_add_clones(const char *dsname, list_t *minors_list)
{
/* Also check if it has clones */
dsl_dir_t *dd = NULL;
dsl_pool_t *dp = NULL;
if (dsl_pool_hold(dsname, FTAG, &dp) != 0)
return;
if (!spa_feature_is_enabled(dp->dp_spa,
SPA_FEATURE_ENCRYPTION))
goto out;
if (dsl_dir_hold(dp, dsname, FTAG, &dd, NULL) != 0)
goto out;
if (dsl_dir_phys(dd)->dd_clones == 0)
goto out;
zap_cursor_t *zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP);
zap_attribute_t *za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
objset_t *mos = dd->dd_pool->dp_meta_objset;
for (zap_cursor_init(zc, mos, dsl_dir_phys(dd)->dd_clones);
zap_cursor_retrieve(zc, za) == 0;
zap_cursor_advance(zc)) {
dsl_dataset_t *clone;
minors_job_t *job;
if (dsl_dataset_hold_obj(dd->dd_pool,
za->za_first_integer, FTAG, &clone) == 0) {
char name[ZFS_MAX_DATASET_NAME_LEN];
dsl_dataset_name(clone, name);
char *n = kmem_strdup(name);
job = kmem_alloc(sizeof (minors_job_t), KM_SLEEP);
job->name = n;
job->list = minors_list;
job->error = 0;
list_insert_tail(minors_list, job);
dsl_dataset_rele(clone, FTAG);
}
}
zap_cursor_fini(zc);
kmem_free(za, sizeof (zap_attribute_t));
kmem_free(zc, sizeof (zap_cursor_t));
out:
if (dd != NULL)
dsl_dir_rele(dd, FTAG);
if (dp != NULL)
dsl_pool_rele(dp, FTAG);
}
/*
* Mask errors to continue dmu_objset_find() traversal
*/
static int
zvol_create_minors_cb(const char *dsname, void *arg)
{
uint64_t snapdev;
int error;
list_t *minors_list = arg;
ASSERT0(MUTEX_HELD(&spa_namespace_lock));
error = dsl_prop_get_integer(dsname, "snapdev", &snapdev, NULL);
if (error)
return (0);
/*
* Given the name and the 'snapdev' property, create device minor nodes
* with the linkages to zvols/snapshots as needed.
* If the name represents a zvol, create a minor node for the zvol, then
* check if its snapshots are 'visible', and if so, iterate over the
* snapshots and create device minor nodes for those.
*/
if (strchr(dsname, '@') == 0) {
minors_job_t *job;
char *n = kmem_strdup(dsname);
if (n == NULL)
return (0);
job = kmem_alloc(sizeof (minors_job_t), KM_SLEEP);
job->name = n;
job->list = minors_list;
job->error = 0;
list_insert_tail(minors_list, job);
/* don't care if dispatch fails, because job->error is 0 */
taskq_dispatch(system_taskq, zvol_prefetch_minors_impl, job,
TQ_SLEEP);
zvol_add_clones(dsname, minors_list);
if (snapdev == ZFS_SNAPDEV_VISIBLE) {
/*
* traverse snapshots only, do not traverse children,
* and skip the 'dsname'
*/
error = dmu_objset_find(dsname,
zvol_create_snap_minor_cb, (void *)job,
DS_FIND_SNAPSHOTS);
}
} else {
dprintf("zvol_create_minors_cb(): %s is not a zvol name\n",
dsname);
}
return (0);
}
/*
* Create minors for the specified dataset, including children and snapshots.
* Pay attention to the 'snapdev' property and iterate over the snapshots
* only if they are 'visible'. This approach allows one to assure that the
* snapshot metadata is read from disk only if it is needed.
*
* The name can represent a dataset to be recursively scanned for zvols and
* their snapshots, or a single zvol snapshot. If the name represents a
* dataset, the scan is performed in two nested stages:
* - scan the dataset for zvols, and
* - for each zvol, create a minor node, then check if the zvol's snapshots
* are 'visible', and only then iterate over the snapshots if needed
*
* If the name represents a snapshot, a check is performed if the snapshot is
* 'visible' (which also verifies that the parent is a zvol), and if so,
* a minor node for that snapshot is created.
*/
void
zvol_create_minors_recursive(const char *name)
{
list_t minors_list;
minors_job_t *job;
if (zvol_inhibit_dev)
return;
/*
* This is the list for prefetch jobs. Whenever we found a match
* during dmu_objset_find, we insert a minors_job to the list and do
* taskq_dispatch to parallel prefetch zvol dnodes. Note we don't need
* any lock because all list operation is done on the current thread.
*
* We will use this list to do zvol_create_minor_impl after prefetch
* so we don't have to traverse using dmu_objset_find again.
*/
list_create(&minors_list, sizeof (minors_job_t),
offsetof(minors_job_t, link));
if (strchr(name, '@') != NULL) {
uint64_t snapdev;
int error = dsl_prop_get_integer(name, "snapdev",
&snapdev, NULL);
if (error == 0 && snapdev == ZFS_SNAPDEV_VISIBLE)
(void) ops->zv_create_minor(name);
} else {
fstrans_cookie_t cookie = spl_fstrans_mark();
(void) dmu_objset_find(name, zvol_create_minors_cb,
&minors_list, DS_FIND_CHILDREN);
spl_fstrans_unmark(cookie);
}
taskq_wait_outstanding(system_taskq, 0);
/*
* Prefetch is completed, we can do zvol_create_minor_impl
* sequentially.
*/
while ((job = list_head(&minors_list)) != NULL) {
list_remove(&minors_list, job);
if (!job->error)
(void) ops->zv_create_minor(job->name);
kmem_strfree(job->name);
kmem_free(job, sizeof (minors_job_t));
}
list_destroy(&minors_list);
}
void
zvol_create_minor(const char *name)
{
/*
* Note: the dsl_pool_config_lock must not be held.
* Minor node creation needs to obtain the zvol_state_lock.
* zvol_open() obtains the zvol_state_lock and then the dsl pool
* config lock. Therefore, we can't have the config lock now if
* we are going to wait for the zvol_state_lock, because it
* would be a lock order inversion which could lead to deadlock.
*/
if (zvol_inhibit_dev)
return;
if (strchr(name, '@') != NULL) {
uint64_t snapdev;
int error = dsl_prop_get_integer(name,
"snapdev", &snapdev, NULL);
if (error == 0 && snapdev == ZFS_SNAPDEV_VISIBLE)
(void) ops->zv_create_minor(name);
} else {
(void) ops->zv_create_minor(name);
}
}
/*
* Remove minors for specified dataset including children and snapshots.
*/
static void
zvol_free_task(void *arg)
{
ops->zv_free(arg);
}
void
zvol_remove_minors_impl(const char *name)
{
zvol_state_t *zv, *zv_next;
int namelen = ((name) ? strlen(name) : 0);
taskqid_t t;
list_t free_list;
if (zvol_inhibit_dev)
return;
list_create(&free_list, sizeof (zvol_state_t),
offsetof(zvol_state_t, zv_next));
rw_enter(&zvol_state_lock, RW_WRITER);
for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
zv_next = list_next(&zvol_state_list, zv);
mutex_enter(&zv->zv_state_lock);
if (name == NULL || strcmp(zv->zv_name, name) == 0 ||
(strncmp(zv->zv_name, name, namelen) == 0 &&
(zv->zv_name[namelen] == '/' ||
zv->zv_name[namelen] == '@'))) {
/*
* By holding zv_state_lock here, we guarantee that no
* one is currently using this zv
*/
/* If in use, leave alone */
if (zv->zv_open_count > 0 ||
atomic_read(&zv->zv_suspend_ref)) {
mutex_exit(&zv->zv_state_lock);
continue;
}
zvol_remove(zv);
/*
* Cleared while holding zvol_state_lock as a writer
* which will prevent zvol_open() from opening it.
*/
ops->zv_clear_private(zv);
/* Drop zv_state_lock before zvol_free() */
mutex_exit(&zv->zv_state_lock);
/* Try parallel zv_free, if failed do it in place */
t = taskq_dispatch(system_taskq, zvol_free_task, zv,
TQ_SLEEP);
if (t == TASKQID_INVALID)
list_insert_head(&free_list, zv);
} else {
mutex_exit(&zv->zv_state_lock);
}
}
rw_exit(&zvol_state_lock);
/* Drop zvol_state_lock before calling zvol_free() */
while ((zv = list_head(&free_list)) != NULL) {
list_remove(&free_list, zv);
ops->zv_free(zv);
}
}
/* Remove minor for this specific volume only */
static void
zvol_remove_minor_impl(const char *name)
{
zvol_state_t *zv = NULL, *zv_next;
if (zvol_inhibit_dev)
return;
rw_enter(&zvol_state_lock, RW_WRITER);
for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
zv_next = list_next(&zvol_state_list, zv);
mutex_enter(&zv->zv_state_lock);
if (strcmp(zv->zv_name, name) == 0) {
/*
* By holding zv_state_lock here, we guarantee that no
* one is currently using this zv
*/
/* If in use, leave alone */
if (zv->zv_open_count > 0 ||
atomic_read(&zv->zv_suspend_ref)) {
mutex_exit(&zv->zv_state_lock);
continue;
}
zvol_remove(zv);
ops->zv_clear_private(zv);
mutex_exit(&zv->zv_state_lock);
break;
} else {
mutex_exit(&zv->zv_state_lock);
}
}
/* Drop zvol_state_lock before calling zvol_free() */
rw_exit(&zvol_state_lock);
if (zv != NULL)
ops->zv_free(zv);
}
/*
* Rename minors for specified dataset including children and snapshots.
*/
static void
zvol_rename_minors_impl(const char *oldname, const char *newname)
{
zvol_state_t *zv, *zv_next;
int oldnamelen;
if (zvol_inhibit_dev)
return;
oldnamelen = strlen(oldname);
rw_enter(&zvol_state_lock, RW_READER);
for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
zv_next = list_next(&zvol_state_list, zv);
mutex_enter(&zv->zv_state_lock);
if (strcmp(zv->zv_name, oldname) == 0) {
ops->zv_rename_minor(zv, newname);
} else if (strncmp(zv->zv_name, oldname, oldnamelen) == 0 &&
(zv->zv_name[oldnamelen] == '/' ||
zv->zv_name[oldnamelen] == '@')) {
char *name = kmem_asprintf("%s%c%s", newname,
zv->zv_name[oldnamelen],
zv->zv_name + oldnamelen + 1);
ops->zv_rename_minor(zv, name);
kmem_strfree(name);
}
mutex_exit(&zv->zv_state_lock);
}
rw_exit(&zvol_state_lock);
}
typedef struct zvol_snapdev_cb_arg {
uint64_t snapdev;
} zvol_snapdev_cb_arg_t;
static int
zvol_set_snapdev_cb(const char *dsname, void *param)
{
zvol_snapdev_cb_arg_t *arg = param;
if (strchr(dsname, '@') == NULL)
return (0);
switch (arg->snapdev) {
case ZFS_SNAPDEV_VISIBLE:
(void) ops->zv_create_minor(dsname);
break;
case ZFS_SNAPDEV_HIDDEN:
(void) zvol_remove_minor_impl(dsname);
break;
}
return (0);
}
static void
zvol_set_snapdev_impl(char *name, uint64_t snapdev)
{
zvol_snapdev_cb_arg_t arg = {snapdev};
fstrans_cookie_t cookie = spl_fstrans_mark();
/*
* The zvol_set_snapdev_sync() sets snapdev appropriately
* in the dataset hierarchy. Here, we only scan snapshots.
*/
dmu_objset_find(name, zvol_set_snapdev_cb, &arg, DS_FIND_SNAPSHOTS);
spl_fstrans_unmark(cookie);
}
static void
zvol_set_volmode_impl(char *name, uint64_t volmode)
{
fstrans_cookie_t cookie;
uint64_t old_volmode;
zvol_state_t *zv;
if (strchr(name, '@') != NULL)
return;
/*
* It's unfortunate we need to remove minors before we create new ones:
* this is necessary because our backing gendisk (zvol_state->zv_disk)
* could be different when we set, for instance, volmode from "geom"
* to "dev" (or vice versa).
*/
zv = zvol_find_by_name(name, RW_NONE);
if (zv == NULL && volmode == ZFS_VOLMODE_NONE)
return;
if (zv != NULL) {
old_volmode = zv->zv_volmode;
mutex_exit(&zv->zv_state_lock);
if (old_volmode == volmode)
return;
zvol_wait_close(zv);
}
cookie = spl_fstrans_mark();
switch (volmode) {
case ZFS_VOLMODE_NONE:
(void) zvol_remove_minor_impl(name);
break;
case ZFS_VOLMODE_GEOM:
case ZFS_VOLMODE_DEV:
(void) zvol_remove_minor_impl(name);
(void) ops->zv_create_minor(name);
break;
case ZFS_VOLMODE_DEFAULT:
(void) zvol_remove_minor_impl(name);
if (zvol_volmode == ZFS_VOLMODE_NONE)
break;
else /* if zvol_volmode is invalid defaults to "geom" */
(void) ops->zv_create_minor(name);
break;
}
spl_fstrans_unmark(cookie);
}
static zvol_task_t *
zvol_task_alloc(zvol_async_op_t op, const char *name1, const char *name2,
uint64_t value)
{
zvol_task_t *task;
/* Never allow tasks on hidden names. */
if (name1[0] == '$')
return (NULL);
task = kmem_zalloc(sizeof (zvol_task_t), KM_SLEEP);
task->op = op;
task->value = value;
strlcpy(task->name1, name1, MAXNAMELEN);
if (name2 != NULL)
strlcpy(task->name2, name2, MAXNAMELEN);
return (task);
}
static void
zvol_task_free(zvol_task_t *task)
{
kmem_free(task, sizeof (zvol_task_t));
}
/*
* The worker thread function performed asynchronously.
*/
static void
zvol_task_cb(void *arg)
{
zvol_task_t *task = arg;
switch (task->op) {
case ZVOL_ASYNC_REMOVE_MINORS:
zvol_remove_minors_impl(task->name1);
break;
case ZVOL_ASYNC_RENAME_MINORS:
zvol_rename_minors_impl(task->name1, task->name2);
break;
case ZVOL_ASYNC_SET_SNAPDEV:
zvol_set_snapdev_impl(task->name1, task->value);
break;
case ZVOL_ASYNC_SET_VOLMODE:
zvol_set_volmode_impl(task->name1, task->value);
break;
default:
VERIFY(0);
break;
}
zvol_task_free(task);
}
typedef struct zvol_set_prop_int_arg {
const char *zsda_name;
uint64_t zsda_value;
zprop_source_t zsda_source;
dmu_tx_t *zsda_tx;
} zvol_set_prop_int_arg_t;
/*
* Sanity check the dataset for safe use by the sync task. No additional
* conditions are imposed.
*/
static int
zvol_set_snapdev_check(void *arg, dmu_tx_t *tx)
{
zvol_set_prop_int_arg_t *zsda = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *dd;
int error;
error = dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL);
if (error != 0)
return (error);
dsl_dir_rele(dd, FTAG);
return (error);
}
/* ARGSUSED */
static int
zvol_set_snapdev_sync_cb(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg)
{
char dsname[MAXNAMELEN];
zvol_task_t *task;
uint64_t snapdev;
dsl_dataset_name(ds, dsname);
if (dsl_prop_get_int_ds(ds, "snapdev", &snapdev) != 0)
return (0);
task = zvol_task_alloc(ZVOL_ASYNC_SET_SNAPDEV, dsname, NULL, snapdev);
if (task == NULL)
return (0);
(void) taskq_dispatch(dp->dp_spa->spa_zvol_taskq, zvol_task_cb,
task, TQ_SLEEP);
return (0);
}
/*
* Traverse all child datasets and apply snapdev appropriately.
* We call dsl_prop_set_sync_impl() here to set the value only on the toplevel
* dataset and read the effective "snapdev" on every child in the callback
* function: this is because the value is not guaranteed to be the same in the
* whole dataset hierarchy.
*/
static void
zvol_set_snapdev_sync(void *arg, dmu_tx_t *tx)
{
zvol_set_prop_int_arg_t *zsda = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *dd;
dsl_dataset_t *ds;
int error;
VERIFY0(dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL));
zsda->zsda_tx = tx;
error = dsl_dataset_hold(dp, zsda->zsda_name, FTAG, &ds);
if (error == 0) {
dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_SNAPDEV),
zsda->zsda_source, sizeof (zsda->zsda_value), 1,
&zsda->zsda_value, zsda->zsda_tx);
dsl_dataset_rele(ds, FTAG);
}
dmu_objset_find_dp(dp, dd->dd_object, zvol_set_snapdev_sync_cb,
zsda, DS_FIND_CHILDREN);
dsl_dir_rele(dd, FTAG);
}
int
zvol_set_snapdev(const char *ddname, zprop_source_t source, uint64_t snapdev)
{
zvol_set_prop_int_arg_t zsda;
zsda.zsda_name = ddname;
zsda.zsda_source = source;
zsda.zsda_value = snapdev;
return (dsl_sync_task(ddname, zvol_set_snapdev_check,
zvol_set_snapdev_sync, &zsda, 0, ZFS_SPACE_CHECK_NONE));
}
/*
* Sanity check the dataset for safe use by the sync task. No additional
* conditions are imposed.
*/
static int
zvol_set_volmode_check(void *arg, dmu_tx_t *tx)
{
zvol_set_prop_int_arg_t *zsda = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *dd;
int error;
error = dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL);
if (error != 0)
return (error);
dsl_dir_rele(dd, FTAG);
return (error);
}
/* ARGSUSED */
static int
zvol_set_volmode_sync_cb(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg)
{
char dsname[MAXNAMELEN];
zvol_task_t *task;
uint64_t volmode;
dsl_dataset_name(ds, dsname);
if (dsl_prop_get_int_ds(ds, "volmode", &volmode) != 0)
return (0);
task = zvol_task_alloc(ZVOL_ASYNC_SET_VOLMODE, dsname, NULL, volmode);
if (task == NULL)
return (0);
(void) taskq_dispatch(dp->dp_spa->spa_zvol_taskq, zvol_task_cb,
task, TQ_SLEEP);
return (0);
}
/*
* Traverse all child datasets and apply volmode appropriately.
* We call dsl_prop_set_sync_impl() here to set the value only on the toplevel
* dataset and read the effective "volmode" on every child in the callback
* function: this is because the value is not guaranteed to be the same in the
* whole dataset hierarchy.
*/
static void
zvol_set_volmode_sync(void *arg, dmu_tx_t *tx)
{
zvol_set_prop_int_arg_t *zsda = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *dd;
dsl_dataset_t *ds;
int error;
VERIFY0(dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL));
zsda->zsda_tx = tx;
error = dsl_dataset_hold(dp, zsda->zsda_name, FTAG, &ds);
if (error == 0) {
dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_VOLMODE),
zsda->zsda_source, sizeof (zsda->zsda_value), 1,
&zsda->zsda_value, zsda->zsda_tx);
dsl_dataset_rele(ds, FTAG);
}
dmu_objset_find_dp(dp, dd->dd_object, zvol_set_volmode_sync_cb,
zsda, DS_FIND_CHILDREN);
dsl_dir_rele(dd, FTAG);
}
int
zvol_set_volmode(const char *ddname, zprop_source_t source, uint64_t volmode)
{
zvol_set_prop_int_arg_t zsda;
zsda.zsda_name = ddname;
zsda.zsda_source = source;
zsda.zsda_value = volmode;
return (dsl_sync_task(ddname, zvol_set_volmode_check,
zvol_set_volmode_sync, &zsda, 0, ZFS_SPACE_CHECK_NONE));
}
void
zvol_remove_minors(spa_t *spa, const char *name, boolean_t async)
{
zvol_task_t *task;
taskqid_t id;
task = zvol_task_alloc(ZVOL_ASYNC_REMOVE_MINORS, name, NULL, ~0ULL);
if (task == NULL)
return;
id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP);
if ((async == B_FALSE) && (id != TASKQID_INVALID))
taskq_wait_id(spa->spa_zvol_taskq, id);
}
void
zvol_rename_minors(spa_t *spa, const char *name1, const char *name2,
boolean_t async)
{
zvol_task_t *task;
taskqid_t id;
task = zvol_task_alloc(ZVOL_ASYNC_RENAME_MINORS, name1, name2, ~0ULL);
if (task == NULL)
return;
id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP);
if ((async == B_FALSE) && (id != TASKQID_INVALID))
taskq_wait_id(spa->spa_zvol_taskq, id);
}
boolean_t
zvol_is_zvol(const char *name)
{
return (ops->zv_is_zvol(name));
}
void
zvol_register_ops(const zvol_platform_ops_t *zvol_ops)
{
ops = zvol_ops;
}
int
zvol_init_impl(void)
{
int i;
list_create(&zvol_state_list, sizeof (zvol_state_t),
offsetof(zvol_state_t, zv_next));
rw_init(&zvol_state_lock, NULL, RW_DEFAULT, NULL);
zvol_htable = kmem_alloc(ZVOL_HT_SIZE * sizeof (struct hlist_head),
KM_SLEEP);
for (i = 0; i < ZVOL_HT_SIZE; i++)
INIT_HLIST_HEAD(&zvol_htable[i]);
return (0);
}
void
zvol_fini_impl(void)
{
zvol_remove_minors_impl(NULL);
/*
* The call to "zvol_remove_minors_impl" may dispatch entries to
* the system_taskq, but it doesn't wait for those entries to
* complete before it returns. Thus, we must wait for all of the
* removals to finish, before we can continue.
*/
taskq_wait_outstanding(system_taskq, 0);
kmem_free(zvol_htable, ZVOL_HT_SIZE * sizeof (struct hlist_head));
list_destroy(&zvol_state_list);
rw_destroy(&zvol_state_lock);
}

File Metadata

Mime Type
application/octet-stream
Expires
Thu, May 23, 11:00 AM (1 d, 23 h)
Storage Engine
chunks
Storage Format
Chunks
Storage Handle
ENnCkR5FZCv_
Default Alt Text
(6 MB)

Event Timeline