/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
- * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2015, Nexenta Systems, Inc. All rights reserved.
+ * Copyright (c) 2013, 2014, Nexenta Systems, Inc. All rights reserved.
+ * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
+ * Copyright 2013 Saso Kiselkov. All rights reserved.
+ * Copyright (c) 2014 Integros [integros.com]
+ * Copyright 2016 Toomas Soome <tsoome@me.com>
+ * Copyright (c) 2016 Actifio, Inc. All rights reserved.
+ * Copyright (c) 2017 Datto Inc.
+ * Copyright 2017 Joyent, Inc.
*/
/*
+ * SPA: Storage Pool Allocator
+ *
* This file contains all the routines used when modifying on-disk SPA state.
* This includes opening, importing, destroying, exporting a pool, and syncing a
* pool.
#include <sys/vdev_disk.h>
#include <sys/metaslab.h>
#include <sys/metaslab_impl.h>
+#include <sys/mmp.h>
#include <sys/uberblock_impl.h>
#include <sys/txg.h>
#include <sys/avl.h>
#include <sys/zfs_ioctl.h>
#include <sys/dsl_scan.h>
#include <sys/zfeature.h>
+#include <sys/dsl_destroy.h>
+#include <sys/zvol.h>
#ifdef _KERNEL
+#include <sys/fm/protocol.h>
+#include <sys/fm/util.h>
#include <sys/bootprops.h>
#include <sys/callb.h>
#include <sys/cpupart.h>
#include "zfs_prop.h"
#include "zfs_comutil.h"
+/*
+ * The interval, in seconds, at which failed configuration cache file writes
+ * should be retried.
+ */
+static int zfs_ccw_retry_interval = 300;
+
typedef enum zti_modes {
- zti_mode_fixed, /* value is # of threads (min 1) */
- zti_mode_online_percent, /* value is % of online CPUs */
- zti_mode_batch, /* cpu-intensive; value is ignored */
- zti_mode_null, /* don't create a taskq */
- zti_nmodes
+ ZTI_MODE_FIXED, /* value is # of threads (min 1) */
+ ZTI_MODE_BATCH, /* cpu-intensive; value is ignored */
+ ZTI_MODE_NULL, /* don't create a taskq */
+ ZTI_NMODES
} zti_modes_t;
-#define ZTI_FIX(n) { zti_mode_fixed, (n) }
-#define ZTI_PCT(n) { zti_mode_online_percent, (n) }
-#define ZTI_BATCH { zti_mode_batch, 0 }
-#define ZTI_NULL { zti_mode_null, 0 }
+#define ZTI_P(n, q) { ZTI_MODE_FIXED, (n), (q) }
+#define ZTI_PCT(n) { ZTI_MODE_ONLINE_PERCENT, (n), 1 }
+#define ZTI_BATCH { ZTI_MODE_BATCH, 0, 1 }
+#define ZTI_NULL { ZTI_MODE_NULL, 0, 0 }
-#define ZTI_ONE ZTI_FIX(1)
+#define ZTI_N(n) ZTI_P(n, 1)
+#define ZTI_ONE ZTI_N(1)
typedef struct zio_taskq_info {
- enum zti_modes zti_mode;
+ zti_modes_t zti_mode;
uint_t zti_value;
+ uint_t zti_count;
} zio_taskq_info_t;
static const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = {
};
/*
- * Define the taskq threads for the following I/O types:
- * NULL, READ, WRITE, FREE, CLAIM, and IOCTL
+ * This table defines the taskq settings for each ZFS I/O type. When
+ * initializing a pool, we use this table to create an appropriately sized
+ * taskq. Some operations are low volume and therefore have a small, static
+ * number of threads assigned to their taskqs using the ZTI_N(#) or ZTI_ONE
+ * macros. Other operations process a large amount of data; the ZTI_BATCH
+ * macro causes us to create a taskq oriented for throughput. Some operations
+ * are so high frequency and short-lived that the taskq itself can become a a
+ * point of lock contention. The ZTI_P(#, #) macro indicates that we need an
+ * additional degree of parallelism specified by the number of threads per-
+ * taskq and the number of taskqs; when dispatching an event in this case, the
+ * particular taskq is chosen at random.
+ *
+ * The different taskq priorities are to handle the different contexts (issue
+ * and interrupt) and then to reserve threads for ZIO_PRIORITY_NOW I/Os that
+ * need to be handled with minimum delay.
*/
const zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = {
/* ISSUE ISSUE_HIGH INTR INTR_HIGH */
- { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL },
- { ZTI_FIX(8), ZTI_NULL, ZTI_BATCH, ZTI_NULL },
- { ZTI_BATCH, ZTI_FIX(5), ZTI_FIX(16), ZTI_FIX(5) },
- { ZTI_PCT(100), ZTI_NULL, ZTI_ONE, ZTI_NULL },
- { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL },
- { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL },
+ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* NULL */
+ { ZTI_N(8), ZTI_NULL, ZTI_P(12, 8), ZTI_NULL }, /* READ */
+ { ZTI_BATCH, ZTI_N(5), ZTI_P(12, 8), ZTI_N(5) }, /* WRITE */
+ { ZTI_P(12, 8), ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* FREE */
+ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* CLAIM */
+ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* IOCTL */
};
-static dsl_syncfunc_t spa_sync_version;
-static dsl_syncfunc_t spa_sync_props;
-static dsl_checkfunc_t spa_change_guid_check;
-static dsl_syncfunc_t spa_change_guid_sync;
+static sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl,
+ const char *name);
+static void spa_event_post(sysevent_t *ev);
+static void spa_sync_version(void *arg, dmu_tx_t *tx);
+static void spa_sync_props(void *arg, dmu_tx_t *tx);
static boolean_t spa_has_active_shared_spare(spa_t *spa);
static inline int spa_load_impl(spa_t *spa, uint64_t, nvlist_t *config,
spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig,
char **ereport);
static void spa_vdev_resilver_done(spa_t *spa);
-uint_t zio_taskq_batch_pct = 100; /* 1 thread per cpu in pset */
+uint_t zio_taskq_batch_pct = 75; /* 1 thread per cpu in pset */
id_t zio_taskq_psrset_bind = PS_NONE;
boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */
uint_t zio_taskq_basedc = 80; /* base duty cycle */
const char *propname = zpool_prop_to_name(prop);
nvlist_t *propval;
- VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+ VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0);
if (strval != NULL)
{
vdev_t *rvd = spa->spa_root_vdev;
dsl_pool_t *pool = spa->spa_dsl_pool;
- uint64_t size;
- uint64_t alloc;
- uint64_t space;
- uint64_t cap, version;
- zprop_source_t src = ZPROP_SRC_NONE;
+ uint64_t size, alloc, cap, version;
+ const zprop_source_t src = ZPROP_SRC_NONE;
spa_config_dirent_t *dp;
- int c;
+ metaslab_class_t *mc = spa_normal_class(spa);
ASSERT(MUTEX_HELD(&spa->spa_props_lock));
spa_prop_add_list(*nvp, ZPOOL_PROP_FREE, NULL,
size - alloc, src);
- space = 0;
- for (c = 0; c < rvd->vdev_children; c++) {
- vdev_t *tvd = rvd->vdev_child[c];
- space += tvd->vdev_max_asize - tvd->vdev_asize;
- }
- spa_prop_add_list(*nvp, ZPOOL_PROP_EXPANDSZ, NULL, space,
- src);
-
+ spa_prop_add_list(*nvp, ZPOOL_PROP_FRAGMENTATION, NULL,
+ metaslab_class_fragmentation(mc), src);
+ spa_prop_add_list(*nvp, ZPOOL_PROP_EXPANDSZ, NULL,
+ metaslab_class_expandable_space(mc), src);
spa_prop_add_list(*nvp, ZPOOL_PROP_READONLY, NULL,
(spa_mode(spa) == FREAD), src);
rvd->vdev_state, src);
version = spa_version(spa);
- if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION))
- src = ZPROP_SRC_DEFAULT;
- else
- src = ZPROP_SRC_LOCAL;
- spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, src);
+ if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION)) {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL,
+ version, ZPROP_SRC_DEFAULT);
+ } else {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL,
+ version, ZPROP_SRC_LOCAL);
+ }
}
if (pool != NULL) {
- dsl_dir_t *freedir = pool->dp_free_dir;
-
/*
* The $FREE directory was introduced in SPA_VERSION_DEADLISTS,
* when opening pools before this version freedir will be NULL.
*/
- if (freedir != NULL) {
+ if (pool->dp_free_dir != NULL) {
spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL,
- freedir->dd_phys->dd_used_bytes, src);
+ dsl_dir_phys(pool->dp_free_dir)->dd_used_bytes,
+ src);
} else {
spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING,
NULL, 0, src);
}
+
+ if (pool->dp_leak_dir != NULL) {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED, NULL,
+ dsl_dir_phys(pool->dp_leak_dir)->dd_used_bytes,
+ src);
+ } else {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED,
+ NULL, 0, src);
+ }
}
spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src);
spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root,
0, ZPROP_SRC_LOCAL);
+ if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
+ MIN(zfs_max_recordsize, SPA_MAXBLOCKSIZE), ZPROP_SRC_NONE);
+ } else {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
+ SPA_OLD_MAXBLOCKSIZE, ZPROP_SRC_NONE);
+ }
+
+ if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
+ DNODE_MAX_SIZE, ZPROP_SRC_NONE);
+ } else {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
+ DNODE_MIN_SIZE, ZPROP_SRC_NONE);
+ }
+
if ((dp = list_head(&spa->spa_config_list)) != NULL) {
if (dp->scd_path == NULL) {
spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
zap_attribute_t za;
int err;
- err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_PUSHPAGE);
+ err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP);
if (err)
- return err;
+ return (err);
mutex_enter(&spa->spa_props_lock);
dsl_dataset_t *ds = NULL;
dp = spa_get_dsl(spa);
- rw_enter(&dp->dp_config_rwlock, RW_READER);
+ dsl_pool_config_enter(dp, FTAG);
if ((err = dsl_dataset_hold_obj(dp,
za.za_first_integer, FTAG, &ds))) {
- rw_exit(&dp->dp_config_rwlock);
+ dsl_pool_config_exit(dp, FTAG);
break;
}
- strval = kmem_alloc(
- MAXNAMELEN + strlen(MOS_DIR_NAME) + 1,
- KM_PUSHPAGE);
+ strval = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN,
+ KM_SLEEP);
dsl_dataset_name(ds, strval);
dsl_dataset_rele(ds, FTAG);
- rw_exit(&dp->dp_config_rwlock);
+ dsl_pool_config_exit(dp, FTAG);
} else {
strval = NULL;
intval = za.za_first_integer;
spa_prop_add_list(*nvp, prop, strval, intval, src);
if (strval != NULL)
- kmem_free(strval,
- MAXNAMELEN + strlen(MOS_DIR_NAME) + 1);
+ kmem_free(strval, ZFS_MAX_DATASET_NAME_LEN);
break;
case 1:
/* string property */
- strval = kmem_alloc(za.za_num_integers, KM_PUSHPAGE);
+ strval = kmem_alloc(za.za_num_integers, KM_SLEEP);
err = zap_lookup(mos, spa->spa_pool_props_object,
za.za_name, 1, za.za_num_integers, strval);
if (err) {
switch ((int)prop) {
case ZPROP_INVAL:
if (!zpool_prop_feature(propname)) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
* Sanitize the input.
*/
if (nvpair_type(elem) != DATA_TYPE_UINT64) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
if (nvpair_value_uint64(elem, &intval) != 0) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
if (intval != 0) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
fname = strchr(propname, '@') + 1;
if (zfeature_lookup_name(fname, NULL) != 0) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
(intval < spa_version(spa) ||
intval > SPA_VERSION_BEFORE_FEATURES ||
has_feature))
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
case ZPOOL_PROP_DELEGATION:
case ZPOOL_PROP_AUTOEXPAND:
error = nvpair_value_uint64(elem, &intval);
if (!error && intval > 1)
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
+ break;
+
+ case ZPOOL_PROP_MULTIHOST:
+ error = nvpair_value_uint64(elem, &intval);
+ if (!error && intval > 1)
+ error = SET_ERROR(EINVAL);
+
+ if (!error && !spa_get_hostid())
+ error = SET_ERROR(ENOTSUP);
+
break;
case ZPOOL_PROP_BOOTFS:
* the bootfs property cannot be set.
*/
if (spa_version(spa) < SPA_VERSION_BOOTFS) {
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
break;
}
* Make sure the vdev config is bootable
*/
if (!vdev_is_bootable(spa->spa_root_vdev)) {
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
break;
}
if (!error) {
objset_t *os;
- uint64_t compress;
+ uint64_t propval;
if (strval == NULL || strval[0] == '\0') {
objnum = zpool_prop_default_numeric(
break;
}
- if ((error = dmu_objset_hold(strval,FTAG,&os)))
+ error = dmu_objset_hold(strval, FTAG, &os);
+ if (error)
break;
- /* Must be ZPL and not gzip compressed. */
+ /*
+ * Must be ZPL, and its property settings
+ * must be supported by GRUB (compression
+ * is not gzip, and large blocks or large
+ * dnodes are not used).
+ */
if (dmu_objset_type(os) != DMU_OST_ZFS) {
- error = ENOTSUP;
- } else if ((error = dsl_prop_get_integer(strval,
+ error = SET_ERROR(ENOTSUP);
+ } else if ((error =
+ dsl_prop_get_int_ds(dmu_objset_ds(os),
zfs_prop_to_name(ZFS_PROP_COMPRESSION),
- &compress, NULL)) == 0 &&
- !BOOTFS_COMPRESS_VALID(compress)) {
- error = ENOTSUP;
+ &propval)) == 0 &&
+ !BOOTFS_COMPRESS_VALID(propval)) {
+ error = SET_ERROR(ENOTSUP);
+ } else if ((error =
+ dsl_prop_get_int_ds(dmu_objset_ds(os),
+ zfs_prop_to_name(ZFS_PROP_DNODESIZE),
+ &propval)) == 0 &&
+ propval != ZFS_DNSIZE_LEGACY) {
+ error = SET_ERROR(ENOTSUP);
} else {
objnum = dmu_objset_id(os);
}
case ZPOOL_PROP_FAILUREMODE:
error = nvpair_value_uint64(elem, &intval);
- if (!error && (intval < ZIO_FAILURE_MODE_WAIT ||
- intval > ZIO_FAILURE_MODE_PANIC))
- error = EINVAL;
+ if (!error && intval > ZIO_FAILURE_MODE_PANIC)
+ error = SET_ERROR(EINVAL);
/*
* This is a special case which only occurs when
*/
if (!error && spa_suspended(spa)) {
spa->spa_failmode = intval;
- error = EIO;
+ error = SET_ERROR(EIO);
}
break;
break;
if (strval[0] != '/') {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
strcmp(slash, "/..") == 0)
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
case ZPOOL_PROP_COMMENT:
break;
for (check = strval; *check != '\0'; check++) {
if (!isprint(*check)) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
- check++;
}
if (strlen(strval) > ZPROP_MAX_COMMENT)
- error = E2BIG;
+ error = SET_ERROR(E2BIG);
break;
case ZPOOL_PROP_DEDUPDITTO:
if (spa_version(spa) < SPA_VERSION_DEDUP)
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
else
error = nvpair_value_uint64(elem, &intval);
if (error == 0 &&
intval != 0 && intval < ZIO_DEDUPDITTO_MIN)
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
default:
return;
dp = kmem_alloc(sizeof (spa_config_dirent_t),
- KM_PUSHPAGE);
+ KM_SLEEP);
if (cachefile[0] == '\0')
dp->scd_path = spa_strdup(spa_config_path);
* read object, the features for write object, or the
* feature descriptions object.
*/
- error = dsl_sync_task_do(spa_get_dsl(spa), NULL,
- spa_sync_version, spa, &ver, 6);
+ error = dsl_sync_task(spa->spa_name, NULL,
+ spa_sync_version, &ver,
+ 6, ZFS_SPACE_CHECK_RESERVED);
if (error)
return (error);
continue;
}
if (need_sync) {
- return (dsl_sync_task_do(spa_get_dsl(spa), NULL, spa_sync_props,
- spa, nvp, 6));
+ return (dsl_sync_task(spa->spa_name, NULL, spa_sync_props,
+ nvp, 6, ZFS_SPACE_CHECK_RESERVED));
}
return (0);
/*ARGSUSED*/
static int
-spa_change_guid_check(void *arg1, void *arg2, dmu_tx_t *tx)
+spa_change_guid_check(void *arg, dmu_tx_t *tx)
{
- spa_t *spa = arg1;
+ spa_t *spa = dmu_tx_pool(tx)->dp_spa;
vdev_t *rvd = spa->spa_root_vdev;
uint64_t vdev_state;
- ASSERTV(uint64_t *newguid = arg2);
+ ASSERTV(uint64_t *newguid = arg);
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
vdev_state = rvd->vdev_state;
spa_config_exit(spa, SCL_STATE, FTAG);
if (vdev_state != VDEV_STATE_HEALTHY)
- return (ENXIO);
+ return (SET_ERROR(ENXIO));
ASSERT3U(spa_guid(spa), !=, *newguid);
}
static void
-spa_change_guid_sync(void *arg1, void *arg2, dmu_tx_t *tx)
+spa_change_guid_sync(void *arg, dmu_tx_t *tx)
{
- spa_t *spa = arg1;
- uint64_t *newguid = arg2;
+ uint64_t *newguid = arg;
+ spa_t *spa = dmu_tx_pool(tx)->dp_spa;
uint64_t oldguid;
vdev_t *rvd = spa->spa_root_vdev;
vdev_config_dirty(rvd);
spa_config_exit(spa, SCL_STATE, FTAG);
- spa_history_log_internal(LOG_POOL_GUID_CHANGE, spa, tx,
- "old=%lld new=%lld", oldguid, *newguid);
+ spa_history_log_internal(spa, "guid change", tx, "old=%llu new=%llu",
+ oldguid, *newguid);
}
/*
int error;
uint64_t guid;
+ mutex_enter(&spa->spa_vdev_top_lock);
mutex_enter(&spa_namespace_lock);
guid = spa_generate_guid(NULL);
- error = dsl_sync_task_do(spa_get_dsl(spa), spa_change_guid_check,
- spa_change_guid_sync, spa, &guid, 5);
+ error = dsl_sync_task(spa->spa_name, spa_change_guid_check,
+ spa_change_guid_sync, &guid, 5, ZFS_SPACE_CHECK_RESERVED);
if (error == 0) {
spa_config_sync(spa, B_FALSE, B_TRUE);
- spa_event_notify(spa, NULL, FM_EREPORT_ZFS_POOL_REGUID);
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_REGUID);
}
mutex_exit(&spa_namespace_lock);
+ mutex_exit(&spa->spa_vdev_top_lock);
return (error);
}
static int
spa_error_entry_compare(const void *a, const void *b)
{
- spa_error_entry_t *sa = (spa_error_entry_t *)a;
- spa_error_entry_t *sb = (spa_error_entry_t *)b;
+ const spa_error_entry_t *sa = (const spa_error_entry_t *)a;
+ const spa_error_entry_t *sb = (const spa_error_entry_t *)b;
int ret;
- ret = bcmp(&sa->se_bookmark, &sb->se_bookmark,
- sizeof (zbookmark_t));
+ ret = memcmp(&sa->se_bookmark, &sb->se_bookmark,
+ sizeof (zbookmark_phys_t));
- if (ret < 0)
- return (-1);
- else if (ret > 0)
- return (1);
- else
- return (0);
+ return (AVL_ISIGN(ret));
}
/*
offsetof(spa_error_entry_t, se_avl));
}
-static taskq_t *
-spa_taskq_create(spa_t *spa, const char *name, enum zti_modes mode,
- uint_t value)
+static void
+spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
{
- uint_t flags = TASKQ_PREPOPULATE;
+ const zio_taskq_info_t *ztip = &zio_taskqs[t][q];
+ enum zti_modes mode = ztip->zti_mode;
+ uint_t value = ztip->zti_value;
+ uint_t count = ztip->zti_count;
+ spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
+ char name[32];
+ uint_t i, flags = 0;
boolean_t batch = B_FALSE;
- switch (mode) {
- case zti_mode_null:
- return (NULL); /* no taskq needed */
+ if (mode == ZTI_MODE_NULL) {
+ tqs->stqs_count = 0;
+ tqs->stqs_taskq = NULL;
+ return;
+ }
+
+ ASSERT3U(count, >, 0);
- case zti_mode_fixed:
+ tqs->stqs_count = count;
+ tqs->stqs_taskq = kmem_alloc(count * sizeof (taskq_t *), KM_SLEEP);
+
+ switch (mode) {
+ case ZTI_MODE_FIXED:
ASSERT3U(value, >=, 1);
value = MAX(value, 1);
+ flags |= TASKQ_DYNAMIC;
break;
- case zti_mode_batch:
+ case ZTI_MODE_BATCH:
batch = B_TRUE;
flags |= TASKQ_THREADS_CPU_PCT;
- value = zio_taskq_batch_pct;
- break;
-
- case zti_mode_online_percent:
- flags |= TASKQ_THREADS_CPU_PCT;
+ value = MIN(zio_taskq_batch_pct, 100);
break;
default:
- panic("unrecognized mode for %s taskq (%u:%u) in "
+ panic("unrecognized mode for %s_%s taskq (%u:%u) in "
"spa_activate()",
- name, mode, value);
+ zio_type_name[t], zio_taskq_types[q], mode, value);
break;
}
- if (zio_taskq_sysdc && spa->spa_proc != &p0) {
- if (batch)
- flags |= TASKQ_DC_BATCH;
+ for (i = 0; i < count; i++) {
+ taskq_t *tq;
+
+ if (count > 1) {
+ (void) snprintf(name, sizeof (name), "%s_%s_%u",
+ zio_type_name[t], zio_taskq_types[q], i);
+ } else {
+ (void) snprintf(name, sizeof (name), "%s_%s",
+ zio_type_name[t], zio_taskq_types[q]);
+ }
+
+ if (zio_taskq_sysdc && spa->spa_proc != &p0) {
+ if (batch)
+ flags |= TASKQ_DC_BATCH;
+
+ tq = taskq_create_sysdc(name, value, 50, INT_MAX,
+ spa->spa_proc, zio_taskq_basedc, flags);
+ } else {
+ pri_t pri = maxclsyspri;
+ /*
+ * The write issue taskq can be extremely CPU
+ * intensive. Run it at slightly less important
+ * priority than the other taskqs. Under Linux this
+ * means incrementing the priority value on platforms
+ * like illumos it should be decremented.
+ */
+ if (t == ZIO_TYPE_WRITE && q == ZIO_TASKQ_ISSUE)
+ pri++;
+
+ tq = taskq_create_proc(name, value, pri, 50,
+ INT_MAX, spa->spa_proc, flags);
+ }
+
+ tqs->stqs_taskq[i] = tq;
+ }
+}
+
+static void
+spa_taskqs_fini(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
+{
+ spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
+ uint_t i;
+
+ if (tqs->stqs_taskq == NULL) {
+ ASSERT3U(tqs->stqs_count, ==, 0);
+ return;
+ }
+
+ for (i = 0; i < tqs->stqs_count; i++) {
+ ASSERT3P(tqs->stqs_taskq[i], !=, NULL);
+ taskq_destroy(tqs->stqs_taskq[i]);
+ }
+
+ kmem_free(tqs->stqs_taskq, tqs->stqs_count * sizeof (taskq_t *));
+ tqs->stqs_taskq = NULL;
+}
+
+/*
+ * Dispatch a task to the appropriate taskq for the ZFS I/O type and priority.
+ * Note that a type may have multiple discrete taskqs to avoid lock contention
+ * on the taskq itself. In that case we choose which taskq at random by using
+ * the low bits of gethrtime().
+ */
+void
+spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
+ task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent)
+{
+ spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
+ taskq_t *tq;
+
+ ASSERT3P(tqs->stqs_taskq, !=, NULL);
+ ASSERT3U(tqs->stqs_count, !=, 0);
+
+ if (tqs->stqs_count == 1) {
+ tq = tqs->stqs_taskq[0];
+ } else {
+ tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count];
+ }
+
+ taskq_dispatch_ent(tq, func, arg, flags, ent);
+}
+
+/*
+ * Same as spa_taskq_dispatch_ent() but block on the task until completion.
+ */
+void
+spa_taskq_dispatch_sync(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
+ task_func_t *func, void *arg, uint_t flags)
+{
+ spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
+ taskq_t *tq;
+ taskqid_t id;
+
+ ASSERT3P(tqs->stqs_taskq, !=, NULL);
+ ASSERT3U(tqs->stqs_count, !=, 0);
- return (taskq_create_sysdc(name, value, 50, INT_MAX,
- spa->spa_proc, zio_taskq_basedc, flags));
+ if (tqs->stqs_count == 1) {
+ tq = tqs->stqs_taskq[0];
+ } else {
+ tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count];
}
- return (taskq_create_proc(name, value, maxclsyspri, 50, INT_MAX,
- spa->spa_proc, flags));
+
+ id = taskq_dispatch(tq, func, arg, flags);
+ if (id)
+ taskq_wait_id(tq, id);
}
static void
for (t = 0; t < ZIO_TYPES; t++) {
for (q = 0; q < ZIO_TASKQ_TYPES; q++) {
- const zio_taskq_info_t *ztip = &zio_taskqs[t][q];
- enum zti_modes mode = ztip->zti_mode;
- uint_t value = ztip->zti_value;
- char name[32];
-
- (void) snprintf(name, sizeof (name),
- "%s_%s", zio_type_name[t], zio_taskq_types[q]);
-
- spa->spa_zio_taskq[t][q] =
- spa_taskq_create(spa, name, mode, value);
+ spa_taskqs_init(spa, t, q);
}
}
}
list_create(&spa->spa_config_dirty_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_config_dirty_node));
+ list_create(&spa->spa_evicting_os_list, sizeof (objset_t),
+ offsetof(objset_t, os_evicting_node));
list_create(&spa->spa_state_dirty_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_state_dirty_node));
- txg_list_create(&spa->spa_vdev_txg_list,
+ txg_list_create(&spa->spa_vdev_txg_list, spa,
offsetof(struct vdev, vdev_txg_node));
avl_create(&spa->spa_errlist_scrub,
avl_create(&spa->spa_errlist_last,
spa_error_entry_compare, sizeof (spa_error_entry_t),
offsetof(spa_error_entry_t, se_avl));
+
+ /*
+ * This taskq is used to perform zvol-minor-related tasks
+ * asynchronously. This has several advantages, including easy
+ * resolution of various deadlocks (zfsonlinux bug #3681).
+ *
+ * The taskq must be single threaded to ensure tasks are always
+ * processed in the order in which they were dispatched.
+ *
+ * A taskq per pool allows one to keep the pools independent.
+ * This way if one pool is suspended, it will not impact another.
+ *
+ * The preferred location to dispatch a zvol minor task is a sync
+ * task. In this context, there is easy access to the spa_t and minimal
+ * error handling is required because the sync task must succeed.
+ */
+ spa->spa_zvol_taskq = taskq_create("z_zvol", 1, defclsyspri,
+ 1, INT_MAX, 0);
+
+ /*
+ * The taskq to upgrade datasets in this pool. Currently used by
+ * feature SPA_FEATURE_USEROBJ_ACCOUNTING.
+ */
+ spa->spa_upgrade_taskq = taskq_create("z_upgrade", boot_ncpus,
+ defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC);
}
/*
ASSERT(spa->spa_async_zio_root == NULL);
ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED);
+ spa_evicting_os_wait(spa);
+
+ if (spa->spa_zvol_taskq) {
+ taskq_destroy(spa->spa_zvol_taskq);
+ spa->spa_zvol_taskq = NULL;
+ }
+
+ if (spa->spa_upgrade_taskq) {
+ taskq_destroy(spa->spa_upgrade_taskq);
+ spa->spa_upgrade_taskq = NULL;
+ }
+
txg_list_destroy(&spa->spa_vdev_txg_list);
list_destroy(&spa->spa_config_dirty_list);
+ list_destroy(&spa->spa_evicting_os_list);
list_destroy(&spa->spa_state_dirty_list);
+ taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
+
for (t = 0; t < ZIO_TYPES; t++) {
for (q = 0; q < ZIO_TASKQ_TYPES; q++) {
- if (spa->spa_zio_taskq[t][q] != NULL)
- taskq_destroy(spa->spa_zio_taskq[t][q]);
- spa->spa_zio_taskq[t][q] = NULL;
+ spa_taskqs_fini(spa, t, q);
}
}
if (error) {
vdev_free(*vdp);
*vdp = NULL;
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
}
for (c = 0; c < children; c++) {
static void
spa_unload(spa_t *spa)
{
- int i;
+ int i, c;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
spa->spa_sync_on = B_FALSE;
}
+ /*
+ * Even though vdev_free() also calls vdev_metaslab_fini, we need
+ * to call it earlier, before we wait for async i/o to complete.
+ * This ensures that there is no async metaslab prefetching, by
+ * calling taskq_wait(mg_taskq).
+ */
+ if (spa->spa_root_vdev != NULL) {
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ for (c = 0; c < spa->spa_root_vdev->vdev_children; c++)
+ vdev_metaslab_fini(spa->spa_root_vdev->vdev_child[c]);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ }
+
+ if (spa->spa_mmp.mmp_thread)
+ mmp_thread_stop(spa);
+
/*
* Wait for any outstanding async I/O to complete.
*/
if (spa->spa_async_zio_root != NULL) {
- (void) zio_wait(spa->spa_async_zio_root);
+ for (i = 0; i < max_ncpus; i++)
+ (void) zio_wait(spa->spa_async_zio_root[i]);
+ kmem_free(spa->spa_async_zio_root, max_ncpus * sizeof (void *));
spa->spa_async_zio_root = NULL;
}
bpobj_close(&spa->spa_deferred_bpobj);
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+
+ /*
+ * Close all vdevs.
+ */
+ if (spa->spa_root_vdev)
+ vdev_free(spa->spa_root_vdev);
+ ASSERT(spa->spa_root_vdev == NULL);
+
/*
* Close the dsl pool.
*/
ddt_unload(spa);
- spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
-
/*
* Drop and purge level 2 cache
*/
spa_l2cache_drop(spa);
- /*
- * Close all vdevs.
- */
- if (spa->spa_root_vdev)
- vdev_free(spa->spa_root_vdev);
- ASSERT(spa->spa_root_vdev == NULL);
-
for (i = 0; i < spa->spa_spares.sav_count; i++)
vdev_free(spa->spa_spares.sav_vdevs[i]);
if (spa->spa_spares.sav_vdevs) {
* validate each vdev on the spare list. If the vdev also exists in the
* active configuration, then we also mark this vdev as an active spare.
*/
- spa->spa_spares.sav_vdevs = kmem_alloc(nspares * sizeof (void *),
- KM_PUSHPAGE);
+ spa->spa_spares.sav_vdevs = kmem_zalloc(nspares * sizeof (void *),
+ KM_SLEEP);
for (i = 0; i < spa->spa_spares.sav_count; i++) {
VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0,
VDEV_ALLOC_SPARE) == 0);
DATA_TYPE_NVLIST_ARRAY) == 0);
spares = kmem_alloc(spa->spa_spares.sav_count * sizeof (void *),
- KM_PUSHPAGE);
+ KM_SLEEP);
for (i = 0; i < spa->spa_spares.sav_count; i++)
spares[i] = vdev_config_generate(spa,
spa->spa_spares.sav_vdevs[i], B_TRUE, VDEV_CONFIG_SPARE);
uint_t nl2cache;
int i, j, oldnvdevs;
uint64_t guid;
- vdev_t *vd, **oldvdevs, **newvdevs = NULL;
+ vdev_t *vd, **oldvdevs, **newvdevs;
spa_aux_vdev_t *sav = &spa->spa_l2cache;
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
- if (sav->sav_config != NULL) {
- VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
- ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0);
- newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_PUSHPAGE);
- } else {
- nl2cache = 0;
- }
-
oldvdevs = sav->sav_vdevs;
oldnvdevs = sav->sav_count;
sav->sav_vdevs = NULL;
sav->sav_count = 0;
+ if (sav->sav_config == NULL) {
+ nl2cache = 0;
+ newvdevs = NULL;
+ goto out;
+ }
+
+ VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
+ ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0);
+ newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_SLEEP);
+
/*
* Process new nvlist of vdevs.
*/
}
}
+ sav->sav_vdevs = newvdevs;
+ sav->sav_count = (int)nl2cache;
+
+ /*
+ * Recompute the stashed list of l2cache devices, with status
+ * information this time.
+ */
+ VERIFY(nvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE,
+ DATA_TYPE_NVLIST_ARRAY) == 0);
+
+ l2cache = kmem_alloc(sav->sav_count * sizeof (void *), KM_SLEEP);
+ for (i = 0; i < sav->sav_count; i++)
+ l2cache[i] = vdev_config_generate(spa,
+ sav->sav_vdevs[i], B_TRUE, VDEV_CONFIG_L2CACHE);
+ VERIFY(nvlist_add_nvlist_array(sav->sav_config,
+ ZPOOL_CONFIG_L2CACHE, l2cache, sav->sav_count) == 0);
+
+out:
/*
* Purge vdevs that were dropped
*/
if (oldvdevs)
kmem_free(oldvdevs, oldnvdevs * sizeof (void *));
- if (sav->sav_config == NULL)
- goto out;
-
- sav->sav_vdevs = newvdevs;
- sav->sav_count = (int)nl2cache;
-
- /*
- * Recompute the stashed list of l2cache devices, with status
- * information this time.
- */
- VERIFY(nvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE,
- DATA_TYPE_NVLIST_ARRAY) == 0);
-
- l2cache = kmem_alloc(sav->sav_count * sizeof (void *), KM_PUSHPAGE);
- for (i = 0; i < sav->sav_count; i++)
- l2cache[i] = vdev_config_generate(spa,
- sav->sav_vdevs[i], B_TRUE, VDEV_CONFIG_L2CACHE);
- VERIFY(nvlist_add_nvlist_array(sav->sav_config,
- ZPOOL_CONFIG_L2CACHE, l2cache, sav->sav_count) == 0);
-out:
for (i = 0; i < sav->sav_count; i++)
nvlist_free(l2cache[i]);
if (sav->sav_count)
nvsize = *(uint64_t *)db->db_data;
dmu_buf_rele(db, FTAG);
- packed = kmem_alloc(nvsize, KM_PUSHPAGE | KM_NODEBUG);
+ packed = vmem_alloc(nvsize, KM_SLEEP);
error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed,
DMU_READ_PREFETCH);
if (error == 0)
error = nvlist_unpack(packed, nvsize, value, 0);
- kmem_free(packed, nvsize);
+ vmem_free(packed, nvsize);
return (error);
}
for (c = 0; c < vd->vdev_children; c++)
spa_check_removed(vd->vdev_child[c]);
- if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd)) {
- zfs_ereport_post(FM_EREPORT_RESOURCE_AUTOREPLACE,
- vd->vdev_spa, vd, NULL, 0, 0);
- spa_event_notify(vd->vdev_spa, vd, FM_EREPORT_ZFS_DEVICE_CHECK);
+ if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd) &&
+ !vd->vdev_ishole) {
+ zfs_post_autoreplace(vd->vdev_spa, vd);
+ spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_CHECK);
+ }
+}
+
+static void
+spa_config_valid_zaps(vdev_t *vd, vdev_t *mvd)
+{
+ uint64_t i;
+
+ ASSERT3U(vd->vdev_children, ==, mvd->vdev_children);
+
+ vd->vdev_top_zap = mvd->vdev_top_zap;
+ vd->vdev_leaf_zap = mvd->vdev_leaf_zap;
+
+ for (i = 0; i < vd->vdev_children; i++) {
+ spa_config_valid_zaps(vd->vdev_child[i], mvd->vdev_child[i]);
}
}
nvlist_t **child, *nv;
uint64_t idx = 0;
- child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t **),
- KM_PUSHPAGE);
- VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+ child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t *),
+ KM_SLEEP);
+ VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
for (c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
vdev_reopen(rvd);
- } else if (mtvd->vdev_islog) {
+ } else {
+ if (mtvd->vdev_islog) {
+ /*
+ * Load the slog device's state from the MOS
+ * config since it's possible that the label
+ * does not contain the most up-to-date
+ * information.
+ */
+ vdev_load_log_state(tvd, mtvd);
+ vdev_reopen(tvd);
+ }
+
/*
- * Load the slog device's state from the MOS config
- * since it's possible that the label does not
- * contain the most up-to-date information.
+ * Per-vdev ZAP info is stored exclusively in the MOS.
*/
- vdev_load_log_state(tvd, mtvd);
- vdev_reopen(tvd);
+ spa_config_valid_zaps(tvd, mtvd);
}
}
+
vdev_free(mrvd);
spa_config_exit(spa, SCL_ALL, FTAG);
/*
* Check for missing log devices
*/
-static int
+static boolean_t
spa_check_logs(spa_t *spa)
{
+ boolean_t rv = B_FALSE;
+ dsl_pool_t *dp = spa_get_dsl(spa);
+
switch (spa->spa_log_state) {
default:
break;
case SPA_LOG_MISSING:
/* need to recheck in case slog has been restored */
case SPA_LOG_UNKNOWN:
- if (dmu_objset_find(spa->spa_name, zil_check_log_chain, NULL,
- DS_FIND_CHILDREN)) {
+ rv = (dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
+ zil_check_log_chain, NULL, DS_FIND_CHILDREN) != 0);
+ if (rv)
spa_set_log_state(spa, SPA_LOG_MISSING);
- return (1);
- }
break;
}
- return (0);
+ return (rv);
}
static boolean_t
int
spa_offline_log(spa_t *spa)
{
- int error = 0;
-
- if ((error = dmu_objset_find(spa_name(spa), zil_vdev_offline,
- NULL, DS_FIND_CHILDREN)) == 0) {
+ int error;
+ error = dmu_objset_find(spa_name(spa), zil_vdev_offline,
+ NULL, DS_FIND_CHILDREN);
+ if (error == 0) {
/*
* We successfully offlined the log device, sync out the
* current txg so that the "stubby" block can be removed
spa_load_error_t *sle = zio->io_private;
dmu_object_type_t type = BP_GET_TYPE(bp);
int error = zio->io_error;
+ spa_t *spa = zio->io_spa;
+ abd_free(zio->io_abd);
if (error) {
if ((BP_GET_LEVEL(bp) != 0 || DMU_OT_IS_METADATA(type)) &&
type != DMU_OT_INTENT_LOG)
- atomic_add_64(&sle->sle_meta_count, 1);
+ atomic_inc_64(&sle->sle_meta_count);
else
- atomic_add_64(&sle->sle_data_count, 1);
+ atomic_inc_64(&sle->sle_data_count);
}
- zio_data_buf_free(zio->io_data, zio->io_size);
+
+ mutex_enter(&spa->spa_scrub_lock);
+ spa->spa_scrub_inflight--;
+ cv_broadcast(&spa->spa_scrub_io_cv);
+ mutex_exit(&spa->spa_scrub_lock);
}
+/*
+ * Maximum number of concurrent scrub i/os to create while verifying
+ * a pool while importing it.
+ */
+int spa_load_verify_maxinflight = 10000;
+int spa_load_verify_metadata = B_TRUE;
+int spa_load_verify_data = B_TRUE;
+
/*ARGSUSED*/
static int
spa_load_verify_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
- arc_buf_t *pbuf, const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg)
+ const zbookmark_phys_t *zb, const dnode_phys_t *dnp, void *arg)
{
- if (bp != NULL) {
- zio_t *rio = arg;
- size_t size = BP_GET_PSIZE(bp);
- void *data = zio_data_buf_alloc(size);
+ zio_t *rio;
+ size_t size;
+
+ if (bp == NULL || BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp))
+ return (0);
+ /*
+ * Note: normally this routine will not be called if
+ * spa_load_verify_metadata is not set. However, it may be useful
+ * to manually set the flag after the traversal has begun.
+ */
+ if (!spa_load_verify_metadata)
+ return (0);
+ if (!BP_IS_METADATA(bp) && !spa_load_verify_data)
+ return (0);
+
+ rio = arg;
+ size = BP_GET_PSIZE(bp);
+
+ mutex_enter(&spa->spa_scrub_lock);
+ while (spa->spa_scrub_inflight >= spa_load_verify_maxinflight)
+ cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
+ spa->spa_scrub_inflight++;
+ mutex_exit(&spa->spa_scrub_lock);
+
+ zio_nowait(zio_read(rio, spa, bp, abd_alloc_for_io(size, B_FALSE), size,
+ spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB,
+ ZIO_FLAG_SPECULATIVE | ZIO_FLAG_CANFAIL |
+ ZIO_FLAG_SCRUB | ZIO_FLAG_RAW, zb));
+ return (0);
+}
+
+/* ARGSUSED */
+int
+verify_dataset_name_len(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg)
+{
+ if (dsl_dataset_namelen(ds) >= ZFS_MAX_DATASET_NAME_LEN)
+ return (SET_ERROR(ENAMETOOLONG));
- zio_nowait(zio_read(rio, spa, bp, data, size,
- spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB,
- ZIO_FLAG_SPECULATIVE | ZIO_FLAG_CANFAIL |
- ZIO_FLAG_SCRUB | ZIO_FLAG_RAW, zb));
- }
return (0);
}
spa_load_error_t sle = { 0 };
zpool_rewind_policy_t policy;
boolean_t verify_ok = B_FALSE;
- int error;
+ int error = 0;
zpool_get_rewind_policy(spa->spa_config, &policy);
if (policy.zrp_request & ZPOOL_NEVER_REWIND)
return (0);
+ dsl_pool_config_enter(spa->spa_dsl_pool, FTAG);
+ error = dmu_objset_find_dp(spa->spa_dsl_pool,
+ spa->spa_dsl_pool->dp_root_dir_obj, verify_dataset_name_len, NULL,
+ DS_FIND_CHILDREN);
+ dsl_pool_config_exit(spa->spa_dsl_pool, FTAG);
+ if (error != 0)
+ return (error);
+
rio = zio_root(spa, NULL, &sle,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE);
- error = traverse_pool(spa, spa->spa_verify_min_txg,
- TRAVERSE_PRE | TRAVERSE_PREFETCH, spa_load_verify_cb, rio);
+ if (spa_load_verify_metadata) {
+ error = traverse_pool(spa, spa->spa_verify_min_txg,
+ TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA,
+ spa_load_verify_cb, rio);
+ }
(void) zio_wait(rio);
if (error) {
if (error != ENXIO && error != EIO)
- error = EIO;
+ error = SET_ERROR(EIO);
return (error);
}
&glist, &gcount) != 0)
return;
- vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_PUSHPAGE);
+ vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_SLEEP);
/* attempt to online all the vdevs & validate */
attempt_reopen = B_TRUE;
nvlist_t *nvl;
if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid))
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
ASSERT(spa->spa_comment == NULL);
if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
if ((state == SPA_LOAD_IMPORT || state == SPA_LOAD_TRYIMPORT) &&
spa_guid_exists(pool_guid, 0)) {
- error = EEXIST;
+ error = SET_ERROR(EEXIST);
} else {
spa->spa_config_guid = pool_guid;
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT,
&nvl) == 0) {
VERIFY(nvlist_dup(nvl, &spa->spa_config_splitting,
- KM_PUSHPAGE) == 0);
+ KM_SLEEP) == 0);
}
nvlist_free(spa->spa_load_info);
mosconfig, &ereport);
}
+ /*
+ * Don't count references from objsets that are already closed
+ * and are making their way through the eviction process.
+ */
+ spa_evicting_os_wait(spa);
spa->spa_minref = refcount_count(&spa->spa_refcount);
if (error) {
if (error != EEXIST) {
return (error);
}
+#ifdef ZFS_DEBUG
/*
- * Load an existing storage pool, using the pool's builtin spa_config as a
- * source of configuration information.
+ * Count the number of per-vdev ZAPs associated with all of the vdevs in the
+ * vdev tree rooted in the given vd, and ensure that each ZAP is present in the
+ * spa's per-vdev ZAP list.
*/
-__attribute__((always_inline))
-static inline int
-spa_load_impl(spa_t *spa, uint64_t pool_guid, nvlist_t *config,
- spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig,
- char **ereport)
+static uint64_t
+vdev_count_verify_zaps(vdev_t *vd)
{
- int error = 0;
- nvlist_t *nvroot = NULL;
- nvlist_t *label;
- vdev_t *rvd;
- uberblock_t *ub = &spa->spa_uberblock;
- uint64_t children, config_cache_txg = spa->spa_config_txg;
- int orig_mode = spa->spa_mode;
- int parse;
- uint64_t obj;
- boolean_t missing_feat_write = B_FALSE;
+ spa_t *spa = vd->vdev_spa;
+ uint64_t total = 0;
+ uint64_t i;
- /*
- * If this is an untrusted config, access the pool in read-only mode.
- * This prevents things like resilvering recently removed devices.
- */
- if (!mosconfig)
- spa->spa_mode = FREAD;
+ if (vd->vdev_top_zap != 0) {
+ total++;
+ ASSERT0(zap_lookup_int(spa->spa_meta_objset,
+ spa->spa_all_vdev_zaps, vd->vdev_top_zap));
+ }
+ if (vd->vdev_leaf_zap != 0) {
+ total++;
+ ASSERT0(zap_lookup_int(spa->spa_meta_objset,
+ spa->spa_all_vdev_zaps, vd->vdev_leaf_zap));
+ }
+
+ for (i = 0; i < vd->vdev_children; i++) {
+ total += vdev_count_verify_zaps(vd->vdev_child[i]);
+ }
+
+ return (total);
+}
+#endif
+
+/*
+ * Determine whether the activity check is required.
+ */
+static boolean_t
+spa_activity_check_required(spa_t *spa, uberblock_t *ub, nvlist_t *label,
+ nvlist_t *config)
+{
+ uint64_t state = 0;
+ uint64_t hostid = 0;
+ uint64_t tryconfig_txg = 0;
+ uint64_t tryconfig_timestamp = 0;
+ nvlist_t *nvinfo;
+
+ if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
+ nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
+ (void) nvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG,
+ &tryconfig_txg);
+ (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
+ &tryconfig_timestamp);
+ }
+
+ (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, &state);
+
+ /*
+ * Disable the MMP activity check - This is used by zdb which
+ * is intended to be used on potentially active pools.
+ */
+ if (spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP)
+ return (B_FALSE);
+
+ /*
+ * Skip the activity check when the MMP feature is disabled.
+ */
+ if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay == 0)
+ return (B_FALSE);
+ /*
+ * If the tryconfig_* values are nonzero, they are the results of an
+ * earlier tryimport. If they match the uberblock we just found, then
+ * the pool has not changed and we return false so we do not test a
+ * second time.
+ */
+ if (tryconfig_txg && tryconfig_txg == ub->ub_txg &&
+ tryconfig_timestamp && tryconfig_timestamp == ub->ub_timestamp)
+ return (B_FALSE);
+
+ /*
+ * Allow the activity check to be skipped when importing the pool
+ * on the same host which last imported it. Since the hostid from
+ * configuration may be stale use the one read from the label.
+ */
+ if (nvlist_exists(label, ZPOOL_CONFIG_HOSTID))
+ hostid = fnvlist_lookup_uint64(label, ZPOOL_CONFIG_HOSTID);
+
+ if (hostid == spa_get_hostid())
+ return (B_FALSE);
+
+ /*
+ * Skip the activity test when the pool was cleanly exported.
+ */
+ if (state != POOL_STATE_ACTIVE)
+ return (B_FALSE);
+
+ return (B_TRUE);
+}
+
+/*
+ * Perform the import activity check. If the user canceled the import or
+ * we detected activity then fail.
+ */
+static int
+spa_activity_check(spa_t *spa, uberblock_t *ub, nvlist_t *config)
+{
+ uint64_t import_intervals = MAX(zfs_multihost_import_intervals, 1);
+ uint64_t txg = ub->ub_txg;
+ uint64_t timestamp = ub->ub_timestamp;
+ uint64_t import_delay = NANOSEC;
+ hrtime_t import_expire;
+ nvlist_t *mmp_label = NULL;
+ vdev_t *rvd = spa->spa_root_vdev;
+ kcondvar_t cv;
+ kmutex_t mtx;
+ int error = 0;
+
+ cv_init(&cv, NULL, CV_DEFAULT, NULL);
+ mutex_init(&mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_enter(&mtx);
+
+ /*
+ * If ZPOOL_CONFIG_MMP_TXG is present an activity check was performed
+ * during the earlier tryimport. If the txg recorded there is 0 then
+ * the pool is known to be active on another host.
+ *
+ * Otherwise, the pool might be in use on another node. Check for
+ * changes in the uberblocks on disk if necessary.
+ */
+ if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
+ nvlist_t *nvinfo = fnvlist_lookup_nvlist(config,
+ ZPOOL_CONFIG_LOAD_INFO);
+
+ if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_TXG) &&
+ fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG) == 0) {
+ vdev_uberblock_load(rvd, ub, &mmp_label);
+ error = SET_ERROR(EREMOTEIO);
+ goto out;
+ }
+ }
+
+ /*
+ * Preferentially use the zfs_multihost_interval from the node which
+ * last imported the pool. This value is stored in an MMP uberblock as.
+ *
+ * ub_mmp_delay * vdev_count_leaves() == zfs_multihost_interval
+ */
+ if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay)
+ import_delay = MAX(import_delay, import_intervals *
+ ub->ub_mmp_delay * MAX(vdev_count_leaves(spa), 1));
+
+ /* Apply a floor using the local default values. */
+ import_delay = MAX(import_delay, import_intervals *
+ MSEC2NSEC(MAX(zfs_multihost_interval, MMP_MIN_INTERVAL)));
+
+ /* Add a small random factor in case of simultaneous imports (0-25%) */
+ import_expire = gethrtime() + import_delay +
+ (import_delay * spa_get_random(250) / 1000);
+
+ while (gethrtime() < import_expire) {
+ vdev_uberblock_load(rvd, ub, &mmp_label);
+
+ if (txg != ub->ub_txg || timestamp != ub->ub_timestamp) {
+ error = SET_ERROR(EREMOTEIO);
+ break;
+ }
+
+ if (mmp_label) {
+ nvlist_free(mmp_label);
+ mmp_label = NULL;
+ }
+
+ error = cv_timedwait_sig(&cv, &mtx, ddi_get_lbolt() + hz);
+ if (error != -1) {
+ error = SET_ERROR(EINTR);
+ break;
+ }
+ error = 0;
+ }
+
+out:
+ mutex_exit(&mtx);
+ mutex_destroy(&mtx);
+ cv_destroy(&cv);
+
+ /*
+ * If the pool is determined to be active store the status in the
+ * spa->spa_load_info nvlist. If the remote hostname or hostid are
+ * available from configuration read from disk store them as well.
+ * This allows 'zpool import' to generate a more useful message.
+ *
+ * ZPOOL_CONFIG_MMP_STATE - observed pool status (mandatory)
+ * ZPOOL_CONFIG_MMP_HOSTNAME - hostname from the active pool
+ * ZPOOL_CONFIG_MMP_HOSTID - hostid from the active pool
+ */
+ if (error == EREMOTEIO) {
+ char *hostname = "<unknown>";
+ uint64_t hostid = 0;
+
+ if (mmp_label) {
+ if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTNAME)) {
+ hostname = fnvlist_lookup_string(mmp_label,
+ ZPOOL_CONFIG_HOSTNAME);
+ fnvlist_add_string(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_HOSTNAME, hostname);
+ }
+
+ if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTID)) {
+ hostid = fnvlist_lookup_uint64(mmp_label,
+ ZPOOL_CONFIG_HOSTID);
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_HOSTID, hostid);
+ }
+ }
+
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_STATE, MMP_STATE_ACTIVE);
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_TXG, 0);
+
+ error = spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO);
+ }
+
+ if (mmp_label)
+ nvlist_free(mmp_label);
+
+ return (error);
+}
+
+/*
+ * Load an existing storage pool, using the pool's builtin spa_config as a
+ * source of configuration information.
+ */
+__attribute__((always_inline))
+static inline int
+spa_load_impl(spa_t *spa, uint64_t pool_guid, nvlist_t *config,
+ spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig,
+ char **ereport)
+{
+ int error = 0;
+ nvlist_t *nvroot = NULL;
+ nvlist_t *label;
+ vdev_t *rvd;
+ uberblock_t *ub = &spa->spa_uberblock;
+ uint64_t children, config_cache_txg = spa->spa_config_txg;
+ int orig_mode = spa->spa_mode;
+ int parse, i;
+ uint64_t obj;
+ boolean_t missing_feat_write = B_FALSE;
+ boolean_t activity_check = B_FALSE;
+ nvlist_t *mos_config;
+
+ /*
+ * If this is an untrusted config, access the pool in read-only mode.
+ * This prevents things like resilvering recently removed devices.
+ */
+ if (!mosconfig)
+ spa->spa_mode = FREAD;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
spa->spa_load_state = state;
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot))
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
parse = (type == SPA_IMPORT_EXISTING ?
VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT);
/*
* Create "The Godfather" zio to hold all async IOs
*/
- spa->spa_async_zio_root = zio_root(spa, NULL, NULL,
- ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER);
+ spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *),
+ KM_SLEEP);
+ for (i = 0; i < max_ncpus; i++) {
+ spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL,
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE |
+ ZIO_FLAG_GODFATHER);
+ }
/*
* Parse the configuration into a vdev tree. We explicitly set the
return (error);
ASSERT(spa->spa_root_vdev == rvd);
+ ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT);
+ ASSERT3U(spa->spa_max_ashift, <=, SPA_MAXBLOCKSHIFT);
if (type != SPA_IMPORT_ASSEMBLE) {
ASSERT(spa_guid(spa) == pool_guid);
return (error);
if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN)
- return (ENXIO);
+ return (SET_ERROR(ENXIO));
}
/*
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO));
}
+ /*
+ * For pools which have the multihost property on determine if the
+ * pool is truly inactive and can be safely imported. Prevent
+ * hosts which don't have a hostid set from importing the pool.
+ */
+ activity_check = spa_activity_check_required(spa, ub, label, config);
+ if (activity_check) {
+ if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay &&
+ spa_get_hostid() == 0) {
+ nvlist_free(label);
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
+ return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
+ }
+
+ error = spa_activity_check(spa, ub, config);
+ if (error) {
+ nvlist_free(label);
+ return (error);
+ }
+
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_STATE, MMP_STATE_INACTIVE);
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_TXG, ub->ub_txg);
+ }
+
/*
* If the pool has an unsupported version we can't open it.
*/
if (spa_version(spa) >= SPA_VERSION_FEATURES) {
boolean_t missing_feat_read = B_FALSE;
nvlist_t *unsup_feat, *enabled_feat;
+ spa_feature_t i;
if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_READ,
&spa->spa_feat_for_read_obj) != 0) {
enabled_feat = fnvlist_alloc();
unsup_feat = fnvlist_alloc();
- if (!feature_is_supported(spa->spa_meta_objset,
- spa->spa_feat_for_read_obj, spa->spa_feat_desc_obj,
+ if (!spa_features_check(spa, B_FALSE,
unsup_feat, enabled_feat))
missing_feat_read = B_TRUE;
if (spa_writeable(spa) || state == SPA_LOAD_TRYIMPORT) {
- if (!feature_is_supported(spa->spa_meta_objset,
- spa->spa_feat_for_write_obj, spa->spa_feat_desc_obj,
+ if (!spa_features_check(spa, B_TRUE,
unsup_feat, enabled_feat)) {
missing_feat_write = B_TRUE;
}
return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
ENOTSUP));
}
+
+ /*
+ * Load refcounts for ZFS features from disk into an in-memory
+ * cache during SPA initialization.
+ */
+ for (i = 0; i < SPA_FEATURES; i++) {
+ uint64_t refcount;
+
+ error = feature_get_refcount_from_disk(spa,
+ &spa_feature_table[i], &refcount);
+ if (error == 0) {
+ spa->spa_feat_refcount_cache[i] = refcount;
+ } else if (error == ENOTSUP) {
+ spa->spa_feat_refcount_cache[i] =
+ SPA_FEATURE_DISABLED;
+ } else {
+ return (spa_vdev_err(rvd,
+ VDEV_AUX_CORRUPT_DATA, EIO));
+ }
+ }
+ }
+
+ if (spa_feature_is_active(spa, SPA_FEATURE_ENABLED_TXG)) {
+ if (spa_dir_prop(spa, DMU_POOL_FEATURE_ENABLED_TXG,
+ &spa->spa_feat_enabled_txg_obj) != 0)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
}
spa->spa_is_initializing = B_TRUE;
VERIFY(nvlist_lookup_string(nvconfig,
ZPOOL_CONFIG_HOSTNAME, &hostname) == 0);
-#ifdef _KERNEL
- myhostid = zone_get_hostid(NULL);
-#else /* _KERNEL */
- /*
- * We're emulating the system's hostid in userland, so
- * we can't use zone_get_hostid().
- */
- (void) ddi_strtoul(hw_serial, NULL, 10, &myhostid);
-#endif /* _KERNEL */
- if (hostid != 0 && myhostid != 0 &&
- hostid != myhostid) {
+ myhostid = spa_get_hostid();
+ if (hostid && myhostid && hostid != myhostid) {
nvlist_free(nvconfig);
- cmn_err(CE_WARN, "pool '%s' could not be "
- "loaded as it was last accessed by "
- "another system (host: %s hostid: 0x%lx). "
- "See: http://zfsonlinux.org/msg/ZFS-8000-EY",
- spa_name(spa), hostname,
- (unsigned long)hostid);
- return (EBADF);
+ return (SET_ERROR(EBADF));
}
}
if (nvlist_lookup_nvlist(spa->spa_config,
return (spa_load(spa, state, SPA_IMPORT_EXISTING, B_TRUE));
}
+ /* Grab the checksum salt from the MOS. */
+ error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_CHECKSUM_SALT, 1,
+ sizeof (spa->spa_cksum_salt.zcs_bytes),
+ spa->spa_cksum_salt.zcs_bytes);
+ if (error == ENOENT) {
+ /* Generate a new salt for subsequent use */
+ (void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes,
+ sizeof (spa->spa_cksum_salt.zcs_bytes));
+ } else if (error != 0) {
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+ }
+
if (spa_dir_prop(spa, DMU_POOL_SYNC_BPOBJ, &obj) != 0)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
error = bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj);
if (error != 0 && error != ENOENT)
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+ /*
+ * Load the per-vdev ZAP map. If we have an older pool, this will not
+ * be present; in this case, defer its creation to a later time to
+ * avoid dirtying the MOS this early / out of sync context. See
+ * spa_sync_config_object.
+ */
+
+ /* The sentinel is only available in the MOS config. */
+ if (load_nvlist(spa, spa->spa_config_object, &mos_config) != 0)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+
+ error = spa_dir_prop(spa, DMU_POOL_VDEV_ZAP_MAP,
+ &spa->spa_all_vdev_zaps);
+
+ if (error == ENOENT) {
+ VERIFY(!nvlist_exists(mos_config,
+ ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
+ spa->spa_avz_action = AVZ_ACTION_INITIALIZE;
+ ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
+ } else if (error != 0) {
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+ } else if (!nvlist_exists(mos_config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) {
+ /*
+ * An older version of ZFS overwrote the sentinel value, so
+ * we have orphaned per-vdev ZAPs in the MOS. Defer their
+ * destruction to later; see spa_sync_config_object.
+ */
+ spa->spa_avz_action = AVZ_ACTION_DESTROY;
+ /*
+ * We're assuming that no vdevs have had their ZAPs created
+ * before this. Better be sure of it.
+ */
+ ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
+ }
+ nvlist_free(mos_config);
+
/*
* If we're assembling the pool from the split-off vdevs of
* an existing pool, we don't want to attach the spares & cache
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
if (error == 0) {
- uint64_t autoreplace;
+ uint64_t autoreplace = 0;
spa_prop_find(spa, ZPOOL_PROP_BOOTFS, &spa->spa_bootfs);
spa_prop_find(spa, ZPOOL_PROP_AUTOREPLACE, &autoreplace);
spa_prop_find(spa, ZPOOL_PROP_DELEGATION, &spa->spa_delegation);
spa_prop_find(spa, ZPOOL_PROP_FAILUREMODE, &spa->spa_failmode);
spa_prop_find(spa, ZPOOL_PROP_AUTOEXPAND, &spa->spa_autoexpand);
+ spa_prop_find(spa, ZPOOL_PROP_MULTIHOST, &spa->spa_multihost);
spa_prop_find(spa, ZPOOL_PROP_DEDUPDITTO,
&spa->spa_dedup_ditto);
spa->spa_autoreplace = (autoreplace != 0);
}
+ /*
+ * If the 'multihost' property is set, then never allow a pool to
+ * be imported when the system hostid is zero. The exception to
+ * this rule is zdb which is always allowed to access pools.
+ */
+ if (spa_multihost(spa) && spa_get_hostid() == 0 &&
+ (spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP) == 0) {
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
+ return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
+ }
+
/*
* If the 'autoreplace' property is set, then post a resource notifying
* the ZFS DE that it should not issue any faults for unopenable
* more toplevel vdevs are faulted.
*/
if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN)
- return (ENXIO);
+ return (SET_ERROR(ENXIO));
- if (spa_check_logs(spa)) {
+ if (spa_writeable(spa) && spa_check_logs(spa)) {
*ereport = FM_EREPORT_ZFS_LOG_REPLAY;
return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG, ENXIO));
}
spa->spa_load_max_txg == UINT64_MAX)) {
dmu_tx_t *tx;
int need_update = B_FALSE;
+ dsl_pool_t *dp = spa_get_dsl(spa);
int c;
ASSERT(state != SPA_LOAD_TRYIMPORT);
*/
spa->spa_claiming = B_TRUE;
- tx = dmu_tx_create_assigned(spa_get_dsl(spa),
- spa_first_txg(spa));
- (void) dmu_objset_find(spa_name(spa),
+ tx = dmu_tx_create_assigned(dp, spa_first_txg(spa));
+ (void) dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
zil_claim, tx, DS_FIND_CHILDREN);
dmu_tx_commit(tx);
spa_set_log_state(spa, SPA_LOG_GOOD);
spa->spa_sync_on = B_TRUE;
txg_sync_start(spa->spa_dsl_pool);
+ mmp_thread_start(spa);
/*
* Wait for all claims to sync. We sync up to the highest
vdev_resilver_needed(rvd, NULL, NULL))
spa_async_request(spa, SPA_ASYNC_RESILVER);
+ /*
+ * Log the fact that we booted up (so that we can detect if
+ * we rebooted in the middle of an operation).
+ */
+ spa_history_log_version(spa, "open", NULL);
+
/*
* Delete any inconsistent datasets.
*/
spa_unload(spa);
spa_deactivate(spa);
- spa->spa_load_max_txg--;
+ spa->spa_load_max_txg = spa->spa_uberblock.ub_txg - 1;
spa_activate(spa, mode);
spa_async_suspend(spa);
spa_set_log_state(spa, SPA_LOG_CLEAR);
} else {
spa->spa_load_max_txg = max_request;
+ if (max_request != UINT64_MAX)
+ spa->spa_extreme_rewind = B_TRUE;
}
load_error = rewind_error = spa_load(spa, state, SPA_IMPORT_EXISTING,
if (config && (rewind_error || state != SPA_LOAD_RECOVER))
spa_config_set(spa, config);
+ else
+ nvlist_free(config);
if (state == SPA_LOAD_RECOVER) {
ASSERT3P(loadinfo, ==, NULL);
spa_load_state_t state = SPA_LOAD_OPEN;
int error;
int locked = B_FALSE;
+ int firstopen = B_FALSE;
*spapp = NULL;
if ((spa = spa_lookup(pool)) == NULL) {
if (locked)
mutex_exit(&spa_namespace_lock);
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
}
if (spa->spa_state == POOL_STATE_UNINITIALIZED) {
zpool_rewind_policy_t policy;
+ firstopen = B_TRUE;
+
zpool_get_rewind_policy(nvpolicy ? nvpolicy : spa->spa_config,
&policy);
if (policy.zrp_request & ZPOOL_DO_REWIND)
spa_remove(spa);
if (locked)
mutex_exit(&spa_namespace_lock);
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
}
if (error) {
*/
if (config != NULL && spa->spa_config) {
VERIFY(nvlist_dup(spa->spa_config, config,
- KM_PUSHPAGE) == 0);
+ KM_SLEEP) == 0);
VERIFY(nvlist_add_nvlist(*config,
ZPOOL_CONFIG_LOAD_INFO,
spa->spa_load_info) == 0);
mutex_exit(&spa_namespace_lock);
}
+ if (firstopen)
+ zvol_create_minors(spa, spa_name(spa), B_TRUE);
+
*spapp = spa;
return (0);
ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
== 0);
vdev_get_stats(vd, vs);
+ vdev_config_generate_stats(vd, l2cache[i]);
+
}
}
}
static void
-spa_add_feature_stats(spa_t *spa, nvlist_t *config)
+spa_feature_stats_from_disk(spa_t *spa, nvlist_t *features)
{
- nvlist_t *features;
zap_cursor_t zc;
zap_attribute_t za;
- ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));
- VERIFY(nvlist_alloc(&features, NV_UNIQUE_NAME, KM_SLEEP) == 0);
-
if (spa->spa_feat_for_read_obj != 0) {
for (zap_cursor_init(&zc, spa->spa_meta_objset,
spa->spa_feat_for_read_obj);
zap_cursor_advance(&zc)) {
ASSERT(za.za_integer_length == sizeof (uint64_t) &&
za.za_num_integers == 1);
- VERIFY3U(0, ==, nvlist_add_uint64(features, za.za_name,
+ VERIFY0(nvlist_add_uint64(features, za.za_name,
za.za_first_integer));
}
zap_cursor_fini(&zc);
zap_cursor_advance(&zc)) {
ASSERT(za.za_integer_length == sizeof (uint64_t) &&
za.za_num_integers == 1);
- VERIFY3U(0, ==, nvlist_add_uint64(features, za.za_name,
+ VERIFY0(nvlist_add_uint64(features, za.za_name,
za.za_first_integer));
}
zap_cursor_fini(&zc);
}
+}
+
+static void
+spa_feature_stats_from_cache(spa_t *spa, nvlist_t *features)
+{
+ int i;
+
+ for (i = 0; i < SPA_FEATURES; i++) {
+ zfeature_info_t feature = spa_feature_table[i];
+ uint64_t refcount;
+
+ if (feature_get_refcount(spa, &feature, &refcount) != 0)
+ continue;
+
+ VERIFY0(nvlist_add_uint64(features, feature.fi_guid, refcount));
+ }
+}
+
+/*
+ * Store a list of pool features and their reference counts in the
+ * config.
+ *
+ * The first time this is called on a spa, allocate a new nvlist, fetch
+ * the pool features and reference counts from disk, then save the list
+ * in the spa. In subsequent calls on the same spa use the saved nvlist
+ * and refresh its values from the cached reference counts. This
+ * ensures we don't block here on I/O on a suspended pool so 'zpool
+ * clear' can resume the pool.
+ */
+static void
+spa_add_feature_stats(spa_t *spa, nvlist_t *config)
+{
+ nvlist_t *features;
+
+ ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));
+
+ mutex_enter(&spa->spa_feat_stats_lock);
+ features = spa->spa_feat_stats;
+
+ if (features != NULL) {
+ spa_feature_stats_from_cache(spa, features);
+ } else {
+ VERIFY0(nvlist_alloc(&features, NV_UNIQUE_NAME, KM_SLEEP));
+ spa->spa_feat_stats = features;
+ spa_feature_stats_from_disk(spa, features);
+ }
- VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS,
- features) == 0);
- nvlist_free(features);
+ VERIFY0(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS,
+ features));
+
+ mutex_exit(&spa->spa_feat_stats_lock);
}
int
return (0);
if (ndev == 0)
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
/*
* Make sure the pool is formatted with a version that supports this
* device type.
*/
if (spa_version(spa) < version)
- return (ENOTSUP);
+ return (SET_ERROR(ENOTSUP));
/*
* Set the pending device list so we correctly handle device in-use
if (!vd->vdev_ops->vdev_op_leaf) {
vdev_free(vd);
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
goto out;
}
- /*
- * The L2ARC currently only supports disk devices in
- * kernel context. For user-level testing, we allow it.
- */
-#ifdef _KERNEL
- if ((strcmp(config, ZPOOL_CONFIG_L2CACHE) == 0) &&
- strcmp(vd->vdev_ops->vdev_op_type, VDEV_TYPE_DISK) != 0) {
- error = ENOTBLK;
- vdev_free(vd);
- goto out;
- }
-#endif
vd->vdev_top = vd;
if ((error = vdev_open(vd)) == 0 &&
nvlist_t **newdevs;
/*
- * Generate new dev list by concatentating with the
+ * Generate new dev list by concatenating with the
* current dev list.
*/
VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, config,
&olddevs, &oldndevs) == 0);
newdevs = kmem_alloc(sizeof (void *) *
- (ndevs + oldndevs), KM_PUSHPAGE);
+ (ndevs + oldndevs), KM_SLEEP);
for (i = 0; i < oldndevs; i++)
VERIFY(nvlist_dup(olddevs[i], &newdevs[i],
- KM_PUSHPAGE) == 0);
+ KM_SLEEP) == 0);
for (i = 0; i < ndevs; i++)
VERIFY(nvlist_dup(devs[i], &newdevs[i + oldndevs],
- KM_PUSHPAGE) == 0);
+ KM_SLEEP) == 0);
VERIFY(nvlist_remove(sav->sav_config, config,
DATA_TYPE_NVLIST_ARRAY) == 0);
* Generate a new dev list.
*/
VERIFY(nvlist_alloc(&sav->sav_config, NV_UNIQUE_NAME,
- KM_PUSHPAGE) == 0);
+ KM_SLEEP) == 0);
VERIFY(nvlist_add_nvlist_array(sav->sav_config, config,
devs, ndevs) == 0);
}
*/
int
spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
- const char *history_str, nvlist_t *zplprops)
+ nvlist_t *zplprops)
{
spa_t *spa;
char *altroot = NULL;
uint64_t version, obj;
boolean_t has_features;
nvpair_t *elem;
- int c;
+ int c, i;
+ char *poolname;
+ nvlist_t *nvl;
+
+ if (nvlist_lookup_string(props, "tname", &poolname) != 0)
+ poolname = (char *)pool;
/*
* If this pool already exists, return failure.
*/
mutex_enter(&spa_namespace_lock);
- if (spa_lookup(pool) != NULL) {
+ if (spa_lookup(poolname) != NULL) {
mutex_exit(&spa_namespace_lock);
- return (EEXIST);
+ return (SET_ERROR(EEXIST));
}
/*
* Allocate a new spa_t structure.
*/
+ nvl = fnvlist_alloc();
+ fnvlist_add_string(nvl, ZPOOL_CONFIG_POOL_NAME, pool);
(void) nvlist_lookup_string(props,
zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
- spa = spa_add(pool, NULL, altroot);
+ spa = spa_add(poolname, nvl, altroot);
+ fnvlist_free(nvl);
spa_activate(spa, spa_mode_global);
if (props && (error = spa_prop_validate(spa, props))) {
return (error);
}
+ /*
+ * Temporary pool names should never be written to disk.
+ */
+ if (poolname != pool)
+ spa->spa_import_flags |= ZFS_IMPORT_TEMP_NAME;
+
has_features = B_FALSE;
for (elem = nvlist_next_nvpair(props, NULL);
elem != NULL; elem = nvlist_next_nvpair(props, elem)) {
spa->spa_uberblock.ub_txg = txg - 1;
spa->spa_uberblock.ub_version = version;
spa->spa_ubsync = spa->spa_uberblock;
+ spa->spa_load_state = SPA_LOAD_CREATE;
/*
* Create "The Godfather" zio to hold all async IOs
*/
- spa->spa_async_zio_root = zio_root(spa, NULL, NULL,
- ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER);
+ spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *),
+ KM_SLEEP);
+ for (i = 0; i < max_ncpus; i++) {
+ spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL,
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE |
+ ZIO_FLAG_GODFATHER);
+ }
/*
* Create the root vdev.
ASSERT(error != 0 || spa->spa_root_vdev == rvd);
if (error == 0 && !zfs_allocatable_devs(nvroot))
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
if (error == 0 &&
(error = vdev_create(rvd, txg, B_FALSE)) == 0 &&
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
&spares, &nspares) == 0) {
VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, NV_UNIQUE_NAME,
- KM_PUSHPAGE) == 0);
+ KM_SLEEP) == 0);
VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, spares, nspares) == 0);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
&l2cache, &nl2cache) == 0) {
VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config,
- NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
tx = dmu_tx_create_assigned(dp, txg);
+ /*
+ * Create the pool's history object.
+ */
+ if (version >= SPA_VERSION_ZPOOL_HISTORY && !spa->spa_history)
+ spa_history_create_obj(spa, tx);
+
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_CREATE);
+ spa_history_log_version(spa, "create", tx);
+
/*
* Create the pool config object.
*/
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION,
sizeof (uint64_t), 1, &version, tx) != 0) {
cmn_err(CE_PANIC, "failed to add pool version");
- }
-
- /* Newly created pools with the right version are always deflated. */
- if (version >= SPA_VERSION_RAIDZ_DEFLATE) {
- spa->spa_deflate = TRUE;
- if (zap_add(spa->spa_meta_objset,
- DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
- sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) {
- cmn_err(CE_PANIC, "failed to add deflate");
- }
- }
-
- /*
- * Create the deferred-free bpobj. Turn off compression
- * because sync-to-convergence takes longer if the blocksize
- * keeps changing.
- */
- obj = bpobj_alloc(spa->spa_meta_objset, 1 << 14, tx);
- dmu_object_set_compress(spa->spa_meta_objset, obj,
- ZIO_COMPRESS_OFF, tx);
- if (zap_add(spa->spa_meta_objset,
- DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPOBJ,
- sizeof (uint64_t), 1, &obj, tx) != 0) {
- cmn_err(CE_PANIC, "failed to add bpobj");
- }
- VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj,
- spa->spa_meta_objset, obj));
-
- /*
- * Create the pool's history object.
- */
- if (version >= SPA_VERSION_ZPOOL_HISTORY)
- spa_history_create_obj(spa, tx);
-
- /*
- * Set pool properties.
- */
- spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS);
- spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
- spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE);
- spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND);
-
- if (props != NULL) {
- spa_configfile_set(spa, props, B_FALSE);
- spa_sync_props(spa, props, tx);
- }
-
- dmu_tx_commit(tx);
-
- spa->spa_sync_on = B_TRUE;
- txg_sync_start(spa->spa_dsl_pool);
-
- /*
- * We explicitly wait for the first transaction to complete so that our
- * bean counters are appropriately updated.
- */
- txg_wait_synced(spa->spa_dsl_pool, txg);
-
- spa_config_sync(spa, B_FALSE, B_TRUE);
-
- if (version >= SPA_VERSION_ZPOOL_HISTORY && history_str != NULL)
- (void) spa_history_log(spa, history_str, LOG_CMD_POOL_CREATE);
- spa_history_log_version(spa, LOG_POOL_CREATE);
-
- spa->spa_minref = refcount_count(&spa->spa_refcount);
-
- mutex_exit(&spa_namespace_lock);
-
- return (0);
-}
-
-#ifdef _KERNEL
-/*
- * Get the root pool information from the root disk, then import the root pool
- * during the system boot up time.
- */
-extern int vdev_disk_read_rootlabel(char *, char *, nvlist_t **);
-
-static nvlist_t *
-spa_generate_rootconf(char *devpath, char *devid, uint64_t *guid)
-{
- nvlist_t *config;
- nvlist_t *nvtop, *nvroot;
- uint64_t pgid;
-
- if (vdev_disk_read_rootlabel(devpath, devid, &config) != 0)
- return (NULL);
-
- /*
- * Add this top-level vdev to the child array.
- */
- VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
- &nvtop) == 0);
- VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
- &pgid) == 0);
- VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, guid) == 0);
-
- /*
- * Put this pool's top-level vdevs into a root vdev.
- */
- VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
- VERIFY(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
- VDEV_TYPE_ROOT) == 0);
- VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) == 0);
- VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, pgid) == 0);
- VERIFY(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
- &nvtop, 1) == 0);
-
- /*
- * Replace the existing vdev_tree with the new root vdev in
- * this pool's configuration (remove the old, add the new).
- */
- VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
- nvlist_free(nvroot);
- return (config);
-}
-
-/*
- * Walk the vdev tree and see if we can find a device with "better"
- * configuration. A configuration is "better" if the label on that
- * device has a more recent txg.
- */
-static void
-spa_alt_rootvdev(vdev_t *vd, vdev_t **avd, uint64_t *txg)
-{
- int c;
-
- for (c = 0; c < vd->vdev_children; c++)
- spa_alt_rootvdev(vd->vdev_child[c], avd, txg);
-
- if (vd->vdev_ops->vdev_op_leaf) {
- nvlist_t *label;
- uint64_t label_txg;
-
- if (vdev_disk_read_rootlabel(vd->vdev_physpath, vd->vdev_devid,
- &label) != 0)
- return;
-
- VERIFY(nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG,
- &label_txg) == 0);
-
- /*
- * Do we have a better boot device?
- */
- if (label_txg > *txg) {
- *txg = label_txg;
- *avd = vd;
- }
- nvlist_free(label);
- }
-}
-
-/*
- * Import a root pool.
- *
- * For x86. devpath_list will consist of devid and/or physpath name of
- * the vdev (e.g. "id1,sd@SSEAGATE..." or "/pci@1f,0/ide@d/disk@0,0:a").
- * The GRUB "findroot" command will return the vdev we should boot.
- *
- * For Sparc, devpath_list consists the physpath name of the booting device
- * no matter the rootpool is a single device pool or a mirrored pool.
- * e.g.
- * "/pci@1f,0/ide@d/disk@0,0:a"
- */
-int
-spa_import_rootpool(char *devpath, char *devid)
-{
- spa_t *spa;
- vdev_t *rvd, *bvd, *avd = NULL;
- nvlist_t *config, *nvtop;
- uint64_t guid, txg;
- char *pname;
- int error;
+ }
- /*
- * Read the label from the boot device and generate a configuration.
- */
- config = spa_generate_rootconf(devpath, devid, &guid);
-#if defined(_OBP) && defined(_KERNEL)
- if (config == NULL) {
- if (strstr(devpath, "/iscsi/ssd") != NULL) {
- /* iscsi boot */
- get_iscsi_bootpath_phy(devpath);
- config = spa_generate_rootconf(devpath, devid, &guid);
+ /* Newly created pools with the right version are always deflated. */
+ if (version >= SPA_VERSION_RAIDZ_DEFLATE) {
+ spa->spa_deflate = TRUE;
+ if (zap_add(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
+ sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) {
+ cmn_err(CE_PANIC, "failed to add deflate");
}
}
-#endif
- if (config == NULL) {
- cmn_err(CE_NOTE, "Cannot read the pool label from '%s'",
- devpath);
- return (EIO);
- }
- VERIFY(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
- &pname) == 0);
- VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, &txg) == 0);
-
- mutex_enter(&spa_namespace_lock);
- if ((spa = spa_lookup(pname)) != NULL) {
- /*
- * Remove the existing root pool from the namespace so that we
- * can replace it with the correct config we just read in.
- */
- spa_remove(spa);
+ /*
+ * Create the deferred-free bpobj. Turn off compression
+ * because sync-to-convergence takes longer if the blocksize
+ * keeps changing.
+ */
+ obj = bpobj_alloc(spa->spa_meta_objset, 1 << 14, tx);
+ dmu_object_set_compress(spa->spa_meta_objset, obj,
+ ZIO_COMPRESS_OFF, tx);
+ if (zap_add(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPOBJ,
+ sizeof (uint64_t), 1, &obj, tx) != 0) {
+ cmn_err(CE_PANIC, "failed to add bpobj");
}
-
- spa = spa_add(pname, config, NULL);
- spa->spa_is_root = B_TRUE;
- spa->spa_import_flags = ZFS_IMPORT_VERBATIM;
+ VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj,
+ spa->spa_meta_objset, obj));
/*
- * Build up a vdev tree based on the boot device's label config.
+ * Generate some random noise for salted checksums to operate on.
*/
- VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
- &nvtop) == 0);
- spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
- error = spa_config_parse(spa, &rvd, nvtop, NULL, 0,
- VDEV_ALLOC_ROOTPOOL);
- spa_config_exit(spa, SCL_ALL, FTAG);
- if (error) {
- mutex_exit(&spa_namespace_lock);
- nvlist_free(config);
- cmn_err(CE_NOTE, "Can not parse the config for pool '%s'",
- pname);
- return (error);
- }
+ (void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes,
+ sizeof (spa->spa_cksum_salt.zcs_bytes));
/*
- * Get the boot vdev.
+ * Set pool properties.
*/
- if ((bvd = vdev_lookup_by_guid(rvd, guid)) == NULL) {
- cmn_err(CE_NOTE, "Can not find the boot vdev for guid %llu",
- (u_longlong_t)guid);
- error = ENOENT;
- goto out;
+ spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS);
+ spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
+ spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE);
+ spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND);
+ spa->spa_multihost = zpool_prop_default_numeric(ZPOOL_PROP_MULTIHOST);
+
+ if (props != NULL) {
+ spa_configfile_set(spa, props, B_FALSE);
+ spa_sync_props(props, tx);
}
+ dmu_tx_commit(tx);
+
+ spa->spa_sync_on = B_TRUE;
+ txg_sync_start(spa->spa_dsl_pool);
+ mmp_thread_start(spa);
+
/*
- * Determine if there is a better boot device.
+ * We explicitly wait for the first transaction to complete so that our
+ * bean counters are appropriately updated.
*/
- avd = bvd;
- spa_alt_rootvdev(rvd, &avd, &txg);
- if (avd != bvd) {
- cmn_err(CE_NOTE, "The boot device is 'degraded'. Please "
- "try booting from '%s'", avd->vdev_path);
- error = EINVAL;
- goto out;
- }
+ txg_wait_synced(spa->spa_dsl_pool, txg);
+
+ spa_config_sync(spa, B_FALSE, B_TRUE);
/*
- * If the boot device is part of a spare vdev then ensure that
- * we're booting off the active spare.
+ * Don't count references from objsets that are already closed
+ * and are making their way through the eviction process.
*/
- if (bvd->vdev_parent->vdev_ops == &vdev_spare_ops &&
- !bvd->vdev_isspare) {
- cmn_err(CE_NOTE, "The boot device is currently spared. Please "
- "try booting from '%s'",
- bvd->vdev_parent->
- vdev_child[bvd->vdev_parent->vdev_children - 1]->vdev_path);
- error = EINVAL;
- goto out;
- }
+ spa_evicting_os_wait(spa);
+ spa->spa_minref = refcount_count(&spa->spa_refcount);
+ spa->spa_load_state = SPA_LOAD_NONE;
- error = 0;
- spa_history_log_version(spa, LOG_POOL_IMPORT);
-out:
- spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
- vdev_free(rvd);
- spa_config_exit(spa, SCL_ALL, FTAG);
mutex_exit(&spa_namespace_lock);
- nvlist_free(config);
- return (error);
+ return (0);
}
-#endif
-
/*
* Import a non-root pool into the system.
*/
int
-spa_import(const char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags)
+spa_import(char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags)
{
spa_t *spa;
char *altroot = NULL;
mutex_enter(&spa_namespace_lock);
if (spa_lookup(pool) != NULL) {
mutex_exit(&spa_namespace_lock);
- return (EEXIST);
+ return (SET_ERROR(EEXIST));
}
/*
spa_configfile_set(spa, props, B_FALSE);
spa_config_sync(spa, B_FALSE, B_TRUE);
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);
mutex_exit(&spa_namespace_lock);
- spa_history_log_version(spa, LOG_POOL_IMPORT);
-
return (0);
}
VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0);
- if (error == 0)
- error = spa_validate_aux(spa, nvroot, -1ULL,
- VDEV_ALLOC_SPARE);
- if (error == 0)
- error = spa_validate_aux(spa, nvroot, -1ULL,
- VDEV_ALLOC_L2CACHE);
spa_config_exit(spa, SCL_ALL, FTAG);
if (props != NULL)
ZPOOL_CONFIG_SPARES, DATA_TYPE_NVLIST_ARRAY) == 0);
else
VERIFY(nvlist_alloc(&spa->spa_spares.sav_config,
- NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, spares, nspares) == 0);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
ZPOOL_CONFIG_L2CACHE, DATA_TYPE_NVLIST_ARRAY) == 0);
else
VERIFY(nvlist_alloc(&spa->spa_l2cache.sav_config,
- NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache) == 0);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
*/
spa_async_request(spa, SPA_ASYNC_AUTOEXPAND);
+ spa_history_log_version(spa, "import", NULL);
+
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);
+
+ zvol_create_minors(spa, pool, B_TRUE);
+
mutex_exit(&spa_namespace_lock);
- spa_history_log_version(spa, LOG_POOL_IMPORT);
return (0);
}
spa->spa_uberblock.ub_timestamp) == 0);
VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO,
spa->spa_load_info) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_ERRATA,
+ spa->spa_errata) == 0);
/*
* If the bootfs property exists on this pool then we
* pools are bootable.
*/
if ((!error || error == EEXIST) && spa->spa_bootfs) {
- char *tmpname = kmem_alloc(MAXPATHLEN, KM_PUSHPAGE);
+ char *tmpname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
/*
* We have to play games with the name since the
if (dsl_dsobj_to_dsname(spa_name(spa),
spa->spa_bootfs, tmpname) == 0) {
char *cp;
- char *dsname = kmem_alloc(MAXPATHLEN, KM_PUSHPAGE);
+ char *dsname;
+
+ dsname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
cp = strchr(tmpname, '/');
if (cp == NULL) {
*oldconfig = NULL;
if (!(spa_mode_global & FWRITE))
- return (EROFS);
+ return (SET_ERROR(EROFS));
mutex_enter(&spa_namespace_lock);
if ((spa = spa_lookup(pool)) == NULL) {
mutex_exit(&spa_namespace_lock);
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
}
/*
spa_open_ref(spa, FTAG);
mutex_exit(&spa_namespace_lock);
spa_async_suspend(spa);
+ if (spa->spa_zvol_taskq) {
+ zvol_remove_minors(spa, spa_name(spa), B_TRUE);
+ taskq_wait(spa->spa_zvol_taskq);
+ }
mutex_enter(&spa_namespace_lock);
spa_close(spa, FTAG);
+ if (spa->spa_state == POOL_STATE_UNINITIALIZED)
+ goto export_spa;
/*
- * The pool will be in core if it's openable,
- * in which case we can modify its state.
+ * The pool will be in core if it's openable, in which case we can
+ * modify its state. Objsets may be open only because they're dirty,
+ * so we have to force it to sync before checking spa_refcnt.
*/
- if (spa->spa_state != POOL_STATE_UNINITIALIZED && spa->spa_sync_on) {
- /*
- * Objsets may be open only because they're dirty, so we
- * have to force it to sync before checking spa_refcnt.
- */
+ if (spa->spa_sync_on) {
txg_wait_synced(spa->spa_dsl_pool, 0);
+ spa_evicting_os_wait(spa);
+ }
- /*
- * A pool cannot be exported or destroyed if there are active
- * references. If we are resetting a pool, allow references by
- * fault injection handlers.
- */
- if (!spa_refcount_zero(spa) ||
- (spa->spa_inject_ref != 0 &&
- new_state != POOL_STATE_UNINITIALIZED)) {
- spa_async_resume(spa);
- mutex_exit(&spa_namespace_lock);
- return (EBUSY);
- }
+ /*
+ * A pool cannot be exported or destroyed if there are active
+ * references. If we are resetting a pool, allow references by
+ * fault injection handlers.
+ */
+ if (!spa_refcount_zero(spa) ||
+ (spa->spa_inject_ref != 0 &&
+ new_state != POOL_STATE_UNINITIALIZED)) {
+ spa_async_resume(spa);
+ mutex_exit(&spa_namespace_lock);
+ return (SET_ERROR(EBUSY));
+ }
+ if (spa->spa_sync_on) {
/*
* A pool cannot be exported if it has an active shared spare.
* This is to prevent other pools stealing the active spare
spa_has_active_shared_spare(spa)) {
spa_async_resume(spa);
mutex_exit(&spa_namespace_lock);
- return (EXDEV);
+ return (SET_ERROR(EXDEV));
}
/*
}
}
- spa_event_notify(spa, NULL, FM_EREPORT_ZFS_POOL_DESTROY);
+export_spa:
+ if (new_state == POOL_STATE_DESTROYED)
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_DESTROY);
+ else if (new_state == POOL_STATE_EXPORTED)
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_EXPORT);
if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
spa_unload(spa);
mutex_enter(&spa_namespace_lock);
spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_VDEV_ADD);
mutex_exit(&spa_namespace_lock);
return (0);
spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing)
{
uint64_t txg, dtl_max_txg;
- ASSERTV(vdev_t *rvd = spa->spa_root_vdev;)
vdev_t *oldvd, *newvd, *newrootvd, *pvd, *tvd;
vdev_ops_t *pvops;
char *oldvdpath, *newvdpath;
int newvd_isspare;
int error;
+ ASSERTV(vdev_t *rvd = spa->spa_root_vdev);
ASSERT(spa_writeable(spa));
if (strcmp(oldvd->vdev_path, newvd->vdev_path) == 0) {
spa_strfree(oldvd->vdev_path);
oldvd->vdev_path = kmem_alloc(strlen(newvd->vdev_path) + 5,
- KM_PUSHPAGE);
+ KM_SLEEP);
(void) sprintf(oldvd->vdev_path, "%s/%s",
newvd->vdev_path, "old");
if (oldvd->vdev_devid != NULL) {
}
/* mark the device being resilvered */
- newvd->vdev_resilvering = B_TRUE;
+ newvd->vdev_resilver_txg = txg;
/*
* If the parent is not a mirror, or if we're replacing, insert the new
newvd->vdev_crtxg = oldvd->vdev_crtxg;
vdev_add_child(pvd, newvd);
+ /*
+ * Reevaluate the parent vdev state.
+ */
+ vdev_propagate_state(pvd);
+
tvd = newvd->vdev_top;
ASSERT(pvd->vdev_top == tvd);
ASSERT(tvd->vdev_parent == rvd);
if (newvd->vdev_isspare) {
spa_spare_activate(newvd);
- spa_event_notify(spa, newvd, FM_EREPORT_ZFS_DEVICE_SPARE);
+ spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_SPARE);
}
oldvdpath = spa_strdup(oldvd->vdev_path);
vdev_dirty(tvd, VDD_DTL, newvd, txg);
/*
- * Restart the resilver
+ * Schedule the resilver to restart in the future. We do this to
+ * ensure that dmu_sync-ed blocks have been stitched into the
+ * respective datasets.
*/
dsl_resilver_restart(spa->spa_dsl_pool, dtl_max_txg);
+ if (spa->spa_bootfs)
+ spa_event_notify(spa, newvd, NULL, ESC_ZFS_BOOTFS_VDEV_ATTACH);
+
+ spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_ATTACH);
+
/*
* Commit the config
*/
(void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0);
- spa_history_log_internal(LOG_POOL_VDEV_ATTACH, spa, NULL,
+ spa_history_log_internal(spa, "vdev attach", NULL,
"%s vdev=%s %s vdev=%s",
replacing && newvd_isspare ? "spare in" :
replacing ? "replace" : "attach", newvdpath,
spa_strfree(oldvdpath);
spa_strfree(newvdpath);
- if (spa->spa_bootfs)
- spa_event_notify(spa, newvd, FM_EREPORT_ZFS_BOOTFS_VDEV_ATTACH);
-
return (0);
}
/*
* Detach a device from a mirror or replacing vdev.
+ *
* If 'replace_done' is specified, only detach if the parent
* is a replacing vdev.
*/
{
uint64_t txg;
int error;
- ASSERTV(vdev_t *rvd = spa->spa_root_vdev;)
vdev_t *vd, *pvd, *cvd, *tvd;
boolean_t unspare = B_FALSE;
uint64_t unspare_guid = 0;
char *vdpath;
int c, t;
-
+ ASSERTV(vdev_t *rvd = spa->spa_root_vdev);
ASSERT(spa_writeable(spa));
txg = spa_vdev_enter(spa);
if (pvd->vdev_ops == &vdev_spare_ops)
cvd->vdev_unspare = B_FALSE;
vdev_remove_parent(cvd);
- cvd->vdev_resilvering = B_FALSE;
}
* But first make sure we're not on any *other* txg's DTL list, to
* prevent vd from being accessed after it's freed.
*/
- vdpath = spa_strdup(vd->vdev_path);
+ vdpath = spa_strdup(vd->vdev_path ? vd->vdev_path : "none");
for (t = 0; t < TXG_SIZE; t++)
(void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t);
vd->vdev_detached = B_TRUE;
vdev_dirty(tvd, VDD_DTL, vd, txg);
- spa_event_notify(spa, vd, FM_EREPORT_ZFS_DEVICE_REMOVE);
+ spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE);
/* hang on to the spa before we release the lock */
spa_open_ref(spa, FTAG);
error = spa_vdev_exit(spa, vd, txg, 0);
- spa_history_log_internal(LOG_POOL_VDEV_DETACH, spa, NULL,
+ spa_history_log_internal(spa, "detach", NULL,
"vdev=%s", vdpath);
spa_strfree(vdpath);
nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0)
return (spa_vdev_exit(spa, NULL, txg, EINVAL));
- vml = kmem_zalloc(children * sizeof (vdev_t *), KM_PUSHPAGE);
- glist = kmem_zalloc(children * sizeof (uint64_t), KM_PUSHPAGE);
+ vml = kmem_zalloc(children * sizeof (vdev_t *), KM_SLEEP);
+ glist = kmem_zalloc(children * sizeof (uint64_t), KM_SLEEP);
/* then, loop over each vdev and validate it */
for (c = 0; c < children; c++) {
spa->spa_root_vdev->vdev_child[c]->vdev_islog) {
continue;
} else {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
}
/* which disk is going to be split? */
if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID,
&glist[c]) != 0) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
/* look it up in the spa */
vml[c] = spa_lookup_by_guid(spa, glist[c], B_FALSE);
if (vml[c] == NULL) {
- error = ENODEV;
+ error = SET_ERROR(ENODEV);
break;
}
vml[c]->vdev_children != 0 ||
vml[c]->vdev_state != VDEV_STATE_HEALTHY ||
c != spa->spa_root_vdev->vdev_child[c]->vdev_id) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
if (vdev_dtl_required(vml[c])) {
- error = EBUSY;
+ error = SET_ERROR(EBUSY);
break;
}
vml[c]->vdev_top->vdev_asize) == 0);
VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT,
vml[c]->vdev_top->vdev_ashift) == 0);
+
+ /* transfer per-vdev ZAPs */
+ ASSERT3U(vml[c]->vdev_leaf_zap, !=, 0);
+ VERIFY0(nvlist_add_uint64(child[c],
+ ZPOOL_CONFIG_VDEV_LEAF_ZAP, vml[c]->vdev_leaf_zap));
+
+ ASSERT3U(vml[c]->vdev_top->vdev_top_zap, !=, 0);
+ VERIFY0(nvlist_add_uint64(child[c],
+ ZPOOL_CONFIG_VDEV_TOP_ZAP,
+ vml[c]->vdev_parent->vdev_top_zap));
}
if (error != 0) {
* Temporarily record the splitting vdevs in the spa config. This
* will disappear once the config is regenerated.
*/
- VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+ VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST,
glist, children) == 0);
kmem_free(glist, children * sizeof (uint64_t));
spa->spa_config_txg) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
spa_generate_guid(NULL)) == 0);
+ VERIFY0(nvlist_add_boolean(config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
(void) nvlist_lookup_string(props,
zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
/* add the new pool to the namespace */
newspa = spa_add(newname, config, altroot);
+ newspa->spa_avz_action = AVZ_ACTION_REBUILD;
newspa->spa_config_txg = spa->spa_config_txg;
spa_set_log_state(newspa, SPA_LOG_CLEAR);
/* if that worked, generate a real config for the new pool */
if (newspa->spa_root_vdev != NULL) {
VERIFY(nvlist_alloc(&newspa->spa_config_splitting,
- NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+ NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(newspa->spa_config_splitting,
ZPOOL_CONFIG_SPLIT_GUID, spa_guid(spa)) == 0);
spa_config_set(newspa, spa_config_generate(newspa, NULL, -1ULL,
if (vml[c] != NULL) {
vdev_split(vml[c]);
if (error == 0)
- spa_history_log_internal(LOG_POOL_VDEV_DETACH,
- spa, tx, "vdev=%s",
- vml[c]->vdev_path);
+ spa_history_log_internal(spa, "detach", tx,
+ "vdev=%s", vml[c]->vdev_path);
+
vdev_free(vml[c]);
}
}
+ spa->spa_avz_action = AVZ_ACTION_REBUILD;
vdev_config_dirty(spa->spa_root_vdev);
spa->spa_config_splitting = NULL;
nvlist_free(nvl);
zio_handle_panic_injection(spa, FTAG, 3);
/* split is complete; log a history record */
- spa_history_log_internal(LOG_POOL_SPLIT, newspa, NULL,
- "split new pool %s from pool %s", newname, spa_name(spa));
+ spa_history_log_internal(newspa, "split", NULL,
+ "from pool %s", spa_name(spa));
kmem_free(vml, children * sizeof (vdev_t *));
static void
spa_vdev_remove_aux(nvlist_t *config, char *name, nvlist_t **dev, int count,
- nvlist_t *dev_to_remove)
+ nvlist_t *dev_to_remove)
{
nvlist_t **newdev = NULL;
int i, j;
if (count > 1)
- newdev = kmem_alloc((count - 1) * sizeof (void *), KM_PUSHPAGE);
+ newdev = kmem_alloc((count - 1) * sizeof (void *), KM_SLEEP);
for (i = 0, j = 0; i < count; i++) {
if (dev[i] == dev_to_remove)
continue;
- VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_PUSHPAGE) == 0);
+ VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_SLEEP) == 0);
}
VERIFY(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY) == 0);
if (vd->vdev_stat.vs_alloc != 0)
error = spa_offline_log(spa);
} else {
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
}
if (error)
* The evacuation succeeded. Remove any remaining MOS metadata
* associated with this vdev, and wait for these changes to sync.
*/
- ASSERT3U(vd->vdev_stat.vs_alloc, ==, 0);
+ ASSERT0(vd->vdev_stat.vs_alloc);
txg = spa_vdev_config_enter(spa);
vd->vdev_removing = B_TRUE;
- vdev_dirty(vd, 0, NULL, txg);
+ vdev_dirty_leaves(vd, VDD_DTL, txg);
vdev_config_dirty(vd);
spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);
* the spa_vdev_config_[enter/exit] functions which allow us to
* grab and release the spa_config_lock while still holding the namespace
* lock. During each step the configuration is synced out.
- */
-
-/*
- * Remove a device from the pool. Currently, this supports removing only hot
- * spares, slogs, and level 2 ARC devices.
+ *
+ * Currently, this supports removing only hot spares, slogs, and level 2 ARC
+ * devices.
*/
int
spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare)
{
vdev_t *vd;
+ sysevent_t *ev = NULL;
metaslab_group_t *mg;
nvlist_t **spares, **l2cache, *nv;
uint64_t txg = 0;
* in this pool.
*/
if (vd == NULL || unspare) {
+ if (vd == NULL)
+ vd = spa_lookup_by_guid(spa, guid, B_TRUE);
+ ev = spa_event_create(spa, vd, NULL,
+ ESC_ZFS_VDEV_REMOVE_AUX);
spa_vdev_remove_aux(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, spares, nspares, nv);
spa_load_spares(spa);
spa->spa_spares.sav_sync = B_TRUE;
} else {
- error = EBUSY;
+ error = SET_ERROR(EBUSY);
}
} else if (spa->spa_l2cache.sav_vdevs != NULL &&
nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
/*
* Cache devices can always be removed.
*/
+ vd = spa_lookup_by_guid(spa, guid, B_TRUE);
+ ev = spa_event_create(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE_AUX);
spa_vdev_remove_aux(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv);
spa_load_l2cache(spa);
ASSERT(!locked);
ASSERT(vd == vd->vdev_top);
- /*
- * XXX - Once we have bp-rewrite this should
- * become the common case.
- */
-
mg = vd->vdev_mg;
/*
/*
* Clean up the vdev namespace.
*/
+ ev = spa_event_create(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE_DEV);
spa_vdev_remove_from_namespace(spa, vd);
} else if (vd != NULL) {
/*
* Normal vdevs cannot be removed (yet).
*/
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
} else {
/*
* There is no vdev of any kind with the specified guid.
*/
- error = ENOENT;
+ error = SET_ERROR(ENOENT);
}
if (!locked)
- return (spa_vdev_exit(spa, NULL, txg, error));
+ error = spa_vdev_exit(spa, NULL, txg, error);
+
+ if (ev)
+ spa_event_post(ev);
return (error);
}
/*
* Find any device that's done replacing, or a vdev marked 'unspare' that's
- * current spared, so we can detach it.
+ * currently spared, so we can detach it.
*/
static vdev_t *
spa_vdev_resilver_done_hunt(vdev_t *vd)
ASSERT(pvd->vdev_ops == &vdev_replacing_ops);
sguid = ppvd->vdev_child[1]->vdev_guid;
}
+ ASSERT(vd->vdev_resilver_txg == 0 || !vdev_dtl_required(vd));
+
spa_config_exit(spa, SCL_ALL, FTAG);
if (spa_vdev_detach(spa, guid, pguid, B_TRUE) != 0)
return;
* SPA Scanning
* ==========================================================================
*/
+int
+spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t cmd)
+{
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
+
+ if (dsl_scan_resilvering(spa->spa_dsl_pool))
+ return (SET_ERROR(EBUSY));
+
+ return (dsl_scrub_set_pause_resume(spa->spa_dsl_pool, cmd));
+}
int
spa_scan_stop(spa_t *spa)
{
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
if (dsl_scan_resilvering(spa->spa_dsl_pool))
- return (EBUSY);
+ return (SET_ERROR(EBUSY));
return (dsl_scan_cancel(spa->spa_dsl_pool));
}
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
if (func >= POOL_SCAN_FUNCS || func == POOL_SCAN_NONE)
- return (ENOTSUP);
+ return (SET_ERROR(ENOTSUP));
/*
* If a resilver was requested, but there is no DTL on a
if (!vd->vdev_ops->vdev_op_leaf || vd->vdev_physpath == NULL)
return;
- spa_event_notify(vd->vdev_spa, vd, FM_EREPORT_ZFS_DEVICE_AUTOEXPAND);
+ spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_AUTOEXPAND);
}
static void
* then log an internal history event.
*/
if (new_space != old_space) {
- spa_history_log_internal(LOG_POOL_VDEV_ONLINE,
- spa, NULL,
+ spa_history_log_internal(spa, "vdev online", NULL,
"pool '%s' size: %llu(+%llu)",
spa_name(spa), new_space, new_space - old_space);
}
mutex_exit(&spa->spa_async_lock);
}
+static boolean_t
+spa_async_tasks_pending(spa_t *spa)
+{
+ uint_t non_config_tasks;
+ uint_t config_task;
+ boolean_t config_task_suspended;
+
+ non_config_tasks = spa->spa_async_tasks & ~SPA_ASYNC_CONFIG_UPDATE;
+ config_task = spa->spa_async_tasks & SPA_ASYNC_CONFIG_UPDATE;
+ if (spa->spa_ccw_fail_time == 0) {
+ config_task_suspended = B_FALSE;
+ } else {
+ config_task_suspended =
+ (gethrtime() - spa->spa_ccw_fail_time) <
+ ((hrtime_t)zfs_ccw_retry_interval * NANOSEC);
+ }
+
+ return (non_config_tasks || (config_task && !config_task_suspended));
+}
+
static void
spa_async_dispatch(spa_t *spa)
{
mutex_enter(&spa->spa_async_lock);
- if (spa->spa_async_tasks && !spa->spa_async_suspended &&
+ if (spa_async_tasks_pending(spa) &&
+ !spa->spa_async_suspended &&
spa->spa_async_thread == NULL &&
- rootdir != NULL && !vn_is_readonly(rootdir))
+ rootdir != NULL)
spa->spa_async_thread = thread_create(NULL, 0,
spa_async_thread, spa, 0, &p0, TS_RUN, maxclsyspri);
mutex_exit(&spa->spa_async_lock);
return (0);
}
+/*
+ * Note: this simple function is not inlined to make it easier to dtrace the
+ * amount of time spent syncing frees.
+ */
+static void
+spa_sync_frees(spa_t *spa, bplist_t *bpl, dmu_tx_t *tx)
+{
+ zio_t *zio = zio_root(spa, NULL, NULL, 0);
+ bplist_iterate(bpl, spa_free_sync_cb, zio, tx);
+ VERIFY(zio_wait(zio) == 0);
+}
+
+/*
+ * Note: this simple function is not inlined to make it easier to dtrace the
+ * amount of time spent syncing deferred frees.
+ */
+static void
+spa_sync_deferred_frees(spa_t *spa, dmu_tx_t *tx)
+{
+ zio_t *zio = zio_root(spa, NULL, NULL, 0);
+ VERIFY3U(bpobj_iterate(&spa->spa_deferred_bpobj,
+ spa_free_sync_cb, zio, tx), ==, 0);
+ VERIFY0(zio_wait(zio));
+}
+
static void
spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx)
{
/*
* Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration
- * information. This avoids the dbuf_will_dirty() path and
+ * information. This avoids the dmu_buf_will_dirty() path and
* saves us a pre-read to get data we don't actually care about.
*/
bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE);
- packed = vmem_alloc(bufsize, KM_PUSHPAGE);
+ packed = vmem_alloc(bufsize, KM_SLEEP);
VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR,
- KM_PUSHPAGE) == 0);
+ KM_SLEEP) == 0);
bzero(packed + nvsize, bufsize - nvsize);
dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx);
&sav->sav_object, tx) == 0);
}
- VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+ VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0);
if (sav->sav_count == 0) {
VERIFY(nvlist_add_nvlist_array(nvroot, config, NULL, 0) == 0);
} else {
- list = kmem_alloc(sav->sav_count * sizeof (void *), KM_PUSHPAGE);
+ list = kmem_alloc(sav->sav_count*sizeof (void *), KM_SLEEP);
for (i = 0; i < sav->sav_count; i++)
list[i] = vdev_config_generate(spa, sav->sav_vdevs[i],
B_FALSE, VDEV_CONFIG_L2CACHE);
sav->sav_sync = B_FALSE;
}
+/*
+ * Rebuild spa's all-vdev ZAP from the vdev ZAPs indicated in each vdev_t.
+ * The all-vdev ZAP must be empty.
+ */
+static void
+spa_avz_build(vdev_t *vd, uint64_t avz, dmu_tx_t *tx)
+{
+ spa_t *spa = vd->vdev_spa;
+ uint64_t i;
+
+ if (vd->vdev_top_zap != 0) {
+ VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
+ vd->vdev_top_zap, tx));
+ }
+ if (vd->vdev_leaf_zap != 0) {
+ VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
+ vd->vdev_leaf_zap, tx));
+ }
+ for (i = 0; i < vd->vdev_children; i++) {
+ spa_avz_build(vd->vdev_child[i], avz, tx);
+ }
+}
+
static void
spa_sync_config_object(spa_t *spa, dmu_tx_t *tx)
{
nvlist_t *config;
- if (list_is_empty(&spa->spa_config_dirty_list))
+ /*
+ * If the pool is being imported from a pre-per-vdev-ZAP version of ZFS,
+ * its config may not be dirty but we still need to build per-vdev ZAPs.
+ * Similarly, if the pool is being assembled (e.g. after a split), we
+ * need to rebuild the AVZ although the config may not be dirty.
+ */
+ if (list_is_empty(&spa->spa_config_dirty_list) &&
+ spa->spa_avz_action == AVZ_ACTION_NONE)
return;
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
+ ASSERT(spa->spa_avz_action == AVZ_ACTION_NONE ||
+ spa->spa_avz_action == AVZ_ACTION_INITIALIZE ||
+ spa->spa_all_vdev_zaps != 0);
+
+ if (spa->spa_avz_action == AVZ_ACTION_REBUILD) {
+ zap_cursor_t zc;
+ zap_attribute_t za;
+
+ /* Make and build the new AVZ */
+ uint64_t new_avz = zap_create(spa->spa_meta_objset,
+ DMU_OTN_ZAP_METADATA, DMU_OT_NONE, 0, tx);
+ spa_avz_build(spa->spa_root_vdev, new_avz, tx);
+
+ /* Diff old AVZ with new one */
+ for (zap_cursor_init(&zc, spa->spa_meta_objset,
+ spa->spa_all_vdev_zaps);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+ uint64_t vdzap = za.za_first_integer;
+ if (zap_lookup_int(spa->spa_meta_objset, new_avz,
+ vdzap) == ENOENT) {
+ /*
+ * ZAP is listed in old AVZ but not in new one;
+ * destroy it
+ */
+ VERIFY0(zap_destroy(spa->spa_meta_objset, vdzap,
+ tx));
+ }
+ }
+
+ zap_cursor_fini(&zc);
+
+ /* Destroy the old AVZ */
+ VERIFY0(zap_destroy(spa->spa_meta_objset,
+ spa->spa_all_vdev_zaps, tx));
+
+ /* Replace the old AVZ in the dir obj with the new one */
+ VERIFY0(zap_update(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP,
+ sizeof (new_avz), 1, &new_avz, tx));
+
+ spa->spa_all_vdev_zaps = new_avz;
+ } else if (spa->spa_avz_action == AVZ_ACTION_DESTROY) {
+ zap_cursor_t zc;
+ zap_attribute_t za;
+
+ /* Walk through the AVZ and destroy all listed ZAPs */
+ for (zap_cursor_init(&zc, spa->spa_meta_objset,
+ spa->spa_all_vdev_zaps);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+ uint64_t zap = za.za_first_integer;
+ VERIFY0(zap_destroy(spa->spa_meta_objset, zap, tx));
+ }
+
+ zap_cursor_fini(&zc);
+
+ /* Destroy and unlink the AVZ itself */
+ VERIFY0(zap_destroy(spa->spa_meta_objset,
+ spa->spa_all_vdev_zaps, tx));
+ VERIFY0(zap_remove(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP, tx));
+ spa->spa_all_vdev_zaps = 0;
+ }
+
+ if (spa->spa_all_vdev_zaps == 0) {
+ spa->spa_all_vdev_zaps = zap_create_link(spa->spa_meta_objset,
+ DMU_OTN_ZAP_METADATA, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_VDEV_ZAP_MAP, tx);
+ }
+ spa->spa_avz_action = AVZ_ACTION_NONE;
+
+ /* Create ZAPs for vdevs that don't have them. */
+ vdev_construct_zaps(spa->spa_root_vdev, tx);
+
config = spa_config_generate(spa, spa->spa_root_vdev,
dmu_tx_get_txg(tx), B_FALSE);
+ /*
+ * If we're upgrading the spa version then make sure that
+ * the config object gets updated with the correct version.
+ */
+ if (spa->spa_ubsync.ub_version < spa->spa_uberblock.ub_version)
+ fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
+ spa->spa_uberblock.ub_version);
+
spa_config_exit(spa, SCL_STATE, FTAG);
- if (spa->spa_config_syncing)
- nvlist_free(spa->spa_config_syncing);
+ nvlist_free(spa->spa_config_syncing);
spa->spa_config_syncing = config;
spa_sync_nvlist(spa, spa->spa_config_object, config, tx);
}
static void
-spa_sync_version(void *arg1, void *arg2, dmu_tx_t *tx)
+spa_sync_version(void *arg, dmu_tx_t *tx)
{
- spa_t *spa = arg1;
- uint64_t version = *(uint64_t *)arg2;
+ uint64_t *versionp = arg;
+ uint64_t version = *versionp;
+ spa_t *spa = dmu_tx_pool(tx)->dp_spa;
/*
* Setting the version is special cased when first creating the pool.
*/
ASSERT(tx->tx_txg != TXG_INITIAL);
- ASSERT(version <= SPA_VERSION);
+ ASSERT(SPA_VERSION_IS_SUPPORTED(version));
ASSERT(version >= spa_version(spa));
spa->spa_uberblock.ub_version = version;
vdev_config_dirty(spa->spa_root_vdev);
+ spa_history_log_internal(spa, "set", tx, "version=%lld", version);
}
/*
* Set zpool properties.
*/
static void
-spa_sync_props(void *arg1, void *arg2, dmu_tx_t *tx)
+spa_sync_props(void *arg, dmu_tx_t *tx)
{
- spa_t *spa = arg1;
+ nvlist_t *nvp = arg;
+ spa_t *spa = dmu_tx_pool(tx)->dp_spa;
objset_t *mos = spa->spa_meta_objset;
- nvlist_t *nvp = arg2;
nvpair_t *elem = NULL;
mutex_enter(&spa->spa_props_lock);
zpool_prop_t prop;
const char *propname;
zprop_type_t proptype;
- zfeature_info_t *feature;
+ spa_feature_t fid;
prop = zpool_name_to_prop(nvpair_name(elem));
switch ((int)prop) {
ASSERT(zpool_prop_feature(nvpair_name(elem)));
fname = strchr(nvpair_name(elem), '@') + 1;
- VERIFY3U(0, ==, zfeature_lookup_name(fname, &feature));
+ VERIFY0(zfeature_lookup_name(fname, &fid));
- spa_feature_enable(spa, feature, tx);
+ spa_feature_enable(spa, fid, tx);
+ spa_history_log_internal(spa, "set", tx,
+ "%s=enabled", nvpair_name(elem));
break;
case ZPOOL_PROP_VERSION:
- VERIFY(nvpair_value_uint64(elem, &intval) == 0);
+ intval = fnvpair_value_uint64(elem);
/*
- * The version is synced seperatly before other
+ * The version is synced separately before other
* properties and should be correct by now.
*/
ASSERT3U(spa_version(spa), >=, intval);
*/
break;
case ZPOOL_PROP_COMMENT:
- VERIFY(nvpair_value_string(elem, &strval) == 0);
+ strval = fnvpair_value_string(elem);
if (spa->spa_comment != NULL)
spa_strfree(spa->spa_comment);
spa->spa_comment = spa_strdup(strval);
* We need to dirty the configuration on all the vdevs
* so that their labels get updated. It's unnecessary
* to do this for pool creation since the vdev's
- * configuratoin has already been dirtied.
+ * configuration has already been dirtied.
*/
if (tx->tx_txg != TXG_INITIAL)
vdev_config_dirty(spa->spa_root_vdev);
+ spa_history_log_internal(spa, "set", tx,
+ "%s=%s", nvpair_name(elem), strval);
break;
default:
/*
if (nvpair_type(elem) == DATA_TYPE_STRING) {
ASSERT(proptype == PROP_TYPE_STRING);
- VERIFY(nvpair_value_string(elem, &strval) == 0);
- VERIFY(zap_update(mos,
+ strval = fnvpair_value_string(elem);
+ VERIFY0(zap_update(mos,
spa->spa_pool_props_object, propname,
- 1, strlen(strval) + 1, strval, tx) == 0);
-
+ 1, strlen(strval) + 1, strval, tx));
+ spa_history_log_internal(spa, "set", tx,
+ "%s=%s", nvpair_name(elem), strval);
} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
- VERIFY(nvpair_value_uint64(elem, &intval) == 0);
+ intval = fnvpair_value_uint64(elem);
if (proptype == PROP_TYPE_INDEX) {
const char *unused;
- VERIFY(zpool_prop_index_to_string(
- prop, intval, &unused) == 0);
+ VERIFY0(zpool_prop_index_to_string(
+ prop, intval, &unused));
}
- VERIFY(zap_update(mos,
+ VERIFY0(zap_update(mos,
spa->spa_pool_props_object, propname,
- 8, 1, &intval, tx) == 0);
+ 8, 1, &intval, tx));
+ spa_history_log_internal(spa, "set", tx,
+ "%s=%lld", nvpair_name(elem), intval);
} else {
ASSERT(0); /* not allowed */
}
spa_async_request(spa,
SPA_ASYNC_AUTOEXPAND);
break;
+ case ZPOOL_PROP_MULTIHOST:
+ spa->spa_multihost = intval;
+ break;
case ZPOOL_PROP_DEDUPDITTO:
spa->spa_dedup_ditto = intval;
break;
}
}
- /* log internal history if this is not a zpool create */
- if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY &&
- tx->tx_txg != TXG_INITIAL) {
- spa_history_log_internal(LOG_POOL_PROPSET,
- spa, tx, "%s %lld %s",
- nvpair_name(elem), intval, spa_name(spa));
- }
}
mutex_exit(&spa->spa_props_lock);
ASSERT(spa->spa_sync_pass == 1);
+ rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
+
if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN &&
spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) {
dsl_pool_create_origin(dp, tx);
spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
spa_feature_create_zap_objects(spa, tx);
}
+
+ /*
+ * LZ4_COMPRESS feature's behaviour was changed to activate_on_enable
+ * when possibility to use lz4 compression for metadata was added
+ * Old pools that have this feature enabled must be upgraded to have
+ * this feature active
+ */
+ if (spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
+ boolean_t lz4_en = spa_feature_is_enabled(spa,
+ SPA_FEATURE_LZ4_COMPRESS);
+ boolean_t lz4_ac = spa_feature_is_active(spa,
+ SPA_FEATURE_LZ4_COMPRESS);
+
+ if (lz4_en && !lz4_ac)
+ spa_feature_incr(spa, SPA_FEATURE_LZ4_COMPRESS, tx);
+ }
+
+ /*
+ * If we haven't written the salt, do so now. Note that the
+ * feature may not be activated yet, but that's fine since
+ * the presence of this ZAP entry is backwards compatible.
+ */
+ if (zap_contains(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_CHECKSUM_SALT) == ENOENT) {
+ VERIFY0(zap_add(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CHECKSUM_SALT, 1,
+ sizeof (spa->spa_cksum_salt.zcs_bytes),
+ spa->spa_cksum_salt.zcs_bytes, tx));
+ }
+
+ rrw_exit(&dp->dp_config_rwlock, FTAG);
}
/*
{
dsl_pool_t *dp = spa->spa_dsl_pool;
objset_t *mos = spa->spa_meta_objset;
- bpobj_t *defer_bpo = &spa->spa_deferred_bpobj;
bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK];
+ metaslab_class_t *mc;
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *vd;
dmu_tx_t *tx;
int error;
+ uint32_t max_queue_depth = zfs_vdev_async_write_max_active *
+ zfs_vdev_queue_depth_pct / 100;
+ uint64_t queue_depth_total;
int c;
VERIFY(spa_writeable(spa));
spa->spa_syncing_txg = txg;
spa->spa_sync_pass = 0;
+ mutex_enter(&spa->spa_alloc_lock);
+ VERIFY0(avl_numnodes(&spa->spa_alloc_tree));
+ mutex_exit(&spa->spa_alloc_lock);
+
/*
* If there are any pending vdev state changes, convert them
* into config changes that go out with this transaction group.
tx = dmu_tx_create_assigned(dp, txg);
+ spa->spa_sync_starttime = gethrtime();
+ taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
+ spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq,
+ spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() +
+ NSEC_TO_TICK(spa->spa_deadman_synctime));
+
/*
* If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg,
* set spa_deflate if we have no raid-z vdevs.
}
/*
- * If anything has changed in this txg, or if someone is waiting
- * for this txg to sync (eg, spa_vdev_remove()), push the
- * deferred frees from the previous txg. If not, leave them
- * alone so that we don't generate work on an otherwise idle
- * system.
+ * Set the top-level vdev's max queue depth. Evaluate each
+ * top-level's async write queue depth in case it changed.
+ * The max queue depth will not change in the middle of syncing
+ * out this txg.
*/
- if (!txg_list_empty(&dp->dp_dirty_datasets, txg) ||
- !txg_list_empty(&dp->dp_dirty_dirs, txg) ||
- !txg_list_empty(&dp->dp_sync_tasks, txg) ||
- ((dsl_scan_active(dp->dp_scan) ||
- txg_sync_waiting(dp)) && !spa_shutting_down(spa))) {
- zio_t *zio = zio_root(spa, NULL, NULL, 0);
- VERIFY3U(bpobj_iterate(defer_bpo,
- spa_free_sync_cb, zio, tx), ==, 0);
- VERIFY3U(zio_wait(zio), ==, 0);
+ queue_depth_total = 0;
+ for (c = 0; c < rvd->vdev_children; c++) {
+ vdev_t *tvd = rvd->vdev_child[c];
+ metaslab_group_t *mg = tvd->vdev_mg;
+
+ if (mg == NULL || mg->mg_class != spa_normal_class(spa) ||
+ !metaslab_group_initialized(mg))
+ continue;
+
+ /*
+ * It is safe to do a lock-free check here because only async
+ * allocations look at mg_max_alloc_queue_depth, and async
+ * allocations all happen from spa_sync().
+ */
+ ASSERT0(refcount_count(&mg->mg_alloc_queue_depth));
+ mg->mg_max_alloc_queue_depth = max_queue_depth;
+ queue_depth_total += mg->mg_max_alloc_queue_depth;
}
+ mc = spa_normal_class(spa);
+ ASSERT0(refcount_count(&mc->mc_alloc_slots));
+ mc->mc_alloc_max_slots = queue_depth_total;
+ mc->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
+
+ ASSERT3U(mc->mc_alloc_max_slots, <=,
+ max_queue_depth * rvd->vdev_children);
/*
* Iterate to convergence.
spa_errlog_sync(spa, txg);
dsl_pool_sync(dp, txg);
- if (pass <= SYNC_PASS_DEFERRED_FREE) {
- zio_t *zio = zio_root(spa, NULL, NULL, 0);
- bplist_iterate(free_bpl, spa_free_sync_cb,
- zio, tx);
- VERIFY(zio_wait(zio) == 0);
+ if (pass < zfs_sync_pass_deferred_free) {
+ spa_sync_frees(spa, free_bpl, tx);
} else {
+ /*
+ * We can not defer frees in pass 1, because
+ * we sync the deferred frees later in pass 1.
+ */
+ ASSERT3U(pass, >, 1);
bplist_iterate(free_bpl, bpobj_enqueue_cb,
- defer_bpo, tx);
+ &spa->spa_deferred_bpobj, tx);
}
ddt_sync(spa, txg);
while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, txg)))
vdev_sync(vd, txg);
- if (pass == 1)
+ if (pass == 1) {
spa_sync_upgrades(spa, tx);
+ ASSERT3U(txg, >=,
+ spa->spa_uberblock.ub_rootbp.blk_birth);
+ /*
+ * Note: We need to check if the MOS is dirty
+ * because we could have marked the MOS dirty
+ * without updating the uberblock (e.g. if we
+ * have sync tasks but no dirty user data). We
+ * need to check the uberblock's rootbp because
+ * it is updated if we have synced out dirty
+ * data (though in this case the MOS will most
+ * likely also be dirty due to second order
+ * effects, we don't want to rely on that here).
+ */
+ if (spa->spa_uberblock.ub_rootbp.blk_birth < txg &&
+ !dmu_objset_is_dirty(mos, txg)) {
+ /*
+ * Nothing changed on the first pass,
+ * therefore this TXG is a no-op. Avoid
+ * syncing deferred frees, so that we
+ * can keep this TXG as a no-op.
+ */
+ ASSERT(txg_list_empty(&dp->dp_dirty_datasets,
+ txg));
+ ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg));
+ ASSERT(txg_list_empty(&dp->dp_sync_tasks, txg));
+ break;
+ }
+ spa_sync_deferred_frees(spa, tx);
+ }
} while (dmu_objset_is_dirty(mos, txg));
+#ifdef ZFS_DEBUG
+ if (!list_is_empty(&spa->spa_config_dirty_list)) {
+ /*
+ * Make sure that the number of ZAPs for all the vdevs matches
+ * the number of ZAPs in the per-vdev ZAP list. This only gets
+ * called if the config is dirty; otherwise there may be
+ * outstanding AVZ operations that weren't completed in
+ * spa_sync_config_object.
+ */
+ uint64_t all_vdev_zap_entry_count;
+ ASSERT0(zap_count(spa->spa_meta_objset,
+ spa->spa_all_vdev_zaps, &all_vdev_zap_entry_count));
+ ASSERT3U(vdev_count_verify_zaps(spa->spa_root_vdev), ==,
+ all_vdev_zap_entry_count);
+ }
+#endif
+
/*
* Rewrite the vdev configuration (which includes the uberblock)
* to commit the transaction group.
if (svdcount == SPA_DVAS_PER_BP)
break;
}
- error = vdev_config_sync(svd, svdcount, txg, B_FALSE);
- if (error != 0)
- error = vdev_config_sync(svd, svdcount, txg,
- B_TRUE);
+ error = vdev_config_sync(svd, svdcount, txg);
} else {
error = vdev_config_sync(rvd->vdev_child,
- rvd->vdev_children, txg, B_FALSE);
- if (error != 0)
- error = vdev_config_sync(rvd->vdev_child,
- rvd->vdev_children, txg, B_TRUE);
+ rvd->vdev_children, txg);
}
if (error == 0)
}
dmu_tx_commit(tx);
+ taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
+ spa->spa_deadman_tqid = 0;
+
/*
* Clear the dirty config list.
*/
spa->spa_config_syncing = NULL;
}
- spa->spa_ubsync = spa->spa_uberblock;
-
dsl_pool_sync_done(dp, txg);
+ mutex_enter(&spa->spa_alloc_lock);
+ VERIFY0(avl_numnodes(&spa->spa_alloc_tree));
+ mutex_exit(&spa->spa_alloc_lock);
+
/*
* Update usable space statistics.
*/
spa->spa_sync_pass = 0;
+ /*
+ * Update the last synced uberblock here. We want to do this at
+ * the end of spa_sync() so that consumers of spa_last_synced_txg()
+ * will be guaranteed that all the processing associated with
+ * that txg has been completed.
+ */
+ spa->spa_ubsync = spa->spa_uberblock;
spa_config_exit(spa, SCL_CONFIG, FTAG);
spa_handle_ignored_writes(spa);
* future version would result in an unopenable pool, this shouldn't be
* possible.
*/
- ASSERT(spa->spa_uberblock.ub_version <= SPA_VERSION);
- ASSERT(version >= spa->spa_uberblock.ub_version);
+ ASSERT(SPA_VERSION_IS_SUPPORTED(spa->spa_uberblock.ub_version));
+ ASSERT3U(version, >=, spa->spa_uberblock.ub_version);
spa->spa_uberblock.ub_version = version;
vdev_config_dirty(spa->spa_root_vdev);
return (B_FALSE);
}
+static sysevent_t *
+spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, const char *name)
+{
+ sysevent_t *ev = NULL;
+#ifdef _KERNEL
+ nvlist_t *resource;
+
+ resource = zfs_event_create(spa, vd, FM_SYSEVENT_CLASS, name, hist_nvl);
+ if (resource) {
+ ev = kmem_alloc(sizeof (sysevent_t), KM_SLEEP);
+ ev->resource = resource;
+ }
+#endif
+ return (ev);
+}
+
+static void
+spa_event_post(sysevent_t *ev)
+{
+#ifdef _KERNEL
+ if (ev) {
+ zfs_zevent_post(ev->resource, NULL, zfs_zevent_post_cb);
+ kmem_free(ev, sizeof (*ev));
+ }
+#endif
+}
+
/*
- * Post a FM_EREPORT_ZFS_* event from sys/fm/fs/zfs.h. The payload will be
+ * Post a zevent corresponding to the given sysevent. The 'name' must be one
+ * of the event definitions in sys/sysevent/eventdefs.h. The payload will be
* filled in from the spa and (optionally) the vdev. This doesn't do anything
* in the userland libzpool, as we don't want consumers to misinterpret ztest
* or zdb as real changes.
*/
void
-spa_event_notify(spa_t *spa, vdev_t *vd, const char *name)
+spa_event_notify(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, const char *name)
{
-#ifdef _KERNEL
- zfs_ereport_post(name, spa, vd, NULL, 0, 0);
-#endif
+ spa_event_post(spa_event_create(spa, vd, hist_nvl, name));
}
#if defined(_KERNEL) && defined(HAVE_SPL)
EXPORT_SYMBOL(spa_open_rewind);
EXPORT_SYMBOL(spa_get_stats);
EXPORT_SYMBOL(spa_create);
-EXPORT_SYMBOL(spa_import_rootpool);
EXPORT_SYMBOL(spa_import);
EXPORT_SYMBOL(spa_tryimport);
EXPORT_SYMBOL(spa_destroy);
/* asynchronous event notification */
EXPORT_SYMBOL(spa_event_notify);
#endif
+
+#if defined(_KERNEL) && defined(HAVE_SPL)
+module_param(spa_load_verify_maxinflight, int, 0644);
+MODULE_PARM_DESC(spa_load_verify_maxinflight,
+ "Max concurrent traversal I/Os while verifying pool during import -X");
+
+module_param(spa_load_verify_metadata, int, 0644);
+MODULE_PARM_DESC(spa_load_verify_metadata,
+ "Set to traverse metadata on pool import");
+
+module_param(spa_load_verify_data, int, 0644);
+MODULE_PARM_DESC(spa_load_verify_data,
+ "Set to traverse data on pool import");
+
+/* CSTYLED */
+module_param(zio_taskq_batch_pct, uint, 0444);
+MODULE_PARM_DESC(zio_taskq_batch_pct,
+ "Percentage of CPUs to run an IO worker thread");
+
+#endif
projects / mirror_zfs-debian.git / blobdiff