*/
/*
- * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
- * Use is subject to license terms.
+ * 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.
*/
/*
#include <sys/spa_impl.h>
#include <sys/zio.h>
#include <sys/zio_checksum.h>
-#include <sys/zio_compress.h>
#include <sys/dmu.h>
#include <sys/dmu_tx.h>
#include <sys/zap.h>
#include <sys/zil.h>
+#include <sys/ddt.h>
#include <sys/vdev_impl.h>
+#include <sys/vdev_disk.h>
#include <sys/metaslab.h>
+#include <sys/metaslab_impl.h>
#include <sys/uberblock_impl.h>
#include <sys/txg.h>
#include <sys/avl.h>
#include <sys/arc.h>
#include <sys/callb.h>
#include <sys/systeminfo.h>
-#include <sys/sunddi.h>
#include <sys/spa_boot.h>
#include <sys/zfs_ioctl.h>
+#include <sys/dsl_scan.h>
+#include <sys/zfeature.h>
#ifdef _KERNEL
+#include <sys/bootprops.h>
+#include <sys/callb.h>
+#include <sys/cpupart.h>
+#include <sys/pool.h>
+#include <sys/sysdc.h>
#include <sys/zone.h>
#endif /* _KERNEL */
#include "zfs_prop.h"
#include "zfs_comutil.h"
-enum zti_modes {
- zti_mode_fixed, /* value is # of threads (min 1) */
- zti_mode_online_percent, /* value is % of online CPUs */
- zti_mode_tune, /* fill from zio_taskq_tune_* */
- zti_nmodes
-};
+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_modes_t;
-#define ZTI_THREAD_FIX(n) { zti_mode_fixed, (n) }
-#define ZTI_THREAD_PCT(n) { zti_mode_online_percent, (n) }
-#define ZTI_THREAD_TUNE { zti_mode_tune, 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_THREAD_ONE ZTI_THREAD_FIX(1)
+#define ZTI_N(n) ZTI_P(n, 1)
+#define ZTI_ONE ZTI_N(1)
typedef struct zio_taskq_info {
- const char *zti_name;
- struct {
- enum zti_modes zti_mode;
- uint_t zti_value;
- } zti_nthreads[ZIO_TASKQ_TYPES];
+ 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] = {
- "issue", "intr"
+ "iss", "iss_h", "int", "int_h"
};
-const zio_taskq_info_t zio_taskqs[ZIO_TYPES] = {
- /* ISSUE INTR */
- { "spa_zio_null", { ZTI_THREAD_ONE, ZTI_THREAD_ONE } },
- { "spa_zio_read", { ZTI_THREAD_FIX(8), ZTI_THREAD_TUNE } },
- { "spa_zio_write", { ZTI_THREAD_TUNE, ZTI_THREAD_FIX(8) } },
- { "spa_zio_free", { ZTI_THREAD_ONE, ZTI_THREAD_ONE } },
- { "spa_zio_claim", { ZTI_THREAD_ONE, ZTI_THREAD_ONE } },
- { "spa_zio_ioctl", { ZTI_THREAD_ONE, ZTI_THREAD_ONE } },
+/*
+ * 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 }, /* NULL */
+ { ZTI_N(8), ZTI_NULL, ZTI_BATCH, ZTI_NULL }, /* READ */
+ { ZTI_BATCH, ZTI_N(5), ZTI_N(16), ZTI_N(5) }, /* WRITE */
+ { ZTI_P(4, 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 */
};
-enum zti_modes zio_taskq_tune_mode = zti_mode_online_percent;
-uint_t zio_taskq_tune_value = 80; /* #threads = 80% of # online CPUs */
-
-static void spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx);
+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 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 */
+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 */
+
+boolean_t spa_create_process = B_TRUE; /* no process ==> no sysdc */
+
+/*
+ * This (illegal) pool name is used when temporarily importing a spa_t in order
+ * to get the vdev stats associated with the imported devices.
+ */
+#define TRYIMPORT_NAME "$import"
/*
* ==========================================================================
const char *propname = zpool_prop_to_name(prop);
nvlist_t *propval;
- VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0);
if (strval != NULL)
static void
spa_prop_get_config(spa_t *spa, nvlist_t **nvp)
{
+ vdev_t *rvd = spa->spa_root_vdev;
+ dsl_pool_t *pool = spa->spa_dsl_pool;
uint64_t size;
- uint64_t used;
+ uint64_t alloc;
+ uint64_t space;
uint64_t cap, version;
zprop_source_t src = ZPROP_SRC_NONE;
spa_config_dirent_t *dp;
+ int c;
ASSERT(MUTEX_HELD(&spa->spa_props_lock));
- if (spa->spa_root_vdev != NULL) {
- size = spa_get_space(spa);
- used = spa_get_alloc(spa);
+ if (rvd != NULL) {
+ alloc = metaslab_class_get_alloc(spa_normal_class(spa));
+ size = metaslab_class_get_space(spa_normal_class(spa));
spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src);
spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src);
- spa_prop_add_list(*nvp, ZPOOL_PROP_USED, NULL, used, src);
- spa_prop_add_list(*nvp, ZPOOL_PROP_AVAILABLE, NULL,
- size - used, src);
+ spa_prop_add_list(*nvp, ZPOOL_PROP_ALLOCATED, NULL, alloc, src);
+ 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_READONLY, NULL,
+ (spa_mode(spa) == FREAD), src);
- cap = (size == 0) ? 0 : (used * 100 / size);
+ cap = (size == 0) ? 0 : (alloc * 100 / size);
spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src);
+ spa_prop_add_list(*nvp, ZPOOL_PROP_DEDUPRATIO, NULL,
+ ddt_get_pool_dedup_ratio(spa), src);
+
spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL,
- spa->spa_root_vdev->vdev_state, src);
+ rvd->vdev_state, src);
version = spa_version(spa);
if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION))
spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL, version, src);
}
+ 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) {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL,
+ freedir->dd_phys->dd_used_bytes, src);
+ } else {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING,
+ NULL, 0, src);
+ }
+ }
+
spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src);
+ if (spa->spa_comment != NULL) {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_COMMENT, spa->spa_comment,
+ 0, ZPROP_SRC_LOCAL);
+ }
+
if (spa->spa_root != NULL)
spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root,
0, ZPROP_SRC_LOCAL);
int
spa_prop_get(spa_t *spa, nvlist_t **nvp)
{
+ objset_t *mos = spa->spa_meta_objset;
zap_cursor_t zc;
zap_attribute_t za;
- objset_t *mos = spa->spa_meta_objset;
int err;
- VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_PUSHPAGE);
+ if (err)
+ return err;
mutex_enter(&spa->spa_props_lock);
spa_prop_get_config(spa, nvp);
/* If no pool property object, no more prop to get. */
- if (spa->spa_pool_props_object == 0) {
+ if (mos == NULL || spa->spa_pool_props_object == 0) {
mutex_exit(&spa->spa_props_lock);
- return (0);
+ goto out;
}
/*
dp = spa_get_dsl(spa);
rw_enter(&dp->dp_config_rwlock, RW_READER);
- if (err = dsl_dataset_hold_obj(dp,
- za.za_first_integer, FTAG, &ds)) {
+ if ((err = dsl_dataset_hold_obj(dp,
+ za.za_first_integer, FTAG, &ds))) {
rw_exit(&dp->dp_config_rwlock);
break;
}
strval = kmem_alloc(
MAXNAMELEN + strlen(MOS_DIR_NAME) + 1,
- KM_SLEEP);
+ KM_PUSHPAGE);
dsl_dataset_name(ds, strval);
dsl_dataset_rele(ds, FTAG);
rw_exit(&dp->dp_config_rwlock);
case 1:
/* string property */
- strval = kmem_alloc(za.za_num_integers, KM_SLEEP);
+ strval = kmem_alloc(za.za_num_integers, KM_PUSHPAGE);
err = zap_lookup(mos, spa->spa_pool_props_object,
za.za_name, 1, za.za_num_integers, strval);
if (err) {
{
nvpair_t *elem;
int error = 0, reset_bootfs = 0;
- uint64_t objnum;
+ uint64_t objnum = 0;
+ boolean_t has_feature = B_FALSE;
elem = NULL;
while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
- zpool_prop_t prop;
- char *propname, *strval;
uint64_t intval;
- objset_t *os;
- char *slash;
+ char *strval, *slash, *check, *fname;
+ const char *propname = nvpair_name(elem);
+ zpool_prop_t prop = zpool_name_to_prop(propname);
+
+ switch ((int)prop) {
+ case ZPROP_INVAL:
+ if (!zpool_prop_feature(propname)) {
+ error = EINVAL;
+ break;
+ }
+
+ /*
+ * Sanitize the input.
+ */
+ if (nvpair_type(elem) != DATA_TYPE_UINT64) {
+ error = EINVAL;
+ break;
+ }
+
+ if (nvpair_value_uint64(elem, &intval) != 0) {
+ error = EINVAL;
+ break;
+ }
- propname = nvpair_name(elem);
+ if (intval != 0) {
+ error = EINVAL;
+ break;
+ }
+
+ fname = strchr(propname, '@') + 1;
+ if (zfeature_lookup_name(fname, NULL) != 0) {
+ error = EINVAL;
+ break;
+ }
- if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL)
- return (EINVAL);
+ has_feature = B_TRUE;
+ break;
- switch (prop) {
case ZPOOL_PROP_VERSION:
error = nvpair_value_uint64(elem, &intval);
if (!error &&
- (intval < spa_version(spa) || intval > SPA_VERSION))
+ (intval < spa_version(spa) ||
+ intval > SPA_VERSION_BEFORE_FEATURES ||
+ has_feature))
error = EINVAL;
break;
error = nvpair_value_string(elem, &strval);
if (!error) {
+ objset_t *os;
uint64_t compress;
if (strval == NULL || strval[0] == '\0') {
break;
}
- if (error = dmu_objset_open(strval, DMU_OST_ZFS,
- DS_MODE_USER | DS_MODE_READONLY, &os))
+ if ((error = dmu_objset_hold(strval,FTAG,&os)))
break;
- /* We don't support gzip bootable datasets */
- if ((error = dsl_prop_get_integer(strval,
+ /* Must be ZPL and not gzip compressed. */
+
+ if (dmu_objset_type(os) != DMU_OST_ZFS) {
+ error = ENOTSUP;
+ } else if ((error = dsl_prop_get_integer(strval,
zfs_prop_to_name(ZFS_PROP_COMPRESSION),
&compress, NULL)) == 0 &&
!BOOTFS_COMPRESS_VALID(compress)) {
} else {
objnum = dmu_objset_id(os);
}
- dmu_objset_close(os);
+ dmu_objset_rele(os, FTAG);
}
break;
strcmp(slash, "/..") == 0)
error = EINVAL;
break;
+
+ case ZPOOL_PROP_COMMENT:
+ if ((error = nvpair_value_string(elem, &strval)) != 0)
+ break;
+ for (check = strval; *check != '\0'; check++) {
+ if (!isprint(*check)) {
+ error = EINVAL;
+ break;
+ }
+ check++;
+ }
+ if (strlen(strval) > ZPROP_MAX_COMMENT)
+ error = E2BIG;
+ break;
+
+ case ZPOOL_PROP_DEDUPDITTO:
+ if (spa_version(spa) < SPA_VERSION_DEDUP)
+ error = ENOTSUP;
+ else
+ error = nvpair_value_uint64(elem, &intval);
+ if (error == 0 &&
+ intval != 0 && intval < ZIO_DEDUPDITTO_MIN)
+ error = EINVAL;
+ break;
+
+ default:
+ break;
}
if (error)
return;
dp = kmem_alloc(sizeof (spa_config_dirent_t),
- KM_SLEEP);
+ KM_PUSHPAGE);
if (cachefile[0] == '\0')
dp->scd_path = spa_strdup(spa_config_path);
spa_prop_set(spa_t *spa, nvlist_t *nvp)
{
int error;
- nvpair_t *elem;
+ nvpair_t *elem = NULL;
boolean_t need_sync = B_FALSE;
- zpool_prop_t prop;
if ((error = spa_prop_validate(spa, nvp)) != 0)
return (error);
- elem = NULL;
while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) {
- if ((prop = zpool_name_to_prop(
- nvpair_name(elem))) == ZPROP_INVAL)
- return (EINVAL);
+ zpool_prop_t prop = zpool_name_to_prop(nvpair_name(elem));
+
+ if (prop == ZPOOL_PROP_CACHEFILE ||
+ prop == ZPOOL_PROP_ALTROOT ||
+ prop == ZPOOL_PROP_READONLY)
+ continue;
+
+ if (prop == ZPOOL_PROP_VERSION || prop == ZPROP_INVAL) {
+ uint64_t ver;
+
+ if (prop == ZPOOL_PROP_VERSION) {
+ VERIFY(nvpair_value_uint64(elem, &ver) == 0);
+ } else {
+ ASSERT(zpool_prop_feature(nvpair_name(elem)));
+ ver = SPA_VERSION_FEATURES;
+ need_sync = B_TRUE;
+ }
+
+ /* Save time if the version is already set. */
+ if (ver == spa_version(spa))
+ continue;
- if (prop == ZPOOL_PROP_CACHEFILE || prop == ZPOOL_PROP_ALTROOT)
+ /*
+ * In addition to the pool directory object, we might
+ * create the pool properties object, the features for
+ * read object, the features for write object, or the
+ * feature descriptions object.
+ */
+ error = dsl_sync_task_do(spa_get_dsl(spa), NULL,
+ spa_sync_version, spa, &ver, 6);
+ if (error)
+ return (error);
continue;
+ }
need_sync = B_TRUE;
break;
}
- if (need_sync)
+ if (need_sync) {
return (dsl_sync_task_do(spa_get_dsl(spa), NULL, spa_sync_props,
- spa, nvp, 3));
- else
- return (0);
+ spa, nvp, 6));
+ }
+
+ return (0);
}
/*
}
}
+/*ARGSUSED*/
+static int
+spa_change_guid_check(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ spa_t *spa = arg1;
+ vdev_t *rvd = spa->spa_root_vdev;
+ uint64_t vdev_state;
+ ASSERTV(uint64_t *newguid = arg2);
+
+ 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);
+
+ ASSERT3U(spa_guid(spa), !=, *newguid);
+
+ return (0);
+}
+
+static void
+spa_change_guid_sync(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ spa_t *spa = arg1;
+ uint64_t *newguid = arg2;
+ uint64_t oldguid;
+ vdev_t *rvd = spa->spa_root_vdev;
+
+ oldguid = spa_guid(spa);
+
+ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
+ rvd->vdev_guid = *newguid;
+ rvd->vdev_guid_sum += (*newguid - oldguid);
+ vdev_config_dirty(rvd);
+ spa_config_exit(spa, SCL_STATE, FTAG);
+
+ spa_history_log_internal(LOG_POOL_GUID_CHANGE, spa, tx,
+ "old=%lld new=%lld", oldguid, *newguid);
+}
+
+/*
+ * Change the GUID for the pool. This is done so that we can later
+ * re-import a pool built from a clone of our own vdevs. We will modify
+ * the root vdev's guid, our own pool guid, and then mark all of our
+ * vdevs dirty. Note that we must make sure that all our vdevs are
+ * online when we do this, or else any vdevs that weren't present
+ * would be orphaned from our pool. We are also going to issue a
+ * sysevent to update any watchers.
+ */
+int
+spa_change_guid(spa_t *spa)
+{
+ int error;
+ uint64_t guid;
+
+ mutex_enter(&spa_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);
+
+ if (error == 0) {
+ spa_config_sync(spa, B_FALSE, B_TRUE);
+ spa_event_notify(spa, NULL, FM_EREPORT_ZFS_POOL_REGUID);
+ }
+
+ mutex_exit(&spa_namespace_lock);
+
+ return (error);
+}
+
/*
* ==========================================================================
* SPA state manipulation (open/create/destroy/import/export)
offsetof(spa_error_entry_t, se_avl));
}
+static void
+spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
+{
+ const zio_taskq_info_t *ztip = &zio_taskqs[t][q];
+ enum zti_modes mode = ztip->zti_mode;
+ uint_t value = ztip->zti_value;
+ uint_t count = ztip->zti_count;
+ spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
+ char name[32];
+ uint_t i, flags = 0;
+ boolean_t batch = B_FALSE;
+
+ if (mode == ZTI_MODE_NULL) {
+ tqs->stqs_count = 0;
+ tqs->stqs_taskq = NULL;
+ return;
+ }
+
+ ASSERT3U(count, >, 0);
+
+ tqs->stqs_count = count;
+ tqs->stqs_taskq = kmem_alloc(count * sizeof (taskq_t *), KM_SLEEP);
+
+ for (i = 0; i < count; i++) {
+ taskq_t *tq;
+
+ switch (mode) {
+ case ZTI_MODE_FIXED:
+ ASSERT3U(value, >=, 1);
+ value = MAX(value, 1);
+ break;
+
+ 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;
+ break;
+
+ default:
+ panic("unrecognized mode for %s_%s taskq (%u:%u) in "
+ "spa_activate()",
+ zio_type_name[t], zio_taskq_types[q], mode, value);
+ break;
+ }
+
+ 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 {
+ tq = taskq_create_proc(name, value, maxclsyspri, 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;
+}
+
/*
- * Activate an uninitialized pool.
+ * 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);
+
+ if (tqs->stqs_count == 1) {
+ tq = tqs->stqs_taskq[0];
+ } else {
+ tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count];
+ }
+
+ id = taskq_dispatch(tq, func, arg, flags);
+ if (id)
+ taskq_wait_id(tq, id);
+}
+
static void
-spa_activate(spa_t *spa, int mode)
+spa_create_zio_taskqs(spa_t *spa)
{
- ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);
+ int t, q;
- spa->spa_state = POOL_STATE_ACTIVE;
- spa->spa_mode = mode;
+ for (t = 0; t < ZIO_TYPES; t++) {
+ for (q = 0; q < ZIO_TASKQ_TYPES; q++) {
+ spa_taskqs_init(spa, t, q);
+ }
+ }
+}
- spa->spa_normal_class = metaslab_class_create(zfs_metaslab_ops);
- spa->spa_log_class = metaslab_class_create(zfs_metaslab_ops);
+#if defined(_KERNEL) && defined(HAVE_SPA_THREAD)
+static void
+spa_thread(void *arg)
+{
+ callb_cpr_t cprinfo;
- for (int t = 0; t < ZIO_TYPES; t++) {
- const zio_taskq_info_t *ztip = &zio_taskqs[t];
- for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
- enum zti_modes mode = ztip->zti_nthreads[q].zti_mode;
- uint_t value = ztip->zti_nthreads[q].zti_value;
- char name[32];
+ spa_t *spa = arg;
+ user_t *pu = PTOU(curproc);
- (void) snprintf(name, sizeof (name),
- "%s_%s", ztip->zti_name, zio_taskq_types[q]);
+ CALLB_CPR_INIT(&cprinfo, &spa->spa_proc_lock, callb_generic_cpr,
+ spa->spa_name);
- if (mode == zti_mode_tune) {
- mode = zio_taskq_tune_mode;
- value = zio_taskq_tune_value;
- if (mode == zti_mode_tune)
- mode = zti_mode_online_percent;
- }
+ ASSERT(curproc != &p0);
+ (void) snprintf(pu->u_psargs, sizeof (pu->u_psargs),
+ "zpool-%s", spa->spa_name);
+ (void) strlcpy(pu->u_comm, pu->u_psargs, sizeof (pu->u_comm));
- switch (mode) {
- case zti_mode_fixed:
- ASSERT3U(value, >=, 1);
- value = MAX(value, 1);
+ /* bind this thread to the requested psrset */
+ if (zio_taskq_psrset_bind != PS_NONE) {
+ pool_lock();
+ mutex_enter(&cpu_lock);
+ mutex_enter(&pidlock);
+ mutex_enter(&curproc->p_lock);
- spa->spa_zio_taskq[t][q] = taskq_create(name,
- value, maxclsyspri, 50, INT_MAX,
- TASKQ_PREPOPULATE);
- break;
+ if (cpupart_bind_thread(curthread, zio_taskq_psrset_bind,
+ 0, NULL, NULL) == 0) {
+ curthread->t_bind_pset = zio_taskq_psrset_bind;
+ } else {
+ cmn_err(CE_WARN,
+ "Couldn't bind process for zfs pool \"%s\" to "
+ "pset %d\n", spa->spa_name, zio_taskq_psrset_bind);
+ }
- case zti_mode_online_percent:
- spa->spa_zio_taskq[t][q] = taskq_create(name,
- value, maxclsyspri, 50, INT_MAX,
- TASKQ_PREPOPULATE | TASKQ_THREADS_CPU_PCT);
- break;
+ mutex_exit(&curproc->p_lock);
+ mutex_exit(&pidlock);
+ mutex_exit(&cpu_lock);
+ pool_unlock();
+ }
- case zti_mode_tune:
- default:
- panic("unrecognized mode for "
- "zio_taskqs[%u]->zti_nthreads[%u] (%u:%u) "
- "in spa_activate()",
- t, q, mode, value);
- break;
+ if (zio_taskq_sysdc) {
+ sysdc_thread_enter(curthread, 100, 0);
+ }
+
+ spa->spa_proc = curproc;
+ spa->spa_did = curthread->t_did;
+
+ spa_create_zio_taskqs(spa);
+
+ mutex_enter(&spa->spa_proc_lock);
+ ASSERT(spa->spa_proc_state == SPA_PROC_CREATED);
+
+ spa->spa_proc_state = SPA_PROC_ACTIVE;
+ cv_broadcast(&spa->spa_proc_cv);
+
+ CALLB_CPR_SAFE_BEGIN(&cprinfo);
+ while (spa->spa_proc_state == SPA_PROC_ACTIVE)
+ cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock);
+ CALLB_CPR_SAFE_END(&cprinfo, &spa->spa_proc_lock);
+
+ ASSERT(spa->spa_proc_state == SPA_PROC_DEACTIVATE);
+ spa->spa_proc_state = SPA_PROC_GONE;
+ spa->spa_proc = &p0;
+ cv_broadcast(&spa->spa_proc_cv);
+ CALLB_CPR_EXIT(&cprinfo); /* drops spa_proc_lock */
+
+ mutex_enter(&curproc->p_lock);
+ lwp_exit();
+}
+#endif
+
+/*
+ * Activate an uninitialized pool.
+ */
+static void
+spa_activate(spa_t *spa, int mode)
+{
+ ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);
+
+ spa->spa_state = POOL_STATE_ACTIVE;
+ spa->spa_mode = mode;
+
+ spa->spa_normal_class = metaslab_class_create(spa, zfs_metaslab_ops);
+ spa->spa_log_class = metaslab_class_create(spa, zfs_metaslab_ops);
+
+ /* Try to create a covering process */
+ mutex_enter(&spa->spa_proc_lock);
+ ASSERT(spa->spa_proc_state == SPA_PROC_NONE);
+ ASSERT(spa->spa_proc == &p0);
+ spa->spa_did = 0;
+
+#ifdef HAVE_SPA_THREAD
+ /* Only create a process if we're going to be around a while. */
+ if (spa_create_process && strcmp(spa->spa_name, TRYIMPORT_NAME) != 0) {
+ if (newproc(spa_thread, (caddr_t)spa, syscid, maxclsyspri,
+ NULL, 0) == 0) {
+ spa->spa_proc_state = SPA_PROC_CREATED;
+ while (spa->spa_proc_state == SPA_PROC_CREATED) {
+ cv_wait(&spa->spa_proc_cv,
+ &spa->spa_proc_lock);
}
+ ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE);
+ ASSERT(spa->spa_proc != &p0);
+ ASSERT(spa->spa_did != 0);
+ } else {
+#ifdef _KERNEL
+ cmn_err(CE_WARN,
+ "Couldn't create process for zfs pool \"%s\"\n",
+ spa->spa_name);
+#endif
}
}
+#endif /* HAVE_SPA_THREAD */
+ mutex_exit(&spa->spa_proc_lock);
+
+ /* If we didn't create a process, we need to create our taskqs. */
+ if (spa->spa_proc == &p0) {
+ spa_create_zio_taskqs(spa);
+ }
list_create(&spa->spa_config_dirty_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_config_dirty_node));
static void
spa_deactivate(spa_t *spa)
{
+ int t, q;
+
ASSERT(spa->spa_sync_on == B_FALSE);
ASSERT(spa->spa_dsl_pool == NULL);
ASSERT(spa->spa_root_vdev == NULL);
list_destroy(&spa->spa_config_dirty_list);
list_destroy(&spa->spa_state_dirty_list);
- for (int t = 0; t < ZIO_TYPES; t++) {
- for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
- taskq_destroy(spa->spa_zio_taskq[t][q]);
- spa->spa_zio_taskq[t][q] = NULL;
+ taskq_cancel_id(system_taskq, spa->spa_deadman_tqid);
+
+ for (t = 0; t < ZIO_TYPES; t++) {
+ for (q = 0; q < ZIO_TASKQ_TYPES; q++) {
+ spa_taskqs_fini(spa, t, q);
}
}
avl_destroy(&spa->spa_errlist_last);
spa->spa_state = POOL_STATE_UNINITIALIZED;
+
+ mutex_enter(&spa->spa_proc_lock);
+ if (spa->spa_proc_state != SPA_PROC_NONE) {
+ ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE);
+ spa->spa_proc_state = SPA_PROC_DEACTIVATE;
+ cv_broadcast(&spa->spa_proc_cv);
+ while (spa->spa_proc_state == SPA_PROC_DEACTIVATE) {
+ ASSERT(spa->spa_proc != &p0);
+ cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock);
+ }
+ ASSERT(spa->spa_proc_state == SPA_PROC_GONE);
+ spa->spa_proc_state = SPA_PROC_NONE;
+ }
+ ASSERT(spa->spa_proc == &p0);
+ mutex_exit(&spa->spa_proc_lock);
+
+ /*
+ * We want to make sure spa_thread() has actually exited the ZFS
+ * module, so that the module can't be unloaded out from underneath
+ * it.
+ */
+ if (spa->spa_did != 0) {
+ thread_join(spa->spa_did);
+ spa->spa_did = 0;
+ }
}
/*
nvlist_t **child;
uint_t children;
int error;
+ int c;
if ((error = vdev_alloc(spa, vdp, nv, parent, id, atype)) != 0)
return (error);
return (EINVAL);
}
- for (int c = 0; c < children; c++) {
+ for (c = 0; c < children; c++) {
vdev_t *vd;
if ((error = spa_config_parse(spa, &vd, child[c], *vdp, c,
atype)) != 0) {
spa->spa_async_zio_root = NULL;
}
+ bpobj_close(&spa->spa_deferred_bpobj);
+
/*
* Close the dsl pool.
*/
if (spa->spa_dsl_pool) {
dsl_pool_close(spa->spa_dsl_pool);
spa->spa_dsl_pool = NULL;
+ spa->spa_meta_objset = NULL;
}
+ ddt_unload(spa);
+
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
/*
}
spa->spa_spares.sav_count = 0;
- for (i = 0; i < spa->spa_l2cache.sav_count; i++)
+ for (i = 0; i < spa->spa_l2cache.sav_count; i++) {
+ vdev_clear_stats(spa->spa_l2cache.sav_vdevs[i]);
vdev_free(spa->spa_l2cache.sav_vdevs[i]);
+ }
if (spa->spa_l2cache.sav_vdevs) {
kmem_free(spa->spa_l2cache.sav_vdevs,
spa->spa_l2cache.sav_count * sizeof (void *));
spa->spa_async_suspended = 0;
+ if (spa->spa_comment != NULL) {
+ spa_strfree(spa->spa_comment);
+ spa->spa_comment = NULL;
+ }
+
spa_config_exit(spa, SCL_ALL, FTAG);
}
* active configuration, then we also mark this vdev as an active spare.
*/
spa->spa_spares.sav_vdevs = kmem_alloc(nspares * sizeof (void *),
- KM_SLEEP);
+ KM_PUSHPAGE);
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_SLEEP);
+ KM_PUSHPAGE);
for (i = 0; i < spa->spa_spares.sav_count; i++)
spares[i] = vdev_config_generate(spa,
- spa->spa_spares.sav_vdevs[i], B_TRUE, B_TRUE, B_FALSE);
+ spa->spa_spares.sav_vdevs[i], B_TRUE, VDEV_CONFIG_SPARE);
VERIFY(nvlist_add_nvlist_array(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, spares, spa->spa_spares.sav_count) == 0);
for (i = 0; i < spa->spa_spares.sav_count; i++)
uint_t nl2cache;
int i, j, oldnvdevs;
uint64_t guid;
- vdev_t *vd, **oldvdevs, **newvdevs;
+ vdev_t *vd, **oldvdevs, **newvdevs = NULL;
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_SLEEP);
+ newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_PUSHPAGE);
} else {
nl2cache = 0;
}
vd = oldvdevs[i];
if (vd != NULL) {
+ ASSERT(vd->vdev_isl2cache);
+
if (spa_l2cache_exists(vd->vdev_guid, &pool) &&
pool != 0ULL && l2arc_vdev_present(vd))
l2arc_remove_vdev(vd);
- (void) vdev_close(vd);
- spa_l2cache_remove(vd);
+ vdev_clear_stats(vd);
+ vdev_free(vd);
}
}
VERIFY(nvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE,
DATA_TYPE_NVLIST_ARRAY) == 0);
- l2cache = kmem_alloc(sav->sav_count * sizeof (void *), KM_SLEEP);
+ 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, B_FALSE, B_TRUE);
+ 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:
int error;
*value = NULL;
- VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db));
+ error = dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db);
+ if (error)
+ return (error);
+
nvsize = *(uint64_t *)db->db_data;
dmu_buf_rele(db, FTAG);
- packed = kmem_alloc(nvsize, KM_SLEEP);
+ packed = kmem_alloc(nvsize, KM_PUSHPAGE | KM_NODEBUG);
error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed,
DMU_READ_PREFETCH);
if (error == 0)
static void
spa_check_removed(vdev_t *vd)
{
- for (int c = 0; c < vd->vdev_children; c++)
+ int c;
+
+ 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_post_autoreplace(vd->vdev_spa, vd);
- spa_event_notify(vd->vdev_spa, vd, ESC_ZFS_VDEV_CHECK);
+ 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);
}
}
/*
- * Load the slog device state from the config object since it's possible
- * that the label does not contain the most up-to-date information.
+ * Validate the current config against the MOS config
*/
-void
-spa_load_log_state(spa_t *spa)
+static boolean_t
+spa_config_valid(spa_t *spa, nvlist_t *config)
{
- nvlist_t *nv, *nvroot, **child;
- uint64_t is_log;
- uint_t children;
- vdev_t *rvd = spa->spa_root_vdev;
+ vdev_t *mrvd, *rvd = spa->spa_root_vdev;
+ nvlist_t *nv;
+ int c, i;
+
+ VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nv) == 0);
+
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ VERIFY(spa_config_parse(spa, &mrvd, nv, NULL, 0, VDEV_ALLOC_LOAD) == 0);
- VERIFY(load_nvlist(spa, spa->spa_config_object, &nv) == 0);
- VERIFY(nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
- VERIFY(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
- &child, &children) == 0);
+ ASSERT3U(rvd->vdev_children, ==, mrvd->vdev_children);
+
+ /*
+ * If we're doing a normal import, then build up any additional
+ * diagnostic information about missing devices in this config.
+ * We'll pass this up to the user for further processing.
+ */
+ if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG)) {
+ nvlist_t **child, *nv;
+ uint64_t idx = 0;
+
+ child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t **),
+ KM_PUSHPAGE);
+ VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+
+ for (c = 0; c < rvd->vdev_children; c++) {
+ vdev_t *tvd = rvd->vdev_child[c];
+ vdev_t *mtvd = mrvd->vdev_child[c];
+
+ if (tvd->vdev_ops == &vdev_missing_ops &&
+ mtvd->vdev_ops != &vdev_missing_ops &&
+ mtvd->vdev_islog)
+ child[idx++] = vdev_config_generate(spa, mtvd,
+ B_FALSE, 0);
+ }
+
+ if (idx) {
+ VERIFY(nvlist_add_nvlist_array(nv,
+ ZPOOL_CONFIG_CHILDREN, child, idx) == 0);
+ VERIFY(nvlist_add_nvlist(spa->spa_load_info,
+ ZPOOL_CONFIG_MISSING_DEVICES, nv) == 0);
+
+ for (i = 0; i < idx; i++)
+ nvlist_free(child[i]);
+ }
+ nvlist_free(nv);
+ kmem_free(child, rvd->vdev_children * sizeof (char **));
+ }
- for (int c = 0; c < children; c++) {
+ /*
+ * Compare the root vdev tree with the information we have
+ * from the MOS config (mrvd). Check each top-level vdev
+ * with the corresponding MOS config top-level (mtvd).
+ */
+ for (c = 0; c < rvd->vdev_children; c++) {
vdev_t *tvd = rvd->vdev_child[c];
+ vdev_t *mtvd = mrvd->vdev_child[c];
+
+ /*
+ * Resolve any "missing" vdevs in the current configuration.
+ * If we find that the MOS config has more accurate information
+ * about the top-level vdev then use that vdev instead.
+ */
+ if (tvd->vdev_ops == &vdev_missing_ops &&
+ mtvd->vdev_ops != &vdev_missing_ops) {
+
+ if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG))
+ continue;
- if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
- &is_log) == 0 && is_log)
- vdev_load_log_state(tvd, child[c]);
+ /*
+ * Device specific actions.
+ */
+ if (mtvd->vdev_islog) {
+ spa_set_log_state(spa, SPA_LOG_CLEAR);
+ } else {
+ /*
+ * XXX - once we have 'readonly' pool
+ * support we should be able to handle
+ * missing data devices by transitioning
+ * the pool to readonly.
+ */
+ continue;
+ }
+
+ /*
+ * Swap the missing vdev with the data we were
+ * able to obtain from the MOS config.
+ */
+ vdev_remove_child(rvd, tvd);
+ vdev_remove_child(mrvd, mtvd);
+
+ vdev_add_child(rvd, mtvd);
+ vdev_add_child(mrvd, tvd);
+
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ vdev_load(mtvd);
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+
+ vdev_reopen(rvd);
+ } 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);
+ }
}
- nvlist_free(nv);
+ vdev_free(mrvd);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+
+ /*
+ * Ensure we were able to validate the config.
+ */
+ return (rvd->vdev_guid_sum == spa->spa_uberblock.ub_guid_sum);
}
/*
* Check for missing log devices
*/
-int
+static int
spa_check_logs(spa_t *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)) {
- spa->spa_log_state = SPA_LOG_MISSING;
+ spa_set_log_state(spa, SPA_LOG_MISSING);
return (1);
}
break;
return (0);
}
+static boolean_t
+spa_passivate_log(spa_t *spa)
+{
+ vdev_t *rvd = spa->spa_root_vdev;
+ boolean_t slog_found = B_FALSE;
+ int c;
+
+ ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER));
+
+ if (!spa_has_slogs(spa))
+ return (B_FALSE);
+
+ for (c = 0; c < rvd->vdev_children; c++) {
+ vdev_t *tvd = rvd->vdev_child[c];
+ metaslab_group_t *mg = tvd->vdev_mg;
+
+ if (tvd->vdev_islog) {
+ metaslab_group_passivate(mg);
+ slog_found = B_TRUE;
+ }
+ }
+
+ return (slog_found);
+}
+
+static void
+spa_activate_log(spa_t *spa)
+{
+ vdev_t *rvd = spa->spa_root_vdev;
+ int c;
+
+ ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER));
+
+ for (c = 0; c < rvd->vdev_children; c++) {
+ vdev_t *tvd = rvd->vdev_child[c];
+ metaslab_group_t *mg = tvd->vdev_mg;
+
+ if (tvd->vdev_islog)
+ metaslab_group_activate(mg);
+ }
+}
+
+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) {
+
+ /*
+ * We successfully offlined the log device, sync out the
+ * current txg so that the "stubby" block can be removed
+ * by zil_sync().
+ */
+ txg_wait_synced(spa->spa_dsl_pool, 0);
+ }
+ return (error);
+}
+
+static void
+spa_aux_check_removed(spa_aux_vdev_t *sav)
+{
+ int i;
+
+ for (i = 0; i < sav->sav_count; i++)
+ spa_check_removed(sav->sav_vdevs[i]);
+}
+
+void
+spa_claim_notify(zio_t *zio)
+{
+ spa_t *spa = zio->io_spa;
+
+ if (zio->io_error)
+ return;
+
+ mutex_enter(&spa->spa_props_lock); /* any mutex will do */
+ if (spa->spa_claim_max_txg < zio->io_bp->blk_birth)
+ spa->spa_claim_max_txg = zio->io_bp->blk_birth;
+ mutex_exit(&spa->spa_props_lock);
+}
+
+typedef struct spa_load_error {
+ uint64_t sle_meta_count;
+ uint64_t sle_data_count;
+} spa_load_error_t;
+
+static void
+spa_load_verify_done(zio_t *zio)
+{
+ blkptr_t *bp = zio->io_bp;
+ spa_load_error_t *sle = zio->io_private;
+ dmu_object_type_t type = BP_GET_TYPE(bp);
+ int error = zio->io_error;
+
+ 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);
+ else
+ atomic_add_64(&sle->sle_data_count, 1);
+ }
+ zio_data_buf_free(zio->io_data, zio->io_size);
+}
+
+/*ARGSUSED*/
+static int
+spa_load_verify_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
+ const zbookmark_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_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);
+}
+
+static int
+spa_load_verify(spa_t *spa)
+{
+ zio_t *rio;
+ spa_load_error_t sle = { 0 };
+ zpool_rewind_policy_t policy;
+ boolean_t verify_ok = B_FALSE;
+ int error;
+
+ zpool_get_rewind_policy(spa->spa_config, &policy);
+
+ if (policy.zrp_request & ZPOOL_NEVER_REWIND)
+ return (0);
+
+ 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);
+
+ (void) zio_wait(rio);
+
+ spa->spa_load_meta_errors = sle.sle_meta_count;
+ spa->spa_load_data_errors = sle.sle_data_count;
+
+ if (!error && sle.sle_meta_count <= policy.zrp_maxmeta &&
+ sle.sle_data_count <= policy.zrp_maxdata) {
+ int64_t loss = 0;
+
+ verify_ok = B_TRUE;
+ spa->spa_load_txg = spa->spa_uberblock.ub_txg;
+ spa->spa_load_txg_ts = spa->spa_uberblock.ub_timestamp;
+
+ loss = spa->spa_last_ubsync_txg_ts - spa->spa_load_txg_ts;
+ VERIFY(nvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_LOAD_TIME, spa->spa_load_txg_ts) == 0);
+ VERIFY(nvlist_add_int64(spa->spa_load_info,
+ ZPOOL_CONFIG_REWIND_TIME, loss) == 0);
+ VERIFY(nvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_LOAD_DATA_ERRORS, sle.sle_data_count) == 0);
+ } else {
+ spa->spa_load_max_txg = spa->spa_uberblock.ub_txg;
+ }
+
+ if (error) {
+ if (error != ENXIO && error != EIO)
+ error = EIO;
+ return (error);
+ }
+
+ return (verify_ok ? 0 : EIO);
+}
+
/*
- * Load an existing storage pool, using the pool's builtin spa_config as a
- * source of configuration information.
+ * Find a value in the pool props object.
+ */
+static void
+spa_prop_find(spa_t *spa, zpool_prop_t prop, uint64_t *val)
+{
+ (void) zap_lookup(spa->spa_meta_objset, spa->spa_pool_props_object,
+ zpool_prop_to_name(prop), sizeof (uint64_t), 1, val);
+}
+
+/*
+ * Find a value in the pool directory object.
*/
static int
-spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig)
+spa_dir_prop(spa_t *spa, const char *name, uint64_t *val)
{
- int error = 0;
- nvlist_t *nvroot = NULL;
- vdev_t *rvd;
- uberblock_t *ub = &spa->spa_uberblock;
- uint64_t config_cache_txg = spa->spa_config_txg;
- uint64_t pool_guid;
- uint64_t version;
- uint64_t autoreplace = 0;
- int orig_mode = spa->spa_mode;
- char *ereport = FM_EREPORT_ZFS_POOL;
+ return (zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
+ name, sizeof (uint64_t), 1, val));
+}
+
+static int
+spa_vdev_err(vdev_t *vdev, vdev_aux_t aux, int err)
+{
+ vdev_set_state(vdev, B_TRUE, VDEV_STATE_CANT_OPEN, aux);
+ return (err);
+}
+
+/*
+ * Fix up config after a partly-completed split. This is done with the
+ * ZPOOL_CONFIG_SPLIT nvlist. Both the splitting pool and the split-off
+ * pool have that entry in their config, but only the splitting one contains
+ * a list of all the guids of the vdevs that are being split off.
+ *
+ * This function determines what to do with that list: either rejoin
+ * all the disks to the pool, or complete the splitting process. To attempt
+ * the rejoin, each disk that is offlined is marked online again, and
+ * we do a reopen() call. If the vdev label for every disk that was
+ * marked online indicates it was successfully split off (VDEV_AUX_SPLIT_POOL)
+ * then we call vdev_split() on each disk, and complete the split.
+ *
+ * Otherwise we leave the config alone, with all the vdevs in place in
+ * the original pool.
+ */
+static void
+spa_try_repair(spa_t *spa, nvlist_t *config)
+{
+ uint_t extracted;
+ uint64_t *glist;
+ uint_t i, gcount;
+ nvlist_t *nvl;
+ vdev_t **vd;
+ boolean_t attempt_reopen;
+
+ if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) != 0)
+ return;
+
+ /* check that the config is complete */
+ if (nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST,
+ &glist, &gcount) != 0)
+ return;
+
+ vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_PUSHPAGE);
+
+ /* attempt to online all the vdevs & validate */
+ attempt_reopen = B_TRUE;
+ for (i = 0; i < gcount; i++) {
+ if (glist[i] == 0) /* vdev is hole */
+ continue;
+
+ vd[i] = spa_lookup_by_guid(spa, glist[i], B_FALSE);
+ if (vd[i] == NULL) {
+ /*
+ * Don't bother attempting to reopen the disks;
+ * just do the split.
+ */
+ attempt_reopen = B_FALSE;
+ } else {
+ /* attempt to re-online it */
+ vd[i]->vdev_offline = B_FALSE;
+ }
+ }
+
+ if (attempt_reopen) {
+ vdev_reopen(spa->spa_root_vdev);
+
+ /* check each device to see what state it's in */
+ for (extracted = 0, i = 0; i < gcount; i++) {
+ if (vd[i] != NULL &&
+ vd[i]->vdev_stat.vs_aux != VDEV_AUX_SPLIT_POOL)
+ break;
+ ++extracted;
+ }
+ }
/*
- * If this is an untrusted config, access the pool in read-only mode.
- * This prevents things like resilvering recently removed devices.
+ * If every disk has been moved to the new pool, or if we never
+ * even attempted to look at them, then we split them off for
+ * good.
*/
- if (!mosconfig)
- spa->spa_mode = FREAD;
+ if (!attempt_reopen || gcount == extracted) {
+ for (i = 0; i < gcount; i++)
+ if (vd[i] != NULL)
+ vdev_split(vd[i]);
+ vdev_reopen(spa->spa_root_vdev);
+ }
- ASSERT(MUTEX_HELD(&spa_namespace_lock));
+ kmem_free(vd, gcount * sizeof (vdev_t *));
+}
- spa->spa_load_state = state;
+static int
+spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type,
+ boolean_t mosconfig)
+{
+ nvlist_t *config = spa->spa_config;
+ char *ereport = FM_EREPORT_ZFS_POOL;
+ char *comment;
+ int error;
+ uint64_t pool_guid;
+ nvlist_t *nvl;
- if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) ||
- nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) {
- error = EINVAL;
- goto out;
- }
+ if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid))
+ return (EINVAL);
+
+ ASSERT(spa->spa_comment == NULL);
+ if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
+ spa->spa_comment = spa_strdup(comment);
/*
* Versioning wasn't explicitly added to the label until later, so if
* it's not present treat it as the initial version.
*/
- if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &version) != 0)
- version = SPA_VERSION_INITIAL;
+ if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
+ &spa->spa_ubsync.ub_version) != 0)
+ spa->spa_ubsync.ub_version = SPA_VERSION_INITIAL;
(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
&spa->spa_config_txg);
if ((state == SPA_LOAD_IMPORT || state == SPA_LOAD_TRYIMPORT) &&
spa_guid_exists(pool_guid, 0)) {
error = EEXIST;
- goto out;
+ } 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);
+ }
+
+ nvlist_free(spa->spa_load_info);
+ spa->spa_load_info = fnvlist_alloc();
+
+ gethrestime(&spa->spa_loaded_ts);
+ error = spa_load_impl(spa, pool_guid, config, state, type,
+ mosconfig, &ereport);
}
- spa->spa_load_guid = pool_guid;
+ spa->spa_minref = refcount_count(&spa->spa_refcount);
+ if (error) {
+ if (error != EEXIST) {
+ spa->spa_loaded_ts.tv_sec = 0;
+ spa->spa_loaded_ts.tv_nsec = 0;
+ }
+ if (error != EBADF) {
+ zfs_ereport_post(ereport, spa, NULL, NULL, 0, 0);
+ }
+ }
+ spa->spa_load_state = error ? SPA_LOAD_ERROR : SPA_LOAD_NONE;
+ spa->spa_ena = 0;
+
+ 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;
+ uint64_t obj;
+ boolean_t missing_feat_write = B_FALSE;
+
+ /*
+ * 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);
+
+ parse = (type == SPA_IMPORT_EXISTING ?
+ VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT);
/*
* Create "The Godfather" zio to hold all async IOs
* configuration requires knowing the version number.
*/
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
- spa->spa_ubsync.ub_version = version;
- error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_LOAD);
+ error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, parse);
spa_config_exit(spa, SCL_ALL, FTAG);
if (error != 0)
- goto out;
+ return (error);
ASSERT(spa->spa_root_vdev == rvd);
- ASSERT(spa_guid(spa) == pool_guid);
+
+ if (type != SPA_IMPORT_ASSEMBLE) {
+ ASSERT(spa_guid(spa) == pool_guid);
+ }
/*
* Try to open all vdevs, loading each label in the process.
error = vdev_open(rvd);
spa_config_exit(spa, SCL_ALL, FTAG);
if (error != 0)
- goto out;
+ return (error);
/*
* We need to validate the vdev labels against the configuration that
* we have in hand, which is dependent on the setting of mosconfig. If
* mosconfig is true then we're validating the vdev labels based on
- * that config. Otherwise, we're validating against the cached config
+ * that config. Otherwise, we're validating against the cached config
* (zpool.cache) that was read when we loaded the zfs module, and then
* later we will recursively call spa_load() and validate against
* the vdev config.
+ *
+ * If we're assembling a new pool that's been split off from an
+ * existing pool, the labels haven't yet been updated so we skip
+ * validation for now.
*/
- spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
- error = vdev_validate(rvd);
- spa_config_exit(spa, SCL_ALL, FTAG);
- if (error != 0)
- goto out;
+ if (type != SPA_IMPORT_ASSEMBLE) {
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ error = vdev_validate(rvd, mosconfig);
+ spa_config_exit(spa, SCL_ALL, FTAG);
- if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) {
- error = ENXIO;
- goto out;
+ if (error != 0)
+ return (error);
+
+ if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN)
+ return (ENXIO);
}
/*
* Find the best uberblock.
*/
- vdev_uberblock_load(NULL, rvd, ub);
+ vdev_uberblock_load(rvd, ub, &label);
/*
* If we weren't able to find a single valid uberblock, return failure.
*/
if (ub->ub_txg == 0) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = ENXIO;
- goto out;
+ nvlist_free(label);
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO));
}
/*
- * If the pool is newer than the code, we can't open it.
+ * If the pool has an unsupported version we can't open it.
*/
- if (ub->ub_version > SPA_VERSION) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_VERSION_NEWER);
- error = ENOTSUP;
- goto out;
+ if (!SPA_VERSION_IS_SUPPORTED(ub->ub_version)) {
+ nvlist_free(label);
+ return (spa_vdev_err(rvd, VDEV_AUX_VERSION_NEWER, ENOTSUP));
+ }
+
+ if (ub->ub_version >= SPA_VERSION_FEATURES) {
+ nvlist_t *features;
+
+ /*
+ * If we weren't able to find what's necessary for reading the
+ * MOS in the label, return failure.
+ */
+ if (label == NULL || nvlist_lookup_nvlist(label,
+ ZPOOL_CONFIG_FEATURES_FOR_READ, &features) != 0) {
+ nvlist_free(label);
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
+ ENXIO));
+ }
+
+ /*
+ * Update our in-core representation with the definitive values
+ * from the label.
+ */
+ nvlist_free(spa->spa_label_features);
+ VERIFY(nvlist_dup(features, &spa->spa_label_features, 0) == 0);
+ }
+
+ nvlist_free(label);
+
+ /*
+ * Look through entries in the label nvlist's features_for_read. If
+ * there is a feature listed there which we don't understand then we
+ * cannot open a pool.
+ */
+ if (ub->ub_version >= SPA_VERSION_FEATURES) {
+ nvlist_t *unsup_feat;
+ nvpair_t *nvp;
+
+ VERIFY(nvlist_alloc(&unsup_feat, NV_UNIQUE_NAME, KM_SLEEP) ==
+ 0);
+
+ for (nvp = nvlist_next_nvpair(spa->spa_label_features, NULL);
+ nvp != NULL;
+ nvp = nvlist_next_nvpair(spa->spa_label_features, nvp)) {
+ if (!zfeature_is_supported(nvpair_name(nvp))) {
+ VERIFY(nvlist_add_string(unsup_feat,
+ nvpair_name(nvp), "") == 0);
+ }
+ }
+
+ if (!nvlist_empty(unsup_feat)) {
+ VERIFY(nvlist_add_nvlist(spa->spa_load_info,
+ ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat) == 0);
+ nvlist_free(unsup_feat);
+ return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
+ ENOTSUP));
+ }
+
+ nvlist_free(unsup_feat);
}
/*
* If the vdev guid sum doesn't match the uberblock, we have an
- * incomplete configuration.
+ * incomplete configuration. We first check to see if the pool
+ * is aware of the complete config (i.e ZPOOL_CONFIG_VDEV_CHILDREN).
+ * If it is, defer the vdev_guid_sum check till later so we
+ * can handle missing vdevs.
*/
- if (rvd->vdev_guid_sum != ub->ub_guid_sum && mosconfig) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_BAD_GUID_SUM);
- error = ENXIO;
- goto out;
+ if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN,
+ &children) != 0 && mosconfig && type != SPA_IMPORT_ASSEMBLE &&
+ rvd->vdev_guid_sum != ub->ub_guid_sum)
+ return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO));
+
+ if (type != SPA_IMPORT_ASSEMBLE && spa->spa_config_splitting) {
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ spa_try_repair(spa, config);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ nvlist_free(spa->spa_config_splitting);
+ spa->spa_config_splitting = NULL;
}
/*
*/
spa->spa_state = POOL_STATE_ACTIVE;
spa->spa_ubsync = spa->spa_uberblock;
- spa->spa_first_txg = spa_last_synced_txg(spa) + 1;
- error = dsl_pool_open(spa, spa->spa_first_txg, &spa->spa_dsl_pool);
- if (error) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- goto out;
- }
+ spa->spa_verify_min_txg = spa->spa_extreme_rewind ?
+ TXG_INITIAL - 1 : spa_last_synced_txg(spa) - TXG_DEFER_SIZE - 1;
+ spa->spa_first_txg = spa->spa_last_ubsync_txg ?
+ spa->spa_last_ubsync_txg : spa_last_synced_txg(spa) + 1;
+ spa->spa_claim_max_txg = spa->spa_first_txg;
+ spa->spa_prev_software_version = ub->ub_software_version;
+
+ error = dsl_pool_init(spa, spa->spa_first_txg, &spa->spa_dsl_pool);
+ if (error)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset;
- if (zap_lookup(spa->spa_meta_objset,
- DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG,
- sizeof (uint64_t), 1, &spa->spa_config_object) != 0) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
+ if (spa_dir_prop(spa, DMU_POOL_CONFIG, &spa->spa_config_object) != 0)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+
+ if (spa_version(spa) >= SPA_VERSION_FEATURES) {
+ boolean_t missing_feat_read = B_FALSE;
+ nvlist_t *unsup_feat, *enabled_feat;
+
+ if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_READ,
+ &spa->spa_feat_for_read_obj) != 0) {
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+ }
+
+ if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_WRITE,
+ &spa->spa_feat_for_write_obj) != 0) {
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+ }
+
+ if (spa_dir_prop(spa, DMU_POOL_FEATURE_DESCRIPTIONS,
+ &spa->spa_feat_desc_obj) != 0) {
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+ }
+
+ 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,
+ 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,
+ unsup_feat, enabled_feat)) {
+ missing_feat_write = B_TRUE;
+ }
+ }
+
+ fnvlist_add_nvlist(spa->spa_load_info,
+ ZPOOL_CONFIG_ENABLED_FEAT, enabled_feat);
+
+ if (!nvlist_empty(unsup_feat)) {
+ fnvlist_add_nvlist(spa->spa_load_info,
+ ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat);
+ }
+
+ fnvlist_free(enabled_feat);
+ fnvlist_free(unsup_feat);
+
+ if (!missing_feat_read) {
+ fnvlist_add_boolean(spa->spa_load_info,
+ ZPOOL_CONFIG_CAN_RDONLY);
+ }
+
+ /*
+ * If the state is SPA_LOAD_TRYIMPORT, our objective is
+ * twofold: to determine whether the pool is available for
+ * import in read-write mode and (if it is not) whether the
+ * pool is available for import in read-only mode. If the pool
+ * is available for import in read-write mode, it is displayed
+ * as available in userland; if it is not available for import
+ * in read-only mode, it is displayed as unavailable in
+ * userland. If the pool is available for import in read-only
+ * mode but not read-write mode, it is displayed as unavailable
+ * in userland with a special note that the pool is actually
+ * available for open in read-only mode.
+ *
+ * As a result, if the state is SPA_LOAD_TRYIMPORT and we are
+ * missing a feature for write, we must first determine whether
+ * the pool can be opened read-only before returning to
+ * userland in order to know whether to display the
+ * abovementioned note.
+ */
+ if (missing_feat_read || (missing_feat_write &&
+ spa_writeable(spa))) {
+ return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
+ ENOTSUP));
+ }
}
+ spa->spa_is_initializing = B_TRUE;
+ error = dsl_pool_open(spa->spa_dsl_pool);
+ spa->spa_is_initializing = B_FALSE;
+ if (error != 0)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+
if (!mosconfig) {
- nvlist_t *newconfig;
uint64_t hostid;
+ nvlist_t *policy = NULL, *nvconfig;
- if (load_nvlist(spa, spa->spa_config_object, &newconfig) != 0) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
+ if (load_nvlist(spa, spa->spa_config_object, &nvconfig) != 0)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
- if (!spa_is_root(spa) && nvlist_lookup_uint64(newconfig,
+ if (!spa_is_root(spa) && nvlist_lookup_uint64(nvconfig,
ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
char *hostname;
unsigned long myhostid = 0;
- VERIFY(nvlist_lookup_string(newconfig,
+ VERIFY(nvlist_lookup_string(nvconfig,
ZPOOL_CONFIG_HOSTNAME, &hostname) == 0);
#ifdef _KERNEL
#endif /* _KERNEL */
if (hostid != 0 && myhostid != 0 &&
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://www.sun.com/msg/ZFS-8000-EY",
+ "See: http://zfsonlinux.org/msg/ZFS-8000-EY",
spa_name(spa), hostname,
(unsigned long)hostid);
- error = EBADF;
- goto out;
+ return (EBADF);
}
}
+ if (nvlist_lookup_nvlist(spa->spa_config,
+ ZPOOL_REWIND_POLICY, &policy) == 0)
+ VERIFY(nvlist_add_nvlist(nvconfig,
+ ZPOOL_REWIND_POLICY, policy) == 0);
- spa_config_set(spa, newconfig);
+ spa_config_set(spa, nvconfig);
spa_unload(spa);
spa_deactivate(spa);
spa_activate(spa, orig_mode);
- return (spa_load(spa, newconfig, state, B_TRUE));
+ return (spa_load(spa, state, SPA_IMPORT_EXISTING, B_TRUE));
}
- if (zap_lookup(spa->spa_meta_objset,
- DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPLIST,
- sizeof (uint64_t), 1, &spa->spa_sync_bplist_obj) != 0) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
+ 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)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
/*
* Load the bit that tells us to use the new accounting function
* (raid-z deflation). If we have an older pool, this will not
* be present.
*/
- error = zap_lookup(spa->spa_meta_objset,
- DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
- sizeof (uint64_t), 1, &spa->spa_deflate);
- if (error != 0 && error != ENOENT) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
+ error = spa_dir_prop(spa, DMU_POOL_DEFLATE, &spa->spa_deflate);
+ if (error != 0 && error != ENOENT)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+
+ error = spa_dir_prop(spa, DMU_POOL_CREATION_VERSION,
+ &spa->spa_creation_version);
+ if (error != 0 && error != ENOENT)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
/*
* Load the persistent error log. If we have an older pool, this will
* not be present.
*/
- error = zap_lookup(spa->spa_meta_objset,
- DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ERRLOG_LAST,
- sizeof (uint64_t), 1, &spa->spa_errlog_last);
- if (error != 0 && error != ENOENT) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
+ error = spa_dir_prop(spa, DMU_POOL_ERRLOG_LAST, &spa->spa_errlog_last);
+ if (error != 0 && error != ENOENT)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
- error = zap_lookup(spa->spa_meta_objset,
- DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ERRLOG_SCRUB,
- sizeof (uint64_t), 1, &spa->spa_errlog_scrub);
- if (error != 0 && error != ENOENT) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
+ error = spa_dir_prop(spa, DMU_POOL_ERRLOG_SCRUB,
+ &spa->spa_errlog_scrub);
+ if (error != 0 && error != ENOENT)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
/*
* Load the history object. If we have an older pool, this
* will not be present.
*/
- error = zap_lookup(spa->spa_meta_objset,
- DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_HISTORY,
- sizeof (uint64_t), 1, &spa->spa_history);
- if (error != 0 && error != ENOENT) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
+ error = spa_dir_prop(spa, DMU_POOL_HISTORY, &spa->spa_history);
+ if (error != 0 && error != ENOENT)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+
+ /*
+ * If we're assembling the pool from the split-off vdevs of
+ * an existing pool, we don't want to attach the spares & cache
+ * devices.
+ */
/*
* Load any hot spares for this pool.
*/
- error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
- DMU_POOL_SPARES, sizeof (uint64_t), 1, &spa->spa_spares.sav_object);
- if (error != 0 && error != ENOENT) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
- if (error == 0) {
+ error = spa_dir_prop(spa, DMU_POOL_SPARES, &spa->spa_spares.sav_object);
+ if (error != 0 && error != ENOENT)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+ if (error == 0 && type != SPA_IMPORT_ASSEMBLE) {
ASSERT(spa_version(spa) >= SPA_VERSION_SPARES);
if (load_nvlist(spa, spa->spa_spares.sav_object,
- &spa->spa_spares.sav_config) != 0) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
+ &spa->spa_spares.sav_config) != 0)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa_load_spares(spa);
spa_config_exit(spa, SCL_ALL, FTAG);
+ } else if (error == 0) {
+ spa->spa_spares.sav_sync = B_TRUE;
}
/*
* Load any level 2 ARC devices for this pool.
*/
- error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
- DMU_POOL_L2CACHE, sizeof (uint64_t), 1,
+ error = spa_dir_prop(spa, DMU_POOL_L2CACHE,
&spa->spa_l2cache.sav_object);
- if (error != 0 && error != ENOENT) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
- if (error == 0) {
+ if (error != 0 && error != ENOENT)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+ if (error == 0 && type != SPA_IMPORT_ASSEMBLE) {
ASSERT(spa_version(spa) >= SPA_VERSION_L2CACHE);
if (load_nvlist(spa, spa->spa_l2cache.sav_object,
- &spa->spa_l2cache.sav_config) != 0) {
- vdev_set_state(rvd, B_TRUE,
- VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
+ &spa->spa_l2cache.sav_config) != 0)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa_load_l2cache(spa);
spa_config_exit(spa, SCL_ALL, FTAG);
+ } else if (error == 0) {
+ spa->spa_l2cache.sav_sync = B_TRUE;
}
- spa_load_log_state(spa);
-
- if (spa_check_logs(spa)) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_BAD_LOG);
- error = ENXIO;
- ereport = FM_EREPORT_ZFS_LOG_REPLAY;
- goto out;
- }
-
-
spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
- error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
- DMU_POOL_PROPS, sizeof (uint64_t), 1, &spa->spa_pool_props_object);
-
- if (error && error != ENOENT) {
- vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
- VDEV_AUX_CORRUPT_DATA);
- error = EIO;
- goto out;
- }
+ error = spa_dir_prop(spa, DMU_POOL_PROPS, &spa->spa_pool_props_object);
+ if (error && error != ENOENT)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
if (error == 0) {
- (void) zap_lookup(spa->spa_meta_objset,
- spa->spa_pool_props_object,
- zpool_prop_to_name(ZPOOL_PROP_BOOTFS),
- sizeof (uint64_t), 1, &spa->spa_bootfs);
- (void) zap_lookup(spa->spa_meta_objset,
- spa->spa_pool_props_object,
- zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE),
- sizeof (uint64_t), 1, &autoreplace);
- (void) zap_lookup(spa->spa_meta_objset,
- spa->spa_pool_props_object,
- zpool_prop_to_name(ZPOOL_PROP_DELEGATION),
- sizeof (uint64_t), 1, &spa->spa_delegation);
- (void) zap_lookup(spa->spa_meta_objset,
- spa->spa_pool_props_object,
- zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE),
- sizeof (uint64_t), 1, &spa->spa_failmode);
- (void) zap_lookup(spa->spa_meta_objset,
- spa->spa_pool_props_object,
- zpool_prop_to_name(ZPOOL_PROP_AUTOEXPAND),
- sizeof (uint64_t), 1, &spa->spa_autoexpand);
+ uint64_t autoreplace;
+
+ 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_DEDUPDITTO,
+ &spa->spa_dedup_ditto);
+
+ spa->spa_autoreplace = (autoreplace != 0);
}
/*
* unopenable vdevs so that the normal autoreplace handler can take
* over.
*/
- if (autoreplace && state != SPA_LOAD_TRYIMPORT)
+ if (spa->spa_autoreplace && state != SPA_LOAD_TRYIMPORT) {
spa_check_removed(spa->spa_root_vdev);
+ /*
+ * For the import case, this is done in spa_import(), because
+ * at this point we're using the spare definitions from
+ * the MOS config, not necessarily from the userland config.
+ */
+ if (state != SPA_LOAD_IMPORT) {
+ spa_aux_check_removed(&spa->spa_spares);
+ spa_aux_check_removed(&spa->spa_l2cache);
+ }
+ }
/*
* Load the vdev state for all toplevel vdevs.
spa_config_exit(spa, SCL_ALL, FTAG);
/*
- * Check the state of the root vdev. If it can't be opened, it
- * indicates one or more toplevel vdevs are faulted.
+ * Load the DDTs (dedup tables).
*/
- if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) {
- error = ENXIO;
- goto out;
+ error = ddt_load(spa);
+ if (error != 0)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+
+ spa_update_dspace(spa);
+
+ /*
+ * Validate the config, using the MOS config to fill in any
+ * information which might be missing. If we fail to validate
+ * the config then declare the pool unfit for use. If we're
+ * assembling a pool from a split, the log is not transferred
+ * over.
+ */
+ if (type != SPA_IMPORT_ASSEMBLE) {
+ nvlist_t *nvconfig;
+
+ if (load_nvlist(spa, spa->spa_config_object, &nvconfig) != 0)
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
+
+ if (!spa_config_valid(spa, nvconfig)) {
+ nvlist_free(nvconfig);
+ return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM,
+ ENXIO));
+ }
+ nvlist_free(nvconfig);
+
+ /*
+ * Now that we've validated the config, check the state of the
+ * root vdev. If it can't be opened, it indicates one or
+ * more toplevel vdevs are faulted.
+ */
+ if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN)
+ return (ENXIO);
+
+ if (spa_check_logs(spa)) {
+ *ereport = FM_EREPORT_ZFS_LOG_REPLAY;
+ return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG, ENXIO));
+ }
}
- if (spa_writeable(spa)) {
+ if (missing_feat_write) {
+ ASSERT(state == SPA_LOAD_TRYIMPORT);
+
+ /*
+ * At this point, we know that we can open the pool in
+ * read-only mode but not read-write mode. We now have enough
+ * information and can return to userland.
+ */
+ return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT, ENOTSUP));
+ }
+
+ /*
+ * We've successfully opened the pool, verify that we're ready
+ * to start pushing transactions.
+ */
+ if (state != SPA_LOAD_TRYIMPORT) {
+ if ((error = spa_load_verify(spa)))
+ return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
+ error));
+ }
+
+ if (spa_writeable(spa) && (state == SPA_LOAD_RECOVER ||
+ spa->spa_load_max_txg == UINT64_MAX)) {
dmu_tx_t *tx;
int need_update = B_FALSE;
+ int c;
ASSERT(state != SPA_LOAD_TRYIMPORT);
/*
* Claim log blocks that haven't been committed yet.
* This must all happen in a single txg.
+ * Note: spa_claim_max_txg is updated by spa_claim_notify(),
+ * invoked from zil_claim_log_block()'s i/o done callback.
+ * Price of rollback is that we abandon the log.
*/
+ spa->spa_claiming = B_TRUE;
+
tx = dmu_tx_create_assigned(spa_get_dsl(spa),
spa_first_txg(spa));
(void) dmu_objset_find(spa_name(spa),
zil_claim, tx, DS_FIND_CHILDREN);
dmu_tx_commit(tx);
- spa->spa_log_state = SPA_LOG_GOOD;
+ spa->spa_claiming = B_FALSE;
+
+ spa_set_log_state(spa, SPA_LOG_GOOD);
spa->spa_sync_on = B_TRUE;
txg_sync_start(spa->spa_dsl_pool);
/*
- * Wait for all claims to sync.
+ * Wait for all claims to sync. We sync up to the highest
+ * claimed log block birth time so that claimed log blocks
+ * don't appear to be from the future. spa_claim_max_txg
+ * will have been set for us by either zil_check_log_chain()
+ * (invoked from spa_check_logs()) or zil_claim() above.
*/
- txg_wait_synced(spa->spa_dsl_pool, 0);
+ txg_wait_synced(spa->spa_dsl_pool, spa->spa_claim_max_txg);
/*
* If the config cache is stale, or we have uninitialized
* metaslabs (see spa_vdev_add()), then update the config.
*
- * If spa_load_verbatim is true, trust the current
+ * If this is a verbatim import, trust the current
* in-core spa_config and update the disk labels.
*/
if (config_cache_txg != spa->spa_config_txg ||
- state == SPA_LOAD_IMPORT || spa->spa_load_verbatim)
+ state == SPA_LOAD_IMPORT ||
+ state == SPA_LOAD_RECOVER ||
+ (spa->spa_import_flags & ZFS_IMPORT_VERBATIM))
need_update = B_TRUE;
- for (int c = 0; c < rvd->vdev_children; c++)
+ for (c = 0; c < rvd->vdev_children; c++)
if (rvd->vdev_child[c]->vdev_ms_array == 0)
need_update = B_TRUE;
/*
* Check all DTLs to see if anything needs resilvering.
*/
- if (vdev_resilver_needed(rvd, NULL, NULL))
+ if (!dsl_scan_resilvering(spa->spa_dsl_pool) &&
+ vdev_resilver_needed(rvd, NULL, NULL))
spa_async_request(spa, SPA_ASYNC_RESILVER);
+
+ /*
+ * Delete any inconsistent datasets.
+ */
+ (void) dmu_objset_find(spa_name(spa),
+ dsl_destroy_inconsistent, NULL, DS_FIND_CHILDREN);
+
+ /*
+ * Clean up any stale temporary dataset userrefs.
+ */
+ dsl_pool_clean_tmp_userrefs(spa->spa_dsl_pool);
}
- error = 0;
-out:
- spa->spa_minref = refcount_count(&spa->spa_refcount);
- if (error && error != EBADF)
- zfs_ereport_post(ereport, spa, NULL, NULL, 0, 0);
- spa->spa_load_state = SPA_LOAD_NONE;
- spa->spa_ena = 0;
+ return (0);
+}
- return (error);
+static int
+spa_load_retry(spa_t *spa, spa_load_state_t state, int mosconfig)
+{
+ int mode = spa->spa_mode;
+
+ spa_unload(spa);
+ spa_deactivate(spa);
+
+ spa->spa_load_max_txg--;
+
+ spa_activate(spa, mode);
+ spa_async_suspend(spa);
+
+ return (spa_load(spa, state, SPA_IMPORT_EXISTING, mosconfig));
+}
+
+/*
+ * If spa_load() fails this function will try loading prior txg's. If
+ * 'state' is SPA_LOAD_RECOVER and one of these loads succeeds the pool
+ * will be rewound to that txg. If 'state' is not SPA_LOAD_RECOVER this
+ * function will not rewind the pool and will return the same error as
+ * spa_load().
+ */
+static int
+spa_load_best(spa_t *spa, spa_load_state_t state, int mosconfig,
+ uint64_t max_request, int rewind_flags)
+{
+ nvlist_t *loadinfo = NULL;
+ nvlist_t *config = NULL;
+ int load_error, rewind_error;
+ uint64_t safe_rewind_txg;
+ uint64_t min_txg;
+
+ if (spa->spa_load_txg && state == SPA_LOAD_RECOVER) {
+ spa->spa_load_max_txg = spa->spa_load_txg;
+ spa_set_log_state(spa, SPA_LOG_CLEAR);
+ } else {
+ spa->spa_load_max_txg = max_request;
+ }
+
+ load_error = rewind_error = spa_load(spa, state, SPA_IMPORT_EXISTING,
+ mosconfig);
+ if (load_error == 0)
+ return (0);
+
+ if (spa->spa_root_vdev != NULL)
+ config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
+
+ spa->spa_last_ubsync_txg = spa->spa_uberblock.ub_txg;
+ spa->spa_last_ubsync_txg_ts = spa->spa_uberblock.ub_timestamp;
+
+ if (rewind_flags & ZPOOL_NEVER_REWIND) {
+ nvlist_free(config);
+ return (load_error);
+ }
+
+ if (state == SPA_LOAD_RECOVER) {
+ /* Price of rolling back is discarding txgs, including log */
+ spa_set_log_state(spa, SPA_LOG_CLEAR);
+ } else {
+ /*
+ * If we aren't rolling back save the load info from our first
+ * import attempt so that we can restore it after attempting
+ * to rewind.
+ */
+ loadinfo = spa->spa_load_info;
+ spa->spa_load_info = fnvlist_alloc();
+ }
+
+ spa->spa_load_max_txg = spa->spa_last_ubsync_txg;
+ safe_rewind_txg = spa->spa_last_ubsync_txg - TXG_DEFER_SIZE;
+ min_txg = (rewind_flags & ZPOOL_EXTREME_REWIND) ?
+ TXG_INITIAL : safe_rewind_txg;
+
+ /*
+ * Continue as long as we're finding errors, we're still within
+ * the acceptable rewind range, and we're still finding uberblocks
+ */
+ while (rewind_error && spa->spa_uberblock.ub_txg >= min_txg &&
+ spa->spa_uberblock.ub_txg <= spa->spa_load_max_txg) {
+ if (spa->spa_load_max_txg < safe_rewind_txg)
+ spa->spa_extreme_rewind = B_TRUE;
+ rewind_error = spa_load_retry(spa, state, mosconfig);
+ }
+
+ spa->spa_extreme_rewind = B_FALSE;
+ spa->spa_load_max_txg = UINT64_MAX;
+
+ if (config && (rewind_error || state != SPA_LOAD_RECOVER))
+ spa_config_set(spa, config);
+
+ if (state == SPA_LOAD_RECOVER) {
+ ASSERT3P(loadinfo, ==, NULL);
+ return (rewind_error);
+ } else {
+ /* Store the rewind info as part of the initial load info */
+ fnvlist_add_nvlist(loadinfo, ZPOOL_CONFIG_REWIND_INFO,
+ spa->spa_load_info);
+
+ /* Restore the initial load info */
+ fnvlist_free(spa->spa_load_info);
+ spa->spa_load_info = loadinfo;
+
+ return (load_error);
+ }
}
/*
* ambiguous state.
*/
static int
-spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t **config)
+spa_open_common(const char *pool, spa_t **spapp, void *tag, nvlist_t *nvpolicy,
+ nvlist_t **config)
{
spa_t *spa;
+ spa_load_state_t state = SPA_LOAD_OPEN;
int error;
int locked = B_FALSE;
mutex_exit(&spa_namespace_lock);
return (ENOENT);
}
+
if (spa->spa_state == POOL_STATE_UNINITIALIZED) {
+ zpool_rewind_policy_t policy;
+
+ zpool_get_rewind_policy(nvpolicy ? nvpolicy : spa->spa_config,
+ &policy);
+ if (policy.zrp_request & ZPOOL_DO_REWIND)
+ state = SPA_LOAD_RECOVER;
spa_activate(spa, spa_mode_global);
- error = spa_load(spa, spa->spa_config, SPA_LOAD_OPEN, B_FALSE);
+ if (state != SPA_LOAD_RECOVER)
+ spa->spa_last_ubsync_txg = spa->spa_load_txg = 0;
+
+ error = spa_load_best(spa, state, B_FALSE, policy.zrp_txg,
+ policy.zrp_request);
if (error == EBADF) {
/*
* information: the state of each vdev after the
* attempted vdev_open(). Return this to the user.
*/
- if (config != NULL && spa->spa_root_vdev != NULL)
- *config = spa_config_generate(spa, NULL, -1ULL,
- B_TRUE);
+ if (config != NULL && spa->spa_config) {
+ VERIFY(nvlist_dup(spa->spa_config, config,
+ KM_PUSHPAGE) == 0);
+ VERIFY(nvlist_add_nvlist(*config,
+ ZPOOL_CONFIG_LOAD_INFO,
+ spa->spa_load_info) == 0);
+ }
spa_unload(spa);
spa_deactivate(spa);
- spa->spa_last_open_failed = B_TRUE;
+ spa->spa_last_open_failed = error;
if (locked)
mutex_exit(&spa_namespace_lock);
*spapp = NULL;
return (error);
- } else {
- spa->spa_last_open_failed = B_FALSE;
}
}
spa_open_ref(spa, tag);
- if (locked)
+ if (config != NULL)
+ *config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
+
+ /*
+ * If we've recovered the pool, pass back any information we
+ * gathered while doing the load.
+ */
+ if (state == SPA_LOAD_RECOVER) {
+ VERIFY(nvlist_add_nvlist(*config, ZPOOL_CONFIG_LOAD_INFO,
+ spa->spa_load_info) == 0);
+ }
+
+ if (locked) {
+ spa->spa_last_open_failed = 0;
+ spa->spa_last_ubsync_txg = 0;
+ spa->spa_load_txg = 0;
mutex_exit(&spa_namespace_lock);
+ }
*spapp = spa;
- if (config != NULL)
- *config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
+ return (0);
+}
- return (0);
+int
+spa_open_rewind(const char *name, spa_t **spapp, void *tag, nvlist_t *policy,
+ nvlist_t **config)
+{
+ return (spa_open_common(name, spapp, tag, policy, config));
}
int
spa_open(const char *name, spa_t **spapp, void *tag)
{
- return (spa_open_common(name, spapp, tag, NULL));
+ return (spa_open_common(name, spapp, tag, NULL, NULL));
}
/*
if (spa_spare_exists(guid, &pool, NULL) &&
pool != 0ULL) {
VERIFY(nvlist_lookup_uint64_array(
- spares[i], ZPOOL_CONFIG_STATS,
+ spares[i], ZPOOL_CONFIG_VDEV_STATS,
(uint64_t **)&vs, &vsc) == 0);
vs->vs_state = VDEV_STATE_CANT_OPEN;
vs->vs_aux = VDEV_AUX_SPARED;
ASSERT(vd != NULL);
VERIFY(nvlist_lookup_uint64_array(l2cache[i],
- ZPOOL_CONFIG_STATS, (uint64_t **)&vs, &vsc) == 0);
+ ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
+ == 0);
vdev_get_stats(vd, vs);
}
}
}
+static void
+spa_add_feature_stats(spa_t *spa, nvlist_t *config)
+{
+ 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_retrieve(&zc, &za) == 0;
+ 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,
+ za.za_first_integer));
+ }
+ zap_cursor_fini(&zc);
+ }
+
+ if (spa->spa_feat_for_write_obj != 0) {
+ for (zap_cursor_init(&zc, spa->spa_meta_objset,
+ spa->spa_feat_for_write_obj);
+ zap_cursor_retrieve(&zc, &za) == 0;
+ zap_cursor_advance(&zc)) {
+ ASSERT(za.za_integer_length == sizeof (uint64_t) &&
+ za.za_num_integers == 1);
+ VERIFY3U(0, ==, nvlist_add_uint64(features, za.za_name,
+ za.za_first_integer));
+ }
+ zap_cursor_fini(&zc);
+ }
+
+ VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS,
+ features) == 0);
+ nvlist_free(features);
+}
+
int
-spa_get_stats(const char *name, nvlist_t **config, char *altroot, size_t buflen)
+spa_get_stats(const char *name, nvlist_t **config,
+ char *altroot, size_t buflen)
{
int error;
spa_t *spa;
*config = NULL;
- error = spa_open_common(name, &spa, FTAG, config);
+ error = spa_open_common(name, &spa, FTAG, NULL, config);
if (spa != NULL) {
/*
spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
if (*config != NULL) {
+ uint64_t loadtimes[2];
+
+ loadtimes[0] = spa->spa_loaded_ts.tv_sec;
+ loadtimes[1] = spa->spa_loaded_ts.tv_nsec;
+ VERIFY(nvlist_add_uint64_array(*config,
+ ZPOOL_CONFIG_LOADED_TIME, loadtimes, 2) == 0);
+
VERIFY(nvlist_add_uint64(*config,
ZPOOL_CONFIG_ERRCOUNT,
spa_get_errlog_size(spa)) == 0);
spa_add_spares(spa, *config);
spa_add_l2cache(spa, *config);
+ spa_add_feature_stats(spa, *config);
}
}
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
&olddevs, &oldndevs) == 0);
newdevs = kmem_alloc(sizeof (void *) *
- (ndevs + oldndevs), KM_SLEEP);
+ (ndevs + oldndevs), KM_PUSHPAGE);
for (i = 0; i < oldndevs; i++)
VERIFY(nvlist_dup(olddevs[i], &newdevs[i],
- KM_SLEEP) == 0);
+ KM_PUSHPAGE) == 0);
for (i = 0; i < ndevs; i++)
VERIFY(nvlist_dup(devs[i], &newdevs[i + oldndevs],
- KM_SLEEP) == 0);
+ KM_PUSHPAGE) == 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_SLEEP) == 0);
+ KM_PUSHPAGE) == 0);
VERIFY(nvlist_add_nvlist_array(sav->sav_config, config,
devs, ndevs) == 0);
}
if (spa_l2cache_exists(vd->vdev_guid, &pool) &&
pool != 0ULL && l2arc_vdev_present(vd))
l2arc_remove_vdev(vd);
- if (vd->vdev_isl2cache)
- spa_l2cache_remove(vd);
- vdev_clear_stats(vd);
- (void) vdev_close(vd);
}
}
uint64_t txg = TXG_INITIAL;
nvlist_t **spares, **l2cache;
uint_t nspares, nl2cache;
- uint64_t version;
+ uint64_t version, obj;
+ boolean_t has_features;
+ nvpair_t *elem;
+ int c;
/*
* If this pool already exists, return failure.
*/
(void) nvlist_lookup_string(props,
zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
- spa = spa_add(pool, altroot);
+ spa = spa_add(pool, NULL, altroot);
spa_activate(spa, spa_mode_global);
- spa->spa_uberblock.ub_txg = txg - 1;
-
if (props && (error = spa_prop_validate(spa, props))) {
spa_deactivate(spa);
spa_remove(spa);
return (error);
}
- if (nvlist_lookup_uint64(props, zpool_prop_to_name(ZPOOL_PROP_VERSION),
- &version) != 0)
+ has_features = B_FALSE;
+ for (elem = nvlist_next_nvpair(props, NULL);
+ elem != NULL; elem = nvlist_next_nvpair(props, elem)) {
+ if (zpool_prop_feature(nvpair_name(elem)))
+ has_features = B_TRUE;
+ }
+
+ if (has_features || nvlist_lookup_uint64(props,
+ zpool_prop_to_name(ZPOOL_PROP_VERSION), &version) != 0) {
version = SPA_VERSION;
- ASSERT(version <= SPA_VERSION);
+ }
+ ASSERT(SPA_VERSION_IS_SUPPORTED(version));
+
+ spa->spa_first_txg = txg;
+ spa->spa_uberblock.ub_txg = txg - 1;
spa->spa_uberblock.ub_version = version;
spa->spa_ubsync = spa->spa_uberblock;
(error = vdev_create(rvd, txg, B_FALSE)) == 0 &&
(error = spa_validate_aux(spa, nvroot, txg,
VDEV_ALLOC_ADD)) == 0) {
- for (int c = 0; c < rvd->vdev_children; c++) {
+ for (c = 0; c < rvd->vdev_children; c++) {
vdev_metaslab_set_size(rvd->vdev_child[c]);
vdev_expand(rvd->vdev_child[c], txg);
}
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
&spares, &nspares) == 0) {
VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, NV_UNIQUE_NAME,
- KM_SLEEP) == 0);
+ KM_PUSHPAGE) == 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_SLEEP) == 0);
+ NV_UNIQUE_NAME, KM_PUSHPAGE) == 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->spa_l2cache.sav_sync = B_TRUE;
}
+ spa->spa_is_initializing = B_TRUE;
spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, txg);
spa->spa_meta_objset = dp->dp_meta_objset;
+ spa->spa_is_initializing = B_FALSE;
+
+ /*
+ * Create DDTs (dedup tables).
+ */
+ ddt_create(spa);
+
+ spa_update_dspace(spa);
tx = dmu_tx_create_assigned(dp, txg);
cmn_err(CE_PANIC, "failed to add pool config");
}
+ if (spa_version(spa) >= SPA_VERSION_FEATURES)
+ spa_feature_create_zap_objects(spa, tx);
+
+ if (zap_add(spa->spa_meta_objset,
+ DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION,
+ sizeof (uint64_t), 1, &version, tx) != 0) {
+ cmn_err(CE_PANIC, "failed to add pool version");
+ }
+
/* Newly created pools with the right version are always deflated. */
if (version >= SPA_VERSION_RAIDZ_DEFLATE) {
spa->spa_deflate = TRUE;
}
/*
- * Create the deferred-free bplist object. Turn off compression
+ * Create the deferred-free bpobj. Turn off compression
* because sync-to-convergence takes longer if the blocksize
* keeps changing.
*/
- spa->spa_sync_bplist_obj = bplist_create(spa->spa_meta_objset,
- 1 << 14, tx);
- dmu_object_set_compress(spa->spa_meta_objset, spa->spa_sync_bplist_obj,
+ 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_BPLIST,
- sizeof (uint64_t), 1, &spa->spa_sync_bplist_obj, tx) != 0) {
- cmn_err(CE_PANIC, "failed to add bplist");
+ 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.
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, CRED(), tx);
+ spa_sync_props(spa, props, tx);
}
dmu_tx_commit(tx);
/*
* Put this pool's top-level vdevs into a root vdev.
*/
- VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ 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);
static void
spa_alt_rootvdev(vdev_t *vd, vdev_t **avd, uint64_t *txg)
{
- for (int c = 0; c < vd->vdev_children; c++)
+ 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) {
/*
* Read the label from the boot device and generate a configuration.
*/
- if ((config = spa_generate_rootconf(devpath, devid, &guid)) == NULL) {
- cmn_err(CE_NOTE, "Can not read the pool label from '%s'",
+ 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);
+ }
+ }
+#endif
+ if (config == NULL) {
+ cmn_err(CE_NOTE, "Cannot read the pool label from '%s'",
devpath);
return (EIO);
}
spa_remove(spa);
}
- spa = spa_add(pname, NULL);
+ spa = spa_add(pname, config, NULL);
spa->spa_is_root = B_TRUE;
- spa->spa_load_verbatim = B_TRUE;
+ spa->spa_import_flags = ZFS_IMPORT_VERBATIM;
/*
* Build up a vdev tree based on the boot device's label config.
!bvd->vdev_isspare) {
cmn_err(CE_NOTE, "The boot device is currently spared. Please "
"try booting from '%s'",
- bvd->vdev_parent->vdev_child[1]->vdev_path);
+ bvd->vdev_parent->
+ vdev_child[bvd->vdev_parent->vdev_children - 1]->vdev_path);
error = EINVAL;
goto out;
}
- VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0);
error = 0;
spa_history_log_version(spa, LOG_POOL_IMPORT);
out:
#endif
/*
- * Take a pool and insert it into the namespace as if it had been loaded at
- * boot.
- */
-int
-spa_import_verbatim(const char *pool, nvlist_t *config, nvlist_t *props)
-{
- spa_t *spa;
- char *altroot = NULL;
-
- mutex_enter(&spa_namespace_lock);
- if (spa_lookup(pool) != NULL) {
- mutex_exit(&spa_namespace_lock);
- return (EEXIST);
- }
-
- (void) nvlist_lookup_string(props,
- zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
- spa = spa_add(pool, altroot);
-
- spa->spa_load_verbatim = B_TRUE;
-
- VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0);
-
- if (props != NULL)
- spa_configfile_set(spa, props, B_FALSE);
-
- spa_config_sync(spa, B_FALSE, B_TRUE);
-
- mutex_exit(&spa_namespace_lock);
- spa_history_log_version(spa, LOG_POOL_IMPORT);
-
- return (0);
-}
-
-/*
* Import a non-root pool into the system.
*/
int
-spa_import(const char *pool, nvlist_t *config, nvlist_t *props)
+spa_import(const char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags)
{
spa_t *spa;
char *altroot = NULL;
+ spa_load_state_t state = SPA_LOAD_IMPORT;
+ zpool_rewind_policy_t policy;
+ uint64_t mode = spa_mode_global;
+ uint64_t readonly = B_FALSE;
int error;
nvlist_t *nvroot;
nvlist_t **spares, **l2cache;
* If a pool with this name exists, return failure.
*/
mutex_enter(&spa_namespace_lock);
- if ((spa = spa_lookup(pool)) != NULL) {
+ if (spa_lookup(pool) != NULL) {
mutex_exit(&spa_namespace_lock);
return (EEXIST);
}
*/
(void) nvlist_lookup_string(props,
zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
- spa = spa_add(pool, altroot);
- spa_activate(spa, spa_mode_global);
+ (void) nvlist_lookup_uint64(props,
+ zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
+ if (readonly)
+ mode = FREAD;
+ spa = spa_add(pool, config, altroot);
+ spa->spa_import_flags = flags;
+
+ /*
+ * Verbatim import - Take a pool and insert it into the namespace
+ * as if it had been loaded at boot.
+ */
+ if (spa->spa_import_flags & ZFS_IMPORT_VERBATIM) {
+ if (props != NULL)
+ spa_configfile_set(spa, props, B_FALSE);
+
+ spa_config_sync(spa, B_FALSE, B_TRUE);
+
+ mutex_exit(&spa_namespace_lock);
+ spa_history_log_version(spa, LOG_POOL_IMPORT);
+
+ return (0);
+ }
+
+ spa_activate(spa, mode);
/*
* Don't start async tasks until we know everything is healthy.
*/
spa_async_suspend(spa);
+ zpool_get_rewind_policy(config, &policy);
+ if (policy.zrp_request & ZPOOL_DO_REWIND)
+ state = SPA_LOAD_RECOVER;
+
/*
* Pass off the heavy lifting to spa_load(). Pass TRUE for mosconfig
* because the user-supplied config is actually the one to trust when
* doing an import.
*/
- error = spa_load(spa, config, SPA_LOAD_IMPORT, B_TRUE);
+ if (state != SPA_LOAD_RECOVER)
+ spa->spa_last_ubsync_txg = spa->spa_load_txg = 0;
+
+ error = spa_load_best(spa, state, B_TRUE, policy.zrp_txg,
+ policy.zrp_request);
+
+ /*
+ * Propagate anything learned while loading the pool and pass it
+ * back to caller (i.e. rewind info, missing devices, etc).
+ */
+ VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO,
+ spa->spa_load_info) == 0);
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
/*
ZPOOL_CONFIG_SPARES, DATA_TYPE_NVLIST_ARRAY) == 0);
else
VERIFY(nvlist_alloc(&spa->spa_spares.sav_config,
- NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ NV_UNIQUE_NAME, KM_PUSHPAGE) == 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_SLEEP) == 0);
+ NV_UNIQUE_NAME, KM_PUSHPAGE) == 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->spa_l2cache.sav_sync = B_TRUE;
}
+ /*
+ * Check for any removed devices.
+ */
+ if (spa->spa_autoreplace) {
+ spa_aux_check_removed(&spa->spa_spares);
+ spa_aux_check_removed(&spa->spa_l2cache);
+ }
+
if (spa_writeable(spa)) {
/*
* Update the config cache to include the newly-imported pool.
return (0);
}
-
-/*
- * This (illegal) pool name is used when temporarily importing a spa_t in order
- * to get the vdev stats associated with the imported devices.
- */
-#define TRYIMPORT_NAME "$import"
-
nvlist_t *
spa_tryimport(nvlist_t *tryconfig)
{
* Create and initialize the spa structure.
*/
mutex_enter(&spa_namespace_lock);
- spa = spa_add(TRYIMPORT_NAME, NULL);
+ spa = spa_add(TRYIMPORT_NAME, tryconfig, NULL);
spa_activate(spa, FREAD);
/*
* Pass TRUE for mosconfig because the user-supplied config
* is actually the one to trust when doing an import.
*/
- error = spa_load(spa, tryconfig, SPA_LOAD_TRYIMPORT, B_TRUE);
+ error = spa_load(spa, SPA_LOAD_TRYIMPORT, SPA_IMPORT_EXISTING, B_TRUE);
/*
* If 'tryconfig' was at least parsable, return the current config.
state) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
spa->spa_uberblock.ub_timestamp) == 0);
+ VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO,
+ spa->spa_load_info) == 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_SLEEP);
+ char *tmpname = kmem_alloc(MAXPATHLEN, KM_PUSHPAGE);
/*
* 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_SLEEP);
+ char *dsname = kmem_alloc(MAXPATHLEN, KM_PUSHPAGE);
cp = strchr(tmpname, '/');
if (cp == NULL) {
if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) {
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
spa->spa_state = new_state;
- spa->spa_final_txg = spa_last_synced_txg(spa) + 1;
+ spa->spa_final_txg = spa_last_synced_txg(spa) +
+ TXG_DEFER_SIZE + 1;
vdev_config_dirty(spa->spa_root_vdev);
spa_config_exit(spa, SCL_ALL, FTAG);
}
}
- spa_event_notify(spa, NULL, ESC_ZFS_POOL_DESTROY);
+ spa_event_notify(spa, NULL, FM_EREPORT_ZFS_POOL_DESTROY);
if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
spa_unload(spa);
int
spa_vdev_add(spa_t *spa, nvlist_t *nvroot)
{
- uint64_t txg;
+ uint64_t txg, id;
int error;
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *vd, *tvd;
nvlist_t **spares, **l2cache;
uint_t nspares, nl2cache;
+ int c;
+
+ ASSERT(spa_writeable(spa));
txg = spa_vdev_enter(spa);
/*
* Transfer each new top-level vdev from vd to rvd.
*/
- for (int c = 0; c < vd->vdev_children; c++) {
+ for (c = 0; c < vd->vdev_children; c++) {
+
+ /*
+ * Set the vdev id to the first hole, if one exists.
+ */
+ for (id = 0; id < rvd->vdev_children; id++) {
+ if (rvd->vdev_child[id]->vdev_ishole) {
+ vdev_free(rvd->vdev_child[id]);
+ break;
+ }
+ }
tvd = vd->vdev_child[c];
vdev_remove_child(vd, tvd);
- tvd->vdev_id = rvd->vdev_children;
+ tvd->vdev_id = id;
vdev_add_child(rvd, tvd);
vdev_config_dirty(tvd);
}
int
spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing)
{
- uint64_t txg, open_txg;
- vdev_t *rvd = spa->spa_root_vdev;
+ 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;
+ ASSERT(spa_writeable(spa));
+
txg = spa_vdev_enter(spa);
oldvd = spa_lookup_by_guid(spa, guid, B_FALSE);
pvd = oldvd->vdev_parent;
if ((error = spa_config_parse(spa, &newrootvd, nvroot, NULL, 0,
- VDEV_ALLOC_ADD)) != 0)
+ VDEV_ALLOC_ATTACH)) != 0)
return (spa_vdev_exit(spa, NULL, txg, EINVAL));
if (newrootvd->vdev_children != 1)
* spares.
*/
if (pvd->vdev_ops == &vdev_spare_ops &&
- pvd->vdev_child[1] == oldvd &&
+ oldvd->vdev_isspare &&
!spa_has_spare(spa, newvd->vdev_guid))
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
* the same (spare replaces spare, non-spare replaces
* non-spare).
*/
- if (pvd->vdev_ops == &vdev_replacing_ops)
+ if (pvd->vdev_ops == &vdev_replacing_ops &&
+ spa_version(spa) < SPA_VERSION_MULTI_REPLACE) {
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
- else if (pvd->vdev_ops == &vdev_spare_ops &&
- newvd->vdev_isspare != oldvd->vdev_isspare)
+ } else if (pvd->vdev_ops == &vdev_spare_ops &&
+ newvd->vdev_isspare != oldvd->vdev_isspare) {
return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
- else if (pvd->vdev_ops != &vdev_spare_ops &&
- newvd->vdev_isspare)
+ }
+
+ if (newvd->vdev_isspare)
pvops = &vdev_spare_ops;
else
pvops = &vdev_replacing_ops;
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_SLEEP);
+ KM_PUSHPAGE);
(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;
+
/*
* If the parent is not a mirror, or if we're replacing, insert the new
* mirror/replacing/spare vdev above oldvd.
*/
vdev_remove_child(newrootvd, newvd);
newvd->vdev_id = pvd->vdev_children;
+ newvd->vdev_crtxg = oldvd->vdev_crtxg;
vdev_add_child(pvd, newvd);
tvd = newvd->vdev_top;
vdev_config_dirty(tvd);
/*
- * Set newvd's DTL to [TXG_INITIAL, open_txg]. It will propagate
- * upward when spa_vdev_exit() calls vdev_dtl_reassess().
+ * Set newvd's DTL to [TXG_INITIAL, dtl_max_txg) so that we account
+ * for any dmu_sync-ed blocks. It will propagate upward when
+ * spa_vdev_exit() calls vdev_dtl_reassess().
*/
- open_txg = txg + TXG_CONCURRENT_STATES - 1;
+ dtl_max_txg = txg + TXG_CONCURRENT_STATES;
- vdev_dtl_dirty(newvd, DTL_MISSING,
- TXG_INITIAL, open_txg - TXG_INITIAL + 1);
+ vdev_dtl_dirty(newvd, DTL_MISSING, TXG_INITIAL,
+ dtl_max_txg - TXG_INITIAL);
if (newvd->vdev_isspare) {
spa_spare_activate(newvd);
- spa_event_notify(spa, newvd, ESC_ZFS_VDEV_SPARE);
+ spa_event_notify(spa, newvd, FM_EREPORT_ZFS_DEVICE_SPARE);
}
oldvdpath = spa_strdup(oldvd->vdev_path);
*/
vdev_dirty(tvd, VDD_DTL, newvd, txg);
- (void) spa_vdev_exit(spa, newrootvd, open_txg, 0);
+ /*
+ * Restart the resilver
+ */
+ dsl_resilver_restart(spa->spa_dsl_pool, dtl_max_txg);
+
+ /*
+ * Commit the config
+ */
+ (void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0);
- spa_history_internal_log(LOG_POOL_VDEV_ATTACH, spa, NULL,
- CRED(), "%s vdev=%s %s vdev=%s",
+ spa_history_log_internal(LOG_POOL_VDEV_ATTACH, spa, NULL,
+ "%s vdev=%s %s vdev=%s",
replacing && newvd_isspare ? "spare in" :
replacing ? "replace" : "attach", newvdpath,
replacing ? "for" : "to", oldvdpath);
spa_strfree(oldvdpath);
spa_strfree(newvdpath);
- /*
- * Kick off a resilver to update newvd.
- */
- VERIFY3U(spa_scrub(spa, POOL_SCRUB_RESILVER), ==, 0);
+ if (spa->spa_bootfs)
+ spa_event_notify(spa, newvd, FM_EREPORT_ZFS_BOOTFS_VDEV_ATTACH);
return (0);
}
{
uint64_t txg;
int error;
- vdev_t *rvd = spa->spa_root_vdev;
+ ASSERTV(vdev_t *rvd = spa->spa_root_vdev;)
vdev_t *vd, *pvd, *cvd, *tvd;
boolean_t unspare = B_FALSE;
- uint64_t unspare_guid;
- size_t len;
+ uint64_t unspare_guid = 0;
+ char *vdpath;
+ int c, t;
+
+ ASSERT(spa_writeable(spa));
txg = spa_vdev_enter(spa);
return (spa_vdev_exit(spa, NULL, txg, EBUSY));
/*
- * If replace_done is specified, only remove this device if it's
- * the first child of a replacing vdev. For the 'spare' vdev, either
- * disk can be removed.
+ * Only 'replacing' or 'spare' vdevs can be replaced.
*/
- if (replace_done) {
- if (pvd->vdev_ops == &vdev_replacing_ops) {
- if (vd->vdev_id != 0)
- return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
- } else if (pvd->vdev_ops != &vdev_spare_ops) {
- return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
- }
- }
+ if (replace_done && pvd->vdev_ops != &vdev_replacing_ops &&
+ pvd->vdev_ops != &vdev_spare_ops)
+ return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
ASSERT(pvd->vdev_ops != &vdev_spare_ops ||
spa_version(spa) >= SPA_VERSION_SPARES);
* check to see if we changed the original vdev's path to have "/old"
* at the end in spa_vdev_attach(). If so, undo that change now.
*/
- if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id == 1 &&
- pvd->vdev_child[0]->vdev_path != NULL &&
- pvd->vdev_child[1]->vdev_path != NULL) {
- ASSERT(pvd->vdev_child[1] == vd);
- cvd = pvd->vdev_child[0];
- len = strlen(vd->vdev_path);
- if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 &&
- strcmp(cvd->vdev_path + len, "/old") == 0) {
- spa_strfree(cvd->vdev_path);
- cvd->vdev_path = spa_strdup(vd->vdev_path);
+ if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id > 0 &&
+ vd->vdev_path != NULL) {
+ size_t len = strlen(vd->vdev_path);
+
+ for (c = 0; c < pvd->vdev_children; c++) {
+ cvd = pvd->vdev_child[c];
+
+ if (cvd == vd || cvd->vdev_path == NULL)
+ continue;
+
+ if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 &&
+ strcmp(cvd->vdev_path + len, "/old") == 0) {
+ spa_strfree(cvd->vdev_path);
+ cvd->vdev_path = spa_strdup(vd->vdev_path);
+ break;
+ }
}
}
* active spare list for the pool.
*/
if (pvd->vdev_ops == &vdev_spare_ops &&
- vd->vdev_id == 0 && pvd->vdev_child[1]->vdev_isspare)
+ vd->vdev_id == 0 &&
+ pvd->vdev_child[pvd->vdev_children - 1]->vdev_isspare)
unspare = B_TRUE;
/*
/*
* Remember one of the remaining children so we can get tvd below.
*/
- cvd = pvd->vdev_child[0];
+ cvd = pvd->vdev_child[pvd->vdev_children - 1];
/*
* If we need to remove the remaining child from the list of hot spares,
spa_spare_remove(cvd);
unspare_guid = cvd->vdev_guid;
(void) spa_vdev_remove(spa, unspare_guid, B_TRUE);
+ cvd->vdev_unspare = B_TRUE;
}
/*
* If the parent mirror/replacing vdev only has one child,
* the parent is no longer needed. Remove it from the tree.
*/
- if (pvd->vdev_children == 1)
+ if (pvd->vdev_children == 1) {
+ if (pvd->vdev_ops == &vdev_spare_ops)
+ cvd->vdev_unspare = B_FALSE;
vdev_remove_parent(cvd);
+ cvd->vdev_resilvering = B_FALSE;
+ }
+
/*
* We don't set tvd until now because the parent we just removed
* But first make sure we're not on any *other* txg's DTL list, to
* prevent vd from being accessed after it's freed.
*/
- for (int t = 0; t < TXG_SIZE; t++)
+ vdpath = spa_strdup(vd->vdev_path);
+ 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, ESC_ZFS_VDEV_REMOVE);
+ spa_event_notify(spa, vd, FM_EREPORT_ZFS_DEVICE_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,
+ "vdev=%s", vdpath);
+ spa_strfree(vdpath);
+
+ /*
+ * If this was the removal of the original device in a hot spare vdev,
+ * then we want to go through and remove the device from the hot spare
+ * list of every other pool.
+ */
+ if (unspare) {
+ spa_t *altspa = NULL;
+
+ mutex_enter(&spa_namespace_lock);
+ while ((altspa = spa_next(altspa)) != NULL) {
+ if (altspa->spa_state != POOL_STATE_ACTIVE ||
+ altspa == spa)
+ continue;
+
+ spa_open_ref(altspa, FTAG);
+ mutex_exit(&spa_namespace_lock);
+ (void) spa_vdev_remove(altspa, unspare_guid, B_TRUE);
+ mutex_enter(&spa_namespace_lock);
+ spa_close(altspa, FTAG);
+ }
+ mutex_exit(&spa_namespace_lock);
+
+ /* search the rest of the vdevs for spares to remove */
+ spa_vdev_resilver_done(spa);
+ }
+
+ /* all done with the spa; OK to release */
+ mutex_enter(&spa_namespace_lock);
+ spa_close(spa, FTAG);
+ mutex_exit(&spa_namespace_lock);
+
+ return (error);
+}
+
+/*
+ * Split a set of devices from their mirrors, and create a new pool from them.
+ */
+int
+spa_vdev_split_mirror(spa_t *spa, char *newname, nvlist_t *config,
+ nvlist_t *props, boolean_t exp)
+{
+ int error = 0;
+ uint64_t txg, *glist;
+ spa_t *newspa;
+ uint_t c, children, lastlog;
+ nvlist_t **child, *nvl, *tmp;
+ dmu_tx_t *tx;
+ char *altroot = NULL;
+ vdev_t *rvd, **vml = NULL; /* vdev modify list */
+ boolean_t activate_slog;
+
+ ASSERT(spa_writeable(spa));
+
+ txg = spa_vdev_enter(spa);
+
+ /* clear the log and flush everything up to now */
+ activate_slog = spa_passivate_log(spa);
+ (void) spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);
+ error = spa_offline_log(spa);
+ txg = spa_vdev_config_enter(spa);
+
+ if (activate_slog)
+ spa_activate_log(spa);
+
+ if (error != 0)
+ return (spa_vdev_exit(spa, NULL, txg, error));
+
+ /* check new spa name before going any further */
+ if (spa_lookup(newname) != NULL)
+ return (spa_vdev_exit(spa, NULL, txg, EEXIST));
+
+ /*
+ * scan through all the children to ensure they're all mirrors
+ */
+ if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvl) != 0 ||
+ nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN, &child,
+ &children) != 0)
+ return (spa_vdev_exit(spa, NULL, txg, EINVAL));
+
+ /* first, check to ensure we've got the right child count */
+ rvd = spa->spa_root_vdev;
+ lastlog = 0;
+ for (c = 0; c < rvd->vdev_children; c++) {
+ vdev_t *vd = rvd->vdev_child[c];
+
+ /* don't count the holes & logs as children */
+ if (vd->vdev_islog || vd->vdev_ishole) {
+ if (lastlog == 0)
+ lastlog = c;
+ continue;
+ }
+
+ lastlog = 0;
+ }
+ if (children != (lastlog != 0 ? lastlog : rvd->vdev_children))
+ return (spa_vdev_exit(spa, NULL, txg, EINVAL));
+
+ /* next, ensure no spare or cache devices are part of the split */
+ if (nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_SPARES, &tmp) == 0 ||
+ nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0)
+ return (spa_vdev_exit(spa, NULL, txg, EINVAL));
+
+ vml = kmem_zalloc(children * sizeof (vdev_t *), KM_PUSHPAGE);
+ glist = kmem_zalloc(children * sizeof (uint64_t), KM_PUSHPAGE);
+
+ /* then, loop over each vdev and validate it */
+ for (c = 0; c < children; c++) {
+ uint64_t is_hole = 0;
+
+ (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
+ &is_hole);
+
+ if (is_hole != 0) {
+ if (spa->spa_root_vdev->vdev_child[c]->vdev_ishole ||
+ spa->spa_root_vdev->vdev_child[c]->vdev_islog) {
+ continue;
+ } else {
+ error = EINVAL;
+ break;
+ }
+ }
+
+ /* which disk is going to be split? */
+ if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID,
+ &glist[c]) != 0) {
+ 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;
+ break;
+ }
+
+ /* make sure there's nothing stopping the split */
+ if (vml[c]->vdev_parent->vdev_ops != &vdev_mirror_ops ||
+ vml[c]->vdev_islog ||
+ vml[c]->vdev_ishole ||
+ vml[c]->vdev_isspare ||
+ vml[c]->vdev_isl2cache ||
+ !vdev_writeable(vml[c]) ||
+ vml[c]->vdev_children != 0 ||
+ vml[c]->vdev_state != VDEV_STATE_HEALTHY ||
+ c != spa->spa_root_vdev->vdev_child[c]->vdev_id) {
+ error = EINVAL;
+ break;
+ }
+
+ if (vdev_dtl_required(vml[c])) {
+ error = EBUSY;
+ break;
+ }
+
+ /* we need certain info from the top level */
+ VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_ARRAY,
+ vml[c]->vdev_top->vdev_ms_array) == 0);
+ VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_SHIFT,
+ vml[c]->vdev_top->vdev_ms_shift) == 0);
+ VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASIZE,
+ vml[c]->vdev_top->vdev_asize) == 0);
+ VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT,
+ vml[c]->vdev_top->vdev_ashift) == 0);
+ }
+
+ if (error != 0) {
+ kmem_free(vml, children * sizeof (vdev_t *));
+ kmem_free(glist, children * sizeof (uint64_t));
+ return (spa_vdev_exit(spa, NULL, txg, error));
+ }
+
+ /* stop writers from using the disks */
+ for (c = 0; c < children; c++) {
+ if (vml[c] != NULL)
+ vml[c]->vdev_offline = B_TRUE;
+ }
+ vdev_reopen(spa->spa_root_vdev);
+
+ /*
+ * Temporarily record the splitting vdevs in the spa config. This
+ * will disappear once the config is regenerated.
+ */
+ VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME, KM_PUSHPAGE) == 0);
+ VERIFY(nvlist_add_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST,
+ glist, children) == 0);
+ kmem_free(glist, children * sizeof (uint64_t));
+
+ mutex_enter(&spa->spa_props_lock);
+ VERIFY(nvlist_add_nvlist(spa->spa_config, ZPOOL_CONFIG_SPLIT,
+ nvl) == 0);
+ mutex_exit(&spa->spa_props_lock);
+ spa->spa_config_splitting = nvl;
+ vdev_config_dirty(spa->spa_root_vdev);
+
+ /* configure and create the new pool */
+ VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, newname) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
+ exp ? POOL_STATE_EXPORTED : POOL_STATE_ACTIVE) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
+ spa_version(spa)) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
+ spa->spa_config_txg) == 0);
+ VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
+ spa_generate_guid(NULL)) == 0);
+ (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_config_txg = spa->spa_config_txg;
+ spa_set_log_state(newspa, SPA_LOG_CLEAR);
+
+ /* release the spa config lock, retaining the namespace lock */
+ spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);
+
+ if (zio_injection_enabled)
+ zio_handle_panic_injection(spa, FTAG, 1);
+
+ spa_activate(newspa, spa_mode_global);
+ spa_async_suspend(newspa);
+
+ /* create the new pool from the disks of the original pool */
+ error = spa_load(newspa, SPA_LOAD_IMPORT, SPA_IMPORT_ASSEMBLE, B_TRUE);
+ if (error)
+ goto out;
+
+ /* 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);
+ 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,
+ B_TRUE));
+ }
+
+ /* set the props */
+ if (props != NULL) {
+ spa_configfile_set(newspa, props, B_FALSE);
+ error = spa_prop_set(newspa, props);
+ if (error)
+ goto out;
+ }
+
+ /* flush everything */
+ txg = spa_vdev_config_enter(newspa);
+ vdev_config_dirty(newspa->spa_root_vdev);
+ (void) spa_vdev_config_exit(newspa, NULL, txg, 0, FTAG);
+
+ if (zio_injection_enabled)
+ zio_handle_panic_injection(spa, FTAG, 2);
+
+ spa_async_resume(newspa);
+
+ /* finally, update the original pool's config */
+ txg = spa_vdev_config_enter(spa);
+ tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error != 0)
+ dmu_tx_abort(tx);
+ for (c = 0; c < children; c++) {
+ if (vml[c] != NULL) {
+ vdev_split(vml[c]);
+ if (error == 0)
+ spa_history_log_internal(LOG_POOL_VDEV_DETACH,
+ spa, tx, "vdev=%s",
+ vml[c]->vdev_path);
+ vdev_free(vml[c]);
+ }
+ }
+ vdev_config_dirty(spa->spa_root_vdev);
+ spa->spa_config_splitting = NULL;
+ nvlist_free(nvl);
+ if (error == 0)
+ dmu_tx_commit(tx);
+ (void) spa_vdev_exit(spa, NULL, txg, 0);
+
+ if (zio_injection_enabled)
+ zio_handle_panic_injection(spa, FTAG, 3);
- error = spa_vdev_exit(spa, vd, txg, 0);
+ /* 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));
- /*
- * If this was the removal of the original device in a hot spare vdev,
- * then we want to go through and remove the device from the hot spare
- * list of every other pool.
- */
- if (unspare) {
- spa_t *myspa = spa;
- spa = NULL;
- mutex_enter(&spa_namespace_lock);
- while ((spa = spa_next(spa)) != NULL) {
- if (spa->spa_state != POOL_STATE_ACTIVE)
- continue;
- if (spa == myspa)
- continue;
- spa_open_ref(spa, FTAG);
- mutex_exit(&spa_namespace_lock);
- (void) spa_vdev_remove(spa, unspare_guid, B_TRUE);
- mutex_enter(&spa_namespace_lock);
- spa_close(spa, FTAG);
- }
- mutex_exit(&spa_namespace_lock);
+ kmem_free(vml, children * sizeof (vdev_t *));
+
+ /* if we're not going to mount the filesystems in userland, export */
+ if (exp)
+ error = spa_export_common(newname, POOL_STATE_EXPORTED, NULL,
+ B_FALSE, B_FALSE);
+
+ return (error);
+
+out:
+ spa_unload(newspa);
+ spa_deactivate(newspa);
+ spa_remove(newspa);
+
+ txg = spa_vdev_config_enter(spa);
+
+ /* re-online all offlined disks */
+ for (c = 0; c < children; c++) {
+ if (vml[c] != NULL)
+ vml[c]->vdev_offline = B_FALSE;
}
+ vdev_reopen(spa->spa_root_vdev);
+ nvlist_free(spa->spa_config_splitting);
+ spa->spa_config_splitting = NULL;
+ (void) spa_vdev_exit(spa, NULL, txg, error);
+
+ kmem_free(vml, children * sizeof (vdev_t *));
return (error);
}
static nvlist_t *
spa_nvlist_lookup_by_guid(nvlist_t **nvpp, int count, uint64_t target_guid)
{
- for (int i = 0; i < count; i++) {
+ int i;
+
+ for (i = 0; i < count; i++) {
uint64_t guid;
VERIFY(nvlist_lookup_uint64(nvpp[i], ZPOOL_CONFIG_GUID,
nvlist_t *dev_to_remove)
{
nvlist_t **newdev = NULL;
+ int i, j;
if (count > 1)
- newdev = kmem_alloc((count - 1) * sizeof (void *), KM_SLEEP);
+ newdev = kmem_alloc((count - 1) * sizeof (void *), KM_PUSHPAGE);
- for (int i = 0, j = 0; i < count; i++) {
+ for (i = 0, j = 0; i < count; i++) {
if (dev[i] == dev_to_remove)
continue;
- VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_SLEEP) == 0);
+ VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_PUSHPAGE) == 0);
}
VERIFY(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY) == 0);
VERIFY(nvlist_add_nvlist_array(config, name, newdev, count - 1) == 0);
- for (int i = 0; i < count - 1; i++)
+ for (i = 0; i < count - 1; i++)
nvlist_free(newdev[i]);
if (count > 1)
}
/*
+ * Evacuate the device.
+ */
+static int
+spa_vdev_remove_evacuate(spa_t *spa, vdev_t *vd)
+{
+ uint64_t txg;
+ int error = 0;
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
+ ASSERT(vd == vd->vdev_top);
+
+ /*
+ * Evacuate the device. We don't hold the config lock as writer
+ * since we need to do I/O but we do keep the
+ * spa_namespace_lock held. Once this completes the device
+ * should no longer have any blocks allocated on it.
+ */
+ if (vd->vdev_islog) {
+ if (vd->vdev_stat.vs_alloc != 0)
+ error = spa_offline_log(spa);
+ } else {
+ error = ENOTSUP;
+ }
+
+ if (error)
+ return (error);
+
+ /*
+ * The evacuation succeeded. Remove any remaining MOS metadata
+ * associated with this vdev, and wait for these changes to sync.
+ */
+ ASSERT0(vd->vdev_stat.vs_alloc);
+ txg = spa_vdev_config_enter(spa);
+ vd->vdev_removing = B_TRUE;
+ vdev_dirty(vd, 0, NULL, txg);
+ vdev_config_dirty(vd);
+ spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);
+
+ return (0);
+}
+
+/*
+ * Complete the removal by cleaning up the namespace.
+ */
+static void
+spa_vdev_remove_from_namespace(spa_t *spa, vdev_t *vd)
+{
+ vdev_t *rvd = spa->spa_root_vdev;
+ uint64_t id = vd->vdev_id;
+ boolean_t last_vdev = (id == (rvd->vdev_children - 1));
+
+ ASSERT(MUTEX_HELD(&spa_namespace_lock));
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
+ ASSERT(vd == vd->vdev_top);
+
+ /*
+ * Only remove any devices which are empty.
+ */
+ if (vd->vdev_stat.vs_alloc != 0)
+ return;
+
+ (void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE);
+
+ if (list_link_active(&vd->vdev_state_dirty_node))
+ vdev_state_clean(vd);
+ if (list_link_active(&vd->vdev_config_dirty_node))
+ vdev_config_clean(vd);
+
+ vdev_free(vd);
+
+ if (last_vdev) {
+ vdev_compact_children(rvd);
+ } else {
+ vd = vdev_alloc_common(spa, id, 0, &vdev_hole_ops);
+ vdev_add_child(rvd, vd);
+ }
+ vdev_config_dirty(rvd);
+
+ /*
+ * Reassess the health of our root vdev.
+ */
+ vdev_reopen(rvd);
+}
+
+/*
+ * Remove a device from the pool -
+ *
+ * Removing a device from the vdev namespace requires several steps
+ * and can take a significant amount of time. As a result we use
+ * 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 and level 2 ARC devices.
+ * spares, slogs, and level 2 ARC devices.
*/
int
spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare)
{
vdev_t *vd;
+ metaslab_group_t *mg;
nvlist_t **spares, **l2cache, *nv;
- uint_t nspares, nl2cache;
uint64_t txg = 0;
+ uint_t nspares, nl2cache;
int error = 0;
boolean_t locked = MUTEX_HELD(&spa_namespace_lock);
+ ASSERT(spa_writeable(spa));
+
if (!locked)
txg = spa_vdev_enter(spa);
ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv);
spa_load_l2cache(spa);
spa->spa_l2cache.sav_sync = B_TRUE;
+ } else if (vd != NULL && vd->vdev_islog) {
+ ASSERT(!locked);
+ ASSERT(vd == vd->vdev_top);
+
+ /*
+ * XXX - Once we have bp-rewrite this should
+ * become the common case.
+ */
+
+ mg = vd->vdev_mg;
+
+ /*
+ * Stop allocating from this vdev.
+ */
+ metaslab_group_passivate(mg);
+
+ /*
+ * Wait for the youngest allocations and frees to sync,
+ * and then wait for the deferral of those frees to finish.
+ */
+ spa_vdev_config_exit(spa, NULL,
+ txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG);
+
+ /*
+ * Attempt to evacuate the vdev.
+ */
+ error = spa_vdev_remove_evacuate(spa, vd);
+
+ txg = spa_vdev_config_enter(spa);
+
+ /*
+ * If we couldn't evacuate the vdev, unwind.
+ */
+ if (error) {
+ metaslab_group_activate(mg);
+ return (spa_vdev_exit(spa, NULL, txg, error));
+ }
+
+ /*
+ * Clean up the vdev namespace.
+ */
+ spa_vdev_remove_from_namespace(spa, vd);
+
} else if (vd != NULL) {
/*
* Normal vdevs cannot be removed (yet).
spa_vdev_resilver_done_hunt(vdev_t *vd)
{
vdev_t *newvd, *oldvd;
+ int c;
- for (int c = 0; c < vd->vdev_children; c++) {
+ for (c = 0; c < vd->vdev_children; c++) {
oldvd = spa_vdev_resilver_done_hunt(vd->vdev_child[c]);
if (oldvd != NULL)
return (oldvd);
}
/*
- * Check for a completed replacement.
+ * Check for a completed replacement. We always consider the first
+ * vdev in the list to be the oldest vdev, and the last one to be
+ * the newest (see spa_vdev_attach() for how that works). In
+ * the case where the newest vdev is faulted, we will not automatically
+ * remove it after a resilver completes. This is OK as it will require
+ * user intervention to determine which disk the admin wishes to keep.
*/
- if (vd->vdev_ops == &vdev_replacing_ops && vd->vdev_children == 2) {
+ if (vd->vdev_ops == &vdev_replacing_ops) {
+ ASSERT(vd->vdev_children > 1);
+
+ newvd = vd->vdev_child[vd->vdev_children - 1];
oldvd = vd->vdev_child[0];
- newvd = vd->vdev_child[1];
if (vdev_dtl_empty(newvd, DTL_MISSING) &&
+ vdev_dtl_empty(newvd, DTL_OUTAGE) &&
!vdev_dtl_required(oldvd))
return (oldvd);
}
/*
* Check for a completed resilver with the 'unspare' flag set.
*/
- if (vd->vdev_ops == &vdev_spare_ops && vd->vdev_children == 2) {
- newvd = vd->vdev_child[0];
- oldvd = vd->vdev_child[1];
+ if (vd->vdev_ops == &vdev_spare_ops) {
+ vdev_t *first = vd->vdev_child[0];
+ vdev_t *last = vd->vdev_child[vd->vdev_children - 1];
+
+ if (last->vdev_unspare) {
+ oldvd = first;
+ newvd = last;
+ } else if (first->vdev_unspare) {
+ oldvd = last;
+ newvd = first;
+ } else {
+ oldvd = NULL;
+ }
- if (newvd->vdev_unspare &&
+ if (oldvd != NULL &&
vdev_dtl_empty(newvd, DTL_MISSING) &&
- !vdev_dtl_required(oldvd)) {
- newvd->vdev_unspare = 0;
+ vdev_dtl_empty(newvd, DTL_OUTAGE) &&
+ !vdev_dtl_required(oldvd))
return (oldvd);
+
+ /*
+ * If there are more than two spares attached to a disk,
+ * and those spares are not required, then we want to
+ * attempt to free them up now so that they can be used
+ * by other pools. Once we're back down to a single
+ * disk+spare, we stop removing them.
+ */
+ if (vd->vdev_children > 2) {
+ newvd = vd->vdev_child[1];
+
+ if (newvd->vdev_isspare && last->vdev_isspare &&
+ vdev_dtl_empty(last, DTL_MISSING) &&
+ vdev_dtl_empty(last, DTL_OUTAGE) &&
+ !vdev_dtl_required(newvd))
+ return (newvd);
}
}
* we need to detach the parent's first child (the original hot
* spare) as well.
*/
- if (ppvd->vdev_ops == &vdev_spare_ops && pvd->vdev_id == 0) {
+ if (ppvd->vdev_ops == &vdev_spare_ops && pvd->vdev_id == 0 &&
+ ppvd->vdev_children == 2) {
ASSERT(pvd->vdev_ops == &vdev_replacing_ops);
- ASSERT(ppvd->vdev_children == 2);
sguid = ppvd->vdev_child[1]->vdev_guid;
}
spa_config_exit(spa, SCL_ALL, FTAG);
}
/*
- * Update the stored path or FRU for this vdev. Dirty the vdev configuration,
- * relying on spa_vdev_enter/exit() to synchronize the labels and cache.
+ * Update the stored path or FRU for this vdev.
*/
int
spa_vdev_set_common(spa_t *spa, uint64_t guid, const char *value,
boolean_t ispath)
{
vdev_t *vd;
- uint64_t txg;
+ boolean_t sync = B_FALSE;
- txg = spa_vdev_enter(spa);
+ ASSERT(spa_writeable(spa));
+
+ spa_vdev_state_enter(spa, SCL_ALL);
if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
- return (spa_vdev_exit(spa, NULL, txg, ENOENT));
+ return (spa_vdev_state_exit(spa, NULL, ENOENT));
if (!vd->vdev_ops->vdev_op_leaf)
- return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
+ return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
if (ispath) {
- spa_strfree(vd->vdev_path);
- vd->vdev_path = spa_strdup(value);
+ if (strcmp(value, vd->vdev_path) != 0) {
+ spa_strfree(vd->vdev_path);
+ vd->vdev_path = spa_strdup(value);
+ sync = B_TRUE;
+ }
} else {
- if (vd->vdev_fru != NULL)
+ if (vd->vdev_fru == NULL) {
+ vd->vdev_fru = spa_strdup(value);
+ sync = B_TRUE;
+ } else if (strcmp(value, vd->vdev_fru) != 0) {
spa_strfree(vd->vdev_fru);
- vd->vdev_fru = spa_strdup(value);
+ vd->vdev_fru = spa_strdup(value);
+ sync = B_TRUE;
+ }
}
- vdev_config_dirty(vd->vdev_top);
-
- return (spa_vdev_exit(spa, NULL, txg, 0));
+ return (spa_vdev_state_exit(spa, sync ? vd : NULL, 0));
}
int
/*
* ==========================================================================
- * SPA Scrubbing
+ * SPA Scanning
* ==========================================================================
*/
int
-spa_scrub(spa_t *spa, pool_scrub_type_t type)
+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 (dsl_scan_cancel(spa->spa_dsl_pool));
+}
+
+int
+spa_scan(spa_t *spa, pool_scan_func_t func)
{
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
- if ((uint_t)type >= POOL_SCRUB_TYPES)
+ if (func >= POOL_SCAN_FUNCS || func == POOL_SCAN_NONE)
return (ENOTSUP);
/*
* If a resilver was requested, but there is no DTL on a
* writeable leaf device, we have nothing to do.
*/
- if (type == POOL_SCRUB_RESILVER &&
+ if (func == POOL_SCAN_RESILVER &&
!vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) {
spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);
return (0);
}
- if (type == POOL_SCRUB_EVERYTHING &&
- spa->spa_dsl_pool->dp_scrub_func != SCRUB_FUNC_NONE &&
- spa->spa_dsl_pool->dp_scrub_isresilver)
- return (EBUSY);
-
- if (type == POOL_SCRUB_EVERYTHING || type == POOL_SCRUB_RESILVER) {
- return (dsl_pool_scrub_clean(spa->spa_dsl_pool));
- } else if (type == POOL_SCRUB_NONE) {
- return (dsl_pool_scrub_cancel(spa->spa_dsl_pool));
- } else {
- return (EINVAL);
- }
+ return (dsl_scan(spa->spa_dsl_pool, func));
}
/*
static void
spa_async_remove(spa_t *spa, vdev_t *vd)
{
+ int c;
+
if (vd->vdev_remove_wanted) {
- vd->vdev_remove_wanted = 0;
+ vd->vdev_remove_wanted = B_FALSE;
+ vd->vdev_delayed_close = B_FALSE;
vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE);
- vdev_clear(spa, vd);
+
+ /*
+ * We want to clear the stats, but we don't want to do a full
+ * vdev_clear() as that will cause us to throw away
+ * degraded/faulted state as well as attempt to reopen the
+ * device, all of which is a waste.
+ */
+ vd->vdev_stat.vs_read_errors = 0;
+ vd->vdev_stat.vs_write_errors = 0;
+ vd->vdev_stat.vs_checksum_errors = 0;
+
vdev_state_dirty(vd->vdev_top);
}
- for (int c = 0; c < vd->vdev_children; c++)
+ for (c = 0; c < vd->vdev_children; c++)
spa_async_remove(spa, vd->vdev_child[c]);
}
static void
spa_async_probe(spa_t *spa, vdev_t *vd)
{
+ int c;
+
if (vd->vdev_probe_wanted) {
- vd->vdev_probe_wanted = 0;
+ vd->vdev_probe_wanted = B_FALSE;
vdev_reopen(vd); /* vdev_open() does the actual probe */
}
- for (int c = 0; c < vd->vdev_children; c++)
+ for (c = 0; c < vd->vdev_children; c++)
spa_async_probe(spa, vd->vdev_child[c]);
}
static void
spa_async_autoexpand(spa_t *spa, vdev_t *vd)
{
- sysevent_id_t eid;
- nvlist_t *attr;
- char *physpath;
+ int c;
if (!spa->spa_autoexpand)
return;
- for (int c = 0; c < vd->vdev_children; c++) {
+ for (c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
spa_async_autoexpand(spa, cvd);
}
if (!vd->vdev_ops->vdev_op_leaf || vd->vdev_physpath == NULL)
return;
- physpath = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
- (void) snprintf(physpath, MAXPATHLEN, "/devices%s", vd->vdev_physpath);
-
- VERIFY(nvlist_alloc(&attr, NV_UNIQUE_NAME, KM_SLEEP) == 0);
- VERIFY(nvlist_add_string(attr, DEV_PHYS_PATH, physpath) == 0);
-
- (void) ddi_log_sysevent(zfs_dip, SUNW_VENDOR, EC_DEV_STATUS,
- ESC_DEV_DLE, attr, &eid, DDI_SLEEP);
-
- nvlist_free(attr);
- kmem_free(physpath, MAXPATHLEN);
+ spa_event_notify(vd->vdev_spa, vd, FM_EREPORT_ZFS_DEVICE_AUTOEXPAND);
}
static void
spa_async_thread(spa_t *spa)
{
- int tasks;
+ int tasks, i;
ASSERT(spa->spa_sync_on);
* See if the config needs to be updated.
*/
if (tasks & SPA_ASYNC_CONFIG_UPDATE) {
- uint64_t oldsz, space_update;
+ uint64_t old_space, new_space;
mutex_enter(&spa_namespace_lock);
- oldsz = spa_get_space(spa);
+ old_space = metaslab_class_get_space(spa_normal_class(spa));
spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
- space_update = spa_get_space(spa) - oldsz;
+ new_space = metaslab_class_get_space(spa_normal_class(spa));
mutex_exit(&spa_namespace_lock);
/*
* If the pool grew as a result of the config update,
* then log an internal history event.
*/
- if (space_update) {
- spa_history_internal_log(LOG_POOL_VDEV_ONLINE,
- spa, NULL, CRED(),
+ if (new_space != old_space) {
+ spa_history_log_internal(LOG_POOL_VDEV_ONLINE,
+ spa, NULL,
"pool '%s' size: %llu(+%llu)",
- spa_name(spa), spa_get_space(spa),
- space_update);
+ spa_name(spa), new_space, new_space - old_space);
}
}
* See if any devices need to be marked REMOVED.
*/
if (tasks & SPA_ASYNC_REMOVE) {
- spa_vdev_state_enter(spa);
+ spa_vdev_state_enter(spa, SCL_NONE);
spa_async_remove(spa, spa->spa_root_vdev);
- for (int i = 0; i < spa->spa_l2cache.sav_count; i++)
+ for (i = 0; i < spa->spa_l2cache.sav_count; i++)
spa_async_remove(spa, spa->spa_l2cache.sav_vdevs[i]);
- for (int i = 0; i < spa->spa_spares.sav_count; i++)
+ for (i = 0; i < spa->spa_spares.sav_count; i++)
spa_async_remove(spa, spa->spa_spares.sav_vdevs[i]);
(void) spa_vdev_state_exit(spa, NULL, 0);
}
* See if any devices need to be probed.
*/
if (tasks & SPA_ASYNC_PROBE) {
- spa_vdev_state_enter(spa);
+ spa_vdev_state_enter(spa, SCL_NONE);
spa_async_probe(spa, spa->spa_root_vdev);
(void) spa_vdev_state_exit(spa, NULL, 0);
}
* Kick off a resilver.
*/
if (tasks & SPA_ASYNC_RESILVER)
- VERIFY(spa_scrub(spa, POOL_SCRUB_RESILVER) == 0);
+ dsl_resilver_restart(spa->spa_dsl_pool, 0);
/*
* Let the world know that we're done.
void
spa_async_request(spa_t *spa, int task)
{
+ zfs_dbgmsg("spa=%s async request task=%u", spa->spa_name, task);
mutex_enter(&spa->spa_async_lock);
spa->spa_async_tasks |= task;
mutex_exit(&spa->spa_async_lock);
* ==========================================================================
*/
-static void
-spa_sync_deferred_frees(spa_t *spa, uint64_t txg)
+static int
+bpobj_enqueue_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
{
- bplist_t *bpl = &spa->spa_sync_bplist;
- dmu_tx_t *tx;
- blkptr_t blk;
- uint64_t itor = 0;
- zio_t *zio;
- int error;
- uint8_t c = 1;
-
- zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
-
- while (bplist_iterate(bpl, &itor, &blk) == 0) {
- ASSERT(blk.blk_birth < txg);
- zio_nowait(zio_free(zio, spa, txg, &blk, NULL, NULL,
- ZIO_FLAG_MUSTSUCCEED));
- }
-
- error = zio_wait(zio);
- ASSERT3U(error, ==, 0);
+ bpobj_t *bpo = arg;
+ bpobj_enqueue(bpo, bp, tx);
+ return (0);
+}
- tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
- bplist_vacate(bpl, tx);
+static int
+spa_free_sync_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
+{
+ zio_t *zio = arg;
- /*
- * Pre-dirty the first block so we sync to convergence faster.
- * (Usually only the first block is needed.)
- */
- dmu_write(spa->spa_meta_objset, spa->spa_sync_bplist_obj, 0, 1, &c, tx);
- dmu_tx_commit(tx);
+ zio_nowait(zio_free_sync(zio, zio->io_spa, dmu_tx_get_txg(tx), bp,
+ zio->io_flags));
+ return (0);
}
static void
* information. This avoids the dbuf_will_dirty() path and
* saves us a pre-read to get data we don't actually care about.
*/
- bufsize = P2ROUNDUP(nvsize, SPA_CONFIG_BLOCKSIZE);
- packed = kmem_alloc(bufsize, KM_SLEEP);
+ bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE);
+ packed = vmem_alloc(bufsize, KM_PUSHPAGE);
VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR,
- KM_SLEEP) == 0);
+ KM_PUSHPAGE) == 0);
bzero(packed + nvsize, bufsize - nvsize);
dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx);
- kmem_free(packed, bufsize);
+ vmem_free(packed, bufsize);
VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db));
dmu_buf_will_dirty(db, tx);
&sav->sav_object, tx) == 0);
}
- VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_PUSHPAGE) == 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_SLEEP);
+ list = kmem_alloc(sav->sav_count * sizeof (void *), KM_PUSHPAGE);
for (i = 0; i < sav->sav_count; i++)
list[i] = vdev_config_generate(spa, sav->sav_vdevs[i],
- B_FALSE, B_FALSE, B_TRUE);
+ B_FALSE, VDEV_CONFIG_L2CACHE);
VERIFY(nvlist_add_nvlist_array(nvroot, config, list,
sav->sav_count) == 0);
for (i = 0; i < sav->sav_count; i++)
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)
spa_sync_nvlist(spa, spa->spa_config_object, config, tx);
}
+static void
+spa_sync_version(void *arg1, void *arg2, dmu_tx_t *tx)
+{
+ spa_t *spa = arg1;
+ uint64_t version = *(uint64_t *)arg2;
+
+ /*
+ * Setting the version is special cased when first creating the pool.
+ */
+ ASSERT(tx->tx_txg != TXG_INITIAL);
+
+ ASSERT(SPA_VERSION_IS_SUPPORTED(version));
+ ASSERT(version >= spa_version(spa));
+
+ spa->spa_uberblock.ub_version = version;
+ vdev_config_dirty(spa->spa_root_vdev);
+}
+
/*
* Set zpool properties.
*/
static void
-spa_sync_props(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
+spa_sync_props(void *arg1, void *arg2, dmu_tx_t *tx)
{
spa_t *spa = arg1;
objset_t *mos = spa->spa_meta_objset;
nvlist_t *nvp = arg2;
- nvpair_t *elem;
- uint64_t intval;
- char *strval;
- zpool_prop_t prop;
- const char *propname;
- zprop_type_t proptype;
+ nvpair_t *elem = NULL;
mutex_enter(&spa->spa_props_lock);
- elem = NULL;
while ((elem = nvlist_next_nvpair(nvp, elem))) {
- switch (prop = zpool_name_to_prop(nvpair_name(elem))) {
+ uint64_t intval;
+ char *strval, *fname;
+ zpool_prop_t prop;
+ const char *propname;
+ zprop_type_t proptype;
+ zfeature_info_t *feature;
+
+ prop = zpool_name_to_prop(nvpair_name(elem));
+ switch ((int)prop) {
+ case ZPROP_INVAL:
+ /*
+ * We checked this earlier in spa_prop_validate().
+ */
+ ASSERT(zpool_prop_feature(nvpair_name(elem)));
+
+ fname = strchr(nvpair_name(elem), '@') + 1;
+ VERIFY3U(0, ==, zfeature_lookup_name(fname, &feature));
+
+ spa_feature_enable(spa, feature, tx);
+ break;
+
case ZPOOL_PROP_VERSION:
+ VERIFY(nvpair_value_uint64(elem, &intval) == 0);
/*
- * Only set version for non-zpool-creation cases
- * (set/import). spa_create() needs special care
- * for version setting.
+ * The version is synced seperatly before other
+ * properties and should be correct by now.
*/
- if (tx->tx_txg != TXG_INITIAL) {
- VERIFY(nvpair_value_uint64(elem,
- &intval) == 0);
- ASSERT(intval <= SPA_VERSION);
- ASSERT(intval >= spa_version(spa));
- spa->spa_uberblock.ub_version = intval;
- vdev_config_dirty(spa->spa_root_vdev);
- }
+ ASSERT3U(spa_version(spa), >=, intval);
break;
case ZPOOL_PROP_ALTROOT:
ASSERT(spa->spa_root != NULL);
break;
+ case ZPOOL_PROP_READONLY:
case ZPOOL_PROP_CACHEFILE:
/*
- * 'cachefile' is also a non-persisitent property.
+ * 'readonly' and 'cachefile' are also non-persisitent
+ * properties.
+ */
+ break;
+ case ZPOOL_PROP_COMMENT:
+ VERIFY(nvpair_value_string(elem, &strval) == 0);
+ 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.
*/
+ if (tx->tx_txg != TXG_INITIAL)
+ vdev_config_dirty(spa->spa_root_vdev);
break;
default:
/*
* Set pool property values in the poolprops mos object.
*/
if (spa->spa_pool_props_object == 0) {
- objset_t *mos = spa->spa_meta_objset;
-
- VERIFY((spa->spa_pool_props_object =
- zap_create(mos, DMU_OT_POOL_PROPS,
- DMU_OT_NONE, 0, tx)) > 0);
-
- VERIFY(zap_update(mos,
+ spa->spa_pool_props_object =
+ zap_create_link(mos, DMU_OT_POOL_PROPS,
DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS,
- 8, 1, &spa->spa_pool_props_object, tx)
- == 0);
+ tx);
}
/* normalize the property name */
break;
case ZPOOL_PROP_AUTOEXPAND:
spa->spa_autoexpand = intval;
- spa_async_request(spa, SPA_ASYNC_AUTOEXPAND);
+ if (tx->tx_txg != TXG_INITIAL)
+ spa_async_request(spa,
+ SPA_ASYNC_AUTOEXPAND);
+ break;
+ case ZPOOL_PROP_DEDUPDITTO:
+ spa->spa_dedup_ditto = intval;
break;
default:
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_internal_log(LOG_POOL_PROPSET,
- spa, tx, cr, "%s %lld %s",
+ spa_history_log_internal(LOG_POOL_PROPSET,
+ spa, tx, "%s %lld %s",
nvpair_name(elem), intval, spa_name(spa));
}
}
}
/*
+ * Perform one-time upgrade on-disk changes. spa_version() does not
+ * reflect the new version this txg, so there must be no changes this
+ * txg to anything that the upgrade code depends on after it executes.
+ * Therefore this must be called after dsl_pool_sync() does the sync
+ * tasks.
+ */
+static void
+spa_sync_upgrades(spa_t *spa, dmu_tx_t *tx)
+{
+ dsl_pool_t *dp = spa->spa_dsl_pool;
+
+ ASSERT(spa->spa_sync_pass == 1);
+
+ if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN &&
+ spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) {
+ dsl_pool_create_origin(dp, tx);
+
+ /* Keeping the origin open increases spa_minref */
+ spa->spa_minref += 3;
+ }
+
+ if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES &&
+ spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) {
+ dsl_pool_upgrade_clones(dp, tx);
+ }
+
+ if (spa->spa_ubsync.ub_version < SPA_VERSION_DIR_CLONES &&
+ spa->spa_uberblock.ub_version >= SPA_VERSION_DIR_CLONES) {
+ dsl_pool_upgrade_dir_clones(dp, tx);
+
+ /* Keeping the freedir open increases spa_minref */
+ spa->spa_minref += 3;
+ }
+
+ if (spa->spa_ubsync.ub_version < SPA_VERSION_FEATURES &&
+ spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
+ spa_feature_create_zap_objects(spa, tx);
+ }
+}
+
+/*
* Sync the specified transaction group. New blocks may be dirtied as
* part of the process, so we iterate until it converges.
*/
{
dsl_pool_t *dp = spa->spa_dsl_pool;
objset_t *mos = spa->spa_meta_objset;
- bplist_t *bpl = &spa->spa_sync_bplist;
+ bpobj_t *defer_bpo = &spa->spa_deferred_bpobj;
+ bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK];
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *vd;
dmu_tx_t *tx;
- int dirty_vdevs;
int error;
+ int c;
+
+ VERIFY(spa_writeable(spa));
/*
* Lock out configuration changes.
}
spa_config_exit(spa, SCL_STATE, FTAG);
- VERIFY(0 == bplist_open(bpl, mos, spa->spa_sync_bplist_obj));
-
tx = dmu_tx_create_assigned(dp, txg);
+ spa->spa_sync_starttime = gethrtime();
+ taskq_cancel_id(system_taskq, spa->spa_deadman_tqid);
+ spa->spa_deadman_tqid = taskq_dispatch_delay(system_taskq,
+ spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() +
+ NSEC_TO_TICK(spa->spa_deadman_synctime));
+
/*
* If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg,
* set spa_deflate if we have no raid-z vdevs.
}
}
- if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN &&
- spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) {
- dsl_pool_create_origin(dp, tx);
-
- /* Keeping the origin open increases spa_minref */
- spa->spa_minref += 3;
- }
-
- if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES &&
- spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) {
- dsl_pool_upgrade_clones(dp, tx);
- }
-
/*
- * If anything has changed in this txg, 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.
+ * 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.
*/
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))
- spa_sync_deferred_frees(spa, 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);
+ VERIFY0(zio_wait(zio));
+ }
/*
* Iterate to convergence.
*/
do {
- spa->spa_sync_pass++;
+ int pass = ++spa->spa_sync_pass;
spa_sync_config_object(spa, tx);
spa_sync_aux_dev(spa, &spa->spa_spares, tx,
spa_errlog_sync(spa, txg);
dsl_pool_sync(dp, txg);
- dirty_vdevs = 0;
- while (vd = txg_list_remove(&spa->spa_vdev_txg_list, txg)) {
- vdev_sync(vd, txg);
- dirty_vdevs++;
+ if (pass < zfs_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);
+ } else {
+ bplist_iterate(free_bpl, bpobj_enqueue_cb,
+ defer_bpo, tx);
}
- bplist_sync(bpl, tx);
- } while (dirty_vdevs);
+ ddt_sync(spa, txg);
+ dsl_scan_sync(dp, tx);
+
+ while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, txg)))
+ vdev_sync(vd, txg);
- bplist_close(bpl);
+ if (pass == 1)
+ spa_sync_upgrades(spa, tx);
- dprintf("txg %llu passes %d\n", txg, spa->spa_sync_pass);
+ } while (dmu_objset_is_dirty(mos, txg));
/*
* Rewrite the vdev configuration (which includes the uberblock)
int children = rvd->vdev_children;
int c0 = spa_get_random(children);
- for (int c = 0; c < children; c++) {
+ for (c = 0; c < children; c++) {
vd = rvd->vdev_child[(c0 + c) % children];
if (vd->vdev_ms_array == 0 || vd->vdev_islog)
continue;
rvd->vdev_children, txg, B_TRUE);
}
+ if (error == 0)
+ spa->spa_last_synced_guid = rvd->vdev_guid;
+
spa_config_exit(spa, SCL_STATE, FTAG);
if (error == 0)
}
dmu_tx_commit(tx);
+ taskq_cancel_id(system_taskq, spa->spa_deadman_tqid);
+ spa->spa_deadman_tqid = 0;
+
/*
* Clear the dirty config list.
*/
spa->spa_ubsync = spa->spa_uberblock;
- /*
- * Clean up the ZIL records for the synced txg.
- */
- dsl_pool_zil_clean(dp);
+ dsl_pool_sync_done(dp, txg);
/*
* Update usable space statistics.
*/
- while (vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)))
+ while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg))))
vdev_sync_done(vd, txg);
+ spa_update_dspace(spa);
+
/*
* It had better be the case that we didn't dirty anything
* since vdev_config_sync().
ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg));
ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg));
ASSERT(txg_list_empty(&spa->spa_vdev_txg_list, txg));
- ASSERT(bpl->bpl_queue == NULL);
+
+ spa->spa_sync_pass = 0;
spa_config_exit(spa, SCL_CONFIG, FTAG);
+ spa_handle_ignored_writes(spa);
+
/*
* If any async tasks have been requested, kick them off.
*/
spa_t *spa = NULL;
mutex_enter(&spa_namespace_lock);
while ((spa = spa_next(spa)) != NULL) {
- if (spa_state(spa) != POOL_STATE_ACTIVE || spa_suspended(spa))
+ if (spa_state(spa) != POOL_STATE_ACTIVE ||
+ !spa_writeable(spa) || spa_suspended(spa))
continue;
spa_open_ref(spa, FTAG);
mutex_exit(&spa_namespace_lock);
void
spa_upgrade(spa_t *spa, uint64_t version)
{
+ ASSERT(spa_writeable(spa));
+
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
/*
* future version would result in an unopenable pool, this shouldn't be
* possible.
*/
- ASSERT(spa->spa_uberblock.ub_version <= SPA_VERSION);
+ ASSERT(SPA_VERSION_IS_SUPPORTED(spa->spa_uberblock.ub_version));
ASSERT(version >= spa->spa_uberblock.ub_version);
spa->spa_uberblock.ub_version = version;
}
/*
- * Post a sysevent corresponding to the given event. The 'name' must be one of
- * the event definitions in sys/sysevent/eventdefs.h. The payload will be
+ * Post a FM_EREPORT_ZFS_* event from sys/fm/fs/zfs.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.
spa_event_notify(spa_t *spa, vdev_t *vd, const char *name)
{
#ifdef _KERNEL
- sysevent_t *ev;
- sysevent_attr_list_t *attr = NULL;
- sysevent_value_t value;
- sysevent_id_t eid;
-
- ev = sysevent_alloc(EC_ZFS, (char *)name, SUNW_KERN_PUB "zfs",
- SE_SLEEP);
-
- value.value_type = SE_DATA_TYPE_STRING;
- value.value.sv_string = spa_name(spa);
- if (sysevent_add_attr(&attr, ZFS_EV_POOL_NAME, &value, SE_SLEEP) != 0)
- goto done;
-
- value.value_type = SE_DATA_TYPE_UINT64;
- value.value.sv_uint64 = spa_guid(spa);
- if (sysevent_add_attr(&attr, ZFS_EV_POOL_GUID, &value, SE_SLEEP) != 0)
- goto done;
-
- if (vd) {
- value.value_type = SE_DATA_TYPE_UINT64;
- value.value.sv_uint64 = vd->vdev_guid;
- if (sysevent_add_attr(&attr, ZFS_EV_VDEV_GUID, &value,
- SE_SLEEP) != 0)
- goto done;
-
- if (vd->vdev_path) {
- value.value_type = SE_DATA_TYPE_STRING;
- value.value.sv_string = vd->vdev_path;
- if (sysevent_add_attr(&attr, ZFS_EV_VDEV_PATH,
- &value, SE_SLEEP) != 0)
- goto done;
- }
- }
-
- if (sysevent_attach_attributes(ev, attr) != 0)
- goto done;
- attr = NULL;
-
- (void) log_sysevent(ev, SE_SLEEP, &eid);
-
-done:
- if (attr)
- sysevent_free_attr(attr);
- sysevent_free(ev);
+ zfs_ereport_post(name, spa, vd, NULL, 0, 0);
#endif
}
+
+#if defined(_KERNEL) && defined(HAVE_SPL)
+/* state manipulation functions */
+EXPORT_SYMBOL(spa_open);
+EXPORT_SYMBOL(spa_open_rewind);
+EXPORT_SYMBOL(spa_get_stats);
+EXPORT_SYMBOL(spa_create);
+EXPORT_SYMBOL(spa_import_rootpool);
+EXPORT_SYMBOL(spa_import);
+EXPORT_SYMBOL(spa_tryimport);
+EXPORT_SYMBOL(spa_destroy);
+EXPORT_SYMBOL(spa_export);
+EXPORT_SYMBOL(spa_reset);
+EXPORT_SYMBOL(spa_async_request);
+EXPORT_SYMBOL(spa_async_suspend);
+EXPORT_SYMBOL(spa_async_resume);
+EXPORT_SYMBOL(spa_inject_addref);
+EXPORT_SYMBOL(spa_inject_delref);
+EXPORT_SYMBOL(spa_scan_stat_init);
+EXPORT_SYMBOL(spa_scan_get_stats);
+
+/* device maniion */
+EXPORT_SYMBOL(spa_vdev_add);
+EXPORT_SYMBOL(spa_vdev_attach);
+EXPORT_SYMBOL(spa_vdev_detach);
+EXPORT_SYMBOL(spa_vdev_remove);
+EXPORT_SYMBOL(spa_vdev_setpath);
+EXPORT_SYMBOL(spa_vdev_setfru);
+EXPORT_SYMBOL(spa_vdev_split_mirror);
+
+/* spare statech is global across all pools) */
+EXPORT_SYMBOL(spa_spare_add);
+EXPORT_SYMBOL(spa_spare_remove);
+EXPORT_SYMBOL(spa_spare_exists);
+EXPORT_SYMBOL(spa_spare_activate);
+
+/* L2ARC statech is global across all pools) */
+EXPORT_SYMBOL(spa_l2cache_add);
+EXPORT_SYMBOL(spa_l2cache_remove);
+EXPORT_SYMBOL(spa_l2cache_exists);
+EXPORT_SYMBOL(spa_l2cache_activate);
+EXPORT_SYMBOL(spa_l2cache_drop);
+
+/* scanning */
+EXPORT_SYMBOL(spa_scan);
+EXPORT_SYMBOL(spa_scan_stop);
+
+/* spa syncing */
+EXPORT_SYMBOL(spa_sync); /* only for DMU use */
+EXPORT_SYMBOL(spa_sync_allpools);
+
+/* properties */
+EXPORT_SYMBOL(spa_prop_set);
+EXPORT_SYMBOL(spa_prop_get);
+EXPORT_SYMBOL(spa_prop_clear_bootfs);
+
+/* asynchronous event notification */
+EXPORT_SYMBOL(spa_event_notify);
+#endif
git://git.camperquake.de / zfs.git / blobdiff