/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
* Copyright (c) 2015, Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2013, 2014, Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
* Copyright 2013 Saso Kiselkov. All rights reserved.
+ * Copyright (c) 2014 Integros [integros.com]
+ * Copyright 2016 Toomas Soome <tsoome@me.com>
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
+ * Copyright (c) 2017 Datto Inc.
+ * Copyright 2017 Joyent, Inc.
*/
/*
#include <sys/vdev_disk.h>
#include <sys/metaslab.h>
#include <sys/metaslab_impl.h>
+#include <sys/mmp.h>
#include <sys/uberblock_impl.h>
#include <sys/txg.h>
#include <sys/avl.h>
#include <sys/zvol.h>
#ifdef _KERNEL
+#include <sys/fm/protocol.h>
+#include <sys/fm/util.h>
#include <sys/bootprops.h>
#include <sys/callb.h>
#include <sys/cpupart.h>
{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* IOCTL */
};
+static sysevent_t *spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl,
+ const char *name);
+static void spa_event_post(sysevent_t *ev);
static void spa_sync_version(void *arg, dmu_tx_t *tx);
static void spa_sync_props(void *arg, dmu_tx_t *tx);
static boolean_t spa_has_active_shared_spare(spa_t *spa);
error = SET_ERROR(EINVAL);
break;
+ case ZPOOL_PROP_MULTIHOST:
+ error = nvpair_value_uint64(elem, &intval);
+ if (!error && intval > 1)
+ error = SET_ERROR(EINVAL);
+
+ if (!error && !spa_get_hostid())
+ error = SET_ERROR(ENOTSUP);
+
+ break;
+
case ZPOOL_PROP_BOOTFS:
/*
* If the pool version is less than SPA_VERSION_BOOTFS,
&propval)) == 0 &&
!BOOTFS_COMPRESS_VALID(propval)) {
error = SET_ERROR(ENOTSUP);
- } else if ((error =
- dsl_prop_get_int_ds(dmu_objset_ds(os),
- zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
- &propval)) == 0 &&
- propval > SPA_OLD_MAXBLOCKSIZE) {
- error = SET_ERROR(ENOTSUP);
} else if ((error =
dsl_prop_get_int_ds(dmu_objset_ds(os),
zfs_prop_to_name(ZFS_PROP_DNODESIZE),
case ZPOOL_PROP_FAILUREMODE:
error = nvpair_value_uint64(elem, &intval);
- if (!error && (intval < ZIO_FAILURE_MODE_WAIT ||
- intval > ZIO_FAILURE_MODE_PANIC))
+ if (!error && intval > ZIO_FAILURE_MODE_PANIC)
error = SET_ERROR(EINVAL);
/*
if (error == 0) {
spa_config_sync(spa, B_FALSE, B_TRUE);
- spa_event_notify(spa, NULL, ESC_ZFS_POOL_REGUID);
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_REGUID);
}
mutex_exit(&spa_namespace_lock);
uint_t count = ztip->zti_count;
spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
char name[32];
- uint_t i, flags = TASKQ_DYNAMIC;
+ uint_t i, flags = 0;
boolean_t batch = B_FALSE;
if (mode == ZTI_MODE_NULL) {
case ZTI_MODE_FIXED:
ASSERT3U(value, >=, 1);
value = MAX(value, 1);
+ flags |= TASKQ_DYNAMIC;
break;
case ZTI_MODE_BATCH:
list_create(&spa->spa_state_dirty_list, sizeof (vdev_t),
offsetof(vdev_t, vdev_state_dirty_node));
- txg_list_create(&spa->spa_vdev_txg_list,
+ txg_list_create(&spa->spa_vdev_txg_list, spa,
offsetof(struct vdev, vdev_txg_node));
avl_create(&spa->spa_errlist_scrub,
*/
spa->spa_zvol_taskq = taskq_create("z_zvol", 1, defclsyspri,
1, INT_MAX, 0);
+
+ /*
+ * The taskq to upgrade datasets in this pool. Currently used by
+ * feature SPA_FEATURE_USEROBJ_ACCOUNTING.
+ */
+ spa->spa_upgrade_taskq = taskq_create("z_upgrade", boot_ncpus,
+ defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC);
}
/*
spa->spa_zvol_taskq = NULL;
}
+ if (spa->spa_upgrade_taskq) {
+ taskq_destroy(spa->spa_upgrade_taskq);
+ spa->spa_upgrade_taskq = NULL;
+ }
+
txg_list_destroy(&spa->spa_vdev_txg_list);
list_destroy(&spa->spa_config_dirty_list);
list_destroy(&spa->spa_evicting_os_list);
list_destroy(&spa->spa_state_dirty_list);
- taskq_cancel_id(system_taskq, spa->spa_deadman_tqid);
+ taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
for (t = 0; t < ZIO_TYPES; t++) {
for (q = 0; q < ZIO_TASKQ_TYPES; q++) {
static void
spa_unload(spa_t *spa)
{
- int i;
+ int i, c;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
spa->spa_sync_on = B_FALSE;
}
+ /*
+ * Even though vdev_free() also calls vdev_metaslab_fini, we need
+ * to call it earlier, before we wait for async i/o to complete.
+ * This ensures that there is no async metaslab prefetching, by
+ * calling taskq_wait(mg_taskq).
+ */
+ if (spa->spa_root_vdev != NULL) {
+ spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
+ for (c = 0; c < spa->spa_root_vdev->vdev_children; c++)
+ vdev_metaslab_fini(spa->spa_root_vdev->vdev_child[c]);
+ spa_config_exit(spa, SCL_ALL, FTAG);
+ }
+
+ if (spa->spa_mmp.mmp_thread)
+ mmp_thread_stop(spa);
+
/*
* Wait for any outstanding async I/O to complete.
*/
ddt_unload(spa);
-
/*
* Drop and purge level 2 cache
*/
if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd) &&
!vd->vdev_ishole) {
zfs_post_autoreplace(vd->vdev_spa, vd);
- spa_event_notify(vd->vdev_spa, vd, ESC_ZFS_VDEV_CHECK);
+ spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_CHECK);
}
}
int error = zio->io_error;
spa_t *spa = zio->io_spa;
+ abd_free(zio->io_abd);
if (error) {
if ((BP_GET_LEVEL(bp) != 0 || DMU_OT_IS_METADATA(type)) &&
type != DMU_OT_INTENT_LOG)
else
atomic_inc_64(&sle->sle_data_count);
}
- zio_data_buf_free(zio->io_data, zio->io_size);
mutex_enter(&spa->spa_scrub_lock);
spa->spa_scrub_inflight--;
{
zio_t *rio;
size_t size;
- void *data;
if (bp == NULL || BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp))
return (0);
*/
if (!spa_load_verify_metadata)
return (0);
- if (BP_GET_BUFC_TYPE(bp) == ARC_BUFC_DATA && !spa_load_verify_data)
+ if (!BP_IS_METADATA(bp) && !spa_load_verify_data)
return (0);
rio = arg;
size = BP_GET_PSIZE(bp);
- data = zio_data_buf_alloc(size);
mutex_enter(&spa->spa_scrub_lock);
while (spa->spa_scrub_inflight >= spa_load_verify_maxinflight)
spa->spa_scrub_inflight++;
mutex_exit(&spa->spa_scrub_lock);
- zio_nowait(zio_read(rio, spa, bp, data, size,
+ zio_nowait(zio_read(rio, spa, bp, abd_alloc_for_io(size, B_FALSE), size,
spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB,
ZIO_FLAG_SPECULATIVE | ZIO_FLAG_CANFAIL |
ZIO_FLAG_SCRUB | ZIO_FLAG_RAW, zb));
}
#endif
+/*
+ * Determine whether the activity check is required.
+ */
+static boolean_t
+spa_activity_check_required(spa_t *spa, uberblock_t *ub, nvlist_t *label,
+ nvlist_t *config)
+{
+ uint64_t state = 0;
+ uint64_t hostid = 0;
+ uint64_t tryconfig_txg = 0;
+ uint64_t tryconfig_timestamp = 0;
+ nvlist_t *nvinfo;
+
+ if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
+ nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
+ (void) nvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG,
+ &tryconfig_txg);
+ (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
+ &tryconfig_timestamp);
+ }
+
+ (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, &state);
+
+ /*
+ * Disable the MMP activity check - This is used by zdb which
+ * is intended to be used on potentially active pools.
+ */
+ if (spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP)
+ return (B_FALSE);
+
+ /*
+ * Skip the activity check when the MMP feature is disabled.
+ */
+ if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay == 0)
+ return (B_FALSE);
+ /*
+ * If the tryconfig_* values are nonzero, they are the results of an
+ * earlier tryimport. If they match the uberblock we just found, then
+ * the pool has not changed and we return false so we do not test a
+ * second time.
+ */
+ if (tryconfig_txg && tryconfig_txg == ub->ub_txg &&
+ tryconfig_timestamp && tryconfig_timestamp == ub->ub_timestamp)
+ return (B_FALSE);
+
+ /*
+ * Allow the activity check to be skipped when importing the pool
+ * on the same host which last imported it. Since the hostid from
+ * configuration may be stale use the one read from the label.
+ */
+ if (nvlist_exists(label, ZPOOL_CONFIG_HOSTID))
+ hostid = fnvlist_lookup_uint64(label, ZPOOL_CONFIG_HOSTID);
+
+ if (hostid == spa_get_hostid())
+ return (B_FALSE);
+
+ /*
+ * Skip the activity test when the pool was cleanly exported.
+ */
+ if (state != POOL_STATE_ACTIVE)
+ return (B_FALSE);
+
+ return (B_TRUE);
+}
+
+/*
+ * Perform the import activity check. If the user canceled the import or
+ * we detected activity then fail.
+ */
+static int
+spa_activity_check(spa_t *spa, uberblock_t *ub, nvlist_t *config)
+{
+ uint64_t import_intervals = MAX(zfs_multihost_import_intervals, 1);
+ uint64_t txg = ub->ub_txg;
+ uint64_t timestamp = ub->ub_timestamp;
+ uint64_t import_delay = NANOSEC;
+ hrtime_t import_expire;
+ nvlist_t *mmp_label = NULL;
+ vdev_t *rvd = spa->spa_root_vdev;
+ kcondvar_t cv;
+ kmutex_t mtx;
+ int error = 0;
+
+ cv_init(&cv, NULL, CV_DEFAULT, NULL);
+ mutex_init(&mtx, NULL, MUTEX_DEFAULT, NULL);
+ mutex_enter(&mtx);
+
+ /*
+ * If ZPOOL_CONFIG_MMP_TXG is present an activity check was performed
+ * during the earlier tryimport. If the txg recorded there is 0 then
+ * the pool is known to be active on another host.
+ *
+ * Otherwise, the pool might be in use on another node. Check for
+ * changes in the uberblocks on disk if necessary.
+ */
+ if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
+ nvlist_t *nvinfo = fnvlist_lookup_nvlist(config,
+ ZPOOL_CONFIG_LOAD_INFO);
+
+ if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_TXG) &&
+ fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG) == 0) {
+ vdev_uberblock_load(rvd, ub, &mmp_label);
+ error = SET_ERROR(EREMOTEIO);
+ goto out;
+ }
+ }
+
+ /*
+ * Preferentially use the zfs_multihost_interval from the node which
+ * last imported the pool. This value is stored in an MMP uberblock as.
+ *
+ * ub_mmp_delay * vdev_count_leaves() == zfs_multihost_interval
+ */
+ if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay)
+ import_delay = MAX(import_delay, import_intervals *
+ ub->ub_mmp_delay * MAX(vdev_count_leaves(spa), 1));
+
+ /* Apply a floor using the local default values. */
+ import_delay = MAX(import_delay, import_intervals *
+ MSEC2NSEC(MAX(zfs_multihost_interval, MMP_MIN_INTERVAL)));
+
+ /* Add a small random factor in case of simultaneous imports (0-25%) */
+ import_expire = gethrtime() + import_delay +
+ (import_delay * spa_get_random(250) / 1000);
+
+ while (gethrtime() < import_expire) {
+ vdev_uberblock_load(rvd, ub, &mmp_label);
+
+ if (txg != ub->ub_txg || timestamp != ub->ub_timestamp) {
+ error = SET_ERROR(EREMOTEIO);
+ break;
+ }
+
+ if (mmp_label) {
+ nvlist_free(mmp_label);
+ mmp_label = NULL;
+ }
+
+ error = cv_timedwait_sig(&cv, &mtx, ddi_get_lbolt() + hz);
+ if (error != -1) {
+ error = SET_ERROR(EINTR);
+ break;
+ }
+ error = 0;
+ }
+
+out:
+ mutex_exit(&mtx);
+ mutex_destroy(&mtx);
+ cv_destroy(&cv);
+
+ /*
+ * If the pool is determined to be active store the status in the
+ * spa->spa_load_info nvlist. If the remote hostname or hostid are
+ * available from configuration read from disk store them as well.
+ * This allows 'zpool import' to generate a more useful message.
+ *
+ * ZPOOL_CONFIG_MMP_STATE - observed pool status (mandatory)
+ * ZPOOL_CONFIG_MMP_HOSTNAME - hostname from the active pool
+ * ZPOOL_CONFIG_MMP_HOSTID - hostid from the active pool
+ */
+ if (error == EREMOTEIO) {
+ char *hostname = "<unknown>";
+ uint64_t hostid = 0;
+
+ if (mmp_label) {
+ if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTNAME)) {
+ hostname = fnvlist_lookup_string(mmp_label,
+ ZPOOL_CONFIG_HOSTNAME);
+ fnvlist_add_string(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_HOSTNAME, hostname);
+ }
+
+ if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTID)) {
+ hostid = fnvlist_lookup_uint64(mmp_label,
+ ZPOOL_CONFIG_HOSTID);
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_HOSTID, hostid);
+ }
+ }
+
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_STATE, MMP_STATE_ACTIVE);
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_TXG, 0);
+
+ error = spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO);
+ }
+
+ if (mmp_label)
+ nvlist_free(mmp_label);
+
+ return (error);
+}
+
/*
* Load an existing storage pool, using the pool's builtin spa_config as a
* source of configuration information.
int parse, i;
uint64_t obj;
boolean_t missing_feat_write = B_FALSE;
+ boolean_t activity_check = B_FALSE;
nvlist_t *mos_config;
/*
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO));
}
+ /*
+ * For pools which have the multihost property on determine if the
+ * pool is truly inactive and can be safely imported. Prevent
+ * hosts which don't have a hostid set from importing the pool.
+ */
+ activity_check = spa_activity_check_required(spa, ub, label, config);
+ if (activity_check) {
+ if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay &&
+ spa_get_hostid() == 0) {
+ nvlist_free(label);
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
+ return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
+ }
+
+ error = spa_activity_check(spa, ub, config);
+ if (error) {
+ nvlist_free(label);
+ return (error);
+ }
+
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_STATE, MMP_STATE_INACTIVE);
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_TXG, ub->ub_txg);
+ }
+
/*
* If the pool has an unsupported version we can't open it.
*/
VERIFY(nvlist_lookup_string(nvconfig,
ZPOOL_CONFIG_HOSTNAME, &hostname) == 0);
-#ifdef _KERNEL
- myhostid = zone_get_hostid(NULL);
-#else /* _KERNEL */
- /*
- * We're emulating the system's hostid in userland, so
- * we can't use zone_get_hostid().
- */
- (void) ddi_strtoul(hw_serial, NULL, 10, &myhostid);
-#endif /* _KERNEL */
- if (hostid != 0 && myhostid != 0 &&
- hostid != myhostid) {
+ myhostid = spa_get_hostid();
+ if (hostid && myhostid && hostid != myhostid) {
nvlist_free(nvconfig);
- cmn_err(CE_WARN, "pool '%s' could not be "
- "loaded as it was last accessed by another "
- "system (host: %s hostid: 0x%lx). See: "
- "http://zfsonlinux.org/msg/ZFS-8000-EY",
- spa_name(spa), hostname,
- (unsigned long)hostid);
return (SET_ERROR(EBADF));
}
}
error = spa_dir_prop(spa, DMU_POOL_VDEV_ZAP_MAP,
&spa->spa_all_vdev_zaps);
- if (error != ENOENT && error != 0) {
+ if (error == ENOENT) {
+ VERIFY(!nvlist_exists(mos_config,
+ ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
+ spa->spa_avz_action = AVZ_ACTION_INITIALIZE;
+ ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
+ } else if (error != 0) {
return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
- } else if (error == 0 && !nvlist_exists(mos_config,
- ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) {
+ } else if (!nvlist_exists(mos_config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) {
/*
* An older version of ZFS overwrote the sentinel value, so
* we have orphaned per-vdev ZAPs in the MOS. Defer their
spa_prop_find(spa, ZPOOL_PROP_DELEGATION, &spa->spa_delegation);
spa_prop_find(spa, ZPOOL_PROP_FAILUREMODE, &spa->spa_failmode);
spa_prop_find(spa, ZPOOL_PROP_AUTOEXPAND, &spa->spa_autoexpand);
+ spa_prop_find(spa, ZPOOL_PROP_MULTIHOST, &spa->spa_multihost);
spa_prop_find(spa, ZPOOL_PROP_DEDUPDITTO,
&spa->spa_dedup_ditto);
spa->spa_autoreplace = (autoreplace != 0);
}
+ /*
+ * If the 'multihost' property is set, then never allow a pool to
+ * be imported when the system hostid is zero. The exception to
+ * this rule is zdb which is always allowed to access pools.
+ */
+ if (spa_multihost(spa) && spa_get_hostid() == 0 &&
+ (spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP) == 0) {
+ fnvlist_add_uint64(spa->spa_load_info,
+ ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
+ return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
+ }
+
/*
* If the 'autoreplace' property is set, then post a resource notifying
* the ZFS DE that it should not issue any faults for unopenable
spa_set_log_state(spa, SPA_LOG_GOOD);
spa->spa_sync_on = B_TRUE;
txg_sync_start(spa->spa_dsl_pool);
+ mmp_thread_start(spa);
/*
* Wait for all claims to sync. We sync up to the highest
* Log the fact that we booted up (so that we can detect if
* we rebooted in the middle of an operation).
*/
- spa_history_log_version(spa, "open");
+ spa_history_log_version(spa, "open", NULL);
/*
* Delete any inconsistent datasets.
ZPOOL_CONFIG_ERRCOUNT,
spa_get_errlog_size(spa)) == 0);
- if (spa_suspended(spa))
+ if (spa_suspended(spa)) {
VERIFY(nvlist_add_uint64(*config,
ZPOOL_CONFIG_SUSPENDED,
spa->spa_failmode) == 0);
+ VERIFY(nvlist_add_uint64(*config,
+ ZPOOL_CONFIG_SUSPENDED_REASON,
+ spa->spa_suspended) == 0);
+ }
spa_add_spares(spa, *config);
spa_add_l2cache(spa, *config);
goto out;
}
- /*
- * The L2ARC currently only supports disk devices in
- * kernel context. For user-level testing, we allow it.
- */
-#ifdef _KERNEL
- if ((strcmp(config, ZPOOL_CONFIG_L2CACHE) == 0) &&
- strcmp(vd->vdev_ops->vdev_op_type, VDEV_TYPE_DISK) != 0) {
- error = SET_ERROR(ENOTBLK);
- vdev_free(vd);
- goto out;
- }
-#endif
vd->vdev_top = vd;
if ((error = vdev_open(vd)) == 0 &&
nvlist_t **newdevs;
/*
- * Generate new dev list by concatentating with the
+ * Generate new dev list by concatenating with the
* current dev list.
*/
VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, config,
spa->spa_uberblock.ub_txg = txg - 1;
spa->spa_uberblock.ub_version = version;
spa->spa_ubsync = spa->spa_uberblock;
+ spa->spa_load_state = SPA_LOAD_CREATE;
/*
* Create "The Godfather" zio to hold all async IOs
tx = dmu_tx_create_assigned(dp, txg);
+ /*
+ * Create the pool's history object.
+ */
+ if (version >= SPA_VERSION_ZPOOL_HISTORY && !spa->spa_history)
+ spa_history_create_obj(spa, tx);
+
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_CREATE);
+ spa_history_log_version(spa, "create", tx);
+
/*
* Create the pool config object.
*/
VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj,
spa->spa_meta_objset, obj));
- /*
- * Create the pool's history object.
- */
- if (version >= SPA_VERSION_ZPOOL_HISTORY)
- spa_history_create_obj(spa, tx);
-
/*
* Generate some random noise for salted checksums to operate on.
*/
spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE);
spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND);
+ spa->spa_multihost = zpool_prop_default_numeric(ZPOOL_PROP_MULTIHOST);
if (props != NULL) {
spa_configfile_set(spa, props, B_FALSE);
spa->spa_sync_on = B_TRUE;
txg_sync_start(spa->spa_dsl_pool);
+ mmp_thread_start(spa);
/*
* We explicitly wait for the first transaction to complete so that our
txg_wait_synced(spa->spa_dsl_pool, txg);
spa_config_sync(spa, B_FALSE, B_TRUE);
- spa_event_notify(spa, NULL, ESC_ZFS_POOL_CREATE);
-
- spa_history_log_version(spa, "create");
/*
* Don't count references from objsets that are already closed
*/
spa_evicting_os_wait(spa);
spa->spa_minref = refcount_count(&spa->spa_refcount);
+ spa->spa_load_state = SPA_LOAD_NONE;
mutex_exit(&spa_namespace_lock);
spa_configfile_set(spa, props, B_FALSE);
spa_config_sync(spa, B_FALSE, B_TRUE);
- spa_event_notify(spa, NULL, ESC_ZFS_POOL_IMPORT);
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);
mutex_exit(&spa_namespace_lock);
return (0);
VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
&nvroot) == 0);
- if (error == 0)
- error = spa_validate_aux(spa, nvroot, -1ULL,
- VDEV_ALLOC_SPARE);
- if (error == 0)
- error = spa_validate_aux(spa, nvroot, -1ULL,
- VDEV_ALLOC_L2CACHE);
spa_config_exit(spa, SCL_ALL, FTAG);
if (props != NULL)
*/
spa_async_request(spa, SPA_ASYNC_AUTOEXPAND);
- spa_history_log_version(spa, "import");
+ spa_history_log_version(spa, "import", NULL);
- spa_event_notify(spa, NULL, ESC_ZFS_POOL_IMPORT);
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);
zvol_create_minors(spa, pool, B_TRUE);
}
export_spa:
- spa_event_notify(spa, NULL, ESC_ZFS_POOL_DESTROY);
+ if (new_state == POOL_STATE_DESTROYED)
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_DESTROY);
+ else if (new_state == POOL_STATE_EXPORTED)
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_EXPORT);
if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
spa_unload(spa);
mutex_enter(&spa_namespace_lock);
spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
- spa_event_notify(spa, NULL, ESC_ZFS_VDEV_ADD);
+ spa_event_notify(spa, NULL, NULL, ESC_ZFS_VDEV_ADD);
mutex_exit(&spa_namespace_lock);
return (0);
newvd->vdev_crtxg = oldvd->vdev_crtxg;
vdev_add_child(pvd, newvd);
+ /*
+ * Reevaluate the parent vdev state.
+ */
+ vdev_propagate_state(pvd);
+
tvd = newvd->vdev_top;
ASSERT(pvd->vdev_top == tvd);
ASSERT(tvd->vdev_parent == rvd);
if (newvd->vdev_isspare) {
spa_spare_activate(newvd);
- spa_event_notify(spa, newvd, ESC_ZFS_VDEV_SPARE);
+ spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_SPARE);
}
oldvdpath = spa_strdup(oldvd->vdev_path);
dsl_resilver_restart(spa->spa_dsl_pool, dtl_max_txg);
if (spa->spa_bootfs)
- spa_event_notify(spa, newvd, ESC_ZFS_BOOTFS_VDEV_ATTACH);
+ spa_event_notify(spa, newvd, NULL, ESC_ZFS_BOOTFS_VDEV_ATTACH);
- spa_event_notify(spa, newvd, ESC_ZFS_VDEV_ATTACH);
+ spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_ATTACH);
/*
* Commit the config
* But first make sure we're not on any *other* txg's DTL list, to
* prevent vd from being accessed after it's freed.
*/
- vdpath = spa_strdup(vd->vdev_path);
+ vdpath = spa_strdup(vd->vdev_path ? vd->vdev_path : "none");
for (t = 0; t < TXG_SIZE; t++)
(void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t);
vd->vdev_detached = B_TRUE;
vdev_dirty(tvd, VDD_DTL, vd, txg);
- spa_event_notify(spa, vd, ESC_ZFS_VDEV_REMOVE);
+ spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE);
/* hang on to the spa before we release the lock */
spa_open_ref(spa, FTAG);
static void
spa_vdev_remove_aux(nvlist_t *config, char *name, nvlist_t **dev, int count,
- nvlist_t *dev_to_remove)
+ nvlist_t *dev_to_remove)
{
nvlist_t **newdev = NULL;
int i, j;
spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare)
{
vdev_t *vd;
+ sysevent_t *ev = NULL;
metaslab_group_t *mg;
nvlist_t **spares, **l2cache, *nv;
uint64_t txg = 0;
* in this pool.
*/
if (vd == NULL || unspare) {
+ if (vd == NULL)
+ vd = spa_lookup_by_guid(spa, guid, B_TRUE);
+ ev = spa_event_create(spa, vd, NULL,
+ ESC_ZFS_VDEV_REMOVE_AUX);
spa_vdev_remove_aux(spa->spa_spares.sav_config,
ZPOOL_CONFIG_SPARES, spares, nspares, nv);
spa_load_spares(spa);
} else {
error = SET_ERROR(EBUSY);
}
- spa_event_notify(spa, vd, ESC_ZFS_VDEV_REMOVE_AUX);
} else if (spa->spa_l2cache.sav_vdevs != NULL &&
nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0 &&
/*
* Cache devices can always be removed.
*/
+ vd = spa_lookup_by_guid(spa, guid, B_TRUE);
+ ev = spa_event_create(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE_AUX);
spa_vdev_remove_aux(spa->spa_l2cache.sav_config,
ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv);
spa_load_l2cache(spa);
spa->spa_l2cache.sav_sync = B_TRUE;
- spa_event_notify(spa, vd, ESC_ZFS_VDEV_REMOVE_AUX);
} else if (vd != NULL && vd->vdev_islog) {
ASSERT(!locked);
ASSERT(vd == vd->vdev_top);
/*
* Clean up the vdev namespace.
*/
+ ev = spa_event_create(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE_DEV);
spa_vdev_remove_from_namespace(spa, vd);
- spa_event_notify(spa, vd, ESC_ZFS_VDEV_REMOVE_DEV);
} else if (vd != NULL) {
/*
* Normal vdevs cannot be removed (yet).
}
if (!locked)
- return (spa_vdev_exit(spa, NULL, txg, error));
+ error = spa_vdev_exit(spa, NULL, txg, error);
+
+ if (ev)
+ spa_event_post(ev);
return (error);
}
* SPA Scanning
* ==========================================================================
*/
+int
+spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t cmd)
+{
+ ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
+
+ if (dsl_scan_resilvering(spa->spa_dsl_pool))
+ return (SET_ERROR(EBUSY));
+
+ return (dsl_scrub_set_pause_resume(spa->spa_dsl_pool, cmd));
+}
int
spa_scan_stop(spa_t *spa)
if (!vd->vdev_ops->vdev_op_leaf || vd->vdev_physpath == NULL)
return;
- spa_event_notify(vd->vdev_spa, vd, ESC_ZFS_VDEV_AUTOEXPAND);
+ spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_AUTOEXPAND);
}
static void
} else {
config_task_suspended =
(gethrtime() - spa->spa_ccw_fail_time) <
- (zfs_ccw_retry_interval * NANOSEC);
+ ((hrtime_t)zfs_ccw_retry_interval * NANOSEC);
}
return (non_config_tasks || (config_task && !config_task_suspended));
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
ASSERT(spa->spa_avz_action == AVZ_ACTION_NONE ||
+ spa->spa_avz_action == AVZ_ACTION_INITIALIZE ||
spa->spa_all_vdev_zaps != 0);
if (spa->spa_avz_action == AVZ_ACTION_REBUILD) {
case ZPOOL_PROP_VERSION:
intval = fnvpair_value_uint64(elem);
/*
- * The version is synced seperatly before other
+ * The version is synced separately before other
* properties and should be correct by now.
*/
ASSERT3U(spa_version(spa), >=, intval);
* We need to dirty the configuration on all the vdevs
* so that their labels get updated. It's unnecessary
* to do this for pool creation since the vdev's
- * configuratoin has already been dirtied.
+ * configuration has already been dirtied.
*/
if (tx->tx_txg != TXG_INITIAL)
vdev_config_dirty(spa->spa_root_vdev);
spa_async_request(spa,
SPA_ASYNC_AUTOEXPAND);
break;
+ case ZPOOL_PROP_MULTIHOST:
+ spa->spa_multihost = intval;
+ break;
case ZPOOL_PROP_DEDUPDITTO:
spa->spa_dedup_ditto = intval;
break;
dsl_pool_t *dp = spa->spa_dsl_pool;
objset_t *mos = spa->spa_meta_objset;
bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK];
+ metaslab_class_t *mc;
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *vd;
dmu_tx_t *tx;
int error;
+ uint32_t max_queue_depth = zfs_vdev_async_write_max_active *
+ zfs_vdev_queue_depth_pct / 100;
+ uint64_t queue_depth_total;
int c;
VERIFY(spa_writeable(spa));
spa->spa_syncing_txg = txg;
spa->spa_sync_pass = 0;
+ mutex_enter(&spa->spa_alloc_lock);
+ VERIFY0(avl_numnodes(&spa->spa_alloc_tree));
+ mutex_exit(&spa->spa_alloc_lock);
+
/*
* If there are any pending vdev state changes, convert them
* into config changes that go out with this transaction group.
tx = dmu_tx_create_assigned(dp, txg);
spa->spa_sync_starttime = gethrtime();
- taskq_cancel_id(system_taskq, spa->spa_deadman_tqid);
- spa->spa_deadman_tqid = taskq_dispatch_delay(system_taskq,
+ taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
+ spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq,
spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() +
NSEC_TO_TICK(spa->spa_deadman_synctime));
}
}
+ /*
+ * Set the top-level vdev's max queue depth. Evaluate each
+ * top-level's async write queue depth in case it changed.
+ * The max queue depth will not change in the middle of syncing
+ * out this txg.
+ */
+ queue_depth_total = 0;
+ for (c = 0; c < rvd->vdev_children; c++) {
+ vdev_t *tvd = rvd->vdev_child[c];
+ metaslab_group_t *mg = tvd->vdev_mg;
+
+ if (mg == NULL || mg->mg_class != spa_normal_class(spa) ||
+ !metaslab_group_initialized(mg))
+ continue;
+
+ /*
+ * It is safe to do a lock-free check here because only async
+ * allocations look at mg_max_alloc_queue_depth, and async
+ * allocations all happen from spa_sync().
+ */
+ ASSERT0(refcount_count(&mg->mg_alloc_queue_depth));
+ mg->mg_max_alloc_queue_depth = max_queue_depth;
+ queue_depth_total += mg->mg_max_alloc_queue_depth;
+ }
+ mc = spa_normal_class(spa);
+ ASSERT0(refcount_count(&mc->mc_alloc_slots));
+ mc->mc_alloc_max_slots = queue_depth_total;
+ mc->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
+
+ ASSERT3U(mc->mc_alloc_max_slots, <=,
+ max_queue_depth * rvd->vdev_children);
+
/*
* Iterate to convergence.
*/
if (error == 0)
break;
- zio_suspend(spa, NULL);
+ zio_suspend(spa, NULL, ZIO_SUSPEND_IOERR);
zio_resume_wait(spa);
}
dmu_tx_commit(tx);
- taskq_cancel_id(system_taskq, spa->spa_deadman_tqid);
+ taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
spa->spa_deadman_tqid = 0;
/*
spa->spa_config_syncing = NULL;
}
- spa->spa_ubsync = spa->spa_uberblock;
-
dsl_pool_sync_done(dp, txg);
+ mutex_enter(&spa->spa_alloc_lock);
+ VERIFY0(avl_numnodes(&spa->spa_alloc_tree));
+ mutex_exit(&spa->spa_alloc_lock);
+
/*
* Update usable space statistics.
*/
spa->spa_sync_pass = 0;
+ /*
+ * Update the last synced uberblock here. We want to do this at
+ * the end of spa_sync() so that consumers of spa_last_synced_txg()
+ * will be guaranteed that all the processing associated with
+ * that txg has been completed.
+ */
+ spa->spa_ubsync = spa->spa_uberblock;
spa_config_exit(spa, SCL_CONFIG, FTAG);
spa_handle_ignored_writes(spa);
return (B_FALSE);
}
+static sysevent_t *
+spa_event_create(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, const char *name)
+{
+ sysevent_t *ev = NULL;
+#ifdef _KERNEL
+ nvlist_t *resource;
+
+ resource = zfs_event_create(spa, vd, FM_SYSEVENT_CLASS, name, hist_nvl);
+ if (resource) {
+ ev = kmem_alloc(sizeof (sysevent_t), KM_SLEEP);
+ ev->resource = resource;
+ }
+#endif
+ return (ev);
+}
+
+static void
+spa_event_post(sysevent_t *ev)
+{
+#ifdef _KERNEL
+ if (ev) {
+ zfs_zevent_post(ev->resource, NULL, zfs_zevent_post_cb);
+ kmem_free(ev, sizeof (*ev));
+ }
+#endif
+}
+
/*
* Post a zevent corresponding to the given sysevent. The 'name' must be one
* of the event definitions in sys/sysevent/eventdefs.h. The payload will be
* or zdb as real changes.
*/
void
-spa_event_notify(spa_t *spa, vdev_t *vd, const char *name)
+spa_event_notify(spa_t *spa, vdev_t *vd, nvlist_t *hist_nvl, const char *name)
{
- zfs_post_sysevent(spa, vd, name);
+ spa_event_post(spa_event_create(spa, vd, hist_nvl, name));
}
#if defined(_KERNEL) && defined(HAVE_SPL)
MODULE_PARM_DESC(spa_load_verify_data,
"Set to traverse data on pool import");
+/* CSTYLED */
module_param(zio_taskq_batch_pct, uint, 0444);
MODULE_PARM_DESC(zio_taskq_batch_pct,
"Percentage of CPUs to run an IO worker thread");