/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
- * Copyright (c) 2011 by Delphix. All rights reserved.
+ * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright 2013 Nexenta Systems, Inc. All rights reserved.
*/
/*
+ * SPA: Storage Pool Allocator
+ *
* This file contains all the routines used when modifying on-disk SPA state.
* This includes opening, importing, destroying, exporting a pool, and syncing a
* pool.
#include <sys/spa_boot.h>
#include <sys/zfs_ioctl.h>
#include <sys/dsl_scan.h>
+#include <sys/zfeature.h>
+#include <sys/dsl_destroy.h>
+#include <sys/zvol.h>
#ifdef _KERNEL
#include <sys/bootprops.h>
#include "zfs_comutil.h"
typedef enum zti_modes {
- zti_mode_fixed, /* value is # of threads (min 1) */
- zti_mode_online_percent, /* value is % of online CPUs */
- zti_mode_batch, /* cpu-intensive; value is ignored */
- zti_mode_null, /* don't create a taskq */
- zti_nmodes
+ ZTI_MODE_FIXED, /* value is # of threads (min 1) */
+ ZTI_MODE_BATCH, /* cpu-intensive; value is ignored */
+ ZTI_MODE_NULL, /* don't create a taskq */
+ ZTI_NMODES
} zti_modes_t;
-#define ZTI_FIX(n) { zti_mode_fixed, (n) }
-#define ZTI_PCT(n) { zti_mode_online_percent, (n) }
-#define ZTI_BATCH { zti_mode_batch, 0 }
-#define ZTI_NULL { zti_mode_null, 0 }
+#define ZTI_P(n, q) { ZTI_MODE_FIXED, (n), (q) }
+#define ZTI_PCT(n) { ZTI_MODE_ONLINE_PERCENT, (n), 1 }
+#define ZTI_BATCH { ZTI_MODE_BATCH, 0, 1 }
+#define ZTI_NULL { ZTI_MODE_NULL, 0, 0 }
-#define ZTI_ONE ZTI_FIX(1)
+#define ZTI_N(n) ZTI_P(n, 1)
+#define ZTI_ONE ZTI_N(1)
typedef struct zio_taskq_info {
- enum zti_modes zti_mode;
+ zti_modes_t zti_mode;
uint_t zti_value;
+ uint_t zti_count;
} zio_taskq_info_t;
static const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = {
};
/*
- * Define the taskq threads for the following I/O types:
- * NULL, READ, WRITE, FREE, CLAIM, and IOCTL
+ * This table defines the taskq settings for each ZFS I/O type. When
+ * initializing a pool, we use this table to create an appropriately sized
+ * taskq. Some operations are low volume and therefore have a small, static
+ * number of threads assigned to their taskqs using the ZTI_N(#) or ZTI_ONE
+ * macros. Other operations process a large amount of data; the ZTI_BATCH
+ * macro causes us to create a taskq oriented for throughput. Some operations
+ * are so high frequency and short-lived that the taskq itself can become a a
+ * point of lock contention. The ZTI_P(#, #) macro indicates that we need an
+ * additional degree of parallelism specified by the number of threads per-
+ * taskq and the number of taskqs; when dispatching an event in this case, the
+ * particular taskq is chosen at random.
+ *
+ * The different taskq priorities are to handle the different contexts (issue
+ * and interrupt) and then to reserve threads for ZIO_PRIORITY_NOW I/Os that
+ * need to be handled with minimum delay.
*/
const zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = {
/* ISSUE ISSUE_HIGH INTR INTR_HIGH */
- { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL },
- { ZTI_FIX(8), ZTI_NULL, ZTI_BATCH, ZTI_NULL },
- { ZTI_BATCH, ZTI_FIX(5), ZTI_FIX(16), ZTI_FIX(5) },
- { ZTI_PCT(100), ZTI_NULL, ZTI_ONE, ZTI_NULL },
- { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL },
- { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL },
+ { ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* NULL */
+ { ZTI_N(8), ZTI_NULL, ZTI_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 */
};
-static dsl_syncfunc_t spa_sync_props;
+static void spa_sync_version(void *arg, dmu_tx_t *tx);
+static void spa_sync_props(void *arg, dmu_tx_t *tx);
static boolean_t spa_has_active_shared_spare(spa_t *spa);
static inline int spa_load_impl(spa_t *spa, uint64_t, nvlist_t *config,
spa_load_state_t state, spa_import_type_t type, boolean_t mosconfig,
char **ereport);
static void spa_vdev_resilver_done(spa_t *spa);
-uint_t zio_taskq_batch_pct = 100; /* 1 thread per cpu in pset */
+uint_t zio_taskq_batch_pct = 75; /* 1 thread per cpu in pset */
id_t zio_taskq_psrset_bind = PS_NONE;
boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */
uint_t zio_taskq_basedc = 80; /* base duty cycle */
const char *propname = zpool_prop_to_name(prop);
nvlist_t *propval;
- VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_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 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) {
+ 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_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);
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);
zap_attribute_t za;
int err;
- err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP);
+ err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_PUSHPAGE);
if (err)
- return err;
+ return (err);
mutex_enter(&spa->spa_props_lock);
dsl_dataset_t *ds = NULL;
dp = spa_get_dsl(spa);
- rw_enter(&dp->dp_config_rwlock, RW_READER);
+ dsl_pool_config_enter(dp, FTAG);
if ((err = dsl_dataset_hold_obj(dp,
za.za_first_integer, FTAG, &ds))) {
- rw_exit(&dp->dp_config_rwlock);
+ dsl_pool_config_exit(dp, FTAG);
break;
}
strval = kmem_alloc(
MAXNAMELEN + strlen(MOS_DIR_NAME) + 1,
- KM_SLEEP);
+ KM_PUSHPAGE);
dsl_dataset_name(ds, strval);
dsl_dataset_rele(ds, FTAG);
- rw_exit(&dp->dp_config_rwlock);
+ dsl_pool_config_exit(dp, FTAG);
} else {
strval = NULL;
intval = za.za_first_integer;
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 = 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 = SET_ERROR(EINVAL);
+ break;
+ }
- propname = nvpair_name(elem);
+ /*
+ * Sanitize the input.
+ */
+ if (nvpair_type(elem) != DATA_TYPE_UINT64) {
+ error = SET_ERROR(EINVAL);
+ break;
+ }
- if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL)
- return (EINVAL);
+ if (nvpair_value_uint64(elem, &intval) != 0) {
+ error = SET_ERROR(EINVAL);
+ break;
+ }
+
+ if (intval != 0) {
+ error = SET_ERROR(EINVAL);
+ break;
+ }
+
+ fname = strchr(propname, '@') + 1;
+ if (zfeature_lookup_name(fname, NULL) != 0) {
+ error = SET_ERROR(EINVAL);
+ break;
+ }
+
+ has_feature = B_TRUE;
+ break;
- switch (prop) {
case ZPOOL_PROP_VERSION:
error = nvpair_value_uint64(elem, &intval);
if (!error &&
- (intval < spa_version(spa) || intval > SPA_VERSION))
- error = EINVAL;
+ (intval < spa_version(spa) ||
+ intval > SPA_VERSION_BEFORE_FEATURES ||
+ has_feature))
+ error = SET_ERROR(EINVAL);
break;
case ZPOOL_PROP_DELEGATION:
case ZPOOL_PROP_AUTOEXPAND:
error = nvpair_value_uint64(elem, &intval);
if (!error && intval > 1)
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
case ZPOOL_PROP_BOOTFS:
* the bootfs property cannot be set.
*/
if (spa_version(spa) < SPA_VERSION_BOOTFS) {
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
break;
}
* Make sure the vdev config is bootable
*/
if (!vdev_is_bootable(spa->spa_root_vdev)) {
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
break;
}
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_hold(strval,FTAG,&os)))
+ error = dmu_objset_hold(strval, FTAG, &os);
+ if (error)
break;
/* 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,
+ error = SET_ERROR(ENOTSUP);
+ } else if ((error =
+ dsl_prop_get_int_ds(dmu_objset_ds(os),
zfs_prop_to_name(ZFS_PROP_COMPRESSION),
- &compress, NULL)) == 0 &&
+ &compress)) == 0 &&
!BOOTFS_COMPRESS_VALID(compress)) {
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
} else {
objnum = dmu_objset_id(os);
}
error = nvpair_value_uint64(elem, &intval);
if (!error && (intval < ZIO_FAILURE_MODE_WAIT ||
intval > ZIO_FAILURE_MODE_PANIC))
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
/*
* This is a special case which only occurs when
*/
if (!error && spa_suspended(spa)) {
spa->spa_failmode = intval;
- error = EIO;
+ error = SET_ERROR(EIO);
}
break;
break;
if (strval[0] != '/') {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
strcmp(slash, "/..") == 0)
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
+ break;
+
+ case ZPOOL_PROP_COMMENT:
+ if ((error = nvpair_value_string(elem, &strval)) != 0)
+ break;
+ for (check = strval; *check != '\0'; check++) {
+ if (!isprint(*check)) {
+ error = SET_ERROR(EINVAL);
+ break;
+ }
+ check++;
+ }
+ if (strlen(strval) > ZPROP_MAX_COMMENT)
+ error = SET_ERROR(E2BIG);
break;
case ZPOOL_PROP_DEDUPDITTO:
if (spa_version(spa) < SPA_VERSION_DEDUP)
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
else
error = nvpair_value_uint64(elem, &intval);
if (error == 0 &&
intval != 0 && intval < ZIO_DEDUPDITTO_MIN)
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
default:
return;
dp = kmem_alloc(sizeof (spa_config_dirent_t),
- KM_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;
+
+ /*
+ * In addition to the pool directory object, we might
+ * create the pool properties object, the features for
+ * read object, the features for write object, or the
+ * feature descriptions object.
+ */
+ error = dsl_sync_task(spa->spa_name, NULL,
+ spa_sync_version, &ver, 6);
+ if (error)
+ return (error);
+ continue;
+ }
+
need_sync = B_TRUE;
break;
}
- if (need_sync)
- return (dsl_sync_task_do(spa_get_dsl(spa), NULL, spa_sync_props,
- spa, nvp, 3));
- else
- return (0);
+ if (need_sync) {
+ return (dsl_sync_task(spa->spa_name, NULL, spa_sync_props,
+ nvp, 6));
+ }
+
+ return (0);
}
/*
}
}
+/*ARGSUSED*/
+static int
+spa_change_guid_check(void *arg, dmu_tx_t *tx)
+{
+ spa_t *spa = dmu_tx_pool(tx)->dp_spa;
+ vdev_t *rvd = spa->spa_root_vdev;
+ uint64_t vdev_state;
+ ASSERTV(uint64_t *newguid = arg);
+
+ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
+ vdev_state = rvd->vdev_state;
+ spa_config_exit(spa, SCL_STATE, FTAG);
+
+ if (vdev_state != VDEV_STATE_HEALTHY)
+ return (SET_ERROR(ENXIO));
+
+ ASSERT3U(spa_guid(spa), !=, *newguid);
+
+ return (0);
+}
+
+static void
+spa_change_guid_sync(void *arg, dmu_tx_t *tx)
+{
+ uint64_t *newguid = arg;
+ spa_t *spa = dmu_tx_pool(tx)->dp_spa;
+ uint64_t oldguid;
+ vdev_t *rvd = spa->spa_root_vdev;
+
+ oldguid = spa_guid(spa);
+
+ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
+ rvd->vdev_guid = *newguid;
+ rvd->vdev_guid_sum += (*newguid - oldguid);
+ vdev_config_dirty(rvd);
+ spa_config_exit(spa, SCL_STATE, FTAG);
+
+ spa_history_log_internal(spa, "guid change", tx, "old=%llu new=%llu",
+ 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
int
spa_change_guid(spa_t *spa)
{
- uint64_t oldguid, newguid;
- uint64_t txg;
-
- if (!(spa_mode_global & FWRITE))
- return (EROFS);
-
- txg = spa_vdev_enter(spa);
-
- if (spa->spa_root_vdev->vdev_state != VDEV_STATE_HEALTHY)
- return (spa_vdev_exit(spa, NULL, txg, ENXIO));
+ int error;
+ uint64_t guid;
- oldguid = spa_guid(spa);
- newguid = spa_generate_guid(NULL);
- ASSERT3U(oldguid, !=, newguid);
+ mutex_enter(&spa->spa_vdev_top_lock);
+ mutex_enter(&spa_namespace_lock);
+ guid = spa_generate_guid(NULL);
- spa->spa_root_vdev->vdev_guid = newguid;
- spa->spa_root_vdev->vdev_guid_sum += (newguid - oldguid);
+ error = dsl_sync_task(spa->spa_name, spa_change_guid_check,
+ spa_change_guid_sync, &guid, 5);
- vdev_config_dirty(spa->spa_root_vdev);
+ if (error == 0) {
+ spa_config_sync(spa, B_FALSE, B_TRUE);
+ spa_event_notify(spa, NULL, FM_EREPORT_ZFS_POOL_REGUID);
+ }
- spa_event_notify(spa, NULL, FM_EREPORT_ZFS_POOL_REGUID);
+ mutex_exit(&spa_namespace_lock);
+ mutex_exit(&spa->spa_vdev_top_lock);
- return (spa_vdev_exit(spa, NULL, txg, 0));
+ return (error);
}
/*
offsetof(spa_error_entry_t, se_avl));
}
-static taskq_t *
-spa_taskq_create(spa_t *spa, const char *name, enum zti_modes mode,
- uint_t value, uint_t flags)
+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;
- switch (mode) {
- case zti_mode_null:
- return (NULL); /* no taskq needed */
+ if (mode == ZTI_MODE_NULL) {
+ tqs->stqs_count = 0;
+ tqs->stqs_taskq = NULL;
+ return;
+ }
- case zti_mode_fixed:
+ ASSERT3U(count, >, 0);
+
+ tqs->stqs_count = count;
+ tqs->stqs_taskq = kmem_alloc(count * sizeof (taskq_t *), KM_SLEEP);
+
+ switch (mode) {
+ case ZTI_MODE_FIXED:
ASSERT3U(value, >=, 1);
value = MAX(value, 1);
break;
- case zti_mode_batch:
+ case ZTI_MODE_BATCH:
batch = B_TRUE;
flags |= TASKQ_THREADS_CPU_PCT;
value = zio_taskq_batch_pct;
break;
- case zti_mode_online_percent:
- flags |= TASKQ_THREADS_CPU_PCT;
- break;
-
default:
- panic("unrecognized mode for %s taskq (%u:%u) in "
+ panic("unrecognized mode for %s_%s taskq (%u:%u) in "
"spa_activate()",
- name, mode, value);
+ zio_type_name[t], zio_taskq_types[q], mode, value);
break;
}
- if (zio_taskq_sysdc && spa->spa_proc != &p0) {
- if (batch)
- flags |= TASKQ_DC_BATCH;
+ for (i = 0; i < count; i++) {
+ taskq_t *tq;
+
+ if (count > 1) {
+ (void) snprintf(name, sizeof (name), "%s_%s_%u",
+ zio_type_name[t], zio_taskq_types[q], i);
+ } else {
+ (void) snprintf(name, sizeof (name), "%s_%s",
+ zio_type_name[t], zio_taskq_types[q]);
+ }
+
+ if (zio_taskq_sysdc && spa->spa_proc != &p0) {
+ if (batch)
+ flags |= TASKQ_DC_BATCH;
+
+ tq = taskq_create_sysdc(name, value, 50, INT_MAX,
+ spa->spa_proc, zio_taskq_basedc, flags);
+ } else {
+ pri_t pri = maxclsyspri;
+ /*
+ * The write issue taskq can be extremely CPU
+ * intensive. Run it at slightly lower priority
+ * than the other taskqs.
+ */
+ if (t == ZIO_TYPE_WRITE && q == ZIO_TASKQ_ISSUE)
+ pri--;
+
+ tq = taskq_create_proc(name, value, pri, 50,
+ INT_MAX, spa->spa_proc, flags);
+ }
+
+ tqs->stqs_taskq[i] = tq;
+ }
+}
+
+static void
+spa_taskqs_fini(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
+{
+ spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
+ uint_t i;
+
+ if (tqs->stqs_taskq == NULL) {
+ ASSERT3U(tqs->stqs_count, ==, 0);
+ return;
+ }
+
+ for (i = 0; i < tqs->stqs_count; i++) {
+ ASSERT3P(tqs->stqs_taskq[i], !=, NULL);
+ taskq_destroy(tqs->stqs_taskq[i]);
+ }
+
+ kmem_free(tqs->stqs_taskq, tqs->stqs_count * sizeof (taskq_t *));
+ tqs->stqs_taskq = NULL;
+}
+
+/*
+ * Dispatch a task to the appropriate taskq for the ZFS I/O type and priority.
+ * Note that a type may have multiple discrete taskqs to avoid lock contention
+ * on the taskq itself. In that case we choose which taskq at random by using
+ * the low bits of gethrtime().
+ */
+void
+spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
+ task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent)
+{
+ spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
+ taskq_t *tq;
+
+ ASSERT3P(tqs->stqs_taskq, !=, NULL);
+ ASSERT3U(tqs->stqs_count, !=, 0);
- return (taskq_create_sysdc(name, value, 50, INT_MAX,
- spa->spa_proc, zio_taskq_basedc, flags));
+ if (tqs->stqs_count == 1) {
+ tq = tqs->stqs_taskq[0];
+ } else {
+ tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count];
}
- return (taskq_create_proc(name, value, maxclsyspri, 50, INT_MAX,
- spa->spa_proc, flags));
+
+ 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
for (t = 0; t < ZIO_TYPES; t++) {
for (q = 0; q < ZIO_TASKQ_TYPES; q++) {
- const zio_taskq_info_t *ztip = &zio_taskqs[t][q];
- enum zti_modes mode = ztip->zti_mode;
- uint_t value = ztip->zti_value;
- uint_t flags = 0;
- char name[32];
-
- if (t == ZIO_TYPE_WRITE)
- flags |= TASKQ_NORECLAIM;
-
- (void) snprintf(name, sizeof (name),
- "%s_%s", zio_type_name[t], zio_taskq_types[q]);
-
- spa->spa_zio_taskq[t][q] =
- spa_taskq_create(spa, name, mode, value, flags);
+ spa_taskqs_init(spa, t, q);
}
}
}
list_destroy(&spa->spa_config_dirty_list);
list_destroy(&spa->spa_state_dirty_list);
+ taskq_cancel_id(system_taskq, spa->spa_deadman_tqid);
+
for (t = 0; t < ZIO_TYPES; t++) {
for (q = 0; q < ZIO_TASKQ_TYPES; q++) {
- if (spa->spa_zio_taskq[t][q] != NULL)
- taskq_destroy(spa->spa_zio_taskq[t][q]);
- spa->spa_zio_taskq[t][q] = NULL;
+ spa_taskqs_fini(spa, t, q);
}
}
if (error) {
vdev_free(*vdp);
*vdp = NULL;
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
}
for (c = 0; c < children; c++) {
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, VDEV_CONFIG_SPARE);
uint_t nl2cache;
int i, j, oldnvdevs;
uint64_t guid;
- vdev_t *vd, **oldvdevs, **newvdevs = NULL;
+ vdev_t *vd, **oldvdevs, **newvdevs;
spa_aux_vdev_t *sav = &spa->spa_l2cache;
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);
if (sav->sav_config != NULL) {
VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0);
- newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_SLEEP);
+ newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_PUSHPAGE);
} else {
nl2cache = 0;
+ newvdevs = NULL;
}
oldvdevs = sav->sav_vdevs;
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, VDEV_CONFIG_L2CACHE);
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 | KM_NODEBUG);
+ 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)
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)) {
+ if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd) &&
+ !vd->vdev_ishole) {
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);
uint64_t idx = 0;
child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t **),
- KM_SLEEP);
- VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
+ 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];
/*
* Check for missing log devices
*/
-static int
+static boolean_t
spa_check_logs(spa_t *spa)
{
+ boolean_t rv = B_FALSE;
+
switch (spa->spa_log_state) {
default:
break;
case SPA_LOG_MISSING:
/* need to recheck in case slog has been restored */
case SPA_LOG_UNKNOWN:
- if (dmu_objset_find(spa->spa_name, zil_check_log_chain, NULL,
- DS_FIND_CHILDREN)) {
+ rv = (dmu_objset_find(spa->spa_name, zil_check_log_chain,
+ NULL, DS_FIND_CHILDREN) != 0);
+ if (rv)
spa_set_log_state(spa, SPA_LOG_MISSING);
- return (1);
- }
break;
}
- return (0);
+ return (rv);
}
static boolean_t
int
spa_offline_log(spa_t *spa)
{
- int error = 0;
-
- if ((error = dmu_objset_find(spa_name(spa), zil_vdev_offline,
- NULL, DS_FIND_CHILDREN)) == 0) {
+ int error;
+ error = dmu_objset_find(spa_name(spa), zil_vdev_offline,
+ NULL, DS_FIND_CHILDREN);
+ if (error == 0) {
/*
* We successfully offlined the log device, sync out the
* current txg so that the "stubby" block can be removed
int error = zio->io_error;
if (error) {
- if ((BP_GET_LEVEL(bp) != 0 || dmu_ot[type].ot_metadata) &&
+ 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
/*ARGSUSED*/
static int
spa_load_verify_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
- arc_buf_t *pbuf, const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg)
+ const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg)
{
if (bp != NULL) {
zio_t *rio = arg;
if (error) {
if (error != ENXIO && error != EIO)
- error = EIO;
+ error = SET_ERROR(EIO);
return (error);
}
&glist, &gcount) != 0)
return;
- vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_SLEEP);
+ vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_PUSHPAGE);
/* attempt to online all the vdevs & validate */
attempt_reopen = B_TRUE;
{
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_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid))
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
+
+ ASSERT(spa->spa_comment == NULL);
+ if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
+ spa->spa_comment = spa_strdup(comment);
/*
* Versioning wasn't explicitly added to the label until later, so if
if ((state == SPA_LOAD_IMPORT || state == SPA_LOAD_TRYIMPORT) &&
spa_guid_exists(pool_guid, 0)) {
- error = EEXIST;
+ error = SET_ERROR(EEXIST);
} else {
spa->spa_config_guid = pool_guid;
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT,
&nvl) == 0) {
VERIFY(nvlist_dup(nvl, &spa->spa_config_splitting,
- KM_SLEEP) == 0);
+ 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);
{
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.
spa->spa_load_state = state;
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvroot))
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
parse = (type == SPA_IMPORT_EXISTING ?
VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT);
return (error);
if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN)
- return (ENXIO);
+ return (SET_ERROR(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)
+ if (ub->ub_txg == 0) {
+ 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)
+ 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
spa->spa_claim_max_txg = spa->spa_first_txg;
spa->spa_prev_software_version = ub->ub_software_version;
- error = dsl_pool_open(spa, spa->spa_first_txg, &spa->spa_dsl_pool);
+ 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 (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) {
uint64_t hostid;
nvlist_t *policy = NULL, *nvconfig;
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",
+ "loaded as it was last accessed by another "
+ "system (host: %s hostid: 0x%lx). See: "
+ "http://zfsonlinux.org/msg/ZFS-8000-EY",
spa_name(spa), hostname,
(unsigned long)hostid);
- return (EBADF);
+ return (SET_ERROR(EBADF));
}
}
if (nvlist_lookup_nvlist(spa->spa_config,
nvlist_free(nvconfig);
/*
- * Now that we've validate the config, check the state of the
+ * 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);
+ return (SET_ERROR(ENXIO));
if (spa_check_logs(spa)) {
*ereport = FM_EREPORT_ZFS_LOG_REPLAY;
}
}
+ 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.
vdev_resilver_needed(rvd, NULL, NULL))
spa_async_request(spa, SPA_ASYNC_RESILVER);
+ /*
+ * Log the fact that we booted up (so that we can detect if
+ * we rebooted in the middle of an operation).
+ */
+ spa_history_log_version(spa, "open");
+
/*
* Delete any inconsistent datasets.
*/
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;
return (load_error);
}
- /* Price of rolling back is discarding txgs, including log */
- if (state == SPA_LOAD_RECOVER)
+ 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;
if (config && (rewind_error || state != SPA_LOAD_RECOVER))
spa_config_set(spa, config);
- return (state == SPA_LOAD_RECOVER ? rewind_error : load_error);
+ 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);
+ }
}
/*
spa_load_state_t state = SPA_LOAD_OPEN;
int error;
int locked = B_FALSE;
+ int firstopen = B_FALSE;
*spapp = NULL;
if ((spa = spa_lookup(pool)) == NULL) {
if (locked)
mutex_exit(&spa_namespace_lock);
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
}
if (spa->spa_state == POOL_STATE_UNINITIALIZED) {
zpool_rewind_policy_t policy;
+ firstopen = B_TRUE;
+
zpool_get_rewind_policy(nvpolicy ? nvpolicy : spa->spa_config,
&policy);
if (policy.zrp_request & ZPOOL_DO_REWIND)
spa_remove(spa);
if (locked)
mutex_exit(&spa_namespace_lock);
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
}
if (error) {
*/
if (config != NULL && spa->spa_config) {
VERIFY(nvlist_dup(spa->spa_config, config,
- KM_SLEEP) == 0);
+ KM_PUSHPAGE) == 0);
VERIFY(nvlist_add_nvlist(*config,
ZPOOL_CONFIG_LOAD_INFO,
spa->spa_load_info) == 0);
mutex_exit(&spa_namespace_lock);
}
+#ifdef _KERNEL
+ if (firstopen)
+ zvol_create_minors(spa->spa_name);
+#endif
+
*spapp = spa;
return (0);
}
}
+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;
spa_add_spares(spa, *config);
spa_add_l2cache(spa, *config);
+ spa_add_feature_stats(spa, *config);
}
}
return (0);
if (ndev == 0)
- return (EINVAL);
+ return (SET_ERROR(EINVAL));
/*
* Make sure the pool is formatted with a version that supports this
* device type.
*/
if (spa_version(spa) < version)
- return (ENOTSUP);
+ return (SET_ERROR(ENOTSUP));
/*
* Set the pending device list so we correctly handle device in-use
if (!vd->vdev_ops->vdev_op_leaf) {
vdev_free(vd);
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
goto out;
}
#ifdef _KERNEL
if ((strcmp(config, ZPOOL_CONFIG_L2CACHE) == 0) &&
strcmp(vd->vdev_ops->vdev_op_type, VDEV_TYPE_DISK) != 0) {
- error = ENOTBLK;
+ error = SET_ERROR(ENOTBLK);
vdev_free(vd);
goto out;
}
&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);
}
*/
int
spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
- const char *history_str, nvlist_t *zplprops)
+ nvlist_t *zplprops)
{
spa_t *spa;
char *altroot = NULL;
nvlist_t **spares, **l2cache;
uint_t nspares, nl2cache;
uint64_t version, obj;
+ boolean_t has_features;
+ nvpair_t *elem;
int c;
/*
mutex_enter(&spa_namespace_lock);
if (spa_lookup(pool) != NULL) {
mutex_exit(&spa_namespace_lock);
- return (EEXIST);
+ return (SET_ERROR(EEXIST));
}
/*
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;
ASSERT(error != 0 || spa->spa_root_vdev == rvd);
if (error == 0 && !zfs_allocatable_devs(nvroot))
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
if (error == 0 &&
(error = vdev_create(rvd, txg, B_FALSE)) == 0 &&
if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
&spares, &nspares) == 0) {
VERIFY(nvlist_alloc(&spa->spa_spares.sav_config, NV_UNIQUE_NAME,
- KM_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).
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) {
if (props != NULL) {
spa_configfile_set(spa, props, B_FALSE);
- spa_sync_props(spa, props, tx);
+ spa_sync_props(props, tx);
}
dmu_tx_commit(tx);
spa_config_sync(spa, B_FALSE, B_TRUE);
- if (version >= SPA_VERSION_ZPOOL_HISTORY && history_str != NULL)
- (void) spa_history_log(spa, history_str, LOG_CMD_POOL_CREATE);
- spa_history_log_version(spa, LOG_POOL_CREATE);
+ spa_history_log_version(spa, "create");
spa->spa_minref = refcount_count(&spa->spa_refcount);
/*
* 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);
}
#endif
if (config == NULL) {
- cmn_err(CE_NOTE, "Can not read the pool label from '%s'",
+ cmn_err(CE_NOTE, "Cannot read the pool label from '%s'",
devpath);
- return (EIO);
+ return (SET_ERROR(EIO));
}
VERIFY(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
if ((bvd = vdev_lookup_by_guid(rvd, guid)) == NULL) {
cmn_err(CE_NOTE, "Can not find the boot vdev for guid %llu",
(u_longlong_t)guid);
- error = ENOENT;
+ error = SET_ERROR(ENOENT);
goto out;
}
if (avd != bvd) {
cmn_err(CE_NOTE, "The boot device is 'degraded'. Please "
"try booting from '%s'", avd->vdev_path);
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
goto out;
}
"try booting from '%s'",
bvd->vdev_parent->
vdev_child[bvd->vdev_parent->vdev_children - 1]->vdev_path);
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
goto out;
}
error = 0;
- spa_history_log_version(spa, LOG_POOL_IMPORT);
out:
spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
vdev_free(rvd);
* Import a non-root pool into the system.
*/
int
-spa_import(const char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags)
+spa_import(char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags)
{
spa_t *spa;
char *altroot = NULL;
mutex_enter(&spa_namespace_lock);
if (spa_lookup(pool) != NULL) {
mutex_exit(&spa_namespace_lock);
- return (EEXIST);
+ return (SET_ERROR(EEXIST));
}
/*
spa_config_sync(spa, B_FALSE, B_TRUE);
mutex_exit(&spa_namespace_lock);
- spa_history_log_version(spa, LOG_POOL_IMPORT);
+ spa_history_log_version(spa, "import");
return (0);
}
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_async_request(spa, SPA_ASYNC_AUTOEXPAND);
mutex_exit(&spa_namespace_lock);
- spa_history_log_version(spa, LOG_POOL_IMPORT);
+ spa_history_log_version(spa, "import");
+
+#ifdef _KERNEL
+ zvol_create_minors(pool);
+#endif
return (0);
}
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;
+
+ dsname = kmem_alloc(MAXPATHLEN, KM_PUSHPAGE);
cp = strchr(tmpname, '/');
if (cp == NULL) {
*oldconfig = NULL;
if (!(spa_mode_global & FWRITE))
- return (EROFS);
+ return (SET_ERROR(EROFS));
mutex_enter(&spa_namespace_lock);
if ((spa = spa_lookup(pool)) == NULL) {
mutex_exit(&spa_namespace_lock);
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
}
/*
new_state != POOL_STATE_UNINITIALIZED)) {
spa_async_resume(spa);
mutex_exit(&spa_namespace_lock);
- return (EBUSY);
+ return (SET_ERROR(EBUSY));
}
/*
spa_has_active_shared_spare(spa)) {
spa_async_resume(spa);
mutex_exit(&spa_namespace_lock);
- return (EXDEV);
+ return (SET_ERROR(EXDEV));
}
/*
spa_vdev_attach(spa_t *spa, uint64_t guid, nvlist_t *nvroot, int replacing)
{
uint64_t txg, dtl_max_txg;
- ASSERTV(vdev_t *rvd = spa->spa_root_vdev;)
vdev_t *oldvd, *newvd, *newrootvd, *pvd, *tvd;
vdev_ops_t *pvops;
char *oldvdpath, *newvdpath;
int newvd_isspare;
int error;
+ ASSERTV(vdev_t *rvd = spa->spa_root_vdev);
ASSERT(spa_writeable(spa));
if (strcmp(oldvd->vdev_path, newvd->vdev_path) == 0) {
spa_strfree(oldvd->vdev_path);
oldvd->vdev_path = kmem_alloc(strlen(newvd->vdev_path) + 5,
- KM_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;
+ newvd->vdev_resilver_txg = txg;
/*
* If the parent is not a mirror, or if we're replacing, insert the new
*/
(void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0);
- spa_history_log_internal(LOG_POOL_VDEV_ATTACH, spa, NULL,
+ spa_history_log_internal(spa, "vdev attach", NULL,
"%s vdev=%s %s vdev=%s",
replacing && newvd_isspare ? "spare in" :
replacing ? "replace" : "attach", newvdpath,
/*
* Detach a device from a mirror or replacing vdev.
+ *
* If 'replace_done' is specified, only detach if the parent
* is a replacing vdev.
*/
{
uint64_t txg;
int error;
- ASSERTV(vdev_t *rvd = spa->spa_root_vdev;)
vdev_t *vd, *pvd, *cvd, *tvd;
boolean_t unspare = B_FALSE;
uint64_t unspare_guid = 0;
char *vdpath;
int c, t;
-
+ ASSERTV(vdev_t *rvd = spa->spa_root_vdev);
ASSERT(spa_writeable(spa));
txg = spa_vdev_enter(spa);
if (pvd->vdev_ops == &vdev_spare_ops)
cvd->vdev_unspare = B_FALSE;
vdev_remove_parent(cvd);
- cvd->vdev_resilvering = B_FALSE;
}
error = spa_vdev_exit(spa, vd, txg, 0);
- spa_history_log_internal(LOG_POOL_VDEV_DETACH, spa, NULL,
+ spa_history_log_internal(spa, "detach", NULL,
"vdev=%s", vdpath);
spa_strfree(vdpath);
nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0)
return (spa_vdev_exit(spa, NULL, txg, EINVAL));
- vml = kmem_zalloc(children * sizeof (vdev_t *), KM_SLEEP);
- glist = kmem_zalloc(children * sizeof (uint64_t), KM_SLEEP);
+ 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++) {
spa->spa_root_vdev->vdev_child[c]->vdev_islog) {
continue;
} else {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
}
/* which disk is going to be split? */
if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID,
&glist[c]) != 0) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
/* look it up in the spa */
vml[c] = spa_lookup_by_guid(spa, glist[c], B_FALSE);
if (vml[c] == NULL) {
- error = ENODEV;
+ error = SET_ERROR(ENODEV);
break;
}
vml[c]->vdev_children != 0 ||
vml[c]->vdev_state != VDEV_STATE_HEALTHY ||
c != spa->spa_root_vdev->vdev_child[c]->vdev_id) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
break;
}
if (vdev_dtl_required(vml[c])) {
- error = EBUSY;
+ error = SET_ERROR(EBUSY);
break;
}
* 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_SLEEP) == 0);
+ 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));
/* 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_SLEEP) == 0);
+ 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,
if (vml[c] != NULL) {
vdev_split(vml[c]);
if (error == 0)
- spa_history_log_internal(LOG_POOL_VDEV_DETACH,
- spa, tx, "vdev=%s",
- vml[c]->vdev_path);
+ spa_history_log_internal(spa, "detach", tx,
+ "vdev=%s", vml[c]->vdev_path);
vdev_free(vml[c]);
}
}
zio_handle_panic_injection(spa, FTAG, 3);
/* split is complete; log a history record */
- spa_history_log_internal(LOG_POOL_SPLIT, newspa, NULL,
- "split new pool %s from pool %s", newname, spa_name(spa));
+ spa_history_log_internal(newspa, "split", NULL,
+ "from pool %s", spa_name(spa));
kmem_free(vml, children * sizeof (vdev_t *));
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 (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);
if (vd->vdev_stat.vs_alloc != 0)
error = spa_offline_log(spa);
} else {
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
}
if (error)
* The evacuation succeeded. Remove any remaining MOS metadata
* associated with this vdev, and wait for these changes to sync.
*/
- ASSERT3U(vd->vdev_stat.vs_alloc, ==, 0);
+ ASSERT0(vd->vdev_stat.vs_alloc);
txg = spa_vdev_config_enter(spa);
vd->vdev_removing = B_TRUE;
vdev_dirty(vd, 0, NULL, txg);
* the spa_vdev_config_[enter/exit] functions which allow us to
* grab and release the spa_config_lock while still holding the namespace
* lock. During each step the configuration is synced out.
- */
-
-/*
- * Remove a device from the pool. Currently, this supports removing only hot
- * spares, slogs, and level 2 ARC devices.
+ *
+ * Currently, this supports removing only hot spares, slogs, and level 2 ARC
+ * devices.
*/
int
spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare)
spa_load_spares(spa);
spa->spa_spares.sav_sync = B_TRUE;
} else {
- error = EBUSY;
+ error = SET_ERROR(EBUSY);
}
} else if (spa->spa_l2cache.sav_vdevs != NULL &&
nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
/*
* Normal vdevs cannot be removed (yet).
*/
- error = ENOTSUP;
+ error = SET_ERROR(ENOTSUP);
} else {
/*
* There is no vdev of any kind with the specified guid.
*/
- error = ENOENT;
+ error = SET_ERROR(ENOENT);
}
if (!locked)
/*
* Find any device that's done replacing, or a vdev marked 'unspare' that's
- * current spared, so we can detach it.
+ * currently spared, so we can detach it.
*/
static vdev_t *
spa_vdev_resilver_done_hunt(vdev_t *vd)
ASSERT(pvd->vdev_ops == &vdev_replacing_ops);
sguid = ppvd->vdev_child[1]->vdev_guid;
}
+ ASSERT(vd->vdev_resilver_txg == 0 || !vdev_dtl_required(vd));
+
spa_config_exit(spa, SCL_ALL, FTAG);
if (spa_vdev_detach(spa, guid, pguid, B_TRUE) != 0)
return;
{
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
if (dsl_scan_resilvering(spa->spa_dsl_pool))
- return (EBUSY);
+ return (SET_ERROR(EBUSY));
return (dsl_scan_cancel(spa->spa_dsl_pool));
}
ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
if (func >= POOL_SCAN_FUNCS || func == POOL_SCAN_NONE)
- return (ENOTSUP);
+ return (SET_ERROR(ENOTSUP));
/*
* If a resilver was requested, but there is no DTL on a
* then log an internal history event.
*/
if (new_space != old_space) {
- spa_history_log_internal(LOG_POOL_VDEV_ONLINE,
- spa, NULL,
+ spa_history_log_internal(spa, "vdev online", NULL,
"pool '%s' size: %llu(+%llu)",
spa_name(spa), new_space, new_space - old_space);
}
return (0);
}
+/*
+ * Note: this simple function is not inlined to make it easier to dtrace the
+ * amount of time spent syncing frees.
+ */
+static void
+spa_sync_frees(spa_t *spa, bplist_t *bpl, dmu_tx_t *tx)
+{
+ zio_t *zio = zio_root(spa, NULL, NULL, 0);
+ bplist_iterate(bpl, spa_free_sync_cb, zio, tx);
+ VERIFY(zio_wait(zio) == 0);
+}
+
+/*
+ * Note: this simple function is not inlined to make it easier to dtrace the
+ * amount of time spent syncing deferred frees.
+ */
+static void
+spa_sync_deferred_frees(spa_t *spa, dmu_tx_t *tx)
+{
+ zio_t *zio = zio_root(spa, NULL, NULL, 0);
+ VERIFY3U(bpobj_iterate(&spa->spa_deferred_bpobj,
+ spa_free_sync_cb, zio, tx), ==, 0);
+ VERIFY0(zio_wait(zio));
+}
+
static void
spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx)
{
* 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 = vmem_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);
&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, VDEV_CONFIG_L2CACHE);
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 *arg, dmu_tx_t *tx)
+{
+ uint64_t *versionp = arg;
+ uint64_t version = *versionp;
+ spa_t *spa = dmu_tx_pool(tx)->dp_spa;
+
+ /*
+ * Setting the version is special cased when first creating the pool.
+ */
+ ASSERT(tx->tx_txg != TXG_INITIAL);
+
+ ASSERT(SPA_VERSION_IS_SUPPORTED(version));
+ ASSERT(version >= spa_version(spa));
+
+ spa->spa_uberblock.ub_version = version;
+ vdev_config_dirty(spa->spa_root_vdev);
+ spa_history_log_internal(spa, "set", tx, "version=%lld", version);
+}
+
/*
* Set zpool properties.
*/
static void
-spa_sync_props(void *arg1, void *arg2, dmu_tx_t *tx)
+spa_sync_props(void *arg, dmu_tx_t *tx)
{
- spa_t *spa = arg1;
+ nvlist_t *nvp = arg;
+ spa_t *spa = dmu_tx_pool(tx)->dp_spa;
objset_t *mos = spa->spa_meta_objset;
- nvlist_t *nvp = arg2;
- nvpair_t *elem;
- 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);
+ spa_history_log_internal(spa, "set", tx,
+ "%s=enabled", nvpair_name(elem));
+ 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:
* 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);
+ spa_history_log_internal(spa, "set", tx,
+ "%s=%s", nvpair_name(elem), strval);
+ break;
default:
/*
* Set pool property values in the poolprops mos object.
*/
if (spa->spa_pool_props_object == 0) {
- 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 */
VERIFY(zap_update(mos,
spa->spa_pool_props_object, propname,
1, strlen(strval) + 1, strval, tx) == 0);
-
+ spa_history_log_internal(spa, "set", tx,
+ "%s=%s", nvpair_name(elem), strval);
} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
VERIFY(nvpair_value_uint64(elem, &intval) == 0);
VERIFY(zap_update(mos,
spa->spa_pool_props_object, propname,
8, 1, &intval, tx) == 0);
+ spa_history_log_internal(spa, "set", tx,
+ "%s=%lld", nvpair_name(elem), intval);
} else {
ASSERT(0); /* not allowed */
}
}
}
- /* log internal history if this is not a zpool create */
- if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY &&
- tx->tx_txg != TXG_INITIAL) {
- spa_history_log_internal(LOG_POOL_PROPSET,
- spa, tx, "%s %lld %s",
- nvpair_name(elem), intval, spa_name(spa));
- }
}
mutex_exit(&spa->spa_props_lock);
ASSERT(spa->spa_sync_pass == 1);
+ rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
+
if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN &&
spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) {
dsl_pool_create_origin(dp, tx);
/* 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);
+ }
+ rrw_exit(&dp->dp_config_rwlock, FTAG);
}
/*
{
dsl_pool_t *dp = spa->spa_dsl_pool;
objset_t *mos = spa->spa_meta_objset;
- bpobj_t *defer_bpo = &spa->spa_deferred_bpobj;
bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK];
vdev_t *rvd = spa->spa_root_vdev;
vdev_t *vd;
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_PUSHPAGE, 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.
!txg_list_empty(&dp->dp_sync_tasks, txg) ||
((dsl_scan_active(dp->dp_scan) ||
txg_sync_waiting(dp)) && !spa_shutting_down(spa))) {
- zio_t *zio = zio_root(spa, NULL, NULL, 0);
- VERIFY3U(bpobj_iterate(defer_bpo,
- spa_free_sync_cb, zio, tx), ==, 0);
- VERIFY3U(zio_wait(zio), ==, 0);
+ spa_sync_deferred_frees(spa, tx);
}
/*
spa_errlog_sync(spa, txg);
dsl_pool_sync(dp, txg);
- if (pass <= SYNC_PASS_DEFERRED_FREE) {
- zio_t *zio = zio_root(spa, NULL, NULL, 0);
- bplist_iterate(free_bpl, spa_free_sync_cb,
- zio, tx);
- VERIFY(zio_wait(zio) == 0);
+ if (pass < zfs_sync_pass_deferred_free) {
+ spa_sync_frees(spa, free_bpl, tx);
} else {
bplist_iterate(free_bpl, bpobj_enqueue_cb,
- defer_bpo, tx);
+ &spa->spa_deferred_bpobj, tx);
}
ddt_sync(spa, txg);
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.
*/
* 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;