static int
vdev_root_open(vdev_t *vd, uint64_t *asize, uint64_t *max_asize,
uint64_t *ashift)
{
spa_t *spa = vd->vdev_spa;
int lasterror = 0;
int numerrors = 0;
int c;
if (vd->vdev_children == 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
vdev_open_children(vd);
for (c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
if (cvd->vdev_open_error && !cvd->vdev_islog) {
lasterror = cvd->vdev_open_error;
numerrors++;
}
}
if (spa_load_state(spa) != SPA_LOAD_NONE)
spa_set_missing_tvds(spa, numerrors);
if (too_many_errors(vd, numerrors)) {
vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
return (lasterror);
}
*asize = 0;
*max_asize = 0;
*ashift = 0;
return (0);
}
/*
* Generate the pool's configuration based on the current in-core state.
*
* We infer whether to generate a complete config or just one top-level config
* based on whether vd is the root vdev.
*/
nvlist_t *
spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
{
nvlist_t *config, *nvroot;
vdev_t *rvd = spa->spa_root_vdev;
unsigned long hostid = 0;
boolean_t locked = B_FALSE;
uint64_t split_guid;
if (vd == NULL) {
vd = rvd;
locked = B_TRUE;
spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
}
ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
(SCL_CONFIG | SCL_STATE));
/*
* If txg is -1, report the current value of spa->spa_config_txg.
*/
if (txg == -1ULL)
txg = spa->spa_config_txg;
VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
spa_version(spa)) == 0);
VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
spa_name(spa)) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
spa_state(spa)) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
txg) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
spa_guid(spa)) == 0);
VERIFY(spa->spa_comment == NULL || nvlist_add_string(config,
ZPOOL_CONFIG_COMMENT, spa->spa_comment) == 0);
#ifdef _KERNEL
hostid = 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, &hostid);
#endif /* _KERNEL */
if (hostid != 0) {
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
hostid) == 0);
}
VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
utsname.nodename) == 0);
if (vd != rvd) {
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
vd->vdev_top->vdev_guid) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
vd->vdev_guid) == 0);
if (vd->vdev_isspare)
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE,
1ULL) == 0);
if (vd->vdev_islog)
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG,
1ULL) == 0);
vd = vd->vdev_top; /* label contains top config */
} else {
/*
* Only add the (potentially large) split information
* in the mos config, and not in the vdev labels
*/
if (spa->spa_config_splitting != NULL)
VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT,
spa->spa_config_splitting) == 0);
}
/*
* Add the top-level config. We even add this on pools which
* don't support holes in the namespace.
*/
vdev_top_config_generate(spa, config);
/*
* If we're splitting, record the original pool's guid.
*/
if (spa->spa_config_splitting != NULL &&
nvlist_lookup_uint64(spa->spa_config_splitting,
ZPOOL_CONFIG_SPLIT_GUID, &split_guid) == 0) {
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID,
split_guid) == 0);
}
nvroot = vdev_config_generate(spa, vd, getstats, 0);
VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
nvlist_free(nvroot);
/*
* Store what's necessary for reading the MOS in the label.
*/
VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
spa->spa_label_features) == 0);
if (getstats && spa_load_state(spa) == SPA_LOAD_NONE) {
ddt_histogram_t *ddh;
ddt_stat_t *dds;
ddt_object_t *ddo;
ddh = kmem_zalloc(sizeof (ddt_histogram_t), KM_SLEEP);
ddt_get_dedup_histogram(spa, ddh);
VERIFY(nvlist_add_uint64_array(config,
ZPOOL_CONFIG_DDT_HISTOGRAM,
(uint64_t *)ddh, sizeof (*ddh) / sizeof (uint64_t)) == 0);
kmem_free(ddh, sizeof (ddt_histogram_t));
ddo = kmem_zalloc(sizeof (ddt_object_t), KM_SLEEP);
ddt_get_dedup_object_stats(spa, ddo);
VERIFY(nvlist_add_uint64_array(config,
ZPOOL_CONFIG_DDT_OBJ_STATS,
(uint64_t *)ddo, sizeof (*ddo) / sizeof (uint64_t)) == 0);
kmem_free(ddo, sizeof (ddt_object_t));
dds = kmem_zalloc(sizeof (ddt_stat_t), KM_SLEEP);
ddt_get_dedup_stats(spa, dds);
VERIFY(nvlist_add_uint64_array(config,
ZPOOL_CONFIG_DDT_STATS,
(uint64_t *)dds, sizeof (*dds) / sizeof (uint64_t)) == 0);
kmem_free(dds, sizeof (ddt_stat_t));
}
if (locked)
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (config);
}
/*
* Generate the pool's configuration based on the current in-core state.
*
* We infer whether to generate a complete config or just one top-level config
* based on whether vd is the root vdev.
*/
nvlist_t *
spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
{
nvlist_t *config, *nvroot;
vdev_t *rvd = spa->spa_root_vdev;
unsigned long hostid = 0;
boolean_t locked = B_FALSE;
uint64_t split_guid;
char *pool_name;
if (vd == NULL) {
vd = rvd;
locked = B_TRUE;
spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
}
ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
(SCL_CONFIG | SCL_STATE));
/*
* If txg is -1, report the current value of spa->spa_config_txg.
*/
if (txg == -1ULL)
txg = spa->spa_config_txg;
/*
* Originally, users had to handle spa namespace collisions by either
* exporting the already imported pool or by specifying a new name for
* the pool with a conflicting name. In the case of root pools from
* virtual guests, neither approach to collision resolution is
* reasonable. This is addressed by extending the new name syntax with
* an option to specify that the new name is temporary. When specified,
* ZFS_IMPORT_TEMP_NAME will be set in spa->spa_import_flags to tell us
* to use the previous name, which we do below.
*/
if (spa->spa_import_flags & ZFS_IMPORT_TEMP_NAME) {
VERIFY0(nvlist_lookup_string(spa->spa_config,
ZPOOL_CONFIG_POOL_NAME, &pool_name));
} else
pool_name = spa_name(spa);
VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
spa_version(spa)) == 0);
VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
pool_name) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
spa_state(spa)) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
txg) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
spa_guid(spa)) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_ERRATA,
spa->spa_errata) == 0);
VERIFY(spa->spa_comment == NULL || nvlist_add_string(config,
ZPOOL_CONFIG_COMMENT, spa->spa_comment) == 0);
#ifdef _KERNEL
hostid = 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, &hostid);
#endif /* _KERNEL */
if (hostid != 0) {
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
hostid) == 0);
}
VERIFY0(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
utsname()->nodename));
if (vd != rvd) {
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
vd->vdev_top->vdev_guid) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
vd->vdev_guid) == 0);
if (vd->vdev_isspare)
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE,
1ULL) == 0);
if (vd->vdev_islog)
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG,
1ULL) == 0);
vd = vd->vdev_top; /* label contains top config */
} else {
/*
* Only add the (potentially large) split information
* in the mos config, and not in the vdev labels
*/
if (spa->spa_config_splitting != NULL)
VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT,
spa->spa_config_splitting) == 0);
}
/*
* Add the top-level config. We even add this on pools which
* don't support holes in the namespace.
*/
vdev_top_config_generate(spa, config);
/*
* If we're splitting, record the original pool's guid.
*/
if (spa->spa_config_splitting != NULL &&
nvlist_lookup_uint64(spa->spa_config_splitting,
ZPOOL_CONFIG_SPLIT_GUID, &split_guid) == 0) {
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID,
split_guid) == 0);
}
nvroot = vdev_config_generate(spa, vd, getstats, 0);
VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
nvlist_free(nvroot);
/*
* Store what's necessary for reading the MOS in the label.
*/
VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
spa->spa_label_features) == 0);
if (getstats && spa_load_state(spa) == SPA_LOAD_NONE) {
ddt_histogram_t *ddh;
ddt_stat_t *dds;
ddt_object_t *ddo;
ddh = kmem_zalloc(sizeof (ddt_histogram_t), KM_SLEEP);
ddt_get_dedup_histogram(spa, ddh);
VERIFY(nvlist_add_uint64_array(config,
ZPOOL_CONFIG_DDT_HISTOGRAM,
(uint64_t *)ddh, sizeof (*ddh) / sizeof (uint64_t)) == 0);
kmem_free(ddh, sizeof (ddt_histogram_t));
ddo = kmem_zalloc(sizeof (ddt_object_t), KM_SLEEP);
ddt_get_dedup_object_stats(spa, ddo);
VERIFY(nvlist_add_uint64_array(config,
ZPOOL_CONFIG_DDT_OBJ_STATS,
(uint64_t *)ddo, sizeof (*ddo) / sizeof (uint64_t)) == 0);
kmem_free(ddo, sizeof (ddt_object_t));
dds = kmem_zalloc(sizeof (ddt_stat_t), KM_SLEEP);
ddt_get_dedup_stats(spa, dds);
VERIFY(nvlist_add_uint64_array(config,
ZPOOL_CONFIG_DDT_STATS,
(uint64_t *)dds, sizeof (*dds) / sizeof (uint64_t)) == 0);
kmem_free(dds, sizeof (ddt_stat_t));
}
if (locked)
spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
return (config);
}
static void
txg_sync_thread(dsl_pool_t *dp)
{
spa_t *spa = dp->dp_spa;
tx_state_t *tx = &dp->dp_tx;
callb_cpr_t cpr;
uint64_t start, delta;
txg_thread_enter(tx, &cpr);
start = delta = 0;
for (;;) {
uint64_t timer, timeout = zfs_txg_timeout * hz;
uint64_t txg;
/*
* We sync when we're scrubbing, there's someone waiting
* on us, or the quiesce thread has handed off a txg to
* us, or we have reached our timeout.
*/
timer = (delta >= timeout ? 0 : timeout - delta);
while ((dp->dp_scrub_func == SCRUB_FUNC_NONE ||
spa_load_state(spa) != SPA_LOAD_NONE ||
spa_shutting_down(spa)) &&
!tx->tx_exiting && timer > 0 &&
tx->tx_synced_txg >= tx->tx_sync_txg_waiting &&
tx->tx_quiesced_txg == 0) {
dprintf("waiting; tx_synced=%llu waiting=%llu dp=%p\n",
tx->tx_synced_txg, tx->tx_sync_txg_waiting, dp);
txg_thread_wait(tx, &cpr, &tx->tx_sync_more_cv, timer);
delta = ddi_get_lbolt() - start;
timer = (delta > timeout ? 0 : timeout - delta);
}
/*
* Wait until the quiesce thread hands off a txg to us,
* prompting it to do so if necessary.
*/
while (!tx->tx_exiting && tx->tx_quiesced_txg == 0) {
if (tx->tx_quiesce_txg_waiting < tx->tx_open_txg+1)
tx->tx_quiesce_txg_waiting = tx->tx_open_txg+1;
cv_broadcast(&tx->tx_quiesce_more_cv);
txg_thread_wait(tx, &cpr, &tx->tx_quiesce_done_cv, 0);
}
if (tx->tx_exiting)
txg_thread_exit(tx, &cpr, &tx->tx_sync_thread);
/*
* Consume the quiesced txg which has been handed off to
* us. This may cause the quiescing thread to now be
* able to quiesce another txg, so we must signal it.
*/
txg = tx->tx_quiesced_txg;
tx->tx_quiesced_txg = 0;
tx->tx_syncing_txg = txg;
cv_broadcast(&tx->tx_quiesce_more_cv);
dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
mutex_exit(&tx->tx_sync_lock);
start = ddi_get_lbolt();
spa_sync(spa, txg);
delta = ddi_get_lbolt() - start;
mutex_enter(&tx->tx_sync_lock);
tx->tx_synced_txg = txg;
tx->tx_syncing_txg = 0;
cv_broadcast(&tx->tx_sync_done_cv);
/*
* Dispatch commit callbacks to worker threads.
*/
txg_dispatch_callbacks(dp, txg);
}
}
static mirror_map_t *
vdev_mirror_map_init(zio_t *zio)
{
mirror_map_t *mm = NULL;
mirror_child_t *mc;
vdev_t *vd = zio->io_vd;
int c;
if (vd == NULL) {
dva_t *dva = zio->io_bp->blk_dva;
spa_t *spa = zio->io_spa;
mm = vdev_mirror_map_alloc(BP_GET_NDVAS(zio->io_bp), B_FALSE,
B_TRUE);
for (c = 0; c < mm->mm_children; c++) {
mc = &mm->mm_child[c];
mc->mc_vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[c]));
mc->mc_offset = DVA_GET_OFFSET(&dva[c]);
}
} else {
/*
* If we are resilvering, then we should handle scrub reads
* differently; we shouldn't issue them to the resilvering
* device because it might not have those blocks.
*
* We are resilvering iff:
* 1) We are a replacing vdev (ie our name is "replacing-1" or
* "spare-1" or something like that), and
* 2) The pool is currently being resilvered.
*
* We cannot simply check vd->vdev_resilver_txg, because it's
* not set in this path.
*
* Nor can we just check our vdev_ops; there are cases (such as
* when a user types "zpool replace pool odev spare_dev" and
* spare_dev is in the spare list, or when a spare device is
* automatically used to replace a DEGRADED device) when
* resilvering is complete but both the original vdev and the
* spare vdev remain in the pool. That behavior is intentional.
* It helps implement the policy that a spare should be
* automatically removed from the pool after the user replaces
* the device that originally failed.
*
* If a spa load is in progress, then spa_dsl_pool may be
* uninitialized. But we shouldn't be resilvering during a spa
* load anyway.
*/
boolean_t replacing = (vd->vdev_ops == &vdev_replacing_ops ||
vd->vdev_ops == &vdev_spare_ops) &&
spa_load_state(vd->vdev_spa) == SPA_LOAD_NONE &&
dsl_scan_resilvering(vd->vdev_spa->spa_dsl_pool);
mm = vdev_mirror_map_alloc(vd->vdev_children, replacing,
B_FALSE);
for (c = 0; c < mm->mm_children; c++) {
mc = &mm->mm_child[c];
mc->mc_vd = vd->vdev_child[c];
mc->mc_offset = zio->io_offset;
}
}
zio->io_vsd = mm;
zio->io_vsd_ops = &vdev_mirror_vsd_ops;
return (mm);
}
static mirror_map_t *
vdev_mirror_map_alloc(zio_t *zio)
{
mirror_map_t *mm = NULL;
mirror_child_t *mc;
vdev_t *vd = zio->io_vd;
int c, d;
if (vd == NULL) {
dva_t *dva = zio->io_bp->blk_dva;
spa_t *spa = zio->io_spa;
c = BP_GET_NDVAS(zio->io_bp);
mm = kmem_zalloc(offsetof(mirror_map_t, mm_child[c]), KM_SLEEP);
mm->mm_children = c;
mm->mm_resilvering = B_FALSE;
mm->mm_preferred = spa_get_random(c);
mm->mm_root = B_TRUE;
/*
* Check the other, lower-index DVAs to see if they're on
* the same vdev as the child we picked. If they are, use
* them since they are likely to have been allocated from
* the primary metaslab in use at the time, and hence are
* more likely to have locality with single-copy data.
*/
for (c = mm->mm_preferred, d = c - 1; d >= 0; d--) {
if (DVA_GET_VDEV(&dva[d]) == DVA_GET_VDEV(&dva[c]))
mm->mm_preferred = d;
}
for (c = 0; c < mm->mm_children; c++) {
mc = &mm->mm_child[c];
mc->mc_vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[c]));
mc->mc_offset = DVA_GET_OFFSET(&dva[c]);
}
} else {
int replacing;
c = vd->vdev_children;
mm = kmem_zalloc(offsetof(mirror_map_t, mm_child[c]), KM_SLEEP);
mm->mm_children = c;
/*
* If we are resilvering, then we should handle scrub reads
* differently; we shouldn't issue them to the resilvering
* device because it might not have those blocks.
*
* We are resilvering iff:
* 1) We are a replacing vdev (ie our name is "replacing-1" or
* "spare-1" or something like that), and
* 2) The pool is currently being resilvered.
*
* We cannot simply check vd->vdev_resilver_txg, because it's
* not set in this path.
*
* Nor can we just check our vdev_ops; there are cases (such as
* when a user types "zpool replace pool odev spare_dev" and
* spare_dev is in the spare list, or when a spare device is
* automatically used to replace a DEGRADED device) when
* resilvering is complete but both the original vdev and the
* spare vdev remain in the pool. That behavior is intentional.
* It helps implement the policy that a spare should be
* automatically removed from the pool after the user replaces
* the device that originally failed.
*/
replacing = (vd->vdev_ops == &vdev_replacing_ops ||
vd->vdev_ops == &vdev_spare_ops);
/*
* If a spa load is in progress, then spa_dsl_pool may be
* uninitialized. But we shouldn't be resilvering during a spa
* load anyway.
*/
if (replacing &&
(spa_load_state(vd->vdev_spa) == SPA_LOAD_NONE) &&
dsl_scan_resilvering(vd->vdev_spa->spa_dsl_pool)) {
mm->mm_resilvering = B_TRUE;
} else {
mm->mm_resilvering = B_FALSE;
}
mm->mm_preferred = mm->mm_resilvering ? 0 :
(zio->io_offset >> vdev_mirror_shift) % c;
mm->mm_root = B_FALSE;
for (c = 0; c < mm->mm_children; c++) {
mc = &mm->mm_child[c];
mc->mc_vd = vd->vdev_child[c];
mc->mc_offset = zio->io_offset;
}
}
zio->io_vsd = mm;
zio->io_vsd_ops = &vdev_mirror_vsd_ops;
return (mm);
}