OpenZFS 7743 - per-vdev-zaps init path for upgrade

Authored by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Matt Ahrens <mahrens@delphix.com>
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Don Brady <don.brady@intel.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Joe Stein <jas14@cs.brown.edu>
Ported-by: Don Brady <don.brady@intel.com>

When loading a pool that had been created before the existance of
per-vdev zaps, on a system that knows about per-vdev zaps, the
per-vdev zaps will not be allocated and initialized.

This appears to be because the logic that would have done so, in
spa_sync_config_object(), is not reached under normal operation. It is
only reached if spa_config_dirty_list is non-empty.

The fix is to add another `AVZ_ACTION_` enum that will allow this code
to be reached when we detect that we're loading an old pool, even when
there are no dirty configs.

OpenZFS-issue: https://www.illumos.org/issues/7743
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/e2d29d0
Closes #5582
This commit is contained in:
Don Brady 2017-01-13 14:50:22 -07:00 committed by Brian Behlendorf
parent fdbaf44ffb
commit 38640550f2
2 changed files with 10 additions and 4 deletions

View File

@ -120,7 +120,8 @@ typedef struct spa_taskqs {
typedef enum spa_all_vdev_zap_action { typedef enum spa_all_vdev_zap_action {
AVZ_ACTION_NONE = 0, AVZ_ACTION_NONE = 0,
AVZ_ACTION_DESTROY, /* Destroy all per-vdev ZAPs and the AVZ. */ AVZ_ACTION_DESTROY, /* Destroy all per-vdev ZAPs and the AVZ. */
AVZ_ACTION_REBUILD /* Populate the new AVZ, see spa_avz_rebuild */ AVZ_ACTION_REBUILD, /* Populate the new AVZ, see spa_avz_rebuild */
AVZ_ACTION_INITIALIZE
} spa_avz_action_t; } spa_avz_action_t;
struct spa { struct spa {

View File

@ -2766,10 +2766,14 @@ spa_load_impl(spa_t *spa, uint64_t pool_guid, nvlist_t *config,
error = spa_dir_prop(spa, DMU_POOL_VDEV_ZAP_MAP, error = spa_dir_prop(spa, DMU_POOL_VDEV_ZAP_MAP,
&spa->spa_all_vdev_zaps); &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)); return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
} else if (error == 0 && !nvlist_exists(mos_config, } else if (!nvlist_exists(mos_config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) {
ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) {
/* /*
* An older version of ZFS overwrote the sentinel value, so * An older version of ZFS overwrote the sentinel value, so
* we have orphaned per-vdev ZAPs in the MOS. Defer their * we have orphaned per-vdev ZAPs in the MOS. Defer their
@ -6135,6 +6139,7 @@ spa_sync_config_object(spa_t *spa, dmu_tx_t *tx)
spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
ASSERT(spa->spa_avz_action == AVZ_ACTION_NONE || ASSERT(spa->spa_avz_action == AVZ_ACTION_NONE ||
spa->spa_avz_action == AVZ_ACTION_INITIALIZE ||
spa->spa_all_vdev_zaps != 0); spa->spa_all_vdev_zaps != 0);
if (spa->spa_avz_action == AVZ_ACTION_REBUILD) { if (spa->spa_avz_action == AVZ_ACTION_REBUILD) {