Fix vdev removal finishing race

This patch fixes a race condition where the end of
vdev_remove_replace_with_indirect(), which holds
svr_lock, would race against spa_vdev_removal_destroy(),
which destroys the same lock and is called asynchronously
via dsl_sync_task_nowait().

Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Issue #6900 
Closes #8083
This commit is contained in:
Tom Caputi 2018-11-07 18:38:10 -05:00 committed by Brian Behlendorf
parent 4021ba4cfa
commit ac53e50f79

View File

@ -1115,19 +1115,16 @@ vdev_remove_replace_with_indirect(vdev_t *vd, uint64_t txg)
ASSERT(!list_link_active(&vd->vdev_state_dirty_node));
tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
dsl_sync_task_nowait(spa->spa_dsl_pool, vdev_remove_complete_sync, svr,
0, ZFS_SPACE_CHECK_NONE, tx);
dmu_tx_commit(tx);
/*
* Indicate that this thread has exited.
* After this, we can not use svr.
*/
mutex_enter(&svr->svr_lock);
svr->svr_thread = NULL;
cv_broadcast(&svr->svr_cv);
mutex_exit(&svr->svr_lock);
/* After this, we can not use svr. */
tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
dsl_sync_task_nowait(spa->spa_dsl_pool, vdev_remove_complete_sync, svr,
0, ZFS_SPACE_CHECK_NONE, tx);
dmu_tx_commit(tx);
}
/*