OpenZFS 7614, 9064 - zfs device evacuation/removal

OpenZFS 7614 - zfs device evacuation/removal
OpenZFS 9064 - remove_mirror should wait for device removal to complete

This project allows top-level vdevs to be removed from the storage pool
with "zpool remove", reducing the total amount of storage in the pool.
This operation copies all allocated regions of the device to be removed
onto other devices, recording the mapping from old to new location.
After the removal is complete, read and free operations to the removed
(now "indirect") vdev must be remapped and performed at the new location
on disk.  The indirect mapping table is kept in memory whenever the pool
is loaded, so there is minimal performance overhead when doing operations
on the indirect vdev.

The size of the in-memory mapping table will be reduced when its entries
become "obsolete" because they are no longer used by any block pointers
in the pool.  An entry becomes obsolete when all the blocks that use
it are freed.  An entry can also become obsolete when all the snapshots
that reference it are deleted, and the block pointers that reference it
have been "remapped" in all filesystems/zvols (and clones).  Whenever an
indirect block is written, all the block pointers in it will be "remapped"
to their new (concrete) locations if possible.  This process can be
accelerated by using the "zfs remap" command to proactively rewrite all
indirect blocks that reference indirect (removed) vdevs.

Note that when a device is removed, we do not verify the checksum of
the data that is copied.  This makes the process much faster, but if it
were used on redundant vdevs (i.e. mirror or raidz vdevs), it would be
possible to copy the wrong data, when we have the correct data on e.g.
the other side of the mirror.

At the moment, only mirrors and simple top-level vdevs can be removed
and no removal is allowed if any of the top-level vdevs are raidz.

Porting Notes:

* Avoid zero-sized kmem_alloc() in vdev_compact_children().

    The device evacuation code adds a dependency that
    vdev_compact_children() be able to properly empty the vdev_child
    array by setting it to NULL and zeroing vdev_children.  Under Linux,
    kmem_alloc() and related functions return a sentinel pointer rather
    than NULL for zero-sized allocations.

* Remove comment regarding "mpt" driver where zfs_remove_max_segment
  is initialized to SPA_MAXBLOCKSIZE.

  Change zfs_condense_indirect_commit_entry_delay_ticks to
  zfs_condense_indirect_commit_entry_delay_ms for consistency with
  most other tunables in which delays are specified in ms.

* ZTS changes:

    Use set_tunable rather than mdb
    Use zpool sync as appropriate
    Use sync_pool instead of sync
    Kill jobs during test_removal_with_operation to allow unmount/export
    Don't add non-disk names such as "mirror" or "raidz" to $DISKS
    Use $TEST_BASE_DIR instead of /tmp
    Increase HZ from 100 to 1000 which is more common on Linux

    removal_multiple_indirection.ksh
        Reduce iterations in order to not time out on the code
        coverage builders.

    removal_resume_export:
        Functionally, the test case is correct but there exists a race
        where the kernel thread hasn't been fully started yet and is
        not visible.  Wait for up to 1 second for the removal thread
        to be started before giving up on it.  Also, increase the
        amount of data copied in order that the removal not finish
        before the export has a chance to fail.

* MMP compatibility, the concept of concrete versus non-concrete devices
  has slightly changed the semantics of vdev_writeable().  Update
  mmp_random_leaf_impl() accordingly.

* Updated dbuf_remap() to handle the org.zfsonlinux:large_dnode pool
  feature which is not supported by OpenZFS.

* Added support for new vdev removal tracepoints.

* Test cases removal_with_zdb and removal_condense_export have been
  intentionally disabled.  When run manually they pass as intended,
  but when running in the automated test environment they produce
  unreliable results on the latest Fedora release.

  They may work better once the upstream pool import refectoring is
  merged into ZoL at which point they will be re-enabled.

Authored by: Matthew Ahrens <[email protected]>
Reviewed-by: Alex Reece <[email protected]>
Reviewed-by: George Wilson <[email protected]>
Reviewed-by: John Kennedy <[email protected]>
Reviewed-by: Prakash Surya <[email protected]>
Reviewed by: Richard Laager <[email protected]>
Reviewed by: Tim Chase <[email protected]>
Reviewed by: Brian Behlendorf <[email protected]>
Approved by: Garrett D'Amore <[email protected]>
Ported-by: Tim Chase <[email protected]>
Signed-off-by: Tim Chase <[email protected]>

OpenZFS-issue: https://www.illumos.org/issues/7614
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/f539f1eb
Closes #6900

1 files changed, 268 insertions, 82 deletions

diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
index 8ab996434..f3e3f90fa 100644
--- a/module/zfs/vdev.c
+++ b/module/zfs/vdev.c
@@ -32,8 +32,10 @@
 #include <sys/fm/fs/zfs.h>
 #include <sys/spa.h>
 #include <sys/spa_impl.h>
+#include <sys/bpobj.h>
 #include <sys/dmu.h>
 #include <sys/dmu_tx.h>
+#include <sys/dsl_dir.h>
 #include <sys/vdev_impl.h>
 #include <sys/uberblock_impl.h>
 #include <sys/metaslab.h>
@@ -86,6 +88,7 @@ static vdev_ops_t *vdev_ops_table[] = {
 	&vdev_file_ops,
 	&vdev_missing_ops,
 	&vdev_hole_ops,
+	&vdev_indirect_ops,
 	NULL
 };
 
@@ -310,17 +313,24 @@ vdev_compact_children(vdev_t *pvd)
 
 	ASSERT(spa_config_held(pvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
 
+	if (oldc == 0)
+		return;
+
 	for (int c = newc = 0; c < oldc; c++)
 		if (pvd->vdev_child[c])
 			newc++;
 
-	newchild = kmem_zalloc(newc * sizeof (vdev_t *), KM_SLEEP);
+	if (newc > 0) {
+		newchild = kmem_zalloc(newc * sizeof (vdev_t *), KM_SLEEP);
 
-	for (int c = newc = 0; c < oldc; c++) {
-		if ((cvd = pvd->vdev_child[c]) != NULL) {
-			newchild[newc] = cvd;
-			cvd->vdev_id = newc++;
+		for (int c = newc = 0; c < oldc; c++) {
+			if ((cvd = pvd->vdev_child[c]) != NULL) {
+				newchild[newc] = cvd;
+				cvd->vdev_id = newc++;
+			}
 		}
+	} else {
+		newchild = NULL;
 	}
 
 	kmem_free(pvd->vdev_child, oldc * sizeof (vdev_t *));
@@ -335,8 +345,10 @@ vdev_t *
 vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops)
 {
 	vdev_t *vd;
+	vdev_indirect_config_t *vic;
 
 	vd = kmem_zalloc(sizeof (vdev_t), KM_SLEEP);
+	vic = &vd->vdev_indirect_config;
 
 	if (spa->spa_root_vdev == NULL) {
 		ASSERT(ops == &vdev_root_ops);
@@ -367,6 +379,11 @@ vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops)
 	vd->vdev_ops = ops;
 	vd->vdev_state = VDEV_STATE_CLOSED;
 	vd->vdev_ishole = (ops == &vdev_hole_ops);
+	vic->vic_prev_indirect_vdev = UINT64_MAX;
+
+	rw_init(&vd->vdev_indirect_rwlock, NULL, RW_DEFAULT, NULL);
+	mutex_init(&vd->vdev_obsolete_lock, NULL, MUTEX_DEFAULT, NULL);
+	vd->vdev_obsolete_segments = range_tree_create(NULL, NULL);
 
 	/*
 	 * Initialize rate limit structs for events.  We rate limit ZIO delay
@@ -385,8 +402,7 @@ vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops)
 	mutex_init(&vd->vdev_scan_io_queue_lock, NULL, MUTEX_DEFAULT, NULL);
 
 	for (int t = 0; t < DTL_TYPES; t++) {
-		vd->vdev_dtl[t] = range_tree_create(NULL, NULL,
-		    &vd->vdev_dtl_lock);
+		vd->vdev_dtl[t] = range_tree_create(NULL, NULL);
 	}
 	txg_list_create(&vd->vdev_ms_list, spa,
 	    offsetof(struct metaslab, ms_txg_node));
@@ -412,6 +428,7 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 	char *type;
 	uint64_t guid = 0, islog, nparity;
 	vdev_t *vd;
+	vdev_indirect_config_t *vic;
 	char *tmp = NULL;
 	int rc;
 
@@ -501,6 +518,7 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 	ASSERT(nparity != -1ULL);
 
 	vd = vdev_alloc_common(spa, id, guid, ops);
+	vic = &vd->vdev_indirect_config;
 
 	vd->vdev_islog = islog;
 	vd->vdev_nparity = nparity;
@@ -541,6 +559,16 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 	    &vd->vdev_wholedisk) != 0)
 		vd->vdev_wholedisk = -1ULL;
 
+	ASSERT0(vic->vic_mapping_object);
+	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_INDIRECT_OBJECT,
+	    &vic->vic_mapping_object);
+	ASSERT0(vic->vic_births_object);
+	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_INDIRECT_BIRTHS,
+	    &vic->vic_births_object);
+	ASSERT3U(vic->vic_prev_indirect_vdev, ==, UINT64_MAX);
+	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_PREV_INDIRECT_VDEV,
+	    &vic->vic_prev_indirect_vdev);
+
 	/*
 	 * Look for the 'not present' flag.  This will only be set if the device
 	 * was not present at the time of import.
@@ -754,6 +782,23 @@ vdev_free(vdev_t *vd)
 	}
 	mutex_exit(&vd->vdev_dtl_lock);
 
+	EQUIV(vd->vdev_indirect_births != NULL,
+	    vd->vdev_indirect_mapping != NULL);
+	if (vd->vdev_indirect_births != NULL) {
+		vdev_indirect_mapping_close(vd->vdev_indirect_mapping);
+		vdev_indirect_births_close(vd->vdev_indirect_births);
+	}
+
+	if (vd->vdev_obsolete_sm != NULL) {
+		ASSERT(vd->vdev_removing ||
+		    vd->vdev_ops == &vdev_indirect_ops);
+		space_map_close(vd->vdev_obsolete_sm);
+		vd->vdev_obsolete_sm = NULL;
+	}
+	range_tree_destroy(vd->vdev_obsolete_segments);
+	rw_destroy(&vd->vdev_indirect_rwlock);
+	mutex_destroy(&vd->vdev_obsolete_lock);
+
 	mutex_destroy(&vd->vdev_queue_lock);
 	mutex_destroy(&vd->vdev_dtl_lock);
 	mutex_destroy(&vd->vdev_stat_lock);
@@ -869,6 +914,7 @@ vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops)
 	mvd->vdev_asize = cvd->vdev_asize;
 	mvd->vdev_min_asize = cvd->vdev_min_asize;
 	mvd->vdev_max_asize = cvd->vdev_max_asize;
+	mvd->vdev_psize = cvd->vdev_psize;
 	mvd->vdev_ashift = cvd->vdev_ashift;
 	mvd->vdev_state = cvd->vdev_state;
 	mvd->vdev_crtxg = cvd->vdev_crtxg;
@@ -960,15 +1006,6 @@ vdev_metaslab_init(vdev_t *vd, uint64_t txg)
 
 	ASSERT(!vd->vdev_ishole);
 
-	/*
-	 * Compute the raidz-deflation ratio.  Note, we hard-code
-	 * in 128k (1 << 17) because it is the "typical" blocksize.
-	 * Even though SPA_MAXBLOCKSIZE changed, this algorithm can not change,
-	 * otherwise it would inconsistently account for existing bp's.
-	 */
-	vd->vdev_deflate_ratio = (1 << 17) /
-	    (vdev_psize_to_asize(vd, 1 << 17) >> SPA_MINBLOCKSHIFT);
-
 	ASSERT(oldc <= newc);
 
 	mspp = vmem_zalloc(newc * sizeof (*mspp), KM_SLEEP);
@@ -984,7 +1021,12 @@ vdev_metaslab_init(vdev_t *vd, uint64_t txg)
 	for (m = oldc; m < newc; m++) {
 		uint64_t object = 0;
 
-		if (txg == 0) {
+		/*
+		 * vdev_ms_array may be 0 if we are creating the "fake"
+		 * metaslabs for an indirect vdev for zdb's leak detection.
+		 * See zdb_leak_init().
+		 */
+		if (txg == 0 && vd->vdev_ms_array != 0) {
 			error = dmu_read(mos, vd->vdev_ms_array,
 			    m * sizeof (uint64_t), sizeof (uint64_t), &object,
 			    DMU_READ_PREFETCH);
@@ -1018,12 +1060,11 @@ vdev_metaslab_init(vdev_t *vd, uint64_t txg)
 void
 vdev_metaslab_fini(vdev_t *vd)
 {
-	uint64_t m;
-	uint64_t count = vd->vdev_ms_count;
-
 	if (vd->vdev_ms != NULL) {
+		uint64_t count = vd->vdev_ms_count;
+
 		metaslab_group_passivate(vd->vdev_mg);
-		for (m = 0; m < count; m++) {
+		for (uint64_t m = 0; m < count; m++) {
 			metaslab_t *msp = vd->vdev_ms[m];
 
 			if (msp != NULL)
@@ -1031,8 +1072,10 @@ vdev_metaslab_fini(vdev_t *vd)
 		}
 		vmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
 		vd->vdev_ms = NULL;
-	}
 
+		vd->vdev_ms_count = 0;
+	}
+	ASSERT0(vd->vdev_ms_count);
 	ASSERT3U(vd->vdev_pending_fastwrite, ==, 0);
 }
 
@@ -1078,6 +1121,8 @@ vdev_probe_done(zio_t *zio)
 			zio->io_error = 0;
 		} else {
 			ASSERT(zio->io_error != 0);
+			zfs_dbgmsg("failed probe on vdev %llu",
+			    (longlong_t)vd->vdev_id);
 			zfs_ereport_post(FM_EREPORT_ZFS_PROBE_FAILURE,
 			    spa, vd, NULL, NULL, 0, 0);
 			zio->io_error = SET_ERROR(ENXIO);
@@ -1256,6 +1301,21 @@ retry_sync:
 }
 
 /*
+ * Compute the raidz-deflation ratio.  Note, we hard-code
+ * in 128k (1 << 17) because it is the "typical" blocksize.
+ * Even though SPA_MAXBLOCKSIZE changed, this algorithm can not change,
+ * otherwise it would inconsistently account for existing bp's.
+ */
+static void
+vdev_set_deflate_ratio(vdev_t *vd)
+{
+	if (vd == vd->vdev_top && !vd->vdev_ishole && vd->vdev_ashift != 0) {
+		vd->vdev_deflate_ratio = (1 << 17) /
+		    (vdev_psize_to_asize(vd, 1 << 17) >> SPA_MINBLOCKSHIFT);
+	}
+}
+
+/*
  * Prepare a virtual device for access.
  */
 int
@@ -1458,6 +1518,14 @@ vdev_open(vdev_t *vd)
 		return (error);
 	}
 
+	if (vd->vdev_top == vd && vd->vdev_ashift != 0 &&
+	    !vd->vdev_isl2cache && !vd->vdev_islog) {
+		if (vd->vdev_ashift > spa->spa_max_ashift)
+			spa->spa_max_ashift = vd->vdev_ashift;
+		if (vd->vdev_ashift < spa->spa_min_ashift)
+			spa->spa_min_ashift = vd->vdev_ashift;
+	}
+
 	/*
 	 * Track the min and max ashift values for normal data devices.
 	 */
@@ -1752,7 +1820,8 @@ void
 vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg)
 {
 	ASSERT(vd == vd->vdev_top);
-	ASSERT(!vd->vdev_ishole);
+	/* indirect vdevs don't have metaslabs or dtls */
+	ASSERT(vdev_is_concrete(vd) || flags == 0);
 	ASSERT(ISP2(flags));
 	ASSERT(spa_writeable(vd->vdev_spa));
 
@@ -1822,10 +1891,10 @@ vdev_dtl_dirty(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
 	ASSERT(vd != vd->vdev_spa->spa_root_vdev);
 	ASSERT(spa_writeable(vd->vdev_spa));
 
-	mutex_enter(rt->rt_lock);
+	mutex_enter(&vd->vdev_dtl_lock);
 	if (!range_tree_contains(rt, txg, size))
 		range_tree_add(rt, txg, size);
-	mutex_exit(rt->rt_lock);
+	mutex_exit(&vd->vdev_dtl_lock);
 }
 
 boolean_t
@@ -1837,10 +1906,21 @@ vdev_dtl_contains(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
 	ASSERT(t < DTL_TYPES);
 	ASSERT(vd != vd->vdev_spa->spa_root_vdev);
 
-	mutex_enter(rt->rt_lock);
+	/*
+	 * While we are loading the pool, the DTLs have not been loaded yet.
+	 * Ignore the DTLs and try all devices.  This avoids a recursive
+	 * mutex enter on the vdev_dtl_lock, and also makes us try hard
+	 * when loading the pool (relying on the checksum to ensure that
+	 * we get the right data -- note that we while loading, we are
+	 * only reading the MOS, which is always checksummed).
+	 */
+	if (vd->vdev_spa->spa_load_state != SPA_LOAD_NONE)
+		return (B_FALSE);
+
+	mutex_enter(&vd->vdev_dtl_lock);
 	if (range_tree_space(rt) != 0)
 		dirty = range_tree_contains(rt, txg, size);
-	mutex_exit(rt->rt_lock);
+	mutex_exit(&vd->vdev_dtl_lock);
 
 	return (dirty);
 }
@@ -1851,9 +1931,9 @@ vdev_dtl_empty(vdev_t *vd, vdev_dtl_type_t t)
 	range_tree_t *rt = vd->vdev_dtl[t];
 	boolean_t empty;
 
-	mutex_enter(rt->rt_lock);
+	mutex_enter(&vd->vdev_dtl_lock);
 	empty = (range_tree_space(rt) == 0);
-	mutex_exit(rt->rt_lock);
+	mutex_exit(&vd->vdev_dtl_lock);
 
 	return (empty);
 }
@@ -1961,7 +2041,7 @@ vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg, int scrub_done)
 		vdev_dtl_reassess(vd->vdev_child[c], txg,
 		    scrub_txg, scrub_done);
 
-	if (vd == spa->spa_root_vdev || vd->vdev_ishole || vd->vdev_aux)
+	if (vd == spa->spa_root_vdev || !vdev_is_concrete(vd) || vd->vdev_aux)
 		return;
 
 	if (vd->vdev_ops->vdev_op_leaf) {
@@ -2076,10 +2156,10 @@ vdev_dtl_load(vdev_t *vd)
 	int error = 0;
 
 	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_dtl_object != 0) {
-		ASSERT(!vd->vdev_ishole);
+		ASSERT(vdev_is_concrete(vd));
 
 		error = space_map_open(&vd->vdev_dtl_sm, mos,
-		    vd->vdev_dtl_object, 0, -1ULL, 0, &vd->vdev_dtl_lock);
+		    vd->vdev_dtl_object, 0, -1ULL, 0);
 		if (error)
 			return (error);
 		ASSERT(vd->vdev_dtl_sm != NULL);
@@ -2158,11 +2238,10 @@ vdev_dtl_sync(vdev_t *vd, uint64_t txg)
 	range_tree_t *rt = vd->vdev_dtl[DTL_MISSING];
 	objset_t *mos = spa->spa_meta_objset;
 	range_tree_t *rtsync;
-	kmutex_t rtlock;
 	dmu_tx_t *tx;
 	uint64_t object = space_map_object(vd->vdev_dtl_sm);
 
-	ASSERT(!vd->vdev_ishole);
+	ASSERT(vdev_is_concrete(vd));
 	ASSERT(vd->vdev_ops->vdev_op_leaf);
 
 	tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
@@ -2196,15 +2275,11 @@ vdev_dtl_sync(vdev_t *vd, uint64_t txg)
 		VERIFY3U(new_object, !=, 0);
 
 		VERIFY0(space_map_open(&vd->vdev_dtl_sm, mos, new_object,
-		    0, -1ULL, 0, &vd->vdev_dtl_lock));
+		    0, -1ULL, 0));
 		ASSERT(vd->vdev_dtl_sm != NULL);
 	}
 
-	mutex_init(&rtlock, NULL, MUTEX_DEFAULT, NULL);
-
-	rtsync = range_tree_create(NULL, NULL, &rtlock);
-
-	mutex_enter(&rtlock);
+	rtsync = range_tree_create(NULL, NULL);
 
 	mutex_enter(&vd->vdev_dtl_lock);
 	range_tree_walk(rt, range_tree_add, rtsync);
@@ -2216,9 +2291,6 @@ vdev_dtl_sync(vdev_t *vd, uint64_t txg)
 
 	range_tree_destroy(rtsync);
 
-	mutex_exit(&rtlock);
-	mutex_destroy(&rtlock);
-
 	/*
 	 * If the object for the space map has changed then dirty
 	 * the top level so that we update the config.
@@ -2311,29 +2383,63 @@ vdev_resilver_needed(vdev_t *vd, uint64_t *minp, uint64_t *maxp)
 	return (needed);
 }
 
-void
+int
 vdev_load(vdev_t *vd)
 {
+	int error = 0;
+
 	/*
 	 * Recursively load all children.
 	 */
-	for (int c = 0; c < vd->vdev_children; c++)
-		vdev_load(vd->vdev_child[c]);
+	for (int c = 0; c < vd->vdev_children; c++) {
+		error = vdev_load(vd->vdev_child[c]);
+		if (error != 0) {
+			return (error);
+		}
+	}
+
+	vdev_set_deflate_ratio(vd);
 
 	/*
 	 * If this is a top-level vdev, initialize its metaslabs.
 	 */
-	if (vd == vd->vdev_top && !vd->vdev_ishole &&
-	    (vd->vdev_ashift == 0 || vd->vdev_asize == 0 ||
-	    vdev_metaslab_init(vd, 0) != 0))
-		vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
-		    VDEV_AUX_CORRUPT_DATA);
+	if (vd == vd->vdev_top && vdev_is_concrete(vd)) {
+		if (vd->vdev_ashift == 0 || vd->vdev_asize == 0) {
+			vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+			    VDEV_AUX_CORRUPT_DATA);
+			return (SET_ERROR(ENXIO));
+		} else if ((error = vdev_metaslab_init(vd, 0)) != 0) {
+			vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+			    VDEV_AUX_CORRUPT_DATA);
+			return (error);
+		}
+	}
+
 	/*
 	 * If this is a leaf vdev, load its DTL.
 	 */
-	if (vd->vdev_ops->vdev_op_leaf && vdev_dtl_load(vd) != 0)
+	if (vd->vdev_ops->vdev_op_leaf && (error = vdev_dtl_load(vd)) != 0) {
 		vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
 		    VDEV_AUX_CORRUPT_DATA);
+		return (error);
+	}
+
+	uint64_t obsolete_sm_object = vdev_obsolete_sm_object(vd);
+	if (obsolete_sm_object != 0) {
+		objset_t *mos = vd->vdev_spa->spa_meta_objset;
+		ASSERT(vd->vdev_asize != 0);
+		ASSERT(vd->vdev_obsolete_sm == NULL);
+
+		if ((error = space_map_open(&vd->vdev_obsolete_sm, mos,
+		    obsolete_sm_object, 0, vd->vdev_asize, 0))) {
+			vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
+			    VDEV_AUX_CORRUPT_DATA);
+			return (error);
+		}
+		space_map_update(vd->vdev_obsolete_sm);
+	}
+
+	return (0);
 }
 
 /*
@@ -2378,14 +2484,42 @@ vdev_validate_aux(vdev_t *vd)
 	return (0);
 }
 
+/*
+ * Free the objects used to store this vdev's spacemaps, and the array
+ * that points to them.
+ */
 void
-vdev_remove(vdev_t *vd, uint64_t txg)
+vdev_destroy_spacemaps(vdev_t *vd, dmu_tx_t *tx)
+{
+	if (vd->vdev_ms_array == 0)
+		return;
+
+	objset_t *mos = vd->vdev_spa->spa_meta_objset;
+	uint64_t array_count = vd->vdev_asize >> vd->vdev_ms_shift;
+	size_t array_bytes = array_count * sizeof (uint64_t);
+	uint64_t *smobj_array = kmem_alloc(array_bytes, KM_SLEEP);
+	VERIFY0(dmu_read(mos, vd->vdev_ms_array, 0,
+	    array_bytes, smobj_array, 0));
+
+	for (uint64_t i = 0; i < array_count; i++) {
+		uint64_t smobj = smobj_array[i];
+		if (smobj == 0)
+			continue;
+
+		space_map_free_obj(mos, smobj, tx);
+	}
+
+	kmem_free(smobj_array, array_bytes);
+	VERIFY0(dmu_object_free(mos, vd->vdev_ms_array, tx));
+	vd->vdev_ms_array = 0;
+}
+
+static void
+vdev_remove_empty(vdev_t *vd, uint64_t txg)
 {
 	spa_t *spa = vd->vdev_spa;
-	objset_t *mos = spa->spa_meta_objset;
 	dmu_tx_t *tx;
 
-	tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);
 	ASSERT(vd == vd->vdev_top);
 	ASSERT3U(txg, ==, spa_syncing_txg(spa));
 
@@ -2412,7 +2546,6 @@ vdev_remove(vdev_t *vd, uint64_t txg)
 			metaslab_group_histogram_remove(mg, msp);
 
 			VERIFY0(space_map_allocated(msp->ms_sm));
-			space_map_free(msp->ms_sm, tx);
 			space_map_close(msp->ms_sm);
 			msp->ms_sm = NULL;
 			mutex_exit(&msp->ms_lock);
@@ -2422,13 +2555,10 @@ vdev_remove(vdev_t *vd, uint64_t txg)
 		metaslab_class_histogram_verify(mg->mg_class);
 		for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++)
 			ASSERT0(mg->mg_histogram[i]);
-
 	}
 
-	if (vd->vdev_ms_array) {
-		(void) dmu_object_free(mos, vd->vdev_ms_array, tx);
-		vd->vdev_ms_array = 0;
-	}
+	tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);
+	vdev_destroy_spacemaps(vd, tx);
 
 	if (vd->vdev_islog && vd->vdev_top_zap != 0) {
 		vdev_destroy_unlink_zap(vd, vd->vdev_top_zap, tx);
@@ -2443,7 +2573,7 @@ vdev_sync_done(vdev_t *vd, uint64_t txg)
 	metaslab_t *msp;
 	boolean_t reassess = !txg_list_empty(&vd->vdev_ms_list, TXG_CLEAN(txg));
 
-	ASSERT(!vd->vdev_ishole);
+	ASSERT(vdev_is_concrete(vd));
 
 	while ((msp = txg_list_remove(&vd->vdev_ms_list, TXG_CLEAN(txg))))
 		metaslab_sync_done(msp, txg);
@@ -2460,10 +2590,33 @@ vdev_sync(vdev_t *vd, uint64_t txg)
 	metaslab_t *msp;
 	dmu_tx_t *tx;
 
-	ASSERT(!vd->vdev_ishole);
+	if (range_tree_space(vd->vdev_obsolete_segments) > 0) {
+		dmu_tx_t *tx;
 
-	if (vd->vdev_ms_array == 0 && vd->vdev_ms_shift != 0) {
+		ASSERT(vd->vdev_removing ||
+		    vd->vdev_ops == &vdev_indirect_ops);
+
+		tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
+		vdev_indirect_sync_obsolete(vd, tx);
+		dmu_tx_commit(tx);
+
+		/*
+		 * If the vdev is indirect, it can't have dirty
+		 * metaslabs or DTLs.
+		 */
+		if (vd->vdev_ops == &vdev_indirect_ops) {
+			ASSERT(txg_list_empty(&vd->vdev_ms_list, txg));
+			ASSERT(txg_list_empty(&vd->vdev_dtl_list, txg));
+			return;
+		}
+	}
+
+	ASSERT(vdev_is_concrete(vd));
+
+	if (vd->vdev_ms_array == 0 && vd->vdev_ms_shift != 0 &&
+	    !vd->vdev_removing) {
 		ASSERT(vd == vd->vdev_top);
+		ASSERT0(vd->vdev_indirect_config.vic_mapping_object);
 		tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
 		vd->vdev_ms_array = dmu_object_alloc(spa->spa_meta_objset,
 		    DMU_OT_OBJECT_ARRAY, 0, DMU_OT_NONE, 0, tx);
@@ -2472,12 +2625,6 @@ vdev_sync(vdev_t *vd, uint64_t txg)
 		dmu_tx_commit(tx);
 	}
 
-	/*
-	 * Remove the metadata associated with this vdev once it's empty.
-	 */
-	if (vd->vdev_stat.vs_alloc == 0 && vd->vdev_removing)
-		vdev_remove(vd, txg);
-
 	while ((msp = txg_list_remove(&vd->vdev_ms_list, txg)) != NULL) {
 		metaslab_sync(msp, txg);
 		(void) txg_list_add(&vd->vdev_ms_list, msp, TXG_CLEAN(txg));
@@ -2486,6 +2633,16 @@ vdev_sync(vdev_t *vd, uint64_t txg)
 	while ((lvd = txg_list_remove(&vd->vdev_dtl_list, txg)) != NULL)
 		vdev_dtl_sync(lvd, txg);
 
+	/*
+	 * Remove the metadata associated with this vdev once it's empty.
+	 * Note that this is typically used for log/cache device removal;
+	 * we don't empty toplevel vdevs when removing them.  But if
+	 * a toplevel happens to be emptied, this is not harmful.
+	 */
+	if (vd->vdev_stat.vs_alloc == 0 && vd->vdev_removing) {
+		vdev_remove_empty(vd, txg);
+	}
+
 	(void) txg_list_add(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg));
 }
 
@@ -2725,7 +2882,7 @@ top:
 			metaslab_group_passivate(mg);
 			(void) spa_vdev_state_exit(spa, vd, 0);
 
-			error = spa_offline_log(spa);
+			error = spa_reset_logs(spa);
 
 			spa_vdev_state_enter(spa, SCL_ALLOC);
 
@@ -2807,6 +2964,12 @@ vdev_clear(spa_t *spa, vdev_t *vd)
 		vdev_clear(spa, vd->vdev_child[c]);
 
 	/*
+	 * It makes no sense to "clear" an indirect vdev.
+	 */
+	if (!vdev_is_concrete(vd))
+		return;
+
+	/*
 	 * If we're in the FAULTED state or have experienced failed I/O, then
 	 * clear the persistent state and attempt to reopen the device.  We
 	 * also mark the vdev config dirty, so that the new faulted state is
@@ -2860,7 +3023,8 @@ vdev_is_dead(vdev_t *vd)
 	 * Instead we rely on the fact that we skip over dead devices
 	 * before issuing I/O to them.
 	 */
-	return (vd->vdev_state < VDEV_STATE_DEGRADED || vd->vdev_ishole ||
+	return (vd->vdev_state < VDEV_STATE_DEGRADED ||
+	    vd->vdev_ops == &vdev_hole_ops ||
 	    vd->vdev_ops == &vdev_missing_ops);
 }
 
@@ -2873,7 +3037,8 @@ vdev_readable(vdev_t *vd)
 boolean_t
 vdev_writeable(vdev_t *vd)
 {
-	return (!vdev_is_dead(vd) && !vd->vdev_cant_write);
+	return (!vdev_is_dead(vd) && !vd->vdev_cant_write &&
+	    vdev_is_concrete(vd));
 }
 
 boolean_t
@@ -2890,7 +3055,7 @@ vdev_allocatable(vdev_t *vd)
 	 * we're asking two separate questions about it.
 	 */
 	return (!(state < VDEV_STATE_DEGRADED && state != VDEV_STATE_CLOSED) &&
-	    !vd->vdev_cant_write && !vd->vdev_ishole &&
+	    !vd->vdev_cant_write && vdev_is_concrete(vd) &&
 	    vd->vdev_mg->mg_initialized);
 }
 
@@ -3033,7 +3198,7 @@ vdev_get_stats_ex(vdev_t *vd, vdev_stat_t *vs, vdev_stat_ex_t *vsx)
 		}
 		vs->vs_esize = vd->vdev_max_asize - vd->vdev_asize;
 		if (vd->vdev_aux == NULL && vd == vd->vdev_top &&
-		    !vd->vdev_ishole) {
+		    vdev_is_concrete(vd)) {
 			vs->vs_fragmentation = vd->vdev_mg->mg_fragmentation;
 		}
 	}
@@ -3196,7 +3361,8 @@ vdev_stat_update(zio_t *zio, uint64_t psize)
 		vs->vs_write_errors++;
 	mutex_exit(&vd->vdev_stat_lock);
 
-	if (type == ZIO_TYPE_WRITE && txg != 0 &&
+	if (spa->spa_load_state == SPA_LOAD_NONE &&
+	    type == ZIO_TYPE_WRITE && txg != 0 &&
 	    (!(flags & ZIO_FLAG_IO_REPAIR) ||
 	    (flags & ZIO_FLAG_SCAN_THREAD) ||
 	    spa->spa_claiming)) {
@@ -3361,8 +3527,9 @@ vdev_config_dirty(vdev_t *vd)
 		ASSERT(vd == vd->vdev_top);
 
 		if (!list_link_active(&vd->vdev_config_dirty_node) &&
-		    !vd->vdev_ishole)
+		    vdev_is_concrete(vd)) {
 			list_insert_head(&spa->spa_config_dirty_list, vd);
+		}
 	}
 }
 
@@ -3403,7 +3570,8 @@ vdev_state_dirty(vdev_t *vd)
 	    (dsl_pool_sync_context(spa_get_dsl(spa)) &&
 	    spa_config_held(spa, SCL_STATE, RW_READER)));
 
-	if (!list_link_active(&vd->vdev_state_dirty_node) && !vd->vdev_ishole)
+	if (!list_link_active(&vd->vdev_state_dirty_node) &&
+	    vdev_is_concrete(vd))
 		list_insert_head(&spa->spa_state_dirty_list, vd);
 }
 
@@ -3437,9 +3605,10 @@ vdev_propagate_state(vdev_t *vd)
 			child = vd->vdev_child[c];
 
 			/*
-			 * Don't factor holes into the decision.
+			 * Don't factor holes or indirect vdevs into the
+			 * decision.
 			 */
-			if (child->vdev_ishole)
+			if (!vdev_is_concrete(child))
 				continue;
 
 			if (!vdev_readable(child) ||
@@ -3642,8 +3811,10 @@ vdev_is_bootable(vdev_t *vd)
 	if (!vd->vdev_ops->vdev_op_leaf) {
 		const char *vdev_type = vd->vdev_ops->vdev_op_type;
 
-		if (strcmp(vdev_type, VDEV_TYPE_MISSING) == 0)
+		if (strcmp(vdev_type, VDEV_TYPE_MISSING) == 0 ||
+		    strcmp(vdev_type, VDEV_TYPE_INDIRECT) == 0) {
 			return (B_FALSE);
+		}
 	}
 
 	for (int c = 0; c < vd->vdev_children; c++) {
@@ -3653,6 +3824,18 @@ vdev_is_bootable(vdev_t *vd)
 	return (B_TRUE);
 }
 
+boolean_t
+vdev_is_concrete(vdev_t *vd)
+{
+	vdev_ops_t *ops = vd->vdev_ops;
+	if (ops == &vdev_indirect_ops || ops == &vdev_hole_ops ||
+	    ops == &vdev_missing_ops || ops == &vdev_root_ops) {
+		return (B_FALSE);
+	} else {
+		return (B_TRUE);
+	}
+}
+
 /*
  * Load the state from the original vdev tree (ovd) which
  * we've retrieved from the MOS config object. If the original
@@ -3709,7 +3892,10 @@ vdev_expand(vdev_t *vd, uint64_t txg)
 	ASSERT(vd->vdev_top == vd);
 	ASSERT(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
 
-	if ((vd->vdev_asize >> vd->vdev_ms_shift) > vd->vdev_ms_count) {
+	vdev_set_deflate_ratio(vd);
+
+	if ((vd->vdev_asize >> vd->vdev_ms_shift) > vd->vdev_ms_count &&
+	    vdev_is_concrete(vd)) {
 		VERIFY(vdev_metaslab_init(vd, txg) == 0);
 		vdev_config_dirty(vd);
 	}