4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2012, 2018 by Delphix. All rights reserved.
28 * Virtual Device Labels
29 * ---------------------
31 * The vdev label serves several distinct purposes:
33 * 1. Uniquely identify this device as part of a ZFS pool and confirm its
34 * identity within the pool.
36 * 2. Verify that all the devices given in a configuration are present
39 * 3. Determine the uberblock for the pool.
41 * 4. In case of an import operation, determine the configuration of the
42 * toplevel vdev of which it is a part.
44 * 5. If an import operation cannot find all the devices in the pool,
45 * provide enough information to the administrator to determine which
46 * devices are missing.
48 * It is important to note that while the kernel is responsible for writing the
49 * label, it only consumes the information in the first three cases. The
50 * latter information is only consumed in userland when determining the
51 * configuration to import a pool.
57 * Before describing the contents of the label, it's important to understand how
58 * the labels are written and updated with respect to the uberblock.
60 * When the pool configuration is altered, either because it was newly created
61 * or a device was added, we want to update all the labels such that we can deal
62 * with fatal failure at any point. To this end, each disk has two labels which
63 * are updated before and after the uberblock is synced. Assuming we have
64 * labels and an uberblock with the following transaction groups:
67 * +------+ +------+ +------+
69 * | t10 | | t10 | | t10 |
71 * +------+ +------+ +------+
73 * In this stable state, the labels and the uberblock were all updated within
74 * the same transaction group (10). Each label is mirrored and checksummed, so
75 * that we can detect when we fail partway through writing the label.
77 * In order to identify which labels are valid, the labels are written in the
80 * 1. For each vdev, update 'L1' to the new label
81 * 2. Update the uberblock
82 * 3. For each vdev, update 'L2' to the new label
84 * Given arbitrary failure, we can determine the correct label to use based on
85 * the transaction group. If we fail after updating L1 but before updating the
86 * UB, we will notice that L1's transaction group is greater than the uberblock,
87 * so L2 must be valid. If we fail after writing the uberblock but before
88 * writing L2, we will notice that L2's transaction group is less than L1, and
89 * therefore L1 is valid.
91 * Another added complexity is that not every label is updated when the config
92 * is synced. If we add a single device, we do not want to have to re-write
93 * every label for every device in the pool. This means that both L1 and L2 may
94 * be older than the pool uberblock, because the necessary information is stored
101 * The vdev label consists of two distinct parts, and is wrapped within the
102 * vdev_label_t structure. The label includes 8k of padding to permit legacy
103 * VTOC disk labels, but is otherwise ignored.
105 * The first half of the label is a packed nvlist which contains pool wide
106 * properties, per-vdev properties, and configuration information. It is
107 * described in more detail below.
109 * The latter half of the label consists of a redundant array of uberblocks.
110 * These uberblocks are updated whenever a transaction group is committed,
111 * or when the configuration is updated. When a pool is loaded, we scan each
112 * vdev for the 'best' uberblock.
115 * Configuration Information
116 * -------------------------
118 * The nvlist describing the pool and vdev contains the following elements:
120 * version ZFS on-disk version
123 * txg Transaction group in which this label was written
124 * pool_guid Unique identifier for this pool
125 * vdev_tree An nvlist describing vdev tree.
127 * An nvlist of the features necessary for reading the MOS.
129 * Each leaf device label also contains the following:
131 * top_guid Unique ID for top-level vdev in which this is contained
132 * guid Unique ID for the leaf vdev
134 * The 'vs' configuration follows the format described in 'spa_config.c'.
137 #include <sys/zfs_context.h>
139 #include <sys/spa_impl.h>
142 #include <sys/vdev.h>
143 #include <sys/vdev_impl.h>
144 #include <sys/uberblock_impl.h>
145 #include <sys/metaslab.h>
146 #include <sys/metaslab_impl.h>
148 #include <sys/dsl_scan.h>
150 #include <sys/fs/zfs.h>
153 * Basic routines to read and write from a vdev label.
154 * Used throughout the rest of this file.
157 vdev_label_offset(uint64_t psize
, int l
, uint64_t offset
)
159 ASSERT(offset
< sizeof (vdev_label_t
));
160 ASSERT(P2PHASE_TYPED(psize
, sizeof (vdev_label_t
), uint64_t) == 0);
162 return (offset
+ l
* sizeof (vdev_label_t
) + (l
< VDEV_LABELS
/ 2 ?
163 0 : psize
- VDEV_LABELS
* sizeof (vdev_label_t
)));
167 * Returns back the vdev label associated with the passed in offset.
170 vdev_label_number(uint64_t psize
, uint64_t offset
)
174 if (offset
>= psize
- VDEV_LABEL_END_SIZE
) {
175 offset
-= psize
- VDEV_LABEL_END_SIZE
;
176 offset
+= (VDEV_LABELS
/ 2) * sizeof (vdev_label_t
);
178 l
= offset
/ sizeof (vdev_label_t
);
179 return (l
< VDEV_LABELS
? l
: -1);
183 vdev_label_read(zio_t
*zio
, vdev_t
*vd
, int l
, abd_t
*buf
, uint64_t offset
,
184 uint64_t size
, zio_done_func_t
*done
, void *private, int flags
)
186 ASSERT(spa_config_held(zio
->io_spa
, SCL_STATE_ALL
, RW_WRITER
) ==
188 ASSERT(flags
& ZIO_FLAG_CONFIG_WRITER
);
190 zio_nowait(zio_read_phys(zio
, vd
,
191 vdev_label_offset(vd
->vdev_psize
, l
, offset
),
192 size
, buf
, ZIO_CHECKSUM_LABEL
, done
, private,
193 ZIO_PRIORITY_SYNC_READ
, flags
, B_TRUE
));
197 vdev_label_write(zio_t
*zio
, vdev_t
*vd
, int l
, abd_t
*buf
, uint64_t offset
,
198 uint64_t size
, zio_done_func_t
*done
, void *private, int flags
)
200 ASSERT(spa_config_held(zio
->io_spa
, SCL_ALL
, RW_WRITER
) == SCL_ALL
||
201 (spa_config_held(zio
->io_spa
, SCL_CONFIG
| SCL_STATE
, RW_READER
) ==
202 (SCL_CONFIG
| SCL_STATE
) &&
203 dsl_pool_sync_context(spa_get_dsl(zio
->io_spa
))));
204 ASSERT(flags
& ZIO_FLAG_CONFIG_WRITER
);
206 zio_nowait(zio_write_phys(zio
, vd
,
207 vdev_label_offset(vd
->vdev_psize
, l
, offset
),
208 size
, buf
, ZIO_CHECKSUM_LABEL
, done
, private,
209 ZIO_PRIORITY_SYNC_WRITE
, flags
, B_TRUE
));
213 root_vdev_actions_getprogress(vdev_t
*vd
, nvlist_t
*nvl
)
215 spa_t
*spa
= vd
->vdev_spa
;
217 if (vd
!= spa
->spa_root_vdev
)
220 /* provide either current or previous scan information */
222 if (spa_scan_get_stats(spa
, &ps
) == 0) {
223 fnvlist_add_uint64_array(nvl
,
224 ZPOOL_CONFIG_SCAN_STATS
, (uint64_t *)&ps
,
225 sizeof (pool_scan_stat_t
) / sizeof (uint64_t));
228 pool_removal_stat_t prs
;
229 if (spa_removal_get_stats(spa
, &prs
) == 0) {
230 fnvlist_add_uint64_array(nvl
,
231 ZPOOL_CONFIG_REMOVAL_STATS
, (uint64_t *)&prs
,
232 sizeof (prs
) / sizeof (uint64_t));
235 pool_checkpoint_stat_t pcs
;
236 if (spa_checkpoint_get_stats(spa
, &pcs
) == 0) {
237 fnvlist_add_uint64_array(nvl
,
238 ZPOOL_CONFIG_CHECKPOINT_STATS
, (uint64_t *)&pcs
,
239 sizeof (pcs
) / sizeof (uint64_t));
244 * Generate the nvlist representing this vdev's config.
247 vdev_config_generate(spa_t
*spa
, vdev_t
*vd
, boolean_t getstats
,
248 vdev_config_flag_t flags
)
251 vdev_indirect_config_t
*vic
= &vd
->vdev_indirect_config
;
253 nv
= fnvlist_alloc();
255 fnvlist_add_string(nv
, ZPOOL_CONFIG_TYPE
, vd
->vdev_ops
->vdev_op_type
);
256 if (!(flags
& (VDEV_CONFIG_SPARE
| VDEV_CONFIG_L2CACHE
)))
257 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ID
, vd
->vdev_id
);
258 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_GUID
, vd
->vdev_guid
);
260 if (vd
->vdev_path
!= NULL
)
261 fnvlist_add_string(nv
, ZPOOL_CONFIG_PATH
, vd
->vdev_path
);
263 if (vd
->vdev_devid
!= NULL
)
264 fnvlist_add_string(nv
, ZPOOL_CONFIG_DEVID
, vd
->vdev_devid
);
266 if (vd
->vdev_physpath
!= NULL
)
267 fnvlist_add_string(nv
, ZPOOL_CONFIG_PHYS_PATH
,
270 if (vd
->vdev_fru
!= NULL
)
271 fnvlist_add_string(nv
, ZPOOL_CONFIG_FRU
, vd
->vdev_fru
);
273 if (vd
->vdev_nparity
!= 0) {
274 ASSERT(strcmp(vd
->vdev_ops
->vdev_op_type
,
275 VDEV_TYPE_RAIDZ
) == 0);
278 * Make sure someone hasn't managed to sneak a fancy new vdev
279 * into a crufty old storage pool.
281 ASSERT(vd
->vdev_nparity
== 1 ||
282 (vd
->vdev_nparity
<= 2 &&
283 spa_version(spa
) >= SPA_VERSION_RAIDZ2
) ||
284 (vd
->vdev_nparity
<= 3 &&
285 spa_version(spa
) >= SPA_VERSION_RAIDZ3
));
288 * Note that we'll add the nparity tag even on storage pools
289 * that only support a single parity device -- older software
290 * will just ignore it.
292 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_NPARITY
, vd
->vdev_nparity
);
295 if (vd
->vdev_wholedisk
!= -1ULL)
296 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
299 if (vd
->vdev_not_present
&& !(flags
& VDEV_CONFIG_MISSING
))
300 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
, 1);
302 if (vd
->vdev_isspare
)
303 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_SPARE
, 1);
305 if (!(flags
& (VDEV_CONFIG_SPARE
| VDEV_CONFIG_L2CACHE
)) &&
306 vd
== vd
->vdev_top
) {
307 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_METASLAB_ARRAY
,
309 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_METASLAB_SHIFT
,
311 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ASHIFT
, vd
->vdev_ashift
);
312 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ASIZE
,
314 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_LOG
, vd
->vdev_islog
);
315 if (vd
->vdev_removing
) {
316 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_REMOVING
,
321 if (vd
->vdev_dtl_sm
!= NULL
) {
322 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_DTL
,
323 space_map_object(vd
->vdev_dtl_sm
));
326 if (vic
->vic_mapping_object
!= 0) {
327 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_OBJECT
,
328 vic
->vic_mapping_object
);
331 if (vic
->vic_births_object
!= 0) {
332 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_BIRTHS
,
333 vic
->vic_births_object
);
336 if (vic
->vic_prev_indirect_vdev
!= UINT64_MAX
) {
337 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_PREV_INDIRECT_VDEV
,
338 vic
->vic_prev_indirect_vdev
);
342 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_CREATE_TXG
, vd
->vdev_crtxg
);
344 if (flags
& VDEV_CONFIG_MOS
) {
345 if (vd
->vdev_leaf_zap
!= 0) {
346 ASSERT(vd
->vdev_ops
->vdev_op_leaf
);
347 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_VDEV_LEAF_ZAP
,
351 if (vd
->vdev_top_zap
!= 0) {
352 ASSERT(vd
== vd
->vdev_top
);
353 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_VDEV_TOP_ZAP
,
361 vdev_get_stats(vd
, &vs
);
362 fnvlist_add_uint64_array(nv
, ZPOOL_CONFIG_VDEV_STATS
,
363 (uint64_t *)&vs
, sizeof (vs
) / sizeof (uint64_t));
365 root_vdev_actions_getprogress(vd
, nv
);
368 * Note: this can be called from open context
369 * (spa_get_stats()), so we need the rwlock to prevent
370 * the mapping from being changed by condensing.
372 rw_enter(&vd
->vdev_indirect_rwlock
, RW_READER
);
373 if (vd
->vdev_indirect_mapping
!= NULL
) {
374 ASSERT(vd
->vdev_indirect_births
!= NULL
);
375 vdev_indirect_mapping_t
*vim
=
376 vd
->vdev_indirect_mapping
;
377 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_SIZE
,
378 vdev_indirect_mapping_size(vim
));
380 rw_exit(&vd
->vdev_indirect_rwlock
);
381 if (vd
->vdev_mg
!= NULL
&&
382 vd
->vdev_mg
->mg_fragmentation
!= ZFS_FRAG_INVALID
) {
384 * Compute approximately how much memory would be used
385 * for the indirect mapping if this device were to
388 * Note: If the frag metric is invalid, then not
389 * enough metaslabs have been converted to have
392 uint64_t seg_count
= 0;
393 uint64_t to_alloc
= vd
->vdev_stat
.vs_alloc
;
396 * There are the same number of allocated segments
397 * as free segments, so we will have at least one
398 * entry per free segment. However, small free
399 * segments (smaller than vdev_removal_max_span)
400 * will be combined with adjacent allocated segments
401 * as a single mapping.
403 for (int i
= 0; i
< RANGE_TREE_HISTOGRAM_SIZE
; i
++) {
404 if (1ULL << (i
+ 1) < vdev_removal_max_span
) {
406 vd
->vdev_mg
->mg_histogram
[i
] <<
410 vd
->vdev_mg
->mg_histogram
[i
];
415 * The maximum length of a mapping is
416 * zfs_remove_max_segment, so we need at least one entry
417 * per zfs_remove_max_segment of allocated data.
419 seg_count
+= to_alloc
/ zfs_remove_max_segment
;
421 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_SIZE
,
423 sizeof (vdev_indirect_mapping_entry_phys_t
));
427 if (!vd
->vdev_ops
->vdev_op_leaf
) {
431 ASSERT(!vd
->vdev_ishole
);
433 child
= kmem_alloc(vd
->vdev_children
* sizeof (nvlist_t
*),
436 for (c
= 0, idx
= 0; c
< vd
->vdev_children
; c
++) {
437 vdev_t
*cvd
= vd
->vdev_child
[c
];
440 * If we're generating an nvlist of removing
441 * vdevs then skip over any device which is
444 if ((flags
& VDEV_CONFIG_REMOVING
) &&
448 child
[idx
++] = vdev_config_generate(spa
, cvd
,
453 fnvlist_add_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
457 for (c
= 0; c
< idx
; c
++)
458 nvlist_free(child
[c
]);
460 kmem_free(child
, vd
->vdev_children
* sizeof (nvlist_t
*));
463 const char *aux
= NULL
;
465 if (vd
->vdev_offline
&& !vd
->vdev_tmpoffline
)
466 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, B_TRUE
);
467 if (vd
->vdev_resilver_txg
!= 0)
468 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_RESILVER_TXG
,
469 vd
->vdev_resilver_txg
);
470 if (vd
->vdev_faulted
)
471 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_FAULTED
, B_TRUE
);
472 if (vd
->vdev_degraded
)
473 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_DEGRADED
, B_TRUE
);
474 if (vd
->vdev_removed
)
475 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_REMOVED
, B_TRUE
);
476 if (vd
->vdev_unspare
)
477 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_UNSPARE
, B_TRUE
);
479 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_HOLE
, B_TRUE
);
481 switch (vd
->vdev_stat
.vs_aux
) {
482 case VDEV_AUX_ERR_EXCEEDED
:
483 aux
= "err_exceeded";
486 case VDEV_AUX_EXTERNAL
:
492 fnvlist_add_string(nv
, ZPOOL_CONFIG_AUX_STATE
, aux
);
494 if (vd
->vdev_splitting
&& vd
->vdev_orig_guid
!= 0LL) {
495 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ORIG_GUID
,
504 * Generate a view of the top-level vdevs. If we currently have holes
505 * in the namespace, then generate an array which contains a list of holey
506 * vdevs. Additionally, add the number of top-level children that currently
510 vdev_top_config_generate(spa_t
*spa
, nvlist_t
*config
)
512 vdev_t
*rvd
= spa
->spa_root_vdev
;
516 array
= kmem_alloc(rvd
->vdev_children
* sizeof (uint64_t), KM_SLEEP
);
518 for (c
= 0, idx
= 0; c
< rvd
->vdev_children
; c
++) {
519 vdev_t
*tvd
= rvd
->vdev_child
[c
];
521 if (tvd
->vdev_ishole
) {
527 VERIFY(nvlist_add_uint64_array(config
, ZPOOL_CONFIG_HOLE_ARRAY
,
531 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_VDEV_CHILDREN
,
532 rvd
->vdev_children
) == 0);
534 kmem_free(array
, rvd
->vdev_children
* sizeof (uint64_t));
538 * Returns the configuration from the label of the given vdev. For vdevs
539 * which don't have a txg value stored on their label (i.e. spares/cache)
540 * or have not been completely initialized (txg = 0) just return
541 * the configuration from the first valid label we find. Otherwise,
542 * find the most up-to-date label that does not exceed the specified
546 vdev_label_read_config(vdev_t
*vd
, uint64_t txg
)
548 spa_t
*spa
= vd
->vdev_spa
;
549 nvlist_t
*config
= NULL
;
553 uint64_t best_txg
= 0;
554 uint64_t label_txg
= 0;
556 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
557 ZIO_FLAG_SPECULATIVE
;
559 ASSERT(spa_config_held(spa
, SCL_STATE_ALL
, RW_WRITER
) == SCL_STATE_ALL
);
561 if (!vdev_readable(vd
))
564 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
565 vp
= abd_to_buf(vp_abd
);
568 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
569 nvlist_t
*label
= NULL
;
571 zio
= zio_root(spa
, NULL
, NULL
, flags
);
573 vdev_label_read(zio
, vd
, l
, vp_abd
,
574 offsetof(vdev_label_t
, vl_vdev_phys
),
575 sizeof (vdev_phys_t
), NULL
, NULL
, flags
);
577 if (zio_wait(zio
) == 0 &&
578 nvlist_unpack(vp
->vp_nvlist
, sizeof (vp
->vp_nvlist
),
581 * Auxiliary vdevs won't have txg values in their
582 * labels and newly added vdevs may not have been
583 * completely initialized so just return the
584 * configuration from the first valid label we
587 error
= nvlist_lookup_uint64(label
,
588 ZPOOL_CONFIG_POOL_TXG
, &label_txg
);
589 if ((error
|| label_txg
== 0) && !config
) {
592 } else if (label_txg
<= txg
&& label_txg
> best_txg
) {
593 best_txg
= label_txg
;
595 config
= fnvlist_dup(label
);
605 if (config
== NULL
&& !(flags
& ZIO_FLAG_TRYHARD
)) {
606 flags
|= ZIO_FLAG_TRYHARD
;
611 * We found a valid label but it didn't pass txg restrictions.
613 if (config
== NULL
&& label_txg
!= 0) {
614 vdev_dbgmsg(vd
, "label discarded as txg is too large "
615 "(%llu > %llu)", (u_longlong_t
)label_txg
,
625 * Determine if a device is in use. The 'spare_guid' parameter will be filled
626 * in with the device guid if this spare is active elsewhere on the system.
629 vdev_inuse(vdev_t
*vd
, uint64_t crtxg
, vdev_labeltype_t reason
,
630 uint64_t *spare_guid
, uint64_t *l2cache_guid
)
632 spa_t
*spa
= vd
->vdev_spa
;
633 uint64_t state
, pool_guid
, device_guid
, txg
, spare_pool
;
640 *l2cache_guid
= 0ULL;
643 * Read the label, if any, and perform some basic sanity checks.
645 if ((label
= vdev_label_read_config(vd
, -1ULL)) == NULL
)
648 (void) nvlist_lookup_uint64(label
, ZPOOL_CONFIG_CREATE_TXG
,
651 if (nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
653 nvlist_lookup_uint64(label
, ZPOOL_CONFIG_GUID
,
654 &device_guid
) != 0) {
659 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
660 (nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_GUID
,
662 nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_TXG
,
671 * Check to see if this device indeed belongs to the pool it claims to
672 * be a part of. The only way this is allowed is if the device is a hot
673 * spare (which we check for later on).
675 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
676 !spa_guid_exists(pool_guid
, device_guid
) &&
677 !spa_spare_exists(device_guid
, NULL
, NULL
) &&
678 !spa_l2cache_exists(device_guid
, NULL
))
682 * If the transaction group is zero, then this an initialized (but
683 * unused) label. This is only an error if the create transaction
684 * on-disk is the same as the one we're using now, in which case the
685 * user has attempted to add the same vdev multiple times in the same
688 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
689 txg
== 0 && vdtxg
== crtxg
)
693 * Check to see if this is a spare device. We do an explicit check for
694 * spa_has_spare() here because it may be on our pending list of spares
695 * to add. We also check if it is an l2cache device.
697 if (spa_spare_exists(device_guid
, &spare_pool
, NULL
) ||
698 spa_has_spare(spa
, device_guid
)) {
700 *spare_guid
= device_guid
;
703 case VDEV_LABEL_CREATE
:
704 case VDEV_LABEL_L2CACHE
:
707 case VDEV_LABEL_REPLACE
:
708 return (!spa_has_spare(spa
, device_guid
) ||
711 case VDEV_LABEL_SPARE
:
712 return (spa_has_spare(spa
, device_guid
));
717 * Check to see if this is an l2cache device.
719 if (spa_l2cache_exists(device_guid
, NULL
))
723 * We can't rely on a pool's state if it's been imported
724 * read-only. Instead we look to see if the pools is marked
725 * read-only in the namespace and set the state to active.
727 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
728 (spa
= spa_by_guid(pool_guid
, device_guid
)) != NULL
&&
729 spa_mode(spa
) == FREAD
)
730 state
= POOL_STATE_ACTIVE
;
733 * If the device is marked ACTIVE, then this device is in use by another
734 * pool on the system.
736 return (state
== POOL_STATE_ACTIVE
);
740 * Initialize a vdev label. We check to make sure each leaf device is not in
741 * use, and writable. We put down an initial label which we will later
742 * overwrite with a complete label. Note that it's important to do this
743 * sequentially, not in parallel, so that we catch cases of multiple use of the
744 * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with
748 vdev_label_init(vdev_t
*vd
, uint64_t crtxg
, vdev_labeltype_t reason
)
750 spa_t
*spa
= vd
->vdev_spa
;
761 uint64_t spare_guid
, l2cache_guid
;
762 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
;
764 ASSERT(spa_config_held(spa
, SCL_ALL
, RW_WRITER
) == SCL_ALL
);
766 for (int c
= 0; c
< vd
->vdev_children
; c
++)
767 if ((error
= vdev_label_init(vd
->vdev_child
[c
],
768 crtxg
, reason
)) != 0)
771 /* Track the creation time for this vdev */
772 vd
->vdev_crtxg
= crtxg
;
774 if (!vd
->vdev_ops
->vdev_op_leaf
|| !spa_writeable(spa
))
778 * Dead vdevs cannot be initialized.
780 if (vdev_is_dead(vd
))
781 return (SET_ERROR(EIO
));
784 * Determine if the vdev is in use.
786 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_SPLIT
&&
787 vdev_inuse(vd
, crtxg
, reason
, &spare_guid
, &l2cache_guid
))
788 return (SET_ERROR(EBUSY
));
791 * If this is a request to add or replace a spare or l2cache device
792 * that is in use elsewhere on the system, then we must update the
793 * guid (which was initialized to a random value) to reflect the
794 * actual GUID (which is shared between multiple pools).
796 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_L2CACHE
&&
797 spare_guid
!= 0ULL) {
798 uint64_t guid_delta
= spare_guid
- vd
->vdev_guid
;
800 vd
->vdev_guid
+= guid_delta
;
802 for (vdev_t
*pvd
= vd
; pvd
!= NULL
; pvd
= pvd
->vdev_parent
)
803 pvd
->vdev_guid_sum
+= guid_delta
;
806 * If this is a replacement, then we want to fallthrough to the
807 * rest of the code. If we're adding a spare, then it's already
808 * labeled appropriately and we can just return.
810 if (reason
== VDEV_LABEL_SPARE
)
812 ASSERT(reason
== VDEV_LABEL_REPLACE
||
813 reason
== VDEV_LABEL_SPLIT
);
816 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_SPARE
&&
817 l2cache_guid
!= 0ULL) {
818 uint64_t guid_delta
= l2cache_guid
- vd
->vdev_guid
;
820 vd
->vdev_guid
+= guid_delta
;
822 for (vdev_t
*pvd
= vd
; pvd
!= NULL
; pvd
= pvd
->vdev_parent
)
823 pvd
->vdev_guid_sum
+= guid_delta
;
826 * If this is a replacement, then we want to fallthrough to the
827 * rest of the code. If we're adding an l2cache, then it's
828 * already labeled appropriately and we can just return.
830 if (reason
== VDEV_LABEL_L2CACHE
)
832 ASSERT(reason
== VDEV_LABEL_REPLACE
);
836 * Initialize its label.
838 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
839 abd_zero(vp_abd
, sizeof (vdev_phys_t
));
840 vp
= abd_to_buf(vp_abd
);
843 * Generate a label describing the pool and our top-level vdev.
844 * We mark it as being from txg 0 to indicate that it's not
845 * really part of an active pool just yet. The labels will
846 * be written again with a meaningful txg by spa_sync().
848 if (reason
== VDEV_LABEL_SPARE
||
849 (reason
== VDEV_LABEL_REMOVE
&& vd
->vdev_isspare
)) {
851 * For inactive hot spares, we generate a special label that
852 * identifies as a mutually shared hot spare. We write the
853 * label if we are adding a hot spare, or if we are removing an
854 * active hot spare (in which case we want to revert the
857 VERIFY(nvlist_alloc(&label
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
859 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_VERSION
,
860 spa_version(spa
)) == 0);
861 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
862 POOL_STATE_SPARE
) == 0);
863 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_GUID
,
864 vd
->vdev_guid
) == 0);
865 } else if (reason
== VDEV_LABEL_L2CACHE
||
866 (reason
== VDEV_LABEL_REMOVE
&& vd
->vdev_isl2cache
)) {
868 * For level 2 ARC devices, add a special label.
870 VERIFY(nvlist_alloc(&label
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
872 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_VERSION
,
873 spa_version(spa
)) == 0);
874 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
875 POOL_STATE_L2CACHE
) == 0);
876 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_GUID
,
877 vd
->vdev_guid
) == 0);
881 if (reason
== VDEV_LABEL_SPLIT
)
882 txg
= spa
->spa_uberblock
.ub_txg
;
883 label
= spa_config_generate(spa
, vd
, txg
, B_FALSE
);
886 * Add our creation time. This allows us to detect multiple
887 * vdev uses as described above, and automatically expires if we
890 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_CREATE_TXG
,
895 buflen
= sizeof (vp
->vp_nvlist
);
897 error
= nvlist_pack(label
, &buf
, &buflen
, NV_ENCODE_XDR
, KM_SLEEP
);
901 /* EFAULT means nvlist_pack ran out of room */
902 return (error
== EFAULT
? ENAMETOOLONG
: EINVAL
);
906 * Initialize uberblock template.
908 ub_abd
= abd_alloc_linear(VDEV_UBERBLOCK_RING
, B_TRUE
);
909 abd_zero(ub_abd
, VDEV_UBERBLOCK_RING
);
910 abd_copy_from_buf(ub_abd
, &spa
->spa_uberblock
, sizeof (uberblock_t
));
911 ub
= abd_to_buf(ub_abd
);
914 /* Initialize the 2nd padding area. */
915 pad2
= abd_alloc_for_io(VDEV_PAD_SIZE
, B_TRUE
);
916 abd_zero(pad2
, VDEV_PAD_SIZE
);
919 * Write everything in parallel.
922 zio
= zio_root(spa
, NULL
, NULL
, flags
);
924 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
926 vdev_label_write(zio
, vd
, l
, vp_abd
,
927 offsetof(vdev_label_t
, vl_vdev_phys
),
928 sizeof (vdev_phys_t
), NULL
, NULL
, flags
);
931 * Skip the 1st padding area.
932 * Zero out the 2nd padding area where it might have
933 * left over data from previous filesystem format.
935 vdev_label_write(zio
, vd
, l
, pad2
,
936 offsetof(vdev_label_t
, vl_pad2
),
937 VDEV_PAD_SIZE
, NULL
, NULL
, flags
);
939 vdev_label_write(zio
, vd
, l
, ub_abd
,
940 offsetof(vdev_label_t
, vl_uberblock
),
941 VDEV_UBERBLOCK_RING
, NULL
, NULL
, flags
);
944 error
= zio_wait(zio
);
946 if (error
!= 0 && !(flags
& ZIO_FLAG_TRYHARD
)) {
947 flags
|= ZIO_FLAG_TRYHARD
;
957 * If this vdev hasn't been previously identified as a spare, then we
958 * mark it as such only if a) we are labeling it as a spare, or b) it
959 * exists as a spare elsewhere in the system. Do the same for
960 * level 2 ARC devices.
962 if (error
== 0 && !vd
->vdev_isspare
&&
963 (reason
== VDEV_LABEL_SPARE
||
964 spa_spare_exists(vd
->vdev_guid
, NULL
, NULL
)))
967 if (error
== 0 && !vd
->vdev_isl2cache
&&
968 (reason
== VDEV_LABEL_L2CACHE
||
969 spa_l2cache_exists(vd
->vdev_guid
, NULL
)))
976 * ==========================================================================
977 * uberblock load/sync
978 * ==========================================================================
982 * Consider the following situation: txg is safely synced to disk. We've
983 * written the first uberblock for txg + 1, and then we lose power. When we
984 * come back up, we fail to see the uberblock for txg + 1 because, say,
985 * it was on a mirrored device and the replica to which we wrote txg + 1
986 * is now offline. If we then make some changes and sync txg + 1, and then
987 * the missing replica comes back, then for a few seconds we'll have two
988 * conflicting uberblocks on disk with the same txg. The solution is simple:
989 * among uberblocks with equal txg, choose the one with the latest timestamp.
992 vdev_uberblock_compare(uberblock_t
*ub1
, uberblock_t
*ub2
)
994 if (ub1
->ub_txg
< ub2
->ub_txg
)
996 if (ub1
->ub_txg
> ub2
->ub_txg
)
999 if (ub1
->ub_timestamp
< ub2
->ub_timestamp
)
1001 if (ub1
->ub_timestamp
> ub2
->ub_timestamp
)
1008 uberblock_t
*ubl_ubbest
; /* Best uberblock */
1009 vdev_t
*ubl_vd
; /* vdev associated with the above */
1013 vdev_uberblock_load_done(zio_t
*zio
)
1015 vdev_t
*vd
= zio
->io_vd
;
1016 spa_t
*spa
= zio
->io_spa
;
1017 zio_t
*rio
= zio
->io_private
;
1018 uberblock_t
*ub
= abd_to_buf(zio
->io_abd
);
1019 struct ubl_cbdata
*cbp
= rio
->io_private
;
1021 ASSERT3U(zio
->io_size
, ==, VDEV_UBERBLOCK_SIZE(vd
));
1023 if (zio
->io_error
== 0 && uberblock_verify(ub
) == 0) {
1024 mutex_enter(&rio
->io_lock
);
1025 if (ub
->ub_txg
<= spa
->spa_load_max_txg
&&
1026 vdev_uberblock_compare(ub
, cbp
->ubl_ubbest
) > 0) {
1028 * Keep track of the vdev in which this uberblock
1029 * was found. We will use this information later
1030 * to obtain the config nvlist associated with
1033 *cbp
->ubl_ubbest
= *ub
;
1036 mutex_exit(&rio
->io_lock
);
1039 abd_free(zio
->io_abd
);
1043 vdev_uberblock_load_impl(zio_t
*zio
, vdev_t
*vd
, int flags
,
1044 struct ubl_cbdata
*cbp
)
1046 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1047 vdev_uberblock_load_impl(zio
, vd
->vdev_child
[c
], flags
, cbp
);
1049 if (vd
->vdev_ops
->vdev_op_leaf
&& vdev_readable(vd
)) {
1050 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
1051 for (int n
= 0; n
< VDEV_UBERBLOCK_COUNT(vd
); n
++) {
1052 vdev_label_read(zio
, vd
, l
,
1053 abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd
),
1054 B_TRUE
), VDEV_UBERBLOCK_OFFSET(vd
, n
),
1055 VDEV_UBERBLOCK_SIZE(vd
),
1056 vdev_uberblock_load_done
, zio
, flags
);
1063 * Reads the 'best' uberblock from disk along with its associated
1064 * configuration. First, we read the uberblock array of each label of each
1065 * vdev, keeping track of the uberblock with the highest txg in each array.
1066 * Then, we read the configuration from the same vdev as the best uberblock.
1069 vdev_uberblock_load(vdev_t
*rvd
, uberblock_t
*ub
, nvlist_t
**config
)
1072 spa_t
*spa
= rvd
->vdev_spa
;
1073 struct ubl_cbdata cb
;
1074 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
1075 ZIO_FLAG_SPECULATIVE
| ZIO_FLAG_TRYHARD
;
1080 bzero(ub
, sizeof (uberblock_t
));
1086 spa_config_enter(spa
, SCL_ALL
, FTAG
, RW_WRITER
);
1087 zio
= zio_root(spa
, NULL
, &cb
, flags
);
1088 vdev_uberblock_load_impl(zio
, rvd
, flags
, &cb
);
1089 (void) zio_wait(zio
);
1092 * It's possible that the best uberblock was discovered on a label
1093 * that has a configuration which was written in a future txg.
1094 * Search all labels on this vdev to find the configuration that
1095 * matches the txg for our uberblock.
1097 if (cb
.ubl_vd
!= NULL
) {
1098 vdev_dbgmsg(cb
.ubl_vd
, "best uberblock found for spa %s. "
1099 "txg %llu", spa
->spa_name
, (u_longlong_t
)ub
->ub_txg
);
1101 *config
= vdev_label_read_config(cb
.ubl_vd
, ub
->ub_txg
);
1102 if (*config
== NULL
&& spa
->spa_extreme_rewind
) {
1103 vdev_dbgmsg(cb
.ubl_vd
, "failed to read label config. "
1104 "Trying again without txg restrictions.");
1105 *config
= vdev_label_read_config(cb
.ubl_vd
, UINT64_MAX
);
1107 if (*config
== NULL
) {
1108 vdev_dbgmsg(cb
.ubl_vd
, "failed to read label config");
1111 spa_config_exit(spa
, SCL_ALL
, FTAG
);
1115 * On success, increment root zio's count of good writes.
1116 * We only get credit for writes to known-visible vdevs; see spa_vdev_add().
1119 vdev_uberblock_sync_done(zio_t
*zio
)
1121 uint64_t *good_writes
= zio
->io_private
;
1123 if (zio
->io_error
== 0 && zio
->io_vd
->vdev_top
->vdev_ms_array
!= 0)
1124 atomic_inc_64(good_writes
);
1128 * Write the uberblock to all labels of all leaves of the specified vdev.
1131 vdev_uberblock_sync(zio_t
*zio
, uint64_t *good_writes
,
1132 uberblock_t
*ub
, vdev_t
*vd
, int flags
)
1134 for (uint64_t c
= 0; c
< vd
->vdev_children
; c
++) {
1135 vdev_uberblock_sync(zio
, good_writes
,
1136 ub
, vd
->vdev_child
[c
], flags
);
1139 if (!vd
->vdev_ops
->vdev_op_leaf
)
1142 if (!vdev_writeable(vd
))
1145 int n
= ub
->ub_txg
& (VDEV_UBERBLOCK_COUNT(vd
) - 1);
1147 /* Copy the uberblock_t into the ABD */
1148 abd_t
*ub_abd
= abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd
), B_TRUE
);
1149 abd_zero(ub_abd
, VDEV_UBERBLOCK_SIZE(vd
));
1150 abd_copy_from_buf(ub_abd
, ub
, sizeof (uberblock_t
));
1152 for (int l
= 0; l
< VDEV_LABELS
; l
++)
1153 vdev_label_write(zio
, vd
, l
, ub_abd
,
1154 VDEV_UBERBLOCK_OFFSET(vd
, n
), VDEV_UBERBLOCK_SIZE(vd
),
1155 vdev_uberblock_sync_done
, good_writes
,
1156 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1161 /* Sync the uberblocks to all vdevs in svd[] */
1163 vdev_uberblock_sync_list(vdev_t
**svd
, int svdcount
, uberblock_t
*ub
, int flags
)
1165 spa_t
*spa
= svd
[0]->vdev_spa
;
1167 uint64_t good_writes
= 0;
1169 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1171 for (int v
= 0; v
< svdcount
; v
++)
1172 vdev_uberblock_sync(zio
, &good_writes
, ub
, svd
[v
], flags
);
1174 (void) zio_wait(zio
);
1177 * Flush the uberblocks to disk. This ensures that the odd labels
1178 * are no longer needed (because the new uberblocks and the even
1179 * labels are safely on disk), so it is safe to overwrite them.
1181 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1183 for (int v
= 0; v
< svdcount
; v
++) {
1184 if (vdev_writeable(svd
[v
])) {
1185 zio_flush(zio
, svd
[v
]);
1189 (void) zio_wait(zio
);
1191 return (good_writes
>= 1 ? 0 : EIO
);
1195 * On success, increment the count of good writes for our top-level vdev.
1198 vdev_label_sync_done(zio_t
*zio
)
1200 uint64_t *good_writes
= zio
->io_private
;
1202 if (zio
->io_error
== 0)
1203 atomic_inc_64(good_writes
);
1207 * If there weren't enough good writes, indicate failure to the parent.
1210 vdev_label_sync_top_done(zio_t
*zio
)
1212 uint64_t *good_writes
= zio
->io_private
;
1214 if (*good_writes
== 0)
1215 zio
->io_error
= SET_ERROR(EIO
);
1217 kmem_free(good_writes
, sizeof (uint64_t));
1221 * We ignore errors for log and cache devices, simply free the private data.
1224 vdev_label_sync_ignore_done(zio_t
*zio
)
1226 kmem_free(zio
->io_private
, sizeof (uint64_t));
1230 * Write all even or odd labels to all leaves of the specified vdev.
1233 vdev_label_sync(zio_t
*zio
, uint64_t *good_writes
,
1234 vdev_t
*vd
, int l
, uint64_t txg
, int flags
)
1242 for (int c
= 0; c
< vd
->vdev_children
; c
++) {
1243 vdev_label_sync(zio
, good_writes
,
1244 vd
->vdev_child
[c
], l
, txg
, flags
);
1247 if (!vd
->vdev_ops
->vdev_op_leaf
)
1250 if (!vdev_writeable(vd
))
1254 * Generate a label describing the top-level config to which we belong.
1256 label
= spa_config_generate(vd
->vdev_spa
, vd
, txg
, B_FALSE
);
1258 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
1259 abd_zero(vp_abd
, sizeof (vdev_phys_t
));
1260 vp
= abd_to_buf(vp_abd
);
1262 buf
= vp
->vp_nvlist
;
1263 buflen
= sizeof (vp
->vp_nvlist
);
1265 if (nvlist_pack(label
, &buf
, &buflen
, NV_ENCODE_XDR
, KM_SLEEP
) == 0) {
1266 for (; l
< VDEV_LABELS
; l
+= 2) {
1267 vdev_label_write(zio
, vd
, l
, vp_abd
,
1268 offsetof(vdev_label_t
, vl_vdev_phys
),
1269 sizeof (vdev_phys_t
),
1270 vdev_label_sync_done
, good_writes
,
1271 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1280 vdev_label_sync_list(spa_t
*spa
, int l
, uint64_t txg
, int flags
)
1282 list_t
*dl
= &spa
->spa_config_dirty_list
;
1288 * Write the new labels to disk.
1290 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1292 for (vd
= list_head(dl
); vd
!= NULL
; vd
= list_next(dl
, vd
)) {
1293 uint64_t *good_writes
= kmem_zalloc(sizeof (uint64_t),
1296 ASSERT(!vd
->vdev_ishole
);
1298 zio_t
*vio
= zio_null(zio
, spa
, NULL
,
1299 (vd
->vdev_islog
|| vd
->vdev_aux
!= NULL
) ?
1300 vdev_label_sync_ignore_done
: vdev_label_sync_top_done
,
1301 good_writes
, flags
);
1302 vdev_label_sync(vio
, good_writes
, vd
, l
, txg
, flags
);
1306 error
= zio_wait(zio
);
1309 * Flush the new labels to disk.
1311 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1313 for (vd
= list_head(dl
); vd
!= NULL
; vd
= list_next(dl
, vd
))
1316 (void) zio_wait(zio
);
1322 * Sync the uberblock and any changes to the vdev configuration.
1324 * The order of operations is carefully crafted to ensure that
1325 * if the system panics or loses power at any time, the state on disk
1326 * is still transactionally consistent. The in-line comments below
1327 * describe the failure semantics at each stage.
1329 * Moreover, vdev_config_sync() is designed to be idempotent: if it fails
1330 * at any time, you can just call it again, and it will resume its work.
1333 vdev_config_sync(vdev_t
**svd
, int svdcount
, uint64_t txg
)
1335 spa_t
*spa
= svd
[0]->vdev_spa
;
1336 uberblock_t
*ub
= &spa
->spa_uberblock
;
1338 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
;
1340 ASSERT(svdcount
!= 0);
1343 * Normally, we don't want to try too hard to write every label and
1344 * uberblock. If there is a flaky disk, we don't want the rest of the
1345 * sync process to block while we retry. But if we can't write a
1346 * single label out, we should retry with ZIO_FLAG_TRYHARD before
1347 * bailing out and declaring the pool faulted.
1350 if ((flags
& ZIO_FLAG_TRYHARD
) != 0)
1352 flags
|= ZIO_FLAG_TRYHARD
;
1355 ASSERT(ub
->ub_txg
<= txg
);
1358 * If this isn't a resync due to I/O errors,
1359 * and nothing changed in this transaction group,
1360 * and the vdev configuration hasn't changed,
1361 * then there's nothing to do.
1363 if (ub
->ub_txg
< txg
&&
1364 uberblock_update(ub
, spa
->spa_root_vdev
, txg
) == B_FALSE
&&
1365 list_is_empty(&spa
->spa_config_dirty_list
))
1368 if (txg
> spa_freeze_txg(spa
))
1371 ASSERT(txg
<= spa
->spa_final_txg
);
1374 * Flush the write cache of every disk that's been written to
1375 * in this transaction group. This ensures that all blocks
1376 * written in this txg will be committed to stable storage
1377 * before any uberblock that references them.
1379 zio_t
*zio
= zio_root(spa
, NULL
, NULL
, flags
);
1382 txg_list_head(&spa
->spa_vdev_txg_list
, TXG_CLEAN(txg
)); vd
!= NULL
;
1383 vd
= txg_list_next(&spa
->spa_vdev_txg_list
, vd
, TXG_CLEAN(txg
)))
1386 (void) zio_wait(zio
);
1389 * Sync out the even labels (L0, L2) for every dirty vdev. If the
1390 * system dies in the middle of this process, that's OK: all of the
1391 * even labels that made it to disk will be newer than any uberblock,
1392 * and will therefore be considered invalid. The odd labels (L1, L3),
1393 * which have not yet been touched, will still be valid. We flush
1394 * the new labels to disk to ensure that all even-label updates
1395 * are committed to stable storage before the uberblock update.
1397 if ((error
= vdev_label_sync_list(spa
, 0, txg
, flags
)) != 0) {
1398 if ((flags
& ZIO_FLAG_TRYHARD
) != 0) {
1399 zfs_dbgmsg("vdev_label_sync_list() returned error %d "
1400 "for pool '%s' when syncing out the even labels "
1401 "of dirty vdevs", error
, spa_name(spa
));
1407 * Sync the uberblocks to all vdevs in svd[].
1408 * If the system dies in the middle of this step, there are two cases
1409 * to consider, and the on-disk state is consistent either way:
1411 * (1) If none of the new uberblocks made it to disk, then the
1412 * previous uberblock will be the newest, and the odd labels
1413 * (which had not yet been touched) will be valid with respect
1414 * to that uberblock.
1416 * (2) If one or more new uberblocks made it to disk, then they
1417 * will be the newest, and the even labels (which had all
1418 * been successfully committed) will be valid with respect
1419 * to the new uberblocks.
1421 if ((error
= vdev_uberblock_sync_list(svd
, svdcount
, ub
, flags
)) != 0) {
1422 if ((flags
& ZIO_FLAG_TRYHARD
) != 0) {
1423 zfs_dbgmsg("vdev_uberblock_sync_list() returned error "
1424 "%d for pool '%s'", error
, spa_name(spa
));
1430 * Sync out odd labels for every dirty vdev. If the system dies
1431 * in the middle of this process, the even labels and the new
1432 * uberblocks will suffice to open the pool. The next time
1433 * the pool is opened, the first thing we'll do -- before any
1434 * user data is modified -- is mark every vdev dirty so that
1435 * all labels will be brought up to date. We flush the new labels
1436 * to disk to ensure that all odd-label updates are committed to
1437 * stable storage before the next transaction group begins.
1439 if ((error
= vdev_label_sync_list(spa
, 1, txg
, flags
)) != 0) {
1440 if ((flags
& ZIO_FLAG_TRYHARD
) != 0) {
1441 zfs_dbgmsg("vdev_label_sync_list() returned error %d "
1442 "for pool '%s' when syncing out the odd labels of "
1443 "dirty vdevs", error
, spa_name(spa
));