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, 2016 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 * Generate the nvlist representing this vdev's config.
216 vdev_config_generate(spa_t
*spa
, vdev_t
*vd
, boolean_t getstats
,
217 vdev_config_flag_t flags
)
220 vdev_indirect_config_t
*vic
= &vd
->vdev_indirect_config
;
222 nv
= fnvlist_alloc();
224 fnvlist_add_string(nv
, ZPOOL_CONFIG_TYPE
, vd
->vdev_ops
->vdev_op_type
);
225 if (!(flags
& (VDEV_CONFIG_SPARE
| VDEV_CONFIG_L2CACHE
)))
226 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ID
, vd
->vdev_id
);
227 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_GUID
, vd
->vdev_guid
);
229 if (vd
->vdev_path
!= NULL
)
230 fnvlist_add_string(nv
, ZPOOL_CONFIG_PATH
, vd
->vdev_path
);
232 if (vd
->vdev_devid
!= NULL
)
233 fnvlist_add_string(nv
, ZPOOL_CONFIG_DEVID
, vd
->vdev_devid
);
235 if (vd
->vdev_physpath
!= NULL
)
236 fnvlist_add_string(nv
, ZPOOL_CONFIG_PHYS_PATH
,
239 if (vd
->vdev_fru
!= NULL
)
240 fnvlist_add_string(nv
, ZPOOL_CONFIG_FRU
, vd
->vdev_fru
);
242 if (vd
->vdev_nparity
!= 0) {
243 ASSERT(strcmp(vd
->vdev_ops
->vdev_op_type
,
244 VDEV_TYPE_RAIDZ
) == 0);
247 * Make sure someone hasn't managed to sneak a fancy new vdev
248 * into a crufty old storage pool.
250 ASSERT(vd
->vdev_nparity
== 1 ||
251 (vd
->vdev_nparity
<= 2 &&
252 spa_version(spa
) >= SPA_VERSION_RAIDZ2
) ||
253 (vd
->vdev_nparity
<= 3 &&
254 spa_version(spa
) >= SPA_VERSION_RAIDZ3
));
257 * Note that we'll add the nparity tag even on storage pools
258 * that only support a single parity device -- older software
259 * will just ignore it.
261 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_NPARITY
, vd
->vdev_nparity
);
264 if (vd
->vdev_wholedisk
!= -1ULL)
265 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
268 if (vd
->vdev_not_present
)
269 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
, 1);
271 if (vd
->vdev_isspare
)
272 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_SPARE
, 1);
274 if (!(flags
& (VDEV_CONFIG_SPARE
| VDEV_CONFIG_L2CACHE
)) &&
275 vd
== vd
->vdev_top
) {
276 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_METASLAB_ARRAY
,
278 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_METASLAB_SHIFT
,
280 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ASHIFT
, vd
->vdev_ashift
);
281 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ASIZE
,
283 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_LOG
, vd
->vdev_islog
);
284 if (vd
->vdev_removing
) {
285 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_REMOVING
,
290 if (vd
->vdev_dtl_sm
!= NULL
) {
291 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_DTL
,
292 space_map_object(vd
->vdev_dtl_sm
));
295 if (vic
->vic_mapping_object
!= 0) {
296 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_OBJECT
,
297 vic
->vic_mapping_object
);
300 if (vic
->vic_births_object
!= 0) {
301 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_BIRTHS
,
302 vic
->vic_births_object
);
305 if (vic
->vic_prev_indirect_vdev
!= UINT64_MAX
) {
306 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_PREV_INDIRECT_VDEV
,
307 vic
->vic_prev_indirect_vdev
);
311 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_CREATE_TXG
, vd
->vdev_crtxg
);
313 if (flags
& VDEV_CONFIG_MOS
) {
314 if (vd
->vdev_leaf_zap
!= 0) {
315 ASSERT(vd
->vdev_ops
->vdev_op_leaf
);
316 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_VDEV_LEAF_ZAP
,
320 if (vd
->vdev_top_zap
!= 0) {
321 ASSERT(vd
== vd
->vdev_top
);
322 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_VDEV_TOP_ZAP
,
330 vdev_get_stats(vd
, &vs
);
331 fnvlist_add_uint64_array(nv
, ZPOOL_CONFIG_VDEV_STATS
,
332 (uint64_t *)&vs
, sizeof (vs
) / sizeof (uint64_t));
334 /* provide either current or previous scan information */
336 if (spa_scan_get_stats(spa
, &ps
) == 0) {
337 fnvlist_add_uint64_array(nv
,
338 ZPOOL_CONFIG_SCAN_STATS
, (uint64_t *)&ps
,
339 sizeof (pool_scan_stat_t
) / sizeof (uint64_t));
342 pool_removal_stat_t prs
;
343 if (spa_removal_get_stats(spa
, &prs
) == 0) {
344 fnvlist_add_uint64_array(nv
,
345 ZPOOL_CONFIG_REMOVAL_STATS
, (uint64_t *)&prs
,
346 sizeof (prs
) / sizeof (uint64_t));
350 * Note: this can be called from open context
351 * (spa_get_stats()), so we need the rwlock to prevent
352 * the mapping from being changed by condensing.
354 rw_enter(&vd
->vdev_indirect_rwlock
, RW_READER
);
355 if (vd
->vdev_indirect_mapping
!= NULL
) {
356 ASSERT(vd
->vdev_indirect_births
!= NULL
);
357 vdev_indirect_mapping_t
*vim
=
358 vd
->vdev_indirect_mapping
;
359 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_SIZE
,
360 vdev_indirect_mapping_size(vim
));
362 rw_exit(&vd
->vdev_indirect_rwlock
);
363 if (vd
->vdev_mg
!= NULL
&&
364 vd
->vdev_mg
->mg_fragmentation
!= ZFS_FRAG_INVALID
) {
366 * Compute approximately how much memory would be used
367 * for the indirect mapping if this device were to
370 * Note: If the frag metric is invalid, then not
371 * enough metaslabs have been converted to have
374 uint64_t seg_count
= 0;
377 * There are the same number of allocated segments
378 * as free segments, so we will have at least one
379 * entry per free segment.
381 for (int i
= 0; i
< RANGE_TREE_HISTOGRAM_SIZE
; i
++) {
382 seg_count
+= vd
->vdev_mg
->mg_histogram
[i
];
386 * The maximum length of a mapping is SPA_MAXBLOCKSIZE,
387 * so we need at least one entry per SPA_MAXBLOCKSIZE
390 seg_count
+= vd
->vdev_stat
.vs_alloc
/ SPA_MAXBLOCKSIZE
;
392 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_SIZE
,
394 sizeof (vdev_indirect_mapping_entry_phys_t
));
398 if (!vd
->vdev_ops
->vdev_op_leaf
) {
402 ASSERT(!vd
->vdev_ishole
);
404 child
= kmem_alloc(vd
->vdev_children
* sizeof (nvlist_t
*),
407 for (c
= 0, idx
= 0; c
< vd
->vdev_children
; c
++) {
408 vdev_t
*cvd
= vd
->vdev_child
[c
];
411 * If we're generating an nvlist of removing
412 * vdevs then skip over any device which is
415 if ((flags
& VDEV_CONFIG_REMOVING
) &&
419 child
[idx
++] = vdev_config_generate(spa
, cvd
,
424 fnvlist_add_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
428 for (c
= 0; c
< idx
; c
++)
429 nvlist_free(child
[c
]);
431 kmem_free(child
, vd
->vdev_children
* sizeof (nvlist_t
*));
434 const char *aux
= NULL
;
436 if (vd
->vdev_offline
&& !vd
->vdev_tmpoffline
)
437 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, B_TRUE
);
438 if (vd
->vdev_resilver_txg
!= 0)
439 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_RESILVER_TXG
,
440 vd
->vdev_resilver_txg
);
441 if (vd
->vdev_faulted
)
442 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_FAULTED
, B_TRUE
);
443 if (vd
->vdev_degraded
)
444 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_DEGRADED
, B_TRUE
);
445 if (vd
->vdev_removed
)
446 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_REMOVED
, B_TRUE
);
447 if (vd
->vdev_unspare
)
448 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_UNSPARE
, B_TRUE
);
450 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_HOLE
, B_TRUE
);
452 switch (vd
->vdev_stat
.vs_aux
) {
453 case VDEV_AUX_ERR_EXCEEDED
:
454 aux
= "err_exceeded";
457 case VDEV_AUX_EXTERNAL
:
463 fnvlist_add_string(nv
, ZPOOL_CONFIG_AUX_STATE
, aux
);
465 if (vd
->vdev_splitting
&& vd
->vdev_orig_guid
!= 0LL) {
466 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ORIG_GUID
,
475 * Generate a view of the top-level vdevs. If we currently have holes
476 * in the namespace, then generate an array which contains a list of holey
477 * vdevs. Additionally, add the number of top-level children that currently
481 vdev_top_config_generate(spa_t
*spa
, nvlist_t
*config
)
483 vdev_t
*rvd
= spa
->spa_root_vdev
;
487 array
= kmem_alloc(rvd
->vdev_children
* sizeof (uint64_t), KM_SLEEP
);
489 for (c
= 0, idx
= 0; c
< rvd
->vdev_children
; c
++) {
490 vdev_t
*tvd
= rvd
->vdev_child
[c
];
492 if (tvd
->vdev_ishole
) {
498 VERIFY(nvlist_add_uint64_array(config
, ZPOOL_CONFIG_HOLE_ARRAY
,
502 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_VDEV_CHILDREN
,
503 rvd
->vdev_children
) == 0);
505 kmem_free(array
, rvd
->vdev_children
* sizeof (uint64_t));
509 * Returns the configuration from the label of the given vdev. For vdevs
510 * which don't have a txg value stored on their label (i.e. spares/cache)
511 * or have not been completely initialized (txg = 0) just return
512 * the configuration from the first valid label we find. Otherwise,
513 * find the most up-to-date label that does not exceed the specified
517 vdev_label_read_config(vdev_t
*vd
, uint64_t txg
)
519 spa_t
*spa
= vd
->vdev_spa
;
520 nvlist_t
*config
= NULL
;
524 uint64_t best_txg
= 0;
526 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
527 ZIO_FLAG_SPECULATIVE
;
529 ASSERT(spa_config_held(spa
, SCL_STATE_ALL
, RW_WRITER
) == SCL_STATE_ALL
);
531 if (!vdev_readable(vd
))
534 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
535 vp
= abd_to_buf(vp_abd
);
538 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
539 nvlist_t
*label
= NULL
;
541 zio
= zio_root(spa
, NULL
, NULL
, flags
);
543 vdev_label_read(zio
, vd
, l
, vp_abd
,
544 offsetof(vdev_label_t
, vl_vdev_phys
),
545 sizeof (vdev_phys_t
), NULL
, NULL
, flags
);
547 if (zio_wait(zio
) == 0 &&
548 nvlist_unpack(vp
->vp_nvlist
, sizeof (vp
->vp_nvlist
),
550 uint64_t label_txg
= 0;
553 * Auxiliary vdevs won't have txg values in their
554 * labels and newly added vdevs may not have been
555 * completely initialized so just return the
556 * configuration from the first valid label we
559 error
= nvlist_lookup_uint64(label
,
560 ZPOOL_CONFIG_POOL_TXG
, &label_txg
);
561 if ((error
|| label_txg
== 0) && !config
) {
564 } else if (label_txg
<= txg
&& label_txg
> best_txg
) {
565 best_txg
= label_txg
;
567 config
= fnvlist_dup(label
);
577 if (config
== NULL
&& !(flags
& ZIO_FLAG_TRYHARD
)) {
578 flags
|= ZIO_FLAG_TRYHARD
;
588 * Determine if a device is in use. The 'spare_guid' parameter will be filled
589 * in with the device guid if this spare is active elsewhere on the system.
592 vdev_inuse(vdev_t
*vd
, uint64_t crtxg
, vdev_labeltype_t reason
,
593 uint64_t *spare_guid
, uint64_t *l2cache_guid
)
595 spa_t
*spa
= vd
->vdev_spa
;
596 uint64_t state
, pool_guid
, device_guid
, txg
, spare_pool
;
603 *l2cache_guid
= 0ULL;
606 * Read the label, if any, and perform some basic sanity checks.
608 if ((label
= vdev_label_read_config(vd
, -1ULL)) == NULL
)
611 (void) nvlist_lookup_uint64(label
, ZPOOL_CONFIG_CREATE_TXG
,
614 if (nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
616 nvlist_lookup_uint64(label
, ZPOOL_CONFIG_GUID
,
617 &device_guid
) != 0) {
622 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
623 (nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_GUID
,
625 nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_TXG
,
634 * Check to see if this device indeed belongs to the pool it claims to
635 * be a part of. The only way this is allowed is if the device is a hot
636 * spare (which we check for later on).
638 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
639 !spa_guid_exists(pool_guid
, device_guid
) &&
640 !spa_spare_exists(device_guid
, NULL
, NULL
) &&
641 !spa_l2cache_exists(device_guid
, NULL
))
645 * If the transaction group is zero, then this an initialized (but
646 * unused) label. This is only an error if the create transaction
647 * on-disk is the same as the one we're using now, in which case the
648 * user has attempted to add the same vdev multiple times in the same
651 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
652 txg
== 0 && vdtxg
== crtxg
)
656 * Check to see if this is a spare device. We do an explicit check for
657 * spa_has_spare() here because it may be on our pending list of spares
658 * to add. We also check if it is an l2cache device.
660 if (spa_spare_exists(device_guid
, &spare_pool
, NULL
) ||
661 spa_has_spare(spa
, device_guid
)) {
663 *spare_guid
= device_guid
;
666 case VDEV_LABEL_CREATE
:
667 case VDEV_LABEL_L2CACHE
:
670 case VDEV_LABEL_REPLACE
:
671 return (!spa_has_spare(spa
, device_guid
) ||
674 case VDEV_LABEL_SPARE
:
675 return (spa_has_spare(spa
, device_guid
));
680 * Check to see if this is an l2cache device.
682 if (spa_l2cache_exists(device_guid
, NULL
))
686 * We can't rely on a pool's state if it's been imported
687 * read-only. Instead we look to see if the pools is marked
688 * read-only in the namespace and set the state to active.
690 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
691 (spa
= spa_by_guid(pool_guid
, device_guid
)) != NULL
&&
692 spa_mode(spa
) == FREAD
)
693 state
= POOL_STATE_ACTIVE
;
696 * If the device is marked ACTIVE, then this device is in use by another
697 * pool on the system.
699 return (state
== POOL_STATE_ACTIVE
);
703 * Initialize a vdev label. We check to make sure each leaf device is not in
704 * use, and writable. We put down an initial label which we will later
705 * overwrite with a complete label. Note that it's important to do this
706 * sequentially, not in parallel, so that we catch cases of multiple use of the
707 * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with
711 vdev_label_init(vdev_t
*vd
, uint64_t crtxg
, vdev_labeltype_t reason
)
713 spa_t
*spa
= vd
->vdev_spa
;
724 uint64_t spare_guid
, l2cache_guid
;
725 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
;
727 ASSERT(spa_config_held(spa
, SCL_ALL
, RW_WRITER
) == SCL_ALL
);
729 for (int c
= 0; c
< vd
->vdev_children
; c
++)
730 if ((error
= vdev_label_init(vd
->vdev_child
[c
],
731 crtxg
, reason
)) != 0)
734 /* Track the creation time for this vdev */
735 vd
->vdev_crtxg
= crtxg
;
737 if (!vd
->vdev_ops
->vdev_op_leaf
|| !spa_writeable(spa
))
741 * Dead vdevs cannot be initialized.
743 if (vdev_is_dead(vd
))
744 return (SET_ERROR(EIO
));
747 * Determine if the vdev is in use.
749 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_SPLIT
&&
750 vdev_inuse(vd
, crtxg
, reason
, &spare_guid
, &l2cache_guid
))
751 return (SET_ERROR(EBUSY
));
754 * If this is a request to add or replace a spare or l2cache device
755 * that is in use elsewhere on the system, then we must update the
756 * guid (which was initialized to a random value) to reflect the
757 * actual GUID (which is shared between multiple pools).
759 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_L2CACHE
&&
760 spare_guid
!= 0ULL) {
761 uint64_t guid_delta
= spare_guid
- vd
->vdev_guid
;
763 vd
->vdev_guid
+= guid_delta
;
765 for (vdev_t
*pvd
= vd
; pvd
!= NULL
; pvd
= pvd
->vdev_parent
)
766 pvd
->vdev_guid_sum
+= guid_delta
;
769 * If this is a replacement, then we want to fallthrough to the
770 * rest of the code. If we're adding a spare, then it's already
771 * labeled appropriately and we can just return.
773 if (reason
== VDEV_LABEL_SPARE
)
775 ASSERT(reason
== VDEV_LABEL_REPLACE
||
776 reason
== VDEV_LABEL_SPLIT
);
779 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_SPARE
&&
780 l2cache_guid
!= 0ULL) {
781 uint64_t guid_delta
= l2cache_guid
- vd
->vdev_guid
;
783 vd
->vdev_guid
+= guid_delta
;
785 for (vdev_t
*pvd
= vd
; pvd
!= NULL
; pvd
= pvd
->vdev_parent
)
786 pvd
->vdev_guid_sum
+= guid_delta
;
789 * If this is a replacement, then we want to fallthrough to the
790 * rest of the code. If we're adding an l2cache, then it's
791 * already labeled appropriately and we can just return.
793 if (reason
== VDEV_LABEL_L2CACHE
)
795 ASSERT(reason
== VDEV_LABEL_REPLACE
);
799 * Initialize its label.
801 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
802 abd_zero(vp_abd
, sizeof (vdev_phys_t
));
803 vp
= abd_to_buf(vp_abd
);
806 * Generate a label describing the pool and our top-level vdev.
807 * We mark it as being from txg 0 to indicate that it's not
808 * really part of an active pool just yet. The labels will
809 * be written again with a meaningful txg by spa_sync().
811 if (reason
== VDEV_LABEL_SPARE
||
812 (reason
== VDEV_LABEL_REMOVE
&& vd
->vdev_isspare
)) {
814 * For inactive hot spares, we generate a special label that
815 * identifies as a mutually shared hot spare. We write the
816 * label if we are adding a hot spare, or if we are removing an
817 * active hot spare (in which case we want to revert the
820 VERIFY(nvlist_alloc(&label
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
822 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_VERSION
,
823 spa_version(spa
)) == 0);
824 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
825 POOL_STATE_SPARE
) == 0);
826 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_GUID
,
827 vd
->vdev_guid
) == 0);
828 } else if (reason
== VDEV_LABEL_L2CACHE
||
829 (reason
== VDEV_LABEL_REMOVE
&& vd
->vdev_isl2cache
)) {
831 * For level 2 ARC devices, add a special label.
833 VERIFY(nvlist_alloc(&label
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
835 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_VERSION
,
836 spa_version(spa
)) == 0);
837 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
838 POOL_STATE_L2CACHE
) == 0);
839 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_GUID
,
840 vd
->vdev_guid
) == 0);
844 if (reason
== VDEV_LABEL_SPLIT
)
845 txg
= spa
->spa_uberblock
.ub_txg
;
846 label
= spa_config_generate(spa
, vd
, txg
, B_FALSE
);
849 * Add our creation time. This allows us to detect multiple
850 * vdev uses as described above, and automatically expires if we
853 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_CREATE_TXG
,
858 buflen
= sizeof (vp
->vp_nvlist
);
860 error
= nvlist_pack(label
, &buf
, &buflen
, NV_ENCODE_XDR
, KM_SLEEP
);
864 /* EFAULT means nvlist_pack ran out of room */
865 return (error
== EFAULT
? ENAMETOOLONG
: EINVAL
);
869 * Initialize uberblock template.
871 ub_abd
= abd_alloc_linear(VDEV_UBERBLOCK_RING
, B_TRUE
);
872 abd_zero(ub_abd
, VDEV_UBERBLOCK_RING
);
873 abd_copy_from_buf(ub_abd
, &spa
->spa_uberblock
, sizeof (uberblock_t
));
874 ub
= abd_to_buf(ub_abd
);
877 /* Initialize the 2nd padding area. */
878 pad2
= abd_alloc_for_io(VDEV_PAD_SIZE
, B_TRUE
);
879 abd_zero(pad2
, VDEV_PAD_SIZE
);
882 * Write everything in parallel.
885 zio
= zio_root(spa
, NULL
, NULL
, flags
);
887 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
889 vdev_label_write(zio
, vd
, l
, vp_abd
,
890 offsetof(vdev_label_t
, vl_vdev_phys
),
891 sizeof (vdev_phys_t
), NULL
, NULL
, flags
);
894 * Skip the 1st padding area.
895 * Zero out the 2nd padding area where it might have
896 * left over data from previous filesystem format.
898 vdev_label_write(zio
, vd
, l
, pad2
,
899 offsetof(vdev_label_t
, vl_pad2
),
900 VDEV_PAD_SIZE
, NULL
, NULL
, flags
);
902 vdev_label_write(zio
, vd
, l
, ub_abd
,
903 offsetof(vdev_label_t
, vl_uberblock
),
904 VDEV_UBERBLOCK_RING
, NULL
, NULL
, flags
);
907 error
= zio_wait(zio
);
909 if (error
!= 0 && !(flags
& ZIO_FLAG_TRYHARD
)) {
910 flags
|= ZIO_FLAG_TRYHARD
;
920 * If this vdev hasn't been previously identified as a spare, then we
921 * mark it as such only if a) we are labeling it as a spare, or b) it
922 * exists as a spare elsewhere in the system. Do the same for
923 * level 2 ARC devices.
925 if (error
== 0 && !vd
->vdev_isspare
&&
926 (reason
== VDEV_LABEL_SPARE
||
927 spa_spare_exists(vd
->vdev_guid
, NULL
, NULL
)))
930 if (error
== 0 && !vd
->vdev_isl2cache
&&
931 (reason
== VDEV_LABEL_L2CACHE
||
932 spa_l2cache_exists(vd
->vdev_guid
, NULL
)))
939 * ==========================================================================
940 * uberblock load/sync
941 * ==========================================================================
945 * Consider the following situation: txg is safely synced to disk. We've
946 * written the first uberblock for txg + 1, and then we lose power. When we
947 * come back up, we fail to see the uberblock for txg + 1 because, say,
948 * it was on a mirrored device and the replica to which we wrote txg + 1
949 * is now offline. If we then make some changes and sync txg + 1, and then
950 * the missing replica comes back, then for a few seconds we'll have two
951 * conflicting uberblocks on disk with the same txg. The solution is simple:
952 * among uberblocks with equal txg, choose the one with the latest timestamp.
955 vdev_uberblock_compare(uberblock_t
*ub1
, uberblock_t
*ub2
)
957 if (ub1
->ub_txg
< ub2
->ub_txg
)
959 if (ub1
->ub_txg
> ub2
->ub_txg
)
962 if (ub1
->ub_timestamp
< ub2
->ub_timestamp
)
964 if (ub1
->ub_timestamp
> ub2
->ub_timestamp
)
971 uberblock_t
*ubl_ubbest
; /* Best uberblock */
972 vdev_t
*ubl_vd
; /* vdev associated with the above */
976 vdev_uberblock_load_done(zio_t
*zio
)
978 vdev_t
*vd
= zio
->io_vd
;
979 spa_t
*spa
= zio
->io_spa
;
980 zio_t
*rio
= zio
->io_private
;
981 uberblock_t
*ub
= abd_to_buf(zio
->io_abd
);
982 struct ubl_cbdata
*cbp
= rio
->io_private
;
984 ASSERT3U(zio
->io_size
, ==, VDEV_UBERBLOCK_SIZE(vd
));
986 if (zio
->io_error
== 0 && uberblock_verify(ub
) == 0) {
987 mutex_enter(&rio
->io_lock
);
988 if (ub
->ub_txg
<= spa
->spa_load_max_txg
&&
989 vdev_uberblock_compare(ub
, cbp
->ubl_ubbest
) > 0) {
991 * Keep track of the vdev in which this uberblock
992 * was found. We will use this information later
993 * to obtain the config nvlist associated with
996 *cbp
->ubl_ubbest
= *ub
;
999 mutex_exit(&rio
->io_lock
);
1002 abd_free(zio
->io_abd
);
1006 vdev_uberblock_load_impl(zio_t
*zio
, vdev_t
*vd
, int flags
,
1007 struct ubl_cbdata
*cbp
)
1009 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1010 vdev_uberblock_load_impl(zio
, vd
->vdev_child
[c
], flags
, cbp
);
1012 if (vd
->vdev_ops
->vdev_op_leaf
&& vdev_readable(vd
)) {
1013 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
1014 for (int n
= 0; n
< VDEV_UBERBLOCK_COUNT(vd
); n
++) {
1015 vdev_label_read(zio
, vd
, l
,
1016 abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd
),
1017 B_TRUE
), VDEV_UBERBLOCK_OFFSET(vd
, n
),
1018 VDEV_UBERBLOCK_SIZE(vd
),
1019 vdev_uberblock_load_done
, zio
, flags
);
1026 * Reads the 'best' uberblock from disk along with its associated
1027 * configuration. First, we read the uberblock array of each label of each
1028 * vdev, keeping track of the uberblock with the highest txg in each array.
1029 * Then, we read the configuration from the same vdev as the best uberblock.
1032 vdev_uberblock_load(vdev_t
*rvd
, uberblock_t
*ub
, nvlist_t
**config
)
1035 spa_t
*spa
= rvd
->vdev_spa
;
1036 struct ubl_cbdata cb
;
1037 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
1038 ZIO_FLAG_SPECULATIVE
| ZIO_FLAG_TRYHARD
;
1043 bzero(ub
, sizeof (uberblock_t
));
1049 spa_config_enter(spa
, SCL_ALL
, FTAG
, RW_WRITER
);
1050 zio
= zio_root(spa
, NULL
, &cb
, flags
);
1051 vdev_uberblock_load_impl(zio
, rvd
, flags
, &cb
);
1052 (void) zio_wait(zio
);
1055 * It's possible that the best uberblock was discovered on a label
1056 * that has a configuration which was written in a future txg.
1057 * Search all labels on this vdev to find the configuration that
1058 * matches the txg for our uberblock.
1060 if (cb
.ubl_vd
!= NULL
) {
1061 vdev_dbgmsg(cb
.ubl_vd
, "best uberblock found for spa %s. "
1062 "txg %llu", spa
->spa_name
, (u_longlong_t
)ub
->ub_txg
);
1064 *config
= vdev_label_read_config(cb
.ubl_vd
, ub
->ub_txg
);
1065 if (*config
== NULL
) {
1066 vdev_dbgmsg(cb
.ubl_vd
, "failed to read label config");
1069 spa_config_exit(spa
, SCL_ALL
, FTAG
);
1073 * On success, increment root zio's count of good writes.
1074 * We only get credit for writes to known-visible vdevs; see spa_vdev_add().
1077 vdev_uberblock_sync_done(zio_t
*zio
)
1079 uint64_t *good_writes
= zio
->io_private
;
1081 if (zio
->io_error
== 0 && zio
->io_vd
->vdev_top
->vdev_ms_array
!= 0)
1082 atomic_inc_64(good_writes
);
1086 * Write the uberblock to all labels of all leaves of the specified vdev.
1089 vdev_uberblock_sync(zio_t
*zio
, uberblock_t
*ub
, vdev_t
*vd
, int flags
)
1091 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1092 vdev_uberblock_sync(zio
, ub
, vd
->vdev_child
[c
], flags
);
1094 if (!vd
->vdev_ops
->vdev_op_leaf
)
1097 if (!vdev_writeable(vd
))
1100 int n
= ub
->ub_txg
& (VDEV_UBERBLOCK_COUNT(vd
) - 1);
1102 /* Copy the uberblock_t into the ABD */
1103 abd_t
*ub_abd
= abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd
), B_TRUE
);
1104 abd_zero(ub_abd
, VDEV_UBERBLOCK_SIZE(vd
));
1105 abd_copy_from_buf(ub_abd
, ub
, sizeof (uberblock_t
));
1107 for (int l
= 0; l
< VDEV_LABELS
; l
++)
1108 vdev_label_write(zio
, vd
, l
, ub_abd
,
1109 VDEV_UBERBLOCK_OFFSET(vd
, n
), VDEV_UBERBLOCK_SIZE(vd
),
1110 vdev_uberblock_sync_done
, zio
->io_private
,
1111 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1116 /* Sync the uberblocks to all vdevs in svd[] */
1118 vdev_uberblock_sync_list(vdev_t
**svd
, int svdcount
, uberblock_t
*ub
, int flags
)
1120 spa_t
*spa
= svd
[0]->vdev_spa
;
1122 uint64_t good_writes
= 0;
1124 zio
= zio_root(spa
, NULL
, &good_writes
, flags
);
1126 for (int v
= 0; v
< svdcount
; v
++)
1127 vdev_uberblock_sync(zio
, ub
, svd
[v
], flags
);
1129 (void) zio_wait(zio
);
1132 * Flush the uberblocks to disk. This ensures that the odd labels
1133 * are no longer needed (because the new uberblocks and the even
1134 * labels are safely on disk), so it is safe to overwrite them.
1136 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1138 for (int v
= 0; v
< svdcount
; v
++) {
1139 if (vdev_writeable(svd
[v
])) {
1140 zio_flush(zio
, svd
[v
]);
1144 (void) zio_wait(zio
);
1146 return (good_writes
>= 1 ? 0 : EIO
);
1150 * On success, increment the count of good writes for our top-level vdev.
1153 vdev_label_sync_done(zio_t
*zio
)
1155 uint64_t *good_writes
= zio
->io_private
;
1157 if (zio
->io_error
== 0)
1158 atomic_inc_64(good_writes
);
1162 * If there weren't enough good writes, indicate failure to the parent.
1165 vdev_label_sync_top_done(zio_t
*zio
)
1167 uint64_t *good_writes
= zio
->io_private
;
1169 if (*good_writes
== 0)
1170 zio
->io_error
= SET_ERROR(EIO
);
1172 kmem_free(good_writes
, sizeof (uint64_t));
1176 * We ignore errors for log and cache devices, simply free the private data.
1179 vdev_label_sync_ignore_done(zio_t
*zio
)
1181 kmem_free(zio
->io_private
, sizeof (uint64_t));
1185 * Write all even or odd labels to all leaves of the specified vdev.
1188 vdev_label_sync(zio_t
*zio
, vdev_t
*vd
, int l
, uint64_t txg
, int flags
)
1196 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1197 vdev_label_sync(zio
, vd
->vdev_child
[c
], l
, txg
, flags
);
1199 if (!vd
->vdev_ops
->vdev_op_leaf
)
1202 if (!vdev_writeable(vd
))
1206 * Generate a label describing the top-level config to which we belong.
1208 label
= spa_config_generate(vd
->vdev_spa
, vd
, txg
, B_FALSE
);
1210 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
1211 abd_zero(vp_abd
, sizeof (vdev_phys_t
));
1212 vp
= abd_to_buf(vp_abd
);
1214 buf
= vp
->vp_nvlist
;
1215 buflen
= sizeof (vp
->vp_nvlist
);
1217 if (nvlist_pack(label
, &buf
, &buflen
, NV_ENCODE_XDR
, KM_SLEEP
) == 0) {
1218 for (; l
< VDEV_LABELS
; l
+= 2) {
1219 vdev_label_write(zio
, vd
, l
, vp_abd
,
1220 offsetof(vdev_label_t
, vl_vdev_phys
),
1221 sizeof (vdev_phys_t
),
1222 vdev_label_sync_done
, zio
->io_private
,
1223 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1232 vdev_label_sync_list(spa_t
*spa
, int l
, uint64_t txg
, int flags
)
1234 list_t
*dl
= &spa
->spa_config_dirty_list
;
1240 * Write the new labels to disk.
1242 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1244 for (vd
= list_head(dl
); vd
!= NULL
; vd
= list_next(dl
, vd
)) {
1245 uint64_t *good_writes
= kmem_zalloc(sizeof (uint64_t),
1248 ASSERT(!vd
->vdev_ishole
);
1250 zio_t
*vio
= zio_null(zio
, spa
, NULL
,
1251 (vd
->vdev_islog
|| vd
->vdev_aux
!= NULL
) ?
1252 vdev_label_sync_ignore_done
: vdev_label_sync_top_done
,
1253 good_writes
, flags
);
1254 vdev_label_sync(vio
, vd
, l
, txg
, flags
);
1258 error
= zio_wait(zio
);
1261 * Flush the new labels to disk.
1263 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1265 for (vd
= list_head(dl
); vd
!= NULL
; vd
= list_next(dl
, vd
))
1268 (void) zio_wait(zio
);
1274 * Sync the uberblock and any changes to the vdev configuration.
1276 * The order of operations is carefully crafted to ensure that
1277 * if the system panics or loses power at any time, the state on disk
1278 * is still transactionally consistent. The in-line comments below
1279 * describe the failure semantics at each stage.
1281 * Moreover, vdev_config_sync() is designed to be idempotent: if it fails
1282 * at any time, you can just call it again, and it will resume its work.
1285 vdev_config_sync(vdev_t
**svd
, int svdcount
, uint64_t txg
)
1287 spa_t
*spa
= svd
[0]->vdev_spa
;
1288 uberblock_t
*ub
= &spa
->spa_uberblock
;
1292 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
;
1296 * Normally, we don't want to try too hard to write every label and
1297 * uberblock. If there is a flaky disk, we don't want the rest of the
1298 * sync process to block while we retry. But if we can't write a
1299 * single label out, we should retry with ZIO_FLAG_TRYHARD before
1300 * bailing out and declaring the pool faulted.
1303 if ((flags
& ZIO_FLAG_TRYHARD
) != 0)
1305 flags
|= ZIO_FLAG_TRYHARD
;
1308 ASSERT(ub
->ub_txg
<= txg
);
1311 * If this isn't a resync due to I/O errors,
1312 * and nothing changed in this transaction group,
1313 * and the vdev configuration hasn't changed,
1314 * then there's nothing to do.
1316 if (ub
->ub_txg
< txg
&&
1317 uberblock_update(ub
, spa
->spa_root_vdev
, txg
) == B_FALSE
&&
1318 list_is_empty(&spa
->spa_config_dirty_list
))
1321 if (txg
> spa_freeze_txg(spa
))
1324 ASSERT(txg
<= spa
->spa_final_txg
);
1327 * Flush the write cache of every disk that's been written to
1328 * in this transaction group. This ensures that all blocks
1329 * written in this txg will be committed to stable storage
1330 * before any uberblock that references them.
1332 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1334 for (vd
= txg_list_head(&spa
->spa_vdev_txg_list
, TXG_CLEAN(txg
)); vd
;
1335 vd
= txg_list_next(&spa
->spa_vdev_txg_list
, vd
, TXG_CLEAN(txg
)))
1338 (void) zio_wait(zio
);
1341 * Sync out the even labels (L0, L2) for every dirty vdev. If the
1342 * system dies in the middle of this process, that's OK: all of the
1343 * even labels that made it to disk will be newer than any uberblock,
1344 * and will therefore be considered invalid. The odd labels (L1, L3),
1345 * which have not yet been touched, will still be valid. We flush
1346 * the new labels to disk to ensure that all even-label updates
1347 * are committed to stable storage before the uberblock update.
1349 if ((error
= vdev_label_sync_list(spa
, 0, txg
, flags
)) != 0)
1353 * Sync the uberblocks to all vdevs in svd[].
1354 * If the system dies in the middle of this step, there are two cases
1355 * to consider, and the on-disk state is consistent either way:
1357 * (1) If none of the new uberblocks made it to disk, then the
1358 * previous uberblock will be the newest, and the odd labels
1359 * (which had not yet been touched) will be valid with respect
1360 * to that uberblock.
1362 * (2) If one or more new uberblocks made it to disk, then they
1363 * will be the newest, and the even labels (which had all
1364 * been successfully committed) will be valid with respect
1365 * to the new uberblocks.
1367 if ((error
= vdev_uberblock_sync_list(svd
, svdcount
, ub
, flags
)) != 0)
1371 * Sync out odd labels for every dirty vdev. If the system dies
1372 * in the middle of this process, the even labels and the new
1373 * uberblocks will suffice to open the pool. The next time
1374 * the pool is opened, the first thing we'll do -- before any
1375 * user data is modified -- is mark every vdev dirty so that
1376 * all labels will be brought up to date. We flush the new labels
1377 * to disk to ensure that all odd-label updates are committed to
1378 * stable storage before the next transaction group begins.
1380 if ((error
= vdev_label_sync_list(spa
, 1, txg
, flags
)) != 0)