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, 2015 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
)
187 spa_config_held(zio
->io_spa
, SCL_STATE
, RW_READER
) == SCL_STATE
||
188 spa_config_held(zio
->io_spa
, SCL_STATE
, RW_WRITER
) == SCL_STATE
);
189 ASSERT(flags
& ZIO_FLAG_CONFIG_WRITER
);
191 zio_nowait(zio_read_phys(zio
, vd
,
192 vdev_label_offset(vd
->vdev_psize
, l
, offset
),
193 size
, buf
, ZIO_CHECKSUM_LABEL
, done
, private,
194 ZIO_PRIORITY_SYNC_READ
, flags
, B_TRUE
));
198 vdev_label_write(zio_t
*zio
, vdev_t
*vd
, int l
, abd_t
*buf
, uint64_t offset
,
199 uint64_t size
, zio_done_func_t
*done
, void *private, int flags
)
202 spa_config_held(zio
->io_spa
, SCL_STATE
, RW_READER
) == SCL_STATE
||
203 spa_config_held(zio
->io_spa
, SCL_STATE
, RW_WRITER
) == SCL_STATE
);
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 stats
216 vdev_config_generate_stats(vdev_t
*vd
, nvlist_t
*nv
)
222 vs
= kmem_alloc(sizeof (*vs
), KM_SLEEP
);
223 vsx
= kmem_alloc(sizeof (*vsx
), KM_SLEEP
);
225 vdev_get_stats_ex(vd
, vs
, vsx
);
226 fnvlist_add_uint64_array(nv
, ZPOOL_CONFIG_VDEV_STATS
,
227 (uint64_t *)vs
, sizeof (*vs
) / sizeof (uint64_t));
229 kmem_free(vs
, sizeof (*vs
));
232 * Add extended stats into a special extended stats nvlist. This keeps
233 * all the extended stats nicely grouped together. The extended stats
234 * nvlist is then added to the main nvlist.
236 nvx
= fnvlist_alloc();
238 /* ZIOs in flight to disk */
239 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE
,
240 vsx
->vsx_active_queue
[ZIO_PRIORITY_SYNC_READ
]);
242 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE
,
243 vsx
->vsx_active_queue
[ZIO_PRIORITY_SYNC_WRITE
]);
245 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE
,
246 vsx
->vsx_active_queue
[ZIO_PRIORITY_ASYNC_READ
]);
248 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE
,
249 vsx
->vsx_active_queue
[ZIO_PRIORITY_ASYNC_WRITE
]);
251 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE
,
252 vsx
->vsx_active_queue
[ZIO_PRIORITY_SCRUB
]);
255 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE
,
256 vsx
->vsx_pend_queue
[ZIO_PRIORITY_SYNC_READ
]);
258 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE
,
259 vsx
->vsx_pend_queue
[ZIO_PRIORITY_SYNC_WRITE
]);
261 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE
,
262 vsx
->vsx_pend_queue
[ZIO_PRIORITY_ASYNC_READ
]);
264 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE
,
265 vsx
->vsx_pend_queue
[ZIO_PRIORITY_ASYNC_WRITE
]);
267 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE
,
268 vsx
->vsx_pend_queue
[ZIO_PRIORITY_SCRUB
]);
271 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO
,
272 vsx
->vsx_total_histo
[ZIO_TYPE_READ
],
273 ARRAY_SIZE(vsx
->vsx_total_histo
[ZIO_TYPE_READ
]));
275 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO
,
276 vsx
->vsx_total_histo
[ZIO_TYPE_WRITE
],
277 ARRAY_SIZE(vsx
->vsx_total_histo
[ZIO_TYPE_WRITE
]));
279 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO
,
280 vsx
->vsx_disk_histo
[ZIO_TYPE_READ
],
281 ARRAY_SIZE(vsx
->vsx_disk_histo
[ZIO_TYPE_READ
]));
283 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO
,
284 vsx
->vsx_disk_histo
[ZIO_TYPE_WRITE
],
285 ARRAY_SIZE(vsx
->vsx_disk_histo
[ZIO_TYPE_WRITE
]));
287 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO
,
288 vsx
->vsx_queue_histo
[ZIO_PRIORITY_SYNC_READ
],
289 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_SYNC_READ
]));
291 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO
,
292 vsx
->vsx_queue_histo
[ZIO_PRIORITY_SYNC_WRITE
],
293 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_SYNC_WRITE
]));
295 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO
,
296 vsx
->vsx_queue_histo
[ZIO_PRIORITY_ASYNC_READ
],
297 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_ASYNC_READ
]));
299 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO
,
300 vsx
->vsx_queue_histo
[ZIO_PRIORITY_ASYNC_WRITE
],
301 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_ASYNC_WRITE
]));
303 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO
,
304 vsx
->vsx_queue_histo
[ZIO_PRIORITY_SCRUB
],
305 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_SCRUB
]));
308 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO
,
309 vsx
->vsx_ind_histo
[ZIO_PRIORITY_SYNC_READ
],
310 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_SYNC_READ
]));
312 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO
,
313 vsx
->vsx_ind_histo
[ZIO_PRIORITY_SYNC_WRITE
],
314 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_SYNC_WRITE
]));
316 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO
,
317 vsx
->vsx_ind_histo
[ZIO_PRIORITY_ASYNC_READ
],
318 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_ASYNC_READ
]));
320 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO
,
321 vsx
->vsx_ind_histo
[ZIO_PRIORITY_ASYNC_WRITE
],
322 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_ASYNC_WRITE
]));
324 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO
,
325 vsx
->vsx_ind_histo
[ZIO_PRIORITY_SCRUB
],
326 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_SCRUB
]));
328 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO
,
329 vsx
->vsx_agg_histo
[ZIO_PRIORITY_SYNC_READ
],
330 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_SYNC_READ
]));
332 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO
,
333 vsx
->vsx_agg_histo
[ZIO_PRIORITY_SYNC_WRITE
],
334 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_SYNC_WRITE
]));
336 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO
,
337 vsx
->vsx_agg_histo
[ZIO_PRIORITY_ASYNC_READ
],
338 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_ASYNC_READ
]));
340 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO
,
341 vsx
->vsx_agg_histo
[ZIO_PRIORITY_ASYNC_WRITE
],
342 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_ASYNC_WRITE
]));
344 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO
,
345 vsx
->vsx_agg_histo
[ZIO_PRIORITY_SCRUB
],
346 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_SCRUB
]));
348 /* Add extended stats nvlist to main nvlist */
349 fnvlist_add_nvlist(nv
, ZPOOL_CONFIG_VDEV_STATS_EX
, nvx
);
352 kmem_free(vsx
, sizeof (*vsx
));
356 * Generate the nvlist representing this vdev's config.
359 vdev_config_generate(spa_t
*spa
, vdev_t
*vd
, boolean_t getstats
,
360 vdev_config_flag_t flags
)
363 vdev_indirect_config_t
*vic
= &vd
->vdev_indirect_config
;
365 nv
= fnvlist_alloc();
367 fnvlist_add_string(nv
, ZPOOL_CONFIG_TYPE
, vd
->vdev_ops
->vdev_op_type
);
368 if (!(flags
& (VDEV_CONFIG_SPARE
| VDEV_CONFIG_L2CACHE
)))
369 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ID
, vd
->vdev_id
);
370 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_GUID
, vd
->vdev_guid
);
372 if (vd
->vdev_path
!= NULL
)
373 fnvlist_add_string(nv
, ZPOOL_CONFIG_PATH
, vd
->vdev_path
);
375 if (vd
->vdev_devid
!= NULL
)
376 fnvlist_add_string(nv
, ZPOOL_CONFIG_DEVID
, vd
->vdev_devid
);
378 if (vd
->vdev_physpath
!= NULL
)
379 fnvlist_add_string(nv
, ZPOOL_CONFIG_PHYS_PATH
,
382 if (vd
->vdev_enc_sysfs_path
!= NULL
)
383 fnvlist_add_string(nv
, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH
,
384 vd
->vdev_enc_sysfs_path
);
386 if (vd
->vdev_fru
!= NULL
)
387 fnvlist_add_string(nv
, ZPOOL_CONFIG_FRU
, vd
->vdev_fru
);
389 if (vd
->vdev_nparity
!= 0) {
390 ASSERT(strcmp(vd
->vdev_ops
->vdev_op_type
,
391 VDEV_TYPE_RAIDZ
) == 0);
394 * Make sure someone hasn't managed to sneak a fancy new vdev
395 * into a crufty old storage pool.
397 ASSERT(vd
->vdev_nparity
== 1 ||
398 (vd
->vdev_nparity
<= 2 &&
399 spa_version(spa
) >= SPA_VERSION_RAIDZ2
) ||
400 (vd
->vdev_nparity
<= 3 &&
401 spa_version(spa
) >= SPA_VERSION_RAIDZ3
));
404 * Note that we'll add the nparity tag even on storage pools
405 * that only support a single parity device -- older software
406 * will just ignore it.
408 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_NPARITY
, vd
->vdev_nparity
);
411 if (vd
->vdev_wholedisk
!= -1ULL)
412 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
415 if (vd
->vdev_not_present
)
416 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
, 1);
418 if (vd
->vdev_isspare
)
419 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_SPARE
, 1);
421 if (!(flags
& (VDEV_CONFIG_SPARE
| VDEV_CONFIG_L2CACHE
)) &&
422 vd
== vd
->vdev_top
) {
423 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_METASLAB_ARRAY
,
425 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_METASLAB_SHIFT
,
427 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ASHIFT
, vd
->vdev_ashift
);
428 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ASIZE
,
430 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_LOG
, vd
->vdev_islog
);
431 if (vd
->vdev_removing
) {
432 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_REMOVING
,
437 if (vd
->vdev_dtl_sm
!= NULL
) {
438 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_DTL
,
439 space_map_object(vd
->vdev_dtl_sm
));
442 if (vic
->vic_mapping_object
!= 0) {
443 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_OBJECT
,
444 vic
->vic_mapping_object
);
447 if (vic
->vic_births_object
!= 0) {
448 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_BIRTHS
,
449 vic
->vic_births_object
);
452 if (vic
->vic_prev_indirect_vdev
!= UINT64_MAX
) {
453 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_PREV_INDIRECT_VDEV
,
454 vic
->vic_prev_indirect_vdev
);
458 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_CREATE_TXG
, vd
->vdev_crtxg
);
460 if (flags
& VDEV_CONFIG_MOS
) {
461 if (vd
->vdev_leaf_zap
!= 0) {
462 ASSERT(vd
->vdev_ops
->vdev_op_leaf
);
463 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_VDEV_LEAF_ZAP
,
467 if (vd
->vdev_top_zap
!= 0) {
468 ASSERT(vd
== vd
->vdev_top
);
469 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_VDEV_TOP_ZAP
,
475 vdev_config_generate_stats(vd
, nv
);
477 /* provide either current or previous scan information */
479 if (spa_scan_get_stats(spa
, &ps
) == 0) {
480 fnvlist_add_uint64_array(nv
,
481 ZPOOL_CONFIG_SCAN_STATS
, (uint64_t *)&ps
,
482 sizeof (pool_scan_stat_t
) / sizeof (uint64_t));
485 pool_removal_stat_t prs
;
486 if (spa_removal_get_stats(spa
, &prs
) == 0) {
487 fnvlist_add_uint64_array(nv
,
488 ZPOOL_CONFIG_REMOVAL_STATS
, (uint64_t *)&prs
,
489 sizeof (prs
) / sizeof (uint64_t));
493 * Note: this can be called from open context
494 * (spa_get_stats()), so we need the rwlock to prevent
495 * the mapping from being changed by condensing.
497 rw_enter(&vd
->vdev_indirect_rwlock
, RW_READER
);
498 if (vd
->vdev_indirect_mapping
!= NULL
) {
499 ASSERT(vd
->vdev_indirect_births
!= NULL
);
500 vdev_indirect_mapping_t
*vim
=
501 vd
->vdev_indirect_mapping
;
502 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_SIZE
,
503 vdev_indirect_mapping_size(vim
));
505 rw_exit(&vd
->vdev_indirect_rwlock
);
506 if (vd
->vdev_mg
!= NULL
&&
507 vd
->vdev_mg
->mg_fragmentation
!= ZFS_FRAG_INVALID
) {
509 * Compute approximately how much memory would be used
510 * for the indirect mapping if this device were to
513 * Note: If the frag metric is invalid, then not
514 * enough metaslabs have been converted to have
517 uint64_t seg_count
= 0;
520 * There are the same number of allocated segments
521 * as free segments, so we will have at least one
522 * entry per free segment.
524 for (int i
= 0; i
< RANGE_TREE_HISTOGRAM_SIZE
; i
++) {
525 seg_count
+= vd
->vdev_mg
->mg_histogram
[i
];
529 * The maximum length of a mapping is SPA_MAXBLOCKSIZE,
530 * so we need at least one entry per SPA_MAXBLOCKSIZE
533 seg_count
+= vd
->vdev_stat
.vs_alloc
/ SPA_MAXBLOCKSIZE
;
535 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_INDIRECT_SIZE
,
537 sizeof (vdev_indirect_mapping_entry_phys_t
));
541 if (!vd
->vdev_ops
->vdev_op_leaf
) {
545 ASSERT(!vd
->vdev_ishole
);
547 child
= kmem_alloc(vd
->vdev_children
* sizeof (nvlist_t
*),
550 for (c
= 0, idx
= 0; c
< vd
->vdev_children
; c
++) {
551 vdev_t
*cvd
= vd
->vdev_child
[c
];
554 * If we're generating an nvlist of removing
555 * vdevs then skip over any device which is
558 if ((flags
& VDEV_CONFIG_REMOVING
) &&
562 child
[idx
++] = vdev_config_generate(spa
, cvd
,
567 fnvlist_add_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
571 for (c
= 0; c
< idx
; c
++)
572 nvlist_free(child
[c
]);
574 kmem_free(child
, vd
->vdev_children
* sizeof (nvlist_t
*));
577 const char *aux
= NULL
;
579 if (vd
->vdev_offline
&& !vd
->vdev_tmpoffline
)
580 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, B_TRUE
);
581 if (vd
->vdev_resilver_txg
!= 0)
582 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_RESILVER_TXG
,
583 vd
->vdev_resilver_txg
);
584 if (vd
->vdev_faulted
)
585 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_FAULTED
, B_TRUE
);
586 if (vd
->vdev_degraded
)
587 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_DEGRADED
, B_TRUE
);
588 if (vd
->vdev_removed
)
589 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_REMOVED
, B_TRUE
);
590 if (vd
->vdev_unspare
)
591 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_UNSPARE
, B_TRUE
);
593 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_HOLE
, B_TRUE
);
595 /* Set the reason why we're FAULTED/DEGRADED. */
596 switch (vd
->vdev_stat
.vs_aux
) {
597 case VDEV_AUX_ERR_EXCEEDED
:
598 aux
= "err_exceeded";
601 case VDEV_AUX_EXTERNAL
:
606 if (aux
!= NULL
&& !vd
->vdev_tmpoffline
) {
607 fnvlist_add_string(nv
, ZPOOL_CONFIG_AUX_STATE
, aux
);
610 * We're healthy - clear any previous AUX_STATE values.
612 if (nvlist_exists(nv
, ZPOOL_CONFIG_AUX_STATE
))
613 nvlist_remove_all(nv
, ZPOOL_CONFIG_AUX_STATE
);
616 if (vd
->vdev_splitting
&& vd
->vdev_orig_guid
!= 0LL) {
617 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ORIG_GUID
,
626 * Generate a view of the top-level vdevs. If we currently have holes
627 * in the namespace, then generate an array which contains a list of holey
628 * vdevs. Additionally, add the number of top-level children that currently
632 vdev_top_config_generate(spa_t
*spa
, nvlist_t
*config
)
634 vdev_t
*rvd
= spa
->spa_root_vdev
;
638 array
= kmem_alloc(rvd
->vdev_children
* sizeof (uint64_t), KM_SLEEP
);
640 for (c
= 0, idx
= 0; c
< rvd
->vdev_children
; c
++) {
641 vdev_t
*tvd
= rvd
->vdev_child
[c
];
643 if (tvd
->vdev_ishole
) {
649 VERIFY(nvlist_add_uint64_array(config
, ZPOOL_CONFIG_HOLE_ARRAY
,
653 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_VDEV_CHILDREN
,
654 rvd
->vdev_children
) == 0);
656 kmem_free(array
, rvd
->vdev_children
* sizeof (uint64_t));
660 * Returns the configuration from the label of the given vdev. For vdevs
661 * which don't have a txg value stored on their label (i.e. spares/cache)
662 * or have not been completely initialized (txg = 0) just return
663 * the configuration from the first valid label we find. Otherwise,
664 * find the most up-to-date label that does not exceed the specified
668 vdev_label_read_config(vdev_t
*vd
, uint64_t txg
)
670 spa_t
*spa
= vd
->vdev_spa
;
671 nvlist_t
*config
= NULL
;
675 uint64_t best_txg
= 0;
677 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
678 ZIO_FLAG_SPECULATIVE
;
680 ASSERT(spa_config_held(spa
, SCL_STATE_ALL
, RW_WRITER
) == SCL_STATE_ALL
);
682 if (!vdev_readable(vd
))
685 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
686 vp
= abd_to_buf(vp_abd
);
689 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
690 nvlist_t
*label
= NULL
;
692 zio
= zio_root(spa
, NULL
, NULL
, flags
);
694 vdev_label_read(zio
, vd
, l
, vp_abd
,
695 offsetof(vdev_label_t
, vl_vdev_phys
),
696 sizeof (vdev_phys_t
), NULL
, NULL
, flags
);
698 if (zio_wait(zio
) == 0 &&
699 nvlist_unpack(vp
->vp_nvlist
, sizeof (vp
->vp_nvlist
),
701 uint64_t label_txg
= 0;
704 * Auxiliary vdevs won't have txg values in their
705 * labels and newly added vdevs may not have been
706 * completely initialized so just return the
707 * configuration from the first valid label we
710 error
= nvlist_lookup_uint64(label
,
711 ZPOOL_CONFIG_POOL_TXG
, &label_txg
);
712 if ((error
|| label_txg
== 0) && !config
) {
715 } else if (label_txg
<= txg
&& label_txg
> best_txg
) {
716 best_txg
= label_txg
;
718 config
= fnvlist_dup(label
);
728 if (config
== NULL
&& !(flags
& ZIO_FLAG_TRYHARD
)) {
729 flags
|= ZIO_FLAG_TRYHARD
;
739 * Determine if a device is in use. The 'spare_guid' parameter will be filled
740 * in with the device guid if this spare is active elsewhere on the system.
743 vdev_inuse(vdev_t
*vd
, uint64_t crtxg
, vdev_labeltype_t reason
,
744 uint64_t *spare_guid
, uint64_t *l2cache_guid
)
746 spa_t
*spa
= vd
->vdev_spa
;
747 uint64_t state
, pool_guid
, device_guid
, txg
, spare_pool
;
754 *l2cache_guid
= 0ULL;
757 * Read the label, if any, and perform some basic sanity checks.
759 if ((label
= vdev_label_read_config(vd
, -1ULL)) == NULL
)
762 (void) nvlist_lookup_uint64(label
, ZPOOL_CONFIG_CREATE_TXG
,
765 if (nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
767 nvlist_lookup_uint64(label
, ZPOOL_CONFIG_GUID
,
768 &device_guid
) != 0) {
773 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
774 (nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_GUID
,
776 nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_TXG
,
785 * Check to see if this device indeed belongs to the pool it claims to
786 * be a part of. The only way this is allowed is if the device is a hot
787 * spare (which we check for later on).
789 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
790 !spa_guid_exists(pool_guid
, device_guid
) &&
791 !spa_spare_exists(device_guid
, NULL
, NULL
) &&
792 !spa_l2cache_exists(device_guid
, NULL
))
796 * If the transaction group is zero, then this an initialized (but
797 * unused) label. This is only an error if the create transaction
798 * on-disk is the same as the one we're using now, in which case the
799 * user has attempted to add the same vdev multiple times in the same
802 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
803 txg
== 0 && vdtxg
== crtxg
)
807 * Check to see if this is a spare device. We do an explicit check for
808 * spa_has_spare() here because it may be on our pending list of spares
809 * to add. We also check if it is an l2cache device.
811 if (spa_spare_exists(device_guid
, &spare_pool
, NULL
) ||
812 spa_has_spare(spa
, device_guid
)) {
814 *spare_guid
= device_guid
;
817 case VDEV_LABEL_CREATE
:
818 case VDEV_LABEL_L2CACHE
:
821 case VDEV_LABEL_REPLACE
:
822 return (!spa_has_spare(spa
, device_guid
) ||
825 case VDEV_LABEL_SPARE
:
826 return (spa_has_spare(spa
, device_guid
));
833 * Check to see if this is an l2cache device.
835 if (spa_l2cache_exists(device_guid
, NULL
))
839 * We can't rely on a pool's state if it's been imported
840 * read-only. Instead we look to see if the pools is marked
841 * read-only in the namespace and set the state to active.
843 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
844 (spa
= spa_by_guid(pool_guid
, device_guid
)) != NULL
&&
845 spa_mode(spa
) == FREAD
)
846 state
= POOL_STATE_ACTIVE
;
849 * If the device is marked ACTIVE, then this device is in use by another
850 * pool on the system.
852 return (state
== POOL_STATE_ACTIVE
);
856 * Initialize a vdev label. We check to make sure each leaf device is not in
857 * use, and writable. We put down an initial label which we will later
858 * overwrite with a complete label. Note that it's important to do this
859 * sequentially, not in parallel, so that we catch cases of multiple use of the
860 * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with
864 vdev_label_init(vdev_t
*vd
, uint64_t crtxg
, vdev_labeltype_t reason
)
866 spa_t
*spa
= vd
->vdev_spa
;
877 uint64_t spare_guid
= 0, l2cache_guid
= 0;
878 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
;
880 ASSERT(spa_config_held(spa
, SCL_ALL
, RW_WRITER
) == SCL_ALL
);
882 for (int c
= 0; c
< vd
->vdev_children
; c
++)
883 if ((error
= vdev_label_init(vd
->vdev_child
[c
],
884 crtxg
, reason
)) != 0)
887 /* Track the creation time for this vdev */
888 vd
->vdev_crtxg
= crtxg
;
890 if (!vd
->vdev_ops
->vdev_op_leaf
|| !spa_writeable(spa
))
894 * Dead vdevs cannot be initialized.
896 if (vdev_is_dead(vd
))
897 return (SET_ERROR(EIO
));
900 * Determine if the vdev is in use.
902 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_SPLIT
&&
903 vdev_inuse(vd
, crtxg
, reason
, &spare_guid
, &l2cache_guid
))
904 return (SET_ERROR(EBUSY
));
907 * If this is a request to add or replace a spare or l2cache device
908 * that is in use elsewhere on the system, then we must update the
909 * guid (which was initialized to a random value) to reflect the
910 * actual GUID (which is shared between multiple pools).
912 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_L2CACHE
&&
913 spare_guid
!= 0ULL) {
914 uint64_t guid_delta
= spare_guid
- vd
->vdev_guid
;
916 vd
->vdev_guid
+= guid_delta
;
918 for (vdev_t
*pvd
= vd
; pvd
!= NULL
; pvd
= pvd
->vdev_parent
)
919 pvd
->vdev_guid_sum
+= guid_delta
;
922 * If this is a replacement, then we want to fallthrough to the
923 * rest of the code. If we're adding a spare, then it's already
924 * labeled appropriately and we can just return.
926 if (reason
== VDEV_LABEL_SPARE
)
928 ASSERT(reason
== VDEV_LABEL_REPLACE
||
929 reason
== VDEV_LABEL_SPLIT
);
932 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_SPARE
&&
933 l2cache_guid
!= 0ULL) {
934 uint64_t guid_delta
= l2cache_guid
- vd
->vdev_guid
;
936 vd
->vdev_guid
+= guid_delta
;
938 for (vdev_t
*pvd
= vd
; pvd
!= NULL
; pvd
= pvd
->vdev_parent
)
939 pvd
->vdev_guid_sum
+= guid_delta
;
942 * If this is a replacement, then we want to fallthrough to the
943 * rest of the code. If we're adding an l2cache, then it's
944 * already labeled appropriately and we can just return.
946 if (reason
== VDEV_LABEL_L2CACHE
)
948 ASSERT(reason
== VDEV_LABEL_REPLACE
);
952 * Initialize its label.
954 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
955 abd_zero(vp_abd
, sizeof (vdev_phys_t
));
956 vp
= abd_to_buf(vp_abd
);
959 * Generate a label describing the pool and our top-level vdev.
960 * We mark it as being from txg 0 to indicate that it's not
961 * really part of an active pool just yet. The labels will
962 * be written again with a meaningful txg by spa_sync().
964 if (reason
== VDEV_LABEL_SPARE
||
965 (reason
== VDEV_LABEL_REMOVE
&& vd
->vdev_isspare
)) {
967 * For inactive hot spares, we generate a special label that
968 * identifies as a mutually shared hot spare. We write the
969 * label if we are adding a hot spare, or if we are removing an
970 * active hot spare (in which case we want to revert the
973 VERIFY(nvlist_alloc(&label
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
975 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_VERSION
,
976 spa_version(spa
)) == 0);
977 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
978 POOL_STATE_SPARE
) == 0);
979 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_GUID
,
980 vd
->vdev_guid
) == 0);
981 } else if (reason
== VDEV_LABEL_L2CACHE
||
982 (reason
== VDEV_LABEL_REMOVE
&& vd
->vdev_isl2cache
)) {
984 * For level 2 ARC devices, add a special label.
986 VERIFY(nvlist_alloc(&label
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
988 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_VERSION
,
989 spa_version(spa
)) == 0);
990 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
991 POOL_STATE_L2CACHE
) == 0);
992 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_GUID
,
993 vd
->vdev_guid
) == 0);
997 if (reason
== VDEV_LABEL_SPLIT
)
998 txg
= spa
->spa_uberblock
.ub_txg
;
999 label
= spa_config_generate(spa
, vd
, txg
, B_FALSE
);
1002 * Add our creation time. This allows us to detect multiple
1003 * vdev uses as described above, and automatically expires if we
1006 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_CREATE_TXG
,
1010 buf
= vp
->vp_nvlist
;
1011 buflen
= sizeof (vp
->vp_nvlist
);
1013 error
= nvlist_pack(label
, &buf
, &buflen
, NV_ENCODE_XDR
, KM_SLEEP
);
1017 /* EFAULT means nvlist_pack ran out of room */
1018 return (SET_ERROR(error
== EFAULT
? ENAMETOOLONG
: EINVAL
));
1022 * Initialize uberblock template.
1024 ub_abd
= abd_alloc_linear(VDEV_UBERBLOCK_RING
, B_TRUE
);
1025 abd_zero(ub_abd
, VDEV_UBERBLOCK_RING
);
1026 abd_copy_from_buf(ub_abd
, &spa
->spa_uberblock
, sizeof (uberblock_t
));
1027 ub
= abd_to_buf(ub_abd
);
1030 /* Initialize the 2nd padding area. */
1031 pad2
= abd_alloc_for_io(VDEV_PAD_SIZE
, B_TRUE
);
1032 abd_zero(pad2
, VDEV_PAD_SIZE
);
1035 * Write everything in parallel.
1038 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1040 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
1042 vdev_label_write(zio
, vd
, l
, vp_abd
,
1043 offsetof(vdev_label_t
, vl_vdev_phys
),
1044 sizeof (vdev_phys_t
), NULL
, NULL
, flags
);
1047 * Skip the 1st padding area.
1048 * Zero out the 2nd padding area where it might have
1049 * left over data from previous filesystem format.
1051 vdev_label_write(zio
, vd
, l
, pad2
,
1052 offsetof(vdev_label_t
, vl_pad2
),
1053 VDEV_PAD_SIZE
, NULL
, NULL
, flags
);
1055 vdev_label_write(zio
, vd
, l
, ub_abd
,
1056 offsetof(vdev_label_t
, vl_uberblock
),
1057 VDEV_UBERBLOCK_RING
, NULL
, NULL
, flags
);
1060 error
= zio_wait(zio
);
1062 if (error
!= 0 && !(flags
& ZIO_FLAG_TRYHARD
)) {
1063 flags
|= ZIO_FLAG_TRYHARD
;
1073 * If this vdev hasn't been previously identified as a spare, then we
1074 * mark it as such only if a) we are labeling it as a spare, or b) it
1075 * exists as a spare elsewhere in the system. Do the same for
1076 * level 2 ARC devices.
1078 if (error
== 0 && !vd
->vdev_isspare
&&
1079 (reason
== VDEV_LABEL_SPARE
||
1080 spa_spare_exists(vd
->vdev_guid
, NULL
, NULL
)))
1083 if (error
== 0 && !vd
->vdev_isl2cache
&&
1084 (reason
== VDEV_LABEL_L2CACHE
||
1085 spa_l2cache_exists(vd
->vdev_guid
, NULL
)))
1086 spa_l2cache_add(vd
);
1092 * ==========================================================================
1093 * uberblock load/sync
1094 * ==========================================================================
1098 * Consider the following situation: txg is safely synced to disk. We've
1099 * written the first uberblock for txg + 1, and then we lose power. When we
1100 * come back up, we fail to see the uberblock for txg + 1 because, say,
1101 * it was on a mirrored device and the replica to which we wrote txg + 1
1102 * is now offline. If we then make some changes and sync txg + 1, and then
1103 * the missing replica comes back, then for a few seconds we'll have two
1104 * conflicting uberblocks on disk with the same txg. The solution is simple:
1105 * among uberblocks with equal txg, choose the one with the latest timestamp.
1108 vdev_uberblock_compare(const uberblock_t
*ub1
, const uberblock_t
*ub2
)
1110 int cmp
= AVL_CMP(ub1
->ub_txg
, ub2
->ub_txg
);
1114 return (AVL_CMP(ub1
->ub_timestamp
, ub2
->ub_timestamp
));
1118 uberblock_t
*ubl_ubbest
; /* Best uberblock */
1119 vdev_t
*ubl_vd
; /* vdev associated with the above */
1123 vdev_uberblock_load_done(zio_t
*zio
)
1125 vdev_t
*vd
= zio
->io_vd
;
1126 spa_t
*spa
= zio
->io_spa
;
1127 zio_t
*rio
= zio
->io_private
;
1128 uberblock_t
*ub
= abd_to_buf(zio
->io_abd
);
1129 struct ubl_cbdata
*cbp
= rio
->io_private
;
1131 ASSERT3U(zio
->io_size
, ==, VDEV_UBERBLOCK_SIZE(vd
));
1133 if (zio
->io_error
== 0 && uberblock_verify(ub
) == 0) {
1134 mutex_enter(&rio
->io_lock
);
1135 if (ub
->ub_txg
<= spa
->spa_load_max_txg
&&
1136 vdev_uberblock_compare(ub
, cbp
->ubl_ubbest
) > 0) {
1138 * Keep track of the vdev in which this uberblock
1139 * was found. We will use this information later
1140 * to obtain the config nvlist associated with
1143 *cbp
->ubl_ubbest
= *ub
;
1146 mutex_exit(&rio
->io_lock
);
1149 abd_free(zio
->io_abd
);
1153 vdev_uberblock_load_impl(zio_t
*zio
, vdev_t
*vd
, int flags
,
1154 struct ubl_cbdata
*cbp
)
1156 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1157 vdev_uberblock_load_impl(zio
, vd
->vdev_child
[c
], flags
, cbp
);
1159 if (vd
->vdev_ops
->vdev_op_leaf
&& vdev_readable(vd
)) {
1160 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
1161 for (int n
= 0; n
< VDEV_UBERBLOCK_COUNT(vd
); n
++) {
1162 vdev_label_read(zio
, vd
, l
,
1163 abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd
),
1164 B_TRUE
), VDEV_UBERBLOCK_OFFSET(vd
, n
),
1165 VDEV_UBERBLOCK_SIZE(vd
),
1166 vdev_uberblock_load_done
, zio
, flags
);
1173 * Reads the 'best' uberblock from disk along with its associated
1174 * configuration. First, we read the uberblock array of each label of each
1175 * vdev, keeping track of the uberblock with the highest txg in each array.
1176 * Then, we read the configuration from the same vdev as the best uberblock.
1179 vdev_uberblock_load(vdev_t
*rvd
, uberblock_t
*ub
, nvlist_t
**config
)
1182 spa_t
*spa
= rvd
->vdev_spa
;
1183 struct ubl_cbdata cb
;
1184 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
1185 ZIO_FLAG_SPECULATIVE
| ZIO_FLAG_TRYHARD
;
1190 bzero(ub
, sizeof (uberblock_t
));
1196 spa_config_enter(spa
, SCL_ALL
, FTAG
, RW_WRITER
);
1197 zio
= zio_root(spa
, NULL
, &cb
, flags
);
1198 vdev_uberblock_load_impl(zio
, rvd
, flags
, &cb
);
1199 (void) zio_wait(zio
);
1202 * It's possible that the best uberblock was discovered on a label
1203 * that has a configuration which was written in a future txg.
1204 * Search all labels on this vdev to find the configuration that
1205 * matches the txg for our uberblock.
1207 if (cb
.ubl_vd
!= NULL
)
1208 *config
= vdev_label_read_config(cb
.ubl_vd
, ub
->ub_txg
);
1209 spa_config_exit(spa
, SCL_ALL
, FTAG
);
1213 * For use when a leaf vdev is expanded.
1214 * The location of labels 2 and 3 changed, and at the new location the
1215 * uberblock rings are either empty or contain garbage. The sync will write
1216 * new configs there because the vdev is dirty, but expansion also needs the
1217 * uberblock rings copied. Read them from label 0 which did not move.
1219 * Since the point is to populate labels {2,3} with valid uberblocks,
1220 * we zero uberblocks we fail to read or which are not valid.
1224 vdev_copy_uberblocks(vdev_t
*vd
)
1228 int locks
= (SCL_L2ARC
| SCL_ZIO
);
1229 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
1230 ZIO_FLAG_SPECULATIVE
;
1232 ASSERT(spa_config_held(vd
->vdev_spa
, SCL_STATE
, RW_READER
) ==
1234 ASSERT(vd
->vdev_ops
->vdev_op_leaf
);
1236 spa_config_enter(vd
->vdev_spa
, locks
, FTAG
, RW_READER
);
1238 ub_abd
= abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd
), B_TRUE
);
1240 write_zio
= zio_root(vd
->vdev_spa
, NULL
, NULL
, flags
);
1241 for (int n
= 0; n
< VDEV_UBERBLOCK_COUNT(vd
); n
++) {
1242 const int src_label
= 0;
1245 zio
= zio_root(vd
->vdev_spa
, NULL
, NULL
, flags
);
1246 vdev_label_read(zio
, vd
, src_label
, ub_abd
,
1247 VDEV_UBERBLOCK_OFFSET(vd
, n
), VDEV_UBERBLOCK_SIZE(vd
),
1250 if (zio_wait(zio
) || uberblock_verify(abd_to_buf(ub_abd
)))
1251 abd_zero(ub_abd
, VDEV_UBERBLOCK_SIZE(vd
));
1253 for (int l
= 2; l
< VDEV_LABELS
; l
++)
1254 vdev_label_write(write_zio
, vd
, l
, ub_abd
,
1255 VDEV_UBERBLOCK_OFFSET(vd
, n
),
1256 VDEV_UBERBLOCK_SIZE(vd
), NULL
, NULL
,
1257 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1259 (void) zio_wait(write_zio
);
1261 spa_config_exit(vd
->vdev_spa
, locks
, FTAG
);
1267 * On success, increment root zio's count of good writes.
1268 * We only get credit for writes to known-visible vdevs; see spa_vdev_add().
1271 vdev_uberblock_sync_done(zio_t
*zio
)
1273 uint64_t *good_writes
= zio
->io_private
;
1275 if (zio
->io_error
== 0 && zio
->io_vd
->vdev_top
->vdev_ms_array
!= 0)
1276 atomic_inc_64(good_writes
);
1280 * Write the uberblock to all labels of all leaves of the specified vdev.
1283 vdev_uberblock_sync(zio_t
*zio
, uberblock_t
*ub
, vdev_t
*vd
, int flags
)
1285 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1286 vdev_uberblock_sync(zio
, ub
, vd
->vdev_child
[c
], flags
);
1288 if (!vd
->vdev_ops
->vdev_op_leaf
)
1291 if (!vdev_writeable(vd
))
1294 /* If the vdev was expanded, need to copy uberblock rings. */
1295 if (vd
->vdev_state
== VDEV_STATE_HEALTHY
&&
1296 vd
->vdev_copy_uberblocks
== B_TRUE
) {
1297 vdev_copy_uberblocks(vd
);
1298 vd
->vdev_copy_uberblocks
= B_FALSE
;
1301 int m
= spa_multihost(vd
->vdev_spa
) ? MMP_BLOCKS_PER_LABEL
: 0;
1302 int n
= ub
->ub_txg
% (VDEV_UBERBLOCK_COUNT(vd
) - m
);
1304 /* Copy the uberblock_t into the ABD */
1305 abd_t
*ub_abd
= abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd
), B_TRUE
);
1306 abd_zero(ub_abd
, VDEV_UBERBLOCK_SIZE(vd
));
1307 abd_copy_from_buf(ub_abd
, ub
, sizeof (uberblock_t
));
1309 for (int l
= 0; l
< VDEV_LABELS
; l
++)
1310 vdev_label_write(zio
, vd
, l
, ub_abd
,
1311 VDEV_UBERBLOCK_OFFSET(vd
, n
), VDEV_UBERBLOCK_SIZE(vd
),
1312 vdev_uberblock_sync_done
, zio
->io_private
,
1313 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1318 /* Sync the uberblocks to all vdevs in svd[] */
1320 vdev_uberblock_sync_list(vdev_t
**svd
, int svdcount
, uberblock_t
*ub
, int flags
)
1322 spa_t
*spa
= svd
[0]->vdev_spa
;
1324 uint64_t good_writes
= 0;
1326 zio
= zio_root(spa
, NULL
, &good_writes
, flags
);
1328 for (int v
= 0; v
< svdcount
; v
++)
1329 vdev_uberblock_sync(zio
, ub
, svd
[v
], flags
);
1331 (void) zio_wait(zio
);
1334 * Flush the uberblocks to disk. This ensures that the odd labels
1335 * are no longer needed (because the new uberblocks and the even
1336 * labels are safely on disk), so it is safe to overwrite them.
1338 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1340 for (int v
= 0; v
< svdcount
; v
++) {
1341 if (vdev_writeable(svd
[v
])) {
1342 zio_flush(zio
, svd
[v
]);
1346 (void) zio_wait(zio
);
1348 return (good_writes
>= 1 ? 0 : EIO
);
1352 * On success, increment the count of good writes for our top-level vdev.
1355 vdev_label_sync_done(zio_t
*zio
)
1357 uint64_t *good_writes
= zio
->io_private
;
1359 if (zio
->io_error
== 0)
1360 atomic_inc_64(good_writes
);
1364 * If there weren't enough good writes, indicate failure to the parent.
1367 vdev_label_sync_top_done(zio_t
*zio
)
1369 uint64_t *good_writes
= zio
->io_private
;
1371 if (*good_writes
== 0)
1372 zio
->io_error
= SET_ERROR(EIO
);
1374 kmem_free(good_writes
, sizeof (uint64_t));
1378 * We ignore errors for log and cache devices, simply free the private data.
1381 vdev_label_sync_ignore_done(zio_t
*zio
)
1383 kmem_free(zio
->io_private
, sizeof (uint64_t));
1387 * Write all even or odd labels to all leaves of the specified vdev.
1390 vdev_label_sync(zio_t
*zio
, vdev_t
*vd
, int l
, uint64_t txg
, int flags
)
1398 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1399 vdev_label_sync(zio
, vd
->vdev_child
[c
], l
, txg
, flags
);
1401 if (!vd
->vdev_ops
->vdev_op_leaf
)
1404 if (!vdev_writeable(vd
))
1408 * Generate a label describing the top-level config to which we belong.
1410 label
= spa_config_generate(vd
->vdev_spa
, vd
, txg
, B_FALSE
);
1412 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
1413 abd_zero(vp_abd
, sizeof (vdev_phys_t
));
1414 vp
= abd_to_buf(vp_abd
);
1416 buf
= vp
->vp_nvlist
;
1417 buflen
= sizeof (vp
->vp_nvlist
);
1419 if (!nvlist_pack(label
, &buf
, &buflen
, NV_ENCODE_XDR
, KM_SLEEP
)) {
1420 for (; l
< VDEV_LABELS
; l
+= 2) {
1421 vdev_label_write(zio
, vd
, l
, vp_abd
,
1422 offsetof(vdev_label_t
, vl_vdev_phys
),
1423 sizeof (vdev_phys_t
),
1424 vdev_label_sync_done
, zio
->io_private
,
1425 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1434 vdev_label_sync_list(spa_t
*spa
, int l
, uint64_t txg
, int flags
)
1436 list_t
*dl
= &spa
->spa_config_dirty_list
;
1442 * Write the new labels to disk.
1444 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1446 for (vd
= list_head(dl
); vd
!= NULL
; vd
= list_next(dl
, vd
)) {
1447 uint64_t *good_writes
;
1449 ASSERT(!vd
->vdev_ishole
);
1451 good_writes
= kmem_zalloc(sizeof (uint64_t), KM_SLEEP
);
1452 zio_t
*vio
= zio_null(zio
, spa
, NULL
,
1453 (vd
->vdev_islog
|| vd
->vdev_aux
!= NULL
) ?
1454 vdev_label_sync_ignore_done
: vdev_label_sync_top_done
,
1455 good_writes
, flags
);
1456 vdev_label_sync(vio
, vd
, l
, txg
, flags
);
1460 error
= zio_wait(zio
);
1463 * Flush the new labels to disk.
1465 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1467 for (vd
= list_head(dl
); vd
!= NULL
; vd
= list_next(dl
, vd
))
1470 (void) zio_wait(zio
);
1476 * Sync the uberblock and any changes to the vdev configuration.
1478 * The order of operations is carefully crafted to ensure that
1479 * if the system panics or loses power at any time, the state on disk
1480 * is still transactionally consistent. The in-line comments below
1481 * describe the failure semantics at each stage.
1483 * Moreover, vdev_config_sync() is designed to be idempotent: if it fails
1484 * at any time, you can just call it again, and it will resume its work.
1487 vdev_config_sync(vdev_t
**svd
, int svdcount
, uint64_t txg
)
1489 spa_t
*spa
= svd
[0]->vdev_spa
;
1490 uberblock_t
*ub
= &spa
->spa_uberblock
;
1494 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
;
1498 * Normally, we don't want to try too hard to write every label and
1499 * uberblock. If there is a flaky disk, we don't want the rest of the
1500 * sync process to block while we retry. But if we can't write a
1501 * single label out, we should retry with ZIO_FLAG_TRYHARD before
1502 * bailing out and declaring the pool faulted.
1505 if ((flags
& ZIO_FLAG_TRYHARD
) != 0)
1507 flags
|= ZIO_FLAG_TRYHARD
;
1510 ASSERT(ub
->ub_txg
<= txg
);
1513 * If this isn't a resync due to I/O errors,
1514 * and nothing changed in this transaction group,
1515 * and the vdev configuration hasn't changed,
1516 * then there's nothing to do.
1518 if (ub
->ub_txg
< txg
) {
1519 boolean_t changed
= uberblock_update(ub
, spa
->spa_root_vdev
,
1520 txg
, spa
->spa_mmp
.mmp_delay
);
1522 if (!changed
&& list_is_empty(&spa
->spa_config_dirty_list
))
1526 if (txg
> spa_freeze_txg(spa
))
1529 ASSERT(txg
<= spa
->spa_final_txg
);
1532 * Flush the write cache of every disk that's been written to
1533 * in this transaction group. This ensures that all blocks
1534 * written in this txg will be committed to stable storage
1535 * before any uberblock that references them.
1537 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1539 for (vd
= txg_list_head(&spa
->spa_vdev_txg_list
, TXG_CLEAN(txg
)); vd
;
1540 vd
= txg_list_next(&spa
->spa_vdev_txg_list
, vd
, TXG_CLEAN(txg
)))
1543 (void) zio_wait(zio
);
1546 * Sync out the even labels (L0, L2) for every dirty vdev. If the
1547 * system dies in the middle of this process, that's OK: all of the
1548 * even labels that made it to disk will be newer than any uberblock,
1549 * and will therefore be considered invalid. The odd labels (L1, L3),
1550 * which have not yet been touched, will still be valid. We flush
1551 * the new labels to disk to ensure that all even-label updates
1552 * are committed to stable storage before the uberblock update.
1554 if ((error
= vdev_label_sync_list(spa
, 0, txg
, flags
)) != 0)
1558 * Sync the uberblocks to all vdevs in svd[].
1559 * If the system dies in the middle of this step, there are two cases
1560 * to consider, and the on-disk state is consistent either way:
1562 * (1) If none of the new uberblocks made it to disk, then the
1563 * previous uberblock will be the newest, and the odd labels
1564 * (which had not yet been touched) will be valid with respect
1565 * to that uberblock.
1567 * (2) If one or more new uberblocks made it to disk, then they
1568 * will be the newest, and the even labels (which had all
1569 * been successfully committed) will be valid with respect
1570 * to the new uberblocks.
1572 if ((error
= vdev_uberblock_sync_list(svd
, svdcount
, ub
, flags
)) != 0)
1575 if (spa_multihost(spa
))
1576 mmp_update_uberblock(spa
, ub
);
1579 * Sync out odd labels for every dirty vdev. If the system dies
1580 * in the middle of this process, the even labels and the new
1581 * uberblocks will suffice to open the pool. The next time
1582 * the pool is opened, the first thing we'll do -- before any
1583 * user data is modified -- is mark every vdev dirty so that
1584 * all labels will be brought up to date. We flush the new labels
1585 * to disk to ensure that all odd-label updates are committed to
1586 * stable storage before the next transaction group begins.
1588 if ((error
= vdev_label_sync_list(spa
, 1, txg
, flags
)) != 0)