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>
147 #include <sys/dsl_scan.h>
149 #include <sys/fs/zfs.h>
152 * Basic routines to read and write from a vdev label.
153 * Used throughout the rest of this file.
156 vdev_label_offset(uint64_t psize
, int l
, uint64_t offset
)
158 ASSERT(offset
< sizeof (vdev_label_t
));
159 ASSERT(P2PHASE_TYPED(psize
, sizeof (vdev_label_t
), uint64_t) == 0);
161 return (offset
+ l
* sizeof (vdev_label_t
) + (l
< VDEV_LABELS
/ 2 ?
162 0 : psize
- VDEV_LABELS
* sizeof (vdev_label_t
)));
166 * Returns back the vdev label associated with the passed in offset.
169 vdev_label_number(uint64_t psize
, uint64_t offset
)
173 if (offset
>= psize
- VDEV_LABEL_END_SIZE
) {
174 offset
-= psize
- VDEV_LABEL_END_SIZE
;
175 offset
+= (VDEV_LABELS
/ 2) * sizeof (vdev_label_t
);
177 l
= offset
/ sizeof (vdev_label_t
);
178 return (l
< VDEV_LABELS
? l
: -1);
182 vdev_label_read(zio_t
*zio
, vdev_t
*vd
, int l
, abd_t
*buf
, uint64_t offset
,
183 uint64_t size
, zio_done_func_t
*done
, void *private, int flags
)
186 spa_config_held(zio
->io_spa
, SCL_STATE
, RW_READER
) == SCL_STATE
||
187 spa_config_held(zio
->io_spa
, SCL_STATE
, RW_WRITER
) == SCL_STATE
);
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
)
201 spa_config_held(zio
->io_spa
, SCL_STATE
, RW_READER
) == SCL_STATE
||
202 spa_config_held(zio
->io_spa
, SCL_STATE
, RW_WRITER
) == SCL_STATE
);
203 ASSERT(flags
& ZIO_FLAG_CONFIG_WRITER
);
205 zio_nowait(zio_write_phys(zio
, vd
,
206 vdev_label_offset(vd
->vdev_psize
, l
, offset
),
207 size
, buf
, ZIO_CHECKSUM_LABEL
, done
, private,
208 ZIO_PRIORITY_SYNC_WRITE
, flags
, B_TRUE
));
212 * Generate the nvlist representing this vdev's stats
215 vdev_config_generate_stats(vdev_t
*vd
, nvlist_t
*nv
)
221 vs
= kmem_alloc(sizeof (*vs
), KM_SLEEP
);
222 vsx
= kmem_alloc(sizeof (*vsx
), KM_SLEEP
);
224 vdev_get_stats_ex(vd
, vs
, vsx
);
225 fnvlist_add_uint64_array(nv
, ZPOOL_CONFIG_VDEV_STATS
,
226 (uint64_t *)vs
, sizeof (*vs
) / sizeof (uint64_t));
228 kmem_free(vs
, sizeof (*vs
));
231 * Add extended stats into a special extended stats nvlist. This keeps
232 * all the extended stats nicely grouped together. The extended stats
233 * nvlist is then added to the main nvlist.
235 nvx
= fnvlist_alloc();
237 /* ZIOs in flight to disk */
238 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE
,
239 vsx
->vsx_active_queue
[ZIO_PRIORITY_SYNC_READ
]);
241 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE
,
242 vsx
->vsx_active_queue
[ZIO_PRIORITY_SYNC_WRITE
]);
244 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE
,
245 vsx
->vsx_active_queue
[ZIO_PRIORITY_ASYNC_READ
]);
247 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE
,
248 vsx
->vsx_active_queue
[ZIO_PRIORITY_ASYNC_WRITE
]);
250 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE
,
251 vsx
->vsx_active_queue
[ZIO_PRIORITY_SCRUB
]);
254 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE
,
255 vsx
->vsx_pend_queue
[ZIO_PRIORITY_SYNC_READ
]);
257 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE
,
258 vsx
->vsx_pend_queue
[ZIO_PRIORITY_SYNC_WRITE
]);
260 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE
,
261 vsx
->vsx_pend_queue
[ZIO_PRIORITY_ASYNC_READ
]);
263 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE
,
264 vsx
->vsx_pend_queue
[ZIO_PRIORITY_ASYNC_WRITE
]);
266 fnvlist_add_uint64(nvx
, ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE
,
267 vsx
->vsx_pend_queue
[ZIO_PRIORITY_SCRUB
]);
270 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO
,
271 vsx
->vsx_total_histo
[ZIO_TYPE_READ
],
272 ARRAY_SIZE(vsx
->vsx_total_histo
[ZIO_TYPE_READ
]));
274 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO
,
275 vsx
->vsx_total_histo
[ZIO_TYPE_WRITE
],
276 ARRAY_SIZE(vsx
->vsx_total_histo
[ZIO_TYPE_WRITE
]));
278 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO
,
279 vsx
->vsx_disk_histo
[ZIO_TYPE_READ
],
280 ARRAY_SIZE(vsx
->vsx_disk_histo
[ZIO_TYPE_READ
]));
282 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO
,
283 vsx
->vsx_disk_histo
[ZIO_TYPE_WRITE
],
284 ARRAY_SIZE(vsx
->vsx_disk_histo
[ZIO_TYPE_WRITE
]));
286 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO
,
287 vsx
->vsx_queue_histo
[ZIO_PRIORITY_SYNC_READ
],
288 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_SYNC_READ
]));
290 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO
,
291 vsx
->vsx_queue_histo
[ZIO_PRIORITY_SYNC_WRITE
],
292 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_SYNC_WRITE
]));
294 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO
,
295 vsx
->vsx_queue_histo
[ZIO_PRIORITY_ASYNC_READ
],
296 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_ASYNC_READ
]));
298 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO
,
299 vsx
->vsx_queue_histo
[ZIO_PRIORITY_ASYNC_WRITE
],
300 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_ASYNC_WRITE
]));
302 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO
,
303 vsx
->vsx_queue_histo
[ZIO_PRIORITY_SCRUB
],
304 ARRAY_SIZE(vsx
->vsx_queue_histo
[ZIO_PRIORITY_SCRUB
]));
307 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO
,
308 vsx
->vsx_ind_histo
[ZIO_PRIORITY_SYNC_READ
],
309 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_SYNC_READ
]));
311 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO
,
312 vsx
->vsx_ind_histo
[ZIO_PRIORITY_SYNC_WRITE
],
313 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_SYNC_WRITE
]));
315 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO
,
316 vsx
->vsx_ind_histo
[ZIO_PRIORITY_ASYNC_READ
],
317 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_ASYNC_READ
]));
319 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO
,
320 vsx
->vsx_ind_histo
[ZIO_PRIORITY_ASYNC_WRITE
],
321 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_ASYNC_WRITE
]));
323 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO
,
324 vsx
->vsx_ind_histo
[ZIO_PRIORITY_SCRUB
],
325 ARRAY_SIZE(vsx
->vsx_ind_histo
[ZIO_PRIORITY_SCRUB
]));
327 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO
,
328 vsx
->vsx_agg_histo
[ZIO_PRIORITY_SYNC_READ
],
329 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_SYNC_READ
]));
331 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO
,
332 vsx
->vsx_agg_histo
[ZIO_PRIORITY_SYNC_WRITE
],
333 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_SYNC_WRITE
]));
335 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO
,
336 vsx
->vsx_agg_histo
[ZIO_PRIORITY_ASYNC_READ
],
337 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_ASYNC_READ
]));
339 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO
,
340 vsx
->vsx_agg_histo
[ZIO_PRIORITY_ASYNC_WRITE
],
341 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_ASYNC_WRITE
]));
343 fnvlist_add_uint64_array(nvx
, ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO
,
344 vsx
->vsx_agg_histo
[ZIO_PRIORITY_SCRUB
],
345 ARRAY_SIZE(vsx
->vsx_agg_histo
[ZIO_PRIORITY_SCRUB
]));
347 /* Add extended stats nvlist to main nvlist */
348 fnvlist_add_nvlist(nv
, ZPOOL_CONFIG_VDEV_STATS_EX
, nvx
);
351 kmem_free(vsx
, sizeof (*vsx
));
355 * Generate the nvlist representing this vdev's config.
358 vdev_config_generate(spa_t
*spa
, vdev_t
*vd
, boolean_t getstats
,
359 vdev_config_flag_t flags
)
362 nv
= fnvlist_alloc();
364 fnvlist_add_string(nv
, ZPOOL_CONFIG_TYPE
, vd
->vdev_ops
->vdev_op_type
);
365 if (!(flags
& (VDEV_CONFIG_SPARE
| VDEV_CONFIG_L2CACHE
)))
366 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ID
, vd
->vdev_id
);
367 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_GUID
, vd
->vdev_guid
);
369 if (vd
->vdev_path
!= NULL
)
370 fnvlist_add_string(nv
, ZPOOL_CONFIG_PATH
, vd
->vdev_path
);
372 if (vd
->vdev_devid
!= NULL
)
373 fnvlist_add_string(nv
, ZPOOL_CONFIG_DEVID
, vd
->vdev_devid
);
375 if (vd
->vdev_physpath
!= NULL
)
376 fnvlist_add_string(nv
, ZPOOL_CONFIG_PHYS_PATH
,
379 if (vd
->vdev_enc_sysfs_path
!= NULL
)
380 fnvlist_add_string(nv
, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH
,
381 vd
->vdev_enc_sysfs_path
);
383 if (vd
->vdev_fru
!= NULL
)
384 fnvlist_add_string(nv
, ZPOOL_CONFIG_FRU
, vd
->vdev_fru
);
386 if (vd
->vdev_nparity
!= 0) {
387 ASSERT(strcmp(vd
->vdev_ops
->vdev_op_type
,
388 VDEV_TYPE_RAIDZ
) == 0);
391 * Make sure someone hasn't managed to sneak a fancy new vdev
392 * into a crufty old storage pool.
394 ASSERT(vd
->vdev_nparity
== 1 ||
395 (vd
->vdev_nparity
<= 2 &&
396 spa_version(spa
) >= SPA_VERSION_RAIDZ2
) ||
397 (vd
->vdev_nparity
<= 3 &&
398 spa_version(spa
) >= SPA_VERSION_RAIDZ3
));
401 * Note that we'll add the nparity tag even on storage pools
402 * that only support a single parity device -- older software
403 * will just ignore it.
405 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_NPARITY
, vd
->vdev_nparity
);
408 if (vd
->vdev_wholedisk
!= -1ULL)
409 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
412 if (vd
->vdev_not_present
)
413 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
, 1);
415 if (vd
->vdev_isspare
)
416 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_SPARE
, 1);
418 if (!(flags
& (VDEV_CONFIG_SPARE
| VDEV_CONFIG_L2CACHE
)) &&
419 vd
== vd
->vdev_top
) {
420 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_METASLAB_ARRAY
,
422 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_METASLAB_SHIFT
,
424 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ASHIFT
, vd
->vdev_ashift
);
425 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ASIZE
,
427 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_LOG
, vd
->vdev_islog
);
428 if (vd
->vdev_removing
)
429 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_REMOVING
,
433 if (vd
->vdev_dtl_sm
!= NULL
) {
434 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_DTL
,
435 space_map_object(vd
->vdev_dtl_sm
));
439 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_CREATE_TXG
, vd
->vdev_crtxg
);
441 if (flags
& VDEV_CONFIG_MOS
) {
442 if (vd
->vdev_leaf_zap
!= 0) {
443 ASSERT(vd
->vdev_ops
->vdev_op_leaf
);
444 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_VDEV_LEAF_ZAP
,
448 if (vd
->vdev_top_zap
!= 0) {
449 ASSERT(vd
== vd
->vdev_top
);
450 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_VDEV_TOP_ZAP
,
458 vdev_config_generate_stats(vd
, nv
);
460 /* provide either current or previous scan information */
461 if (spa_scan_get_stats(spa
, &ps
) == 0) {
462 fnvlist_add_uint64_array(nv
,
463 ZPOOL_CONFIG_SCAN_STATS
, (uint64_t *)&ps
,
464 sizeof (pool_scan_stat_t
) / sizeof (uint64_t));
468 if (!vd
->vdev_ops
->vdev_op_leaf
) {
472 ASSERT(!vd
->vdev_ishole
);
474 child
= kmem_alloc(vd
->vdev_children
* sizeof (nvlist_t
*),
477 for (c
= 0, idx
= 0; c
< vd
->vdev_children
; c
++) {
478 vdev_t
*cvd
= vd
->vdev_child
[c
];
481 * If we're generating an nvlist of removing
482 * vdevs then skip over any device which is
485 if ((flags
& VDEV_CONFIG_REMOVING
) &&
489 child
[idx
++] = vdev_config_generate(spa
, cvd
,
494 fnvlist_add_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
498 for (c
= 0; c
< idx
; c
++)
499 nvlist_free(child
[c
]);
501 kmem_free(child
, vd
->vdev_children
* sizeof (nvlist_t
*));
504 const char *aux
= NULL
;
506 if (vd
->vdev_offline
&& !vd
->vdev_tmpoffline
)
507 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, B_TRUE
);
508 if (vd
->vdev_resilver_txg
!= 0)
509 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_RESILVER_TXG
,
510 vd
->vdev_resilver_txg
);
511 if (vd
->vdev_faulted
)
512 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_FAULTED
, B_TRUE
);
513 if (vd
->vdev_degraded
)
514 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_DEGRADED
, B_TRUE
);
515 if (vd
->vdev_removed
)
516 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_REMOVED
, B_TRUE
);
517 if (vd
->vdev_unspare
)
518 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_UNSPARE
, B_TRUE
);
520 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_IS_HOLE
, B_TRUE
);
522 /* Set the reason why we're FAULTED/DEGRADED. */
523 switch (vd
->vdev_stat
.vs_aux
) {
524 case VDEV_AUX_ERR_EXCEEDED
:
525 aux
= "err_exceeded";
528 case VDEV_AUX_EXTERNAL
:
533 if (aux
!= NULL
&& !vd
->vdev_tmpoffline
) {
534 fnvlist_add_string(nv
, ZPOOL_CONFIG_AUX_STATE
, aux
);
537 * We're healthy - clear any previous AUX_STATE values.
539 if (nvlist_exists(nv
, ZPOOL_CONFIG_AUX_STATE
))
540 nvlist_remove_all(nv
, ZPOOL_CONFIG_AUX_STATE
);
543 if (vd
->vdev_splitting
&& vd
->vdev_orig_guid
!= 0LL) {
544 fnvlist_add_uint64(nv
, ZPOOL_CONFIG_ORIG_GUID
,
553 * Generate a view of the top-level vdevs. If we currently have holes
554 * in the namespace, then generate an array which contains a list of holey
555 * vdevs. Additionally, add the number of top-level children that currently
559 vdev_top_config_generate(spa_t
*spa
, nvlist_t
*config
)
561 vdev_t
*rvd
= spa
->spa_root_vdev
;
565 array
= kmem_alloc(rvd
->vdev_children
* sizeof (uint64_t), KM_SLEEP
);
567 for (c
= 0, idx
= 0; c
< rvd
->vdev_children
; c
++) {
568 vdev_t
*tvd
= rvd
->vdev_child
[c
];
570 if (tvd
->vdev_ishole
)
575 VERIFY(nvlist_add_uint64_array(config
, ZPOOL_CONFIG_HOLE_ARRAY
,
579 VERIFY(nvlist_add_uint64(config
, ZPOOL_CONFIG_VDEV_CHILDREN
,
580 rvd
->vdev_children
) == 0);
582 kmem_free(array
, rvd
->vdev_children
* sizeof (uint64_t));
586 * Returns the configuration from the label of the given vdev. For vdevs
587 * which don't have a txg value stored on their label (i.e. spares/cache)
588 * or have not been completely initialized (txg = 0) just return
589 * the configuration from the first valid label we find. Otherwise,
590 * find the most up-to-date label that does not exceed the specified
594 vdev_label_read_config(vdev_t
*vd
, uint64_t txg
)
596 spa_t
*spa
= vd
->vdev_spa
;
597 nvlist_t
*config
= NULL
;
601 uint64_t best_txg
= 0;
603 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
604 ZIO_FLAG_SPECULATIVE
;
606 ASSERT(spa_config_held(spa
, SCL_STATE_ALL
, RW_WRITER
) == SCL_STATE_ALL
);
608 if (!vdev_readable(vd
))
611 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
612 vp
= abd_to_buf(vp_abd
);
615 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
616 nvlist_t
*label
= NULL
;
618 zio
= zio_root(spa
, NULL
, NULL
, flags
);
620 vdev_label_read(zio
, vd
, l
, vp_abd
,
621 offsetof(vdev_label_t
, vl_vdev_phys
),
622 sizeof (vdev_phys_t
), NULL
, NULL
, flags
);
624 if (zio_wait(zio
) == 0 &&
625 nvlist_unpack(vp
->vp_nvlist
, sizeof (vp
->vp_nvlist
),
627 uint64_t label_txg
= 0;
630 * Auxiliary vdevs won't have txg values in their
631 * labels and newly added vdevs may not have been
632 * completely initialized so just return the
633 * configuration from the first valid label we
636 error
= nvlist_lookup_uint64(label
,
637 ZPOOL_CONFIG_POOL_TXG
, &label_txg
);
638 if ((error
|| label_txg
== 0) && !config
) {
641 } else if (label_txg
<= txg
&& label_txg
> best_txg
) {
642 best_txg
= label_txg
;
644 config
= fnvlist_dup(label
);
654 if (config
== NULL
&& !(flags
& ZIO_FLAG_TRYHARD
)) {
655 flags
|= ZIO_FLAG_TRYHARD
;
665 * Determine if a device is in use. The 'spare_guid' parameter will be filled
666 * in with the device guid if this spare is active elsewhere on the system.
669 vdev_inuse(vdev_t
*vd
, uint64_t crtxg
, vdev_labeltype_t reason
,
670 uint64_t *spare_guid
, uint64_t *l2cache_guid
)
672 spa_t
*spa
= vd
->vdev_spa
;
673 uint64_t state
, pool_guid
, device_guid
, txg
, spare_pool
;
680 *l2cache_guid
= 0ULL;
683 * Read the label, if any, and perform some basic sanity checks.
685 if ((label
= vdev_label_read_config(vd
, -1ULL)) == NULL
)
688 (void) nvlist_lookup_uint64(label
, ZPOOL_CONFIG_CREATE_TXG
,
691 if (nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
693 nvlist_lookup_uint64(label
, ZPOOL_CONFIG_GUID
,
694 &device_guid
) != 0) {
699 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
700 (nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_GUID
,
702 nvlist_lookup_uint64(label
, ZPOOL_CONFIG_POOL_TXG
,
711 * Check to see if this device indeed belongs to the pool it claims to
712 * be a part of. The only way this is allowed is if the device is a hot
713 * spare (which we check for later on).
715 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
716 !spa_guid_exists(pool_guid
, device_guid
) &&
717 !spa_spare_exists(device_guid
, NULL
, NULL
) &&
718 !spa_l2cache_exists(device_guid
, NULL
))
722 * If the transaction group is zero, then this an initialized (but
723 * unused) label. This is only an error if the create transaction
724 * on-disk is the same as the one we're using now, in which case the
725 * user has attempted to add the same vdev multiple times in the same
728 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
729 txg
== 0 && vdtxg
== crtxg
)
733 * Check to see if this is a spare device. We do an explicit check for
734 * spa_has_spare() here because it may be on our pending list of spares
735 * to add. We also check if it is an l2cache device.
737 if (spa_spare_exists(device_guid
, &spare_pool
, NULL
) ||
738 spa_has_spare(spa
, device_guid
)) {
740 *spare_guid
= device_guid
;
743 case VDEV_LABEL_CREATE
:
744 case VDEV_LABEL_L2CACHE
:
747 case VDEV_LABEL_REPLACE
:
748 return (!spa_has_spare(spa
, device_guid
) ||
751 case VDEV_LABEL_SPARE
:
752 return (spa_has_spare(spa
, device_guid
));
759 * Check to see if this is an l2cache device.
761 if (spa_l2cache_exists(device_guid
, NULL
))
765 * We can't rely on a pool's state if it's been imported
766 * read-only. Instead we look to see if the pools is marked
767 * read-only in the namespace and set the state to active.
769 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
770 (spa
= spa_by_guid(pool_guid
, device_guid
)) != NULL
&&
771 spa_mode(spa
) == FREAD
)
772 state
= POOL_STATE_ACTIVE
;
775 * If the device is marked ACTIVE, then this device is in use by another
776 * pool on the system.
778 return (state
== POOL_STATE_ACTIVE
);
782 * Initialize a vdev label. We check to make sure each leaf device is not in
783 * use, and writable. We put down an initial label which we will later
784 * overwrite with a complete label. Note that it's important to do this
785 * sequentially, not in parallel, so that we catch cases of multiple use of the
786 * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with
790 vdev_label_init(vdev_t
*vd
, uint64_t crtxg
, vdev_labeltype_t reason
)
792 spa_t
*spa
= vd
->vdev_spa
;
803 uint64_t spare_guid
= 0, l2cache_guid
= 0;
804 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
;
806 ASSERT(spa_config_held(spa
, SCL_ALL
, RW_WRITER
) == SCL_ALL
);
808 for (int c
= 0; c
< vd
->vdev_children
; c
++)
809 if ((error
= vdev_label_init(vd
->vdev_child
[c
],
810 crtxg
, reason
)) != 0)
813 /* Track the creation time for this vdev */
814 vd
->vdev_crtxg
= crtxg
;
816 if (!vd
->vdev_ops
->vdev_op_leaf
|| !spa_writeable(spa
))
820 * Dead vdevs cannot be initialized.
822 if (vdev_is_dead(vd
))
823 return (SET_ERROR(EIO
));
826 * Determine if the vdev is in use.
828 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_SPLIT
&&
829 vdev_inuse(vd
, crtxg
, reason
, &spare_guid
, &l2cache_guid
))
830 return (SET_ERROR(EBUSY
));
833 * If this is a request to add or replace a spare or l2cache device
834 * that is in use elsewhere on the system, then we must update the
835 * guid (which was initialized to a random value) to reflect the
836 * actual GUID (which is shared between multiple pools).
838 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_L2CACHE
&&
839 spare_guid
!= 0ULL) {
840 uint64_t guid_delta
= spare_guid
- vd
->vdev_guid
;
842 vd
->vdev_guid
+= guid_delta
;
844 for (vdev_t
*pvd
= vd
; pvd
!= NULL
; pvd
= pvd
->vdev_parent
)
845 pvd
->vdev_guid_sum
+= guid_delta
;
848 * If this is a replacement, then we want to fallthrough to the
849 * rest of the code. If we're adding a spare, then it's already
850 * labeled appropriately and we can just return.
852 if (reason
== VDEV_LABEL_SPARE
)
854 ASSERT(reason
== VDEV_LABEL_REPLACE
||
855 reason
== VDEV_LABEL_SPLIT
);
858 if (reason
!= VDEV_LABEL_REMOVE
&& reason
!= VDEV_LABEL_SPARE
&&
859 l2cache_guid
!= 0ULL) {
860 uint64_t guid_delta
= l2cache_guid
- vd
->vdev_guid
;
862 vd
->vdev_guid
+= guid_delta
;
864 for (vdev_t
*pvd
= vd
; pvd
!= NULL
; pvd
= pvd
->vdev_parent
)
865 pvd
->vdev_guid_sum
+= guid_delta
;
868 * If this is a replacement, then we want to fallthrough to the
869 * rest of the code. If we're adding an l2cache, then it's
870 * already labeled appropriately and we can just return.
872 if (reason
== VDEV_LABEL_L2CACHE
)
874 ASSERT(reason
== VDEV_LABEL_REPLACE
);
878 * Initialize its label.
880 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
881 abd_zero(vp_abd
, sizeof (vdev_phys_t
));
882 vp
= abd_to_buf(vp_abd
);
885 * Generate a label describing the pool and our top-level vdev.
886 * We mark it as being from txg 0 to indicate that it's not
887 * really part of an active pool just yet. The labels will
888 * be written again with a meaningful txg by spa_sync().
890 if (reason
== VDEV_LABEL_SPARE
||
891 (reason
== VDEV_LABEL_REMOVE
&& vd
->vdev_isspare
)) {
893 * For inactive hot spares, we generate a special label that
894 * identifies as a mutually shared hot spare. We write the
895 * label if we are adding a hot spare, or if we are removing an
896 * active hot spare (in which case we want to revert the
899 VERIFY(nvlist_alloc(&label
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
901 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_VERSION
,
902 spa_version(spa
)) == 0);
903 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
904 POOL_STATE_SPARE
) == 0);
905 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_GUID
,
906 vd
->vdev_guid
) == 0);
907 } else if (reason
== VDEV_LABEL_L2CACHE
||
908 (reason
== VDEV_LABEL_REMOVE
&& vd
->vdev_isl2cache
)) {
910 * For level 2 ARC devices, add a special label.
912 VERIFY(nvlist_alloc(&label
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
914 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_VERSION
,
915 spa_version(spa
)) == 0);
916 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_POOL_STATE
,
917 POOL_STATE_L2CACHE
) == 0);
918 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_GUID
,
919 vd
->vdev_guid
) == 0);
923 if (reason
== VDEV_LABEL_SPLIT
)
924 txg
= spa
->spa_uberblock
.ub_txg
;
925 label
= spa_config_generate(spa
, vd
, txg
, B_FALSE
);
928 * Add our creation time. This allows us to detect multiple
929 * vdev uses as described above, and automatically expires if we
932 VERIFY(nvlist_add_uint64(label
, ZPOOL_CONFIG_CREATE_TXG
,
937 buflen
= sizeof (vp
->vp_nvlist
);
939 error
= nvlist_pack(label
, &buf
, &buflen
, NV_ENCODE_XDR
, KM_SLEEP
);
943 /* EFAULT means nvlist_pack ran out of room */
944 return (SET_ERROR(error
== EFAULT
? ENAMETOOLONG
: EINVAL
));
948 * Initialize uberblock template.
950 ub_abd
= abd_alloc_linear(VDEV_UBERBLOCK_RING
, B_TRUE
);
951 abd_zero(ub_abd
, VDEV_UBERBLOCK_RING
);
952 abd_copy_from_buf(ub_abd
, &spa
->spa_uberblock
, sizeof (uberblock_t
));
953 ub
= abd_to_buf(ub_abd
);
956 /* Initialize the 2nd padding area. */
957 pad2
= abd_alloc_for_io(VDEV_PAD_SIZE
, B_TRUE
);
958 abd_zero(pad2
, VDEV_PAD_SIZE
);
961 * Write everything in parallel.
964 zio
= zio_root(spa
, NULL
, NULL
, flags
);
966 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
968 vdev_label_write(zio
, vd
, l
, vp_abd
,
969 offsetof(vdev_label_t
, vl_vdev_phys
),
970 sizeof (vdev_phys_t
), NULL
, NULL
, flags
);
973 * Skip the 1st padding area.
974 * Zero out the 2nd padding area where it might have
975 * left over data from previous filesystem format.
977 vdev_label_write(zio
, vd
, l
, pad2
,
978 offsetof(vdev_label_t
, vl_pad2
),
979 VDEV_PAD_SIZE
, NULL
, NULL
, flags
);
981 vdev_label_write(zio
, vd
, l
, ub_abd
,
982 offsetof(vdev_label_t
, vl_uberblock
),
983 VDEV_UBERBLOCK_RING
, NULL
, NULL
, flags
);
986 error
= zio_wait(zio
);
988 if (error
!= 0 && !(flags
& ZIO_FLAG_TRYHARD
)) {
989 flags
|= ZIO_FLAG_TRYHARD
;
999 * If this vdev hasn't been previously identified as a spare, then we
1000 * mark it as such only if a) we are labeling it as a spare, or b) it
1001 * exists as a spare elsewhere in the system. Do the same for
1002 * level 2 ARC devices.
1004 if (error
== 0 && !vd
->vdev_isspare
&&
1005 (reason
== VDEV_LABEL_SPARE
||
1006 spa_spare_exists(vd
->vdev_guid
, NULL
, NULL
)))
1009 if (error
== 0 && !vd
->vdev_isl2cache
&&
1010 (reason
== VDEV_LABEL_L2CACHE
||
1011 spa_l2cache_exists(vd
->vdev_guid
, NULL
)))
1012 spa_l2cache_add(vd
);
1018 * ==========================================================================
1019 * uberblock load/sync
1020 * ==========================================================================
1024 * Consider the following situation: txg is safely synced to disk. We've
1025 * written the first uberblock for txg + 1, and then we lose power. When we
1026 * come back up, we fail to see the uberblock for txg + 1 because, say,
1027 * it was on a mirrored device and the replica to which we wrote txg + 1
1028 * is now offline. If we then make some changes and sync txg + 1, and then
1029 * the missing replica comes back, then for a few seconds we'll have two
1030 * conflicting uberblocks on disk with the same txg. The solution is simple:
1031 * among uberblocks with equal txg, choose the one with the latest timestamp.
1034 vdev_uberblock_compare(const uberblock_t
*ub1
, const uberblock_t
*ub2
)
1036 int cmp
= AVL_CMP(ub1
->ub_txg
, ub2
->ub_txg
);
1040 return (AVL_CMP(ub1
->ub_timestamp
, ub2
->ub_timestamp
));
1044 uberblock_t
*ubl_ubbest
; /* Best uberblock */
1045 vdev_t
*ubl_vd
; /* vdev associated with the above */
1049 vdev_uberblock_load_done(zio_t
*zio
)
1051 vdev_t
*vd
= zio
->io_vd
;
1052 spa_t
*spa
= zio
->io_spa
;
1053 zio_t
*rio
= zio
->io_private
;
1054 uberblock_t
*ub
= abd_to_buf(zio
->io_abd
);
1055 struct ubl_cbdata
*cbp
= rio
->io_private
;
1057 ASSERT3U(zio
->io_size
, ==, VDEV_UBERBLOCK_SIZE(vd
));
1059 if (zio
->io_error
== 0 && uberblock_verify(ub
) == 0) {
1060 mutex_enter(&rio
->io_lock
);
1061 if (ub
->ub_txg
<= spa
->spa_load_max_txg
&&
1062 vdev_uberblock_compare(ub
, cbp
->ubl_ubbest
) > 0) {
1064 * Keep track of the vdev in which this uberblock
1065 * was found. We will use this information later
1066 * to obtain the config nvlist associated with
1069 *cbp
->ubl_ubbest
= *ub
;
1072 mutex_exit(&rio
->io_lock
);
1075 abd_free(zio
->io_abd
);
1079 vdev_uberblock_load_impl(zio_t
*zio
, vdev_t
*vd
, int flags
,
1080 struct ubl_cbdata
*cbp
)
1082 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1083 vdev_uberblock_load_impl(zio
, vd
->vdev_child
[c
], flags
, cbp
);
1085 if (vd
->vdev_ops
->vdev_op_leaf
&& vdev_readable(vd
)) {
1086 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
1087 for (int n
= 0; n
< VDEV_UBERBLOCK_COUNT(vd
); n
++) {
1088 vdev_label_read(zio
, vd
, l
,
1089 abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd
),
1090 B_TRUE
), VDEV_UBERBLOCK_OFFSET(vd
, n
),
1091 VDEV_UBERBLOCK_SIZE(vd
),
1092 vdev_uberblock_load_done
, zio
, flags
);
1099 * Reads the 'best' uberblock from disk along with its associated
1100 * configuration. First, we read the uberblock array of each label of each
1101 * vdev, keeping track of the uberblock with the highest txg in each array.
1102 * Then, we read the configuration from the same vdev as the best uberblock.
1105 vdev_uberblock_load(vdev_t
*rvd
, uberblock_t
*ub
, nvlist_t
**config
)
1108 spa_t
*spa
= rvd
->vdev_spa
;
1109 struct ubl_cbdata cb
;
1110 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
1111 ZIO_FLAG_SPECULATIVE
| ZIO_FLAG_TRYHARD
;
1116 bzero(ub
, sizeof (uberblock_t
));
1122 spa_config_enter(spa
, SCL_ALL
, FTAG
, RW_WRITER
);
1123 zio
= zio_root(spa
, NULL
, &cb
, flags
);
1124 vdev_uberblock_load_impl(zio
, rvd
, flags
, &cb
);
1125 (void) zio_wait(zio
);
1128 * It's possible that the best uberblock was discovered on a label
1129 * that has a configuration which was written in a future txg.
1130 * Search all labels on this vdev to find the configuration that
1131 * matches the txg for our uberblock.
1133 if (cb
.ubl_vd
!= NULL
)
1134 *config
= vdev_label_read_config(cb
.ubl_vd
, ub
->ub_txg
);
1135 spa_config_exit(spa
, SCL_ALL
, FTAG
);
1139 * For use when a leaf vdev is expanded.
1140 * The location of labels 2 and 3 changed, and at the new location the
1141 * uberblock rings are either empty or contain garbage. The sync will write
1142 * new configs there because the vdev is dirty, but expansion also needs the
1143 * uberblock rings copied. Read them from label 0 which did not move.
1145 * Since the point is to populate labels {2,3} with valid uberblocks,
1146 * we zero uberblocks we fail to read or which are not valid.
1150 vdev_copy_uberblocks(vdev_t
*vd
)
1154 int locks
= (SCL_L2ARC
| SCL_ZIO
);
1155 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
|
1156 ZIO_FLAG_SPECULATIVE
;
1158 ASSERT(spa_config_held(vd
->vdev_spa
, SCL_STATE
, RW_READER
) ==
1160 ASSERT(vd
->vdev_ops
->vdev_op_leaf
);
1162 spa_config_enter(vd
->vdev_spa
, locks
, FTAG
, RW_READER
);
1164 ub_abd
= abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd
), B_TRUE
);
1166 write_zio
= zio_root(vd
->vdev_spa
, NULL
, NULL
, flags
);
1167 for (int n
= 0; n
< VDEV_UBERBLOCK_COUNT(vd
); n
++) {
1168 const int src_label
= 0;
1171 zio
= zio_root(vd
->vdev_spa
, NULL
, NULL
, flags
);
1172 vdev_label_read(zio
, vd
, src_label
, ub_abd
,
1173 VDEV_UBERBLOCK_OFFSET(vd
, n
), VDEV_UBERBLOCK_SIZE(vd
),
1176 if (zio_wait(zio
) || uberblock_verify(abd_to_buf(ub_abd
)))
1177 abd_zero(ub_abd
, VDEV_UBERBLOCK_SIZE(vd
));
1179 for (int l
= 2; l
< VDEV_LABELS
; l
++)
1180 vdev_label_write(write_zio
, vd
, l
, ub_abd
,
1181 VDEV_UBERBLOCK_OFFSET(vd
, n
),
1182 VDEV_UBERBLOCK_SIZE(vd
), NULL
, NULL
,
1183 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1185 (void) zio_wait(write_zio
);
1187 spa_config_exit(vd
->vdev_spa
, locks
, FTAG
);
1193 * On success, increment root zio's count of good writes.
1194 * We only get credit for writes to known-visible vdevs; see spa_vdev_add().
1197 vdev_uberblock_sync_done(zio_t
*zio
)
1199 uint64_t *good_writes
= zio
->io_private
;
1201 if (zio
->io_error
== 0 && zio
->io_vd
->vdev_top
->vdev_ms_array
!= 0)
1202 atomic_inc_64(good_writes
);
1206 * Write the uberblock to all labels of all leaves of the specified vdev.
1209 vdev_uberblock_sync(zio_t
*zio
, uberblock_t
*ub
, vdev_t
*vd
, int flags
)
1211 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1212 vdev_uberblock_sync(zio
, ub
, vd
->vdev_child
[c
], flags
);
1214 if (!vd
->vdev_ops
->vdev_op_leaf
)
1217 if (!vdev_writeable(vd
))
1220 /* If the vdev was expanded, need to copy uberblock rings. */
1221 if (vd
->vdev_state
== VDEV_STATE_HEALTHY
&&
1222 vd
->vdev_copy_uberblocks
== B_TRUE
) {
1223 vdev_copy_uberblocks(vd
);
1224 vd
->vdev_copy_uberblocks
= B_FALSE
;
1227 int m
= spa_multihost(vd
->vdev_spa
) ? MMP_BLOCKS_PER_LABEL
: 0;
1228 int n
= ub
->ub_txg
% (VDEV_UBERBLOCK_COUNT(vd
) - m
);
1230 /* Copy the uberblock_t into the ABD */
1231 abd_t
*ub_abd
= abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd
), B_TRUE
);
1232 abd_zero(ub_abd
, VDEV_UBERBLOCK_SIZE(vd
));
1233 abd_copy_from_buf(ub_abd
, ub
, sizeof (uberblock_t
));
1235 for (int l
= 0; l
< VDEV_LABELS
; l
++)
1236 vdev_label_write(zio
, vd
, l
, ub_abd
,
1237 VDEV_UBERBLOCK_OFFSET(vd
, n
), VDEV_UBERBLOCK_SIZE(vd
),
1238 vdev_uberblock_sync_done
, zio
->io_private
,
1239 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1244 /* Sync the uberblocks to all vdevs in svd[] */
1246 vdev_uberblock_sync_list(vdev_t
**svd
, int svdcount
, uberblock_t
*ub
, int flags
)
1248 spa_t
*spa
= svd
[0]->vdev_spa
;
1250 uint64_t good_writes
= 0;
1252 zio
= zio_root(spa
, NULL
, &good_writes
, flags
);
1254 for (int v
= 0; v
< svdcount
; v
++)
1255 vdev_uberblock_sync(zio
, ub
, svd
[v
], flags
);
1257 (void) zio_wait(zio
);
1260 * Flush the uberblocks to disk. This ensures that the odd labels
1261 * are no longer needed (because the new uberblocks and the even
1262 * labels are safely on disk), so it is safe to overwrite them.
1264 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1266 for (int v
= 0; v
< svdcount
; v
++)
1267 zio_flush(zio
, svd
[v
]);
1269 (void) zio_wait(zio
);
1271 return (good_writes
>= 1 ? 0 : EIO
);
1275 * On success, increment the count of good writes for our top-level vdev.
1278 vdev_label_sync_done(zio_t
*zio
)
1280 uint64_t *good_writes
= zio
->io_private
;
1282 if (zio
->io_error
== 0)
1283 atomic_inc_64(good_writes
);
1287 * If there weren't enough good writes, indicate failure to the parent.
1290 vdev_label_sync_top_done(zio_t
*zio
)
1292 uint64_t *good_writes
= zio
->io_private
;
1294 if (*good_writes
== 0)
1295 zio
->io_error
= SET_ERROR(EIO
);
1297 kmem_free(good_writes
, sizeof (uint64_t));
1301 * We ignore errors for log and cache devices, simply free the private data.
1304 vdev_label_sync_ignore_done(zio_t
*zio
)
1306 kmem_free(zio
->io_private
, sizeof (uint64_t));
1310 * Write all even or odd labels to all leaves of the specified vdev.
1313 vdev_label_sync(zio_t
*zio
, vdev_t
*vd
, int l
, uint64_t txg
, int flags
)
1321 for (int c
= 0; c
< vd
->vdev_children
; c
++)
1322 vdev_label_sync(zio
, vd
->vdev_child
[c
], l
, txg
, flags
);
1324 if (!vd
->vdev_ops
->vdev_op_leaf
)
1327 if (!vdev_writeable(vd
))
1331 * Generate a label describing the top-level config to which we belong.
1333 label
= spa_config_generate(vd
->vdev_spa
, vd
, txg
, B_FALSE
);
1335 vp_abd
= abd_alloc_linear(sizeof (vdev_phys_t
), B_TRUE
);
1336 abd_zero(vp_abd
, sizeof (vdev_phys_t
));
1337 vp
= abd_to_buf(vp_abd
);
1339 buf
= vp
->vp_nvlist
;
1340 buflen
= sizeof (vp
->vp_nvlist
);
1342 if (!nvlist_pack(label
, &buf
, &buflen
, NV_ENCODE_XDR
, KM_SLEEP
)) {
1343 for (; l
< VDEV_LABELS
; l
+= 2) {
1344 vdev_label_write(zio
, vd
, l
, vp_abd
,
1345 offsetof(vdev_label_t
, vl_vdev_phys
),
1346 sizeof (vdev_phys_t
),
1347 vdev_label_sync_done
, zio
->io_private
,
1348 flags
| ZIO_FLAG_DONT_PROPAGATE
);
1357 vdev_label_sync_list(spa_t
*spa
, int l
, uint64_t txg
, int flags
)
1359 list_t
*dl
= &spa
->spa_config_dirty_list
;
1365 * Write the new labels to disk.
1367 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1369 for (vd
= list_head(dl
); vd
!= NULL
; vd
= list_next(dl
, vd
)) {
1370 uint64_t *good_writes
;
1372 ASSERT(!vd
->vdev_ishole
);
1374 good_writes
= kmem_zalloc(sizeof (uint64_t), KM_SLEEP
);
1375 zio_t
*vio
= zio_null(zio
, spa
, NULL
,
1376 (vd
->vdev_islog
|| vd
->vdev_aux
!= NULL
) ?
1377 vdev_label_sync_ignore_done
: vdev_label_sync_top_done
,
1378 good_writes
, flags
);
1379 vdev_label_sync(vio
, vd
, l
, txg
, flags
);
1383 error
= zio_wait(zio
);
1386 * Flush the new labels to disk.
1388 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1390 for (vd
= list_head(dl
); vd
!= NULL
; vd
= list_next(dl
, vd
))
1393 (void) zio_wait(zio
);
1399 * Sync the uberblock and any changes to the vdev configuration.
1401 * The order of operations is carefully crafted to ensure that
1402 * if the system panics or loses power at any time, the state on disk
1403 * is still transactionally consistent. The in-line comments below
1404 * describe the failure semantics at each stage.
1406 * Moreover, vdev_config_sync() is designed to be idempotent: if it fails
1407 * at any time, you can just call it again, and it will resume its work.
1410 vdev_config_sync(vdev_t
**svd
, int svdcount
, uint64_t txg
)
1412 spa_t
*spa
= svd
[0]->vdev_spa
;
1413 uberblock_t
*ub
= &spa
->spa_uberblock
;
1417 int flags
= ZIO_FLAG_CONFIG_WRITER
| ZIO_FLAG_CANFAIL
;
1421 * Normally, we don't want to try too hard to write every label and
1422 * uberblock. If there is a flaky disk, we don't want the rest of the
1423 * sync process to block while we retry. But if we can't write a
1424 * single label out, we should retry with ZIO_FLAG_TRYHARD before
1425 * bailing out and declaring the pool faulted.
1428 if ((flags
& ZIO_FLAG_TRYHARD
) != 0)
1430 flags
|= ZIO_FLAG_TRYHARD
;
1433 ASSERT(ub
->ub_txg
<= txg
);
1436 * If this isn't a resync due to I/O errors,
1437 * and nothing changed in this transaction group,
1438 * and the vdev configuration hasn't changed,
1439 * then there's nothing to do.
1441 if (ub
->ub_txg
< txg
) {
1442 boolean_t changed
= uberblock_update(ub
, spa
->spa_root_vdev
,
1443 txg
, spa
->spa_mmp
.mmp_delay
);
1445 if (!changed
&& list_is_empty(&spa
->spa_config_dirty_list
))
1449 if (txg
> spa_freeze_txg(spa
))
1452 ASSERT(txg
<= spa
->spa_final_txg
);
1455 * Flush the write cache of every disk that's been written to
1456 * in this transaction group. This ensures that all blocks
1457 * written in this txg will be committed to stable storage
1458 * before any uberblock that references them.
1460 zio
= zio_root(spa
, NULL
, NULL
, flags
);
1462 for (vd
= txg_list_head(&spa
->spa_vdev_txg_list
, TXG_CLEAN(txg
)); vd
;
1463 vd
= txg_list_next(&spa
->spa_vdev_txg_list
, vd
, TXG_CLEAN(txg
)))
1466 (void) zio_wait(zio
);
1469 * Sync out the even labels (L0, L2) for every dirty vdev. If the
1470 * system dies in the middle of this process, that's OK: all of the
1471 * even labels that made it to disk will be newer than any uberblock,
1472 * and will therefore be considered invalid. The odd labels (L1, L3),
1473 * which have not yet been touched, will still be valid. We flush
1474 * the new labels to disk to ensure that all even-label updates
1475 * are committed to stable storage before the uberblock update.
1477 if ((error
= vdev_label_sync_list(spa
, 0, txg
, flags
)) != 0)
1481 * Sync the uberblocks to all vdevs in svd[].
1482 * If the system dies in the middle of this step, there are two cases
1483 * to consider, and the on-disk state is consistent either way:
1485 * (1) If none of the new uberblocks made it to disk, then the
1486 * previous uberblock will be the newest, and the odd labels
1487 * (which had not yet been touched) will be valid with respect
1488 * to that uberblock.
1490 * (2) If one or more new uberblocks made it to disk, then they
1491 * will be the newest, and the even labels (which had all
1492 * been successfully committed) will be valid with respect
1493 * to the new uberblocks.
1495 if ((error
= vdev_uberblock_sync_list(svd
, svdcount
, ub
, flags
)) != 0)
1499 if (spa_multihost(spa
))
1500 mmp_update_uberblock(spa
, ub
);
1503 * Sync out odd labels for every dirty vdev. If the system dies
1504 * in the middle of this process, the even labels and the new
1505 * uberblocks will suffice to open the pool. The next time
1506 * the pool is opened, the first thing we'll do -- before any
1507 * user data is modified -- is mark every vdev dirty so that
1508 * all labels will be brought up to date. We flush the new labels
1509 * to disk to ensure that all odd-label updates are committed to
1510 * stable storage before the next transaction group begins.
1512 if ((error
= vdev_label_sync_list(spa
, 1, txg
, flags
)) != 0)