]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
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. | |
7 | * | |
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. | |
12 | * | |
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] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
9ae529ec | 21 | |
34dc7c2f | 22 | /* |
428870ff | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
e9aa730c | 24 | * Copyright (c) 2012, 2015 by Delphix. All rights reserved. |
34dc7c2f BB |
25 | */ |
26 | ||
34dc7c2f BB |
27 | /* |
28 | * Virtual Device Labels | |
29 | * --------------------- | |
30 | * | |
31 | * The vdev label serves several distinct purposes: | |
32 | * | |
33 | * 1. Uniquely identify this device as part of a ZFS pool and confirm its | |
34 | * identity within the pool. | |
35 | * | |
36 | * 2. Verify that all the devices given in a configuration are present | |
37 | * within the pool. | |
38 | * | |
39 | * 3. Determine the uberblock for the pool. | |
40 | * | |
41 | * 4. In case of an import operation, determine the configuration of the | |
42 | * toplevel vdev of which it is a part. | |
43 | * | |
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. | |
47 | * | |
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. | |
52 | * | |
53 | * | |
54 | * Label Organization | |
55 | * ------------------ | |
56 | * | |
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. | |
59 | * | |
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: | |
65 | * | |
66 | * L1 UB L2 | |
67 | * +------+ +------+ +------+ | |
68 | * | | | | | | | |
69 | * | t10 | | t10 | | t10 | | |
70 | * | | | | | | | |
71 | * +------+ +------+ +------+ | |
72 | * | |
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. | |
76 | * | |
77 | * In order to identify which labels are valid, the labels are written in the | |
78 | * following manner: | |
79 | * | |
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 | |
83 | * | |
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. | |
90 | * | |
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 | |
95 | * on another vdev. | |
96 | * | |
97 | * | |
98 | * On-disk Format | |
99 | * -------------- | |
100 | * | |
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. | |
104 | * | |
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. | |
108 | * | |
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. | |
113 | * | |
114 | * | |
115 | * Configuration Information | |
116 | * ------------------------- | |
117 | * | |
118 | * The nvlist describing the pool and vdev contains the following elements: | |
119 | * | |
120 | * version ZFS on-disk version | |
121 | * name Pool name | |
122 | * state Pool state | |
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. | |
9ae529ec CS |
126 | * features_for_read |
127 | * An nvlist of the features necessary for reading the MOS. | |
34dc7c2f BB |
128 | * |
129 | * Each leaf device label also contains the following: | |
130 | * | |
131 | * top_guid Unique ID for top-level vdev in which this is contained | |
132 | * guid Unique ID for the leaf vdev | |
133 | * | |
134 | * The 'vs' configuration follows the format described in 'spa_config.c'. | |
135 | */ | |
136 | ||
137 | #include <sys/zfs_context.h> | |
138 | #include <sys/spa.h> | |
139 | #include <sys/spa_impl.h> | |
140 | #include <sys/dmu.h> | |
141 | #include <sys/zap.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/zio.h> | |
428870ff | 147 | #include <sys/dsl_scan.h> |
a6255b7f | 148 | #include <sys/abd.h> |
34dc7c2f BB |
149 | #include <sys/fs/zfs.h> |
150 | ||
151 | /* | |
152 | * Basic routines to read and write from a vdev label. | |
153 | * Used throughout the rest of this file. | |
154 | */ | |
155 | uint64_t | |
156 | vdev_label_offset(uint64_t psize, int l, uint64_t offset) | |
157 | { | |
158 | ASSERT(offset < sizeof (vdev_label_t)); | |
159 | ASSERT(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t) == 0); | |
160 | ||
161 | return (offset + l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? | |
162 | 0 : psize - VDEV_LABELS * sizeof (vdev_label_t))); | |
163 | } | |
164 | ||
b128c09f BB |
165 | /* |
166 | * Returns back the vdev label associated with the passed in offset. | |
167 | */ | |
168 | int | |
169 | vdev_label_number(uint64_t psize, uint64_t offset) | |
170 | { | |
171 | int l; | |
172 | ||
173 | if (offset >= psize - VDEV_LABEL_END_SIZE) { | |
174 | offset -= psize - VDEV_LABEL_END_SIZE; | |
175 | offset += (VDEV_LABELS / 2) * sizeof (vdev_label_t); | |
176 | } | |
177 | l = offset / sizeof (vdev_label_t); | |
178 | return (l < VDEV_LABELS ? l : -1); | |
179 | } | |
180 | ||
34dc7c2f | 181 | static void |
a6255b7f | 182 | vdev_label_read(zio_t *zio, vdev_t *vd, int l, abd_t *buf, uint64_t offset, |
e9aa730c | 183 | uint64_t size, zio_done_func_t *done, void *private, int flags) |
34dc7c2f | 184 | { |
0091d66f OF |
185 | ASSERT( |
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); | |
b128c09f | 188 | ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); |
34dc7c2f BB |
189 | |
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, | |
b128c09f | 193 | ZIO_PRIORITY_SYNC_READ, flags, B_TRUE)); |
34dc7c2f BB |
194 | } |
195 | ||
379ca9cf | 196 | void |
a6255b7f | 197 | vdev_label_write(zio_t *zio, vdev_t *vd, int l, abd_t *buf, uint64_t offset, |
e9aa730c | 198 | uint64_t size, zio_done_func_t *done, void *private, int flags) |
34dc7c2f | 199 | { |
0091d66f OF |
200 | ASSERT( |
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); | |
b128c09f | 203 | ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); |
34dc7c2f BB |
204 | |
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)); | |
209 | } | |
210 | ||
193a37cb TH |
211 | /* |
212 | * Generate the nvlist representing this vdev's stats | |
213 | */ | |
214 | void | |
215 | vdev_config_generate_stats(vdev_t *vd, nvlist_t *nv) | |
216 | { | |
217 | nvlist_t *nvx; | |
218 | vdev_stat_t *vs; | |
219 | vdev_stat_ex_t *vsx; | |
220 | ||
221 | vs = kmem_alloc(sizeof (*vs), KM_SLEEP); | |
222 | vsx = kmem_alloc(sizeof (*vsx), KM_SLEEP); | |
223 | ||
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)); | |
227 | ||
228 | kmem_free(vs, sizeof (*vs)); | |
229 | ||
230 | /* | |
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. | |
234 | */ | |
235 | nvx = fnvlist_alloc(); | |
236 | ||
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]); | |
240 | ||
241 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE, | |
242 | vsx->vsx_active_queue[ZIO_PRIORITY_SYNC_WRITE]); | |
243 | ||
244 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE, | |
245 | vsx->vsx_active_queue[ZIO_PRIORITY_ASYNC_READ]); | |
246 | ||
247 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE, | |
248 | vsx->vsx_active_queue[ZIO_PRIORITY_ASYNC_WRITE]); | |
249 | ||
250 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE, | |
251 | vsx->vsx_active_queue[ZIO_PRIORITY_SCRUB]); | |
252 | ||
253 | /* ZIOs pending */ | |
254 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE, | |
255 | vsx->vsx_pend_queue[ZIO_PRIORITY_SYNC_READ]); | |
256 | ||
257 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE, | |
258 | vsx->vsx_pend_queue[ZIO_PRIORITY_SYNC_WRITE]); | |
259 | ||
260 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE, | |
261 | vsx->vsx_pend_queue[ZIO_PRIORITY_ASYNC_READ]); | |
262 | ||
263 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE, | |
264 | vsx->vsx_pend_queue[ZIO_PRIORITY_ASYNC_WRITE]); | |
265 | ||
266 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE, | |
267 | vsx->vsx_pend_queue[ZIO_PRIORITY_SCRUB]); | |
268 | ||
269 | /* Histograms */ | |
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])); | |
273 | ||
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])); | |
277 | ||
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])); | |
281 | ||
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])); | |
285 | ||
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])); | |
289 | ||
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])); | |
293 | ||
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])); | |
297 | ||
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])); | |
301 | ||
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])); | |
305 | ||
7e945072 TH |
306 | /* Request sizes */ |
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])); | |
310 | ||
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])); | |
314 | ||
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])); | |
318 | ||
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])); | |
322 | ||
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])); | |
326 | ||
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])); | |
330 | ||
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])); | |
334 | ||
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])); | |
338 | ||
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])); | |
342 | ||
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])); | |
346 | ||
193a37cb TH |
347 | /* Add extended stats nvlist to main nvlist */ |
348 | fnvlist_add_nvlist(nv, ZPOOL_CONFIG_VDEV_STATS_EX, nvx); | |
349 | ||
6a796725 | 350 | fnvlist_free(nvx); |
193a37cb TH |
351 | kmem_free(vsx, sizeof (*vsx)); |
352 | } | |
353 | ||
34dc7c2f BB |
354 | /* |
355 | * Generate the nvlist representing this vdev's config. | |
356 | */ | |
357 | nvlist_t * | |
358 | vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats, | |
428870ff | 359 | vdev_config_flag_t flags) |
34dc7c2f BB |
360 | { |
361 | nvlist_t *nv = NULL; | |
79c76d5b | 362 | nv = fnvlist_alloc(); |
34dc7c2f | 363 | |
5d1f7fb6 | 364 | fnvlist_add_string(nv, ZPOOL_CONFIG_TYPE, vd->vdev_ops->vdev_op_type); |
428870ff | 365 | if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE))) |
5d1f7fb6 GW |
366 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ID, vd->vdev_id); |
367 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_GUID, vd->vdev_guid); | |
34dc7c2f BB |
368 | |
369 | if (vd->vdev_path != NULL) | |
5d1f7fb6 | 370 | fnvlist_add_string(nv, ZPOOL_CONFIG_PATH, vd->vdev_path); |
34dc7c2f BB |
371 | |
372 | if (vd->vdev_devid != NULL) | |
5d1f7fb6 | 373 | fnvlist_add_string(nv, ZPOOL_CONFIG_DEVID, vd->vdev_devid); |
34dc7c2f BB |
374 | |
375 | if (vd->vdev_physpath != NULL) | |
5d1f7fb6 GW |
376 | fnvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH, |
377 | vd->vdev_physpath); | |
34dc7c2f | 378 | |
1bbd8770 TH |
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); | |
382 | ||
9babb374 | 383 | if (vd->vdev_fru != NULL) |
5d1f7fb6 | 384 | fnvlist_add_string(nv, ZPOOL_CONFIG_FRU, vd->vdev_fru); |
9babb374 | 385 | |
34dc7c2f BB |
386 | if (vd->vdev_nparity != 0) { |
387 | ASSERT(strcmp(vd->vdev_ops->vdev_op_type, | |
388 | VDEV_TYPE_RAIDZ) == 0); | |
389 | ||
390 | /* | |
391 | * Make sure someone hasn't managed to sneak a fancy new vdev | |
392 | * into a crufty old storage pool. | |
393 | */ | |
394 | ASSERT(vd->vdev_nparity == 1 || | |
45d1cae3 BB |
395 | (vd->vdev_nparity <= 2 && |
396 | spa_version(spa) >= SPA_VERSION_RAIDZ2) || | |
397 | (vd->vdev_nparity <= 3 && | |
398 | spa_version(spa) >= SPA_VERSION_RAIDZ3)); | |
34dc7c2f BB |
399 | |
400 | /* | |
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. | |
404 | */ | |
5d1f7fb6 | 405 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY, vd->vdev_nparity); |
34dc7c2f BB |
406 | } |
407 | ||
408 | if (vd->vdev_wholedisk != -1ULL) | |
5d1f7fb6 GW |
409 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, |
410 | vd->vdev_wholedisk); | |
34dc7c2f BB |
411 | |
412 | if (vd->vdev_not_present) | |
5d1f7fb6 | 413 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1); |
34dc7c2f BB |
414 | |
415 | if (vd->vdev_isspare) | |
5d1f7fb6 | 416 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1); |
34dc7c2f | 417 | |
428870ff BB |
418 | if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)) && |
419 | vd == vd->vdev_top) { | |
5d1f7fb6 GW |
420 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY, |
421 | vd->vdev_ms_array); | |
422 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT, | |
423 | vd->vdev_ms_shift); | |
424 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT, vd->vdev_ashift); | |
425 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASIZE, | |
426 | vd->vdev_asize); | |
427 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_LOG, vd->vdev_islog); | |
428870ff | 428 | if (vd->vdev_removing) |
5d1f7fb6 GW |
429 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVING, |
430 | vd->vdev_removing); | |
34dc7c2f BB |
431 | } |
432 | ||
93cf2076 | 433 | if (vd->vdev_dtl_sm != NULL) { |
5d1f7fb6 | 434 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_DTL, |
93cf2076 GW |
435 | space_map_object(vd->vdev_dtl_sm)); |
436 | } | |
34dc7c2f | 437 | |
428870ff | 438 | if (vd->vdev_crtxg) |
5d1f7fb6 | 439 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG, vd->vdev_crtxg); |
428870ff | 440 | |
e0ab3ab5 JS |
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, | |
445 | vd->vdev_leaf_zap); | |
446 | } | |
447 | ||
448 | if (vd->vdev_top_zap != 0) { | |
449 | ASSERT(vd == vd->vdev_top); | |
450 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP, | |
451 | vd->vdev_top_zap); | |
452 | } | |
453 | } | |
454 | ||
34dc7c2f | 455 | if (getstats) { |
428870ff BB |
456 | pool_scan_stat_t ps; |
457 | ||
193a37cb | 458 | vdev_config_generate_stats(vd, nv); |
428870ff BB |
459 | |
460 | /* provide either current or previous scan information */ | |
461 | if (spa_scan_get_stats(spa, &ps) == 0) { | |
5d1f7fb6 | 462 | fnvlist_add_uint64_array(nv, |
428870ff | 463 | ZPOOL_CONFIG_SCAN_STATS, (uint64_t *)&ps, |
5d1f7fb6 | 464 | sizeof (pool_scan_stat_t) / sizeof (uint64_t)); |
428870ff | 465 | } |
34dc7c2f BB |
466 | } |
467 | ||
468 | if (!vd->vdev_ops->vdev_op_leaf) { | |
469 | nvlist_t **child; | |
428870ff BB |
470 | int c, idx; |
471 | ||
472 | ASSERT(!vd->vdev_ishole); | |
34dc7c2f BB |
473 | |
474 | child = kmem_alloc(vd->vdev_children * sizeof (nvlist_t *), | |
79c76d5b | 475 | KM_SLEEP); |
34dc7c2f | 476 | |
428870ff BB |
477 | for (c = 0, idx = 0; c < vd->vdev_children; c++) { |
478 | vdev_t *cvd = vd->vdev_child[c]; | |
34dc7c2f | 479 | |
428870ff BB |
480 | /* |
481 | * If we're generating an nvlist of removing | |
482 | * vdevs then skip over any device which is | |
483 | * not being removed. | |
484 | */ | |
485 | if ((flags & VDEV_CONFIG_REMOVING) && | |
486 | !cvd->vdev_removing) | |
487 | continue; | |
34dc7c2f | 488 | |
428870ff BB |
489 | child[idx++] = vdev_config_generate(spa, cvd, |
490 | getstats, flags); | |
491 | } | |
492 | ||
493 | if (idx) { | |
5d1f7fb6 GW |
494 | fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, |
495 | child, idx); | |
428870ff BB |
496 | } |
497 | ||
498 | for (c = 0; c < idx; c++) | |
34dc7c2f BB |
499 | nvlist_free(child[c]); |
500 | ||
501 | kmem_free(child, vd->vdev_children * sizeof (nvlist_t *)); | |
502 | ||
503 | } else { | |
428870ff BB |
504 | const char *aux = NULL; |
505 | ||
34dc7c2f | 506 | if (vd->vdev_offline && !vd->vdev_tmpoffline) |
5d1f7fb6 GW |
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); | |
34dc7c2f | 511 | if (vd->vdev_faulted) |
5d1f7fb6 | 512 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_FAULTED, B_TRUE); |
34dc7c2f | 513 | if (vd->vdev_degraded) |
5d1f7fb6 | 514 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_DEGRADED, B_TRUE); |
34dc7c2f | 515 | if (vd->vdev_removed) |
5d1f7fb6 | 516 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVED, B_TRUE); |
34dc7c2f | 517 | if (vd->vdev_unspare) |
5d1f7fb6 | 518 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_UNSPARE, B_TRUE); |
428870ff | 519 | if (vd->vdev_ishole) |
5d1f7fb6 | 520 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_HOLE, B_TRUE); |
428870ff | 521 | |
4a283c7f | 522 | /* Set the reason why we're FAULTED/DEGRADED. */ |
428870ff BB |
523 | switch (vd->vdev_stat.vs_aux) { |
524 | case VDEV_AUX_ERR_EXCEEDED: | |
525 | aux = "err_exceeded"; | |
526 | break; | |
527 | ||
528 | case VDEV_AUX_EXTERNAL: | |
529 | aux = "external"; | |
530 | break; | |
531 | } | |
532 | ||
4a283c7f | 533 | if (aux != NULL && !vd->vdev_tmpoffline) { |
5d1f7fb6 | 534 | fnvlist_add_string(nv, ZPOOL_CONFIG_AUX_STATE, aux); |
4a283c7f TH |
535 | } else { |
536 | /* | |
537 | * We're healthy - clear any previous AUX_STATE values. | |
538 | */ | |
539 | if (nvlist_exists(nv, ZPOOL_CONFIG_AUX_STATE)) | |
540 | nvlist_remove_all(nv, ZPOOL_CONFIG_AUX_STATE); | |
541 | } | |
428870ff BB |
542 | |
543 | if (vd->vdev_splitting && vd->vdev_orig_guid != 0LL) { | |
5d1f7fb6 GW |
544 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ORIG_GUID, |
545 | vd->vdev_orig_guid); | |
428870ff | 546 | } |
34dc7c2f BB |
547 | } |
548 | ||
549 | return (nv); | |
550 | } | |
551 | ||
428870ff BB |
552 | /* |
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 | |
556 | * exist. | |
557 | */ | |
558 | void | |
559 | vdev_top_config_generate(spa_t *spa, nvlist_t *config) | |
560 | { | |
561 | vdev_t *rvd = spa->spa_root_vdev; | |
562 | uint64_t *array; | |
563 | uint_t c, idx; | |
564 | ||
79c76d5b | 565 | array = kmem_alloc(rvd->vdev_children * sizeof (uint64_t), KM_SLEEP); |
428870ff BB |
566 | |
567 | for (c = 0, idx = 0; c < rvd->vdev_children; c++) { | |
568 | vdev_t *tvd = rvd->vdev_child[c]; | |
569 | ||
570 | if (tvd->vdev_ishole) | |
571 | array[idx++] = c; | |
572 | } | |
573 | ||
574 | if (idx) { | |
575 | VERIFY(nvlist_add_uint64_array(config, ZPOOL_CONFIG_HOLE_ARRAY, | |
576 | array, idx) == 0); | |
577 | } | |
578 | ||
579 | VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN, | |
580 | rvd->vdev_children) == 0); | |
581 | ||
582 | kmem_free(array, rvd->vdev_children * sizeof (uint64_t)); | |
583 | } | |
584 | ||
9ae529ec | 585 | /* |
3bc7e0fb GW |
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 | |
591 | * 'txg' value. | |
9ae529ec | 592 | */ |
34dc7c2f | 593 | nvlist_t * |
3bc7e0fb | 594 | vdev_label_read_config(vdev_t *vd, uint64_t txg) |
34dc7c2f BB |
595 | { |
596 | spa_t *spa = vd->vdev_spa; | |
597 | nvlist_t *config = NULL; | |
598 | vdev_phys_t *vp; | |
a6255b7f | 599 | abd_t *vp_abd; |
34dc7c2f | 600 | zio_t *zio; |
3bc7e0fb GW |
601 | uint64_t best_txg = 0; |
602 | int error = 0; | |
9babb374 BB |
603 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | |
604 | ZIO_FLAG_SPECULATIVE; | |
d6320ddb | 605 | int l; |
34dc7c2f | 606 | |
b128c09f | 607 | ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL); |
34dc7c2f BB |
608 | |
609 | if (!vdev_readable(vd)) | |
610 | return (NULL); | |
611 | ||
a6255b7f DQ |
612 | vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE); |
613 | vp = abd_to_buf(vp_abd); | |
34dc7c2f | 614 | |
9babb374 | 615 | retry: |
d6320ddb | 616 | for (l = 0; l < VDEV_LABELS; l++) { |
3bc7e0fb | 617 | nvlist_t *label = NULL; |
34dc7c2f | 618 | |
b128c09f | 619 | zio = zio_root(spa, NULL, NULL, flags); |
34dc7c2f | 620 | |
a6255b7f | 621 | vdev_label_read(zio, vd, l, vp_abd, |
34dc7c2f | 622 | offsetof(vdev_label_t, vl_vdev_phys), |
b128c09f | 623 | sizeof (vdev_phys_t), NULL, NULL, flags); |
34dc7c2f BB |
624 | |
625 | if (zio_wait(zio) == 0 && | |
626 | nvlist_unpack(vp->vp_nvlist, sizeof (vp->vp_nvlist), | |
3bc7e0fb GW |
627 | &label, 0) == 0) { |
628 | uint64_t label_txg = 0; | |
629 | ||
630 | /* | |
631 | * Auxiliary vdevs won't have txg values in their | |
632 | * labels and newly added vdevs may not have been | |
633 | * completely initialized so just return the | |
634 | * configuration from the first valid label we | |
635 | * encounter. | |
636 | */ | |
637 | error = nvlist_lookup_uint64(label, | |
638 | ZPOOL_CONFIG_POOL_TXG, &label_txg); | |
639 | if ((error || label_txg == 0) && !config) { | |
640 | config = label; | |
641 | break; | |
642 | } else if (label_txg <= txg && label_txg > best_txg) { | |
643 | best_txg = label_txg; | |
644 | nvlist_free(config); | |
645 | config = fnvlist_dup(label); | |
646 | } | |
647 | } | |
34dc7c2f | 648 | |
3bc7e0fb GW |
649 | if (label != NULL) { |
650 | nvlist_free(label); | |
651 | label = NULL; | |
34dc7c2f BB |
652 | } |
653 | } | |
654 | ||
9babb374 BB |
655 | if (config == NULL && !(flags & ZIO_FLAG_TRYHARD)) { |
656 | flags |= ZIO_FLAG_TRYHARD; | |
657 | goto retry; | |
658 | } | |
659 | ||
a6255b7f | 660 | abd_free(vp_abd); |
34dc7c2f BB |
661 | |
662 | return (config); | |
663 | } | |
664 | ||
665 | /* | |
666 | * Determine if a device is in use. The 'spare_guid' parameter will be filled | |
667 | * in with the device guid if this spare is active elsewhere on the system. | |
668 | */ | |
669 | static boolean_t | |
670 | vdev_inuse(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason, | |
671 | uint64_t *spare_guid, uint64_t *l2cache_guid) | |
672 | { | |
673 | spa_t *spa = vd->vdev_spa; | |
674 | uint64_t state, pool_guid, device_guid, txg, spare_pool; | |
675 | uint64_t vdtxg = 0; | |
676 | nvlist_t *label; | |
677 | ||
678 | if (spare_guid) | |
679 | *spare_guid = 0ULL; | |
680 | if (l2cache_guid) | |
681 | *l2cache_guid = 0ULL; | |
682 | ||
683 | /* | |
684 | * Read the label, if any, and perform some basic sanity checks. | |
685 | */ | |
3bc7e0fb | 686 | if ((label = vdev_label_read_config(vd, -1ULL)) == NULL) |
34dc7c2f BB |
687 | return (B_FALSE); |
688 | ||
689 | (void) nvlist_lookup_uint64(label, ZPOOL_CONFIG_CREATE_TXG, | |
690 | &vdtxg); | |
691 | ||
692 | if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, | |
693 | &state) != 0 || | |
694 | nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, | |
695 | &device_guid) != 0) { | |
696 | nvlist_free(label); | |
697 | return (B_FALSE); | |
698 | } | |
699 | ||
700 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
701 | (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, | |
702 | &pool_guid) != 0 || | |
703 | nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG, | |
704 | &txg) != 0)) { | |
705 | nvlist_free(label); | |
706 | return (B_FALSE); | |
707 | } | |
708 | ||
709 | nvlist_free(label); | |
710 | ||
711 | /* | |
712 | * Check to see if this device indeed belongs to the pool it claims to | |
713 | * be a part of. The only way this is allowed is if the device is a hot | |
714 | * spare (which we check for later on). | |
715 | */ | |
716 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
717 | !spa_guid_exists(pool_guid, device_guid) && | |
b128c09f | 718 | !spa_spare_exists(device_guid, NULL, NULL) && |
34dc7c2f BB |
719 | !spa_l2cache_exists(device_guid, NULL)) |
720 | return (B_FALSE); | |
721 | ||
722 | /* | |
723 | * If the transaction group is zero, then this an initialized (but | |
724 | * unused) label. This is only an error if the create transaction | |
725 | * on-disk is the same as the one we're using now, in which case the | |
726 | * user has attempted to add the same vdev multiple times in the same | |
727 | * transaction. | |
728 | */ | |
729 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
730 | txg == 0 && vdtxg == crtxg) | |
731 | return (B_TRUE); | |
732 | ||
733 | /* | |
734 | * Check to see if this is a spare device. We do an explicit check for | |
735 | * spa_has_spare() here because it may be on our pending list of spares | |
736 | * to add. We also check if it is an l2cache device. | |
737 | */ | |
b128c09f | 738 | if (spa_spare_exists(device_guid, &spare_pool, NULL) || |
34dc7c2f BB |
739 | spa_has_spare(spa, device_guid)) { |
740 | if (spare_guid) | |
741 | *spare_guid = device_guid; | |
742 | ||
743 | switch (reason) { | |
744 | case VDEV_LABEL_CREATE: | |
745 | case VDEV_LABEL_L2CACHE: | |
746 | return (B_TRUE); | |
747 | ||
748 | case VDEV_LABEL_REPLACE: | |
749 | return (!spa_has_spare(spa, device_guid) || | |
750 | spare_pool != 0ULL); | |
751 | ||
752 | case VDEV_LABEL_SPARE: | |
753 | return (spa_has_spare(spa, device_guid)); | |
e75c13c3 BB |
754 | default: |
755 | break; | |
34dc7c2f BB |
756 | } |
757 | } | |
758 | ||
759 | /* | |
760 | * Check to see if this is an l2cache device. | |
761 | */ | |
762 | if (spa_l2cache_exists(device_guid, NULL)) | |
763 | return (B_TRUE); | |
764 | ||
572e2857 BB |
765 | /* |
766 | * We can't rely on a pool's state if it's been imported | |
767 | * read-only. Instead we look to see if the pools is marked | |
768 | * read-only in the namespace and set the state to active. | |
769 | */ | |
485c581c RY |
770 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && |
771 | (spa = spa_by_guid(pool_guid, device_guid)) != NULL && | |
572e2857 BB |
772 | spa_mode(spa) == FREAD) |
773 | state = POOL_STATE_ACTIVE; | |
774 | ||
34dc7c2f BB |
775 | /* |
776 | * If the device is marked ACTIVE, then this device is in use by another | |
777 | * pool on the system. | |
778 | */ | |
779 | return (state == POOL_STATE_ACTIVE); | |
780 | } | |
781 | ||
782 | /* | |
783 | * Initialize a vdev label. We check to make sure each leaf device is not in | |
784 | * use, and writable. We put down an initial label which we will later | |
785 | * overwrite with a complete label. Note that it's important to do this | |
786 | * sequentially, not in parallel, so that we catch cases of multiple use of the | |
787 | * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with | |
788 | * itself. | |
789 | */ | |
790 | int | |
791 | vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason) | |
792 | { | |
793 | spa_t *spa = vd->vdev_spa; | |
794 | nvlist_t *label; | |
795 | vdev_phys_t *vp; | |
a6255b7f DQ |
796 | abd_t *vp_abd; |
797 | abd_t *pad2; | |
34dc7c2f | 798 | uberblock_t *ub; |
a6255b7f | 799 | abd_t *ub_abd; |
34dc7c2f | 800 | zio_t *zio; |
34dc7c2f BB |
801 | char *buf; |
802 | size_t buflen; | |
803 | int error; | |
d4ed6673 | 804 | uint64_t spare_guid = 0, l2cache_guid = 0; |
b128c09f | 805 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; |
d6320ddb BB |
806 | int c, l; |
807 | vdev_t *pvd; | |
34dc7c2f | 808 | |
b128c09f | 809 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); |
34dc7c2f | 810 | |
d6320ddb | 811 | for (c = 0; c < vd->vdev_children; c++) |
34dc7c2f BB |
812 | if ((error = vdev_label_init(vd->vdev_child[c], |
813 | crtxg, reason)) != 0) | |
814 | return (error); | |
815 | ||
428870ff BB |
816 | /* Track the creation time for this vdev */ |
817 | vd->vdev_crtxg = crtxg; | |
818 | ||
dda12da9 | 819 | if (!vd->vdev_ops->vdev_op_leaf || !spa_writeable(spa)) |
34dc7c2f BB |
820 | return (0); |
821 | ||
822 | /* | |
823 | * Dead vdevs cannot be initialized. | |
824 | */ | |
825 | if (vdev_is_dead(vd)) | |
2e528b49 | 826 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
827 | |
828 | /* | |
829 | * Determine if the vdev is in use. | |
830 | */ | |
428870ff | 831 | if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPLIT && |
34dc7c2f | 832 | vdev_inuse(vd, crtxg, reason, &spare_guid, &l2cache_guid)) |
2e528b49 | 833 | return (SET_ERROR(EBUSY)); |
34dc7c2f | 834 | |
34dc7c2f BB |
835 | /* |
836 | * If this is a request to add or replace a spare or l2cache device | |
837 | * that is in use elsewhere on the system, then we must update the | |
838 | * guid (which was initialized to a random value) to reflect the | |
839 | * actual GUID (which is shared between multiple pools). | |
840 | */ | |
841 | if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_L2CACHE && | |
842 | spare_guid != 0ULL) { | |
b128c09f | 843 | uint64_t guid_delta = spare_guid - vd->vdev_guid; |
34dc7c2f | 844 | |
b128c09f | 845 | vd->vdev_guid += guid_delta; |
34dc7c2f | 846 | |
d6320ddb | 847 | for (pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) |
b128c09f | 848 | pvd->vdev_guid_sum += guid_delta; |
34dc7c2f BB |
849 | |
850 | /* | |
851 | * If this is a replacement, then we want to fallthrough to the | |
852 | * rest of the code. If we're adding a spare, then it's already | |
853 | * labeled appropriately and we can just return. | |
854 | */ | |
855 | if (reason == VDEV_LABEL_SPARE) | |
856 | return (0); | |
428870ff BB |
857 | ASSERT(reason == VDEV_LABEL_REPLACE || |
858 | reason == VDEV_LABEL_SPLIT); | |
34dc7c2f BB |
859 | } |
860 | ||
861 | if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPARE && | |
862 | l2cache_guid != 0ULL) { | |
b128c09f | 863 | uint64_t guid_delta = l2cache_guid - vd->vdev_guid; |
34dc7c2f | 864 | |
b128c09f | 865 | vd->vdev_guid += guid_delta; |
34dc7c2f | 866 | |
d6320ddb | 867 | for (pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) |
b128c09f | 868 | pvd->vdev_guid_sum += guid_delta; |
34dc7c2f BB |
869 | |
870 | /* | |
871 | * If this is a replacement, then we want to fallthrough to the | |
872 | * rest of the code. If we're adding an l2cache, then it's | |
873 | * already labeled appropriately and we can just return. | |
874 | */ | |
875 | if (reason == VDEV_LABEL_L2CACHE) | |
876 | return (0); | |
877 | ASSERT(reason == VDEV_LABEL_REPLACE); | |
878 | } | |
879 | ||
880 | /* | |
881 | * Initialize its label. | |
882 | */ | |
a6255b7f DQ |
883 | vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE); |
884 | abd_zero(vp_abd, sizeof (vdev_phys_t)); | |
885 | vp = abd_to_buf(vp_abd); | |
34dc7c2f BB |
886 | |
887 | /* | |
888 | * Generate a label describing the pool and our top-level vdev. | |
889 | * We mark it as being from txg 0 to indicate that it's not | |
890 | * really part of an active pool just yet. The labels will | |
891 | * be written again with a meaningful txg by spa_sync(). | |
892 | */ | |
893 | if (reason == VDEV_LABEL_SPARE || | |
894 | (reason == VDEV_LABEL_REMOVE && vd->vdev_isspare)) { | |
895 | /* | |
896 | * For inactive hot spares, we generate a special label that | |
897 | * identifies as a mutually shared hot spare. We write the | |
898 | * label if we are adding a hot spare, or if we are removing an | |
899 | * active hot spare (in which case we want to revert the | |
900 | * labels). | |
901 | */ | |
79c76d5b | 902 | VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); |
34dc7c2f BB |
903 | |
904 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, | |
905 | spa_version(spa)) == 0); | |
906 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, | |
907 | POOL_STATE_SPARE) == 0); | |
908 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, | |
909 | vd->vdev_guid) == 0); | |
910 | } else if (reason == VDEV_LABEL_L2CACHE || | |
911 | (reason == VDEV_LABEL_REMOVE && vd->vdev_isl2cache)) { | |
912 | /* | |
913 | * For level 2 ARC devices, add a special label. | |
914 | */ | |
79c76d5b | 915 | VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); |
34dc7c2f BB |
916 | |
917 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, | |
918 | spa_version(spa)) == 0); | |
919 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, | |
920 | POOL_STATE_L2CACHE) == 0); | |
921 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, | |
922 | vd->vdev_guid) == 0); | |
923 | } else { | |
428870ff BB |
924 | uint64_t txg = 0ULL; |
925 | ||
926 | if (reason == VDEV_LABEL_SPLIT) | |
927 | txg = spa->spa_uberblock.ub_txg; | |
928 | label = spa_config_generate(spa, vd, txg, B_FALSE); | |
34dc7c2f BB |
929 | |
930 | /* | |
931 | * Add our creation time. This allows us to detect multiple | |
932 | * vdev uses as described above, and automatically expires if we | |
933 | * fail. | |
934 | */ | |
935 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG, | |
936 | crtxg) == 0); | |
937 | } | |
938 | ||
939 | buf = vp->vp_nvlist; | |
940 | buflen = sizeof (vp->vp_nvlist); | |
941 | ||
79c76d5b | 942 | error = nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP); |
34dc7c2f BB |
943 | if (error != 0) { |
944 | nvlist_free(label); | |
a6255b7f | 945 | abd_free(vp_abd); |
34dc7c2f | 946 | /* EFAULT means nvlist_pack ran out of room */ |
ecb2b7dc | 947 | return (SET_ERROR(error == EFAULT ? ENAMETOOLONG : EINVAL)); |
34dc7c2f BB |
948 | } |
949 | ||
34dc7c2f BB |
950 | /* |
951 | * Initialize uberblock template. | |
952 | */ | |
a6255b7f DQ |
953 | ub_abd = abd_alloc_linear(VDEV_UBERBLOCK_RING, B_TRUE); |
954 | abd_zero(ub_abd, VDEV_UBERBLOCK_RING); | |
955 | abd_copy_from_buf(ub_abd, &spa->spa_uberblock, sizeof (uberblock_t)); | |
956 | ub = abd_to_buf(ub_abd); | |
34dc7c2f BB |
957 | ub->ub_txg = 0; |
958 | ||
9babb374 | 959 | /* Initialize the 2nd padding area. */ |
a6255b7f DQ |
960 | pad2 = abd_alloc_for_io(VDEV_PAD_SIZE, B_TRUE); |
961 | abd_zero(pad2, VDEV_PAD_SIZE); | |
9babb374 | 962 | |
34dc7c2f BB |
963 | /* |
964 | * Write everything in parallel. | |
965 | */ | |
9babb374 | 966 | retry: |
34dc7c2f BB |
967 | zio = zio_root(spa, NULL, NULL, flags); |
968 | ||
d6320ddb | 969 | for (l = 0; l < VDEV_LABELS; l++) { |
34dc7c2f | 970 | |
a6255b7f | 971 | vdev_label_write(zio, vd, l, vp_abd, |
34dc7c2f BB |
972 | offsetof(vdev_label_t, vl_vdev_phys), |
973 | sizeof (vdev_phys_t), NULL, NULL, flags); | |
974 | ||
9babb374 BB |
975 | /* |
976 | * Skip the 1st padding area. | |
977 | * Zero out the 2nd padding area where it might have | |
978 | * left over data from previous filesystem format. | |
979 | */ | |
980 | vdev_label_write(zio, vd, l, pad2, | |
981 | offsetof(vdev_label_t, vl_pad2), | |
982 | VDEV_PAD_SIZE, NULL, NULL, flags); | |
34dc7c2f | 983 | |
a6255b7f | 984 | vdev_label_write(zio, vd, l, ub_abd, |
45d1cae3 BB |
985 | offsetof(vdev_label_t, vl_uberblock), |
986 | VDEV_UBERBLOCK_RING, NULL, NULL, flags); | |
34dc7c2f BB |
987 | } |
988 | ||
989 | error = zio_wait(zio); | |
990 | ||
9babb374 BB |
991 | if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) { |
992 | flags |= ZIO_FLAG_TRYHARD; | |
993 | goto retry; | |
994 | } | |
995 | ||
34dc7c2f | 996 | nvlist_free(label); |
a6255b7f DQ |
997 | abd_free(pad2); |
998 | abd_free(ub_abd); | |
999 | abd_free(vp_abd); | |
34dc7c2f BB |
1000 | |
1001 | /* | |
1002 | * If this vdev hasn't been previously identified as a spare, then we | |
1003 | * mark it as such only if a) we are labeling it as a spare, or b) it | |
1004 | * exists as a spare elsewhere in the system. Do the same for | |
1005 | * level 2 ARC devices. | |
1006 | */ | |
1007 | if (error == 0 && !vd->vdev_isspare && | |
1008 | (reason == VDEV_LABEL_SPARE || | |
b128c09f | 1009 | spa_spare_exists(vd->vdev_guid, NULL, NULL))) |
34dc7c2f BB |
1010 | spa_spare_add(vd); |
1011 | ||
1012 | if (error == 0 && !vd->vdev_isl2cache && | |
1013 | (reason == VDEV_LABEL_L2CACHE || | |
1014 | spa_l2cache_exists(vd->vdev_guid, NULL))) | |
1015 | spa_l2cache_add(vd); | |
1016 | ||
1017 | return (error); | |
1018 | } | |
1019 | ||
1020 | /* | |
1021 | * ========================================================================== | |
1022 | * uberblock load/sync | |
1023 | * ========================================================================== | |
1024 | */ | |
1025 | ||
1026 | /* | |
1027 | * Consider the following situation: txg is safely synced to disk. We've | |
1028 | * written the first uberblock for txg + 1, and then we lose power. When we | |
1029 | * come back up, we fail to see the uberblock for txg + 1 because, say, | |
1030 | * it was on a mirrored device and the replica to which we wrote txg + 1 | |
1031 | * is now offline. If we then make some changes and sync txg + 1, and then | |
9ae529ec | 1032 | * the missing replica comes back, then for a few seconds we'll have two |
34dc7c2f BB |
1033 | * conflicting uberblocks on disk with the same txg. The solution is simple: |
1034 | * among uberblocks with equal txg, choose the one with the latest timestamp. | |
1035 | */ | |
1036 | static int | |
ee36c709 | 1037 | vdev_uberblock_compare(const uberblock_t *ub1, const uberblock_t *ub2) |
34dc7c2f | 1038 | { |
ee36c709 GN |
1039 | int cmp = AVL_CMP(ub1->ub_txg, ub2->ub_txg); |
1040 | if (likely(cmp)) | |
1041 | return (cmp); | |
34dc7c2f | 1042 | |
ee36c709 | 1043 | return (AVL_CMP(ub1->ub_timestamp, ub2->ub_timestamp)); |
34dc7c2f BB |
1044 | } |
1045 | ||
9ae529ec CS |
1046 | struct ubl_cbdata { |
1047 | uberblock_t *ubl_ubbest; /* Best uberblock */ | |
1048 | vdev_t *ubl_vd; /* vdev associated with the above */ | |
9ae529ec CS |
1049 | }; |
1050 | ||
34dc7c2f BB |
1051 | static void |
1052 | vdev_uberblock_load_done(zio_t *zio) | |
1053 | { | |
9ae529ec | 1054 | vdev_t *vd = zio->io_vd; |
428870ff | 1055 | spa_t *spa = zio->io_spa; |
b128c09f | 1056 | zio_t *rio = zio->io_private; |
a6255b7f | 1057 | uberblock_t *ub = abd_to_buf(zio->io_abd); |
9ae529ec | 1058 | struct ubl_cbdata *cbp = rio->io_private; |
34dc7c2f | 1059 | |
9ae529ec | 1060 | ASSERT3U(zio->io_size, ==, VDEV_UBERBLOCK_SIZE(vd)); |
34dc7c2f BB |
1061 | |
1062 | if (zio->io_error == 0 && uberblock_verify(ub) == 0) { | |
b128c09f | 1063 | mutex_enter(&rio->io_lock); |
428870ff | 1064 | if (ub->ub_txg <= spa->spa_load_max_txg && |
9ae529ec CS |
1065 | vdev_uberblock_compare(ub, cbp->ubl_ubbest) > 0) { |
1066 | /* | |
3bc7e0fb GW |
1067 | * Keep track of the vdev in which this uberblock |
1068 | * was found. We will use this information later | |
1069 | * to obtain the config nvlist associated with | |
9ae529ec CS |
1070 | * this uberblock. |
1071 | */ | |
1072 | *cbp->ubl_ubbest = *ub; | |
1073 | cbp->ubl_vd = vd; | |
9ae529ec | 1074 | } |
b128c09f | 1075 | mutex_exit(&rio->io_lock); |
34dc7c2f BB |
1076 | } |
1077 | ||
a6255b7f | 1078 | abd_free(zio->io_abd); |
34dc7c2f BB |
1079 | } |
1080 | ||
9ae529ec CS |
1081 | static void |
1082 | vdev_uberblock_load_impl(zio_t *zio, vdev_t *vd, int flags, | |
1083 | struct ubl_cbdata *cbp) | |
34dc7c2f | 1084 | { |
379ca9cf | 1085 | for (int c = 0; c < vd->vdev_children; c++) |
9ae529ec | 1086 | vdev_uberblock_load_impl(zio, vd->vdev_child[c], flags, cbp); |
34dc7c2f | 1087 | |
b128c09f | 1088 | if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) { |
379ca9cf OF |
1089 | for (int l = 0; l < VDEV_LABELS; l++) { |
1090 | for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) { | |
b128c09f | 1091 | vdev_label_read(zio, vd, l, |
a6255b7f DQ |
1092 | abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd), |
1093 | B_TRUE), VDEV_UBERBLOCK_OFFSET(vd, n), | |
b128c09f BB |
1094 | VDEV_UBERBLOCK_SIZE(vd), |
1095 | vdev_uberblock_load_done, zio, flags); | |
1096 | } | |
34dc7c2f BB |
1097 | } |
1098 | } | |
9ae529ec CS |
1099 | } |
1100 | ||
1101 | /* | |
1102 | * Reads the 'best' uberblock from disk along with its associated | |
1103 | * configuration. First, we read the uberblock array of each label of each | |
1104 | * vdev, keeping track of the uberblock with the highest txg in each array. | |
3bc7e0fb | 1105 | * Then, we read the configuration from the same vdev as the best uberblock. |
9ae529ec CS |
1106 | */ |
1107 | void | |
1108 | vdev_uberblock_load(vdev_t *rvd, uberblock_t *ub, nvlist_t **config) | |
1109 | { | |
9ae529ec CS |
1110 | zio_t *zio; |
1111 | spa_t *spa = rvd->vdev_spa; | |
1112 | struct ubl_cbdata cb; | |
1113 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | | |
1114 | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD; | |
1115 | ||
1116 | ASSERT(ub); | |
1117 | ASSERT(config); | |
b128c09f | 1118 | |
9ae529ec CS |
1119 | bzero(ub, sizeof (uberblock_t)); |
1120 | *config = NULL; | |
1121 | ||
1122 | cb.ubl_ubbest = ub; | |
1123 | cb.ubl_vd = NULL; | |
1124 | ||
1125 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
1126 | zio = zio_root(spa, NULL, &cb, flags); | |
1127 | vdev_uberblock_load_impl(zio, rvd, flags, &cb); | |
1128 | (void) zio_wait(zio); | |
3bc7e0fb GW |
1129 | |
1130 | /* | |
1131 | * It's possible that the best uberblock was discovered on a label | |
1132 | * that has a configuration which was written in a future txg. | |
1133 | * Search all labels on this vdev to find the configuration that | |
1134 | * matches the txg for our uberblock. | |
1135 | */ | |
1136 | if (cb.ubl_vd != NULL) | |
1137 | *config = vdev_label_read_config(cb.ubl_vd, ub->ub_txg); | |
9ae529ec | 1138 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1139 | } |
1140 | ||
9d3f7b87 OF |
1141 | /* |
1142 | * For use when a leaf vdev is expanded. | |
1143 | * The location of labels 2 and 3 changed, and at the new location the | |
1144 | * uberblock rings are either empty or contain garbage. The sync will write | |
1145 | * new configs there because the vdev is dirty, but expansion also needs the | |
1146 | * uberblock rings copied. Read them from label 0 which did not move. | |
1147 | * | |
1148 | * Since the point is to populate labels {2,3} with valid uberblocks, | |
1149 | * we zero uberblocks we fail to read or which are not valid. | |
1150 | */ | |
1151 | ||
1152 | static void | |
1153 | vdev_copy_uberblocks(vdev_t *vd) | |
1154 | { | |
1155 | abd_t *ub_abd; | |
1156 | zio_t *write_zio; | |
1157 | int locks = (SCL_L2ARC | SCL_ZIO); | |
1158 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | | |
1159 | ZIO_FLAG_SPECULATIVE; | |
1160 | ||
1161 | ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_READER) == | |
1162 | SCL_STATE); | |
1163 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
1164 | ||
1165 | spa_config_enter(vd->vdev_spa, locks, FTAG, RW_READER); | |
1166 | ||
1167 | ub_abd = abd_alloc(VDEV_UBERBLOCK_SIZE(vd), B_TRUE); | |
1168 | ||
1169 | write_zio = zio_root(vd->vdev_spa, NULL, NULL, flags); | |
1170 | for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) { | |
1171 | const int src_label = 0; | |
1172 | zio_t *zio; | |
1173 | ||
1174 | zio = zio_root(vd->vdev_spa, NULL, NULL, flags); | |
1175 | vdev_label_read(zio, vd, src_label, ub_abd, | |
1176 | VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd), | |
1177 | NULL, NULL, flags); | |
1178 | ||
1179 | if (zio_wait(zio) || uberblock_verify(abd_to_buf(ub_abd))) | |
1180 | abd_zero(ub_abd, VDEV_UBERBLOCK_SIZE(vd)); | |
1181 | ||
1182 | for (int l = 2; l < VDEV_LABELS; l++) | |
1183 | vdev_label_write(write_zio, vd, l, ub_abd, | |
1184 | VDEV_UBERBLOCK_OFFSET(vd, n), | |
1185 | VDEV_UBERBLOCK_SIZE(vd), NULL, NULL, | |
1186 | flags | ZIO_FLAG_DONT_PROPAGATE); | |
1187 | } | |
1188 | (void) zio_wait(write_zio); | |
1189 | ||
1190 | spa_config_exit(vd->vdev_spa, locks, FTAG); | |
1191 | ||
1192 | abd_free(ub_abd); | |
1193 | } | |
1194 | ||
34dc7c2f BB |
1195 | /* |
1196 | * On success, increment root zio's count of good writes. | |
1197 | * We only get credit for writes to known-visible vdevs; see spa_vdev_add(). | |
1198 | */ | |
1199 | static void | |
1200 | vdev_uberblock_sync_done(zio_t *zio) | |
1201 | { | |
1202 | uint64_t *good_writes = zio->io_private; | |
1203 | ||
1204 | if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0) | |
bc89ac84 | 1205 | atomic_inc_64(good_writes); |
34dc7c2f BB |
1206 | } |
1207 | ||
1208 | /* | |
1209 | * Write the uberblock to all labels of all leaves of the specified vdev. | |
1210 | */ | |
1211 | static void | |
b128c09f | 1212 | vdev_uberblock_sync(zio_t *zio, uberblock_t *ub, vdev_t *vd, int flags) |
34dc7c2f | 1213 | { |
379ca9cf | 1214 | for (int c = 0; c < vd->vdev_children; c++) |
b128c09f | 1215 | vdev_uberblock_sync(zio, ub, vd->vdev_child[c], flags); |
34dc7c2f BB |
1216 | |
1217 | if (!vd->vdev_ops->vdev_op_leaf) | |
1218 | return; | |
1219 | ||
b128c09f | 1220 | if (!vdev_writeable(vd)) |
34dc7c2f BB |
1221 | return; |
1222 | ||
9d3f7b87 OF |
1223 | /* If the vdev was expanded, need to copy uberblock rings. */ |
1224 | if (vd->vdev_state == VDEV_STATE_HEALTHY && | |
1225 | vd->vdev_copy_uberblocks == B_TRUE) { | |
1226 | vdev_copy_uberblocks(vd); | |
1227 | vd->vdev_copy_uberblocks = B_FALSE; | |
1228 | } | |
1229 | ||
379ca9cf OF |
1230 | int m = spa_multihost(vd->vdev_spa) ? MMP_BLOCKS_PER_LABEL : 0; |
1231 | int n = ub->ub_txg % (VDEV_UBERBLOCK_COUNT(vd) - m); | |
34dc7c2f | 1232 | |
a6255b7f | 1233 | /* Copy the uberblock_t into the ABD */ |
379ca9cf | 1234 | abd_t *ub_abd = abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd), B_TRUE); |
a6255b7f DQ |
1235 | abd_zero(ub_abd, VDEV_UBERBLOCK_SIZE(vd)); |
1236 | abd_copy_from_buf(ub_abd, ub, sizeof (uberblock_t)); | |
34dc7c2f | 1237 | |
379ca9cf | 1238 | for (int l = 0; l < VDEV_LABELS; l++) |
a6255b7f | 1239 | vdev_label_write(zio, vd, l, ub_abd, |
b128c09f | 1240 | VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd), |
34dc7c2f | 1241 | vdev_uberblock_sync_done, zio->io_private, |
b128c09f | 1242 | flags | ZIO_FLAG_DONT_PROPAGATE); |
34dc7c2f | 1243 | |
a6255b7f | 1244 | abd_free(ub_abd); |
34dc7c2f BB |
1245 | } |
1246 | ||
e49f1e20 | 1247 | /* Sync the uberblocks to all vdevs in svd[] */ |
34dc7c2f BB |
1248 | int |
1249 | vdev_uberblock_sync_list(vdev_t **svd, int svdcount, uberblock_t *ub, int flags) | |
1250 | { | |
1251 | spa_t *spa = svd[0]->vdev_spa; | |
34dc7c2f BB |
1252 | zio_t *zio; |
1253 | uint64_t good_writes = 0; | |
d6320ddb | 1254 | int v; |
34dc7c2f BB |
1255 | |
1256 | zio = zio_root(spa, NULL, &good_writes, flags); | |
1257 | ||
d6320ddb | 1258 | for (v = 0; v < svdcount; v++) |
b128c09f | 1259 | vdev_uberblock_sync(zio, ub, svd[v], flags); |
34dc7c2f BB |
1260 | |
1261 | (void) zio_wait(zio); | |
1262 | ||
1263 | /* | |
1264 | * Flush the uberblocks to disk. This ensures that the odd labels | |
1265 | * are no longer needed (because the new uberblocks and the even | |
1266 | * labels are safely on disk), so it is safe to overwrite them. | |
1267 | */ | |
1268 | zio = zio_root(spa, NULL, NULL, flags); | |
1269 | ||
d6320ddb | 1270 | for (v = 0; v < svdcount; v++) |
34dc7c2f BB |
1271 | zio_flush(zio, svd[v]); |
1272 | ||
1273 | (void) zio_wait(zio); | |
1274 | ||
1275 | return (good_writes >= 1 ? 0 : EIO); | |
1276 | } | |
1277 | ||
1278 | /* | |
1279 | * On success, increment the count of good writes for our top-level vdev. | |
1280 | */ | |
1281 | static void | |
1282 | vdev_label_sync_done(zio_t *zio) | |
1283 | { | |
1284 | uint64_t *good_writes = zio->io_private; | |
1285 | ||
1286 | if (zio->io_error == 0) | |
bc89ac84 | 1287 | atomic_inc_64(good_writes); |
34dc7c2f BB |
1288 | } |
1289 | ||
1290 | /* | |
1291 | * If there weren't enough good writes, indicate failure to the parent. | |
1292 | */ | |
1293 | static void | |
1294 | vdev_label_sync_top_done(zio_t *zio) | |
1295 | { | |
1296 | uint64_t *good_writes = zio->io_private; | |
1297 | ||
1298 | if (*good_writes == 0) | |
2e528b49 | 1299 | zio->io_error = SET_ERROR(EIO); |
34dc7c2f BB |
1300 | |
1301 | kmem_free(good_writes, sizeof (uint64_t)); | |
1302 | } | |
1303 | ||
b128c09f BB |
1304 | /* |
1305 | * We ignore errors for log and cache devices, simply free the private data. | |
1306 | */ | |
1307 | static void | |
1308 | vdev_label_sync_ignore_done(zio_t *zio) | |
1309 | { | |
1310 | kmem_free(zio->io_private, sizeof (uint64_t)); | |
1311 | } | |
1312 | ||
34dc7c2f BB |
1313 | /* |
1314 | * Write all even or odd labels to all leaves of the specified vdev. | |
1315 | */ | |
1316 | static void | |
b128c09f | 1317 | vdev_label_sync(zio_t *zio, vdev_t *vd, int l, uint64_t txg, int flags) |
34dc7c2f BB |
1318 | { |
1319 | nvlist_t *label; | |
1320 | vdev_phys_t *vp; | |
a6255b7f | 1321 | abd_t *vp_abd; |
34dc7c2f BB |
1322 | char *buf; |
1323 | size_t buflen; | |
d6320ddb | 1324 | int c; |
34dc7c2f | 1325 | |
d6320ddb | 1326 | for (c = 0; c < vd->vdev_children; c++) |
b128c09f | 1327 | vdev_label_sync(zio, vd->vdev_child[c], l, txg, flags); |
34dc7c2f BB |
1328 | |
1329 | if (!vd->vdev_ops->vdev_op_leaf) | |
1330 | return; | |
1331 | ||
b128c09f | 1332 | if (!vdev_writeable(vd)) |
34dc7c2f BB |
1333 | return; |
1334 | ||
1335 | /* | |
1336 | * Generate a label describing the top-level config to which we belong. | |
1337 | */ | |
1338 | label = spa_config_generate(vd->vdev_spa, vd, txg, B_FALSE); | |
1339 | ||
a6255b7f DQ |
1340 | vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE); |
1341 | abd_zero(vp_abd, sizeof (vdev_phys_t)); | |
1342 | vp = abd_to_buf(vp_abd); | |
34dc7c2f BB |
1343 | |
1344 | buf = vp->vp_nvlist; | |
1345 | buflen = sizeof (vp->vp_nvlist); | |
1346 | ||
79c76d5b | 1347 | if (!nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP)) { |
34dc7c2f | 1348 | for (; l < VDEV_LABELS; l += 2) { |
a6255b7f | 1349 | vdev_label_write(zio, vd, l, vp_abd, |
34dc7c2f BB |
1350 | offsetof(vdev_label_t, vl_vdev_phys), |
1351 | sizeof (vdev_phys_t), | |
1352 | vdev_label_sync_done, zio->io_private, | |
b128c09f | 1353 | flags | ZIO_FLAG_DONT_PROPAGATE); |
34dc7c2f BB |
1354 | } |
1355 | } | |
1356 | ||
a6255b7f | 1357 | abd_free(vp_abd); |
34dc7c2f BB |
1358 | nvlist_free(label); |
1359 | } | |
1360 | ||
1361 | int | |
b128c09f | 1362 | vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags) |
34dc7c2f | 1363 | { |
b128c09f | 1364 | list_t *dl = &spa->spa_config_dirty_list; |
34dc7c2f BB |
1365 | vdev_t *vd; |
1366 | zio_t *zio; | |
1367 | int error; | |
1368 | ||
1369 | /* | |
1370 | * Write the new labels to disk. | |
1371 | */ | |
1372 | zio = zio_root(spa, NULL, NULL, flags); | |
1373 | ||
1374 | for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) { | |
d6320ddb BB |
1375 | uint64_t *good_writes; |
1376 | zio_t *vio; | |
428870ff BB |
1377 | |
1378 | ASSERT(!vd->vdev_ishole); | |
1379 | ||
79c76d5b | 1380 | good_writes = kmem_zalloc(sizeof (uint64_t), KM_SLEEP); |
d6320ddb | 1381 | vio = zio_null(zio, spa, NULL, |
b128c09f BB |
1382 | (vd->vdev_islog || vd->vdev_aux != NULL) ? |
1383 | vdev_label_sync_ignore_done : vdev_label_sync_top_done, | |
34dc7c2f | 1384 | good_writes, flags); |
b128c09f | 1385 | vdev_label_sync(vio, vd, l, txg, flags); |
34dc7c2f BB |
1386 | zio_nowait(vio); |
1387 | } | |
1388 | ||
1389 | error = zio_wait(zio); | |
1390 | ||
1391 | /* | |
1392 | * Flush the new labels to disk. | |
1393 | */ | |
1394 | zio = zio_root(spa, NULL, NULL, flags); | |
1395 | ||
1396 | for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) | |
1397 | zio_flush(zio, vd); | |
1398 | ||
1399 | (void) zio_wait(zio); | |
1400 | ||
1401 | return (error); | |
1402 | } | |
1403 | ||
1404 | /* | |
1405 | * Sync the uberblock and any changes to the vdev configuration. | |
1406 | * | |
1407 | * The order of operations is carefully crafted to ensure that | |
1408 | * if the system panics or loses power at any time, the state on disk | |
1409 | * is still transactionally consistent. The in-line comments below | |
1410 | * describe the failure semantics at each stage. | |
1411 | * | |
1412 | * Moreover, vdev_config_sync() is designed to be idempotent: if it fails | |
1413 | * at any time, you can just call it again, and it will resume its work. | |
1414 | */ | |
1415 | int | |
b6fcb792 | 1416 | vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg) |
34dc7c2f BB |
1417 | { |
1418 | spa_t *spa = svd[0]->vdev_spa; | |
1419 | uberblock_t *ub = &spa->spa_uberblock; | |
1420 | vdev_t *vd; | |
1421 | zio_t *zio; | |
b6fcb792 | 1422 | int error = 0; |
b128c09f | 1423 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; |
34dc7c2f | 1424 | |
b6fcb792 | 1425 | retry: |
9babb374 BB |
1426 | /* |
1427 | * Normally, we don't want to try too hard to write every label and | |
1428 | * uberblock. If there is a flaky disk, we don't want the rest of the | |
1429 | * sync process to block while we retry. But if we can't write a | |
1430 | * single label out, we should retry with ZIO_FLAG_TRYHARD before | |
1431 | * bailing out and declaring the pool faulted. | |
1432 | */ | |
b6fcb792 BB |
1433 | if (error != 0) { |
1434 | if ((flags & ZIO_FLAG_TRYHARD) != 0) | |
1435 | return (error); | |
9babb374 | 1436 | flags |= ZIO_FLAG_TRYHARD; |
b6fcb792 | 1437 | } |
9babb374 | 1438 | |
34dc7c2f BB |
1439 | ASSERT(ub->ub_txg <= txg); |
1440 | ||
1441 | /* | |
1442 | * If this isn't a resync due to I/O errors, | |
1443 | * and nothing changed in this transaction group, | |
1444 | * and the vdev configuration hasn't changed, | |
1445 | * then there's nothing to do. | |
1446 | */ | |
379ca9cf OF |
1447 | if (ub->ub_txg < txg) { |
1448 | boolean_t changed = uberblock_update(ub, spa->spa_root_vdev, | |
1449 | txg, spa->spa_mmp.mmp_delay); | |
1450 | ||
1451 | if (!changed && list_is_empty(&spa->spa_config_dirty_list)) | |
1452 | return (0); | |
1453 | } | |
34dc7c2f BB |
1454 | |
1455 | if (txg > spa_freeze_txg(spa)) | |
1456 | return (0); | |
1457 | ||
1458 | ASSERT(txg <= spa->spa_final_txg); | |
1459 | ||
1460 | /* | |
1461 | * Flush the write cache of every disk that's been written to | |
1462 | * in this transaction group. This ensures that all blocks | |
1463 | * written in this txg will be committed to stable storage | |
1464 | * before any uberblock that references them. | |
1465 | */ | |
1466 | zio = zio_root(spa, NULL, NULL, flags); | |
1467 | ||
1468 | for (vd = txg_list_head(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)); vd; | |
1469 | vd = txg_list_next(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg))) | |
1470 | zio_flush(zio, vd); | |
1471 | ||
1472 | (void) zio_wait(zio); | |
1473 | ||
1474 | /* | |
1475 | * Sync out the even labels (L0, L2) for every dirty vdev. If the | |
1476 | * system dies in the middle of this process, that's OK: all of the | |
1477 | * even labels that made it to disk will be newer than any uberblock, | |
1478 | * and will therefore be considered invalid. The odd labels (L1, L3), | |
1479 | * which have not yet been touched, will still be valid. We flush | |
1480 | * the new labels to disk to ensure that all even-label updates | |
1481 | * are committed to stable storage before the uberblock update. | |
1482 | */ | |
b128c09f | 1483 | if ((error = vdev_label_sync_list(spa, 0, txg, flags)) != 0) |
b6fcb792 | 1484 | goto retry; |
34dc7c2f BB |
1485 | |
1486 | /* | |
1487 | * Sync the uberblocks to all vdevs in svd[]. | |
1488 | * If the system dies in the middle of this step, there are two cases | |
1489 | * to consider, and the on-disk state is consistent either way: | |
1490 | * | |
1491 | * (1) If none of the new uberblocks made it to disk, then the | |
1492 | * previous uberblock will be the newest, and the odd labels | |
1493 | * (which had not yet been touched) will be valid with respect | |
1494 | * to that uberblock. | |
1495 | * | |
1496 | * (2) If one or more new uberblocks made it to disk, then they | |
1497 | * will be the newest, and the even labels (which had all | |
1498 | * been successfully committed) will be valid with respect | |
1499 | * to the new uberblocks. | |
1500 | */ | |
1501 | if ((error = vdev_uberblock_sync_list(svd, svdcount, ub, flags)) != 0) | |
b6fcb792 | 1502 | goto retry; |
34dc7c2f | 1503 | |
379ca9cf OF |
1504 | |
1505 | if (spa_multihost(spa)) | |
1506 | mmp_update_uberblock(spa, ub); | |
1507 | ||
34dc7c2f BB |
1508 | /* |
1509 | * Sync out odd labels for every dirty vdev. If the system dies | |
1510 | * in the middle of this process, the even labels and the new | |
1511 | * uberblocks will suffice to open the pool. The next time | |
1512 | * the pool is opened, the first thing we'll do -- before any | |
1513 | * user data is modified -- is mark every vdev dirty so that | |
1514 | * all labels will be brought up to date. We flush the new labels | |
1515 | * to disk to ensure that all odd-label updates are committed to | |
1516 | * stable storage before the next transaction group begins. | |
1517 | */ | |
b6fcb792 BB |
1518 | if ((error = vdev_label_sync_list(spa, 1, txg, flags)) != 0) |
1519 | goto retry; | |
1520 | ||
1521 | return (0); | |
34dc7c2f | 1522 | } |