]>
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. |
2e528b49 | 24 | * Copyright (c) 2013 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> |
34dc7c2f BB |
148 | #include <sys/fs/zfs.h> |
149 | ||
150 | /* | |
151 | * Basic routines to read and write from a vdev label. | |
152 | * Used throughout the rest of this file. | |
153 | */ | |
154 | uint64_t | |
155 | vdev_label_offset(uint64_t psize, int l, uint64_t offset) | |
156 | { | |
157 | ASSERT(offset < sizeof (vdev_label_t)); | |
158 | ASSERT(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t) == 0); | |
159 | ||
160 | return (offset + l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? | |
161 | 0 : psize - VDEV_LABELS * sizeof (vdev_label_t))); | |
162 | } | |
163 | ||
b128c09f BB |
164 | /* |
165 | * Returns back the vdev label associated with the passed in offset. | |
166 | */ | |
167 | int | |
168 | vdev_label_number(uint64_t psize, uint64_t offset) | |
169 | { | |
170 | int l; | |
171 | ||
172 | if (offset >= psize - VDEV_LABEL_END_SIZE) { | |
173 | offset -= psize - VDEV_LABEL_END_SIZE; | |
174 | offset += (VDEV_LABELS / 2) * sizeof (vdev_label_t); | |
175 | } | |
176 | l = offset / sizeof (vdev_label_t); | |
177 | return (l < VDEV_LABELS ? l : -1); | |
178 | } | |
179 | ||
34dc7c2f BB |
180 | static void |
181 | vdev_label_read(zio_t *zio, vdev_t *vd, int l, void *buf, uint64_t offset, | |
b128c09f | 182 | uint64_t size, zio_done_func_t *done, void *private, int flags) |
34dc7c2f | 183 | { |
b128c09f BB |
184 | ASSERT(spa_config_held(zio->io_spa, SCL_STATE_ALL, RW_WRITER) == |
185 | SCL_STATE_ALL); | |
186 | ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); | |
34dc7c2f BB |
187 | |
188 | zio_nowait(zio_read_phys(zio, vd, | |
189 | vdev_label_offset(vd->vdev_psize, l, offset), | |
190 | size, buf, ZIO_CHECKSUM_LABEL, done, private, | |
b128c09f | 191 | ZIO_PRIORITY_SYNC_READ, flags, B_TRUE)); |
34dc7c2f BB |
192 | } |
193 | ||
194 | static void | |
195 | vdev_label_write(zio_t *zio, vdev_t *vd, int l, void *buf, uint64_t offset, | |
196 | uint64_t size, zio_done_func_t *done, void *private, int flags) | |
197 | { | |
b128c09f BB |
198 | ASSERT(spa_config_held(zio->io_spa, SCL_ALL, RW_WRITER) == SCL_ALL || |
199 | (spa_config_held(zio->io_spa, SCL_CONFIG | SCL_STATE, RW_READER) == | |
200 | (SCL_CONFIG | SCL_STATE) && | |
201 | dsl_pool_sync_context(spa_get_dsl(zio->io_spa)))); | |
202 | ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); | |
34dc7c2f BB |
203 | |
204 | zio_nowait(zio_write_phys(zio, vd, | |
205 | vdev_label_offset(vd->vdev_psize, l, offset), | |
206 | size, buf, ZIO_CHECKSUM_LABEL, done, private, | |
207 | ZIO_PRIORITY_SYNC_WRITE, flags, B_TRUE)); | |
208 | } | |
209 | ||
193a37cb TH |
210 | /* |
211 | * Generate the nvlist representing this vdev's stats | |
212 | */ | |
213 | void | |
214 | vdev_config_generate_stats(vdev_t *vd, nvlist_t *nv) | |
215 | { | |
216 | nvlist_t *nvx; | |
217 | vdev_stat_t *vs; | |
218 | vdev_stat_ex_t *vsx; | |
219 | ||
220 | vs = kmem_alloc(sizeof (*vs), KM_SLEEP); | |
221 | vsx = kmem_alloc(sizeof (*vsx), KM_SLEEP); | |
222 | ||
223 | vdev_get_stats_ex(vd, vs, vsx); | |
224 | fnvlist_add_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, | |
225 | (uint64_t *)vs, sizeof (*vs) / sizeof (uint64_t)); | |
226 | ||
227 | kmem_free(vs, sizeof (*vs)); | |
228 | ||
229 | /* | |
230 | * Add extended stats into a special extended stats nvlist. This keeps | |
231 | * all the extended stats nicely grouped together. The extended stats | |
232 | * nvlist is then added to the main nvlist. | |
233 | */ | |
234 | nvx = fnvlist_alloc(); | |
235 | ||
236 | /* ZIOs in flight to disk */ | |
237 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE, | |
238 | vsx->vsx_active_queue[ZIO_PRIORITY_SYNC_READ]); | |
239 | ||
240 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE, | |
241 | vsx->vsx_active_queue[ZIO_PRIORITY_SYNC_WRITE]); | |
242 | ||
243 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE, | |
244 | vsx->vsx_active_queue[ZIO_PRIORITY_ASYNC_READ]); | |
245 | ||
246 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE, | |
247 | vsx->vsx_active_queue[ZIO_PRIORITY_ASYNC_WRITE]); | |
248 | ||
249 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE, | |
250 | vsx->vsx_active_queue[ZIO_PRIORITY_SCRUB]); | |
251 | ||
252 | /* ZIOs pending */ | |
253 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE, | |
254 | vsx->vsx_pend_queue[ZIO_PRIORITY_SYNC_READ]); | |
255 | ||
256 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE, | |
257 | vsx->vsx_pend_queue[ZIO_PRIORITY_SYNC_WRITE]); | |
258 | ||
259 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE, | |
260 | vsx->vsx_pend_queue[ZIO_PRIORITY_ASYNC_READ]); | |
261 | ||
262 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE, | |
263 | vsx->vsx_pend_queue[ZIO_PRIORITY_ASYNC_WRITE]); | |
264 | ||
265 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE, | |
266 | vsx->vsx_pend_queue[ZIO_PRIORITY_SCRUB]); | |
267 | ||
268 | /* Histograms */ | |
269 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO, | |
270 | vsx->vsx_total_histo[ZIO_TYPE_READ], | |
271 | ARRAY_SIZE(vsx->vsx_total_histo[ZIO_TYPE_READ])); | |
272 | ||
273 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO, | |
274 | vsx->vsx_total_histo[ZIO_TYPE_WRITE], | |
275 | ARRAY_SIZE(vsx->vsx_total_histo[ZIO_TYPE_WRITE])); | |
276 | ||
277 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO, | |
278 | vsx->vsx_disk_histo[ZIO_TYPE_READ], | |
279 | ARRAY_SIZE(vsx->vsx_disk_histo[ZIO_TYPE_READ])); | |
280 | ||
281 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO, | |
282 | vsx->vsx_disk_histo[ZIO_TYPE_WRITE], | |
283 | ARRAY_SIZE(vsx->vsx_disk_histo[ZIO_TYPE_WRITE])); | |
284 | ||
285 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO, | |
286 | vsx->vsx_queue_histo[ZIO_PRIORITY_SYNC_READ], | |
287 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_SYNC_READ])); | |
288 | ||
289 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO, | |
290 | vsx->vsx_queue_histo[ZIO_PRIORITY_SYNC_WRITE], | |
291 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_SYNC_WRITE])); | |
292 | ||
293 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO, | |
294 | vsx->vsx_queue_histo[ZIO_PRIORITY_ASYNC_READ], | |
295 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_ASYNC_READ])); | |
296 | ||
297 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO, | |
298 | vsx->vsx_queue_histo[ZIO_PRIORITY_ASYNC_WRITE], | |
299 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_ASYNC_WRITE])); | |
300 | ||
301 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO, | |
302 | vsx->vsx_queue_histo[ZIO_PRIORITY_SCRUB], | |
303 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_SCRUB])); | |
304 | ||
7e945072 TH |
305 | /* Request sizes */ |
306 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO, | |
307 | vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_READ], | |
308 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_READ])); | |
309 | ||
310 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO, | |
311 | vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_WRITE], | |
312 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_WRITE])); | |
313 | ||
314 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO, | |
315 | vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_READ], | |
316 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_READ])); | |
317 | ||
318 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO, | |
319 | vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_WRITE], | |
320 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_WRITE])); | |
321 | ||
322 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO, | |
323 | vsx->vsx_ind_histo[ZIO_PRIORITY_SCRUB], | |
324 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_SCRUB])); | |
325 | ||
326 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO, | |
327 | vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_READ], | |
328 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_READ])); | |
329 | ||
330 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO, | |
331 | vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_WRITE], | |
332 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_WRITE])); | |
333 | ||
334 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO, | |
335 | vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_READ], | |
336 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_READ])); | |
337 | ||
338 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO, | |
339 | vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_WRITE], | |
340 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_WRITE])); | |
341 | ||
342 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO, | |
343 | vsx->vsx_agg_histo[ZIO_PRIORITY_SCRUB], | |
344 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_SCRUB])); | |
345 | ||
193a37cb TH |
346 | /* Add extended stats nvlist to main nvlist */ |
347 | fnvlist_add_nvlist(nv, ZPOOL_CONFIG_VDEV_STATS_EX, nvx); | |
348 | ||
6a796725 | 349 | fnvlist_free(nvx); |
193a37cb TH |
350 | kmem_free(vsx, sizeof (*vsx)); |
351 | } | |
352 | ||
34dc7c2f BB |
353 | /* |
354 | * Generate the nvlist representing this vdev's config. | |
355 | */ | |
356 | nvlist_t * | |
357 | vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats, | |
428870ff | 358 | vdev_config_flag_t flags) |
34dc7c2f BB |
359 | { |
360 | nvlist_t *nv = NULL; | |
79c76d5b | 361 | nv = fnvlist_alloc(); |
34dc7c2f | 362 | |
5d1f7fb6 | 363 | fnvlist_add_string(nv, ZPOOL_CONFIG_TYPE, vd->vdev_ops->vdev_op_type); |
428870ff | 364 | if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE))) |
5d1f7fb6 GW |
365 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ID, vd->vdev_id); |
366 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_GUID, vd->vdev_guid); | |
34dc7c2f BB |
367 | |
368 | if (vd->vdev_path != NULL) | |
5d1f7fb6 | 369 | fnvlist_add_string(nv, ZPOOL_CONFIG_PATH, vd->vdev_path); |
34dc7c2f BB |
370 | |
371 | if (vd->vdev_devid != NULL) | |
5d1f7fb6 | 372 | fnvlist_add_string(nv, ZPOOL_CONFIG_DEVID, vd->vdev_devid); |
34dc7c2f BB |
373 | |
374 | if (vd->vdev_physpath != NULL) | |
5d1f7fb6 GW |
375 | fnvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH, |
376 | vd->vdev_physpath); | |
34dc7c2f | 377 | |
9babb374 | 378 | if (vd->vdev_fru != NULL) |
5d1f7fb6 | 379 | fnvlist_add_string(nv, ZPOOL_CONFIG_FRU, vd->vdev_fru); |
9babb374 | 380 | |
34dc7c2f BB |
381 | if (vd->vdev_nparity != 0) { |
382 | ASSERT(strcmp(vd->vdev_ops->vdev_op_type, | |
383 | VDEV_TYPE_RAIDZ) == 0); | |
384 | ||
385 | /* | |
386 | * Make sure someone hasn't managed to sneak a fancy new vdev | |
387 | * into a crufty old storage pool. | |
388 | */ | |
389 | ASSERT(vd->vdev_nparity == 1 || | |
45d1cae3 BB |
390 | (vd->vdev_nparity <= 2 && |
391 | spa_version(spa) >= SPA_VERSION_RAIDZ2) || | |
392 | (vd->vdev_nparity <= 3 && | |
393 | spa_version(spa) >= SPA_VERSION_RAIDZ3)); | |
34dc7c2f BB |
394 | |
395 | /* | |
396 | * Note that we'll add the nparity tag even on storage pools | |
397 | * that only support a single parity device -- older software | |
398 | * will just ignore it. | |
399 | */ | |
5d1f7fb6 | 400 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY, vd->vdev_nparity); |
34dc7c2f BB |
401 | } |
402 | ||
403 | if (vd->vdev_wholedisk != -1ULL) | |
5d1f7fb6 GW |
404 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, |
405 | vd->vdev_wholedisk); | |
34dc7c2f BB |
406 | |
407 | if (vd->vdev_not_present) | |
5d1f7fb6 | 408 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1); |
34dc7c2f BB |
409 | |
410 | if (vd->vdev_isspare) | |
5d1f7fb6 | 411 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1); |
34dc7c2f | 412 | |
428870ff BB |
413 | if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)) && |
414 | vd == vd->vdev_top) { | |
5d1f7fb6 GW |
415 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY, |
416 | vd->vdev_ms_array); | |
417 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT, | |
418 | vd->vdev_ms_shift); | |
419 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT, vd->vdev_ashift); | |
420 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASIZE, | |
421 | vd->vdev_asize); | |
422 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_LOG, vd->vdev_islog); | |
428870ff | 423 | if (vd->vdev_removing) |
5d1f7fb6 GW |
424 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVING, |
425 | vd->vdev_removing); | |
34dc7c2f BB |
426 | } |
427 | ||
93cf2076 | 428 | if (vd->vdev_dtl_sm != NULL) { |
5d1f7fb6 | 429 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_DTL, |
93cf2076 GW |
430 | space_map_object(vd->vdev_dtl_sm)); |
431 | } | |
34dc7c2f | 432 | |
428870ff | 433 | if (vd->vdev_crtxg) |
5d1f7fb6 | 434 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG, vd->vdev_crtxg); |
428870ff | 435 | |
e0ab3ab5 JS |
436 | if (flags & VDEV_CONFIG_MOS) { |
437 | if (vd->vdev_leaf_zap != 0) { | |
438 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
439 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_LEAF_ZAP, | |
440 | vd->vdev_leaf_zap); | |
441 | } | |
442 | ||
443 | if (vd->vdev_top_zap != 0) { | |
444 | ASSERT(vd == vd->vdev_top); | |
445 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP, | |
446 | vd->vdev_top_zap); | |
447 | } | |
448 | } | |
449 | ||
34dc7c2f | 450 | if (getstats) { |
428870ff BB |
451 | pool_scan_stat_t ps; |
452 | ||
193a37cb | 453 | vdev_config_generate_stats(vd, nv); |
428870ff BB |
454 | |
455 | /* provide either current or previous scan information */ | |
456 | if (spa_scan_get_stats(spa, &ps) == 0) { | |
5d1f7fb6 | 457 | fnvlist_add_uint64_array(nv, |
428870ff | 458 | ZPOOL_CONFIG_SCAN_STATS, (uint64_t *)&ps, |
5d1f7fb6 | 459 | sizeof (pool_scan_stat_t) / sizeof (uint64_t)); |
428870ff | 460 | } |
34dc7c2f BB |
461 | } |
462 | ||
463 | if (!vd->vdev_ops->vdev_op_leaf) { | |
464 | nvlist_t **child; | |
428870ff BB |
465 | int c, idx; |
466 | ||
467 | ASSERT(!vd->vdev_ishole); | |
34dc7c2f BB |
468 | |
469 | child = kmem_alloc(vd->vdev_children * sizeof (nvlist_t *), | |
79c76d5b | 470 | KM_SLEEP); |
34dc7c2f | 471 | |
428870ff BB |
472 | for (c = 0, idx = 0; c < vd->vdev_children; c++) { |
473 | vdev_t *cvd = vd->vdev_child[c]; | |
34dc7c2f | 474 | |
428870ff BB |
475 | /* |
476 | * If we're generating an nvlist of removing | |
477 | * vdevs then skip over any device which is | |
478 | * not being removed. | |
479 | */ | |
480 | if ((flags & VDEV_CONFIG_REMOVING) && | |
481 | !cvd->vdev_removing) | |
482 | continue; | |
34dc7c2f | 483 | |
428870ff BB |
484 | child[idx++] = vdev_config_generate(spa, cvd, |
485 | getstats, flags); | |
486 | } | |
487 | ||
488 | if (idx) { | |
5d1f7fb6 GW |
489 | fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, |
490 | child, idx); | |
428870ff BB |
491 | } |
492 | ||
493 | for (c = 0; c < idx; c++) | |
34dc7c2f BB |
494 | nvlist_free(child[c]); |
495 | ||
496 | kmem_free(child, vd->vdev_children * sizeof (nvlist_t *)); | |
497 | ||
498 | } else { | |
428870ff BB |
499 | const char *aux = NULL; |
500 | ||
34dc7c2f | 501 | if (vd->vdev_offline && !vd->vdev_tmpoffline) |
5d1f7fb6 GW |
502 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_OFFLINE, B_TRUE); |
503 | if (vd->vdev_resilver_txg != 0) | |
504 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_RESILVER_TXG, | |
505 | vd->vdev_resilver_txg); | |
34dc7c2f | 506 | if (vd->vdev_faulted) |
5d1f7fb6 | 507 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_FAULTED, B_TRUE); |
34dc7c2f | 508 | if (vd->vdev_degraded) |
5d1f7fb6 | 509 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_DEGRADED, B_TRUE); |
34dc7c2f | 510 | if (vd->vdev_removed) |
5d1f7fb6 | 511 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVED, B_TRUE); |
34dc7c2f | 512 | if (vd->vdev_unspare) |
5d1f7fb6 | 513 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_UNSPARE, B_TRUE); |
428870ff | 514 | if (vd->vdev_ishole) |
5d1f7fb6 | 515 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_HOLE, B_TRUE); |
428870ff BB |
516 | |
517 | switch (vd->vdev_stat.vs_aux) { | |
518 | case VDEV_AUX_ERR_EXCEEDED: | |
519 | aux = "err_exceeded"; | |
520 | break; | |
521 | ||
522 | case VDEV_AUX_EXTERNAL: | |
523 | aux = "external"; | |
524 | break; | |
525 | } | |
526 | ||
527 | if (aux != NULL) | |
5d1f7fb6 | 528 | fnvlist_add_string(nv, ZPOOL_CONFIG_AUX_STATE, aux); |
428870ff BB |
529 | |
530 | if (vd->vdev_splitting && vd->vdev_orig_guid != 0LL) { | |
5d1f7fb6 GW |
531 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ORIG_GUID, |
532 | vd->vdev_orig_guid); | |
428870ff | 533 | } |
34dc7c2f BB |
534 | } |
535 | ||
536 | return (nv); | |
537 | } | |
538 | ||
428870ff BB |
539 | /* |
540 | * Generate a view of the top-level vdevs. If we currently have holes | |
541 | * in the namespace, then generate an array which contains a list of holey | |
542 | * vdevs. Additionally, add the number of top-level children that currently | |
543 | * exist. | |
544 | */ | |
545 | void | |
546 | vdev_top_config_generate(spa_t *spa, nvlist_t *config) | |
547 | { | |
548 | vdev_t *rvd = spa->spa_root_vdev; | |
549 | uint64_t *array; | |
550 | uint_t c, idx; | |
551 | ||
79c76d5b | 552 | array = kmem_alloc(rvd->vdev_children * sizeof (uint64_t), KM_SLEEP); |
428870ff BB |
553 | |
554 | for (c = 0, idx = 0; c < rvd->vdev_children; c++) { | |
555 | vdev_t *tvd = rvd->vdev_child[c]; | |
556 | ||
557 | if (tvd->vdev_ishole) | |
558 | array[idx++] = c; | |
559 | } | |
560 | ||
561 | if (idx) { | |
562 | VERIFY(nvlist_add_uint64_array(config, ZPOOL_CONFIG_HOLE_ARRAY, | |
563 | array, idx) == 0); | |
564 | } | |
565 | ||
566 | VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN, | |
567 | rvd->vdev_children) == 0); | |
568 | ||
569 | kmem_free(array, rvd->vdev_children * sizeof (uint64_t)); | |
570 | } | |
571 | ||
9ae529ec | 572 | /* |
3bc7e0fb GW |
573 | * Returns the configuration from the label of the given vdev. For vdevs |
574 | * which don't have a txg value stored on their label (i.e. spares/cache) | |
575 | * or have not been completely initialized (txg = 0) just return | |
576 | * the configuration from the first valid label we find. Otherwise, | |
577 | * find the most up-to-date label that does not exceed the specified | |
578 | * 'txg' value. | |
9ae529ec | 579 | */ |
34dc7c2f | 580 | nvlist_t * |
3bc7e0fb | 581 | vdev_label_read_config(vdev_t *vd, uint64_t txg) |
34dc7c2f BB |
582 | { |
583 | spa_t *spa = vd->vdev_spa; | |
584 | nvlist_t *config = NULL; | |
585 | vdev_phys_t *vp; | |
586 | zio_t *zio; | |
3bc7e0fb GW |
587 | uint64_t best_txg = 0; |
588 | int error = 0; | |
9babb374 BB |
589 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | |
590 | ZIO_FLAG_SPECULATIVE; | |
d6320ddb | 591 | int l; |
34dc7c2f | 592 | |
b128c09f | 593 | ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL); |
34dc7c2f BB |
594 | |
595 | if (!vdev_readable(vd)) | |
596 | return (NULL); | |
597 | ||
598 | vp = zio_buf_alloc(sizeof (vdev_phys_t)); | |
599 | ||
9babb374 | 600 | retry: |
d6320ddb | 601 | for (l = 0; l < VDEV_LABELS; l++) { |
3bc7e0fb | 602 | nvlist_t *label = NULL; |
34dc7c2f | 603 | |
b128c09f | 604 | zio = zio_root(spa, NULL, NULL, flags); |
34dc7c2f BB |
605 | |
606 | vdev_label_read(zio, vd, l, vp, | |
607 | offsetof(vdev_label_t, vl_vdev_phys), | |
b128c09f | 608 | sizeof (vdev_phys_t), NULL, NULL, flags); |
34dc7c2f BB |
609 | |
610 | if (zio_wait(zio) == 0 && | |
611 | nvlist_unpack(vp->vp_nvlist, sizeof (vp->vp_nvlist), | |
3bc7e0fb GW |
612 | &label, 0) == 0) { |
613 | uint64_t label_txg = 0; | |
614 | ||
615 | /* | |
616 | * Auxiliary vdevs won't have txg values in their | |
617 | * labels and newly added vdevs may not have been | |
618 | * completely initialized so just return the | |
619 | * configuration from the first valid label we | |
620 | * encounter. | |
621 | */ | |
622 | error = nvlist_lookup_uint64(label, | |
623 | ZPOOL_CONFIG_POOL_TXG, &label_txg); | |
624 | if ((error || label_txg == 0) && !config) { | |
625 | config = label; | |
626 | break; | |
627 | } else if (label_txg <= txg && label_txg > best_txg) { | |
628 | best_txg = label_txg; | |
629 | nvlist_free(config); | |
630 | config = fnvlist_dup(label); | |
631 | } | |
632 | } | |
34dc7c2f | 633 | |
3bc7e0fb GW |
634 | if (label != NULL) { |
635 | nvlist_free(label); | |
636 | label = NULL; | |
34dc7c2f BB |
637 | } |
638 | } | |
639 | ||
9babb374 BB |
640 | if (config == NULL && !(flags & ZIO_FLAG_TRYHARD)) { |
641 | flags |= ZIO_FLAG_TRYHARD; | |
642 | goto retry; | |
643 | } | |
644 | ||
34dc7c2f BB |
645 | zio_buf_free(vp, sizeof (vdev_phys_t)); |
646 | ||
647 | return (config); | |
648 | } | |
649 | ||
650 | /* | |
651 | * Determine if a device is in use. The 'spare_guid' parameter will be filled | |
652 | * in with the device guid if this spare is active elsewhere on the system. | |
653 | */ | |
654 | static boolean_t | |
655 | vdev_inuse(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason, | |
656 | uint64_t *spare_guid, uint64_t *l2cache_guid) | |
657 | { | |
658 | spa_t *spa = vd->vdev_spa; | |
659 | uint64_t state, pool_guid, device_guid, txg, spare_pool; | |
660 | uint64_t vdtxg = 0; | |
661 | nvlist_t *label; | |
662 | ||
663 | if (spare_guid) | |
664 | *spare_guid = 0ULL; | |
665 | if (l2cache_guid) | |
666 | *l2cache_guid = 0ULL; | |
667 | ||
668 | /* | |
669 | * Read the label, if any, and perform some basic sanity checks. | |
670 | */ | |
3bc7e0fb | 671 | if ((label = vdev_label_read_config(vd, -1ULL)) == NULL) |
34dc7c2f BB |
672 | return (B_FALSE); |
673 | ||
674 | (void) nvlist_lookup_uint64(label, ZPOOL_CONFIG_CREATE_TXG, | |
675 | &vdtxg); | |
676 | ||
677 | if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, | |
678 | &state) != 0 || | |
679 | nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, | |
680 | &device_guid) != 0) { | |
681 | nvlist_free(label); | |
682 | return (B_FALSE); | |
683 | } | |
684 | ||
685 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
686 | (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, | |
687 | &pool_guid) != 0 || | |
688 | nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG, | |
689 | &txg) != 0)) { | |
690 | nvlist_free(label); | |
691 | return (B_FALSE); | |
692 | } | |
693 | ||
694 | nvlist_free(label); | |
695 | ||
696 | /* | |
697 | * Check to see if this device indeed belongs to the pool it claims to | |
698 | * be a part of. The only way this is allowed is if the device is a hot | |
699 | * spare (which we check for later on). | |
700 | */ | |
701 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
702 | !spa_guid_exists(pool_guid, device_guid) && | |
b128c09f | 703 | !spa_spare_exists(device_guid, NULL, NULL) && |
34dc7c2f BB |
704 | !spa_l2cache_exists(device_guid, NULL)) |
705 | return (B_FALSE); | |
706 | ||
707 | /* | |
708 | * If the transaction group is zero, then this an initialized (but | |
709 | * unused) label. This is only an error if the create transaction | |
710 | * on-disk is the same as the one we're using now, in which case the | |
711 | * user has attempted to add the same vdev multiple times in the same | |
712 | * transaction. | |
713 | */ | |
714 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
715 | txg == 0 && vdtxg == crtxg) | |
716 | return (B_TRUE); | |
717 | ||
718 | /* | |
719 | * Check to see if this is a spare device. We do an explicit check for | |
720 | * spa_has_spare() here because it may be on our pending list of spares | |
721 | * to add. We also check if it is an l2cache device. | |
722 | */ | |
b128c09f | 723 | if (spa_spare_exists(device_guid, &spare_pool, NULL) || |
34dc7c2f BB |
724 | spa_has_spare(spa, device_guid)) { |
725 | if (spare_guid) | |
726 | *spare_guid = device_guid; | |
727 | ||
728 | switch (reason) { | |
729 | case VDEV_LABEL_CREATE: | |
730 | case VDEV_LABEL_L2CACHE: | |
731 | return (B_TRUE); | |
732 | ||
733 | case VDEV_LABEL_REPLACE: | |
734 | return (!spa_has_spare(spa, device_guid) || | |
735 | spare_pool != 0ULL); | |
736 | ||
737 | case VDEV_LABEL_SPARE: | |
738 | return (spa_has_spare(spa, device_guid)); | |
e75c13c3 BB |
739 | default: |
740 | break; | |
34dc7c2f BB |
741 | } |
742 | } | |
743 | ||
744 | /* | |
745 | * Check to see if this is an l2cache device. | |
746 | */ | |
747 | if (spa_l2cache_exists(device_guid, NULL)) | |
748 | return (B_TRUE); | |
749 | ||
572e2857 BB |
750 | /* |
751 | * We can't rely on a pool's state if it's been imported | |
752 | * read-only. Instead we look to see if the pools is marked | |
753 | * read-only in the namespace and set the state to active. | |
754 | */ | |
485c581c RY |
755 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && |
756 | (spa = spa_by_guid(pool_guid, device_guid)) != NULL && | |
572e2857 BB |
757 | spa_mode(spa) == FREAD) |
758 | state = POOL_STATE_ACTIVE; | |
759 | ||
34dc7c2f BB |
760 | /* |
761 | * If the device is marked ACTIVE, then this device is in use by another | |
762 | * pool on the system. | |
763 | */ | |
764 | return (state == POOL_STATE_ACTIVE); | |
765 | } | |
766 | ||
767 | /* | |
768 | * Initialize a vdev label. We check to make sure each leaf device is not in | |
769 | * use, and writable. We put down an initial label which we will later | |
770 | * overwrite with a complete label. Note that it's important to do this | |
771 | * sequentially, not in parallel, so that we catch cases of multiple use of the | |
772 | * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with | |
773 | * itself. | |
774 | */ | |
775 | int | |
776 | vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason) | |
777 | { | |
778 | spa_t *spa = vd->vdev_spa; | |
779 | nvlist_t *label; | |
780 | vdev_phys_t *vp; | |
9babb374 | 781 | char *pad2; |
34dc7c2f BB |
782 | uberblock_t *ub; |
783 | zio_t *zio; | |
34dc7c2f BB |
784 | char *buf; |
785 | size_t buflen; | |
786 | int error; | |
d4ed6673 | 787 | uint64_t spare_guid = 0, l2cache_guid = 0; |
b128c09f | 788 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; |
d6320ddb BB |
789 | int c, l; |
790 | vdev_t *pvd; | |
34dc7c2f | 791 | |
b128c09f | 792 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); |
34dc7c2f | 793 | |
d6320ddb | 794 | for (c = 0; c < vd->vdev_children; c++) |
34dc7c2f BB |
795 | if ((error = vdev_label_init(vd->vdev_child[c], |
796 | crtxg, reason)) != 0) | |
797 | return (error); | |
798 | ||
428870ff BB |
799 | /* Track the creation time for this vdev */ |
800 | vd->vdev_crtxg = crtxg; | |
801 | ||
dda12da9 | 802 | if (!vd->vdev_ops->vdev_op_leaf || !spa_writeable(spa)) |
34dc7c2f BB |
803 | return (0); |
804 | ||
805 | /* | |
806 | * Dead vdevs cannot be initialized. | |
807 | */ | |
808 | if (vdev_is_dead(vd)) | |
2e528b49 | 809 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
810 | |
811 | /* | |
812 | * Determine if the vdev is in use. | |
813 | */ | |
428870ff | 814 | if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPLIT && |
34dc7c2f | 815 | vdev_inuse(vd, crtxg, reason, &spare_guid, &l2cache_guid)) |
2e528b49 | 816 | return (SET_ERROR(EBUSY)); |
34dc7c2f | 817 | |
34dc7c2f BB |
818 | /* |
819 | * If this is a request to add or replace a spare or l2cache device | |
820 | * that is in use elsewhere on the system, then we must update the | |
821 | * guid (which was initialized to a random value) to reflect the | |
822 | * actual GUID (which is shared between multiple pools). | |
823 | */ | |
824 | if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_L2CACHE && | |
825 | spare_guid != 0ULL) { | |
b128c09f | 826 | uint64_t guid_delta = spare_guid - vd->vdev_guid; |
34dc7c2f | 827 | |
b128c09f | 828 | vd->vdev_guid += guid_delta; |
34dc7c2f | 829 | |
d6320ddb | 830 | for (pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) |
b128c09f | 831 | pvd->vdev_guid_sum += guid_delta; |
34dc7c2f BB |
832 | |
833 | /* | |
834 | * If this is a replacement, then we want to fallthrough to the | |
835 | * rest of the code. If we're adding a spare, then it's already | |
836 | * labeled appropriately and we can just return. | |
837 | */ | |
838 | if (reason == VDEV_LABEL_SPARE) | |
839 | return (0); | |
428870ff BB |
840 | ASSERT(reason == VDEV_LABEL_REPLACE || |
841 | reason == VDEV_LABEL_SPLIT); | |
34dc7c2f BB |
842 | } |
843 | ||
844 | if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPARE && | |
845 | l2cache_guid != 0ULL) { | |
b128c09f | 846 | uint64_t guid_delta = l2cache_guid - vd->vdev_guid; |
34dc7c2f | 847 | |
b128c09f | 848 | vd->vdev_guid += guid_delta; |
34dc7c2f | 849 | |
d6320ddb | 850 | for (pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) |
b128c09f | 851 | pvd->vdev_guid_sum += guid_delta; |
34dc7c2f BB |
852 | |
853 | /* | |
854 | * If this is a replacement, then we want to fallthrough to the | |
855 | * rest of the code. If we're adding an l2cache, then it's | |
856 | * already labeled appropriately and we can just return. | |
857 | */ | |
858 | if (reason == VDEV_LABEL_L2CACHE) | |
859 | return (0); | |
860 | ASSERT(reason == VDEV_LABEL_REPLACE); | |
861 | } | |
862 | ||
863 | /* | |
864 | * Initialize its label. | |
865 | */ | |
866 | vp = zio_buf_alloc(sizeof (vdev_phys_t)); | |
867 | bzero(vp, sizeof (vdev_phys_t)); | |
868 | ||
869 | /* | |
870 | * Generate a label describing the pool and our top-level vdev. | |
871 | * We mark it as being from txg 0 to indicate that it's not | |
872 | * really part of an active pool just yet. The labels will | |
873 | * be written again with a meaningful txg by spa_sync(). | |
874 | */ | |
875 | if (reason == VDEV_LABEL_SPARE || | |
876 | (reason == VDEV_LABEL_REMOVE && vd->vdev_isspare)) { | |
877 | /* | |
878 | * For inactive hot spares, we generate a special label that | |
879 | * identifies as a mutually shared hot spare. We write the | |
880 | * label if we are adding a hot spare, or if we are removing an | |
881 | * active hot spare (in which case we want to revert the | |
882 | * labels). | |
883 | */ | |
79c76d5b | 884 | VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); |
34dc7c2f BB |
885 | |
886 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, | |
887 | spa_version(spa)) == 0); | |
888 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, | |
889 | POOL_STATE_SPARE) == 0); | |
890 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, | |
891 | vd->vdev_guid) == 0); | |
892 | } else if (reason == VDEV_LABEL_L2CACHE || | |
893 | (reason == VDEV_LABEL_REMOVE && vd->vdev_isl2cache)) { | |
894 | /* | |
895 | * For level 2 ARC devices, add a special label. | |
896 | */ | |
79c76d5b | 897 | VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); |
34dc7c2f BB |
898 | |
899 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, | |
900 | spa_version(spa)) == 0); | |
901 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, | |
902 | POOL_STATE_L2CACHE) == 0); | |
903 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, | |
904 | vd->vdev_guid) == 0); | |
905 | } else { | |
428870ff BB |
906 | uint64_t txg = 0ULL; |
907 | ||
908 | if (reason == VDEV_LABEL_SPLIT) | |
909 | txg = spa->spa_uberblock.ub_txg; | |
910 | label = spa_config_generate(spa, vd, txg, B_FALSE); | |
34dc7c2f BB |
911 | |
912 | /* | |
913 | * Add our creation time. This allows us to detect multiple | |
914 | * vdev uses as described above, and automatically expires if we | |
915 | * fail. | |
916 | */ | |
917 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG, | |
918 | crtxg) == 0); | |
919 | } | |
920 | ||
921 | buf = vp->vp_nvlist; | |
922 | buflen = sizeof (vp->vp_nvlist); | |
923 | ||
79c76d5b | 924 | error = nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP); |
34dc7c2f BB |
925 | if (error != 0) { |
926 | nvlist_free(label); | |
927 | zio_buf_free(vp, sizeof (vdev_phys_t)); | |
928 | /* EFAULT means nvlist_pack ran out of room */ | |
929 | return (error == EFAULT ? ENAMETOOLONG : EINVAL); | |
930 | } | |
931 | ||
34dc7c2f BB |
932 | /* |
933 | * Initialize uberblock template. | |
934 | */ | |
45d1cae3 BB |
935 | ub = zio_buf_alloc(VDEV_UBERBLOCK_RING); |
936 | bzero(ub, VDEV_UBERBLOCK_RING); | |
34dc7c2f BB |
937 | *ub = spa->spa_uberblock; |
938 | ub->ub_txg = 0; | |
939 | ||
9babb374 BB |
940 | /* Initialize the 2nd padding area. */ |
941 | pad2 = zio_buf_alloc(VDEV_PAD_SIZE); | |
942 | bzero(pad2, VDEV_PAD_SIZE); | |
943 | ||
34dc7c2f BB |
944 | /* |
945 | * Write everything in parallel. | |
946 | */ | |
9babb374 | 947 | retry: |
34dc7c2f BB |
948 | zio = zio_root(spa, NULL, NULL, flags); |
949 | ||
d6320ddb | 950 | for (l = 0; l < VDEV_LABELS; l++) { |
34dc7c2f BB |
951 | |
952 | vdev_label_write(zio, vd, l, vp, | |
953 | offsetof(vdev_label_t, vl_vdev_phys), | |
954 | sizeof (vdev_phys_t), NULL, NULL, flags); | |
955 | ||
9babb374 BB |
956 | /* |
957 | * Skip the 1st padding area. | |
958 | * Zero out the 2nd padding area where it might have | |
959 | * left over data from previous filesystem format. | |
960 | */ | |
961 | vdev_label_write(zio, vd, l, pad2, | |
962 | offsetof(vdev_label_t, vl_pad2), | |
963 | VDEV_PAD_SIZE, NULL, NULL, flags); | |
34dc7c2f | 964 | |
45d1cae3 BB |
965 | vdev_label_write(zio, vd, l, ub, |
966 | offsetof(vdev_label_t, vl_uberblock), | |
967 | VDEV_UBERBLOCK_RING, NULL, NULL, flags); | |
34dc7c2f BB |
968 | } |
969 | ||
970 | error = zio_wait(zio); | |
971 | ||
9babb374 BB |
972 | if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) { |
973 | flags |= ZIO_FLAG_TRYHARD; | |
974 | goto retry; | |
975 | } | |
976 | ||
34dc7c2f | 977 | nvlist_free(label); |
9babb374 | 978 | zio_buf_free(pad2, VDEV_PAD_SIZE); |
45d1cae3 | 979 | zio_buf_free(ub, VDEV_UBERBLOCK_RING); |
34dc7c2f BB |
980 | zio_buf_free(vp, sizeof (vdev_phys_t)); |
981 | ||
982 | /* | |
983 | * If this vdev hasn't been previously identified as a spare, then we | |
984 | * mark it as such only if a) we are labeling it as a spare, or b) it | |
985 | * exists as a spare elsewhere in the system. Do the same for | |
986 | * level 2 ARC devices. | |
987 | */ | |
988 | if (error == 0 && !vd->vdev_isspare && | |
989 | (reason == VDEV_LABEL_SPARE || | |
b128c09f | 990 | spa_spare_exists(vd->vdev_guid, NULL, NULL))) |
34dc7c2f BB |
991 | spa_spare_add(vd); |
992 | ||
993 | if (error == 0 && !vd->vdev_isl2cache && | |
994 | (reason == VDEV_LABEL_L2CACHE || | |
995 | spa_l2cache_exists(vd->vdev_guid, NULL))) | |
996 | spa_l2cache_add(vd); | |
997 | ||
998 | return (error); | |
999 | } | |
1000 | ||
1001 | /* | |
1002 | * ========================================================================== | |
1003 | * uberblock load/sync | |
1004 | * ========================================================================== | |
1005 | */ | |
1006 | ||
1007 | /* | |
1008 | * Consider the following situation: txg is safely synced to disk. We've | |
1009 | * written the first uberblock for txg + 1, and then we lose power. When we | |
1010 | * come back up, we fail to see the uberblock for txg + 1 because, say, | |
1011 | * it was on a mirrored device and the replica to which we wrote txg + 1 | |
1012 | * is now offline. If we then make some changes and sync txg + 1, and then | |
9ae529ec | 1013 | * the missing replica comes back, then for a few seconds we'll have two |
34dc7c2f BB |
1014 | * conflicting uberblocks on disk with the same txg. The solution is simple: |
1015 | * among uberblocks with equal txg, choose the one with the latest timestamp. | |
1016 | */ | |
1017 | static int | |
ee36c709 | 1018 | vdev_uberblock_compare(const uberblock_t *ub1, const uberblock_t *ub2) |
34dc7c2f | 1019 | { |
ee36c709 GN |
1020 | int cmp = AVL_CMP(ub1->ub_txg, ub2->ub_txg); |
1021 | if (likely(cmp)) | |
1022 | return (cmp); | |
34dc7c2f | 1023 | |
ee36c709 | 1024 | return (AVL_CMP(ub1->ub_timestamp, ub2->ub_timestamp)); |
34dc7c2f BB |
1025 | } |
1026 | ||
9ae529ec CS |
1027 | struct ubl_cbdata { |
1028 | uberblock_t *ubl_ubbest; /* Best uberblock */ | |
1029 | vdev_t *ubl_vd; /* vdev associated with the above */ | |
9ae529ec CS |
1030 | }; |
1031 | ||
34dc7c2f BB |
1032 | static void |
1033 | vdev_uberblock_load_done(zio_t *zio) | |
1034 | { | |
9ae529ec | 1035 | vdev_t *vd = zio->io_vd; |
428870ff | 1036 | spa_t *spa = zio->io_spa; |
b128c09f | 1037 | zio_t *rio = zio->io_private; |
34dc7c2f | 1038 | uberblock_t *ub = zio->io_data; |
9ae529ec | 1039 | struct ubl_cbdata *cbp = rio->io_private; |
34dc7c2f | 1040 | |
9ae529ec | 1041 | ASSERT3U(zio->io_size, ==, VDEV_UBERBLOCK_SIZE(vd)); |
34dc7c2f BB |
1042 | |
1043 | if (zio->io_error == 0 && uberblock_verify(ub) == 0) { | |
b128c09f | 1044 | mutex_enter(&rio->io_lock); |
428870ff | 1045 | if (ub->ub_txg <= spa->spa_load_max_txg && |
9ae529ec CS |
1046 | vdev_uberblock_compare(ub, cbp->ubl_ubbest) > 0) { |
1047 | /* | |
3bc7e0fb GW |
1048 | * Keep track of the vdev in which this uberblock |
1049 | * was found. We will use this information later | |
1050 | * to obtain the config nvlist associated with | |
9ae529ec CS |
1051 | * this uberblock. |
1052 | */ | |
1053 | *cbp->ubl_ubbest = *ub; | |
1054 | cbp->ubl_vd = vd; | |
9ae529ec | 1055 | } |
b128c09f | 1056 | mutex_exit(&rio->io_lock); |
34dc7c2f BB |
1057 | } |
1058 | ||
1059 | zio_buf_free(zio->io_data, zio->io_size); | |
1060 | } | |
1061 | ||
9ae529ec CS |
1062 | static void |
1063 | vdev_uberblock_load_impl(zio_t *zio, vdev_t *vd, int flags, | |
1064 | struct ubl_cbdata *cbp) | |
34dc7c2f | 1065 | { |
d6320ddb | 1066 | int c, l, n; |
b128c09f | 1067 | |
d6320ddb | 1068 | for (c = 0; c < vd->vdev_children; c++) |
9ae529ec | 1069 | vdev_uberblock_load_impl(zio, vd->vdev_child[c], flags, cbp); |
34dc7c2f | 1070 | |
b128c09f | 1071 | if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) { |
d6320ddb BB |
1072 | for (l = 0; l < VDEV_LABELS; l++) { |
1073 | for (n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) { | |
b128c09f BB |
1074 | vdev_label_read(zio, vd, l, |
1075 | zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd)), | |
1076 | VDEV_UBERBLOCK_OFFSET(vd, n), | |
1077 | VDEV_UBERBLOCK_SIZE(vd), | |
1078 | vdev_uberblock_load_done, zio, flags); | |
1079 | } | |
34dc7c2f BB |
1080 | } |
1081 | } | |
9ae529ec CS |
1082 | } |
1083 | ||
1084 | /* | |
1085 | * Reads the 'best' uberblock from disk along with its associated | |
1086 | * configuration. First, we read the uberblock array of each label of each | |
1087 | * vdev, keeping track of the uberblock with the highest txg in each array. | |
3bc7e0fb | 1088 | * Then, we read the configuration from the same vdev as the best uberblock. |
9ae529ec CS |
1089 | */ |
1090 | void | |
1091 | vdev_uberblock_load(vdev_t *rvd, uberblock_t *ub, nvlist_t **config) | |
1092 | { | |
9ae529ec CS |
1093 | zio_t *zio; |
1094 | spa_t *spa = rvd->vdev_spa; | |
1095 | struct ubl_cbdata cb; | |
1096 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | | |
1097 | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD; | |
1098 | ||
1099 | ASSERT(ub); | |
1100 | ASSERT(config); | |
b128c09f | 1101 | |
9ae529ec CS |
1102 | bzero(ub, sizeof (uberblock_t)); |
1103 | *config = NULL; | |
1104 | ||
1105 | cb.ubl_ubbest = ub; | |
1106 | cb.ubl_vd = NULL; | |
1107 | ||
1108 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
1109 | zio = zio_root(spa, NULL, &cb, flags); | |
1110 | vdev_uberblock_load_impl(zio, rvd, flags, &cb); | |
1111 | (void) zio_wait(zio); | |
3bc7e0fb GW |
1112 | |
1113 | /* | |
1114 | * It's possible that the best uberblock was discovered on a label | |
1115 | * that has a configuration which was written in a future txg. | |
1116 | * Search all labels on this vdev to find the configuration that | |
1117 | * matches the txg for our uberblock. | |
1118 | */ | |
1119 | if (cb.ubl_vd != NULL) | |
1120 | *config = vdev_label_read_config(cb.ubl_vd, ub->ub_txg); | |
9ae529ec | 1121 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1122 | } |
1123 | ||
1124 | /* | |
1125 | * On success, increment root zio's count of good writes. | |
1126 | * We only get credit for writes to known-visible vdevs; see spa_vdev_add(). | |
1127 | */ | |
1128 | static void | |
1129 | vdev_uberblock_sync_done(zio_t *zio) | |
1130 | { | |
1131 | uint64_t *good_writes = zio->io_private; | |
1132 | ||
1133 | if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0) | |
bc89ac84 | 1134 | atomic_inc_64(good_writes); |
34dc7c2f BB |
1135 | } |
1136 | ||
1137 | /* | |
1138 | * Write the uberblock to all labels of all leaves of the specified vdev. | |
1139 | */ | |
1140 | static void | |
b128c09f | 1141 | vdev_uberblock_sync(zio_t *zio, uberblock_t *ub, vdev_t *vd, int flags) |
34dc7c2f | 1142 | { |
34dc7c2f | 1143 | uberblock_t *ubbuf; |
d6320ddb | 1144 | int c, l, n; |
34dc7c2f | 1145 | |
d6320ddb | 1146 | for (c = 0; c < vd->vdev_children; c++) |
b128c09f | 1147 | vdev_uberblock_sync(zio, ub, vd->vdev_child[c], flags); |
34dc7c2f BB |
1148 | |
1149 | if (!vd->vdev_ops->vdev_op_leaf) | |
1150 | return; | |
1151 | ||
b128c09f | 1152 | if (!vdev_writeable(vd)) |
34dc7c2f BB |
1153 | return; |
1154 | ||
1155 | n = ub->ub_txg & (VDEV_UBERBLOCK_COUNT(vd) - 1); | |
1156 | ||
1157 | ubbuf = zio_buf_alloc(VDEV_UBERBLOCK_SIZE(vd)); | |
1158 | bzero(ubbuf, VDEV_UBERBLOCK_SIZE(vd)); | |
1159 | *ubbuf = *ub; | |
1160 | ||
d6320ddb | 1161 | for (l = 0; l < VDEV_LABELS; l++) |
34dc7c2f | 1162 | vdev_label_write(zio, vd, l, ubbuf, |
b128c09f | 1163 | VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd), |
34dc7c2f | 1164 | vdev_uberblock_sync_done, zio->io_private, |
b128c09f | 1165 | flags | ZIO_FLAG_DONT_PROPAGATE); |
34dc7c2f BB |
1166 | |
1167 | zio_buf_free(ubbuf, VDEV_UBERBLOCK_SIZE(vd)); | |
1168 | } | |
1169 | ||
e49f1e20 | 1170 | /* Sync the uberblocks to all vdevs in svd[] */ |
34dc7c2f BB |
1171 | int |
1172 | vdev_uberblock_sync_list(vdev_t **svd, int svdcount, uberblock_t *ub, int flags) | |
1173 | { | |
1174 | spa_t *spa = svd[0]->vdev_spa; | |
34dc7c2f BB |
1175 | zio_t *zio; |
1176 | uint64_t good_writes = 0; | |
d6320ddb | 1177 | int v; |
34dc7c2f BB |
1178 | |
1179 | zio = zio_root(spa, NULL, &good_writes, flags); | |
1180 | ||
d6320ddb | 1181 | for (v = 0; v < svdcount; v++) |
b128c09f | 1182 | vdev_uberblock_sync(zio, ub, svd[v], flags); |
34dc7c2f BB |
1183 | |
1184 | (void) zio_wait(zio); | |
1185 | ||
1186 | /* | |
1187 | * Flush the uberblocks to disk. This ensures that the odd labels | |
1188 | * are no longer needed (because the new uberblocks and the even | |
1189 | * labels are safely on disk), so it is safe to overwrite them. | |
1190 | */ | |
1191 | zio = zio_root(spa, NULL, NULL, flags); | |
1192 | ||
d6320ddb | 1193 | for (v = 0; v < svdcount; v++) |
34dc7c2f BB |
1194 | zio_flush(zio, svd[v]); |
1195 | ||
1196 | (void) zio_wait(zio); | |
1197 | ||
1198 | return (good_writes >= 1 ? 0 : EIO); | |
1199 | } | |
1200 | ||
1201 | /* | |
1202 | * On success, increment the count of good writes for our top-level vdev. | |
1203 | */ | |
1204 | static void | |
1205 | vdev_label_sync_done(zio_t *zio) | |
1206 | { | |
1207 | uint64_t *good_writes = zio->io_private; | |
1208 | ||
1209 | if (zio->io_error == 0) | |
bc89ac84 | 1210 | atomic_inc_64(good_writes); |
34dc7c2f BB |
1211 | } |
1212 | ||
1213 | /* | |
1214 | * If there weren't enough good writes, indicate failure to the parent. | |
1215 | */ | |
1216 | static void | |
1217 | vdev_label_sync_top_done(zio_t *zio) | |
1218 | { | |
1219 | uint64_t *good_writes = zio->io_private; | |
1220 | ||
1221 | if (*good_writes == 0) | |
2e528b49 | 1222 | zio->io_error = SET_ERROR(EIO); |
34dc7c2f BB |
1223 | |
1224 | kmem_free(good_writes, sizeof (uint64_t)); | |
1225 | } | |
1226 | ||
b128c09f BB |
1227 | /* |
1228 | * We ignore errors for log and cache devices, simply free the private data. | |
1229 | */ | |
1230 | static void | |
1231 | vdev_label_sync_ignore_done(zio_t *zio) | |
1232 | { | |
1233 | kmem_free(zio->io_private, sizeof (uint64_t)); | |
1234 | } | |
1235 | ||
34dc7c2f BB |
1236 | /* |
1237 | * Write all even or odd labels to all leaves of the specified vdev. | |
1238 | */ | |
1239 | static void | |
b128c09f | 1240 | vdev_label_sync(zio_t *zio, vdev_t *vd, int l, uint64_t txg, int flags) |
34dc7c2f BB |
1241 | { |
1242 | nvlist_t *label; | |
1243 | vdev_phys_t *vp; | |
1244 | char *buf; | |
1245 | size_t buflen; | |
d6320ddb | 1246 | int c; |
34dc7c2f | 1247 | |
d6320ddb | 1248 | for (c = 0; c < vd->vdev_children; c++) |
b128c09f | 1249 | vdev_label_sync(zio, vd->vdev_child[c], l, txg, flags); |
34dc7c2f BB |
1250 | |
1251 | if (!vd->vdev_ops->vdev_op_leaf) | |
1252 | return; | |
1253 | ||
b128c09f | 1254 | if (!vdev_writeable(vd)) |
34dc7c2f BB |
1255 | return; |
1256 | ||
1257 | /* | |
1258 | * Generate a label describing the top-level config to which we belong. | |
1259 | */ | |
1260 | label = spa_config_generate(vd->vdev_spa, vd, txg, B_FALSE); | |
1261 | ||
1262 | vp = zio_buf_alloc(sizeof (vdev_phys_t)); | |
1263 | bzero(vp, sizeof (vdev_phys_t)); | |
1264 | ||
1265 | buf = vp->vp_nvlist; | |
1266 | buflen = sizeof (vp->vp_nvlist); | |
1267 | ||
79c76d5b | 1268 | if (!nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP)) { |
34dc7c2f BB |
1269 | for (; l < VDEV_LABELS; l += 2) { |
1270 | vdev_label_write(zio, vd, l, vp, | |
1271 | offsetof(vdev_label_t, vl_vdev_phys), | |
1272 | sizeof (vdev_phys_t), | |
1273 | vdev_label_sync_done, zio->io_private, | |
b128c09f | 1274 | flags | ZIO_FLAG_DONT_PROPAGATE); |
34dc7c2f BB |
1275 | } |
1276 | } | |
1277 | ||
1278 | zio_buf_free(vp, sizeof (vdev_phys_t)); | |
1279 | nvlist_free(label); | |
1280 | } | |
1281 | ||
1282 | int | |
b128c09f | 1283 | vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags) |
34dc7c2f | 1284 | { |
b128c09f | 1285 | list_t *dl = &spa->spa_config_dirty_list; |
34dc7c2f BB |
1286 | vdev_t *vd; |
1287 | zio_t *zio; | |
1288 | int error; | |
1289 | ||
1290 | /* | |
1291 | * Write the new labels to disk. | |
1292 | */ | |
1293 | zio = zio_root(spa, NULL, NULL, flags); | |
1294 | ||
1295 | for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) { | |
d6320ddb BB |
1296 | uint64_t *good_writes; |
1297 | zio_t *vio; | |
428870ff BB |
1298 | |
1299 | ASSERT(!vd->vdev_ishole); | |
1300 | ||
79c76d5b | 1301 | good_writes = kmem_zalloc(sizeof (uint64_t), KM_SLEEP); |
d6320ddb | 1302 | vio = zio_null(zio, spa, NULL, |
b128c09f BB |
1303 | (vd->vdev_islog || vd->vdev_aux != NULL) ? |
1304 | vdev_label_sync_ignore_done : vdev_label_sync_top_done, | |
34dc7c2f | 1305 | good_writes, flags); |
b128c09f | 1306 | vdev_label_sync(vio, vd, l, txg, flags); |
34dc7c2f BB |
1307 | zio_nowait(vio); |
1308 | } | |
1309 | ||
1310 | error = zio_wait(zio); | |
1311 | ||
1312 | /* | |
1313 | * Flush the new labels to disk. | |
1314 | */ | |
1315 | zio = zio_root(spa, NULL, NULL, flags); | |
1316 | ||
1317 | for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) | |
1318 | zio_flush(zio, vd); | |
1319 | ||
1320 | (void) zio_wait(zio); | |
1321 | ||
1322 | return (error); | |
1323 | } | |
1324 | ||
1325 | /* | |
1326 | * Sync the uberblock and any changes to the vdev configuration. | |
1327 | * | |
1328 | * The order of operations is carefully crafted to ensure that | |
1329 | * if the system panics or loses power at any time, the state on disk | |
1330 | * is still transactionally consistent. The in-line comments below | |
1331 | * describe the failure semantics at each stage. | |
1332 | * | |
1333 | * Moreover, vdev_config_sync() is designed to be idempotent: if it fails | |
1334 | * at any time, you can just call it again, and it will resume its work. | |
1335 | */ | |
1336 | int | |
b6fcb792 | 1337 | vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg) |
34dc7c2f BB |
1338 | { |
1339 | spa_t *spa = svd[0]->vdev_spa; | |
1340 | uberblock_t *ub = &spa->spa_uberblock; | |
1341 | vdev_t *vd; | |
1342 | zio_t *zio; | |
b6fcb792 | 1343 | int error = 0; |
b128c09f | 1344 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; |
34dc7c2f | 1345 | |
b6fcb792 | 1346 | retry: |
9babb374 BB |
1347 | /* |
1348 | * Normally, we don't want to try too hard to write every label and | |
1349 | * uberblock. If there is a flaky disk, we don't want the rest of the | |
1350 | * sync process to block while we retry. But if we can't write a | |
1351 | * single label out, we should retry with ZIO_FLAG_TRYHARD before | |
1352 | * bailing out and declaring the pool faulted. | |
1353 | */ | |
b6fcb792 BB |
1354 | if (error != 0) { |
1355 | if ((flags & ZIO_FLAG_TRYHARD) != 0) | |
1356 | return (error); | |
9babb374 | 1357 | flags |= ZIO_FLAG_TRYHARD; |
b6fcb792 | 1358 | } |
9babb374 | 1359 | |
34dc7c2f BB |
1360 | ASSERT(ub->ub_txg <= txg); |
1361 | ||
1362 | /* | |
1363 | * If this isn't a resync due to I/O errors, | |
1364 | * and nothing changed in this transaction group, | |
1365 | * and the vdev configuration hasn't changed, | |
1366 | * then there's nothing to do. | |
1367 | */ | |
1368 | if (ub->ub_txg < txg && | |
1369 | uberblock_update(ub, spa->spa_root_vdev, txg) == B_FALSE && | |
b128c09f | 1370 | list_is_empty(&spa->spa_config_dirty_list)) |
34dc7c2f BB |
1371 | return (0); |
1372 | ||
1373 | if (txg > spa_freeze_txg(spa)) | |
1374 | return (0); | |
1375 | ||
1376 | ASSERT(txg <= spa->spa_final_txg); | |
1377 | ||
1378 | /* | |
1379 | * Flush the write cache of every disk that's been written to | |
1380 | * in this transaction group. This ensures that all blocks | |
1381 | * written in this txg will be committed to stable storage | |
1382 | * before any uberblock that references them. | |
1383 | */ | |
1384 | zio = zio_root(spa, NULL, NULL, flags); | |
1385 | ||
1386 | for (vd = txg_list_head(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)); vd; | |
1387 | vd = txg_list_next(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg))) | |
1388 | zio_flush(zio, vd); | |
1389 | ||
1390 | (void) zio_wait(zio); | |
1391 | ||
1392 | /* | |
1393 | * Sync out the even labels (L0, L2) for every dirty vdev. If the | |
1394 | * system dies in the middle of this process, that's OK: all of the | |
1395 | * even labels that made it to disk will be newer than any uberblock, | |
1396 | * and will therefore be considered invalid. The odd labels (L1, L3), | |
1397 | * which have not yet been touched, will still be valid. We flush | |
1398 | * the new labels to disk to ensure that all even-label updates | |
1399 | * are committed to stable storage before the uberblock update. | |
1400 | */ | |
b128c09f | 1401 | if ((error = vdev_label_sync_list(spa, 0, txg, flags)) != 0) |
b6fcb792 | 1402 | goto retry; |
34dc7c2f BB |
1403 | |
1404 | /* | |
1405 | * Sync the uberblocks to all vdevs in svd[]. | |
1406 | * If the system dies in the middle of this step, there are two cases | |
1407 | * to consider, and the on-disk state is consistent either way: | |
1408 | * | |
1409 | * (1) If none of the new uberblocks made it to disk, then the | |
1410 | * previous uberblock will be the newest, and the odd labels | |
1411 | * (which had not yet been touched) will be valid with respect | |
1412 | * to that uberblock. | |
1413 | * | |
1414 | * (2) If one or more new uberblocks made it to disk, then they | |
1415 | * will be the newest, and the even labels (which had all | |
1416 | * been successfully committed) will be valid with respect | |
1417 | * to the new uberblocks. | |
1418 | */ | |
1419 | if ((error = vdev_uberblock_sync_list(svd, svdcount, ub, flags)) != 0) | |
b6fcb792 | 1420 | goto retry; |
34dc7c2f BB |
1421 | |
1422 | /* | |
1423 | * Sync out odd labels for every dirty vdev. If the system dies | |
1424 | * in the middle of this process, the even labels and the new | |
1425 | * uberblocks will suffice to open the pool. The next time | |
1426 | * the pool is opened, the first thing we'll do -- before any | |
1427 | * user data is modified -- is mark every vdev dirty so that | |
1428 | * all labels will be brought up to date. We flush the new labels | |
1429 | * to disk to ensure that all odd-label updates are committed to | |
1430 | * stable storage before the next transaction group begins. | |
1431 | */ | |
b6fcb792 BB |
1432 | if ((error = vdev_label_sync_list(spa, 1, txg, flags)) != 0) |
1433 | goto retry; | |
1434 | ||
1435 | return (0); | |
34dc7c2f | 1436 | } |