]>
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 | |
1d3ba0bf | 9 | * or https://opensource.org/licenses/CDDL-1.0. |
34dc7c2f BB |
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. |
108a454a | 24 | * Copyright (c) 2012, 2020 by Delphix. All rights reserved. |
cc99f275 | 25 | * Copyright (c) 2017, Intel Corporation. |
34dc7c2f BB |
26 | */ |
27 | ||
34dc7c2f BB |
28 | /* |
29 | * Virtual Device Labels | |
30 | * --------------------- | |
31 | * | |
32 | * The vdev label serves several distinct purposes: | |
33 | * | |
34 | * 1. Uniquely identify this device as part of a ZFS pool and confirm its | |
35 | * identity within the pool. | |
36 | * | |
0dc2f70c | 37 | * 2. Verify that all the devices given in a configuration are present |
34dc7c2f BB |
38 | * within the pool. |
39 | * | |
0dc2f70c | 40 | * 3. Determine the uberblock for the pool. |
34dc7c2f | 41 | * |
0dc2f70c | 42 | * 4. In case of an import operation, determine the configuration of the |
34dc7c2f BB |
43 | * toplevel vdev of which it is a part. |
44 | * | |
0dc2f70c | 45 | * 5. If an import operation cannot find all the devices in the pool, |
34dc7c2f BB |
46 | * provide enough information to the administrator to determine which |
47 | * devices are missing. | |
48 | * | |
49 | * It is important to note that while the kernel is responsible for writing the | |
50 | * label, it only consumes the information in the first three cases. The | |
51 | * latter information is only consumed in userland when determining the | |
52 | * configuration to import a pool. | |
53 | * | |
54 | * | |
55 | * Label Organization | |
56 | * ------------------ | |
57 | * | |
58 | * Before describing the contents of the label, it's important to understand how | |
59 | * the labels are written and updated with respect to the uberblock. | |
60 | * | |
61 | * When the pool configuration is altered, either because it was newly created | |
62 | * or a device was added, we want to update all the labels such that we can deal | |
63 | * with fatal failure at any point. To this end, each disk has two labels which | |
64 | * are updated before and after the uberblock is synced. Assuming we have | |
65 | * labels and an uberblock with the following transaction groups: | |
66 | * | |
67 | * L1 UB L2 | |
68 | * +------+ +------+ +------+ | |
69 | * | | | | | | | |
70 | * | t10 | | t10 | | t10 | | |
71 | * | | | | | | | |
72 | * +------+ +------+ +------+ | |
73 | * | |
74 | * In this stable state, the labels and the uberblock were all updated within | |
75 | * the same transaction group (10). Each label is mirrored and checksummed, so | |
76 | * that we can detect when we fail partway through writing the label. | |
77 | * | |
78 | * In order to identify which labels are valid, the labels are written in the | |
79 | * following manner: | |
80 | * | |
0dc2f70c MA |
81 | * 1. For each vdev, update 'L1' to the new label |
82 | * 2. Update the uberblock | |
83 | * 3. For each vdev, update 'L2' to the new label | |
34dc7c2f BB |
84 | * |
85 | * Given arbitrary failure, we can determine the correct label to use based on | |
86 | * the transaction group. If we fail after updating L1 but before updating the | |
87 | * UB, we will notice that L1's transaction group is greater than the uberblock, | |
88 | * so L2 must be valid. If we fail after writing the uberblock but before | |
89 | * writing L2, we will notice that L2's transaction group is less than L1, and | |
90 | * therefore L1 is valid. | |
91 | * | |
92 | * Another added complexity is that not every label is updated when the config | |
93 | * is synced. If we add a single device, we do not want to have to re-write | |
94 | * every label for every device in the pool. This means that both L1 and L2 may | |
95 | * be older than the pool uberblock, because the necessary information is stored | |
96 | * on another vdev. | |
97 | * | |
98 | * | |
99 | * On-disk Format | |
100 | * -------------- | |
101 | * | |
102 | * The vdev label consists of two distinct parts, and is wrapped within the | |
103 | * vdev_label_t structure. The label includes 8k of padding to permit legacy | |
104 | * VTOC disk labels, but is otherwise ignored. | |
105 | * | |
106 | * The first half of the label is a packed nvlist which contains pool wide | |
107 | * properties, per-vdev properties, and configuration information. It is | |
108 | * described in more detail below. | |
109 | * | |
110 | * The latter half of the label consists of a redundant array of uberblocks. | |
111 | * These uberblocks are updated whenever a transaction group is committed, | |
112 | * or when the configuration is updated. When a pool is loaded, we scan each | |
113 | * vdev for the 'best' uberblock. | |
114 | * | |
115 | * | |
116 | * Configuration Information | |
117 | * ------------------------- | |
118 | * | |
119 | * The nvlist describing the pool and vdev contains the following elements: | |
120 | * | |
0dc2f70c MA |
121 | * version ZFS on-disk version |
122 | * name Pool name | |
123 | * state Pool state | |
124 | * txg Transaction group in which this label was written | |
125 | * pool_guid Unique identifier for this pool | |
126 | * vdev_tree An nvlist describing vdev tree. | |
9ae529ec CS |
127 | * features_for_read |
128 | * An nvlist of the features necessary for reading the MOS. | |
34dc7c2f BB |
129 | * |
130 | * Each leaf device label also contains the following: | |
131 | * | |
0dc2f70c MA |
132 | * top_guid Unique ID for top-level vdev in which this is contained |
133 | * guid Unique ID for the leaf vdev | |
34dc7c2f BB |
134 | * |
135 | * The 'vs' configuration follows the format described in 'spa_config.c'. | |
136 | */ | |
137 | ||
138 | #include <sys/zfs_context.h> | |
139 | #include <sys/spa.h> | |
140 | #include <sys/spa_impl.h> | |
141 | #include <sys/dmu.h> | |
142 | #include <sys/zap.h> | |
143 | #include <sys/vdev.h> | |
144 | #include <sys/vdev_impl.h> | |
5caeef02 | 145 | #include <sys/vdev_raidz.h> |
b2255edc | 146 | #include <sys/vdev_draid.h> |
34dc7c2f BB |
147 | #include <sys/uberblock_impl.h> |
148 | #include <sys/metaslab.h> | |
a1d477c2 | 149 | #include <sys/metaslab_impl.h> |
34dc7c2f | 150 | #include <sys/zio.h> |
428870ff | 151 | #include <sys/dsl_scan.h> |
a6255b7f | 152 | #include <sys/abd.h> |
34dc7c2f | 153 | #include <sys/fs/zfs.h> |
1db9e6e4 TS |
154 | #include <sys/byteorder.h> |
155 | #include <sys/zfs_bootenv.h> | |
34dc7c2f BB |
156 | |
157 | /* | |
158 | * Basic routines to read and write from a vdev label. | |
159 | * Used throughout the rest of this file. | |
160 | */ | |
161 | uint64_t | |
162 | vdev_label_offset(uint64_t psize, int l, uint64_t offset) | |
163 | { | |
164 | ASSERT(offset < sizeof (vdev_label_t)); | |
165 | ASSERT(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t) == 0); | |
166 | ||
167 | return (offset + l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? | |
168 | 0 : psize - VDEV_LABELS * sizeof (vdev_label_t))); | |
169 | } | |
170 | ||
b128c09f BB |
171 | /* |
172 | * Returns back the vdev label associated with the passed in offset. | |
173 | */ | |
174 | int | |
175 | vdev_label_number(uint64_t psize, uint64_t offset) | |
176 | { | |
177 | int l; | |
178 | ||
179 | if (offset >= psize - VDEV_LABEL_END_SIZE) { | |
180 | offset -= psize - VDEV_LABEL_END_SIZE; | |
181 | offset += (VDEV_LABELS / 2) * sizeof (vdev_label_t); | |
182 | } | |
183 | l = offset / sizeof (vdev_label_t); | |
184 | return (l < VDEV_LABELS ? l : -1); | |
185 | } | |
186 | ||
34dc7c2f | 187 | static void |
a6255b7f | 188 | vdev_label_read(zio_t *zio, vdev_t *vd, int l, abd_t *buf, uint64_t offset, |
e9aa730c | 189 | uint64_t size, zio_done_func_t *done, void *private, int flags) |
34dc7c2f | 190 | { |
0091d66f OF |
191 | ASSERT( |
192 | spa_config_held(zio->io_spa, SCL_STATE, RW_READER) == SCL_STATE || | |
193 | spa_config_held(zio->io_spa, SCL_STATE, RW_WRITER) == SCL_STATE); | |
b128c09f | 194 | ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); |
34dc7c2f BB |
195 | |
196 | zio_nowait(zio_read_phys(zio, vd, | |
197 | vdev_label_offset(vd->vdev_psize, l, offset), | |
198 | size, buf, ZIO_CHECKSUM_LABEL, done, private, | |
b128c09f | 199 | ZIO_PRIORITY_SYNC_READ, flags, B_TRUE)); |
34dc7c2f BB |
200 | } |
201 | ||
379ca9cf | 202 | void |
a6255b7f | 203 | vdev_label_write(zio_t *zio, vdev_t *vd, int l, abd_t *buf, uint64_t offset, |
e9aa730c | 204 | uint64_t size, zio_done_func_t *done, void *private, int flags) |
34dc7c2f | 205 | { |
0091d66f OF |
206 | ASSERT( |
207 | spa_config_held(zio->io_spa, SCL_STATE, RW_READER) == SCL_STATE || | |
208 | spa_config_held(zio->io_spa, SCL_STATE, RW_WRITER) == SCL_STATE); | |
b128c09f | 209 | ASSERT(flags & ZIO_FLAG_CONFIG_WRITER); |
34dc7c2f BB |
210 | |
211 | zio_nowait(zio_write_phys(zio, vd, | |
212 | vdev_label_offset(vd->vdev_psize, l, offset), | |
213 | size, buf, ZIO_CHECKSUM_LABEL, done, private, | |
214 | ZIO_PRIORITY_SYNC_WRITE, flags, B_TRUE)); | |
215 | } | |
216 | ||
193a37cb TH |
217 | /* |
218 | * Generate the nvlist representing this vdev's stats | |
219 | */ | |
220 | void | |
221 | vdev_config_generate_stats(vdev_t *vd, nvlist_t *nv) | |
222 | { | |
223 | nvlist_t *nvx; | |
224 | vdev_stat_t *vs; | |
225 | vdev_stat_ex_t *vsx; | |
226 | ||
227 | vs = kmem_alloc(sizeof (*vs), KM_SLEEP); | |
228 | vsx = kmem_alloc(sizeof (*vsx), KM_SLEEP); | |
229 | ||
230 | vdev_get_stats_ex(vd, vs, vsx); | |
231 | fnvlist_add_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, | |
232 | (uint64_t *)vs, sizeof (*vs) / sizeof (uint64_t)); | |
233 | ||
193a37cb TH |
234 | /* |
235 | * Add extended stats into a special extended stats nvlist. This keeps | |
236 | * all the extended stats nicely grouped together. The extended stats | |
237 | * nvlist is then added to the main nvlist. | |
238 | */ | |
239 | nvx = fnvlist_alloc(); | |
240 | ||
241 | /* ZIOs in flight to disk */ | |
242 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE, | |
243 | vsx->vsx_active_queue[ZIO_PRIORITY_SYNC_READ]); | |
244 | ||
245 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE, | |
246 | vsx->vsx_active_queue[ZIO_PRIORITY_SYNC_WRITE]); | |
247 | ||
248 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE, | |
249 | vsx->vsx_active_queue[ZIO_PRIORITY_ASYNC_READ]); | |
250 | ||
251 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE, | |
252 | vsx->vsx_active_queue[ZIO_PRIORITY_ASYNC_WRITE]); | |
253 | ||
254 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE, | |
255 | vsx->vsx_active_queue[ZIO_PRIORITY_SCRUB]); | |
256 | ||
1b939560 BB |
257 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE, |
258 | vsx->vsx_active_queue[ZIO_PRIORITY_TRIM]); | |
259 | ||
00888c08 TB |
260 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE, |
261 | vsx->vsx_active_queue[ZIO_PRIORITY_REBUILD]); | |
262 | ||
193a37cb TH |
263 | /* ZIOs pending */ |
264 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE, | |
265 | vsx->vsx_pend_queue[ZIO_PRIORITY_SYNC_READ]); | |
266 | ||
267 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE, | |
268 | vsx->vsx_pend_queue[ZIO_PRIORITY_SYNC_WRITE]); | |
269 | ||
270 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE, | |
271 | vsx->vsx_pend_queue[ZIO_PRIORITY_ASYNC_READ]); | |
272 | ||
273 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE, | |
274 | vsx->vsx_pend_queue[ZIO_PRIORITY_ASYNC_WRITE]); | |
275 | ||
276 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE, | |
277 | vsx->vsx_pend_queue[ZIO_PRIORITY_SCRUB]); | |
278 | ||
1b939560 BB |
279 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_TRIM_PEND_QUEUE, |
280 | vsx->vsx_pend_queue[ZIO_PRIORITY_TRIM]); | |
281 | ||
00888c08 TB |
282 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_REBUILD_PEND_QUEUE, |
283 | vsx->vsx_pend_queue[ZIO_PRIORITY_REBUILD]); | |
284 | ||
193a37cb TH |
285 | /* Histograms */ |
286 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO, | |
287 | vsx->vsx_total_histo[ZIO_TYPE_READ], | |
288 | ARRAY_SIZE(vsx->vsx_total_histo[ZIO_TYPE_READ])); | |
289 | ||
290 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO, | |
291 | vsx->vsx_total_histo[ZIO_TYPE_WRITE], | |
292 | ARRAY_SIZE(vsx->vsx_total_histo[ZIO_TYPE_WRITE])); | |
293 | ||
294 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO, | |
295 | vsx->vsx_disk_histo[ZIO_TYPE_READ], | |
296 | ARRAY_SIZE(vsx->vsx_disk_histo[ZIO_TYPE_READ])); | |
297 | ||
298 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO, | |
299 | vsx->vsx_disk_histo[ZIO_TYPE_WRITE], | |
300 | ARRAY_SIZE(vsx->vsx_disk_histo[ZIO_TYPE_WRITE])); | |
301 | ||
302 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO, | |
303 | vsx->vsx_queue_histo[ZIO_PRIORITY_SYNC_READ], | |
304 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_SYNC_READ])); | |
305 | ||
306 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO, | |
307 | vsx->vsx_queue_histo[ZIO_PRIORITY_SYNC_WRITE], | |
308 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_SYNC_WRITE])); | |
309 | ||
310 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO, | |
311 | vsx->vsx_queue_histo[ZIO_PRIORITY_ASYNC_READ], | |
312 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_ASYNC_READ])); | |
313 | ||
314 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO, | |
315 | vsx->vsx_queue_histo[ZIO_PRIORITY_ASYNC_WRITE], | |
316 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_ASYNC_WRITE])); | |
317 | ||
318 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO, | |
319 | vsx->vsx_queue_histo[ZIO_PRIORITY_SCRUB], | |
320 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_SCRUB])); | |
321 | ||
1b939560 BB |
322 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO, |
323 | vsx->vsx_queue_histo[ZIO_PRIORITY_TRIM], | |
324 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_TRIM])); | |
325 | ||
00888c08 TB |
326 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO, |
327 | vsx->vsx_queue_histo[ZIO_PRIORITY_REBUILD], | |
328 | ARRAY_SIZE(vsx->vsx_queue_histo[ZIO_PRIORITY_REBUILD])); | |
329 | ||
7e945072 TH |
330 | /* Request sizes */ |
331 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO, | |
332 | vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_READ], | |
333 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_READ])); | |
334 | ||
335 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO, | |
336 | vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_WRITE], | |
337 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_SYNC_WRITE])); | |
338 | ||
339 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO, | |
340 | vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_READ], | |
341 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_READ])); | |
342 | ||
343 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO, | |
344 | vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_WRITE], | |
345 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_ASYNC_WRITE])); | |
346 | ||
347 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO, | |
348 | vsx->vsx_ind_histo[ZIO_PRIORITY_SCRUB], | |
349 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_SCRUB])); | |
350 | ||
1b939560 BB |
351 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO, |
352 | vsx->vsx_ind_histo[ZIO_PRIORITY_TRIM], | |
353 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_TRIM])); | |
354 | ||
00888c08 TB |
355 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_IND_REBUILD_HISTO, |
356 | vsx->vsx_ind_histo[ZIO_PRIORITY_REBUILD], | |
357 | ARRAY_SIZE(vsx->vsx_ind_histo[ZIO_PRIORITY_REBUILD])); | |
358 | ||
7e945072 TH |
359 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO, |
360 | vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_READ], | |
361 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_READ])); | |
362 | ||
363 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO, | |
364 | vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_WRITE], | |
365 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_SYNC_WRITE])); | |
366 | ||
367 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO, | |
368 | vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_READ], | |
369 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_READ])); | |
370 | ||
371 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO, | |
372 | vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_WRITE], | |
373 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_ASYNC_WRITE])); | |
374 | ||
375 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO, | |
376 | vsx->vsx_agg_histo[ZIO_PRIORITY_SCRUB], | |
377 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_SCRUB])); | |
378 | ||
1b939560 BB |
379 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_AGG_TRIM_HISTO, |
380 | vsx->vsx_agg_histo[ZIO_PRIORITY_TRIM], | |
381 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_TRIM])); | |
382 | ||
00888c08 TB |
383 | fnvlist_add_uint64_array(nvx, ZPOOL_CONFIG_VDEV_AGG_REBUILD_HISTO, |
384 | vsx->vsx_agg_histo[ZIO_PRIORITY_REBUILD], | |
385 | ARRAY_SIZE(vsx->vsx_agg_histo[ZIO_PRIORITY_REBUILD])); | |
386 | ||
ad796b8a TH |
387 | /* IO delays */ |
388 | fnvlist_add_uint64(nvx, ZPOOL_CONFIG_VDEV_SLOW_IOS, vs->vs_slow_ios); | |
389 | ||
193a37cb TH |
390 | /* Add extended stats nvlist to main nvlist */ |
391 | fnvlist_add_nvlist(nv, ZPOOL_CONFIG_VDEV_STATS_EX, nvx); | |
392 | ||
6a796725 | 393 | fnvlist_free(nvx); |
c8fd652c | 394 | kmem_free(vs, sizeof (*vs)); |
193a37cb TH |
395 | kmem_free(vsx, sizeof (*vsx)); |
396 | } | |
397 | ||
d2734cce SD |
398 | static void |
399 | root_vdev_actions_getprogress(vdev_t *vd, nvlist_t *nvl) | |
400 | { | |
401 | spa_t *spa = vd->vdev_spa; | |
402 | ||
403 | if (vd != spa->spa_root_vdev) | |
404 | return; | |
405 | ||
406 | /* provide either current or previous scan information */ | |
407 | pool_scan_stat_t ps; | |
408 | if (spa_scan_get_stats(spa, &ps) == 0) { | |
409 | fnvlist_add_uint64_array(nvl, | |
410 | ZPOOL_CONFIG_SCAN_STATS, (uint64_t *)&ps, | |
411 | sizeof (pool_scan_stat_t) / sizeof (uint64_t)); | |
412 | } | |
413 | ||
414 | pool_removal_stat_t prs; | |
415 | if (spa_removal_get_stats(spa, &prs) == 0) { | |
416 | fnvlist_add_uint64_array(nvl, | |
417 | ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t *)&prs, | |
418 | sizeof (prs) / sizeof (uint64_t)); | |
419 | } | |
420 | ||
421 | pool_checkpoint_stat_t pcs; | |
422 | if (spa_checkpoint_get_stats(spa, &pcs) == 0) { | |
423 | fnvlist_add_uint64_array(nvl, | |
424 | ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t *)&pcs, | |
425 | sizeof (pcs) / sizeof (uint64_t)); | |
426 | } | |
5caeef02 DB |
427 | |
428 | pool_raidz_expand_stat_t pres; | |
429 | if (spa_raidz_expand_get_stats(spa, &pres) == 0) { | |
430 | fnvlist_add_uint64_array(nvl, | |
431 | ZPOOL_CONFIG_RAIDZ_EXPAND_STATS, (uint64_t *)&pres, | |
432 | sizeof (pres) / sizeof (uint64_t)); | |
433 | } | |
d2734cce SD |
434 | } |
435 | ||
9a49d3f3 BB |
436 | static void |
437 | top_vdev_actions_getprogress(vdev_t *vd, nvlist_t *nvl) | |
438 | { | |
439 | if (vd == vd->vdev_top) { | |
440 | vdev_rebuild_stat_t vrs; | |
441 | if (vdev_rebuild_get_stats(vd, &vrs) == 0) { | |
442 | fnvlist_add_uint64_array(nvl, | |
443 | ZPOOL_CONFIG_REBUILD_STATS, (uint64_t *)&vrs, | |
444 | sizeof (vrs) / sizeof (uint64_t)); | |
445 | } | |
446 | } | |
447 | } | |
448 | ||
34dc7c2f BB |
449 | /* |
450 | * Generate the nvlist representing this vdev's config. | |
451 | */ | |
452 | nvlist_t * | |
453 | vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats, | |
428870ff | 454 | vdev_config_flag_t flags) |
34dc7c2f BB |
455 | { |
456 | nvlist_t *nv = NULL; | |
a1d477c2 MA |
457 | vdev_indirect_config_t *vic = &vd->vdev_indirect_config; |
458 | ||
79c76d5b | 459 | nv = fnvlist_alloc(); |
34dc7c2f | 460 | |
5d1f7fb6 | 461 | fnvlist_add_string(nv, ZPOOL_CONFIG_TYPE, vd->vdev_ops->vdev_op_type); |
428870ff | 462 | if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE))) |
5d1f7fb6 GW |
463 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ID, vd->vdev_id); |
464 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_GUID, vd->vdev_guid); | |
34dc7c2f BB |
465 | |
466 | if (vd->vdev_path != NULL) | |
5d1f7fb6 | 467 | fnvlist_add_string(nv, ZPOOL_CONFIG_PATH, vd->vdev_path); |
34dc7c2f BB |
468 | |
469 | if (vd->vdev_devid != NULL) | |
5d1f7fb6 | 470 | fnvlist_add_string(nv, ZPOOL_CONFIG_DEVID, vd->vdev_devid); |
34dc7c2f BB |
471 | |
472 | if (vd->vdev_physpath != NULL) | |
5d1f7fb6 GW |
473 | fnvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH, |
474 | vd->vdev_physpath); | |
34dc7c2f | 475 | |
1bbd8770 TH |
476 | if (vd->vdev_enc_sysfs_path != NULL) |
477 | fnvlist_add_string(nv, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH, | |
478 | vd->vdev_enc_sysfs_path); | |
479 | ||
9babb374 | 480 | if (vd->vdev_fru != NULL) |
5d1f7fb6 | 481 | fnvlist_add_string(nv, ZPOOL_CONFIG_FRU, vd->vdev_fru); |
9babb374 | 482 | |
b2255edc BB |
483 | if (vd->vdev_ops->vdev_op_config_generate != NULL) |
484 | vd->vdev_ops->vdev_op_config_generate(vd, nv); | |
34dc7c2f | 485 | |
b2255edc | 486 | if (vd->vdev_wholedisk != -1ULL) { |
5d1f7fb6 GW |
487 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, |
488 | vd->vdev_wholedisk); | |
b2255edc | 489 | } |
34dc7c2f | 490 | |
6cb8e530 | 491 | if (vd->vdev_not_present && !(flags & VDEV_CONFIG_MISSING)) |
5d1f7fb6 | 492 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1); |
34dc7c2f BB |
493 | |
494 | if (vd->vdev_isspare) | |
5d1f7fb6 | 495 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1); |
34dc7c2f | 496 | |
8af1104f GA |
497 | if (flags & VDEV_CONFIG_L2CACHE) |
498 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT, vd->vdev_ashift); | |
499 | ||
428870ff BB |
500 | if (!(flags & (VDEV_CONFIG_SPARE | VDEV_CONFIG_L2CACHE)) && |
501 | vd == vd->vdev_top) { | |
5d1f7fb6 GW |
502 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY, |
503 | vd->vdev_ms_array); | |
504 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT, | |
505 | vd->vdev_ms_shift); | |
506 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASHIFT, vd->vdev_ashift); | |
507 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ASIZE, | |
508 | vd->vdev_asize); | |
509 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_LOG, vd->vdev_islog); | |
2a673e76 AJ |
510 | if (vd->vdev_noalloc) { |
511 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_NONALLOCATING, | |
512 | vd->vdev_noalloc); | |
513 | } | |
460d887c GW |
514 | |
515 | /* | |
516 | * Slog devices are removed synchronously so don't | |
517 | * persist the vdev_removing flag to the label. | |
518 | */ | |
519 | if (vd->vdev_removing && !vd->vdev_islog) { | |
5d1f7fb6 GW |
520 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVING, |
521 | vd->vdev_removing); | |
a1d477c2 | 522 | } |
cc99f275 DB |
523 | |
524 | /* zpool command expects alloc class data */ | |
525 | if (getstats && vd->vdev_alloc_bias != VDEV_BIAS_NONE) { | |
526 | const char *bias = NULL; | |
527 | ||
528 | switch (vd->vdev_alloc_bias) { | |
529 | case VDEV_BIAS_LOG: | |
530 | bias = VDEV_ALLOC_BIAS_LOG; | |
531 | break; | |
532 | case VDEV_BIAS_SPECIAL: | |
533 | bias = VDEV_ALLOC_BIAS_SPECIAL; | |
534 | break; | |
535 | case VDEV_BIAS_DEDUP: | |
536 | bias = VDEV_ALLOC_BIAS_DEDUP; | |
537 | break; | |
538 | default: | |
539 | ASSERT3U(vd->vdev_alloc_bias, ==, | |
540 | VDEV_BIAS_NONE); | |
541 | } | |
542 | fnvlist_add_string(nv, ZPOOL_CONFIG_ALLOCATION_BIAS, | |
543 | bias); | |
544 | } | |
34dc7c2f BB |
545 | } |
546 | ||
93cf2076 | 547 | if (vd->vdev_dtl_sm != NULL) { |
5d1f7fb6 | 548 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_DTL, |
93cf2076 GW |
549 | space_map_object(vd->vdev_dtl_sm)); |
550 | } | |
34dc7c2f | 551 | |
a1d477c2 MA |
552 | if (vic->vic_mapping_object != 0) { |
553 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_OBJECT, | |
554 | vic->vic_mapping_object); | |
555 | } | |
556 | ||
557 | if (vic->vic_births_object != 0) { | |
558 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_BIRTHS, | |
559 | vic->vic_births_object); | |
560 | } | |
561 | ||
562 | if (vic->vic_prev_indirect_vdev != UINT64_MAX) { | |
563 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_PREV_INDIRECT_VDEV, | |
564 | vic->vic_prev_indirect_vdev); | |
565 | } | |
566 | ||
428870ff | 567 | if (vd->vdev_crtxg) |
5d1f7fb6 | 568 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_CREATE_TXG, vd->vdev_crtxg); |
428870ff | 569 | |
d48091de | 570 | if (vd->vdev_expansion_time) |
571 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_EXPANSION_TIME, | |
572 | vd->vdev_expansion_time); | |
573 | ||
e0ab3ab5 JS |
574 | if (flags & VDEV_CONFIG_MOS) { |
575 | if (vd->vdev_leaf_zap != 0) { | |
576 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
577 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_LEAF_ZAP, | |
578 | vd->vdev_leaf_zap); | |
579 | } | |
580 | ||
581 | if (vd->vdev_top_zap != 0) { | |
582 | ASSERT(vd == vd->vdev_top); | |
583 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP, | |
584 | vd->vdev_top_zap); | |
585 | } | |
80a91e74 | 586 | |
3e4ed421 RW |
587 | if (vd->vdev_ops == &vdev_root_ops && vd->vdev_root_zap != 0 && |
588 | spa_feature_is_active(vd->vdev_spa, SPA_FEATURE_AVZ_V2)) { | |
589 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_VDEV_ROOT_ZAP, | |
590 | vd->vdev_root_zap); | |
591 | } | |
592 | ||
80a91e74 TC |
593 | if (vd->vdev_resilver_deferred) { |
594 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
595 | ASSERT(spa->spa_resilver_deferred); | |
596 | fnvlist_add_boolean(nv, ZPOOL_CONFIG_RESILVER_DEFER); | |
597 | } | |
e0ab3ab5 JS |
598 | } |
599 | ||
34dc7c2f | 600 | if (getstats) { |
193a37cb | 601 | vdev_config_generate_stats(vd, nv); |
428870ff | 602 | |
d2734cce | 603 | root_vdev_actions_getprogress(vd, nv); |
9a49d3f3 | 604 | top_vdev_actions_getprogress(vd, nv); |
a1d477c2 MA |
605 | |
606 | /* | |
607 | * Note: this can be called from open context | |
608 | * (spa_get_stats()), so we need the rwlock to prevent | |
609 | * the mapping from being changed by condensing. | |
610 | */ | |
611 | rw_enter(&vd->vdev_indirect_rwlock, RW_READER); | |
612 | if (vd->vdev_indirect_mapping != NULL) { | |
613 | ASSERT(vd->vdev_indirect_births != NULL); | |
614 | vdev_indirect_mapping_t *vim = | |
615 | vd->vdev_indirect_mapping; | |
616 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_SIZE, | |
617 | vdev_indirect_mapping_size(vim)); | |
618 | } | |
619 | rw_exit(&vd->vdev_indirect_rwlock); | |
620 | if (vd->vdev_mg != NULL && | |
621 | vd->vdev_mg->mg_fragmentation != ZFS_FRAG_INVALID) { | |
622 | /* | |
623 | * Compute approximately how much memory would be used | |
624 | * for the indirect mapping if this device were to | |
625 | * be removed. | |
626 | * | |
627 | * Note: If the frag metric is invalid, then not | |
628 | * enough metaslabs have been converted to have | |
629 | * histograms. | |
630 | */ | |
631 | uint64_t seg_count = 0; | |
0dc2f70c | 632 | uint64_t to_alloc = vd->vdev_stat.vs_alloc; |
a1d477c2 MA |
633 | |
634 | /* | |
635 | * There are the same number of allocated segments | |
636 | * as free segments, so we will have at least one | |
0dc2f70c MA |
637 | * entry per free segment. However, small free |
638 | * segments (smaller than vdev_removal_max_span) | |
639 | * will be combined with adjacent allocated segments | |
640 | * as a single mapping. | |
a1d477c2 MA |
641 | */ |
642 | for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) { | |
a8fc1b87 CT |
643 | if (i + 1 < highbit64(vdev_removal_max_span) |
644 | - 1) { | |
0dc2f70c MA |
645 | to_alloc += |
646 | vd->vdev_mg->mg_histogram[i] << | |
647 | (i + 1); | |
648 | } else { | |
649 | seg_count += | |
650 | vd->vdev_mg->mg_histogram[i]; | |
651 | } | |
a1d477c2 MA |
652 | } |
653 | ||
654 | /* | |
0dc2f70c MA |
655 | * The maximum length of a mapping is |
656 | * zfs_remove_max_segment, so we need at least one entry | |
657 | * per zfs_remove_max_segment of allocated data. | |
a1d477c2 | 658 | */ |
53dce5ac | 659 | seg_count += to_alloc / spa_remove_max_segment(spa); |
a1d477c2 MA |
660 | |
661 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_INDIRECT_SIZE, | |
662 | seg_count * | |
663 | sizeof (vdev_indirect_mapping_entry_phys_t)); | |
664 | } | |
34dc7c2f BB |
665 | } |
666 | ||
667 | if (!vd->vdev_ops->vdev_op_leaf) { | |
668 | nvlist_t **child; | |
460d887c | 669 | uint64_t c; |
428870ff BB |
670 | |
671 | ASSERT(!vd->vdev_ishole); | |
34dc7c2f BB |
672 | |
673 | child = kmem_alloc(vd->vdev_children * sizeof (nvlist_t *), | |
79c76d5b | 674 | KM_SLEEP); |
34dc7c2f | 675 | |
460d887c GW |
676 | for (c = 0; c < vd->vdev_children; c++) { |
677 | child[c] = vdev_config_generate(spa, vd->vdev_child[c], | |
428870ff BB |
678 | getstats, flags); |
679 | } | |
680 | ||
460d887c GW |
681 | fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, |
682 | (const nvlist_t * const *)child, vd->vdev_children); | |
428870ff | 683 | |
460d887c | 684 | for (c = 0; c < vd->vdev_children; c++) |
34dc7c2f BB |
685 | nvlist_free(child[c]); |
686 | ||
687 | kmem_free(child, vd->vdev_children * sizeof (nvlist_t *)); | |
688 | ||
689 | } else { | |
428870ff BB |
690 | const char *aux = NULL; |
691 | ||
34dc7c2f | 692 | if (vd->vdev_offline && !vd->vdev_tmpoffline) |
5d1f7fb6 GW |
693 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_OFFLINE, B_TRUE); |
694 | if (vd->vdev_resilver_txg != 0) | |
695 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_RESILVER_TXG, | |
696 | vd->vdev_resilver_txg); | |
9a49d3f3 BB |
697 | if (vd->vdev_rebuild_txg != 0) |
698 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_REBUILD_TXG, | |
699 | vd->vdev_rebuild_txg); | |
34dc7c2f | 700 | if (vd->vdev_faulted) |
5d1f7fb6 | 701 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_FAULTED, B_TRUE); |
34dc7c2f | 702 | if (vd->vdev_degraded) |
5d1f7fb6 | 703 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_DEGRADED, B_TRUE); |
34dc7c2f | 704 | if (vd->vdev_removed) |
5d1f7fb6 | 705 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_REMOVED, B_TRUE); |
34dc7c2f | 706 | if (vd->vdev_unspare) |
5d1f7fb6 | 707 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_UNSPARE, B_TRUE); |
428870ff | 708 | if (vd->vdev_ishole) |
5d1f7fb6 | 709 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_IS_HOLE, B_TRUE); |
428870ff | 710 | |
4a283c7f | 711 | /* Set the reason why we're FAULTED/DEGRADED. */ |
428870ff BB |
712 | switch (vd->vdev_stat.vs_aux) { |
713 | case VDEV_AUX_ERR_EXCEEDED: | |
714 | aux = "err_exceeded"; | |
715 | break; | |
716 | ||
717 | case VDEV_AUX_EXTERNAL: | |
718 | aux = "external"; | |
719 | break; | |
720 | } | |
721 | ||
4a283c7f | 722 | if (aux != NULL && !vd->vdev_tmpoffline) { |
5d1f7fb6 | 723 | fnvlist_add_string(nv, ZPOOL_CONFIG_AUX_STATE, aux); |
4a283c7f TH |
724 | } else { |
725 | /* | |
726 | * We're healthy - clear any previous AUX_STATE values. | |
727 | */ | |
728 | if (nvlist_exists(nv, ZPOOL_CONFIG_AUX_STATE)) | |
729 | nvlist_remove_all(nv, ZPOOL_CONFIG_AUX_STATE); | |
730 | } | |
428870ff BB |
731 | |
732 | if (vd->vdev_splitting && vd->vdev_orig_guid != 0LL) { | |
5d1f7fb6 GW |
733 | fnvlist_add_uint64(nv, ZPOOL_CONFIG_ORIG_GUID, |
734 | vd->vdev_orig_guid); | |
428870ff | 735 | } |
34dc7c2f BB |
736 | } |
737 | ||
738 | return (nv); | |
739 | } | |
740 | ||
428870ff BB |
741 | /* |
742 | * Generate a view of the top-level vdevs. If we currently have holes | |
743 | * in the namespace, then generate an array which contains a list of holey | |
744 | * vdevs. Additionally, add the number of top-level children that currently | |
745 | * exist. | |
746 | */ | |
747 | void | |
748 | vdev_top_config_generate(spa_t *spa, nvlist_t *config) | |
749 | { | |
750 | vdev_t *rvd = spa->spa_root_vdev; | |
751 | uint64_t *array; | |
752 | uint_t c, idx; | |
753 | ||
79c76d5b | 754 | array = kmem_alloc(rvd->vdev_children * sizeof (uint64_t), KM_SLEEP); |
428870ff BB |
755 | |
756 | for (c = 0, idx = 0; c < rvd->vdev_children; c++) { | |
757 | vdev_t *tvd = rvd->vdev_child[c]; | |
758 | ||
a1d477c2 | 759 | if (tvd->vdev_ishole) { |
428870ff | 760 | array[idx++] = c; |
a1d477c2 | 761 | } |
428870ff BB |
762 | } |
763 | ||
764 | if (idx) { | |
765 | VERIFY(nvlist_add_uint64_array(config, ZPOOL_CONFIG_HOLE_ARRAY, | |
766 | array, idx) == 0); | |
767 | } | |
768 | ||
769 | VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN, | |
770 | rvd->vdev_children) == 0); | |
771 | ||
772 | kmem_free(array, rvd->vdev_children * sizeof (uint64_t)); | |
773 | } | |
774 | ||
9ae529ec | 775 | /* |
3bc7e0fb GW |
776 | * Returns the configuration from the label of the given vdev. For vdevs |
777 | * which don't have a txg value stored on their label (i.e. spares/cache) | |
778 | * or have not been completely initialized (txg = 0) just return | |
779 | * the configuration from the first valid label we find. Otherwise, | |
780 | * find the most up-to-date label that does not exceed the specified | |
781 | * 'txg' value. | |
9ae529ec | 782 | */ |
34dc7c2f | 783 | nvlist_t * |
3bc7e0fb | 784 | vdev_label_read_config(vdev_t *vd, uint64_t txg) |
34dc7c2f BB |
785 | { |
786 | spa_t *spa = vd->vdev_spa; | |
787 | nvlist_t *config = NULL; | |
67874d54 AS |
788 | vdev_phys_t *vp[VDEV_LABELS]; |
789 | abd_t *vp_abd[VDEV_LABELS]; | |
790 | zio_t *zio[VDEV_LABELS]; | |
3bc7e0fb | 791 | uint64_t best_txg = 0; |
38a19edd | 792 | uint64_t label_txg = 0; |
3bc7e0fb | 793 | int error = 0; |
9babb374 BB |
794 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | |
795 | ZIO_FLAG_SPECULATIVE; | |
34dc7c2f | 796 | |
cf0977ad AS |
797 | ASSERT(vd->vdev_validate_thread == curthread || |
798 | spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL); | |
34dc7c2f BB |
799 | |
800 | if (!vdev_readable(vd)) | |
801 | return (NULL); | |
802 | ||
b2255edc BB |
803 | /* |
804 | * The label for a dRAID distributed spare is not stored on disk. | |
805 | * Instead it is generated when needed which allows us to bypass | |
806 | * the pipeline when reading the config from the label. | |
807 | */ | |
808 | if (vd->vdev_ops == &vdev_draid_spare_ops) | |
809 | return (vdev_draid_read_config_spare(vd)); | |
810 | ||
67874d54 AS |
811 | for (int l = 0; l < VDEV_LABELS; l++) { |
812 | vp_abd[l] = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE); | |
813 | vp[l] = abd_to_buf(vp_abd[l]); | |
814 | } | |
34dc7c2f | 815 | |
9babb374 | 816 | retry: |
1c27024e | 817 | for (int l = 0; l < VDEV_LABELS; l++) { |
67874d54 | 818 | zio[l] = zio_root(spa, NULL, NULL, flags); |
34dc7c2f | 819 | |
67874d54 AS |
820 | vdev_label_read(zio[l], vd, l, vp_abd[l], |
821 | offsetof(vdev_label_t, vl_vdev_phys), sizeof (vdev_phys_t), | |
822 | NULL, NULL, flags); | |
823 | } | |
824 | for (int l = 0; l < VDEV_LABELS; l++) { | |
825 | nvlist_t *label = NULL; | |
34dc7c2f | 826 | |
67874d54 AS |
827 | if (zio_wait(zio[l]) == 0 && |
828 | nvlist_unpack(vp[l]->vp_nvlist, sizeof (vp[l]->vp_nvlist), | |
3bc7e0fb | 829 | &label, 0) == 0) { |
3bc7e0fb GW |
830 | /* |
831 | * Auxiliary vdevs won't have txg values in their | |
832 | * labels and newly added vdevs may not have been | |
833 | * completely initialized so just return the | |
834 | * configuration from the first valid label we | |
835 | * encounter. | |
836 | */ | |
837 | error = nvlist_lookup_uint64(label, | |
838 | ZPOOL_CONFIG_POOL_TXG, &label_txg); | |
839 | if ((error || label_txg == 0) && !config) { | |
840 | config = label; | |
67874d54 AS |
841 | for (l++; l < VDEV_LABELS; l++) |
842 | zio_wait(zio[l]); | |
3bc7e0fb GW |
843 | break; |
844 | } else if (label_txg <= txg && label_txg > best_txg) { | |
845 | best_txg = label_txg; | |
846 | nvlist_free(config); | |
847 | config = fnvlist_dup(label); | |
848 | } | |
849 | } | |
34dc7c2f | 850 | |
3bc7e0fb GW |
851 | if (label != NULL) { |
852 | nvlist_free(label); | |
853 | label = NULL; | |
34dc7c2f BB |
854 | } |
855 | } | |
856 | ||
9babb374 BB |
857 | if (config == NULL && !(flags & ZIO_FLAG_TRYHARD)) { |
858 | flags |= ZIO_FLAG_TRYHARD; | |
859 | goto retry; | |
860 | } | |
861 | ||
38a19edd PZ |
862 | /* |
863 | * We found a valid label but it didn't pass txg restrictions. | |
864 | */ | |
865 | if (config == NULL && label_txg != 0) { | |
866 | vdev_dbgmsg(vd, "label discarded as txg is too large " | |
867 | "(%llu > %llu)", (u_longlong_t)label_txg, | |
868 | (u_longlong_t)txg); | |
869 | } | |
870 | ||
67874d54 AS |
871 | for (int l = 0; l < VDEV_LABELS; l++) { |
872 | abd_free(vp_abd[l]); | |
873 | } | |
34dc7c2f BB |
874 | |
875 | return (config); | |
876 | } | |
877 | ||
878 | /* | |
879 | * Determine if a device is in use. The 'spare_guid' parameter will be filled | |
880 | * in with the device guid if this spare is active elsewhere on the system. | |
881 | */ | |
882 | static boolean_t | |
883 | vdev_inuse(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason, | |
884 | uint64_t *spare_guid, uint64_t *l2cache_guid) | |
885 | { | |
886 | spa_t *spa = vd->vdev_spa; | |
887 | uint64_t state, pool_guid, device_guid, txg, spare_pool; | |
888 | uint64_t vdtxg = 0; | |
889 | nvlist_t *label; | |
890 | ||
891 | if (spare_guid) | |
892 | *spare_guid = 0ULL; | |
893 | if (l2cache_guid) | |
894 | *l2cache_guid = 0ULL; | |
895 | ||
896 | /* | |
897 | * Read the label, if any, and perform some basic sanity checks. | |
898 | */ | |
3bc7e0fb | 899 | if ((label = vdev_label_read_config(vd, -1ULL)) == NULL) |
34dc7c2f BB |
900 | return (B_FALSE); |
901 | ||
902 | (void) nvlist_lookup_uint64(label, ZPOOL_CONFIG_CREATE_TXG, | |
903 | &vdtxg); | |
904 | ||
905 | if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, | |
906 | &state) != 0 || | |
907 | nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, | |
908 | &device_guid) != 0) { | |
909 | nvlist_free(label); | |
910 | return (B_FALSE); | |
911 | } | |
912 | ||
913 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
914 | (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, | |
915 | &pool_guid) != 0 || | |
916 | nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG, | |
917 | &txg) != 0)) { | |
918 | nvlist_free(label); | |
919 | return (B_FALSE); | |
920 | } | |
921 | ||
922 | nvlist_free(label); | |
923 | ||
924 | /* | |
925 | * Check to see if this device indeed belongs to the pool it claims to | |
926 | * be a part of. The only way this is allowed is if the device is a hot | |
927 | * spare (which we check for later on). | |
928 | */ | |
929 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
930 | !spa_guid_exists(pool_guid, device_guid) && | |
b128c09f | 931 | !spa_spare_exists(device_guid, NULL, NULL) && |
34dc7c2f BB |
932 | !spa_l2cache_exists(device_guid, NULL)) |
933 | return (B_FALSE); | |
934 | ||
935 | /* | |
936 | * If the transaction group is zero, then this an initialized (but | |
937 | * unused) label. This is only an error if the create transaction | |
938 | * on-disk is the same as the one we're using now, in which case the | |
939 | * user has attempted to add the same vdev multiple times in the same | |
940 | * transaction. | |
941 | */ | |
942 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
943 | txg == 0 && vdtxg == crtxg) | |
944 | return (B_TRUE); | |
945 | ||
946 | /* | |
947 | * Check to see if this is a spare device. We do an explicit check for | |
948 | * spa_has_spare() here because it may be on our pending list of spares | |
49d42425 | 949 | * to add. |
34dc7c2f | 950 | */ |
b128c09f | 951 | if (spa_spare_exists(device_guid, &spare_pool, NULL) || |
34dc7c2f BB |
952 | spa_has_spare(spa, device_guid)) { |
953 | if (spare_guid) | |
954 | *spare_guid = device_guid; | |
955 | ||
956 | switch (reason) { | |
957 | case VDEV_LABEL_CREATE: | |
34dc7c2f BB |
958 | return (B_TRUE); |
959 | ||
960 | case VDEV_LABEL_REPLACE: | |
961 | return (!spa_has_spare(spa, device_guid) || | |
962 | spare_pool != 0ULL); | |
963 | ||
964 | case VDEV_LABEL_SPARE: | |
965 | return (spa_has_spare(spa, device_guid)); | |
e75c13c3 BB |
966 | default: |
967 | break; | |
34dc7c2f BB |
968 | } |
969 | } | |
970 | ||
971 | /* | |
972 | * Check to see if this is an l2cache device. | |
973 | */ | |
49d42425 FU |
974 | if (spa_l2cache_exists(device_guid, NULL) || |
975 | spa_has_l2cache(spa, device_guid)) { | |
976 | if (l2cache_guid) | |
977 | *l2cache_guid = device_guid; | |
978 | ||
979 | switch (reason) { | |
980 | case VDEV_LABEL_CREATE: | |
981 | return (B_TRUE); | |
982 | ||
983 | case VDEV_LABEL_REPLACE: | |
984 | return (!spa_has_l2cache(spa, device_guid)); | |
985 | ||
986 | case VDEV_LABEL_L2CACHE: | |
987 | return (spa_has_l2cache(spa, device_guid)); | |
988 | default: | |
989 | break; | |
990 | } | |
991 | } | |
34dc7c2f | 992 | |
572e2857 BB |
993 | /* |
994 | * We can't rely on a pool's state if it's been imported | |
995 | * read-only. Instead we look to see if the pools is marked | |
996 | * read-only in the namespace and set the state to active. | |
997 | */ | |
485c581c RY |
998 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && |
999 | (spa = spa_by_guid(pool_guid, device_guid)) != NULL && | |
da92d5cb | 1000 | spa_mode(spa) == SPA_MODE_READ) |
572e2857 BB |
1001 | state = POOL_STATE_ACTIVE; |
1002 | ||
34dc7c2f BB |
1003 | /* |
1004 | * If the device is marked ACTIVE, then this device is in use by another | |
1005 | * pool on the system. | |
1006 | */ | |
1007 | return (state == POOL_STATE_ACTIVE); | |
1008 | } | |
1009 | ||
1010 | /* | |
1011 | * Initialize a vdev label. We check to make sure each leaf device is not in | |
1012 | * use, and writable. We put down an initial label which we will later | |
1013 | * overwrite with a complete label. Note that it's important to do this | |
1014 | * sequentially, not in parallel, so that we catch cases of multiple use of the | |
1015 | * same leaf vdev in the vdev we're creating -- e.g. mirroring a disk with | |
1016 | * itself. | |
1017 | */ | |
1018 | int | |
1019 | vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason) | |
1020 | { | |
1021 | spa_t *spa = vd->vdev_spa; | |
1022 | nvlist_t *label; | |
1023 | vdev_phys_t *vp; | |
a6255b7f | 1024 | abd_t *vp_abd; |
108a454a | 1025 | abd_t *bootenv; |
34dc7c2f | 1026 | uberblock_t *ub; |
a6255b7f | 1027 | abd_t *ub_abd; |
34dc7c2f | 1028 | zio_t *zio; |
34dc7c2f BB |
1029 | char *buf; |
1030 | size_t buflen; | |
1031 | int error; | |
d4ed6673 | 1032 | uint64_t spare_guid = 0, l2cache_guid = 0; |
b128c09f | 1033 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; |
34dc7c2f | 1034 | |
b128c09f | 1035 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); |
34dc7c2f | 1036 | |
1c27024e | 1037 | for (int c = 0; c < vd->vdev_children; c++) |
34dc7c2f BB |
1038 | if ((error = vdev_label_init(vd->vdev_child[c], |
1039 | crtxg, reason)) != 0) | |
1040 | return (error); | |
1041 | ||
428870ff BB |
1042 | /* Track the creation time for this vdev */ |
1043 | vd->vdev_crtxg = crtxg; | |
1044 | ||
dda12da9 | 1045 | if (!vd->vdev_ops->vdev_op_leaf || !spa_writeable(spa)) |
34dc7c2f BB |
1046 | return (0); |
1047 | ||
1048 | /* | |
1049 | * Dead vdevs cannot be initialized. | |
1050 | */ | |
1051 | if (vdev_is_dead(vd)) | |
2e528b49 | 1052 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1053 | |
1054 | /* | |
1055 | * Determine if the vdev is in use. | |
1056 | */ | |
428870ff | 1057 | if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPLIT && |
34dc7c2f | 1058 | vdev_inuse(vd, crtxg, reason, &spare_guid, &l2cache_guid)) |
2e528b49 | 1059 | return (SET_ERROR(EBUSY)); |
34dc7c2f | 1060 | |
34dc7c2f BB |
1061 | /* |
1062 | * If this is a request to add or replace a spare or l2cache device | |
1063 | * that is in use elsewhere on the system, then we must update the | |
1064 | * guid (which was initialized to a random value) to reflect the | |
1065 | * actual GUID (which is shared between multiple pools). | |
1066 | */ | |
1067 | if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_L2CACHE && | |
1068 | spare_guid != 0ULL) { | |
b128c09f | 1069 | uint64_t guid_delta = spare_guid - vd->vdev_guid; |
34dc7c2f | 1070 | |
b128c09f | 1071 | vd->vdev_guid += guid_delta; |
34dc7c2f | 1072 | |
1c27024e | 1073 | for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) |
b128c09f | 1074 | pvd->vdev_guid_sum += guid_delta; |
34dc7c2f BB |
1075 | |
1076 | /* | |
1077 | * If this is a replacement, then we want to fallthrough to the | |
1078 | * rest of the code. If we're adding a spare, then it's already | |
1079 | * labeled appropriately and we can just return. | |
1080 | */ | |
1081 | if (reason == VDEV_LABEL_SPARE) | |
1082 | return (0); | |
428870ff BB |
1083 | ASSERT(reason == VDEV_LABEL_REPLACE || |
1084 | reason == VDEV_LABEL_SPLIT); | |
34dc7c2f BB |
1085 | } |
1086 | ||
1087 | if (reason != VDEV_LABEL_REMOVE && reason != VDEV_LABEL_SPARE && | |
1088 | l2cache_guid != 0ULL) { | |
b128c09f | 1089 | uint64_t guid_delta = l2cache_guid - vd->vdev_guid; |
34dc7c2f | 1090 | |
b128c09f | 1091 | vd->vdev_guid += guid_delta; |
34dc7c2f | 1092 | |
1c27024e | 1093 | for (vdev_t *pvd = vd; pvd != NULL; pvd = pvd->vdev_parent) |
b128c09f | 1094 | pvd->vdev_guid_sum += guid_delta; |
34dc7c2f BB |
1095 | |
1096 | /* | |
1097 | * If this is a replacement, then we want to fallthrough to the | |
1098 | * rest of the code. If we're adding an l2cache, then it's | |
1099 | * already labeled appropriately and we can just return. | |
1100 | */ | |
1101 | if (reason == VDEV_LABEL_L2CACHE) | |
1102 | return (0); | |
1103 | ASSERT(reason == VDEV_LABEL_REPLACE); | |
1104 | } | |
1105 | ||
1106 | /* | |
1107 | * Initialize its label. | |
1108 | */ | |
a6255b7f DQ |
1109 | vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE); |
1110 | abd_zero(vp_abd, sizeof (vdev_phys_t)); | |
1111 | vp = abd_to_buf(vp_abd); | |
34dc7c2f BB |
1112 | |
1113 | /* | |
1114 | * Generate a label describing the pool and our top-level vdev. | |
1115 | * We mark it as being from txg 0 to indicate that it's not | |
1116 | * really part of an active pool just yet. The labels will | |
1117 | * be written again with a meaningful txg by spa_sync(). | |
1118 | */ | |
1119 | if (reason == VDEV_LABEL_SPARE || | |
1120 | (reason == VDEV_LABEL_REMOVE && vd->vdev_isspare)) { | |
1121 | /* | |
1122 | * For inactive hot spares, we generate a special label that | |
1123 | * identifies as a mutually shared hot spare. We write the | |
1124 | * label if we are adding a hot spare, or if we are removing an | |
1125 | * active hot spare (in which case we want to revert the | |
1126 | * labels). | |
1127 | */ | |
79c76d5b | 1128 | VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); |
34dc7c2f BB |
1129 | |
1130 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, | |
1131 | spa_version(spa)) == 0); | |
1132 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, | |
1133 | POOL_STATE_SPARE) == 0); | |
1134 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, | |
1135 | vd->vdev_guid) == 0); | |
1136 | } else if (reason == VDEV_LABEL_L2CACHE || | |
1137 | (reason == VDEV_LABEL_REMOVE && vd->vdev_isl2cache)) { | |
1138 | /* | |
1139 | * For level 2 ARC devices, add a special label. | |
1140 | */ | |
79c76d5b | 1141 | VERIFY(nvlist_alloc(&label, NV_UNIQUE_NAME, KM_SLEEP) == 0); |
34dc7c2f BB |
1142 | |
1143 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_VERSION, | |
1144 | spa_version(spa)) == 0); | |
1145 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_POOL_STATE, | |
1146 | POOL_STATE_L2CACHE) == 0); | |
1147 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_GUID, | |
1148 | vd->vdev_guid) == 0); | |
fe4d055b GA |
1149 | |
1150 | /* | |
1151 | * This is merely to facilitate reporting the ashift of the | |
1152 | * cache device through zdb. The actual retrieval of the | |
1153 | * ashift (in vdev_alloc()) uses the nvlist | |
1154 | * spa->spa_l2cache->sav_config (populated in | |
1155 | * spa_ld_open_aux_vdevs()). | |
1156 | */ | |
1157 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_ASHIFT, | |
1158 | vd->vdev_ashift) == 0); | |
d9885b37 AH |
1159 | |
1160 | /* | |
1161 | * When spare or l2cache (aux) vdev is added during pool | |
1162 | * creation, spa->spa_uberblock is not written until this | |
1163 | * point. Write it on next config sync. | |
1164 | */ | |
1165 | if (uberblock_verify(&spa->spa_uberblock)) | |
1166 | spa->spa_aux_sync_uber = B_TRUE; | |
34dc7c2f | 1167 | } else { |
428870ff BB |
1168 | uint64_t txg = 0ULL; |
1169 | ||
1170 | if (reason == VDEV_LABEL_SPLIT) | |
1171 | txg = spa->spa_uberblock.ub_txg; | |
1172 | label = spa_config_generate(spa, vd, txg, B_FALSE); | |
34dc7c2f BB |
1173 | |
1174 | /* | |
1175 | * Add our creation time. This allows us to detect multiple | |
1176 | * vdev uses as described above, and automatically expires if we | |
1177 | * fail. | |
1178 | */ | |
1179 | VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG, | |
1180 | crtxg) == 0); | |
1181 | } | |
1182 | ||
1183 | buf = vp->vp_nvlist; | |
1184 | buflen = sizeof (vp->vp_nvlist); | |
1185 | ||
79c76d5b | 1186 | error = nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP); |
34dc7c2f BB |
1187 | if (error != 0) { |
1188 | nvlist_free(label); | |
a6255b7f | 1189 | abd_free(vp_abd); |
34dc7c2f | 1190 | /* EFAULT means nvlist_pack ran out of room */ |
ecb2b7dc | 1191 | return (SET_ERROR(error == EFAULT ? ENAMETOOLONG : EINVAL)); |
34dc7c2f BB |
1192 | } |
1193 | ||
34dc7c2f BB |
1194 | /* |
1195 | * Initialize uberblock template. | |
1196 | */ | |
a6255b7f | 1197 | ub_abd = abd_alloc_linear(VDEV_UBERBLOCK_RING, B_TRUE); |
a6255b7f | 1198 | abd_copy_from_buf(ub_abd, &spa->spa_uberblock, sizeof (uberblock_t)); |
22c8c33a AM |
1199 | abd_zero_off(ub_abd, sizeof (uberblock_t), |
1200 | VDEV_UBERBLOCK_RING - sizeof (uberblock_t)); | |
a6255b7f | 1201 | ub = abd_to_buf(ub_abd); |
34dc7c2f BB |
1202 | ub->ub_txg = 0; |
1203 | ||
9babb374 | 1204 | /* Initialize the 2nd padding area. */ |
108a454a PD |
1205 | bootenv = abd_alloc_for_io(VDEV_PAD_SIZE, B_TRUE); |
1206 | abd_zero(bootenv, VDEV_PAD_SIZE); | |
9babb374 | 1207 | |
34dc7c2f BB |
1208 | /* |
1209 | * Write everything in parallel. | |
1210 | */ | |
9babb374 | 1211 | retry: |
34dc7c2f BB |
1212 | zio = zio_root(spa, NULL, NULL, flags); |
1213 | ||
1c27024e | 1214 | for (int l = 0; l < VDEV_LABELS; l++) { |
34dc7c2f | 1215 | |
a6255b7f | 1216 | vdev_label_write(zio, vd, l, vp_abd, |
34dc7c2f BB |
1217 | offsetof(vdev_label_t, vl_vdev_phys), |
1218 | sizeof (vdev_phys_t), NULL, NULL, flags); | |
1219 | ||
9babb374 BB |
1220 | /* |
1221 | * Skip the 1st padding area. | |
1222 | * Zero out the 2nd padding area where it might have | |
1223 | * left over data from previous filesystem format. | |
1224 | */ | |
108a454a PD |
1225 | vdev_label_write(zio, vd, l, bootenv, |
1226 | offsetof(vdev_label_t, vl_be), | |
9babb374 | 1227 | VDEV_PAD_SIZE, NULL, NULL, flags); |
34dc7c2f | 1228 | |
a6255b7f | 1229 | vdev_label_write(zio, vd, l, ub_abd, |
45d1cae3 BB |
1230 | offsetof(vdev_label_t, vl_uberblock), |
1231 | VDEV_UBERBLOCK_RING, NULL, NULL, flags); | |
34dc7c2f BB |
1232 | } |
1233 | ||
1234 | error = zio_wait(zio); | |
1235 | ||
9babb374 BB |
1236 | if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) { |
1237 | flags |= ZIO_FLAG_TRYHARD; | |
1238 | goto retry; | |
1239 | } | |
1240 | ||
34dc7c2f | 1241 | nvlist_free(label); |
108a454a | 1242 | abd_free(bootenv); |
a6255b7f DQ |
1243 | abd_free(ub_abd); |
1244 | abd_free(vp_abd); | |
34dc7c2f BB |
1245 | |
1246 | /* | |
1247 | * If this vdev hasn't been previously identified as a spare, then we | |
1248 | * mark it as such only if a) we are labeling it as a spare, or b) it | |
1249 | * exists as a spare elsewhere in the system. Do the same for | |
1250 | * level 2 ARC devices. | |
1251 | */ | |
1252 | if (error == 0 && !vd->vdev_isspare && | |
1253 | (reason == VDEV_LABEL_SPARE || | |
b128c09f | 1254 | spa_spare_exists(vd->vdev_guid, NULL, NULL))) |
34dc7c2f BB |
1255 | spa_spare_add(vd); |
1256 | ||
1257 | if (error == 0 && !vd->vdev_isl2cache && | |
1258 | (reason == VDEV_LABEL_L2CACHE || | |
1259 | spa_l2cache_exists(vd->vdev_guid, NULL))) | |
1260 | spa_l2cache_add(vd); | |
1261 | ||
1262 | return (error); | |
1263 | } | |
1264 | ||
108a454a PD |
1265 | /* |
1266 | * Done callback for vdev_label_read_bootenv_impl. If this is the first | |
1267 | * callback to finish, store our abd in the callback pointer. Otherwise, we | |
1268 | * just free our abd and return. | |
1269 | */ | |
1270 | static void | |
1271 | vdev_label_read_bootenv_done(zio_t *zio) | |
1272 | { | |
1273 | zio_t *rio = zio->io_private; | |
1274 | abd_t **cbp = rio->io_private; | |
1275 | ||
1276 | ASSERT3U(zio->io_size, ==, VDEV_PAD_SIZE); | |
1277 | ||
1278 | if (zio->io_error == 0) { | |
1279 | mutex_enter(&rio->io_lock); | |
1280 | if (*cbp == NULL) { | |
1281 | /* Will free this buffer in vdev_label_read_bootenv. */ | |
1282 | *cbp = zio->io_abd; | |
1283 | } else { | |
1284 | abd_free(zio->io_abd); | |
1285 | } | |
1286 | mutex_exit(&rio->io_lock); | |
1287 | } else { | |
1288 | abd_free(zio->io_abd); | |
1289 | } | |
1290 | } | |
1291 | ||
1292 | static void | |
1293 | vdev_label_read_bootenv_impl(zio_t *zio, vdev_t *vd, int flags) | |
1294 | { | |
1295 | for (int c = 0; c < vd->vdev_children; c++) | |
1296 | vdev_label_read_bootenv_impl(zio, vd->vdev_child[c], flags); | |
1297 | ||
1298 | /* | |
1299 | * We just use the first label that has a correct checksum; the | |
1300 | * bootloader should have rewritten them all to be the same on boot, | |
1301 | * and any changes we made since boot have been the same across all | |
1302 | * labels. | |
108a454a PD |
1303 | */ |
1304 | if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd)) { | |
1db9e6e4 | 1305 | for (int l = 0; l < VDEV_LABELS; l++) { |
108a454a PD |
1306 | vdev_label_read(zio, vd, l, |
1307 | abd_alloc_linear(VDEV_PAD_SIZE, B_FALSE), | |
1308 | offsetof(vdev_label_t, vl_be), VDEV_PAD_SIZE, | |
1309 | vdev_label_read_bootenv_done, zio, flags); | |
1310 | } | |
1311 | } | |
1312 | } | |
1313 | ||
1314 | int | |
1db9e6e4 | 1315 | vdev_label_read_bootenv(vdev_t *rvd, nvlist_t *bootenv) |
108a454a | 1316 | { |
1db9e6e4 | 1317 | nvlist_t *config; |
108a454a PD |
1318 | spa_t *spa = rvd->vdev_spa; |
1319 | abd_t *abd = NULL; | |
1320 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | | |
1321 | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD; | |
1322 | ||
1db9e6e4 | 1323 | ASSERT(bootenv); |
108a454a PD |
1324 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); |
1325 | ||
1326 | zio_t *zio = zio_root(spa, NULL, &abd, flags); | |
1327 | vdev_label_read_bootenv_impl(zio, rvd, flags); | |
1328 | int err = zio_wait(zio); | |
1329 | ||
1330 | if (abd != NULL) { | |
1db9e6e4 | 1331 | char *buf; |
108a454a | 1332 | vdev_boot_envblock_t *vbe = abd_to_buf(abd); |
1db9e6e4 TS |
1333 | |
1334 | vbe->vbe_version = ntohll(vbe->vbe_version); | |
1335 | switch (vbe->vbe_version) { | |
1336 | case VB_RAW: | |
1337 | /* | |
1338 | * if we have textual data in vbe_bootenv, create nvlist | |
1339 | * with key "envmap". | |
1340 | */ | |
1341 | fnvlist_add_uint64(bootenv, BOOTENV_VERSION, VB_RAW); | |
1342 | vbe->vbe_bootenv[sizeof (vbe->vbe_bootenv) - 1] = '\0'; | |
1343 | fnvlist_add_string(bootenv, GRUB_ENVMAP, | |
1344 | vbe->vbe_bootenv); | |
1345 | break; | |
1346 | ||
1347 | case VB_NVLIST: | |
1348 | err = nvlist_unpack(vbe->vbe_bootenv, | |
1349 | sizeof (vbe->vbe_bootenv), &config, 0); | |
1350 | if (err == 0) { | |
1351 | fnvlist_merge(bootenv, config); | |
1352 | nvlist_free(config); | |
1353 | break; | |
1354 | } | |
9a70e97f | 1355 | zfs_fallthrough; |
1db9e6e4 TS |
1356 | default: |
1357 | /* Check for FreeBSD zfs bootonce command string */ | |
1358 | buf = abd_to_buf(abd); | |
1359 | if (*buf == '\0') { | |
1360 | fnvlist_add_uint64(bootenv, BOOTENV_VERSION, | |
1361 | VB_NVLIST); | |
1362 | break; | |
1363 | } | |
1364 | fnvlist_add_string(bootenv, FREEBSD_BOOTONCE, buf); | |
108a454a | 1365 | } |
1db9e6e4 TS |
1366 | |
1367 | /* | |
1368 | * abd was allocated in vdev_label_read_bootenv_impl() | |
1369 | */ | |
108a454a | 1370 | abd_free(abd); |
1db9e6e4 TS |
1371 | /* |
1372 | * If we managed to read any successfully, | |
1373 | * return success. | |
1374 | */ | |
108a454a PD |
1375 | return (0); |
1376 | } | |
1377 | return (err); | |
1378 | } | |
1379 | ||
1380 | int | |
1db9e6e4 | 1381 | vdev_label_write_bootenv(vdev_t *vd, nvlist_t *env) |
108a454a PD |
1382 | { |
1383 | zio_t *zio; | |
1384 | spa_t *spa = vd->vdev_spa; | |
1385 | vdev_boot_envblock_t *bootenv; | |
1386 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; | |
1db9e6e4 TS |
1387 | int error; |
1388 | size_t nvsize; | |
1389 | char *nvbuf; | |
d1807f16 | 1390 | const char *tmp; |
108a454a | 1391 | |
1db9e6e4 TS |
1392 | error = nvlist_size(env, &nvsize, NV_ENCODE_XDR); |
1393 | if (error != 0) | |
1394 | return (SET_ERROR(error)); | |
1395 | ||
1396 | if (nvsize >= sizeof (bootenv->vbe_bootenv)) { | |
108a454a PD |
1397 | return (SET_ERROR(E2BIG)); |
1398 | } | |
1399 | ||
1400 | ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL); | |
1401 | ||
1db9e6e4 | 1402 | error = ENXIO; |
108a454a | 1403 | for (int c = 0; c < vd->vdev_children; c++) { |
1db9e6e4 TS |
1404 | int child_err; |
1405 | ||
1406 | child_err = vdev_label_write_bootenv(vd->vdev_child[c], env); | |
108a454a PD |
1407 | /* |
1408 | * As long as any of the disks managed to write all of their | |
1409 | * labels successfully, return success. | |
1410 | */ | |
1411 | if (child_err == 0) | |
1412 | error = child_err; | |
1413 | } | |
1414 | ||
1415 | if (!vd->vdev_ops->vdev_op_leaf || vdev_is_dead(vd) || | |
1416 | !vdev_writeable(vd)) { | |
1417 | return (error); | |
1418 | } | |
1419 | ASSERT3U(sizeof (*bootenv), ==, VDEV_PAD_SIZE); | |
1420 | abd_t *abd = abd_alloc_for_io(VDEV_PAD_SIZE, B_TRUE); | |
1421 | abd_zero(abd, VDEV_PAD_SIZE); | |
1db9e6e4 | 1422 | |
108a454a | 1423 | bootenv = abd_borrow_buf_copy(abd, VDEV_PAD_SIZE); |
1db9e6e4 TS |
1424 | nvbuf = bootenv->vbe_bootenv; |
1425 | nvsize = sizeof (bootenv->vbe_bootenv); | |
1426 | ||
1427 | bootenv->vbe_version = fnvlist_lookup_uint64(env, BOOTENV_VERSION); | |
1428 | switch (bootenv->vbe_version) { | |
1429 | case VB_RAW: | |
d1807f16 RY |
1430 | if (nvlist_lookup_string(env, GRUB_ENVMAP, &tmp) == 0) { |
1431 | (void) strlcpy(bootenv->vbe_bootenv, tmp, nvsize); | |
1db9e6e4 TS |
1432 | } |
1433 | error = 0; | |
1434 | break; | |
1435 | ||
1436 | case VB_NVLIST: | |
1437 | error = nvlist_pack(env, &nvbuf, &nvsize, NV_ENCODE_XDR, | |
1438 | KM_SLEEP); | |
1439 | break; | |
108a454a | 1440 | |
1db9e6e4 TS |
1441 | default: |
1442 | error = EINVAL; | |
1443 | break; | |
1444 | } | |
1445 | ||
1446 | if (error == 0) { | |
1447 | bootenv->vbe_version = htonll(bootenv->vbe_version); | |
1448 | abd_return_buf_copy(abd, bootenv, VDEV_PAD_SIZE); | |
1449 | } else { | |
1450 | abd_free(abd); | |
1451 | return (SET_ERROR(error)); | |
1452 | } | |
108a454a PD |
1453 | |
1454 | retry: | |
1455 | zio = zio_root(spa, NULL, NULL, flags); | |
1db9e6e4 | 1456 | for (int l = 0; l < VDEV_LABELS; l++) { |
108a454a PD |
1457 | vdev_label_write(zio, vd, l, abd, |
1458 | offsetof(vdev_label_t, vl_be), | |
1459 | VDEV_PAD_SIZE, NULL, NULL, flags); | |
1460 | } | |
1461 | ||
1462 | error = zio_wait(zio); | |
1463 | if (error != 0 && !(flags & ZIO_FLAG_TRYHARD)) { | |
1464 | flags |= ZIO_FLAG_TRYHARD; | |
1465 | goto retry; | |
1466 | } | |
1467 | ||
1468 | abd_free(abd); | |
1469 | return (error); | |
1470 | } | |
1471 | ||
34dc7c2f BB |
1472 | /* |
1473 | * ========================================================================== | |
1474 | * uberblock load/sync | |
1475 | * ========================================================================== | |
1476 | */ | |
1477 | ||
1478 | /* | |
1479 | * Consider the following situation: txg is safely synced to disk. We've | |
1480 | * written the first uberblock for txg + 1, and then we lose power. When we | |
1481 | * come back up, we fail to see the uberblock for txg + 1 because, say, | |
1482 | * it was on a mirrored device and the replica to which we wrote txg + 1 | |
1483 | * is now offline. If we then make some changes and sync txg + 1, and then | |
9ae529ec | 1484 | * the missing replica comes back, then for a few seconds we'll have two |
34dc7c2f BB |
1485 | * conflicting uberblocks on disk with the same txg. The solution is simple: |
1486 | * among uberblocks with equal txg, choose the one with the latest timestamp. | |
1487 | */ | |
1488 | static int | |
ee36c709 | 1489 | vdev_uberblock_compare(const uberblock_t *ub1, const uberblock_t *ub2) |
34dc7c2f | 1490 | { |
ca577779 | 1491 | int cmp = TREE_CMP(ub1->ub_txg, ub2->ub_txg); |
060f0226 | 1492 | |
ee36c709 GN |
1493 | if (likely(cmp)) |
1494 | return (cmp); | |
34dc7c2f | 1495 | |
ca577779 | 1496 | cmp = TREE_CMP(ub1->ub_timestamp, ub2->ub_timestamp); |
060f0226 OF |
1497 | if (likely(cmp)) |
1498 | return (cmp); | |
1499 | ||
1500 | /* | |
1501 | * If MMP_VALID(ub) && MMP_SEQ_VALID(ub) then the host has an MMP-aware | |
d0249a4b | 1502 | * ZFS, e.g. OpenZFS >= 0.7. |
060f0226 OF |
1503 | * |
1504 | * If one ub has MMP and the other does not, they were written by | |
1505 | * different hosts, which matters for MMP. So we treat no MMP/no SEQ as | |
1506 | * a 0 value. | |
1507 | * | |
1508 | * Since timestamp and txg are the same if we get this far, either is | |
1509 | * acceptable for importing the pool. | |
1510 | */ | |
1511 | unsigned int seq1 = 0; | |
1512 | unsigned int seq2 = 0; | |
1513 | ||
1514 | if (MMP_VALID(ub1) && MMP_SEQ_VALID(ub1)) | |
1515 | seq1 = MMP_SEQ(ub1); | |
1516 | ||
1517 | if (MMP_VALID(ub2) && MMP_SEQ_VALID(ub2)) | |
1518 | seq2 = MMP_SEQ(ub2); | |
1519 | ||
ca577779 | 1520 | return (TREE_CMP(seq1, seq2)); |
34dc7c2f BB |
1521 | } |
1522 | ||
9ae529ec | 1523 | struct ubl_cbdata { |
5caeef02 DB |
1524 | uberblock_t ubl_latest; /* Most recent uberblock */ |
1525 | uberblock_t *ubl_ubbest; /* Best uberblock (w/r/t max_txg) */ | |
9ae529ec | 1526 | vdev_t *ubl_vd; /* vdev associated with the above */ |
9ae529ec CS |
1527 | }; |
1528 | ||
34dc7c2f BB |
1529 | static void |
1530 | vdev_uberblock_load_done(zio_t *zio) | |
1531 | { | |
9ae529ec | 1532 | vdev_t *vd = zio->io_vd; |
428870ff | 1533 | spa_t *spa = zio->io_spa; |
b128c09f | 1534 | zio_t *rio = zio->io_private; |
a6255b7f | 1535 | uberblock_t *ub = abd_to_buf(zio->io_abd); |
9ae529ec | 1536 | struct ubl_cbdata *cbp = rio->io_private; |
34dc7c2f | 1537 | |
9ae529ec | 1538 | ASSERT3U(zio->io_size, ==, VDEV_UBERBLOCK_SIZE(vd)); |
34dc7c2f BB |
1539 | |
1540 | if (zio->io_error == 0 && uberblock_verify(ub) == 0) { | |
b128c09f | 1541 | mutex_enter(&rio->io_lock); |
5caeef02 DB |
1542 | if (vdev_uberblock_compare(ub, &cbp->ubl_latest) > 0) { |
1543 | cbp->ubl_latest = *ub; | |
1544 | } | |
428870ff | 1545 | if (ub->ub_txg <= spa->spa_load_max_txg && |
9ae529ec CS |
1546 | vdev_uberblock_compare(ub, cbp->ubl_ubbest) > 0) { |
1547 | /* | |
3bc7e0fb GW |
1548 | * Keep track of the vdev in which this uberblock |
1549 | * was found. We will use this information later | |
1550 | * to obtain the config nvlist associated with | |
9ae529ec CS |
1551 | * this uberblock. |
1552 | */ | |
1553 | *cbp->ubl_ubbest = *ub; | |
1554 | cbp->ubl_vd = vd; | |
9ae529ec | 1555 | } |
b128c09f | 1556 | mutex_exit(&rio->io_lock); |
34dc7c2f BB |
1557 | } |
1558 | ||
a6255b7f | 1559 | abd_free(zio->io_abd); |
34dc7c2f BB |
1560 | } |
1561 | ||
9ae529ec CS |
1562 | static void |
1563 | vdev_uberblock_load_impl(zio_t *zio, vdev_t *vd, int flags, | |
1564 | struct ubl_cbdata *cbp) | |
34dc7c2f | 1565 | { |
379ca9cf | 1566 | for (int c = 0; c < vd->vdev_children; c++) |
9ae529ec | 1567 | vdev_uberblock_load_impl(zio, vd->vdev_child[c], flags, cbp); |
34dc7c2f | 1568 | |
b2255edc BB |
1569 | if (vd->vdev_ops->vdev_op_leaf && vdev_readable(vd) && |
1570 | vd->vdev_ops != &vdev_draid_spare_ops) { | |
379ca9cf OF |
1571 | for (int l = 0; l < VDEV_LABELS; l++) { |
1572 | for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) { | |
b128c09f | 1573 | vdev_label_read(zio, vd, l, |
a6255b7f DQ |
1574 | abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd), |
1575 | B_TRUE), VDEV_UBERBLOCK_OFFSET(vd, n), | |
b128c09f BB |
1576 | VDEV_UBERBLOCK_SIZE(vd), |
1577 | vdev_uberblock_load_done, zio, flags); | |
1578 | } | |
34dc7c2f BB |
1579 | } |
1580 | } | |
9ae529ec CS |
1581 | } |
1582 | ||
1583 | /* | |
1584 | * Reads the 'best' uberblock from disk along with its associated | |
1585 | * configuration. First, we read the uberblock array of each label of each | |
1586 | * vdev, keeping track of the uberblock with the highest txg in each array. | |
3bc7e0fb | 1587 | * Then, we read the configuration from the same vdev as the best uberblock. |
9ae529ec CS |
1588 | */ |
1589 | void | |
1590 | vdev_uberblock_load(vdev_t *rvd, uberblock_t *ub, nvlist_t **config) | |
1591 | { | |
9ae529ec CS |
1592 | zio_t *zio; |
1593 | spa_t *spa = rvd->vdev_spa; | |
1594 | struct ubl_cbdata cb; | |
1595 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | | |
1596 | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_TRYHARD; | |
1597 | ||
1598 | ASSERT(ub); | |
1599 | ASSERT(config); | |
b128c09f | 1600 | |
861166b0 | 1601 | memset(ub, 0, sizeof (uberblock_t)); |
5caeef02 | 1602 | memset(&cb, 0, sizeof (cb)); |
9ae529ec CS |
1603 | *config = NULL; |
1604 | ||
1605 | cb.ubl_ubbest = ub; | |
9ae529ec CS |
1606 | |
1607 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); | |
1608 | zio = zio_root(spa, NULL, &cb, flags); | |
1609 | vdev_uberblock_load_impl(zio, rvd, flags, &cb); | |
1610 | (void) zio_wait(zio); | |
3bc7e0fb GW |
1611 | |
1612 | /* | |
1613 | * It's possible that the best uberblock was discovered on a label | |
1614 | * that has a configuration which was written in a future txg. | |
1615 | * Search all labels on this vdev to find the configuration that | |
1616 | * matches the txg for our uberblock. | |
1617 | */ | |
4a0ee12a PZ |
1618 | if (cb.ubl_vd != NULL) { |
1619 | vdev_dbgmsg(cb.ubl_vd, "best uberblock found for spa %s. " | |
1620 | "txg %llu", spa->spa_name, (u_longlong_t)ub->ub_txg); | |
1621 | ||
5caeef02 DB |
1622 | if (ub->ub_raidz_reflow_info != |
1623 | cb.ubl_latest.ub_raidz_reflow_info) { | |
1624 | vdev_dbgmsg(cb.ubl_vd, | |
1625 | "spa=%s best uberblock (txg=%llu info=0x%llx) " | |
1626 | "has different raidz_reflow_info than latest " | |
1627 | "uberblock (txg=%llu info=0x%llx)", | |
1628 | spa->spa_name, | |
1629 | (u_longlong_t)ub->ub_txg, | |
1630 | (u_longlong_t)ub->ub_raidz_reflow_info, | |
1631 | (u_longlong_t)cb.ubl_latest.ub_txg, | |
1632 | (u_longlong_t)cb.ubl_latest.ub_raidz_reflow_info); | |
1633 | memset(ub, 0, sizeof (uberblock_t)); | |
1634 | spa_config_exit(spa, SCL_ALL, FTAG); | |
1635 | return; | |
1636 | } | |
1637 | ||
3bc7e0fb | 1638 | *config = vdev_label_read_config(cb.ubl_vd, ub->ub_txg); |
6cb8e530 PZ |
1639 | if (*config == NULL && spa->spa_extreme_rewind) { |
1640 | vdev_dbgmsg(cb.ubl_vd, "failed to read label config. " | |
1641 | "Trying again without txg restrictions."); | |
1642 | *config = vdev_label_read_config(cb.ubl_vd, UINT64_MAX); | |
1643 | } | |
4a0ee12a PZ |
1644 | if (*config == NULL) { |
1645 | vdev_dbgmsg(cb.ubl_vd, "failed to read label config"); | |
1646 | } | |
1647 | } | |
9ae529ec | 1648 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
1649 | } |
1650 | ||
9d3f7b87 OF |
1651 | /* |
1652 | * For use when a leaf vdev is expanded. | |
1653 | * The location of labels 2 and 3 changed, and at the new location the | |
1654 | * uberblock rings are either empty or contain garbage. The sync will write | |
1655 | * new configs there because the vdev is dirty, but expansion also needs the | |
1656 | * uberblock rings copied. Read them from label 0 which did not move. | |
1657 | * | |
1658 | * Since the point is to populate labels {2,3} with valid uberblocks, | |
1659 | * we zero uberblocks we fail to read or which are not valid. | |
1660 | */ | |
1661 | ||
1662 | static void | |
1663 | vdev_copy_uberblocks(vdev_t *vd) | |
1664 | { | |
1665 | abd_t *ub_abd; | |
1666 | zio_t *write_zio; | |
1667 | int locks = (SCL_L2ARC | SCL_ZIO); | |
1668 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL | | |
1669 | ZIO_FLAG_SPECULATIVE; | |
1670 | ||
1671 | ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_READER) == | |
1672 | SCL_STATE); | |
1673 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
1674 | ||
b2255edc BB |
1675 | /* |
1676 | * No uberblocks are stored on distributed spares, they may be | |
1677 | * safely skipped when expanding a leaf vdev. | |
1678 | */ | |
1679 | if (vd->vdev_ops == &vdev_draid_spare_ops) | |
1680 | return; | |
1681 | ||
9d3f7b87 OF |
1682 | spa_config_enter(vd->vdev_spa, locks, FTAG, RW_READER); |
1683 | ||
eea2e241 | 1684 | ub_abd = abd_alloc_linear(VDEV_UBERBLOCK_SIZE(vd), B_TRUE); |
9d3f7b87 OF |
1685 | |
1686 | write_zio = zio_root(vd->vdev_spa, NULL, NULL, flags); | |
1687 | for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) { | |
1688 | const int src_label = 0; | |
1689 | zio_t *zio; | |
1690 | ||
1691 | zio = zio_root(vd->vdev_spa, NULL, NULL, flags); | |
1692 | vdev_label_read(zio, vd, src_label, ub_abd, | |
1693 | VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd), | |
1694 | NULL, NULL, flags); | |
1695 | ||
1696 | if (zio_wait(zio) || uberblock_verify(abd_to_buf(ub_abd))) | |
1697 | abd_zero(ub_abd, VDEV_UBERBLOCK_SIZE(vd)); | |
1698 | ||
1699 | for (int l = 2; l < VDEV_LABELS; l++) | |
1700 | vdev_label_write(write_zio, vd, l, ub_abd, | |
1701 | VDEV_UBERBLOCK_OFFSET(vd, n), | |
1702 | VDEV_UBERBLOCK_SIZE(vd), NULL, NULL, | |
1703 | flags | ZIO_FLAG_DONT_PROPAGATE); | |
1704 | } | |
1705 | (void) zio_wait(write_zio); | |
1706 | ||
1707 | spa_config_exit(vd->vdev_spa, locks, FTAG); | |
1708 | ||
1709 | abd_free(ub_abd); | |
1710 | } | |
1711 | ||
34dc7c2f BB |
1712 | /* |
1713 | * On success, increment root zio's count of good writes. | |
1714 | * We only get credit for writes to known-visible vdevs; see spa_vdev_add(). | |
1715 | */ | |
1716 | static void | |
1717 | vdev_uberblock_sync_done(zio_t *zio) | |
1718 | { | |
1719 | uint64_t *good_writes = zio->io_private; | |
1720 | ||
1721 | if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0) | |
bc89ac84 | 1722 | atomic_inc_64(good_writes); |
34dc7c2f BB |
1723 | } |
1724 | ||
1725 | /* | |
1726 | * Write the uberblock to all labels of all leaves of the specified vdev. | |
1727 | */ | |
1728 | static void | |
7f96cc23 MA |
1729 | vdev_uberblock_sync(zio_t *zio, uint64_t *good_writes, |
1730 | uberblock_t *ub, vdev_t *vd, int flags) | |
34dc7c2f | 1731 | { |
7f96cc23 MA |
1732 | for (uint64_t c = 0; c < vd->vdev_children; c++) { |
1733 | vdev_uberblock_sync(zio, good_writes, | |
1734 | ub, vd->vdev_child[c], flags); | |
1735 | } | |
34dc7c2f BB |
1736 | |
1737 | if (!vd->vdev_ops->vdev_op_leaf) | |
1738 | return; | |
1739 | ||
b128c09f | 1740 | if (!vdev_writeable(vd)) |
34dc7c2f BB |
1741 | return; |
1742 | ||
b2255edc BB |
1743 | /* |
1744 | * There's no need to write uberblocks to a distributed spare, they | |
1745 | * are already stored on all the leaves of the parent dRAID. For | |
1746 | * this same reason vdev_uberblock_load_impl() skips distributed | |
1747 | * spares when reading uberblocks. | |
1748 | */ | |
1749 | if (vd->vdev_ops == &vdev_draid_spare_ops) | |
1750 | return; | |
1751 | ||
9d3f7b87 OF |
1752 | /* If the vdev was expanded, need to copy uberblock rings. */ |
1753 | if (vd->vdev_state == VDEV_STATE_HEALTHY && | |
1754 | vd->vdev_copy_uberblocks == B_TRUE) { | |
1755 | vdev_copy_uberblocks(vd); | |
1756 | vd->vdev_copy_uberblocks = B_FALSE; | |
1757 | } | |
1758 | ||
5caeef02 DB |
1759 | /* |
1760 | * We chose a slot based on the txg. If this uberblock has a special | |
1761 | * RAIDZ expansion state, then it is essentially an update of the | |
1762 | * current uberblock (it has the same txg). However, the current | |
1763 | * state is committed, so we want to write it to a different slot. If | |
1764 | * we overwrote the same slot, and we lose power during the uberblock | |
1765 | * write, and the disk does not do single-sector overwrites | |
1766 | * atomically (even though it is required to - i.e. we should see | |
1767 | * either the old or the new uberblock), then we could lose this | |
1768 | * txg's uberblock. Rewinding to the previous txg's uberblock may not | |
1769 | * be possible because RAIDZ expansion may have already overwritten | |
1770 | * some of the data, so we need the progress indicator in the | |
1771 | * uberblock. | |
1772 | */ | |
379ca9cf | 1773 | int m = spa_multihost(vd->vdev_spa) ? MMP_BLOCKS_PER_LABEL : 0; |
5caeef02 DB |
1774 | int n = (ub->ub_txg - (RRSS_GET_STATE(ub) == RRSS_SCRATCH_VALID)) % |
1775 | (VDEV_UBERBLOCK_COUNT(vd) - m); | |
34dc7c2f | 1776 | |
a6255b7f | 1777 | /* Copy the uberblock_t into the ABD */ |
379ca9cf | 1778 | abd_t *ub_abd = abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd), B_TRUE); |
a6255b7f | 1779 | abd_copy_from_buf(ub_abd, ub, sizeof (uberblock_t)); |
22c8c33a AM |
1780 | abd_zero_off(ub_abd, sizeof (uberblock_t), |
1781 | VDEV_UBERBLOCK_SIZE(vd) - sizeof (uberblock_t)); | |
34dc7c2f | 1782 | |
379ca9cf | 1783 | for (int l = 0; l < VDEV_LABELS; l++) |
a6255b7f | 1784 | vdev_label_write(zio, vd, l, ub_abd, |
b128c09f | 1785 | VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd), |
7f96cc23 | 1786 | vdev_uberblock_sync_done, good_writes, |
b128c09f | 1787 | flags | ZIO_FLAG_DONT_PROPAGATE); |
34dc7c2f | 1788 | |
a6255b7f | 1789 | abd_free(ub_abd); |
34dc7c2f BB |
1790 | } |
1791 | ||
e49f1e20 | 1792 | /* Sync the uberblocks to all vdevs in svd[] */ |
5caeef02 | 1793 | int |
34dc7c2f BB |
1794 | vdev_uberblock_sync_list(vdev_t **svd, int svdcount, uberblock_t *ub, int flags) |
1795 | { | |
1796 | spa_t *spa = svd[0]->vdev_spa; | |
34dc7c2f BB |
1797 | zio_t *zio; |
1798 | uint64_t good_writes = 0; | |
1799 | ||
7f96cc23 | 1800 | zio = zio_root(spa, NULL, NULL, flags); |
34dc7c2f | 1801 | |
1c27024e | 1802 | for (int v = 0; v < svdcount; v++) |
7f96cc23 | 1803 | vdev_uberblock_sync(zio, &good_writes, ub, svd[v], flags); |
34dc7c2f | 1804 | |
d9885b37 AH |
1805 | if (spa->spa_aux_sync_uber) { |
1806 | for (int v = 0; v < spa->spa_spares.sav_count; v++) { | |
1807 | vdev_uberblock_sync(zio, &good_writes, ub, | |
1808 | spa->spa_spares.sav_vdevs[v], flags); | |
1809 | } | |
1810 | for (int v = 0; v < spa->spa_l2cache.sav_count; v++) { | |
1811 | vdev_uberblock_sync(zio, &good_writes, ub, | |
1812 | spa->spa_l2cache.sav_vdevs[v], flags); | |
1813 | } | |
1814 | } | |
34dc7c2f BB |
1815 | (void) zio_wait(zio); |
1816 | ||
1817 | /* | |
1818 | * Flush the uberblocks to disk. This ensures that the odd labels | |
1819 | * are no longer needed (because the new uberblocks and the even | |
1820 | * labels are safely on disk), so it is safe to overwrite them. | |
1821 | */ | |
1822 | zio = zio_root(spa, NULL, NULL, flags); | |
1823 | ||
a1d477c2 MA |
1824 | for (int v = 0; v < svdcount; v++) { |
1825 | if (vdev_writeable(svd[v])) { | |
1826 | zio_flush(zio, svd[v]); | |
1827 | } | |
1828 | } | |
d9885b37 AH |
1829 | if (spa->spa_aux_sync_uber) { |
1830 | spa->spa_aux_sync_uber = B_FALSE; | |
1831 | for (int v = 0; v < spa->spa_spares.sav_count; v++) { | |
1832 | if (vdev_writeable(spa->spa_spares.sav_vdevs[v])) { | |
1833 | zio_flush(zio, spa->spa_spares.sav_vdevs[v]); | |
1834 | } | |
1835 | } | |
1836 | for (int v = 0; v < spa->spa_l2cache.sav_count; v++) { | |
1837 | if (vdev_writeable(spa->spa_l2cache.sav_vdevs[v])) { | |
1838 | zio_flush(zio, spa->spa_l2cache.sav_vdevs[v]); | |
1839 | } | |
1840 | } | |
1841 | } | |
34dc7c2f BB |
1842 | |
1843 | (void) zio_wait(zio); | |
1844 | ||
1845 | return (good_writes >= 1 ? 0 : EIO); | |
1846 | } | |
1847 | ||
1848 | /* | |
1849 | * On success, increment the count of good writes for our top-level vdev. | |
1850 | */ | |
1851 | static void | |
1852 | vdev_label_sync_done(zio_t *zio) | |
1853 | { | |
1854 | uint64_t *good_writes = zio->io_private; | |
1855 | ||
1856 | if (zio->io_error == 0) | |
bc89ac84 | 1857 | atomic_inc_64(good_writes); |
34dc7c2f BB |
1858 | } |
1859 | ||
1860 | /* | |
1861 | * If there weren't enough good writes, indicate failure to the parent. | |
1862 | */ | |
1863 | static void | |
1864 | vdev_label_sync_top_done(zio_t *zio) | |
1865 | { | |
1866 | uint64_t *good_writes = zio->io_private; | |
1867 | ||
1868 | if (*good_writes == 0) | |
2e528b49 | 1869 | zio->io_error = SET_ERROR(EIO); |
34dc7c2f BB |
1870 | |
1871 | kmem_free(good_writes, sizeof (uint64_t)); | |
1872 | } | |
1873 | ||
b128c09f BB |
1874 | /* |
1875 | * We ignore errors for log and cache devices, simply free the private data. | |
1876 | */ | |
1877 | static void | |
1878 | vdev_label_sync_ignore_done(zio_t *zio) | |
1879 | { | |
1880 | kmem_free(zio->io_private, sizeof (uint64_t)); | |
1881 | } | |
1882 | ||
34dc7c2f BB |
1883 | /* |
1884 | * Write all even or odd labels to all leaves of the specified vdev. | |
1885 | */ | |
1886 | static void | |
7f96cc23 MA |
1887 | vdev_label_sync(zio_t *zio, uint64_t *good_writes, |
1888 | vdev_t *vd, int l, uint64_t txg, int flags) | |
34dc7c2f BB |
1889 | { |
1890 | nvlist_t *label; | |
1891 | vdev_phys_t *vp; | |
a6255b7f | 1892 | abd_t *vp_abd; |
34dc7c2f BB |
1893 | char *buf; |
1894 | size_t buflen; | |
34dc7c2f | 1895 | |
7f96cc23 MA |
1896 | for (int c = 0; c < vd->vdev_children; c++) { |
1897 | vdev_label_sync(zio, good_writes, | |
1898 | vd->vdev_child[c], l, txg, flags); | |
1899 | } | |
34dc7c2f BB |
1900 | |
1901 | if (!vd->vdev_ops->vdev_op_leaf) | |
1902 | return; | |
1903 | ||
b128c09f | 1904 | if (!vdev_writeable(vd)) |
34dc7c2f BB |
1905 | return; |
1906 | ||
b2255edc BB |
1907 | /* |
1908 | * The top-level config never needs to be written to a distributed | |
1909 | * spare. When read vdev_dspare_label_read_config() will generate | |
1910 | * the config for the vdev_label_read_config(). | |
1911 | */ | |
1912 | if (vd->vdev_ops == &vdev_draid_spare_ops) | |
1913 | return; | |
1914 | ||
34dc7c2f BB |
1915 | /* |
1916 | * Generate a label describing the top-level config to which we belong. | |
1917 | */ | |
1918 | label = spa_config_generate(vd->vdev_spa, vd, txg, B_FALSE); | |
1919 | ||
a6255b7f DQ |
1920 | vp_abd = abd_alloc_linear(sizeof (vdev_phys_t), B_TRUE); |
1921 | abd_zero(vp_abd, sizeof (vdev_phys_t)); | |
1922 | vp = abd_to_buf(vp_abd); | |
34dc7c2f BB |
1923 | |
1924 | buf = vp->vp_nvlist; | |
1925 | buflen = sizeof (vp->vp_nvlist); | |
1926 | ||
79c76d5b | 1927 | if (!nvlist_pack(label, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP)) { |
34dc7c2f | 1928 | for (; l < VDEV_LABELS; l += 2) { |
a6255b7f | 1929 | vdev_label_write(zio, vd, l, vp_abd, |
34dc7c2f BB |
1930 | offsetof(vdev_label_t, vl_vdev_phys), |
1931 | sizeof (vdev_phys_t), | |
7f96cc23 | 1932 | vdev_label_sync_done, good_writes, |
b128c09f | 1933 | flags | ZIO_FLAG_DONT_PROPAGATE); |
34dc7c2f BB |
1934 | } |
1935 | } | |
1936 | ||
a6255b7f | 1937 | abd_free(vp_abd); |
34dc7c2f BB |
1938 | nvlist_free(label); |
1939 | } | |
1940 | ||
65c7cc49 | 1941 | static int |
b128c09f | 1942 | vdev_label_sync_list(spa_t *spa, int l, uint64_t txg, int flags) |
34dc7c2f | 1943 | { |
b128c09f | 1944 | list_t *dl = &spa->spa_config_dirty_list; |
34dc7c2f BB |
1945 | vdev_t *vd; |
1946 | zio_t *zio; | |
1947 | int error; | |
1948 | ||
1949 | /* | |
1950 | * Write the new labels to disk. | |
1951 | */ | |
1952 | zio = zio_root(spa, NULL, NULL, flags); | |
1953 | ||
1954 | for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) { | |
d6320ddb | 1955 | uint64_t *good_writes; |
428870ff BB |
1956 | |
1957 | ASSERT(!vd->vdev_ishole); | |
1958 | ||
79c76d5b | 1959 | good_writes = kmem_zalloc(sizeof (uint64_t), KM_SLEEP); |
1c27024e | 1960 | zio_t *vio = zio_null(zio, spa, NULL, |
b128c09f BB |
1961 | (vd->vdev_islog || vd->vdev_aux != NULL) ? |
1962 | vdev_label_sync_ignore_done : vdev_label_sync_top_done, | |
34dc7c2f | 1963 | good_writes, flags); |
7f96cc23 | 1964 | vdev_label_sync(vio, good_writes, vd, l, txg, flags); |
34dc7c2f BB |
1965 | zio_nowait(vio); |
1966 | } | |
1967 | ||
1968 | error = zio_wait(zio); | |
1969 | ||
1970 | /* | |
1971 | * Flush the new labels to disk. | |
1972 | */ | |
1973 | zio = zio_root(spa, NULL, NULL, flags); | |
1974 | ||
1975 | for (vd = list_head(dl); vd != NULL; vd = list_next(dl, vd)) | |
1976 | zio_flush(zio, vd); | |
1977 | ||
1978 | (void) zio_wait(zio); | |
1979 | ||
1980 | return (error); | |
1981 | } | |
1982 | ||
1983 | /* | |
1984 | * Sync the uberblock and any changes to the vdev configuration. | |
1985 | * | |
1986 | * The order of operations is carefully crafted to ensure that | |
1987 | * if the system panics or loses power at any time, the state on disk | |
1988 | * is still transactionally consistent. The in-line comments below | |
1989 | * describe the failure semantics at each stage. | |
1990 | * | |
1991 | * Moreover, vdev_config_sync() is designed to be idempotent: if it fails | |
1992 | * at any time, you can just call it again, and it will resume its work. | |
1993 | */ | |
1994 | int | |
b6fcb792 | 1995 | vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg) |
34dc7c2f BB |
1996 | { |
1997 | spa_t *spa = svd[0]->vdev_spa; | |
1998 | uberblock_t *ub = &spa->spa_uberblock; | |
b6fcb792 | 1999 | int error = 0; |
b128c09f | 2000 | int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL; |
34dc7c2f | 2001 | |
d2734cce | 2002 | ASSERT(svdcount != 0); |
b6fcb792 | 2003 | retry: |
9babb374 BB |
2004 | /* |
2005 | * Normally, we don't want to try too hard to write every label and | |
2006 | * uberblock. If there is a flaky disk, we don't want the rest of the | |
2007 | * sync process to block while we retry. But if we can't write a | |
2008 | * single label out, we should retry with ZIO_FLAG_TRYHARD before | |
2009 | * bailing out and declaring the pool faulted. | |
2010 | */ | |
b6fcb792 BB |
2011 | if (error != 0) { |
2012 | if ((flags & ZIO_FLAG_TRYHARD) != 0) | |
2013 | return (error); | |
9babb374 | 2014 | flags |= ZIO_FLAG_TRYHARD; |
b6fcb792 | 2015 | } |
9babb374 | 2016 | |
34dc7c2f BB |
2017 | ASSERT(ub->ub_txg <= txg); |
2018 | ||
2019 | /* | |
2020 | * If this isn't a resync due to I/O errors, | |
2021 | * and nothing changed in this transaction group, | |
2022 | * and the vdev configuration hasn't changed, | |
2023 | * then there's nothing to do. | |
2024 | */ | |
379ca9cf OF |
2025 | if (ub->ub_txg < txg) { |
2026 | boolean_t changed = uberblock_update(ub, spa->spa_root_vdev, | |
2027 | txg, spa->spa_mmp.mmp_delay); | |
2028 | ||
2029 | if (!changed && list_is_empty(&spa->spa_config_dirty_list)) | |
2030 | return (0); | |
2031 | } | |
34dc7c2f BB |
2032 | |
2033 | if (txg > spa_freeze_txg(spa)) | |
2034 | return (0); | |
2035 | ||
2036 | ASSERT(txg <= spa->spa_final_txg); | |
2037 | ||
2038 | /* | |
2039 | * Flush the write cache of every disk that's been written to | |
2040 | * in this transaction group. This ensures that all blocks | |
2041 | * written in this txg will be committed to stable storage | |
2042 | * before any uberblock that references them. | |
2043 | */ | |
d2734cce | 2044 | zio_t *zio = zio_root(spa, NULL, NULL, flags); |
34dc7c2f | 2045 | |
d2734cce SD |
2046 | for (vdev_t *vd = |
2047 | txg_list_head(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)); vd != NULL; | |
34dc7c2f BB |
2048 | vd = txg_list_next(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg))) |
2049 | zio_flush(zio, vd); | |
2050 | ||
2051 | (void) zio_wait(zio); | |
2052 | ||
2053 | /* | |
2054 | * Sync out the even labels (L0, L2) for every dirty vdev. If the | |
2055 | * system dies in the middle of this process, that's OK: all of the | |
2056 | * even labels that made it to disk will be newer than any uberblock, | |
2057 | * and will therefore be considered invalid. The odd labels (L1, L3), | |
2058 | * which have not yet been touched, will still be valid. We flush | |
2059 | * the new labels to disk to ensure that all even-label updates | |
2060 | * are committed to stable storage before the uberblock update. | |
2061 | */ | |
d2734cce SD |
2062 | if ((error = vdev_label_sync_list(spa, 0, txg, flags)) != 0) { |
2063 | if ((flags & ZIO_FLAG_TRYHARD) != 0) { | |
2064 | zfs_dbgmsg("vdev_label_sync_list() returned error %d " | |
2065 | "for pool '%s' when syncing out the even labels " | |
2066 | "of dirty vdevs", error, spa_name(spa)); | |
2067 | } | |
b6fcb792 | 2068 | goto retry; |
d2734cce | 2069 | } |
34dc7c2f BB |
2070 | |
2071 | /* | |
2072 | * Sync the uberblocks to all vdevs in svd[]. | |
2073 | * If the system dies in the middle of this step, there are two cases | |
2074 | * to consider, and the on-disk state is consistent either way: | |
2075 | * | |
2076 | * (1) If none of the new uberblocks made it to disk, then the | |
2077 | * previous uberblock will be the newest, and the odd labels | |
2078 | * (which had not yet been touched) will be valid with respect | |
2079 | * to that uberblock. | |
2080 | * | |
2081 | * (2) If one or more new uberblocks made it to disk, then they | |
2082 | * will be the newest, and the even labels (which had all | |
2083 | * been successfully committed) will be valid with respect | |
2084 | * to the new uberblocks. | |
2085 | */ | |
d2734cce SD |
2086 | if ((error = vdev_uberblock_sync_list(svd, svdcount, ub, flags)) != 0) { |
2087 | if ((flags & ZIO_FLAG_TRYHARD) != 0) { | |
2088 | zfs_dbgmsg("vdev_uberblock_sync_list() returned error " | |
2089 | "%d for pool '%s'", error, spa_name(spa)); | |
2090 | } | |
b6fcb792 | 2091 | goto retry; |
d2734cce | 2092 | } |
34dc7c2f | 2093 | |
379ca9cf OF |
2094 | if (spa_multihost(spa)) |
2095 | mmp_update_uberblock(spa, ub); | |
2096 | ||
34dc7c2f BB |
2097 | /* |
2098 | * Sync out odd labels for every dirty vdev. If the system dies | |
2099 | * in the middle of this process, the even labels and the new | |
2100 | * uberblocks will suffice to open the pool. The next time | |
2101 | * the pool is opened, the first thing we'll do -- before any | |
2102 | * user data is modified -- is mark every vdev dirty so that | |
2103 | * all labels will be brought up to date. We flush the new labels | |
2104 | * to disk to ensure that all odd-label updates are committed to | |
2105 | * stable storage before the next transaction group begins. | |
2106 | */ | |
d2734cce SD |
2107 | if ((error = vdev_label_sync_list(spa, 1, txg, flags)) != 0) { |
2108 | if ((flags & ZIO_FLAG_TRYHARD) != 0) { | |
2109 | zfs_dbgmsg("vdev_label_sync_list() returned error %d " | |
2110 | "for pool '%s' when syncing out the odd labels of " | |
2111 | "dirty vdevs", error, spa_name(spa)); | |
2112 | } | |
b6fcb792 | 2113 | goto retry; |
d2734cce | 2114 | } |
b6fcb792 BB |
2115 | |
2116 | return (0); | |
34dc7c2f | 2117 | } |